Author response:

The following is the authors’ response to the previous reviews

Reviewer #1 (Public review):

Summary:

The crystal structure of the Sld3CBD-Cdc45 complex presented by Li et al. is a significant contribution that enhances our understanding of CMG formation during the rate-limiting step of DNA replication initiation. This structure provides crucial insights into the intermediate steps of CMG formation, and the particle analysis and model predictions compellingly describe the mechanism of Cdc45 loading. Building upon previously known Sld3 and Cdc45 structures, this study offers new perspectives on how Cdc45 is recruited to MCM DH through the Sld3-Sld7 complex. The most notable finding is the structural rearrangement of Sld3CBD upon Cdc45 binding, particularly the α8-helix conformation, which is essential for Cdc45 interaction and may also be relevant to its metazoan counterpart, Treslin. Additionally, the conformational shift in the DHHA1 domain of Cdc45 suggests a potential mechanism for its binding to Mcm2NTD. Furthermore, Sld3's ssDNA-binding experiments provide evidence of its novel functions in the DNA replication process in yeast, expanding our understanding of its role beyond Cdc45 recruitment.

Strengths:

The manuscript is generally well-written, with a precise structural analysis and a solid methodological section that will significantly advance future studies in the field. The predictions based on structural alignments are intriguing and provide a new direction for exploring CMG formation, potentially shaping the future of DNA replication research. This research also opens up several new opportunities to utilize structural biology to unravel the molecular details of the model presented in the paper.

Weaknesses:

The main weakness of the manuscript lies in the lack of detailed structural validation for the proposed Sld3-Sld7-Cdc45 model, and its CMG bound models, which could be done in the future using advanced structural biology techniques such as single particle cryo-electron microscopy. It would also be interesting to explore how Sld7 interacts with the MCM helicase, and this would help to build a detailed long-flexible model of Sld3-Sld7-Cdc45 binding to MCM DH and to show where Sld7 will lie on the structure. This will help us to understand how Sld7 functions in the complex. Also, future experiments would be needed to understand the molecular details of how Sld3 and Sld7 release from CMG is associated with ssARS1 binding.

The proposals based on this study provide new knowledge of the CMG formation process. We agree that our Sld3-Sld7-Cdc45 model will be further confirmed by cryo-EM. We improved our ssARS1-binding assay and quantified data (See the response to Recommendations for the authors of #3 review).

Reviewer #2 (Public review):

Summary

The manuscript presents valuable findings, particularly in the crystal structure of the Sld3CBD-Cdc45 interaction and the identification of additional sequences involved in their binding. The modeling of the Sld7-Sld3CBD-CDC45 subcomplex is novel, and the results provide insights into potential conformational changes that occur upon interaction. Although the single-stranded DNA binding data from Sld3 of different species is a minor weakness, the experiments support a model in which the release of Sld3 from the complex may be promoted by its binding to origin single-stranded DNA exposed by the helicase.

Strengths

The Sld3CBD-Cdc45 structure is a novel contribution, revealing critical residues involved in the interaction.

The model structures generated from the crystal data are well presented and provide valuable insights into the interaction sequences between Sld3 and Cdc45.

The experiments testing the requirements for interaction sequences are thorough and conducted well, with clear figures supporting the conclusions.

The conformational changes observed in Sld3 and Cdc45 upon binding are interesting and enhance our understanding of the interaction.

The modeling of the Sld7-Sld3CBD-CDC45 subcomplex is a new and valuable addition to the field.

The proposed model of Sld3 release from the complex through binding to single stranded DNA at the origin is intriguing.

Weaknesses

The section on the binding of Sld3 complexes to origin single-stranded DNA is somewhat weakened by the use of Sld3 proteins from different species. The comparisons between Sld3-CBD, Sld3CBD-Cdc45, and Sld7-Sld3CBD-Cdc45 involve complexes from different species, limiting the comparisons' value.

Although the study reveals that Sld3 binds to different residues of Cdc45 than those previously shown to bind Mcm or GINS, the data in the paper do not shed any additional light on how GINS and Sld3 binding to Cdc45 or Mcms. would affect each other. Other previous research has suggested that the binding of GINS and Sld3 to Mcm or Cdc45 may be mutually exclusive. The authors acknowledge that a structural investigation of Sld3, Sld7, Cdc45, and MCM during the stage of GINS recruitment will be a significant goal for future research.

We agree that it is better to use all samples from a source; however, due to limitations in protein expression, we used Sld7-Sld3CBD-Cdc45 from a different source. The two sources used in this study belong to the same family, and the proteins Sld7, Sld3 and Cdc45 share sequence conservation with similar structures predicted by Alphafold3 (RMSD = 0.356, 1.392, and 0.891 for Ca atoms of Sld7CTD, Sld7NTD-Sld3NTD, and Sld3CBD-Cdc45). Such similarity in source and proteins allows us to do the comparison. We also mentioned that a cryo-EM study of Sld3-Sld7-Cdc45-MCM and Sld3-Sld7-CMG structures will be a significant goal for future research in our manuscript.

Reviewer #3 (Public review):

Summary:

The paper by Li et al. describes the crystal structure of a complex of Sld3-Cdc45-binding domain (CBD) with Cdc45 and a model of the dimer of an Sld3-binding protein, Sld7, with two Sld3-CBD-Cdc45 for the tethering. In addition, the authors showed the genetic analysis of the amino acid substitution of residues of Sld3 in the interface with Cdc45 and biochemical analysis of the protein interaction between Sld3 and Cdc45 as well as DNA binding activity of Sld3 to the single-strand DNAs of the ARS sequence.

Strengths:

The authors provided a nice model of an intermediate step in the assembly of an active Cdc45-MCM-GINS (CMG) double hexamers at the replication origin, which is mediated by the Sld3-Sld7 complex. The dimer of the Sld3-Sld7 complexes tethers two MCM hexamers together for the recruitment of GINS-Pol epsilon on the replication origin.

Weaknesses:

The biochemical analysis should be carefully evaluated with more quantitative ways to strengthen the authors' conclusion even in the revised version.

In this revision, we improved our ssARS1-binding assay in more quantitative ways (See the response to Recommendations for the authors).

Reviewer #1 (Recommendations for the authors):

I thank the authors for all their replies to my previous questions and for doing all the necessary corrections. I am satisfied with most of their replies, however, upon second reading I have a few more suggestions which could help to improve the manuscript further and make an impact in the field. My comments are listed below.

(1) In general, the manuscript is well structured, but I feel that it requires professional English correction. In many places it was difficult to understand the sentences and I had to read it several times to understand it. Also, very long sentences should be avoided. The flow should be easy to read and understand, and that is why I feel it requires professional English correction.

Following the comment, we checked English carefully and shortened the very long sentences.

(2) Page 5, line 103, please include molecule after the word complex to make it like- "Only one complex molecule exists within an asymmetric unit."

We revised this sentence (P5/L103).

(3) Line 113- more than the N-terminal half of the protruding long helix α7 113 was disordered in the Sld3CBD-Cdc45 complex. This sentence is not clear. What does it mean more than the N-terminal half? Please rewrite it.

We revised this sentence to give the corresponding residue number “(D219–H231)” (P5/L114).

(4) Page 5, result 2- Conformation changes in Sld3CBD and Cdc45 for binding each other, this section may require a little restructuring. Line 130-131- "Therefore, the helix α8CTP seems to be an intrinsically disordered segment when Sld3 alone but 130 folds into a helix coupled to the binding partner Cdc45 in the Sld3CBD-Cdc45 complex." This statement is the crux of the structural finding and therefore, I feel it should move after the first sentence.

Thank you for your comments. We rewrote this part (P5/L128-131).

(5) Line 121-122: Compared to the isolated form (PDBIDs: 5DGO 121 for huCdc45 [31] and 6CC2 for EhCdc45 [33]) and the CMG form (PDBID: 3JC6. Write it in the same format. Make 6CC2 in bracket like other PDB IDs. Restructure this sentence.

We revised this sentence (P5/122-123).

(6) Line 127-129: This sentence is also not very clear.

We revised this sentence together with above No (4). (P5/L128-131)

(7) In my question 4- "Can authors add a supplementary figure showing the probability of disordernes..."., I meant to use a disorder prediction tool like IUPred for the protein sequences and show that α8 is predicted to be a disordered upon sequence analysis. This will help to show the inherent property of α8 helix, and it could add up to the understanding that a disordered region is being structured in the complex structure.

The structures showed that α8CTP is stabilized by binding with Cdc45, but disordered in Sld3CBD alone, indicating that this part is flexible, like an intrinsically disordered segment. We have deposited the structure to PDB, so predictions like IUPred cannot show meaningful information.

(8) Question 9 regarding Supplementary Figure 8- Please include your statement in the figure legend - "WT Sld3CBD was prepared in a complex with Cdc45, while the mutants of Sld3CBD existed alone, we calculated the elements of secondary structure from the crystal structure of Sld3CBD-Cdc45. The concentration of samples was controlled to the same level for CD measurement."

Following the comment, we optimized the figure legend of Supplementary Figure 8.

(9) Question 13- I understand that negative staining and SEC-SAXS experiments could be very tricky for such protein complexes, which have very long loops and are flexible. Did authors try a GraFix cross-linking before doing the negative staining TEM? If it is not being tried, then it might be a good idea to try it and it may help to get much cleaner particles and easier class averaging. Although I completely understand the technical challenges the authors describe and I agree with them, I still feel that one good experiment that shows this dimer model would be very helpful to strengthen the claim. I am concerned because if people start using a similar DLS experiment to calculate intermolecular distances, citing your paper, in many cases it might be a wrong interpretation. In case the negative staining still does not work, at least discuss your technical challenges in the discussion section and mention that SEC-SAXS showed a similar length of the complex and show the Guinier plot and Porod plots in the supplementary data.

We believe that DLS is one of the methods for analyzing the single particle size. Of course, the confirmation by multiple methods will give compelling evidence. Following the comment, we added SEC-SAXS data in the [Results] (P7/L194-196) (Cdc45 recruitment to MCM DH by Sld3 with partner Sld7) and Supplementary Figure 11. The Sld7-Sld3-Cdc45 forms a flexible, long shape. Each binding domain is rigid but linked by the long loops. The flexibility problems are caused by the long loop linkers, but not by binding. So, we did not try to use the cross-linking method for analysis experiments.  

(10) Page 8, line 221- litter sequence specificity: Correct the word "litter" with little. Also, the word shaped is written as sharped at a few places in the manuscript. Please correct it.

We apologize for making such mistakes. We have modified these words.

(11) Page 9, line 237-238: Would it be possible to add a lane showing Sld7 binding to the ssDNA in figure 4. I recommend showing this to understand the ssDNA binding affinity of Sld7 by itself and it will also help us to compare when it is in complex with Sld3.

Considering that Sld7 on CMG is always a complex with Sld3, the ssDNA binding affinity should use the Sld3-Sld7 complex. Additionally, we attempted to overexpress Sld7, but could not obtain the target protein.

Reviewer #2 (Recommendations for the authors):

Thank you for the improved manuscript. The following sentence is unclear: "Cdc45 binds tighter to long ssDNA (>60 bases) with a litter sequence specificity".

We apologize for making such a mistake. We modified “litter” to “little”.

I found it challenging to understand which species were used while reading the results section and figure legends. I recommend that the authors revise the text in both the results and figure legends to clearly indicate when proteins from different species are being compared. Additionally, it would be valuable to explicitly acknowledge this limitation in the text.

Following the comment, we added a description for using different species in results (P8/L224-225) and figure legends (Supplementary Figure 14). We added more information in the Methods to explain why we used two species for preparing proteins.

Reviewer #3 (Recommendations for the authors):

Major points:

(1) The current title is not appropriate for the general readers. At least, DNA replication or DNA replication initiation should be added and abbreviations such as CBD should be avoided.

Following the comment, we added “DNA replication” into the title. Regarding “CBD”, since the full name of “Cdc45 binding domain” is too long, we continue to use Sld3CBD.

(2) As in my previous review, I asked for quantification of the EMSA assay shown in Figure 4 and Supplemental Figures 13 and 14. Since some signals of the bands are very weak, it is hard to conclude something. Given different protein concentrations used in the experiment, the authors should provide any kinds of value. For example, Sld3CBD-CDC45 shows weaker DNA binding than Sld3CBD alone (line 231). Is this true (or reproducible)? It is hard to conclude without any quantification.

We have repeated the EMSA assay four or more times with different rods of overexpression, purification and DNA synthesis, indicating that the EMSA assay is reproducible. In this revision, we changed the DNA stain and adjusted the ratio between the protein and ssDNA with increasing concentrations. The smeared bands of ssDNA with Sld7–Sld3ΔC–Cdc45 or Sld7–Sld3ΔC exhibit enhanced discernibility, and the ssDNA bands are intense enough for grayscale calculations (Figure 4 in the second revised version). We used a series of t-tests to confirm a significantly ssDNA residual level between Sld3CBD–Cdc45 to Sld3CBD, Sld7–Sld3ΔC–Cdc45, and Sld7–Sld3ΔCS (t-test, ****: P<0.0001). We also carefully controlled the sample amount in the EMAS assay and described it in the [Methods].

Moreover, in this EMSA assay (in Figure 4), the authors suggest that the disappearance of ssDNA bands corresponds with the binding of the protein to the DNA. However, it is also possible that the DNA is degraded. It is very important to show the band of protein-DNA complexes on the gel (a whole gel, not the parts of the gel shown in Figure). Why did the authors use this "insensitive" assay using SyberGreen, not radio-labelled ssDNA?

In this revision, we added a negative control of no ssDNA-binding by using ssARS1-3_3 for all protein samples (Sld3CBD, Sld3CBD–Cdc45, Sld7–Sld3ΔC–Cdc45 and Sld7–Sld3ΔC), which were the same rod of expression and purification for bound to ssARS1s (ssARS1-2 and ssARS1-5) (Figure 4), showing that the disappearance of ssDNA bands is caused by binding to proteins, not degradation. Moreover, this time, by changing the DNA stain and increasing the concentration of the samples, the smeared ssDNA bands exhibit enhanced discernibility in the high molecular weight regions when mixed with Sld7–Sld3ΔC–Cdc45 or Sld7–Sld3ΔC, whereas no bands appeared in the NC (ssARS1-3_1). The positions of smeared ssDNA bonds correspond to those of protein in the protein-stain pages, indicating that ssARS1 were complexed with proteins. Following the comment, we show all bands on the gel in Figure 4 and Supplementary Figure 14. Compared to Sld7–Sld3ΔC–Cdc45 or Sld7–Sld3ΔC, Sld3CBD and ssDNA bonds could not be observed because the pI value of Sld3CBD, which affects the entry of the samples into the gel.

We agree that using radio-labelled ssDNA can obtain a sensitive binding assay. However, current laboratory constraints did not allow us to use radio-labelled ssDNA. Furthermore, considering the characteristics of our target proteins, Sld3CBD, Sld3CBD–Cdc45, Sld7–Sld3ΔC–Cdc45, and Sld7–Sld3ΔC, we planned to perform the binding assay in a more natural state without any modifications, labelling or linkers. Additionally, we have attempted to use ITC experiments but failed in the measurements. Presumably, the conformational flexibility of Sld7-Sld3-Cdc45 and Sld7-Sld3 caused a thermodynamic anomaly.

Minor points:

(1) Line 215, 80b: This should be "80 nucleotides(nt)". Throughout the text, nucleotides is better than base to show the length of ssDNAs.

Thank you for your comments. We modified these words throughout the text.

Author response:

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

Reviewer #1 (Public review):

“This is an exploratory study that doesn't explore quite enough. Critically, the authors make a point of mentioning that neuronal firing properties vary across cell types, but only use baseline firing rate as a proxy metric for cell type. This leaves several important explorations on the table, not limited to the following:”

1a: “Do waveform shape features, which can also be informative of cell type, predict the effect of stimulation?”

To address this question, we modeled our approach to cell type classification after Peyrache et al. 2012. More specifically, we extracted two features from the mean unit waveforms—the valley-to-peak time (VP) and the peak half-width (PHW). These features were then used to classify units into two distinct clusters (k-means, clusters = 2, based on a strong prior from existing literature), representing putative excitatory and inhibitory neurons. Our approach recapitulated many of the same observations in Peyrache et al. 2012, namely (1) identification of two clusters (low PHW/VP: inhibitory, high PHW/VP: excitatory), (2) an ~80/20 ratio of excitatory/inhibitory neurons, and (3) greater baseline firing rates in the inhibitory vs. excitatory neurons. However, we did not observe a preferential modulation of one cell type compared to another (see newly created Figure 4). A description of this analysis and its takeaways has been incorporated into the manuscript.

Change to Text:

Created Figure 4 (Separation of presumed excitatory and inhibitory neurons by waveform morphology).

Caption: (A) Two metrics were calculated using the averaged waveforms for each detected unit: the valley-to-peak width (VP) and peak half-width (PHW). (B) Scatterplot of the relationship between VP and PHW; note that units with identical metrics are overlaid. Using k-means clustering, we identified two distinct response clusters, representing presumed excitatory (E, blue) and inhibitory (I, red) neurons. The units from which the example waveforms were taken are outlined in black. Probability distributions for each metric are shown along the axes. (C) Total number of units within each cluster, separated by region. (D) Comparison of baseline firing rates, separated by cluster. (E) Percent of modulated units in each cluster. * p < 0.05, NS = not significant.

Added a description of clustering methodology to lines 132-137: “We calculated two metrics from the averaged waveform from each detected unit: the valley-to-peak-width (VP) and the peak half-width (PHW) (Figure 4A); previously, these two properties of waveform morphology have been used to discriminate pyramidal cells (excitatory) from interneurons (inhibitory) in human intracranial recordings (Peyrache et al., 2012). Next, we performed k-means clustering (n = 2 clusters) on the waveform metrics, in line with previous approaches to cell type classification.

Added a section in the Results titled “Theta Burst Stimulation Modulates Excitatory and Inhibitory Neurons Equally”. Lines 370-378: “Using k-means clustering, we grouped neurons into two distinct clusters based on waveform morphology, representing neurons that were presumed to be excitatory (E) and inhibitory (I) (Figure 4B). Inhibitory (fast-spiking) neurons exhibited shorter waveform VP and PHW, compared with excitatory (regular-spiking) neurons (I cluster centroid: VP = 0.50ms, PHW = 0.51ms; E cluster centroid: VP = 0.32ms, PHW = 0.31ms), and greater baseline firing rates (U(N_I = 23, N<_E = 133) = 1074.50, p = 0.023) (Figure 4D). Although we observed a much greater proportion of excitatory vs. inhibitory neurons (E: 85.3%, I: 14.7%), stimulation appeared to affect excitatory and inhibitory neurons equally, suggesting that one cell type is not preferentially activated over another (Figure 4E).

Modified discussion of the effects of stimulation on different cell types. Lines 475-483: “…To test these hypotheses directly, we clustered neurons into presumed excitatory and inhibitory neurons based on waveform morphology. In doing so, we observed ~85% excitatory and ~15% inhibitory neurons, which is very similar what has been reported previously in human intracranial recordings (Cowan et al. 2024, Peyrache et al., 2012). Interestingly, stimulation appeared to modulate approximately the same proportion of neurons for each cell type (~30%), despite the differently-sized groups. Recent reports, however, have suggested that the extent to which electrical fields entrain neuronal spiking, particularly with respect to phase-locking, may be specific to distinct classes of cells (Lee et al., 2024).”

1b:  “Is the autocorrelation of spike timing, which can be informative about temporal dynamics, altered by stimulation? This is especially interesting if theta-burst stimulation either entrains theta-rhythmic spiking or is more modulatory of endogenously theta-modulated units.”

The reviewer is correct in suggesting that rate-modulation represents only one of many possible ways by which exogenous theta burst stimulation may influence neuronal activity. Indeed, intracranial theta burst stimulation has previously been shown to evoke theta-frequency oscillatory responses in local field potentials (Solomon et al. 2021), and other forms of stimulation (i.e., transcranial alternating current stimulation) may modulate the rhythm, rather than the rate, of neuronal spiking (Krause et al. 2019).

To investigate whether stimulation altered rhythmicity in neuronal firing, we contrasted the spike timing autocorrelograms, as suggested. More specifically, we computed the pairwise differences in spike timing for each trial, separating spikes into the same pre-, during-, and post-stimulation epochs described in the manuscript (bin size = 5 ms, max lag = 250 ms), grouped neurons by whether they were modulated, and then contrasted the differences in the latencies of the peak normalized autocorrelation value between epochs. Only neurons with a firing rate of ≥ 1 Hz (n = 70/203, 34.5%) were included in this analysis since sparse firing resulted in noisy autocorrelation estimates. Subsequent statistical testing of the peak latency differences between pre-/during- and pre-/post-stimulation did not reveal any group-level differences (Mann-Whitney U tests, p > 0.05). Thus, we were not able to identify neuronal responses suggestive of altered rhythmicity (see Figure S5). A description of this analysis and its takeaways has been incorporated into the manuscript.

Of note, there are two elements of the data that constrain our ability to detect modulation in the rhythm of firing. First, the baseline activity recorded across neurons modulated by stimulation was relatively low (i.e., median firing rate = 1.77 Hz). Second, stimulation often resulted in a suppression, rather than an enhancement, of firing rate. Taken together, the sparse firing afforded limited opportunity to characterize changes to subtle patterns of spiking. 

Change to Text:

Created Figure S5 (Analysis of modulation in spiking rhythmicity)

Caption: (A) Representative autocorrelograms ACG) for a single neuron. The pairwise differences in spike timing were computed for each trial and epoch (bin size = 5 ms, max lag = 250 ms), then smoothed with a Gaussian kernel. The peak in the normalized ACG across trials was computed for each epoch. (B) Kernel density estimate of the peak ACG lag, separated by epoch. (C) The peak ACG lags were split by whether the neuron was modulated (Mod) or unaffected by stimulation (NS = not significant) for each of the two contrasts: pre- vs. during-stim (left) and pre- vs. post-stim (right).

Details about the autocorrelation methodology have been incorporated. Lines 166-172: “To investigate whether stimulation altered rhythmicity in neuronal firing, we analyzed the spike timing autocorrelograms. More specifically, we computed the pairwise differences in spike timing for each trial (bin size = 5 ms, max lag = 250 ms) and then contrasted the differences in the latencies of the peak normalized autocorrelation value between epochs (pre-, during-, post-stimulation). Only neurons with a firing rate of ≥ 1 Hz (n = 70/203, 34.5%) were included in this analysis since sparse firing resulted in noisy autocorrelation estimates.

The results from contrasting the autocorrelograms are now mentioned briefly. Lines 297-298: “Stimulation, however, did not appear to alter the rhythmicity in neuronal firing, as measured by spiking autocorrelograms (Figure S5).”

1c: “The authors reference the relevance of spike-field synchrony (30-55 Hz) in animal work, but ignore it here. Does spike-field synchrony (comparing the image presentation to post-stimulation) change in this frequency range? This does not seem beyond the scope of investigation here.”

We agree that a further characterization of spike-field and spike-phase relationships may provide rich insights into more complex regional and interregional dynamics that may be altered by stimulation. Given that many metrics are biased by sample size (e.g., number of spikes), which can vary considerably, computing the pairwise phase consistency (PPC) between spikes and LFP is a preferred metric (Vinck et al. 2010). Although PPC is unbiased, its variance nonetheless increases considerably with low spike counts; pooling spike counts across trials, however, decouples the temporal relationship between spiking and the LFP phase for each trial, confounding results and yielding an unstable estimate.

To determine whether such an analysis is indeed possible, we calculated the percentage of stimulation trials with ≥ 10 spikes in both the 1s pre- and post-stimulation epochs (a relatively low threshold for inclusion). Only a very small proportion of the total number of trials across all neurons met this criterion (2.5%). Thus, because of the sparse spiking in our data, we are unable to reliably characterize spike-field or spike-phase modulation in detected neurons.

Change to Text:

In the manuscript, we have added a description of why our data is not well-suited to investigate these relationships.

Lines 532-538: “The present study did not investigate interactions between spiking activity and local field potentials because neuronal spiking was sparse at baseline and often further suppressed by stimulation; only a very small proportion of the total number of trials across all neurons exhibited ≥ 10 spikes in both the 1s pre- and post-stimulation epochs (~2.5%). Although certain metrics are not biased by sample size (e.g., pairwise phase consistency), low spike counts can dramatically affect variance and, therefore, result in unstable estimates (Vinck et al., 2011).

1d: “How does multi-unit activity respond to stimulation? At this somewhat low count of neurons (total n=156 included) it would be valuable to provide input on multi-unit responses to stimulation as well.”

We thank the reviewer for this suggestion. We have incorporated an analysis of multiunit activity (MUA), which similarly identifies robust modulation via permutation-based statistical testing and characterizes the different profiles of responses (i.e., increased vs. decreased MUA threshold crossings pre- vs. post-stimulation).

Change to Text:

Created Figure S8 (Analysis of multiunit activity response to stimulation)

Caption: (A) Example trace of multiunit activity (MUA) in one channel during a single stimulation trial. Threshold crossings are highlighted with a pink dot overlaid on the MUA signal with a corresponding hash below. (B) The percentage of channels with significantly modulated MUA, separated by the direction of effect. (C) The percentage of channels with significantly modulated MUA, separated by direction effect and region. Inc (red; post > pre) vs. Dec (blue; post < pre). HIP = hippocampus, OFC = orbitofrontal cortex, AMY = amygdala, ACC = anterior cingulate cortex. *** p < 0.001, NS = not significant.

Details about the MUA methodology have been incorporated. Lines 174-180: “Finally, we measured modulation in multiunit activity (MUA) by filtering the microleectrode signals in a 300-3,000 Hz window and counting the number of threshold crossings. Thresholds were determined on a per-channel basis and defined as -3.5 times the root mean square of the signal during the baseline period; activity during stimulation was excluded since stimulation artifact is difficult to separate from MUA in the absence of spike sorting.

MUA results are now incorporated. Lines 365-367: “Additional characterization of MUA revealed a dominant signature of increased activity post- vs. pre-stimulation, in line with these trends observed at the single-neuron level (Figure S8).”

1e: “Several intracranial studies have implicated proximity to white matter in determining the effects of stimulation on LFPs; do the authors see an effect of white matter proximity here?”

We thank the reviewer for the interesting question. Subsequent characterization revealed only small differences in the proximity of stimulation contacts to white matter (range 1.5-8.0 mm), likely because the chosen target (i.e., basolateral amygdala) has several nearby white matter structures (e.g., stria terminalis). Nonetheless, we performed a linear regression between the proximity to white matter and the stimulation-induced effect on behavior (stimulation vs. no-stimulation d’ difference), the results of which indicate no clear association (p > 0.05; see Figure S9). Critically, this is not to suggest that white matter proximity has no interaction with the reported behavioral effects, but rather, that we could not identify such an association within our data.

Change to Text:

Created Figure S9 (The effect of stimulation proximity to white matter and distance to recorded neurons).

Caption: (A) Kernel density estimate of the Euclidean distance from stimulation contacts to nearest WM structure (in mm); hash marks represent individual observations. (B) The change in memory performance (Δd’) was linearly regressed onto the distance from the stimulated contacts to white matter.

The following has been added to lines 405-426: “Proximity to white matter has been shown to influence the effects of stimulation on behavior and the strength of evoked responses (Mankin et al., 2021; Mohan et al., 2020; Paulk et al., 2022). Across all stimulated contacts, we observed only small differences in the proximity of stimulation contacts to white matter (median = 4.5 mm, range = 1.5-8.0 mm), likely because the chosen target (i.e., basolateral amygdala) has several nearby white matter structures (e.g., stria terminalis). Nonetheless, we performed a linear regression between the proximity to white matter and the stimulation-induced effect on behavior (stimulation vs. no-stimulation d’ difference), the results of which indicate no clear association (p > 0.05; see Figure S9).

Comment 2: “It is a little confusing to interpret stimulation-induced modulation of neuronal spiking in the absence of stimulation-induced change in behavior. How do the authors findings tell us anything about the neural mechanisms of stimulation-modulated memory if memory isn't altered? In line with point #1, I would suggest a deeper dive into behavior (e.g. reaction time? Or focus on individual sessions that do change in Figure 4A?) to make a stronger statement connecting the neural results to behavioral relevance.”

We agree that the connection between the observed stimulation-induced neuronal modulation and effects on behavior is unclear and has proven challenging to elucidate. Per the reviewer’s suggestion, we further focused our analyses on the neuronal modulation effects in the individual sessions that resulted in a robust change in memory performance (stimulation vs. no-stimulation d’ difference threshold of ± 0.5, based on a moderate effect size for Cohen’s d); both a positive and negative threshold were used to capture robust changes in memory performance associated with firing rate modulation, whether enhancement or suppression. To this end, we contrasted the proportion of modulated neurons in the sessions where stimulation resulted in a robust behavioral change (Δd’) with those that did not (~d’). We did not observe a difference in the proportions between groups when collapsed across all sampled regions, or when separately evaluated (Fisher’s exact tests, p > 0.05; see Figure 5C).

Given that this approach did not further clarify the connection between our neural and behavioral results, we believe it is most appropriate to deemphasize claims in the manuscript regarding the potential insights for behavioral modulation (e.g., memory enhancement), and have done so.

Change to Text:

Toned down reference to the memory-related effects of stimulation in the abstract by removing the following lines from the abstract: “Previously, we demonstrated that intracranial theta burst stimulation (TBS) of the basolateral amygdala (BLA) can enhance declarative memory, likely by modulating hippocampal-dependent memory consolidation…” and “…and motivate future neuromodulatory therapies that aim to recapitulate specific patterns of activity implicated in cognition and memory.”

Changed Figure 4 to Figure 5

Created Figure 5C (Interaction between behavioral effects and neuronal modulation)(C)  Change in recognition memory performance was split into two categories using a d’ difference threshold of ± 0.5: responder (positive or negative; Δd’, pink) and non-responder (~d’, grey). Individual d’ scores are shown (left) with points colored by outcome category; dotted lines demarcate category boundaries, and the grey-shaded region represents negligible change. The number of sessions within each outcome category (middle) and the proportion of modulated units as a function of outcome category, separated by region (right). NS = not significant.

The description of the behavioral results has been updated. Lines 394-403: “At the level of individual sessions, we observed enhanced memory (Δd’ > +0.5) in 36.7%, impaired memory (Δd’ < -0.5) in 20.0%, and negligible change (-0.5 ≤ Δd’ ≤ 0.5) in 43.3% when comparing performance between the stim and no-stim conditions; a threshold of Δd’ ± 0.5 was chosen for this classification based on the defined range of a “medium effect” for Cohen’s d. To test our hypothesis that neuronal modulation would be associated with changes in memory performance, we combined the sessions that resulted in either memory enhancement or impairment and contrasted the proportion of modulated units across regions sampled. We did not, however, observe a meaningful difference in the proportion of modulated units when grouped by behavioral outcome (all contrasts p > 0.05) (Figure 5C).

Lines 213-214 and 394-397 have been edited to reflect a change in the d’ threshold used for categorizing behavioral results (from Δd’ ± 0.2 to Δd’ ± 0.5).

Comment 3: “It is not clear to me why the assessment of firing rates after image onset and after stim offset is limited to one second - this choice should be more theoretically justified, particularly for regions that spike as sparsely as these.”

We thank the reviewer for this question and acknowledge that no clear justification was provided for this decision in the manuscript. Our decision to limit each of the analysis epochs to 1s was chosen for two reasons. First, the maximum possible length of the during-stimulation epoch was 1 s (stim on for 1 s). Although the pre- and post-stimulation epochs could be extended without issue, we were concerned that variable time windows could introduce a bias, for instance, resulting in different variances between epochs. Second, we anticipated, both from empirical observations and prior literature, that the neural response following stimulation or task features (e.g., image onset/offset) was likely to be transient, rather than sustained for a period of many seconds. By keeping the windows short, we ensured that our approach to detecting modulation (i.e., contrasting trial-wise spike counts between each pair of epochs) captured the intended effect rather than random noise. We have incorporated a discussion of this rationale in the Peri-Stimulation Modulation Analyses section.

Change to Text:

Lines 156-158 have been added: “Each epoch was constrained to 1 s to ensure that subsequent firing rate contrasts were unbiased and to capture potential transient effects (e.g., image onset/offset).”

Comment 4: “This work coincides with another example of human intracranial stimulation investigating the effect on firing rates (doi: https://doi.org/10.1101/2024.11.28.625915). Given how incredibly rare this type of work is, I think the authors should discuss how their work converges with this work (or doesn't).”

Thank you for bringing this highly relevant work to our attention. We were unaware of this recent preprint and have incorporated a discussion of its main findings into the manuscript.

Change to Text:

New citations: van der Plas et al. 2024 (bioRxiv), Cowan et al. 2024 (bioRxiv)

The discussion of related studies has been updated. Lines 447-457: “Few studies, however, have characterized the impact of electrical stimulation via macroelectrodes on the spiking activity of human cortical neurons, none of which involve intracranial theta burst stimulation. One study reported a long-lasting reduction in neural excitability among parietal neurons, with variable onset time and recovery following continuous transcranial TBS in non-human primates (Romero et al., 2022). In a similar vein, it was recently shown that human neurons are largely suppressed by single-pulse electrical stimulation (Cowan et al., 2024; Plas et al., 2024). Other emerging evidence suggests that transcranial direct current stimulation may entrain the rhythm rather than rate of neuronal spiking (Krause et al., 2019) and that stimulation-evoked modulation of spiking may meaningfully impact behavioral performance on cognitive tasks (Fehring et al., 2024).”

Comment 5: “What information does the pseudo-population analysis add? It's not totally clear to me.”

We recognize the need to further contextualize the motivation for the exploratory pseudo-population analysis and appreciate the reviewer for bringing the lack of detail to our attention. In brief, the analysis allowed us to observe trends in activity across populations of neurons, which, in principle, are not visible by characterizing modulation solely in discrete neurons. Additional details have been incorporated into the manuscript, as suggested.

Change to Text:

Additional justification has been incorporated in the description of the methodology. Lines 185-187: “…This approach enables the identification of dominant patterns of coordinated neural activity that may not be apparent when examining individual neurons in isolation.”, lines 192-194: “…By collapsing across subjects into a common pseudo-population, this analysis provides a mesoscale view of how stimulation modulates shared activity patterns across anatomically distributed neural populations.”

A summary interpretation has been added to the paragraph describing the results. Lines 326-328: “Taken together, these analyses reveal global structure in the state space of responses to BLA stimulation within hippocampal circuits.”

Reviewer #2 (Public review):

Comment 1 “Authors suggest that the units modulated by stimulation are largely distinct from those responsive to image offset during trials without stimulation. The subpopulation that responds strongly also tends to have a higher baseline of firing rate. It's important to add that the chosen modulation index is more likely to be significant in neurons with higher firing rates.”

This is an important point that was not previously addressed in our manuscript. We suspect there are likely two factors at play worth considering with respect to our chosen nonparametric modulation index: neurons with lower activity require smaller changes in spike counts to be significantly modulated (easier to flip ranks), and neurons with higher activity empirically exhibit greater absolute shifts in the number of spikes. Our further use of permutation testing, while mitigating false positives, may also somewhat constrain the ability to detect modulation in sparsely active neurons. Nonetheless, given that many trials entailed few or no spikes, we believe this approach is preferable to alternatives that may be more susceptible to noise (e.g., percent change in trial-averaged firing rate from baseline).

To better understand the tradeoffs with detection probability, we performed a sensitivity analysis. We generated synthetic data with different baseline firing rates (0.1-5.0 Hz) and effect sizes (± 0.1-0.7 Hz) and simulated the likelihood of detection with our given modulation index across neurons. The results of the simulation support the notion that the probability of detecting modulation is lower for sparsely active neurons (Figure S8C). Further discussion of this consideration for the chosen modulation index, as well as details regarding the sensitivity analysis, have been incorporated into the manuscript.

Change to Text:

Created Figure S7C (Detection probability analysis)

Caption: The same permutation-based analyses reported in the manuscript were repeated under different control conditions… (C) Visualization of the predicted probability of detecting modulation across synthetic neurons with variable firing rates and modulation effect sizes; FR = firing rate.

Lines 223-224 have been added to the Methods section titled “Firing Rate Control Analyses”: “We performed a series of control analyses to test whether our approach to firing rate detection was robust…”

A description of the simulation has been incorporated into the same section as above. Lines 234-237: “Finally, to better understand the tradeoffs with our statistical approach, we generated synthetic data with different baseline firing rates (0.1-5.0 Hz) and effect sizes (± 0.1-0.7 Hz), then simulated the likelihood of detecting modulation across variable conditions (Figure S7C).”

The description of the results from the control analyses has been updated. Lines 330-339: “Finally, we performed three supplementary analyses to evaluate the robustness of our approach to detecting firing rate modulation: a sensitivity analysis assessing the proportion of modulated units at different firing rate thresholds for inclusion/exclusion, a data dropout analysis designed to control for the possibility that non-physiological stimulation artifacts may preclude the detection of temporally adjacent spiking, and a synthetic detection probability analysis. These results recapitulate our observation that units with higher baseline firing are most likely to exhibit modulation (though the probability of detecting modulation is lower for sparsely active neurons) and suggest that suppression in firing rate is not solely attributable to amplifier saturation following stimulation (Figure S7).

Comment 2: “Readers can benefit from understanding with more details the locations chosen for stimulation - in light of previous studies that found differences between effects based on proximity to white matter (For example - PMID 32446925, Mohan et al, Brain Stimul. 2020 and PMID 33279717 Mankin et al Brain Stimul. 2021).”

This has been addressed in the above response to Reviewer’s 1 comment 1.1e.

Change to Text:

See changes related to Reviewer 1 comment 1.1e.

Comment 3: “Missing information in the manuscript…”

3a: “Images of stimulation anatomical locations for all subjects included in this study. Ideally information about the impedance of the contacts to be able to calculate the actual current used.”

As requested, we have provided an image from the coronal T1 MRI sequence, which highlights the position of the stimulated contacts for each of the 16 patients. Though we did not measure the impedances directly, the stimulation was current-controlled, which ensured that the desired current and charge density were consistent regardless of the tissue or electrode impedance.

Change to Text:

Created Figure S1 (Anatomical location of stimulated electrodes).

Caption: A coronal slice from the T1-weighted MRI scan is shown for each patient who participated in the study (n = 16). Electrode contacts within the same plane of the image are shown with blue circles, and the bipolar pair of stimulated contacts within the basolateral amygdala is highlighted in red.

Lines 144-145 have been edited to reflect that the delivered stimulation was current-controlled: “Specifically, we administered current-controlled, charge-balanced, …”

3b: “The studied population is epilepsy patients, and the manuscript lacks description of their condition, proximity to electrodes included in the study to pathological areas, and the number of units from each patient/hemisphere.”

We agree that additional information regarding patient demographics, experimental details, and clinical characteristics would further contextualize this unique patient population. A new table has been included, which contains the following information: patient ID, sex, age, # experimental session, # SEEG leads (and # microelectrodes), # detected units (L vs. R hemisphere), and suspected seizure onset zone.

Change to Text:

Created Table S1 (Patient demographics and clinical characteristics).

Lines 258-259 have been added: “…(see Table S1 for patient demographics).”

3c: “I haven't seen any comments on code availability (calculating modulation indices and statistics) and data sharing.”

For clarification, a section titled Resource Availability is already appended to the end of the manuscript following the Conclusion, which describes the data and code availability.

Change to Text:

None

3d: “Small comment - Figure legend 3E - Define gray markers (non-modulated units?)”

Thank you for highlighting this omission. We have updated the relevant figure caption.

Change to Text:

The following has been added to the Figure 3 caption: “…whereas units without a significant change in activity are shown in grey.”

Reviewer #2 (Public review):

Summary:

Soham Mukhopadhyay et al. investigated the protein folding of the secretome from gall-forming microbes using the AI-based structure-modeling tool AlphaFold2. Their study analyzed six gall-forming species, including two Plasmodiophorid species and four others spanning different kingdoms, along with one non-gall-forming Plasmodiophorid species, Polymyxa betae. The authors found no effector fold specifically conserved among gall-forming pathogens, leading to the conclusion that their virulence strategies are likely achieved through diverse mechanisms. However, they identified an expansion of the Ankyrin repeat family in two gall-forming Plasmodiophorid species, with a less pronounced presence in the non-gall-forming Polymyxa betae. Additionally, the study revealed that known effectors such as CCG and AvrSen1 belong to sequence-unrelated but structurally similar (SUSS) effector clusters.

Strengths:

(1) The bioinformatics analyses presented in this study are robust, and the AlphaFold2-derived resources deposited in Zenodo provide valuable resources for researchers studying plant-microbe interactions. The manuscript is also logically organized and easy to follow.

(2) The inclusion of the non-gall-forming Polymyxa betae strengthens the conclusion that no effector fold is specifically conserved in gall-forming pathogens and highlights the specific expansion of the Ankyrin repeat family in gall-forming Plasmodiophorids.

(3) Figure 4a and 4b effectively illustrate the SUSS effector clusters, providing a clear visual representation of this finding.

(4) Figure 1 is a well-designed, comprehensive summary of the number and functional annotations of putative secretomes in gall-forming pathogens. Notably, it reveals that more than half of the analyzed effectors lack known protein domains in some pathogens, yet some were annotated based on their predicted structures, despite the absence of domain annotations.

Weaknesses:

(1) The effector families discussed in this paper remain hypothetical in terms of their functional roles, which is understandable given the challenges of demonstrating their functions experimentally. However, this highlights the need for experimental validation as a next step.

Authors' response: Thank you. Yes, there is a lot of work to do in the coming years.

Reviewer's response: Incorporating experimental validation substantially strengthened the manuscript. Did you try the AlphaFold-Multimer prediction of the interaction between PBTT_00818 and the GroES-like protein? Does the model indicate a high-confidence interface?

(2) Some analyses, such as those in Figure 4e, emphasize motifs derived from sequence alignments of SUSS effector clusters. Since these effectors are sequence-unrelated, sequence alignments might be unreliable. It would be more rigorous to perform structure-based alignments in addition to sequence-based ones for motif confirmation. For instance, methods described in Figure 3E of de Guillen et al. (2015, https://doi.org/10.1371/journal.ppat.1005228) or tools like Foldseek could be useful for aligning structures of multiple sequences.

Authors' response: In Fig. 4e, we highlight the conserved cysteine residues. While there is no clearly conserved overall motif, the figure illustrates that despite the high sequence divergence, the key cysteines involved in disulfide-bridge formation are consistently conserved across the sequences.

Reviewer's response: Understood. Nevertheless, if a reliable sequence alignment can indeed be generated, I would interpret this to mean that the CCG effectors constitute a highly diversified family rather than being truly sequence unrelated. By comparison, members of the MAX effector family share a common fold, yet their sequences are so divergent that sequence alignment is impossible.

(3) When presenting AlphaFold-generated structures, it is essential to include confidence scores such as pLDDT and PAE. For example, in Figure 1D of Derbyshire and Raffaele (2023, https://doi.org/10.1038/s41467-023-40949-9), the structural representations were colored red due to their high pLDDT scores, emphasizing their reliability.

Authors' response: Thank you for the observation. Due to the restrictive parameters used in our analysis, over 90 % of the structure would appear red. For this reason, we chose not to include the color scale, as it would not provide additional informative value in this context.

Reviewer's response: Understood.

Author response:

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

Reviewer #1 (Public review):

Summary:

This manuscript presents a comprehensive structure-guided secretome analysis of gall-forming microbes, providing valuable insights into effector diversity and evolution. The authors have employed AlphaFold2 to predict the 3D structures of the secretome from selected pathogens and conducted a thorough comparative analysis to elucidate commonalities and unique features of effectors among these phytopathogens.

Strengths:

The discovery of conserved motifs such as 'CCG' and 'RAYH' and their central role in maintaining the overall fold is an insightful finding. Additionally, the discovery of a nucleoside hydrolase-like fold conserved among various gall-forming microbes is interesting.

Weaknesses:

Important conclusions are not verified by experiments.

Thank you very much. There are many aspects of this study that could be further validated, each potentially requiring years of work. Therefore, we chose to focus on two specific hypotheses: are AlphaFol-Multimer predictions accurate? Can ANK target more than one host protein? Particularly, we focused on the identification of putative targets for one of the ankyrin repeat proteins, PBTT_00818 (Fig. 6). Using one-by-one yeast two-hybrid (Y2H) assays, we tested the AlphaFold-Multimer prediction of an interaction between PBTT_00818 and MPK3. The interaction did not occur in yeast, suggesting it might not take place under those conditions.

This negative result led us to perform a Y2H screen using an Arabidopsis cDNA library, which identified a GroES-like protein, highly expressed in roots, as a potential target of the ANK effector. Surprisingly, both the PBTT_00818–MPK3 and PBTT_00818–GroES-like protein interactions were later confirmed in planta using BiFC assays. These findings suggest two key points: (1) AlphaFold predictions can be accurate for ANK proteins, and (2) ANK domains, known for mediating protein-protein interactions, may enable these effectors to target multiple host proteins.

Although the precise biological implications remain unclear, it is possible that ANK proteins act as scaffolds or adaptors for other effectors during infection. The validations presented here open exciting avenues for further research into the role of ANK proteins in Plasmodiophorid pathogenesis and gall formation. This is presented in the corrected preprint and Fig. 7, Table S12, Fig. S7-S8.

Reviewer #2 (Public review):

Summary:

Soham Mukhopadhyay et al. investigated the protein folding of the secretome from gall-forming microbes using the AI-based structure modeling tool AlphaFold2. Their study analyzed six gall-forming species, including two Plasmodiophorid species and four others spanning different kingdoms, along with one non-gall-forming Plasmodiophorid species, Polymyxa betae. The authors found no effector fold specifically conserved among gall-forming pathogens, leading to the conclusion that their virulence strategies are likely achieved through diverse mechanisms. However, they identified an expansion of the Ankyrin repeat family in two gall-forming Plasmodiophorid species, with a less pronounced presence in the non-gall-forming Polymyxa betae. Additionally, the study revealed that known effectors such as CCG and AvrSen1 belong to sequence-unrelated but structurally similar (SUSS) effector clusters.

Strengths:

(1) The bioinformatics analyses presented in this study are robust, and the AlphaFold2-derived resources deposited in Zenodo provide valuable resources for researchers studying plant-microbe interactions. The manuscript is also logically organized and easy to follow.

(2) The inclusion of the non-gall-forming Polymyxa betae strengthens the conclusion that no effector fold is specifically conserved in gall-forming pathogens and highlights the specific expansion of the Ankyrin repeat family in gall-forming Plasmodiophorids.

(3) Figure 4a and 4b effectively illustrate the SUSS effector clusters, providing a clear visual representation of this finding.

(4) Figure 1 is a well-designed, comprehensive summary of the number and functional annotations of putative secretomes in gall-forming pathogens. Notably, it reveals that more than half of the analyzed effectors lack known protein domains in some pathogens, yet some were annotated based on their predicted structures, despite the absence of domain annotations.

Weaknesses:

(1) The effector families discussed in this paper remain hypothetical in terms of their functional roles, which is understandable given the challenges of demonstrating their functions experimentally. However, this highlights the need for experimental validation as a next step.

Thank you. Yes, there is a lot of work to do in the coming years.

(2) Some analyses, such as those in Figure 4e, emphasize motifs derived from sequence alignments of SUSS effector clusters. Since these effectors are sequence-unrelated, sequence alignments might be unreliable. It would be more rigorous to perform structure-based alignments in addition to sequence-based ones for motif confirmation. For instance, methods described in Figure 3E of de Guillen et al. (2015, https://doi.org/10.1371/journal.ppat.1005228) or tools like Foldseek could be useful for aligning structures of multiple sequences.

In Fig. 4e, we highlight the conserved cysteine residues. While there is no clearly conserved overall motif, the figure illustrates that despite the high sequence divergence, the key cysteines involved in disulfide bridge formation are consistently conserved across the sequences.

(3) When presenting AlphaFold-generated structures, it is essential to include confidence scores such as pLDDT and PAE. For example, in Figure 1D of Derbyshire and Raffaele (2023, https://doi.org/10.1038/s41467-023-40949-9), the structural representations were colored red due to their high pLDDT scores, emphasizing their reliability.

Thank you for the observation. Due to the restrictive parameters used in our analysis, over 90% of the structure would appear red. For this reason, we chose not to include the color scale, as it would not provide additional informative value in this context.

Reviewer #1 (Recommendations for the authors):

Experimental validation of the significance of 'CCG' and 'RAYH' motifs would further strengthen this study.

Regarding the Mig1-like protein in Ustilago maydis, the presence of four conserved cysteine residues that are pivotal for maintaining the stability of its folded structure raises an intriguing question. Specifically, while many Mig cluster effectors contain four cysteine residues that form two conserved disulfide bridges, this structure is notably absent in the Mig protein itself. The author has speculated that these four cysteine residues form two conserved disulfide bonds, which are crucial for the stability of Mig protein folding. However, this hypothesis remains unvalidated. To test this prediction, it would be prudent to simulate mutations in the cysteine residues corresponding to the disulfide bonds in Mig and employ molecular dynamics simulations to assess the stability of folding before and after the mutation.

Mig-1 does contain the four conserved cysteine residues responsible for forming disulfide bridges. However, due to the high divergence among Mig-1-like sequences, the alignment software was unable to properly align all the cysteine residues. As a result, Mig-1 may appear to lack these conserved cysteines in the alignment, although they are indeed present upon individual inspection. This is an area that research groups working with U. maidis as a model could explore further to expand our understanding of this effector family.

Could you please clarify why talking about Ankyrins and LRR in Arabidopsis thaliana (line 252)? Additionally, what are the structural and functional differences between the LRR sequences of P. brassicae and those of the host plants?

This sentence refers to the identification of the ANK motif in P. brassicae and S. spongospora, not in Arabidopsis thaliana. While the hydrophobic core of the ANK domains appears conserved between the host and the pathogen, the surface residues are highly polymorphic.

The evidence supporting the interaction between the ANK effector and Arabidopsis immunity-related proteins, as validated using AlphaFold-Multimer, is currently limited. To enhance the reliability of these data, it is advisable for the author to select several pairs of proteins predicted to interact for further experimental verification.

We conducted a large-scale yeast two-hybrid (Y2H) screen using the ANK domain effector PBTT_00818, which was selected due to its high iPTM+pTM score. The Y2H interactions were subsequently validated through BiFC assays. Our results show that PBTT_00818 interacts with Arabidopsis MPK3 in the nucleus, consistent with predictions from the AlphaFold2-multimer model. In addition, PBTT_00818 was also found to target AT3G56460, a GroES-like zinc-binding alcohol dehydrogenase, also localized in the nucleus.

While the manuscript is well-composed, certain sections could be enhanced for clarity and readability. For example, the discussion section could be expanded to include a more in-depth analysis of the implications of the findings for understanding the virulence mechanisms of gall-forming microbes. Additionally, a comparison of the findings with previous studies on related pathogens would provide a more comprehensive perspective.

Certain sections of the discussion have been expanded. However, we chose to focus on the novel aspects of the study and to avoid comparisons with other plant pathogens, as those mechanisms are already well known and extensively studied. Studies using AlphaFold in plant pathology are also limited.

*Reviewer #2 (Recommendations for the authors):*

The results of clustering analyses are highly dependent on the chosen thresholds. Given that the authors provide clear and well-designed visualizations of SUSS effectors in Figures 4a and 4b, applying the same presentation methods to Figures 5a and 5b could make these analyses more convincing.

We were able to generate the all-vs-all matrix for Figures 4a and 4b because it involved only 13 proteins. However, Figure 5b includes over 40 effectors, making it impractical to visualize the data in the same way. Instead, we presented the sequence-based clusters as nodes and connected them based on structural similarity.

dweb.link@ http://bafybeihda4gloeygr5moflptlfedkbhkuysutw7ulbomreqfx6fywro4xa.ipfs.localhost:8080/?filename=%EF%BC%82display%20metaphor%20scripting%20language%EF%BC%82%20gyuri%20lajos%20dime%20-%20Brave%20Search%20(8_12_2025%209%EF%BC%9A35%EF%BC%9A01%20AM).html

for = wikify myself

CSP vs Actor model for concurrency -

my comment

my comment on this

1. Being chill with Robert's teenager-like ways
2. Cuddling Alan and kissing his cheeks like Grandma Tracy would
3. Applying cooling cream on Alan's back
4. Cuddling up under a soft blanket
5. Lady Penelope sleeping in a treetop nest with me in her arms
6. Being away from the loud, raucous ways of Scott, Virgil and Gordon
7. Being surrounded with ylang ylang and lotus flowers
8. Lying on the sofa with Lady Penelope beside me
9. Sleepy days in
10. Being allowed to feel small and weak

1. Tin Tin being the peaceful, animal-loving daughter of Eve
2. Floating on Lady Penelope's smooth, soft body in the river
3. Endless cuddles from Alan when I snuggle into his bed
4. Floating on a soft, fluffy cloud
5. Modelling a cloud into a teddy bear to cuddle
6. Lady Penelope being a woman of nature
7. Tin Tin stroking my hair and whispering reassuring words to me
8. The cosiness and warmth of being with friends
9. Lady Penelope gently mollycoddling me
10. Gentle foot tickles

Taxa de segurança preventiva relativa a eventos não gratuitos e a emissão de certidões para defesa de direitos - ADI 3.717/PR Relator: Ministro Nunes Marques

RESUMO - É constitucional a instituição de taxa por serviços prestados por órgãos de segurança pública relativos (i) à segurança preventiva em eventos esportivos e de lazer com cobrança de ingresso, bem como (ii) à emissão de certidões e atestados, desde que não se destinem à defesa de direitos ou ao esclarecimento de interesse pessoal (CF/1988, art. 5º, XXXIV, b).

• Conforme jurisprudência desta Corte (1), o serviço de segurança pública e as atividades a ela inerentes, como policiamento ostensivo e vigilância, não podem ser financiados mediante taxas, dada a impossibilidade de que sua prestação ocorra de forma individualizada. Assim, por constituir serviço geral e indivisível, prestado a toda a coletividade, este deve ser remunerado por meio de impostos. Contudo, há situações em que os serviços, apesar de prestados por órgãos de segurança pública, são efetivamente oferecidos de modo específico e divisível. Nesse contexto, prestações oferecidas atipicamente pelos órgãos de segurança pública e que são usufruídas de modo particular pelos administrados podem ser custeadas por meio de taxas (2).

• Na espécie, a operação logística necessária para garantir a segurança em eventos de grande porte, com finalidade lucrativa, não pode ser imputada à sociedade como um todo através de um financiamento indistinto, arrecadado pelo poder público via impostos. Também não é cabível partilhar, entre toda a sociedade, os custos de serviços prestados pelos órgãos da Administração Policial Militar estadual para fornecimento, entre outros, de “cópias (xerox) autenticadas (por folha)”, “diárias/permanência de veículos apreendidos nas unidades policiais militares”, fotografias e inscrição em cursos e exames. Por expressa vedação constitucional (3), a cobrança de taxa não é válida apenas para o fornecimento de certidões e atestados direcionados à defesa de direitos ou ao esclarecimento de interesse pessoal (4). Com base nesses entendimentos, o Plenário, por unanimidade, julgou parcialmente procedente a ação para (i) declarar a inconstitucionalidade dos itens 1.1.1 e 1.2 (1.2.1 a 1.2.5) da tabela anexa à Lei nº 10.236/1992 do Estado do Paraná; e (ii) dar interpretação conforme aos itens 2.1 e 2.3 da mesma lista, no sentido de impossibilitar a cobrança de taxa para emissão de certidões/atestados solicitados com o propósito de defender direitos e esclarecer situações de interesse pessoal.

• Também não é cabível partilhar, entre toda a sociedade, os custos de serviços prestados pelos órgãos da Administração Policial Militar estadual para fornecimento, entre outros, de “cópias (xerox) autenticadas (por folha)”, “diárias/permanência de veículos apreendidos nas unidades policiais militares”, fotografias e inscrição em cursos e exames. Por expressa vedação constitucional (3), a cobrança de taxa não é válida apenas para o fornecimento de certidões e atestados direcionados à defesa de direitos ou ao esclarecimento de interesse pessoal (4). Com base nesses entendimentos, o Plenário, por unanimidade, julgou parcialmente procedente a ação para (i) declarar a inconstitucionalidade dos itens 1.1.1 e 1.2 (1.2.1 a 1.2.5) da tabela anexa à Lei nº 10.236/1992 do Estado do Paraná; e (ii) dar interpretação conforme aos itens 2.1 e 2.3 da mesma lista, no sentido de impossibilitar a cobrança de taxa para emissão de certidões/atestados solicitados com o propósito de defender direitos e esclarecer situações de interesse pessoal.

Reviewer #1 (Public review):

Summary:

The study by Druker et al. shows that siRNA depletion of PHD1, but not PHD2, increases H3T3 phosphorylation in cells arrested in prometaphase. Additionally, the expression of wild-type RepoMan, but not the RepoMan P604A mutant, restored normal H3T3 phosphorylation localization in cells arrested in prometaphase. Furthermore, the study demonstrates that expression of the RepoMan P604A mutant leads to defects in chromosome alignment and segregation, resulting in increased cell death. These data support a role for PHD1-mediated prolyl hydroxylation in controlling progression through mitosis. This occurs, at least in part, by hydroxylating RepoMan at P604, which regulates its interaction with PP2A during chromosome alignment.

Strengths:

The data support most of the conclusions made. However, some issues need to be addressed.

Weaknesses:

(1) Although ectopically expressed PHD1 interacts with ectopically expressed RepoMan, there is no evidence that endogenous PHD1 binds to endogenous RepoMan or that PHD1 directly binds to RepoMan.

(2) There is no genetic evidence indicating that PHD1 controls progression through mitosis by catalyzing the hydroxylation of RepoMan.

(3) Data demonstrating the correlation between dynamic changes in RepoMan hydroxylation and H3T3 phosphorylation throughout the cell cycle are needed.

(4) The authors should provide biochemical evidence of the difference in binding ability between RepoMan WT/PP2A and RepoMan P604A/PP2A.

(5) PHD2 is the primary proline hydroxylase in cells. Why does PHD1, but not PHD2, affect RepoMan hydroxylation and subsequent control of mitotic progression? The authors should discuss this issue further.

Author response:

Public Reviews:

Reviewer #1 (Public review):

We appreciate the reviewer’s agreement that our data, "support most of the conclusions made”.

With respect to Concerns raised by reviewer 1:

(1) Although ectopically expressed PHD1 interacts with ectopically expressed RepoMan, there is no evidence that endogenous PHD1 binds to endogenous RepoMan or that PHD1 directly binds to RepoMan.

We do not fully agree that this comment is accurate - the implication is that we only show interaction between two exogenously expressed proteins, i.e. both exogenous  PHD1 and RepoMan, when in fact we show that tagged PHD1 interacts with endogenous RepoMan. The major technical challenge here is the well known difficulty of detetcing endogenous PHD1 in such cell lines. We agree that co-IP studies do not prove that this interaction is direct and never claim to have shown this, though we do feel that a direct interaction is most likely, albeit not proven.

(2) There is no genetic evidence indicating that PHD1 controls progression through mitosis by catalyzing the hydroxylation of RepoMan.

We agree that our current study is primarily a biochemical and cell biological study, rather than a genetic study. Nonetheless, similar biochemical and cellular approaches have been widely used and validated in previous studies in mechanisms regulating cell cycle progression and we are confident in the conclusions drawn based on the data obtained so far.

(3) Data demonstrating the correlation between dynamic changes in RepoMan hydroxylation and H3T3 phosphorylation throughout the cell cycle are needed.

We agree that it will be very interesting to analyse in more detail the cell cycle dynamics of RepoMan hydroxylation and H3T3 phosphorylation - along with other cell cycle parameters. We view this as outside the scope of our present study and are actively engaged in raising the additional funding needed to pursue such future experiments.

(4) The authors should provide biochemical evidence of the difference in binding ability between RepoMan WT/PP2A and RepoMan P604A/PP2A.

Here again we agree that it will be very interesting to analyse in future the detailed binding interactions between wt and mutant RepoMan and other interacting proteins, including PP2A. We view this as outside the scope of our present study and are actively engaged in raising the additional funding needed to pursue such future experiments.

(5) PHD2 is the primary proline hydroxylase in cells. Why does PHD1, but not PHD2, affect RepoMan hydroxylation and subsequent control of mitotic progression? The authors should discuss this issue further.

We agree with the main point underlining this comment, i.e., that there are still many things to be learned concerning the specific roles and mechanisms of the different PHD enzymes in vivo. We look forward to addressing these questions in future studies.

Reviewer #2 (Public review):

We appreciate the reviewer’s comments that our manuscript uses biochemical and imaging tools to delineate a key mechanism in the regulation of the progression of the cell cycle and their appreciation that our experiments performed are, 'conclusive with well-designed controls.'

With respect to the specific Concern raised by reviewer 2:

Lack of in vitro reconstitution and binding data.

We agree that it will be very interesting to pursue in vitro reconstitution studies and detailed binding data. We view this as outside the scope of our present study and are actively engaged in raising the additional funding needed to pursue such future experiments.

Reviewer #3 (Public review):

We appreciate the reviewer’s comments that our study, “is a comprehensive molecular and cell biological characterisation of the effects of P604 hydroxylation by PHD1 on RepoMan, a regulatory subunit of the PPIgamma complex” and their conclusion that, “we should have no question about the validity of the PHD1-mediated hydroxylation”.

With respect to the specific Concern raised by reviewer 3:

Reliance on a Proline-Alanine mutation in RepoMan to mimic an unhydroxylatable protein. The mutation will introduce structural alterations, and inhibition or knockdown of PHD1 would be necessary to strengthen the data on how hydroxylates regulate chromatin loading and interactions with B56/PP2A.

We do not agree that we rely solely on analysis of the single site pro-ala mutatin in RepoMan for our conclusions, since we also present a raft of additional experimental evidence, including knock-down data and experiments using both fumarate and FG. We would also reference the data we present on RepoMan in the parallel study by Jiang et al, which has also been reviewed by eLife and is currently available on biorxiv (doi: https://doi.org/10.1101/2025.05.06.652400). Of course we agree with the reviewer that even although the muatnt RepoMan features only a single amino acid change, this could still result in undetermined structural effects on the RepoMan protein that could conceivably contribute, at least in part, to some of the phenotypic effects observed. Hopefully future studies will help to clarify this.

Author response:

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

Reviewer #1 (Public review):

When you search for something, you need to maintain some representation (a "template") of that target in your mind/brain. Otherwise, how would you know what you were looking for? If your phone is in a shocking pink case, you can guide your attention to pink things based on a target template that includes the attribute 'pink'. That guidance should get you to the phone pretty effectively if it is in view. Most real-world searches are more complicated. If you are looking for the toaster, you will make use of your knowledge of where toasters can be. Thus, if you are asked to find a toaster, you might first activate a template of a kitchen or a kitchen counter. You might worry about pulling up the toaster template only after you are reasonably sure you have restricted your attention to a sensible part of the scene.

Zhou and Geng are looking for evidence of this early stage of guidance by information about the surrounding scene in a search task. They train Os to associate four faces with four places. Then, with Os in the scanner, they show one face - the target for a subsequent search. After an 8 sec delay, they show a search display where the face is placed on the associated scene 75% of the time. Thus, attending to the associated scene is a good idea. The questions of interest are "When can the experimenters decode which face Os saw from fMRI recording?" "When can the experimenters decode the associated scene?" and "Where in the brain can the experimenters see evidence of this decoding? The answer is that the face but not the scene can be read out during the face's initial presentation. The key finding is that the scene can be read out (imperfectly but above chance) during the subsequent delay when Os are looking at just a fixation point. Apparently, seeing the face conjures up the scene in the mind's eye.

This is a solid and believable result. The only issue, for me, is whether it is telling us anything specifically about search. Suppose you trained Os on the face-scene pairing but never did anything connected to the search. If you presented the face, would you not see evidence of recall of the associated scene? Maybe you would see the activation of the scene in different areas and you could identify some areas as search specific. I don't think anything like that was discussed here.

You might also expect this result to be asymmetric. The idea is that the big scene gives the search information about the little face. The face should activate the larger useful scene more than the scene should activate the more incidental face, if the task was reversed. That might be true if the finding is related to a search where the scene context is presumed to be the useful attention guiding stimulus. You might not expect an asymmetry if Os were just learning an association.

It is clear in this study that the face and the scene have been associated and that this can be seen in the fMRI data. It is also clear that a valid scene background speeds the behavioral response in the search task. The linkage between these two results is not entirely clear but perhaps future research will shed more light.

It is also possible that I missed the clear evidence of the search-specific nature of the activation by the scene during the delay period. If so, I apologize and suggest that the point be underlined for readers like me.

We have added text related to this issue, particularly in the discussion (page 19, line 6), and have also added citations of studies in humans and non-human primates showing a causal relationship between preparatory activity in prefrontal and visual cortex and visual search performance (page 6, line 16).

Reviewer #2 (Public review):

Summary:

This work is one of the best instances of a well-controlled experiment and theoretically impactful findings within the literature on templates guiding attentional selection. I am a fan of the work that comes out of this lab and this particular manuscript is an excellent example as to why that is the case. Here, the authors use fMRI (employing MVPA) to test whether during the preparatory search period, a search template is invoked within the corresponding sensory regions, in the absence of physical stimulation. By associating faces with scenes, a strong association was created between two types of stimuli that recruit very specific neural processing regions - FFA for faces and PPA for scenes. The critical results showed that scene information that was associated with a particular cue could be decoded from PPA during the delay period. This result strongly supports the invoking of a very specific attentional template.

Strengths:

There is so much to be impressed with in this report. The writing of the manuscript is incredibly clear. The experimental design is clever and innovative. The analysis is sophisticated and also innovative. The results are solid and convincing.

Weaknesses:

I only have a few weaknesses to point out.<br /> This point is not so much of a weakness, but a further test of the hypothesis put forward by the authors. The delay period was long - 8 seconds. It would be interesting to split the delay period into the first 4seconds and the last 4seconds and run the same decoding analyses. The hypothesis here is that semantic associations take time to evolve, and it would be great to show that decoding gets stronger in the second delay period as opposed to the period right after the cue. I don't think this is necessary for publication, but I think it would be a stronger test of the template hypothesis.

We conducted the suggested analysis, and we did not find clear evidence of differences in decoding scene information between the earlier and later portions of the delay period. This may be due to insufficient power when the data are divided, individual differences in when preparatory activation is the strongest, or truly no difference in activation over the delay period. More details of this analysis can be found in the supplementary materials (page 12, line 16; Figure S1).

Type in the abstract "curing" vs "during."

Fixed.

It is hard to know what to do with significant results in ROIs that are not motivated by specific hypotheses. However, for Figure 3, what are the explanations for ROIs that show significant differences above and beyond the direct hypotheses set out by the authors?

We added reasoning for the other a priori ROIs in the introduction (page 4, line 26). There is substantial evidence suggesting that frontoparietal areas are involved in cognitive control, attentional control, and working memory. The ROIs we selected from frontal and parietal cortex are based on parcels within resting state networks defined by the s17-network atlases (Schaefer et al., 2018). The IFJ was defined by the HCP-MMP1 (Glasser et al., 2016). These regions are commonly used in studies of attention and cognitive control, and the exact ROIs selected are described in the section on “Regions of interest (ROI) definition”. While we have the strongest hypothesis for IFJ based on relatively recent work from the Desimone lab, the other ROIs in lateral frontal cortex and parietal cortex, are also well documented in similar studies, although the exact computation being done by these regions during tasks can be hard to differentiate with fMRI.\

Reviewer #3 (Public review):

The manuscript contains a carefully designed fMRI study, using MVPA pattern analysis to investigate which high-level associate cortices contain target-related information to guide visual search. A special focus is hereby on so-called 'target-associated' information, that has previously been shown to help in guiding attention during visual search. For this purpose the author trained their participants and made them learn specific target-associations, in order to then test which brain regions may contain neural representations of those learnt associations. They found that at least some of the associations tested were encoded in prefrontal cortex during the cue and delay period.

The manuscript is very carefully prepared. As far as I can see, the statistical analyses are all sound and the results integrate well with previous findings.

I have no strong objections against the presented results and their interpretation.

Reviewer #1 (Recommendations for the authors):

One bit of trivia. In the abstract, you should define IFJ on its first appearance in the text. You get to that a bit later.

Fixed.

Reviewer #2 (Recommendations for the authors):

I really don't have much to suggest, as I thought that this was a clearly written report that offered a clever paradigm and data that supported the conclusions. My only suggestion would be to split the delay period activity and test whether the strength of the template evolves over time. Even though fMRI is not the best tool for this, still you would predict stronger decoding in the second half of the delay period

Please see above for our response to the same comment.

Reviewer #3 (Recommendations for the authors):

I would just like to point out some minor aspects that might be worth improving before publishing this work.

Abstract: While in general, the writing is clear and concise, I felt that the abstract of the manuscript was particularly hard to follow, probably because the authors at some point re-arranged individual sentences. For example, they write in line 12 about 'the preparatory period', but explain only in the following sentence that the preparatory period ensues 'before search begins'. This made it a bit hard to follow the overall logic and I think could easily be fixed. 

We have addressed this comment and updated the abstract.

Also in the abstract: 'The CONTENTS of the template typically CONTAIN...' sounds weird, no? Also, 'information is used to modulate sensory processing in preparation for guiding attention during search' sounds like a very over-complicated description of attentional facilitation. I'm not convinced either whether the sequence is correct here. Is the information really used to (first) modulate sensory processing (which is a sort of definition of attention in itself) to (then) prepare the guidance of attention in visual search?

We have addressed this comment and updated the abstract.

The sentence in line 7, 'However, many behavioral studies have shown that target-associated information is used to guide attention,...' (and the following sentence) assumes that the reader is somewhat familiar with the term 'target-associations'. I'm afraid that, for a naive reader, this term may only become fully understandable once the idea is introduced a bit later when mentioning that participants of the study were trained on face-scene pairings. I think it could help to give some very short explanation of 'target-associations' already when it is first mentioned. The term 'statistically co-occurring object pairs', for example, could be of great help here.

Thank you for the suggestion. We have added it to the abstract.

page 2, line 22: 'prefrotnal'

Fixed.

page 2, line 24/25: 'information ... can SUPPLANT (?) ... information'. (That's also a somewhat unfortunate repetition of 'information')

Fixed.

page 4, line 23-25: 'Working memory representations in lateral prefrontal and parietal regions are engaged in cognitive control computations that ARE (?) task non-specific but essential to their functioning'

Fixed.

page 7, line 1: maybe a comma before 'suggesting'?

Fixed.

page 7, line 14-16: Something seems wrong with this sentence: 'The distractor face was a race-gender match, which we previously FOUND MADE (?) target discrimination difficult enough to make the scene useful for guiding attention'

We have addressed this comment and rewritten this part (now on page 7, line 18).

Results / Discussion sections:

In several figures, like in Fig3A, the three different IFJ regions, are grouped separately from the other frontal areas, which makes sense given the special role IFJ plays for representing task-related templates. However, IFJ is still part of PFC. I think it would be more correct to group the other frontal areas (like FEF vLPFC etc.) as 'Other Frontal' or even 'Other PFC'.

We have made the changes based on the reviewer’s suggestion.

In some of the Figures, e.g. Fig 3 and 5, I had the impression that the activation patterns of some conditions in vLPFC were rather close to the location of IFJ, which is just a bit posterior. I think I remember that functional localisers of IFJ can actually vary quite a bit in localisation (see e.g. in the Baldauf/Desimone paper). Also, I think it has been shown in the context of other regions, like the human FEF that its position when defined by localisation tasks is not always nicely and fully congruent with the respective labels in an atlas like the Glasser atlas. It might help to take this in consideration when discussing the results, particularly since the term vLPFC is a rather vague collection of several brain parcels and not a parcel name in the Glasser atlas. Some people might even argue that vLPFC in the broad sense contains IFJ, similar to how 'Frontal' contains IFJ (see above). How strong of a point do the authors want to make about activation in IFJ versus in vlPFC?

We have now added text discussing the inability to truly differentiate between subregions of IFJ and other parts of vLPFC in the methods section on ROIs (page 25, line 13) and in the discussion (page 18, line 25). However, one might think that it is even more surprising given the likely imprecision of ROI boundaries that we see distinct patterns between the subregions of IFG defined by Glasser HCP-MMP1 and the other vLPFC regions defined by the 17-network atlases. We do not wish to overstate the precision of IFJ regions, but note the ROI results within the context of the larger literature. We are sure that our findings will have to be reinterpreted when newer methods allow for better localization of functional subregions of the vLPFC in individuals.

Given that the authors nicely explain in the introduction how important templates are in visual search, and given that FEF has such an important role in serially guiding saccades through visual search templates, I think it would be worth discussing the finding that FEF did not hold representation of these targets. Of course, this could be in part due to the specific task at hand, but it may still be interesting to note in the Discussion section that here FEF, although important for some top-down attention signals, did not keep representations of the 'search' templates. Is it because there is no spatial component to the task at hand (like proposed in Bedini 2021)?

We have now added text directly addressing this point and citing the Bedini et al. paper in the discussion (page 18, line 18). Besides our current findings, the relationship between IFJ and FEF is really interesting and will hopefully be investigated more in the future.

Page 18, line 5: 'we the(N) associated...'

Fixed.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Manuscript number: RC-2025-02879 Corresponding author(s): Matteo Allegretti; Alia dos Santos

1. General Statements

In this study, we investigated the effects of paclitaxel on both healthy and cancerous cells, focusing on alterations in nuclear architecture. Our novel findings show that:

• Paclitaxel-induced microtubule reorganisation during interphase alters the perinuclear distribution of actin and vimentin. The formation of extensive microtubule bundles, in paclitaxel or following GFP-Tau overexpression, coincides with nuclear shape deformation, loss of regulation of nuclear envelope spacing, and alteration of the nuclear lamina.

• Paclitaxel treatment reduces Lamin A/C protein levels via a SUN2-dependent mechanism. SUN2, which links the lamina to the cytoskeleton, undergoes ubiquitination and consequent degradation following paclitaxel exposure.

• Lamin A/C expression, frequently dysregulated in cancer cells, is a key determinant of cellular sensitivity to, and recovery from, paclitaxel treatment.

Collectively, our data support a model in which paclitaxel disrupts nuclear architecture through two mechanisms: (i) aberrant nuclear-cytoskeletal coupling during interphase, and (ii) multimicronucleation following defective mitotic exit. This represents an additional mode of action for paclitaxel beyond its well-established mechanism of mitotic arrest.

We thank the reviewers for their time and constructive feedback. We have carefully considered all comments and have carried out a full revision. The updated manuscript now includes additional data showing:

• Overexpression of microtubule-associated protein Tau causes similar nuclear aberration phenotypes to paclitaxel. This supports our hypothesis that increased microtubule bundling directly leads to nuclear disruption in paclitaxel during interphase.

• Paclitaxel's effects on nuclear shape and Lamin A/C and SUN2 expression levels occur independently of cell division.

• Reduced levels of Lamin A/C and SUN2 upon paclitaxel treatment occur at the protein level via ubiquitination of SUN2.

• The effects of paclitaxel on the nucleus are conserved in breast cancer cells.

Full Revision

We have also edited our text and added further detail to clarify points raised by the reviewers. We believe that our revised manuscript is overall more complete, solid and compelling thanks to the reviewers' comments.

1. Point-by-point description of the revisions

Reviewer #1 Evidence, reproducibility and clarity

This description of the down-regulation of the expression of lamin A/C upon treatment with paclitaxel and its sensitivity to SUN2 is quite interesting but still somehow preliminary. It is unclear whether this effect involves the regulation of gene expression, or of the stability of the proteins. How SUN2 mediates this effect is still unknown.

We thank the reviewer for this valuable comment. To elucidate the mechanism behind the decrease in Lamin A/C and SUN2 levels, we have now performed several additional experiments. First, we performed RT-qPCR to quantify mRNA levels of these genes, relative to the housekeeping gene GAPDH (Supplementary Figure 3B and O). The levels of SUN2 and LMNA mRNA remained the same between control and paclitaxel-treated cells, indicating that this effect instead occurs at the protein level. We have also tested post-translational modifications as a potential regulatory mechanism for Lamin A/C and SUN2. In addition to the phosphorylation of Ser404 which we had already tested (Supplementary Figure 3C), we have now included additional Phos-tag gel and Western blotting data showing that the overall phosphorylation status of Lamin A/C is not affected by paclitaxel (Supplementary Figure 3E and F). We also pulled-down Lamin A/C from cell lysates and then Western blotted for polyubiquitin and acetyl-lysine, which showed that the ubiquitination and acetylation states of Lamin A/C are also not affected by paclitaxel (Supplementary Figure 3G-I). However, Western blots for polyubiquitin of SUN2 pulled down from cell lysates showed that paclitaxel treatment results in significant SUN2 ubiquitination (Figure 3M and N). Therefore, we propose that the downregulation of SUN2 following paclitaxel treatment occurs by ubiquitin-mediated proteolysis.

The roles of free tubulins and polymerized microtubules, and thus the potential role of paclitaxel, need to be uncovered.

We addressed this important point by using an alternative method to stabilise/bundle microtubules in interphase, namely by overexpressing GFP-Tau, as suggested by reviewer 2. Following GFP- Tau overexpression, large microtubule bundles were observed throughout the cytoplasm (Figure 4A), and this resulted in a significant decrease in nuclear solidity (Figure 4B). Furthermore, in cells where microtubule bundles extensively contacted the nucleus, the nuclear lamina became unevenly distributed and appeared patchy (Figure 4C). This supports our hypothesis that the aberrations to nuclear shape and Lamin A/C localisation in paclitaxel-treated cells are due to the presence of microtubules bundles surrounding the nucleus.

The doses of paclitaxel at which occur the effects described in the paper are not fully consistent with all the conclusions. Most experiments have been done at 5 nM. However, at this dose the effect of lamin A/C over or down expression on the growth (differences in the slopes of the curves in Figure 4A) are not fully convincing and not fully consistent with the clear effect on viability as well (in addition, duration of treatments before assessing vialbility are not specified). At 1 nM, cell growth is reduced and the rescuing effect of lamin over-expression is much more clear (Fig 4A), and the nucleus deformation clear (Fig 2A) but this dose has no effect on lamin A/C expression (Fig 3C), which questions how lamins impact nucleus shape and cell survival. Cytoskeleton reorganisation in these conditions is not described although it could clarify the respective role of force production (suggested in figure 1) and nuclei resistance (shown in figure 2) in paclitaxel sensitivity.

We thank the reviewer for raising this important point. We have addressed this by conducting additional repeats for the cell confluency measurements to increase the statistical power of our experiments (Figure 5A). Our data now show that GFP-lamin A/C had a statistically significant effect on rescuing cell growth at both 1 nM and 5 nM paclitaxel, while Lamin A/C knockdown exacerbated the inhibition of cell growth at 5 nM paclitaxel but not 1 nM paclitaxel (Figure 5A). In addition, we note that the duration of paclitaxel treatment before assessing viability was specified in the figure legend: "Bar graph comparing cell viability between wild-type (red), GFP-Lamin A/C overexpression (green), and Lamin A/C knockdown (blue) cells following 20 h incubation in 0, 1, 5, or 10 nM paclitaxel." We also repeated cell viability analysis after 48 h incubation in paclitaxel instead of 20 h to allow for a longer time for differences to take effect (Figure 5B).

We also added figures showing the cytoskeletal reorganisation at both 1 and 10 nM in addition to 0 and 5 nM (Supplementary Figure 1A) showing that microtubule bundling and condensation of actin into puncta correlated with increased paclitaxel concentration. Vimentin colocalised well with microtubules at all concentrations.

We have also included in our results section further clarification for the use of 5nM paclitaxel in this study. The new section reads as follows: "Experiments were performed at 5 nM paclitaxel (with additional experiments to determine dose relationships at 1 and 10 nM) because this aligns with previous studies7,14,24. Furthermore, previous analysis of patient plasma reveals that typical concentrations are within the low nanomolar range8, and concentrations of 5-10 nM are required in cell culture to reach the same intracellular concentrations observed in vivo in patient tumours9. This aligns with in vitro cytotoxic studies of paclitaxel in eight human tumour cell lines which show that paclitaxel's IC50 ranges between 2.5 and 7.5 nM41."

Finally, although the absence of role of mitotic arrest is clear from the data, the defective reorganisation of the nucleus after mitosis still suggest that the effect of paclitaxel is not independent of mitosis.

We thank the reviewer for pointing out the need for clarification in the wording of our manuscript. We have reworded the title and relevant sections of our abstract, introduction, and discussion to make it clearer that the effects of paclitaxel on the nucleus are due to a combination of aberrant nuclear cytoskeletal coupling during interphase and multimicronucleation following mitotic slippage. We have also added additional data in support of the effect of paclitaxel on nuclear architecture during interphase. For this, we used serum-starved cells (which divide only very slowly such that the majority of cells do not pass through mitosis during the 16 h incubation in paclitaxel [Supplementary Figure 2D]). Our new data confirmed that paclitaxel's effects on nuclear solidity, and Lamin A/C and SUN2 proteins levels can occur independently of cell division (Figure 2C; Figure 3H-J). Finally, when we overexpressed GFP-Tau (as discussed above) we observed similar aberrations to nuclear solidity and Lamin A/C localisation. This indicates that these effects occur due to microtubule bundling in interphase, especially as in our study GFP-Tau did not lead to multimicronucleation or appear to affect mitosis (Figure 4).

Below are the main changes to the text regarding the interphase effect of paclitaxel:

• Title: "Paclitaxel compromises nuclear integrity in interphase through SUN2-mediated cytoskeletal coupling"

• Abstract: "Overall, our data supports nuclear architecture disruption, caused by both aberrant nuclear-cytoskeletal coupling during interphase and exit from defective mitosis, as an additional mechanism for paclitaxel beyond mitotic arrest."

• Introduction: "Here we propose that cancer cells have increased vulnerability to paclitaxel both during interphase and following aberrant mitosis due to pre-existing defects in their NE and nuclear lamina."

• Discussion: "Overall, our work builds on previous studies investigating loss of nuclear integrity as an anti-cancer mechanism of paclitaxel separate from mitotic arrest14,20,21. We propose that cancer cells show increased sensitivity to nuclear deformation induced by aberrant nuclear-cytoskeletal coupling and multimicronucleation following mitotic slippage. Therefore, we conclude that paclitaxel functions in interphase as well as mitosis, elucidating how slowly growing tumours are targeted."

minor: a more thorough introduction of known data about dose response of cells in culture and in vivo would help understanding the range of concentrations used in this study.

As mentioned above, we have now included additional information in our Results section to clarify our paclitaxel dose range: "Experiments were performed at 5 nM paclitaxel (with additional experiments to determine dose relationships at 1 and 10 nM) because this aligns with previous studies7,14,24. Furthermore, previous analysis of patient plasma reveals that typical concentrations are within the low nanomolar range8, and concentrations of 5-10 nM are required in cell culture to reach the same intracellular concentrations observed in vivo in patient tumours9. This aligns with in vitro cytotoxic studies of paclitaxel in eight human tumour cell lines which show that paclitaxel's IC50 ranges between 2.5 and 7.5 nM41."

Significance

In this manuscript, Hale and colleagues describe the effect of paclitaxel on nucleus deformation and cell survival. They showed that 5nM of paclitaxel induces nucleus fragmentation, cytoskeleton reorganisation, reduced expression of LaminA/C and SUN2, and reduced cell growth and viability. They also showed that these effects could be at least partly compensated by the over-expression of lamin A/C. As fairly acknowledged by the authors, the induction of nuclear deformation in paclitaxel-treated cells, and the increased sensitivity to paclitaxel of cells expressing low level of lamin A/C are not novel (reference #14). Here the authors provided more details on the cytoskeleton changes and nuclear membrane deformation upon paclitaxel treatment. The effect of lamin A/C over and down expression on cell growth and survival are not fully convincing, as further discussed below. The most novel part is the observation that paclitaxel can induce the down-regulation of the expression of lamin A/C and that this effect is mediated by SUN2.

We appreciate the reviewer's summary and thank them for their time. We believe our comprehensive revisions have addressed all comments, strengthening the manuscript and making it more robust and compelling.

Reviewer #2 Evidence, reproducibility and clarity This study investigates the effects of the chemotherapeutic drug paclitaxel on nuclear-cytoskeletal coupling during interphase, claiming a novel mechanism for its anti-cancer activity. The study uses hTERT-immortalized human fibroblasts. After paclitaxel exposure, a suite of state- of-the-art imaging modalities visualizes changes in the cytoskeleton and nuclear architecture. These include STORM imaging and a large number of FIB-SEM tomograms.

