Author response:

We thank the three reviewers for their critical and in-depth assessment of our study. Below you find our comments to the public reviews and our revision plans.

Public Reviews:

Reviewer #1 (Public review):

This manuscript adds to the recent, exciting developments in our understanding of the MmpL/S transporters from mycobacteria. This work provides solid support for the trimeric/hexameric arrangement of subunits in the complex, and reveals a possible pathway for substrate translocation.Overall, I think this manuscript is a solid body of work that adds to several recent studies from this team and others on the structure and mechanism of the MmpL/S transporter family, particularly MmpL4/S4. The combination of AF, disulfide engineering, and experimental structure is good, though it is a bit puzzling that the experimental structure based on disulfide stabilization of the AF prediction does not recapitulate key elements (MmpS periplasmic domain docking to MmpL, and altered CCD configuration).

I have no major concerns about this manuscript.

We thank reviewer#1 for this positive assessment of our work. The deviation of the AF prediction from the experimental structure is , in our view, not puzzling. AF does not take the physical properties of proteins into account, but predicts structures based on strong sequence alignments. It therefore does not have “knowledge” about the general flexibility of domains such as the CCD, which is also observed in the corresponding MmpL5 structures, nor does it have knowledge about preferred conformational states. Rather than “failing” to confirm the AF predictions, our cryo-EM structure revealed an unexpected tilted conformation of the CCD. As we outline in comments below, the physiological relevance of the tilted CCD is unclear. Its flexibility might be required to interact with (still elusive) outer membrane protein components to form the fully assembled efflux machinery.

Reviewer #2 (Public review):

Summary:

The manuscript describes the structure of the Mycobacterium tuberculosis (MmpS4)3-(MmpL4)3 hetero-heximeric transporter complex. The structure was obtained by cryogenic electron microscopy using an engineered construct that cross-links MmpS4 to MmpL4 via a disulfide bond. The position of the disulfide bond was determined using an Alphafold2 model of the hetero-heximer. Although Alphafold2 predicts a symmetric hetero-heximer, the author found that the structure of the coiled-coil domain (CCD) is asymmetric, tilted at about 60° relative to the membrane domains, and only contains two of the three alpha helical hairpins, with the third being disordered.

Strengths:

The strategy of using Alphafold2 models to guide construct design for experimental structure determination is state-of-the-art, and this work provides a great example of its applications and limitations. I.e., the experimental structure does not fully recapitulate the prediction but provides unexpected results.

The comparisons between the authors' structures and the previously published structures of the MmpL4 monomer and MmpL5 trimers strengthen the authors' findings.

We thank reviewer#2 for this positive assessment of our work and agree that it is interesting that the experimental structures do not fully agree with the AF predictions (see also comment to reviewer#1).

Weaknesses:

A more detailed description of the current mechanistic hypothesis would strengthen the manuscript. The authors state that the two periplasmic domains "are expected to undergo rigid body movements that allow substrate transport through these periplasmic domains similar to the conformational changes observed in the E. coli multidrug efflux pump AcrB". A schematic of the proposed transport cycle, as a supplemental figure that shows the current hypothesis regarding transport, would be beneficial for understanding the previous structures and putting the current structure in context. Outside of "the mechanistic basis of how these conformational changes are coupled to protonation of the DY-pairs", what are the major controversies/open questions regarding the mechanism?

We thank the reviewer for this valuable comment. We will add a new figure with the model of the MmpL4 transport cycle based on our new data and discuss the proposed molecular transport mechanism in more detail in the main.

The authors provide evidence that the cysteine-depleted S4L4 construct is functional, but do not show that the construct with the introduced disulfide bond #5 (D39C MmpS4 and S434C MmpL4) is also functional. Demonstrating this would allow the authors to better interpret their resulting structures.

In the revised version, we will include additional data to assess the functional consequences of cross-linking.

The analysis presented in Figure 5 and Supplementary Figure 7 seems to suggest that the authors are proposing that the CCD central cavity acts as a transport pathway for the transported substrate, but I am not sure that this hypothesis is explicitly stated. This makes the reasoning behind the analysis presented unclear. Clarity could be improved by stating that the hypothesis of direct transport of substrate through the CCD central channel is being examined using the structure prediction, and what the implications are for the structure solved with the incompletely formed CCD.

We state clearly in the discussion that the channel through the CCD seems too narrow to let large molecules like mycobactin and bedaquiline pass:[AG1]

Line 318ff: “ The channel radius of the MmpL4 CCD is very narrow with a minimum of 1.1 Å according to the AlphaFold3 predition (Fig. 5). This is much smaller than the smallest axis of a molecular model of mycobactin molecule of ?? nm as determined from a model of iron-free mycobactin. In addition, the cryo-EM structure of MSMEG_1382 revealed a constriction in the CCD channel [21]. Even though the methionine side chains lining the channel wall are considered to be flexible{Aledo, 2019 #69594}, large conformational changes of the α-helical hairpins relative to each other would be required to allow passage of molecules as large as mycobactin and bedaquiline. The AcrAB-TolC efflux machinery provides an example for such large conformational changes to enable transport of large molecules by iris-like opening and closing movements the outer membrane channel-tunnel TolC [33]. Similar helical twisting may widen the channel of the CCD. Alternatively, it is conceivable that the substrates of MmpL4/MmpL5 are transported along the CCD surface, potentially requiring further protein partners. It is interesting to note that siderophore secretion and drug efflux by MmpL4/MmpL5 systems involves at least two additional proteins, namely the periplasmic protein Rv0455, which was shown to be essential for mycobactin efflux [34] and an outer membrane channel, whose identity remains elusive. A complete molecular understanding of the transport mechanism through the MmpL4/MmpL5 systems hence requires the identification of the missing components and structural information about their interactions.”

The channel radius of the MmpL4 CCD is very narrow (minimum of 1.1 Å) according to the AlphaFold3 prediction (Fig. 5), and the cryo-EM [AG2] [MN3] structure of MSMEG_1382 revealed a further constriction in the CCD channel [21]. We therefore consider direct substrate transport through the CCD central channel to be physically implausible for molecules of the size of mycobactin and bedaquiline. Even accounting for the flexibility of the methionine side chains lining the channel wall, the large conformational changes of the α-helical hairpins relative to each other would be required to accommodate such large substrates. While iris-like opening movements have been described for TolC in the AcrAB-TolC system [33], those movements widen an already substantially larger channel, and even such dramatic conformational changes would be insufficient to open a channel as narrow as that of the MmpL4 CCD to a diameter permissive for substrate passage. We instead favor a model in which substrates are transported along the outer surface of the CCD, potentially with the assistance of additional protein partners. This is consistent with the observation that MmpL4/MmpL5-mediated siderophore secretion and drug efflux involves at least two further proteins: the periplasmic protein Rv0455, shown to be essential for mycobactin efflux [34], and an as-yet-unidentified outer membrane channel. In this context, the overall flexibility of the CCD - illustrated here by the tilted, incompletely formed conformation - may reflect the conformational dynamics required for interaction with these partner proteins, rather than being directly involved in forming a transport conduit. A complete mechanistic understanding will require identification of the missing components and structural characterization of the fully assembled efflux machinery.

We do not think that the incompletely formed CCD represents a conformation that is relevant for transport. But it is a demonstration of the overall flexibility of the CCD, which may be required to further open the channel in case the substrates are transported within the CCD tube. Further in-depth experiments will be needed to clarify this interesting question, which is beyond the scope of this paper.

Given that the results emphasize the flexibility of the CCD, the manuscript would be strengthened by 3D variability analysis either in cryoSPARC or using cryoDRGN (or both). This would allow the authors to better quantify the degree of motion in the CCD and how it may correlate to flexibility in other regions. Further 3D flex reconstruction in cryoSPARC may improve the map quality of the CCD.

This is a great suggestion. We will include a 3D variability analysisin the revised manuscript.

Reviewer #3 (Public review):

Summary:

This manuscript by Earp et al reports cryoEM structures of the hexameric (MmpS4)₃-(MmpL4) )₃ complex from Mycobacterium tuberculosis, which belongs to the RND family of transporters and is known to have a role in the export of siderophores and contribute to drug resistance. The experimental workflow showcased involves the design of disulfide pairs using distance constraints obtained from the AlphaFold predicted structure of the hexameric complex. One such disulfide pair was used to determine the ~3.0 Å structures. The structure reveals density for the previously unresolved coiled-coil domain (CCD), a tilted CCD arrangement, and a cavity within the periplasmic domain, which the authors assert is occupied by detergent. Comparison of this complex with the monomer structure of MmpL4 shows conformational variations interpreted to implicate different domains and conserved residues involved in proton coupling, which might be related to the transport mechanism. While the methodological aspects of the manuscript are solid, enthusiasm for the overall advance/significance is less so, with doubts about the relevance of the tilted CCD structure, considering disulfide trapping and an incomplete validation of the claim that the titled CCD represents a stable intermediate conformation. A clear, updated transport mechanism is largely missing from the manuscript.

We thank reviewer#3 for these useful comments, which we will address during the revision of the manuscript. In particular, we plan to include a scheme of an updated transport model.

Strengths:

Beautiful structures, AF prediction-experimental validation nexus that could be fine-tuned for different systems/difficult to target complexes.

Weaknesses:

Physiological relevance of the tilted CCD conformation. No clear mechanistic model for the transport. While the CCD may indeed be a stable intermediate, the fact that the rest of the trimeric arrangement is unaffected does not fully rule out disulfide trapping as a factor in promoting this. The findings would be strengthened if the same tilted conformation is seen using a different set of disulfides. The significance of the detergent molecule and the new cavity observed could also be better discussed in terms of an updated transport model.

We believe that there was a misunderstanding about our interpretation of the tilted CCD. As a matter of fact, it must be a stable intermediate, otherwise no density would have been observed for it in the cryo-EM maps. Despite being a stable intermediate, it is indeed unlikely that it represents a conformational state that is relevant/required for transport. Firstly, only the upright, complete CCD can bridge the periplasm. because . Secondly, the structure was determined in detergent and lacks additional protein binder partners, which might stabilize the upright conformation of the CCD . It is also conceivable, as the reviewer pointed out, that disulfide cross-linking may have caused the tilt. However, as we wrote in the manuscript, we do not think that cross-linking caused this striking asymmetry of the CCD, because the three MmpL4 and MmpS4 chains are basically symmetrical in the C1-processed data (see also Figure 2E):

Line 182 ff: “To assess whether there are asymmetries in other parts of the structure, we superimposed the individual protomers of the (MmpS4)3-(MmpL4)3 complex analyzed using C1 symmetry (Fig. 2E). Apart from the two resolved α-helical hairpins, the MmpL4 core domains and the resolved parts of MmpS4 differ by a RMSD of less than 0.6 Å and are therefore structurally identical considering the map resolution of around 3 Å. The fact that the core domains of MmpS4 and MmpL4 do not deviate between the protomers argues against the possibility that the cross-links established between them cause the (asymmetric) tilt of the CCD.”

Regarding the DDM binding site, we will indeed include an updated transport model. That said, we wish to be cautious, because we lack experimental proof that MmpL4 can in fact transport DDM.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

In "Drift in Individual Behavioral Phenotype as a Strategy for Unpredictable Worlds," Maloney et al. (2024) investigate changes in individual responses over time, referred to as behavioral drift within the lifespan of an animal. Drift, as defined in the paper, complements stable behavioral variation (animal individuality/personality within a lifetime) over shorter timeframes, which the authors associate with an underlying bet-hedging strategy. The third timeframe of behavioral variability that the authors discuss occurs within seasons (across several generations of some insects), termed "adaptive tracking." This division of "adaptive" behavioral variability over different timeframes is intuitively logical and adds valuable depth to the theoretical framework concerning the ecological role of individual behavioral differences in animals.

Strengths:

While the theoretical foundations of the study are strong, the connection between the experimental data (Figure 1) and the modeling work (Figure 2-4) is less convincing.

Weaknesses:

In the experimental data (Figure 1), the authors describe the changes in behavioral preferences over time. While generally plausible, I identify three significant issues with the experiments:

(1) All of the subsequent theoretical/simulation data is based on changing environments, yet all the experiments are conducted in unchanging environments. While this may suffice to demonstrate the phenomenon of behavioral instability (drift) over time, it does not properly link to the theory-driven work in changing environments. An experiment conducted in a changing environment and its effects on behavioral drift would improve the manuscript's internal consistency and clarify some points related to (3) below.

We have added further discussion of this to the discussion section.

(2) The temporal aspect of behavioral instability. While the analysis demonstrates behavioral instability, the temporal dynamics remain unclear. It would be helpful for the authors to clarify (based on graphs and text) whether the behavioral changes occur randomly over time or follow a pattern (e.g., initially more right turns, then more left turns). A proper temporal analysis and clearer explanations are currently missing from the manuscript.

We have added a figure (1F to better visualize the changes in handedness over days). We have also pointed out the connection between the power spectrum and the autoregressive model given by the Wiener-Khinchen theorem (which states that the autocorrelation function of a wide-sense stationary process has a spectral decomposition of its power spectrum).

(3) The temporal dimension leads directly into the third issue: distinguishing between drift and learning (e.g., line 56). In the neutral stimuli used in the experimental data, changes should either occur randomly (drift) or purposefully, as in a neutral environment, previous strategies do not yield a favorable outcome. For instance, the animal might initially employ strategy A, but if no improvement in the food situation occurs, it later adopts strategy B (learning). In changing environments, this distinction between drift and learning should be even more pronounced (e.g., if bananas are available, I prefer bananas; once they are gone, I either change my preference or face negative consequences). Alternatively, is my random choice of grapes the substrate for the learning process towards grapes in a changing environment? Further clarification is needed to resolve these potential conflicts.

We have discussed this further in the discussion.

Reviewer #2 (Public review):

Summary:

This is an inspired study that merges the concept of individuality with evolutionary processes to uncover a new strategy that diversifies individual behavior that is also potentially evolutionarily adaptive.

The authors use a time-resolved measurement of spontaneous, innate behavior, namely handedness or turn bias in individual, isogenic flies, across several genetic backgrounds.

They find that an individual's behavior changes over time, or drifts. This has been observed before, but what is interesting here is that by looking at multiple genotypes, the authors find the amount of drift is consistent within genotype i.e., genetically regulated, and thus not entirely stochastic. This is not in line with what is known about innate, spontaneous behaviors. Normally, fluctuations in behavior would be ascribed to a response to environmental noise. However, here, the authors go on to find what is the pattern or rule that determines the rate of change of the behavior over time within individuals. Using modeling of behavior and environment in the context of evolutionarily important timeframes such as lifespan or reproductive age, they could show when drift is favored over bet-hedging and that there is an evolutionary purpose to behavioral drift. Namely, drift diversifies behaviors across individuals of the same genotype within the timescale of lifespan, so that the genotype's chance for expressing beneficial behavior is optimally matched with potential variation of environment experienced prior to reproduction. This ultimately increases the fitness of the genotype. Because they find that behavioral drift is genetically variable, they argue it can also evolve.

Strengths:

Unlike most studies of individuality, in this study, the authors consider the impact of individuality on evolution. This is enabled by the use of multiple natural genetic backgrounds and an appropriately large number of individuals to come to the conclusions presented in the study. I thought it was really creative to study how individual behavior evolves over multiple timescales. And indeed this approach yielded interesting and important insight into individuality. Unlike most studies so far, this one highlights that behavioral individuality is not a static property of an individual, but it dynamically changes. Also, placing these findings in the evolutionary context was beneficial. The conclusion that individual drift and bet-hedging are differently favored over different timescales is, I think, a significant and exciting finding.

Overall, I think this study highlights how little we know about the fundamental, general concepts behind individuality and why behavioral individuality is an important trait. They also show that with simple but elegant behavioral experiments and appropriate modeling, we could uncover fundamental rules underlying the emergence of individual behavior. These rules may not at all be apparent using classical approaches to studying individuality, using individual variation within a single genotype or within a single timeframe.

Weaknesses:

I am unconvinced by the claim that serotonin neuron circuits regulate behavioral drift, especially because of its bidirectional effect and lack of relative results for other neuromodulators. Without testing other neuromodulators, it will remain unclear if serotonin intervention increases behavioral noise within individuals, or if any other pharmacological or genetic intervention would do the same. Another issue is that the amount of drugs that the individuals ingested was not tracked. Variable amounts can result in variable changes in behavior that are more consistent with the interpretation of environmental plasticity, rather than behavioral drift. With the current evidence presented, individual behavior may change upon serotonin perturbation, but this does not necessarily mean that it changes or regulates drift.

However, I think for the scope of this study, finding out whether serotonin regulates drift or not is less important. I understand that today there is a strong push to find molecular and circuit mechanisms of any behavior, and other peers may have asked for such experiments, perhaps even simply out of habit. Fortunately, the main conclusions derived from behavioral data across multiple genetic backgrounds and the modeling are anyway novel, interesting, and in fact more fundamental than showing if it is serotonin that does it or not.

We have adjusted our wording and contextualized our claims based on previous literature.

To this point, one thing that was unclear from the methods section is whether genotypes that were tested were raised in replicate vials and how was replication accounted for in the analyses. This is a crucial point - the conclusion that genotypes have different amounts of behavioral drift cannot be drawn without showing that the difference in behavioral drift does not stem from differences in developmental environment.

We have reanalyzed the behavioral data in a hierarchical model to account for batch effects. Accounting for batch effects (Fig 1G, S1G) we still observe differences between genotypes and for pharmaceutical manipulations of serotonin, though our data provides more equivocal evidence for the effects of trhⁿ on drift.

Reviewer #3 (Public review):

Summary:

The paper begins by analyzing the drift in individual behavior over time. Specifically, it quantifies the circling direction of freely walking flies in an arena. The main takeaway from this dataset is that while flies exhibit an individual turning bias (when averaged over time), their preferences fluctuate over slow timescales.

To understand whether genetic or neuromodulatory mechanisms influence the drift in individual preference, the authors test different fly strains concluding that both genetic background and the neuromodulator serotonin contribute to the degree of drift.

Finally, the authors use theoretical approaches to identify the range of environmental conditions under which drift in individual bias supports population growth.

Strengths:

The model provides a clear prediction of the environmental fluctuations under which a drift in bias should be beneficial for population growth.

The approach attempts to identify genetic and neurophysiological mechanisms underlying drift in bias.

Weaknesses:

Different behavioral assays are used and are differently analysed, with little discussion on how these behaviors and analyses compare to each other.

We have added text indicating that these two behavioral responses have previously been shown to be correlated to each other and that the spectral power analysis and autoregressive model are conceptually linked.

Some of the model assumptions should be made more explicit to better understand which aspects of the behaviors are covered.

We have added a table in the supplemental clarifying all of the parameters of modeling for each figure.

Recommendations for the authors:

Reviewing Editor Comments:

Highlights of the Consultation Session of 3 Reviewers

In the consultation session, the reviewers discussed as particularly important the relative contribution of genotype and variable environment. Further analyses of the replicates of the genotypes were suggested to exclude the environment as the source of difference in the extent of drift between genotypes. If the difference in the extent of drift between replicates is greater than the difference in the extent of drift between genotypes, then one cannot really say that there is a genetic control over drift and that it would evolve (which is still an interesting result, but would be less exciting for a follow-up evolution experiment). If replicates differ, testing whether the relative difference in the extent of drift between genotypes is maintained across environments would also be strong evidence that the extent of behavioral drift is a property of a genotype and not a mere result of a fluctuating/variable environment. The authors do present two behavior paradigms that can serve the purpose of comparing the relative extent of drift between genotypes across the two paradigms that they already have. The authors might consider whether experimental data could be brought closer to theory by including an experiment in a variable environment (e.g temp or diet changes etc.).

Reviewers also agreed in the consultation session that methods and definitions were somewhat cryptic, and it would be very helpful if they were more detailed. For example, linking the free walking analysis to the Ymaze and then the model1 to the model2 was not straightforward.

We have added text to make more explicit the theoretical connection between the freewalking analysis, the ymaze analysis, and the model. We have added text and a supplemental table to clarify the methods.

Reviewer #1 (Recommendations for the authors):

(1) Line 161: The authors state in the supplement that they used DGRP strains, which are inbred and not isogenic. According to the original authors, they possess 99.3% genetic identity. The isoD1 strain has no known crossing scheme, so complete chromosome isogeneity remains questionable, especially after 12 or more years since its creation. The authors should refer to the strains as "near-isogenic" or a similar term.

We have adjusted the language as suggested to be more accurate.

(2) Lines 276, 338: The manuscript contains some unfinished sentences or remnants from the drafting process (e.g., "REFREF"). A thorough editorial review is recommended to eliminate such errors.

We have cleaned up all references and made additional passes to adjust text.

Reviewer #2 (Recommendations for the authors):

(1) If the authors want to claim that serotonin is a regulator of drift, they should provide a negative control experiment, using equivalent perturbations of another neuromodulator and non-modulator. Alternatively, they could simply soften the claims revolving around serotonin and its putative direct role in modulating drift.

We have softened the claims as suggested to avoid claiming our results show a specific role for serotonin.

(2) I would suggest always using "behavioral drift" when referring to drift, especially in the context of modeling, because it can be easily confused with genetic drift and cause confusion when reading.

We have adjusted the language throughout the manuscript per this suggestion.

(3) It would be good to see in the methods if the 2-hour assays were always done at the same time of the fly's subjective day and when (e.g. how many hours after lights on).

We have clarified this.

(4) I understand that many experiments use methodology replicated from the group's previous work, but I would recommend elaborating the experimental methods section in the supplementary such that the reader can understand and reproduce the methods without having to sift through and look for them in previous papers.

We have expanded on our discussion of the methodology in the methods section.

Reviewer #3 (Recommendations for the authors):

The paper begins by analyzing the drift in individual behavior over time. Specifically, it quantifies the circling direction of freely walking flies in an arena. The main takeaway from this dataset is that flies exhibit an individual turning bias (when averaged over time), yet their preferences fluctuate over slow timescales. However, it's unclear why the authors chose to switch to a different assay to compare strains. In particular, it's ambiguous whether the behavioral measure in one setup is comparable to that in the other; specifically, whether a bias in one setup reflects the same type of bias in the other. The behavior is also sampled differently across setups (though the details are unclear; see comments below) and analyzed using different methods. Consequently, it remains uncertain whether the slow fluctuations observed in the arena setup are also present in the Y maze. It appears that the analysis of the Y maze data only addresses individual behavioral variance or, at most, day-to-day changes, without accounting for longer-term correlations in bias-which I understood to be the primary interest in the arena setup. Some clarification is needed here (see specific comments below).

In Figure 2, the authors attempt to show the potential advantage of individual drift for survival in unpredictable, fluctuating environments. They demonstrate that while bet-hedging provides an advantage over timescales matching the generation time (since reproduction is required), it offers less benefit on shorter timescales, where an increased individual drift could be advantageous. This approach is well-conceived, and the findings are convincing, though the model would benefit from further clarification and additional explanation in the text.

Here are some more specific comments:

PART 1:

(1) L 223 one probably cannot see a circadian peak at 24h if the data were filtered at 24h, did they look with another low pass cutoff?

We clarified in the text that the power spectrum analysis was performed on unfiltered data.

(2) L 243 the spread in standard deviation is said to be consistent with drifting bias, however, I do not agree with this. The variation could be stochastic but independent across days, and show no temporal correlation. As done with the circular arena, a drift should be estimated as a temporal correlation in the behavior.

It is consistent insofar as seeing a non-zero standard deviation is a necessary condition for drift. While it does not show that there is any consistency over time, this can be inferred from the autoregressive model (as well as previous work). We have added text to make this clearer.

(3) In the autoregressive model this temporal aspect seems to be incorporated only to the first order (from day to day). Therefore, from what I understand, the drift term is not correlated over time. This seems very different from the spectral analysis done in the circular assay, and I wonder if it fits at all the initial definition of drift. For example, is the model compatible with a fixed mean and a similar power spectrum as in Figure 1C? The text should clarify that.

can be made clear in the case of σ = 0 and ϕ = 1, where values wouldϕ ≠ be0 In an AR(1) process, datapoints day to day are correlated as long as . This perfectly correlated with each other across time. The AR(1) model and the PSD of circling can be related via the Wiener-Khinchin theorem. We have added text to make this connection clear.

(4) Did serotonin have no role in turning bias? My understanding of previous work was that serotonin should affect the bet-hedg variance as well - the authors should discuss what is expected or not, especially given that the pharmacological and genetic approaches do not have the same effect on bet-edging (Figure 1H-I).

As the pharmacological methods were only applied after eclosion, we do not find it surprising that we do not measure differences in the initially measured distribution of handedness in that case. We do see more evidence of it in the mutations, though the trhⁿ experiments provide a less clear effect after our adjustments to account for batch effects.

(5) Methods: It is unclear how flies were handled across days; e.g. in Y mazes: 2h each day for how many days? In the arena flies were imaged either twice daily for 2h per session, or continuously for 24h (L138) - but which data are used where?

We will make this more clear, but all data in figure 1 was the continuous 24h data

This part of the methods is not well explained and I think it should be described in more detail.

(6) How many flies per genotype were tested in fig 1E?

Information was added to the caption to duplicate information in the table.

PART 2:

(7) In Figure 2B I do not understand the formulation N(50−ϕ: 50, σ), N(phi-et: et, σ) or in general N(x: m, s): does this mean that the variable x has normal distribution with mean m and variance s? Usually this would be written as N(x|m, s) or N(x; m, s)

If so then: N(50−ϕ: 50, σ) = N(ϕ: 0, σ) which has mean=0 while the figure caption says "from a normal distribution centred on the long term environmental mean" - what is the long term environmental mean?

If this is correct, and, therefore, we are just centering the mean, what about N(et-phi: et, σ)?

Et is the environment at the time, not the mean of the environment (which is 50). We have added more detail in supplementary methods to address this.

(8) Should ϕ vary between 1-100? And is the environmental parameter in Figure 2C also varying between 1-100? These ranges should be written somewhere.

While implied in the sigma notation, we have added more detail in supplementary methods to explain the situation.

(9) As far as I understand the bounding envelope in Figure 2B is necessary to contain the drift model. In Figure 1F, a bounding effect was generated by the "tendency to revert to no bias." It is unclear to me whether these two formulations are equivalent. Moreover, none of these two models might be able to recapitulate the correlations observed in the circular arena and analyzed spectrally in Figure 1C. It would be necessary that the author make an effort to relate these models/quantifications one to another. My understanding of Figure 1B is that there are slow fluctuations around the mean. Is the bounded drift model in 2B not returning to the same mean? And do these models generate slow fluctuations? Further explanation could help clarify these points.

We have added additional explanation to explain the connection between the power spectrum and the two methods of (phi and bounding envelop) of establishing stationarity.

(10) Expanding on the above: I thought that the definition of individuality is based on some degree of stability over days. However, both models assume drift to occur from day to day (and also the analysis of the DGRP lines assumes so). Some clarification here could help: is the initial bet-edging variation maintained in the population? And is the mean individual bias still a thing or it is just drifting away all the time?

The initial bet-hedging is maintained to some degree, based on the parameter of phi and the bounding envelope. We have added text to make this clearer.

(11) In both Figures 2C and 2E the populations are always shrinking, is that correct? And if so, is it expected? Does the model allow growth in a constant environment?

As the plotted values are the log, the optimal environments do allow growth (visible more clearly in 2D). We have added some text to make this clearer.

(12) Growth is quantified only across 100 days (Figure 2D) but at day 100 there is not something like a steady state, how is 100 chosen? Would it make sense to check longer times to see if the system eventually takes off? And if not, why?

(13) Related to the above: what is the growth range achieved in Figure 3A-B? Is the heatmap normalized to the same value across conditions? I think it would be important to consider the absolute range of variation of growth or at least the upper value across conditions.

Moreover: is growth quantified at day 100? What happens at longer times? Does the temporal profile of the growth curve differ across environmental conditions? (I'm referring to a Figure as 2D).

As we are plotting the log change, we are ultimately showing the growth rate. While a more realistic model would involve carrying capacity, we believe a simplified model showing growth or no growth captures the difference in growth rate between different strategies. We have added some text to make this clearer.

(14) Suddenly at line 502, sexual maturity is introduced as a parameter, which was never mentioned before, called a_min in the figure legend of panel 3a, but it is unclear where this is in the model. And please also clarify if sex maturity is the same as generation time.

Sexual maturity is the same as generation time, we have standardized terminology throughout the paper.

(15) Regarding lines 505-508, could one simply conclude that in this model formulation, the generation time has the effect of a low pass filter on environmental fluctuation? The question is: is this filtering effect the only effect of generation time?

While this seems to capture the high-frequency effect we see, it does not explain the shift from bet-hedging->drift we see at lower-frequency environmental fluctuations.

(16) What reproductive rate is used for the PCA analysis? Is the variance associated with the drift so low because of choosing a fast reproductive rate? A comment in the main text would be helpful.

We have clarified that these plots were done at 10 days.

Author response:

Thank you for the eLife assessment and the constructive reviews. We appreciate the reviewers’ valuable insights and the time they dedicated to providing such thoughtful feedback on our manuscript. The reviewers highlighted the technical rigor of our study, specifically the tracking of individual neurons across both anesthetized and awake states using two-photon imaging. They also emphasized the importance of our cell-type-specific analysis (excitatory, PV, and SOM neurons) and noted that the study provides solid evidence for isoflurane-induced shifts in preferred spatial frequency (SF).

Based on our team's evaluation of the reviewers' comments, we would like to outline our planned revisions.

(1) Expanded Population and Single-Neuron Analysis

We will re-analyze our dataset to include all neurons that were responsive under anesthesia, in the awake state, or both. This will ensure our findings accurately represent the entire population of visually responsive neurons. We will also provide examples of individual tuning curves to clarify the relationship between tuning shape and SF shifts in individual neurons.

(2) Addressing Methodological Scope and Behavioral Metrics

Receptive Field Size and Dynamics: While we did not utilize a stimulus set specifically designed to map receptive field (RF) sizes, we intend to examine how other functional parameters co-varied with the shift in preferred SF within each cell type. Furthermore, although characterizing the precise temporal dynamics during anesthesia onset presents technical challenges, we will attempt to analyze the time-dependence of the observed changes to provide deeper insight into the transition between states.

Behavioral Metrics: While pupil size is a well-established proxy for brain state, we will explore the inclusion of other available behavioral parameters.

(3) Cell-type Specificity (SOM, PV, and VIP)

SOM vs. PV Comparison: We will perform a detailed comparison of preferred SFs between SOM and PV interneurons, including those responsive only under anesthesia or only in the awake state.

VIP Neurons: While VIP neurons are known to play critical roles in cortical circuits, such as disinhibition, we have decided not to conduct new recordings for VIP interneurons in the present study. Based on existing literature, the proportion of visually responsive VIP cells is too low to yield statistically reliable conclusions for this specific study (de Vries et al., Nature Neuroscience 23, 138-151, 2020). Additionally, we intend to focus our analysis on inhibitory interneuron subtypes that provide direct input to pyramidal cells.

Histology: We will provide additional histological validation.

(4) Refined Framing

As suggested, we will focus the manuscript strictly on isoflurane anesthesia. This includes updating the title and abstract to reflect this specificity and discussing how our results compare with other anesthetics like urethane. Furthermore, we will substantially deepen our discussion on the potential mechanisms by which anesthesia induces a downward shift in preferred spatial frequency.

We believe these additions will significantly strengthen the manuscript.

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

In this work, Huang et al. revealed the complex regulatory functions and transcription network of 172 unknown transcriptional factors (TFs) in Pseudomonas aeruginosa PAO1. They have built a global TF-DNA binding landscape and elucidated binding preferences and functional roles of these TFs. More specifically, the authors established a hierarchical regulatory network and identified ternary regulatory motifs, and co-association modules. Since P. aeruginosa is a well known pathogen, the authors thus identified key TFs associated with virulence pathways (e.g., quorum sensing [QS], motility, biofilm formation), which could be potential drug targets for future development. The authors also explored the TF conservation and functional evolution through pan-genome and phylogenetic analyses. For the easy searching by other researchers, the authors developed a publicly accessible database (PATF_Net) integrating ChIP-seq and HT-SELEX data.

Strengths:

(1) The authors performed ChIP-seq analysis of 172 TFs (nearly half of the 373 predicted TFs in P. aeruginosa) and identified 81,009 significant binding peaks, representing one of the largest TF-DNA interaction studies in the field. Also, The integration of HT-SELEX, pan-genome, and phylogenetic analyses provided multi-dimensional insights into TF conservation and function.

(2) The authors provided informative analytical Framework for presenting the TFs, where a hierarchical network model based on the "hierarchy index (h)" classified TFs into top, middle, and bottom levels. They identified 13 ternary regulatory motifs and co-association clusters, which deepened our understanding of complex regulatory interactions.

(3) The PATF_Net database provides TF-target network visualization and data-sharing capabilities, offering practical utility for researchers especially for the P. aeruginosa field.

Thank you for your positive feedback!

Weaknesses:

(1) There is very limited experimental validation for this study. Although 24 virulence-related master regulators (e.g., PA0815 regulating motility, biofilm, and QS) were identified, functional validation (e.g., gene knockout or phenotypic assays) is lacking, leaving some conclusions reliant on bioinformatic predictions. Another approach for validation is checking the mutations of these TFs from clinical strains of P. aeruginosa, where chronically adapted isolates often gain mutations in virulence regulators.

Thank you for this valuable suggestion. We have performed the EMSA experiment to validate the binding result and also constructed the mutants for further functional validation. The details can be found in Figure S5.