We thank the reviewer for the summary and for highlighting our efforts in using the latest imaging technical advances.

Major comments:

The authors make a major claim that in addition to the somewhat well-described mechanism of paclitaxel on mitosis, they have discovered 'an alternative, poorly characterised mechanism in interphase'.

However, none of the data proves that the effects shown are independent of mitosis. To the contrary, measurements are presented 48 hours after paclitaxel treatment starts, after which it can be assumed that 100% of cells have completed at least one mitotic event. The appearance of micronuclei evidences this, as discussed by the authors shortly. It looks like most of the results shown are based on botched mitosis or, more specifically, errors on nuclear assembly upon exit from mitosis rather than a specific effect of paclitaxel on interphase. The readouts the authors show just happen to be measurements while the cells are in interphase.

Alternative hypotheses are missing throughout the manuscript, and so are critical controls and interpretations.

We thank the reviewer for highlighting the lack of clarity in our wording. We have revised the title, abstract and relevant sections of the introduction and discussion to clarify our message that the effects of paclitaxel on the nucleus arise from a combination of aberrant nuclear-cytoskeletal coupling during interphase and multimicronucleation following exit from defective mitosis. We have also included additional data where we used slow-dividing, serum-starved cells (under these conditions, the majority of cells do not undergo mitosis during the 16 h incubation in paclitaxel [Supplementary Figure 2D]). Our new data show that even in these cells there is a clear effect of paclitaxel on nuclear solidity, and Lamin A/C and SUN2 protein levels, further supporting our hypothesis that these phenotypes can occur independently of cell division (Figure 2C; Figure 3H-J). Furthermore, we performed additional experiments where we used overexpression of GFP-Tau as an alternative method of stabilising microtubules in interphase and observed similar aberrations to nuclear solidity and Lamin A/C localisation. As GFP-Tau overexpression did not lead to micronucleation or appear to affect mitosis, these data support the hypothesis that nuclear aberrations occur due to microtubule bundling in interphase (Figure 4). We discuss these experiments in more detail below. Finally, we have reworded the introduction to better introduce alternative hypotheses and mechanisms for paclitaxel's activity.

The authors claim that 'Previously, the anti-cancer activity of paclitaxel was thought to rely mostly on the activation of the mitotic checkpoint through disruption of microtubule dynamics, ultimately resulting in apoptosis.' The authors may have overlooked much of the existing literature on the topic, including many recent manuscripts from Xiang-Xi Xu's and another lab.

We would like to note that the paper from Xiang-Xi Xu's lab (Smith et al, 2021) was cited in our original manuscript (reference 14 in both the original and revised manuscripts). We have now also included additional review articles from the Xiang-Xi Xu lab (PMID:36368286 20 and PMID: 35048083 21). Furthermore, we have clarified the wording in both the introduction and discussion to better reflect the current understanding of paclitaxel's mechanism and alternative hypotheses.

The data, e.g. in Figure 1, does not hold up to the first alternative hypothesis, e.g. that paclitaxel stabilizes microtubules and that excessive mechanical bundling of microtubules induces major changes to cell shape and mechanical stress on the nucleus. Even the simplest controls for this effect (the application of an alternative MT stabilizing drug or the overexpression of an MT stabilizer, e.g., tau).

We thank the reviewer for suggesting this control experiment using the microtubule stabiliser Tau. We have now included these experiments in the revised version of the manuscript (Figure 4). The overexpression of GFP-Tau supports our hypothesis that cytoskeletal reorganisation in paclitaxel exerts mechanical stress on the nucleus during interphase, resulting in nuclear deformation and aberrations to the nuclear lamina. In particular, GFP-Tau overexpression resulted in large microtubule bundles throughout the cytoplasm (Figure 4A). Notably, in cells where these bundles extensively contacted the nucleus, we observed a significant decrease in nuclear solidity (Figure 4B) accompanied by changes in nuclear lamina organisation, including a patchy lamina phenotype, similar to that induced by paclitaxel (Figure 4C).

The focus on nuclear lamina seems somewhat arbitrary and adjacent to previously published work by other groups. What would happen if the authors stained for focal adhesion markers? There would probably be a major change in number and distribution. Would the authors conclude that paclitaxel exerts a specific effect on focal adhesions? Or would the conclusion be that microtubule stabilization and the following mechanical disruption induce pleiotropic effects in cells? Which effects are significant for paclitaxel function on cancer cells?

We thank the reviewer for raising important points regarding the specificity of paclitaxel's effects. We agree that microtubule stabilisation can induce myriad cellular changes, including alterations to focal adhesions and other cytoskeletal components. Our focus on Lamin A/C and nuclear morphology is grounded both in the established clinical relevance of nuclear mechanics in cancer and builds on mechanistic work from other groups.

Lamin A/C expression is commonly altered in cancer, and nuclear morphology is frequently used in cancer diagnosis35. Lamin A/C also plays a crucial role in regulating nuclear mechanics32 and, importantly, determines cell sensitivity to paclitaxel14. However, the mechanism by which Lamin A/C determines sensitivity of cancer cells to paclitaxel is unclear.

Our data are consistent with Lamin A/C being a determinant of paclitaxel survival sensitivity. We also provide evidence that paclitaxel itself reduces Lamin A/C protein levels and disrupts its organisation at the nuclear envelope. We directly link these effects to microtubule bundling around the nucleus and degradation of force-sensing LINC component SUN2, highlighting the importance of nuclear architecture and mechanics to overall cellular function. Furthermore, we show that recovery from paclitaxel treatment depends on Lamin A/C expression levels. This has clinical relevance, as unlike cancer cells, healthy tissue with non-aberrant lamina would be able to selectively recover from paclitaxel treatment.

Minor comments:

While I understand the difficulty of the experiments and the effort the authors have put into producing FIB-SEM tomograms, I am not sure they are helping their study or adding anything beyond the light microscopy images. Some of the images may even be in the way, such as supplementary Figure 6, which lacks in quality, controls, and interpretation. Do I see a lot of mitochondria in that slice?

We agree with the reviewer that Supplementary Figure 6 does not add significant value to the manuscript and thank the reviewer for pointing this out. We have removed it from the manuscript accordingly.

I may have overlooked it, but has the number of cells from which lamellae have been produced been stated?

We thank the reviewer for pointing out the missing information. For our cryo-ET experiments, we collected data from 9 lamellae from paclitaxel-treated cells and 6 lamellae from control cells, with each lamella derived from a single cell. This information has now been added to the figure legend (Figure 2F).

Significance

The significance of studying the effect of paclitaxel, the most successful chemotherapy drug, should be broad and of interest to basic researchers and clinicians.

As outlined above, I believe that major concerns about the design and interpretation of the study hamper its significance and advancements.

We appreciate the reviewer's concerns and have performed major revisions to strengthen the significance of our study. Specifically, we conducted two key sets of experiments to validate our original conclusions: serum starvation to control for the effects of cell division, and overexpression of the microtubule stabiliser Tau to demonstrate that paclitaxel can affect the nucleus via its microtubule bundling activity in interphase.

By elucidating the mechanistic link between microtubule stabilisation and nuclear-cytoskeletal coupling, our findings contribute to our understanding of paclitaxel's multifaceted actions in cancer cells.

My areas of expertise could be broadly defined as Cell Biology, Cytoskeleton, Microtubules, and Structural Biology.

Reviewer #3 Evidence, reproducibility and clarity The manuscript presents interesting new ideas for the mechanism of an old drug, taxol, which has been studied for the last 40 years.

We thank the reviewer for the positive feedback.

Although similar ideas are published, which may be suitable to be cited? • Paclitaxel resistance related to nuclear envelope structural sturdiness. Smith ER, Wang JQ, Yang DH, Xu XX. Drug Resist Updat. 2022 Dec;65:100881. doi: 10.1016/j.drup.2022.100881. Epub 2022 Oct 15. PMID: 36368286 Review. • Breaking malignant nuclei as a non-mitotic mechanism of taxol/paclitaxel. Smith ER, Xu XX. J Cancer Biol. 2021;2(4):86-93. doi: 10.46439/cancerbiology.2.031. PMID: 35048083 Free PMC article.

We thank the reviewer for bringing to our attention these important review articles. In our initial manuscript, we only cited the original paper (14, also reference 14 in the original manuscript). We have now included citations to the suggested publications (20,21).

We would also like to emphasise how our manuscript distinguishes itself from the work of Smith et al.14,20,21:

• Cell-type focus: In their study 14, Smith et al. examined the effect of paclitaxel on malignant ovarian cancer cells and proposed that paclitaxel's effects on the nucleus are limited to cancer cells. However, our data extends these findings by demonstrating paclitaxel's effects in both cancerous and non-cancerous backgrounds.

• Cytoskeletal reorganisation: Smith et al. show reorganisation of microtubules in paclitaxel-treated cells14. Our data show re-organisation of other cytoskeletal components, including F-actin and vimentin.

• Multimicronucleation: Smith et al. propose that paclitaxel-induced multimicronucleation occurs independently of cell division14. Although we observe progressive nuclear abnormalities during interphase over the course of paclitaxel treatment, our data do not support this conclusion; we find that multimicronucleation occurs only following mitosis.

• Direct link between microtubule bundling and nuclear aberrations: We show that nuclear aberrations caused by paclitaxel during interphase (distinct from multimicronucleation) are directly linked to microtubule bundling around the nucleus, suggesting they result from mechanical disruption and altered force propagation.

• Lamin A/C regulation: Consistent with Smith et al.14, we show that Lamin A/C depletion leads to increased sensitivity to paclitaxel treatment. However, we further demonstrate that paclitaxel itself leads to reduced levels of Lamin A/C and that this effect occurs independently of mitosis and is mediated via force-sensing LINC component SUN2. Upon SUN2 knockdown, Lamin A/C levels are no longer affected by paclitaxel treatment.

• Recovery: Finally, our work reveals that cells expressing low levels of Lamin A/C recover less efficiently after paclitaxel removal. This might help explain how cancer cells could be more susceptible to paclitaxel.

Only one cell line was used in all the experiments? "Human telomerase reverse transcriptase (hTERT) immortalised human fibroblasts" ? The cells used are not very relevant to cancer cells (carcinomas) that are treated with paclitaxel. It is not clear if the observations and conclusions will be able to be generalized to cancer cells.

We thank the reviewer for this comment. Our initial study aimed to understand the effects of paclitaxel on nuclear architecture in non-aberrant backgrounds. To show that the observed effects of paclitaxel are also applicable to cancer cells, we have now repeated our main experiments using MDA-MB-231 human breast cancer cells (Supplementary Figure 1B; Supplementary Figure 3P-T). Similar to our findings in human fibroblasts, paclitaxel treatment of MDA-MB-231 led to cytoskeletal reorganisation (Supplementary Figure 1B), a decrease in nuclear solidity (Supplementary Figure 3P), aberrant (patchy) localisation of Lamin A/C (Supplementary Figure 3Q), and a reduction in Lamin A/C and SUN2 levels (Supplementary Figure 3R-T).

"Fig. 1. (B) STORM imaging of α-tubulin immunofluorescence in cells fixed after 16 h incubation in control media or 5 nM paclitaxel. Lower panels show α-tubulin clusters generated with HDBSCAN analysis. Scale bars = 10 μm." It needs explanation of what is meaning of the different color lines in the lower panels, just different filaments?

We have added further detail to the figure legend for clarification: "Lower panels show α-tubulin clusters generated with HDBSCAN analysis. Different colours distinguish individual α-tubulin clusters, representing individual microtubule filaments or filament bundles."

Generally, the figures need additional description to be clear.

We have added further clarification and detail to our figure legends.

"Figure 3 - Paclitaxel results in aberrations to the nuclear lamina." The sentence seems not to be well constructed. "Paclitaxel treatment causes ..."?

We changed this sentence to: "Figure 3 - Paclitaxel treatment results in aberrant organisation of the nuclear lamina and decreased Lamin A/C levels via SUN2."

Lamin A and C levels are different in different images (Fig. 3B, H): some Lamin A is higher, and sometime Lamin C is higher? This may possibly due to culture condition or subtle difference in sample handling?.

We thank the reviewer for pointing this out and we agree that the ratio of Lamin A to Lamin C can vary with culture conditions. To confirm that paclitaxel treatment reduces total Lamin A/C levels regardless of this ratio, we repeated the Western blot analysis in three additional biological replicates using cells in which Lamin C levels exceeded Lamin A levels. These experiments confirmed a comparable decrease in total Lamin A/C levels. Figure 3B and 3C have been updated accordingly.

Also, the effect on Lamin A/C and SUN2 levels are not significant of robust.

Decreased Lamin A/C and SUN2 levels following paclitaxel treatment were consistently seen across three or more biological repeats (Figure 3B-C), and this could be replicated in a different cell type (MDA-MB-231) (Supplementary Figure 3R-T). Furthermore, Western blotting results are consistent with the patchy Lamin A/C distribution observed using confocal and STORM following paclitaxel treatment (Figure 3A; Supplementary Figure 3A), where Lamin A/C appears to be absent from discrete areas of the lamina.

Any mechanisms are speculated for the reason for the reduction?

We have now included additional data which aims to shed light on the mechanism behind the decrease in Lamin A/C and SUN2 levels following paclitaxel treatment. We found that SUN2 is selectively degraded during paclitaxel treatment. Immunoprecipitation of SUN2 followed by Western blotting against Polyubiquitin C showed increased SUN2 ubiquitination in paclitaxel (Figure 3M and N). Furthermore, in our original manuscript, we showed that Lamina A/C levels remained unaltered during paclitaxel treatment in cells where SUN2 had been knocked down. We propose that changes in microtubule organisation affect force propagation to Lamin A/C specifically via SUN2 and that this leads to Lamina A/C removal and depletion. Future work will be needed to fully understand this mechanism.

In addition to the findings described above, we report no significant changes in mRNA levels for LMNA or SUN2 in paclitaxel (Supplementary Figure 3B and O). Phos-tag gels followed by Western blotting analysis for Lamin A/C also did not detect changes to the overall phosphorylation status of Lamin A/C due to paclitaxel treatment. This is in agreement with our initial data showing no changes to Lamin A/C Ser 404 phosphorylation levels (Supplementary Figure 3E and F). Finally, Lamin A/C immunoprecipitation experiments followed by Western blotting for Polyubiquitin C and acetyl-lysine showed no significant changes in the ubiquitination and acetylation state of Lamin A/C in paclitaxel-treated cells (Supplementary Figure 3G-I).

Also, the about 50% reduction in protein level is difficult to be convincing as an explanation of nuclear disruption.

The nuclear lamina and LINC complex proteins play a critical role in regulating nuclear integrity, stiffness and mechanical responsiveness to external forces28,31-33,54,75, as well as in maintaining the nuclear intermembrane distance69,74. In particular, SUN-domain proteins physically bridge the nuclear lamina to the cytoskeleton through interactions with Nesprins, thereby preserving the perinuclear space distance30,69,74. Mutations in Lamins have been shown to disrupt chromatin organization, alter gene expression, and compromise nuclear structural integrity, and experiments with LMNA knockout cells reveal that nuclear mechanical fragility is closely coupled to nuclear deformation47. Furthermore, nuclear-cytoskeletal coupling is essential during processes such as cell migration, where cells undergo stretching and compression of the nucleus; weakening or loss of the lamina in such cases compromises cell movement47,73. In our work, we show that alterations to nuclear Lamin A/C and SUN2 by paclitaxel treatment coincide with nuclear deformations (Figure 2A-D, F, G; Figure 3A-D, F, G; Supplementary Figure 3A, P-T) and that these deformations are reversible following paclitaxel removal (Supplementary Figure 4B-D). Our experiments also demonstrate that Lamin A/C expression levels significantly influence cell growth, cell viability, and cell recovery in paclitaxel (Figure 5). Therefore, drawing on current literature and our results, we propose that, during interphase, paclitaxel induces severe nuclear aberrations through the combined effects of: i) increased cytoskeletal forces on the NE caused by microtubule bundling; ii) loss of ~50% Lamin A/C and SUN2; iii) reorganisation of nucleo-cytoskeletal components.

Significance

The manuscript presents interesting new ideas for the mechanism of an old drug, taxol, which has been studied for the last 40 years.

The data may be improved to provide stronger support.

Additional cell lines (of cancer or epithelial origin) may be repeated to confirm the generality of the observation and conclusions.?

We thank the reviewer for the feedback and valuable suggestions. In response, we have included experiments using human breast cancer cell line MDA-MB-231 to further corroborate our findings and interpretations. We believe these additions have improved the clarity, robustness and impact of our manuscript, and we are grateful for the reviewer's contributions to its improvement.

Author response:

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

Reviewer #1 (Public review):

Summary:

In this study by Li et al., the authors re-investigated the role of cDC1 for atherosclerosis progression using the ApoE model. First, the authors confirmed the accumulation of cDC1 in atherosclerotic lesions in mice and humans. Then, in order to examine the functional relevance of this cell type, the authors developed a new mouse model to selectively target cDC1. Specifically, they inserted the Cre recombinase directly after the start codon of the endogenous XCR1 gene, thereby avoiding off-target activity. Following validation of this model, the authors crossed it with ApoE-deficient mice and found a striking reduction of aortic lesions (numbers and size) following a high-fat diet. The authors further characterized the impact of cDC1 depletion on lesional T cells and their activation state. Also, they provide in-depth transcriptomic analyses of lesional in comparison to splenic and nodal cDC1. These results imply cellular interactions between lesion T cells and cDC1. Finally, the authors show that the chemokine XCL1, which is produced by activated CD8 T cells (and NK cells), plays a key role in the interaction with XCR1-expressing cDC1 and particularly in the atherosclerotic disease progression.<br /> Strengths:

The surprising results on XCL1 represent a very important gain in knowledge. The role of cDC1 is clarified with a new genetic mouse model.

Thank you

Weaknesses:

My criticism is limited to the analysis of the scRNAseq data of the cDC1. I think it would be important to match these data with published data sets on cDC1. In particular, the data set by Sophie Janssen's group on splenic cDC1 might be helpful here (PMID: 37172103; https://www.single-cell.be/spleen_cDC_homeostatic_maturation/datasets/cdc1). It would be good to assign a cluster based on the categories used there (early/late, immature/mature, at least for splenic DC).

Thank you very much for your help. Using the scRNA seq data of Xcr1⁺ cDC1 sorted from ApoE^–/– mice, we re-annotated the populations, following the methodology proposed by Sophie Janssen's group. These results are presented in Figure S9 and Figure S10 and described in detail in the Results and Discussion section.

Please refer to the Results section from line 264 to 284: “Using the scRNA seq data of Xcr1⁺ cDC1 sorted from hyperlipidemic mice, we annotated the 10 populations as shown in Figure S9A, following the methodology from a previous study [41]. Ccr7⁺ mature cDC1s (Cluster 3, 7 and 9) and Ccr7- immature cDC1s (remaining clusters) were identified across cDC1 cells sorted from aorta, spleen and lymph nodes (Figure S9B). Further stratification based on marker genes reveals that Cluster 10 is the pre-cDC1, with high expression level of CD62L (Sell) and low expression level of CD8a (Figure S9C). Cluster 6 and 8 are the proliferating cDC1s, which express high level of cell cycling genes Stmn1 and Top2a (Figure S9D). Cluster 1 and 4 are early immature cDC1s, and cluster 2 and 5 are late immature cDC1s, according to the expression pattern of Itgae, Nr4a2 (Figure S9E). Cluster 9 cells are early mature cDC1s, with elevated expression of Cxcl9 and Cxcl10 (Figure S9F). Cluster 3 and 7 as late mature cDC1s, characterized by the expression of Cd63 and Fscn1 (Figure S9G). As shown in Figure 5C and Figure S9, the 10 populations displayed a major difference of aortic cDC1 cells that lack in pre-cDC1s (cluster 10) and mature cells (cluster 3, 7 and 9). Interestingly, in hyperlipidemic mice splenic cDC1 possess only Cluster 3 as the late mature cells while the lymph node cDC1 cells have two late mature populations namely Cluster 3 and Cluster 7. In further analysis, we also compared splenic cDC1 cells from HFD mice to those from ND mice. As shown in Figure S10, HFD appears to impact early immature cDC1-1 cells (Cluster 1) and increases the abundance of late immature cDC1 cells (Cluster 2 and 5), regardless of the fact that all 10 populations are present in two origins of samples. We also found that Tnfaip3 and Serinc3 are among the most upregulated genes, while Apol7c and Tifab are downregulated in splenic cDC1 cells sorted from HFD mice”.  

Please refer to the Discussion section from line 380 to 385: “Based on the maturation analysis of the cDC1 scRNA seq data [41], our findings suggest that the aortic cDC1 cells display a major difference from those of spleen and lymph nodes by lacking the mature clusters, whereas lymph node cDC1 cells contain an additional Fabp5⁺ S100a4⁺ late mature Cluster. Our results also suggest that hyperlipidemia contributes to alteration in early immature cDC1 and in the abundance of late immature cDC1 cells, which was associated with dramatic change in gene expression of Tnfaip3, Serinc3, Apol7c and Tifab”.

Reviewer #2 (Public review):

This study investigates the role of cDC1 in atherosclerosis progression using Xcr1Cre-Gfp Rosa26LSL-DTA ApoE-/- mice. The authors demonstrate that selective depletion of cDC1 reduces atherosclerotic lesions in hyperlipidemic mice. While cDC1 depletion did not alter macrophage populations, it suppressed T cell activation (both CD4+ and CD8+ subsets) within aortic plaques. Further, targeting the chemokine Xcl1 (ligand of Xcr1) effectively inhibits atherosclerosis. The manuscript is well-written, and the data are clearly presented. However, several points require clarification:

(1) In Figure 1C (upper plot), it is not clear what the Xcr1 single-positive region in the aortic root represents, or whether this is caused by unspecific staining. So I wonder whether Xcr1 single-positive staining can reliably represent cDC1. For accurate cDC1 gating in Figure 1E, Xcr1+CD11c+ co-staining should be used instead.

The observed false-positive signal in the wavy structures within immunofluorescence Figure 1C (upper panel) results from the strong autofluorescence of elastic fibers, a major vascular wall component (alongside collagen). This intrinsic property of elastic fibers is a well-documented confounder in immunofluorescence studies [A, B].

In contrast, immunohistochemistry (IHC) employs an enzymatic chromogenic reaction (HRP with DAB substrate) that generates a brown precipitate exclusively at antigen-antibody binding sites. Importantly, vascular elastic fibers lack endogenous enzymatic activity capable of catalyzing the DAB reaction, thereby preventing this source of false positivity in IHC.

Given that Xcr1 is exclusively expressed on conventional type 1 dendritic cells [C], and considering that IHC lacks the multiplexing capability inherent to immunofluorescence for antigen co-localization, single-positive Xcr1 staining reliably identifies cDC1s in IHC results.

[A] König, K et al. “Multiphoton autofluorescence imaging of intratissue elastic fibers.” Biomaterials vol. 26,5 (2005): 495-500. doi:10.1016/j.biomaterials.2004.02.059

[B] Andreasson, Anne-Christine et al. “Confocal scanning laser microscopy measurements of atherosclerotic lesions in mice aorta. A fast evaluation method for volume determinations.” Atherosclerosis vol. 179,1 (2005): 35-42. doi:10.1016/j.atherosclerosis.2004.10.040

[C] Dorner, Brigitte G et al. “Selective expression of the chemokine receptor XCR1 on cross-presenting dendritic cells determines cooperation with CD8+ T cells.” Immunity vol. 31,5 (2009): 823-33. doi:10.1016/j.immuni.2009.08.027

(2) Figure 4D suggests that cDC1 depletion does not affect CD4+/CD8+ T cells. However, only the proportion of these subsets within total T cells is shown. To fully interpret effects, the authors should provide:

(a) Absolute numbers of total T cells in aortas.

(b) Absolute counts of CD4+ and CD8+ T cells.

Thanks for your suggestions. We agree that assessing both proportions and absolute numbers in Figure 4 provides a more complete picture of the effects of cDC1 depletion on T cell populations. Furthermore, we also add the absolute count of cDC1 cells and total T cells, and CD44 MFI (mean fluorescence intensity) in CD4⁺ and CD8⁺ T cells in Figure 4, and supplemented corresponding textual descriptions in the revised manuscript.

Please refer to the Results section from line 183 to 187: “Subsequently, we assessed T cell phenotype in the two groups of mice. While neither the frequencies nor absolute counts of aortic CD4⁺ and CD8⁺ T cells differed significantly between two groups of mice (Figure 4D-F), CD69 frequency and CD44 MFI (Mean Fluorescence Intensity), the T cell activation markers, were significantly reduced in both CD4⁺ and CD8⁺ T cells from Xcr1⁺ cDC1 depleted mice compared to controls (Figure 4G and H)”.

(3) How does T cell activation mechanistically influence atherosclerosis progression? Why was CD69 selected as the sole activation marker? Were other markers (e.g., KLRG1, ICOS, CD44) examined to confirm activation status?

We sincerely appreciate these insightful comments. As extensively documented in the literature, activated effector T cells (both CD4+ and CD8+) critically promote plaque inflammation and instability through their production of pro-inflammatory cytokines (particularly IFN-γ and TNF-α), which drive endothelial activation, exacerbate macrophage inflammatory responses, and impair smooth muscle cell function [A].

In our study, we specifically investigated the role of cDC1 cells in atherosclerosis progression. Our key findings demonstrate that cDC1 depletion attenuates T cell activation (as shown by reduced CD69/CD44 expression) and that this reduction in activation is functionally linked to the observed decrease in atherosclerosis burden in our model. 

Regarding CD44 as an activation marker, we performed quantitative analyses of CD44 mean fluorescence intensity (MFI) in aortic T cells (Figure 4). Importantly, the MFI of CD44 was significantly lower on both CD4+ and CD8+ T cells from Xcr1^Cre-Gfp Rosa26^LSL-DTA ApoE^–/– mice compared to the control ApoE^–/– mice (data shown below), which is consistent with the result of CD69 in Figure 4. We added the related description in the Result section.

Please refer to the Results section from line 185 to 187 “CD69 frequency and CD44 MFI (Mean Fluorescence Intensity), the T cell activation markers, were significantly reduced in both CD4+ and CD8+ T cells from Xcr1+ cDC1 depleted mice compared to controls (Figure 4G and H)”.

Similarly, MFI of CD44 was significantly lower on both CD4⁺ and CD8⁺ T cells from Xcl1^–/– ApoE^–/– mice compared to the control ApoE^–/– mice (data shown below), which is consistent with the result of CD69 in Figure 7. We also added the related description in the Result section.

Please refer to the Results section from line 308 to 309 “Crucially, CD69⁺ frequency and CD44 MFI remained comparable in both aortic CD4⁺ and CD8⁺ T cells between two groups (Figure 7D-F).”

[A] Hansson, Göran K, and Andreas Hermansson. “The immune system in atherosclerosis.” Nature immunology vol. 12,3 (2011): 204-12. doi:10.1038/ni.2001

(4) Figure 7B: Beyond cDC1/2 proportions within cDCs, please report absolute counts of: Total cDCs, cDC1, and cDC2 subsets. Figure 7D: In addition to CD4+/CD8+ T cell proportions, the following should be included:

(a) Total T cell numbers in aortas

(b) Absolute counts of CD4+ and CD8+ T cells.

Thanks for your suggestions. We have now included in Figure 7 the absolute counts of cDC, cDC1, and cDC2 cells, along with CD4⁺ and CD8⁺ T cells in aortic tissues. Additionally, we provide the corresponding CD44 mean fluorescence intensity (MFI) measurements for both CD4⁺ and CD8⁺ T cell populations. We added the related description in the Result section.

Please refer to the Results section from line 303 to 311: “The flow cytometric results illustrated that both frequencies and absolute counts of Xcr1⁺ cDC1 cells in the aorta were significantly reduced, but cDCs and cDC2 cells from Xcl1^–/– ApoE^–/– were comparable with that from ApoE^–/– (Figure 7A-C). Moreover, in both lymph node and spleen, the absolute numbers of pDC, cDC1 and cDC2 from Xcl1^–/– ApoE^–/– were comparable with that from ApoE^–/– (Figure S11). Crucially, CD69⁺ frequency and CD44 MFI remained comparable in both aortic CD4⁺ and CD8⁺ T cells between two groups (Figure 7D-F). However, aortic CD8⁺ T cells exhibited reduced frequency and absolute count, while CD4⁺ T cells showed increased frequency but unchanged counts in Xcl1^–/– ApoE^–/– mouse versus controls (Figure 7G and H).”

(5) cDC1 depletion reduced CD69+CD4+ and CD69+CD8+ T cells, whereas Xcl1 depletion decreased Xcr1+ cDC1 cells without altering activated T cells. How do the authors explain these different results? This discrepancy needs explanation.

We sincerely appreciate your professional and insightful comments regarding the mechanistic relationship between cDC1 depletion and T cell activation. Direct cDC1 depletion in the Xcr1^Cre-Gfp Rosa26^LSL-DTA ApoE^–/– micmodel removes both recruited and tissue-resident cDC1s, eliminating their multifunctional roles in antigen presentation, co-stimulation and cytokine secretion essential for T cell activation. In contrast, Xcl1 depletion reduces, but does not eliminate cDC1 migration into plaques. Furthermore, alternative chemokine axes (e.g., CCL5/CCR5, CXCL9/CXCR3, BCL9/BCL9L) may partially rescue cDC1 recruitment [13, 68, 69], and non-cDC1 APCs (e.g., monocytes, cDC2s) may compensate for T cell activation [55, 70]. We emphasize that Xcl1 depletion specifically failed to alter T cell activation in hyperlipidemic ApoE^–/– mice. However, its impact may differ in other pathophysiological contexts due to compensatory mechanisms. We thank you again for highlighting this nuance, which strengthens our mechanistic interpretation. We have added these points to the discussion section and included new references.

Please refer to the Discussion section from line 407 to 413: “Notably, while complete ablation of Xcr1⁺ cDC1s impaired T cell activation, reduction of Xcr1⁺ cDC1 recruitment via Xcl1 deletion did not significantly compromise this process. This discrepancy may arise through compensatory mechanisms: alternative chemokine axes (e.g., CCL5/CCR5, CXCL9/CXCR3, BCL9/BCL9L) may partially rescue Xcr1⁺ cDC1 homing [13, 68, 69], while non-cDC1 antigen-presenting cells (e.g., monocytes, cDC2s) may sustain T cell activation [55, 70]. Furthermore, tissue-specific microenvironment factors could potentially modulate its role in other diseases.”. [13] Eisenbarth, S C. “Dendritic cell subsets in T cell programming: location dictates function.” Nature reviews. Immunology vol. 19,2 (2019): 89-103. doi:10.1038/s41577-018-0088-1 [55] Brewitz, Anna et al. “CD8+ T Cells Orchestrate pDC-XCR1+ Dendritic Cell Spatial and Functional Cooperativity to Optimize Priming.” Immunity vol. 46,2 (2017): 205-219. doi:10.1016/j.immuni.2017.01.003 [68] de Oliveira, Carine Ervolino et al. “CCR5-Dependent Homing of T Regulatory Cells to the Tumor Microenvironment Contributes to Skin Squamous Cell Carcinoma Development.” Molecular cancer therapeutics vol. 16,12 (2017): 2871-2880. doi:10.1158/1535-7163.MCT-17-0341.[69] He F, Wu Z, Liu C, Zhu Y, Zhou Y, Tian E, et al. Targeting BCL9/BCL9L enhances antigen presentation by promoting conventional type 1 dendritic cell (cDC1) activation and tumor infiltration. Signal Transduct Target Ther. 2024;9(1):139. Epub 2024/05/30. doi: 10.1038/s41392-024-01838-9. PubMed PMID: 38811552; PubMed Central PMCID: PMCPMC11137111.[70] Böttcher, Jan P et al. “Functional classification of memory CD8(+) T cells by CX3CR1 expression.” Nature communications vol. 6 8306. 25 Sep. 2015, doi:10.1038/ncomms9306.

Reviewer #1 (Recommendations for the authors):

(1) Line 32 - The authors might want to add that the mouse model leads to a "constitutive" depletion of cDC1.

Thanks for your advice, we have revised the sentence as follows.

Please refer to the Results section from line 31 to 33: “we established Xcr1^Cre-Gfp Rosa26^LSL-DTA ApoE^–/– mice, a novel and complex genetic model, in which cDC1 was constitutively depleted in vivo during atherosclerosis development”.

(2) Line 187-188: The authors claim that T cell activation was "inhibited" if cDC1 was depleted. The data shows that the T cells were less activated, but there is no indication of any kind of inhibition; this should be corrected.

Thanks for your advice, we have revised the sentence as follows.

Please refer to the Results section from line 183 to 187: “Subsequently, we assessed T cell phenotype in the two groups of mice. While neither the frequencies nor absolute counts of aortic CD4⁺ and CD8⁺ T cells differed significantly between two groups of mice (Figure 4D-F), CD69 frequency and CD44 MFI (Mean Fluorescence Intensity), the T cell activation markers, were significantly reduced in both CD4⁺ and CD8⁺ T cells from Xcr1⁺ cDC1 depleted mice compared to controls (Figure 4G and H)”.

(3) Why are some splenic DC clusters absent in LNs and vice versa? This is not obvious to this reviewer and should at least be discussed.

We appreciate the insightful question regarding the absence of certain splenic DC clusters in LNs. This phenomenon in Figure 5 aligns with the 'division of labor' paradigm in dendritic cell biology: tissue microenvironments evolve specialized DC subsets to address local immunological challenges. The absence of universal clusters reflects functional adaptation, not technical artifacts. We acknowledge that this tissue-specific heterogeneity warrants further discussion and have expanded our analysis to address this point in the discussion part of our manuscript.

Please refer to the Discussion section from line 375 to 385: “This pronounced tissue-specific compartmentalization of Xcr1⁺ cDC1 subsets may related to multiple mechanisms including developmental imprinting that instructs precursor differentiation into transcriptionally distinct subpopulations [62], and microenvironmental filtering through organ-specific chemokine axes (e.g., CCL2/CCR2 in spleen) selectively recruits receptor-matched subsets [63, 64]. This spatial specialization optimizes pathogen surveillance for local immunological challenges. Based on the maturation analysis of the cDC1 scRNA seq data [41], our findings suggest that the aortic cDC1 cells display a major difference from those of spleen and lymph nodes by lacking the mature clusters, whereas lymph node cDC1 cells contain an additional Fabp5⁺ S100a4⁺ late mature Cluster. Our results also suggest that hyperlipidemia contributes to alteration in early immature cDC1 and in the abundance of late immature cDC1 cells, which was associated with dramatic change in gene expression of Tnfaip3, Serinc3, Apol7c and Tifab”.

[62]. Liu Z, Gu Y, Chakarov S, Bleriot C, Kwok I, Chen X, et al. Fate Mapping via Ms4a3-Expression History Traces Monocyte-Derived Cells. Cell. 2019;178(6):1509-25 e19. Epub 2019/09/07. doi: 10.1016/j.cell.2019.08.009. PubMed PMID: 31491389.

[63]. Bosmans LA, van Tiel CM, Aarts S, Willemsen L, Baardman J, van Os BW, et al. Myeloid CD40 deficiency reduces atherosclerosis by impairing macrophages' transition into a pro-inflammatory state. Cardiovasc Res. 2023;119(5):1146-60. Epub 2022/05/20. doi: 10.1093/cvr/cvac084. PubMed PMID: 35587037; PubMed Central PMCID: PMCPMC10202633.

[64]. Mildner A, Schonheit J, Giladi A, David E, Lara-Astiaso D, Lorenzo-Vivas E, et al. Genomic Characterization of Murine Monocytes Reveals C/EBPbeta Transcription Factor Dependence of Ly6C(-) Cells. Immunity. 2017;46(5):849-62 e7. Epub 2017/05/18. doi: 10.1016/j.immuni.2017.04.018. PubMed PMID: 28514690.

[41]. Bosteels V, Marechal S, De Nolf C, Rennen S, Maelfait J, Tavernier SJ, et al. LXR signaling controls homeostatic dendritic cell maturation. Sci Immunol. 2023;8(83):eadd3955. Epub 2023/05/12. doi: 10.1126/sciimmunol.add3955. PubMed PMID: 37172103.

(4) The authors should discuss how XCL1 could impact lesional cDC1 and T cell abundance. Notably, preDCs do not express XCR1, and T cells express XCL1 following TCR activation. Is there a recruitment or local proliferation defect of cDC1 in the absence of XCL1? Could there also be a role for NK cells as a potential source of XCL1?

We appreciate your insightful questions regarding the differential effects of Xcl1 on cDC1s and T cells. Xcl1 primarily mediates the recruitment of mature cDC1s. Our data demonstrate that Xcl1 deletion significantly reduces aortic cDC1 abundance, which correlates with a concomitant decrease in CD8⁺ T cell numbers within the aorta. These findings strongly suggest that the Xcl1-Xcr1 axis plays a regulatory role in T cell accumulation in aortic plaques.

Consistent with prior studies [A, B], cDC1 recruitment can occur in the absence of Xcl1 which echoes our findings that cDC1 cells were still found in Xcl1 knockout aortic plaque but in lower abundance. It is very true that further studies are required to address how the Xcl1 dependent and independent cDC1 cells activate T cells and if they possess capability of proliferation in tissue differentially. We have added these points in discussion section.

Please refer to the Discussion section from line 407 to 415: “Notably, while complete ablation of Xcr1⁺ cDC1s impaired T cell activation, reduction of Xcr1⁺ cDC1 recruitment via Xcl1 deletion did not significantly compromise this process. This discrepancy may arise through compensatory mechanisms: alternative chemokine axes (e.g., CCL5/CCR5, CXCL9/CXCR3, BCL9/BCL9L) may partially rescue Xcr1⁺ cDC1 homing [13, 68, 69], while non-cDC1 antigen-presenting cells (e.g., monocytes, cDC2s) may sustain T cell activation [55, 70]. Furthermore, tissue-specific microenvironment factors could potentially modulate its role in other diseases. In summary, our findings identify Xcl1 as a potential therapeutic target for atherosclerosis therapy, though its cellular origins and regulation of lesional Xcr1⁺ cDC1 and T cells dynamics require further studies”.

In literatures, Xcl1 are expressed in NK cells and subsects of T cells, and NK cells can be a potential source of Xcl1 during atherosclerosis which deserve further investigations [A, C, D].

[A] Böttcher, Jan P et al. “NK Cells Stimulate Recruitment of cDC1 into the Tumor Microenvironment Promoting Cancer Immune Control.” Cell vol. 172,5 (2018): 1022-1037.e14. doi:10.1016/j.cell.2018.01.004

[B] He, Fenglian et al. “Targeting BCL9/BCL9L enhances antigen presentation by promoting conventional type 1 dendritic cell (cDC1) activation and tumor infiltration.” Signal transduction and targeted therapy vol. 9,1 139. 29 May. 2024, doi:10.1038/s41392-024-01838-9

[C] Woo, Yeon Duk et al. “The invariant natural killer T cell-mediated chemokine X-C motif chemokine ligand 1-X-C motif chemokine receptor 1 axis promotes allergic airway hyperresponsiveness by recruiting CD103+ dendritic cells.” The Journal of allergy and clinical immunology vol. 142,6 (2018): 1781-1792.e12. doi:10.1016/j.jaci.2017.12.1005

[D] Winkels, Holger et al. “Atlas of the Immune Cell Repertoire in Mouse Atherosclerosis Defined by Single-Cell RNA-Sequencing and Mass Cytometry.” Circulation research vol. 122,12 (2018): 1675-1688. doi:10.1161/CIRCRESAHA.117.312513

Reviewer #2 (Recommendations for the authors):

There is a logical error in line 298. I suggest revising to: "Collectively, these data suggest that Xcl1 promotes atherosclerosis by recruiting Xcr1+ cDC1 cells, which subsequently drive T cell activation in lesions."

Thanks for your advice. Since Xcl1 deficiency reduced both the frequencies and absolute counts of Xcr1+ cDC1 and CD8+ T cells in lesions without affecting T cell activation, we revised the sentence as you suggested.

Please refer to the Results section from line 314 to 315: “Collectively, these data suggest that Xcl1 promotes atherosclerosis by recruiting Xcr1⁺ cDC1 cells, and facilitating CD8⁺ T cell accumulation in lesions”.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Manuscript number: RC-2025-02946

Corresponding author(s): Margaret, Frame

Roza, Masalmeh

[Please use this template only if the submitted manuscript should be considered by the affiliate journal as a full revision in response to the points raised by the reviewers.

If you wish to submit a preliminary revision with a revision plan, please use our "Revision Plan" template. It is important to use the appropriate template to clearly inform the editors of your intentions.]

1. General Statements [optional]

This section is optional. Insert here any general statements you wish to make about the goal of the study or about the reviews.

We thank the reviewers for recognizing the significance of our work and for their constructive feedback and suggestions, most of which we have implemented in our revised manuscript.

2. Point-by-point description of the revisions

This section is mandatory. *Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. *

Reviewer #1

Evidence, reproducibility and clarity

Review of Masalmeh et al. Title: "FAK modulates glioblastoma stem cell energetics..."

Previous studies have implicated FAK and the related tyrosine kinase PYK2 in glioblastoma growth, cell migration, and invasion. Herein, using a murine stem cell model of glioblastoma, the authors used CRISPR to inactivate FAK, FAK-null cells selected and cloned, and lentiviral re-expression of murine FAK in the FAK-null cells (termed FAK Rx) was accomplished. FAK-/- cells were shown to possess epithelial characteristics whereas FAK Rx cells expressed mesenchymal markers and increased cell migration/invasion in vitro. Comparisons between FAK-/- and FAK Rx cells showed that FAK re-expressed increased mitochondrial respiration and amino acid uptake. This was associated with FAK Rx cells exhibiting filamentous mitochondrial morphology (potentially an OXPHOS phenotype) and decreased levels of MTFR1L S235 phosphorylation (implicated in mito morphology fragmentation). Mito and epithelial cell morphology of FAK-/- cells was reversed by treatment with Rho-kinase inhibitors that also increased mito metabolism and cell viability. Last, FAK-dependent glioblastoma tumor growth was shown by comparisons of FAK-/- and FAK Rx implantation studies.

The studies by Masalmeh provide interesting findings associating FAK expression with changes in mitochondrial morphology, energy metabolism, and glutamate uptake. According to the authors model, FAK expression is supporting a glioblastoma stem cell like phenotype in vitro and tumor growth in vivo. What remains unclear is the mechanistic connection to cell changes and whether or not these are be dependent on intrinsic FAK activity or as the Frame group has previously published, potentially FAK nuclear localization. The associations with MTFR1L phosphorylation and effects by Rho kinase inhibition are likely indirect and remind this reviewer of long-ago studies with FAK-null fibroblasts that exhibit epithelial characteristics, still express PYK2, exhibited elevated RhoA GTPase activity. Some of these phenotypes were linked to changes in RhoGEF and RhoGAP signaling with FAK and/or Pyk2. At a minimum, it would be informative to know whether Pyk2 signaling is relevant for observed phenotypes and whether the authors can further support their associations with FAK-targeted or FAK-Pyk2-targeted inhibitors or PROTACs.

Some questions that would enhance potential impact. 1. Cell generation. Please describe the analysis of FAK-/- clones in more detail. The "low viability" phenotype needs further explanation with regard to clonal expansion and growth characteristics?

Response:

• We included a better description and a supplementary figure in our revised manuscript to indicate that we have examined several FAK -/- clones and confirmed that our observations were not due to clonal variation; multiple clones displayed similar morphological changes (Figure S1D). We also show that the elongated mesenchymal-like morphology was observed at 48 h after nucleofecting the cells with the FAK‑expressing vector, before beginning G418 selection to enrich for cells expressing FAK (Figure S1C). We also included experiments to acutely modulate FAK signalling (detaching and seeding cells on fibronectin) (Figure S2D, E, F and Figure S3) to exclude the possibility that the profound effects are due to protocols/selection we used for generating FAK-deleted cells.
• Regarding the term "low viability", we have clarified in the text that there is no significant difference in cell number (Figure S1A) or 'cell viability' when it is assessed by trypan blue exclusion (a non-mitochondria-dependent read-out) (Figure S1B) between FAK-expressing FAK Rx and FAK-/- cells cultured for three days under normal conditions. Therefore, we agree the term 'cell viability' in this context could be confusing and have replace "cell viability" with "metabolic activity as measured by Alamar Blue." in Figure 1D and Figure 5B, and the corresponding text in the original manuscript. This wording more accurately reflects the data.

Figure 1F: need further support of MET change upon FAK KO and EMT reversion.

Response: We have added a heatmap (Figure S1E) illustrating the changes in protein expression of core-enriched EMT/MET genes products (by proteomics) after FAK gene deletion (EMT genes as defined in Howe et al., 2018) ; this strengthens the conclusion that the MET reversion morphological phenotype is accompanied by recognised MET protein changes.

Fig. 2: Need further support if FAK effects impact glycolysis or oxidative phosphorylation in particular as implicated by the stem cell model.

Response: We show that FAK impacts both glycolysis (Figure 2A, 2E, and 2F) and mitochondrial oxidative phosphorylation on the basis of the oxygen consumption rate (OCR) (Figure 2B, and 2D), showing both are contributing pathways to FAK-dependent energy production. We have clarified this in the text.

Is there a combinatorial potential between FAKi and chemotherapies used for glioblastoma. Need to build upon past studies.

Response: Yes, previous studies suggest that inhibiting FAK can sensitize GBM cells to chemotherapy (Golubovskaya et al., 2012; Ortiz-Rivera et al., 2023). We have included a paragraph in the discussion section to make sure this is clearer. Although it is not the subject of this study, we appreciate it is useful context.

The notation of changes in glucose transporter expression should be followed up with regard to the potential that FAK-expressing cells may have different uptake of carbon sources and other amino acids. Altered uptake could be one potential explanation for increase glycolysis and glutamine flux.

Response: We agree with the reviewer that glucose uptake could be contributing and we include data that 2 glucose transporters are indeed FAK-regulated namely Glucose transporter 1 (GLUT1, encoded by Slc2a1 gene) and Glucose transporter 3 (GLUT 3, encoded by Slc2a3 gene) (shown in Figure S2B and C).

It would be helpful to support the confocal microscopy of mitos with EM.

Response:

We are concerned (and in our experience) that Electron microscopy (EM) may introduce artefacts during sample preparation. In contrast, immunofluorescence sample preparation is less susceptible to artefacts. The SORA system we used is not a conventional point-scanning confocal microscope, but is a super-resolution module based on a spinning disk confocal platform (CSU-W1; Yokogawa) using optical pixel reassignment with confocal detection. This method enhances resolution in all dimensions with resolution in our samples measured at 120nm. This has been instructive in defining a new level of changes in mitochondrial morphology upon FAK gene deletion.

Lack of FAK expression with increased MTFR1 phosphorylation is difficult to interpret.

Response: We do not directly show that this phosphorylation event is causal in our experiments; however, we think it important to document this change since it has been published that phosphorylation of MTFR1 has been causally linked to the mitochondrial morphology we observed in other systems (Tilokani et al., 2022).

Need to have better support between loss of FAK and the increase in Rho signaling. Use of Rho kinase inhibitors is very limited and the context to FAK (and or Pyk2) remains unclear. Past studies have linked integrin adhesion to ECM as a linkage between FAK activation and the transient inhibition of RhoA GTP binding. Is integrin signaling and FAK involved in the cell and metabolism phenotypes in this new model?

Response: To better support the antagonistic effect of FAK on Rho-kinase (ROCK) signalling, we included a new experiment in which the integrin-FAK signalling pathway has been disrupted by treating FAK WT cells with an agent that causes detachment from the substratum, Accutase, and growing the cells in suspension in laminin-free medium. We present ROCK activity data, as judged by phosphorylated MLC2 at serine 19 (pMLC2 S19), relating this to induced FAK phosphorylation at Y397 (a surrogate for FAK activity) that is supressed after integrin disengagement. These measurements have been compared with conditions whereby integrin-FAK signalling is activated by growing the cells on laminin coated surfaces. We observed a time-dependent decrease in pFAK(Y397) levels (normalised to total FAK) in suspended cells compared to those spread on laminin, while pMLC2(S19) levels increased in a reciprocal manner over time in detached cells relative to spread cells (S4A and B). There is therefore an inverse relationship between integrin-FAK signalling and ROCK-MLC2 activity, consistent with findings from FAK gene deletion experiments. In the former case, we do not rely on gene deletion cell clones.

Significance

The studies by Masalmeh provide interesting findings associating FAK expression with changes in mitochondrial morphology, energy metabolism, and glutamate uptake. According to the authors model, FAK expression is supporting a glioblastoma stem cell like phenotype in vitro and tumor growth in vivo. What remains unclear is the mechanistic connection to cell changes and whether or not these are be dependent on intrinsic FAK activity or as the Frame group has previously published, potentially FAK nuclear localization. The associations with MTFR1L phosphorylation and effects by Rho kinase inhibition are likely indirect and remind this reviewer of long-ago studies with FAK-null fibroblasts that exhibit epithelial characteristics, still express PYK2, exhibited elevated RhoA GTPase activity. Some of these phenotypes were linked to changes in RhoGEF and RhoGAP signaling with FAK and/or Pyk2. At a minimum, it would be informative to know whether Pyk2 signaling is relevant for observed phenotypes and whether the authors can further support their associations with FAK-targeted or FAK-Pyk2-targeted inhibitors or PROTACs.

__Response: __

Deleting the gene encoding FAK in mouse embryonic fibroblasts leads to elevated Pyk2 expression (Sieg, 2000). However, in the GBM stem cell model we used here, Pyk2 was not expressed (determined by both transcriptomics and proteomics). We have included Figure S1E to show that PYK2 expression was undetectable in FAK -/- and FAK Rx cells at the RNA level (Figure S1F). We conclude that there is no compensatory increase in Pyk2 upon FAK loss in these cells. In the transformed neural stem cell model of GBM, we do not consistently or robustly detect nuclear FAK.

Review #2

Masalmeh and colleagues employ a neural stem/progenitor cell-based glioma model (NPE cells) to investigate the role of Focal Adhesion Kinase (FAK) in GBM, with a focus on potential links between the regulation of morphological/adhesive and metabolic GBM cell properties. For this, the authors employ wt cells alongside newly generated FAK-KO and -reexpressing cells, as well as pharmacological interventions to probe the relevance of specific signaling pathways. The authors´ main claims are that FAK crucially modulates glioma cell morphology, cell-cell and cell-substrate interactions and motility, as well as their metabolism, and that these effects translate to changes to relevant in vivo properties such as invasion and tumor growth.

My main issues are with the model chosen by the authors.

As per the methods section, generation of FAK-KO and -"Rx" NPE cells entailed protracted selection/expansion processes, which may have resulted in inadvertent selection for cellular/molecular properties unrelated to the desired one (loss or gain of FAK expression) and which may have had cascading effects on NPE cells. The authors nonetheless repeatedly claim the parameters they quantify, such as mitochondrial or cytoskeletal properties or metabolic features, to have directly resulted from FAK loss or reintroduction. Examples of such causal inferences are to be found in lines 123, 134/135, 165, 181. Such causal claims are, in my view, unsupported.

Acute perturbation of FAK expression/activity, genetically or pharmacologically, followed by a rapid assessment of the processes under investigation, would be needed to begin to assess causality, even if acute genetic perturbations may be technically challenging as sufficient gene expression reduction or restoration to physiologically relevant levels may be hard to achieve.

Response:

We would like to first comment on the model we used here, which we think will clarify the validity of our approach. The model is a transformed stem cell model of GBM that was published in (Gangoso et al., Cell, 2021) and is now used regularly in the GBM field. As mentioned in the response to Reviewer 1, we have added text (page 4 and 5 in the revised manuscript) and a new supplementary figure (Figure S1D) clarifying that the morphological changes we observed were consistent across multiple FAK -/- clones, showing this was not due to any inter-clonal variability. We also added images showing that the morphological changes were apparent at 48 h after nucleofecting FAK -/- cells with the FAK‑expressing vector specifically (not the empty vector), prior to starting G418 selection to enrich for FAK‑expressing cells (Figure S1C), addressing the worry that clonal variation and selection was the cause of the FAK-dependent phenotypes we observed. We believe that our model provides a type of well controlled, clean genetic cancer cell system of a type that is commonly used in cancer cell biology, allowing us to attribute phenotypes to individual proteins.

We have also carried out a more acute treatment by using the FAK inhibitor VS4718 to perturb FAK kinase activity and assessed the effects on glycolysis and glutamine oxidation after 48h treatment (Figure S2D, E and F). We found that treating the transformed neural stem cells (parental population) with FAK inhibitor (300nM VS4718) decreases glucose incorporation into glycolysis intermediates and glutamine incorporation into TCA cycle intermediates, consistent with a role for FAK's kinase activity in maintaining glycolysis and glutamine oxidation.

The employed pharmacological modulation of ROCK activity is the only approach that, given the presumably acute nature of the treatment, may have allowed the authors to probe the proposed functional links. The methods section of the manuscript does not however comprise details as to the duration of these treatments, which leaves open the possibility of long-term treatment having been carried out (data shown in Figure 5B refers to 72hr treatment).

__Response: __

We have added the duration of the treatment to the Methods section and Figure Legends, to clarify that cells were treated with ROCK inhibitors for 24h, before assessing the effects on mictochondria (Figure 4C, D, S4C and D) and glutamine oxidation (Figure 5A, and S5). For metabolic activity by AlamarBlue assay, cells were treated with ROCK inhibitors for 72h (Figure 5B).

Even in the case of ROCK inhibitor experiments, it is however unclear if and how the effects on cell morphology and adhesion, mitochondrial organization and metabolic activity may be connected to each other and, if at all, to FAK expression.

Given the above uncertainties due to the nature of the model and experimental approaches, it is hard to assess the reliability and thus the relevance of the findings.

Response:

FAK suppresses ROCK activity (as judged by pMLC2 S19, Figure 4A and B). Treating FAK -/- cells with two different ROCK inhibitors restored mesenchymal-like cell morphology, mitochondrial morphology and glutamine oxidation. As mentioned above, to strengthen our evidence for the antagonistic role of FAK in ROCK-MLC2 signalling, we have now introduced an experiment whereby integrin-FAK signalling was disrupted through treatment with a detachment agent (Accutase), and subsequently maintaining the cells in suspension in laminin-free medium. We assessed pMLC2 S19 levels (a measure of ROCK activity) relating this to FAK phosphorylation that is supressed after integrin disengagement. These results were evaluated relative to spread wild type cells growing on laminin where Integrin-FAK signalling was active (Figure S4A and B). We observed an inverse relationship between Integrin-FAK signalling and ROCK-MLC2 activity in keeping with our conclusions (Figure 4A and B).

Experimental support for the ability of cell-substrate interaction modulation to concomitantly impact cellular metabolism and motility/invasion would be significant both in terms of advancing our understanding of glioma cell biology and of its translational potential, but the evidence being provided is at best compatible with the proposed model.

Response: We carried out a new experiment to support the ability of cell-substrate interaction modulation to impact metabolism; specifically, we inhibited cell-substrate interactions by plating the cells on Poly-2-hydroxyethyl methacrylate (Poly 2-HEMA)-coated dishes. This suppressed FAK phosphorylation at Y397, as expected, with concomitant reduction in glutamine utilisation in the TCA cycle (Figure S3A, B and C).

My background/expertise is in developmental and adult neurogenesis, in vivo modelling of gliomagenesis and cell fate control/reprogramming, with a focus on molecular mechanisms of differentiation and quantitative aspects of lineage dynamics; molecular details of the control of cellular metabolism, cell-cell adhesion and cytoskeletal dynamics are not core expertise of mine.

We appreciate this reviewer's expertise are not necessarily in the cancer cell biology and genetic intervention aspects of our study. We hope that the explanations we have provided satisfy the reviewer that our conclusions are valid.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Review of Masalmeh et al.

Title: "FAK modulates glioblastoma stem cell energetics..."

Previous studies have implicated FAK and the related tyrosine kinase PYK2 in glioblastoma growth, cell migration, and invasion. Herein, using a murine stem cell model of glioblastoma, the authors used CRISPR to inactivate FAK, FAK-null cells selected and cloned, and lentiviral re-expression of murine FAK in the FAK-null cells (termed FAK Rx) was accomplished. FAK-/- cells were shown to possess epithelial characteristics whereas FAK Rx cells expressed mesenchymal markers and increased cell migration/invasion in vitro. Comparisons between FAK-/- and FAK Rx cells showed that FAK re-expressed increased mitochondrial respiration and amino acid uptake. This was associated with FAK Rx cells exhibiting filamentous mitochondrial morphology (potentially an OXPHOS phenotype) and decreased levels of MTFR1L S235 phosphorylation (implicated in mito morphology fragmentation). Mito and epithelial cell morphology of FAK-/- cells was reversed by treatment with Rho-kinase inhibitors that also increased mito metabolism and cell viability. Last, FAK-dependent glioblastoma tumor growth was shown by comparisons of FAK-/- and FAK Rx implantation studies.

The studies by Masalmeh provide interesting findings associating FAK expression with changes in mitochondrial morphology, energy metabolism, and glutamate uptake. According to the authors model, FAK expression is supporting a glioblastoma stem cell like phenotype in vitro and tumor growth in vivo. What remains unclear is the mechanistic connection to cell changes and whether or not these are be dependent on intrinsic FAK activity or as the Frame group has previously published, potentially FAK nuclear localization. The associations with MTFR1L phosphorylation and effects by Rho kinase inhibition are likely indirect and remind this reviewer of long-ago studies with FAK-null fibroblasts that exhibit epithelial characteristics, still express PYK2, exhibited elevated RhoA GTPase activity. Some of these phenotypes were linked to changes in RhoGEF and RhoGAP signaling with FAK and/or Pyk2. At a minimum, it would be informative to know whether Pyk2 signaling is relevant for observed phenotypes and whether the authors can further support their associations with FAK-targeted or FAK-Pyk2-targeted inhibitors or PROTACs.

Some questions that would enhance potential impact.

1. Cell generation. Please describe the analysis of FAK-/- clones in more detail. The "low viability" phenotype needs further explanation with regard to clonal expansion and growth characteristics?
2. Figure 1F: need further support of MET change upon FAK KO and EMT reversion.
3. Fig. 2: Need further support if FAK effects impact glycolysis or oxidative phosphorylation in particular as implicated by the stem cell model.
4. Is there a combinatorial potential between FAKi and chemotherapies used for glioblastoma. Need to build upon past studies.
5. The notation of changes in glucose transporter expression should be followed up with regard to the potential that FAK-expressing cells may have different uptake of carbon sources and other amino acids. Altered uptake could be one potential explanation for increase glycolysis and glutamine flux.
6. It would be helpful to support the confocal microscopy of mitos with EM.
7. Lack of FAK expression with increased MTFR1 phosphorylation is difficult to interpret.
8. Need to have better support between loss of FAK and the increase in Rho signaling. Use of Rho kinase inhibitors is very limited and the context to FAK (and or Pyk2) remains unclear. Past studies have linked integrin adhesion to ECM as a linkage between FAK activation and the transient inhibition of RhoA GTP binding. Is integrin signaling and FAK involved in the cell and metabolism phenotypes in this new model?

Significance

The studies by Masalmeh provide interesting findings associating FAK expression with changes in mitochondrial morphology, energy metabolism, and glutamate uptake. According to the authors model, FAK expression is supporting a glioblastoma stem cell like phenotype in vitro and tumor growth in vivo. What remains unclear is the mechanistic connection to cell changes and whether or not these are be dependent on intrinsic FAK activity or as the Frame group has previously published, potentially FAK nuclear localization. The associations with MTFR1L phosphorylation and effects by Rho kinase inhibition are likely indirect and remind this reviewer of long-ago studies with FAK-null fibroblasts that exhibit epithelial characteristics, still express PYK2, exhibited elevated RhoA GTPase activity. Some of these phenotypes were linked to changes in RhoGEF and RhoGAP signaling with FAK and/or Pyk2. At a minimum, it would be informative to know whether Pyk2 signaling is relevant for observed phenotypes and whether the authors can further support their associations with FAK-targeted or FAK-Pyk2-targeted inhibitors or PROTACs.

Author response:

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

Reviewer #1 (Public Review):

Major points

(1) The authors discovered a novel regulation of the Hippo-YAP pathway by SARS-CoV-2 infection but did not address the pathological significance of this finding. It remains unclear why YAP downstream gene transcription needs to be inhibited in response to SARS-CoV-2 infection. Is this inhibition crucial for the innate immune response to SARS-CoV-2? The authors should re-analyze their snRNA-seq and bulk RNA-seq data described in Figure 1 to determine whether any of the affected YAP downstream genes are involved in this process.

We appreciate the reviewer’s suggestion to clarify the pathological significance of YAP pathway inhibition in SARS-CoV-2 infection. To address this, we re-analyzed our snRNA-seq and bulk RNA-seq datasets to determine whether YAP target genes overlap with known mediators of the innate immune response. As described in Fig. 1C, bulk RNA-seq revealed decreased expression of multiple YAP downstream targets linked to innate immune regulation (e.g., Thbs1, Ccl2, Axl, and Csf1) in SARS-CoV-2–infected cells in vitro.

snRNA-seq of alveolar type I (AT1) cells from COVID-19 patients revealed a more complex landscape: While we observed reduced YAP activity overall (Fig. 1G), multiple YAP target genes involved in innate immunity and cytokine signaling were paradoxically elevated (Supplemental Fig. 1E). Several factors likelt explain these conflicting observations: 1. In the lung, AT1 cells (which are critical for gas exchange) may cell specifically respond to virus infection by upregulating genes related to immune response by other signaling pathway(s); 2. In vivo, SARS-CoV-2 infection triggers a surge in cytokines, chemokines, and other local factors that can differentially modulate YAP binding sites and thus affect its downstream targets, a complexity not fully captured in vitro; 3. YAP is highly sensitive to mechanical signals and tissue architecture. The 3D structure of altered cell–cell junctions in infected lung tissue, and fluid shear stress in the alveolar space could shape YAP target gene transcription differently from simplified monolayer cell cultures.

We have expanded the results section of the new version to include the above points. We also acknowledge that ongoing and future work is needed to delineate the exact molecular and tissue-specific pathways through which YAP inhibition confers a potential advantage in combating SARS-CoV-2.

(2) The authors concluded that helicase activity is required for NSP13-induced inhibition of YAP transcriptional activity based on mutation studies (Figure 3B). This finding is somewhat confusing, as K131, K345/K347, and R567 are all essential residues for NSP13 helicase activity while mutating K131 did not affect NSP13's ability to inhibit YAP (Figure 3B). Additionally, there are no data showing exactly how NSP13 inhibits the YAP/TEAD complex through its helicase function. This point was also not reflected in their proposed working model (Figure 4H).

We appreciate the reviewer’s concerns regarding the helicase‐dependent inhibition of YAP by NSP13, particularly the roles of K131, K345/K347, and R567. Based on published structural and biochemical studies, each of these residues uniquely supports helicase function (1): K131 is crucial for stabilizing the NSP13 stalk region by interacting with S424. Substituting K131 with alanine (K131A) reduces helicase efficiency but does not completely abolish it; K345/K347 are key DNA‐binding residues, and mutating both (K345A/K347A) largely prevents NSP13 from binding DNA, thus eliminating unwinding. R567 is critical for ATP hydrolysis, and the R567A mutant retains DNA binding capacity but fails to unwind it. In Fig. 3B, K131A suppresses YAP transactivation to nearly the same extent as wild‐type NSP13, suggesting that partial helicase activity is sufficient for complete YAP/TEAD inhibition. Conversely, the K345A/K347A and R567A mutants show markedly diminished repression, underscoring the importance of DNA binding and ATP hydrolysis.

As the new Fig. 4J illustrates, NSP13 must bind DNA and hydrolyze ATP to unwind nucleic acids. This helicase‐dependent process likely enables NSP13 to remodel chromatin structure by binding TEAD and properly organize YAP repressors at YAP/TEAD complex to prevent YAP/TEAD transactivation. In support of this mechanism, the K345A/K347A mutant, unable to anchor to DNA, fails to repress YAP and slightly increases YAP‐driven transcription (Fig. 3B), presumably by mislocalizing YAP repressors. Likewise, the ATPase‐dead R567A can bind DNA but does not unwind and remodel chromatin to recruit YAP repressors, resulting in a loss of YAP suppression (Fig. 3B and 3F). Our revised model demonstrates that both DNA binding and ATP‐dependent unwinding are essential for NSP13 to suppress YAP transcriptional activity. We have updated the results, discussion, and model accordingly.

(3) The proposed model that NSP13 binds TEAD4 to recruit repressor proteins and inhibits YAP/TEAD downstream gene transcription (Figure 4H) needs further characterization. Second, NSP13 is a DNA-binding protein, and its nucleic acid-binding mutant K345A/K347A failed to inhibit YAP transcriptional activity (Figure 3B). The authors should investigate whether NSP13 could bind to the TEAD binding sequence or the nearby sequence on the genome to modulate TEAD's DNA binding ability. Third, regarding the identified nuclear repressors, the authors should validate the interaction of NSP13 with the ones whose loss activates YAP transcriptional activity (Figure 4G). Lastly, why can't NSP13 bind TEAD4 in the cytoplasmic fractionation if both NSP13 and TEAD4 are detected there (Figure 3B)? This finding indicates their interaction is not a direct protein-protein interaction but is mediated by something in the nucleus, such as genomic DNA.

(1) Low TEAD expression in HEK293T cells: Our IP-MS experiments were performed in HEK293T cells, which, according to the Human Protein Atlas, express TEAD1–4 at comparatively low levels (TEAD1: 16.5, TEAD2: 16.4, TEAD3: 4.9, TEAD4: 38.7 nTPM). In contrast, HeLa cells, where we successfully validated NSP13-mediated YAP suppression (Fig. 4H, Supplementary Fig.5B-D), show higher expression of these TEAD isoforms (TEAD1: 97.1, TEAD2: 27.3, TEAD3: 12.2, TEAD4: 48.1 nTPM). Therefore, insufficient TEAD abundance in HEK293T cells may limit the sensitivity needed to detect TEAD–NSP13 interactions in our proteomic screens.

(2) Transience and potential DNA dependence: Our co-immunoprecipitation (co-IP) experiments (Fig. 4B, Supplementary Fig.4C-E) indicated that NSP13–TEAD4 binding is low-affinity. Under standard IP-MS conditions (which typically do not include chemical cross-linkers or nucleic acids to stabilize transient complexes), weak or short-lived interactions can be lost during washes or sample processing.

(3). Additional supporting evidence: We carefully checked our IP-MS data and found that the well-known TEAD binding proteins, including CTBP1/2 and GATA4, were pulled down, suggesting TEAD’s absence does not rule out an NSP13–TEAD association.

(3a) We acknowledge that our NSP13 immunoprecipitation–mass spectrometry (IP-MS) did not identify any TEAD proteins (Fig. 4G and IP-MS tables). Several factors likely contributed to this outcome:

(3b) We sincerely appreciate the reviewer’s insightful suggestion. While we agree that mapping NSP13 occupancy at individual TEAD-binding motifs is valuable, we respectfully consider this to be beyond the scope of the current study. Biochemical and structural work on coronavirus NSP13 shows that it recognizes nucleic‑acid substrates primarily through their 5′ single‑stranded overhang and duplex architecture, not through a defined base sequence(2, 3). Accordingly, our data (Fig. 3B and 3F) indicate that DNA binding ability, rather than recognition of a specific motif, enables NSP13 to perform its helicase activity in proximity to TEAD and recruit repressors. Moreover, the DNA‑binding mutant K345A/K347A and the ATPase‑dead mutant R567A both fail to suppress YAP/TEAD transcription despite retaining the ability to interact with TEAD (Fig. 3B). These loss‑of‑function phenotypes demonstrate that NSP13’s chromatin engagement and unwinding activity, rather than sequence‑restricted targeting, are essential for repression. For these reasons, motif‑specific binding assays were not pursued in this revision, but we clarified in the discussion that NSP13’s DNA engagement is likely structural or TEAD-dependent, rather than sequence‑directed. We also highlighted this as an important avenue for future investigation.

(3c) To validate the NSP13 interacting proteins from our IP-MS data, we generated plasmids expressing several candidates (CCT3, SMARCD1, EIF4A1, LMNA, TTF2, and YY2) and performed co-IP assays. As predicted, we confirmed the robust interaction between NSP13 and TEAD (Supplemental Fig. 5E). However, these putative nuclear repressors exhibited weak binding to NSP13 compared with TEAD4, suggesting that NSP13 associates with them indirectly, possibly as part of a larger multiprotein complex or depending on the chromatin structure, rather than via direct protein–protein interaction (Fig. 4J).

(3d) We appreciate the reviewer’s question. To investigate whether their association might be DNA‐dependent, we performed co‐IP experiments using nuclear lysates in the presence or absence of various nucleases: Universal Nuclease (which degrades all forms of DNA and RNA), DNase I (which cleaves both single‐ and double‐stranded DNA), and RNase H (which selectively cleaves the RNA strand in RNA/DNA hybrids). Our findings revealed that nucleic acid removal did not disrupt the NSP13/TEAD4 interaction (Supplemental Fig.4E), indicating that their binding is not solely mediated by DNA or RNA.

Reviewer #2 (Public Review):

Specific comments and suggestions for improvement of the manuscript:

(1) NSP13 has been reported to block, in a helicase-dependent manner, episomal DNA transcription (PMID: 37347173), raising questions about the effects observed on the data shown from the HOP-Flash and 8xGTIIC assays. It would be valuable to demonstrate the specificity of the proposed effect of NSP13 on TEAD activation by YAP (versus broad effects on reporter assays) and also to show that NSP13 reduces the function of endogenous YAP-TEAD transcriptional activity (i.e., does ectopic NSP13 expression reduce the expression of YAP induced TEAD target genes in cells).

We appreciate the reviewer’s comments and have carefully revisited the conclusions from the published paper(4) (PMID: 37347173), which reported that NSP13 suppresses episomal DNA transcription, as evidenced by reduced Renilla luciferase (driven by the herpes simplex virus thymidine kinase promoter) and GFP expression upon co‐expression with NSP13. For our experiments, we used a dual‐luciferase assay with Renilla luciferase (under the same promoter) as an internal control. After re-examining our raw Renilla luciferase data (now provided in the supplemental Excel file “Supporting data value”), we found that while 100 ng of NSP13 did not affect Renilla luciferase levels, 400 ng of NSP13 reduced them by approximately 50% relative to the YAP5SA‐only group (Supplemental Fig.2B, Fig.3C-D). We observed a similar reduction with NSP13 truncation mutants—an outcome not fully consistent with the published study (Supplemental Fig.3D, PMID: 37347173). However, unlike their finding of robust episomal DNA suppression, our data indicate that the K345A/K347A mutant of NSP13, which lacks DNA‐binding ability, completely lost its suppressive effect (Fig.3B).

We performed additional Notch reporter assays to address the concern that NSP13 might nonspecifically inhibit episomal DNA transcription (including the HOP‑Flash and 8×GTIIC reporters). These experiments revealed that co‑expression of NSP13 with NICD (Notch intracellular domain) does not suppress Notch signaling (Supplemental Fig. 2C), indicating that NSP13 does not globally block all reporter systems. To evaluate whether NSP13 reduces endogenous YAP‑TEAD activity, we transiently overexpressed NSP13 WT and its R567A mutant in HeLa cells. However, bulk RNA‑seq and qPCR analyses did not reveal a clear decrease in YAP target genes, possibly due to the low transfection efficiency (< 50%, Supplemental Fig.4D). Interestingly, we observed that YAP5SA was predominantly retained in the nucleus upon NSP13 or R567A co‑expression, suggesting that NSP13 (or together with its interacting partners) restricts YAP5SA cytoplasmic shuttling. Future studies will involve stable cell lines expressing NSP13 WT or R567A to better characterize the mechanisms driving YAP5SA nuclear retention and clarify how NSP13 specifically suppresses YAP activity.

(2) While the IP-MS experiment may have revealed new regulators of TEAD activity, the data presented are preliminary and inconclusive. No interactions are validated and beyond slight changes in TEAD reporter activity following knockdown, no direct links to YAP-TEAD are demonstrated, and no link to NPS13 was shown. Also, no details are provided about the methods used for the IP-MS experiment, raising some concerns about potential false positive associations within the data.

We appreciate the reviewer’s feedback regarding our IP-MS findings and acknowledge that additional validation is required to establish definitive links between the identified putative regulators, YAP-TEAD, and NSP13. We have taken the following steps (and plan further experiments) to address these concerns:

(2a) Co-IP validation: Same with the answer for Reviewer #1 (3c), we generated plasmids expressing several top candidate interactors from the IP-MS data (CCT3, SMARCD1, EIF4A1, LMNA, TTF2, and YY2) and performed direct co-IP assays in a more controlled setting. The results indicated that these putative NSP13 interactors had weaker binding compared to TEAD4, implying that NSP13 may associate with them as part of a larger complex or depending on the chromatin structure rather than through a direct protein–protein interaction (Fig. 4J).

(2b) qPCR validation: Beyond reporter assays for evaluating YAP transactivation after the candidate YAP suppressor knockdown (Fig. 4H and Supplemental Fig. 5C), we performed qPCR to detect YAP activation on endogenous YAP-TEAD target genes (e.g., CTGF CYR61, and AMOTL2) after CCT3 knockdown. Expression of CTGF and CYR61 was higher compared to control (Supplemental Fig. 5D), strengthening the case for an interaction relevant to YAP-TEAD signaling.

(2c) To investigate how NSP13‐interacting proteins link to the YAP/TEAD complex, we examined the IP‑MS dataset and identified several well‐known YAP and TEAD binding partners, including CTBP1/2 (TEAD‐binding), GATA4 (TEAD‐binding), and multiple 14‐3‐3 isoforms (YWHAZ/YWHAB/YWHAH/YWHAQ, YAP binding). These findings suggest that NSP13 may form a larger nuclear complex with YAP/TEAD and associated cofactors. In the future, we will determine whether these putative TEAD regulators also interact with NSP13 under various conditions (e.g., in the presence or absence of DNA) and whether co‐expression of NSP13 influences their association with YAP or TEAD. This approach will clarify how NSP13 might leverage these factors to regulate YAP‐TEAD function.

(2e) For the mass spectrometry experiments, HEK293T cells were transfected with Flag‐YAP1, HA‐NSP13, or Flag‐YAP1 + HA‐NSP13 according to the manufacturer’s standard protocols. After nuclear extraction and lysis, the supernatant was incubated with HA magnetic beads to immunoprecipitate (IP) NSP13. The IP samples were subsequently analyzed by mass spectrometry to identify NSP13‐associated proteins (Fig. 4F). Each experimental condition was performed in duplicate to ensure reproducibility. We included an appropriate negative control (Flag‐YAP1) and stringent data‐filtering criteria to minimize false positives. We apologize for not including these details in our original Methods section; in this revised manuscript, we have fully described the number of replicates, the controls used, and our data analysis pipelines.

Reviewer #2 (Public review):

Summary:

This study explores how gene regulatory networks that include intra- and extracellular signaling can give rise to spatial patterns of gene expression in cells. The authors investigate this question in a simplified theoretical framework, where all cells are assumed to respond identically to signals, and spatial details such as cell boundaries and extensions are abstracted away. Within this setting, they identify three distinct signaling topologies, referred to as L and H types, and combine them into three minimal subnetworks capable of generating patterns. The study analyzes possible combinations of these topologies and examines how each subnetwork behaves under three different initial conditions. Combining the analyses with mathematical proofs and heuristic arguments, the authors define necessary conditions under which such networks can produce non-trivial spatial patterns.

Strengths:

The authors break down larger gene regulatory networks into smaller subnetworks, which allows for a more tractable analysis of pattern formation. These minimal subnetworks are examined under different initial conditions, providing a range of examples for how patterns can emerge in simplified settings. The study also proposes necessary conditions for pattern formation, which may be useful for identifying relevant network structures. In addition, the manuscript offers heuristic explanations for the emergence of patterns in each subnetwork, which help to interpret the simulation results and analytical criteria.

Weaknesses:

(1) We have serious concerns regarding the validity of the simulation results presented in the manuscript. Rather than simulating the full nonlinear system described by Equation (1), the authors base their results on a truncated expansion (Equation S.8.2) that captures only the time evolution of small deviations around a spatially homogeneous steady state. However, it remains unclear how this reduced system is derived from the full equations - specifically, which terms are retained or neglected and why - and how the expansion of the nonlinear function can be steady-state independent, as claimed. Additionally, in simulations involving the spike plus homogeneous initial condition, it is not evident - or, where equations are provided, it is not correct - that the assumed global homogeneous background actually corresponds to a steady state of the full dynamics. We elaborate on these concerns in the following:

It is assumed that the homogeneous steady states are given by g_i=0 and g_i=c_i, where 1/c_i = \mu_i or \hat{\mu}_i​, independently of the specific network structure. However, the basis for this assumption is unclear, especially since some of the functions do not satisfy this condition - for example, f5​ as defined below Eq. S8.10.5. Moreover, if g_i=c_i does not correspond to a true steady state, then the time evolution of deviations from this state is not correctly described by Eq. S8.2, as the zeroth-order terms do not vanish in that case.

Additionally, the equations used contain only linear terms and a cubic degradation term for each species g_i, while neglecting all quadratic terms and cubic terms involving cross-species interactions (i≠j). An explanation for this selective truncation is not provided, and without knowledge of the full equation (f), it is impossible to assess whether this expansion is mathematically justified. If, as suggested in the Supplementary Information, the linear and cubic terms are derived from f, then at the very least, the Jacobian matrix should depend on the background steady-state concentration. However, the equations for the small deviation around a steady state (including the Jacobian matrix) used in the simulations appear to be independent of the particular steady state concentration.

This is why we believe that the differences observed between the spike-only initial condition and the spike superimposed on a homogeneous background are not due to the initial conditions themselves, but rather result from a modified reaction scheme introduced through a questionable cutoff.

"In simulations with spike initial patterns, the reference value g≡0 represents an actual concentration of 0 and therefore, we must add to (S8.2) a Heaviside function Φ acting of f (i.e., Φ(f(g))=f(g) if f(g)>0 , Φ(f(g))=0 if f(g){less than or equal to}0 ) to prevent the existence of negative concentrations for any gene product (i.e., g_i<0 for some i )." (SI chapter S8).

This cutoff alters the dynamics (no inhibition) and introduces a different reaction scheme between the two simulations. The need for this correction may itself reflect either a problem in the original equations (which should fulfill the necessary conditions and prevent negative concentrations (R4 in main text)) or the inappropriateness of using an expanded approximation which assumes independence on the steady state concentration. It is already questionable if the linearized equations with a cubic degradation term are valid for the spike initial conditions (with different background concentration values), as the amplitude of this perturbation seems rather large.

Lastly, we note that under the current simulation scheme, it is not possible to meaningfully assess criteria RH2a and RH2b, as they rely on nonlinear interactions that are absent from the implemented dynamics.

(2) Most of the proofs presented in the Supplementary Information rely on linearized versions of the governing equations, and it remains unclear how these results extend to the fully nonlinear system. We are concerned that the generality of the conclusions drawn from the linear analysis may be overstated in the main text. For example, in Section S3, the authors introduce the concept of dynamic equivalence of transitive chains (Proposition S3.1) and intracellular transitive M-branching (Proposition S3.2), which pertains to the system's steady-state behavior. However, the proof is based solely on the linearized equations, without additional justification for why the result should hold in the presence of nonlinearities. Moreover, the linearized system is used to analyze the response to a "spike initial pattern of arbitrary height C" (SI Chapter S5.1), yet it is not clear how conclusions derived from the linear regime can be valid for large perturbations, where nonlinear effects are expected to play a significant role. We encourage the authors to clarify the assumptions under which the linearized analysis remains valid and to discuss the potential limitations of applying these results to the nonlinear regime.

(3) Several statements in the main text are presented without accompanying proof or sufficient explanation, which makes it difficult to assess their validity. In some cases, the lack of justification raises serious doubts about whether the claims are generally true. Examples are:

"For the purpose of clarity we will explain our results as if these cells have a simple arrangement in space (e.g., a 1D line or a 2D square lattice) but, as we will discuss, our results shall apply with the same logic to any distribution of cells in space." (Main text l.145-l.148).

"For any non-trivial pattern transformation (as long as it is symmetric around the initial spike), there exists an H gene network capable of producing it from a spike initial pattern." (Main text l.366f).

"In 2D there are no peaks but concentric rings of high gene product concentration centered around the spike, while in 3D there are concentric spherical shells." (Main text l. 447ff).

(4) The study identifies one-signal networks and examines how combinations of these structures can give rise to minimal pattern-forming subnetworks. However, the analysis of the combinations of these minimal pattern-forming subnetworks remains relatively brief, and the manuscript does not explore how the results might change if the subnetworks were combined in upstream and downstream configurations. In our view, it is not evident that all possible gene regulatory networks can be fully characterized by these categories, nor that the resulting patterns can be reliably predicted. Rather, the approach appears more suited to identifying which known subnetworks are present within a larger network, without necessarily capturing the full dynamics of more complex configurations.

(5) The definition of non-trivial pattern formation is provided only in the Supplementary Information, despite its central importance for interpreting the main results. It would significantly improve clarity if this definition were included and explained in the main text. Additionally, it remains unclear how the definition is consistently applied across the different initial conditions. In particular, the authors should clarify how slope-based measures are determined for both the random noise and sharp peak/step function initial states. Furthermore, the authors do not specify how the sign function is evaluated at zero. If the standard mathematical definition sgn(0)=0 is used, then even a simple widening of a peak could fulfill the criterion for non-trivial pattern transformation.

(6) The manuscript lacks a clear and detailed explanation of the underlying model and its assumptions. In particular, it is not well-defined what constitutes a "cell" in the context of the model, nor is it justified why spatial features of cells - such as their size or boundaries - can be neglected. Furthermore, the concept of the extracellular space in the one-dimensional model remains ambiguous, making it unclear which gene products are assumed to diffuse.

Reviewer #2 (Public review):

Summary:

In this work, the authors introduce DIRseq, a fast, sequence-based method that predicts drug-interacting residues (DIRs) in IDPs without requiring structural or drug information. DIRseq builds on the authors' prior work looking at NMR relaxation rates, and presumes that those residues that show enhanced R2 values are the residues that will interact with drugs, allowing these residues to be nominated from the sequence directly. By making small modifications to their prior tool, DIRseq enables the prediction of residues seen to interact with small molecules in vivo.

Strengths:

The preprint is well written and easy to follow

Weaknesses:

(1) The DIRseq method is based on SeqDYN, which itself is a simple (which I do not mean as a negative - simple is good!) statistical predictor for R2 relaxation rates. The challenge here is that R2 rates cover a range of timescales, so the physical intuition as to what exactly elevated R2 values mean is not necessarily consistent with "drug interacting". Presumably, the authors are not using the helix boost component of SeqDYN here (it would be good to explicitly state this). This is not necessarily a weakness, but I think it would behove the authors to compare a few alternative models before settling on the DIRseq method, given the somewhat ad hoc modifications to SeqDYN to get DIRseq.

Specifically, the authors previously showed good correlation between the stickiness parameter of Tesei et al and the inferred "q" parameter for SeqDYN; as such, I am left wondering if comparable accuracy would be obtained simply by taking the stickiness parameters directly and using these to predict "drug interacting residues", at which point I'd argue we're not really predicting "drug interacting residues" as much as we're predicting "sticky" residues, using the stickiness parameters. It would, I think, be worth the authors comparing the predictive power obtained from DIRseq with the predictive power obtained by using the lambda coefficients from Tesei et al in the model, local density of aromatic residues, local hydrophobicity (note that Tesei at al have tabulated a large set of hydrophobicity scores!) and the raw SeqDYN predictions. In the absence of lots of data to compare against, this is another way to convince readers that DIRseq offers reasonable predictive power.

(2) Second, the DIRseq is essentially SeqDYN with some changes to it, but those changes appear somewhat ad hoc. I recognize that there is very limited data, but the tweaking of parameters based on physical intuition feels a bit stochastic in developing a method; presumably (while not explicitly spelt out) those tweaks were chosen to give better agreement with the very limited experimental data (otherwise why make the changes?), which does raise the question of if the DIRseq implementation of SeqDYN is rather over-parameterized to the (very limited) data available now? I want to be clear, the authors should not be critiqued for attempting to develop a model despite a paucity of data, and I'm not necessarily saying this is a problem, but I think it would be really important for the authors to acknowledge to the reader the fact that with such limited data it's possible the model is over-fit to specific sequences studied previously, and generalization will be seen as more data are collected.

(3) Third, perhaps my biggest concern here is that - implicit in the author's assumptions - is that all "drugs" interact with IDPs in the same way and all drugs are "small" (motivating the change in correlation length). Prescribing a specific lengthscale and chemistry to all drugs seems broadly inconsistent with a world in which we presume drugs offer some degree of specificity. While it is perhaps not unexpected that aromatic-rich small molecules tend to interact with aromatic residues, the logical conclusion from this work, if one assumes DIRseq has utility, is that all IDRs bind drugs with similar chemical biases. This, at the very least, deserves some discussion.

(4) Fourth, the authors make some general claims in the introduction regarding the state of the art, which appear to lack sufficient data to be made. I don't necessarily disagree with the author's points, but I'm not sure the claims (as stated) can be made absent strong data to support them. For example, the authors state: "Although an IDP can be locked into a specific conformation by a drug molecule in rare cases, the prevailing scenario is that the protein remains disordered upon drug binding." But is this true? The authors should provide evidence to support this assertion, both examples in which this happens, and evidence to support the idea that it's the "prevailing view" and specific examples where these types of interactions have been biophysically characterized.

Similarly, they go on to say:

"Consequently, the IDP-drug complex typically samples a vast conformational space, and the drug molecule only exhibits preferences, rather than exclusiveness, for interacting with subsets of residues." But again, where is the data to support this assertion? I don't necessarily disagree, but we need specific empirical studies to justify declarative claims like this; otherwise, we propagate lore into the scientific literature. The use of "typically" here is a strong claim, implying most IDP complexes behave in a certain way, yet how can the authors make such a claim?

Finally, they continue to claim:

"Such drug interacting residues (DIRs), akin to binding pockets in structured proteins, are key to optimizing compounds and elucidating the mechanism of action." But again, is this a fact or a hypothesis? If the latter, it must be stated as such; if the former, we need data and evidence to support the claim.

Author response:

Reviewer #1 (Public review):

Summary:

The authors developed a sequence-based method to predict drug-interacting residues in IDP, based on their recent work, to predict the transverse relaxation rates (R2) of IDP trained on 45 IDP sequences and their corresponding R2 values. The discovery is that the IDPs interact with drugs mostly using aromatic residues that are easy to understand, as most drugs contain aromatic rings. They validated the method using several case studies, and the predictions are in accordance with chemical shift perturbations and MD simulations. The location of the predicted residues serves as a starting point for ligand optimization.

Strengths:

This work provides the first sequence-based prediction method to identify potential drug-interacting residues in IDP. The validity of the method is supported by case studies. It is easy to use, and no time-consuming MD simulations and NMR studies are needed.

Weaknesses:

The method does not depend on the information of binding compounds, which may give general features of IDP-drug binding. However, due to the size and chemical structures of the compounds (for example, how many aromatic rings), the number of interacting residues varies, which is not considered in this work. Lacking specific information may restrict its application in compound optimization, aiming to derive specific and potent binding compounds.

We fully recognize that different compounds may have different interaction propensity profiles along the IDP sequence. In future studies, we will investigate compound-specific parameter values. The limiting factor is training data, but such data are beginning to be available.

Reviewer #2 (Public review):

Summary:

In this work, the authors introduce DIRseq, a fast, sequence-based method that predicts drug-interacting residues (DIRs) in IDPs without requiring structural or drug information. DIRseq builds on the authors' prior work looking at NMR relaxation rates, and presumes that those residues that show enhanced R2 values are the residues that will interact with drugs, allowing these residues to be nominated from the sequence directly. By making small modifications to their prior tool, DIRseq enables the prediction of residues seen to interact with small molecules in vivo.

Strengths:

The preprint is well written and easy to follow

Weaknesses:

(1) The DIRseq method is based on SeqDYN, which itself is a simple (which I do not mean as a negative - simple is good!) statistical predictor for R2 relaxation rates. The challenge here is that R2 rates cover a range of timescales, so the physical intuition as to what exactly elevated R2 values mean is not necessarily consistent with "drug interacting". Presumably, the authors are not using the helix boost component of SeqDYN here (it would be good to explicitly state this). This is not necessarily a weakness, but I think it would behove the authors to compare a few alternative models before settling on the DIRseq method, given the somewhat ad hoc modifications to SeqDYN to get DIRseq.

Actually, the factors that elevate R2 are well-established. These are local interactions and residual secondary structures (if any). The basic assumption of our method is that intra-IDP interactions that elevate R2 convert to IDP-drug interactions. This assumption was supported by our initial observation that the drug interaction propensity profiles predicted using the original SeqDYN parameters already showed good agreement with CSP profiles. We only made relatively small adjustments to the parameters to improve the agreement. Indeed we did not apply the helix boost portion of SeqDYN to DIRseq, and will state as such. We will also compare DIRseq with several alternative models.

Specifically, the authors previously showed good correlation between the stickiness parameter of Tesei et al and the inferred "q" parameter for SeqDYN; as such, I am left wondering if comparable accuracy would be obtained simply by taking the stickiness parameters directly and using these to predict "drug interacting residues", at which point I'd argue we're not really predicting "drug interacting residues" as much as we're predicting "sticky" residues, using the stickiness parameters. It would, I think, be worth the authors comparing the predictive power obtained from DIRseq with the predictive power obtained by using the lambda coefficients from Tesei et al in the model, local density of aromatic residues, local hydrophobicity (note that Tesei at al have tabulated a large set of hydrophobicity scores!) and the raw SeqDYN predictions. In the absence of lots of data to compare against, this is another way to convince readers that DIRseq offers reasonable predictive power.

We will compare predictions of these various parameter sets, and summarize the results in a table.

(2) Second, the DIRseq is essentially SeqDYN with some changes to it, but those changes appear somewhat ad hoc. I recognize that there is very limited data, but the tweaking of parameters based on physical intuition feels a bit stochastic in developing a method; presumably (while not explicitly spelt out) those tweaks were chosen to give better agreement with the very limited experimental data (otherwise why make the changes?), which does raise the question of if the DIRseq implementation of SeqDYN is rather over-parameterized to the (very limited) data available now? I want to be clear, the authors should not be critiqued for attempting to develop a model despite a paucity of data, and I'm not necessarily saying this is a problem, but I think it would be really important for the authors to acknowledge to the reader the fact that with such limited data it's possible the model is over-fit to specific sequences studied previously, and generalization will be seen as more data are collected.

We have explained the rationale for the parameter tweaks, which were limited to q values for four amino-acid types, i.e., to deemphasize hydrophobic interactions and slightly enhance electrostatic interactions (p. 4-5). We will add that these tweaks were motivated by observations from MD simulations of drug interactions with a-syn (ref 20). As already noted in the response to the preceding comment, we will also present results for the original parameter values as well as for when the four q values are changed one at a time.

(3) Third, perhaps my biggest concern here is that - implicit in the author's assumptions - is that all "drugs" interact with IDPs in the same way and all drugs are "small" (motivating the change in correlation length). Prescribing a specific lengthscale and chemistry to all drugs seems broadly inconsistent with a world in which we presume drugs offer some degree of specificity. While it is perhaps not unexpected that aromatic-rich small molecules tend to interact with aromatic residues, the logical conclusion from this work, if one assumes DIRseq has utility, is that all IDRs bind drugs with similar chemical biases. This, at the very least, deserves some discussion.

The reviewer raises a very important point. In Discussion, we will add that it is important to further develop DIRseq to include drug-specific parameters when data for training become available.

(4) Fourth, the authors make some general claims in the introduction regarding the state of the art, which appear to lack sufficient data to be made. I don't necessarily disagree with the author's points, but I'm not sure the claims (as stated) can be made absent strong data to support them. For example, the authors state: "Although an IDP can be locked into a specific conformation by a drug molecule in rare cases, the prevailing scenario is that the protein remains disordered upon drug binding." But is this true? The authors should provide evidence to support this assertion, both examples in which this happens, and evidence to support the idea that it's the "prevailing view" and specific examples where these types of interactions have been biophysically characterized.

We will cite several studies showing that IDPs remain disordered upon drug binding.

Similarly, they go on to say:

"Consequently, the IDP-drug complex typically samples a vast conformational space, and the drug molecule only exhibits preferences, rather than exclusiveness, for interacting with subsets of residues." But again, where is the data to support this assertion? I don't necessarily disagree, but we need specific empirical studies to justify declarative claims like this; otherwise, we propagate lore into the scientific literature. The use of "typically" here is a strong claim, implying most IDP complexes behave in a certain way, yet how can the authors make such a claim? 

Here again we will add citations to support the statement.

Finally, they continue to claim:

"Such drug interacting residues (DIRs), akin to binding pockets in structured proteins, are key to optimizing compounds and elucidating the mechanism of action." But again, is this a fact or a hypothesis? If the latter, it must be stated as such; if the former, we need data and evidence to support the claim. 

We will add citations to both compound optimization and mechanism of action.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

This manuscript addresses the question of whether inhibitors of the phosphatases Eya1-4 and of the kinase PLK1 provide an effective therapeutic approach to a range of cancers. Both Eyas and PLK1 have well documented roles in development, and have been implicated in a subset of tumors. Moreover, the authors have previously shown that PLK1 is a substrate of Eya phosphatase activity. Building on these previous findings, the authors assess the possibility of combining an Eya inhibitor, benzarone, with a PLK1 inhibitor, BI2536.

There are several concerns with the study: 1. The authors suggest that these two drugs are synergistic. Synergy is usually taken as indicative of a greater than additive effect of the two drugs. The ZIP synergy score tested here indicates that the combination of the two drugs has a synergy score between 0 and 10 (figure1, and figure 5). According to "Synergy Finder", "A ZIP synergy score of greater than 10 often indicates a strong synergistic effect, while a score less than -10 suggests a strong antagonistic effect. Scores between -10 and 10 are typically considered additive or near-additive." The data in figure 2 on mitotic cell fraction and on cell death also seems to be more of an additive effect of the two drugs than synergy. The data in figure 3 are also additive effects on RAD51. Therefore a conclusion that "These data indicate that the drug combination was broadly synergistic" seems unwarranted.

There is a general lack of nomenclature standardisation for defining synergy. Furthermore, multiple synergy models exist, with discrepancies between them. However, as the reviewer states, the prevailing view is that synergy is a combination effect that is stronger than the additive effect of the two drugs. Synergy scores derived from dose-response matrices using different synergy scoring models with scores that fall above 5 are considered truly synergistic (Malyutina A et al., 2019). To strengthen our conclusion of synergy between PLK1 and EYA inhibitors, we have calculated synergy scores using additional synergy models for both benzarone + BI2536 and benzarone + volasertib in H4 and T98G cell lines. Specifically, we find robust synergy (>5) using ZIP, HSA and Bliss calculations with the Benzarone + BI2536 drug combination in H4 cells and with Benzarone + Volasertib in H4 and T98G cells. Synergy scores for Benzarone + BI2536 fell just below 5 in T98G cells. These data are now included in Supplemental Fig S1G of the revised manuscript.

The discovery of synergistic drug combinations can be further strengthened by evaluating synergy across multiple cellular models. In this study, we have tested a total of 27 different cancer models that universally support synergy.

Regarding the phenotypic outcomes (mitotic cell fraction, cell death, RAD51 foci), we agree that the observed effects are additive. This is consistent with overall synergistic effects on viability being caused by a combination of additive mechanistic effects. We have amended the text in the revised manuscript to clarify this point.

There was no statistical difference in the synergy scores of the "high expressing" versus "low expressing cells". So the conclusion that the drug combination "t was effective at lower doses in cell lines with high levels of EYA1 and/or EYA4" seems unwarranted based on the data. Moreover, since there was no statistical difference in synergy between high and low expressing cells, stating that "the potential utility of the combination treatment depends on the specific overexpression of EYA1 and/or EYA4 in cancer cells," seems unwarranted by the data.

Synergy scores quantify the interaction between drugs, but do not capture absolute treatment effectiveness or dose sensitivity, both of which are crucial for therapeutic considerations. We have included the following sentence in the revised manuscript to clarify this distinction: “While synergy scores did not significantly differ between high and low EYA expressors, high EYA1/4 expression was associated with increased sensitivity to the combination treatment at lower doses, as evidenced by decreased cell viability.” We have also amended the conclusions in the Abstract and Discussion to reflect that the potential utility of the combination therapy in EYA1/4-high cancers is supported by potency rather than synergy scores alone.

Benzarone and benzbromarone and their derivatives have been shown to bind and inhibit Eya phosphatases, albeit at fairly high doses. However, these two compounds also have a number of other, unrelated targets. The only demonstration that Eyas are a target of benzarone in this study are the CETSA data in supplemental figure 1. The data here seem to represent an n of 1, with no error bars shown. Even more importantly, there is no control. Looking at the blot of actin, it seems as if there may be a benzarone- temperature effect on this protein as well. It would be very helpful to show some evidence that knockdown of Eya similarly synergizes with the PLK1 inhibitor, show data that benzarone is in fact inhibiting Eya activity in these cells by looking at known targets (ie the carboxyterminal tyrosine of H2AX), and other evidence of specificity.

The specificity of benzarone to the EYA proteins has been demonstrated previously using both in vitro phosphatase assays and the assessment of EYA-mediated pathways (Tadjuidje et al., 2012; Wang et al., 2021; Nelson et al., 2024). These publications have been cited in the manuscript. In addition, benzarone produces phenotypes consistent with the known functions of the EYAs (ie, reduction of PLK1 activity, reduction in RAD51 foci, G2/M arrest, and apoptosis). To further validate EYA target specificity, we have performed viability assays on control and EYA4-depleted HeLa, H4 and T98G cells in response to BI2536 treatment, demonstrating EYA4 depletion-mediated sensitization to BI2536. These data are now included in Fig 1H of the revised manuscript.

To strengthen our CETSA data, we have now included: (i) densitometry of actin, demonstrating a lack of benzarone-temperature effect, (ii) CETSA analysis for an additional cell line (T98G), demonstrating enhanced thermal stability of the EYAs in the presence of benzarone, and (iii) CETSA analysis of an additional protein (BUB1) to demonstrate target specificity. These data are now included in Supplemental Fig S1E and F of the revised manuscript.

The proteomic and transcriptomic data of cell lines that were vulnerable to the combination of BI2536 and benzarone implicate overall changes in chromatin with sensitivity. These findings call into question the idea that these two compounds are acting selectively on PLK1 and Eyas. The authors don't really provide any model for explaining this correlation of Nurd complex components with targeting Eyas and PLK1.

The proteomic and transcriptomic data demonstrate that sensitivity to the combination treatment is associated with higher expression of NuRD complex members and other chromatin regulators. This suggests that cell lines with certain chromatin configurations might be more susceptible to the combined inhibition of PLK1 and EYA. This does not undermine the demonstrated on-target effects of the two compounds, but rather suggests a potential contextual dependence of drug efficacy on chromatin state. Our data thereby implicate NuRD complex expression as a predictive biomarker for tumours that are likely to respond to EYA and PLK1 combination therapy. This has now been clarified in the discussion section of the revised manuscript.

Specificity of antibodies: I would like to see validation of the Eya antibodies, given the difficulty with such reagents in the field.

All EYA antibodies have now been validated by western blot analysis following siRNA-mediated depletion. These data are presented in Supplemental Fig S1A of the revised manuscript.

Reviewer #1 (Significance (Required)):

New therapies targeting glioblastoma would be welcome. It is not clear that the combination tested here is an effective approach to therapy. It would be necessary to know the targets of the combination and understand the mechanism so that the approach could be pursued further,

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

This study explores the sensitivity of cancer cell lines, particularly GBM cells, to dual inhibition of EYA and PLK1, aiming to uncover the connection between these pathways and the cancer stem cell state. Additionally, it investigates whether the NuRD complex modulates GBM cell responses to EYA and PLK1 inhibition. While the findings are interesting, further clarification is needed to establish the mechanistic links between EYA, PLK1, and NuRD, as well as a stronger rationale for their targeted inhibition in GBM therapy- this can be better clarified.

Some key comments and recommendations: The findings demonstrate that the combination of Benzarone (EYAi) and volasertib (PLKi) significantly reduced cell proliferation in H4 and T98G GBM cell lines, both of which show high expression of EYA. In contrast, the low EYA-expressing A172 cells exhibited limited response. A possible explanation is the inherently slower proliferation rate of A172 cells, which may reduce their dependence on G2/M arrest, thereby diminishing the impact of PLK1i. Does A172 line show a similar growth or cell division rate to H4 and T98G lines.

A172 cells have a slower proliferation rate than H4 or T98G cells, which may diminish their response to EYA/PLK1 inhibitors. However, in this study we have tested a total of 15 cancer cell lines and 12 GBM stem cell line models. No clear correlation between cell growth rate and sensitivity was observed. As a specific example, the low EYA expressing SJSA-1 cell line has a high proliferation rate but is a low responder to EYA1/PLK1 inhibitors.

Additionally, although protein expression levels of EYA were assessed across these cell lines, the activity and expression levels of PLK1 were not fully characterized. Since PLK1 is a crucial regulator of mitotic entry and DNA damage repair, its activity across cell lines may contribute to the observed variations in drug sensitivity. Could the authors investigate levels of PLK in these cell lines?

To address this point, we compared PLK1 expression levels across the panel of cancer cell lines used in our study. These data are now included in Supplemental Fig S1D of the revised manuscript, and show that PLK1 levels are comparable across the cell lines, indicating that baseline PLK1 abundance does not fully explain the observed differential sensitivity.

The study describes the combination treatment as synergistic in H4 and T98G cells, however this synergy is unclear in Fig 2A and Supplemental Fig S2A. The data suggest that H4 and T98G cells exhibit sensitivity to either EYA or PLK1 inhibition alone, with combined treatment showing enhanced effects rather than synergy. This distinction is evident as BI2536 alone induces robust G2/M arrest with decreased G1 and S phase cells. To validate these findings, combination treatment should be tested in additional GBM cell lines. Additionally, repeating FUCCI cell cycle assays in A172 and H4 cells, particularly in H4, where increased γH2AX and phospho-H3 were detected in response to individual inhibitors, would provide more definitive insights into treatment-induced cell cycle dynamics.

We agree that several of the phenotypic outcomes, for example G2/M arrest (Fig 2A) and micronuclei formation (Supplemental Fig S2A), produce additive rather than synergistic effects in the combination treated cells. The major claim of the study is that the combination treatment results in potent loss of cell viability in EYA1/EYA4 overexpressing cancer cell models. This is consistent with a combination of additive mechanistic effects causing overall synergistic effects on cancer cell viability. We have clarified this point in the revised manuscript.

We have previously struggled to get adequate FUCCI sensor expression in H4 cells. However, to address this point, we have quantified cell cycle phase distribution in H4 cells treated with benzarone, BI2536, and the drug combination, using our quantitative image-based cytometry data (Fig 3A, B). These data demonstrate an accumulation of H4 cells in G2/M following combination treatment, consistent with the FUCCI data from T98G cells. Cell cycle dynamics of H4 cells are now included in Supplemental Fig S2A of the revised manuscript.

A notable inconsistency: Figure 1 utilizes volasertib, whereas Figure 2 employs BI2536. Given that both inhibitors target PLK1 why these specific inhibitors were chosen for each experiment.

This is not the case. To clarify, BI2536 is used in both Fig 1 and 2. Volasertib is used in Supplemental Fig S1 to reproduce the synergy matrix, thereby demonstrating consistent results with a second PLK1 inhibitor.

The observation of increased Rad52 foci and sister chromatid exchange (SCE) upon EYA and PLK1 inhibition (Figure 3) is interesting. These findings suggest that dual inhibition impairs homologous recombination (HR), reinforcing the role of EYA and PLK1 in maintaining genomic stability.

We agree.

Figure 4 suggests that SJH1 cells, with low EYA expression, exhibit increased sensitivity to EYA inhibition - does this cell line show high expression of PLK or NuRD?

To clarify, Fig 4 shows that SJH1 cells, which display moderate levels of EYA expression, are highly sensitive to EYA/PLK1 inhibition. Consistent with the observed positive correlation between NuRD protein expression and EYA/PLK1 inhibitor sensitivity, SJH1 cells exhibit the highest levels of NuRD components relative to the other GBM stem cell lines. Expression levels of NuRD components across the slightly sensitive, moderately sensitive, and highly sensitive GBM stem cell lines from publicly available proteomic data and western blot analysis have now been included in Supplemental Fig S5A and B of the revised manuscript, further demonstrating this positive correlation.

It seems like EYA1 (HW1) and EY4 (SB2B and PB1) expression levels are better predictors of sensitivity to treatment, but not EYA2 and 3 (which is high in H4)- can the authors comment on this?

Overall, EYA1 and EYA4 expression levels are the major predictors of EYA/PLK1 inhibitor sensitivity in both the cancer cell lines (Fig 1) and the GBM stem cell models (Fig 4). EYA3 levels are also positively associated with sensitivity in the GBM stem cell models, but not in the cancer cell lines. Despite being consistently high, EYA2 expression levels were not associated with sensitivity in either model. These intricacies are likely to reflect functional differences between the proteins, and their ability to form different sub-complexes with each other. We have now clarified these points in the discussion of the revised manuscript.

Reviewer #2 (Significance (Required)):

It remains unclear whether NuRD complex involvement is independent of EYA expression levels. Since EYA and PLK1 regulate cell cycle progression and DNA repair, further investigation is needed to delineate their connection to NuRD-mediated chromatin remodeling and differentiation programs. Overall, this study provides some interesting evidence for targeting transcriptional and mitotic vulnerabilities in GBM but requires further validation of synergistic mechanisms, differential inhibitor effects, and NuRD complex involvement in regulating the EYA-PLK1 axis.

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

This manuscript extends the findings of the interactions between EYA family members and PLK1. The idea to combine EYA inhibitors and PLK1 inhibitors is a thoughtful approach. The effects on proliferation and DNA damage are useful. This effort is a combination of preclinical efforts and some mechanistic efforts and will require additional efforts to support the conclusions drawn.

Major concerns: 1. The preclinical studies will absolutely require in vivo studies. All brain tumor treatments are limited by delivery across the blood-brain barrier. It is critical to have intracranial survival studies to support the significance of the findings.

In this study, we have focused on in vitro models including cancer cell lines, GBM stem cell models and 3D tumor spheroids, to establish proof-of-principle as well as mechanistic insight for combined EYA/PLK1 inhibition. We recognize that blood-brain-barrier penetration and therapeutic efficacy in vivo are key translational steps; however, we feel that benzarone is a suboptimal drug candidate for in vivo evaluation. Future development of second-generation EYA inhibitors with higher potency, improved selectivity, and better blood-brain-barrier permeability, is currently underway by ourselves and other groups. These compounds are likely to be more suitable for future in vivo studies, including pharmacokinetic profiling, blood-brain-barrier penetration assays, and orthotopic intracranial tumour models to assess their therapeutic potential more rigorously.

Likewise, cancer stem cell studies require in vivo studies.

As outlined above, we feel that in vivo studies fall beyond the scope of this study.

The proper studies of sphere formation would include in vitro limiting dilution assays. I would suggest greater depth in stem cell and differentiation marker studies to understand what the connection to stemness is.

The limiting dilution assay is used to measure the self-renewal potential of cancer stem cells, and would be used in this context to determine whether the treatments impact cellular differentiation. This is not the focus of this study. Rather, we are interested in comparing drug sensitivity in cancer stem cells versus differentiated cancer cells. Nevertheless, this is a great suggestion for future investigation as part of a more detailed evaluation of stemness and how these drugs and drug combinations impact self-renewal.

DNA damage responses differ between cancer stem cells and differentiated tumor cells. I would suggest comparison of effects between matched cells with different cell states.

We agree that cancer stem cells and their differentiated counterparts often display distinct DNA damage responses. We have tried to mimimise the impact of these differences on the overall conclusions by using multiple cancer cell lines and GBM stem models. To address this comment, we performed western blot analysis of DNA damage response proteins in matched PB1 stem cells and differentiated cells, demonstrating comparable expression of DNA damage response proteins. These data have now been included in Supplemental Fig S5C of the revised manuscript.

While the inhibitors used may have general specificity for the molecular targets, I would suggest that the authors use genetic loss-of-function and gain-of-function studies to validate the findings. It is particularly important because the primary targets do not predict treatment responses. I would suggest that rescues with PLK1 phosphorylation mutants would be helpful.

Our data demonstrate that EYA expression levels are predictive of treatment response in both cancer cell lines and GBM stem cell models. To further validate EYA target specificity, we have used a genetic loss-of-function approach. Specifically, we performed viability assays on control and EYA4-depleted HeLa, H4 and T98G cells in response to BI2536 treatment, demonstrating EYA4 depletion-mediated sensitization to BI2536. These data are now included in Fig 1H of the revised manuscript.

We have previously performed comprehensive rescue experiments with PLK1 phosphorylation mutants (Fig 5C–K; Nelson et al., Nat. Commun. 2024). These experiments demonstrated that cell death in response to EYA depletion or inhibition is attributable to the phosphorylation status of pY445 on PLK1, with an accumulation of Y445 phosphorylation reducing PLK1 activity and functionality, culminating in the potent induction of mitotic cell death.

Figure 5 should be performed with several lines across different response groups.

Our study currently includes cell viability and proliferation data from multiple models including 15 cancer cell lines and 12 GBM stem cell line models, spanning different EYA expression levels, and displaying varying sensitivities to both single agents and the EYA/PLK1 combination treatment. We then narrowed the number of models significantly for follow-up analysis. In Fig 5, we selected the highly sensitive PB1 GBM stem cell line based on its ability to form and grow as spheroids. While we appreciate the suggestion to expand these analyses to additional lines, we would like to respectfully decline growing additional spheroids at this time due to limitations inherent in the expansion of these models. We believe that the current dataset adequately demonstrates the reproducibility and relevance of our findings across different response groups.

The molecular associations are currently just associations. I would suggest greater analysis using genetic manipulation to test causation.

To address this concern, we have performed additional experiments using siRNA-mediated knockdown of EYA4 in HeLa, H4 and T98G cells. These experiments demonstrate that depletion of EYA4 sensitizes cells to PLK1 inhibition, mimicking the effects observed with pharmacological EYA inhibition. These data have been included in Fig 1H of the revised manuscript, and provide additional functional evidence supporting a causal relationship between EYA activity and sensitivity to PLK1 inhibition.

Figure 6 should be better developed to include protein testing and validation.

To address this point, expression levels of NuRD components have been compared using publicly available proteomic datasets and western blot analysis across the slightly sensitive, moderately sensitive and highly sensitive GBM stem cell lines, supporting a positive correlation with sensitivity. These data have been included in Supplemental Fig S5A and B of the revised manuscript.

Reviewer #3 (Significance (Required)):

This is a modest advance in understanding how EYA family members may function with PLK1.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

This manuscript extends the findings of the interactions between EYA family members and PLK1. The idea to combine EYA inhibitors and PLK1 inhibitors is a thoughtful approach. The effects on proliferation and DNA damage are useful. This effort is a combination of preclinical efforts and some mechanistic efforts and will require additional efforts to support the conclusions drawn.

Major concerns:

1. The preclinical studies will absolutely require in vivo studies. All brain tumor treatments are limited by delivery across the blood-brain barrier. It is critical to have intracranial survival studies to support the significance of the findings.

2. Likewise, cancer stem cell studies require in vivo studies.

3. The proper studies of sphere formation would include in vitro limiting dilution assays. I would suggest greater depth in stem cell and differentiation marker studies to understand what the connection to stemness is.

4. DNA damage responses differ between cancer stem cells and differentiated tumor cells. I would suggest comparison of effects between matched cells with different cell states.

5. While the inhibitors used may have general specificity for the molecular targets, I would suggest that the authors use genetic loss-of-function and gain-of-function studies to validate the findings. It is particularly important because the primary targets do not predict treatment responses. I would suggest that rescues with PLK1 phosphorylation mutants would be helpful.

6. Figure 5 should be performed with several lines across different response groups.

7. The molecular associations are currently just associations. I would suggest greater analysis using genetic manipulation to test causation.

8. Figure 6 should be better developed to include protein testing and validation.

Significance

This is a modest advance in understanding how EYA family members may function with PLK1.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

This study explores the sensitivity of cancer cell lines, particularly GBM cells, to dual inhibition of EYA and PLK1, aiming to uncover the connection between these pathways and the cancer stem cell state. Additionally, it investigates whether the NuRD complex modulates GBM cell responses to EYA and PLK1 inhibition. While the findings are interesting, further clarification is needed to establish the mechanistic links between EYA, PLK1, and NuRD, as well as a stronger rationale for their targeted inhibition in GBM therapy- this can be better clarified.

Some key comments and recommendations:

• The findings demonstrate that the combination of Benzarone (EYAi) and volasertib (PLKi) significantly reduced cell proliferation in H4 and T98G GBM cell lines, both of which show high expression of EYA. In contrast, the low EYA-expressing A172 cells exhibited limited response. A possible explanation is the inherently slower proliferation rate of A172 cells, which may reduce their dependence on G2/M arrest, thereby diminishing the impact of PLK1i. Does A172 line show a similar growth or cell division rate to H4 and T98G lines.

• Additionally, although protein expression levels of EYA were assessed across these cell lines, the activity and expression levels of PLK1 were not fully characterized. Since PLK1 is a crucial regulator of mitotic entry and DNA damage repair, its activity across cell lines may contribute to the observed variations in drug sensitivity. Could the authors investigate levels of PLK in these cell lines?

• The study describes the combination treatment as synergistic in H4 and T98G cells, however this synergy is unclear in Figure 2A and EV 2A. The data suggest that H4 and T98G cells exhibit sensitivity to either EYA or PLK1 inhibition alone, with combined treatment showing enhanced effects rather than synergy. This distinction is evident as BI2536 alone induces robust G2/M arrest with decreased G1 and S phase cells. To validate these findings, combination treatment should be tested in additional GBM cell lines. Additionally, repeating FUCCI cell cycle assays in A172 and H4 cells, particularly in H4, where increased γH2AX and phospho-H3 were detected in response to individual inhibitors, would provide more definitive insights into treatment-induced cell cycle dynamics.

• A notable inconsistency: Figure 1 utilizes volasertib, whereas Figure 2 employs BI2536. Given that both inhibitors target PLK1 why these specific inhibitors were chosen for each experiment.

• The observation of increased Rad52 foci and sister chromatid exchange (SCE) upon EYA and PLK1 inhibition (Figure 3) is interesting. These findings suggest that dual inhibition impairs homologous recombination (HR), reinforcing the role of EYA and PLK1 in maintaining genomic stability.

• Figure 4 suggests that SJH1 cells, with low EYA expression, exhibit increased sensitivity to EYA inhibition - does this cell line show high expression of PLK or NuRD?

• It seems like EYA1 (HW1) and EY4 (SB2B and PB1) expression levels are better predictors of sensitivity to treatment, but not EYA2 and 3 (which is high in H4)- can the authors comment on this?

Significance

It remains unclear whether NuRD complex involvement is independent of EYA expression levels. Since EYA and PLK1 regulate cell cycle progression and DNA repair, further investigation is needed to delineate their connection to NuRD-mediated chromatin remodeling and differentiation programs.

Overall, this study provides some interesting evidence for targeting transcriptional and mitotic vulnerabilities in GBM but requires further validation of synergistic mechanisms, differential inhibitor effects, and NuRD complex involvement in regulating the EYA-PLK1 axis.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

This manuscript addresses the question of whether inhibitors of the phosphatases Eya1-4 and of the kinase PLK1 provide an effective therapeutic approach to a range of cancers. Both Eyas and PLK1 have well documented roles in development, and have been implicated in a subset of tumors. Moreover, the authors have previously shown that PLK1 is a substrate of Eya phosphatase activity. Building on these previous findings, the authors assess the possibility of combining an Eya inhibitor,benzarone, with a PLK1 inhibitor, BI2536.

There are several concerns with the study:

1. The authors suggest that these two drugs are synergistic. Synergy is usually taken as indicative of a greater than additive effect of the two drugs. The ZIP synergy score tested here indicates that the combination of the two drugs has a synergy score between 0 and 10 (figure1, and figure 5) . According to "Synergy Finder" , "A ZIP synergy score of greater than 10 often indicates a strong synergistic effect, while a score less than -10 suggests a strong antagonistic effect. Scores between -10 and 10 are typically considered additive or near-additive." The data in figure 2 on mitotic cell fraction and on cell death also seems to be more of an additive effect of the two drugs than synergy. The data in figure 3 are also additive effects on RAD51. Therefore a conclusion that "These data indicate that the drug combination was broadly synergistic" seems unwarranted. Indeed, the data form

2. There was no statistical difference in the synergy scores of the "high expressing" versus "low expressing cells". So the conclusion that the drug combination "t was effective at lower doses in cell lines with high levels of EYA1 and/or EYA4" seems unwarranted based on the data. Moreover, since there was no statistical difference in synergy between high and low expressing cells, stating that "the potential utility of the combination treatment depends on the specific overexpression of EYA1 and/or EYA4 in cancer cells," seems unwarranted by the data.

3. Benzarone and benzbromarone and their derivatives have been shown to bind and inhibit Eya phosphatases, albeit at fairly high doses. However, these two compounds also have a number of other, unrelated targets. The only demonstration that Eyas are a target of benzarone in this study are the CETSA data in supplemental figure 1. The data here seem to represent an n of 1, with no error bars shown. Even more importantly, there is no control. Looking at the blot of actin, it seems as if there may be a benzarone- temperature effect on this protein as well. It would be very helpful to show some evidence that knockdown of Eya similarly synergizes with the PLK1 inhibitor, show data that benzarone is in fact inhibiting Eya activity in these cells by looking at known targets (ie the carboxyterminal tyrosine of H2AX), and other evidence of specificity.

4. The proteomic and transcriptomic data of cell lines that were vulnerable to the combination of BI2536 and benzarone implicate overall changes in chromatin with sensitivity. These findings call into question the idea that these two compounds are acting selectively on PLK1 and Eyas. The authors don't really provide any model for explaining this correlation of Nurd complex components with targeting Eyas and PLK1.

5. Specificity of antibodies: I would like to see validation of the Eya antibodies, given the difficulty with such reagents in the field.

Significance

New therapies targeting glioblastoma would be welcome. It is not clear that the combination tested here is an effective approach to therapy. It would be necessary to know the targets of the combination and understand the mechanism so that the approach could be pursued further,

Reviewer #1 (Public review):

The authors responded to multiple criticisms with additional data and more detailed statistics, in some instances improving the quality of the work. However, I had difficulty understanding some of the authors' responses. The logic was not always apparent, the writing was occasionally confusing or would benefit from more careful wording, and some of the provided responses were superficial or raised new concerns. In some cases, the underlying data needed to support their responses were not shown. Thus, the current version of the manuscript does not sufficiently resolve the following critical issues raised by myself and other reviewers.

(1) A clear new insight into a physiological process or cellular behavior remains lacking. The study largely confirms prior observations of MCAK binding to both the microtubule wall and end. However, it is still unclear whether direct binding to the tip-as opposed to accumulation via wall diffusion or interaction with other tip-binding proteins-is a significant mechanism.

(2) The newly revealed adenosine-nucleotide-dependent binding preferences do not help clarify MCAK's catalytic function or its mechanisms of tip recognition. Consequently, the final summary figure remains speculative and is not convincingly supported by the data. It is also unclear what exactly is meant by the "working model" (figure title), or by the claim of "a simple rule of how the end-binding regulators coordinate their activities" (abstract).

(3) As noted in my previous review, the effects of adding different adenosine nucleotides on MCAK binding to microtubules are much more pronounced than the differences in MCAK binding to tubulin with various guanosine-containing nucleotides, or to lattice versus tip (e.g., Fig. 5E). Therefore, the manuscript title-"MCAK recognizes the nucleotide-dependent feature at growing microtubule ends"-does not do justice to the scale of these effects.

(4) The title implies that MCAK selectively recognizes a feature determined by the tubulin-bound guanosine nucleotide. However, the authors frequently claim that MCAK binds to the "entire GTP cap." It appears that they exclude structural protrusions from their definition of the cap, which is debatable. Even using their definition, the conclusion that MCAK recognizes a specific "nucleotide-dependent feature" seems inconsistent with the claim that it binds uniformly across the cap. These distinctions were not made clear.

(5) Some important technical details are still absent. For example, when reading the authors' response to another reviewer's question, I could not find an explanation of how the kon values for end and wall binding were calculated. These calculations clearly require assumptions, e.g. about the number of binding sites, but these details are not described. In addition, the binding data are expressed in units per tubulin dimer, which are non-standard and make comparisons to other published results difficult. There are other instances where more technical detail would be desirable, but they are too numerous to list here.

(6) Several aspects of data presentation as graphs will make it difficult for other researchers to analyze or interpret the findings. Numerical Excel-style data sheets should be provided for all measurements, including raw data-not just the ratios or derived values shown in plots. Other, more significant issues include use of mean values for non-Gaussian distributions (e.g., dwell times); binding affinities inferred from single-concentration measurements, often under varying conditions (e.g., Figs. 3C, 4); and absence of side-by-side plotted controls (e.g., Fig. 6).

(7) While the authors have added some quantitative values and descriptive detail, the manuscript still lacks a critical comparison of their findings with existing literature. This weakens the impact of the study and limits the reader's ability to place the results in a broader context.

Reviewer #4 (Public review):

The revised manuscript from Chen et al. implements many of the changes requested by the 3 reviewers of the initial submission. These changes are well-described in the corresponding Response to Reviews document. Of course, not every request from the reviewers was addressed, and the following major concerns remain:

(1) The authors argue that MCAK binds to the same region as EB proteins, which they refer to as the "EB cap". Reviewers asked for experiments that would increase the size of the EB cap to create "comets" (e.g. by increasing the microtubule growth rate); the prediction is that the MCAK signal should increase in size as well. The authors declined to pursue these experiments. As a result, the EB signals and MCAK signals are diffraction-limited spots, as opposed to the predicted exponential decay signals characteristic of EB comets. The various diffraction-limited spots are then aligned with the diffraction-limited signal of the microtubule end. These alignments and sub-pixel comparisons are technically challenging. The revised manuscript does not go far enough to provide compelling evidence that all technical challenges were overcome. Thus, while the authors can safely conclude that MCAK, EBs, and the microtubule end do occupy the same diffraction-limited spot, more precise conclusions are not supported.

(2) The reviewers criticized the initial manuscript for neglecting key references, particularly Kinoshita et al., Science 2001. Indeed, I cannot fathom writing a manuscript about MCAK and XMAP215 without putting a citation to such a landmark paper front and center. The authors have responded by including more discussion of the relevant literature (and citing Kinoshita et al.). However, the revised manuscript is often still cursory in giving credit where credit is due, contextualizing the new data, and generally engaging with the scholarship on MCAK.

(3) The data presented does not include a simple measurement of the impact of MCAK on the catastrophe frequency of microtubules. The authors explain this absence by pointing out that their movies are short (5 min) and high frame rate (10 fps). While I understand that such imaging parameters are necessary to capture single molecule end-binding events, I do not understand why a separate set of experiments could not be performed. This type of "positive control" is often missing, as pointed out by the 3 reviewers.

(4) Salt conditions, protein concentrations, and other key experimental parameters are not varied, even when varying them would provide excellent tests of the authors' hypotheses.

In summary, the revised manuscript is improved in many ways, but the interested reader should look carefully at the previous reviews and compare the measurements presented here with those of other labs.

Reviewer #2 (Public Review):

Summary:

This paper describes a new approach to detecting directed causal interactions between two genes without directly perturbing either gene. To check whether gene X influences gene Z, a reporter gene (Y) is engineered into the cell in such a way that (1) Y is under the same transcriptional control as X, and (2) Y does not influence Z. Then, under the null hypothesis that X does not affect Z, the authors derive an equation that describes the relationship between the covariance of X and Z and the covariance of Y and Z. Violation of this relationship can then be used to detect causality.

The authors benchmark their approach experimentally in several synthetic circuits. In 4 positive control circuits, X is a TetR-YFP fusion protein that represses Z, which is an RFP reporter. The proposed approach detected the repression interaction in 2 of the 4 positive control circuits. The authors constructed 16 negative control circuit designs in which X was again TetR-YFP, but where Z was either a constitutively expressed reporter, or simply the cellular growth rate. The proposed method detected a causal effect in two of the 16 negative controls, which the authors argue is perhaps not a false positive, but due to an unexpected causal effect. Overall, the data support the potential value of the proposed approach.

Strengths:

The idea of a "no-causality control" in the context of detected directed gene interactions is a valuable conceptual advance that could potentially see play in a variety of settings where perturbation-based causality detection experiments are made difficult by practical considerations.

By proving their mathematical result in the context of a continuous-time Markov chain, the authors use a more realistic model of the cell than, for instance, a set of deterministic ordinary differential equations.

The authors have improved the clarity and completeness of their proof compared to a previous version of the manuscript.

Limitations:

The authors themselves clearly outline the primary limitations of the study: The experimental benchmark is a proof of principle, and limited to synthetic circuits involving a handful of genes expressed on plasmids in E. coli. As acknowledged in the Discussion, negative controls were chosen based on the absence of known interactions, rather than perturbation experiments. Further work is needed to establish that this technique applies to other organisms and to biological networks involving a wider variety of genes and cellular functions. It seems to me that this paper's objective is not to delineate the technique's practical domain of validity, but rather to motivate this future work, and I think it succeeds in that.

Might your new "Proposed additional tests" subsection be better housed under Discussion rather than Results?

I may have missed this, but it doesn't look like you ran simulation benchmarks of your bootstrap-based test for checking whether the normalized covariances are equal. It would be useful to see in simulations how the true and false positive rates of that test vary with the usual suspects like sample size and noise strengths.

It looks like you estimated the uncertainty for eta_xz and eta_yz separately. Can you get the joint distribution? If you can do that, my intuition is you might be able to improve the power of the test (and maybe detect positive control #3?). For instance, if you can get your bootstraps for eta_xz and eta_yz together, could you just use a paired t-test to check for equality of means?

The proof is a lot better, and it's great that you nailed down the requirement on the decay of beta, but the proof is still confusing in some places:

On pg 29, it says "That is, dividing the right equation in Eq. 5.8 with alpha, we write the ..." but the next equation doesn't obviously have anything to do with Eq. 5.8, and instead (I think) it comes from Eq 5.5. This could be clarified.

Later on page 29, you write "We now evoke the requirement that the averages xt and yt are stationary", but then you just repeat Eq. 5.11 and set it to zero. Clearly you needed the limit condition to set Eq. 5.11 to zero, but it's not clear what you're using stationarity for. I mean, if you needed stationarity for 5.11 presumably you would have referenced it at that step.

It could be helpful for readers if you could spell out the practical implications of the theorem's assumptions (other than the no-causality requirement) by discussing examples of setups where it would or wouldn't hold.

Author response:

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

Reviewer #1 (Public review): 

Summary: 

This study addresses a critical gap in veterinary diagnostics by developing a CRISPR-based diagnostic toolbox (SHERLOCK4AAT) for detecting animal African trypanosomosis. It describes the development and field deployment of SHERLOCK4AAT, a CRISPR-Cas13-based diagnostic toolbox for the eco-epidemiological surveillance of animal African trypanosomosis (AAT) in West Africa.The authors successfully created and validated species-specific assays for multiple trypanosomes, including T. congolense, T. vivax, T. theileri, T. simiae, and T. suis, alongside pan-trypanosomatid and pan-Trypanozoon assays. The field validation in pigs from Guinea and Côte d'Ivoire revealed high trypanosome prevalence (62.7%), frequent co-infections, and importantly identified T. b. gambiense in one animal at each site, suggesting pigs may serve as potential reservoirs for this human-infective parasite.

A major strength of the study lies in its methodological innovation. By adapting SHERLOCK to target both conserved and species-discriminating sequences, the authors achieved high sensitivity and specificity in detecting Trypanosoma species. Their use of dried blood spots, validated thresholds through ROC analyses, and statistical robustness (e.g., Bayesian latent class modeling) provides a strong foundation for their conclusions.

The results are significant: over 60% of pigs tested positive for at least one trypanosome species, with co-infections observed frequently and T. b. gambiense detected in pigs at both sites. These findings have direct implications for the role of animal reservoirs in human disease transmission and underscore the value of pigs as sentinel hosts in gHAT elimination efforts.

The limitations are well acknowledged, particularly the suboptimal sensitivity of the T. vivax assay and the reliance on synthetic controls for T. suis and T. simiae. However, these limitations do not undermine the overall conclusions, and the paper provides a clear roadmap for further assay refinement and implementation.

This study offers a timely, impactful, and well-substantiated contribution to the field. The SHERLOCK4AAT toolbox holds promise for improving AAT diagnostics in resource-limited settings and advancing One Health surveillance frameworks.

Thank you

Strengths: 

(1) The adaptation of SHERLOCK technology for AAT represents a significant technical advancement, offering higher sensitivity than traditional parasitological methods and the ability to detect multiple species simultaneously.

(2) Rigorously performed with validation using appropriate controls, ROC curve analyses, and Bayesian latent class modelling, establishing clear analytical sensitivity and specificity for most assays.

(3) Testing 424 pig samples across two countries provides robust evidence of the tool's utility and reveals important epidemiological insights about trypanosome diversity and prevalence.

(4) The identification of T. b. gambiense in pigs at both sites has significant implications for HAT elimination strategies and highlights the need for integrated One Health approaches.

(5) The use of dried blood spots and RNA detection for active infections makes the approach practical for field surveillance in resource-limited settings.

Thank you

Weaknesses: 

(1) The manuscript would benefit from more detailed discussion of practical considerations such as cost, equipment requirements, and training needs for implementing SHERLOCK in endemic areas and rural settings which would improve applicability.

This is now adressed in the revised discussion (end of the first section).

(2) Limited discussion of pig selection criteria: More justification for choosing pigs as sentinel animals and discussion of potential limitations of this approach would strengthen the manuscript.

Yes, this is now more clearly explained in the revised discussion (beginning of the first section).

(3) More details on why certain genes were targeted would strengthen the methods.

The first result section ‘Selection of targets for broad and species-specific SHERLOCK assays targeting AAT species (SHERLOCK4AAT)’ is already dedicated to extensively explaining target selection, hence we’re afraid we don’t know what could be added.  

(4) Table formatting could be improved for readability. 

(5) Some figures are complex and would benefit from additional explanations in the legends.

We have tried to improve these two aspects as much as possible in the revised manuscript.

Reviewer #2 (Public review): 

Summary: 

The manuscript is important due to the significance of the findings. The strength of evidence is convincing.

Thank you

Strengths: 

(1) Using a Novel SHERLOCK4AAT toolkit for diagnosis. 

(2) Identification of various sub-species of Trypanosomes. 

(3) Differentiating the animal subspecies from the human one. 

Thank you

Weaknesses: 

(1) The title is too long, and the use of definite articles should be reduced in the title.

The title has been improved in the revised version.

(2) The route of blood sample collection in the animals should be well defined and explained.

This has been more clearly explained in the revised method section.

Reviewer #3 (Public review):

Summary: 

The study adapts CRISPR-based detection toolkit (SHERLOCK assay) using conserved and species-specific targets for the detection of some members of the Trypanosomatidae family of veterinary importance and species-specific assays to differentiate between the six most common animal trypanosome species responsible for AAT (SHERLOCK4AAT). The assays were able to discriminate between Trypanozoon (T. b. brucei, T. evansi, and T. equiperdum), T. congolense (Savanah, Forest Kilifi, and Dzanga sangha), T. vivax, T. theileri, T. simiae, and T. suis. The design of both broad and species-specific assays was based primarily on sequences of the 18S rRNA, GAPDH (Glyceraldehyde-3-phosphate dehydrogenase), and invariant flagellum antigen (IFX) genes for species identification. Most importantly, the authors showed varying limits of detection for the different SHERLOCK assays, which is somewhat comparable to PCR-derived molecular techniques currently used for detecting animal trypanosomes, even though some of these methodologies have used other primers that target genes such as ITS1 and 7SL sRNA. <br /> The data presented in the study are particularly useful and of significant interest for the diagnosis of AAT in affected areas.

Thank you

Strengths: 

The assays convincingly allow for the analysis and detection of most trypanosomes in AAT.

Thank you

Weaknesses: 

Inability for the assay to distinguish T. b. brucei, T. evansi, and T. equiperdum using the 18S rRNA gene, as well as the IFX gene, not achieving the sensitivity requirements for detection of T. vivax.  Both T. brucei brucei and T. vivax are the most predominant infective species in animals (in addition to T. congolense), therefore, a reliable assay should be able to convincingly detect these to allow for proper use of the diagnostic assay.

We agree with this point and aim to improve the toolbox for future studies.

Reviewer #1 (Recommendations for the authors):

(1) Provide additional details on the practicality of SHERLOCK deployment in the field, including training, costs, and infrastructure (potential challenges for field deployment, including suggestions for how to overcome these barriers).

This is now adressed in the revised discussion (end of the first section).

(2) Provide more detailed justification for choosing pigs as the main study species and discuss potential benefits and limitations of extending the approach to other livestock species.

Yes, this is now more clearly explained in the revised discussion (beginning of the first section).

(3) Add a comparison table comparing SHERLOCK4AAT performance metrics (sensitivity, specificity, LoD) with existing molecular diagnostic methods for AAT for ease of reference.

There are dozens of different serological, immunological and molecular approaches with highlty variable levels of sensitivity and specificities already reviewed and compared in detail in two references from 2022 (Desquesnes et al. a and b), which we have cited, as well as in a newly added reference (EBHODAGHE F acta trop 2018). Hence, we decided to only refer to the most comparable studies in the present article.

(4) Review complex figures and improve legends for better readability and interpretation.

We have tried to improve this as much as possible in the revised manuscript.

Reviewer #2 (Recommendations for the authors): 

(1) Reduce the number of words in the title from 28 to not more than 20.

The title has been improved in the revised version.

(2) Specify the particular route of collection of blood samples in the various animals.

Yes, this is now more clearly explained in the revised method section.

(3) Correct all typographical errors. 

We have tried to improve this as much as possible in the revised manuscript.

Thanks. I wish you the best in your publication process. 

Thank you

Reviewer #3 (Recommendations for the authors): 

Minor comments 

(1) The authors can expand the discussion to include other recent diagnostic assays for Animal trypanosomiasis, such as those that target other genes like tubulin.

Please see response to Review 1 point #3 above.

(2) The cost-effectiveness of the use of the assay can be discussed since the assay is expected to be used for work in some resource-deprived areas. For example, will it cost a researcher less to do a diagnosis with this assay relative to what is already available?

This is now adressed in the revised discussion (end of the first section).

(3) Is Cote d'Ivoire more endemic for AAT than Guinea? Will this account for the apparently consistent differences in the percentage of positive samples, or just because of the type of samples used from the two locations?

As the sampling method, sample preservation and sample analysis were the same for both groups - yes, it appears that pigs, at least for domesticated ones, in the study region of Cote d'Ivoire were more frequently infected than those in the study region of Guinea. It is however risky to extrapolate these observations to the AAT prevalence in the entire countries and/or to other mammals.

(4) Can the authors comment on how long one can store the samples for an effective and reliable assay?

The samples can be stored for several months at ambient temperature in a sealed bag with silica gel packages to reduce humidity. We have added this detail in the revised methods section.

(5) It is not clear whether the authors used conventional molecular diagnostics to compare the data obtained from this particular cohort of animals as reference is made to published data. It is not surprising that the SHERLOCK performed better than using parasitology-based methodology.

This is now adressed in the revised discussion.

(6) (Figure 4D-5D) should be 4D and 5D.

Thank you, this has been corrected.

Author response:

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

Reviewer #1 (Public Review):

Summary:

In this manuscript, the authors used a coarse-grained DNA model (cgNA+) to explore how DNA sequences and CpG methylation/hydroxymethylation influence nucleosome wrapping energy and the probability density of optimal nucleosomal configuration. Their findings indicate that both methylated and hydroxymethylated cytosines lead to increased nucleosome wrapping energy. Additionally, the study demonstrates that methylation of CpG islands increases the probability of nucleosome formation.

Strengths:

The major strength of this method is that the model explicitly includes elastic constraints on the positions of phosphate groups facing a histone octamer, as DNA-histone binding site constraints. The authors claim that their model enhances the accuracy and computational efficiency and allows comprehensive calculations of DNA mechanical properties and deformation energies.

Weaknesses:

A significant limitation of this study is that the parameter sets for the methylated and hydroxymethylated CpG steps in the cgNA+ model are derived from all-atom molecular dynamics (MD) simulations that suggest that both methylated and hydroxymethylated cytosines increase DNA stiffness and nucleosome wrapping energy (P´erez A, et al. Biophys J. 2012; Battistini, et al. PLOS Comput Biol. 2021). It could predispose the coarse-grained model to replicate these findings. Notably, conflicting results from other all-atom MD simulations, such as those by Ngo T in Nat. Commun. 2016, shows that hydroxymethylated cytosines increase DNA flexibility, contrary to methylated cytosines. If the cgNA+ model was trained on these later parameters or other all-atom force fields, different conclusions might be obtained regarding the effects of methylated and hydroxymethylation on nucleosome formation.

Despite the training parameters of the cgNA+ model, the results presented in the manuscript indicate that methylated cytosines increase both DNA stiffness and nucleosome wrapping energy. However, when comparing nucleosome occupancy scores with predicted nucleosome wrapping energies and optimal configurations, the authors find that methylated CGIs exhibit higher nucleosome occupancies than unmethylated ones, which seems to contradict their findings from the same paper which showed that increased stiffness should reduce nucleosome formation affinity. In the manuscript, the authors also admit that these conclusions “apparently runs counter to the (perhaps naive) intuition that high nucleosome forming affinity should arise for fragments with low wrapping energy”. Previous all-atom MD simulations (P´erez A, et al. Biophys J. 2012; Battistini, et al. PLOS Comput Biol. 202; Ngo T, et al. Nat. Commun. 20161) show that the stiffer DNA upon CpG methylation reduces the affinity of DNA to assemble into nucleosomes or destabilizes nucleosomes. Given these findings, the authors need to address and reconcile these seemingly contradictory results, as the influence of epigenetic modifications on DNA mechanical properties and nucleosome formation are critical aspects of their study. Understanding the influence of sequence-dependent and epigenetic modifications of DNA on mechanical properties and nucleosome formation is crucial for comprehending various cellular processes. The authors’ study, focusing on these aspects, will definitely garner interest from the DNA methylation research community.

Training the cgNA+ model on alternative MD simulation datasets is certainly of interest to us. However, due to the significant computational cost, this remains a goal for future work. The relationship between nucleosome occupancy scores and nucleosome wrapping energy is still debated, with conflicting findings reported in the literature, as noted in our Discussion section. Interestingly, we find that our predicted log probability density of DNA spontaneously acquiring a nucleosomal configuration is a better indicator of nucleosome occupancy than our predicted DNA nucleosome wrapping energy.

Reviewer #2 (Public Review):

Summary:

This study uses a coarse-grained model for double-stranded DNA, cgNA+, to assess nucleosome sequence affinity. cgNA+ coarse-grains DNA on the level of bases and accounts also explicitly for the positions of the backbone phosphates. It has been proven to reproduce all-atom MD data very accurately. It is also ideally suited to be incorporated into a nucleosome model because it is known that DNA is bound to the protein core of the nucleosome via the phosphates.

It is still unclear whether this harmonic model parametrized for unbound DNA is accurate in describing DNA inside the nucleosome. Previous models by other authors, using more coarse-grained models of DNA, have been rather successful in predicting base pair sequence-dependent nucleosome behavior. This is at least the case as far as DNA shape is concerned whereas assessing the role of DNA bendability (something this paper focuses on) has been consistently challenging in all nucleosome models, to my knowledge.

It is thus of major interest whether this more sophisticated model is also more successful in handling this issue. As far as I can tell the work is technically sound and properly accounts for not only the energy required in wrapping DNA but also entropic effects, namely the change in entropy that DNA experiences when going from the free state to the bound state. The authors make an approximation here which seems to me to be a reasonable first step.

Of interest is also that the authors have the parameters at hand to study the effect of methylation of CpG-steps. This is especially interesting as it allows us to study a scenario where changes in the physical properties of base pair steps via methylation might influence nucleosome positioning and stability in a cell-type-specific way.

Overall, this is an important contribution to the question of how the sequence affects nucleosome positioning and affinity. The findings suggest that cgNA+ has something new to offer. But the problem is complex, also on the experimental side, so many questions remain open.

Strengths:

The authors use their state-of-the-art coarse-grained DNA model which seems ideally suited to be applied to nucleosomes as it accounts explicitly for the backbone phosphates.

Weaknesses:

(1) According to the abstract the authors consider two “scalar measures of the sequence-dependent propensity of DNA to wrap into nucleosomes”. One is the bending energy and the other, is the free energy. Specifically in the latter, the authors take the difference between the free energies of the wrapped and the free DNA. Whereas the entropy of the latter can be calculated exactly, they assume that the bound DNA always has the same entropy (independent of sequence) in its more confined state. The problem is the way in which this is written (e.g. below Eq. 6) which is hard to understand. The authors should mention that the negative of Eq. 6 is what physicists call free energy, namely especially the free energy difference between bound and free DNA.

We have included the necessary clarifications in the revised manuscript, below Eq. 6.

(2) In Eq. 5 the authors introduce penalty coefficients c_i. They write that values are “set by numerical experiment to keep distances ... within the ranges observed in the PDB structure, while avoiding sterical clashes in DNA.” This is rather vague, especially since it is unclear to me what type of sterical clashes might occur. Figure 1 shows then a comparison between crystal structures and simulated structures. They are reasonably similar but standard deviations in the fluctuations of the simulation are smaller than in the experiments. Why did the authors not choose smaller c_i-values to have a better fit? Do smaller values lead to unwanted large fluctuations that would lead to steric clashes between the two DNA turns? I also wonder what side views of the nucleosomes look like (experiments and simulations) and whether in this side view larger fluctuations of the phosphates can be observed in the simulation that would eventually lead to turn-turn clashes for smaller c_i-values.

The side view plots of the experimental and predicted nucleosome structures are now added to Supplementary material (Figure S8). Indeed, smaller c_i values lead to steric clashes between the two turns of DNA – this is now specified in the Methods section. A possible improvement of our optimisation method and a direction of future work would be adding a penalty which prevents steric clashes to the objective function. Then the c_i values could be reduced to have bigger fluctuations that are even closer to the experimental structures. We added this explanation to the Results section.

Reviewer #3 (Public Review):

Summary:

In this study, the authors utilize biophysical modeling to investigate differences in free energies and nucleosomal configuration probability density of CpG islands and nonmethylated regions in the genome. Toward this goal, they develop and apply the cgNA+ coarse-grained model, an extension of their prior molecular modeling framework.

Strengths:

The study utilizes biophysical modeling to gain mechanistic insight into nucleosomal occupancy differences in CpG and nonmethylated regions in the genome.

Weaknesses:

Although the overall study is interesting, the manuscripts need more clarity in places. Moreover, the rationale and conclusion for some of the analyses are not well described.

We edited the manuscript according to the reviewer’s suggestions and hopefully improved its readability.

Reviewer #1 (Recommendations For The Authors):

(1) The cgNA+ model parameters are derived from all-atom molecular dynamics (MD) simulations, yet there is no consensus within all-atom MD simulations regarding the impact of CpG methylation on DNA mechanical properties. The authors could consider fitting the coarsegrained model with a different all-atom force field to verify whether the conclusions regarding the effects of methylation and hydroxymethylation on DNA nucleosome wrapping energies still hold. For further details on MD simulations related to CpG methylation effects, the authors are advised to consult the review paper by Li et al. (2022) titled “DNA methylation: Precise modulation of chromatin structure and dynamics” published in Current Opinion in Structural Biology.

Parametrizing the cgNA+ model using MD simulations with various force fields is certainly of interest to us. However, due to the computational cost involved, it remains a goal for future work.

(2) Beyond DNA mechanical properties, which are directly linked to nucleosome wrapping energies in this study, the authors might also consider other factors such as geometric properties that could influence nucleosome formation. This approach might help the authors to reconcile the observed higher nucleosome occupancy scores for methylated CpGs. The authors are encouraged to review the aforementioned paper for additional experimental and MD simulation studies that could support this perspective.

Geometric properties of DNA are directly incorporated into our method through the cgNA+ model equilibrium shape prediction µ. We compute the mechanical energy needed deform µ to a nucleosomal configuration. Notably, the equilibrium shape µ is sensitive to methylation, as demonstrated in Figure 3.

(3) There are some issues with citation accuracy in the manuscript. For instance, in the Discussion section, the authors attribute a statement to Collings et al. and Anderson (2017), claiming that “methylated regions, known to have high wrapping energy, are among the highest nucleosome occupied elements in the genome.” However, upon reviewing this paper, it appears that it does not make any claims about the high wrapping energy of methylated regions.

The paragraph is now edited and a separate citation, P´erez et al. (2012), is given for the statement that methylation regions have high wrapping energy.

Reviewer #2 (Recommendations For The Authors):

Please improve the readability by:

(1) making clear that -ln ρ in Eq. 6 on page 4 is actually the free energy. Also, the word entropy comes too late (on page 7) where the best explanation of Eq. 6 is presented.

We added a comment about -ln ρ being the free energy after Eq. 6 and also included an equation, relating ln ρ and entropy.

(2) page 12 and 13 show two sets of experimental data. They are quite different from each other. When reading this, I wondered why there is this difference. But only on page 16, you explain that these are different cell types. The difference should be explained already when the papers are introduced on page 12.

A corresponding sentence already appeared in page 12: “The observations about nucleosome occupancy should be regarded as preliminary, and be treated with caution, as they are based on experimental data obtained for the cancerous HeLa cells Schwartz et al. (2019) and human genome embryonic stem cells Yazdi et al. (2015)”. Now we also added this information to the first paragraph of the subsection for clarity.

Finally, I add here some general thoughts that came up when reading the paper, comparing your findings with earlier findings in the field. This is not a strict one-to-one comparison and thus does not have to find its way into this manuscript but might give ideas for future studies. Experiments suggest that nucleosomes prefer DNA with a high content of C’s and G’s. Figure 2 does not look at the GC content but at the number of CpG’s. But in any case, let’s use this as a proxy for GC content. Figure 2a suggests that there is not a strong dependence of the bending energy on the number CpG steps. This is consistent with earlier work with the rigid basepair model which shows the same behavior for GC content (for both MD and crystal parametrizations). Figure 2c (related to the negative free energy) shows that with an increasing number of CpG steps the propensity to bind goes down. This suggests that the entropic cost to confine CpG-rich DNA increases, which in turn reflects that these DNA stretches are softer. This is rather interesting since in the case of the rigid basepair model this effect is observed only when stiffnesses are extracted from crystal data not MD data (however, this refers again to CG content). This might indicate a difference between the rigid bp model and cgNA+ which will be interesting to study in the future. Interesting is also the effect of CpG methylation. The stiffer methylated steps lead to an increase in the energy with the number of such steps (Figure 2a). The entropic cost for binding is thus expected to be smaller and this is indeed observed in Figure 2c when compared to the non-methylated steps.

We thank the reviewer for this comment. As for the GC content, the energy and lnp plots are indeed very similar to those in Figure 2.

Reviewer #3 (Recommendations For The Authors):

(1) The formulation of the cgNA+ model in the method section was not easy to follow and can be described better to improve clarity.

We have revised the model description and hope that its clarity has been improved

(2) The authors mention utilizing 100 human genome sequences with 100 configurations from DB. It would be helpful to clarify the source of these 100 human genome sequences. Are these 100 distinct regions on the human reference genome, or are they from a specific dataset or database?

We now include an explanation about the origin of sequences: “The human genome sequences are a random subset of our sequence sample for the CGI and NMI intersection in the Chromosome 1, but the following observations remain unchanged for sequence samples from different genomic regions.”

(3) The authors mention the lack of tail unwrapping in their model. It would be beneficial to understand the magnitude of this issue and its potential impact on the overall results. How significant is the lack of unwrapping events in their current model?

We observed the unwrapping of approximately five base-pairs at each end of our predicted nucleosome configurations, in comparison to the experimental configurations (Figure 1). This issue could be solved by adding additional constraints at the ends of the 147 bp sequence. The wrapping energy would increase marginally, as only about 10 of 147 bp would be affected. We added this remark to the main text.

(4) Observations from Figure 3 are not described properly. Are these differences statistically significant? Why is twist higher for CpG sites but lower for a roll?

We added an explanation of how the statistics was computed into the caption of Figure 3. In fact, we didn’t use statistical estimates here, but generated all the possible cases and computed the exact statistics (for the given set of our model parameters). Regarding the changes in twist and roll, we have added the following comment on page 7: “The ground state changes resulting from cytosine modifications – primarily characterized by an average increase in roll and a decrease in twist – may be linked to steric hindrance caused by the cytosine 5-substituent (Battistini et al. (2021)). Notably, the negative coupling between twist and roll has already been observed in X-ray crystallography data (Olson et al. (1998)).”

(5) Figure 4 does not clarify the authors’ conclusion of higher stiffness for ApT and TpA dinucleotides. The authors should provide further explanation for this observation.

We revised the text to clarify that the statement regarding ApT and TpA being the most stiff and the most flexible dinucleotides is not a conclusion derived from Figure 4, but rather from earlier work that we cite.

(6) In Figure 7, the authors note that methylated CGIs have higher nucleosome occupancy on average than unmethylated sequences. Is this observation statistically significant?

We observe that methylated sequences have a higher average occupancy than unmethylated sequences in Yazdi et al. data, when the CpG count falls into the intervals from 5 to 14 and from 15 to 24. For each of the two intervals this difference is statistically significant: the permutation test, used due to the lack of normality, yields a p-value of 0.0001 for both cases. The differences in mean scores shown in Figure 8 are also statistically significant. Such test results are expected, given the large sample sizes and the observed differences in means, therefore we prefer not to include this discussion in main text.

(7) The authors note that their analyses to correlate nucleosome occupancy profile with the methylation state of underlying sequences are preliminary, as different cell lines were used to perform these analyses. Given this inconsistency, it needs to be clarified why this analysis was performed and what the takeaway is.

We added the following comment at the end of the Results section: “Although comparing data from different cell lines is not optimal, to the best of our knowledge, no publicly available methylation and nucleosome occupancy data exist for the entire human genome within the same cell type. Nevertheless, since the lowest log probability densities in the human genome are predicted for CpG-rich sequences regardless of their methylation state (Figure 2d), and the same holds for both sets of the nucleosome occupancy scores (Figure 7), we conclude that the lowest occupancies occur for sequences with the lowest log probability densities.”

The value of the article lies not in its too-roughly calculated estimate but in its attempt to highlight both an important yet unanswered question and the difficulties that hinder our ability to reliably answer it. The article also provides a useful summary of the bourgeoning literature and the challenges of drawing broad inferences from it. The author should consider highlighting these points rather than the rough estimate of the rate of falsification and fabrication. The author should change the article’s title to reflect the skepticism about the estimate that runs throughout the article so as not to confuse readers about what we can reliably take away from the article.

The following are specific suggestions:

1. Adding reference or links to Table 1 would help readers find details related to the listed items.

2. p. 6 (“The following (Table 1) is a selection of events which took place after the figure above was established.”): Clarify why 2005 is the first year of interest in the events table (e.g., change sentence to “The following (Table 1) is a selection of events that took place during or after 2005, the final year of publication of the studies Fanelli used to compute the 2% figure.”).

3. p. 7 (“Significantly, all of the above happened after the figure of 2% was collected.”): change to “… after publication of the studies on which the 2% figure is based.”

4. The link in footnote 5 no longer works. The report can be found at https://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1262&context=scholcom. Consider changing all links to permalinks.

5. p. 16: list the 29 observations of data sleuth estimates in a footnote.

6. Footnote 6: the link pulls up the Retraction Watch Database for Nature Publishing Group. I accessed the link on Jan 15, 2025, and the database found 1,610 items. It’s not clear how you computed 667 (12,000 / 667 = 18). If “Retracted Article” is chosen for “Article Type(s),” the count is 1.

7. p. 18 (“the presumably higher number of papers containing questionable research practices (which are far more commonly admitted to) is presumably higher still.”): Consider citing to published estimates, which are mostly produced using surveys, and note that this literature suffers from the same problems as the literature that estimates FF.

8. p. 18: add citations to articles that address each of the harms caused by false results.

9. Bottom of p. 19 (“In particular, it seems likely that FF rates change by individual field – in doing so, they may present specific rather than general threats to human health and scientific progress.”): Providing some explanation for field-specific rates might help the reader assess the claim. For example, it’s possible that rates are similar across fields because those willing to commit fraud or to fabricate data likely randomly distribute themselves across fields, and journal editors and referees are roughly equally likely to fail to detect falsification and fabrication. Does any evidence call these possibilities into question?

This manuscript attempts to provide an answer to the proportion of scientific papers that are fake. The presence of fake scientific papers in the literature is a serious problem, as the author outlines. Papers of variable quality and significance will inevitably be published, but most researchers assess manuscripts and papers based on the assumption that the described research took place. Papers that disguise their identities as fake papers can therefore be highly damaging to research efforts, by preventing accurate assessments of research quality and significance, and by encouraging future research that could consume time and other resources. As the manuscript describes, fake papers are also damaging to science by eroding trust in the scientific method and communities of scientists.

It is therefore clear that knowing the proportion of fake scientific papers is important, that the author is concerned about the problem, and that the author wants to arrive at an answer. However, as the manuscript partly recognises, the question of the overall proportion of fake scientific papers is currently difficult to answer.

The overall proportion of fake papers in science will represent the individual proportions of fake papers in different scientific disciplines. In turn, the proportions of fake papers in any single discipline will reflect many factors, including (i) researcher incentives to produce fake papers, (ii) the ease with which fake papers can be produced and (iii) published, (iv) the ease or likelihood of fake papers being detected, before or (v) after publication, and (vi) the consequences for authors if they are found to have published fake papers. Some of these factors are likely to vary between different disciplines and in different research settings. For example, it has been suggested that it is similarly difficult to invent some research results as it is to produce genuine data. However, in other fields, it is easier to invent data than to generate data through experiments that remain difficult, expensive and/or slow. It is also likely that factors such as the capacity to invent fake papers, detect fake papers, as well as incentives and consequences for researchers could vary over time, particularly in response to generative AI.

As someone who studies errors in scientific papers, I don’t believe that we currently have a good understanding of the proportions of fake papers in any individual scientific field, at any time. There are some fields where we have estimates of individual error types, but these error types are likely to wrongly estimate the overall proportions of fake papers. Rather than attempting to answer the question of the overall proportion of fake scientific papers in the absence of the necessary data, it seems preferable to describe how we could obtain the data that we need to answer this question. While the overall proportion of fake scientific papers is an important statistic, most scientists will also be more concerned about how many fake papers exist in their own fields. We could therefore start by trying to obtain reliable estimates of fake papers in individual fields, working out how we need to do this, and then carrying out the necessary research. In the absence of reliable data, it’s perhaps most important that researchers are aware that fake papers could exist in their fields, so that all researchers can assess papers more carefully.

Beyond these broad considerations, the following manuscript elements could be reconsidered.

1. Fake science is defined as fabricated or falsified, yet this definition is sometimes expanded to include plagiarism (page 8, Table 2). However, plagiarism doesn’t equate with faking or falsifying data, and some plagiarised articles could describe sound data. Including plagiarised articles as fake articles will inevitably inflate estimates of fake papers, particularly in fields with higher rates of plagiarism.

2. Table 1 was stated to represent “a selection of events that took place after the figure above (ie the figure published by Fanelli (2009)) was established”, yet some listed references/ events were published/ occurred between 2005 and 2008.

3. It is reasonable to expect that increased capacity to autogenerate text and images will increase the numbers of fake papers, but I’m not aware of any evidence to support this. No reference is cited.

4. Table 2; “similar survey results”: it’s not clear how the listed studies are similar.

5. There are many unreferenced statements, eg page 9, “most rejected papers are published, just elsewhere”, page 19.

6. Some estimates of fake papers arise from small sample sizes (eg page 13).

7. The statement “The accumulation of papers assembled here is, frankly, haphazard” doesn’t inspire confidence in the resulting estimate.

8. “…it would be prudent to immediately reproduce the result presented here as a formal systematic review”- any systematic review seems premature without reliable estimates.

9. “The false positive rate (FPR) of detecting fake science is almost certainly quite low”- this seems unlikely to be correct. False positive rates depend on the methods used. Different methods will be required to detect fake papers in different disciplines, and these different methods could have very different false positive rates, particularly when comparing the application of manual versus automated methods that are applied without manual checking.

10. Page 2: I could not see the n=12 studies summarised in a single Table.

11. Page 10: “All relevant studies were included”…. “The list below is comprehensive but not necessarily exhaustive”- these statements contradict each other.

The title of the article makes a simple striking claim about the state of the scientific literature with a numerical estimate of the proportion of “fake” articles. Yet, by contrast to this title, in the text of the article, Heathers is highly critical of his own work.

James’ peer review of Heathers’ article

James Heathers often mentions the limitations of his research thus “peer-reviewing” his own article to the extent that he admits that this work is “incomplete”, “unsystematic” and “far flung”.

“This work is too incomplete to support responsible meta-analysis, and research that could more accurately define this figure does not exist yet. ~1 in 7 papers being fake represents an existential threat to the scientific enterprise.”

“While this is highly unsystematic, it produced a substantially higher figure. Correspondents reliably estimated 1-5% of all papers contain fabricated data, and 2-10% contain falsified results.”

“These values are too disparate to meta-analyze responsibly, and support only the briefest form of numerical summary: n=12 papers return n=16 individual estimates; these have a median of 13.95%, and 9 out of 16 of these estimates are between 13.4% and 16.9%. Given this, a rough approximation is that for any given corpus of papers, 1 in 7 (i.e. 14.3%) contain errors consistent with faking in at least one identifiable element.”

“The accumulation of papers collected here is, frankly, haphazard. It does not represent a mature body of literature. The papers use different methods of analyzing figures, data, or other features of scientific publications. They do not distinguish well between papers that have small problematic elements which are fake, or fake in their entirety. They analyze both small and large corpora of papers, which are in different areas of study and in journals of different scientific quality – and this greatly changes base rates;…”

“As a consequence, it would be prudent to immediately reproduce the result presented here as a formal systematic review. It is possible further figures are available after an exhaustive search, and also that pre registered analytical assumptions would modify the estimations presented.”

Heathers has also in an interview published in Retraction Watch (Chawla 2024) acknowledged pitfalls in this article such as:

“Heathers said he decided to conduct his study as a meta-analysis because his figures are “far flung.””

“They are a little bit from everywhere; it’s wildly nonsystematic as a piece of work,” he said.”

“Heathers acknowledged those limitations but argued that he had to conduct the analysis with the data that exist. “If we waited for the resources necessary to be able to do really big systematic treatments of a problem like this within a specific area, I think we’d be waiting far too long,” he said. “This is crucially underfunded.”

Built in opposition to Fanelli 2009, but it’s illogical

Heathers states in the abstract that his article is “in opposition” to Fanelli’s 2009 PloS One article (Fanelli 2009), yet that opposition is illogical and artificially constructed since there is no contradiction between 2% of scientists self-reporting having taking part in fabrication or falsification and an eventual much higher proportion of “fake scientific outputs”. Like most of what is wrong with Heather’s article, this is in fact acknowledged by the author who notes that the 2% figure “leaves us with no estimate of how much scientific output is fake” (bias in self-reporting, possibility of prolific authors, etc).

Fanelli 2009 is not cited in the way JH says it is cited

Whilst the opposition discussed above is illogical, it could be that the 2% figure is mis-cited by others as representing an estimate of fake scientific outputs thus probably underestimating the extent of fraud. Heathers suggests that this may indeed be the case, but also contradicts himself about how (Fanelli 2009), or the 2% figure coming from that publication, is typically used.

In one sentence, he writes that “the figure is overwhelmingly the salient cited fact in its 1513 citations” and that “this generally appears as some variant of “about 2% of scientists admitted to have fabricated, falsified or modified data or results at least once” (Frank et al. 2023)

whilst and in another sentence, he writes that “the typical phraseology used to express it – e.g. “the most serious types of misconduct, fabrication and falsification (i.e., data fraud), are relatively rare” (George 2016).

Those two sentences cited by Heathers are fundamentally different, the first one accurately reports that the 2% figure relates to individuals self-reporting, whilst the second one appears to relate to the prevalence of misconducts in the literature itself. How Fanelli 2009 is cited in the literature is an empirical question that can be studied by looking at citation contexts beyond the two examples given by Heathers. Given that a central justification for Heathers’ piece appears to be the misuse of this 2% figure, we sought to test whether this was the case.

A first surprise was that whilst the sentence attributed to (George 2016) can indeed be found in that publication (in the abstract), first it is not in a sentence citing (Fanelli 2009) nor the 2% figure, and, second, it is quoted selectively omitting a part of the sentence that nuances it considerably: “The evidence on prevalence is unreliable and fraught with definitional problems and with study design issues. Nevertheless, the evidence taken as a whole seems to suggest that cases of the most serious types of misconduct, fabrication and falsification (i.e., data fraud), are relatively rare but that other types of questionable research practices are quite common.” (Fanelli 2009) is discussed extensively by (George 2016), and some of the caveats, e.g. on self-reporting, are highlighted.

To go beyond those two examples, we constructed a comprehensive corpus of citation contexts, defined as the textual environment surrounding a paper's citation, including several words or sentences before and after the citation (see Methods section below). 737 citation contexts could be analysed. Out of those, the vast majority (533, or 72%) did not cite the 2% figure. Instead, they often referred to this article as a general reference together with other articles to make a broad point, or, focused on other numbers in particular those related to questionable research practices (Bordignon, Said, and Levy 2024). The 28% (204) citation contexts that did mention the 2% figure did so accurately in the majority of cases: 83% (170) of those did mention that it was self-reporting by scientists whilst 17% (34) of those, or 5% of the total citation contexts analysed were either ambiguous or misleading in that they suggested or claimed that the 2% figure related to scientific outputs.

Although the analysis above does not include all citation contexts, it is possible to conclude unambiguously that the 2% figure is not overwhelmingly the salient cited fact in relation to Fanelli 2009, and that when it is cited it is often accurately, i.e. as representing self-reporting by scientists. Whilst an exhaustive analysis is beyond the scope of this peer review, it is not uncommon to find in this corpus citations contexts that have an alarming tone about the seriousness of the problem of FFPs, e.g. “…a meta-analysis (Fanelli 2009) suggest that the few cases that do surface represent only the tip of a large iceberg." [DOI: 10.1177/0022034510384627]

Thus, the rationale for Heathers’ study appears to be misguided. The supposed lack of attention for the very serious problem of FFPs is not due to a minimisation of the situation fueled by a misinterpretation of Fanelli 2009. Importantly, even if that was the case, an attempt to draw attention by claiming that 1 in 7 papers are fake, a claim which according to the author himself is not grounded in solid facts, is not how the scientific literature should be used.

Methods for the construction of the corpus of citation contexts

We used Semantic Scholar, an academic database encompassing over 200 million scholarly documents from diverse sources including publishers, data providers, and web crawlers. Using the specific paper identifier for Fanelli's 2009 publication (d9db67acc223c9bd9b8c1d4969dc105409c6dfef), we queried the Semantic Scholar API to retrieve available citation contexts. Citation contexts were extracted from the "contexts" field within the JSON response pages, (see technical specifications).

The query looks like this: semanticscholar.org

The broad coverage of Semantic Scholar does not imply that citation contexts are always retrieved. The Semantic Scholar API provided citation contexts for only 48% of the 1452 documents citing the paper. To get more, we identified open access papers among the remaining 52% citing papers, retrieved their PDF location and downloaded the files. We used Unpaywall API, which is a database to be queried with a DOI in order to get open access information about a document. The query looks like this.

We downloaded 266 PDF files and converted them to text format using an online bulk PDF-to-text converter. These files were then processed using TXM, a specialized textual analysis tool. We used its concordancer function to identify the term "Fanelli" as a pivot term and check the reference being the good one (the 2009 paper in PlosOne). We did manual cleaning and appended the citation contexts to the previous corpus.

Through this comprehensive methodology, we ultimately identified 824 citation contexts, representing 54% (784) of all documents citing Fanelli's 2009 paper. This corpus comprised 48% of contexts retrieved from Semantic Scholar and an additional 6% obtained through semi-manual extraction from open access documents. 87 of those contexts were excluded from the analysis for a range of reasons including: context too short to conclude, language neither English nor French (shared languages of the authors of this review), duplicate documents (e.g. preprints), etc, leaving us with 737 contexts. They were first classified manually in two categories, those mentioning the 2% figure and those which did not. Then, for the first category, they were further classified manually in two categories depending on whether the figure was appropriately assigned to self-reporting of researchers or rather misleadingly suggesting that the 2% applied to research outputs.

Contributions

Investigation: FB collected the citation contexts.<br /> Data curation and formal analysis: RL and MS<br /> Writing – review & editing: RL, MS and FB

References

Bordignon, Frederique, Maha Said, and Raphael Levy. 2024. “Citation Contexts of [How Many Scientists Fabricate and Falsify Research? A Systematic Review and Meta-Analysis of Survey Data, DOI: 10.1371/Journal.Pone.0005738].” Zenodo. https://doi.org/10.5281/zenodo.14417422.

Chawla, Dalmeet Singh. 2024. “1 in 7 Scientific Papers Is Fake, Suggests Study That Author Calls ‘Wildly Nonsystematic.’” Retraction Watch (blog). September 24, 2024. https://retractionwatch.com/2024/09/24/1-in-7-scientific-papers-is-fake-suggests-study-that-author-calls-wildly-nonsystematic/.

Fanelli, Daniele. 2009. “How Many Scientists Fabricate and Falsify Research? A Systematic Review and Meta-Analysis of Survey Data.” PLOS ONE 4 (5): e5738. https://doi.org/10.1371/journal.pone.0005738.

Frank, Fabrice, Nans Florens, Gideon Meyerowitz-Katz, Jérôme Barriere, Éric Billy, Véronique Saada, Alexander Samuel, Jacques Robert, and Lonni Besançon. 2023. “Raising Concerns on Questionable Ethics Approvals - a Case Study of 456 Trials from the Institut Hospitalo-Universitaire Méditerranée Infection.” Research Integrity and Peer Review 8 (1): 9. https://doi.org/10.1186/s41073-023-00134-4.

George, Stephen L. 2016. “Research Misconduct and Data Fraud in Clinical Trials: Prevalence and Causal Factors.” International Journal of Clinical Oncology 21 (1): 15–21. https://doi.org/10.1007/s10147-015-0887-3.

Reviewer #1 (Public review):

Summary:

In this manuscript, Cho et al. present a comprehensive and multidimensional analysis of glutamine metabolism in the regulation of B cell differentiation and function during immune responses. They further demonstrate how glutamine metabolism interacts with glucose uptake and utilization to modulate key intracellular processes. The manuscript is clearly written, and the experimental approaches are informative and well-executed. The authors provide a detailed mechanistic understanding through the use of both in vivo and in vitro models. The conclusions are well supported by the data, and the findings are novel and impactful. I have only a few, mostly minor, concerns related to data presentation and the rationale for certain experimental choices.

Detailed Comments:

(1) In Figure 1b, it is unclear whether total B cells or follicular B cells were used in the assay. Additionally, the in vitro class-switch recombination and plasma cell differentiation experiments were conducted without BCR stimulation, which makes the system appear overly artificial and limits physiological relevance. Although the effects of glutamine concentration on the measured parameters are evident, the results cannot be confidently interpreted as true plasma cell generation or IgG1 class switching under these conditions. The authors should moderate these claims or provide stronger justification for the chosen differentiation strategy. Incorporating a parallel assay with anti-BCR stimulation would improve the rigor and interpretability of these findings.

(2) In Figure 1c, the DMK alone condition is not presented. This hinders readers' ability to properly asses the glutaminolysis dependency of the cells for the measured readouts. Also, CD138+ in developing PCs goes hand in hand with decreased B220 expression. A representative FACS plot showing the gating strategy for the in vitro PCs should be added as a supplementary figure. Similarly, division number (going all the way to #7) may be tricky to gate and interpret. A representative FACS plot showing the separation of B cells according to their division numbers and a subsequent gating of CD138 or IgG1 in these gates would be ideal for demonstrating the authors' ability to distinguish these populations effectively.

(3) A brief explanation should be provided for the exclusive use of IgG1 as the readout in class-switching assays, given that naïve B cells are capable of switching to multiple isotypes. Clarifying why IgG1 was preferentially selected would aid in the interpretation of the results.

(4) The immunization experiments presented in Figures 1 and 2 are well designed, and the data are comprehensively presented. However, to prevent potential misinterpretation, it should be clarified that the observed differences between NP and OVA immunizations cannot be attributed solely to the chemical nature of the antigens - hapten versus protein. A more significant distinction lies in the route of administration (intraperitoneal vs. intranasal) and the resulting anatomical compartment of the immune response (systemic vs. lung-restricted). This context should be explicitly stated to avoid overinterpretation of the comparative findings.