(2) ChIP-seq in bacteria may suffer from low-abundance TF signals and off-target effects. The functional implications of non-promoter binding peaks (e.g., coding regions) were not discussed.

Thank you for this insightful comment regarding ChIP-seq data quality and non-promoter binding events. While we acknowledge that completely eliminating all non-specific binding signals is technically challenging in bacterial ChIP-seq experiments, we implemented stringent quality control measures including replicates, negative controls, and FDR cutoffs to minimize false positives.

Although the coding binding peaks represent a smaller fraction of total binding events, they are functionally significant rather than mere technical artifacts. Our previous work systematically demonstrated that bacterial TFs can bind to coding sequences and regulate gene expression through multiple mechanisms, including modulating cryptic promoter activity and antisense RNA transcription, hindering transcriptional elongation, and influencing translational efficiency[1]. We have now expanded the Discussion section to address these regulatory mechanisms.

(3) PATF_Net currently supports basic queries but lacks advanced tools (e.g., dynamic network modeling or cross-species comparisons). User experience and accessibility remain underevaluated. But this could be improved in the future.

Thank you for this constructive feedback on PATF_Net. We acknowledge that more advanced features would further enhance the platform’s utility. To enhance the utility of PA_TFNet, we have implemented two new features: (1) a virulence pathway browser that allows users to explore TF binding across curated gene sets for key virulence pathways (quorum sensing, secretion systems, biofilm, motility, etc.), and (2) a target gene search function that enables rapid identification of all TFs regulating any gene of interest by locus tag query.

Achievement of Aims and Support for Conclusions

(1) The authors successfully mapped global P. aeruginosa TF binding sites, constructed hierarchical networks and co-association modules, and identified virulence-related TFs, fulfilling the primary objectives. The database and pan-genome analysis provide foundational resources for future studies.

(2) The hierarchical model aligns with known virulence mechanisms (e.g., LasR and ExsA at the bottom level directly regulating virulence genes). Co-association findings (e.g., PA2417 and PA2718 co-regulating pqsH) resonate with prior studies, though experimental confirmation of synergy is needed.

Thank you for your positive feedback! We have added experimental validation in the Results section.

Impact on the Field and Utility of Data/Methods

(1) This study fills critical gaps in TF functional annotation in P. aeruginosa, offering new insights into pathogenicity mechanisms (e.g., antibiotic resistance, host adaptation). The hierarchical and co-association frameworks are transferable to other pathogens, advancing comparative studies of bacterial regulatory networks.

(2) PATF_Net enables rapid exploration of TF-target interactions, accelerating candidate regulator discovery.

Thank you for your positive feedback!

Reviewer #3 (Public review):

Summary:

The authors utilized ChIP-seq on strains containing tagged transcription factor (TF)-overexpression plasmids to identify binding sites for 172 transcription factors in P. aeruginosa. High-quality binding site data provides a rich resource for understanding regulation in this critical pathogen. These TFs were selected to fill gaps in prior studies measuring TF binding sites in P. aeruginosa. The authors further perform a structured analysis of the resulting transcriptional regulatory network, focusing on regulators of virulence and metabolism, in addition to performing a pangenomic analysis of the TFs. The resulting dataset has been made available through an online database. While the implemented approach to determining functional TF binding sites has limitations, the resulting dataset still has substantial value to P. aeruginosa research.

Strengths:

The generated TF binding site database fills an important gap in regulatory data in the key pathogen P. aeruginosa. Key analyses of this dataset presented include an analysis of TF interactions and regulators of virulence and metabolism, which should provide important context for future studies into these processes. The online database containing this data is well organized and easy to access. As a data resource, this work should be of significant value to the infectious disease community.

Thank you for your positive feedback!

Weaknesses:

Drawbacks of the study include 1) challenges interpreting binding site data obtained from TF overexpression due to unknown activity state of the TFs on the measured conditions, 2) limited practical value of the presented TRN topological analysis, and 3) lack of independent experimental validation of the proposed master regulators of virulence and metabolism.

We thank the reviewer for summarizing these key concerns. We acknowledge the limitations raised regarding TF overexpression, TRN topological analysis interpretation, and experimental validation. We provide detailed point-by-point responses to each of these concerns in our replies to the specific comments below, where we explain our rationale, the measures taken to address these limitations, and our plans for improvement.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Future Directions for the authors to consider for next steps:

(1) Key TFs (e.g., PA1380, PA5428) should be validated via gene knock out experiments, fluorescent reporter assays, or animal models to confirm roles in virulence pathways.

Thank you for this important suggestion. We agree that experimental validation is essential to confirm their regulatory roles and biological functions.

Firstly, we selected a subset of key TFs, including PA0167, PA1380, PA0815, and PA3094, and performed Electrophoretic Mobility Shift Assays (EMSA) experiments to validate their direct binding to target promoters. These results confirmed the ChIP-seq-identified interactions and are now included as Figure S5A-F.

We also constructed a clean deletion mutant of PA1380 and PA 3094 (ΔPA1380 and ΔPA3094) and their complementary strains (ΔPA1380/p and ΔPA3094/p). We then performed RT-qPCR analysis to validate their regulatory effects on key target genes. We found that PA1380 positively regulate the expression of cupB1 and cupB3 genes (Figure S5F). While the CupB cluster was known not be as important as CupA cluster in the biofilm information, so we did not find significant difference in biofilm formation between WT and ΔPA1380. Additionally, we found TF PA3094 also positively regulate lecA expression, which were shown in Figure S5G.

We agree that comprehensive functional validation, including animal model studies, would further strengthen the biological significance of these findings. Such experiments are currently underway in our laboratory and will be the subject of follow-up studies.

We have revised the Results section and Method section to include these validation experiments and their implications. Please see Figure S5 and Lines 283-300.

“To experimentally validate the regulatory interactions identified by ChIP-seq, we performed biochemical and genetic analyses on selected TFs. First, we conducted Electrophoretic Mobility Shift Assays (EMSA) for four TFs, including PA0167, PA0815, PA1380, and PA3094, using DNA fragments containing their predicted binding sites from target gene promoters. These TFs showed specific binding to their cognate DNA sequences (Figure S5A-D), confirming the direct binding of the ChIP-seq-identified interactions.

To further validate the functional regulatory roles of these TFs, we constructed clean deletion mutants of PA1380 and PA3094 (ΔPA1380 and ΔPA3094) along with their complemented strains (ΔPA1380/p and ΔPA3094/p). RT-qPCR analysis revealed that PA1380 positively regulates the expression of cupB1 and cupB3 (Figure S5E), two genes within the CupB fimbrial cluster identified as ChIP-seq targets. Similarly, PA3094 was confirmed to positively regulate lecA expression (Figure S5F), which encodes a lectin involved in biofilm formation and host interactions[2]. Expression of these target genes was restored to wild-type (WT) levels in the complemented strains, validating the regulatory relationships predicted by ChIP-seq. These combined biochemical and genetic validations demonstrate the accuracy and biological relevance of our TF binding data.”

(2) Non-promoter binding events (e.g., coding regions) may regulate RNA stability, warranting integration with translatomics or epigenomics data.

Thank you for this suggestion. We have now expanded the Discussion section to address this comment. Please see Lines 478-482.

“Our analysis revealed that TF binding events occur within coding regions, which is consistent with our previous study demonstrating that bacterial TFs possess binding capabilities for coding regions and can regulate transcription through multiple mechanisms [1]. Besides, it may also regulate RNA stability, warranting integration with translatomics or epigenomics data.”

(3) Incorporate strain-specific TF data (e.g., clinical isolates) and dynamic visualization tools to broaden PATF_Net's applicability.

Thank you for this constructive suggestion. To enhance the utility of PA_TFNet, we have implemented two new features: (1) a virulence pathway browser that allows users to explore TF binding across curated gene sets for key virulence pathways (quorum sensing, secretion systems, biofilm, motility, etc.), and (2) a target gene search function that enables rapid identification of all TFs regulating any gene of interest by locus tag query. These features are now live on the database and described in the revised manuscript.

Regarding strain-specific TF data, we agree this would be valuable for understanding regulatory diversity in clinical isolates. However, such an expansion would require ChIP-seq profiling across multiple strains. The current dataset is based on the reference strain PAO1, which serves as the foundation for most P. aeruginosa research and allows direct comparison with existing genomic and functional studies. We have added a statement in the revised manuscript acknowledging this limitation and highlighting strain-specific TF analysis as an important future direction for the field. Please see Lines 372-390.

“The database offers multiple search modalities to facilitate data exploration: users can perform TF-centric searches to query binding sites, target genes, and regulatory networks for individual TFs, or utilize the target gene search function to identify all TFs that regulate any gene of interest by entering its locus tag. To connect regulatory data with biological function, we have implemented a virulence pathway browser that allows users to explore TF binding patterns across curated gene sets for major P. aeruginosa virulence pathways. Interactive visualization tools, including network graphs and binding profile plots, facilitate intuitive exploration of regulatory relationships. The primary purpose of PATF_Net is to store, search, and mine valuable information on P. aeruginosa TFs for researchers investigating P. aeruginosa infection. The current resource is based on the reference strain PAO1, which serves as the foundation for most P. aeruginosa molecular studies and allows direct integration with existing genomic annotations and functional data. However, P. aeruginosa exhibits substantial genomic diversity across clinical isolates, and strain-specific differences in TF binding patterns may contribute to phenotypic variation in virulence, antibiotic resistance, and host adaptation. Extension of this resource to include strain-specific regulatory maps from diverse clinical isolates would provide valuable insights into the regulatory basis and represents an important direction for future investigation.”

(4) Phylogenetic analysis highlights TF conservation in bacteria; future work could explore functional homology in other Gram-negative pathogens (e.g., E. coli).

Thank for this insightful suggestion. Our phylogenetic analysis revealed that P. aeruginosa TFs exhibit varying degrees of conservation across bacterial species, with some showing broad distribution across Gram-negative pathogens while others are lineage-specific.

We agree that exploring functional homology of orthologous TFs across species would be highly valuable. Such comparative studies could address whether conserved TFs regulate similar target genes and biological processes across species, or whether regulatory networks have been rewired during evolution. For example, comparative ChIP-seq analysis of P. aeruginosa TFs and their orthologs in Klebsiella pneumoniae or even Gram-positive pathogen like Bacillus cereus could reveal conserved regulatory modules governing universal virulence or metabolic strategies versus species-specific adaptations. This represents an important direction for future investigation and would be facilitated by the comprehensive TF binding dataset we provide here. We have expanded the Discussion section to highlight this future direction. Please see Lines 539-550.

“While our phylogenetic analysis reveals varying degrees of TF conservation across bacterial species, the functional implications of this conservation remain to be fully explored. Many P. aeruginosa TFs have clear orthologs in both Gram-negative (e.g., Klebsiella pneumoniae) and Gram-positive pathogens (e.g., Bacillus cereus), yet whether these orthologs regulate similar target genes and biological processes is largely unknown. Future comparative ChIP-seq profiling of orthologous TFs could reveal the extent to which regulatory network architecture is conserved versus rewired during bacterial evolution, potentially identifying core regulatory modules governing universal bacterial strategies versus species-specific innovations. Such cross-species comparisons would enhance our understanding of regulatory network evolution and enable functional prediction in less well-characterized pathogens based on homology to experimentally validated P. aeruginosa regulators.”

Reviewer #3 (Recommendations for the authors):

Major comments

- Limitations of the ChIP-seq approach: With overexpression plasmids as an approach to TRN elucidation, there are always a set of concerns. First, TF expression is not enough to ensure regulatory activity - metabolite effects must be such that the TF is active which requires growing the cells in activating conditions. Second, the presence of a binding event does not mean that the binding has a regulatory effect - the authors are clearly aware of this as they specify binding sites in promoter regions, which should be helpful, but they also mention the possibility of regulatory binding events in coding regions. These issues should be listed as weaknesses of the approach in the Discussion.

Thank you for these important suggestions. We agree that these limitations should be explicitly discussed. We have now added a dedicated paragraph in the Discussion section addressing these concerns. Please see Lines 492-501.

“However, several limitations of the ChIP-seq approach should be acknowledged. Firstly, TF overexpression ensures sufficient protein levels for ChIP-seq signal detection but does not guarantee that all TFs are in their active conformational states, as many bacterial TFs require allosteric activation by metabolites, cofactors, or post-translational modifications. The cells under standard laboratory conditions which may not activate all TFs to their maximal regulatory states, potentially leading to underestimation of condition-specific binding peaks. Secondly, while we observed TF binding at thousands of genomic sites, binding per se does not equate to functional regulation, as chromatin context, cofactor availability, and competitive binding all influence regulatory outcomes.”

- Lack of independent validation: The study seems to lack substantial independent validation of either the functional nature of the binding sites as well as the proposed physiological regulatory role of the TFs. For example, for the 103 identified TF motifs, do any of these agree with existing motifs in motif databases that may be homologous to P. aeruginosa TFs? The authors claim to have discovered master regulators of virulence and associated core regulatory clusters - but there does not seem to be any independent validation of the proposed associations. The authors selected the TF targets to cover TFs that had not yet been characterized; however, it would have been nice to have some overlap with previous studies so that consistency and data quality could be assessed.

Thank you for raising these critical points about validation.

As for motif validation, we compared the existing motifs in the RegPrecise database[3] and we found that the motif of PA3587 show significant similarity to homologous TFs in Pseudomonadaceae. We have added the related description in the Results section. Please see Figure S3B and Lines 228-231.

As for the validation of master regulators, we have performed EMSA experiments for validating the binding events and constructed the mutants for function validation. We have added the related contents in Results section. Please see Figure S5 and Lines 283-300.

We have discussed the overlap between our results and previous studies in the Discussion section. Please see Lines 530-538.

“PA0797 is known to regulate the pqs system and pyocyanin production[4]. In the present study, it was also found to bind to the pqsH promoter region and its motif was visualised. PA5428 was found to bind to the promoter regions of aceA and glcB genes[5], which was also demonstrated in our ChIP-seq results. PA4381 (CloR) was found to be associated with polymyxin resistance in a previous study[6] and to be possibly related to ROS resistance in the present study. Furthermore, PA5032 plays a putative role in biofilm regulation and also forms an operon with PA5033, an HP associated with biofilm formation[7].”

- Uncertain value of TRN topology analysis: The relationship between ternary motifs and pathogenicity of P. aeruginosa, and why the authors argue these results motivated TF-targeting drugs (the topic of the last paragraph of the Discussion), are unclear to me. The authors allude to possible connections between pathogenicity, growth, and drug resistance, but I don't see concrete examples here of related TF interactions that clearly represent these relationships. The sections "Hierarchical networks of TFs based on pairwise interactions" and "Ternary regulatory motifs show flexible relationships among TFs in P. aeruginosa" seem to not say much in terms of results that are actionable or possible to validate. A topological graph is constructed based on observed TF-TF connections in measured binding sites - however, it's unclear if any of these connections are physiologically meaningful. Line 178 - Why would there be any connection between the structural family of TF and its location in the proposed TRN hierarchy?

Thank you for this valuable comment on TRN topology analysis. It is hard to quantify precisely how much this resource will accelerate P. aeruginosa research or drug development, but we believe providing this foundational network architecture has inherent value for the community, which is valued for enabling hypothesis generation even before comprehensive functional validation. We would like to clarify our perspective on these findings and have added the discussion in the revised manuscript to better describe their nature and value. Please see Lines 517-528.

“Additionally, although the TRN analysis revealed organizational patterns in P. aeruginosa regulatory network, the functional significance these topological features, including their specific contributions to pathogenicity, metabolic adaptation, and antibiotic resistance remains to be experimentally determined in the future work. The hierarchical structure and regulatory motifs we identified represent objective network properties derived from our binding data, but translating these structural observations into mechanistic understanding will require condition-specific functional studies, genetic validation, and phenotypic characterization. Our analysis provided a systematic framework and generating testable hypotheses rather than definitive functional conclusions. Nevertheless, these network-level organizational principles provided value to the community as a foundational reference, similar to other regulatory network maps[8] that were useful even before comprehensive validation.”

- Identification of "master" regulators: Line 527 on virulence regulators: "We first generated gene lists associated with nine pathways" - is this not somewhat circular, i.e. using gene lists generated from (I assume) co-regulated gene sets to identify regulators of those gene lists? I can't tell from the cited reference (80), which is their own prior review article, what the original source of these gene lists was. Somewhat related to this point - Line 32: 24 "master regulators" - if there are that many, is it still considered a master regulator? Line 270: This term "master regulator" would seem to require some quantitative justification. Identifying 24 (a large number of) "master" regulators of virulence would seem to dilute the implied power of the term.

We apologize for the lack of clarity regarding the virulence pathway gene lists, and we have provided complete gene lists for virulence-related pathways, which were compiled from functional annotations, in our online PA_TFNet database.