(5) NP immunization is known to be an inducer of an IgG1-dominant Th2-type immune response in mice. IgG2c is not a major player unless a nanoparticle delivery system is used. However, the authors arbitrarily included IgG2c in their assays in Figures 2 and 3. This may be confusing for the readers. The authors should either justify the IgG2c-mediated analyses or remove them from the main figures. (It can be added as supplemental information with proper justification).

(6) Similarly, in affinity maturation analyses, including IgM is somewhat uncommon. I do not see any point in showing high affinity (NP2/NP20) IgMs (Figure 3d), since that data probably does not mean much.

(7) Following on my comment for the PC generation in Figure 1 (see above), in Figure 4, a strategy that relies solely on CD40L stimulation is performed. This is highly artificial for the PC generation and needs to be justified, or more physiologically relevant PC generation strategies involving anti-BCR, CD40L, and various cytokines should be shown.

(8) The effects of CB839 and UK5099 on cell viability are not shown. Including viability data under these treatment conditions would be a valuable addition to the supplementary materials, as it would help readers more accurately interpret the functional outcomes observed in the study.

(9) It is not clear how the RNA seq analysis in Figure 4h was generated. The experimental strategy and the setup need to be better explained.

Reviewer #2 (Public review):

Summary:

In this manuscript, the authors investigate the functional requirements for glutamine and glutaminolysis in antibody responses. The authors first demonstrate that the concentrations of glutamine in lymph nodes are substantially lower than in plasma, and that at these levels, glutamine is limiting for plasma cell differentiation in vitro. The authors go on to use genetic mouse models in which B cells are deficient in glutaminase 1 (Gls), the glucose transporter Slc2a1, and/or mitochondrial pyruvate carrier 2 (Mpc2) to test the importance of these pathways in vivo.

Interestingly, deficiency of Gls alone showed clear antibody defects when ovalbumin was used as the immunogen, but not the hapten NP. For the latter response, defects in antibody titers and affinity were observed only when both Gls and either Mpc2 or Slc2a1 were deleted. These latter findings form the basis of the synthetic auxotrophy conclusion. The authors go on to test these conclusions further using in vitro differentiations, Seahorse assays, pharmacological inhibitors, and targeted quantification of specific metabolites and amino acids. Finally, the authors document reduced STAT3 and STAT1 phosphorylation in response to IL-21 and interferon (both type 1 and 2), respectively, when both glutaminolysis and mitochondrial pyruvate metabolism are prevented.

Strengths:

(1) The main strength of the manuscript is the overall breadth of experiments performed. Orthogonal experiments are performed using genetic models, pharmacological inhibitors, in vitro assays, and in vivo experiments to support the claims. Multiple antigens are used as test immunogens--this is particularly important given the differing results.

(2) B cell metabolism is an area of interest but understudied relative to other cell types in the immune system.

(3) The importance of metabolic flexibility and caution when interpreting negative results is made clear from this study.

Weaknesses:

(1) All of the in vivo studies were done in the context of boosters at 3 weeks and recall responses 1 week later. This makes specific results difficult to interpret. Primary responses, including germinal centers, are still ongoing at 3 weeks after the initial immunization. Thus, untangling what proportion of the defects are due to problems in the primary vs. memory response is difficult.

(2) Along these lines, the defects shown in Figure 3h-i may not be due to the authors' interpretation that Gls and Mpc2 are required for efficient plasma cell differentiation from memory B cells. This interpretation would only be correct if the absence of Gls/Mpc2 leads to preferential recruitment of low-affinity memory B cells into secondary plasma cells. The more likely interpretation is that ongoing primary germinal centers are negatively impacted by Gls and Mpc2 deficiency, and this, in turn, leads to reduced affinities of serum antibodies.

(3) The gating strategies for germinal centers and memory B cells in Supplemental Figure 2 are problematic, especially given that these data are used to claim only modest and/or statistically insignificant differences in these populations when Gls and Mpc2 are ablated. Neither strategy shows distinct flow cytometric populations, and it does not seem that the quantification focuses on antigen-specific cells.

(4) Along these lines, the conclusions in Figure 6a-d may need to be tempered if the analysis was done on polyclonal, rather than antigen-specific cells. Alum induces a heavily type 2-biased response and is not known to induce much of an interferon signature. The authors' observations might be explained by the inclusion of other ongoing GCs unrelated to the immunization.

Reviewer #3 (Public review):

Summary:

In their manuscript, the authors investigate how glutaminolysis (GLS) and mitochondrial pyruvate import (MPC2) jointly shape B cell fate and the humoral immune response. Using inducible knockout systems and metabolic inhibitors, they uncover a "synthetic auxotrophy": When GLS activity/glutaminolysis is lost together with either GLUT1-mediated glucose uptake or MPC2, B cells fail to upregulate mitochondrial respiration, IL 21/STAT3 and IFN/STAT1 signaling is impaired, and the plasma cell output and antigen-specific antibody titers drop significantly. This work thus demonstrates the promotion of plasma cell differentiation and cytokine signaling through parallel activation of two metabolic pathways. The dataset is technically comprehensive and conceptually novel, but some aspects leave the in vivo and translational significance uncertain.

Strengths:

(1) Conceptual novelty: the study goes beyond single-enzyme deletions to reveal conditional metabolic vulnerabilities and fate-deciding mechanisms in B cells.

(2) Mechanistic depth: the study uncovers a novel "metabolic bottleneck" that impairs mitochondrial respiration and elevates ROS, and directly ties these changes to cytokine-receptor signaling. This is both mechanistically compelling and potentially clinically relevant.

(3) Breadth of models and methods: inducible genetics, pharmacology, metabolomics, seahorse assay, ELISpot/ELISA, RNA-seq, two immunization models.

(4) Potential clinical angle: the synergy of CB839 with UK5099 and/or hydroxychloroquine hints at a druggable pathway targeting autoantibody-driven diseases.

Weaknesses:

(1) Physiological relevance of "synthetic auxotrophy"

The manuscript demonstrates that GLS loss is only crippling when glucose influx or mitochondrial pyruvate import is concurrently reduced, which the authors name "synthetic auxotrophy". I think it would help readers to clarify the terminology more and add a concise definition of "synthetic auxotrophy" versus "synthetic lethality" early in the manuscript and justify its relevance for B cells.

While the overall findings, especially the subset specificity and the clinical implications, are generally interesting, the "synthetic auxotrophy" condition feels a little engineered. Therefore, the findings strongly raise the question of the likelihood of such a "double hit" in vivo and whether there are conditions, disease states, or drug regimens that would realistically generate such a "bottleneck". Hence, the authors should document or at least discuss whether GC or inflamed niches naturally show simultaneous downregulation/lack of glutamine and/or pyruvate. The authors should also aim to provide evidence that infections (e.g., influenza), hypoxia, treatments (e.g., rapamycin), or inflammatory diseases like lupus co-limit these pathways.

It would hence also be beneficial to test the CB839 + UK5099/HCQ combinations in a short, proof-of-concept treatment in vivo, e.g., shortly before and after the booster immunization or in an autoimmune model. Likewise, it may also be insightful to discuss potential effects of existing treatments (especially CB839, HCQ) on human memory B cell or PC pools.

(2) Cell survival versus differentiation phenotype

Claims that the phenotypes (e.g., reduced PC numbers) are "independent of death" and are not merely the result of artificial cell stress would benefit from Annexin-V/active-caspase 3 analyses of GC B cells and plasmablasts. Please also show viability curves for inhibitor-treated cells.

(3) Subset specificity of the metabolic phenotype

Could the metabolic differences, mitochondrial ROS, and membrane-potential changes shown for activated pan-B cells (Figure 5) also be demonstrated ex vivo for KO mouse-derived GC B cells and plasma cells? This would also be insightful to investigate following NP-immunization (e.g., NP+ GC B cells 10 days after NP-OVA immunization).

(4) Memory B cell gating strategy

I am not fully convinced that the memory-B-cell gate in Supplementary Figure 2d is appropriate. The legend implies the population is defined simply as CD19+GL7-CD38+ (or CD19+CD38++?), with no further restriction to NP-binding cells. Such a gate could also capture naïve or recently activated B cells. From the descriptions in the figure and the figure legend, it is hard to verify that the events plotted truly represent memory B cells. Please clarify the full gating hierarchy and, ideally, restrict the MBC gate to NP+CD19+GL7-CD38+ B cells (or add additional markers such as CD80 and CD273). Generally, the manuscript would benefit from a more transparent presentation of gating strategies.

(5) Deletion efficiency

mRNA data show residual GLS/MPC2 transcripts (Supplementary Figure 8). Please quantify deletion efficiency in GC B cells and plasmablasts.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

We would like to thank the reviewers for taking the time to review our manuscript and for providing valuable comments on how to improve it. We are pleased to see that both reviewers recognize the novelty and importance of our study, its conceptual advance and potential clinical significance. They also noted the novelty and value of our functional mechanistic approach using epigenetic editing. Below, we provide a point-by-point response to their questions and points raised. The changes introduced in response to their feedback are highlighted in yellow in the revised manuscript file.

Point-by-point description of the revisions

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

Summary This study by Prada et al. aimed to explore DNA methylation and gene expression in primary EpCAMhigh/PDPNlow cells, consisting of for (probably) the largest part of AT2 cells, to understand the molecular mechanisms behind the impaired regeneration of alveolar epithelial progenitor cells in COPD. They found that higher or lower promoter methylation in COPD-associated cells was inversely correlated with changes in gene expression, with interferon signaling emerging as one of the most upregulated pathways in COPD. IRF9 was identified as the master regulator of interferon signaling in COPD. Targeted DNA demethylation of IRF9 in an A549 cell line resulted in a robust activation of its downstream target genes, including OAS1, OAS3, PSMB8, PSMB9, MX2 and IRF7, demonstrating that demethylation of IRF9 is sufficient to activate the IFN signaling pathway, validating IRF9 as a master regulator of IFN signaling in (alveolar) epithelial cells.

Major comments:

• To remove airways (and blood vessels) completely from the lung tissue is difficult, if not impossible. This means that the assumption that the sorted EpCAMpos/PDPNlow cells primarily consisted of AT2 cells remains valid only if a quantitative analysis is conducted on the proportion of HT2-280pos cells in all samples in cytospins to exclude any significant contamination from bronchial epithelial cells. If authors cannot demonstrate >95% pure HT-280-positive cells, then the key conclusions suggesting that the epigenetic regulation of the IFN pathway might be crucial in AT2 progenitor cell regeneration could also potentially apply to bronchial progenitor cells. In addition, if >95% purity cannot be demonstrated, the data should be adjusted to account for differences in cell type composition.

__Response: __

We thank the reviewer for raising this important point. Although, as pointed out by the reviewer, we cannot guarantee that our sorted cells do not contain a minor contamination from respiratory / terminal bronchial cells, we carefully selected donors, tissue regions, and sorting strategy to ensure the highest possible enrichment of AT2 cells, as we explain below. We have now expanded the methods and results section and covered this point in the manuscript discussion.

• The lung tissue pieces we received were distal, as evidenced by the presence of pleura. We collected representative tissue pieces for histology to validate sample quality. Our protocol includes a dissection of all visible airways and vessels using a dissecting microscope, which were cryopreserved separately from distal parenchyma. Hence, the starting material for tissue dissociation was depleted from airways and vessels. The importance of vessel/airway removal for enrichment of distal alveolar cells was established by Tata's group (PMID: 35712012).
• We selected the AT2 sorting protocol (EpCAMpos/PDPNlow) based on previous publications that used tissue from both healthy and COPD lungs to separate AT2 cells from AT1 and airway basal cells, as AT1 and basal cells are both PDPNhigh (PMID: 22033268, PMID: 23117565; PMID: 35078977). This protocol was favoured due to the lack of information about HT2-280 expression and distribution in COPD lungs.
• The sort quality for each sample was assessed by the FACS analysis (back sorting) of the sorted cells, where we observed 95-97% purity (EpCAMpos/PDPNlow, __ 1G __shown below). In addition, we validated the sorting protocol and high AT2 enrichment from both no COPD and COPD tissues by immunostaining the FACS-sorted cells with HT2-280, an AT2 marker widely used in the field (strategy suggested by the reviewer) and observed that close to 100% of cells were positive for this marker (__Fig. 1H __shown below). However, we could not do it retrospectively for those patients, where we didn't have enough material. Sorting primary AT2 from small tissue pieces is challenging, and we need at least 20.000 cells to obtain high-quality methylation & RNA-seq data.

• AT2 marker genes (ABCA3, LPCAT1, LAMP3 and the surfactant genes SFTPA2, SFTPB and SFTPC) were among the top highly expressed genes in our RNA-seq data and were not significantly changed in COPD (please see expression data in __ S2A__ in the manuscript, and below for convenience), as well as Table 6, providing further evidence that the sorted cells carry a strong AT2 transcriptional signature. Fig. 1G* FACS plot examples showing the analysis of sorted AT2 cells (back sorting) from control (blue) and COPD (green) donors displayed over total cell lung suspensions (grey) H Representative IF staining of HT2-280 expression in sorted AT2 cells from no COPD (top) and COPD (bottom) donors. Nuclei (blue) were stained with DAPI, scale bars=20µm __Fig. S2A __Normalized read counts from RNA-seq data for AT2-specific genes in sorted AT2 cells from each donor (dots). Data points represent normalised counts from no COPD (blue), COPD I (light green) and COPD II-IV (dark green). Group median is shown as a black bar. *

• In agreement with a previous study which profiled bulk AT2 using expression arrays (PMID: 23117565), we also observed upregulation of IFN signaling pathway in COPD AT2s. The enrichment of IFNα/β signature was also observed in COPD in the inflammatory AT2 cluster (iAT2) in a recent scRNA-seq study (PMID: 36108172). As part of the revision, we compared the IFN gene signature identified in our bulk AT2 RNA-seq with a recent scRNA-seq study (published after the submission of our manuscript, PMID: 39147413) that profiled EpCAMpos cells from COPD and non-smoker donor lungs. We observed an upregulation of our IFN signature genes in AT2 in COPD (mostly in AT2c and rbAT2 subsets), suggesting that similar signatures were observed in COPD AT2s in this dataset as well (please see __ S4E-F__ below). ____Figure S4E Expression values for the indicated genes of the IFN pathway from an external scRNA-seq dataset of AT2 cells from COPD patients and healthy controls (Hu et al, 2024). Y-axis shows log-normalized gene expression levels. F. Combined gene set score of the genes shown in (E) in different subsets of AT2 cells from Hu et al, 2024. The IFN signature genes were identified in our integrative analysis of TWGBS and RNA-seq in sorted AT2 cells.

• We have also carefully examined DNA methylation profiles across all samples. The density plots of our T-WGBS DNA methylation data are very similar among the individual samples in all 3 groups, indicating that the sorted cells consist mostly of a single cell type, as there are no obvious intermediate (25-75%) methylation peaks, as observed in cell mixtures ( 2A and the panel below). No reference DNA methylation profiles are available for respiratory or terminal bronchial cells; hence, we cannot compare how epigenetically different these cells would be from AT2 nor perform a deconvolution for potential minor contamination with distal airway cells. *Figure: DNA methylation density plots of sorted EpCAMpos/PDPNneg cells from no COPD (blue, n=3), COPD I (light green, n=3) and COPD II-IV (dark green, n=5) showing a homogeneous methylation pattern and low abundance at intermediate (25%-75%) methylation values across all profiled samples, indicating that the sorted cells were mostly of a single cell type. *

• We have now added a sentence to the limitations section of the discussion to cover that point specifically. CHANGES IN THE MANUSCRIPT:

AT2 cells were isolated by fluorescence-activated cell sorting (FACS) from cryopreserved distal lung parenchyma, depleted of visible airways and vessels of three no COPD controls, three COPD I and five COPD II-IV patients as previously described (24, 52, 53)

The isolated cells were positive for HT2-280, a known AT2 marker (54)*, as confirmed by immunofluorescence (Fig. 1H), validating the identity and high enrichment of the isolated AT2 populations. ** *

*Known AT2-specific genes, including ABCA3, LAMP3 and surfactant genes (SFTPA2, SFTPB and SFTPC) were among the top highly expressed genes and were not significantly changed in COPD AT2s (Fig. S2A, Table 6), further confirming the AT2-characteristic transcriptional signature of our isolated cells. *

However, 5-AZA is a global demethylating agent, and the observed effects may not be direct. To validate the epigenetic regulation of central AT2 pathways further, we took advantage of locus-specific epigenetic editing technology *(73). We focused on the IFN pathway because it was the most significantly enriched Gene Ontology (GO) term in our integrative analysis of TWGBS and RNA-seq data. Several IFN pathway members had associated hypomethylated DMRs within promoter-proximal regions and concomitant increased gene expression (Fig. 4C and S2C). Additionally, we confirmed the elevated expression of IFN-related genes with associated DMRs identified in our study in AT2 cells and AT2 cell subclusters from a recently published scRNA-seq cohort (74) (Fig. S4E-F). *

We observed upregulation of multiple IFN genes in AT2 in COPD, consistent with a previous expression array study (24). IFNα/β signaling was also enriched in COPD patients in the inflammatory AT2 cluster (iAT2) in a recent scRNA-seq study (84) and our INF signature genes were also upregulated in AT2c and AT2rb subsets in COPD, identified by another scRNA-seq study recently (74)*. ** *

Finally, despite careful removal of airways from distal lung tissue using a dissecting microscope, we cannot exclude the presence of some terminal/respiratory bronchiole cells in our FACS-isolated EpCAMpos/PDPNlow population. Recent scRNA-seq studies provided an unprecedented resolution and identified several epithelial subpopulations and transitional cells residing in the terminal/respiratory bronchioles and alveoli, including respiratory airway secretory cells (93), terminal airway-enriched secretory cells (28), terminal bronchiole-specific alveolar type-0 (AT0) (70), and emphysema-specific AT2 cells (74). These cells may contribute to alveolar repair in healthy and COPD lungs; however, with our bulk DNA methylation and RNA-seq study, we are unable to resolve all these subpopulations. Future development of single-cell methylation and non-reference-based algorithms for DNA methylation deconvolution will enable deeper epigenetic phenotyping of specific AT2 and bronchiolar cell subsets.

(Methods) Validation of IFN gene upregulation in a published scRNA-seq dataset

scRNA-seq data from (74), generously provided by M. Köningshoff, were processed using the default Seurat workflow (117). Expression of IFN-related genes was extracted and plotted as log-normalised gene expression levels in AT2 cells from control and COPD donors. Seurat's AddModuleScore() function was used to compute a gene set score for a custom IFN program using the genes listed in __Fig. S4E __and to analyse the IFN gene set scores in AT2 cell subclusters identified in (74). Briefly, average gene expression scores were computed for the gene set of interest, and the expression of control features (randomly selected) was subtracted as described in (118).

Fig. S4E and F: E. Expression values for the indicated genes of the IFN pathway from an external scRNA-seq dataset of AT2 cells from COPD patients and healthy controls (74). Y-axis shows log-normalized gene expression levels. F. Combined gene set score of the genes shown in (E) in different subsets of AT2 cells from (74). The IFN signature genes were identified in our integrative analysis of TWGBS and RNA-seq in sorted AT2 cells.

• The overrepresentation of several keratins (KRT5, KRT14, KRT16, KRT17), mucins (MUC12, MUC13, MUC16, MUC20) and the transcription factor FoxJ1 is now attributed by the authors to a possible dysregulation of AT2 identity and differentiation in COPD (lines 282 - 284) where they cite refs 28, 69, 70. Authors try to support this with IF double stains for KRT5 and HT-280 to identify co-expression of KRT5 and HT2-280 in lung tissue (Figure S2H). However, the evidence for the co-expression of both markers could be presented more convincingly.

__Response: __

We found the potential co-expression of airway and alveolar markers in COPD lungs interesting and hence included it in the original manuscript. The initial discovery came from our bulk RNA-seq data, where we observed upregulation of several genes typically found in more proximal airways in COPD (mentioned above by the reviewer). Of note, some of them (e.g., FoxJ1) are expressed at very low levels. Following reviewer's comments, to validate possible colocalization of AT2 and airway markers on protein level, we performed further IF analysis. We took Z-stack images to demonstrate the co-localization of HT2-280 and Krt5 more convincingly and co-stained the same tissue regions with SCGB3A2 (a TASC/distal airway cell marker, PMID 36796082). Even though these are rare events, we were able to reproduce the existence of HT2-280/Krt5 positive, SCGB3A2 negative cells in the alveoli of COPD patients on the protein level (__Fig. S2H __and panels below). Although interesting, we decided to keep this finding in the supplement and did not include it in the discussion to focus the story on the epigenetic regulation of the IFN pathway, which is the main discovery of our study. We will investigate this observation in future studies.

Figure S2H and here: Examples of HT2-280/Krt5 double positive cells. Top, immunofluorescence staining of the alveolar region of a COPD II donor showing the existence of AT2 cells (HT2-280 positive (red), which are SCGB3A2 negative (green, left) but KRT5 positive (green, right). In conclusion, double-positive HT2-280/KRT5 cells are rare but present in the alveoli of COPD patients. Magnification: 20x. Scale bar: 50 µm. Bottom, Z-stack images highlighting HT2-280 (red) and KRT5 (green) double-positive cells at 63x magnification. Scale bar: 5 µm.

CHANGES IN THE MANUSCRIPT:

In addition, we observed an upregulation of several keratins (KRT5, KRT14, KRT16, KRT17) and mucins (MUC12, MUC13, MUC16, MUC20), suggesting a potential dysregulation of alveolar epithelial cell differentiation programs in COPD (Table 6, Fig. S2F). Immunofluorescence staining confirmed the presence of KRT5-positive cells in the distal lung in COPD and identified cells positive for both KRT5 and HT2-280 (Fig. S2H). Collectively, these results indicate a dysregulation of stemness and identity in the alveolar epithelial cells in COPD.

Fig. S2H legend: The zoomed-in panel (right corner, bottom) demonstrates the presence of rare HT2-280/KRT5 double-positive cells in the alveoli of COPD patients.* Slides were counterstained with DAPI, scale bars = 50µm, 20µm or 5µm, as displayed in images. *

• Double staining for KRT5 and HT2-280 did highlight the proximity of both cell types in lung tissue, underscoring the challenge of removing airways (including the smaller and terminal bronchi) from the tissue. In addition, HT-280/KRT5 co-expression is not consistent with recent studies from refs 28, 69, 70 where other markers for distal airway cell transition, such as SCGB3A2 and BPIFB1, have been demonstrated, which were not investigated in this study.

Response:

We provided a general overview of the different signatures observed in our data, but we could not validate every deregulated pathway or gene. We include the relevant tables detailing all differentially expressed genes and differentially methylated regions to enable and encourage the community to follow up on the data in subsequent studies.

As demonstrated above, we detect the co-occurrence of HT2-280/KRT5 staining on the protein level in the same cells in the alveoli of COPD patients. We would like to emphasize that alveolar epithelial cell identity in CODP lungs has not been investigated in detail on the protein or RNA level, and HT2-280/KRT5 co-expression/co-localization has not been directly tested in the studies mentioned by the reviewer since, among other reasons, the gene encoding HT2-280 has not been identified. Notably, a recent study (published after the submission of our manuscript) focusing on enriched epithelial cells from the distal lungs of COPD patients (PMID 35078977), identified an emphysema-specific AT2 subtype co-expressing the AT2 marker SFTPC and distal airway cell transition marker SCGB3A2, indicating that disease-specific AT2 populations with possible co-occurrence of AT2 and airway markers exist. In our dataset, SCGB3A2 was not deregulated (log2 fold change=0.22, adj p-value= 0.47), as shown in Table 6, and the HT2-280/Krt5 positive cells were negative for SCGB3A2 in our IF staining (see above).

BPIFB1 is one of the antimicrobial peptides genes with an associated DMR and is significantly upregulated in COPD cells in our study (log2 fold change=1.17, adj p-value=0.0016), as shown in the supplementary figure Fig S4C and here below for convenience.

Figure S4C Fold-change in gene expression of BPIFB1 in AT2 cells in COPD (RNA-seq) and A549 cells treated with 0.5µM AZA (RT-qPCR) compared to control samples. Left, RNA-seq data from AT2 cells (no COPD, blue, n=3; COPD II-IV, green, n=5). Right, A549 treated with AZA (orange, n=3) compared to control DMSO-treated cells (grey, n=3). The group median is shown as a black bar.

• The small (and not evenly divided) sample size of both COPD and non-COPD specimens may lead to a higher risk for false positive results as adjustments for multiple testing typically rely on the number of comparisons, and small sample sizes may not provide enough data points to adequately control for this.

__Response: __

We acknowledge the problem of testing for multiple traits with relatively small numbers of samples. The availability of donor tissue, especially from non-COPD and COPD-I donors, was limited, and we applied very strict donor matching and quality control criteria for sample inclusion to avoid additional variability and confounding factors. The importance of strict quality control in selecting appropriate control samples was highlighted in our previous study (PMID: 33630765), where we demonstrated that approximately 50% of distal lung tissue from cancer patients with normal spirometry has pathological changes. Hence, we believe that the quality of the tissue was paramount to the reliability of the data. Strict quality control and sample matching for multiple parameters, including age, BMI, smoking status and smoking history (critical for DNA methylation studies), and cancer type (for background tissue), is a key strength of our approach, but it inevitably limited our sample size.

First, all samples were cryopreserved and then processed in parallel in groups of 1 non-COPD and 2-3 COPD samples. This process included tissue dissociation, FACS sorting, back sorting (always), and immunofluorescence staining (when enough material was available). Cell pellets were stored at -80{degree sign}C until the entire cohort was ready for sequencing. This was done to limit the potential variation introduced by processing and sorting. RNA and DNA isolations were performed in parallel for all the sorted cell pellets, which were then sequenced as a single batch.

During data analysis, we applied stringent cutoffs for DMR detection to reduce the risk of false positives due to multiple comparisons and a small sample size. Specifically, we filtered for regions with at least 10% methylation difference and containing at least 3 CpGs. Additionally, we applied a non-parametric Wilcoxon test using average DMR methylation levels to remove potentially false-positive regions, as the t-statistic is not well suited for non-normally distributed values, as expected at very low/high (close to 0% / 100%) methylation levels. A significance level of 0.1 has been used. Therefore, we are confident that the rigorous analysis and strict criteria applied in this study allowed us to detect trustworthy DMRs that we could further functionally validate using epigenetic editing. All the details of the DMR analysis are provided in the methods section. To address this point and limitation, we have added the following paragraphs in the discussion section of the manuscript:

CHANGE IN THE MANUSCRIPT:

*The strengths of our study include the use of purified human alveolar type 2 epithelial progenitor cells from a well-matched and carefully validated cohort of human samples, including mild and severe COPD patients, providing high relevance to human COPD. *

However, we acknowledge several limitations of our study that warrant further investigation. First, the sample size was small. The use of strict quality criteria for donor selection limited the available samples, particularly for the ex-smoker control group. This resulted in an unequal distribution of COPD and control samples. This impacts the power of statistical analysis, particularly in the WGBS analysis, where millions of regions genome-wide are tested. Nevertheless, the clear negative correlation between promoter methylation and corresponding gene expression highlights the robustness of the DMR selection. Additionally, we were able to experimentally validate interferon-associated DMRs using epigenetic editing, highlighting the power of integrated epigenetic profiling in identifying disease-relevant regulators.

__Minor suggestions for improvement __

__Introduction __ • In general, refer to the actual experimental studies rather than review papers where appropriate.

Response:

We have now carefully checked all the references and amended them to refer to experimental studies when required.

• Clearly specify whether a study was conducted in mice or humans, as this distinction is crucial for understanding the relevance of the findings to COPD.

__Response: __

All our experiments were performed with human lung cells and tissues. No mouse samples were used. As suggested, we have now clearly stated that our study was performed using human tissue samples and cells in different parts of the manuscript, including the discussion, where we now explicitly highlight the strengths and limitations of our study.

CHANGES IN THE MANUSCRIPT:

...we generated whole-genome DNA methylation and transcriptome maps of sorted human primary alveolar type 2 cells (AT2) at different disease stages.

However, the regulatory circuits that drive aberrant gene expression programs in human AT2 cells in COPD are poorly understood

Therefore, we set out to profile DNA methylation of human AT2 cells at single CpG-resolution across COPD stages.

...*suggesting that aberrant epigenetic changes may drive COPD phenotypes in human AT2. *

To identify genome-wide DNA methylation changes associated with COPD in purified human AT2 cells...

The similarity of the methylation and gene expression profiles in the PCAs suggested that epigenetic and transcriptomic changes in human AT2 cells during COPD might be interrelated ...

*In this work, we demonstrate that genome-wide DNA methylation changes occurring in human AT2 cells may drive COPD pathology by dysregulating key pathways that control inflammation, viral immunity and AT2 regeneration. *

*Using high-resolution epigenetic profiling, we uncovered widespread alterations of the DNA methylation landscape in human AT2 cells in COPD that were associated with global gene expression changes. *

*Currently, it is unclear how cigarette smoking leads to changes in DNA methylation patterns in human AT2 *

The strengths of our study include the use of purified human alveolar epithelial progenitor cells from a well-matched and carefully validated cohort of human samples, including mild and severe COPD patients, providing high relevance to human COPD.

__Methods __ • Line 473, here is meant 3 ex-smoker controls instead of smoker controls?

__Response: __

All donors (no COPD and COPD) used in our study are ex-smokers. Matching the samples with regard to smoking status and history is critical for epigenetic studies, as cigarette smoke profoundly affects DNA methylation genome-wide (PMID: 38199042, PMID: 27651444). This has now been clarified in the revised manuscript.

CHANGE IN THE MANUSCRIPT____:

Of note, we included only ex-smokers in our profiling to avoid acute smoking-induced inflammation as a confounding factor (50)*. *

Importantly, we matched the smoking status and smoking history of all donors, which is key in epigenetic studies, as cigarette smoking profoundly impacts the DNA methylation landscape of tissues (96).

In total, 3 ex-smoker controls (no COPD), 3 mild COPD donors ex-smokers (GOLD I, COPD I) and 5 moderate-to-severe COPD donors ex-smokers (GOLD II-IV, COPD II-IV) were profiled (Fig. 1A-C, Table 1)

__Discussion __ • A list of limitation should be added to the discussion. One is the use of the alveolar cell line A549, which produces mucus, a characteristic more commonly associated with bronchial epithelial cells. (ref 43)l530:

__Response: __

The profiling was performed using purified primary human alveolar epithelial progenitor cells. For technical reasons, A549 cells were only used for validation of the results using epigenetic editing. The A549 phenotype depends on the growth medium used, in our case, Ham's F-12 medium, which is recommended for long-term A549 culture and promotes multilamellar body formation and differentiation toward an AT2-like phenotype (PMID: 27792742)__. __We are developing epigenetic editing technology for use in primary lung cells; however, the approach currently relies on the high efficiency of transient transfections, which cannot yet be achieved with primary adult AT2 cells. We were positively surprised by how well the methylation data obtained from patient AT2s translated into mechanistic insights when using A549 cells, despite being a cancer cell line. This suggests that the fundamental mechanisms of epigenetic regulation of IRF9 and the IFN signaling pathway are conserved between A549 and primary AT2 cells.

• Another limitation to consider is that cells were isolated primarily from individuals with lung cancer, except for patients with COPD stage IV. In particular as COPD stage II and IV samples were taken together. And discuss the small and unevenly divided sample size

__Response: __

We thank the reviewer for bringing up this important point, which we carefully considered when designing our study. To match our samples across the cohort, all the no-COPD, COPD I, and two of the COPD II-IV samples were obtained from cancer resections. In addition to other characteristics, like age, BMI and smoking status, we also matched the donors by cancer type (all profiled donors had squamous cell carcinoma). We collected lung tissue as far away from the carcinoma as possible and sent representative pieces for histological analysis by an experienced lung pathologist to confirm the absence of visible tumours. In addition, to ensure that our data represents COPD-relevant signatures, we intentionally included samples from three COPD donors undergoing lung resections (without a cancer background) in the profiling.

Following the reviewer's suggestion, to investigate the potential impact of non-cancer samples on driving the observed differences, we carefully checked the PCAs for both DNA methylation and RNA-seq. We could not identify a clear separation of no-cancer COPD samples from the cancer COPD samples (or other cancer samples) in any examined PCs, indicating no cofounding effect of cancer background in the samples. We observed that one sample contributing to PC2 is a non-cancer sample, but this was a rather sample-specific effect, as the other two non-cancer samples clustered together with the other severe COPD samples with a cancer background. Notably, in our DNA methylation data, we do not observe typical features of cancer methylomes, like global loss of DNA methylation or aberrant methylation of CpG islands (e.g., in tumour suppressor genes) (see Fig 2A), further suggesting that we do not "pick up" confounding cancer signatures in our data.

Following the comments from both reviewers, to clarify that point, we added the information about cancer and non-cancer samples to the PCA figures for DNA methylation (new Fig. 2B) and RNA-seq (new Fig. 3A) data in the revised manuscript, as shown below

CHANGE IN THE MANUSCRIPT____:

COPD samples from donors with a cancer background clustered together with the COPD samples from lung resections, confirming that we detected COPD-relevant signatures (Fig. 2B).

Fig.2B* Principal component analysis (PCA) of methylation levels at CpG sites with > 4-fold coverage in all samples. COPD I and COPD II-IV samples are represented in light and dark green triangles, respectively, and no COPD samples as blue circles. COPD samples without a cancer background are displayed with a black contour. The percentage indicates the proportion of variance explained by each component. *

Unsupervised principal component analysis (PCA) on the top 500 variable genes revealed a clear influence of the COPD phenotype in separating no COPD and COPD II-IV samples, as previously observed with the DNA methylation analysis, irrespective of the cancer background of COPD samples (Fig.3A, Fig. S2B).

*Principal component analysis (PCA) of 500 most variable genes in RNA-seq analysis. PCA 1 and 2 are shown in Fig.3A, PCA 1 and 4 in Fig.S2B. COPD I and COPD II-IV samples are represented in light and dark green triangles, respectively, and no COPD samples as blue circles. COPD samples without a cancer background are displayed with a black contour. The percentage indicates the proportion of variance explained by each component. *

__Response: __

We thank the reviewer for suggestions on how to improve the discussion of our manuscript. We have now added a strength/limitation section to our discussion and included the points suggested by both reviewers.

CHANGE IN THE MANUSCRIPT____:

The strengths of our study include the use of purified human alveolar epithelial progenitor cells from a well-matched and carefully validated cohort of human samples, including mild and severe COPD patients, providing high relevance to human COPD. Importantly, we matched the smoking status and smoking history of all donors, which is key in epigenetic studies, as cigarette smoking profoundly impacts the DNA methylation landscape of tissues (96). With the first genome-wide high-resolution methylation profiles of isolated cells across COPD stages, we offer novel insights into the epigenetic regulation of gene expression in epithelial progenitor cells in COPD, expanding our understanding of how alterations in regulatory regions and specific genes could contribute to disease development. We identified IRF9 as a key IFN transcription factor regulated by DNA methylation. Notably, by targeting IRF9 through epigenetic modifications, we modulated the activity of the IFN pathway, which plays a crucial role in the immune response and lung tissue regeneration. Epigenetic editing techniques could offer a novel therapeutic strategy for COPD by downregulating IFN pathway activation and promoting the regeneration of epithelial progenitor cells in the lungs. Further preclinical and clinical studies are needed to validate the efficacy and safety of epigenetic editing approaches in COPD treatment (33)*. *

*However, we acknowledge several limitations to our study that warrant further investigation. First is the small sample size and replication difficulty due to the lack of available data, common challenges for studies working with sparse human material and hard-to-purify cell populations. The use of strict quality criteria in donor selection limited the available samples, especially for the ex-smoker control group, leading to an unequal distribution of COPD and control samples. Overall, this impacts the power of statistical analysis, especially in the WGBS analysis, where millions of regions genome-wide are tested. Nevertheless, the clear negative correlation of promoter methylation to the corresponding gene expression highlights the robustness of the DMR selection. Furthermore, we could experimentally validate interferon-associated DMRs using epigenetic editing, highlighting the power of integrated epigenetic profiling for the discovery of disease-relevant regulators. *

Overall, we detected a higher number of correlated DMR-DEG associations using our simple promoter-proximal linkage compared to the GeneHancer approach. Assigning enhancers to their target genes with high confidence is a complex and challenging task. Enhancers are often located far from the genes they regulate and can interact with their target genes through three-dimensional chromatin loops. Furthermore, enhancers can operate in a highly context-dependent manner, with the same enhancer regulating different genes depending on the cell type, developmental stage, or environmental signals. Determining which enhancer is active under specific conditions remains a hurdle in the field, especially since the AT2-specific chromatin profiles of enhancer marks are not yet available.

In addition, while WGBS provides unprecedented resolution and high coverage of the DNA methylation sites across the genome, it does not allow distinguishing 5-methylcytosine from 5-hydroxymethylcytosine. Therefore, we cannot exclude that some methylated sites we detected are 5-hydroxymethylated. However, as 5-hydroxymethylcytosine is present at very low levels in the lung tissue (97)*, its effect is likely marginal. *

Finally, despite careful removal of airways from distal lung tissue using a dissecting microscope, we cannot exclude the presence of some terminal/respiratory bronchiole cells in our FACS-isolated EpCAMpos/PDPNlow population. Recent scRNA-seq studies provided an unprecedented resolution and identified several epithelial subpopulations and transitional cells residing in the terminal/respiratory bronchioles and alveoli, including respiratory airway secretory cells (93), terminal airway-enriched secretory cells (28), terminal bronchiole-specific alveolar type-0 (AT0) (70), and emphysema-specific AT2 cells (74). These cells may contribute to alveolar repair in healthy and COPD lungs; however, with our bulk DNA methylation and RNA-seq study, we are unable to resolve all these subpopulations. Future development of single-cell methylation and non-reference-based algorithms for DNA methylation deconvolution will enable deeper epigenetic phenotyping of specific AT2 and bronchiolar cell subsets.

__References __ • Check references. For instance, there is no reference in the text to ref 43.

• Align format of references

__Response: __

We thank the reviewer for spotting this inconsistency. We have carefully checked and aligned the format of all references. The (old) reference 43 is now mentioned in the discussion part.

__Reviewer #1 (Significance (Required)): __

The strength of this study lies in its focus on the molecular mechanisms underlying the impaired regeneration of epithelial progenitor cells in COPD. The discovery of IRF9, which regulates IFN signaling and is prominently upregulated in COPD, together with the convincing validation of the epigenetic control of the IFN pathway by targeted DNA demethylation of the IRF9 gene, adds significant value to the COPD research field.

Main limitations of the study are the relatively small sample size of both COPD and non-COPD specimens and the claim that the sorted EpCAMpos/PDPNlow cells primarily consisted of AT2 cells.

__- Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. __

The nature and significance of the advance in epigenetic editing of IRF9 in COPD can be described as both conceptual and potentially clinical:

Conceptual Advance: The epigenetic editing of IRF9 enhances our understanding of the molecular mechanisms underlying COPD pathogenesis. By targeting IRF9 through epigenetic modifications, researchers were able to modulate the activity of the IFN pathway, which plays a crucial role in the immune response and lung tissue regeneration. This approach offers insights into the epigenetic regulation of gene expression in epithelial progenitor cells in COPD and expands our understanding of how alterations in specific gene methylation could contribute to disease progression.

Clinical Significance: The potential clinical significance of epigenetic editing of IRF9 lies in its implications for COPD therapy. If successful, epigenetic editing techniques could offer a novel therapeutic strategy for COPD by downregulating IFN pathway activation and promoting regeneration of epithelial progenitor cells in the lungs. Obviously, further preclinical and clinical studies are needed to validate the efficacy and safety of epigenetic editing approaches in COPD treatment.

__Response: __We thank the reviewer for recognising the importance of our study, its conceptual advance and potential clinical significance. We are pleased to see that the reviewer highlights the promise of epigenetic editing in both furthering our basic understanding of molecular mechanisms of chronic diseases and its future potential as a therapeutic strategy.

__- Place the work in the context of the existing literature (provide references, where appropriate). __ Few experimental papers have been published on epigenetic editing in lung diseases, with limited research available beyond the study referenced in citation 43. Song J, Cano-Rodriquez D, Winkle M, Gjaltema RA, Goubert D, Jurkowski TP, Heijink IH, Rots MG, Hylkema MN. Targeted epigenetic editing of SPDEF reduces mucus production in lung epithelial cells. Am J Physiol Lung Cell Mol Physiol. 2017 Mar 1;312(3):L334-L347. doi: 10.1152/ajplung.00059.2016. Epub 2016 Dec 23. PMID: 28011616.

Response:

We thank the reviewer for recognising the uniqueness and novelty of our study and the lack of research on the functional understanding of DNA methylation in the context of lung and lung diseases.

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

This study is of broad interest to researchers investigating the pathogenesis and treatment of COPD.

__- Define your field of expertise with a few keywords to help the authors contextualize your point of view. __

Expertise in: Lung pathology, Immunology, COPD, Epigenetics

- Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. Less expertise in: Epigenetic Editing

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

__Summary: __

This study aim to understand the molecular mechanisms underlying dysfunction in AT2 cells in COPD, by profiling bulk genome wide DNA methylation using Tagmentation-based whole-genome bisulfite sequencing (T-WGBS) and RNA sequencing in selectively sorted primary AT2 cells. The study stands out in it's sequencing breadth and use of an incredibly difficult cell population, and has the potential to add substantially to our mechanistic understanding of epigenetic contributions to COPD. A further highlight is the concluding aspect of the study where the authors undertook targeted modification of specific CpG methylation, provided direct, site-specific evidence for transcriptional regulation by CpG methylation.

Response:

We thank the reviewer for recognizing the conceptual and methodological advance of our study and for noting the value of our functional mechanistic approach.

__Major comments: __

The authors clearly show that there is DNA methylation alteration in AT2 cells from COPD individuals that links functional to gene expression at some level. However, I think the statement "to identify genome-wide changes associated with COPD development and progression..." and similar other references to disease development understanding is not accurate given the DNA methylation primary comparison is between control and moderate to severe COPD, with no temporal detail or evidence that they drive progression rather than are a result of COPD development. The paragraph starting on line 186 where this is a addressed to some extent is quite vague and doesn't really provide confidence that DNAm dysregulation occurs at an early stage in this context. This can be addressed by changing the focus/style of the text.

__Response: __

Thank you for raising this point. We agree with the reviewer that our cross-sectional study describes the association of methylation changes with either COPD I or more established disease (COPD II-IV) and that the observed changes may be either the driver or a result of COPD development. This has been clarified in the revised manuscript, and we removed the statements about disease initiation and progression. This is an important point; hence, we added an extra line to the discussion to make that clear.

__CHANGE IN THE MANUSCRIPT____: __

Therefore, we set out to profile DNA methylation of human AT2 cells at single CpG-resolution across COPD stages to identify epigenetic changes associated with disease and combine this with RNA-seq expression profiles.

To identify epigenetic changes associated with COPD, we collected lung tissue from patients with different stages of COPD,

....to identify methylation changes associated with mild disease, we included TWGBS data from AT2 isolated from COPD I patients (n=3) in the analysis.

Currently, we do not know whether the identified DNA methylation changes are the cause or the consequence of the disease process and not much is known about the correlation of DNA methylation with disease severity.

*However, our study is cross-sectional, our cohort included only 3 COPD I donors, and we did not have any follow-up data on the patients, so future large-scale profiling of mild disease (or even pre-COPD cohorts) in an extended patient cohort will be crucial for a better understanding of early disease and its progression trajectories. *

__Results comments and suggestions: __

For the integrated analysis, there is a focus on DMRs in promoters with very little analysis on other regions. The paragraph starting on line 317 describes some analysis on enhancers but is very brief, doesn't include information on how many/which DMRs were included, making it hard to interpret the impact of the 147 DMRs and 93 genes identified - is this nearly all DMRs and genes analysed or very few? A comparison to the promoter analysis would be of interest. Especially as the targeted region followed up with lovely functional assessment in the last sections is a gene body DMR, not a promoter DMR.

__Response: __

We thank the reviewer for pointing out the importance of changes in enhancers. We agree that extending the enhancer analysis is very interesting. However, assigning enhancers to their target genes with high confidence is a complex and challenging task. Enhancers are often located far from the gene they regulate, sometimes spanning hundreds of kilobases. They can interact with their target genes through three-dimensional chromatin loops, potentially bypassing nearby genes to activate more distant ones, making it difficult to confidently link specific enhancers to their target genes. Furthermore, enhancers can operate in a highly context-dependent manner. The same enhancer can regulate different genes depending on the cell type, developmental stage, or environmental signals. Another challenge is that enhancers often work in clusters or "enhancer landscapes," where multiple enhancers contribute to the regulation of a single gene. Disentangling the contribution of individual enhancers within such clusters and determining which enhancer is active under specific conditions remains an ongoing hurdle in the field, especially since the AT2-specific chromatin profiles of enhancer marks are not yet available.

One approach we tried to account for more distal regulatory regions was to assign DMRs to the nearest gene with a maximum distance of up to 100 kb using GREAT (Genomic Regions Enrichment of Annotations Tool) and simultaneously perform gene enrichment analysis of the associated genes. The old Figure S1C (now S1D) shows the top 10 enriched terms of either hyper- or hypomethylated DMRs, and Table 4 shows the full list of enriched terms. However, in this analysis, we did not integrate the results of the RNA-seq analysis. To demonstrate that we can correlate methylation with gene expression associations in this analysis, we then took a closer look at the WNT/b-catenin pathway, which contains 147 DMRs associated with 93 genes from the respective pathway (old Figure S3D, now S3G). Here, we showed that distal DMRs up to 100 kb away from the TSS show a high correlation with gene expression. We are including the two figures below for convenience:

*Left panels, functional annotation of genes located next to hypermethylated (top) and hypomethylated (bottom) DMRs using GREAT. Hits were sorted according to the binominal adjusted p-value and the top 10 hits are shown. The adjusted p-value is indicated by the color code and the number of DMR associated genes is indicated by the node size. Right panel, scatter plot showing distal DMR-DEG pairs associated with Wnt-signaling. Pairs were extracted from GREAT analysis (hypermethylated, DMR-DEG distance Following the reviewer's suggestion, we have now extended the enhancer analysis using the GeneHancer database, the most comprehensive, integrated resource of enhancer/promoter-gene associations. We used the GeneHancer version 5.14, which annotates 392,372 regulatory genomic elements (GeneHancer element) on the hg19 reference genome. Of the 25,028 DMRs, 18,289 DMRs (73% of all DMRs) coincided with at least one GeneHancer element, resulting in 19,661 DMR-GeneHancer associations. Next, we extracted the GeneHancer elements associated with protein-coding or long-non-coding RNAs genes, which left us with 2,144 DMR-GeneHancer associations. Next, we used only high-scoring gene GeneHancer associations ("Elite"), leaving 1,485 DMR-GeneHancer associations. Of those, we selected the GeneHancer elements, which are linked to genes differentially expressed in our RNA-seq analysis resulting in a final table of 376 DMR-GeneHancer associations (Table 9 DMR_DEG_GeneHancer, Tab 2). Similar to the promoter-proximal analysis, we analysed the correlation of expression and methylation changes of the DMR-GeneHancer associations, demonstrating a high number of negatively and positively correlated events (Fig.S3D). Finally, we performed the gene enrichment analysis for positively and negatively correlating genes. We detected significant GO term enrichments only for negatively correlating genes (Fig.S3E and Table 10_Enrichment_results, Tab2).

CHANGE IN THE MANUSCRIPT

To harness the full resolution of our whole-genome DNA methylation data, we extended the analysis beyond promoter-proximal regions and assessed how epigenetic changes in distal regulatory regions (enhancers) may relate to transcriptional differences in COPD. As the assignment of enhancer elements to the corresponding genes is challenging, we tried two different approaches. First, we used the GeneHancer database (72) to link DMRs to regulatory genomic elements (GeneHancer element). Of the 25,028 DMRs, 18,289 DMRs (73%) coincided with at least one GeneHancer element. Of those 2,144 DMR-GeneHancer associations were linked either to protein-coding or lncRNA genes. Next, we filtered for high-scoring gene GeneHancer associations ("Elite"), leaving 1,485 DMR-GeneHancer Elite associations. Of those, we selected the GeneHancer elements, which are linked to genes differentially expressed in our RNA-seq analysis, resulting in 376 DMR-GeneHancer associations (Table 9). Similar to the promoter-proximal analysis, we assessed the correlation of expression and methylation changes of the DMR-GeneHancer associations, demonstrating a high proportion of negatively and positively correlated events (Fig. S3E). Finally, we performed gene enrichment analysis for positively and negatively correlated genes. We detected significant GO term enrichments for negatively correlating genes only (Fig. S3F and Table 10), with the most pronounced term "regulation of tumor necrosis factor". In an alternative approach, we linked proximal and distal (within 100 kb from TSS) DMRs to the next gene using GREAT (57) (Fig S1C, Table 4) *and calculated Spearman correlation between DMRs and associated DEGs__. 147 DMRs were associated with high correlation rates with 93 genes from the WNT/β-catenin pathway (Fig. S3G)__, suggesting that DNA methylation may also drive the expression of genes of the WNT/β-catenin family. *

Figure S3E and F: E. Spearman correlation between gene expression and DMR methylation of DMRs assigned to gene regulatory elements using the GeneHancer database. F. GO-Term over-representation analysis of DEGs negatively correlated to DMRs in gene regulatory elements. The adjusted p-value is indicated by the color code and the percentage number of associated DEGs is indicated by the node size.

(Methods) For enhancer analysis, the GeneHancer database version 5.14, which annotates 392,372 regulatory genomic elements (GeneHancer element) on the hg19 reference genome, was used (72). Of the 25,028 DMRs 18,289 DMRs coincided with at least one GeneHancer element, resulting in 19,661 DMR-GeneHancer associations. Next, the GeneHancer elements were filtered for association with protein-coding or long-non-coding RNAs genes and high-scoring gene GeneHancer associations ("Elite"), leaving 1,485 DMR-GeneHancer associations. Of those, the GeneHancer elements were selected, which are linked to differentially expressed genes in COPD resulting in a final table of 376 DMR-GeneHancer associations. Similar to the promoter-proximal analysis, the Spearman correlation of expression and methylation changes of the DMR-GeneHancer associations was assessed. GO gene enrichment analysis for positively and negatively correlating genes was done using Metascape (111).

A comparison to the promoter analysis would be of interest.

Response:

We detected more highly correlated (|correlation coefficient| > 0.5) DMR-DEG associations using our simple promoter proximal linkage (n=643) in comparison with the GeneHancer approach comprising annotated enhancer elements (n=327/2,144). Gene enrichment results pointed to the interferon pathway, which we could confirm using epigenetic editing. This pathway was not present in the GeneHancer analysis, indicating that regulation of the IFN pathway may be controlled by proximal elements.

CHANGE IN THE MANUSCRIPT____:

Overall, we detected a higher number of correlated DMR-DEG associations using our simple promoter-proximal linkage compared to the GeneHancer approach. Assigning enhancers to their target genes with high confidence is a complex and challenging task. Enhancers are often located far from the genes they regulate and can interact with their target genes through three-dimensional chromatin loops. Furthermore, enhancers can operate in a highly context-dependent manner, with the same enhancer regulating different genes depending on the cell type, developmental stage, or environmental signals. Determining which enhancer is active under specific conditions remains a hurdle in the field, especially since the AT2-specific chromatin profiles of enhancer marks are not yet available.

Especially as the targeted region followed up with lovely functional assessment in the last sections is a gene body DMR, not a promoter DMR.

Response:

We thank the reviewer for bringing up that point. To clarify, we defined the promoter regions for the analysis as regions located {plus minus} 6 kb (upstream and downstream) from the transcriptional start site (TSS). Since the term "promoter" often refers to the region upstream of the transcriptional start site, its use may have been misleading. For clarity, we changed the text correspondingly to __promoter proximal methylation __and explained in the methods how the regions for analysis were defined.

__CHANGE IN THE MANUSCRIPT____: __

"DMR association per gene promoter" was changed to "Gene promoter proximal DMRs"

Fig. S3B: "DMR in promoter" was changed to "promoter proximal DMR(s)"

"by DNA methylation changes in promoters" was changed to "by DNA methylation changes in promoter proximity"

"regulated by promoter methylation" was changed to "regulated by promoter-proximal methylation"

"analysis of the promoter DMRs" was changed to "analysis of the promoter-proximal DMRs"

"between promoter methylation" was changed to "between promoter proximal methylation"

Cytoscape was used to analyse negatively or positively correlated DMR DEG pairs. ClueGO (v2.5.6) analysis was conducted using all DEG associated with a promoter proximal DMR (+/- 6 kb from TSS) and the Spearman correlation coefficient 0.5 (112).

• Lines 299-301 - I'm not sure the graph in Fig S3A support the conclusion that there was a preferential negative relationship between DNAm and gene expression. Looks like there are a substantial number of cases where a positive relationship is observed and this needs to be acknowledged.

Response:

In this part, we refer to Fig S3C. In the left panel, downregulated genes clearly show higher counts for the hypermethylated DMRs, whereas the hypomethylated DMRs are enriched at upregulated genes (right panel), indicating a preference for negative correlation: lower methylation, higher gene expression. If there were no preference, we would expect a 50:50 ratio of hypo- and hypermethylated DMRs, and we observed a 77:23 ratio. Nevertheless, we agree that there is a substantial number of cases (n=151) with a high positive correlation, which we now highlight in the text. For clarity, we also modified the figure legend to indicate that a stacked histogram is represented in the panel.

__CHANGE IN THE MANUSCRIPT____: __

L303: Interestingly, 23.5% of the identified DMR DEG pairs (n=151) showed a positive correlation between gene expression and DNA methylation.

*Figure legend in Fig. S3C was changed to: C Stacked histogram showing location of hyper- and hypomethylated DMRs relative to the TSS of DEGs in downregulated (left) and upregulated (right) genes. *

• Line 307 - what are the "analysed DEGs"? Are they the methylation associated genes?

Response:

Those are the DEGs we identified in RNA-seq analysis. To clarify, we changed the text to "identified DEGs".

__CHANGE IN THE MANUSCRIPT____: __

• "analysed DEGs" was changed to "identified DEGs"*

• Line 307-309 - "Among the analyzed DEGs, 76.5% (492) displayed a negative correlation (16.8% of the total DEGs), indicating a possible direct regulation by DNA methylation, while 23.5% (151) showed a positive correlation between gene expression and DNA methylation" - are the authors suggesting the positive correlation doesn't indicate direct regulation?

__Response: __

Thank you for highlighting this point. We did not intend to suggest that negative correlation indicates direct regulation, while positive correlation suggests a lack thereof. To clarify that point, we have reformulated this sentence.

__CHANGE IN THE MANUSCRIPT____: __

Among the identified DEGs, 76.5% (n=492) displayed a negative correlation (16.8% of the total DEGs), consistent with a repressive role of promoter DNA methylation. Interestingly, 23.5% of the identified DEG (n=151) showed a positive correlation between gene expression and DNA methylation.

• Line 313 - why did the authors focus on only negatively correlated genes to identify their top dysregulated pathway of IFN signalling? Why not do pathway analysis on the DNAm associated genes separately to identify DNAm associated pathways?

Response:

We have also performed a pathway enrichment analysis using the positively correlated genes but did not identify any significantly enriched pathways/process/terms. When we examined the top hit of the gene set enrichment analysis, the interferon signaling pathway, we observed only negatively correlated DMR gene associations (Fig. 5B). Therefore, we decided to use only the negatively correlated DMRs, as using all correlated genes would give a higher background and dilute our results.

CHANGE IN THE MANUSCRIPT____:

Cytoscape was used to analyse negatively or positively correlated DMR DEG pairs. ClueGO (v2.5.6) analysis was conducted using all DEG associated with a promoter proximal DMR (+/- 6 kb from TSS) and the Spearman correlation coefficient 0.5 (113).

• A comparison of the gene expression data with previous data in AT2 cell/single cell data would strengthen the gene expression section.

__Response: __

We compared our gene expression signatures with the study of Fujino et al., who profiled sorted AT2 cells (EpCAMhighPDPNlow) from COPD/controls using expression arrays (PMID: 23117565). Consistent with our study, the authors also observed the upregulation of interferon signalling (among other pathways) in COPD AT2s. However, no raw data was available in the published manuscript for a more in-depth analysis.

Several recent scRNA-seq studies identified transcriptional signatures of COPD and control cells (e.g., PMIDs: 36108172, 35078977, 36796082, 39147413__). However, most studies did not match the smoking status of the control and COPD donors and looked at the whole lung tissue, with limited power to detect gene expression changes in distal alveolar cells. It is difficult to directly compare our data to the gene expression data from non-smokers vs COPD patients, as cigarette smoking profoundly remodels the epigenome and transcriptional signatures of cells. In addition, differences in technologies and depth of sequencing make such comparisons challenging. However, one study (PMID: 36108172) performed scRNA-seq analysis on 3 non-smokers, 4 ex-smokers and 7 COPD ex-smoker lungs. Despite relatively limited coverage of epithelial cells in the dataset (We also compared the main AT2 IFN signature identified in the integration of our DNA methylation in promoter-proximal regions and RNA-seq with a recent study (published after the submission of our manuscript, PMID: 39147413) that profiled EpCAMpos cells from COPD and control lungs (non-smokers) using scRNA-seq. We observed an upregulation of our IFN signature genes in AT2 in COPD (specifically in AT2-c and rbAT2 subsets), suggesting that similar signatures were observed in this dataset as well. However, ex-smokers were not included in this study, making direct comparisons difficult. We have now included the panels shown below as __Figure S4E and S4F:

Figure S4E and F: Expression values for the indicated genes of the IFN pathway from an external scRNA-seq dataset of AT2 cells from COPD patients and healthy controls (74). Y-axis shows log-normalized gene expression levels. F. Combined gene set score of the genes shown in (E) in different subsets of AT2 cells from (74)*. The IFN signature genes were identified in our integrative analysis of TWGBS and RNA-seq in sorted AT2 cells. *

CHANGES IN THE MANUSCRIPT:

However, 5-AZA is a global demethylating agent, and the observed effects may not be direct. To validate the epigenetic regulation of central AT2 pathways further, we took advantage of locus-specific epigenetic editing technology (73). We focused on the IFN pathway because it was the most significantly enriched Gene Ontology (GO) term in our integrative analysis of TWGBS and RNA-seq data. Several IFN pathway members had associated hypomethylated DMRs within promoter-proximal regions and concomitant increased gene expression (Fig. 4C and Fig.S2C). Additionally, we confirmed the elevated expression of IFN-related genes with associated DMRs identified in our study in AT2 cells and AT2 cell subclusters from a recently published scRNA-seq cohort (74)* (Fig. S4E-F). *

(Methods) Validation of IFN gene upregulation in a published scRNA-seq dataset

scRNA-seq data from (74), generously provided by M. Köningshoff, were processed using the default Seurat workflow (117). Expression of IFN-related genes was extracted and plotted as log-normalised gene expression levels in AT2 cells from control and COPD donors. Seurat's AddModuleScore() function was used to compute a gene set score for a custom IFN program using the genes listed in __Fig. S4E __and to analyse the IFN gene set scores in AT2 cell subclusters identified in (74). Briefly, average gene expression scores were computed for the gene set of interest, and the expression of control features (randomly selected) was subtracted as described in (118).

Fig. S4 E and F. E. Expression values for the indicated genes of the IFN pathway from an external scRNA-seq dataset of AT2 cells from COPD patients and healthy controls (74). Y-axis shows log-normalized gene expression levels. F. Combined gene set score of the genes shown in (E) in different subsets of AT2 cells from (74). The IFN signature genes were identified in our integrative analysis of TWGBS and RNA-seq in sorted AT2 cells. __ __

• The paragraph starting on line 173 feels a little redundant when we know there is RNA available to test if the differential DNAm links to altered gene expression - this selected of example regions/genes would be better placed after the gene expression has been reported, at which point you could say whether the linked genes displayed altered transcription.

Response:

The current structure (with DNA methylation, followed by RNA-seq and integration) is intentional and serves several important purposes. As this is the first genome-wide high-resolution COPD DNA methylation study of AT2, we aimed to describe the methylation landscape independently of gene expression (noting the limitation of current understanding of how DNA methylation regulates expression). This early focus on DMRs lays clear groundwork by highlighting potential regulatory elements and pathways that could be disrupted, independent of or even before corroborative transcriptional data. Additionally, positioning these examples early in the narrative helps to frame subsequent gene expression analyses. Once RNA data are introduced later, the reader can directly compare the methylation patterns with transcriptional outcomes, thereby enhancing the overall story. In other words, by first showcasing disease-relevant methylation changes, we underscore a hypothesis that these epigenetic modifications are functionally meaningful. The later integration of gene expression data then serves as a confirmatory or complementary layer, rather than the sole basis for inferring biological significance. This is important as we still do not fully understand the function of DNA methylation outside promoters, and its role is also important for splicing, 3D genome organisation, non-coding RNA regulation, enhancer regulation, etc.

• Similarly, the TF enrichment analysis is great but maybe would have added value to be done on DNA regions later shown to be linked to differential expression - was there different enrichment at DNA regions that are vs are not associated with altered expression? And could you test in vitro whether changing methylation of DNA (maybe a blunt too like 5-aza would be ok) alters TF binding (cut+run/ChIP?). Furthermore, it would be interesting to understand the TF sensitivity analysis within the context of positive versus negative DNA methylation:gene expression correlations.

Response:

As suggested by the reviewer, we now performed the TF enrichment analysis using the DMRs with a high correlation (|correlation coefficient|>0.5) between methylation and expression (Figure S3D) and expanded the method section to include TF analysis. We observed ETS domain motifs enriched at hypomethylated regions. They prefer unmethylated DNA (MethylMinus) and are therefore expected to bind with higher affinity to the respective DMRs in COPD. We agree with the reviewer that further verifying altered TF binding using cut&run or ChIP assays would be very interesting, but it is out of the scope of this manuscript. Such analysis is technically very challenging to perform with low numbers of primary AT2 cells and will be the focus of our follow-up mechanistic studies.

CHANGE IN THE MANUSCRIPT____:

Additionally, motif analysis of DMRs that were highly correlated (|Spearman correlation coefficient| > 0.5) with DEGs revealed a prominent enrichment of the cognate motif for ETS family transcription factors, such as ELF5, SPIB, ELF1 and ELF2 at hypomethylated DMRs (Fig. S3D). Interestingly, SPIB was shown to facilitate the recruitment of IRF7, activating interferon signaling (71)*, and our WGBS data uncovers SPIB motifs at hypomethylated DMRs, which aligns with its binding preferences at unmethylated DNA (methyl minus, Fig. S3D). *

Figure S3D: Enrichment of methylation-sensitive binding motifs at hypo- (right) and hypermethylated (left) DMRs, using DMRs with a high correlation (|Spearman correlation coefficient| > 0.5) between methylation and gene expression. Methylation-sensitive motifs were derived from Yin et al (64). Transcription factors, whose binding affinity is impaired upon methylation of their DNA binding motif, are shown in red (Methyl Minus), and transcription factors, whose binding affinity upon CpG methylation is increased, are shown in blue (Methyl Plus).

(Methods) To obtain information about methylation-dependent binding for transcription factor motifs which are enriched at DMRs, the results of a recent SELEX study (64)* were integrated into the analysis. They categorised transcription factors based on the binding affinity of their corresponding DNA motif to methylated or unmethylated motifs. Those whose affinity was impaired by methylation were categorised as MethylMinus, while those whose affinity increased were categorised as MethylPlus. A motif database of 1,787 binding motifs with associated methylation dependency was constructed. The log odds detection threshold was calculated for the HOMER motif search as follows. Bases with a probability > 0.7 got a score of log(base probability/0.25); otherwise, the score was set to 0. The final threshold was calculated as the sum of the scores of all bases in the motif. Motif enrichment analysis was carried out against a sampled background of 50,000 random regions with matching GC content using the findMotifsGenome.pl script of the HOMER software suite, omitting CG correction and setting the generated SELEX motifs as the motif database. *

__Methods: __ • The authors should include more detail of the TWGBS rather than directing the reader to a previous publication. Also DNA concentration post bisulfite conversion would be a useful metric to provide.

__Response: __

Following the suggestion, we have now expanded the details of TWGBS in the methods part of the manuscript. Due to limited space, we did not include a detailed protocol but instead referred to a published step-by-step protocol (55). Of note, we do not measure DNA concentration post-bisulfite conversion but consistently use the starting input of 30 ng of genomic DNA across all samples.

__CHANGE IN THE MANUSCRIPT____: __

(Methods): 15 pg of unmethylated DNA phage lambda was spiked in as a control for bisulfite conversion. Tagmentation was performed in TAPS buffer using an in-house purified Tn5 assembled with load adapter oligos (55) at 55 {degree sign}C for 8 min. Tagmentation was followed by purification using AMPure beads, oligo replacement and gap repair as described (55). Bisulfite treatment was performed using EZ DNA Methylation kit (Zymo) following the manufacturer's protocol.

*The T-WGBS library preparations were performed for all donors in parallel and sequenced in a single batch to minimize batch effects and technical variability. *

• Differential DNA methylation analysis: It is stated that DNA regions had to contain 3 CpG sites but was this within a defined DNA size range?

Response:

The maximum distance between individual CpGs within DMR was set to 300 bp. To clarify, we added that information to the methods part.

__CHANGE IN THE MANUSCRIPT____: __

*"regions with at least 10% methylation difference and containing at least 3 CpGs with a maximum distance of 300 bp between them. *

• Refence genome only provided for RNAseq not TWGBS?

__Response: __We used hg19 as the reference genome. The information on the reference genome for DNA methylation analysis was provided in the methods L574 (original manuscript_: "The reads were aligned to the transformed strands of the hg19 reference genome using BWA MEM")

• The tables do not appear in the PDF and I struggled to tally to the "Dataset" files provided if that is what they were referring to?

Response:

Full tables (uploaded as Datasets in the manuscript central due to their size) were uploaded together with the manuscript files. They are quite large and will not convert to pdf, so they may not have been included in the merged pdf file. We assume that they should be available to the reviewers with the other files and will clarify that with the editorial staff in the resubmission cover letter.

• For the gene expression analysis, can it be made clearer that a full analysis was done on COPD I samples. It is a little confusing to the reader as this was not done for DNAm so might be assumed the same targeted analysis on only genes found to be differentially expressed between control and COPD II-IV, but that cannot be the case as an overlap of COPD1 vs COPD II-IV genes if provided. For this overlap, do genes show the same effect direction?

__Response: __

To clarify, for the RNA-seq analysis, we performed DEG analysis for no-COPD versus COPD II-IV, as well as no-COPD versus COPD I. We then took all differentially expressed genes (presented in the Venn diagram) and plotted them for all samples as a heatmap. To split the genes into groups displaying similar effect directions, we applied a clustering approach and identified 3 main signatures. Cluster 3 primarily comprises genes unique to COPD I samples, which are associated with the adaptive immune system and hemostasis (Fig. 4E). In the other two clusters, we mainly observe a transitioning pattern from control to severe COPD samples, correlating with the FEV1 values of the patients. This has now been clarified in the manuscript.

• Replication is difficult on these studies as the samples are so difficult to come by. Also limited by sample size for the same reason. It doesn't mean the study is not worth doing and the data are still valuable. However, it may be pertinent to include technical validation of a few regions of interest, acknowledge the limitation (along side strengths) in the discussion, and perhaps provide actual p value rather than blanket Response:

We thank the reviewer for acknowledging the replication challenges for studies working with sparse human material and hard-to-purify cell populations. Following the reviewer's suggestion, we have now included a strengths and limitations section in the discussion where we summarised the points highlighted by both reviewers.

Regarding technical validation, we would like to note that the whole genome bisulfite sequencing (WGBS) technology, as well as the tagmentation-based WGBS (T-WGBS), have been validated in the past few years in several publications (e.g., PMID: 24071908) and shown to yield reliable DNA methylation quantification in comparison to other technologies (PMID: 27347756). For us, technical validation using alternative methods (e.g. bisulfite sequencing or pyrosequencing) is difficult as it requires significantly more input DNA than the low-input T-WGBS we have performed and obtaining sufficient amounts of material from primary human AT2 cells (especially from severe COPD) is not possible with the size of tissue we can access. However, while establishing the T-WGBS for this project, we initially validated our approach using Mass Array, a sequencing-independent method. For this, we performed T-WGBS on the commercially available smoker and COPD lung fibroblasts and selected 9 regions with different methylation levels for validation using a Mass Array. We obtained an excellent correlation between both methods, providing technical validation of T-WGBS and our analysis workflow. This validation was published in our earlier manuscript (PMID: 37143403), but we provided the data below for convenience.

Scatter plots showing correlation of average methylation obtained with T-WGBS and Mass Array from COPD and smoker fibroblasts. Each dot represents one region with varying methylation levels. The blue diagonal represents the linear regression. Shaded areas are confidence intervals of the correlation coefficient at 95%. Correlation coefficients and P values were calculated by the Pearson correlation method.

To enable further validation and follow-up by the community, we included the full list of DMRs, associated p-values and additional information for DNA methylation analysis (DMR width, n.CpGs, MethylDiff, etc) in Table 3 (Table_3_wgbs_dmr_info.xlsx) and the information about DEGs from RNA-seq in Table 6 (Table_6_RNAseq_DEG_info.xlsx).

• It isn't clear to me if DNA and RNA are from the same cells? The results say "cells matching those used for T-WGBS" but the methods suggest separate extractions so not the same cells? If they are not the same cells a comment on the implications of this should be included in the discussion for example, potentially some differences in cell type composition, storage time etc.

Response:

Lung tissue samples were freshly cryopreserved, and H&E slides derived from exemplary pieces of the tissue analyzed. Once we had a group of at least 3 samples comprising one non-COPD and 2 COPD samples, we processed them in parallel to limit sorting variation between control and disease samples. The sorted cells were counted, aliquoted and pelleted at 4{degree sign}C before flash freezing and storing at -80{degree sign}C. The storage time of the cell pellets varied between the donors. RNA and DNA were isolated from cell pellets collected from the same FACS sorting experiment; therefore, we do not expect differences in cell type composition. In addition, RNA and DNA isolation were performed for all sorted pellets in parallel. All library preparations for TWGBS and RNA-seq were performed for all donors in parallel and sequenced in a single batch to minimise batch effects and technical variability. This has now been clarified in the methods part of the manuscript.

__CHANGE IN THE MANUSCRIPT____: __

To minimize potential technical bias, samples from no COPD and COPD donors were processed in parallel in groups of 3 (one no COPD and 2 COPD samples).

RNA and genomic DNA for RNA-seq and TWGBS were isolated from identical aliquots of sorted cell pellets.

Genomic DNA was extracted from 1-2x104 sorted alveolar epithelial cells isolated from cryopreserved lung parenchyma from 11 different donors in parallel using QIAamp Micro Kit

The TWGBS library preparations were performed for all donors in parallel and sequenced in a single batch to minimize batch effects and technical variability.* *

RNA was isolated from flash-frozen pellets of 2x104 sorted AT2 cells from 11 different donors in parallel.

The RNA-seq library preparation for all donors was performed in parallel and all samples were sequenced in a single batch to minimize batch effects and technical variability.

• Line 193 the authors say "Since DMRs were overrepresented at cis-regulatory sites...." - "cis" needs to be defined. If you link DNAm regions to gene via "closest gene" does this not automatically mean you're outputs will be cis? Just needs better definition/explanation.

Response:

The term "cis‐regulatory sites" in our manuscript is intended to denote regulatory elements-such as enhancers, promoters, and other nearby control regions-that reside on the same chromosome and close to the genes they regulate. While it's true that linking a DMR to its closest gene captures a cis association, our phrasing emphasises that the DMRs are enriched specifically at these functional regulatory elements (Fig. 2E) rather than being randomly distributed. This usage aligns with established conventions in the field. To avoid any misunderstandings, we have now changed the term to gene regulatory sites.

__CHANGE IN THE MANUSCRIPT____: __

*We changed the "cis-regulatory sites" to "gene regulatory sites" *

__Minor comments: __

Line 157: "we identified site-specific differences....". Change to region specific?

Response:

This has now been corrected as suggested.

Line 102-103: needs a reference for the statement "Alterations in DNA methylation patterns have been implicated......"

Response:

Following the reviewer's suggestion, we added the relevant references (34-36) to this statement.

Line 266 - what does "strong dysregulation" mean? Large fold change, very significant?

Response:

We removed the word "strong" from this sentence.

Lines 423-425 - statement needs a reference

Response:

Following the reviewer's suggestion, we added the relevant reference to this statement.

Line 428 - word missing between "epigenetic , we"?

Response:

This has now been corrected. The text reads: "Through treatment with a demethylating drug and targeted epigenetic editing, we demonstrated the ability to modulate..."

Prior studies are well references, text and figures are clear and accurate.

__Reviewer #2 (Significance (Required)): __

This study has several strengths:

1) Sample collection and characterisation. AT2 cells are incredibly hard to come by and the authors should be commended to generating the samples. However, proximity to cancer is always a potential issue, especially in epigenetic studies. Is it feasible to include any analysis to show the samples derived from those with cancer don't drive the changes observed? Even a high level PCA or an edit of fig 2A with non-cancer in a different colour in supplemental - looks like there is one outlier, is that a non-cancer? Or a correlation of change in beta between control and cancer/COPD and control and non-cancer:COPD (for want a better phrase!). just an indicator that the non-cancer COPD samples are not driving differences.

Response:

We thank the reviewer for highlighting the value of generating data from hard-to-work-with AT2 populations and bringing up the important point of cancer proximity, which we considered very carefully when designing our study. To match our samples across the cohort, all the no-COPD, COPD I, and two of the COPD II-IV distal lung samples were obtained from cancer resections. In addition to other characteristics, like age, BMI and smoking status, we also matched the donors by cancer type (all profiled donors had squamous cell carcinoma). We collected lung tissue as far away from the carcinoma as possible and sent representative pieces for histological analysis by an experienced lung pathologist to confirm the absence of visible tumours. In addition, to ensure that our data represents COPD-relevant signatures, we intentionally included samples from three COPD donors undergoing lung resections (without a cancer background) in the profiling.

Following the reviewer's suggestion, to investigate the potential impact of non-cancer samples on driving the observed differences, we carefully checked the PCAs for both DNA methylation and RNA-seq. We could not identify a clear separation of no-cancer COPD samples from the cancer COPD samples (or other cancer samples) in any examined PCs, indicating no cofounding effect of cancer samples. We observed that one sample contributing to PC2 is a non-cancer sample, but this was a rather sample-specific effect, as the other two non-cancer samples clustered together with the other severe COPD samples with a cancer background. Notably, in our DNA methylation data, we do not observe typical features of cancer methylomes, like global loss of DNA methylation or aberrant methylation of CpG islands (e.g., in tumour suppressor genes) (see Fig. 2A), further suggesting that we do not "pick up" confounding cancer signatures in our data.

Following the comments from both reviewers, to clarify that point, we added the information about cancer and non-cancer samples to the PCA figures for DNA methylation (new Fig. 2B) and RNA-seq (new Fig. 3A) data in the revised manuscript, as shown below

CHANGE IN THE MANUSCRIPT____:

COPD samples from donors with a cancer background clustered together with the COPD samples from lung resections, confirming that we detected COPD-relevant signatures (Fig. 2B).

Fig. 2B.* Principal component analysis (PCA) of methylation levels at CpG sites with > 4-fold coverage in all samples. COPD I and COPD II-IV samples are represented in light and dark green triangles, respectively, and no COPD samples as blue circles. COPD samples without a cancer background are displayed with a black contour. The percentage indicates the proportion of variance explained by each component. *

Unsupervised principal component analysis (PCA) on the top 500 variable genes revealed a clear influence of the COPD phenotype in separating no COPD and COPD II-IV samples, as previously observed with the DNA methylation analysis, irrespective of the cancer background of COPD samples (Fig.3A, Fig. S2B).

*Principal component analysis (PCA) of 500 most variable genes in RNA-seq analysis. PCA 1 and 2 are shown in Fig.3A, PCA 1 and 4 in Fig.S2B. COPD I and COPD II-IV samples are represented in light and dark green triangles, respectively, and no COPD samples as blue circles. COPD samples without a cancer background are displayed with a black contour. The percentage indicates the proportion of variance explained by each component. *

2) This is the first time DNAm has been profiled in AT2 cells. It is incredibly difficult, valuable and novel data that will increase the fields capability technically, their understanding of functional mechanisms and potential translation considerably. It's audience will be primarily translational respiratory however the fundamental science aspect of gene expression regulation by DNA methylation with have wider reach across developmental and disease science.

Response:

We thank the reviewer for recognising the uniqueness and novelty of our study and highlighting the value and potential impact of our datasets for the lung field.

3) the functional analysis using targeted CRISPR-Cas9 is very well done and adds impact.

Response:

We thank the reviewer for recognising the strengths and added value of the functional analysis using epigenetic editing.

__Potential weaknesses/areas for development __

I feel the main weakness is the in the section integrating DNA methylation and gene expression. The rationale for a focus on various aspects, for example inversely related DNAm/gene expression pairs, the IFN pathway and IRF9, are not clear. Also further understanding of the differences between DNAm associated genes and non-DNAm associated genes could be expanded, at the pathway level, TF regulation level, effect size level (are DNAm associated changes to gene expression larger, enriched for earlier differential expression)

Response:

Our rationale for focusing on the inversely related DNAm/gene expression pairs in promoter proximal is purely data-driven, as they represent the biggest group in our data (Fig. 4A-B). Among those negatively correlated genes, we observed the strongest enrichment for the IFN pathway (Fig. C), making it an obvious, data-driven target for further studies. The negative correlation of expression and methylation for IFN pathway genes could be validated in 5-AZA assays in A549 cells (Fig. 5A). Next, we made an interaction network analysis showing IRF9 and STAT2 as master regulators (Fig. 5B) of the negatively correlated IFN genes. As IRF9 itself displayed a negative correlation between DNA methylation and expression (Fig. 5C), we used the associated DMR for further epigenetic editing (Fig. 5D-E). We performed the additional requested analyses of the enhancer-associated changes and genes, as described above. We fully agree with the reviewer that our data sets are a great resource and can be further used to elaborate on other relationships of DNA methylation and RNA expression or other pathways, but this is out of the scope of this study. To enable further studies by the research community, we provide all necessary information about DMRs and DEGs in the associated supplementary tables and the raw data through the EGA, as well as the CRISPRa editing assay.

The authors could comment on potential masking of differences between 5hmC and mC and the implications it may have

Response:

We thank the reviewer for bringing up this important point. Indeed, bisulfite sequencing cannot differentiate between methylated and hydroxymethylated cytosines; hence, some of the methylated sites may be hydroxymethylated. However, the overall levels of hydromethylation in differentiated adult tissues are very low (except for the brain), orders of magnitude lower compared to DNA methylation. Following the reviewer's suggestion, we have added a sentence in the limitation section of the discussion to clarify that point.

__CHANGE IN THE MANUSCRIPT: __

In addition, while WGBS provides unprecedented resolution and high coverage of the DNA methylation sites across the genome, it does not allow distinguishing 5-methylcytosine from 5-hydroxymethylcytosine. Therefore, we cannot exclude that some methylated sites we detected are 5-hydroxymethylated. However, the 5-hydroxymethylcytosine is present at very low levels in the lung tissue (97)*. ** *

Furthermore, while the rationale for looking at DMRs is clear, especially given the sample number, I am interested to understand what proportion of the assayed CpGs "fit" within the cut off stipulations of the DMR analysis - that is, is their potentially COPD effects at sparse CpG regions/individual CpG sites that are not being identified. A comment on this would be useful and seems the strength of profiling genome wide. I'm happy genome wide is beneficial it just feels a little circular that the authors have chosen whole genome to avoid the bias of the Illumina array and a focus on promotors, but have primarily reported promoter DNAm. This caught my attention again in the discussion where the authors state that cis-regulatory regions were also identified in their fibroblast data .....is this finding a factor of the analysis performed? (also a comparison of regions Identified in AT2 cells versus fibroblasts would be really interesting for a future paper)

Response:

We decided to focus our analysis on regions rather than individual CpG sites when looking at differential methylation, as DNA methylation is spatially correlated, and methylation changes in larger regions are more likely to have a biological function. Extending the analysis to single CpG sites would require a higher number of samples for a reliable analysis compared to the DMR analysis (as mentioned by the reviewer).

Of note, we addressed the platform comparison between Illumina array technology and WGBS in our previous fibroblast study (PMID: 37143403), where we compared our WGBS data with the published 450k array data of COPD parenchymal fibroblasts (Clifford et al., 2018). We observed only a marginal overlap between the CpGs from our DMRs and the CpGs probes available on the array (which was due to the differences in technologies used and the limited coverage of the 450K array in comparison to our genome-wide approach, in which we covered 18 million CpGs). Out of the 6279 DMRs identified in our fibroblast study, only 1509 DMRs overlapped with at least one CpG probe on the 450K array, and after removing low-quality CpGs from the array data, only 1419 DMRs were left. This comparison highlighted the increased resolution of the WGBS compared to Illumina arrays.

The reason why we focused on promoter proximal DMRs are the following: 1) the assignment of the enhancer elements in AT2 to the corresponding gene is still too inaccurate in the absence of AT2 specific enhancer chromatin maps 2) regulation at enhancers by DNA methylation might be more complex and might change (increase or attenuate) binding affinities of certain transcription factors (Fig.2H), which might lead to gene expression changes or 3) methylation changes might be an indirect effect of differential TF binding PMID: 22170606). However, we agree with the reviewer that despite these limitations, expanding the analysis beyond promoters adds value to the manuscript; hence, as described above, we expanded the analysis of non-promoter regions, including enhancers, in the revised manuscript.

We thank the reviewer for the suggestion to compare the regions identified in AT2 cells and fibroblasts in a future paper.

My expertise:Respiratory, cell biology, epigenetics.

Author response:

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

Reviewer #1 (Public Review): 

In this manuscript, Gruber et al perform serial EM sections of the antennal lobe and reconstruct the neurites innervating two types of glomeruli one that is narrowly tuned to geosmin and one that is broadly tuned to other odours. They quantify and describe various aspects of the innervations of olfactory sensory neurons (OSNs), uniglomerlular projection neurons (uPNs), and the multiglomerular Local interneurons (LNs) and PNs (mPNs). They find that narrowly tuned glomeruli had stronger connectivity from OSNs to PNs and LNs, and considerably more connections between sister OSNs and sister PNs than the broadly tuned glomeruli. They also had less connectivity with the contralateral glomeruli. These observations are suggestive of strong feed-forward information flow with minimal presynaptic inhibition in narrowly tuned glomeruli, which might be ecologically relevant, for example, while making quick decisions such as avoiding a geosmin-laden landing site. In contrast, information flow in more broadly tuned glomeruli show much more lateralisation of connectivity to the contralateral glomerulus, as well as to other ipsilateral glomeruli. 

The data are well presented, the manuscript clearly written, and the results will be useful to the olfaction community. I wonder, given the hemibrain and FAFB datasets exist, whether the authors have considered verifying whether the trends they observe in connectivity hold across three brains? Is it stereotypic? 

We appreciate the reviewer’s positive view of our study and their thoughtful and relevant comment on the issue of individual variation. We agree in that this is a very important question and notice that it was also asked for by the second Reviewer. It reflects both our limited understanding of the range of individual variation in synaptic connectivity—whether in flies, humans, or other species—and the challenge of determining which of the differences observed in our study are stereotypical features of each glomerulus type. Undoubtedly this criticism addresses a crucial problem of practically all connectome studies so far and for which there is no immediate solution. This type of studies requires so much time, efforts and money that increasing the number of samples is seldom feasible. The Reviewer wonders if we could compare our data with that made available by two of the largest connectome studies of Drosophila. This appeared to us to be a very good idea and we have tried to follow the advice but, unfortunately, it was impracticable because of the reasons we explain below. The hemibrain data cannot be used for this purpose because it does not contain the full glomerulus DA2 (Schlegel et al., 2021). A different problem hindered us from using the FAFB dataset, the other dataset mentioned by the Reviewer. In this case the three glomeruli were sectioned and reconstructed but the dataset lacks an annotated list of all synaptic connections corresponding to each glomerulus. Such annotation (a compendium of all synaptic connections inside each glomerulus informing for each connection which type of neuron provides the presynaptic site and which the postsynaptic site) is essential for direct comparison with our data. It is important to keep in mind that the current analytical tools available for the use of these datasets (e.g., NeuPrint, FlyWire and CATMAID) do not offer the ability to extract data on synapses exclusively from the glomerular volume of DA2 or DL5. In this case, it certainly is theoretically possible to obtain the data by doing ourselves the annotation. However, such a study will demand so much time, efforts and financial resources, which we believe would not be justified solely to increase the number of individuals from one to two. Instead, our manuscript includes a comparison of the OSN connectivity in VA1v and DL5 using the hemibrain dataset published by Schlegel et al. (2021) (see revised manuscript: lines 311–315; 431–434; 558–562; 602–606).