Additionally, we appreciate your concern about the use of “master” regulator. The usage is based on previous studies[9,10], and the master regulator is commonly known in the development of multicellular organisms as a subset of TFs that control the expression of multiple downstream genes and govern lineage commitment or key biological processes. We employed the term "master regulator" in an analogous manner to specify a class of functionally crucial TFs that participate in a pathway or biological event by regulating multiple downstream genes statistically enriched in that pathway. In line with this definition, we identified TFs whose targets were significantly enriched in genes associated with specific virulence pathways (hypergeometric test, P < 0.05).

We understand the concern that identifying 24 master regulators might seem to dilute the term. However, we would like to clarify that each of these 24 TFs is a "master regulator" with respect to specific virulence pathways based on statistical criteria, not necessarily a global master regulator of multiple pathways of P. aeruginosa. We have revised the Method section. Please see Lines 604-612.

- Line 234: "Genome-wide synergistic co-association of TFs in P. aeruginosa." This section was an interesting analysis. As I mention above, the weakness of an overexpression approach is not knowing whether the TF is active on the examined conditions. By looking at shared binding peaks across overexpression of different TFs, it should indeed be possible to glean some regulatory connections across TFs. Furthermore, the authors discuss specific examples that appear physiologically reasonable, which is appreciated.

We thank the reviewer for this positive assessment of our co-association analysis. We agree with the limitation of the overexpression approach, which have been discussed in the Discussion section. We are pleased that the reviewer found the approach and specific examples valuable.

Minor comments

- Line 35 - "high-throughput systematic evolution of ligands by exponential enrichment" - no idea what this means. Is this related to the web-based database, or why is it mentioned in the same sentence?

We apologize for the unclear presentation. To clarify: “High-throughput systematic evolution of ligands by exponential enrichment” (HT-SELEX) is an in vitro technique for determining TF DNA-binding motifs, which our group previously applied to a subset of P. aeruginosa TFs in a prior publication[11]. In the current study, we performed ChIP-seq for 172 TFs, which represent the majority of TFs not covered by the previous HT-SELEX study. Together, these two complementary approaches (HT-SELEX for in vitro binding motifs, ChIP-seq for in vivo genomic binding sites) provide near-complete coverage of the P. aeruginosa TF repertoire. Both datasets are integrated into our PA_TFNet database.

Due to space constraints in the abstract, we could not provide detailed explanation of HT-SELEX, but we have now improved the clarity in the Introduction to better explain the relationship between our previous HT-SELEX work and the current ChIP-seq study, and why both are mentioned together in the context of the database. Please see Lines 99-105.

- Line 193 - Only 9 auto-regulating TFs seems like a low number, given the frequency of negative auto-regulation in other organisms like E. coli. Could the authors comment on their expectations based on well-curated TRNs?

Thank you for this comment. We agree that 9 auto-regulating TFs is lower than might be expected based on E. coli, where auto-regulation is more prevalent. This likely reflects technical limitations of ChIP-seq approach that our detection was limited to standard growth conditions rather than the diverse physiological states where auto-regulation often occurs. Therefore, the 9 TFs we report represent a high-confidence subset, and the true frequency of auto-regulation in P. aeruginosa likely is higher. We added the content in the revised manuscript. Please see Lines 193-196.

“This number likely represents a conservative estimate, as experiments may not optimally capture auto-regulatory events that depend on native expression levels or specific physiological conditions.”

- Line 230 - "This conservation suggests that TFs within the same cluster co-regulate similar sets of genes." - Why would clustering of TF binding site motifs need to be done to make this assessment? Couldn't the shared set of regulated genes be identified directly from the binding site data? Computing TF binding site motifs has obvious value, but I am struggling to understand the point of clustering the motifs. Is there some implied evolutionary or physiological connection here? No specific physiological roles or hypotheses are discussed in this section.

Thank you for this important question. We agree that shared target genes can be identified directly from ChIP-seq binding data, which we also analyzed (co-association analysis). The motif clustering analysis serves a complementary and distinct purpose that provides information not directly obtainable from overlapped targets alone. Specifically, target overlap is inherently condition dependent, and motif clustering captures this intrinsic binding specificity, which reflects the structural similarity of DBDs, evolutionary relationships, and potential for functional redundancy or cooperativity under specific conditions. We have revised the related content in the manuscript, and please see Lines 236-242.

“Clustering of TF binding motifs identified groups of TFs with similar intrinsic DNA-binding specificities. As expected, many clusters contained TFs from the same DBD families, reflecting evolutionary conservation and potential functional redundancy or competitive binding at shared regulatory elements. Notably, the clustering also uncovered associations between TFs from different DBD families, suggesting convergent evolution of binding specificity or novel regulatory interactions that warrant further investigation.”

- Line 284 - should "metabolomic" be "metabolic"? I didn't see metabolomic data

Yes, we have revised. Please see Line 311.

- Several of the figures are too small (e.g. Fig S4A) or complex (Fig 2A) to see clearly or glean information from.

Thank you for this comment. We acknowledge that Figure 2A and Figure S4A contain dense information due to the comprehensive nature of the regulatory network and the large number of TFs analyzed. We believe these overview figures serve an important purpose in conveying the scale and organization of the regulatory network, while the tables (Table S6 for Fig. S4A and Table S3 for Fig. 2A) provide the granular data needed for specific inquiries. We have also made the figures available in higher resolution and increased font sizes where possible without compromising the overall layout.

- I don't understand the organization of the "Ternary regulatory motifs" in Supplementary Data File 4 - A table of contents explaining the tabs and columns would be welcome (for this as well as other supplementary files, some of which are more straightforward than others).

Thank you for this suggestion. We have now revised all supplementary data files to include header and necessary annotations in the first row. Specifically for Supplementary Data File 4, the three columns (Top, Middle, Bottom) represent the left, middle, and right node, respectively, in each ternary regulatory motif.

- I would have expected genomic locations of TF binding sites would have been one of the Supplementary Tables, to increase the accessibility of the data. However, the data is made available through their website, https://jiadhuang0417.shinyapps.io/PATF_Net/, which was easy to access and download the full dataset, so this is a minor issue.

Thank for accessing our PA_TFNet database and for the positive feedback on data accessibility. We agree that providing genomic locations of TF binding sites is crucial. These data are fully available and downloadable through the web interface, which allows flexible searching, filtering, and batch download of binding sites. We felt that the interactive and database format provides more functionality than static supplementary tables (e.g., dynamic filtering by TF, genomic region, or binding strength), given the large scale of this dataset.

References

(1) Hua, C., Huang, J., Wang, T., Sun, Y., Liu, J., Huang, L. et al. Bacterial Transcription Factors Bind to Coding Regions and Regulate Internal Cryptic Promoters. Mbio 13, e0164322 (2022).

(2) Chemani, C., Imberty, A., de Bentzmann, S., Pierre, M., Wimmerová, M., Guery, B. P. et al. Role of LecA and LecB lectins in Pseudomonas aeruginosa-induced lung injury and effect of carbohydrate ligands. Infect Immun 77, 2065-2075 (2009).

(3) Novichkov, P. S., Kazakov, A. E., Ravcheev, D. A., Leyn, S. A., Kovaleva, G. Y., Sutormin, R. A. et al. RegPrecise 3.0–a resource for genome-scale exploration of transcriptional regulation in bacteria. Bmc Genomics 14, 745 (2013).

(4) Cui, G. Y., Zhang, Y. X., Xu, X. J., Liu, Y. Y., Li, Z., Wu, M. et al. PmiR senses 2-methylisocitrate levels to regulate bacterial virulence in Pseudomonas aeruginosa. Sci Adv 8 (2022).

(5) Hwang, W., Yong, J. H., Min, K. B., Lee, K.-M., Pascoe, B., Sheppard, S. K. et al. Genome-wide association study of signature genetic alterations among pseudomonas aeruginosa cystic fibrosis isolates. Plos Pathog 17, e1009681 (2021).

(6) Gutu, A. D., Sgambati, N., Strasbourger, P., Brannon, M. K., Jacobs, M. A., Haugen, E. et al. Polymyxin resistance of Pseudomonas aeruginosa phoQ mutants is dependent on additional two-component regulatory systems. Antimicrob Agents Chemother 57, 2204-2215 (2013).

(7) Zhang, L., Fritsch, M., Hammond, L., Landreville, R., Slatculescu, C., Colavita, A. et al. Identification of genes involved in Pseudomonas aeruginosa biofilm-specific resistance to antibiotics. PLoS One 8, e61625 (2013).

(8) Galan-Vasquez, E., Luna, B. & Martinez-Antonio, A. The Regulatory Network of Pseudomonas aeruginosa. Microb Inform Exp 1, 3 (2011).

(9) Fan, L. G., Wang, T. T., Hua, C. F., Sun, W. J., Li, X. Y., Grunwald, L. et al. A compendium of DNA-binding specificities of transcription factors in Pseudomonas syringae. Nat Commun 11 (2020).

(10) Chan, S. S.-K. & Kyba, M. What is a master regulator? Journal of stem cell research & therapy 3, 114 (2013).

(11) Wang, T. T., Sun, W. J., Fan, L. G., Hua, C. F., Wu, N., Fan, S. R. et al. An atlas of the binding specificities of transcription factors in Pseudomonas aeruginosa directs prediction of novel regulators in virulence. Elife 10 (2021).

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #1 (Public review):

There are a few remaining issues:

(1) The manuscript quantifies changes over learning in prefrontal goal-selective cells (equated to "splitter" place cells in hippocampus) and task-phase selective cells (similar to non-splitter place cells that are not goal modulated). A subset of these task cells remain stable throughout learning, and are equated to schema representations in the study. In the memory literature, schemas are generally described as relational networks of abstract and generalized information, that enable adapting to novel context and inference by enabling retrieval of related information from previous contexts. The task-phase selective cells that stay stable throughout learning clearly will have a role in organizing task representations, but to this reviewer, denoting them as forming a schema is an unwarranted interpretation. By this definition, hippocampal non-splitter place cells that emerge early in learning and are stable over days would also form a schema. Therefore, schema notation cannot just be based on stability, it requires further evidence of abstraction such as cross-condition generalization.

We agree with the reviewer that task phase selective cells (“non-splitter cells”) alone do not fulfill the “relationality” criterion of schemas. We found only few of them, and so we cannot really say something about how they covary. We, however, would like to stress that our finding that task phase selective cells have stable firing field comparing learned (task) and habituation (no-task) conditions can be considered as “cross-condition generalization.” We have further specified our discussion of schemas with a particular emphasis on a potential interpretation of the generalizing task phase cells as “potential building blocks of schemas.”

(2) The quantification of prefrontal replay sequences during reward is useful, but it is still unconvincing that the distinction between existence of sequences in the odor sampling phase and reward phase is not trivially expected based on prior literature. This is odor guided task, not a spatial exploration task with no cues, and it is very well-established (as noted in citations in the previous review) that during odor sampling, animals' will sniff in an exploratory stage, resulting in strong beta and respiratory rhythms in prefrontal cortex. Not having LFP recordings in this task does not preclude considering prior literature that clearly shows that odor sampling results in a unique internal state network state, when animals are retrieving the odor-associated goal, vastly different from a reward sampling phase. The authors argue that this is not trivial since they see some sequences during sampling, although they also argue the opposite in response to a question from Reviewer 2 about shuffling controls for sequences, that 'not' seeing these sequences in the sampling phase is an internal control. The bigger issue here is equating these sequences during sampling to replay/ preplay or reactivation sequences similar to the reward phase, since the prefrontal network dynamics are engaged in odor-driven retrieval of associated goals during sampling, as has been shown in previous studies.

We agree with the reviewer that sampling and reward phase represent two very different behavioral states. Nevertheless, correlations on short time scales could be similar, which we show is not the case and therefore we do not consider this result trivial. Regarding the interpretation of sequences, we apologize that we have not been sufficiently clear on distinguishing replay with pure sequences. While we find such sequences in the sampling phase (indicative of fast temporal correlation structure beyond cofiring quantified in Figure 3) they are NOT pre/replaying any task related information. Otherwise, our results are fully in line with previous literature on oscillations that we have included in the previous round of revisions. We added a similar explanation at multiple instances in the Results and Discussion section.

Reviewer #2 (Public review):

Comments on revisions:

Further changes are needed to improve the description of the methods and the discussion needs to be extended to contrast the results with previously published results of the group. Some control figures would also be needed to quantitatively demonstrate, across the entire dataset, that sequence detection did not identify random events as sequences, even if the detection method was designed to exclude such sequences. For example, showing that sequences are not detected in randomised data with the current method would better convince readers of the method's validity.

We have added control quantifications from time randomized sequences which produce a much lower amount of detected sequences. See response below.

Although differences in the classification scheme relative to the Muysers et al. (2025) paper have been explained, the similarity (perhaps equivalence of results) is not sufficiently acknowledged - e.g., at the beginning of the discussion.

We have added a paragraph at the beginning of the Discussion on how our results align with the Muysers et al. 2025 paper.

Although the control of spurious sequences may have been built into the method, this is not sufficiently explained in the method. It is also not clear what kind of randomization was performed. Importantly, I do not see a quantification that shows that the detected sequences are significantly better than the sequence quality measure on randomized events. Or that randomized data do not lead to sequence clusters.

In response to this question, we have added the requested shuffling control (Supplement 1B to Figure 4). In the shuffled data the amount of detected recurring sequence clusters is only about half of those in the original data. The amount of bursts assigned to clusters in the shuffled data only remains 46% of the originally assigned bursts on average, clearly indicating that the detected sequences in the non-randomized data cannot be explained without assuming stable temporal order.

Some clusters, however, are still detected in randomized data, which, however, is expected if participation of cells is heterogeneous with some highly active cells occurring in more than half of the bursts. Then random sequences spuriously occur above chance level representing the clusters of random order of few highly active cells. In line with this interpretation, we see that

(1) Bursts that were removed after shuffling have exactly 0 high-firing cells

(2) Clusters derived from shuffled sequence have a less sparse contribution of high firing cells, i.e., high firing cells contribute to significantly more clusters in randomized data than in nonrandomized data.

The difference in the distribution of high firing cells further indicates that sequences obtained with and without randomization are of different quality.

The spurious (false positive) clusters detected after randomization nevertheless may have a physiological meaning as they identify rate coactivation patterns that were also picked up by analysis in Figure 3.

Also, it is still not clear how the number of clusters was established. I understand that the previously published paper may have covered these questions; these should be explained here as well.

The Methods sections states “The [cluster merging] procedure was repeated until no pair [of clusters] satisfied the merging criterion.”

Also, the sequence similarity description is still confusing in the method; please correct this sentence "Only the l neurons active in both sequences of a pair were taken into account."

We do not see what is wrong with this sentence. To avoid confusion.” we have replaced lower case l with upper case L as sequence length.

Reviewer #3 (Public review):

One comment is that the threshold for extracting burst events (0.5 standard deviations, presumably above the mean) seems lower than what one usually sees as a threshold for population burst detection, and the authors show (in Supplementary Fig 1) that this means bursts cover ~20-40% of the data. However, it is potentially a strength of this work that their results are found by using this more permissive threshold.

We have added further specifications following the Reviewer’s suggestion and now mention that the threshold is permissive and “capturing large amount cofiring structure.”

Author response:

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

Most importantly, in accordance with questions raised by Reviewer 1, we now include a detailed comparison of the cell type frequencies between the two examined time points as well as comparison of the pseudotimes along those lineages. This is detailed in the new section “Many cell types are shared between day 8 and day 16 EBs” and illustrated in Supplementary Figure 6c and Supplementary Figures 7-8.

Besides this new chapter and its accompanying methods part, we mainly edited the language and to clarify methods and assumptions according to the Reviewer suggestions.

The main concern of Reviewer 2 was our use of the liftoff gene annotation. We explained our reasoning for this choice extensively in our public response to the Reviewer, but did not incorporate this into our manuscript because even though this is an important subject it is not within the main scope of our paper.

Public Reviews:

Reviewer #1 (Public review):

Summary:

Jocher, Janssen, et al examine the robustness of comparative functional genomics studies in primates that make use of induced pluripotent stem cell-derived cells. Comparative studies in primates, especially amongst the great apes, are generally hindered by the very limited availability of samples, and iPSCs, which can be maintained in the laboratory indefinitely and defined into other cell types, have emerged as promising model systems because they allow the generation of data from tissues and cells that would otherwise be unobservable.

Undirected differentiation of iPSCs into many cell types at once, using a method known as embryoid body differentiation, requires researchers to manually assign all cell types in the dataset so they can be correctly analysed. Typically, this is done using marker genes associated with a specific cell type. These are defined a priori, and have historically tended to be characterised in mice and humans and then employed to annotate other species. Jocher, Janssen, et al ask if the marker genes and features used to define a given cell type in one species are suitable for use in a second species, and then quantify the degree of usefulness of these markers. They find that genes that are informative and cell type specific in a given species are less valuable for cell type identification in other species, and that this value, or transferability, drops off as the evolutionary distance between species increases.

This paper will help guide future comparative studies of gene expression in primates (and more broadly) as well as add to the growing literature on the broader challenges of selecting powerful and reliable marker genes for use in single-cell transcriptomics.