Beyond the opinion, that we share in full with the Reviewer, that a comparison including three flies will be better than a comparison made with one glomerulus of each type we are still challenged by the question of which -if any- of the differences are stereotypic. The clarification of what are stereotypical differences between particular glomeruli in features as those discussed in our study and what is simply differences within the normal range of individual variation is basically a statistical problem. A first attempt at a comprehensive comparison focusing on intra- and inter-individual variability was recently made by comparing two connectome datasets from two different Drosophila individuals (Dorkenwald et al., 2024; Schlegel et al., 2024). At present, it is still unclear how many samples are needed to make a statistically robust comparison of olfactory synaptic circuits in adult flies—perhaps 3, 6, or even 18 individuals?  

Reviewer #2 (Public Review):

The chemoreceptor proteins expressed by olfactory sensory neurons differ in their selectivity such that glomeruli vary in the breadth of volatile chemicals to which they respond. Prior work assessing the relationship between tuning breadth and the demographics of principal neuron types that innervate a glomerulus demonstrated that narrowly tuned glomeruli are innervated more projection neurons (output neurons) and fewer local interneurons relative to more broadly tuned glomeruli. The present study used high-resolution electron microscopy to determine which synaptic relationships between principal cell types also vary with glomerulus tuning breadth using a narrowly tuned glomerulus (DA2) and a broadly tuned glomerulus (DL5). The strength of this study lies in the comprehensive, synapse-level resolution of the approach. Furthermore, the authors implement a very elegant approach of using a 2-photon microscope to score the upper and lower bounds of each glomerulus, thus defining the bounds of their restricted regions of interest. There were several interesting differences including greater axo-axonic afferent synapses and dendrodentric output neuron synapses in the narrowly tuned glomerulus, and greater synapses upon sensory afferents from multiglomerular neurons and output neuron autapses in the broadly tuned glomerulus.     The study is limited by a few factors. There was a technical need to group all local interneurons, centrifugal neurons, and multiglomerular projection neurons into one category ("multiglomerular neurons") which complicates any interpretations as even multiglomerular projection neurons are very diverse. Additionally, there were as many differences between the two narrowly tuned glomeruli as there were comparing the narrowly and broadly tuned glomeruli. Architecture differences may therefore not reflect differences in tuning breadth, but rather the ecological significance of the odors detected by cognate sensory afferents. Finally, some synaptic relationships are described as differing and others as being the same between glomeruli, but with only one sample from each glomerulus, it is difficult to determine when measures differ when there is no measure of inter-animal variability. If these caveats are kept in mind, this work reveals some very interesting potential differences in circuit architecture associated with glomerular tuning breadth.

This work establishes specific hypotheses about network function within the olfactory system that can be pursued using targeted physiological approaches. It also identifies key traits that can be explored using other high-resolution EM datasets and other glomeruli that vary in their tuning selectivity. Finally, the laser "branding" technique used in this study establishes a reduced-cost procedure for obtaining smaller EM datasets from targeted volumes of interest by leveraging the ability to transgenically label brain regions in Drosophila.

CLASSIFICATION OF NEURONAL TYPES

We agree that grouping diverse types of interneurons into a single category (referred to as MGNs) limits the ability to make interpretations about synaptic similarities and differences between specific neuronal types. This was, however, an unavoidable compromise resulting from our decision to generate a comprehensive, synapse-level reconstruction of the restricted regions encompassing the DA2 and DL5 glomeruli. As both reviewers have noted, this approach offers significant value and we hope the Editor will also recognize that this limitation does not prevent readers from gaining important and novel insights into the synaptic circuitry of these two glomeruli.  

Similar to the approach taken by Tobin at al. (2017) we prioritized producing a densely reconstructed neuropile, in which no synapses were omitted (Tobin et al., 2017). The downside of this method is that not all synaptic connections could be reliably assigned to specific neuronal types, with about 12% remaining unassigned." We anticipate that future research, supported by advances in semi-automated tracing methods, improved imaging technologies, and increased personnel resources, will allow not only for the generation of more complete connectomes of the entire brain (Scheffer et al., 2020; Zheng et al., 2018), but also, for the accurate reconstruction and classification of individual synapses—even in highly complex regions such as the olfactory glomeruli. We also expect that a second complete connectome of a male Drosophila will soon become available, which will provide valuable opportunities for comparisons across individuals and between male and female brains in future studies.

INTERGLOMERULAR DIFFERENCES

Thank you for this insightful comment. It is indeed true that despite both DA2 and VA1v being narrowly tuned glomeruli, they exhibit considerable differences in specific connectivity features (e.g., relative synaptic strengths above certain thresholds) and that those differences can be as pronounced as those observed between DA2 and the broadly tuned DL5. For this reason, comparing each individual glomerulus to every other is not a practical or informative approach. To derive robust interpretations, we focused instead on whether two glomeruli that share a particular functional characteristic—namely, being narrowly tuned for single odorants—also share connectivity patterns that distinguish them from a broadly tuned reference glomerulus.

Our results support this. Furthermore, additional connectomics data reinforce our conclusions.

For example, OSN-OSN connectivity is stronger in the two narrowly tuned glomeruli (DA2 and VA1v) relative to the broadly tuned glomerulus (DL5). While these pairwise differences alone are not conclusive, the finding that the two narrowly tuned glomeruli studied here share features that distinguish them from the broadly tuned glomerulus supports our interpretation. We found further support for this idea in the data reported by Schlegel et al. (2021) further. In that dataset, other narrowly tuned glomeruli (DA1, DL3, and DL4) also exhibit stronger OSNOSN connectivity than other broadly tuned glomeruli (DM1 or DM4).

We do not deny that there are many differences between any given pair of glomeruli, regardless of whether they are narrowly or broadly tunned. Instead, we propose that our findings on circuit features indicate that most of the observed differences actually grouped the two narrowly tuned glomeruli together relative to the broadly tuned glomerulus. A more concise summary is now provided in the newly added Figure 8. We also added explanatory lines of text in the beginning of the chapter ‘specific features of narrowly tuned glomerular circuits. 

ECOLOGICAL SIGNIFICANCE

This is an interesting point. However, it is difficult to disentangle the "ecological significance" of processed odorants from the "tuning breadth" of a glomerulus. In the Drosophila olfactory system, glomerular circuits that respond to ecologically important odorants—such as those involved in reproduction or danger—tend to be more narrowly tuned. Moreover, while we refer to odorants with specific ecological significance as those linked to survival or reproductive behaviors, defining the significance of an odorant with precision is inherently challenging, as it can vary depending on context and environmental conditions.

What both circuits share is their narrow tuning breadth. We therefore propose that the common circuit features of VA1v and DA2, highlighted in this study, are functionally related to the fact that each circuit processes single odorants. Consequently, their specificity is most likely determined at the level of the receptor. 

INDIVIDUAL VARIABILITY

We agree that accounting for inter-animal variability would strengthen the study. However, we are confident that even a modest statistically sound assessment of this variability would require a larger sample size, certainly more than just two or three flies, which is presently not feasible.

We refer the reviewer to our response to Reviewer #1 regarding this important issue.

Initial insights into variability between flies have been provided through comparative analyses of the two most comprehensive female Drosophila melanogaster connectomes—the FAFB and hemibrain datasets (Schlegel et al., 2024). For more detailed quantitative comparisons regarding inter-animal variability, please refer to our response to the second major point raised by Reviewer #2. As highlighted by Schlegel et al. (2024), making definitive statements about the stereotypy of neuron numbers, unitary cell-cell connections (edges), or synaptic strengths (weights) remains a complex challenge."

While appreciating the rigour of this work we were surprised to notice the omission of a comparison of their observations with the two other existing datasets. This would not only have addressed the technical limitation of this particular study - the inability to identify specific neuron types due to imaging a small part of the brain - but would also have shed light on inter-animal variability 

We strongly recommend that the authors do make this comparison - the datasets are currently extremely user friendly and so we don't estimate the replication of their key findings will be too onerous. This will be particularly important to resolve the issue of having to classify all multiglomerular local interneurons and multiglomerular projection neurons - broadly into "MGN. Such a comparison will dramatically strengthen this study that poses very interesting questions, but in its current form, has this striking shortcoming. 

INDIVIDUAL VARIABILITY AS EXPRESSED HERE:

Earlier on we were of the same opinion that the Reviewer express here but, unfortunately, it was not possible to follow his advice. As far as it was possible, we have compared some of our results to the values of the two datasets that the Reviewer refers to, but the absence of glomerulus DA2 in one of the datasets and the absence of synapse annotation for all the relevant glomeruli in the other dataset prevented us from making a full comparison. Moreover, believe that the problem of individual variation most probably cannot be solved by increasing the comparison with one or two more flies.

Reviewer #1 (Recommendations for The Authors): 

The lines 270 - 282 confused me in the backdrop of Figure 3B. 

The concern may stem from our inclusion of a comparison between the uPNs of glomerulus DA2 and the single uPN of glomerulus DL5 in the statistical analysis presented in Figure 3. This comparison was included to ensure a comprehensive representation of the data, highlighting the variability across all major cell groups. We have clarified this rationale in the revised manuscript (see lines 274-282).

Reviewer #2 (Recommendations for The Authors): 

I commend the authors for taking such a thorough approach to advance an interesting topic in olfaction. The following suggestions are intended to strengthen this study: 

Major points: 

A color-blind-friendly palette should be used for all figures. Currently, five of seven figures use red and green, and in particular, Figure 5 will be uninterpretable for red/green color-blind readers. 

We are thankful for this important comment. We changed the color palette as suggested by the reviewer, and replaced Red with Magenta and changed the figure legend accordingly.

This level of analysis is extremely resource and time-consuming, so even obtaining this information at this resolution is an impressive achievement. However, this study would be well served by strategically supplementing the analysis of this dataset with information from other publicly available connectomics datasets. For instance, some interpretations are limited because there is information from only a single DL5 and DA2 glomerulus. Any claims in which one glomerulus has more, less, or the same of a metric must be tempered because without replicates, there are no measures of inter-animal variability. As an example, on lines 386-387 the authors state "The relative synaptic strength between MGN>uPN was stronger in DA2 (12%) than DL5 (10%)". It is difficult to assess whether this represents a difference that is outside of the range of inter-animal variability inherent to the olfactory system. Taking select measures from the Hemibrain and FAFB (via FlyWire) datasets could help strengthen these claims. 

We fully agree with the Reviewer’s opinion that since our data is from one glomerulus of each type “It is difficult to assess whether this represents a difference that is outside of the range of inter-animal variability inherent to the olfactory system.” This is a weakness of practically all connectome studies based on electron microscopy in both Drosophila and other animals We cannot be sure that measurements from the Hemibrain and FAFB datasets could help strengthen our claims, because the magnitude of the range of individual variation is presently not known and most probably solving this problem will require more than one or two more flies. In any case, it is not possible to follow this advice and compare our data with that of the hemibrain because the DA2 was not included in that study. We ask the Reviewer to read our more detailed explanation in our response to Reviewer 1.

In the particular case commented by the Reviewer above, the relative difference in synaptic strength exceeds 20%. Whether such a difference has functional relevance remains an open question but Schlegel et al. (2024) support our interpretation. They showed that synaptic weights with differences larger than 20% tend to be consistent across individuals, with strong correlations within and between animals (Pearson’s R = 0.97 and R = 0.8; Fig. 4).

Grouping all local interneurons, centrifugal neurons response and multiglomerular PNs into one category limits the ability to make interpretations about similarities or differences in the synaptic relationships involving MGNs. The authors could get an estimate of the number of multiglomerular PNs in DL5, VA1v, and DA2 from Hemibrain and FlyWire platforms to get a better sense of differences between glomeruli in the MGN category. 

We agree in that grouping a variety of interneurons into a single category (called MGNs) limits the ability to make interpretations about similarities or differences in the synaptic relationships involving different neurons. This was the unavoidable price to be paid once we decided to register a “comprehensive, synapse-level resolution” map of these two glomeruli. It appears to us that both reviewers have clearly recognized the intrinsic value of this approach and we hope that the Editor will share this opinion. 

Consistent with the assumptions of Tobin et al., (2017) our hypothesis on LN connectivity differences is based on the fact that they are the most numerous and broadly arborizing neurons of the class that we call multiglomerular neurons in the AL (Chou et al., 2010; Lin et al., 2012; Tanaka et al., 2012). Recent connectome studies confirm this feature across all glomeruli (Bates et al., 2020; Horne et al., 2018; Scheffer et al., 2020; Schlegel et al., 2021; Zheng et al., 2018).  

In response to the reviewer’s question, we conducted a case-specific reanalysis of the data from Horne (2018), which provides comprehensive connectivity information for the VA1v glomerulus. This allowed us to quantify the proportional contributions of LNs (n = 56) and mPNs (n = 13) to all MGN connections (MGN-MGN, MGN>OSN, MGN>uPN, uPN>MGN, OSN>MGN).

Our analysis showed that 84% of MGN output originates from LNs. 57% of the input to MGN comes from LNs and 43% from mPNs, largely due to strong OSN>mPN input. Thus, for the filtered MGN connections relevant to distinguishing narrowly from broadly tuned circuits (e.g., MGN>OSN, uPN>MGN; see Fig. 8), LNs are the dominant contributors in VA1v. (These data are not included in the resubmitted manuscript.) This supports our interpretation that the LN are responsible for the majority of MGN connections underlying the observed differences between glomeruli.

For instance, prior work has reported fewer local interneurons innervating DA2, but in this study there was an unexpected result that there was greater MGN innervation density and synapse # for DA2 relative to DL5 This discrepancy could be due to differences in the number of multiglomerular PNs innervating each glomerulus, which would be obscured when these PNs are combined with local interneurons in the MGN category. 

"We agree that the greater MGN innervation density in DA2 in our study could reflect a stronger contribution from mPNs. However, innervation density alone does not indicate how many mPNs actually innervate DA2 or DL5. Alternatively, increased innervation and/or synaptic frequency of local interneurons (LNs) could also account for this observation. In our view, neuron number does not necessarily correlate with branching complexity or synaptic density. 

For example, the dendritic length of the single uPN in glomerulus DL5 is approximately equal to the combined dendritic length of the multiple uPNs of the DA2. Similarly, Tobin et al. (2017) reported that when comparing uPNs in glomerulus DM6 between the left and right brain hemispheres, they found variability in cell number but not in dendritic length. More recently, the FAFB and hemibrain datasets showed a similar pattern in another neuronal type. A substantial variation in cell number was observed for Kenyon cells between the two Drosophila individuals, but this cell type consistently makes and receives, in both individuals, similar presynapses and post-synapses (Schlegel et al., 2024).

On line 33 the authors cannot claim that DA2-OSNs experience less presynaptic inhibition based on the data in this study. Even without the limitations of the MGN category (described above), presynaptic inhibition depends on more than just the number of synapses, rather it is affected by GABA B receptor expression levels and the second messenger components downstream of this receptor. Physiological experiments are needed to justify this claim, so I recommend adjusting accordingly.

We agree with the Reviewer and have adjusted the text on line 33 and in the main body of the text by referring to this finding as “presynaptic input”, which is what we have quantified, instead of “less presynaptic inhibition”.

Figures 5 and 6 seek to distill the wealth of information from this study into broad takehome points for the reader, while still providing a good amount of detail. I think a final more concise graphic summary (similar to the graphical abstract or Figure 6 of Grabe et al 2016) depicting the most critical differences between glomeruli would further clarify the broad findings of this study. 

We appreciate this comment and we have added a “graphic summary” as the Reviewer proposed. We made a new figure that becomes Figure 8 and summarizes our results and highlights differences between narrowly and broadly tuned glomeruli in a more concise graphical abstract format.

Minor points: 

Much of the manuscript provides details about synapse fractions or % synapses for a given synaptic relationship. Please ensure that it is clear which principal cell types are being described, as it can be easy to get lost.  - Should line 284 say "...than DL5 as it has been reported that DA2 is innervated by fewer LNs..."?

We appreciate the reviewer’s comment and we have corrected this sentence that now reads as follows: (see text: beginning at line 290).  

Taisz et al.  has been published, so the citation should be updated. 

We have updated the corresponding citation.  

On line 233, the authors ascribe the small electron-dense vesicles as likely housing sNPF released by MGNs. However, Carlsson et al. (2010) demonstrated that sNPF is released by OSNs, which was further functionally characterized by Root et al. (2011) and Ko et al. (2014). In terms of MGNs that release neuropeptides, Carlsson et al. 2010 demonstrated that local interneurons immunolabel for tachykinin, myoinhibitory peptide, and allatostatin-A, while two extrinsic neurons release SIFamide. In theory, aminergic neurons could also have small electron-dense vesicles, but this can be variable. 

The Reviewer is completely right in his criticism. The MGN certainly contain neurons that have been reported to contain neuropeptides other than sNPF. We have corrected this sentence and it now reads as follows (page7, line 236): “Interestingly, besides the abundant clear small vesicles..

On line 636, the Berck and Schlegel studies demonstrated that panglomerular local interneurons synapse upon OSN, but not that they induce presynaptic inhibition (which was demonstrated in the studies cited in the next sentence). I recommend adjusting this sentence.

We agree and we have corrected the text following the Reviewers advice. It now reads as follows (page 19. Line 663): “We also observed that OSNs received less MGN feedback.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

This is a revision of a manuscript previously submitted to Review Commons. The authors have partially addressed my comments, mainly by expanding the introduction and discussion sections. Sandy Schmid, a leading expert on the AP2 adaptor and CME, has been added as a co-corresponding author. The main message of the manuscript remains unchanged. Through overexpression of fluorescently tagged CCDC32, the authors propose that, in addition to its established role in AP2 assembly, CCDC32 also follows AP2 to the plasma membrane and regulates CCP maturation. The manuscript presents some interesting ideas, but there are still concerns regarding data inconsistencies and gaps in the evidence.

With due respect, we would argue that a role for CCDC32 in AP2 assembly is hardly ‘established’.  Rather a single publication reporting its role as a co-chaperone for AAGAP appeared while our manuscript was under review.  We find some similar and some conflicting results, which are described in our revised manuscript.  However, in combination our two papers clearly show that CCDC32, a previously unrecognized endocytic accessory protein, deserves further study.

(1) eGFP-CCDC32 was expressed at 5-10 times higher levels than endogenous CCDC32. This high expression can artificially drive CCDC32 to the cell surface via binding to the alpha appendage domain (AD)-an interaction that may not occur under physiological conditions.

While we acknowledge that overexpression of eGFP-CCDC32 could result in artificially driving it to CCPs, we do not believe this is the case for the following reasons:

i. The bulk of our studies (Figures 2-4) demonstrate the effects of siRNA knockdown on CCDC32 on CCP early stages of CME, and so it is likely that these functions require the presence of endogenous CCDC32 at nascent CCPs as detected with overexpressed eGFP-CCDC32 by TIRF imaging.

ii. At these levels of overexpression eGFP-CCDC32 fully rescues the effects of siRNA KD of endogenous CCCDC32 of Tfn uptake and CCP dynamics (Figure 6F,G). If the protein was artificially recruited to the AP2 appendage domain, one would expect it to compete with the recruitment of other EAPS to CCPs and hence exhibit defects in CCP dynamics. Indeed, we see the opposite: CCPs that are positive for eGFP-CCDC32 show normal dynamics and maturation rates, while CCPs lacking eGFP-CCDC32 are short-lived and more likely to be aborted (Figure 1C).

iii. We have identified two modes of binding of CCDC32 to AP2 adaptors: one is through canonical AP2-AD binding motifs, the second is through an a-helix in CCDC32 that, by modeling, docks only to the open conformation of AP2.  Overexpressed CCDC32 lacking this a-helix is not recruited to CCPs (Fig. 6 D,E), indicating that the canonical AP2 binding motifs are not sufficient to recruit CCDC32 to CCPs, even when overexpressed.

(2) Which region of CCDC32 mediates alpha AD binding? Strangely, the only mutant tested in this work, Δ78-98, still binds AP2, but shifts to binding only mu and beta. If the authors claim that CCDC32 is recruited to mature AP2 via the alpha AD, then a mutant deficient in alpha AD binding should not bind AP2 at all. Such a mutant is critical for establish the model proposed in this work.

We understand the reviewer’s confusion and thus devoted a paragraph in the discussion to this issue.  As revealed by AlphaFold 3.0 modeling (Figure S6) binding of CCDC32 to the alpha AD likely occurs via the 2 canonical AP2-AD binding motifs encoded in CCDC32. Given the highly divergent nature of AP2-AD binding motifs, we did not identify these motifs without the AlphaFold 3.0 modeling. While these interactions could be detected by GST-pull downs, they are apparently not of sufficient affinity to recruit CCDC32 to CCPs in cells. In the text, we now describe the a-helix we identified as being essential of CCP recruitment as ‘a’ AP2 binding site on CCDC32 rather than ‘the’ AP2 binding site.  Interestingly, and also discussed, Alphafold 3.0 identifies a highly predicted docking site on a-adaptin that is only accessible in the open, cargo-bound conformation of intact AP2.  This is also consistent with the inability of CCDC32(D78-99) to bind the a:µ2 hemi-complex in cell lysates.

We agree that further structural studies on CCDC32’s interactions with AP2 and its targeting to CCPs will be of interest for future work.

(3) The concept of hemicomplexes is introduced abruptly. What is the evidence that such hemicomplexes exist? If CCDC32 binds to hemicomplexes, this must occur in the cytosol, as only mature AP2 tetramers are recruited to the plasma membrane. The authors state that CCDC32 binds the AD of alpha but not beta, so how can the Δ78-98 mutant bind mu and beta?

We introduced the concept of hemicomplexes based on our unexpected (and now explicitly stated as such) finding that the CCDC32(D78-99) mutant efficiently co-IPs with a b2:µ2 hemicomplex.  As stated, the efficiency of this pulldown suggests that the presumed stable AP2 heterotetramer must indeed exist in equilibrium between the two a:s2 and b2:µ2 hemicomplexes, such that CCDC32(D78-99) can sequester and efficiently co-IP with the b2:µ2 hemicomplex.  A previous study, now cited, had shown that the b2:µ2 hemicomplex could partially rescue null mutations of a in C. elegans (PMID: 23482940).  We do not know how CCDC32 binds to the b2:µ2 hemicomplex and we did not detect these interactions using AlphaFold 3.0. However, these interactions could be indirect and involve the AAGAB chaperone.  It is also likely, based on the results of Wan et al. (PMID: 39145939), that the binding is through the µ2 subunit rather than b2. As mentioned above, and in our Discussion, further studies are needed to define the complex and multi-faceted nature of CCDC32-AP2 interactions.

(4) The reported ability of CCDC32 to pull down AP2 beta is puzzling. Beta is not found in the CCDC32 interactome in two independent studies using 293 and HCT116 cells (BioPlex). In addition, clathrin is also absent in the interactome of CCDC32, which is difficult to reconcile with a proposed role in CCPs. Can the authors detect CCDC32 binding to clathrin?

Based on the studies of Wan et al. (PMID: 39145939), it is likely that CCDC32 binds to µ2, rather than to the b2 in the b2:µ2 hemicomplex.  As to clathrin being absent from the CCDC32 pull down, this is as expected since the interactions of clathrin even with AP2 are weak in solution (as shown in Figure 5C, clathrin is not detected in our AP2 pull down) so as not to have spontaneous assembly of clathrin coats in the cytosol. Rather these interactions are strengthened by both the reduction in dimensionality that occurs on the membrane and by avidity of multivalent interactions.  For example, Kirchausen reported that 2 AP2 complexes are required to recruit one clathrin triskelion to the PM.

(5) Figure 5B appears unusual-is this a chimera?

Figure 5B shows an internal insertion of the eGFP tag into an unstructured region in the AP2 hinge. As we have previously shown (PMID: 32657003), this construct, unique among other commonly used AP2 tags, is fully functional.  We have rearranged the text in the Figure legend to make this clearer.

Figure 5C likely reflects a mixture of immature and mature AP2 adaptor complexes.

This is possible, but mature heterotetramers are by far the dominant species, otherwise the 4 subunits would not be immuno-precipitated at near stoichiometric levels with the a subunit.  Near stoichiometric IP with antibodies to the a-AD have been shown by many others in many cell types. 

(6) CCDC32 is reduced by about half in siRNA knockdown. Why not use CRISPR to completely eliminate CCDC32 expression?

Fortuitously, partial knockdown was essential to reveal this second function of CCDC32, as we have emphasized in our Discussion.  Wan et al, used CRISPR to knockout CCDC32 and reveal its essential role as a AAGAB co-chaperone.  In the complete absence of CCDC32 mature AP2 complexes fail to form.  However, under our conditions of partial CCDC32 depletion, the expression of AP2 heterotetramers is unaffected revealing a second function of CCDC32 at early stages of CME.  We expect that the co-chaperone function of CCDC32 is catalytic, while its role in CME is more structural; hence the different concentration dependencies, the former being less sensitive to KD than the latter.  This is one reason that many researchers are turning to CRISPRi for whole genome perturbation studies as many proteins play multiple roles that can be masked in KO studies.

Reviewer #2 (Public review):

Yang et al. describes CCDC32 as a new clathrin mediated endocytosis (CME) accessory protein. The authors show that CCDC32 binds directly to AP2 via a small alpha helical region and cells depleted for this protein show defective CME. Finally, the authors show that the CCDC32 nonsense mutations found in patients with cardio-facial-neuro-developmental syndrome (CFNDS) disrupt the interaction of this protein to the AP2 complex. The results presented suggest that CCDC32 may act as both a chaperone (as recently published) and a structural component of the AP2 complex.

Strengths:

The conclusions presented are generally well supported by experimental data and the authors carefully point out the differences between their results and the results by Wan et al. (PNAS 2024).

Weaknesses:

The experiments regarding the role of CCDC32 in CFNDS still require some clarifications to make them clearer to scientists working on this disease. The authors fail to describe that the CCDC32 isoform they use in their studies is different from the one used when CFNDS patient mutations were described. This may create some confusion. Also, the authors did not discuss that the frame-shift mutations in patients may be leading to nonsense mediated decay.

As requested we have more clearly described our construct with regard to the human mutations and added the possibility of NMD in the context of the human mutations.

Reviewer #3 (Public review):

In this manuscript, Yang et al. characterize the endocytic accessory protein CCDC32, which has implications in cardio-facio-neuro-developmental syndrome (CFNDS). The authors clearly demonstrate that the protein CCDC32 has a role in the early stages of endocytosis, mainly through the interaction with the major endocytic adaptor protein AP2, and they identify regions taking part in this recognition. Through live cell fluorescence imaging and electron microscopy of endocytic pits, the authors characterize the lifetimes of endocytic sites, the formation rate of endocytic sites and pits and the invagination depth, in addition to transferrin receptor (TfnR) uptake experiments. Binding between CCDC32 and CCDC32 mutants to the AP2 alpha appendage domain is assessed by pull down experiments. While interaction between CCDC32 and the alpha appendage domain of AP2 is clearly described, a discussion of potential association with other AP2 domains would be beneficial to understand the impact of CCDC32 in endocytosis.

The reviewer is correct. That CCDC32 also interacts with other subunits of AP2, is evident from the findings of Wan et al. and by the fact that the CCDC32(D78-99) mutant efficiently co-IPs with the b2:µ2 hemicomplex.  We expanded our discussion around this point. CCDC32 remains an, as yet, poorly characterized, but we now believe very interesting EAP worth further study.

Together, these experiments allow deriving a phenotype of CCDC32 knock-down and CCDC32 mutants within endocytosis, which is a very robust system, in which defects are not so easily detected. A mutation of CCDC32, mimicking CFNDS mutations, is also addressed in this study and shown to have endocytic defects.

In summary, the authors present a strong combination of techniques, assessing the impact of CCDC32 in clathrin mediated endocytosis and its binding to AP2.

Recommendations for the authors:

Reviewer #2 (Recommendations for the authors):

(1) The authors must be clear about the differences between the CCDC32 isoform they used in their manuscript and the one used to describe the patient mutations. This could be done, for example, in the methods. This is essential for the capacity of other labs to reproduce, follow up and correctly cite these results.

We have added this information to the Methods. 

(2) I believe the authors have misunderstood what nonsense mediated decay is. NMD occurs at the mRNA level and requires a full genome context to occur (introns and exons). The fact that a mutant protein is expressed normally from a construct by no means prove that it does not happen. I believe that adding the possibility of NMD occurring would enrich the discussion.

Thank you, we have now done more homework and have added this possibility into our discussion of the mutant phenotype.  However, if a robust NMD mechanism resulted in a complete loss of CCDC42 protein, then the essential co-chaperone function reported by Wan et al, would result in complete loss of AP2.  A more detailed characterization of the cellular phenotype of these mutations, including assessing the expression levels of AP2 would be informative.

Reviewer #3 (Recommendations for the authors):

- It is not clear what the authors mean by '~30s lifetime cohort' (line 159). They refer to Figure 2H, which shows the % of CCPs. Can the authors explain exactly what kind of tracks they used for this analysis, for example which lifetime variations were accepted? Do they refer to the cohorts in Figure S4? In Figure S4, the most frequent tracks have lifetimes < 20 s (in contrast to what is stated in the main text). Why was this cohort not used?

The ‘30s cohort’ refers to CCPs with lifetimes between 25-35s which encompasses the most abundant species in control cells and CCDC32 KD cells, as shown by the probability curves in Figure 2H. Given the large number of CCPs analyzed we still have large numbers for our analyses n=5998 and 4418, for control and siRNA treated conditions, respectively.  Figure 2H shows the frequency of CCPs in cells treated with CCDC32 siRNA are shifted to shorter lifetimes. We have clarified this in the text.

- Figure S1: It is now clear, why the mutant versions of CCDC32 are not detected in this western blot. However, data that show the resistance of these proteins to siCCDC32 is still missing (S1 A is in the absence of siCCSC32 I assume, as the legend suggests). A western blot using an anti-GFP antibody, as the one used in Figure S1, after siRNA knock-known would provide clarity.

That these constructs all contain the same mutation in the siRNA target sequence gives us confidence that they are indeed resistant to siRNA.

- Note that the anti-CCDC32 antibody does not detect the eGFP-CCDC32(∆78-98) as well as full-length and is unable to detect eGFP-CCDC32(1-54)'. This phrase should belong to Figure S1 (B), not (A)

Corrected.

- The immunoprecipitations of CCDC32 and its mutants with AP2 and its subunits are partially confusing. In Figure 5, the authors show that CCDC32 interacts specifically with the alpha-AD, but not with the beta-AD of AP2. In Figure 6B and C, on the other hand, Co-IPs are shown also with the beta and the mu domain of AP2. This is understandable in the context of the full AP2. However, when interaction with the alpha domain (and sigma) is abolished through mutation of helix 78-98, why would beta and mu still interact, when the beta-AD cannot interact with CCDC32 on its own. Are there interaction sites expected outside the ADs in the beta or mu domains?

See responses to reviewer 1 above.  This result likely reflects the co-chaperone activity of CCDC32 as reported by Wan et al it likely due to their reported interactions of CCDC32 with the µ2 subnit of b2:µ2 hemicomplexes.

- Figure S6 D, E and F: How much confidence do the authors have on the AlphaFold predictions? Have the same binding poses been obtained repeatedly by independent predictions?

We provide, with a color scale, the confidence score for each interaction, which is very high (>90%). Of course, this is still a prediction that will need to be verified by further structural studies as we have stated.

Author response:

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

Reviewer #1 (Public review):

Summary:

Cook et al. have presented an important study on the transcriptomic and epigenomic signature underlying craniofacial development in marsupials. Given the lack of a dunnart genome, the authors also prepared long and short-read sequence datasets to assemble and annotate a novel genome to allow for the mapping of RNAseq and ChIPseq data against H3K4me3 and H3K27ac, which allowed for the identification of putative promoter and enhancer sites in dunnart. They found that genes proximal to these regulatory loci were enriched for functions related to bone, skin, muscle and embryonic development, highlighting the precocious state of newborn dunnart facial tissue. When compared with mouse, the authors found a much higher proportion of promoter regions aligned between species than for enhancer regions, and subsequent profiling identified regulatory elements conserved across species and are important for mammalian craniofacial development. In contrast, the identification of dunnart-specific enhancers and patterns of RNA expression further confirm the precocious state of muscle development, as well as for sensory system development, in dunnart suggesting that early formation of these features are critical for neonate marsupials likely to assist with detecting and responding to cues that direct the joeys to the mother's teat after birth. This is one of the few epigenomic studies performed in marsupials (of any organ) and the first performed in fat-tailed dunnart (also of any organ). Marsupials are emerging as an important model for studying mammalian development and evolution and the authors have performed a novel and thorough analysis, impressively including the assembly of a new marsupial reference genome that will benefit many future studies.

Strengths:

The study provides multiple pieces of evidence supporting the important role enhancer elements play in mammalian phenotypic evolution, namely the finding of a lower proportion of peaks present in both dunnart and mouse for enhancers than for promoters, and dunnart showing more genes uniquely associated with it's active enhancers than any other combination of mouse and dunnart samples, whereas this pattern was less pronounced than for promoter-associated genes. In addition, rigorous parameters were used for the cross-species analyses to identify the conserved regulatory elements and the dunnart-specific enhancers. For example, for the results presented in Figure 1, I agree that it is a little surprising that the average promoter-TSS distance is greater than that for enhancers, but that this could be related to the possible presence of unannotated transcripts between genes. The authors addressed this well by examining the distribution of promoter-TSS distances and using proximal promoters (cluster #1) as high confidence promoters for downstream analyses.

The genome assembly method was thorough, using two different long read methods (Pacbio and ONT) to generate the long reads for contig and scaffold construction, increasing the quality of the final assembled genome.

Weaknesses:

Biological replicates of facial tissue were collected at a single developmental time point of the fat-tailed dunnart within the first postnatal day (P0), and analysed this in the context of similar mouse facial samples from the ENCODE consortium at six developmental time points, where previous work from the authors have shown that the younger mouse samples (E11.5-12.5) approximately corresponds to the dunnart developmental stage (Cook et al. 2021). However, it would be useful to have samples from at least one older dunnart time point, for example, at a developmental stage equivalent to mouse E15.5. This would provide additional insight into the extent of accelerated face development in dunnart relative to mouse, i.e. how long do the regulatory elements that activated early in dunnart remain active for and does their function later influence other aspects of craniofacial development?

We thank the reviewer for their feedback and agree that the inclusion of multiple postnatal stages in the dunnart would give further valuable insights to the comparative analyses. Unfortunately, we were limited by the pouch young available and prioritized ensuring robust data at a single stage for this study. We hope to expand this work to more stages in future studies.

The authors refer to the development of the CNS being delayed in marsupials relative to placental mammals, however, evidence shows how development of the dunnart brain (whole brain or cortex) is protracted compared to mouse, by a factor of at least 2 times, rather than delayed per se (Workman et al. 2013; Paolino et al. 2023). In addition, there is evidence that cortical formation and cell birth may begin at approximately the same stage across species equivalent to the neonate period in dunnart (E10.5 in mouse), and that shortly after this at the stage equivalent to mouse E12.5, the dunnart cortex shows signs of advanced neurogenesis followed by a protracted phase of neuronal maturation (Paolino et al. 2023). Therefore, it is possible that marsupial CNS development appears delayed relative to mouse but instead begins at the same stage and then proceeds to develop on a different timing scale.

The comparison here is not directly between CNS development in placental and marsupials but CNS development relative to development of a subset of structures of the cranial skeleton and musculature (as first proposed by Kathleen Smith 1997). For example, Smith 1997 found that in eutherians, evagination of the telencephalon and appearance of the pigment in the eye occur before the ossification of the premaxilla, maxilla, and dentary. However, in marsupials, evagination of the telencephalon and appearance of the pigment in the eye occur concurrently with condensation of cartilage in the basicranium and the ossification of the premaxilla, maxilla, and dentary. Smith 1997 reports both a delay in the initiation of CNS development in marsupials relative to craniofacial ossification and a protraction of CNS development compared to placental mammals.

This also highlights the challenges of correlating different staging systems between placentals and marsupials as stages determined as equivalent can change depending on which developmental events are used. The protracted development of the CNS in marsupials (Smith 1997, Workman et al. 2013; Paolino et al. 2023) still supports the hypothesis that during the short gestation period in marsupials structures required for life outside the womb in an embryonic-like state, such as the orofacial region, are likely prioritized.

We have clarified this based on the reviewers feedback and added text referring to the protraction of marsupial CNS development to the Discussion section.

[New text]: Marsupials display advanced development of the orofacial region relative to development of the central nervous system when compared to placental mammals[3,6].

[New text]: Although development of the central nervous system is protracted in marsupials compared to placentals, marsupials have well-developed peripheral motor nerves and sensory nerves (eg. the trigeminal) at birth [5].

Reviewer #2 (Public review):

This study by Cook and colleagues utilizes genomic techniques to examine gene regulation in the craniofacial region of the fat-tailed dunnart at perinatal stages. Their goal is to understand how accelerated craniofacial development is achieved in marsupials compared to placental mammals.

The authors employ state-of-the-art genomic techniques, including ChIP-seq, transcriptomics, and high-quality genome assembly, to explore how accelerated craniofacial development is achieved in marsupials compared to placental mammals. This work addresses an important biological question and contributes a valuable dataset to the field of comparative developmental biology. The study represents a commendable effort to expand our understanding of marsupial development, a group often underrepresented in genomic studies.

The dunnart's unique biology, characterized by a short gestation and rapid craniofacial development, provides a powerful model for examining developmental timing and gene regulation. The authors successfully identified putative regulatory elements in dunnart facial tissue and linked them to genes involved in key developmental processes such as muscle, skin, bone, and blood formation. Comparative analyses between dunnart and mouse chromatin landscapes suggest intriguing differences in deployment of regulatory elements and gene expression patterns.

Strengths

(1) The authors employ a broad range of cutting-edge genomic tools to tackle a challenging model organism. The data generated - particularly ChIP-seq and RNA-seq from craniofacial tissue - are a valuable resource for the community, which can be employed for comparative studies. The use of multiple histone marks in the ChIP-seq experiments also adds to the utility of the datasets.

(2) Marsupial occupy an important phylogenetic position, but they remain an understudied group. By focusing on the dunnart, this study addresses a significant gap in our understanding of mammalian development and evolution. Obtaining enough biological specimens for these experiments studies was likely a big challenge that the authors were able to overcome.

(3) The comparison of enhancer landscapes and transcriptomes between dunnarts and can serve as the basis of subsequent studies that will examine the mechanisms of developmental timing shifts. The authors also carried out liftover analyses to identify orthologous enhancers and promoters in mice and dunnart.

Weaknesses and Recommendations

(1) The absence of genome browser tracks for ChIP-seq data makes it difficult to assess the quality of the datasets, including peak resolution and signal-to-noise ratios. Including browser tracks would significantly strengthen the paper by provide further support for adequate data quality.

We have put together an IGV session with the dunnart genome, annotation and ChIP-seq tracks. This is now available in the FigShare data repository (10.7554/eLife.103592.1).

(2) The first two figures of the paper heavily rely in gene orthology analysis, motif enrichment, etc, to describe the genomic data generated from the dunnart. The main point of these figures is to demonstrate that the authors are capturing the epigenetic signature of the craniofacial region, but this is not clearly supported in the results. The manuscript should directly state what these analyses aim to accomplish - and provide statistical tests that strengthen confidence on the quality of the datasets.

As this is the first epigenomic profiling for this species we performed extensive data quality control (See Supplementary Tables 2-3, 18, 20-23 and Supplementary Figures 1-3, 6-11). These figures and corresponding Supplementary Tables show the robustness of the data, including well-described metrics for assessing promoters and enhancers, GO terms relevant to craniofacial development and binding motifs for key developmental TF families.

We have emphasised this aspect of the work more strongly in the results section, particularly in [Defining craniofacial putative enhancer- and promoter regions in the dunnart].

(3) The observation that "promoters are located on average 106 kb from the nearest TSS" raises significant concerns about the quality of the ChIP-seq data and/or genome annotation. The results and supplemental information suggest a combination of factors, including unannotated transcripts and enhancer-associated H3K4me3 peaks - but this issue is not fully resolved in the manuscript. The authors should confirm that this is not caused by spurious peaks in the CHIP-seq analysis - and possibly improve genome annotation with the transcriptomic datasets presented in the study.

Spurious ChIP-seq peaks could be possible as there is no “blacklisted regions” database for the dunnart to filter on, however we used a no-IP control, a stringent FDR of 0.01 and peaks had to be reproducible in two biological replicates when calling peaks - all of which should reduce the likelihood of false positives.

H3K4me3 activity at enhancers is well-established, in particular when enhancer sequences are also bound by RNA Pol II ((Koch and Andrau, 2011; Pekowska et al., 2011). However, compared to H3K4me3 activity at promoters, H3K4me3 levels at enhancers are low (Calo and Wysocka, 2013). This is in line with our observations that H3K4me3 levels at enhancers are much lower than observed at promoter regions (see Supplementary Note 2). We found that H3K4me3 peaks located closer to the TSS had a stronger peak signal (mean = 46.10) than distal H3K4me3 peaks (mean = 6.95; Wilcoxon FDR-adjusted p < 2.2 x 10^-16). This suggests that although some distal promoter peaks may be due to missingness in the annotation, the majority likely represent peaks associated with enhancer regions. We have emphasized this finding more strongly in the results section:

[New text]: H3K4me3 activity at enhancers is well-established[25,26], however, compared to H3K4me3 activity at promoters, H3K4me3 levels at enhancers are low[27]. This is in line with our observations where H3K4me3 levels at distal enhancer peaks are nearly 7 times lower than those observed at promoter regions (see SupNote2).

(4) The comparison of gene regulation between a single dunnart stage (P1) and multiple mouse stages lacks proper benchmarking. Morphological and gene expression comparisons should be integrated to identify equivalent developmental stages. This "alignment" is essential for interpreting observed differences as true heterochrony rather than intrinsic regulatory differences.

Given the developmental differences between eutherian and marsupial mammals it is challenging to assign the dunnart a precise “equivalent” developmental stage to the mouse. From our morphological and developmental characterisation (see Cook et al. 2020 Nat Comms Bio) based on ossification patterns the dunnart orofacial region on the day of birth appears to be similar to that of an E12.5 mouse embryo (just prior to the observation of ossified craniofacial bones). However, when we compared both regulatory elements and expressed genes between the dunnart at this stage (P1) and 5 developmental stages in the mouse, there is no obvious equivalent stage. For example, when we simply compare genes linked to enhancer peaks, the group with the largest intersection between dunnart and any mouse stage are ~500 genes that are present in dunnart, and mouse stages E10.5, E12.5 - E15.5, Figure 5B). When we then compare genes expressed in the dunnart to temporal gene expression dynamics during mouse development we find that the largest overlap is with genes highly expressed at E14.5 or E15.5 in the mouse (Figure 6, Supplementary Figure 5). We have strengthened the rationale for the selected mouse stages in the comparative analyses section of the results.

(5) The low conservation of putative enhancers between mouse and dunnart (0.74-6.77%) is surprising given previous reports of higher tissue-specific enhancer conservation across mammals. The authors should address whether this low conservation reflects genuine biological divergence or methodological artifacts (e.g., peak-calling parameters or genome quality). Comparisons with published studies could contextualize these findings.

The reported range (0.74 - 6.77%) refers to the number regions called as an active enhancer peak in both species (conserved activity) divided by the total number of dunnart peaks alignable to the mouse genome, which we expect to be low given sequence turnover rates and the evolutionary distance separating dunnart and mice. The alignability (conserved sequence) for dunnart enhancers to the mouse genome was ~13% for 100bp regions and can be found in Supplementary Table 22, we have now clarified this in the main text.

[New Text]: After building dunnart-mm10 liftover chains (see Methods and SupNote5) we compared mouse and dunnart regulatory elements. The alignability (conserved sequence) for dunnart enhancers to the mouse genome was ~13% for 100bp regions (Supplementary Table 22).

The activity conservation range reported here is consistent with previously reported for marsupial-placental enhancer comparisons (Villar et al. 2015), where ~1% of conserved liver-specific human enhancers had conserved activity to opossum. Follow up studies in Berthelot et al 2018 also found that approximately 1% of human liver enhancers were conserved across the placental mammals included in the study.

(6) Focusing only on genes associated with shared enhancers excludes potentially relevant genes without clear regulatory conservation. A broader analysis incorporating all orthologous genes may reveal additional insights into craniofacial heterochrony.

We appreciate the reviewers comment, we understand that a broader analysis may provide some additional insights to this question however in this study our focus was understanding the enhancers driving craniofacial development in these species. We linked enhancers with gene expression data as additional evidence of regulatory programs involved in craniofacial development. The majority (~70%) of genes reproducibly expressed were linked to an active enhancer and/or promoter.   This has now been highlighted in the result section.

[New Text]: There were 12,153 genes reproducibly expressed at a level > 1 TPM across three biological replicates, with the majority of genes 67% of genes expressed (67%; 8158/12153) associated with near an active enhancer and/or promoter peak.

In conclusion, this study provides an important dataset for understanding marsupial craniofacial development and highlights the potential of genomic approaches in non-traditional model organisms. However, methodological limitations, including incomplete genome annotation and lack of developmental benchmarking weaken the robustness and of the findings. Addressing these issues would significantly enhance the study's utility to the field and its ability to support the study's central conclusion that dunnart-specific enhancers drive accelerated craniofacial development.

Reviewer #1 (Recommendations for the authors):

Minor comments and corrections:

(1) ChIP-seq FRiP fractions were much higher in dunnart samples than in mouse. Is this related to any differences in sample preparation they are aware of in the ENCODE datasets of mouse, such as different anti-histone antibodies used (and therefore different efficiency of binding to the same histone markers across species)? The authors appear to have addressed something similar with respect to the much lower enriched peak number observed in the mouse sample relative to dunnart in Supp note 4. I suspect the "technical cofounder" they refer to there is affecting both the FRiP scores and the higher correlation coefficients between IP and input in mouse.

We chose the same antibodies used in the mouse craniofacial tissue ENCODE experiments however, the procedure is slightly different. We used the MAGnify Chromatin Immunoprecipitation System while in the ENCODE assays performed by Bing Ren’s group in 2012 was an in-house lab protocol for MicroChIP. Given that the samples for mouse and dunnart were not processed together, by the same researcher, with the same protocol there could be any number of technical cofounders impacting enrichment. A low FRiP score suggests low specificity as the majority of reads are in non-specific regions (low enrichment), consistent with the higher correlation between IP and input in mouse. The data quality also appears to vary between H3K27ac and H3K4me3 in the mouse (Supplementary Table 21), with H3K4me3 FRiP scores more similar to those observed in our dunnart experiments. This suggests a potential confounder specific to the mouse H3K27ac IP. QC metrics (FRiP, bam correlation) are consistent between H3K27ac and H3K4me3 IPs in our experiments (Supplementary Table 20).

(2) Some of the promoter peak numbers in Supp table 1 do not match the numbers in the main text.

We have corrected the incorrect number reported in the text for promoter peaks with orthologous genes (8590 -> 8597).

(3) In Supp tables 2 and 3, the number of GO terms similar across tables is 466, which is ~42% of total number of enriched GO terms. However the authors mention that only 23% of terms were the same between promoters and enhancers, and a value of 42% was applied to the proportion of terms uniquely enriched for terms associated with genes assigned to promoters only. Unless I'm reading these Supp tables incorrectly, is it possible the proportions were mixed up?

Thanks for catching this. The lists provided in Supplementary Table 2 were incorrect. The Supplementary Tables and in text description has been corrected to reflect this.

(4) Would be helpful to add a legend for the mouse samples in Supp Figure 10.

We have added the labels to the plot.

(5) In Supp note 5, regarding the percentage of alignable peaks recovered, the percentages mentioned for the 50bp and 500bp peak summit lengths for enhancers and promoters do not seem to match the values in Supp tables 22 and 23.

Thank you for catching this - we have corrected the Supplementary Tables and in text.

(6) Please provide additional information to explain how dunnart RNA expression was associated with the five temporal expression clusters found in the mouse data shown in Figure 6 given there is only one dunnart time point and so the species temporal pattern's could not be compared, i.e. how was the odds ratio calculated and was this applied iteratively for dunnart against each mouse age and within each temporal cluster?

The TCseq package takes the mouse expression data across all 6 stages and calls differentially expressed genes with an absolute log₂ fold-change > 2 compared to the starting time-point (E10.5). The mouse gene expression patterns were clustered into 5 clusters that each show distinct temporal expression patterns (see Supplementary Figure 5D). The output from this is 5 lists where within each list are unique genes that share a temporal pattern. These lists of mouse genes were then each compared to the orthologous genes expressed in the dunnart using a Fishers Exact test with corrections for multiple testing using the Holm method. We have added additional details in the methods:

[New text]: Orthologous genes reproducibly expressed >1 TPM in the dunnart were compared to the list of genes for each cluster using Fisher’s Exact Test followed by p-value corrections for multiple testing with the Holm method.

(7) SupFile1 and SupFile2 - which supplementary note or figure are these referring to?

Apologies for this error. These items were meant to link to the FigShare repository where the supplementary files can be found. We have corrected this using the DOI for the repository.

Reviewer #2 (Recommendations for the authors):

(1) Authors should clarify that the mouse ENCODE data used for the comparisons was obtained from craniofacial tissue.

This has now been corrected to clarify that the mouse ENCODE data used was from craniofacial tissues. ENCODE mouse embryonic facial prominence ChIP-seq and gene expression quantification file accession numbers and details used in study can be found in Supplementary Table 17.

(2) Given the large differences in TPM for highly expressed genes shown in Figure 5, a MA or volcano plot would provide a more comprehensive view of global transcriptome differences between species.

We have added this plot as Supplementary Figure 13.

(3) It is unclear whether the enrichment analysis was performed for mouse genes, dunnart genes, or both.

In reference to Figure 5, Gene Ontology enrichment analysis was performed on the top 500 highly expressed genes in dunnart. Because there is not an ontology database for dunnart gene IDs, these top 500 dunnart gene IDs were converted to the orthologous gene ID in mouse before performing the enrichment analysis. We apologise for the lack of clarity and have added additional text in the results section to make this clearer. In addition, the relevant methods section now reads:

[New text]: As there is no equivalent gene ontology database for dunnart, we converted the Tasmanian devil RefSeq IDs to Ensembl v103 using biomaRt v2.46.3 and then converted these to mouse Ensembl v103 IDs. In this way we were able to use the mouse Ensembl Gene Ontology annotations for the dunnart gene domains. All gene ontology analyses were performed using clusterProfiler v4.1.4[117], with Gene Ontology from the org.Mm.eg.db v3.12.0 database[118], setting an FDR-corrected p-value threshold of 0.01 for statistical significance.

Reviewer #1 (Public review):

Ono et al. compared the activity of prime editor Nickase PE2 and prime editor nuclease PEn in introducing SNPs and short exogenous DNA sequences into the zebrafish genome to model human disease variants. They find the nickase PE2 prime editor had a higher rate of precise integration for introducing single-nucleotide substitutions, whereas the nuclease PEn prime editor showed improved precision of integration of short DNA sequences. In somatic tissue, the percentage of SNP variant precision edits improved when using PE2 RNP injection instead of mRNA injection, but increased precision editing correlated with elevated indel formation. While PEn overall had higher rates of precision edits, the indel rate was also elevated. Similar rates were observed when introducing a 3 bp stop codon into the ror gene using a standard pegRNA with a 13-nucleotide homology arm, or a springRNA lacking the homology arm that drives integration via NHEJ. Inclusion of an abasic sequence in the springRNA prevented imprecise edits caused by scaffold incorporation, but did not improve the overall percentage of precise edits in somatic tissue. Recovery of a germline ror-TGA integration allele using PEn with RNP was robust, resulting in 5 out of 10 founders transmitting a precise allele. Lastly, the authors demonstrate that PEn was effective at the integration of a 30 bp nuclear localization signal into the 5' end of GFP in an existing muscle-specific reporter line. However, the undefined number of cassettes in this multicopy transgene complicates accurate measurements of editing frequency. Integration of the NLS or other longer sequences at an endogenous locus would demonstrate the broad utility of this approach. From the work presented, it is unclear how prime editing could be used to transiently model human pathogenic variants, given the low frequency of precision edits in somatic tissue, or to isolate stable germline alleles of variants that are potentially dominant negative or gain-of-function in nature. Without a direct comparison with CRISPR/Cas9 nuclease HDR-based methods that use oligonucleotide templates to introduce edits, the advantage of prime editing is unclear. A cost comparison between prime editing and HDR methods would also be of interest, particularly for integration of longer DNA sequences.

The conclusions of the paper are mostly well supported, but some changes to the text and additional analyses would strengthen the conclusion that PE2 vs. PEn is preferred for introducing variants, short or long DNA sequences.

(1) In Figure 3, the data indicate a significant increase in precise edits of the 3 bp TGA using PE2 RNP (11.5%) vs. PE2 mRNA (1.3%). At the adgrf3b locus, only PEn mRNA was tested for introducing the 3 bp and 12 bp insertions. The previous study testing PE2 for 3 and 12 bp insertions was mentioned, but the frequency was not listed, and the study wasn't cited (lines 204 - 207). A comparison of germline transmission rates using PE2 vs. PEn would support the conclusion that PEn allows precise integration of longer templates and recovery of germline integration alleles.

(2) Figure 4 shows the results of introducing a TGA stop codon that is predicted to result in nonsense-mediated decay. Testing the ability to also isolate different substitution mutations in the germline would be useful information for identifying the most effective approach for generating human disease variant models.

(3) A comparison with the prime editing variant knock-in frequencies reported in the recent publication by Vanhooydonck et al., 2025, Lab Animal should be included in the Discussion.

Reviewer #3 (Public review):

The manuscript by Ono et al describes the application of prime editors to introduce precise genetic changes in the zebrafish model system. Probably the most important observation is that, compared to the "standard" PE2, the prime editor with full nuclease activity appears to be more efficient at introducing insertions into the genome. Although many laboratories around the world have successfully used oligonucleotide-mediated HDR to insert short exogenous sequences such as epitope tags or loxP sites into the zebrafish genome, the method suffers from a high frequency of indels at the edit site. Thus, additional tools are badly needed, making this manuscript very important. Length of the longer reported insertion (+30) is quite close to the range of V5 (14 amino acids) and ALFA (12 amino acids without "spacer" prolines) epitope tags, as well as loxP site (34 nucleotides). Conclusions drawn in the paper are supported by compelling evidence. I only have a few minor comments:

(1) The logic for introducing two nucleotide changes (at +3 and +10) to change a single amino acid (I378) should be explicitly explained in the main body of the manuscript. It is indeed self-explanatory when looking at Supplementary Figure 1. One way of doing it could be to include Supplementary Figure 1a in Figure 1.

(2) It is not clear why a 3-nucleotide insertion was used to generate W722X. The human W720X is a single-nucleotide polymorphism, and it should be possible to make a corresponding zebrafish mutant by introducing two nucleotide changes.

(3) Lines 137-138: T7 Endonuclease assay used in Figure 2d detects all polymorphisms, both precise changes and indels. Thus, if this assay were performed on embryos shown in Figure 1c-d, the overall percentage of modified alleles would be similarly higher for PEn over PE2 (add up precise prime edits and indels). The conclusion in the last sentence of the paragraph is, therefore, incorrect, I believe.

(4) Use of terminology. "Germline transmission" is typically used to refer to the fraction of F0s transmitting desired changes (or transgenes) to their progeny, while "germline mosaicism" refers to the fraction of F1s with the desired change in the progeny of a given F0. "Germline transmission" in line 217 should be replaced with "germline mosaicism".

(5) Lines 253-255: The fraction of injected embryos that had mosaic nuclear expression of GFP, indicative of NLS insertion, should be clarified. It should also be clarified whether embryos positive for nuclear GFP were preselected for amplicon sequencing and germline transmission analyses. This is extremely important for extrapolation to scenarios like epitope tagging, where preselection is not possible.

(6) Statistical analyses. It would be helpful to clarify why different statistical tests are sometimes used to assess seemingly very similar datasets (Figures 1c, 1d, 2b, 2c, 2f).

(7) Discussion. Since authors suggest that PEn might be especially beneficial for insertion of additional sequences, it is important to stress locus-to-locus variability of success. While the precise +3 insertion was indeed tremendously efficient at both tested loci (ror2 and adgrf3b), +12 addition into adgrf3b was over 10 times less efficient (lines 193-194). In contrast, +30 into smyhc:GFP using the shorter pegRNA was highly efficient again with an average of 8.5% of sequence reads indicating precise integration (line 257, Figure 5c). Longer pegRNA did not work nearly as well (Figure 5c), but was still much better than +12 into adgrf3b. As dangerous as it is to extrapolate from small datasets, perhaps these observations indicate that optimization of RT template and PBS may be needed for each new locus in order to significantly outperform oligonucleotide-mediated HDR? If so, would the cost of ordering several pegRNAs and the effort needed to compare them factor in when deciding which method to use? Reported germline transmission rates for both ror2 W722X (+3, Figure 4a) and smyhc:NLS-GFP (+30, Figure 5f) are tantalizingly high.

Reviewer #2 (Public review):

Summary:

This work by Kadeřábková and Furniss et al. demonstrates the importance of a specific protein folding system to effectively folding β-lactamase proteins, which are responsible for resistance to β-lactam antibiotics, and shows that inhibition of this system sensitize multidrug-resistant pathogens to β-lactam treatment. In addition, the authors extend these observations to a two-species co-culture model where β-lactamases provided by one pathogen can protect another, sensitive pathogen from β-lactam treatment. In this model, disrupting the protein folding system also disrupted protection of the sensitive pathogen from antibiotic killing. Overall, the data presented provide a convincing foundation for subsequent investigations and development of inhibitors for β-lactamases and other resistance determinants. This and similar strategies may have application to polymicrobial contexts when molecular interactions are suspected to confer resistance to natively antibiotic-sensitive pathogens.

Strengths:

The authors use clear and reliable molecular biology strategies to show that β-lactamase proteins from P. aeruginosa and Burkholderia species, expressed in E. coli in the absence of the dsbA protein folding system, are variably less capable of resisting the effects of different β-lactam antibiotics compared to the dsbA-competent parent strain (Figure 1). The appropriate control is included in the supplemental materials to demonstrate that this effect is specifically dependent on dsbA, since complementing the mutant with an intact dsbA gene restores antibiotic resistance (Figure S1). The authors subsequently show that this lack of activity can be explained by significantly reduced protein levels and loss-of-function protein misfolding in the dsbA mutant background (Figure 2). These data support the importance of this protein folding mechanism in the activity of multiple clinically relevant β-lactamases.

Native bacterial species are used for subsequent experiments, and the authors provide important context for their antibiotic choices and concentrations by referencing the breakpoints that guide clinical practice. In Figure 4, the authors show that loss of the DsbA system in P. aeruginosa significantly sensitizes clinical isolates expressing different classes of β-lactamases to clinically relevant antibiotics. The appropriate control showing that the dsbA1 mutation does not result in sensitivity to a non-β-lactam antibiotic is included in Figure S2. The authors further show, using an in vivo model for antibiotic treatment, that treatment of a dsbA1 mutant results in moderate and near-complete survival of the infected organisms. The importance of this system in S. maltophilia is then investigated similarly (Figure 5), showing that a dsbA dsbL mutant is also sensitive to β-lactams and colistin, another antibiotic whose resistance mechanism is dependent on the DsbA protein folding system. Importantly, the authors show that a small-molecule inhibitor that disrupts the DsbA system, rather than genetic mutations, is also capable of sensitizing S. maltophilia to these antibiotics. It should be noted that while the sensitization is less pronounced, this molecule has not been optimized for S. maltophilia and would be expected to increase in efficacy following optimization. Together, the data support that interference with the DsbA system in native hosts can sensitize otherwise resistant pathogens to clinically relevant antibiotic therapy.

Finally, the authors investigate the effects of co-culturing S. maltophilia and P. aeruginosa (Figure 5E). These assays are performed in synthetic cystic fibrosis sputum medium (SCFM), which provides a nutritional context similar to that in CF but without the presence of more complex components such as mucin. The authors show that while P. aeruginosa alone is sensitive to the antibiotic, it can survive moderate concentrations in the presence of S. maltophilia and even grow in higher concentrations where S. maltophilia appears to overproduce its β-lactamases. However, this protection is lost in S. maltophilia without the DsbA protein folding system, showing that the protective effect depends on functional production of β-lactamase in the presence of viable S. maltophilia. The authors further achieved the difficult task of labeling these multi-drug resistant pathogens with selection markers to determine co-infection CFUs in the supplemental materials. Overall, the data support a protective role for DsbA-dependent β-lactamase under these co-culture conditions.

Weaknesses:

No significant weaknesses are noted beyond the limitations identified and discussed by the authors.

Author response:

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

Reviewer #1 (Recommendation For the Authors):

Thanks to the authors for addressing my suggestions. I think these modifications have improved the clarity of the data and the overall presentation of the manuscript. The methods are now more clearly explained, and the additional details help make the results easier to interpret. Where addressing the comment wasn't feasible, the authors gave reasonable explanations. Overall, the revisions strengthen the paper, and I have no further concerns.

Thank you for your recommendations, which have significantly improved our paper.

Reviewer #2 (Recommendation For the Authors):

The additional work conducted by the authors is greatly appreciated. All concerns (and beyond) have been thoroughly addressed by the authors and I am thankful for their consideration and attention to detail. Only one possible issue with the revisions is described below for consideration:

Regarding the CFU counts and/or axis labels in Figure S3B, some of the listed "CFU per 1 mL" values (in both the figure itself and File S2B) are extraordinarily high. For example, the greatest CFU for PA14 observed in Figure 4E is ~1x10^9. However, PA14 at 0 ug/mL Ceftazidime reaches nearly 1x10^16 in Figure S3B. From what I can tell, this should be beyond the capacity of bacteria in this space by several orders of magnitude. (E.g., a cubic centimeter [~1 mL] is ~1x10^12 cubic micrometers. At their smallest dimensions and volume, a maximum of ~1x10^13 cells could theoretically fit in this space assuming no liquid and perfect organization.) Similarly, both "AMM" and "AMM (+PA14)" consistently reach CFUs between 1x10^12 and 1x10^14 in this assay. Are the authors confident in the values and/or depiction of CFUs for this figure? It seems like this could be a labeling or dilutioncounting issue.

Thank you for your positive remarks on our revised manuscript and for your constructive comments that have strengthened our work.

We agree with the concern regarding the CFU counts in Figure S3B. The very high values (>10¹²CFU) reflect a technical enumeration artifact that, due to the nature of the assay, cannot be fully avoided. The origin of these inflated counts is described in more detail below:

Following competition assays between Pseudomonas aeruginosa and Stenotrophomonas maltophilia in liquid culture with antibiotics, we enumerate survivors for each species by colony forming unit (CFU) counts. Because two different bacterial species must be quantified from mixed cultures, we use a gentamicin resistance marker carried by one species at a time.

Each condition is therefore enumerated twice, as we alternate which species harbors the gentamicin cassette.

During coculture in antibiotics and minimal medium, clinical isolates of P. aeruginosa and S. maltophilia, like those used here, can transiently increase their tolerance to antibiotics, including aminoglycosides. This reduces the effectiveness of gentamicin selection at the plating step necessary for CFU enumeration. For the data presented in Figure S3B, in a subset of highOD₆₀₀ conditions in the competition assay, this tolerance produces artificially inflated CFU values that exceed the biological carrying capacity during the CFU enumeration step.

We evaluated alternative enumeration strategies (e.g., fluorescent protein markers with a nonselective medium), but these proved unsuitable for these strains due to differences in growth rates and media compatibility, introducing other large biases. Given these constraints, selective plating remains the only feasible approach for this work, and the associated artifact cannot be eliminated entirely.

Importantly, transient resistance (tolerance), although common, is not a universal occurrence (e.g., we did not observe it when we performed the experiments shown in Figure 4E). When it does arise, it occurs reproducibly under the same experimental high-OD₆₀₀ conditions and does not obscure any of the relative comparisons that underpin our conclusions.

For transparency, we have retained the measured values in Figure S3B and we note in the legend that counts above ~10¹² CFU represent a technical overestimation due to transient gentamicin tolerance. Counts below 10¹² CFU are accurately enumerated.

Reviewer #3 (Recommendation For the Authors):

All concerns have been satisfied and the manuscript is ready for publishing.

Thank you for your recommendations, which have significantly improved our paper.

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

Reviewer #1 (Public review):

The study would benefit from presenting raw data in some cases, such as MIC values and SDS-PAGE gels, by clarifying the number of independent experiments used, as well as further clarification on statistical significance for some of the data.

All original data used to generate Fig. 1, Fig. 4E, Fig. S3 and Fig. S4A are presented in File S2. Tab (A) is dedicated to data used for Fig. 1 and Fig. S4A, while tabs (B) and (C) show the data used for Fig. 4E and S3, respectively. This information is indicated in the legends of the relevant figures.

All experiments in this study were performed in three independent (biological) experiments (with the exception of the complementation data shown in Fig. S1 and Fig. S5, which were performed in two independent (biological) experiments). The number of biological and technical replicates for each experiment is stated in the figure legends, as well as in the “Statistical analysis of experimental data” part of the “Materials and Methods” section of the paper. Specifically, for antibiotic MIC assays we have not performed statistical analyses as per recommended practice. The reason for this is stated in the following section from the “Statistical analysis of experimental data” part of the “Materials and Methods” section of the paper (lines 699-711 of the revised manuscript):

“Antibiotic MIC values were determined in biological triplicate, except for MIC values recorded for dsbA complementation experiments in our E. coli K-12 inducible system that were carried out in duplicate. All ETEST MICs were determined as a single technical replicate, and all BMD MICs were determined in technical triplicate. All recorded MIC values are displayed in the relevant graphs; for MIC assays where three or more biological experiments were performed, the bars indicate the median value, while for assays where two biological experiments were performed the bars indicate the most conservative of the two values (i.e., for increasing trends, the value representing the smallest increase and for decreasing trends, the value representing the smallest decrease). We note that in line with recommended practice, our MIC results were not averaged. This should be avoided because of the quantized nature of MIC assays, which only inform on bacterial survival for specific antibiotic concentrations and do not provide information for antibiotic concentrations that lie in-between the tested values.”

Reviewer #2 (Public review):

While Figure 5E demonstrates a protective effect of DsbA-dependent β-lactamase, the omission of CFU data for S. maltophilia makes it difficult to assess the applicability of the polymicrobial strategy. Since S. maltophilia is pre-cultured prior to the addition of P. aeruginosa and antibiotics, it is unclear whether the protective effect is dependent on high S. maltophilia CFU. It is also unclear what the fate of the S. maltophilia dsbA dsbL mutant is under these conditions. If DsbA-deficient S. maltophilia CFU is not impacted, then this treatment will result in the eradication of only one of the pathogens of interest. If the mutant is lost during treatment, then it is not clear whether the loss of protection is due specifically to the production of non-functional β-lactamase or simply the absence of S. maltophilia.

We have simultaneously tracked the abundance of P. aeruginosa and S. maltophilia strains in our cross-protection experiment for select antibiotic concentrations. To be able to perform this experiment, we had to label two extremely-drug-resistant strains of S. maltophilia with an antibiotic resistance marker that allowed us to quantify them in mixtures with P. aeruginosa. Our results can be found in Fig. S3 of our revised manuscript and, in a nutshell, show that ceftazidime treatment leads to eradication of both P. aeruginosa and S. maltophilia when disulfide bond formation is impaired in S. maltophilia.

The following text was added to address the questions of the reviewer:

“Due to the naturally different growth rates of these two species (S. maltophilia grows much slower than P. aeruginosa) especially in laboratory conditions, the protocol we followed [1] requires S. maltophilia to be grown for 6 hours prior to co-culturing it with P. aeruginosa. To ensure that at this point in the experiment our two S. maltophilia strains, with and without dsbA, had grown comparatively to each other, we determined their cell densities (Fig. S3A). We found that S. maltophilia AMM dsbA dsbL had grown at a similar level as the wild-type strain, and both were at a higher cell density [~10⁷ colony forming units (CFUs)] compared to the P. aeruginosa PA14 inoculum (5 x 10⁴ CFUs)” (lines 353-361 of the revised manuscript).

“To ensure that ceftazidime treatment leads to eradication of both P. aeruginosa and S. maltophilia when disulfide bond formation is impaired in S. maltophilia, we monitored the abundance of both strains in each synthetic community for select antibiotic concentrations (Fig. S3B). In this experiment we largely observed the same trends as in Fig. 4E. At low antibiotic concentrations, for example 4 μg/mL of ceftazidime, S. maltophilia AMM is fully resistant and thrives, thus outcompeting P. aeruginosa PA14 (dark pink and dark blue bars in Fig. S3B). The same can also be seen in Fig. 4E, whereby decreased P. aeruginosa PA14 CFUs are recorded. By contrast S. maltophilia AMM dsbA dsbL already displays decreased growth at 4 μg/mL of ceftazidime because of its non-functional L1-1 enzyme, allowing comparatively higher growth of P. aeruginosa (light pink and light blue bars in Fig. S3B). Despite the competition between the two strains, P. aeruginosa PA14 benefits from S. maltophilia AMM’s high hydrolytic activity against ceftazidime, which allows it to survive and grow in high antibiotic concentrations even though it is not resistant (see 128 μg/mL; dark pink and dark blue bars in Fig. S3B). In stark opposition, without its disulfide bond in S. maltophilia AMM dsbA dsbL, L1-1 cannot confer resistance to ceftazidime, resulting in killing of S. maltophilia AMM dsbA dsbL and, consequently, also of P. aeruginosa PA14 (see 128 μg/mL; light pink and light blue bars in Fig. S3B).

The data presented here show that, at least under laboratory conditions, targeting protein homeostasis pathways in specific recalcitrant pathogens has the potential to not only alter their own antibiotic resistance profiles (Fig. 3 and 4A-D), but also to influence the antibiotic susceptibility profiles of other bacteria that co-occur in the same conditions (Fig. 5). Admittedly, the conditions in a living host are too complex to draw direct conclusions from this experiment. That said, our results show promise for infections, where pathogen interactions affect treatment outcomes, and whereby their inhibition might facilitate treatment” (lines 381406 of the revised manuscript).

The alleged clinical relevance and immediate, theoretical application of this approach should be properly contextualized. At multiple junctures, the authors state or suggest that interactions between S. maltophilia and P. aeruginosa are known to occur in disease or have known clinical relevance related to treatment failure and disease states. For instance, the citations provided for S. maltophilia protection of P. aeruginosa in the CF lung environment both describe simplified laboratory experiments rather than clinical or in vivo observations. Similarly, the citations provided for both the role of S. maltophilia in treatment failure and CF disease severity do not support either claim. The role of S. maltophilia in CF is currently unsettled, with more recent work reporting conflicting results that support S. maltophilia as a marker, rather than cause, of severe disease. These citations also do not support the suggestion that S. maltophilia specifically contributes to treatment failure. While it is reasonable to pursue these ideas as a hypothesis or potential concern, there is no evidence provided that these specific interactions occur in vivo or that they have clinical relevance.

Thank you for your comment. You are entirely correct. We have amended the test throughout our revised manuscript to avoid overstating the role of S. maltophilia in CF infections and to reference additional relevant works in the literature. Please find below representative examples of such passages:

“On the other hand, CF microbiomes are increasingly found to encompass S. maltophilia [2-4], a globally distributed opportunistic pathogen that causes serious nosocomial respiratory and bloodstream infections [5-7]. S. maltophilia is one of the most prevalent emerging pathogens [6] and it is intrinsically resistant to almost all antibiotics, including β-lactams like penicillins, cephalosporins and carbapenems, as well as macrolides, fluoroquinolones, aminoglycosides, chloramphenicol, tetracyclines and colistin. As a result, the standard treatment option for lung infections, i.e., broad-spectrum β-lactam antibiotic therapy, is rarely successful in countering S. maltophilia [7,8], creating a definitive need for approaches that will be effective in eliminating both pathogens” (lines 33-41 of the revised manuscript).

“Of the organisms studied in this work, S. maltophilia deserves further discussion because of its unique intrinsic resistance profile. The prognosis of CF patients with S. maltophilia lung carriage is still debated [4,9-16], largely because studies with extensive and well-controlled patient cohorts are lacking. This notwithstanding, the therapeutic options against this pathogen are currently limited to one non-β-lactam antibiotic-adjuvant combination, , which is not always effective, trimethoprim-sulfamethoxazole [17-20], and a few last-line β-lactam drugs, like the fifth-generation cephalosporin cefiderocol and the combination aztreonam-avibactam. Resistance to commonly used antibiotics causes many problems during treatment and, as a result, infections that harbor S. maltophilia have high case fatality rates [7]. This is not limited to CF patients, as S. maltophilia is a major cause of death in children with bacteremia [5]” (lines 440-450 of the revised manuscript).

Reviewer #3 (Public review):

The impact of the work can be strengthened by demonstrating increased efficacy of antibiotics in mice models or wound models for Pseudomonas infections. Worm models are relevant, but still distant from investigations in animal models.

Thank you for this comment. We appreciate the sentiment, and we would have liked to be able to perform experiments in a murine model of infection. There are several reasons that made this not possible, and as a result we used G. mellonella as an informative preliminary in vivo infection model. The DSB proteins have been shown to play a central role in bacterial virulence. Because of this our P. aeruginosa and S. maltophilia mutant strains are not efficient in establishing an infection, even in a wound model. This could be overcome had we been able to use the chemical inhibitor of the DSB system in vivo, however this also is not possible This is due to the fact that the chemical compound that we use to inhibit the function of DsbA acts on DsbB. Inhibition of DsbB blocks the re-oxidation of DsbA and leads to its accumulation in its inactive reduced form. However, the action of the inhibitor can be bypassed through reoxidation and re-activation of DsbA by small-molecule oxidants such as L-cystine, which are abundant in rich growth media or animal tissues. This makes the inhibitor only suitable for in vitro assays that can be performed in minimal media, where the presence of small-molecule oxidants can be strictly avoided, but entirely unsuitable for an insect or a vertebrate animal model.

Reviewer #1 (Recommendation For the Authors):

(1) The analysis of the role of DsbA in the assembly of cysteine-containing β-lactamases is a significant finding. However, in addition to showing the MIC fold difference, I think, it would be important to show the raw data for the actual MIC values obtained for each β-lactamase enzyme/antibiotic combination and in both strains (+ and - dsbA).

Also, can the authors clarify whether these experiments were conducted on 3 independent samples (there seems to be some contradicting information in the paper and the supplementary figures). If possible, I would also recommend showing in the figure whether the MIC differences observed were statistically significant.

All original data used to generate Fig. 1, Fig. 4E, Fig. S3 and Fig. S4A are presented in File S2. Tab (A) is dedicated to data used for Fig. 1 and Fig. S4A, while tabs (B) and (C) show the data used for Fig. 4E and S3, respectively. This information is indicated in the legends of the relevant figures.

All experiments in this study were performed in three independent (biological) experiments (with the exception of the complementation data shown in Fig. S1 and Fig. S5, which were performed in two independent (biological) experiments). The number of biological and technical replicates for each experiment is stated in the figure legends, as well as in the “Statistical analysis of experimental data” part of the “Materials and Methods” section of the paper. Specifically, for antibiotic MIC assays we have not performed statistical analyses as per recommended practice. The reason for this is stated in the following section from the “Statistical analysis of experimental data” part of the “Materials and Methods” section of the paper (lines 699-711 of the revised manuscript):

“Antibiotic MIC values were determined in biological triplicate, except for MIC values recorded for dsbA complementation experiments in our E. coli K-12 inducible system that were carried out in duplicate. All ETEST MICs were determined as a single technical replicate, and all BMD MICs were determined in technical triplicate. All recorded MIC values are displayed in the relevant graphs; for MIC assays where three or more biological experiments were performed, the bars indicate the median value, while for assays where two biological experiments were performed the bars indicate the most conservative of the two values (i.e., for increasing trends, the value representing the smallest increase and for decreasing trends, the value representing the smallest decrease). We note that in line with recommended practice, our MIC results were not averaged. This should be avoided because of the quantized nature of MIC assays, which only inform on bacterial survival for specific antibiotic concentrations and do not provide information for antibiotic concentrations that lie in-between the tested values.”

(2) For Figure 2A, can the authors provide the full Westerns and ideally the SDS-PAGE gel corresponding to the Westerns where the Β-lactamases and the control DNA-K were detected.

Thank you for this comment. Full immunoblots and SDS PAGE analysis of the immunoblot samples for total protein content are shown in File S3 of our revised manuscript.

(3) For the enzymatic assays, was the concentration of enzyme used "normalised " based on the amount detected in the westerns where possible or was only the total amount of protein considered. When similar amounts of enzyme were added, was the activity still compromised?

The β-lactam hydrolysis assay was normalized based on the weight of the cell pellets (wet cell pellet mass) of the tested strains. This means, that for each enzyme expressed in cells with and without DsbA, strains were normalized to the same weight to volume ratio, and thus strains expressing the same enzyme were only compared to each other.

Because enzyme degradation in the absence of DsbA is a key factor underlying the effects we describe for most of the tested β-lactamases (see Fig. 2A and S4A; no protein band is detected for 5 of the 7 enzymes in the dsbA mutant), it was not possible to normalize our samples based on enzyme levels detected by immunoblot. Normalization based on enzyme amounts would be feasible had we purified each β-lactamase after expression in the two different strain backgrounds (+/- dsbA) assuming sufficient protein amounts could be isolated from the dsbA mutant strain. Nonetheless, we feel that such a comparison would be misleading, since enzyme degradation likely plays the biggest role in the lack of activity observed for most of these enzymes in the absence of DsbA.

(4) Not sure whether Fig 3 is very informative. Perhaps it could be redesigned to better encapsulate the findings in this manuscript (combine figurer 3 and 6 into one). I would also include the chemical structure of the inhibitors used and perhaps include how they block the system by binding to DsbB.

Thank you for this comment. Fig. 3 was combined with Fig. 6 of the submitted manuscript. The new model figure is Fig. 5 in our revised manuscript.

The inhibitor compound used in our study has been extensively characterized in a previous publication [21]. Considering that this inhibitor is not the main focus of our paper, we have avoided showing its chemical structure in any of the main display items. That said, its structure can be found in File S5 of our revised manuscript, which contains the quality control information on this compound. As suggested, we included the following sentence to describe the mode of action of this inhibitor: “Compound 36 was previously shown to inhibit disulfide bond formation in P. aeruginosa via covalently binding onto one of the four essential cysteine residues of DsbB in the DsbA-DsbB complex [21]” (lines 309-311 of the revised manuscript).

(5) Figure 4: Similar to my comment above showing in the figure whether the differences observed in Figure 4, particularly A-C, are statistically significant (i.e. galleria survival difference in the presence and absence of dsbA) would be beneficial.

As mentioned in our answer to comment 1 above, we have not performed statistical analyses for antibiotic MIC assays because, in line with recommended practice, our MIC results were not averaged (Fig. 3A,B,D,E of our revised manuscript). This should be avoided because of the quantized nature of MIC assays, which only inform on bacterial survival for specific antibiotic concentrations and do not provide information for antibiotic concentrations that lie in-between the tested values. Statistical analysis of G. mellonella survival data (Fig. 3C,F of our revised manuscript) was performed and is described fully in the legend of Fig. 3, as well as in the “Statistical analysis of experimental data” part of the “Materials and Methods” section of the paper (lines 729-738 of the revised manuscript). Finally, the statistical analyses for the most important comparisons in panels (C) and (F) of Fig. 3 are also marked directly on the figure.

(6) Were the authors able to test the redox state of DsbA upon addition of the DsbB inhibitor to further demonstrate that the effects observed were indeed due to the obstruction of the Dsb machinery and not due to off target effects.

Thank you for the opportunity to clarify this. In previous work from our lab, we have used a DSB system inhibitor termed “compound 12” in [22] with activity against DsbB proteins from Enterobacteria. In our previous study [23] we, indeed, tested the redox state of DsbA in the presence of this inhibitor compound. We could not perform the same experiment here with “compound 36” from [21], because we do not have an antibody against the DsbA protein of S. maltophilia. That said, we have carried out experiments that confirm that our results are due to specific inhibition of the DSB system and not because of off-target effects. In particular, we show that the gentamicin MIC values of S. maltophilia AMM remain unchanged in the presence of the inhibitor and treatment of S. maltophilia AMM dsbA dsbL with the compound does not affects its colistin MIC value (Fig. S2E and lines 317-320 of the revised manuscript).

(7) Given the remarkable effects shown by the DsbB inhibitor, did the authors use this compound to assess whether inhibition of the Dsb system with small molecules would block cross-resistance in S. maltophilia - P. aeruginosa mixed communities (Fig 5D).

Unfortunately, this was not possible. The decrease in the ceftazidime MIC value of S. maltophilia AMM in the presence of the DSB inhibitor compound is more modest than the effects we observed when the dsbA dsbL mutant is used (compare Fig. 4D (left) with Fig.4A of the revised manuscript). This means that in the presence of the DSB inhibitor there are still sufficient amounts of functional β-lactamase present and we expect that they would contribute to cross-protection of P. aeruginosa. While the use of the DSB inhibitor does have a drastic impact on the colistin resistance profile of S. maltophilia AMM (Fig. 4D of the revised manuscript), unlike β-lactamases, which act as common goods, MCR enzymes act solely on the lipopolysaccharide of their producer and do not contribute to bacterial interactions, precluding the use of colistin for a cross-protection experiment.

Reviewer #2 (Recommendation For the Authors):

(1) The acronym used for synthetic cystic fibrosis sputum medium (lines 523, 531, 535, 601, and 603) is defined in the manuscript as 'SCF', but the common formulation is 'SCFM', including in the provided citation. Suggest changing to SCFM for consistency.

Thank you for this comment. This has been amended throughout our revised manuscript.

(2) In Figure 1, while the legend states that "No changes in MIC values are observed for strains harboring the empty vector control (pDM1)[...]" (lines 729-30), the median of ceftazidime in the pDM1 control appears to indicate a 2-fold decrease in MIC. This would not seem to significantly impact the other results since the MIC decreases observed for other conditions are all 3-fold or greater, but this should be addressed and/or explained in the text.

You are correct. Thank you for the opportunity to clarify this. Generally, since MIC assays have a degree of variability, we have only followed decreases in MIC values that are greater than 2fold. Generally, for most of our controls, the recorded MIC fold changes are below 2-fold. The only exception to this is the ceftazidime MIC drop of the empty-vector control, showing a 2fold change, which we do not consider significant.

To ensure that this is clear in our text and figure legends the following changes were made:

The clause “only differences larger than 2-fold were considered” was added to the text (lines 110-111 of the revised manuscript).

We amended the legend of Fig. 1 accordingly: “No changes in MIC values are observed for the aminoglycoside antibiotic gentamicin (white bars) confirming that absence of DsbA does not compromise the general ability of this strain to resist antibiotic stress. Minor changes in MIC values (≤ 2-fold) are observed for strains harboring the empty vector control (pDM1) or those expressing the class A β-lactamases L2-1 and LUT-1, which contain two or more cysteines (Table S1), but no disulfide bonds (top row)”.

(3) Similarly, in Fig S1E, there appears to be only partial complementation for BPS-1m. Do the authors hypothesize that this observation is related to a folding defect, rather than degradation of protein, as described for BPS-1m for Figure 2?

Thank you for the opportunity to clarify this. You are correct that we only achieve partial complementation for the E. coli strain expressing the BPS-1m enzyme from the Burkholderia complex. Despite the fact that the gene for this enzyme was codon optimized, we observed that its expression in E. coli is sub-optimal and incurs fitness effects. In fact, to record the data presented in our manuscript the E. coli strains had to be transformed anew every time. Considering that the related enzyme BPS-6 does not present any of these challenges, we attribute the partial complementation to technical difficulties with the expression of the bps-1m gene in E. coli. 

We clarified this by adding the following clause to our manuscript: “we only achieve partial complementation for the dsbA mutant expressing BPS-1m, which we attribute to the fact that expression of this enzyme in E. coli is sub-optimal” (lines 132-134 of the revised manuscript).

(4) Lines 204-206: "[...]we deleted the principal dsbA gene, dsbA1 (pathogenic bacteria often encode multiple DsbA analogues [24,25]), in several multidrug-resistant (MDR) P. aeruginosa clinical strains (Table S2)". That multiple DsbA analogues are often encoded is good information to provide, but it was unclear from quickly looking at the citations whether Pa is counted among these. Is it expected that all oxidative protein folding in Pa functions through DsbA1? Conveying this information, if possible, may make the impact of the results in this model clearer.

Thank you for this comment. To address it we added the following text to our manuscript:

“To determine whether the effects on β-lactam MICs observed in our inducible system (Fig. 1 and [23]) can be reproduced in the presence of other resistance determinants in a natural context with endogenous enzyme expression levels, we deleted the principal dsbA gene, dsbA1, in several multidrug-resistant (MDR) P. aeruginosa clinical strains (Table S2). Pathogenic bacteria often encode multiple DsbA analogues [24,25] and P. aeruginosa is no exception. It encodes two DsbAs, but DsbA1 has been found to catalyze the vast majority of the oxidative protein folding reactions taking place in its cell envelope [26]” (lines 172-178 of the revised manuscript).

(5) Regarding the clinical Pa isolates G4R7 and G6R7, have the authors performed any phenotypic testing on these strains to identify differences that might explain the substantial difference in piperacillin MIC? I.e., can these isolates be distinguished by growth rate, genetic markers or expression levels, early or late infection, mucoidy, etc. This is not essential for the current work, but could weigh on the efficacy of this treatment strategy for AIM1expressing clinical isolates. (E.g., the G4R7 dsbA1 strain exhibits a piperacillin MIC still ~2fold higher than WT G6R7).

Thank you for the opportunity to clarify this. For clinical strains used in our study, we have evaluated their antibiotic resistance profiles, but we have not performed any additional phenotypic characterization. There are many reasons that contribute to differences in antibiotic resistance, starting simply from β-lactamase expression levels and extending to organismal effects, like the ones mentioned by the reviewer. Such characterization would fall outside the scope of our paper, especially since we sensitize our tested P. aeruginosa clinical isolates for the majority of the β-lactams antibiotics tested. 

We acknowledged this by adding the following sentence to our revised manuscript: 

“Despite the fact that P. aeruginosa G4R7 dsbA1 was not sensitized for piperacillintazobactam, possibly due to the high level of piperacillin-tazobactam resistance of the parent clinical strain, our results across these two isolates show promise for DsbA as a target against β-lactam resistance in P. aeruginosa” (lines 191-194 of the revised manuscript).

(6) Lines 180-2: "This shows that without their disulfide bonds, these proteins are unstable and are ultimately degraded by other cell envelope proteostasis components [33]". While it is clear that protein is significantly lost in all cases except for BPS-1m in 2A, the dsbA pDM1bla constructs in 2B appear to all retain non-trivial (>10-fold) nitrocefin hydrolysis activity compared to the dsbA pDM1 control. This does not impact the other results in 2B, but it would seem that a loss-of-function folding defect, as described subsequently for BPS-1m, is also part of the explanation for the observed MIC decreases, and this was not necessarily clear from the quoted passage. This could simply be clarified in the final sentence - that both mechanisms are potentially in play - if the authors agree with that interpretation.

You are correct, thank you for your comment. We amended the text in our revised manuscript as follows: 

The data presented so far (Fig. 1 and 2) demonstrate that disulfide bond formation is essential for the biogenesis (stability and/or protein folding) and, in turn, activity of an expanded set of clinically important β-lactamases, including enzymes that currently lack inhibitor options” (lines 158-161 of the revised manuscript).

(7) While it is clear from Figure S2 that the various dsb mutants do not have a general growth defect or collateral sensitivity to another antibiotic, it does not appear that there is an analogous control for the DSB inhibitor demonstrating no growth/toxic effects at the concentration used. This could be provided similarly to Figure S2, using gentamicin as a control antibiotic.

We have carried out experiments that confirm that our results are due to specific inhibition of the DSB system and not because of off-target effects. In particular, we show that the gentamicin MIC values of S. maltophilia AMM remain unchanged in the presence of the inhibitor and treatment of S. maltophilia AMM dsbA dsbL with the compound does not affects its colistin MIC value (Fig. S2E and lines 317-320 of the revised manuscript).

(8) Complementation is appropriately provided for experiments with E. coli, but are not provided for P. aeruginosa or S. maltophilia. It should be straightforward to complement in Pa, but is also probably less critical considering the evidence from E. coli. However, since the Sm mutant is a gene cluster with two genes, it would seem more imperative to complement this strain. This reviewer is not familiar enough with Sm to know if complementation is routine or feasible with this organism; if not, the controls for the DSB inhibitor should at least be provided.

As mentioned in our response to comment 7 above, we have carried out experiments that confirm that our DSB inhibitor results are due to specific inhibition of the DSB system and not because of off-target effects.

Moreover, in response to this comment, we have further demonstrated that our results are due to the specific interaction of DsbA with β-lactamase enzymes by complementing dsbA deletions in representative clinical strains of multidrug-resistant Pseudomonas aeruginosa and extremely-drug-resistant Stenotrophomonas maltophilia. We would like to note here that gene complementation in clinical isolates remains very rare in the literature due to their high levels of resistance and limited genetic tractability. Most of the few complementation examples reported for these two organisms are limited to strains that, although pathogenic, are commonly used in the lab, or to complementation efforts in non-clinical strain systems (for example use of P. aeruginosa PA14 for complementation, instead of the focal clinical isolate).

We tested three different complementation strategies, two of which ended up being unsuccessful. After approximately 9 months of work, we succeeded in complementing a representative clinical strain for each organism (P. aeruginosa CDC #769 dsbA1 and S. maltophilia AMM dsbA dsbL) by inserting the dsbA1 gene from P. aeruginosa PAO1 into the Tn7 site on the chromosome. Both clinical strains show full complementation for every antibiotic tested; our complementation results can be found in Fig. S2B,D of the revised manuscript.

The following text was added for P. aeruginosa clinical isolates:

We have demonstrated the specific interaction of DsbA with the tested β-lactamase enzymes in our E. coli K-12 inducible system using gentamicin controls (Fig. 1 and File S2A) and gene complementation (Fig. S1). To confirm the specificity of this interaction in P. aeruginosa, we performed representative control experiments in one of our clinical strains, P. aeruginosa CDC #769. We first tested the general ability of P. aeruginosa CDC #769 dsbA1 to resist antibiotic stress by recording MIC values against gentamicin, and found it unchanged compared to its parent (Fig. S2A). Gene complementation in clinical isolates is especially challenging and rarely attempted due to the high levels of resistance and lack of genetic tractability in these strains. Despite these challenges, to further ensure the specificity of the interaction of DsbA with tested β-lactamases in P. aeruginosa, we have complemented dsbA1 from P. aeruginosa PAO1 into P. aeruginosa CDC #769 dsbA1. We found that complementation of dsbA1 restores MICs to wild-type values for both tested β-lactam compounds (Fig. S2B) further demonstrating that our results in P. aeruginosa clinical strains are not confounded by off-target effects” (lines 226-239 of the revised manuscript).

The following text was added for S. maltophilia clinical isolates: 

“Since the dsbA and dsbL are organized in a gene cluster in S. maltophilia, we wanted to ensure that our results reported above were exclusively due to disruption of disulfide bond formation in this organism. First, we recorded gentamicin MIC values for S. maltophilia AMM dsbA dsbL and found them to be unchanged compared to the gentamicin MICs of the parent strain (Fig. S2C). This confirms that disruption of disulfide bond formation does not compromise the general ability of this organism to resist antibiotic stress. Next, we complemented S. maltophilia AMM dsbA dsbL. The specific oxidative roles and exact regulation of DsbA and DsbL in S. maltophilia remain unknown. For this reason and considering that genetic manipulation of extremely-drug-resistant organisms is challenging, we used our genetic construct optimized for complementing P. aeruginosa CDC #769 dsbA1 with dsbA1 from P. aeruginosa PAO1 (Fig. S2B) to also complement S. maltophilia AMM dsbA dsbL. We based this approach on the fact that DsbA proteins from one species have been commonly shown to be functional in other species [27-30]. Indeed, we found that complementation of S. maltophilia AMM dsbA dsbL with P. aeruginosa PAO1 dsbA1 restores MICs to wild-type values for both ceftazidime and colistin (Fig. S2D), conclusively demonstrating that our results in S. maltophilia are not confounded by off-target effects” (lines 282-297 of the revised manuscript).

(9) In Figure 5E, the growth inhibition and loss of Pa CFU in 4 ug/mL ceftazidime for the Sm co-culture condition, which is subsequently lost in the Sm dsbA dsbL co-culture, does not appear to be discussed. As Pa is shown to grow fine in monoculture at this concentration, this result should be discussed in relation to the co-culture dynamics. Is it expected or observed that WT Sm is out-competing Pa under this condition and growing to a high CFU/mL? This would seem to have parallels to citation 49.

As requested by this reviewer (see comment 10 below), we simultaneously tracked the abundance of P. aeruginosa and S. maltophilia strains in our cross-protection experiment. During this process we probed the abundances of the two organisms at 4 µg/mL of ceftazidime. Our results can be seen in Fig. S3B of the revised manuscript. The reviewer is correct and these effects are due to competition between P. aeruginosa and S. maltophilia with the latter being able to reach very high CFUs in this antibiotic concentration. 

The following text on co-culture dynamics was added to our revised manuscript: 

At low antibiotic concentrations, for example 4 μg/mL of ceftazidime, S. maltophilia AMM is fully resistant and thrives, thus outcompeting P. aeruginosa PA14 (dark pink and dark blue bars in Fig. S3B). The same can also be seen in Fig. 4E, whereby decreased P. aeruginosa PA14 CFUs are recorded. By contrast S. maltophilia AMM dsbA dsbL already displays decreased growth at 4 μg/mL of ceftazidime because of its non-functional L1-1 enzyme, allowing comparatively higher growth of P. aeruginosa (light pink and light blue bars in Fig. S3B)” (lines 384-390 of the revised manuscript).

(10) The data presented in Figure 5E would be augmented by the inclusion of, for at least a few representative cases, the Sm CFUs relative to the Pa CFUs. In describing the protective effects of Sm on Pa for imipenem treatment, the authors of citation 12 note that the effect was dependent on Sm cell density. This raises the immediate question of whether the protection observed in this work is similarly dependent on cell density of Sm. It is unclear if the authors expect Sm to persist under these conditions, and it seems Sm CFU should be expected to be relatively high considering it is pre-incubated for 6 hours prior to the assay. What is the physiological state of these cells, and how are they affected by ceftazidime? While many other variables are likely relevant to the translation of this protection, the relative abundance and localization of Sm and Pa commonly observed in CF patients, as well as the effective concentration of antibiotic observed in vivo, is likely worth consideration.

As mentioned in our response to comment 9 above, we have simultaneously tracked the abundance of P. aeruginosa and S. maltophilia strains in our cross-protection experiment for select antibiotic concentrations. To be able to perform this experiment, we had to label two extremely-drug-resistant strains of S. maltophilia with an antibiotic resistance marker that allowed us to quantify them in mixtures with P. aeruginosa. Our results can be found in Fig. S3 of our revised manuscript and, in a nutshell, show that ceftazidime treatment leads to eradication of both P. aeruginosa and S. maltophilia when disulfide bond formation is impaired in S. maltophilia.

The following text was added to address the questions of the reviewer:

“Due to the naturally different growth rates of these two species (S. maltophilia grows much slower than P. aeruginosa) especially in laboratory conditions, the protocol we followed [1] requires S. maltophilia to be grown for 6 hours prior to co-culturing it with P. aeruginosa. To ensure that at this point in the experiment our two S. maltophilia strains, with and without dsbA, had grown comparatively to each other, we determined their cell densities (Fig. S3A). We found that S. maltophilia AMM dsbA dsbL had grown at a similar level as the wild-type strain, and both were at a higher cell density [~10⁷ colony forming units (CFUs)] compared to the P.aeruginosa PA14 inoculum (5 x 10⁴ CFUs)” (lines 353-361 of the revised manuscript).

“To ensure that ceftazidime treatment leads to eradication of both P. aeruginosa and S. maltophilia when disulfide bond formation is impaired in S. maltophilia, we monitored the abundance of both strains in each synthetic community for select antibiotic concentrations (Fig. S3B). In this experiment we largely observed the same trends as in Fig. 4E. At low antibiotic concentrations, for example 4 μg/mL of ceftazidime, S. maltophilia AMM is fully resistant and thrives, thus outcompeting P. aeruginosa PA14 (dark pink and dark blue bars in Fig. S3B). The same can also be seen in Fig. 4E, whereby decreased P. aeruginosa PA14 CFUs are recorded. By contrast S. maltophilia AMM dsbA dsbL already displays decreased growth at 4 μg/mL of ceftazidime because of its non-functional L1-1 enzyme, allowing comparatively higher growth of P. aeruginosa (light pink and light blue bars in Fig. S3B). Despite the competition between the two strains, P. aeruginosa PA14 benefits from S. maltophilia AMM’s high hydrolytic activity against ceftazidime, which allows it to survive and grow in high antibiotic concentrations even though it is not resistant (see 128 μg/mL; dark pink and dark blue bars in Fig. S3B). In stark opposition, without its disulfide bond in S. maltophilia AMM dsbA dsbL, L1-1 cannot confer resistance to ceftazidime, resulting in killing of S. maltophilia AMM dsbA dsbL and, consequently, also of P. aeruginosa PA14 (see 128 μg/mL; light pink and light blue bars in Fig. S3B).

The data presented here show that, at least under laboratory conditions, targeting protein homeostasis pathways in specific recalcitrant pathogens has the potential to not only alter their own antibiotic resistance profiles (Fig. 3 and 4A-D), but also to influence the antibiotic susceptibility profiles of other bacteria that co-occur in the same conditions (Fig. 5). Admittedly, the conditions in a living host are too complex to draw direct conclusions from this experiment. That said, our results show promise for infections, where pathogen interactions affect treatment outcomes, and whereby their inhibition might facilitate treatment” (lines 381406 of the revised manuscript).

(11) Regarding the role of microbial interactions in CF and other disease/infection contexts, the authors should temper their descriptions in accordance with citations provided. As an example, lines 96-99: "For example, in the CF lung, highly drug-resistant S. maltophilia strains actively protect susceptible P. aeruginosa from β-lactam antibiotics [12], and ultimately facilitate the evolution of β-lactam resistance in P. aeruginosa [14]."

Neither citation provided here attests to Sm protection of Pa "in the CF lung". Both papers use a simplified in vitro co-culture model to assess Sm protection of Pa from antibiotics and the evolution of Pa antibiotic resistance in the presence or absence of Sm, respectively. In the latter case, it should also be noted that while the authors observed somewhat faster Pa resistance evolution in one co-culture condition, they did not observe it in the other, and that resistance evolution in general was observed regardless of co-culture condition. There are also statements in the ultimate and penultimate paragraphs of the Discussion section that repeat these points. The authors could re-frame this aspect of their investigation as part of a working hypothesis related to potential interactions of these pathogens, and should appropriately caveat what is and is not known from in vitro and in vivo/clinical work.

Thank you for your comment. You are entirely correct. We have amended the test throughout our revised manuscript to avoid overstating these finding and to be clear about the fact that they originate from experimental studies. Please find below representative examples of such passages:

“In particular, some antibiotic resistance proteins, like β-lactamases, which decrease the quantities of active drug present, function akin to common goods, since their benefits are not limited to the pathogen that produces them but can be shared with the rest of the bacterial community. This means that their activity enables pathogen cross-resistance when multiple species are present [1,31], something that was demonstrated in recent work investigating the interactions between pathogens that naturally co-exist in CF infections. More specifically, it was shown that in laboratory co-culture conditions, highly drug-resistant S. maltophilia strains actively protect susceptible P. aeruginosa from β-lactam antibiotics [1]. Moreover, this crossprotection was found to facilitate, at least under specific conditions, the evolution of β-lactam resistance in P. aeruginosa [32]” (lines 47-57 of the revised manuscript).

“The antibiotic resistance mechanisms of S. maltophilia impact the antibiotic tolerance profiles of other organisms that are found in the same infection environment. S. maltophilia hydrolyses all β-lactam drugs through the action of its L1 and L2 β-lactamases [7,8]. In doing so, it has been experimentally shown to protect other pathogens that are, in principle, susceptible to treatment, such as P. aeruginosa [1]. This protection, in turn, allows active growth of otherwise treatable P. aeruginosa in the presence of complex β-lactams, like imipenem [1], and, at least in some conditions, increases the rate of resistance evolution of P. aeruginosa against these antibiotics [32]” (lines 332-340 of the revised manuscript).

(12) Regarding the role of S. maltophilia in CF disease, the authors should either discuss clinical associations more completely or note the conflicting data on its role in disease. As an example, lines 84-87: "As a result, the standard treatment option, i.e., broad-spectrum βlactam antibiotic therapy, constitutes a severe risk for CF patients carrying both P. aeruginosa and S. maltophilia [10,11], creating an urgent need for antimicrobial approaches that will be effective in eliminating both pathogens."

It is unclear how this treatment results in a "severe risk" for CF patients colonized by both Sm and Pa. Citation 10 suggests an association between anti-pseudomonal antibiotic use and increased prevalence of Sm, but neither citation supports a worsening clinical outcome from this treatment. Citation 10 further notes that clinical scores between Sm-positive and control cohorts could not be distinguished statistically. Citation 11 is a review that makes note of this conflicting data regarding Sm, including reference to a more recent (at the time) result using multivariate analysis showing no independent affect of Sm on survival.

The above point similarly applies to other statements in the manuscript, for example at lines 266-267: "Considering the contribution of S. maltophilia strains to treatment failure in CF lung infections [8,10,11][...]" As well as lines 79-80: "Pulmonary exacerbations and severe disease states are also associated with the presence of S. maltophilia [8]"

Again, the provided citations do not support the implication that Sm specifically 'contributes to treatment failure in CF lung infections' or that Sm is specifically associated with severe disease states. In addition to the previously discussed citations, citation 8 describes broad "pulmotypes" composed of 10 species/genera that could be associated with particular clinical (e.g., exacerbation) or treatment (e.g., antibiotic therapy) characteristics, but these cannot, without further analysis, be associated with, or causally linked to, a specific pathogen. While pulmotype 2 in citation 8 was associated with a more severe clinical state and appeared to have the highest relative abundance of Sm compared to other pulmotypes, Sm was not identified (Figure 4A) as an independent factor that distinguishes between moderate and severe disease, unlike Pa and some anaerobes (4F-H). The authors also observed that decreasing relative abundance of Pa, in particuar, is correlated with subsequent exacerbation, but did not correlate this with the presence of any other species or genera. Again, this should be re-framed with the appropriate caveat that this is a hypothesis with possible clinical significance.

Several suggested papers are included below on Sm association with clinical characteristics to incorporate into the manuscript if the authors choose to do so:

https://doi.org/10.1177/14782715221088909

https://doi.org/10.1016/j.prrv.2010.07.003

https://doi.org/10.1016/j.jcf.2013.05.009 https://doi.org/10.1002/ppul.23943

https://doi.org/10.1002/14651858.CD005405.pub2

https://doi.org/10.1164/rccm.2109078 http://dx.doi.org/10.1136/thx.2003.017707

https://erj.ersjournals.com/content/23/1/98.short

Thank you for your comment. You are entirely correct. We have amended the test throughout our revised manuscript to avoid overstating the role of S. maltophilia in CF infections and to reference additional relevant works in the literature. Please find below representative examples of such passages:

“On the other hand, CF microbiomes are increasingly found to encompass S. maltophilia [2-4], a globally distributed opportunistic pathogen that causes serious nosocomial respiratory and bloodstream infections [5-7]. S. maltophilia is one of the most prevalent emerging pathogens [6] and it is intrinsically resistant to almost all antibiotics, including β-lactams like penicillins, cephalosporins and carbapenems, as well as macrolides, fluoroquinolones, aminoglycosides, chloramphenicol, tetracyclines and colistin. As a result, the standard treatment option for lung infections, i.e., broad-spectrum β-lactam antibiotic therapy, is rarely successful in countering S. maltophilia [7,8], creating a definitive need for approaches that will be effective in eliminating both pathogens” (lines 33-41 of the revised manuscript).

“Of the organisms studied in this work, S. maltophilia deserves further discussion because of its unique intrinsic resistance profile. The prognosis of CF patients with S. maltophilia lung carriage is still debated [4,9-16], largely because studies with extensive and well-controlled patient cohorts are lacking. This notwithstanding, the therapeutic options against this pathogen are currently limited to one non-β-lactam antibiotic-adjuvant combination, , which is not always effective, trimethoprim-sulfamethoxazole [17-20], and a few last-line β-lactam drugs, like the fifth-generation cephalosporin cefiderocol and the combination aztreonam-avibactam. Resistance to commonly used antibiotics causes many problems during treatment and, as a result, infections that harbor S. maltophilia have high case fatality rates [7]. This is not limited to CF patients, as S. maltophilia is a major cause of death in children with bacteremia [5]” (lines 440-450 of the revised manuscript).

Reviewer #3 (Recommendation For the Authors):

(1) The referencing of supplemental figures does not follow a sequential order. For example, Figure S2 appears in the text before S1. The sequential ordering of figure numbers improves the readability and can be considered while editing the manuscript for revision.

Thank you for this comment. This is amended in our revised manuscript and supplemental figures and files are cited in order.

(2 )It will be useful to provide a brief description of ambler classes since these are important to study design (for a broader audience).

Thank you for this suggestion. This has been added and can be found in lines 91-101 of the revised manuscript.

(3) The rationale for using K12 strain for E. coli should be provided. It appears that is a model system that is well established in their lab, but a scientific rationale can be listed. Maybe this strain does not have any lactamases in its genome other than the one being expressed as compared to pathogenic E. coli?

Thank you for this suggestion. This has been added and can be found in lines 104-106 of the revised manuscript.

(4) The reviewers used worm model to test their observations, which is relevant. Given the significant implications of their work in overcoming resistance to clinically used antibiotics and availability of already generated dsbA mutants in clinical strains, it will be useful to investigate survival in animal models or at least wound models of Pseudomonas infections. The reviewer does not deem this necessary, but it will significantly increase the impact of their seminal work.

Thank you for this comment. We appreciate the sentiment, and we would have liked to be able to perform experiments in a murine model of infection. There are several reasons that made this not possible, and as a result we used G. mellonella as an informative preliminary in vivo infection model. The DSB proteins have been shown to play a central role in bacterial virulence. Because of this our P. aeruginosa and S. maltophilia mutant strains are not efficient in establishing an infection, even in a wound model. This could be overcome had we been able to use the chemical inhibitor of the DSB system in vivo, however this also is not possible This is due to the fact that the chemical compound that we use to inhibit the function of DsbA acts on DsbB. Inhibition of DsbB blocks the re-oxidation of DsbA and leads to its accumulation in its inactive reduced form. However, the action of the inhibitor can be bypassed through reoxidation and re-activation of DsbA by small-molecule oxidants such as L-cystine, which are abundant in rich growth media or animal tissues. This makes the inhibitor only suitable for in vitro assays that can be performed in minimal media, where the presence of small-molecule oxidants can be strictly avoided, but entirely unsuitable for an insect or a vertebrate animal model.

lighthouse

a map of this new territory

unique wise choices together

inserted my page in the middle of their site

new page comes up

walls become joint property

link open doors instead of putting up walls

but if the link is into a walled garden?

Reviewer #1 (Public review):

Dixit, Noe, and Weikl apply coarse-grained and all-atom molecular dynamics to determine the response of the mechanosensitive proteins Piezo 1 and Piezo 2 proteins to tension. Cryo-EM structures in micelles show a high curvature of the protein whereas structures in lipid bilayers show lower curvature. Is the zero-stress state of the protein closer to the micelle structure or the bilayer structure? Moreover, while the tension sensitivity of channel function can be inferred from experiment, molecular details are not clearly available. How much does the protein's height and effective area change in response to tension? With these in hand, a quantitative model of its function follows that can be related to the properties of the membrane and the effect of external forces.

Simulations indicate that in a bilayer the protein relaxes from the highly curved cryo-EM dome (Figure 1).

Under applied tension the dome flattens (Figure 2) including the underlying lipid bilayer. The shape of the system is a combination of the membrane mechanical and protein conformational energies (Eq. 1). The membrane mechanical energy is well-characterized. It requires only the curvature and bending modulus as inputs. They determine membrane curvature and the local area metric (Eq. 4) by averaging the height on a grid and computing second derivatives (Eqs. 7, 8) consistent with known differential geometric formulas.

While I am still critical generally of a precise estimate of the energy from simulated membrane shapes (after all it is not trivial to precisely determine even the bending modulus from a simulation), I believe with their revision the authors have convinced me that their estimate is a high quality one, without obvious issues. Although there appears to have been a miscommunication about increasing the density of grain or lowering the density of grain, the authors have tried two grains and determined a similar deformation energy, which addresses my concern. Furthermore, they have computed a dramatically reduced simplification of the curve and determined a similar value.

In summary, this paper uses molecular dynamics simulations to quantify the force of the Piezo 1 and Piezo 2 proteins on a lipid bilayer using simulations under controlled tension, observing the membrane deformation, and using that data to infer protein mechanics. While much of the physical mechanism was previously known, the study itself is a valuable quantification.

Author response:

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

Reviewer #1 (Public review): 

Dixit, Noe, and Weikl apply coarse-grained and all-atom molecular dynamics to determine the response of the mechanosensitive proteins Piezo 1 and Piezo 2 proteins to tension. Cryo-EM structures in micelles show a high curvature of the protein whereas structures in lipid bilayers show lower curvature. Is the zero-stress state of the protein closer to the micelle structure or the bilayer structure? Moreover, while the tension sensitivity of channel function can be inferred from the experiment, molecular details are not clearly available. How much does the protein's height and effective area change in response to tension? With these in hand, a quantitative model of its function follows that can be related to the properties of the membrane and the effect of external forces. 

Simulations indicate that in a bilayer the protein relaxes from the highly curved cryo-EM dome (Figure 1). 

Under applied tension, the dome flattens (Figure 2) including the underlying lipid bilayer. The shape of the system is a combination of the membrane mechanical and protein conformational energies (Equation 1). The membrane's mechanical energy is well-characterized. It requires only the curvature and bending modulus as inputs. They determine membrane curvature and the local area metric (Equation 4) by averaging the height on a grid and computing second derivatives (Equations 7, 8) consistent with known differential geometric formulas. 

The bending energy can be limited to the nano dome but this implies that the noise in the membrane energy is significant. Where there is noise outside the dome there is noise inside the dome. At the least, they could characterize the noisy energy due to inadequate averaging of membrane shape. 

My concern for this paper is that they are significantly overestimating the membrane deformation energy based on their numerical scheme, which in turn leads to a much stiffer model of the protein itself.

We agree that “thermal noise” is intrinsic to MD simulations, as in “real” systems, leading to thermally excited shape fluctuations of membranes and conformational fluctuations of proteins. However, for our coarse-grained simulations, the thermally excited membrane shape fluctuations can be averaged out quite well, and the resulting average shapes are smooth, see e.g. the shapes and lines of the contour plots in Fig. 1 and 2. For our atomistic simulations, the averaged shapes are not as smooth, see Fig. 3a and the lines of the contour plots in Fig. 3b. Therefore, we do not report bending energies for the nanodome shapes determined from atomistic simulations, because bending energy calculations are sensitive to remaining “noise” on small scales (due to the scale invariance of the bending energy), in contrast to calculations of excess areas, which we state now on lines 620ff.

For our coarse-grained simulations, we now corroborate our bending energy calculations based on averaged 3d shapes by comparing to bending energy values obtained from highly smoothened 2d mean curvature profiles (see Fig. 1c for mean curvature profiles in tensionless membranes). We discuss this in detail from line 323 on, starting with:

“To corroborate our bending energy calculations for these averaged three-dimensional nanodome shapes, we note that essentially identical bending energies can be obtained from the highly smoothened mean curvatures M of the two-dimensional membrane profiles. …”

Two things would address this: 

(1) Report the membrane energy under different graining schemes (e.g., report schemes up to double the discretization grain). 

There are two graining schemes in the modeling, and we have followed the reviewer’s recommendation regarding the second scheme. In the first, more central graining scheme, we use quadratic membrane patches with a sidelength of about 2 nm to determine membrane midplane shapes and lipid densities of each simulation conformation. This graining scheme has also been previously employed in Hu, Lipowsky, Weikl, PNAS 38, 15283 (2013) to determine the shape and thermal roughness of coarse-grained membranes. A sidelength of 2 nm is necessary to have sufficiently many lipid headgroups in the upper and lower leaflet in the membrane patches for estimating the local height of these leaflets, and the local membrane midplane height as average of these leaflet heights (see subsection “Membrane shape of simulation conformation” in the Methods section for details).  However, we strongly believe that doubling the sidelength of membrane patches in this discretization is not an option, because a discretization length of 4 nm is too coarse to resolve the membrane deformations in the nanodome, see e.g. the profiles in Fig. 1b. Moreover, any “noise” from this discretization is rather completely smoothened out in the averaging process used in the analysis of the membrane shapes, at least for the coarse-grained simulations. This averaging process requires rotations of membrane conformations to align the protein orientations of the conformations (see subsection “Average membrane shapes and lipid densities” for details). Because of these rotations, the original discretization is “lost” in the averaging, and a continuous membrane shape is generated. To calculate the excess areas and bending energies for this smooth, continuous membrane shape, we use a discretization of the Monge plane into a square lattice with lattice parameter 1 nm. As a response to the referee’s suggestion, we now report that the results for the excess area do not change significantly when doubling this lattice parameter to 2 nm. On line 597, we write:

“For a lattice constant of a=2 nm, we obtain extrapolated values of the excess area Delta A from the coarse-grained simulations that are 2 to 3% lower than the values for a=1 nm, which is a small compared to statistical uncertainties with relative errors of around 10%.”

On lines 614ff, we now state that the bending energy results are about 10% to 13% lower for a=2 nm, likely because of the lower resolution of the curvature in the nanodome compared to a=1 nm, rather than incomplete averaging and remaining roughness of the coarse-grained nanodome shapes.

(2) For a Gaussian bump with sigma=6 nm I obtained a bending energy of 0.6 kappa, so certainly in the ballpark with what they are reporting but significantly lower (compared to 2 kappa, Figure 5 lower left). It would be simpler to use the Gaussian approximation to their curves in Figure 3 - and I would argue more accurate, especially since they have not reported the variation of the membrane energy with respect to the discretization size and so I cannot judge the dependence of the energy on discretization. I view reporting the variation of the membrane energy with respect to discretization as being essential for the analysis if their goal is to provide a quantitative estimate for the force of Piezo. The Helfrich energy computed from an analytical model with a membrane shape closely resembling the simulated shapes would be very helpful. According to my intuition, finite-difference estimates of curvatures will tend to be overestimates of the true membrane deformation energy because white noise tends to lead to high curvature at short-length scales, which is strongly penalized by the bending energy. 

Instead of Gaussian bumps, we now calculate the membrane bending energy also from the two-dimensional, continuous mean curvature profiles (see Fig. 1c). These mean curvature profiles are highly smoothened (see figure caption for details). Nonetheless, we obtain essentially the same bending energies as in our discrete calculations of averaged, smoothened threedimensional membrane shapes, see new text on lines 326ff. We believe that this agreement corroborates our bending energy calculations. We still focus on values obtained for threedimensional membrane shapes, because of incomplete rotational symmetry. The three-dimensional membrane shapes exhibit variations with the three-fold symmetry of the Piezo proteins, see Figure 2a and b.

We agree that the bending energy of thermally rough membranes depends on the discretization scheme, because the discretization length of any discretization scheme leads to a cut-off length for fluctuation modes in a Fourier analysis. But again, we average out the thermal noise, for reasons given in the Results section, and analyse smooth membrane shapes.  

The fitting of the system deformation to the inverse time appears to be incredibly ad hoc ... Nor is it clear that the quantified model will be substantially changed without extrapolation. The authors should either justify the extrapolation more clearly (sorry if I missed it!) or also report the unextrapolated numbers alongside the extrapolated ones. 

We report the values of the excess area and bending energy in the different time intervals of our analysis as data points in Fig. 4 with supplement. We find it important to report the time dependence of these quantities, because the intended equilibration of the membrane shapes in our simulations is not “complete” within a certain time window of the simulations. So, just “cutting” the first 20 and 50% of the simulation trajectories, and analysing the remaining parts as “equilibrated” does not seem to be a reasonable choice here, at least for the membrane properties, i.e. for the excess area and bending energy. We agree that the linear extrapolation used in our analysis is a matter of choice. At least for the coarse-grained simulations, the extrapolated values of excess areas and bending energies are rather close to the values obtained in the last time windows (see Figure 4). 

In summary, this paper uses molecular dynamics simulations to quantify the force of the Piezo 1 and Piezo 2 proteins on a lipid bilayer using simulations under controlled tension, observing the membrane deformation, and using that data to infer protein mechanics. While much of the physical mechanism was previously known, the study itself is a valuable quantification. I identified one issue in the membrane deformation energy analysis that has large quantitative repercussions for the extracted model. 

Reviewer #2 (Public review): 

Summary: 

In this study, the authors suggest that the structure of Piezo2 in a tensionless simulation is flatter compared to the electron microscopy structure. This is an interesting observation and highlights the fact that the membrane environment is important for Piezo2 curvature. Additionally, the authors calculate the excess area of Piezo2 and Piezo1, suggesting that it is significantly smaller compared to the area calculated using the EM structure or simulations with restrained Piezo2. Finally, the authors propose an elastic model for Piezo proteins. Those are very important findings, which would be of interest to the mechanobiology field. 

Whilst I like the suggestion that the membrane environment will change Piezo2 flatness, could this be happening because of the lower resolution of the MARTINI simulations? In other words, would it be possible that MARTINI is not able to model such curvature due to its lower resolution? 

Related to my comment above, the authors say that they only restrained the secondary structure using an elastic network model. Whilst I understand why they did this, Piezo proteins are relatively large. How can the authors know that this type of elastic network model restrains, combined with the fact that MARTINI simulations are perhaps not very accurate in predicting protein conformations, can accurately represent the changes that happen within the Piezo channel during membrane tension? 

These questions regarding the reliability of the Martini model are very reasonable and are the reason why we include also results from atomistic simulations, at least for Piezo 2, and compare the results. In the Martini model, secondary structure constraints are standard. In addition, constraints on the tertiary structure (e.g. via an elastic network model) are also typically used in simulations of soluble, globular proteins. However, such tertiary constraints would make it impossible to simulate the tension-induced flattening of the Piezo proteins. So instead, as we write on lines 427ff, “we relied on the capabilities of the Martini coarse-grained force field for modeling membrane systems with TM helix assemblies (Sharma and Juffer, 2013; Chavent et al., 2014; Majumder and Straub, 2021).” In these refences, Martini simulations were used to study the assembly of transmembrane helices, leading to agreement with experimentally observed structures. As we state in our article, our atomistic simulations corroborate the Martini simulations, with the caveats that are now more extensively discussed in the new last paragraph of the Discussion section starting on line 362.

Modelling or Piezo1, seems to be based on homology to Piezo2. However, the authors need to further evaluate their model, e.g. how it compares with an Alphafold model. 

We understand the question, but see it beyond the scope of our article, also because of the computational demand of the simulations. The question is: Do coarse-grained simulations of Piezo1 based on an Alphafold model as starting structure lead to different results? It is important to note that we only model the rather flexible 12 TM helices at the outer ends of the Piezo 1 monomers via homology modeling to the Piezo 2 structure, which includes these TM helices. For the inner 26 TM helices, including the channel, we use the high-quality cryo-EM structure of Piezo 1. Alphafold may be an alternative for modeling the outer 12 helices, but we don’t think this would lead to statistically significant differences in simulations – e.g. because of the observed overall agreement of membrane shapes in all our Piezo 1 and Piezo 2 simulation systems.

To calculate the tension-induced flattening of the Piezo channel, the authors "divide all simulation trajectories into 5 equal intervals and determine the nanodome shape in each interval by averaging over the conformations of all independent simulation runs in this interval.". However, probably the change in the flattening of Piezo channel happens very quickly during the simulations, possibly within the same interval. Is this the case? and if yes does this affect their calculations? 

Unfortunately, the flattening is not sufficiently quick, so is not complete within the first time windows, see data points in Figure 4. We therefore report the time dependence with the plots in Figure 4 and extrapolate, see also our response above to reviewer 1.

Finally, the authors use a specific lipid composition, which is asymmetric. Is it possible that the asymmetry of the membrane causes some of the changes in the curvature that they observe? Perhaps more controls, e.g. with a symmetric POPC bilayer are needed to identify whether membrane asymmetry plays a role in the membrane curvature they observe. 

Because of the rather high computational demands, such controls are beyond our scope. We don’t expect statistically significant differences for symmetric POPC/cholesterol bilayers. On lines 229ff, we now state:

“Our modelling assumes that any spontaneous curvature from asymmetries in the lipid composition is small compared to the curvature of the nanodome and, thus, negligible, which is plausible for the rather slight lipid asymmetry of our simulated membranes (see Methods).”

Reviewer #3 (Public review): 

Strengths: 

This work focuses on a problem of deep significance: quantifying the structure-tension relationship and underlying mechanism for the mechanosensitive Piezo 1 and 2 channels. This objective presents a few technical challenges for molecular dynamics simulations, due to the relatively large size of each membrane-protein system. Nonetheless, the technical approach chosen is based on the methodology that is, in principle, established and widely accessible. Therefore, another group of practitioners would likely be able to reproduce these findings with reasonable effort. 

Weaknesses: 

The two main results of this paper are (1) that both channels exhibit a flatter structure compared to cryo-EM measurements, and (2) their estimated force vs. displacement relationship. Although the former correlates at least quantitatively with prior experimental work, the latter relies exclusively on simulation results and model parameters. 

Below is a summary of the key points we recommend addressing in a revised version of the manuscript: 

(1) The authors should report and discuss controls for the membrane energy calculations, specifically by increasing the density of the discretization graining. We also suggest validating the bending modulus used in the energy calculations for the specific lipid mixture employed in the study. 

We have addressed both points, see our response to the reviewer’s comments for further details.

(2) The authors should consider and discuss the potential limitations of the coarse-grained simulation force field and clarify how atomistic simulations validate the reported results, with a more detailed explanation of the potential interdependencies between the two. 

We now discuss the caveats in the comparison of coarse-grained and atomistic simulations in more detail in a new paragraph starting on line 362.

(3) The authors should provide further clarification on other points raised in the reviewers' comments, for instance, the potential role of membrane asymmetry. 

We have done this – see above. We now further explain on lines 437ff why we use an asymmetric membrane. On lines 230ff, we discuss that any spontaneous membrane curvature due to lipid asymmetry is likely small compared to the nanodome curvature and, thus, negligible.

Reviewer #1 (Recommendations for the authors): 

(1) Report discretization dependence of the membrane energy (up to double the density of the current discretization graining). 

We have added several text pieces in the paragraph “Excess area and bending energy” starting on line 583 in which we state how the results depend on the lattice constant a of the calculations.

(2) Evaluate an analytical energy of a membrane bump with a shape similar to the simulation. This would be free of all sampling and discretization artifacts and would thus be an excellent lower bound of the energy. 

We have done this for the curvature profile in Figure 1c and corresponding curvature profiles of the shape profiles in Figure 2d, see next text on lines 326ff.

Minor: 

(1)  The lipid density (Figure 1 right, 2c, 3c) is not interesting nor is it referred to. It can be dropped. 

We think the lipid density maps are important for two reasons: First, they show the protein shape obtained after averaging conformations, as low-lipid-density regions. Second, the lipid densities are used in the calculation of the bending energies, to limit the bending energy calculations to the membrane in the nanodome, see Eq. 9. We therefore prefer to keep them.

(2) Figure 7 is attractive but not used in a meaningful way. I suggest inserting the protein graphic from Figure 7 into Figure 1 with the 4-helix bundles numbered alongside the structure. Figure 7 could then be dropped. 

Figure 7 is a figure of the Methods section. We need it to illustrate and explain aspects of the setup (numbering of helices, missing loops) and analysis (numbering scheme of 4-TM helix units).

(3) Some editing of the use of the English language would be helpful. "Exemplary" is a bit of a funny word choice, it implies that the conformation is excellent, and not simply representative. I'd suggest "Representative conformation". 

We agree and have replaced “exemplary” by “representative”.

(4) Typos: 

Equation 4 - Missing parentheses before squared operator inside the square root. 

We have corrected this mistake.

Reviewer #2 (Recommendations for the authors): 

This study focuses mainly on Piezo2; the authors do not perform any atomistic simulations of Piezo1, and the coarse-grained simulations for Piezo1 are shorter. As a result, their analysis for Piezo2 seems more complete. It would be good if the authors did similar studies with Piezo1 as with Piezo2. 

We agree that atomistic simulations of Piezo 1 would be interesting, too. However, because the atomistic simulations are particularly demanding, this is beyond our scope.

Reviewer #3 (Recommendations for the authors): 

(1) At line 63, a very large tension from the previous work by De Vecchis et al is reported (68 mN/m). The authors are sampling values up to about 21 mN/m, which is considerably smaller. However, these values greatly exceed what typical lipid membranes can sustain (about 10 mN/m) before rupturing. When mentioning these large tensions, the authors should emphasize that these values are not physiologically significant, because they would rupture most plasma membranes. That said, their use in simulation could be justified to magnify the structural changes compared to experiments. 

We agree that our largest membrane tension values are unphysiological. However, we see a main novelty and relevance of our simulations in the fact that we obtain a response of the nanodome in the physiological range of membrane tensions, see e.g. the 3^rd sentence of the abstract. Yes, we include simulations at tensions of 21 mN/m, but most of our simulated tension values are in the range from 0 to 10 mN/m (see e.g. Fig. 3e), in contrast to previous simulation studies.   

(2) At line 78 and in the Methods, only the reference paper is for the CHARMM protein force field, but not for the lipid force field. 

We have added the reference Klauda et al., 2010 for the CHARMM36 lipid force field in both spots.

(3) (Line 83) Acknowledging that the authors needed to use the structure from micelles (because it has atomic resolution), how closely do their relaxed Piezo structures compare with the lowerresolution data from the MacKinnon and Patapoutian papers? 

There are no structures reported in these papers to compare with, only a clear flattening as stated.  

(4) (Line 99) The authors chose a slightly asymmetric lipid membrane composition to capture some specific plasma-membrane features. However, they do not discuss which features are described by this particular composition, which doesn't include different acyl-chain unsaturations between leaflets. Further, they do not seem to comment on whether there is enrichment of certain lipid species coupled to curvature, or whether there is any "scrambling" occurring when the dome section and the planar membrane are stitched together in the preparation phase (Figure 8). 

Enrichment of lipids in contact with the protein is addressed in the reference Buyan et al., 2020, based on Martini simulations with Piezo 1. We have a different focus, but still wanted to keep an asymmetric membrane as in essentially all previous simulation studies as now stated also on lines 439ff, to mimic the native Piezo membrane environment. There is no apparent “scrambling” in the setup of our membrane systems. We also did not explore any coupling between curvature and lipid composition, but will publish the simulation trajectories to enable such studies.  

(5) (Caption of Figure 2). Please comment briefly in the text why the tensionless simulation required a longer simulation run (e.g. larger fluctuations?) 

We added as explanation on line 500 as explanation: “ … to explore the role of the long-range shape fluctuations in tensionless membranes for the relaxation into equilibrium”. The relaxation time of membrane shape fluctuations strongly increases with the wave length, which is only limited by the simulation box size in the absence of tensions. However, also for 8 microsecond trajectories, we do not observe complete equilibriation and therefore decided to extrapolate the excess area and bending energy values obtained for different time intervals of the trajectories.

(6) (Caption of Figure 3). Please clarify in the Methods how the atomistic simulations were initialized were they taken from independent CG simulation snapshots? If not, the use of the adjective "independent" would be questionable given the very short atomistic simulation time length. 

We now added that the production simulations started from the same structure. On lines 386, we now discuss the starting structure of the atomistic simulations in more detail.

(7) (Line 202). The approach of discretizing the bilayer shape is reasonable, but no justification was provided for the 1-nm grid spacing. In my opinion, there should be a supporting figure showing how the bending energy varies with the grid spacing. 

We now report also the effect of a 2-nm grid spacing on the results, see new text passages on page 18, and provide an explanation for the smaller 1-nm grid spacing on lines 587ff, where we write:

“This lattice constant [a = 1 nm] is chosen to be smaller than the bin width of about 2nm used in determining the membrane shape of the simulation conformations, to take into account that the averaging of these membrane shapes can lead to a higher resolution compared to the 2 nm resolution of the individual membrane shapes.”

(8) (Line 211). The choice by the authors to use a mixed lipid composition complicates the task of defining a reasonable bending modulus. Experimentally and in atomistic simulations, lipids with one saturated tail (like POPC or SOPC) are much stiffer when they are mixed with cholesterol (https://doi.org/10.1529/biophysj.105.067652, https://doi.org/10.1103/PhysRevE.80.021931, https://doi.org/10.1093/pnasnexus/pgad269). On the other hand, MARTINI seems to predict a slight *softening* for POPC mixed with cholesterol (https://doi.org/10.1038/s41467-023-43892-x). Further complicating this matter, mixtures of phospholipids with different preferred curvatures are predicted to be softer than pure bilayers (e.g. https://doi.org/10.1021/acs.jpcb.3c08117), but asymmetric bilayers are stiffer than symmetric ones in some circumstances (https://doi.org/10.1016/j.bpj.2019.11.3398). 

This issue can be quite thorny: therefore, my recommendation would be to either: (a) directly compute k for their lipid composition, which is straightforward when using large CG bilayers (as was done in Fowler et al, 2016), but it would also require more advanced methods for the atomistic ones; (b) use a reasonable *experimental* value for k, based on a similar enough lipid composition. 

We now justify in somewhat more detail why we use an asymmetric membrane, but agree that his complicates the bending energy estimates. We only aim to estimate the bending energy in the Martini 2.2 force field, because our elasticity model is based on and, thus, limited to results obtained with this force field. We have included the two further references using the Martini 2.2 force field suggested by the reviewer on line 213, and discuss now in more detail how the bending rigidity estimate enters and affects the modeling, see lines 226ff.  

(9) (Line 224). Does this closing statement imply that all experimental work from ex-vivo samples describe Piezo states under some small but measurable tension? 

We compare here to the cryo-EM structure in detergent micelles. So, there is no membrane tension, there may be a surface tension of the micelle, but we assume here that Piezo proteins are essentially force free in detergent micelles. Membrane embedding, in contrast, leads to strong forces on Piezo proteins already in the absence of membrane tension, because of the membrane bending energy.

(10) (Line 304). The Discussion concludes with a reasonable point, albeit on a down note: could the authors elaborate on what kind of experimental approach may be able to verify their modeling results? 

Very good question, but this is somewhat beyond our expertise. We don’t have a clear recommendation – it is complicated. What can be verified is the flattening, i.e. the height and curvature of the nanodome in lower-resolution experiments. We see our results in line with these experiments, see Introduction. 

(11) (Line 331). The very title of the Majumder and Straub paper addresses the problem of excessive binding strength between protein beads in the MARTINI force field, which should be mentioned. Figure 3(d) shows that the atomistic systems have larger excess areas than the CG ones. This could be related to MARTINI's "stickiness", or just statistical sampling. Characterizing the grid spacing (see point 7 above) might help illuminate this. 

We discuss now the larger excess area values of the atomistic simulations on lines 381ff.  

(12) (Lines 367, 375). Are the harmonic restraints absolute position restraints or additional bonds?

Note also that the schedule at which the restraints are released (10-ns intervals) is relatively quick. Does the membrane have enough time to equilibrate the number of lipids in each leaflet? 

These are standard, absolute position restraints. The 10-ns intervals may be too short to fully equilibrate the numbers of lipids, we have not explored this. The main point in the setup was to have a reasonable TM helix embedding with a smooth membrane, without any rupturing. This turned out to be tricky, with the procedures illustrated in Figure 8 as solution. If the membrane is smooth, the lipid numbers quickly equilibrate either in the final relaxation or in the initial nanoseconds of the production runs.

(13) (Line 387) The use of an isotropic barostat for equilibration further impedes the system's ability to relax its structure. I feel that the authors should validate more strongly their protocol to rule out the possibility that incomplete equilibration could bias dynamics towards flatter membranes, which is one of the main results of this paper. 

We don’t see how choices in the initial relaxation steps could have affected our results, at least for the coarse-grained simulations. There is more and more flattening throughout all simulation trajectories, see e.g. the extrapolations in Figure 4. All initial simulation structures are significantly less flattened than the final structures in the production runs.

(14) (Line 403). What is the protocol for reducing the membrane size for atomistic simulation? This is even more important to mention than for CG simulations. 

We just cut lipids beyond the intended box size of the atomistic simulations. As a technical point, we now have also added on line 507 how PIP2 lipids were converted.

(15) (Line 423). The CHARMM force field requires a cut-off distance of 12 Å for van der Waals forces, with a force-based continuous switching scheme. The authors should briefly comment on this deviation and its possible impact on membrane properties. Quick test simulations of very small atomistic bilayers with the chosen composition could be used as a comparison. 

We don’t expect any relevant effect on membrane properties within the statistical accuracies of the quantities of interest here (i.e. excess areas).

(16) (Equation 4). There are some mismatched parentheses: please check. 

We have corrected this mistake.

(17) (Equations 7-8). Why did the authors use finite-differences derivatives of z(x,y) instead of using cubic splines and the corresponding analytical derivatives? 

In our experience, second derivatives of standard cubic splines can be problematic. The continuous membrane shapes we obtain in our analysis are averages of such splines. We find standard finite differences more reliable, and therefore discretize these shapes. Already for the 2d membrane profiles of Figure 1b and 2d, calculating curvatures from interpolations using splines is problematic.

Reviewer #2 (Public review):

Summary:

This study introduces an exciting dataset of single-unit responses in humans during a naturalistic and dynamic movie stimulus, with recordings from multiple regions within the medial temporal lobe. The authors use both a traditional firing-rate analysis as well as a sophisticated decoding analysis to connect these neural responses to the visual content of the movie, such as which character is currently on screen.

Strengths:

The results reveal some surprising similarities and differences between these two kinds of analyses. For visual transitions (such as camera angle cuts), the neurons identified in the traditional response analysis (looking for changes in firing rate of an individual neuron at a transition) were the most useful for doing population-level decoding of these cuts. Interestingly, this wasn't true for character decoding; excluding these "responsive" neurons largely did not impact population-level decoding, suggesting that the population representation is distributed and not well-captured by individual-neuron analyses.

The methods and results are well-described both in the text and in the figures. This work could be an excellent starting point for further research on this topic to understand the complex representational dynamics of single neurons during naturalistic perception.

Weaknesses:

(1) I am unsure what the central scientific questions of this work are, and how the findings should impact our understanding of neural representations. Among the questions listed in the introduction is "Which brain regions are informative for specific stimulus categories?". This is a broad research area that has been addressed in many neuroimaging studies for decades, and it's not clear that the results tell us new information about region selectivity. "Is the relevant information distributed across the neuronal population?" is also a question with a long history of work in neuroscience about localist vs distributed representations, so I did not understand what specific claim was being made and tested here. Responses in individual neurons were found for all features across many regions (e.g., Table S1), but decodable information was also spread across the population.

(2) The character and indoor/outdoor labels seem fundamentally different from the scene/camera cut labels, and I was confused by the way that the cuts were put into the decoding framework. The decoding analyses took a 1600ms window around a frame of the video (despite labeling these as frame "onsets" like the feature onsets in the responsive-neuron analysis, I believe this is for any frame regardless of whether it is the onset of a feature), with the goal of predicting a binary label for that frame. Although this makes sense for the character and indoor/outdoor labels, which are a property of a specific frame, it is confusing for the cut labels since these are inherently about a change across frames. The way the authors handle this is by labeling frames as cuts if they are in the 520ms following a cut (there is no justification given for this specific value). Since the input to a decoder is 1600ms, this seems like a challenging decoding setup; the model must respond that an input is a "cut" if there is a cut-specific pattern present approximately in the middle of the window, but not if the pattern appears near the sides of the window. A more straightforward approach would be, for example, to try to discriminate between windows just after a cut versus windows during other parts of the video. It is also unclear how neurons "responsive" to cuts were defined, since the authors state that this was determined by looking for times when a feature was absent for 1000ms to continuously present for 1000ms, which would never happen for cuts (unless this definition was different for cuts?).

(3) The architecture of the decoding model is interesting but needs more explanation. The data is preprocessed with "a linear layer of same size as the input" (is this a layer added to the LSTM that is also trained for classification, or a separate step?), and the number of linear layers after the LSTM is "adapted" for each label type (how many were used for each label?). The LSTM also gets to see data from 800 ms before and after the labeled frame, but usually LSTMs have internal parameters that are the same for all timesteps; can the model know when the "critical" central frame is being input versus the context, i.e., are the inputs temporally tagged in some way? This may not be a big issue for the character or location labels, which appear to be contiguous over long durations and therefore the same label would usually be present for all 1600ms, but this seems like a major issue for the cut labels since the window will include a mix of frames with opposite labels.

(4) Because this is a naturalistic stimulus, some labels are very imbalanced ("Persons" appears in almost every frame), and the labels are correlated. The authors attempt to address the imbalance issue by oversampling the minority class during training, though it's not clear this is the right approach since the test data does not appear to be oversampled; for example, training the Persons decoder to label 50% of training frames as having people seems like it could lead to poor performance on a test set with nearly 100% Persons frames, versus a model trained to be biased toward the most common class. There is no attempt to deal with correlated features, which is especially problematic for features like "Summer Faces" and "Summer Presence", which I would expect to be highly overlapping, making it more difficult to interpret decoding performance for specific features.

(5) Are "responsive" neurons defined as only those showing firing increases at a feature onset, or would decreased activity also count as responsive? If only positive changes are labeled responsive, this would help explain how non-responsive neurons could be useful in a decoding analysis.

(6) Line 516 states that the scene cuts here are analogous to the hard boundaries in Zheng et al. (2022), but the hard boundaries are transitions between completely unrelated movies rather than scenes within the same movie. Previous work has found that within-movie and across-movie transitions may rely on different mechanisms, e.g., see Lee & Chen, 2022 (10.7554/eLife.73693).

Reviewer #3 (Public review):

This is an excellent, very interesting paper. There is a groundbreaking analysis of the data, going from typical picture presentation paradigms to more realistic conditions. I would like to ask the authors to consider a few points in the comments below.

(1) From Figure 2, I understand that there are 7 neurons responding to the character Summer, but then in line 157, we learn that there are 46. Are the other 39 from other areas (not parahippocampal)? If this is the case, it would be important to see examples of these responses, as one of the main claims is that it is possible to decode as good or better with non-responsive compared to single responsive neurons, which is, in principle, surprising.

(2) Also in Figure 2, there seem to be relatively very few neurons responding to Summer (1.88%) and to outdoor scenes (1.07%). Is this significant? Isn't it also a bit surprising, particularly for outdoor scenes, considering a previous paper of Mormann showing many outdoor scene responses in this area? It would be nice if the authors could comment on this.

(3) I was also surprised to see that there are many fewer responses to scene cuts (6.7%) compared to camera cuts (51%) because every scene cut involves a camera cut. Could this have been a result of the much larger number of camera cuts? (A way to test this would be to subsample the camera cuts.)

(4) Line 201. The analysis of decoding on a per-patient basis is important, but it should be done on a per-session basis - i.e., considering only simultaneously recorded neurons, without any pooling. This is because pooling can overestimate decoding performances (see e.g. Quian Quiroga and Panzeri NRN 2009). If there was only one session per patient, then this should be called 'per-session' rather than 'per-patient' to make it clear that there was no pooling.

(5) In general, the decoding results are quite interesting, and I was wondering if the authors could give a bit more insight by showing confusion matrices, with the predictions of the appearance of each of the characters, etc. Some of the characters may appear together, so this could be another entry of the decoder (say, predicting person A, B, C, A&B, A&C, B&C, A&B&C). I guess this could also show the power of analyzing the population activity.

(6) Lines 406-407. The claim that stimulus-selective responses to characters did not account for the decoding of the same character is very surprising. If I understood it correctly, the response criterion the authors used gives 'responsiveness' but not 'selectivity'. So, were people's responses selective (e.g., firing only to Summer) or non-selective (firing to a few characters)? This could explain why they didn't get good decoding results with responsive neurons. Again, it would be nice to see confusion matrices with the decoding of the characters. Another reason for this is that what are labelled as responsive neurons have relatively weak and variable responses.

(7) Line 455. The claim that 500 neurons drive decoding performance is very subjective. 500 neurons gives a performance of 0.38, and 50 neurons gives 0.33.

(8) Lines 492-494. I disagree with the claim that "character decoding does not rely on individual cells, as removing neurons that responded strongly to character onset had little impact on performance". I have not seen strong responses to characters in the paper. In particular, the response to Summer in Figure 2 looks very variable and relatively weak. If there are stronger responses to characters, please show them to make a convincing argument. It is fine to argue that you can get information from the population, but in my view, there are no good single-cell responses (perhaps because the actors and the movie were unknown to the subjects) to make this claim. Also, an older paper (Quian Quiroga et al J. Neurophysiol. 2007) showed that the decoding of individual stimuli in a picture presentation paradigm was determined by the responsive neurons and that the non-responsive neurons did not add any information. The results here could be different due to the use of movies instead of picture presentations, but most likely due to the fact that, in the picture presentation paradigm, the pictures were of famous people for which there were strong single neuron responses, unlike with the relatively unknown persons in this paper.

Author response:

Reviewer #1 (Public review):

Summary:

In this manuscript, Gerken et al examined how neurons in the human medial temporal lobe respond to and potentially code dynamic movie content. They had 29 patients watch a long-form movie while neurons within their MTL were monitored using depth electrodes. They found that neurons throughout the region were responsive to the content of the movie. In particular, neurons showed significant responses to people, places, and to a lesser extent, movie cuts. Modeling with a neural network suggests that neural activity within the recorded regions was better at predicting the content of the movies as a population, as opposed to individual neural representations. Surprisingly, a subpopulation of unresponsive neurons performed better than the responsive neurons at decoding the movie content, further suggesting that while classically nonresponsive, these neurons nonetheless provided critical information about the content of the visual world. The authors conclude from these results that low-level visual features, such as scene cuts, may be coded at the neuronal level, but that semantic features rely on distributed population-level codes.

Strengths:

Overall, the manuscript presents an interesting and reasonable argument for their findings and conclusions. Additionally, the large number of patients and neurons that were recorded and analyzed makes this data set unique and potentially very powerful. On the whole, the manuscript was very well written, and as it is, presents an interesting and useful set of data about the intricacies of how dynamic naturalistic semantic information may be processed within the medial temporal lobe.

We thank the reviewer for their comments on our manuscript and for describing the strengths of our presented work

Weaknesses:

There are a number of concerns I have based on some of the experimental and statistical methods employed that I feel would help to improve our understanding of the current data.

In particular, the authors do not address the issue of superposed visual features very well throughout the manuscript. Previous research using naturalistic movies has shown that low-level visual features, particularly motion, are capable of driving much of the visual system (e.g, Bartels et al 2005; Bartels et al 2007; Huth et al 2012; Çukur et al 2013; Russ et al 2015; Nentwich et al 2023). In some of these papers, low-level features were regressed out to look at the influence of semantics, in others, the influence of low-level features was explicitly modeled. The current manuscript, for the most part, appears to ignore these features with the exception of scene cuts. Based on the previous evidence that low-level features continue to drive later cortical regions, it seems like including these as regressors of no interest or, more ideally, as additional variables, would help to determine how well MTL codes for semantic features over top of these lower-order variables.

We thank the reviewer for this insightful comment and for the relevant literature regarding visual motion in not only the primary visual system but in cortical areas as well. While we agree that the inclusion of visual motion as a regressor of no interest or as an additional variable would be overall informative in determining if single neurons in the MTL are driven by this level of feature, we would argue that our analyses already provide some insight into its role and that only the parahippocampal cortical neurons would robustly track this feature.

As noted by the reviewer, our model includes two features derived from visual motion: Camera Cuts (directly derived from frame-wise changes in pixel values)  and Scene Cuts (a subset of Camera Cuts restricted to changes in scene). As shown in Fig. 5a, decoding performance for these features was strongest in the parahippocampal cortex (~20%), compared to other MTL areas (~10%). While the entorhinal cortex also showed some performance for Scene Cuts (15%), we interpret this as being driven by the changes in location that define a scene, rather than by motion itself.

These findings suggest that while motion features are tracked in the MTL, the effect may be most robust in the parahippocampal cortex. We believe that quantifying more complex 3D motion in a naturalistic stimulus like a full-length movie is a significant challenge that would likely require a dedicated study. We agree this is an interesting future research direction and will update the manuscript to highlight this for the reader.

A few more minor points that would help to clarify the current results involve the selection of data for particular analyses. For some analyses, the authors chose to appropriately downsample their data sets to compare across variables. However, there are a few places where similar downsampling would be informative, but was not completed. In particular, the analyses for patients and regions may have a more informative comparison if the full population were downsampled to match the size of the population for each patient or region of interest. This could be done with the Monte Carlo sampling that is used in other analyses, thus providing a control for population size while still sampling the full population.

We thank the reviewer for raising this important methodological point. The decision not to downsample the patient- and region-specific analyses was deliberate, and we appreciate the opportunity to clarify our rationale.

Generally, we would like to emphasize that due to technical and ethical limitations of human single-neuron recordings, it is currently not possible to record large populations of neurons simultaneously in individual patients. The limited and variable number of recorded neurons per subject (Fig. S1) generally requires pooling neurons into a pseudo-populations for decoding, which is a well‐established standard in human single‐neuron studies (see e.g., (Jamali et al., 2021; Kamiński et al., 2017; Minxha et al., 2020; Rutishauser et al., 2015; Zheng et al., 2022)).

For the patient-specific analysis, our primary goal was to show that no single patient's data could match the performance of the complete pseudo-population. Crucially, we found no direct relationship between the number of recorded neurons and decoding performance; patients with the most neurons (patients 4, 13) were not top performers, and those with the fewest (patients 11, 14) were not the worst (see Fig. 4). This indicates that neuron count was not the primary limiting factor and that downsampling would be unlikely to provide additional insight.

Similarly, for the region-specific analysis, regions with larger neural populations did not systematically outperform those with fewer neurons (Fig. 5). Given the inherent sparseness of single-neuron data, we concluded that retaining the full dataset was more informative than excluding neurons simply to equalize population sizes.

We agree that this methodological choice should be transparent and explicitly justified in the text. We will add an explanation to the revised manuscript to justify why this approach was taken and how it differs from the analysis in Fig. 6.

Reviewer #2 (Public review):

Summary:

This study introduces an exciting dataset of single-unit responses in humans during a naturalistic and dynamic movie stimulus, with recordings from multiple regions within the medial temporal lobe. The authors use both a traditional firing-rate analysis as well as a sophisticated decoding analysis to connect these neural responses to the visual content of the movie, such as which character is currently on screen.

Strengths:

The results reveal some surprising similarities and differences between these two kinds of analyses. For visual transitions (such as camera angle cuts), the neurons identified in the traditional response analysis (looking for changes in firing rate of an individual neuron at a transition) were the most useful for doing population-level decoding of these cuts. Interestingly, this wasn't true for character decoding; excluding these "responsive" neurons largely did not impact population-level decoding, suggesting that the population representation is distributed and not well-captured by individual-neuron analyses.

The methods and results are well-described both in the text and in the figures. This work could be an excellent starting point for further research on this topic to understand the complex representational dynamics of single neurons during naturalistic perception.

We thank the reviewer for their feedback and for summarizing the results of our work.

(1) I am unsure what the central scientific questions of this work are, and how the findings should impact our understanding of neural representations. Among the questions listed in the introduction is "Which brain regions are informative for specific stimulus categories?". This is a broad research area that has been addressed in many neuroimaging studies for decades, and it's not clear that the results tell us new information about region selectivity. "Is the relevant information distributed across the neuronal population?" is also a question with a long history of work in neuroscience about localist vs distributed representations, so I did not understand what specific claim was being made and tested here. Responses in individual neurons were found for all features across many regions (e.g., Table S1), but decodable information was also spread across the population.

We thank the reviewer for this important point, which gets to the core of our study's contribution. While concepts like regional specificity are well-established from studies on the blood-flow level, their investigation at the single-neuron level in humans during naturalistic, dynamic stimulation remains a critical open question. The type of coding (sparse vs. distributed) on the other hand cannot be investigated with blood-flow studies as the technology lacks the spatial and temporal resolution.

Our study addresses this gap directly. The exceptional temporal resolution of single-neuron recordings allows us to move beyond traditional paradigms and examine cellular-level dynamics as they unfold in neuronal response on a frame-by-frame basis to a more naturalistic and ecologically valid stimulus. It cannot be assumed that findings from other modalities or simplified stimuli will generalize to this context.

To meet this challenge, we employed a dual analytical strategy: combining a classic single-unit approach with a machine learning-based population analysis. This allowed us to create a bridge between prior work and our more naturalistic data. A key result is that our findings are often consistent with the existing literature, which validates the generalizability of those principles. However, the differences we observe between these two analytical approaches are equally informative, providing new insights into how the brain processes continuous, real-world information.

We will revise the introduction and discussion to more explicitly frame our work in this context, emphasizing the specific scientific question driving this study, while also highlighting the strengths of our experimental design and recording methods.

(2) The character and indoor/outdoor labels seem fundamentally different from the scene/camera cut labels, and I was confused by the way that the cuts were put into the decoding framework. The decoding analyses took a 1600ms window around a frame of the video (despite labeling these as frame "onsets" like the feature onsets in the responsive-neuron analysis, I believe this is for any frame regardless of whether it is the onset of a feature), with the goal of predicting a binary label for that frame. Although this makes sense for the character and indoor/outdoor labels, which are a property of a specific frame, it is confusing for the cut labels since these are inherently about a change across frames. The way the authors handle this is by labeling frames as cuts if they are in the 520ms following a cut (there is no justification given for this specific value). Since the input to a decoder is 1600ms, this seems like a challenging decoding setup; the model must respond that an input is a "cut" if there is a cut-specific pattern present approximately in the middle of the window, but not if the pattern appears near the sides of the window. A more straightforward approach would be, for example, to try to discriminate between windows just after a cut versus windows during other parts of the video. It is also unclear how neurons "responsive" to cuts were defined, since the authors state that this was determined by looking for times when a feature was absent for 1000ms to continuously present for 1000ms, which would never happen for cuts (unless this definition was different for cuts?).

We thank the reviewer for the valuable comment regarding specifically the cut labels. The choice to label frames that lie in a time window of 520ms following a cut as positive was selected based on prior research and is intended to include the response onsets across all regions within the MTL (Mormann et al., 2008). We agree that this explanation is currently missing from the manuscript, and we will add a brief clarification in the revised version.

As correctly noted, the decoding analysis does not rely on feature onset but instead continuously decodes features throughout the entire movie. Thus, all frames are included, regardless of whether they correspond to a feature onset.

Our treatment of cut labels as sustained events is a deliberate methodological choice. Neural responses to events like cuts often unfold over time, and by extending the label, we provide our LSTM network with the necessary temporal window to learn this evolving signature. This approach not only leverages the sequential processing strengths of the LSTM (Hochreiter et al., 1997) but also ensures a consistent analytical framework for both event-based (cuts) and state-based (character or location) features.

(3) The architecture of the decoding model is interesting but needs more explanation. The data is preprocessed with "a linear layer of same size as the input" (is this a layer added to the LSTM that is also trained for classification, or a separate step?), and the number of linear layers after the LSTM is "adapted" for each label type (how many were used for each label?). The LSTM also gets to see data from 800 ms before and after the labeled frame, but usually LSTMs have internal parameters that are the same for all timesteps; can the model know when the "critical" central frame is being input versus the context, i.e., are the inputs temporally tagged in some way? This may not be a big issue for the character or location labels, which appear to be contiguous over long durations and therefore the same label would usually be present for all 1600ms, but this seems like a major issue for the cut labels since the window will include a mix of frames with opposite labels.

We thank the reviewer for their insightful comments regarding the decoding architecture. The model consists of an LSTM followed by 1–3 linear readout layers, where the exact number of layers is treated as a hyperparameter and selected based on validation performance for each label type. The initial linear layer applied to the input is part of the trainable model and serves as a projection layer to transform the binned neural activity into a suitable feature space before feeding it into the LSTM. The model is trained in an end-to-end fashion on the classification task.

Regarding temporal context, the model receives a 1600 ms window (800 ms before and after the labeled frame), and as correctly pointed out by the reviewer, LSTM parameters are shared across time steps. We do not explicitly tag the temporal position of the central frame within the sequence. While this may have limited impact for labels that persist over time (e.g., characters or locations), we agree this could pose a challenge for cut labels, which are more temporally localized.

This is an important point, and we will clarify this limitation in the revised manuscript and consider incorporating positional encoding in future work to better guide the model’s focus within the temporal window. Additionally, we will add a data table, specifying the ranges of hyperparameters in our decoding networks. Hyperparameters were optimized for each feature and split individually, but we agree that some more details on how these parameters were chosen are important and we will provide a data table in our revised manuscript giving more insights into the ranges of hyperparameters.

We thank the reviewer for this important point. We will clarify this limitation in the revised manuscript and note that positional encoding is a valuable direction to better guide the model’s focus within the temporal window. To improve methodological transparency, we will also add a supplementary table detailing the hyperparameter ranges used for our optimization process.

(4) Because this is a naturalistic stimulus, some labels are very imbalanced ("Persons" appears in almost every frame), and the labels are correlated. The authors attempt to address the imbalance issue by oversampling the minority class during training, though it's not clear this is the right approach since the test data does not appear to be oversampled; for example, training the Persons decoder to label 50% of training frames as having people seems like it could lead to poor performance on a test set with nearly 100% Persons frames, versus a model trained to be biased toward the most common class. [...]

We thank the reviewer for this critical and thoughtful comment. We agree that the imbalanced and correlated nature of labels in naturalistic stimuli is a key challenge.

To address this, we follow a standard machine learning practice: oversampling is applied exclusively to the training data. This technique helps the model learn from underrepresented classes by creating more balanced training batches, thus preventing it from simply defaulting to the majority class. Crucially, the test set remains unaltered to ensure our evaluation reflects the model's true generalization performance on the natural data distribution.

For the “Persons” feature, which appears in nearly all frames, defining a meaningful negative class is particularly challenging. The decoder must learn to identify subtle variations within a highly skewed distribution. Oversampling during training helps provide a more balanced learning signal, while keeping the test distribution intact ensures proper evaluation of generalization.

The reviewer’s comment—that we are “training the Persons decoder to label 50% of training frames as having people”—may suggest that labels were modified. We want to emphasize this is not the case. Our oversampling strategy does not alter the labels; it simply increases the exposure of the rare, underrepresented class during training to ensure the model can learn its pattern despite its low frequency.

We will revise the Methods section to describe this standard procedure more explicitly, clarifying that oversampling is a training-only strategy to mitigate class imbalance.

(5) Are "responsive" neurons defined as only those showing firing increases at a feature onset, or would decreased activity also count as responsive? If only positive changes are labeled responsive, this would help explain how non-responsive neurons could be useful in a decoding analysis.

We define responsive neurons as those showing increased firing rates at feature onset; we did not test for decreases in activity. We thank the reviewer for this valuable comment and will address this point in the revised manuscript by assessing responseness without a restriction on the direction of the firing rate.

(6) Line 516 states that the scene cuts here are analogous to the hard boundaries in Zheng et al. (2022), but the hard boundaries are transitions between completely unrelated movies rather than scenes within the same movie. Previous work has found that within-movie and across-movie transitions may rely on different mechanisms, e.g., see Lee & Chen, 2022 (10.7554/eLife.73693).

We thank the reviewer for pointing out this distinction and for including the relevant work from Lee & Chan (2022) which further contextualizes this distinction. Indeed, the hard boundaries defined in the cited paper differ slightly from ours. The study distinguishes between (1) hard boundaries—transitions between unrelated movies—and (2) soft boundaries—transitions between related events within the same movie. While our camera cuts resemble their soft boundaries, our scene cuts do not fully align with either category. We defined scene cuts to be more similar to the study’s hard boundaries, but we recognize this correspondence is not exact. We will clarify the distinctions between our scene cuts and the hard boundaries described in Zheng et al. (2022) in the revised manuscript, and will update our text to include the finding from Lee & Chan (2022).

Reviewer #3 (Public review):

This is an excellent, very interesting paper. There is a groundbreaking analysis of the data, going from typical picture presentation paradigms to more realistic conditions. I would like to ask the authors to consider a few points in the comments below.

(1) From Figure 2, I understand that there are 7 neurons responding to the character Summer, but then in line 157, we learn that there are 46. Are the other 39 from other areas (not parahippocampal)? If this is the case, it would be important to see examples of these responses, as one of the main claims is that it is possible to decode as good or better with non-responsive compared to single responsive neurons, which is, in principle, surprising.

We thank the reviewer for pointing out this ambiguity in the text. Yes, the other 39 units are responsive neurons from other areas. We will clarify to which neuronal sets the number of responsive neurons corresponds. We will also include response plots depicting the unit activity for the mentioned units.

(2) Also in Figure 2, there seem to be relatively very few neurons responding to Summer (1.88%) and to outdoor scenes (1.07%). Is this significant? Isn't it also a bit surprising, particularly for outdoor scenes, considering a previous paper of Mormann showing many outdoor scene responses in this area? It would be nice if the authors could comment on this.

We thank the reviewer for this insightful point. While a low response to the general 'outdoor scene' label seems surprising at first, our findings align with the established role of the parahippocampal cortex (PHC) in processing scenes and spatial layouts. In previous work using static images, each image introduces a new spatial context. In our movie stimulus, new spatial contexts specifically emerge at scene cuts. Accordingly, our data show a strong PHC response precisely at these moments. We will revise the discussion to emphasize this interpretation, highlighting the consistency with prior work.

Regarding the first comment, we did not originally test if the proportion of the units is significant using e.g. a binomial test. We will include the results of a binomial test for each region and feature pair in the revised manuscript.

(3) I was also surprised to see that there are many fewer responses to scene cuts (6.7%) compared to camera cuts (51%) because every scene cut involves a camera cut. Could this have been a result of the much larger number of camera cuts? (A way to test this would be to subsample the camera cuts.)

The decrease in responsive units for scene cuts relative to camera cuts could indeed be due to the overall decrease in “trials” from one label to the other. To test this, we will follow the reviewer’s suggestion and perform tests using sets of randomly subsampled camera cuts and will include the results in the revised manuscript.

(4) Line 201. The analysis of decoding on a per-patient basis is important, but it should be done on a per-session basis - i.e., considering only simultaneously recorded neurons, without any pooling. This is because pooling can overestimate decoding performances (see e.g. Quian Quiroga and Panzeri NRN 2009). If there was only one session per patient, then this should be called 'per-session' rather than 'per-patient' to make it clear that there was no pooling.

The per-patient decoding was indeed also a per-session decoding, as each patient contributed only a single session to the dataset. We will make note of this explicitly in the text to resolve the ambiguity.

(6) Lines 406-407. The claim that stimulus-selective responses to characters did not account for the decoding of the same character is very surprising. If I understood it correctly, the response criterion the authors used gives 'responsiveness' but not 'selectivity'. So, were people's responses selective (e.g., firing only to Summer) or non-selective (firing to a few characters)? This could explain why they didn't get good decoding results with responsive neurons. Again, it would be nice to see confusion matrices with the decoding of the characters. Another reason for this is that what are labelled as responsive neurons have relatively weak and variable responses.

We thank the reviewer for pointing out the importance of selectivity in addition to responsiveness. Indeed, our response criterion does not take stimulus selectivity into account and exclusively measures increases in firing activity after feature onsets for a given feature irrespective of other features.

We will adjust the text to reflect this shortcoming of the response-detection approach used here. To clarify the relationship between neural populations, we will add visualizations of the overlap of responsive neurons across labels for each subregion. These figures will be included in the revised manuscript.

In our approach, we trained separate networks for each feature to effectively mitigate the issue of correlated feature labels within the dataset (see earlier discussion). While this strategy effectively deals with the correlated features, it precluded the generation of standard confusion matrices, as classification was performed independently for each feature.

To directly assess the feature selectivity of responsive neurons, we will fit generalized linear models to predict their firing rates from the features. This approach will enable us to quantify their selectivity and compare it to that of the broader neuronal population.

(7) Line 455. The claim that 500 neurons drive decoding performance is very subjective. 500 neurons gives a performance of 0.38, and 50 neurons gives 0.33.

We agree with the reviewer that the phrasing is unclear. We will adjust our summary of this analysis as given in Line 455 to reflect that the logistic regression-derived neuronal rankings produce a subset which achieve comparable performance.

(8) Lines 492-494. I disagree with the claim that "character decoding does not rely on individual cells, as removing neurons that responded strongly to character onset had little impact on performance". I have not seen strong responses to characters in the paper. In particular, the response to Summer in Figure 2 looks very variable and relatively weak. If there are stronger responses to characters, please show them to make a convincing argument. It is fine to argue that you can get information from the population, but in my view, there are no good single-cell responses (perhaps because the actors and the movie were unknown to the subjects) to make this claim. Also, an older paper (Quian Quiroga et al J. Neurophysiol. 2007) showed that the decoding of individual stimuli in a picture presentation paradigm was determined by the responsive neurons and that the non-responsive neurons did not add any information. The results here could be different due to the use of movies instead of picture presentations, but most likely due to the fact that, in the picture presentation paradigm, the pictures were of famous people for which there were strong single neuron responses, unlike with the relatively unknown persons in this paper.

This is an important point and we thank the reviewer for highlighting a previous paradigm in which responsive neurons did drive decoding performance. Indeed, the fact that the movie, its characters and the corresponding actors were novel to patients could explain the disparity in decoding performance by way of weaker and more variable responses. We will include additional examples in the supplement of responses to features. Additionally, we will modify the text to emphasize the point that reliable decoding is possible even in the absence of a robust set of neuronal responses. It could indeed be the case that a decoder would place more weight on responsive units if they were present (as shown in the mentioned paper and in our decoding from visual transitions in the parahippocampal cortex).

能否帮我将这个目录中涉及到的内容在开发中实际使用频率进行排序整理

好的！我帮你根据在实际前端/全栈开发中Zod各类类型和特性的使用频率做一个大致的排序整理。排序原则主要基于：

• 该类型/特性在典型业务中出现的频率
• 其重要性和基础性
• 实际开发场景的普遍性

Zod类型及功能按“开发中实际使用频率”大致排序

1. 最常用（基础且必备）

• Primitives（原始值） 字符串（Strings）、数字（Numbers）、布尔（Booleans） → 几乎所有接口字段都会用到基础类型。

• Strings（字符串） 及其常见扩展，如 .min(), .max(), .email(), .url(), .regex()

• Optionals（可选类型） 很多接口字段是可选的，必用。

• Nullables（可空类型） & Nullish（空值） 处理null、undefined的字段，常见于接口数据。

• Objects（对象） 业务数据模型基本都是对象，必不可少。

• Arrays（数组） 常见的数据集合，比如列表、数组参数。

• Unions（联合类型） 处理多个可能类型的字段（比如字符串或数字），较常用。

2. 较常用（中级复杂度）

• Tuples（元组） 用于固定长度且类型固定的数组，稍复杂但常见。

• Enums（枚举） 用于限定字段取固定几种值，业务逻辑常用。

• Date（日期） 处理日期时间字段，业务常见。

• Refinements（精炼） 自定义更复杂校验规则，用于特殊业务逻辑。

• Transforms（转换器） 对数据进行转换（比如字符串转数字），常见于数据预处理。

• Readonly（只读） 用于限定字段不可修改，常见于状态管理。

3. 偶尔用到（复杂或特殊场景）

• Records（记录） 对象的键值对映射，某些业务用。

• Maps & Sets（映射和集合） JS内置复杂数据结构，较少在接口层面直接用。

• Recursive objects（递归对象） 用于树形结构数据，较复杂场景。

• Discriminated unions（区分联合） 处理多个类型的判别型联合，复杂状态机等场景。

• Instanceof 针对类实例校验，使用场景有限。

• Promises（承诺） 验证异步结果，偶尔用于接口响应类型。

• Stringbools（字符串布尔） 字符串表现的布尔值，较少见。

4. 很少用或特殊用途

• Template literals（模板字符串） 用于字面量模板字符串类型校验，较少用。

• BigInts 大整数类型，业务少见。

• Files（文件） 上传文件场景，前端较特殊。

• Defaults & Prefaults（默认值/预设置） 辅助用法，非校验核心。

• Catch（捕获） 错误捕获，较底层。

• Branded types（品牌类型） 用于类型区分，较高级用法。

• Custom（自定义） 用户自定义校验，视业务需求使用频率不确定。

• JSON 处理JSON结构，使用场景较特殊。

• Functions（函数） 验证函数类型，极少在普通业务中用。

• Pipes（管道） 多阶段转换/验证流程，较高级用法。

综上所述

| 频率等级 | 主要类型/功能 | | ---- | ------------------------------------------------------------------------------------------------- | | 高频 | Primitives, Strings, Optionals, Nullables, Objects, Arrays, Unions | | 中频 | Tuples, Enums, Dates, Refinements, Transforms, Readonly | | 低频 | Records, Maps, Sets, Recursive objects, Discriminated unions, Instanceof, Promises, Stringbools | | 很低频 | Template literals, BigInts, Files, Defaults, Catch, Branded types, Custom, JSON, Functions, Pipes |

如果你需要，我可以帮你结合具体开发场景，举例说明某些类型的典型应用。你觉得呢？

为什么感觉zod把js中涉及到的所有数据类型好像都进行了一层映射

你观察得非常精准，Zod确实对JavaScript中的各种数据类型、数据结构和类型组合，做了一个全面且系统的“映射和封装”，这个设计背后有几个核心原因和考量：

为什么Zod对所有JS数据类型都做了一层映射？

1. 统一的验证与声明语法

• JavaScript本身是弱类型语言，数据类型灵活但不严谨，运行时很容易出错。
• Zod的目标是给JS/TS加一层强类型校验层，用声明式的schema来描述数据结构和类型。
• 为了做到这一点，必须支持JS里绝大多数常见的数据类型和结构。
• 这样开发者可以用统一的方式声明和验证各种类型，避免到处写繁琐的校验代码。

2. 增强类型安全和开发体验

• TS的静态类型检查只能做到编译时，运行时依然需要验证数据正确性（比如接口返回、用户输入）。
• Zod提供对应类型的schema，不仅让TS类型推导准确，还保证运行时能自动验证对应类型。
• 这要求它对JS各种类型和组合都有“映射”，确保类型声明和实际数据验证同步。

3. 支持丰富的数据结构和复杂类型组合

• 真实项目中，数据结构复杂多变：对象嵌套、递归结构、联合类型、交叉类型、枚举等，
• 如果只支持基本类型，无法满足大部分场景。
• 所以Zod设计了丰富的类型组合器（Union、Intersection、Tuple、Record、Recursive Object等），支持复杂类型声明和验证。

4. 让schema更具表达力和灵活性

• 不同类型需要不同验证规则（数字范围，字符串格式，日期有效性等）。
• Zod提供对应类型专属API，让校验更精准，规则更易读。
• 例如，字符串格式（email、url）、数字精度、日期范围、文件类型、promise结果类型等，都有专门的schema。

5. 契合TypeScript类型系统

• Zod设计时深度整合TS类型系统，让schema可以和TS类型自动映射（类型推断），
• 这就要求它在类型层面做“映射”，每个Zod schema对应一个TS类型。
• 这保证类型和验证完全一致，避免类型错配。

6. 实用性与扩展性

• 内置丰富类型映射，可以满足绝大多数应用需求，减少用户自定义验证。
• 对于特殊需求，也支持“Custom”、“Refinements”等自定义扩展。
• 并且Zod设计模块化，方便未来继续支持更多类型或扩展功能。

总结

| 原因 | 说明 | | --------- | --------------------------- | | 统一声明与验证接口 | 用schema统一声明所有JS数据类型和结构，简化开发 | | 运行时类型安全 | 结合TS类型推断和运行时验证，提高安全性 | | 支持复杂类型 | 满足真实项目中复杂数据结构需求 | | 表达力强 | 类型专属API支持精准校验 | | TS整合 | 与TS类型系统无缝对接，类型推断准确 | | 扩展性好 | 预留自定义和扩展接口 |

打个比方

Zod就像一个“数据类型的标准语言”，把JS里原本分散、松散的类型和数据结构，通过schema规范化，既支持声明又支持验证，做到**“描述与校验一体化”，且覆盖全面**。

如果你想，我可以帮你结合几个类型讲讲Zod是怎么映射和验证的，或者帮你看具体某些类型的设计细节。你感兴趣吗？

我怎么知道zod提供这些方法 怎么知道对字面的处理需求是这些 我感觉这种api的列表应该是相对基础和完备的

我其实是希望能解释一下这些需求是如何坍缩到这种程度的

为什么Zod的API是“基础且完备”的？ 基础：设计的函数几乎都是最常见的，比如长度校验、正则校验、内置格式校验，这些是任何字符串验证不可或缺的。

完备：为了适应绝大多数前端或后端数据校验场景，不能缺失关键的校验方法。Zod把常见场景都涵盖了，且可以通过链式调用组合使用。

明白了，你是想知道，为什么字符串的各种实际需求，经过抽象和总结，最终“坍缩”成了Zod这类库里那些看起来简洁、基础但完备的API集合，换句话说，字符串校验需求背后的本质和设计逻辑是怎样的？

需求如何坍缩成Zod里的字符串验证API？

1. 现实中字符串验证需求的多样性和共性

现实中，我们对字符串的需求很丰富，比如：

• 长度要求（密码至少8位，用户名最多20位）
• 格式要求（必须是邮箱、URL、数字串、手机号）
• 内容过滤（不允许空格、不允许特殊字符）
• 变换需求（去空格、转大小写）
• 组合复杂条件（长度+格式+内容）

但这些多样的需求，本质上是对字符串的几个核心属性和行为的约束和转换，它们之间是有共通点的。

2. 这些需求背后的本质属性可以归纳为：

| 核心属性 | 说明 | 典型API对应 | | ------------- | ----------------- | --------------------------------------------- | | 长度限制 | 字符串的最短、最长长度 | .min(), .max() | | 格式/模式匹配 | 字符串是否符合特定规则（正则） | .regex(), .email(), .url()等 | | 内容清理/转换 | 去除空白、统一大小写等 | .trim(), .toLowerCase(), .toUpperCase() | | 字符类别限制 | 限制字符是否属于某类（数字、字母） | 通过 .regex() 实现 | | 是否必填/空字符串 | 是否允许空串 | .nonempty() |

3. 为什么能用这几个维度覆盖绝大多数需求？

• 长度限制是最常见、最基础的约束，保证数据不太短也不太长。
• 格式匹配满足几乎所有对结构有要求的场景（邮箱、手机号、UUID、日期等），而且用正则表达式统一表达方便扩展。
• 内容转换是常见的预处理手段，确保数据符合后续处理需求。
• 字符类别限制本质上是格式匹配的一种特例。
• 空值处理在数据验证中非常关键，区分“空”和“非空”，保证字段有效。

通过把各种具体需求映射到这些核心维度，能大幅度简化设计。

4. 设计坍缩的过程其实就是归纳、抽象、模块化

• 归纳：观察大量实际案例，总结出共同点。比如，很多校验本质上是“长度”和“格式”的问题。
• 抽象：用抽象的API接口来表示这些共性，如.min(), .max(), .regex()，而不是每种具体格式都写一套。
• 模块化：提供组合式API，通过链式调用组合各种规则，实现复杂校验。

5. 额外说明：为什么不是直接写一堆不同函数？

• 如果为每个业务需求写一个专门函数（比如validateEmail(), validatePhoneNumber(), validatePasswordComplexity()），

• 不易扩展，容易臃肿

• 复用度低

• 不利于组合更复杂的验证逻辑

• 而用“基础规则+组合”的模式，更灵活，满足更多场景，维护简单。

总结：

| 现实需求的多样性 | 通过归纳抽象总结成的几个核心维度 | Zod等库里的基础API | | ------------------ | --------------------- | ------------------------------------------------------ | | 长度限制、格式复杂度、内容过滤、转换 | 长度限制、正则匹配、空白处理、大小写转换等 | .min(), .max(), .regex(), .trim(), .email()等 |

换句话说，Zod字符串API的设计就是对现实中五花八门的字符串需求，通过“抽象核心属性+组合调用”这套设计思路做了极致简化和统一。

如果你想，我还可以帮你结合实际需求举例，说明它们是怎么映射到这些API上的。你觉得怎么样？

Summary of the Discussion on SwiftUI: Understanding Identity, Lifetime, and Dependencies

1. Introduction to SwiftUI and Its Declarative Nature

2. SwiftUI operates as a declarative UI framework where you describe UI states, and the framework manages their actualization. "That means that you describe what you want for your app at a high level, and SwiftUI decides exactly how to make it happen."