Strengths:

Marker gene selection and cell type annotation is a challenging problem in scRNA studies, and successful classification of cells often requires manual expert input. This can be hard to reproduce across studies, as, despite general agreement on the identity of many cell types, different methods for identifying marker genes will return different sets of genes. The rise of comparative functional genomics complicates this even further, as a robust marker gene in one species need not always be as useful in a different taxon. The finding that so many marker genes have poor transferability is striking, and by interrogating the assumption of transferability in a thorough and systematic fashion, this paper reminds us of the importance of systematically validating analytical choices. The focus on identifying how transferability varies across different types of marker genes (especially when comparing TFs to lncRNAs), and on exploring different methods to identify marker genes, also suggests additional criteria by which future researchers could select robust marker genes in their own data.

The paper is built on a substantial amount of clearly reported and thoroughly considered data, including EBs and cells from four different primate species - humans, orangutans, and two macaque species. The authors go to great lengths to ensure the EBs are as comparable as possible across species, and take similar care with their computational analyses, always erring on the side of drawing conservative conclusions that are robustly supported by their data over more tenuously supported ones that could be impacted by data processing artefacts such as differences in mappability, etc. For example, I like the approach of using liftoff to robustly identify genes in non-human species that can be mapped to and compared across species confidently, rather than relying on the likely incomplete annotation of the non-human primate genomes. The authors also provide an interactive data visualisation website that allows users to explore the dataset in depth, examine expression patterns of their own favourite marker genes and perform the same kinds of analyses on their own data if desired, facilitating consistency between comparative primate studies.

We thank the Reviewer for their kind assessment of our work.

Weaknesses and recommendations:

(1) Embryoid body generation is known to be highly variable from one replicate to the next for both technical and biological reasons, and the authors do their best to account for this, both by their testing of different ways of generating EBs, and by including multiple technical replicates/clones per species. However, there is still some variability that could be worth exploring in more depth. For example, the orangutan seems to have differentiated preferentially towards cardiac mesoderm whereas the other species seemed to prefer ectoderm fates, as shown in Figure 2C. Likewise, Supplementary Figure 2C suggests a significant unbalance in the contributions across replicates within a species, which is not surprising given the nature of EBs, while Supplementary Figure 6 suggests that despite including three different clones from a single rhesus macaque, most of the data came from a single clone. The manuscript would be strengthened by a more thorough exploration of the intra-species patterns of variability, especially for the taxa with multiple biological replicates, and how they impact the number of cell types detected across taxa, etc.

You are absolutely correct in pointing out that the large clonal variability in cell type composition is a challenge for our analysis. We also noted the odd behavior of the orangutan EBs, and their underrepresentation of ectoderm. There are many possible sources for these variable differentiation propensities: clone, sample origin (in this case urine) and individual. However, unfortunately for the orangutan, we have only one individual and one sample origin and thus cannot say whether this germ layer preference says something about the species or is due to our specific sample. Because of this high variability from multiple sources, getting enough cell types with an appreciable overlap between species was limiting to analyses. In order to be able to derive meaningful conclusions from intra-species analyses and the impact of different sources of variation on cell type propensity, we would need to sequence many more EBs with an experimental design that balances possible sources of variation. This would go beyond the scope of this study.

Instead, here we control for intra-species variation in our analyses as much as possible: For the analysis of cell type specificity and conservation the comparison is relative for the different specificity degrees (Figure 3C). For the analysis of marker gene conservation, we explicitly take intra-species variation into account (Figure 4D).

The same holds for the temporal aspect of the data, which is not really discussed in depth despite being a strength of the design. Instead, days 8 and 16 are analysed jointly, without much attention being paid to the possible differences between them.

Concerning the temporal aspect, indeed we knowingly omitted to include an explicit comparison of day 8 and day 16 EBs, because we felt that it was not directly relevant to our main message. Our pseudotime analysis showed that the differences of the two time points were indeed a matter of degree and not so much of quality. All major lineages were already present at day 8 and even though day 8 cells had on average earlier pseudotimes, there was a large overlap in the pseudotime distributions between the two sampling time points (Author response image 1). That is why we decided to analyse the data together.

Are EBs at day 16 more variable between species than at day 8? Is day 8 too soon to do these kinds of analyses?

When we started the experiment, we simply did not know what to expect. We were worried that cell types at day 8 might be too transient, but longer culture can also introduce biases. That is why we wanted to look at two time points, however as mentioned above the differences are in degree.

Concerning the cell type composition: yes, day 16 EBs are more heterogeneous than day 8 EBs. Firstly, older EBs have more distinguishable cell types and hence even if all EBs had identical composition, the sampling variance would be higher given that we sampled a similar number of cells from both time points. Secondly, in order to grow EBs for a longer time, we moved them from floating to attached culture on day 8 and it is unclear how much variance is added by this extra handling step.

Are markers for earlier developmental progenitors better/more transferable than those for more derived cell types?

We did not see any differences in the marker conservation between early and late cell types, but we have too little data to say whether this carries biological meaning.

Author response image 1.

Pseudotime analysis for a differentiation trajectory towards neurons. Single cells were first aggregated into metacells per species using SEACells (Persad et al. 2023). Pluripotent and ectoderm metacells were then integrated across all four species using Harmony and a combined pseudotime was inferred with Slingshot (Street et al. 2018), specifying iPSCs as the starting cluster. Here, lineage 3 is shown, illustrating a differentiation towards neurons. (A) PHATE embedding colored by pseudotime (Moon et al. 2019). (B) PHATE embedding colored by celltype. (C) Pseudotime distribution across the sampling timepoints (day 8 and day 16) in different species.

(2) Closely tied to the point above, by necessity the authors collapse their data into seven fairly coarse cell types and then examine the performance of canonical marker genes (as well as those discovered de novo) across the species. However some of the clusters they use are somewhat broad, and so it is worth asking whether the lack of specificity exhibited by some marker genes and driving their conclusions is driven by inter-species heterogeneity within a given cluster.

Author response image 2.

UMAP visualization for the Harmony-integrated dataset across all four species for the seven shared cell types, colored by cell type identity (A) and species (B).

Good point, if we understand correctly, the concern is that in our relatively broadly defined cell types, species are not well mixed and that this in turn is partly responsible for marker gene divergence. This problem is indeed difficult to address, because most approaches to evaluate this require integration across species which might lead to questionable results (see our Discussion).

Nevertheless, we attempted an integration across all four species. To this end, we subset the cells for the 7 cell types that we found in all four species and visualized cell types and species in the UMAPs above (Author response image 2).

We see that cardiac fibroblasts appear poorly integrated in the UMAP, but they still have very transferable marker genes across species. We quantified integration quality using the cell-specific mixing score (cms) (Lütge et al. 2021) and indeed found that the proportion of well integrated cells is lowest for cardiac fibroblasts (Author response image 3A). On the other end of the cms spectrum, neural crest cells appear to have the best integration across species, but their marker transferability between species is rather worse than for cardiac fibroblasts (Supplementary Figure 9). Cell-type wise calculated rank-biased overlap scores that we use for marker gene conservation show the same trends (Author response image 3B) as the F1 scores for marker gene transferability. Hence, given our current dataset we do not see any indication that the low marker gene conservation is a result of too broadly defined cell types.

Author response image 3.

(A) Evaluation of species mixing per cell type in the Harmony-integrated dataset, quantified by the fraction of cells with an adjusted cell-specific mixing score (cms) above 0.05. (B) Summary of rank-biased overlap (RBO) scores per cell type to assess concordance of marker gene rankings for all species pairs.

Reviewer #2 (Public review):

Summary:

The authors present an important study on identifying and comparing orthologous cell types across multiple species. This manuscript focuses on characterizing cell types in embryoid bodies (EBs) derived from induced pluripotent stem cells (iPSCs) of four primate species, humans, orangutans, cynomolgus macaques, and rhesus macaques, providing valuable insights into cross-species comparisons.

Strengths:

To achieve this, the authors developed a semi-automated computational pipeline that integrates classification and marker-based cluster annotation to identify orthologous cell types across primates. This study makes a significant contribution to the field by advancing cross-species cell type identification.

We thank the reviewer for their positive and thoughtful feedback.

Weaknesses:

However, several critical points need to be addressed.

(1) Use of Liftoff for GTF Annotation

The authors used Liftoff to generate GTF files for Pongo abelii, Macaca fascicularis, and Macaca mulatta by transferring the hg38 annotation to the corresponding primate genomes. However, it is unclear why they did not use species-specific GTF files, as all these genomes have existing annotations. Why did the authors choose not to follow this approach?

As Reviewer 1 also points out, also we have observed that the annotation of non-human primates often has truncated 3’UTRs. This is especially problematic for 3’ UMI transcriptome data as the ones in the 10x dataset that we present here. To illustrate this we compared the Liftoff annotation derived from Gencode v32, that we also used throughout our manuscript to the Ensembl gene annotation Macaca_fascicularis_6.0.111. We used transcriptomes from human and cynomolgus iPSC bulk RNAseq (Kliesmete et al. 2024) using the Prime-seq protocol (Janjic et al. 2022) which is very similar to 10x in that it also uses 3’ UMIs. On average using Liftoff produces higher counts than the Ensembl annotation (Author response image 4A). Moreover, when comparing across species, using Ensembl for the macaque leads to an asymmetry in differentially expressed genes, with apparently many more up-regulated genes in humans. In contrast, when we use the Liftoff annotation, we detect fewer DE-genes and a similar number of genes is up-regulated in macaques as in humans (Author response image 4B). We think that the many more DE-genes are artifacts due to mismatched annotation in human and cynomolgus macaques. We illustrate this for the case of the transcription factor SALL4 in Author response image 4C, D. The Ensembl annotation reports 2 transcripts, while Liftoff from Gencode v32 suggests 5 transcripts, one of which has a longer 3’UTR. This longer transcript is also supported by Nanopore data from macaque iPSCs. The truncation of the 3’UTR in this case leads to underestimation of the expression of SALL4 in macaques and hence SALL4 is detected as up-regulated in humans (DESeq2: LFC= 1.34, p-adj<2e-9). In contrast, when using the Liftoff annotation SALL4 does not appear to be DE between humans and macaques (LFC=0.33, p.adj=0.20).

Author response image 4.

(A) UMI-counts/ gene for the same cynomolgus macaque iPSC samples. On the x-axis the gtf file from Ensembl Macaca_fascicularis_6.0.111 was used to count and on the y-axis we used our filtered Liftoff annotation that transferred the human gene models from Gencode v32. (B) The # of DE-genes between human and cynomolgus iPSCs detected with DESeq2. In Liftoff, we counted human samples using Gencode v32 and compared it to the Liftoff annotation of the same human gene models to macFas6. In Ensembl, we use Gencode v32 for the human and Ensembl Macaca_fascicularis_6.0.111 for the Macaque. For both comparisons we subset the genes to only contain one-to-one orthologs as annotated in biomart. Up and down regulation is relative to human expression. C) Read counts for one example gene SALL4. Here we used in addition to the Liftoff and Ensembl annotation also transcripts derived from Nanopore cDNA sequencing of cynomolgus iPSCs. D) Gene models for SALL4 in the space of MacFas6 and a coverage for iPSC-Prime-seq bulk RNA-sequencing.

(2) Transcript Filtering and Potential Biases

The authors excluded transcripts with partial mapping (<50%), low sequence identity (<50%), or excessive length differences (>100 bp and >2× length ratio). Such filtering may introduce biases in read alignment. Did the authors evaluate the impact of these filtering choices on alignment rates?

We excluded those transcripts from analysis in both species, because they present a convolution of sequence-annotation differences and expression. The focus in our study is on regulatory evolution and we knowingly omit marker differences that are due to a marker being mutated away, we will make this clearer in the text of a revised version.

(3) Data Integration with Harmony

The methods section does not specify the parameters used for data integration with Harmony. Including these details would clarify how cross-species integration was performed.

We want to stress that none of our conservation and marker gene analyses relies on cross-species integration. We only used the Harmony integrated data for visualisation in Figure 1 and the rough germ-layer check up in Supplementary Figure S3. We will add a better description in the revised version.

Reference

Janjic, Aleksandar, Lucas E. Wange, Johannes W. Bagnoli, Johanna Geuder, Phong Nguyen, Daniel Richter, Beate Vieth, et al. 2022. “Prime-Seq, Efficient and Powerful Bulk RNA Sequencing.” Genome Biology 23 (1): 88.

Kliesmete, Zane, Peter Orchard, Victor Yan Kin Lee, Johanna Geuder, Simon M. Krauß, Mari Ohnuki, Jessica Jocher, Beate Vieth, Wolfgang Enard, and Ines Hellmann. 2024. “Evidence for Compensatory Evolution within Pleiotropic Regulatory Elements.” Genome Research 34 (10): 1528–39.

Lütge, Almut, Joanna Zyprych-Walczak, Urszula Brykczynska Kunzmann, Helena L. Crowell, Daniela Calini, Dheeraj Malhotra, Charlotte Soneson, and Mark D. Robinson. 2021. “CellMixS: Quantifying and Visualizing Batch Effects in Single-Cell RNA-Seq Data.” Life Science Alliance 4 (6): e202001004.

Moon, Kevin R., David van Dijk, Zheng Wang, Scott Gigante, Daniel B. Burkhardt, William S. Chen, Kristina Yim, et al. 2019. “Visualizing Structure and Transitions in High-Dimensional Biological Data.” Nature Biotechnology 37 (12): 1482–92.

Persad, Sitara, Zi-Ning Choo, Christine Dien, Noor Sohail, Ignas Masilionis, Ronan Chaligné, Tal Nawy, et al. 2023. “SEACells Infers Transcriptional and Epigenomic Cellular States from Single-Cell Genomics Data.” Nature Biotechnology 41 (12): 1746–57.

Street, Kelly, Davide Risso, Russell B. Fletcher, Diya Das, John Ngai, Nir Yosef, Elizabeth Purdom, and Sandrine Dudoit. 2018. “Slingshot: Cell Lineage and Pseudotime Inference for Single-Cell Transcriptomics.” BMC Genomics 19 (1): 477.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Figure 1B: the orangutan tubulin stain looks a bit unusual - just confirming that this is indeed the right image the authors want to include here.

We agree, this unfortunately also reflects the findings from the scRNA-seq analysis in that we found hardly any cells that we would classify as proper neurons.

(2) Typo on line 90: 'loosing' should be 'losing'.

Fixed

(3) Line 118: why do the authors believe that using singleR will give better results than MetaNeighbour? This certainly seems supported by the data in S4 and S5, but the reasoning is not clear.

We think that this might depend on the signal to noise ratio, which is a property specific to each dataset. Here we just wanted to state that our approach seems to work better for our developmental data, but we didn’t test out other data and thus cannot generalize.

(4) Figure 2B: there are some coloured lines on the first filled black bar from the left - do they mean anything? I couldn't work it out from looking at the figure.

Indeed this is a bit misleading the colors on the left represent the species identity: this was to illustrate the mixing of the of species for each cell type: The legend reads now: “Each line represents a cell which are colored by their species of origin on the left and by their current cell type assignment during the annotation procedure on the right.”

(5) Figure 3: I did not understand how the seven bins of the cell type specificity metric were derived until much later - it is just the number of cell types in which a gene is expressed, yes? Might be worth making this clearer earlier in the text.

We made this more explicit in the legend. “Boxplot of expression conservation of genes according to the number of different cell types in which a gene is expressed in humans (cell type specificity).”

(6) It would be great to provide a bit more thorough documentation for the shiny app, so it can serve as a stand-alone resource and not require going back and forth with the paper to make sure one knows what one is doing at every point.

Agree, this would be a good idea. We are on it.

(7) Line 477: I think this is unclear - the authors retain over 11000 cells per species but then set the maximum number of cells in a cluster for pairwise comparison to 250... which is a lot fewer. What happens to all the other cells? This probably needs some rewriting to clarify it.

We did this to minimize the power differences due to cell numbers and thus make the results more comparable across species. We added this explanation to the methods section for Marker gene detection.

Reviewer #2 (Recommendations for the authors):

How was the clustering resolution (0.1) determined?

This resolution was only used for the initial rough check up of the germ layers as reported in Figure 1 and Supplementary Figures S3. We chose this resolution because it yielded roughly the same number of clusters as the number of cell types that we got from classification with the Rhodes et al data.

Author response:

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

eLife Assessment

This study provides evidence that cerebellar projections to the thalamus are required for learning and execution of motor skills in the accelerating rotarod task. This important study adds to a growing body of literature on the interactions between the cerebellum, motor cortex, and basal ganglia during motor learning. The data presentation is generally sound, especially the main observations, with some limitations in describing the statistical methods and a lack of support for two separate cerebello-thalamic pathways, which is incomplete in supporting the overall claim.

We completed the MS by adding a double retrograde labelling study showing that the two pathways have limited overlap and by addressing the other concerns.

Public Reviews:

Reviewer #1 (Public review):

This is an interesting manuscript tackling the issue of whether subcircuits of the cerebellum are differentially involved in processes of motor performance, learning, or learning consolidation. The authors focus on cerebellar outputs to the ventrolateral thalamus (VL) and to the centrolateral thalamus (CL), since these thalamic nuclei project to the motor cortex and striatum respectively, and thus might be expected to participate in diverse components of motor control and learning. In mice challenged with an accelerating rotarod, the investigators reduce cerebellar output either broadly, or in projection-specific populations, with CNO targeting DREADD-expressing neurons. They first establish that there are not major control deficits with the treatment regime, finding no differences in basic locomotor behavior, grid test, and fixed-speed rotarod. This is interpreted to allow them to differentiate control from learning, and their inter-relationships. These manipulations are coupled with chronic electrophysiological recordings targeted to the cerebellar nuclei (CN) to control for the efficacy of the CNO manipulation. I found the manuscript intriguing, offering much food for thought, and am confident that it will influence further work on motor learning consolidation. The issue of motor consolidation supported by the cerebellum is timely and interesting, and the claims are novel. There are some limitations to the data presentation and claims, highlighted below, which, if amended, would improve the manuscript.

We thank the reviewer for the positive comments and insightful critics.

(1) Statistical analyses: There is too little information provided about how the Deming regressions, mean points, slopes, and intercepts were compared across conditions. This is important since in the heart of the study when the effects of inactivating CL- vs VL- projecting neurons are being compared to control performance, these statistical methods become paramount. Details of these comparisons and their assumptions should be added to the Methods section. As it stands I barely see information about these tests, and only in the figure legends. I would also like the authors to describe whether there is a criterion for significance in a given correlation to be then compared to another. If I have a weak correlation for a regression model that is non-significant, I would not want to 'compare' that regression to another one since it is already a weak model. The authors should comment on the inclusion criteria for using statistics on regression models.

We thank the reviewer for pointing out this weakness of description. The description of the Methods has thus been expanded and better justified in the “Quantification and statistical analysis” section.

We agree with the reviewer that comparison between Deming regressions would be fragile due to the weakness of these regression in treatment groups (while they are quite robust for control groups) and they are not included in the MS, although Deming regression coefficients with their confidence intervals are now provided for all groups in the statistical tables. As now more clearly explained in the Methods, the comparisons between groups are based on the distribution of residuals around regressions of the control regression lines. If we understand correctly the reviewer’s request, the control groups are all included.

(2) The introduction makes the claim that the cerebellar feedback to the forebrain and cortex are functionally segregated. I interpreted this to mean that the cerebellar output neurons are known to project to either VL or CL exclusively (i.e. they do not collateralize). I was unaware of this knowledge and could find no support for the claim in the references provided (Proville 2014; Hintzer 2018; Bosan 2013). Either I am confused as to the authors' meaning or the claim is inaccurate. This point is broader however than some confusion about citation.

The references are not cited in the context of collaterals from the DCN but for the output channels of the basal ganglia and cerebellum: “They [basal ganglia and cerebellum] send projections back to the cortex via anatomically and functionally segregated channels, which are relayed by predominantly non-overlapping thalamic regions (Bostan, Dum et al. 2013, Proville, Spolidoro et al. 2014, Hintzen, Pelzer et al. 2018).” Indeed, the thalamic compartments targeted by the basal ganglia and cerebellum are distinct, and in the Proville 2014, we showed some functional segregation of the cerebello-cortical projections (whisker vs orofacial ascending projections). Hintzen et al. have indeed performed an extensive review indicating the limited overlap between cerebellar- and basal ganglia-recipient territories. The sentence has been corrected to clarify what the “They” referred to.

The study assumes that the CN-CL population and CN-VL population are distinct cells, but to my knowledge, this has not been established. It is difficult to make sense of the data if they are entirely the same populations, unless projection topography differs, but in any event, it is critical to clarify this point: are these different cell types from the nuclei? how has that been rigorously established?; is there overlap? No overlap? Etc. Results should be interpreted in light of the level of this knowledge of the anatomy in the mouse or rat.

There is indeed a paragraph devoted to the discussion of this point (last part of the section “A specific impact on learning of CL-projecting CN neurons.”). Briefly, we actually know from the literature that there is a degree of collateralization (CN neurons projecting to both VAL and CL, see refs cited above), but as the reviewer says, it does not seem logically possible that the exact same population would have different effects, which are very distinct during the first learning days. The only possible explanation is the CN-CL and CN-VAL infections recruit somewhat different populations of neurons. We have now added more experiments to support our finding using retrograde infections using two rAAV viruses expressing red and green fluorescent reporter. These experiments confirm the limited overlap of the two populations of interest obtained by retrograde infection. We feel thus confident that while some CN neurons may project to both structures, retrograde infection strategies thus appear to differentially infect CN populations.

(3) It is commendable that the authors perform electrophysiology to validate DREADD/CNO. So many investigators don't bother and I really appreciate these data. Would the authors please show the 'wash' in Figure 1a, so that we can see the recovery of the spiking hash after CNO is cleared from the system? This would provide confidence that the signal is not disappearing for reasons of electrode instability or tissue damage/ other.

The recordings were not extended to the wash period, but examination of the firing rate before CNO on successive days did not evidence major changes in the population firing rate (this is now shown in a new supplementary figure 6).

(4) I don't think that the "Learning" and "Maintenance" terminology is very helpful and in fact may sow confusion. I would recommend that the authors use a day range " Days 1-3 vs 4-7" or similar, to refer to these epochs. The terminology chosen begs for careful validation, definitions, etc, and seems like it is unlikely uniform across all animals, thus it seems more appropriate to just report it straight, defining the epochs by day. Such original terminology could still be used in the Discussion, with appropriate caveats.

Since reference to these time windows is repeatedly used in the text we have shifted to “Early” and “Late” phase terminology.

(5) Minor, but, on the top of page 14 in the Results, the text states, "Suggesting the presence of a 'critical period' in the consolidation of the task." I think this is a non-standard use of 'critical period' and should be removed. If kept, the authors must define what they mean specifically and provide sufficient additional analyses to support the idea. As it stands, the point will sow confusion.

This has been corrected to: “suggesting the cerebellar contribution to the consolidation of the task is critical early in the learning process and cannot be easily reinstated later”

Reviewer #2 (Public review):

Summary:

This study examines the contribution of cerebello-thalamic pathways to motor skill learning and consolidation in an accelerating rotarod task. The authors use chemogenetic silencing to manipulate the activity of cerebellar nuclei neurons projecting to two thalamic subregions that target the motor cortex and striatum. By silencing these pathways during different phases of task acquisition (during the task vs after the task), the authors report valuable findings of the involvement of these cerebellar pathways in learning and consolidation.

Strengths:

The experiments are well-executed. The authors perform multiple controls and careful analysis to solidly rule out any gross motor deficits caused by their cerebellar nuclei manipulation. The finding that cerebellar projections to the thalamus are required for learning and execution of the accelerating rotarod task adds to a growing body of literature on the interactions between the cerebellum, motor cortex, and basal ganglia during motor learning. The finding that silencing the cerebellar nuclei after a task impairs the consolidation of the learned skill is interesting.

We thank the reviewer for the positive comments and insightful critics below.

Weaknesses:

While the controls for a lack of gross motor deficit are solid, the data seem to show some motor execution deficit when cerebellar nuclei are silenced during task performance. This deficit could potentially impact learning when cerebellar nuclei are silenced during task acquisition.

One of our key controls are the tests of the treatment on fixed speed rotarod, which provides the closest conditions to the ones found in the accelerating rotarod (the main difference between the protocols being the slow steady acceleration of rod rotation in the accelerating version). Indeed, small but measurable deficits are found at the highest speed in the fixed speed rotarod in the CN-VAL group, while there was no measurable effect on the CN-CL group, which actually shows lower performances from the second day of learning; we believe this supports our claim that the CN-CL inhibition impacted more the learning process than the motor coordination. In contrast, the CN-VAL group only showed significantly lower performance on day 4 consistent with intact learning abilities. Yet, under CNO, CN-VAL mice could stay for more than a minute and half at 20rpm, while in average they fell from the accelerating rotarod as soon as the rotarod reached the speed of ~19rpm (130s). Overall, we focused our argument on the first days of learning where the differences between the groups are more pronounced. We clarified the discussion (section “A specific impact on learning of CL-projecting CN neurons.”)

Separately, I find the support for two separate cerebello-thalamic pathways incomplete. The data presented do not clearly show the two pathways are anatomically parallel. The difference in behavioral deficits caused by manipulating these pathways also appears subtle.

There is indeed a paragraph devoted to the discussion of this point (last part of the section “A specific impact on learning of CL-projecting CN neurons.”). Briefly, we actually know from the literature that there is a degree of collateralization (CN neurons projecting to both VAL and CL, see refs cited above), but it does not seem logically possible that the exact same population would have different effects, which are very distinct during the first learning days. The only possible explanation is the CN-CL and CN-VAL infections recruit somewhat different populations of neurons. We have now added more experiments to support our finding using retrograde infections using two rAAV viruses expressing red and green fluorescent reporter. These experiments confirm the limited overlap of the two populations of interest obtained by retrograde infection. We feel thus confident that while some CN neurons may project to both structures, retrograde infection strategies thus appear to differentially infect CN populations.

While we agree that after 3-4 days of learning the difference between the groups becomes elusive, we respectfully disagree with the reviewer that in the early stages these differences are negligible.

Reviewer #3 (Public review):

Summary:

Varani et al present important findings regarding the role of distinct cerebellothalamic connections in motor learning and performance. Their key findings are that:

(1) Cerebellothalamic connections are important for learning motor skills

(2) Cerebellar efferents specifically to the central lateral (CL) thalamus are important for shortterm learning

(3) Cerebellar efferents specifically to the ventral anterior lateral (VAL) complex are important for offline consolidation of learned skills, and

(4) That once a skill is acquired, cerebellothalamic connections become important for online task performance.

The authors went to great lengths to separate effects on motor performance from learning, for the most part successfully. While one could argue about some of the specifics, there is little doubt that the CN-CL and CN-VAL pathways play distinct roles in motor learning and performance. An important next step will be to dissect the downstream mechanisms by which these cerebellothalamic pathways mediate motor learning and adaptation.

Strengths:

(1) The dissociation between online learning through CN-CL and offline consolidation through CN-VAL is convincing.

(2) The ability to tease learning apart from performance using their titrated chemogenetic approach is impressive. In particular, their use of multiple motor assays to demonstrate preserved motor function and balance is an important control.

(3) The evidence supporting the main claims is convincing, with multiple replications of the findings and appropriate controls.

We thank the reviewer for the positive comments and insightful critics below.

Weaknesses:

(1) Despite the care the authors took to demonstrate that their chemogenetic approach does not impair online performance, there is a trend towards impaired rotarod performance at higher speeds in Supplementary Figure 4f, suggesting that there could be subtle changes in motor performance below the level of detection of their assays.

This is now better acknowledged in the discussion in the section “A specific impact on learning of CL-projecting CN neurons.” However, we want to underline that the strongest deficit in learning is found in animals with CN->CL inhibition which latency to fall saturates at about 100s on the rotarod; this indicates that mice fall as soon as the accelerating rotarod speed reaches about 16rpm. In fixed speed rotarod, the inhibition of CN->CL neurons shows not even a trend of difference at 15rpm with control mice, and the animals run 2 minutes without falling at this speed. This makes us confident that the CN->CL pathway interfers more with the learning than with the actual locomotor function on the rotarod.

(2) There is likely some overlap between CN neurons projecting to VAL and CL, somewhat limiting the specificity of their conclusions.

This issue is treated in the discussion. (see also replies to reviewers 1 and 2 above). We added experiments with simultaneous retro-AAV infections in CL and VAL and the data are presented in Supplementary Figure 5. We found that retrograde infection targeted different populations of CN neurons; although collaterals in both CL and VAL may be present for (some of) these two populations of neurons, they are likely strongly biased toward one or the other thalamic regions, explaining the differential retrograde labelling in the CN. We hope these experiments will answer the reviewer’ s concern.

Recommendations for the authors:

Reviewer #2 (Recommendations for the authors):

(1) Multiple studies have reported on the effect of cerebellar nuclei (CN) manipulation on locomotion. Here the authors perform several controls and careful analysis to rule out gross motor deficits caused by DREADD-mediated CN silencing. As the authors point out in the discussion, part of the difference from prior studies could be the mild degree of inhibition here. However, it is possible that the CN inhibition here induces a subtle motor deficit and the accelerating rotarod task is challenging and more readily reveals this motor deficit, rather than a deficit in motor learning per se. Two pieces of data seem to suggest this:

(a) under CN inhibition during the task (Figure 1i), mice could never achieve the level of performance as mice under CN inhibition after the task, even after several days of training, which suggests the CN inhibition is interfering with task performance;

(b) in highly trained mice (after learning), applying the CN inhibition impaired performance to a similar extend as mice in Figure 1i (Figure 4).

Can the authors rule out the possibility that CN inhibition during the task is impairing motor execution rather than motor learning?

We do not rule out a contribution of impaired motor coordination at the highest speed (last paragraph of the section “A specific impact on learning of CL-projecting CN neurons.”). Indeed, most of our argument in favor of deficit in learning is primarily in the first days (Early phase), particularly for the CN->CL CNO group (Fig 3h). A crucial control in our work is the use of fixed speed rotarod, where no deficit is observed. The difference between the fixed and accelerating rotarod is rather minimal since the acceleration of the rotarod is rather small (0.12rpm/s for speed up to >20 rpm).

Interpreting the effect of treatment reversal is challenging. If the only effect of CNO was a motor deficit, the animals who learned under CNO should rapidly regain higher performance under saline, which is not observed. When switching from CNO to Saline after 7 days of training, it is difficult to disentangle which part is due to a crude motor deficit (which would not show in fixed speed rotarod), and which part is due to an unability to resume motor learning after the task has been (mis-)consolidated.

(2) The separation of the cerebellar pathways to the intralaminar thalamus (IL) and ventral thalamus (VAL) is not clear to me. It is not clear the CN neurons projecting to these nuclei are distinct. In addition, although IL projects to the striatum and VAL does not, both IL and VAL project to motor cortex. It is unclear to what extent these pathways can be separated. The argument for distinct pathways (as laid out in the discussion) is the distinct behavior deficits when manipulating these two pathways, but this difference seems subtle (point 3).

We now clarify that CN populations are different help to retrograde labelling experiments (new Suppl Fig 5). A discussion on the differences in IL and VAL projections is now discussed in the last paragraph of the section “A specific impact on learning of CL-projecting CN neurons.” Briefly, we argue that the despite some overlap of their targets, the profiles of the CL and VAL differ substantially.

(3) The pattern of behavioral deficits induced by CN->CL and CN->VAL neurons appear similar in Figure 3b-c and e-f. I have difficulty seeing how these data lead to the differences in the regression fits in panels 3g-k, which seem to show distinct patterns of performance change within and across sessions. One notable difference in Figure 3b-c and e-f seems to be that CN->VAL CNO treated mice exhibit lower performance on the very first trial for most days. Somehow, this pattern is present even after the CNO treatment is switched to saline (Figure 3f). I wonder if this data point is driving the difference. One control analysis the authors could do is to exclude the 1st trial and test if the effects are preserved.

Since the learning is cumulative and involves varying degree of consolidation it is indeed difficult to substantiate the difference from the average performance: a performance on day 3 may be limited by slow learning and perfect consolidation or good learning and imperfect consolidation. That is why we designed an analysis which takes into account the observed relationships between initial performance, within session gain of performance and acrosssession carry-over of this gain of performance (Fig 2). This analysis focuses on the first days of learning, before the performance plateau is reached in the CNO groups. While a clear deficit in consolidation is observed with full CN inhibition, this is not the case for the CN→CL CNO groups, despite their weaker performance after 3 days, similar to that seen with full CN inhibition. In contrast, normal learning is observed in the CN→VAL CNO group during these three days. The consolidation deficit in the CN→VAL CNO group is more subtle than in the CN CNO group and is indeed largely driven by the first data point. This is consistent with the idea that CN→VAL inhibition only partially impairs consolidation (compared to full CN inhibition), leaving some “savings” that allow rapid reacquisition.

(4) The quantification of locomotion in Figure S2 needs more information. What is linear movement? What is sigma? What is the alternation coefficient? These are not defined in the legends or the Methods as far as I can tell. Related to point 1 above, the authors should provide some analysis of the stride length and hindlimb to forelimb distance as measures of locomotion execution.

These measures were taken from Simon J Neurosci 2004 24(8):1987-1995 which is now cited and their description is now provided in the Methods.

Minor:

(5) To help readers follow the logic of experimental design, please explain why CNO was switched to saline after day 4 in Figures 1j, 3c, and f. Specifically, is the saline manipulation meant to test something as opposed to applying CNO throughout the entire course of the behavioral test?