3. Understanding Identity in SwiftUI

4. SwiftUI views have identity to distinguish elements as the same or different across updates, critical for UI transitions and state management. "Identity is how SwiftUI recognizes elements as the same or distinct across multiple updates of your app."

5. Concept of View Identity Using Practical Examples

6. Demonstrated using the "Good Dog, Bad Dog" app example, explaining how identity influences view transitions and behavior. "That distinction actually matters a great deal because it changes how our interface transitions from one state to another."

7. Explicit vs. Structural Identity

8. Discussed two types of identity:

   • Explicit identity is assigned using identifiers like tags. "Explicit identity is powerful and flexible, but does require that someone, somewhere keeps track of all those names."
   • Structural identity is derived from the view's type and position in the hierarchy. "SwiftUI uses the structure of your view hierarchy to generate implicit identities for your views so you don't have to."

9. Role of Lifetime in SwiftUI

10. Explained how SwiftUI manages the life cycle of views and data by associating views' identity over time. "Lifetime is how SwiftUI tracks the existence of views and data over time."

11. Impact of Dependencies on UI Updates

12. Dependencies are inputs like state variables or environmental settings that trigger UI updates when they change. "Dependencies are how SwiftUI understands when your interface needs to be updated and why."

13. How SwiftUI Manages State and Identity

14. Discussed how State and StateObject help preserve state across the lifetime of views tied to their identity. "State and StateObject are the persistent storage associated with your view's identity."

15. Advanced Use of Identity with SwiftUI's ForEach

16. ForEach leverages identifiers for efficient updates and animations, showing how identity can impact performance and correctness. "Choosing a good identifier is your opportunity to control the lifetime of your views and data."

17. Best Practices for Using Identity

18. Emphasized the importance of stable and unique identifiers to improve performance and prevent state loss. "An identifier that isn't stable can result in a shorter view lifetime."

19. Troubleshooting and Optimization Techniques

    • Discussed common pitfalls with AnyView and alternatives using view builders to optimize SwiftUI’s understanding and performance. "Having too many AnyViews will often make code harder to read and understand."

Key Takeaways

• Identity, lifetime, and dependencies are core concepts that determine how SwiftUI manages and updates the UI.
• Effective management of these properties can significantly enhance the performance and predictability of SwiftUI applications.
• Developers are encouraged to use stable and unique identifiers and understand the implications of explicit and structural identities on their code.

This summary focuses on the critical aspects discussed in the tech talk, ensuring a comprehensive understanding of the primary themes and practical implications for SwiftUI developers.

Reviewer #2 (Public review):

Summary:

This study examines how activating specific G protein-coupled receptors (GPCRs) affects the microRNA (miRNA) profiles within extracellular vesicles (EVs). The authors seek to identify whether different GPCRs produce unique EV miRNA signatures and what these signatures could indicate about downstream cellular processes and pathological processes.

Methods:

(1) Used U2OS human osteosarcoma cells, which naturally express multiple GPCR types.

(2) Stimulated four distinct GPCRs (ADORA1, HRH1, FZD4, ACKR3) using selective agonists.

(3) Isolated EVs from culture media and characterized them via size exclusion chromatography, immunoblotting, and microscopy.

(4) Employed qPCR-based miRNA profiling and bioinformatics analyses (e.g., KEGG, PPI networks) to interpret expression changes.

Key Findings:

(1) No significant change in EV quantity or size following GPCR activation.

(2) Each GPCR triggered a distinct EV miRNA expression profile.

(3) miRNAs differentially expressed post-stimulation were linked to pathways involved in cancer, insulin resistance, neurodegenerative diseases, and other physiological/pathological processes.

(4) miRNAs such as miR-550a-5p, miR-502-3p, miR-137, and miR-422a emerged as major regulators following specific receptor activation.

Conclusions:

The study offers evidence that GPCR activation can regulate intercellular communication through miRNAs encapsulated within extracellular vesicles (EVs). This finding paves the way for innovative drug-targeting strategies and enhances understanding of drug side effects that are mediated via GPCR-related EV signaling.

Strengths:

(1) Innovative concept: The idea of linking GPCR signaling to EV miRNA content is novel and mechanistically important.

(2) Robust methodology: The use of multiple validation methods (biochemical, biophysical, and statistical) lends credibility to the findings.

(3) Relevance: GPCRs are major drug targets, and understanding off-target or systemic effects via EVs is highly valuable for pharmacology and medicine.

Weaknesses:

(1) Sample Size & Scope: The analysis included only four GPCRs. Expanding to more receptor types or additional cell lines would enhance the study's applicability.

(2) Exploratory Nature: This study is primarily descriptive and computational. It lacks functional validation, such as assessing phenotypic effects in recipient cells, which is acknowledged as a future step.

(3) EV heterogeneity: The authors recognize that they did not distinguish EV subpopulations, potentially confounding the origin and function of miRNAs.

Author response:

Reviewer #1 (Public review):

Summary:

In this manuscript, the authors explore a novel concept: GPCR-mediated regulation of miRNA release via extracellular vesicles (EVs). They perform an EV miRNA cargo profiling approach to investigate how specific GPCR activations influence the selective secretion of particular miRNAs. Given that GPCRs are highly diverse and orchestrate multiple cellular pathways - either independently or collectively - to regulate gene expression and cellular functions under various conditions, it is logical to expect alterations in gene and miRNA expression within target cells.

Strengths:

The novel idea of GPCRs-mediated control of EV loading of miRNAs.

Weaknesses:

Incomplete findings failed to connect and show evidence of any physiological parameters that are directly related to the observed changes. The mechanical detail is lacking.

We appreciate the reviewer's acknowledgment of the novelty of this study. We agree with the reviewer that further mechanistic insights would strengthen the manuscript. The mechanisms by which miRNA is sorted into EVs remain poorly understood. Various factors, including RNA-binding protein, sequence motifs, and cellular location, can influence this sorting process(Garcia-Martin et al., 2022; Liu & Halushka, 2025; Villarroya-Beltri et al., 2013; Yoon et al., 2015). Ago2, a key component of the RNA-induced silencing complexes, binds to miRNA and facilitates miRNA sorting. Ago2 has been found in the EVs and can be regulated by the cellular signaling pathway.  For instance, McKenzie et al. demonstrated that KRAS-dependent activation of MEK-ERK can phosphorylate Ago2 protein, thereby regulating the sorting of specific miRNAs into EVs(McKenzie et al., 2016). In the differentiated PC12 cells, Gαq activation leads to the formation of Ago2-associated granules, which selectively sequester unique transcripts(Jackson et al., 2022). Investigating GPCR, G protein, and GPCR signaling on Ago2 expression, location, and phosphorylation states could provide valuable insights into how GPCRs regulate specific miRNAs within EVs. We have expanded these potential mechanisms and future research in the discussion section.

The manuscript falls short of providing a comprehensive understanding. Identifying changes in cellular and EV-associated miRNAs without elucidating their physiological significance or underlying regulatory mechanisms limits the study's impact. Without demonstrating whether these miRNA alterations have functional consequences, the findings alone are insufficient. The findings may be suitable for more specialized journals.

Thank you for the feedback. We acknowledge that validating the target genes of the top candidate miRNAs is an important next step. In response to the reviewer's concerns, we have expanded the discussion of future research in the manuscript. Although this initial study is primarily descriptive, it establishes a novel conceptual link between GPCR signaling and EV-mediated communication.

Furthermore, a critical analysis of the relationship between cellular miRNA levels and EV miRNA cargo is essential. Specifically, comparing the intracellular and EV-associated miRNA pools could reveal whether specific miRNAs are preferentially exported, a behavior that should be inversely related to their cellular abundance if export serves a beneficial function by reducing intracellular levels. This comparison is vital to strengthen the biological relevance of the findings and support the proposed regulatory mechanisms by GPCRs.

We appreciate the valuable suggestions from the reviewer. EV miRNA and cell miRNAs may exhibit distinct profiles as miRNAs can be selectively sorted into or excluded from EVs(Pultar et al., 2024; Teng et al., 2017; Zubkova et al., 2021). Investigating the difference between cellular miRNA levels and EV miRNA cargo would provide insight into the mechanism of miRNA sorting and the functions of miRNAs in the recipient cells. The expression of the cellular miRNAs is a highly dynamic process. To accurately compare the miRNA expression levels, profiling of EV miRNA and cellular miRNA should be conducted simultaneously. However, as a pilot study, we were unable to measure the cellular miRNAs without conducting the entire experiment again.

Reviewer #2 (Public review):

Summary:

This study examines how activating specific G protein-coupled receptors (GPCRs) affects the microRNA (miRNA) profiles within extracellular vesicles (EVs). The authors seek to identify whether different GPCRs produce unique EV miRNA signatures and what these signatures could indicate about downstream cellular processes and pathological processes.

Methods:

(1) Used U2OS human osteosarcoma cells, which naturally express multiple GPCR types.

(2) Stimulated four distinct GPCRs (ADORA1, HRH1, FZD4, ACKR3) using selective agonists.

(3) Isolated EVs from culture media and characterized them via size exclusion chromatography, immunoblotting, and microscopy.

(4) Employed qPCR-based miRNA profiling and bioinformatics analyses (e.g., KEGG, PPI networks) to interpret expression changes.

Key Findings:

(1) No significant change in EV quantity or size following GPCR activation.

(2) Each GPCR triggered a distinct EV miRNA expression profile.

(3) miRNAs differentially expressed post-stimulation were linked to pathways involved in cancer, insulin resistance, neurodegenerative diseases, and other physiological/pathological processes.

(4) miRNAs such as miR-550a-5p, miR-502-3p, miR-137, and miR-422a emerged as major regulators following specific receptor activation.

Conclusions:

The study offers evidence that GPCR activation can regulate intercellular communication through miRNAs encapsulated within extracellular vesicles (EVs). This finding paves the way for innovative drug-targeting strategies and enhances understanding of drug side effects that are mediated via GPCR-related EV signaling.

Strengths:

(1) Innovative concept: The idea of linking GPCR signaling to EV miRNA content is novel and mechanistically important.

(2) Robust methodology: The use of multiple validation methods (biochemical, biophysical, and statistical) lends credibility to the findings.

(3) Relevance: GPCRs are major drug targets, and understanding off-target or systemic effects via EVs is highly valuable for pharmacology and medicine.

Weaknesses:

(1) Sample Size & Scope: The analysis included only four GPCRs. Expanding to more receptor types or additional cell lines would enhance the study's applicability.

We are encouraged that the reviewer recognized the novelty, methodological rigor, and significance of our work. We recognize the limitations of our current model system and emphasize the need to test additional GPCR families and cell lines in the future studies, as detailed in the discussion section.

(2) Exploratory Nature: This study is primarily descriptive and computational. It lacks functional validation, such as assessing phenotypic effects in recipient cells, which is acknowledged as a future step.

We appreciate the feedback. We recognize the importance of validating the function of the top candidate miRNAs in the recipient cells, and this will be included in future studies. 

(3) EV heterogeneity: The authors recognize that they did not distinguish EV subpopulations, potentially confounding the origin and function of miRNAs.

Thank you for the comment. EV isolation and purification are major challenges in EV research. Current isolation techniques are often ineffective at separating vesicles produced by different biogenetic pathways. Furthermore, the lack of specific markers to differentiate EV subtypes adds to this complexity. We recognize that the presence of various subpopulations can complicate the interpretation of EV cargos. In our study, we used a combined approach of ultrafiltration followed by size-exclusion chromatography to achieve a balance between EV purity and yield. We adhere to the MISEV (Minimal Information for Studies of Extracellular Vesicles 2023) guidelines by reporting detailed isolation methods, assessing both positive and negative protein markers, and characterizing EVs by electron microscopy to confirm vesicle structure, as well as nanoparticle tracking analysis to verify particle size distribution(Welsh et al., 2024). By following these guidelines, we can ensure the quality of our study and enhance the ability to compare our findings with other studies.

Reviewer #1 (Public review):

Summary:

Felipe and colleagues try to answer an important question in Sarbecovirus Orf9b-mediated interferon signaling suppression, given that this small viral protein adopts two distinct conformations, a dimeric β-sheet-rich fold and a helix-rich monomeric fold when bound by Tom70 protein. Two Orf9b structures determined by X-ray crystallography and Cryo-EM suggest an equilibrium between the two Orf9b conformations, and it is important to understand how this equilibrium relates to its functions. To answer these questions, the authors developed a series of ordinary differential equations (ODE) describing the Orf9b conformation equilibrium between homodimers and monomers binding to Tom70. They used SPR and a fluorescent polarization (FP) peptide displacement assay to identify parameters for the equilibrium and create a theoretical model. They then used the model to characterize the effect of lipid-binding and the effects of Orf9b mutations in homodimer stability, lipid binding, and dimer-monomer equilibrium. They used their model to further analyze dimerization, lipid binding, and Orf9b-Tom70 interactions for truncated Orf9b, Orf9b fusion mutant S53E (blocking Tom70 binding), and Orf9b from a set of Sars-CoV-2 VOCs. They evaluated the ability of different Orf9b variants for binding Tom70 using Co-IP experiments and assessed their activity in suppressing IFN signaling in cells.

Overall, this work is well designed, the results are of high quality and well-presented; the results support their conclusions.

Strengths:

(1) They developed a working biophysical model for analyzing Orf9b monomer-dimer equilibrium and Tom70 binding based on SPR and FP experiments; this is an important tool for future investigation.

(2) They prepared lipid-free Orf9b homodimer and determined its crystal structure.

(3) They designed and purified obligate Orf9b monomer, fused-dimer, etc., a very important Orf9b variant for further investigations.

(4) They identified the lipid bound by Orf9b homodimer using mass spectra data.

(5) They proposed a working model of Orf9b-Tom70 equilibrium.

Weaknesses:

(1) It is difficult to understand why the obligate Orf9b dimer has similar IFN inhibition activity as the WT protein and obligate Orf9b monomer truncations.

(2) The role of Orf9b homodimer and the role of Orf9b-bound lipid in virus infection, remains unknown.

Comments on revisions:

In the revised manuscript, the authors have addressed my concerns.

Author response:

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

Reviewer #1 (Public review):

Summary:

Felipe and colleagues try to answer an important question in Sarbecovirus Orf9b-mediated interferon signaling suppression, given that this small viral protein adopts two distinct conformations, a dimeric β-sheet-rich fold and a helix-rich monomeric fold when bound by Tom70 protein. Two Orf9b structures determined by X-ray crystallography and Cryo-EM suggest an equilibrium between the two Orf9b conformations, and it is important to understand how this equilibrium relates to its functions. To answer these questions, the authors developed a series of ordinary differential equations (ODE) describing the Orf9b conformation equilibrium between homodimers and monomers binding to Tom70. They used SPR and a fluorescent polarization (FP) peptide displacement assay to identify parameters for the equilibrium and create a theoretical model. They then used the model to characterize the effect of lipid-binding and the effects of Orf9b mutations in homodimer stability, lipid binding, and dimer-monomer equilibrium. They used their model to further analyze dimerization, lipid binding, and Orf9b-Tom70 interactions for truncated Orf9b, Orf9b fusion mutant S53E (blocking Tom70 binding), and Orf9b from a set of Sars-CoV-2 VOCs. They evaluated the ability of different Orf9b variants for binding Tom70 using Co-IP experiments and assessed their activity in suppressing IFN signaling in cells.

Overall, this work is well designed, the results are of high quality and well-presented; the results support their conclusions.

We thank reviewer #1 for their thoughtful assessment of our work and their constructive feedback.

Strengths:

(1) They developed a working biophysical model for analyzing Orf9b monomer-dimer equilibrium and Tom70 binding based on SPR and FP experiments; this is an important tool for future investigation.

(2) They prepared lipid-free Orf9b homodimer and determined its crystal structure.

(3) They designed and purified obligate Orf9b monomer, fused-dimer, etc., a very important Orf9b variant for further investigations.

(4) They identified the lipid bound by Orf9b homodimer using mass spectra data.

(5) They proposed a working model of Orf9b-Tom70 equilibrium.

Weaknesses:

(1) It is difficult to understand why the obligate Orf9b dimer has similar IFN inhibition activity as the WT protein and obligate Orf9b monomer truncations.

We thank the reviewer for their observation and agree that the obligate homodimer IFN results were not what we expected to observe given our FP kinetic results with the purified obligate homodimer and noted our surprise in the discussion. We also note that we have two possible hypotheses for why this is the case.

In our discussion, we noted the possible introduction of an increased avidity effect with fused homodimer and have improved it as follows with additions in red:

“This result was unexpected as we had anticipated the obligate homodimer results to resemble the phosphomimetic. We hypothesize that this may be explained by two possible factors. First, we can’t exclude the introduction of an increased avidity between Orf9b and Tom70 when using the fused homodimer. Although our modeled decrease in the association rate of Orf9b:Tom70 (which increases the K_D of the complex) suggests that fusing two copies of Orf9b decreases the affinity to Tom70, one copy of the fusion construct could also be capable of either binding to two copies of Tom70, or, one copy of the fusion could undergo rapid rebinding to Tom70. These effects would lead to a much tighter interaction in cellular assays than we modeled in vitro. A second possible explanation is that our assumptions about high lipid binding are not valid for cell based assays.”

We also noted that a second possible explanation is due to our limitations in isolating the apo-fused homodimer to compare to the lipid-bound fused homodimer and possible differences this could have on our assays and briefly expanded upon this. Again, we improved this with additions in red:

“As we have shown with both WT and fusion constructs, recombinantly expressed and purified Orf9b is lipid-bound and this can stabilize the homodimer to slow or inhibit the binding to Tom70. For the Orf9b fusion construct, we attempted to isolate the lipid-free species through protein refolding as previously described to compare the effect of lipid-binding on the homodimer fusion (similar to our WT experiments); however, we could not recover the stably folded homodimer. We hypothesize that the discrepancy between our kinetic results and Co-IP/IFN results could be due to subsaturation of the Orf9b fusion homodimers by lipids in cell based assays. While we have shown that lipid-binding occurs in recombinant expression systems, it is possible that in our cell based signaling assays that lipid-binding only affects a minor population of Orf9b. Given that we were unable to isolate the apo-fusion homodimer, we could not directly compare whether there are differences in fusion homodimer stability in the presence or absence of lipid-binding. Therefore, it is possible that the apo-fusion homodimer undergoes unfolding and refolding into alpha helices that lead to Tom70 binding similar to the WT construct.”

(2) The role of Orf9b homodimer and the role of Orf9b-bound lipid in virus infection, remains unknown.

We agree that we did not try to directly test for the role of the homodimer during infection and this remains an open area of exploration for future studies. We have included this caveat in our discussion but suggested possible experiments and future directions that could help shed light on this:

“Although we have not directly tested for the role the homodimer conformation plays during infection, we have demonstrated that lipid-binding to the homodimer can bias the equilibrium away from Tom70. Lipids including palmitate have been shown to act as both a signaling molecule as well as a post-translational modification during antiviral innate immune signaling (S Mesquita et al. 2024; Wen et al. 2022; S. Yang et al. 2019). As a post-translational modification (referred to as S-acylation), MAVS, a mitochondrial type 1 IFN signaling protein that associates with Tom70 (X.-Y. Liu et al. 2010; McWhirter, Tenoever, and Maniatis 2005; Seth et al. 2005), has been shown to be post-translationally palmitoylated which affects its ability to localize to the mitochondrial outer membrane during viral infection and is a known target of Orf9b (Bu et al. 2024; Lee et al. 2024). When this is impaired (either by mutation or by depletion of the palmitoylation enzyme ZDHHC24), IFN activation is impaired (Bu et al. 2024). Therefore, future investigations should consider if the homodimer conformation of Orf9b is capable of antagonizing other IFN signaling factors such as MAVS by binding to palmitoyl groups. Indeed, Orf9b has already been shown to be capable of binding to MAVS by Co-IP (Han et al. 2021), however, whether or not this occurs through the palmitoyl modification remains unknown.”

Reviewer #2 (Public review):

Summary:

This study focuses on Orf9b, a SARS-COV1/2 protein that regulates innate signaling through interaction with Tom70. San Felipe et al use a combination of biophysical methods to characterize the coupling between lipid-binding, dimerization, conformational change, and protein-protein-interaction equilibria for the Orf9b-Tom70 system. Their analysis provides a detailed explanation for previous observations of Orf9b function. In a cellular context, they find other factors may also be important for the biological functioning of Orf9b.

Strengths:

San Felipe et al elegantly combine structural biology, biophysics, kinetic modelling, and cellular assays, allowing detailed analysis of the Orf9b-Tom70 system. Such complex systems involving coupled equilibria are prevalent in various aspects of biology, and a quantitative description of them, while challenging, provides a detailed understanding and prediction of biological outcomes. Using SPR to guide initial estimates of the rate constants for solution measurements is an interesting approach.

Weaknesses:

This study would benefit from a more quantitative description of uncertainties in the numerous rate constants of the models, either through a detailed presentation of the sensitivity analysis or another approach such as MCMC. Quantitative uncertainty analysis, such as MCMC is not trivial for ODEs, particularly when they involve many parameters and are to be fitted to numerous data points, as is the case for this study. The authors use sensitivity analysis as an alternative, however, the results of the sensitivity analysis are not presented in detail, and I believe the authors should consider whether there is a way to present this analysis more quantitatively. For example, could the residuals for each +/-10% parameter change for the peptide model be presented as a supplementary figure, and similarly for the more complex models? Further details of the range of rate constants tested would be useful, particularly for the ka and kB parameters.

We thank the reviewer for their constructive feedback and have generated supplemental figures providing a deeper analysis of the residuals for each model parameter adjusted +/- 10% from the reported values which we have added to our supplemental figures as Figure 1 - Supplemental 3 and Figure 4 - Supplemental 5  .

We note that there are modest improvements in residual plots where model parameters are individually lowered by 10% from their reported value when considering this single dataset, however, our choice of using the reported values was driven by finding values that were suitable for improving model behavior across multiple concentration series in different datasets. Specifically, we have also included the RMSD values for each model parameter subjected to a +/-10% change from a single concentration time course as well as the percent change in RMSD relative to the RMSD generated by our reported model parameters to illustrate this. We have also included text that makes note of the observed pattern in the residuals from Figure 4 - Supplement 5 and provided some explanations for why this may occur.

“Inspection of the residuals from the 5uM apo-Orf9b homodimer time course showed clear patterns when individual model parameters were subjected to a 10% increase or decrease from the reported values. While our proposed model qualitatively describes the concentration dependent change in kinetic behavior, the residual plots may suggest that additional binding reactions may also be occurring that are not captured by our model.”

Figure 1 - Supplemental 3. Plots of residuals from Orf9b peptide model showing effect of an increase or decrease by 10% on each model parameter. All residuals and reporting are with respect to the100uM of unlabeled Orf9b peptide condition. Blue dots: reported value. Red dots: 10% increase in reported value. Green dots: 10% decrease in reported value. Table reporting of RMSD values for model fitsafter +/-10% change to model parameter (Left column) and percent change in RMSD relative to reported model RMSD (Right column).

“As an alternative to attempting to place CIs on the parameters, we performed sensitivity analysis to determine which parameters the model was most sensitive to (see methods and Figure 1 - Supplemental 3). Additionally, we note that the model parameters were derived from the fit of only one concentration (100uM), but fit the other concentrations equally well. We observed that the model parameter that was most sensitive to change was the rate of Orf9b-FITC:Tom70 ([PT]) dissociation when subjected to a 10% increase or decrease whereas all other model parameters showed no sensitivity to change (Figure 1 - Supplemental 3).”

Figure 4 - Supplemental 5: Plot of residuals showing the effect of increasing or decreasing individual model parameters 10% compared to the reported values. All residual plots are with respect to the 5uM apo-Orf9b homodimer condition. Blue dots: reported value. Red dot: 10% increase in reported value. Green dot: 10% decrease in reported value. (Left columns) Table of RMSD values calculated from model fits showing the effect of both +/-10% change to individual model parameters. (Right columns) Percent change in RMSD values subjected to +/-10% change for individual model parameters relative to the RMSD of the reported model.

We have also included the following revised text to accompany this figure.

“Further, we repeated the sensitivity analysis described previously for the peptide model and also considered the sensitivity of model parameters by inspecting each individually (Figure 4- figure supplemental 5). We found that when examining the residuals of the lowest concentration of 5uM, the model was most sensitive to changes in three parameters: the rate of homodimer association and dissociation and the conversion from β to α-monomers.”

“Therefore, under low concentrations of Orf9b homodimer, binding to Tom70 is limited by the rate of homodimer association and dissociation as well as the conversion of Orf9b monomers to the α-helical conformation.”

We have also included a supplemental figure showing how changes in the model parameters ka and kB affect the models behavior to help illustrate the range of values tested as Figure 4 - Supplemental 4.

Figure 4 - Supplemental 4: Plots of model behavior showing the effect of changes to alpha-beta and beta-alpha monomer  interconversion rates compared to experimental values. Data is modeled with respect to the apo-Orf9b homodimer 5uM condition. Black line represents reported model fit and values used.

We have also incorporated the following revised text.

“The model parameters k_a and k_B describe the rate of interchange between the β-sheet and α-helix monomer conformations. These parameters must be estimated by modeling because our assays do not allow us to directly measure the folding rates between these conformations. To identify these values, we performed a scan of k_a and k_B values that yielded the best agreement between the model and the experimental conditions (Figure 4 - figure supplemental 4).”

The authors build a model that incorporates an α-helix-β-sheet conformational change, but the rate constant for the conversion to the α-helix conformation is required to be second order. Although the authors provide some rationale, I do not find this satisfactorily convincing given the large number of adjustable parameters in the model and the use of manual model fitting. The authors should discuss whether there is any precedence for second-order rate constants for conformational changes in the literature. On page 14, the authors state this rate constant "had to be non-linear in the monomer β-sheet concentration" - how many other models did the authors explore? For example, would αT↔α↔αα↔ββ (i.e., conformational change before dimer dissociation) or α↔βαT↔ββ (i.e., Tom70 binding driving dimer dissociation) be other plausible models for the conformational change that do not require assumptions of second-order rate constants for the conformational change?

We thank the reviewer for their feedback. During our studies, we tested several models prior to the final one presented in Figure 4A. The first model that we tested as described in Figure 4 - Supplemental 3 described ββ↔α↔αT with no conformational change. We tested several models that integrated the existing structural data for both Orf9b and Tom70 and found that while these models could fit individual time series, they did not explain the concentration dependent changes in subsequent time series nor did they explain changes induced by lipid-binding and mutations in VOC.

With respect to the possibilities of αT↔α↔αα↔ββ and α↔βαT↔ββ models, we have revised our manuscript to mention that we did test additional models before we settled on the model that we presented.

“We tested different reaction schemes that incorporated the interconversion between β-sheet to α-helix conformations by considering models that described a conformational change in the homodimer leading to Tom70 binding rather than monomers. None of these models adequately described our experimental results, therefore we continued developing our model as outlined in Figure 4D”

With respect to the second-order rate describing the fold change from β to α, we have added the revised text to the manuscript:

“We initially tested the impact of keeping the rate constant k_a first order, just like k_B which did yield the sigmoidal behavior we observed in the 5uM apo-homodimer condition. However, this assumption failed to describe the data at other concentrations resulting in a substantial overestimation compared to our experimental results when holding k_B at a constant value throughout. We found that when the β-sheet to α-helix rate (k_a ) was made a second order rate constant, we were able to hold the rate constant across all concentrations tested suggesting a non-linearity in the monomer β-sheet concentration.”

While this was surprising to us, we reasoned that a biological explanation for why the conversion from β to α was second order was that the β-monomers may transiently self-associate to cooperatively fold into the α-helical conformation. We did acknowledge this choice to make the β to α parameter non-linear (unlike the α to β conversion which was single order).

We concede that we could not find specific examples describing non-linear kinetics comparable to the system we described in literature, however, such systems have been reported for proteins that exhibit high structural plasticity where transient interactions with another copy of the protein or another protein altogether drive folding changes and we have revised this manuscript to include some additional citations to papers that describe such systems (Zuber et al. 2022; Tuinstra et al. 2008).

Overall, this study progresses the analysis of coupled equilibria and provides insights into Orf9b function.

Reviewer #1 (Recommendations for the authors):

(1) What was the unlabeled Orf9b peptide is added to the pre-equilibrated Orf9b-FITC:Tom70 solution as a competitor? Figure 1D illustrates that the competitor was full-length Orf9b.

We have revised the figure to illustrate that in this experiment, the competitor is the unlabeled FITC peptide and not the full length Orf9b sequence

(2) Figure 2B, what is the higher Mw peak from refolded Orf9b homodimer.

We have added the following revised text (highlighted in red) to the manuscript to clarify Figure 2B.

“The SEC elution profile and retention volume of refolded Orf9b directly overlapped with natively folded homodimeric Orf9b and suggested a high recovery of the refolded homodimer with the early eluting peaks corresponding to either a chaperone-bound species (natively folded) or misfolded protein (refolded) as judged by SDS-PAGE (Figure 2B). Together, the overlap in elution peaks corresponding to the folded homodimer suggested a high recovery of the homodimer from the refolding conditions.”

(3) Figure 2C, in the main text, the authors state that "...observed that the refolded homodimer structure closely aligned with the lipid-bound reference structure, which shows that the homodimer fold can be recovered after denaturing". Please provide structural comparison details here, software used? Rmsd and Dali Z-score.

We have added the following revised text (highlighted in red) to the manuscript to clarify Figure 2C.

“Aligning the structure of the Orf9b homodimer (PDB 6Z4U) with our structure of the refolded Orf9b homodimer (9N55) in Pymol resulted in an RMSD of 1.1Å. Further, we also searched our structures of the refolded Orf9b homodimer on the Dali server against the existing structures of the lipid-bound Orf9b homodimer which yielded a Z-score of 2.2 which shows good correspondence between the structures.”

(4) To prove the refolded Orf9b homodimer did not contain lipid, could the authors provide mass spectra data for the refolded Orf9b sample and compare it with the results in Figure 2 - Supplemental 1.

We do not have complete mass spectra data for the refolded homodimer samples, however, we feel that the native mass spectrometry data provides a good orthogonal comparison between natively folded and refolded samples for the presence or absence of lipids. We concede that we only used mass spectrometry to characterize the four peaks that were unique to the natively folded deconvoluted spectra which confirmed that shift in mass relative to the expected homodimer molecular weight corresponded to the two lipids we presented. However, we would expect that performing mass spectrometry on the refolded sample would only further confirm our observations from the crystal structures and the native mass spectrometry.

(5) Have the authors tried to use analytical ultracentrifugation to analyze the Orf9b dimer-monomer equilibrium, given that AUC provides a much more accurate measurement of molecular mass?

We thank the reviewer for this suggestion and agree that AUC could be an additional useful strategy for monitoring the dimer-monomer equilibrium and provide additional validation of the molecule weights of both the monomer and homodimer.

While we have not performed AUC, we have revised our manuscript to include more discussion about the determination of molecular weights by SEC.

“For the Orf9b homodimer, the retention volume was consistent with molecular weight standards based on the expected molecular weight of the homodimer (~21kDa) and the standard (~29kDa). In the case of the Orf9b monomer, although we would expect the retention volume of the monomer (~10.6kDA) to be between the molecular weight standards of 13.4kDa and 6.5kDa, the greater retention volume could be explained by non-specific hydrophobic interactions between the monomeric Orf9b and the column.”

(6) The authors used truncation of 7 C-terminal amino acids to generate an obligate Orf9b monomer for their assays. It would be interesting to mutate residues at the homodimer interface to generate Orf9b monomers rather than deleting residues. For example, mutate 91-96aa (FVVVTV) to negatively charged residues, which will not only disrupt the dimerization interface, but also impair lipid binding. The dimer interface mutant should then be tested in their SPR, FP assays, as well as IFN inhibition assays.

We thank the reviewer for their suggestion and agree that mutation of the 7 C-terminal amino acids into negatively charged residues could be an interesting alternative strategy to generating an obligate Orf9b monomer without the need for truncating the residues. Our choice of using the truncated construct we proposed was driven by our analysis of the structure of the homodimer which reveals that a significant portion of the dimer interface is composed of backbone-backbone hydrogen bonding between the two chains of Orf9b. We reasoned that truncating these residues would be the most effective way to compromise the interface between the two chains and drive a predominantly monomeric behavior, however, compromising the interface with multiple mutations is an intriguing alternative.

Reviewer #2 (Recommendations for the authors):

(1) The authors could comment on the slow monomer-dimer exchange observed by SEC and how it fits with their other analysis.

We thank the reviewer for their comment and concede that the slow exchange may be a limitation of this experimental setup. Our observations from our SPR experiments and modeling showed us that the homodimer may be fast to dissociate into monomer given the off rate which would suggest a half-life for the homodimer to be on the order of seconds, however, we still observe a noticeable dimer species on the chromatograms. We initially allowed the diluted samples to reach equilibrium prior to injection onto the analytical sizing column, however, it is possible that the system is still in a pre-equilibrium prior to injection onto the column. This could be driven by interactions between the protein and the column that prevents full dissociation of the homodimer. While this is a limitation, we note that we did not use the Kd value that we determined by non-linear regression fitting to the equilibrium observed on the chromatograms for downstream experiments but instead used the value to get a ballpark estimate for the homodimer Kd which is on the same order as the Kd determined by SPR.

(2) It might be useful to include the rate constants on the reaction arrows of the schematic representation of the models.

We have revised Figure 4D to include the rates for both Orf9b monomer binding to Tom70 and Orf9b binding to Orf9b as derived from the SPR experiments as well as the modeled values for the interconversion between α and β monomers. We also revised Figure 7 to include these values as well as the modeled dissociation rate for homodimer when lipid-bound.

(3) I couldn't find how the sensitivity analysis was performed for the more complex models. Was this the same +/- 10% as per the peptide model?

We used the same +/- 10% sensitivity analysis for the peptide model in the more complex equilibrium model and have revised our manuscript to clearly reflect that.

(4) Further clarification of "inspection of residuals suggested that the fits were accurate". In Figure 1B, the residues look to have systematic errors, perhaps indicating other processes occurring.

We agree that in the SPR kinetic fitting results for the Orf9b peptide binding to Tom70 in Figure 1B that there are some regions where the fit over or under estimates the experimental results. This is partially the result of limitations in the number of different binding models that we can fit in the analysis software which is why we reported using a 1:1 langmuir binding model. It is certainly possible that there may be some additional binding reactions that occur, however, we limited our use of these specific kinetic results to the peptide model that we proposed in Figure 1D. We did note in the manuscript text that it was necessary for us to change the model parameter values to some extent in order to fit our experimental results which may be partially explained by the SPR fitting errors.

“With the parameter set obtained from the 100µM condition, we then held all parameters fixed and simply changed the peptide concentrations in the model to fit the remaining conditions by hand. We note that this process saw the model parameter values change between 3% at the lowest end up to 70% at the highest end from the experimentally derived values but remained within an order of magnitude of the experimental SPR values. We speculate that this arises due to the differences in experimental setup between SPR and FP-based methods of measuring kinetics.”

(5) The manuscript builds logically, but given the sophisticated nature of the system and the modelling could benefit from more clarity/streamlining in the descriptions/illustrations.

We have revised our manuscript in response to both reviewers comments and hope that the clarity of the work is improved as a result.

(6) Figure 4 Supplement 3 - where did the rate constants for Model 1 come from? Was there any attempt to alter them to fit the data better?

We have clarified in the figure description that the rate constants used in Model 1 were the same values used in Figure 4B (but without the interconversion between beta and alpha rates).

“Comparison of kinetic model 1 and 2 in describing experimental results from the kinetic binding assay. Experimental results using 10uM of refolded Orf9b homodimer are shown as rings with the predicted behavior of model 1 (equilibrium exchange) shown as a dark blue line. The predicted behavior of model 2 (equilibrium exchange with a conformational change between β-sheet and ɑ-helical monomers) is shown as the light blue line. Model parameter values were the same as described in Figure 4D and kept constant in both model comparisons.”

(7) What are and [PT] in the second set of equations (page 13)?

[‘PT] refers to the concentration of “fluorescent probe” (Orf9b-FITC) and Tom70.

(8) "Additionally, the fused homodimer association rate (which can be viewed as a rate of tertiary complex formation)" - can the authors provide a mathematical proof for this?

In the case of the fused homodimer kinetic data, we did not develop a separate model to explicitly take into account the differences between using a fused construct versus the WT construct that can dissociate into monomers. We have clarified our interpretation of this in the manuscript.

“Although our model explicitly describes homodimer dissociation into monomers as a requisite step for Orf9b binding to Tom70, we adapted it for the fusion experimental data. In this case, all model parameters other than the association and dissociation kinetics of the fluorescent probe and Tom70 were adjusted to achieve the best agreement with the experimental data. When applied to the fusion homodimer, the parameters describing homodimer dissociation into separate monomers could instead describe the dissociation of the two β-sheet domains away from each other in the tertiary structure but remaining physically linked through the linker region.”

(9) "For Lambda and Omicron, the P10S mutation results in the serine being positioned to form several hydrogen bonds between R13 and the backbone carbonyl of A11 and L48 within the same chain..." is this taken from AlphaFold predicted structures of the mutants? If so, it should be made clear that this is derived from predicted structures. And even so, AlphaFold can be poor at determining structures of mutants, and so there is greater uncertainty in the prediction of the bonds.

For Lambda, Omicron, and Delta mutations, we used Pymol to examine how the placement of mutations could structurally explain the kinetic differences we observed in our model. We have gone back and clarified in the figure description that these predictions are not derived from AlphaFold.

(10) "biological replicates" - is this different protein purifications?

Yes, in this case biological replicates refer to different protein purifications for all variants described and tested.

(11) Are any of the authors involved in the Berkeley Madonna commercial software used in the manuscript? If so, should this be in the conflict of interest statement?

Yes, Michael Grabe is an owner of Berkeley Madonna, and we have updated our conflicts of interest statement to reflect this.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Summary:

Cells need to adjust their gene expression pattern, including nutrient transporters and enzymes to process the available nutrient. How cells maintain the coordination between these processes is one of the most critical questions in biology. In this work authors elegantly combined a range of relevant experimental techniques, ranging from time-lapse microscopy, microfluidics, and mathematical modelling to address this question. Combining these methods, authors proposed a push-pull like mechanism, involving two pairs of repressors (Mth1, Std1 and Migs) in the glucose sensing network. In budding yeast there are multiple hexose transporter genes with varying affinity and transport rate. Authors postulated that on sensing glucose, cells switch between expressing high affinity glucose transporters (when extracellular glucose is low), and low affinity glucose transporters (in high extracellular glucose), and these processes are mediated by the pairs of repressors as mentioned earlier. Following the expressing patterns of fluorescently tagged hexose transporters and varying the extracellular glucose concentrations in media, authors proposed that pairs of repressors switch their activity depending on extracellular glucose level, and which is matched by the promoters of the hexose transporter genes to achieve optimality of glucose transport.

This study is elegantly designed and addressed an interesting question. The mechanism (push-pull involving two pairs of repressors) is plausible and justified by the data. Authors also presented a mathematical model and made predictions, which are also verified. We will recommend the publication of this work with minor modifications.

Major comments:

This study is well designed and experiments performed accordingly. We have only minor comments for revision.

Minor comments:

1. Although authors covered a wide array of literature, but while discussing tradeoffs and nutrient sensing, it will be good to include bacterial growth law and related literature, and physiological level tradeoffs should be discussed. Moreover, authors vouched that the push-pull mechanism helps to circumvent the rate-affinity tradeoff of the transporter, whereas expressing genes to more precisely corelate with the extracellular glucose level brings out physiological optimality. This rate-affinity tradeoff and its physiological role should be discussed clearly.
2. Authors described the ALCATRAS device in their previous publication, but for better clarity, a supplementary figure with schematic diagram and experimental plan should be included.
3. Microscopic images of transporter expression pattern should be shown as kymographs in the supplementary, in this version of the manuscript plots from processed microscopy images are shown only.
4. GFP was used to tag HXT1-7 as mentioned by the authors and expression of these genes are evaluated in separate experiments. We suggest including a schematic diagram describing the experimental design while using the microfluidic device and the experimental plan should be written in more detail in general. We found this part confusing. Did authors considered tagging two separate transporters with different fluorescent tag from either end of the affinity spectrum and showing the expression pattern in one experiment? Authors mentioned co expression of receptors at a particular glucose concentration over time, is this inferred from separate timelapse experiments? This need to be more clearly stated.
5. Please mark the second phase of media glucose concentration in panel 1C, 1% glucose phase is marked, please mark the other phases for clarity.
6. For the repressors to sense glucose and to initiate the push pull mechanism, there should be baseline glucose flux, which is not clearly mentioned in the manuscript. Authors mentioned that minimal intracellular glucose in absence of extracellular glucose and deployed a logistic function to increase intracellular glucose. The baseline glucose level is crucial, and authors should comment on this. Also, glucose mediated protection of HXT4 should be discussed in this context.
7. Figure 3B and 3C, details of the error bars should be mentioned in the figure legend.

Referee cross-commenting

All other reviewers also identified this study insightful and interesting, similar to our comments. We also agree with the suggestions made by other reviewers. Suggested changes and modifications can be addressed within a month as mentioned by most of the reviewers. Excellent point raised by other reviewers on technicalities and addressing those points will improve the readability of this work even more.

Significance

General assessment:

Use of innovative microfluidics platform to trap mother cells and following the gene expression pattern by fluorescence microscopy and combining the experimental approach with mathematical model are the strengths of this work. Whereas the proposed push-pull mechanism is not generalizable to other carbons. Model is merely used to fit the data, rather than making interesting predictions. Also how does the mechanism holds when cells are switched from other nutrient sources is also not clear in this work, which are the limitations of this work.

Advance

This work involves experimental technique and mathematical model to test the hypothesis. Use of custom-built microfluidics set up and live cell imaging to track gene expression levels in varying nutrient condition. This study links single cell level gene expression pattern to model and predict system level behavior. Nutrient sensing and subsequent rearrangement of gene regulatory network is an important question to address, and the proposed push-pull mechanism in this study adds up to the existing body of literature.

Audience:

This work is interdisciplinary and researchers across multiple fields will be interested in this work, including researchers interested in microbial nutrient sensing, systems biology, topology of gene regulatory network, metabolism, and general microbiology.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Summary:

The yeast Saccharomyces cerevisiae possesses a large family of hexose transporters, the HXT genes. Some of these transporters play known roles in transport to feed metabolism while others seem to respond to glucose levels but have differing cellular functions, acting more as sensors than as drivers of carbon and energy production. The authors use single cell fluidics to monitor steady state expression of specific transporters under controlled glucose levels. The authors then used published information on the regulatory network of HXT4 gene expression to predict expression levels and confirm the role of the prior identified regulators. Thus, this work confirms prior work as to the levels of substrate leading to optimized expression of transporters and confirms the role of the identified regulatory network. The fact that the main single cell fluidics findings confirm the prior culture analyses affirms the utility of the prior work.

Major Comments:

1. The analysis uses protein expression levels (HXT4-GFP) as a proxy for transcriptional regulation. This study assumes no regulation of protein expression beyond transcription under steady state conditions. This seems like a reasonable assumption. However, for the dynamic change analysis (Figure 1 C, lines 70-78) loss of GFP-tagged protein from a single cell would be due not just to absence of transcription but also to differential rates of endocytosis and degradation, which could vary across the different HXTs. In cell populations the plasma membrane composition of the bud can be dynamically different from that of the mother cell and will reflect changes in transcription patterns. Meaning that cells with buds might have reduced expression due to the presence of the bud versus non-budding cells. And if buds are washed away during the time course of the experiment this could impact assessment of GFP signal - I am assuming controls were done to address this and should be included in the presentation. Did the authors consider this in their experimental design and interpretation?
2. The modeling was based upon the assumption of the validity of prior work and observations and authors show that models based upon that prior knowledge work to explain the single cell data. One wonders what perturbing prior modes of action would do to fit the data. That is, if the role of one regulator was downplayed or modified in concert with another would data still fit in a reasonable way? My concern again is that loss of signal (protein) is equated exclusively with transcription and not post-transcriptional regulation. This timeline in 1C and in fig 2 of 20 hours certainly would accommodate post-transcriptional regulation of protein levels.
3. Lines 142-150: two models are proposed: Std1 activating Snf1 with std1 deletion therefore hyperactivating Mig1. The second model is for Std1 to repress Mig2 with deletion of std1 then leading to hyperactivation of Mig2. It seems this could be directly tested using multiple deletant strains, or modified repressor proteins. For example, is the effect lost in a std1 mig1 double mutant?
4. Lines 121-122 the comment that comparing expressing GFP from the HXT4 promoter to GFP tagged HXT4 protein allows glucose to protect HXT4 from degradation needs to be explained.
5. Line 180-186: this is an important analysis - I assume binding sites for repressors/inducers of the HXT genes have been mapped -then the comparison to known promoter structure (lines 214-246) is a great test of the model. It seems the finding are consistent with previously published data on differential regulation of these promoters in full-culture studies.
6. Lines 293-299: one thing the authors should highlight is the contrast between these single cell studies and prior population studies that are influenced strongly by the heterogeneity between bud and mother cell plasma membrane composition. The mother cell can of course benefit from the differential expression in the daughter cell and the daughter cell benefits from the differential composition of the mother cell. This study shows that mother cells adapt membrane composition as well, but perhaps the potential role of cell membrane protein turnover should also be included.

no Minor Comments

Significance

It has been known for quite some time that glucose transport in the yeast Saccharomyces cerevisiae is dynamically regulated to optimize sugar depletion to sugar metabolism. This intricate system involves a family of hexose transporters of differing affinities for substrate, the timing and level of expression of which is regulated by both eternal hexose levels and internal ability to metabolize keeping cytoplasmic sugar levels low. Since facilitated diffusion systems can transport in both directions, the consumption of substrate assures the direction of uptake will be dominant. The authors demonstrate in this paper that differential expression of the known major regulators of HXT gene expression work in concert to adjust the expression patterns of transporters of differing affinities leading to optimization of hexose uptake. The study monitored changes in single cells and findings confirm prior work conducted in cell populations. One assumption has always been that the mother cell might "sacrifice" itself by not being able to dynamically clear the membrane of environmentally unmatched hexose transporters relying on the altered membrane composition of the bud. This work's focus on "mother cells" demonstrates that regulation still occurs if cells are allowed to reach a steady state. The timeline may be slower than bud adaptation, but these authors confirm that mother cells respond dynamically to glucose levels.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Summary:

This is a very insightful work showing how to disentangle one of the most complex transcriptional networks in yeast (S. cerevisiae) by combining single-cell dynamics, dynamical-systems modeling, Bayesian-style inference, and genetic perturbations. The authors tackle a problem that has eluded quantitative resolution for over two decades-how yeast regulates its seven primary glucose importer genes (HXT1-HXT7) in response to both steady and temporally changing extracellular [glucose]. Their integrated experimental-theoretical approach delivers the most satisfying mechanistic and quantitative explanation to date, and I enthusiastically recommend this manuscript for publication.

Yeast relies on seven passive hexose transporters (Hxt1-Hxt7) to import glucose, its preferred sugar; deleting all seven abolishes growth on glucose. The underlying regulatory network is exceptionally intricate, reflecting yeast's evolutionary priority for glucose. Two membrane sensors-Snf3 (high affinity) and Rgt2 (low affinity)-detect extracellular glucose and thereby inactivate two co-repressors, Mth1 and Std1, which modulate the DNA-binding factor Rgt1. Concurrently, intracellular glucose activates the SNF1 kinase, phosphorylating and exporting the repressor Mig1, while Mth1/Std1 also govern the transcription and stability of Mig2, another DNA-binding repressor. Together, Rgt1, Mig1, and Mig2 integrate these inputs to control HXT promoter activity (Fig. 2A). Importantly, Mth1 and Std1 do not directly bind to DNA and this complication - the protein-protein interaction that one cannot get from DNA sequence - is just one source of difficulty that the authors overcame.

To map the network's behavior, the authors used microfluidic "cages" housing single cells expressing GFP-tagged HXTs, monitoring fluorescence under three constant glucose levels-low (0.01%), medium (0.1%), and high (1%) (Fig. 1B-C). The authors confirm that steady-state Hxt abundances rank by transporter affinity. But the more important and surprising discovery is that when the cells were subjected to gradual glucose up-shifts and down-shifts, they discovered that some transporters transiently spike only when [glucose] rises and others only when [glucose] falls (Fig. 1C and Fig. S1F). This discovery establishes that the HXT network not only "senses" the absolute external [glucose] concentration but also the direction of the temporal change in external [glucose].

To understand how the regulatory network yields such intricate temporal changes in HXT expression, the authors first focused on the medium-affinity transporter, Hxt4. Targeted knockouts of Mig1/Mig2 versus Mth1/Std1 confirmed that Hxt4 dynamics arise from differential repressor kinetics. To formalize these findings, the authors built an ODE model grounded in literature-based constraints (pg. 13 of the Supplement) with explicit separation of repressor timescales. They rigorously fit the model to wild-type and knockout time series-exploring parameter sensitivity in depth (Fig. S5).

The authors discovered that their model and experiments converged on a push-pull mechanism: fast-acting Mig1/Mig2 dominate during glucose up-shifts, while slower Mth1/Std1 govern down-shifts, determining whether each HXT gene is repressed or de-repressed (i.e., "who gets there first"). Extending this analysis across all seven HXTs via approximate Bayesian computation revealed the most likely repressor-promoter interactions for each transporter, reducing a vast parameter space to unique or small sets of plausible regulatory schemes. The authors thus revealed what could be happening and which regulations are improbable - a more nuanced and comprehensive view than giving just one outcome for each HXT.

Overall, this work represents a role model - textbook-worthy - for quantitative systems biology. Beyond the rigor and novelty of its findings, the authors explain complex mathematical concepts with clarity, and the narrative flows logically from experiment to model to inference. This study provides a definitive mechanistic resolution of the HXT network and establishes a broadly applicable framework for dissecting dynamic and complex gene circuits.

Major points:

I don't recommend any new experiments or modeling; the major claims are already well supported by the data and models. Below are comments and questions intended to improve clarity and facilitate the reader's understanding. Please feel free to disregard any that you find not sensible or beyond the scope of the current work.

1. Preconditioning (Fig. 1B-C): What medium were cells in immediately before t = 0? Were they in log-phase or stationary-phase growth just prior to the glucose addition?
2. Transporter ranking in medium glucose: In the medium [glucose] regime, why is a low-affinity Hxt the second-most highly expressed, rather than the next-highest-affinity transporter? Could co-expression of multiple affinities (e.g., as a bet-hedging strategy) be advantageous? The Discussion section already mentions bet-hedging but I think you could further discuss ideas such as evolutionarily trained "Pavlovian" response or what the 2nd-ranking says about what the yeast anticipates as an upcoming change in the environment.
3. Defining low/medium/high regimes: Low = 0.01%, Medium = 0.1%, and High = 1%. This is indeed in line with the standard classification of [glucose] in the literature regarding HXTs. But how might your results change at intermediate concentrations - those between these three levels. Using the model, could you comment on whether HXT expression dynamics "sharply" change as a function of either the [glucose]/time or the final concentration of [glucose] after the ramping-up phase?
4. Rate-affinity trade-off (Lines 18-20): Give a brief explanation of the rate-affinity trade-off. Why does higher affinity necessarily entail a lower maximal transport rate (Vₘₐₓ) for passive transporters? Perhaps you can give an intuitive explanation backed by mass-action kinetics (e.g., to attain a higher affinity, the glucose-binding pocket on Hxt cannot be flipping rapidly back-and-forth between facing cytoplasm and extracellular space -- the binding pocket must allow sufficient time for molecule to find and bind it).
5. Single-transporter expression (Lines 39-40): It's unclear to me why cells would express only the "optimal" Hxt and suppress all others. For instance, a bet-hedging strategy might favor simultaneous expression of multiple affinities. Consider revising these lines or adding a brief explanation. Related to above is a subtle point I think that was glossed over: there must be a fitness cost associated with making too many copies of Hxtn. After all, why not make as many transporters as possible? Is the cell operating near the upper limit of Hxt abundance, beyond which there's a fitness cost? Is there a pareto-optimal-type front in the space of expression level and another axis? I think this could go into the Discussion section.
6. Hxt5 exception (Fig. 1B): Although Hxt5 follows a distinct regulatory scheme, it is most highly expressed at medium [glucose] (0.1%), consistent with its affinity like the other Hxts. I think you could mention this in lines 51-58.
7. Glucose-ramp details (Fig. 1C; Lines 66-67): You state that [glucose] rises from 0 to 1 % over 15 min and reaches 1 % at t = 3 h. However, the actual ramp slope ([glucose]/time) and when the [glucose] starts to increase from zero aren't specified. The Hxt5-GFP behavior and differing Hxt6/7 levels at t = 0 vs. t = 20 h suggest the ramp may begin later than t = 0. Please clarify these details in the caption and main text, and consider adding a [glucose] vs. time schematic above the panel in Fig. 1C (like in Fig. 1B).
8. Pre-t < 0 incubation (Fig. 1C): Related to point 1, how long were the cells incubated in pyruvate (or other medium) before t < 0? The Hxt6-GFP level at t = 20 h does not match that at t = 0; what is the timescale for Hxt6-GFP and Hxt7-GFP decay to steady state after glucose removal?
9. Hxt-GFP localization: Does the reported Hxt#-GFP level include fluorescence from both the plasma membrane and internal compartments (e.g., vacuole)? Clarifying which pools of fluorescence are quantified would help interpretation, even if they don't change the main conclusions are unchanged.
10. Predominantly transcriptional" wording (Lines 90-92): The phrase "...the regulation is predominantly transcriptional" should specify that it refers to the induction of HXT4 transcription during glucose down-ramping, rather than the subsequent decrease in Hxt4-GFP. The experiments do not rule out post-translational regulation (e.g., endocytosis) once glucose levels fall below a threshold.
11. Glucose "protection" of Hxt4 (Lines 121-122): The statement "we allowed glucose to protect Hxt4 from degradation" is unclear. First, Hxt4-GFP likely degrades at a different rate than free GFP-you could estimate its half-life from Fig. S3. Second, please explain precisely what "protection" means in the model or experiment.
12. Quantifying repressor kinetics (Lines 158-162): The push-pull mechanism is compelling, but it would be helpful to report the quantitative separation of timescales-e.g., how much faster do Mig1/Mig2 respond compared to Mth1/Std1? Including fold-difference would strengthen this explanation.
13. Mechanism of repressor regulation (Lines 197-213): Be clearer about whether and how changes in extracellular glucose alter the expression levels of Mth1, Std1, Mig1, and Mig2, as opposed to modulating say, how Mth1 and Std1 bind to Rgt2 protein. I think you could be clearer here about which regulatory steps (transcriptional, post-translational, or binding-affinity changes) are assumed in the model and supported by the data.

Minor points:

1. Abstract: Original: "...how an HXT for a medium-affinity transporter can be made to respond like the HXTs for the other transporters." Suggestion: "...how the gene-expression regulation of a medium-affinity HXT can be rewired to respond like that of any other HXT." (You might also generalize beyond "medium-affinity" if the converse holds.)
2. Lines 64-66: Please emphasize that the "synthetic complete medium" used for pre-conditioning contains no glucose.
3. Line 143: The phrase "low expression of the std1\Delta strain in glucose" is ambiguous-low expression of which gene or reporter? Please specify.
4. Line 240: Change "should weakened" to "should weaken."
5. Fig. S9 caption (typo) Change "Rtg1 sites are..." to "Rgt1 sites are...."

Hyun Youk.

Referee cross-commenting

I agree with the other reviewers' comments. The other reviewers noticed important points I have missed. But like them, I'm still supportive of the work being published with < 1 month spent on revision. I still don't recommend any further experiments or modeling.

Significance

This is a very insightful work showing how to disentangle one of the most complex transcriptional networks in yeast (S. cerevisiae) by combining single-cell dynamics, dynamical-systems modeling, Bayesian-style inference, and genetic perturbations. The authors tackle a problem that has eluded quantitative resolution for over two decades-how yeast regulates its seven primary glucose importer genes (HXT1-HXT7) in response to both steady and temporally changing extracellular [glucose]. Their integrated experimental-theoretical approach delivers the most satisfying mechanistic and quantitative explanation to date, and I enthusiastically recommend this manuscript for publication via Review Commons.

Reviewer #2 (Public review):

In the present manuscript, Dannulat Frazier et al. provide a novel and advanced protocol for obtaining almost pure populations of neural rosette stem cells (NRSCs) expressing the general markers NES and SOX2. These NSCs are expandable and exhibit dorsal forebrain properties and markers that are maintained throughout passages in culture (at least until passage 12). The authors also demonstrate the multipotency of these NSCs by their ability to differentiate into functional neurons, and precursors of astrocytes and oligodendrocytes.

This method does not require the usual step of manual rosette selection and allows a greater homogeneity of the NSCs obtained and the standardization of the protocol, which will allow greater advances in the applications of these NSCs in research and as models of disease or compound testing. The manuscript is of great interest for the research area, since it describes a new methodology that can facilitate the research and therapeutic application of NSCs.

The manuscript is well-written; the results are clear, robust, and well-explained. The conclusions reached in this paper are well-supported by the data, but some aspects could be better clarified.

(1) The results presented in the present manuscript of the NSCS are performed up to passage 12; it would be interesting to know up to which passages these cells can be expanded, maintaining their initial properties. Have the authors analyzed passages beyond 12?

(2) In Figure 2A, where different markers are shown in NSCs at different passages, it seems that at passage 12, there is a decrease in TJP1+ zones in relation to earlier passages, which could indicate a reduction in the potential to generate rosettes. Have the authors done any quantification along these lines? Could this be the case, or is it just an effect of the image chosen?

(3) In Figure 3A, it is very striking and intriguing that the decrease in the expression of the PAX6 gene in passage 8 in relation to passage 2, which does not correspond to what is observed at the protein level. Have the authors verified this result using another technique, such as for example RT-q-PCR?

(4) In Figure 5B, the labeling for GFAP, appears rather nuclear, despite being a cytoskeleton protein. How can the authors explain this?

Author response:

Reviewer #1 (Public review):

Thank you for your thoughtful and constructive feedback on our manuscript. We greatly appreciate your insights regarding our work, as they are invaluable in refining our research.

We are very happy to hear that you recognize the strengths of our method, particularly the elimination of manual rosette picking, which significantly enhances throughput and reduces variability. We are also pleased that our validation efforts—through flow cytometry, immunocytochemistry, single-cell RNA-sequencing, and functional MEA recordings—effectively demonstrate both the identity and functionality of our derived dorsal forebrain neural rosette stem cells (NRSCs).

Regarding the identified weaknesses, we agree that a direct comparison with conventional manual-selection protocols, specifically those utilizing dual-SMAD inhibition, would be a significant improvement. To address this, we have initiated additional experiments that will directly compare our single-SMAD inhibition approach (RepSox) with dual-SMAD inhibition (SB/LDN), aiming for a comprehensive evaluation of both protocols.

In terms of statistical rigor, we appreciate your suggestion on improving our quantitative assays. All data were collected from at least three independent experiments and presented as mean ±standard deviation unless otherwise specified. Due to the qualitative nature of the data, no formal statistical tests were performed for most of the experiments and the mean and standard deviation were calculated for some quantitative measurements obtained, providing a descriptive summary of the data. When possible, we will incorporate appropriate statistical tests, to present our data in a more robust manner, rather than merely reporting mean ± SD.

Finally, we recognize the importance of situating our work within the broader landscape of neural stem cell research. We aim to elucidate the potential downstream applications for our protocol, which we believe will significantly impact neurodevelopmental and neurodegenerative disorder studies.

Thank you again for your valuable suggestions. We look forward to refining our manuscript and enhancing the contribution of our research to the field.

Reviewer #2 (Public review):

Thank you for your thoughtful and constructive feedback on our manuscript. We appreciate your recognition of the novelty and potential impact of our protocol for obtaining neural rosette stem cells (NRSCs). Your comments are invaluable in improving our work.

We are pleased that you found our methodology to be a significant advancement in the field, particularly the elimination of the manual rosette selection step, which hopefully will enhance homogeneity and standardization. We agree that this development has implications for research, disease modelling, and compound testing.

Regarding your specific points:

Passage expansion: Thank you for your insightful suggestion regarding the analysis beyond passage 12. We have continued passaging our NRSC line for more than 12 passages while maintaining the rosette structure. Although we do not yet have comprehensive and detailed analyses at these later passages, we will include some data and relevant information on our findings in the revised manuscript.

TJP1+ zones: We appreciate your observation regarding the decreased TJP1+ zones at passage 12. We have not consistently detected a reduction in the number of rosettes or TJP1+ lumens across our cultures between passages. While some variability has been noted, we occasionally observe minor reductions at specific time points, followed by a recovery of rosettes in subsequent passages. This suggests that monitoring the number of rosettes is indeed a useful indicator of cell culture health. Cultures should be discarded if rosettes are completely lost. We will take a closer look at this aspect and report the findings in the revised manuscript.

PAX6 Gene expression verification: Thank you for highlighting the discrepancy between PAX6 gene expression levels and protein levels. Unfortunately, we have not yet validated these results using an alternative technique. One potential explanation for this discrepancy may be the phenomenon of negative autoregulation, where increased levels of PAX6 protein can inhibit its own mRNA expression (Manuel et al., 2007). Moreover, Hsieh and Yang (2009) observed that during neurogenesis, PAX6 protein levels may not correlate linearly with mRNA levels, particularly in variable cellular environments. Additionally, post-transcriptional regulatory mechanisms, such as translation initiation mediated by Internal Ribosome Entry Sites (IRES), have been documented in various contexts involving PAX6, suggesting that mRNA levels may not fully represent functional protein levels in developing tissues (Li et al., 2023). We will go deeper into this discussion in the revised manuscript.

GFAP Labeling: We appreciate your comments regarding the nuclear labeling of GFAP. In our astrocyte cultures, we have indeed observed GFAP localization in both the nucleus and the cytoplasm (Figure 5B). We will investigate this phenomenon further and provide a clearer explanation, supported by relevant literature, in the revised version. Although GFAP is primarily categorized as an intermediate filament protein localized in the cytoplasm, evidence suggests its nuclear localization may indicate additional regulatory roles during astrocyte development, activation, and pathology. This finding highlights the potential complexity of GFAP's role during fetal development and cellular stress, suggesting a broader functional scope that may extend into the nuclear space.

Once again, thank you for your insightful feedback and for recognizing the potential of our research. We are committed to addressing your comments and enhancing the quality of our manuscript.

Manuel, M. et al. (2007) ‘Controlled overexpression of Pax6 in vivo negatively autoregulates the Pax6 locus, causing cell-autonomous defects of late cortical progenitor proliferation with little effect on cortical arealization’, Development, 134(3), pp. 545–555. Available at: https://doi.org/10.1242/dev.02764.

Hsieh, Y.-W. and Yang, X.-J. (2009) ‘Dynamic Pax6 expression during the neurogenic cell cycle influences proliferation and cell fate choices of retinal progenitors’, Neural Development, 4(1), p. 32. Available at: https://doi.org/10.1186/1749-8104-4-32.

Li, Q. et al. (2023) ‘Translation of paired box 6 (PAX6) mRNA is IRES-mediated and inhibited by cymarin in breast cancer cells’, Genes & Genetic Systems, 98(4), pp. 161–169. Available at: https://doi.org/10.1266/ggs.23-00039.

4 groups (Active/non-active by femaleMenstruating vs not), 45*4 = 180 participants assuming equal representation of all groups, ignoring race

Reviewer #2 (Public review):

Summary:

This study explores how gene expression evolves in response to seasonal environments, using four evergreen Fagaceae species growing in similar habitats in Japan. By combining chromosome-scale genome assemblies with a two-year RNA-seq time series in leaves and buds, the authors identify seasonal rhythms in gene expression and examine both conserved and divergent patterns. A central result is that winter bud expression is highly conserved across species, likely due to shared physiological demands under cold conditions. One of the intriguing implications of this study is that seasonal cycles might play a role similar to ontogenetic stages in animals. The authors touch on this by comparing their findings to the developmental hourglass model, and indeed, the recurrence of phenological states such as winter dormancy may act as a cyclic form of developmental canalization, shaping expression evolution in a way analogous to embryogenesis in animals.

Strengths:

(1) The evolutionary effects of seasonal environments on gene expression are rarely studied at this scale. This paper fills that gap.

(2) The dataset is extensive, covering two years, two tissues, and four tree species, and is well suited to the questions being asked.

(3) Transcriptome clustering across species (Figure 2) shows strong grouping by season and tissue rather than species, suggesting that the authors effectively controlled for technical confounders such as batch effects and mapping bias.

(4) The idea that winter imposes a shared constraint on gene expression, especially in buds, is well argued and supported by the data.

(5) The discussion links the findings to known concepts like phenological synchrony and the developmental hourglass model, which helps frame the results.

Weaknesses:

(1) While the hierarchical clustering shown in Figure 2A largely supports separation by tissue type and season, one issue worth noting is that some leaf samples appear to cluster closely with bud samples. The authors do not comment on this pattern, which raises questions about possible biological overlap between tissues during certain seasonal transitions or technical artifacts such as sample contamination. Clarifying this point would improve confidence in the interpretation of tissue-specific seasonal expression patterns.

(2) While the study provides compelling evidence of conserved and divergent seasonal gene expression, it does not directly examine the role of cis-regulatory elements or chromatin-level regulatory architecture. Including regulatory genomic or epigenomic data would considerably strengthen the mechanistic understanding of expression divergence.

(3) The manuscript includes a thoughtful analysis of flowering-related genes and seasonal GO enrichment (e.g., Figure 3C-D), providing an initial link between gene expression timing and phenological functions. However, the analysis remains largely gene-centric, and the study does not incorporate direct measurements of phenological traits (e.g., flowering or bud break dates). As a result, the connection between molecular divergence and phenotypic variation, while suggestive, remains indirect.

(4) Although species were sampled from similar habitats, one species (Q. acuta) was collected at a higher elevation, and factors such as microclimate or local photoperiod conditions could influence expression patterns. These potential confounding variables are not fully accounted for, and their effects should be more thoroughly discussed or controlled in future analyses.

(5) Statistical and Interpretive Concerns Regarding Δφ and dN/dS Correlation (Figures 5E and 5F):

(a) Statistical Inappropriateness: Δφ is a discrete ordinal variable (likely 1-11), making it unsuitable for Pearson correlation, which assumes continuous, normally distributed variables. This undermines the statistical validity of the analysis.

(b) Biological Interpretability: Even with the substantial statistical power afforded by genome-wide analysis, the observed correlations are extremely weak. This suggests that the relationship, if any, between temporal divergence in expression and protein-coding evolution is negligible.

Taken together, these issues weaken the case for any biologically meaningful association between Δφ and dN/dS. I recommend either omitting these panels or clearly reframing them as exploratory and statistically limited observations.

Author response:

Reviewer #1 (Public review):

Summary:

The authors performed genome assemblies for two Fagaceae species and collected transcriptome data from four natural tree species every month over two years. They identified seasonal gene expression patterns and further analyzed species-specific differences.

Strengths:

The study of gene expression patterns in natural environments, as opposed to controlled chambers, is gaining increasing attention. The authors collected RNA-seq data monthly for two years from four tree species and analyzed seasonal expression patterns. The data are novel. The authors could revise the manuscript to emphasize seasonal expression patterns in three species (with one additional species having more limited data). Furthermore, the chromosome-scale genome assemblies for the two Fagaceae species represent valuable resources, although the authors did not cite existing assemblies from closely related species.

Thank you for your careful assessment of our manuscript.

Weaknesses:

Comment; The study design has a fundamental flaw regarding the evaluation of genetic or evolutionary effects. As a basic principle in biology, phenotypes, including gene expression levels, are influenced by genetics, environmental factors, and their interaction. This principle is well-established in quantitative genetics.

In this study, the four species were sampled from three different sites (see Materials and Methods, lines 543-546), and additionally, two species were sampled from 2019-2021, while the other two were sampled from 2021-2023 (see Figure S2). This critical detail should be clearly described in the Results and Materials and Methods. Due to these variations in sampling sites and periods, environmental conditions are not uniform across species.

Even in studies conducted in natural environments, there are ways to design experiments that allow genetic effects to be evaluated. For example, by studying co-occurring species, or through transplant experiments, or in common gardens. To illustrate the issue, imagine an experiment where clones of a single species were sampled from three sites and two time periods, similar to the current design. RNA-seq analysis would likely detect differences that could qualitatively resemble those reported in this manuscript.

One example is in line 197, where genus-specific expression patterns are mentioned. While it may be true that the authors' conclusions (e.g., winter synchronization, phylogenetic constraints) reflect real biological trends, these conclusions are also predictable even without empirical data, and the current dataset does not provide quantitative support.

If the authors can present a valid method to disentangle genetic and environmental effects from their dataset, that would significantly strengthen the manuscript. However, I do not believe the current study design is suitable for this purpose.

Unless these issues are addressed, the use of the term "evolution" is inappropriate in this context. The title should be revised, and the result sections starting from "Peak months distribution..." should be either removed or fundamentally revised. The entire Discussion section, which is based on evolutionary interpretation, should be deleted in its current form.

If the authors still wish to explore genetic or evolutionary analyses, the pair of L. edulis and L. glaber, which were sampled at the same site and over the same period, might be used to analyze "seasonal gene expression divergence in relation to sequence divergence." Nevertheless, the manuscript would benefit from focusing on seasonal expression patterns without framing the study in evolutionary terms.

We sincerely thank the reviewer for the detailed and thoughtful comments. We fully recognize the importance of carefully distinguishing genetic and environmental contributions in transcriptomic studies, particularly when addressing evolutionary questions. The reviewer identified two major concerns regarding our study design: (1) the use of different monitoring periods across species, and (2) the use of samples collected from different study sites. We addressed both concerns with additional analyses using 112 new samples and now present new evidence that supports the robustness of our conclusions.

(1) Monitoring period variation does not bias our conclusions

To address concerns about the differing monitoring periods, we added new RNA-seq data (42 samples each for bud and leaf samples for L. glaber and 14 samples each for bud and leaf samples for L. edulis) collected from November 2021 to November 2022, enabling direct comparison across species within a consistent timeframe. Hierarchical clustering of this expanded dataset (Fig. S6) yielded results consistent with our original findings: winter-collected samples cluster together regardless of species identity. This strongly supports our conclusion that the seasonal synchrony observed in winter is not an artifact of the monitoring period and demonstrates the robustness of our conclusions across datasets.

(2) Site variation is limited and does not confound our findings

Although the study included three sites, two of them (Imajuku and Ito Campus) are only 7.3 km apart, share nearly identical temperature profiles (see Fig. S2), and are located at the edge of similar evergreen broadleaf forests. Only Q. acuta was sampled from a higher-altitude, cooler site. To assess whether the higher elevation site of Q. acuta introduced confounding environmental effects, we reanalyzed the data after excluding this species. Hierarchical clustering still revealed that winter bud samples formed a distinct cluster regardless of species identity (Fig. S7), consistent with our original finding.

Furthermore, we recalculated the molecular phenology divergence index D (Fig. 4C) and the interspecific Pearson’s correlation coefficients (Fig. 5A) without including Q. acuta. These analyses produced results that were similar to those obtained from the full dataset (Fig. S12; Fig. S14), indicating that the observed patterns are not driven by environmental differences associated with elevation.

(3) Justification for our approach in natural systems

We agree with the reviewer that experimental approaches such as common gardens, reciprocal transplants, and the use of co-occurring species are valuable for disentangling genetic and environmental effects. In fact, we have previously implemented such designs in studies using the perennial herb Arabidopsis halleri (Komoto et al., 2022, https://doi.org/10.1111/pce.14716) and clonal Someiyoshino cherry trees (Miyawaki-Kuwakado et al., 2024, https://doi.org/10.1002/ppp3.10548) to examine environmental effects on gene expression. However, extending these approaches to long-lived tree species in diverse natural ecosystems poses significant logistical and biological challenges. In this study, we addressed this limitation by including three co-occurring species at the same site, which allowed us to evaluate interspecific differences under comparable environmental conditions. Importantly, even when we limited our analyses to these co-occurring species, the results remained consistent, indicating that the observed variation in transcriptomic profiles cannot be attributed to environmental factors alone and likely reflects underlying genetic influences.

Accordingly, we added four new figures (Fig. S6, Fig. S7, Fig. S12 and Fig. S14) and revised the manuscript to clarify the limitations and strengths of our design, to tone down the evolutionary claims where appropriate, and to more explicitly define the scope of our conclusions in light of the data. We hope that these efforts sufficiently address the reviewer’s concerns and strengthen the manuscript.

To better support the seasonal expression analysis, the early RNA-seq analysis sections should be strengthened. There is little discussion of biological replicate variation or variation among branches of the same individual. These could be important factors to analyze. In line 137, the mapping rate for two species is mentioned, but the rates for each species should be clearly reported. One RNA-seq dataset is based on a species different from the reference genome, so a lower mapping rate is expected. While this likely does not hinder downstream analysis, quantification is important.

We thank the reviewer 1 for the helpful comment. To evaluate the variation among biological replicates, we compared the expression level of each gene across different individuals. We observed high correlation between each pair of individuals (Q. glauca (n=3): an average correlation coefficient r = 0.947; Q. acuta (n=3): r = 0.948; L. glaber (n=3): r = 0.948)). This result suggests that the seasonal gene expression pattern is highly synchronized across individuals within the same species. We mentioned this point in the Result section in the revised manuscript. We also calculated the mean mapping rates for each species. As the reviewer expected, the mapping rate was slightly lower in Q. acuta (88.6 ± 2.3%) and L. glaber (84.3 ± 5.4%), whose RNA-Seq data were mapped to reference genomes of related but different species, compared to that in Q. glauca (92.6 ± 2.2%) and L. edulis (89.3 ± 2.7%). However, we minimized the impact of these differences on downstream analysis. These details have been included in the revised main text.

In Figures 2A and 2B, clustering is used to support several points discussed in the Results section (e.g., lines 175-177). However, clustering is primarily a visualization method or a hypothesis-generating tool; it cannot serve as a statistical test. Stronger conclusions would require further statistical testing.

We thank the reviewer for the helpful comment. As noted, we acknowledge that hierarchical clustering (Fig. 2A) is primarily a visualization and hypothesis-generating method. To assess the biological relevance of the clusters identified, we conducted a Mann-Whitney U test or the Steel-Dwass test to evaluate whether the environmental temperatures at the time of sample collection differed significantly among the clusters. This analysis (Fig. 2B) revealed statistically significant differences in temperature in the cluster B3 (p < 0.01), indicating that the gene expression clusters are associated with seasonal thermal variation. These results support the interpretation that the clusters reflect coordinated transcriptional responses to environmental temperature. We revised the Results section to clarify this point.

The quality of the genome assemblies appears adequate, but related assemblies should be cited and discussed. Several assemblies of Fagaceae species already exist, including Quercus mongolica (Ai et al., Mol Ecol Res, 2022), Q. gilva (Front Plant Sci, 2022), and Fagus sylvatica (GigaScience, 2018), among others. Is there any novelty here? Can you compare your results with these existing assemblies?

We agree that genome assemblies of Fagaceae species are becoming increasing available. However, our study does not aim to emphasize the novelty of the genome assemblies per se. Rather, with the increasing availability of chromosome-level genomes, we regard genome assembly as a necessary foundation for more advanced analyses. The main objective of our study is to investigate how each gene is expressed in response to seasonal environmental changes, and to link genome information with seasonal transcriptomic dynamics. To address the reviewer’s comment in line with this objective, we added a discussion on the syntenic structure of eight genome assemblies spanning four genera within the Fagaceae, including a species from the genus Fagus (Ikezaki et al. 2025, https://doi.org/10.1101/2025.07.31.667835). This addition helps to position our work more clearly within the context of existing genomic resources.

Most importantly, Figure 1B-D shows synteny between the two genera but also indicates homology between different chromosomes. Does this suggest paleopolyploidy or another novel feature? These chromosome connections should be interpreted in the main text-even if they could be methodological artifacts.

A previous study on genome size variation in Fagaceae suggested that, given the consistent ploidy level across the family, genome expansion likely occurred through relatively small segmental duplications rather than whole-genome duplications. Because Figure 1B-D supports this view, we cited the following reference in the revised version of the manuscript.

Chen et al. (2014)  https://doi.org/10.1007/s11295-014-0736-y

In both the Results and Materials and Methods sections, descriptions of genome and RNA-seq data are unclear. In line 128, a paragraph on genome assembly suddenly introduces expression levels. RNA-seq data should be described before this. Similarly, in line 238, the sentence "we assembled high-quality reference genomes" seems disconnected from the surrounding discussion of expression studies. In line 632, Illumina short-read DNA sequencing is mentioned, but it's unclear how these data were used.

We relocated the explanation regarding the expression levels of single-copy and multi-copy genes to the section titled “Seasonal gene expression dynamics.” Additionally, we clarified in the Materials and Methods section that short-read sequencing data were used for both genome size estimation and phylogenetic reconstruction.

Reviewer #2 (Public review):

Summary:

This study explores how gene expression evolves in response to seasonal environments, using four evergreen Fagaceae species growing in similar habitats in Japan. By combining chromosome-scale genome assemblies with a two-year RNA-seq time series in leaves and buds, the authors identify seasonal rhythms in gene expression and examine both conserved and divergent patterns. A central result is that winter bud expression is highly conserved across species, likely due to shared physiological demands under cold conditions. One of the intriguing implications of this study is that seasonal cycles might play a role similar to ontogenetic stages in animals. The authors touch on this by comparing their findings to the developmental hourglass model, and indeed, the recurrence of phenological states such as winter dormancy may act as a cyclic form of developmental canalization, shaping expression evolution in a way analogous to embryogenesis in animals.

Strengths:

(1) The evolutionary effects of seasonal environments on gene expression are rarely studied at this scale. This paper fills that gap.

(2) The dataset is extensive, covering two years, two tissues, and four tree species, and is well suited to the questions being asked.

(3) Transcriptome clustering across species (Figure 2) shows strong grouping by season and tissue rather than species, suggesting that the authors effectively controlled for technical confounders such as batch effects and mapping bias.

(4) The idea that winter imposes a shared constraint on gene expression, especially in buds, is well argued and supported by the data.

(5) The discussion links the findings to known concepts like phenological synchrony and the developmental hourglass model, which helps frame the results.

We are grateful for the reviewer for the detailed and thoughtful review of our manuscript.

Weaknesses:

(1) While the hierarchical clustering shown in Figure 2A largely supports separation by tissue type and season, one issue worth noting is that some leaf samples appear to cluster closely with bud samples. The authors do not comment on this pattern, which raises questions about possible biological overlap between tissues during certain seasonal transitions or technical artifacts such as sample contamination. Clarifying this point would improve confidence in the interpretation of tissue-specific seasonal expression patterns.

Leaf samples clustered into the bud are newly flushed leaves collected in April for Q. glauca, May for Q. acuta, May and June for L. edulis, and August and September for L. glaber. To clarify this point, we highlighted these newly flushed leaf samples as asterisk in the revised figure (Fig. 2A).

comment; (2) While the study provides compelling evidence of conserved and divergent seasonal gene expression, it does not directly examine the role of cis-regulatory elements or chromatin-level regulatory architecture. Including regulatory genomic or epigenomic data would considerably strengthen the mechanistic understanding of expression divergence.

We thank the reviewer for this insightful comment. As noted in the Discussion section, we hypothesize that such genome-wide seasonal expression patterns—and their divergence across species—are likely mediated by cis-regulatory elements and chromatin-level mechanisms. While a direct investigation of regulatory architecture was beyond the scope of the present study, we fully agree that incorporating regulatory genomic and epigenomic data would significantly deepen the mechanistic understanding of expression divergence. In this regard, we are currently working to identify putative cis-regulatory elements in non-coding regions and are collecting epigenetic data from the same tree species using ChIP-seq. We believe the current study provide a foundation for these future investigations into the regulatory basis of seasonal transcriptome variation. We made a minor revision to the Discussion to note that an important future direction is to investigate the evolution of non-coding sequences that regulate gene expression in response to seasonal environmental changes.

(3) The manuscript includes a thoughtful analysis of flowering-related genes and seasonal GO enrichment (e.g., Figure 3C-D), providing an initial link between gene expression timing and phenological functions. However, the analysis remains largely gene-centric, and the study does not incorporate direct measurements of phenological traits (e.g., flowering or bud break dates). As a result, the connection between molecular divergence and phenotypic variation, while suggestive, remains indirect.

We would like to note that phenological traits have been observed in the field on a monthly basis throughout the sampling period and the phenological data were plotted together with molecular phenology (e.g. Fig. 2A, C; Fig. 3C, D). Although the temporal resolution is limited, these observations captured species-specific differences in key phenological events such as leaf flushing and flowering times. We revised the manuscript to clarify this point.

(4) Although species were sampled from similar habitats, one species (Q. acuta) was collected at a higher elevation, and factors such as microclimate or local photoperiod conditions could influence expression patterns. These potential confounding variables are not fully accounted for, and their effects should be more thoroughly discussed or controlled in future analyses.

We fully agree with the reviewer that local environmental conditions, including microclimate and photoperiod differences, could potentially influence gene expression patterns. To assess whether the higher elevation site of Q. acuta introduced confounding environmental effects, we reanalyzed the data after excluding this species. Hierarchical clustering still revealed that winter bud samples formed a distinct cluster regardless of species identity (Fig. S7), consistent with our original finding.

Furthermore, we recalculated the molecular phenology divergence index D (Fig. 4C) and the interspecific Pearson’s correlation coefficients (Fig. 5A) without including Q. acuta. These analyses produced results that were qualitatively similar to those obtained from the full dataset (Fig. S12; Fig. S14), indicating that the observed patterns are not driven by environmental differences associated with elevation.

We believe these additional analyses help to decouple the effects of environment and genetics, and support our conclusion that both seasonal synchrony and phylogenetic constraints play key roles in shaping transcriptome dynamics. We added four new figures (Fig. S6, Fig. S7, Fig. S12 and Fig. S14) and revised the text accordingly to clarify this point and to acknowledge the potential impact of site-specific environmental variation.

(5) Statistical and Interpretive Concerns Regarding Δφ and dN/dS Correlation (Figures 5E and 5F):

(a) Statistical Inappropriateness: Δφ is a discrete ordinal variable (likely 1-11), making it unsuitable for Pearson correlation, which assumes continuous, normally distributed variables. This undermines the statistical validity of the analysis.

We thank the reviewer for the insightful comment. We would like to clarify that the analysis presented in Figures 5E and 5F was based on linear regression, not Pearson’s correlation. Although Δφ is a discrete variable, it takes values from 0 to 6 in 0.5 increments, resulting in 13 levels. We treated it as a quasi-continuous variable for the purposes of linear regression analysis. This approach is commonly adopted in practice when a discrete variable has sufficient resolution and ordering to approximate continuity. To enhance clarity, we revised the manuscript to explicitly state that linear regression was used, and we now reported the regression coefficient and associated p-value to support the interpretation of the observed trend.

(b) Biological Interpretability: Even with the substantial statistical power afforded by genome-wide analysis, the observed correlations are extremely weak. This suggests that the relationship, if any, between temporal divergence in expression and protein-coding evolution is negligible.

Taken together, these issues weaken the case for any biologically meaningful association between Δφ and dN/dS. I recommend either omitting these panels or clearly reframing them as exploratory and statistically limited observations.

We agree with the reviewer’s comment. While we retained the original panels, we reframed our interpretation to emphasize that, despite statistical significance, the observed correlation is very weak—suggesting that coding region variation is unlikely to be the primary driver of seasonal gene expression patterns. Accordingly, we revised the “Relating seasonal gene expression divergence to sequence divergence” section in the Results, as well as the relevant part of the Discussion.

Reviewer #1 (Public review):

Summary:

This is a strong paper that presents a clear advance in multi-animal tracking. The authors introduce an updated version of idtracker.ai that reframes identity assignment as a contrastive learning problem rather than a classification task requiring global fragments. This change leads to gains in speed and accuracy. The method eliminates a known bottleneck in the original system, and the benchmarking across species is comprehensive and well executed. I think the results are convincing and the work is significant.

Strengths:

The main strengths are the conceptual shift from classification to representation learning, the clear performance gains, and the fact that the new version is more robust. Removing the need for global fragments makes the software more flexible in practice, and the accuracy and speed improvements are well demonstrated. The software appears thoughtfully implemented, with GUI updates and integration with pose estimators.

Weaknesses:

I don't have any major criticisms, but I have identified a few points that should be addressed to improve the clarity and accuracy of the claims made in the paper.

(1) The title begins with "New idtracker.ai," which may not age well and sounds more promotional than scientific. The strength of the work is the conceptual shift to contrastive representation learning, and it might be more helpful to emphasize that in the title rather than branding it as "new."

(2) Several technical points regarding the comparison between TRex (a system evaluated in the paper) and idtracker.ai should be addressed to ensure the evaluation is fair and readers are fully informed.

(2.1) Lines 158-160: The description of TRex as based on "Protocol 2 of idtracker.ai" overlooks several key additions in TRex, such as posture image normalization, tracklet subsampling, and the use of uniqueness feedback during training. These features are not acknowledged, and it's unclear whether TRex was properly configured - particularly regarding posture estimation, which appears to have been omitted but isn't discussed. Without knowing the actual parameters used to make comparisons, it's difficult to assess how the method was evaluated.

(2.2) Lines 162-163: The paper implies that TRex gains speed by avoiding Protocol 3, but in practice, idtracker.ai also typically avoids using Protocol 3 due to its extremely long runtime. This part of the framing feels more like a rhetorical contrast than an informative one.

(2.3) Lines 277-280: The contrastive loss function is written using the label l, but since it refers to a pair of images, it would be clearer and more precise to write it as l_{I,J}. This would help readers unfamiliar with contrastive learning understand the formulation more easily.

(2.4) Lines 333-334: The manuscript states that TRex can fail to track certain videos, but this may be inaccurate depending on how the authors classify failures. TRex may return low uniqueness scores if training does not converge well, but this isn't equivalent to tracking failure. Moreover, the metric reported by TRex is uniqueness, not accuracy. Equating the two could mislead readers. If the authors did compare outputs to human-validated data, that should be stated more explicitly.

(2.5) Lines 339-341: The evaluation approach defines a "successful run" and then sums the runtime across all attempts up to that point. If success is defined as simply producing any output, this may not reflect how experienced users actually interact with the software, where parameters are iteratively refined to improve quality.

(2.6) Lines 344-346: The simulation process involves sampling tracking parameters 10,000 times and selecting the first "successful" run. If parameter tuning is randomized rather than informed by expert knowledge, this could skew the results in favor of tools that require fewer or simpler adjustments. TRex relies on more tunable behavior, such as longer fragments improving training time, which this approach may not capture.

(2.7) Line 354 onward: TRex was evaluated using two varying parameters (threshold and track_max_speed), while idtracker.ai used only one (intensity_threshold). With a fixed number of samples, this asymmetry could bias results against TRex. In addition, users typically set these parameters based on domain knowledge rather than random exploration.

(2.8) Figure 2-figure supplement 3: The memory usage comparison lacks detail. It's unclear whether RAM or VRAM was measured, whether shared or compressed memory was included, or how memory was sampled. Since both tools dynamically adjust to system resources, the relevance of this comparison is questionable without more technical detail.

(3) While the authors cite several key papers on contrastive learning, they do not use the introduction or discussion to effectively situate their approach within related fields where similar strategies have been widely adopted. For example, contrastive embedding methods form the backbone of modern facial recognition and other image similarity systems, where the goal is to map images into a latent space that separates identities or classes through clustering. This connection would help emphasize the conceptual strength of the approach and align the work with well-established applications. Similarly, there is a growing literature on animal re-identification (ReID), which often involves learning identity-preserving representations across time or appearance changes. Referencing these bodies of work would help readers connect the proposed method with adjacent areas using similar ideas, and show that the authors are aware of and building on this wider context.

(4) Some sections of the Results text (e.g., lines 48-74) read more like extended figure captions than part of the main narrative. They include detailed explanations of figure elements, sorting procedures, and video naming conventions that may be better placed in the actual figure captions or moved to supplementary notes. Streamlining this section in the main text would improve readability and help the central ideas stand out more clearly.

Overall, though, this is a high-quality paper. The improvements to idtracker.ai are well justified and practically significant. Addressing the above comments will strengthen the work, particularly by clarifying the evaluation and comparisons.

Reviewer #2 (Public review):

This work introduces a new version of the state-of-the-art idtracker.ai software for tracking multiple unmarked animals. The authors aimed to solve a critical limitation of their previous software, which relied on the existence of "global fragments" (video segments where all animals are simultaneously visible) to train an identification classifier network, in addition to addressing concerns with runtime speed. To do this, the authors have both re-implemented the backend of their software in PyTorch (in addition to numerous other performance optimizations) as well as moving from a supervised classification framework to a self-supervised, contrastive representation learning approach that no longer requires global fragments to function. By defining positive training pairs as different images from the same fragment and negative pairs as images from any two co-existing fragments, the system cleverly takes advantage of partial (but high-confidence) tracklets to learn a powerful representation of animal identity without direct human supervision. Their formulation of contrastive learning is carefully thought out and comprises a series of empirically validated design choices that are both creative and technically sound. This methodological advance is significant and directly leads to the software's major strengths, including exceptional performance improvements in speed and accuracy and a newfound robustness to occlusion (even in severe cases where no global fragments can be detected). Benchmark comparisons show the new software is, on average, 44 times faster (up to 440 times faster on difficult videos) while also achieving higher accuracy across a range of species and group sizes. This new version of idtracker.ai is shown to consistently outperform the closely related TRex software (Walter & Couzin, 2021\), which, together with the engineering innovations and usability enhancements (e.g., outputs convenient for downstream pose estimation), positions this tool as an advancement on the state-of-the-art for multi-animal tracking, especially for collective behavior studies.

Despite these advances, we note a number of weaknesses and limitations that are not well addressed in the present version of this paper:

(1) The contrastive representation learning formulation

Contrastive representation learning using deep neural networks has long been used for problems in the multi-object tracking domain, popularized through ReID approaches like DML (Yi et al., 2014\) and DeepReID (Li et al., 2014). More recently, contrastive learning has become more popular as an approach for scalable self-supervised representation learning for open-ended vision tasks, as exemplified by approaches like SimCLR (Chen et al., 2020), SimSiam (Chen et al., 2020\), and MAE (He et al., 2021\) and instantiated in foundation models for image embedding like DINOv2 (Oquab et al., 2023). Given their prevalence, it is useful to contrast the formulation of contrastive learning described here relative to these widely adopted approaches (and why this reviewer feels it is appropriate):

(1.1) No rotations or other image augmentations are performed to generate positive examples. These are not necessary with this approach since the pairs are sampled from heuristically tracked fragments (which produces sufficient training data, though see weaknesses discussed below) and the crops are pre-aligned egocentrically (mitigating the need for rotational invariance).

(1.2) There is no projection head in the architecture, like in SimCLR. Since classification/clustering is the only task that the system is intended to solve, the more general "nuisance" image features that this architectural detail normally affords are not necessary here.

(1.3) There is no stop gradient operator like in BYOL (Grill et al., 2020\) or SimSiam. Since the heuristic tracking implicitly produces plenty of negative pairs from the fragments, there is no need to prevent representational collapse due to class asymmetry. Some care is still needed, but the authors address this well through a pair sampling strategy (discussed below).

(1.4) Euclidean distance is used as the distance metric in the loss rather than cosine similarity as in most contrastive learning works. While cosine similarity coupled with L2-normalized unit hypersphere embeddings has proven to be a successful recipe to deal with the curse of dimensionality (with the added benefit of bounded distance limits), the authors address this through a cleverly constructed loss function that essentially allows direct control over the intra- and inter-cluster distance (D\_pos and D\_neg). This is a clever formulation that aligns well with the use of K-means for the downstream assignment step.

No concerns here, just clarifications for readers who dig into the review. Referencing the above literature would enhance the presentation of the paper to align with the broader computer vision literature.

(2) Network architecture for image feature extraction backbone

As most of the computations that drive up processing time happen in the network backbone, the authors explored a variety of architectures to assess speed, accuracy, and memory requirements. They land on ResNet18 due to its empirically determined performance. While the experiments that support this choice are solid, the rationale behind the architecture selection is somewhat weak. The authors state that:

"\[W\]e tested 23 networks from 8 different families of state-of-the-art convolutional neural network architectures, selected for their compatibility with consumer-grade GPUs and ability to handle small input images (20 × 20 to 100 × 100 pixels) typical in collective animal behavior videos."

(2.1) Most modern architectures have variants that are compatible with consumer-grade GPUs. This is true of, for example, HRNet (Wang et al., 2019), ViT (Dosovitskiy et al., 2020), SwinT (Liu et al., 2021), or ConvNeXt (Liu et al., 2022), all of which report single GPU training and fast runtime speeds through lightweight configuration or subsequent variants, e.g., MobileViT (Mehta et al., 2021). The authors may consider revising that statement or providing additional support for that claim (e.g., empirical experiments) given that these have been reported to outperform ResNet18 across tasks.

(2.2) The compatibility of different architectures with small image sizes is configurable. Most convolutional architectures can be readily adapted to work with smaller image sizes, including 20x20 crops. With their default configuration, they lose feature map resolution through repeated pooling and downsampling steps, but this can be readily mitigated by swapping out standard convolutions with dilated convolutions and/or by setting the stride of pooling layers to 1, preserving feature map resolution across blocks. While these are fairly straightforward modifications (and are even compatible with using pretrained weights), an even more trivial approach is to pad and/or resize the crops to the default image size, which is likely to improve accuracy at a possibly minimal memory and runtime cost. These techniques may even improve the performance with the architectures that the authors did test out.

(2.3) The authors do not report whether the architecture experiments were done with pretrained or randomly initialized weights.

(2.4) The authors do not report some details about their ResNet18 design, specifically whether a global pooling layer is used and whether the output fully connected layer has any activation function. Additionally, they do not report the version of ResNet18 employed here, namely, whether the BatchNorm and ReLU are applied after (v1) or before (v2) the conv layers in the residual path.

(3) Pair sampling strategy

The authors devised a clever approach for sampling positive and negative pairs that is tailored to the nature of the formulation. First, since the positive and negative labels are derived from the co-existence of pretracked fragments, selection has to be done at the level of fragments rather than individual images. This would not be the case if one of the newer approaches for contrastive learning were employed, but it serves as a strength here (assuming that fragment generation/first pass heuristic tracking is achievable and reliable in the dataset). Second, a clever weighted sampling scheme assigns sampling weights to the fragments that are designed to balance "exploration and exploitation". They weigh samples both by fragment length and by the loss associated with that fragment to bias towards different and more difficult examples.

(3.1) The formulation described here resembles and uses elements of online hard example mining (Shrivastava et al., 2016), hard negative sampling (Robinson et al., 2020\), and curriculum learning more broadly. The authors may consider referencing this literature (particularly Robinson et al., 2020\) for inspiration and to inform the interpretation of the current empirical results on positive/negative balancing.

(4) Speed and accuracy improvements

The authors report considerable improvements in speed and accuracy of the new idTracker (v6) over the original idTracker (v4?) and TRex. It's a bit unclear, however, which of these are attributable to the engineering optimizations (v5?) versus the representation learning formulation.

(4.1) Why is there an improvement in accuracy in idTracker v5 (L77-81)? This is described as a port to PyTorch and improvements largely related to the memory and data loading efficiency. This is particularly notable given that the progression went from 97.52% (v4; original) to 99.58% (v5; engineering enhancements) to 99.92% (v6; representation learning), i.e., most of the new improvement in accuracy owes to the "optimizations" which are not the central emphasis of the systematic evaluations reported in this paper.

(4.2) What about the speed improvements? Relative to the original (v4), the authors report average speed-ups of 13.6x in v5 and 44x in v6. Presumably, the drastic speed-up in v6 comes from a lower Protocol 2 failure rate, but v6 is not evaluated in Figure 2 - figure supplement 2.

(5) Robustness to occlusion

A major innovation enabled by the contrastive representation learning approach is the ability to tolerate the absence of a global fragment (contiguous frames where all animals are visible) by requiring only co-existing pairs of fragments owing to the paired sampling formulation. While this removes a major limitation of the previous versions of idtracker.ai, its evaluation could be strengthened. The authors describe an ablation experiment where an arc of the arena is masked out to assess the accuracy under artificially difficult conditions. They find that the v6 works robustly up to significant proportions of occlusions, even when doing so eliminates global fragments.

(5.1) The experiment setup needs to be more carefully described.<br /> What does the masking procedure entail? Are the pixels masked out in the original video or are detections removed after segmentation and first pass tracking is done?<br /> What happens at the boundary of the mask? (Partial segmentation masks would throw off the centroids, and doing it after original segmentation does not realistically model the conditions of entering an occlusion area.)<br /> Are fragments still linked for animals that enter and then exit the mask area?<br /> How is the evaluation done? Is it computed with or without the masked region detections?

(5.2) The circular masking is perhaps not the most appropriate for the mouse data, which is collected in a rectangular arena.

(5.3) The number of co-existing fragments, which seems to be the main determinant of performance that the authors derive from this experiment, should be reported for these experiments. In particular, a "number of co-existing fragments" vs accuracy plot would support the use of the 0.25(N-1) heuristic and would be especially informative for users seeking to optimize experimental and cage design. Additionally, the number of co-existing fragments can be artificially reduced in other ways other than a fixed occlusion, including random dropout, which would disambiguate it from potential allocentric positional confounds (particularly relevant in arenas where egocentric pose is correlated with allocentric position).

(6) Robustness to imaging conditions

The authors state that "the new idtracker.ai can work well with lower resolutions, blur and video compression, and with inhomogeneous light (Figure 2 - figure supplement 4)." (L156).

Despite this claim, there are no speed or accuracy results reported for the artificially corrupted data, only examples of these image manipulations in the supplementary figure.

(7) Robustness across longitudinal or multi-session experiments

The authors reference idmatcher.ai as a compatible tool for this use case (matching identities across sessions or long-term monitoring across chunked videos), however, no performance data is presented to support its usage.

This is relevant as the innovations described here may interact with this setting. While deep metric learning and contrastive learning for ReID were originally motivated by these types of problems (especially individuals leaving and entering the FOV), it is not clear that the current formulation is ideally suited for this use case. Namely, the design decisions described in point 1 of this review are at times at odds with the idea of learning generalizable representations owing to the feature extractor backbone (less scalable), low-dimensional embedding size (less representational capacity), and Euclidean distance metric without hypersphere embedding (possible sensitivity to drift).

It's possible that data to support point 6 can mitigate these concerns through empirical results on variations in illumination, but a stronger experiment would be to artificially split up a longer video into shorter segments and evaluate how generalizable and stable the representations learned in one segment are across contiguous ("longitudinal") or discontiguous ("multi-session") segments.

Reviewer #1 (Public review):

The authors' goal was to arrest PsV capsids on the extracellular matrix using cytochalasin D. The cohort was then released, and interaction with the cell surface, specifically with CD151, was assessed.

The model that fragmented HS associated with released virions mediates the dominant mechanism of infectious entry has only been suggested by research from a single laboratory and has not been verified in the 10+ years since publication. The authors are basing this study on the assumption that this model is correct, and these data are referred to repeatedly as the accepted model despite much evidence to the contrary. The discussion in lines 65-71 concerning virion and HSPG affinity changes is greatly simplified. The structural changes in the capsid induced by HS interaction and the role of this priming for KLK8 and furin cleavage have been well researched. Multiple laboratories have independently documented this. If this study aims to verify the shedding model, additional data need to be provided. The model should be fitted into established entry events, or at minimum, these conflicting data, a subset of which is noted below, need to be acknowledged.

(1) The Sapp lab (Richards et al., 2013) found that HSPG-mediated conformational changes in L1 and L2 allowed the release of the virus from primary binding and allowing secondary receptor engagements in the absence of HS shedding.

(2) Becker et al. found that furin-precleaved capsids could infect cells independently of HSPG interaction, but this infection was still inhibited with cytochalasin D.

(3) Other work from the Schelhaas lab showed that cytochalasin D inhibition of infection resulted in the accumulation of capsids in deep invaginations from the cell surface, not on the ECM.

(4) Selinka et al., 2007, showed that preventing HSPG-induced conformational changes in the capsid surface resulted in noninfectious uptake that was not prevented with cytochalasin D.

(5) The well-described capsid processing events by KLK8 and furin need to be mechanistically linked to the proposed model. Does inhibition of either of these cleavages prevent engagement with CD151?

The authors need to consider an explanation for these discrepancies.

Other issues:

(1) Line 110-111. The statement about PsVs in the ECM being too far away from the cell surface to make physical contact with the cell surface entry receptors is confusing. ECM binding has not been shown to be an obligatory step for in vitro infection. This idea is referred to again on lines 158-159 and 199. The claim (line 158) that PsV does not interact with the cell within an hour needs to be demonstrated experimentally and seems at odds with multiple laboratories' data. PsV has been shown to directly interact with HSPG on the cell surface in addition to the ECM. Why are these PsVs not detected?

(2) The experiments shown in Figure 5 need to be better controlled. Why is there no HS staining of the cell surface at the early timepoints? This antibody has been shown to recognize N-sulfated glucosamine residues on HS and, therefore, detects HSPG on the ECM and cell surface. Therefore, the conclusion that this confirms HS coating of PsV during release from the ECM (line 430-431) is unfounded. How do the authors distinguish between "HS-coated virions" and HSPG-associated virions?

It is difficult to comprehend how the addition of 50 vge/cell of PsV could cause such a global change in HS levels. The claim that the HS levels are decreased in the non-cytochalasin-treated cells due to PsV-induced shedding needs to be demonstrated. If HS is actually shed, staining of the cell periphery could increase with the antibody 3G10, which detects the HS neoepitope created following heparinase cleavage.

Author response:

Reviewer #1 (Public review):

The authors' goal was to arrest PsV capsids on the extracellular matrix using cytochalasin D. The cohort was then released, and interaction with the cell surface, specifically with CD151, was assessed.

The model that fragmented HS associated with released virions mediates the dominant mechanism of infectious entry has only been suggested by research from a single laboratory and has not been verified in the 10+ years since publication. The authors are basing this study on the assumption that this model is correct, and these data are referred to repeatedly as the accepted model despite much evidence to the contrary.

Please note that we state in the introduction on line 65/66 ´Two release mechanisms are discussed, that mutually are not exclusive´. This is implying that we do not consider the shedding model as the one accepted model. HS may associate with PsVs despite of a decreased affinity and only after priming (see below the ‘priming model’) may translocate to the cell body.

Furthermore, we do not state in the discussion either that the shedding model is the preferred one; although it is correct that we refer to the shedding model more extensively, simply because we find HS associated with transferred PsVs, which is in line with this model and requires its citation.

The discussion in lines 65-71 concerning virion and HSPG affinity changes is greatly simplified. The structural changes in the capsid induced by HS interaction and the role of this priming for KLK8 and furin cleavage have been well researched. Multiple laboratories have independently documented this. If this study aims to verify the shedding model, additional data need to be provided.

As outlined above, our finding is compatible with both models, and we do not aim to verify the shedding model or disprove the priming model.

It appears that the referee wishes more visibility of the priming model. Inhibition of KLK8 and furin should reduce the translocation to the cell body, no matter whether PsVs carry HS on their surface or not. For revision, we plan an experiment as in Figure 3 (CytD), testing whether either KLK8 or furin inhibition blocks the transfer to the cell body. Then, our data can be discussed also in the context of the priming model and by this increase its visibility.

The model should be fitted into established entry events, or at minimum, these conflicting data, a subset of which is noted below, need to be acknowledged.

(1) The Sapp lab (Richards et al., 2013) found that HSPG-mediated conformational changes in L1 and L2 allowed the release of the virus from primary binding and allowing secondary receptor engagements in the absence of HS shedding.

(2) Becker et al. found that furin-precleaved capsids could infect cells independently of HSPG interaction, but this infection was still inhibited with cytochalasin D.

(3) Other work from the Schelhaas lab showed that cytochalasin D inhibition of infection resulted in the accumulation of capsids in deep invaginations from the cell surface, not on the ECM

(4) Selinka et al., 2007, showed that preventing HSPG-induced conformational changes in the capsid surface resulted in noninfectious uptake that was not prevented with cytochalasin D.

(5) The well-described capsid processing events by KLK8 and furin need to be mechanistically linked to the proposed model. Does inhibition of either of these cleavages prevent engagement with CD151?

The authors need to consider an explanation for these discrepancies.

That PsVs carry HS-cleavage products doesn´t imply that HS cleavage is sufficient or required for infection. Therefore, we do not view our data as being in conflict with the priming model. In fact, our observations are compatible with aspects of both the shedding and the priming model.

Yet, we acknowledge that the study would gain importance by directly testing the priming model within our experimental system. As requested by the referee, we will discuss the above papers, and further plan to test KLK8 and furin inhibitors.

Other issues:

(1) Line 110-111. The statement about PsVs in the ECM being too far away from the cell surface to make physical contact with the cell surface entry receptors is confusing. ECM binding has not been shown to be an obligatory step for in vitro infection.

Not obligatory, but strongly supportive (Bienkowska-Haba et al., Plos Path., 2018; Surviladze et al., J. Gen. Viro., 2015). As recently published by the Sapp lab (Bienkowska-Haba et al., Plos Path., 2018), ´Direct binding of HPV16 to primary keratinocytes yields very inefficient infection rates for unknown reasons.´ Moreover, the paper shows that HaCaT cell ECM binding of PsVs increases the infection of NHEK by 10-fold and of HFK by almost 50-fold.

This idea is referred to again on lines 158-159 and 199. The claim (line 158) that PsV does not interact with the cell within an hour needs to be demonstrated experimentally and seems at odds with multiple laboratories' data. PsV has been shown to directly interact with HSPG on the cell surface in addition to the ECM. Why are these PsVs not detected?

We do not question that in many cellular systems PsVs interact with heparan sulfate proteoglycans (HSPGs) present on the cell surface, or both on the cell surface and the ECM. We stated in the manuscript on line 59 ´While in cell culture virions bind to HS of the cell surface and the ECM, it has been suggested that in vivo they bind predominantly to HS of the extracellular basement membrane (Day and Schelhaas, 2014; Kines et al., 2009; Schiller et al., 2010).´

Moreover, we ourselves detect these PsVs, for example, in Figure 5A (CytD, 0 min time point), a handful of PsVs localize to the cell body area. However, the large majority overlaps with the strong HS staining at the cell periphery, likely the ECM. An accurate quantification of the fractions of PsVs bound to the ECM/cell body is for the following reasons very difficult. First, the ECM PsVs are very dense and therefore not microscopically resolved into single PsVs, at least not completely (see Figure 1C; the high intensity spots are non-resolved PsVs, please see our discussion on line 148 - 152). For this reason, by just counting spots we strongly underestimate the ECM PsVs versus the cell body PsVs. Second, with the available immunostainings we cannot exactly delineate the ECM from the cell body. In particular, at the cell border region (for example see Figure 4B) we often observe PsV accumulations. Assigning these ´cell border region PsVs´ entirely to the cell body fraction, a preliminary analysis (correcting for the limitation of non-resolved ECM PsVs) suggests that about a quarter of the PsVs bind to the cell body. On the other hand, assigning them to the ECM, the cell body fraction would be much below 10%. Third, we observe that in regions devoid of ECM and cells PsVs apparently adhere unspecifically to the glass-coverslip. This suggests that some of the cell body PsVs are just unspecific background. Subtraction of a background PsV density from the ECM and cell body PsV density will reduce relatively more the cell body PsVs, and consequently decreases the fraction of cell body PsVs even more.

Moreover, in the course of the project we wondered whether at the basolateral membrane there are not many binding sites anyway. To address this question, in an unpublished experiment, we detached HaCaT cells with trypsin, incubated them with PsVs, and then allowed reattachment to assess the binding in suspension. We detected minimal to no binding, which, however, could also result from apical membrane adherence to the coverslip or trypsin-mediated cleavage of HSPGs. As suggested by the reviewing editor, we agree that repeating this experiment using EDTA for detachment—thus preserving HSPGs—would offer more definitive insight into binding efficiency in the absence of accessibility constraints. In summary, the reason why in our cellular system most PsVs do not bind to the cell surface could be a combination of several factors:

(1) The primary binding partners are more abundant in the ECM and the polarized HaCaT cells secrete more ECM when compared to other cultured cells used to study HPV infection. This promotes ECM binding.

(2) In the polarized HaCaT cells, the apical membrane is largely devoid of syndecan-1, CD151 and Itga6, wherefore PsVs infect the cell via the basolateral membrane. However, the accessibility to the basolateral membrane is restricted, PsVs must diffuse through a narrow slit between the glass coverslip and the attached cell to reach HS on the cell surface. This limits cell surface binding.

(3) If HaCaT cells secrete large amounts of ECM, the may become depleted from cell surface HS. As outlined above, we will try to find out how many PsVs bind to the basolateral membrane in the absence of restricted accessibility. If it turns out that HaCaT cells have not many binding sites anyway, this would additionally promote binding to the ECM.

The outcome of the above issues, and how we will mention them in the revised version of the manuscript, is open. In any case, we would like to point out that PsVs bound to the cell body do not weaken our main conclusion. Still, we recognize that this point merits attention and plan several modifications of the manuscript. We did already, but now we will mention more explicitly that PsVs have been shown to directly interact with HSPG on the cell surface, in addition to the ECM, but that it also has been shown that the ECM strongly supports infection in NHEK and HFK (Bienkowska-Haba et al., Plos Path., 2018). The following is a draft version of a paragraph we plan to incorporate, explaining the above issue and why we used in our experiments HaCaT cells:

´In vitro, PsVs bind to both the cell surface and the ECM, as has been widely documented. In vivo, however, it has been proposed that initial binding occurs predominantly to the basement membrane ECM, rather than directly to the cell surface (Day and Schelhaas, 2014; Kines et al., 2009; Schiller et al., 2010). This distinction reinforces the physiological relevance of ECM-bound particles in the early steps of HPV infection. Support for a functional role of ECM-mediated entry comes from a study showing that PsV binding to ECM derived from HaCaT cells significantly enhances infection of primary keratinocytes (Bienkowska-Haba et al., 2018). For these reasons, we specifically chose polarized HaCaT cells as a model system. These cells secrete abundant ECM from which the cells readily collect bound PsVs. On the other hand, the polarization limits the access of PsVs to basolateral receptors such as CD151 and Itgα6, and also cell body resident Syndecan-1, the most abundant HSPG in keratinocytes (Rapraeger et al., 1986; Hayashi et al., 1987; Kim et al., 1994). Hence, as polarization limits direct cell surface accessibility it biases binding toward the ECM, that in this culture system is abundant. Hence, in the HaCaT cell culture system, like probably in vivo, PsVs cannot circumvent binding to the ECM what they can do in unpolarized cell cultures that may not even secrete significant amounts of ECM. Altogether, this experimental situation closely mimics the in vivo situation where PsVs bind preferentially to the ECM (Day and Schelhaas, 2014; Kines et al., 2009; Schiller et al., 2010).´

We appreciate the reviewer’s input and believe these additions will strengthen the manuscript with regard to the relevance of the used cellular model system.

(2) The experiments shown in Figure 5 need to be better controlled. Why is there no HS staining of the cell surface at the early timepoints? This antibody has been shown to recognize N-sulfated glucosamine residues on HS and, therefore, detects HSPG on the ECM and cell surface.

We have shown all images at the same adjustments of brightness and contrast. As the staining at the periphery is stronger, the impression is given that the cell surface is not stained, although there is some staining. Specific staining is documented in Figure 5D, showing the PCC between PsVs and HS only of the cell body. If there was no HS staining, the PCC would be zero, which is not the case. Yet, it is lower when compared to the PCC measured at the cell border region, with more strongly stained HS.

We will provide images at different contrast and brightness adjustments enabling the reader to see the staining on the cell surface. We will provide also more overview images to illustrate the strong variability of the HS staining between cells.

Therefore, the conclusion that this confirms HS coating of PsV during release from the ECM (line 430-431) is unfounded. How do the authors distinguish between "HS-coated virions" and HSPG-associated virions?

The HS intensity transiently increases on the cell body (Fig. 5D) only after releasing a cohort of PsVs, which can be only explained by PsVs that carry HS from the ECM to the cell body. However, the effect is not significant. Using the antibody 3G10 detecting the HS neoepitope (see the referees’ suggestion below) we will reanalyze this point. This should help clarifying the issue.

It is difficult to comprehend how the addition of 50 vge/cell of PsV could cause such a global change in HS levels.

The distribution of bound PsVs largely varies between cells. Some areas are covered with essentially confluent cells, to which hardly any PsVs are bound, because accessing the basolateral membrane of confluent cells is nearly impossible, and PsVs do not bind to the exposed apical membrane. This is different in cultures of unpolarized cells where we expect that PsVs distribute more equally over cells.

This means that in our experiments the vge/cell is not a suitable parameter for relating the magnitude of an effect to a defined number of PsVs. In the ECM, the PsV density is very high, enabling one cell to collect several hundred PsVs, much more than expected from the 50 vge/cell. We will point this out in the revised version.

The claim that the HS levels are decreased in the non-cytochalasin-treated cells due to PsV-induced shedding needs to be demonstrated.

We did not claim that PsVs induce shedding, we rather believe they just take shedded HS with them. Without PsVs, the shedded HS likely remains in the ECM or is washed out very slowly.

If HS is actually shed, staining of the cell periphery could increase with the antibody 3G10, which detects the HS neoepitope created following heparinase cleavage.

As outlined above, we plan to test the suggested antibody 3G10. We also plan to repeat the 0 min time point (with and without PsVs, with and without CytD) to find out whether in the PsV absence the HS intensity (at 0 min) is unchanged between control and CytD.

Reviewer #2 (Public review):

Summary:

Massenberg and colleagues aimed to understand how Human papillomavirus particles that bind to the extracellular matrix (ECM) transfer to the cell body for later uptake, entry, and infection. The binding to ECM is key for getting close to the virus's host cell (basal keratinocytes) after a wounding scenario for later infection in a mouse vaginal challenge model, indicating that this is an important question in the field.

Strengths:

The authors take on a conceptually interesting and potentially very important question to understand how initial infection occurs in vivo. The authors confirm previous work that actin-based processes contribute to virus transport to the cell body. The superresolution microscopy methods and data collection are state-of-the art and provide an interesting new way of analysing the interaction with host cell proteins on the cell surface in certain infection scenarios. The proposed hypothesis is interesting and, if substantiated, could significantly advance the field.

Weaknesses:

As a study design, the authors use infection of HaCaT keratinocytes, and follow virus localisation with and without inhibition of actin polymerisation by cytochalasin D (cytoD) to analyse transfer of virions from the ECM to the cell by filopodial structures using important cellular proteins for cell entry as markers.

First, the data is mostly descriptive besides the use of cytoD, and does not test the main claim of their model, in which virions that are still bound to heparan sulfate proteoglycans are transferred by binding to tetraspanins along filopodia to the cell body.

The study identifies a rapid translocation step from the ECM to the cell body. We have no data that demonstrates a physical interaction between PsVs and CD151. In the model figure, we draw CD151 as part of the secondary receptor complex. We are sorry for having raised the impression that PsVs would bind directly to CD151 and will rephrase the respective section.

Second, using cytoD is a rather broad treatment that not only affects actin retrograde flow, but also virus endocytosis and further vesicular transport in cells, including exocytosis. Inhibition of myosin II, e.g., by blebbistatin, would have been a better choice as it, for instance, does not interfere with endocytosis of the virus.

We agree, and plan to test whether blebbistatin is equally efficient in blocking the transfer.

Third, the authors aim to study transfer from ECM to the cell body and the effects thereof. However, there are substantial, if not the majority of, viruses that bind to the cell body compared to ECM-bound viruses in close vicinity to the cells.

We agree that in multiple cell culture systems viruses bind preferentially to the cell directly. But we respectfully disagree with the assertion that the majority of PsVs bind to the cell body of HaCaT keratinocytes. As noted above (e.g., Figure 5A, CytD, 0 min), only a small fraction of PsVs localize to the cell body, whereas the vast majority overlap with intense HS staining at the cell periphery, consistent with ECM association, as the accessibility to the basolateral expressed HSPG is limited (see above). Based on quantitative estimation from multiple images, ECM-bound PsVs largely outnumber cell-bound particles (see above). These features make HaCaT cells a suitable in vitro model for mimicking in vivo conditions, where HPV has been proposed to bind predominantly to the basement membrane ECM rather than the cell surface (Day and Schelhaas, 2014; Kines et al., 2009; Schiller et al., 2010) which also strongly enhances infection of primary keratinocytes in vitro (Bienkowska-Haba et al., 2018).

Thus, we believe our system appropriately models the physiologically relevant scenario of ECM-to-cell transfer, and the observed predominance of ECM binding supports the validity of our experimental focus.

This is in part obscured by the small subcellular regions of interest that are imaged by STED microscopy, or by the use of plasma membrane sheets. As a consequence, the obtained data from time point experiments is skewed, and remains for the most part unconvincing due to the fact that the origin of virions in time and space cannot be taken into account. This is particularly important when interpreting association with HS, the tetraspanin CD151, and integral alpha 6, as the low degree of association could originate from cell-bound and ECM-transferred virions alike.

As stated above, we observe massive binding of PsVs to the ECM, in contrast to very few PsVs that diffuse beneath the basolateral membrane of the polarized HaCaT cells and do bind directly to the cell surface (or maybe they are simply trapped between glass and basolateral membrane). PsVs are not expected to bind to the apical membrane that is depleted from CD151 and Itga6. In other cellular systems, cells may hardly secrete ECM, are not polarized, and do not adhere so tightly to the substrate. In other cultures, where virions can easily circumvent ECM binding, the large majority of PsVs will likely bind directly to the cell surface.

As outlined above, in order to quantify PsVs that can bind without restricted accessibility, we plan to detach HaCaT cells by EDTA from the substrate, incubate them with PsVs, and let them adhere again (please see above).

No matter what is the outcome, the fraction of PsVs that binds directly to the cell surface does not weaken our conclusion that we have identified a very fast and efficient transfer step from the ECM to the cell body.

Fourth, the use of fixed images in a time course series also does not allow for understanding the issue of a potential contribution of cell membrane retraction upon cytoD treatment due to destabilisation of cortical actin. Or, of cell spreading upon cytoD washout.

If blebbistatin works as expected, we can safely conclude that we observe the very same process as described in Scheelhas et al., PLoS Pathogens, 2008, showing that the PsVs migrate by retrograde transport to the cell surface and not that the cell spreads out and by this reaches the PsVs.

The microscopic analysis uses an extension of a plasma membrane stain as a marker for ECM-bound virions, which may introduce a bias and skew the analysis.

Our plasma membrane stain does not stain the ECM. Please see Figure 1. The stain is actually used to distinguish the cell body from the ECM area.

Fifth, while the use of randomisation during image analysis is highly recommended to establish significance (flipping), it should be done using only ROIs that have a similar density of objects for which correlations are being established.

We agree that the way of how randomization is done is very important. Regarding the association of PsVs with CD151 and HS, based on flipped images, we generated a calibration curve used for the correction of random background. For details, please see Supplementary Figures 3 and 5.

For instance, if one flips an image with half of the image showing the cell body, and half of the image ECM, it is clear that association with cell membrane structures will only be significant in the original. I am rather convinced that using randomisation only on the plasma membrane ROIs will not establish any clear significance of the correlating signals.

Figure 5D shows the PCC specifically of the cell body. In flipped images (not shown in the manuscript for clarity, but can be added) we obtain a PCC of around zero.  For CytD, the flipped images always have a significantly lower PCC compared to the original images. In the control, the PCC of the flipped images are significantly lower only for the 30 min and 60 min time point. The non-significance of the 0 min and 180 min time point is due to low PCCs also in the original images.

Also, there should be a higher n for the measurements.

One n is the average of 15 cells. We realize that with n = 3 we find significant effects only if the effect is very strong or moderate with very low variance.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Manuscript number: RC-2025-03083 Corresponding author(s): David Fay General Statements [optional] This section is optional. Insert here any general statements you wish to make about the goal of the study or about the reviews.

We greatly appreciate the input of the four reviewers, all of whom carried out a careful reading of our manuscript, provided useful suggestions for improvements, and were enthusiastic about the study including its thoroughness and utility to the field. Because the reviewers required no additional experiments, we were able to address their comments in writing.

However, in response to a comment from reviewer #4 we decided to add an additional new biological finding to our study given that our functional validation of proximity labeling targets was not extensive. Namely, we now show that a missense mutation affecting BCC-1, one of the top NEKL-MLT interactors identified by our proximity labeling screen, is a causative mutation (together with catp-1) in a strain isolated through a forward genetic screen for suppressors of nekl molting defects (new Fig 9C). This finding, combined with our genetic enhancer tests, further strengthens the functional relevance of proteins identified though our proximity labeling approach and highlights the synergy of proteomics combined with classical genetics.

Positive statements from reviewers include: Reviewer #1: Overall, this is an outstanding study that will be of great interest to those interested in using proximity labeling to identify interactors of their favorite protein. The experiments are well executed and the data presented in a mostly clear manner.

Reviewer #2: The key conclusions are convincing, and the work is rigorous. The work provides a clear roadmap to reproducing the data. The experiments are adequately replicated, and statistical analysis is adequate... In many papers, TurboID seems very trivial but this paper clearly highlights the limitations and will be an invaluable resource for labs that want to get proximity labeling established in their labs.

Reviewer #3: Overall, the claims are solid and conclusions supported. The data and methods are substantial to enable reproducibility in other labs. The experiments have been repeated multiple times with particular attention to statistical analysis. ...This manuscript represents a methodological advance that will likely become an oft-cited reference for members of the C. elegans community and a springboard for other basic biomedical scientists wanting to adapt rigorous proximity labeling techniques to their system.

Reviewer #4: Fay et al. present a solid, clear and comprehensive BioID-based proteomics study that takes into account and discusses decisive aspects for the (re)production and analysis of high-quality TurboID-based mass spectrometry data. Claims and conclusions are generally well and sufficiently supported by the presented data and illustrated with figures (throughout the text as well as with plenty of supplementary data)... Basic consideration and thoughts for the experimental design and MS data analysis are given in detail and can serve as another guideline for future studies.

Based on these reviews and comments, we believe that our manuscript is suitable for publication in a high-impact journal. 1. Point-by-point description of the revisions This section is mandatory. Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript.

*Reviewer #1 (Evidence, reproducibility and clarity (Required)): *

*Proximity labeling has become a powerful tool for defining protein interaction networks and has been utilized in a growing number of multicellular model systems. However, while such an approach can efficiently generate a list of potential interactors, knowledge of the most appropriate controls and standardized metrics to judge the quality of the data are lacking. The study by Fay systematically investigates these questions using the C. elegans NIMA kinase family members NEKL-2 and NEKL-2 and their known binding partners MLT-2, MLT-3 and MLT-4. The authors perform eight TurboID experiments each with multiple NEKL and MLT proteins and explore general metrics for assessing experimental outcomes as well as how each of the individual metrics correlates with one another. They also compare technical and biological replicates, explore strategies for identifying false positives and investigate a number of variations in the experimental approach, such as the use of N- versus C-terminal tags, depletion of endogenous biotinylated proteins, combining auxin-inducible degradation, and the use of gene ontology analysis to identify physiological interactors. Finally, the authors validate their findings by demonstrating that a number of the candidate identified functionally interact with NEKL-2 or components of the WASH complex. *

Overall this is an outstanding study that will be of great interest to those interested in using proximity labeling to identify interactors of their favorite protein. The experiments are well executed and the data presented in a mostly clear manner. I really like this study (particularly because I plan to do a proximity labeling study of my own), but I did come away less than impressed with some of the analysis. This is a data-dense manuscript, and it appears to me that the authors tried to cover so much ground that in some cases very little insight was provided. For instance, the authors promote the use of data independent acquisition (DIA) as compared to the more commonly used data dependent acquisition (DDA). However the authors do not provide any analysis to indicate one approach is better than the other. Likewise the combined use of auxin-induced degradation and proximity labeling is explored but there is very little to take away from these experiments. Despite these issues, I am very enthusiastic about the study as a whole. Below I list major and minor concerns.

Major concerns * 1. My biggest issue with the manuscript is that a lot is made of the use of data independent acquisition (DIA) as compared to the more commonly used data dependent acquisition (DDA). The authors perform experiments using DIA and DDA approaches but do not directly compare the outcomes. As a result there is really no way to know if one approach is better than the other. I would suggest the authors either perform the necessary analysis to compare the two approaches or tone down their promotion of DIA.* We agree and have scaled back any statements comparing DDA to DIA as our manuscript did not address this directly. We also now point out this caveat in our closing thoughts section, while referencing other studies that compared the two (lines 926-929). Our main point was to convey that DIA worked well for our proximity labeling studies but has seen little use by the model organism field. Surprising (to us), DIA was also considerably less expensive than DDA options.

2. Line 75, The authors promote the use of data-independent acquisition (DIA) without defining what this approach is and how it differs from the more conventional data-dependent acquisition. As a non-mass spectroscopist, I found myself with lots of question concerning DIA, what it is and how it differs from DDA. I think it would really be helpful to expand the description of DIA and its comparison with DDA in the introduction. As non-mass-spectroscopists ourselves, we understand the reviewer's point. Because the paper is quite long, we were trying to avoid non-essential information. We have now added some information to explain some of the key differences between DDA and DIA. We have also included references for readers who may want to learn more. (lines 77-80)

Minor concerns: * Line 92 typo. I believe the authors meant to say NEKL-2-MLT-2-MLT-4. * Corrected. (line 95)

Line169. Is exogenous the correct word to use here? It suggests that you are talking about non-worm proteins, but I know you are not. Corrected. Changed to "Moreover, the detection of biotinylated proteins may be difficult if the bait-TurboID fusion is expressed at low levels..." (line 181).

Line 177 typo (D) should be (C). Corrected. (line 1122)

Figure 1C: Lucky Charms may sue you for infringement of their trademarked marshmallow treats. Thank you for picking up on this. The authors accept full responsibility for any resulting lawsuits.

Figure 1D. The NEKL-2::TurboID band is indicated with a green triangle in the figure but the figure legend states that green triangles indicate mNG::TurboID control. I know this triangle is a shade off the triangle that indicates mNG::TurboID but it's really hard to see the difference. All of the differently colored triangles in panel F are unnecessary. I would either just pick one color for all non-control bait proteins or better yet, only use a triangle to point to bands that are not obvious. For instance I don't need the triangles that point to NEKL-2 -3 and -4 fusion proteins. These are just distracting. We understand the reviewer's point. We colored the triangles to match the colors used for the proteins in the figures. We have now added "bright green triangles with white outlines" (Fig 1 legend) to indicate the Pdpy-7::mNG::TurboID control" and changed triangles in the corresponding figures. Although we would be fine with removing or changing the triangles, we think that they may aid somewhat with clarity.

Line: 316: Conceivably, another factor that could contribute to the counterintuitive upregulation of some proteins in the N2 samples is related to the fusion proteins that are being expressed in the TurboID lines. A partially functional bait protein (one with a level of activity similar to nekl-2(fd81) that may not result in an obvious phenotype) could directly or indirectly affect gene expression leading to lower levels of a subset of proteins in the TurboID samples. The same could be said for fusion proteins with a gain-of-function effect. This is an interesting idea, and we tested this possibility by looking for consistent overlap between N2-up proteins between biological replates of individual bait proteins. We now include a representative Venn diagram in S3C Fig to highlight this comparison. In summary, although we cannot rule out this possibility, our analysis did not support the widespread occurrence of this effect in our study. We also made certain that our statement regarding N2 up proteins was not too definitive. (lines 285-288)

*Fig 3 B-E. I am a little confused how the data in these graphs is normalized. For instance, I would have expected that for NEKL-3 in panel B, that the normalized (log2) intensity value in N2 be set at 0 as it is for NEKL-2. Maybe I just don't have enough information on how these plots were generated. * The difference is that in the N2 sample, NEKL-3 was detected but NEKL-2 was not. The numbers themselves are assigned by the Spectronaut software used to quantify the DIA results but are not meaningful beyond indicating relative amounts (intensity values) of a given protein within an individual biological experiment. We've added some lines to the figure legend to make this clearer. (lines 1165-1169)

*Figure 6C legend is not correct. * Corrected. (line 1214)

Line 575: Figure reference should be Fig. S5G. The authors should check to make sure all references to supplemental figures include correct panel information. Corrected. (line 464) In addition, we have now gone through the manuscript and added panel numbers references where applicable. Note that the addition of a new supplemental file has shifted the numbering.

Line 576. The authors reference a study by Artan and colleagues and report a weak correlation between their study and that of Artan. They reference figure S4 but it should be Fig S5H. Apologies and many thanks to the reviewer for catching these errors. (line 464)

Line 652. The authors note that numerous proteins were present at substantially reduced levels in the mNG::TurboID samples and suggest that sticky proteins may have been outcompeted or otherwise excluded from beads incubated with the mNG::TurboID lysates. Why would sticky proteins only be a problem in these samples? The reasoning is not clear to me. The idea was that in the sample with very high levels of biotinylated proteins (mNG::TurboID), the surface of the beads might become saturated with high-affinity biotinylated proteins. This could prevent or out complete the binding of random proteins that are not biotinylated but nevertheless have some affinity to the beads ("sticky" proteins). We have reworded this section to make this clearer. (lines 546-550)

Line 745: The term "bait overlaps" is a bit vague. Ultimately, I figured out what it meant but it was not immediately obvious. We have changed this to "overlap between baits" and made this section clearer. (line 624-628)

*S7B Fig. Why is actin missing from the eluate? * In S7B we refer to the purified eluate as the "eluate", which may have caused some confusion. In other sections of the manuscript, we refer to the bead-bound proteins as the "purified eluate" (Figs 1 and 5). For the purified eluate a portion of the streptavidin beads are boiled in sample buffer to elute the bound proteins before running a western. Actin would not be expected in these samples because it's (presumably) not biotinylated in our samples and doesn't detectably bind the beads. This result was seen in all relevant westerns in S1 Data. For consistency, however, we've gone through all our files to make sure we consistently use the term "purified eluate" versus "eluate", which is less specific.

L*ine 873: The authors state the extent of overlap in GO terms between the various experiments and provide percentages. I tried to extract this information from Figure 8C and came up with different values. For instance, in the case of Molecular Function, they state that they observed a 54% overlap between NEKL-2 and NEKL-3 but in the Venn diagram in Figure 8C I see that the NEKL-2 and NEKL-3 experiments had 71 (25+46) GO terms in common. Out of 98 GO terms for NEKL-2 or 104 for NEKL-3 the percentage I got is closer to 72. Am I analyzing this correctly? * Thanks for checking this. We believe our method for calculating the percent overlap is correct. In the case of NEKL-2/NEKL-3 overlap for Molecular Function, there are 131 total unique terms, of which 71 overlap, giving a 54% overlap. In the case of NEKL-2/NEKL-3 overlap for Biological Process, however, we made an error in arithmetic (415 unique, 239 overlap), such that the correct percentage is 58%, which we have corrected in the text.

*Reviewer #1 (Significance (Required)): *

*Overall this is an outstanding study that will be of great interest to those interested in using proximity labeling to identify interactors of their favorite protein. The experiments are well executed and the data presented in a mostly clear manner. I really like this study (particularly because I plan to do a proximity labeling study of my own), but I did come away less than impressed with some of the analysis. This is a data-dense manuscript, and it appears to me that the authors tried to cover so much ground that in some cases very little insight was provided. For instance, the authors promote the use of data independent acquisition (DIA) as compared to the more commonly used data dependent acquisition (DDA). However the authors do not provide any analysis to indicate one approach is better than the other. Likewise the combined use of auxin-induced degradation and proximity labeling is explored but there is very little to take away from these experiments. Despite these issues, I am very enthusiastic about the study as a whole. *

*Reviewer #2 (Evidence, reproducibility and clarity (Required)): *

*This study expanded the use of data-independent acquisition-mass spectrometry (DIA-MS) in TurboID proximity-labeling proteomics to identify novel interactors of NEKL-2, NEKL-3, MLT-2, MLT-3, and MLT-4 complexes in C. elegans. The authors described several useful metrics to evaluate the quality of TurboID experiments, such as using the percentage of upregulated genes, the percentage of proteins present only in bait-TurboID experiments as compared to N2 controls, and the percentage of endogenously biotinylated carboxylases as internal controls. Further, the authors introduced methodological variability across 23 TurboID experiments and evaluated any improvement to the resulting data, such as N-terminally tagging bait proteins with TurboID, depleting endogenous carboxylases, and auxin-inducible degradation of known complex members. Finally, this study identified the kinase folding chaperone CDC-37 and the WASH complex component DDL-2 as novel interactors with the NEKL-MLT complexes through an RNAi-based enhancer approach following their identification by TurboID. *

Major comments: * The key conclusions are convincing, and the work is rigorous. The work provides a clear roadmap to reproducing the data. The experiments are adequately replicated, and statistical analysis is adequate. We only have minor comments.*

Minor comments: * •In the western blot in Fig 1 why does the mNG::Turbo have two bands? * Thank you for point this out. To our knowledge this is a breakdown product that was especially prevalent in replicate 3 (also see S1 Data), which we chose to shown because all the NEKL-MLTs were clearly visible in this western. The expected size of the mNeonGreen::TurboID (including linker and tags) is ~68 kDa and our blots are roughly consistent those of Artan et al., (2001). This lower band was not evident in Exp 8. We have now included a statement in the figure legend to indicate that the upper band is the full-length protein whereas the lower band is likely to be a breakdown product (lines 1141-1142).

•Fig 2B is difficult to parse as a reader. Columns labeled "Upreg," "Downreg," "TurboID only," "N2 only," "Filter-1," "Filter-2," and "Epi %" could be moved to Supplemental. Fold change vs N2 could be represented as a bar chart, allowing for trends between fold change and the metrics Upreg %, Turbo %, and Carboxylase % to be seen more clearly. Further, rows headed "Carboxylase depletion," "DDA," and "Auxin treated" could be presented as separate panels to better match the distinct points made in the text. After serious consideration we have made several changes including the addition of S2 Fig, which may provide readers with a better visual representation of the bait and prey fold changes observed in all our experiments. However, we feel that the detailed data embedded in Fig 2 is the most concise and accurate means by which to convey our full results and is key to our methodological conclusions. As such we did not want to relegate this information to a supplemental table. We note that this figure was not found to be problematic by other reviewers, although we do understand the points made by this reviewer.

•Line 179: in vivo should be italicized Because journals differ in their stylistic practices, we are currently waiting before doing our final formatting. We did keep our use of Latin phrases consistently non-italicized in the draft.

•Lines 215-217: The comparison between Western blot expression levels and prior fluorescent reporter levels is unclear. Could be reformatted to make it clearer that relative expression of the different NEKL-MLTs in this study is consistent with prior data. We reformatted this sentence to improve clarity. (lines 205-207)

*•Lines 267-268: The final line of the passage is unclear and can be removed. * This sentence has been removed.

•Lines 311-313: This study is able to use the recovery of bait and known interactor proteins as internal controls to determine the quality of each experiment, but this may not always be the case for other users' experiments. The authors should comment on how Upreg %, a value influenced by many factors, can actually be used as a quality check when a bait protein has no known interactors. We have added language to highlight this point. (lines 344-348)

*•Line 702: There is a [new REF] that should be removed * As described above, we have now included this finding on bcc-1 as part of this manuscript (Fig 9C).

•The approach used mixed stage animals, but some genes oscillate or are transiently expressed. Please discuss cost-benefit of mixed stage vs syncing. This is an important point. We have added a discussion on the benefits and drawbacks of using mixed stages to the discussion. (lines 901-911)

*•Authors were working on hypodermally expressed proteins. It would be valuable to discuss what tissues are amenable to TurboID. Ie are the cases where there are few cells (anchor cell, glial sockets, etc) that it will be extremely challenging to perform this technique * We agree that certain tissues/proteins will not be amenable to proximity labeling. We believe that we have addressed this point together with the above comment throughout the manuscript and now on lines 936-940.

•Authors mention approaches such as nanobodies, split Turbo. Based on their experiences it would be valuable to add Discussion on strengths and weaknesses of these approaches to guide folks considering TurboID and DIA-MS experiments in C. elegans Because we have not tested these methods, we feel that we cannot provide a great deal of insight into these alternate approaches. We mention and reference these methods in the introduction so that readers are aware of them.

*Reviewer #2 (Significance (Required)): *

•Advance in technique: This study expands the use cases of data-independent acquisition MS method (DIA-MS) in C. elegans, which fragments all ions independent of the initial MS1 data. The benefits of this approach include better reproducibility across technical replicates and better recovery of low abundance peptides, which are critical for advancing our ability to capture weak and transient interactions.

•The use of DIA-MS in this study has improved our understanding of the partners of these NEKL-MLTs in membrane trafficking, molting, and cell adhesion within the epidermis.

•In many papers, TurboID seems very trivial but this paper clearly highlights the limitations and will be an invaluable resource for labs that want to get proximity labeling established in their labs.

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

*Summary: *

Fay and colleagues perform a series of proximity labeling experiments in C. elegans followed by thorough and rational analysis of the resulting biotinylated proteins identified by LC-MS/MS. The overall goals of the study are to evaluate different techniques and provide practical guidance on how to achieve success. The major takeaways are that integration of data-independent acquisition (DIA) along with comparison of endogenously tagged TurboID alleles to soluble TurboID expressed in the same tissue results in improved detection of bona-fide interactors and reduced numbers of false-positives.

*Major comments: *

Overall the claims are solid and conclusions supported. The data and methods are substantial to enable reproducibility in other labs. The experiments have been repeated multiple times with particular attention to statistical analysis. I have no major concerns with the manuscript and focus primarily on improving the accessibility of this important contribution to the scientific community. As such, I suggest that the authors:

1) Provide more explanation of and rationale for using DIA. This is not yet a standard technique and most basic biomedical scientists will be unaware of the jargon. As I expect many labs in the C. elegans community and beyond will be interested in the guidance provided in this manuscript, the introduction offers a great opportunity to bring the reader up to speed, as opposed to sending them to the complicated proteomics analysis literature. We have added some additional context (lines 77-80) as well as new references. We note that getting into the technical differences between DIA and DDA, beyond what we briefly mention, would take a substantial amount of space, may not be of interest to many readers, and can be found through standard internet and (sigh) AI-based searches.

*2) Provide a better overview of the various protocols tested (Experiments 1-8). Maybe at the beginning of the results, and maybe with an accompanying schematic. As currently written, it is difficult to figure out details regarding how the experiments vary and why. * We have now added a short paragraph to better inform the reader at the front end regarding the major experiments. (lines 139-146).

3) As to be expected, expression of TurboID tags at endogenous levels via low abundance proteins in a complex multicellular system results in somewhat weak signals that flirt with the limit of detection. Perhaps by combining tagged alleles within the same complex (NEKL-3/MLT-3 or NEKL-2/MLT-2/MLT-4) the signals could be boosted? Tandem tags, either on one end or multiple ends of proteins might help as well. As the authors point out, a benefit of tagging the two NEKL-MLT complexes is that there are strong loss-of-function phenotypes (lethal molting defects) to help evaluate whether a tagging strategy results in a non-functional complex. THESE EXPERIMENTS ARE OPTIONAL and might simply be discussed at the authors discretion. These are interesting ideas that we have now incorporated into our discussion. (lines 936-940)

*Minor Comments: *

*1) Figure 3A is cropped on the right. * Thank you for catching this. Corrected.

*2) Better define [new REF] on line 702. * We have added new results (Fig 9C), obviating the need for this reference.

***Referee cross-comments** *

Overall, I am in agreement with, and supportive of, the other reviewers' comments.

*Reviewer #3 (Significance (Required)): *

*Significance: *

Proximity labeling is often proposed as a technique to determine interaction networks of proteins in vivo, but in practice it remains challenging for most labs to execute a successful experiment, especially within the context of multicellular model organisms. Fay and colleagues provide a much needed roadmap for how to best approach proximity labeling experiments in C. elegans that will likely apply to other model systems.

They establish a rigorous approach by choosing to endogenously tag components of two essential NEKL-MLT complexes required for C. elegans molting. These complexes are relatively low abundance as they are only expressed in a single cell type, the hyp7 epidermal syncytium. In addition, as inactivation of any member of the complexes results in molting defects, they have a powerful selection for functional tags. Thus, they have set a high bar for themselves in order to discern whether a given variation on the experimental approach results in improved detection of interactors and fewer false positives.

*Potential areas for improvement include lowering the expression level of the skin-specific soluble TurboID used to determine non-specific biotinylation events. This control results in much higher levels of biotinylation compared to the TurboID-tagged NEKL-MLT alleles and likely affects their analysis, which they openly admit. In addition, to reduce the high level of background biotinylation signals generated by endogenous carboxylases, they adopt a depletion strategy pioneered by other researchers but this does not offer major improvements in detection of specific signals. The source of these conflicting results remains to be determined. It is also curious that auxin-inducible degradation of components of the NEKL-MLT complexes did not robustly alter the resulting biotinylating capacity of other members. This approach should be evaluated in subsequent studies. Finally, as mentioned in Major Comment #3 (above), it would be interesting to see if combining TurboID tags within the same complex might improve signal-to-background ratios. *

This manuscript represents a methodological advance that will likely become an oft-cited reference for members of the C. elegans community and a springboard for other basic biomedical scientists wanting to adapt rigorous proximity labeling techniques to their system. I am a cell biologist that uses a variety of genetic, molecular and biochemical approaches, mostly centered around C. elegans. I have used LC/MS-MS in our studies but have relatively little expertise in evaluating all aspects of proteomic pipelines.

*Reviewer #4 (Evidence, reproducibility and clarity (Required)): *

*Fay et al. describe an extensive proximity labeling BioID study in C. elegans with TurboID and DIA-LCMS analysis. They chose the NEKL-2/3 kinases and their known interactors MLT-2/3/4 as TurboID-fused bait proteins (C- and partially N-terminal fusions encoded from CRISPR-mediated genome edited genes). With eight biological replicates (and three to four technical replicates each) and with the unmodified wildtype or mNeonGreen-TurboID expressing worms as controls, a comprehensive dataset was generated. Although starting from quite different abundances of the bait-fusions within the cell lysates all bait proteins and known complex-binding partners were convincingly enriched with capturing streptavidin beads after only one hour of incubation with the lysate. This confirms the general applicability of TurboID-BioID approach in C. elegans. The BioID method typically gives rise to large proteomics datasets (up to more than thousand proteins identified after biotin capture) with several tens to hundreds enriched proteins (against negative control strains) as potential proteins that localize proximal to the bait-TurboID protein. However, substantial variations of candidates between biological replicates are frequently observed in BioID experiments. The authors scrutinized their dataset towards indicative metrics, filters and cutoffs in order to separate high-confidence from low-confidence candidates. With the workflow applied the authors melt down the number of candidates to 15 proteins that were grouped in four functional groups reasonably associated to NEKL-MLT function. *

Successful BioID experiments depend on reliable enrichment quantification with mass spectrometry using control cell lines that require a carefully bait-tailored design. Those must adequately express TurboID controls matching the abundance of the bait-TurboID fusion protein and its biotinylation activity. After affinity capture, sample preparation and LCMS data acquisition there is no silver bullet towards the identification true bait neighbors. Fay et al. elaborately describe their considerations and workflow towards high-confidence candidates. The workflow considered (i) data analysis with Volcano plots to account for statistical reproducibility of biological replicates against negative controls, (ii) fraction of proteins only detected in the positive or negative controls thus evading the fold-enrichment quantification approach, (iii) evaluation of variations in carboxylase enrichment as a measure for variations in the general biotin capture quality between experiments, (iv) an assessment of technical reproducibility with scatter plots and Venn diagrams, (v) exclusion of potentially false positives, e.g. promiscuously biotinylated non-proximal proteins, through comparisons with control worms expressing a non-localized mNeonGreen-TurboID fusion protein, (vi) batch effects, (vii) the impact of endogenous biotinylated carboxylases through depletion, (viii) gene ontology analysis of enriched proteins, (ix) weighing data according to the quality of individual experiments according to the afore mentioned metrics, and finally (x) genetic interaction studies to functionally associate high-confidence candidates with the bait.

*Major comments: *

Fay et al. present a solid, clear and comprehensive BioID-based proteomics study that takes into account and discusses decisive aspects for the (re)production and analysis of high-quality TurboID-based mass spectrometry data. Claims and conclusions are generally well and sufficiently supported by the presented data and illustrated with figures (throughout the text as well as with plenty of supplementary data). However, although the authors claim to seek for substrates of the kinase complex they drew no further attention to the phosphorylation status of the captured proteins. Haven't the MS data been analyzed in this respect? Information regarding this issue would enhance the manuscript. Data generation and method description appear reproducible for readers. Also, the statistical analyses appear adequate. The authors should also consider to deposit their MS raw and analysis data in a public repository (e.g. PRIDE) for future reviewing processes and as reference data for readers and followers. Our raw MS data have been deposited by the Arkansas Proteomics Facility. I have followed up to ensure that they are publicly available.

*Minor comments: *

The authors should combine supplementary data files to reduce the number of single files readers have to deal with. We have combined these files as suggested.

The authors should avoid the term "upregulation" or "increased biotinylation" when capture enrichment is meant. We agree with reviewer's point. We now use the terms enriched versus reduced or up versus down, depending on the context, and clearly define these terms. These changes have been incorporated throughout the manuscript.

*Reviewer #4 (Significance (Required)): *

The manuscript presents a robust BioID proteomics screening for co-localizing proteins of NEKL-2/3 kinases and their known interactors MLT-2/3/4. The ongoing validation of their functional interactions and whether the protein candidates reflect phosphorylation substrates or else remains elusive and is announced for upcoming manuscripts. The knowledge gain in terms of molecular mechanisms with NEKL-2/3 MLT-2/3/4 involvement in C. elegans is therefore limited to a table of - promising - interacting candidates that have to be studied further. Information about the phosphorylation status of the captured proteins from the MS data are not given. However, knowing the protein candidates will be of interest for groups working with these complexes (or the identified potentially interacting proteins) either in C. elegans or any other organism. Also, in-depth proteomics screenings with novel approaches such as BioID have to be established for individual organisms. For C. elegans there is only one prior BioID publication (Holzer et al. 2022). Many of the aspects discussed here have also been addressed earlier for BioIDs in other organisms and are not principally new. However, the presented study can be of conceptual interest for labs delving into or entangled with the BioID method in C. elegans or other organisms. The study addresses especially proteomics groups working on protein-protein interactions using proximity labeling/MS approaches. Basic consideration and thoughts for the experimental design and MS data analysis are given in detail and can serve as another guideline for future studies.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #4

Evidence, reproducibility and clarity

Fay et al. describe an extensive proximity labeling BioID study in C. elegans with TurboID and DIA-LCMS analysis. They chose the NEKL-2/3 kinases and their known interactors MLT-2/3/4 as TurboID-fused bait proteins (C- and partially N-terminal fusions encoded from CRISPR-mediated genome edited genes). With eight biological replicates (and three to four technical replicates each) and with the unmodified wildtype or mNeonGreen-TurboID expressing worms as controls, a comprehensive dataset was generated. Although starting from quite different abundances of the bait-fusions within the cell lysates all bait proteins and known complex-binding partners were convincingly enriched with capturing streptavidin beads after only one hour of incubation with the lysate. This confirms the general applicability of TurboID-BioID approach in C. elegans. The BioID method typically gives rise to large proteomics datasets (up to more than thousand proteins identified after biotin capture) with several tens to hundreds enriched proteins (against negative control strains) as potential proteins that localize proximal to the bait-TurboID protein. However, substantial variations of candidates between biological replicates are frequently observed in BioID experiments. The authors scrutinized their dataset towards indicative metrics, filters and cutoffs in order to separate high-confidence from low-confidence candidates. With the workflow applied the authors melt down the number of candidates to 15 proteins that were grouped in four functional groups reasonably associated to NEKL-MLT function.

Successful BioID experiments depend on reliable enrichment quantification with mass spectrometry using control cell lines that require a carefully bait-tailored design. Those must adequately express TurboID controls matching the abundance of the bait-TurboID fusion protein and its biotinylation activity. After affinity capture, sample preparation and LCMS data acquisition there is no silver bullet towards the identification true bait neighbors. Fay et al. elaborately describe their considerations and workflow towards high-confidence candidates. The workflow considered (i) data analysis with Volcano plots to account for statistical reproducibility of biological replicates against negative controls, (ii) fraction of proteins only detected in the positive or negative controls thus evading the fold-enrichment quantification approach, (iii) evaluation of variations in carboxylase enrichment as a measure for variations in the general biotin capture quality between experiments, (iv) an assessment of technical reproducibility with scatter plots and Venn diagrams, (v) exclusion of potentially false positives, e.g. promiscuously biotinylated non-proximal proteins, through comparisons with control worms expressing a non-localized mNeonGreen-TurboID fusion protein, (vi) batch effects, (vii) the impact of endogenous biotinylated carboxylases through depletion, (viii) gene ontology analysis of enriched proteins, (ix) weighing data according to the quality of individual experiments according to the afore mentioned metrics, and finally (x) genetic interaction studies to functionally associate high-confidence candidates with the bait.

Major comments:

Fay et al. present a solid, clear and comprehensive BioID-based proteomics study that takes into account and discusses decisive aspects for the (re)production and analysis of high-quality TurboID-based mass spectrometry data. Claims and conclusions are generally well and sufficiently supported by the presented data and illustrated with figures (throughout the text as well as with plenty of supplementary data). However, although the authors claim to seek for substrates of the kinase complex they drew no further attention to the phosphorylation status of the captured proteins. Haven't the MS data been analyzed in this respect? Information regarding this issue would enhance the manuscript. Data generation and method description appear reproducible for readers. Also, the statistical analyses appear adequate. The authors should also consider to deposit their MS raw and analysis data in a public repository (e.g. PRIDE) for future reviewing processes and as reference data for readers and followers.

Minor comments:

The authors should combine supplementary data files to reduce the number of single files readers have to deal with. The authors should avoid the term "upregulation" or "increased biotinylation" when capture enrichment is meant.

Significance

The manuscript presents a robust BioID proteomics screening for co-localizing proteins of NEKL-2/3 kinases and their known interactors MLT-2/3/4. The ongoing validation of their functional interactions and whether the protein candidates reflect phosphorylation substrates or else remains elusive and is announced for upcoming manuscripts. The knowledge gain in terms of molecular mechanisms with NEKL-2/3 MLT-2/3/4 involvement in C. elegans is therefore limited to a table of - promising - interacting candidates that have to be studied further. Information about the phosphorylation status of the captured proteins from the MS data are not given. However, knowing the protein candidates will be of interest for groups working with these complexes (or the identified potentially interacting proteins) either in C. elegans or any other organism. Also, in-depth proteomics screenings with novel approaches such as BioID have to be established for individual organisms. For C. elegans there is only one prior BioID publication (Holzer et al. 2022). Many of the aspects discussed here have also been addressed earlier for BioIDs in other organisms and are not principally new. However, the presented study can be of conceptual interest for labs delving into or entangled with the BioID method in C. elegans or other organisms. The study addresses especially proteomics groups working on protein-protein interactions using proximity labeling/MS approaches. Basic consideration and thoughts for the experimental design and MS data analysis are given in detail and can serve as another guideline for future studies.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Summary:

Fay and colleagues perform a series of proximity labeling experiments in C. elegans followed by thorough and rational analysis of the resulting biotinylated proteins identified by LC-MS/MS. The overall goals of the study are to evaluate different techniques and provide practical guidance on how to achieve success. The major takeaways are that integration of data-independent acquisition (DIA) along with comparison of endogenously tagged TurboID alleles to soluble TurboID expressed in the same tissue results in improved detection of bona-fide interactors and reduced numbers of false-positives.

Major comments:

Overall the claims are solid and conclusions supported. The data and methods are substantial to enable reproducibility in other labs. The experiments have been repeated multiple times with particular attention to statistical analysis. I have no major concerns with the manuscript and focus primarily on improving the accessibility of this important contribution to the scientific community. As such, I suggest that the authors:

1. Provide more explanation of and rationale for using DIA. This is not yet a standard technique and most basic biomedical scientists will be unaware of the jargon. As I expect many labs in the C. elegans community and beyond will be interested in the guidance provided in this manuscript, the introduction offers a great opportunity to bring the reader up to speed, as opposed to sending them to the complicated proteomics analysis literature.
2. Provide a better overview of the various protocols tested (Experiments 1-8). Maybe at the beginning of the results, and maybe with an accompanying schematic. As currently written, it is difficult to figure out details regarding how the experiments vary and why.
3. As to be expected, expression of TurboID tags at endogenous levels via low abundance proteins in a complex multicellular system results in somewhat weak signals that flirt with the limit of detection. Perhaps by combining tagged alleles within the same complex (NEKL-3/MLT-3 or NEKL-2/MLT-2/MLT-4) the signals could be boosted? Tandem tags, either on one end or multiple ends of proteins might help as well. As the authors point out, a benefit of tagging the two NEKL-MLT complexes is that there are strong loss-of-function phenotypes (lethal molting defects) to help evaluate whether a tagging strategy results in a non-functional complex. THESE EXPERIMENTS ARE OPTIONAL and might simply be discussed at the authors discretion.

Minor Comments:

1. Figure 3A is cropped on the right.
2. Better define [new REF] on line 702.

Referee cross-comments

Overall, I am in agreement with, and supportive of, the other reviewers' comments.

Significance

Proximity labeling is often proposed as a technique to determine interaction networks of proteins in vivo, but in practice it remains challenging for most labs to execute a successful experiment, especially within the context of multicellular model organisms. Fay and colleagues provide a much needed roadmap for how to best approach proximity labeling experiments in C. elegans that will likely apply to other model systems.

They establish a rigorous approach by choosing to endogenously tag components of two essential NEKL-MLT complexes required for C. elegans molting. These complexes are relatively low abundance as they are only expressed in a single cell type, the hyp7 epidermal syncytium. In addition, as inactivation of any member of the complexes results in molting defects, they have a powerful selection for functional tags. Thus, they have set a high bar for themselves in order to discern whether a given variation on the experimental approach results in improved detection of interactors and fewer false positives.

Potential areas for improvement include lowering the expression level of the skin-specific soluble TurboID used to determine non-specific biotinylation events. This control results in much higher levels of biotinylation compared to the TurboID-tagged NEKL-MLT alleles and likely affects their analysis, which they openly admit. In addition, to reduce the high level of background biotinylation signals generated by endogenous carboxylases, they adopt a depletion strategy pioneered by other researchers but this does not offer major improvements in detection of specific signals. The source of these conflicting results remains to be determined. It is also curious that auxin-inducible degradation of components of the NEKL-MLT complexes did not robustly alter the resulting biotinylating capacity of other members. This approach should be evaluated in subsequent studies. Finally, as mentioned in Major Comment #3 (above), it would be interesting to see if combining TurboID tags within the same complex might improve signal-to-background ratios.

This manuscript represents a methodological advance that will likely become an oft-cited reference for members of the C. elegans community and a springboard for other basic biomedical scientists wanting to adapt rigorous proximity labeling techniques to their system. I am a cell biologist that uses a variety of genetic, molecular and biochemical approaches, mostly centered around C. elegans. I have used LC/MS-MS in our studies but have relatively little expertise in evaluating all aspects of proteomic pipelines.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

This study expanded the use of data-independent acquisition-mass spectrometry (DIA-MS) in TurboID proximity-labeling proteomics to identify novel interactors of NEKL-2, NEKL-3, MLT-2, MLT-3, and MLT-4 complexes in C. elegans. The authors described several useful metrics to evaluate the quality of TurboID experiments, such as using the percentage of upregulated genes, the percentage of proteins present only in bait-TurboID experiments as compared to N2 controls, and the percentage of endogenously biotinylated carboxylases as internal controls. Further, the authors introduced methodological variability across 23 TurboID experiments and evaluated any improvement to the resulting data, such as N-terminally tagging bait proteins with TurboID, depleting endogenous carboxylases, and auxin-inducible degradation of known complex members. Finally, this study identified the kinase folding chaperone CDC-37 and the WASH complex component DDL-2 as novel interactors with the NEKL-MLT complexes through an RNAi-based enhancer approach following their identification by TurboID.

Major comments:

The key conclusions are convincing, and the work is rigorous. The work provides a clear roadmap to reproducing the data. The experiments are adequately replicated, and statistical analysis is adequate. We only have minor comments.

Minor comments:

• In the western blot in Fig 1 why does the mNG::Turbo have two bands?
• Fig 2B is difficult to parse as a reader. Columns labeled "Upreg," "Downreg," "TurboID only," "N2 only," "Filter-1," "Filter-2," and "Epi %" could be moved to Supplemental. Fold change vs N2 could be represented as a bar chart, allowing for trends between fold change and the metrics Upreg %, Turbo %, and Carboxylase % to be seen more clearly. Further, rows headed "Carboxylase depletion," "DDA," and "Auxin treated" could be presented as separate panels to better match the distinct points made in the text.
• Line 179: in vivo should be italicized
• Lines 215-217: The comparison between Western blot expression levels and prior fluorescent reporter levels is unclear. Could be reformatted to make it clearer that relative expression of the different NEKL-MLTs in this study is consistent with prior data.
• Lines 267-268: The final line of the passage is unclear and can be removed.
• Lines 311-313: This study is able to use the recovery of bait and known interactor proteins as internal controls to determine the quality of each experiment, but this may not always be the case for other users' experiments. The authors should comment on how Upreg %, a value influenced by many factors, can actually be used as a quality check when a bait protein has no known interactors.
• Line 702: There is a [new REF] that should be removed
• The approach used mixed stage animals, but some genes oscillate or are transiently expressed. Please discuss cost-benefit of mixed stage vs syncing.
• Authors were working on hypodermally expressed proteins. It would be valuable to discuss what tissues are amenable to TurboID. Ie are the cases where there are few cells (anchor cell, glial sockets, etc) that it will be extremely challenging to perform this technique
• Authors mention approaches such as nanobodies, split Turbo. Based on their experiences it would be valuable to add Discussion on strengths and weaknesses of these approaches to guide folks considering TurboID and DIA-MS experiments in C. elegans

Significance

• Advance in technique: This study expands the use cases of data-independent acquisition MS method (DIA-MS) in C. elegans, which fragments all ions independent of the initial MS1 data. The benefits of this approach include better reproducibility across technical replicates and better recovery of low abundance peptides, which are critical for advancing our ability to capture weak and transient interactions.
• The use of DIA-MS in this study has improved our understanding of the partners of these NEKL-MLTs in membrane trafficking, molting, and cell adhesion within the epidermis.
• In many papers, TurboID seems very trivial but this paper clearly highlights the limitations and will be an invaluable resource for labs that want to get proximity labeling established in their labs

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

Proximity labeling has become a powerful tool for defining protein interaction networks and has been utilized in a growing number of multicellular model systems. However, while such an approach can efficiently generate a list of potential interactors, knowledge of the most appropriate controls and standardized metrics to judge the quality of the data are lacking. The study by Fay systematically investigates these questions using the C. elegans NIMA kinase family members NEKL-2 and NEKL-2 and their known binding partners MLT-2, MLT-3 and MLT-4. The authors perform eight TurboID experiments each with multiple NEKL and MLT proteins and explore general metrics for assessing experimental outcomes as well as how each of the individual metrics correlates with one another. They also compare technical and biological replicates, explore strategies for identifying false positives and investigate a number of variations in the experimental approach, such as the use of N- versus C-terminal tags, depletion of endogenous biotinylated proteins, combining auxin-inducible degradation, and the use of gene ontology analysis to identify physiological interactors. Finally, the authors validate their findings by demonstrating that a number of the candidate identified functionally interact with NEKL-2 or components of the WASH complex.

Overall this is an outstanding study that will be of great interest to those interested in using proximity labeling to identify interactors of their favorite protein. The experiments are well executed and the data presented in a mostly clear manner. I really like this study (particularly because I plan to do a proximity labeling study of my own), but I did come away less than impressed with some of the analysis. This is a data-dense manuscript, and it appears to me that the authors tried to cover so much ground that in some cases very little insight was provided. For instance, the authors promote the use of data independent acquisition (DIA) as compared to the more commonly used data dependent acquisition (DDA). However the authors do not provide any analysis to indicate one approach is better than the other. Likewise the combined use of auxin-induced degradation and proximity labeling is explored but there is very little to take away from these experiments. Despite these issues, I am very enthusiastic about the study as a whole. Below I list major and minor concerns.

Major concerns

1. My biggest issue with the manuscript is that a lot is made of the use of data independent acquisition (DIA) as compared to the more commonly used data dependent acquisition (DDA). The authors perform experiments using DIA and DDA approaches but do not directly compare the outcomes. As a result there is really no way to know if one approach is better than the other. I would suggest the authors either perform the necessary analysis to compare the two approaches or tone down their promotion of DIA.
2. Line 75, The authors promote the use of data-independent acquisition (DIA) without defining what this approach is and how it differs from the more conventional data-dependent acquisition. As a non-mass spectroscopist, I found myself with lots of question concerning DIA, what it is and how it differs from DDA. I think it would really be helpful to expand the description of DIA and its comparison with DDA in the introduction.

Minor concerns:

Line 92 typo. I believe the authors meant to say NEKL-2-MLT-2-MLT-4.

Line169. Is exogenous the correct word to use here? It suggests that you are talking about non-worm proteins, but I know you are not.

Line 177 typo (D) should be (C).

Figure 1C: Lucky Charms may sue you for infringement of their trademarked marshmallow treats.

Figure 1D The NEKL-2::TurboID band is indicated with a green triangle in the figure but the figure legend states that green triangles indicate mNG::TurboID control. I know this triangle is a shade off the triangle that indicates mNG::TurboID but it's really hard to see the difference. All of the differently colored triangles in panel F are unnecessary. I would either just pick one color for all non-control bait proteins or better yet, only use a triangle to point to bands that are not obvious. For instance I don't need the triangles that point to NEKL-2 -3 and -4 fusion proteins. These are just distracting.

Line: 316: Conceivably, another factor that could contribute to the counterintuitive upregulation of some proteins in the N2 samples is related to the fusion proteins that are being expressed in the TurboID lines. A partially functional bait protein (one with a level of activity similar to nekl-2(fd81) that may not result in an obvious phenotype) could directly or indirectly affect gene expression leading to lower levels of a subset of proteins in the TurboID samples. The same could be said for fusion proteins with a gain-of-function effect.

Fig 3 B-E. I am a little confused how the data in these graphs is normalized. For instance, I would have expected that for NEKL-3 in panel B, that the normalized (log2) intensity value in N2 be set at 0 as it is for NEKL-2. Maybe I just don't have enough information on how these plots were generated.

Figure 6C legend is not correct.

Line 575: Figure reference should be Fig. S5G. The authors should check to make sure all references to supplemental figures include correct panel information.

Line 576. The authors reference a study by Artan and colleagues and report a weak correlation between their study and that of Artan. They reference figure S4 but it should be Fig S5H.

Line 652. The authors note that numerous proteins were present at substantially reduced levels in the mNG::TurboID samples and suggest that sticky proteins may have been outcompeted or otherwise excluded from beads incubated with the mNG::TurboID lysates. Why would sticky proteins only be a problem in these samples? The reasoning is not clear to me.

Line 745: The term "bait overlaps" is a bit vague. Ultimately, I figured out what it meant but it was not immediately obvious.

S7B Fig. Why is actin missing from the eluate?

Line 873: The authors state the extent of overlap in GO terms between the various experiments and provide percentages. I tried to extract this information from Figure 8C and came up with different values. For instance, in the case of Molecular Function, they state that they observed a 54% overlap between NEKL-2 and NEKL-3 but in the Venn diagram in Figure 8C I see that the NEKL-2 and NEKL-3 experiments had 71 (25+46) GO terms in common. Out of 98 GO terms for NEKL-2 or 104 for NEKL-3 the percentage I got is closer to 72. Am I analyzing this correctly?

Significance

Overall this is an outstanding study that will be of great interest to those interested in using proximity labeling to identify interactors of their favorite protein. The experiments are well executed and the data presented in a mostly clear manner. I really like this study (particularly because I plan to do a proximity labeling study of my own), but I did come away less than impressed with some of the analysis. This is a data-dense manuscript, and it appears to me that the authors tried to cover so much ground that in some cases very little insight was provided. For instance, the authors promote the use of data independent acquisition (DIA) as compared to the more commonly used data dependent acquisition (DDA). However the authors do not provide any analysis to indicate one approach is better than the other. Likewise the combined use of auxin-induced degradation and proximity labeling is explored but there is very little to take away from these experiments. Despite these issues, I am very enthusiastic about the study as a whole.