Since we had no difference between the groups at the end of the Early phase, we decided to test whether the skill consolidated under CNO remained available when the CNO was removed (and it indeed was). This is now more clearly stated in the Results.

(6) I have difficulty understanding what is plotted in Figure 4b and d. The legend says the change in performance is calculated the same way as in Figure 2a, so the changes are presumably the regression slopes. But how are the regression slopes calculated for daily start (1st trial) and daily end (last trial)?

Skill level at the beginning and end of each trial correspond to the values of the regression line for abscissae values of trial 1 and trial 7 (green points). This has been added to the figure legend.

(7) Do CN-CL and CN-VAL neurons also project to other brain regions besides the thalamus? Might these pathways also contribute to learning and consolidation of the accelerating rotarod task? Please discuss.

This is now discussed in more detail in the last paragraph of the section “A specific impact on learning of CL-projecting CN neurons.”

Reviewer #3 (Recommendations for the authors):

(1) Please check the anatomic evidence for the strict dichotomy between intralaminar (specifically central lateral nucleus) nuclei projecting to the striatum and the ventral-anteriorlateral (VAL) complex projecting to the cortex. For example, while the Chen et al paper shows that there are cerebellar-intralaminar-striatal projections, it does not exclude intralaminar cortex projections, which have at least been demonstrated in rats. Similarly, VAL has projections to striatum (see, e.g., Smith et al, "The thalamostriatal system in normal and diseased states", Frontiers in Systems Neuroscience, 2014). It may be that some of these projections are stronger, but I don't think it's true that these pathways are as well-separated as the authors suggest. I also don't think this changes the fundamental conclusions but is important for potential mechanisms by which differential learning could occur and necessitate modification of Figure 5.

We have toned down the interpretation of CL and VAL relaying specifically to different brain structures and mostly put forward the duality of the pathways. The connections with the cortex are now discussed at the end of the section “A specific impact on learning of CL-projecting CN neurons.”

(2) Please provide more details on the spike sorting. By what metrics were single units declared to be well-separated? How many units were identified under each condition? What was the distribution of firing rates with and without CNO treatment? Are the units shown in panel 1f from before and after CNO as in panel E or are just 2 examples of isolated units? The units by themselves are not very helpful to the reader. Showing sample auto and/or crosscorrelograms for units recorded on the same electrode would be more helpful to show how well-isolated the units are.

Single units were considered well-isolated based on quantitative quality metrics computed after MountainSort 4 spike sorting (Phyton 3.8). Units were required to have a signal-to-noise ratio (SNR) greater than 5, inter-spike interval (ISI) violations less than 1%, an amplitude cutoff below 0.1, a presence ratio above 0.9, a firing rate greater than 0.1 Hz, and at least 50 detected spikes. In addition, units were assessed for temporal stability across the recording using autocorrelograms and presence over the recording, ensuring there were no prolonged periods of total inactivity. Units meeting these criteria were deemed well-separated and reliable for further analysis. This has been added to the Methods.

Cell numbers are provided with the statistics in the supplementary table for fig panel 1g. Panels are from the same unit before and after CNO. Example of auto- crosscorr- are provided in the new Supplementary Figure 6.

(3) Panel 2g - "firing rate modulation" is unclear. I think the authors are showing the mean firing rate with DREADD+CNO treatment divided by the mean firing rate in the pre-CNO condition for the same group (I couldn't find that in the Methods, my apologies if I missed it)? However, firing rate modulation to me means variability in firing rate within a recording. Perhaps "relative firing rate" or "% pre-CNO firing rate" would be clearer?

The definition has been added to the Method and the axis has been changed to ‘Change in FR induced by SAL/CNO’

(4) Figure 3f - why does consolidation appear to be impaired after the transition from CNO to saline between sessions, when in panel 1j suppressing the CN does not have a similar effect once CNO is switched to saline? Could this be driven by a small number of mice? Since a central conclusion of the paper is that CN-VAL connections are uniquely important for posttraining consolidation, this discrepancy is important to explain - if the results post-saline are spurious, how do we know that the results post-CNO aren't also spurious? Panels similar to Figure 4b and d showing all the data from the last/first trial of each session I think would be convincing.

Our results overall indicate that the overnight consolidation of the improvement in performance seem only effective in the early phase (as pointed out on the summary figure 5). We do not believe then that the saline results are spurious.

It can be seen indeed in the control groups of the figure 1; to make this more visible, we plot in Author response image 1 the difference between trial 7 and trial 1 the next day. An overnight drop in performance becomes visible in the late phase.

Author response image 1.

The decrement on the first trial in the first 3 days is visible for the majority of the mice. The plot asked by the reviewer is represented in the Author response image 2.

Author response image 2.

Minor points:

(5) In panel 1a, the solid yellow line obscures a lot of the image and I don't think adds anything.

We assume this was referring to a line on fig1d, which has been removed.

(6) Panel 2a - color selection could present problems for those with red-green color blindness.

This has been fixed.

(7) Supplementary Figure 3 - what are the arrows and arrowheads indicating?

These have been removed.

(8) In the Discussion: "Studies of cerebellar synaptic plasticity provide clearly support the involvement of cerebellum in rotarod learning..." Delete the word "provide"

This has been fixed

(9) "This indicates that either the distinct functional roles of VAL-projecting or CLprojecting." The second "of" should be "or", I think.

This has been fixed.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

Fecal virome transfer (FVT) has the potential to take advantage of microbiome associated phages to treat diseases such as NEC. However, FVT is also associated with toxicity due to the presence of eukaryotic viruses in the mixture, which are difficult to filter out. The authors use a chemostat propagation system to reduce the presence of eukaryotic viruses (these become lost over time during culture). They show in pig models of NEC that chemostat propagation reduce the incidence of diarrhea induced by FVTs.

Strengths:

The authors report an innovative yet simple approach that has the potential to be useful for future applications. Most of the experiments are easy to follow and performed well.

Weaknesses:

The biggest weakness is that the authors show that their technique addresses safety, but they are unable to demonstrate that they retain efficacy in their NEC model. This could be due to technical issues or perhaps the efficacy of FVT reported in the literature is not robust. If they cannot demonstrate efficacy of the chemostat propagated virome mixture, the value of the study is compromised.

We appreciate the reviewer’s assessment and fully acknowledge that our inability to demonstrate NEC protection by FVT is a limitation to the study. If technical issues cover the variability in disease phenotype in our animal model, which is of a spontaneous nature, then yes we fully agree. Issues with FVT preparation are however unlikely, as this is performed per protocol. The effect of FVT on NEC has hitherto only been demonstrated by our research group in two individual studies using separate donor fecal material, so it is indeed too early to speculate about robustness in FVT response. We have briefly mentioned this in the results (lines 563-565) and discussion (lines 777-779), but agree that it needs further elaboration. We have now revised the discussion and conclusion to better emphasize the extent and consequences of this limitation (lines 793-797 + lines 817-818). Importantly, we show that inclusion of specific nutrients, such as milk oligosaccharides, impacts the resulting propagated fecal-derived virome. One can argue that this is not surprising, but it has nevertheless not been shown before – and it opens up possibilities for future “tailor-made” fecal-derived viromes with predictable profiles and effects.

Even though we do not demonstrate an effect of the chemostat-propagated virome, we still believe that the study provides valuable insights as a proof-of-concept. Specifically, we demonstrate that in vitro chemostat propagation can significantly modulate the safety profile of FVT, while still driving changes in the microbiome, e.g., by decreasing C. perfringens.

The above issue is especially concerning because the chemostat propagation selected for bacteria that may not necessarily be the ones that harbor the beneficial phages. Without an understanding of exactly how FVT works, is it possible to make any conclusion about the usefulness of the chemostat approach?

The chemostat work was based on the idea that if we culture a fecal inoculum under suitable conditions, then the phageome would propagate alongside and allow for a scalable production method for standardized donor-independent FVT. We are cognizant that the chemostat end-culture diverged quite markedly from the fecal inoculum. In reality, such divergence is unavoidable when performing in vitro simulation of intestinal growth conditions. On the positive side, we showed that we could drive an expansion of Bacteroides spp. by supplementing the media with human milk oligosaccharides. We have previously shown that Bacteroides spp. engraft FMT recipients that are in turn protected from NEC. However, there is much room for refinement of the chemostat culture condition; i.e. to preserve the rich repertoire of lactobacilli from the inoculum e.g. by means of lowering the pH. Moreover, the loss of viral diversity in the chemostat end-culture also needs to be addressed, potentially by lowering the chemostat dilution-rate to allow the time for phage propagation. Based on these insights, we will in the near future invest heavily in improving the chemostat procedure to end up with a propagated fecal virome with better resemblance to the fecal inoculum.

Finally, can the authors rule out that their observations in THP-1 cells are driven by LPS or some other bacterial product in the media?

We thank the reviewer for raising this point. To minimize the influence of bacterial contaminants such as LPS or other small bacterial products, we implemented several steps during sample preparation. Specifically, we performed ultrafiltration using a 300 kDa molecular weight cut-off, which should remove small molecules, including LPS, bacterial metabolites, and other potential soluble immunomodulators. Hereafter, all viral preparations underwent endotoxin removal procedures prior to cell exposure. These precautions reduce the likelihood that our observed effects in THP-1 cells are attributable to bacterial products rather than viral components. This is explained in the referenced article (20), but we have now added the clarification to the Methods section of the revised manuscript (lines 222 and 227). The immune expression profile differs markedly between the viral preparations and the E. coli control, e.g. IFNG, TLR3, TLR8, making it highly likely that viral epitopes are the major drivers of the viral preparations with less impact by any potential bacterial epitope contaminant. This is now mentioned in the results section (line 541-543):

Reviewer #2 (Public review):

Major revision

(1) As authors state that the aim of the research is 'We hypothesized that chemostat propagated viromes could modulate the GM and reduce NEC lesions while avoiding potential side effects, such as earlier onset of diarrhea'.

(a) For the efficacy, in Fig 5, there are no significance in stomach pathology and enterocolitis between groups, even between control group and experimental groups, is it because of the low incidence of NEC? This may affect the statistical power of the conclusions. Therefore, it is unclear how one can draw the conclusion that chemostat can reduce NEC lesions?

Thank you for highlighting this important point. We fully agree and would like to clarify that it is not our intention to conclude that chemostat propagation reduces NEC lesions under the experimental settings within this paper. Rather, this was our initial hypothesis, which could not be confirmed. The unexpectedly low incidence of NEC across groups in Piglet Experiment 1 did not allow for a clear conclusion, but the second Piglet Experiment 2 failed to show a NEC-reducing effect. We have stated this important point in the following sections:

- Abstract (line 42-44): “However, these signatures were lost in recipients of chemostat-propagated viromes, and only minor microbiome effects and no NEC prevention were observed.”

- Results (line 699): “This highlights that while chemostat propagation effectively mitigates virus-associated diarrhea, the method needs further optimization to targt NEC.”

- Discussion (lines 773–775): “However, the MO-propagated chemostat virome did not increase Bacteroides or Parabacteroides spp. in the recipient’s gut, nor did it provide NEC protection.”

- We have rephrased this to emphasize the importance of Experiment 2.

- To avoid any potential misinterpretation, we have rephrased line 598 to reflect that we observed “a difference in the clinical side effect pattern” rather than implying efficacy.

- Furthermore, we have updated the summary title for Figure 8 (line 704) to clearly state: “MO-propagated virome modestly exacerbates gastric injury and fails to improve NEC.”

- Also, we have added the following section to the discussion (lines 793-797): “However, we acknowledge that the absence of demonstrated NEC prevention by the native donor virome is a significant limitation to conclusions regarding efficacy. Without a protective baseline, we cannot assess whether the virome efficacy was lost during chemostat propagation. Consequently, we cannot confirm or dismiss the hypothesis that chemostats can preserve a phage community capable of preventing NEC.”

- Lastly, we have updated the conclusion (lines 817-818): “However, as neither the chemostat-propagated viromes nor the native donor virome demonstrated NEC prevention, the efficacy of the chemostat approach remains inconclusive.”

- These changes should clarify that while the study demonstrates improved safety via reduced diarrhea, NEC efficacy was not obtained.

(b) More convincing pathology images would be helpful.

Since we did not observe a protective effect against NEC with either of the treatments, we opted not to include pathology images. However, extensive examples can be found in the cited paper (reference 37), which describes our NEC scoring methodology in the Methods section (lines 268-271): https://doi.org/10.1016/j.yexmp.2024.104936.

(c) For the safety, such as body weight development, FVT had no statistical significance difference from control, CVT, and CVT-MO, so how can you drawn the conclusion that chemostat can avoid potential side effects?

We appreciate the reviewer’s observation. To clarify, we do not claim that chemostat propagation completely avoids all potential side effects, but rather that it mitigates them. As shown in Fig. 5G, FVT recipients exhibited significantly reduced body weight gain compared to controls, CVT, and CVT-MO specifically on day 4, but not on day 5. This transient effect suggests that side effects such as reduced growth and early-onset diarrhea are delayed, not entirely prevented, by chemostat propagation. This is stated in the results section in lines 593-595. We also believe that this is consistent with the paper title and the conclusion that the chemostat process minimizes the adverse effects associated with native FVT (line 813).

(d) There is lack of evidence to convince the reader that there is a decrease of eukaryotic viruses. More quantitative data here would be useful.

Apart from the fact that it is impossible for eukaryotic viruses to shed in a system devoid of eukaryotic cells, and that the chemostat runs continuously exchanges the culture, thereby diluting any substance incapable of propagation, we agree that quantitative data to demonstrate a reduction of eukaryotic virus load is lacking.

However, in this case we believe the relative viral abundance data are almost as convincing. To make this even clearer, we have produced new graphs showing 1) the eukaryotic viral abundance relative to total viral abundance and 2) observed eukaryotic viral species, both after medium subtraction. Eukaryotic viral relative abundances decrease from around 0.4% to approach zero already in the batch phase, and similarly number of eukaryotic viral species decrease from around 10 in the fecal inoculum to zero midway through the chemostat phase. These new graphs are now part of Supplementary figure S3 B-C. Moreover, an error in the eukaryotic viral heatmaps presented in Figure 3F now means that the relative abundance of each sample (column) now sums up to 100%. Please also notice from the lower heatmap (where the virome signature of the medium is subtracted) that no eukaryotic viruses are identified from the sequencing data of the samples from the chemostat from 50 hours and onwards.

However, for future experiments we will consider adding a known quantity of a marker virus to the inoculum and monitoring its concentration (e.g., by qPCR) throughout the culture process. Importantly, if the resulting virome is meant for in vivo testing, this marker virus should be inert to the receiving organism.

(2) Questions regarding Fig 3F,

(a) How can the medium have 'the baseline viral content' ?

As we have previously seen persistent eukaryotic viral signals in metagenomics sequencing data from chemostat experiments, we sampled and sequenced the culture medium. As is seen from Figure 3F, this only concerns Dicistroviridae, as the patterns of the remaining eukaryotic viral signals before and after medium subtraction are virtually similar. For some reason, a component of the culture medium contains a genetic signal from this entity. Since all culture components are sterilized, it is most likely genomic traces that are then continuously supplied with the medium and appears in all culture samples. As it is unlikely to derive from intact viruses, the in vivo implications are deemed minimal.

(b) What is the statistical significance of relative abundance of specific eukaryotic viruses?

The same as any statistical comparison on single OTU level in a nucleotide sequencing dataset. As commented above, it does not prove a quantitative depletion of eukaryotic virus throughout the chemostat process but given the context a reduction in relative abundance supports the notion that eukaryotic viruses are indeed depleted when the culture medium is exchanged. The relevant question to us is: What is the magnitude of depletion? Which is particularly relevant since the clinical data indicates a delay and not a prevention of side effects after transplantation. Hence, as proposed above, the use of a marker virus would provide us with that answer.

(c) The hosts for some of the listed eukaryotic viruses are neither pigs or human, as such the significance of a decrease in these viruses to humans is unclear.

Dicistroviridae is not present in the inoculum and shows up only when medium is added. Picobirnavirus and Astrovirus are relevant mammalian intestinal viruses, whereas Smacoviridae is less well described (dois: 10.3389/fvets.2020.615293 and 10.3390/v8020042). Genomoviridae as a fungal virus indeed appears to be less relevant in the case of the mammalian intestine. Indeed, at any given time point in any given individual, be it a pig or a human, it would carry with it several viral species that are incapable of infecting it, most likely transiting after being ingested with food, or in the case of pigs through rummaging. It is no secret that we have been searching for a causative agent responsible for the clinical side effect patterns related with FVT, but there seems to be no consistent viral agent that is overabundant in diarrheal piglets. Hence, in this study, we are mostly interested in the proof-of-concept for overall eukaryotic virus reduction through chemostat propagation, and we believe we have presented data in support of this.

(3) In this study, pH 6.5 was selected as the pH value for chemostat cultivation, but considering the different adaptability of different bacteria to pH, it is recommended to further explore the effect of pH on bacteria and virus groups. In particular, it was optimized to maintain the growth of beneficial bacteria such as Lactobacillaceae and Bacteroides in order to improve the effect of chemostat cultivation.

We agree that pH is a key parameter in shaping microbial communities during chemostat cultivation. As noted, we selected pH 6.5 to balance physiological relevance and bacterial viability, but we acknowledge that this pH may not be optimal for supporting the growth of certain potentially beneficial taxa such as Lactobacillaceae. We explicitly address this in the discussion (lines 736–741), where we state that the selected pH may have limited engraftment and that future studies should investigate pH optimization to better support bacterial groups and improve the overall effectiveness of the cultivation system.

(4) Please improve the quality of the images, charts, error bars and statistical significance markers throughout and mark the n's. used in each experiment.

We have carefully reviewed all figures and could not identify any general image quality issues. If some specific images or panels appear unclear or problematic, we would appreciate it if the reviewer could point them out so we can address them directly.

Regarding sample sizes, the number of animals (n) is indicated in Fig. 5A and its legend, as well as in Fig. 8A. We have now also added this information to the legend of Fig. 8 for clarity.

To improve the clarity of statistical findings, we have added asterisks to denote significance in panels 6A, 6F, and 7A, as requested.

To improve the clarity of Fig. 3B, we have added a dashed line to separate LAC and LAC-MO.

Reviewer #3 (Public review):

Major revisions

This study investigated the in vitro amplification of donor fecal virus using chemostat culturing technology, aiming to reduce eukaryotic virus load while preserving bacteriophage community diversity, thereby optimizing the safety and efficacy of FVT. The research employed a preterm pig model to evaluate the effects of chemostat-propagated viromes (CVT) in preventing necrotizing enterocolitis (NEC) and mitigating adverse effects such as diarrhea.

Strengths:

Enhanced Safety Profile: Chemostat cultivation effectively reduced eukaryotic virus load, thereby minimizing the potential infection risks associated with virome transplantation and offering a safer virome preparation method for clinical applications.

Process Reproducibility: The chemostat system achieved stable amplification of bacteriophage communities (Bray-Curtis similarity >70%), mitigating the impact of donor fecal variability on therapeutic efficacy.

Weaknesses:

Loss of Phage Functionality: The chemostat cultivation resulted in a reduction in phage diversity (e.g., the loss of Lactobacillaceae phages), which may compromise their protective effects against NEC (potentially linked to the immunomodulatory functions of Lactobacilli). The authors should explicitly address this limitation in the discussion section, particularly if additional experiments cannot be conducted to resolve it within the current study.

We appreciate the reviewer’s concern and agree that the loss of phage diversity during chemostat cultivation, especially phages targeting Lactobacillaceae, is an important limitation with potential implications for NEC protection.

We already described the depletion of Lactobacillaceae in the chemostat and its implications in the discussion (lines 742-751 + 787-793), along with our plans to address this in future work by adjusting culture pH. However, we acknowledge that the significance of losing phage diversity deserves more explicit attention. Accordingly, we have expanded the discussion to highlight the possible consequences of this loss and its impact on phage functionality (see lines 758–762), as suggested by the reviewer.

Limitations in Experimental Design: The low incidence of NEC lesions in the control group reduced the statistical power of the study. This limitation undermines the ability to conclusively evaluate the efficacy and safety of the chemostat-propagated virome as a novel intervention for NEC. Future studies should optimize experimental conditions (e.g., using a more NEC-susceptible model or diet) to ensure adequate disease incidence for robust statistical comparisons.

We agree that the low NEC incidence in Experiment 1 limited the statistical power to evaluate efficacy. To address this, we designed Experiment 2 using a more NEC-inducing diet (formula 2), which resulted in a higher level of baseline lesions. This allowed for a more conclusive assessment, demonstrating that the MO-propagated chemostat virome did not provide NEC protection when using the donor feces and culture conditions applied in this experiment.

We acknowledge that this was too unclear in the original manuscript. Please see the response to the first comment by Reviewer 2, where we have highlighted several revisions to improve clarity.

However, we do believe the data are robust enough to conclude that the level of diarrhea — and thereby safety — was improved in the piglet model, which is why we chose to focus on this aspect in the paper’s title.

Recommendations for the authors:

Reviewer #3 (Recommendations for the authors):

The manuscript presents a well-structured study investigating the feasibility of using chemostat-based culturing of the fecal virome to reduce the transfer of eukaryotic viruses during fecal virome transfer (FVT). Utilizing both in vitro fermentation systems and a preterm piglet model, the authors explore whether this method could be a safer and equally effective alternative to raw FVT for treating neonatal intestinal diseases, such as necrotizing enterocolitis (NEC). This study introduces a novel mitigation strategy for FVT through chemostat fermentation. However, a significant revision is recommended before the manuscript can be considered for publication.

Major Changes:

- A central aim of the study was to assess whether chemostat-cultured viromes maintain protective effects against NEC. However, this key outcome remains "unresolved" due to the low incidence of NEC in the control group. The discussion should address this limitation.

We fully acknowledge this limitation and agree that our study cannot conclude whether the NEC effect of FVT was maintained without demonstrating an effect of this native virome. Please see our response to a similar concern raised by Reviewer 1, where we describe the revisions made to the discussion (lines 793-797) and conclusion (lines 817-818).

- The section on viral particle enrichment should be expanded and discussed in more detail. It would be beneficial to examine its efficiency in separating bacteria from viral-like particles (VLPs) compared to findings from previously reported studies. The authors should clarify the rationale behind the selected dose of VLPs used in the experiments and their role in virus engraftment results.

We selected the virome isolation method based on previous experiments within our lab, demonstrating efficient separation of bacteria and virus particles, using a 0.45 um filter syringe. Filtrates were quality assessed by fluorescence microscopy, showing absence of intact bacteria. Using a diverse mock virus community, we also showed a high degree of preservation of infective viruses in the FVT following the isolation procedures. We have now expanded the description of the separation method in the results section with a reference to this work (lines 188-190). We did however choose to increase the molecular weight cut off (MWCO) to enhance the exclusion of non-viral components.

We acknowledge that the rationale and importance of the VLP dose was lacking in the discussion. This has now been added (line 758-762).

- The viral richness of chemostat viromes was significantly lower than that of native feces. The authors should discuss how this may impact microbiome and virome outcomes.

We have included this point in the new section about VLP dose in the discussion. Please see lines 758-762.

- The immune response was assessed through THP-1 cells and a limited piglet cytokine panel. These may not fully represent the intestinal epithelial or mucosal immune responses. Thus, authors should acknowledge these limitations in the discussion section.

Thank you for the comment. The limitation of using THP-1 cells as an in vitro model is already acknowledged in the results section (line 545): “Since fecal-derived eukaryotic viruses mainly infect intestinal cells, an

in vivo stimulation may reveal a different response pattern. ”

The limited panel of porcine cytokines was not intended as a comprehensive assessment of the mucosal immune response, but rather as supportive data for NEC-associated inflammation, as we have previously demonstrated (reference 37: https://doi.org/10.1016/j.yexmp.2024.104936). To obtain a comprehensive view of the immune response, a few days after diarrhoea onset, we additionally performed RNA-Seq analyses of the intestinal lymph node.

- While the manuscript is comprehensive, it is also lengthy and text-heavy. Some sections could be condensed for clarity.

The manuscript has been through multiple revisions by authors. While it is indeed lengthy, we have removed non-essential information and redundancies and now feel that the balance between data, text, figures, and supplementary information is acceptable.

- Several figures (e.g., Figs. 1-5) contain significant data but need clearer summaries in their captions.

We appreciate the suggestion and have revised the captions for Figs. 1-8 to provide clearer, more informative summaries of the data they present.

Author response:

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

We sincerely thank the reviewer for the thorough and constructive evaluation of our manuscript. We greatly appreciate the recognition of our work's strengths, particularly the integration of experiments and mathematical modeling, the stochastic framework for describing sloughing events, and the insights into pressure-driven detachment dynamics.

We have carefully considered each point raised and provide detailed responses below. In response to the reviewer's comments, we have revised the Methods section to better clarify our approach to three-dimensional assessment. We believe these revisions have improved the clarity of the manuscript.

Below, we address each of the specific concerns raised by the reviewer:

Public Reviews:

Reviewer #1 (Public review):

Weaknesses:<br /> The study achieves its primary goal of integrating experiments and modeling to understand the coupling between flow and biofilm growth and detachment in a microfluidic channel, but it should have highlighted the weaknesses of the methods. I list the ones that, in my opinion, are the main ones:

The study does not consider biofilm porosity, which could significantly affect the flow and forces exerted on the biofilm. Porosity could impact the boundary conditions, such as the no-slip condition, which should be validated experimentally.

Porosity is indeed a key component of biofilm structures, resulting from the polymeric nature of the EPS matrix, mechanical forces, and biological processes such as cell death or predation. When considering flow-biofilm interactions, this porosity may allow fluid flow through the biofilm, with reported permeability values spanning an extremely broad range from 1015 to 10-7 m2 (Kurz et al., 2023).

However, we argue that biofilm permeability is not the primary driver in our system:

(1) In microscopy visualization, our biofilms form dense structures where flow around the biofilm through narrow channels dominates over flow through the porous biofilm matrix.

(2) We performed microrheology experiments in these biofilms by imaging the Brownian motion of nanoparticles in the biofilm. Their trajectories indicate that, in our conditions, the viscoelastic flow of the biofilm itself largely dominates over the flow of culture medium through the biofilm matrix.

(3) We argue that the extreme variability in reported permeability values (spanning several orders of magnitude, Kurz et al., 2023) reflects not only differences in experimental systems, but also fundamental challenges in defining and measuring permeability for viscoelastoplastic biofilms (the biofilm itself is actually flowing). Given this uncertainty, incorporating permeability into our model would introduce parameters that cannot be reliably constrained from literature or independently measured in our setup. Our approach (i.e. treating the biofilm as impermeable and focusing on flow obstruction) avoids this parametrization complexity while successfully capturing the observed dynamics.

(4) Our model successfully predicts the observed scaling laws (φmax ∝ Q1/2, Fig. 7f) and hydraulic resistance dynamics (Fig. 3) without invoking permeability, suggesting that flow obstruction rather than flow penetration is the dominant mechanism.

Reference: Kurz, D. L.; Secchi, E.; Stocker, R.; Jimenez-Martinez, J. Morphogenesis of biofilms in porous media and control on hydrodynamics. Environ. Sci. Technol. 2023, 57 (14), 5666−5677.

The research suggests EPS development as a stage in biofilm growth but does not probe it using lectin staining. This makes it impossible to accurately assess the role of EPS in biofilm development and detachment processes.

We respectfully disagree that lectin staining is necessary to assess the role of EPS in our system, and we argue that our approach using genetic mutants is superior for the following reasons. Lectin staining has significant limitations. While widely used, lectin staining (e.g., concanavalin A) is non-specific (binding not only to EPS polysaccharides but also to bacterial cell surfaces) and is non-quantitative. It can confirm the presence of polysaccharides but cannot establish causal relationships between specific EPS components and mechanical properties or detachment dynamics. We performed preliminary experiments with ConA-rhodamine (data not shown), which showed widespread presence of polysaccharides. However, this provided limited insight beyond confirming EPS production, which is well-established for P. aeruginosa PAO1 biofilms. We employed a more rigorous genetic approach to directly assess the role of EPS composition. We used Δpel and Δpsl mutants (strains lacking key exopolysaccharides that are the primary structural components of the PAO1 matrix). Our results demonstrate that both mutants show significantly reduced maximum clogging compared to wild-type. The Δpsl mutant is particularly affected, with near-complete detachment at certain flow rates. These differences directly link EPS composition to mechanical stability and detachment dynamics. This genetic approach provides causal, quantitative evidence for the role of specific EPS components in biofilm development and detachment, information that lectin staining cannot provide. We believe this addresses the reviewer's concern more rigorously than lectin staining would.

While the force and flow are three-dimensional, the images are taken in two dimensions. The paper does not clearly explain how the 2D images are extrapolated to make 3D assessments, which could lead to inaccuracies.

We thank the reviewer for this important observation. We would like to clarify our methodological approach. Our primary three-dimensional measurement is the hydraulic resistance R(t), obtained from pressure drop measurements across the biofilm-containing channel section. This pressure-based measurement inherently captures the three-dimensional flow obstruction caused by the biofilm. We then employ a geometric model (uniform biofilm layer on all channel walls) to convert R(t) into volume fraction φ(t).

The two-dimensional fluorescence imaging serves to validate this model-based approach rather than being the basis for three-dimensional extrapolation. The uniform layer assumption is supported by three independent lines of evidence: (i) the excellent quantitative agreement between predicted and measured scaling laws (φmax ∝ Q1/2, Fig. 7f), obtained without adjustable parameters; (ii) the high reproducibility of φmax values across different flow rates and replicates; and (iii) the strong correlation between model-derived φ(t) from pressure measurements and integrated fluorescence intensity (Fig. 3b-d).

We have added clarifying text in the Methods section (subsection "Data analysis for the calculation of the hydraulic resistance and volume fraction") to better explain this approach and emphasize that pressure measurements provide the three-dimensional information, with the geometric model serving as the link to volume fraction.

Although the findings are tested using polysaccharide-deficient mutants, the results could have been analyzed in greater detail. A more thorough analysis would help to better understand the role of matrix composition on the stochastic model of detachment.

We thank the reviewer for this suggestion. Our mutant analysis demonstrates that Δpsl and Δpel strains have significantly reduced φmax and altered detachment dynamics compared to wild-type (Fig. 8), directly linking EPS composition to mechanical stability as predicted by our model. A rigorous quantitative connection between matrix composition and the stochastic parameters (interevent times, jump amplitudes) would require: (i) substantially more sloughing events for statistical power, (ii) independent mechanical characterization of each mutant, and (iii) a mechanistic model linking EPS composition to detachment parameters. We are currently developing microrheology approaches to characterize mutant mechanical properties, which could enable such refinement in future work.

However, this represents a substantial study beyond the scope of the current manuscript, which establishes the self-sustained sloughing-regrowth cycle and its stochastic nature. The mutant results serve their intended purpose: demonstrating that EPS composition affects detachment, consistent with our model's framework.

Reviewer #2 (Public review):

This manuscript develops well-controlled microfluidic experiments and mathematical modelling to resolve how the temporal development of P. aeruginosa biofilms is shaped by ambient flow. The experiment considers a simple rectangular channel on which a constant flow rate is applied and UV LEDs are used to confine the biofilm to a relatively small length of device. While there is often considerable geometrical complexity in confined environments and feedback between biofilm/flow (e.g. in porous media), these simplified conditions are much more amenable to analysis. A non-dimensional mathematical model that considers nutrient transport, biofilm growth and detachment is developed and used to interpret experimental data. Regimes with both gradual detachment and catastrophic sloughing are considered. The concentration of nutrients in the media is altered to resolve the effect of nutrient limitation. In addition, the role of a couple of major polysaccharide EPS components are explored with mutants, which leads results in line with previous studies.

There has been a vast amount of experimental and modelling work done on biofilms, but relatively rarely are the two linked together so tightly as in this paper. Predictions on influence of the non-dimensional Damkohler number on the longitudinal distribution of biofilm and functional dependence of flow on the maximum amount of biofilm (𝜙max) are demonstrated. The study reconfirms a number of previous works that showed the gradual detachment rate of biofilms scales with the square root of the shear stress. More challenging are the rapid biofilm detachment events where a large amount of biofilm is detached at once. These events occur are identified experimentally using an automated analysis pipeline and are fitted with probability distributions. The time between detachment events was fitted with a Gamma distribution and the amplitude of the detachment events was fitted with a log-normal distribution, however, it is not clear how good these fits are. Experimental data was then used as an input for a stochastic differential equation, but the output of this model is compared only qualitatively to that of the experiments. Overall, this paper does an admirable job of developing a well-constrained experiments and a tightly integrated mathematical framework through which to interpret them. However, the new insights this provides the underlying physical/biological mechanisms are relatively limited.

We thank the reviewer for the thorough evaluation of our work and for highlighting the tight integration between experiments and modeling. We appreciate the constructive feedback regarding the goodness-of-fit for the probability distributions.

To address the concern that "it is not clear how good these fits are," we have added quantile-quantile (Q-Q) plots for the Gamma distribution fits of inter-event times to the Supplementary Materials (Supplementary Figure S20). These plots demonstrate that the sample quantiles track the theoretical Gamma quantiles across all flow rates (0.2, 2, and 20 μL/min), indicating that the Gamma distribution provides a reasonable approximation of the overall distributional behavior. For detachment amplitudes, we selected the lognormal distribution based on the observed high skewness and kurtosis in the data, which are characteristic signatures of lognormal processes.

Formal goodness-of-fit tests (chi-square, Kolmogorov-Smirnov) yielded mixed results across datasets, passing for some while failing for others. This variability reflects inherent noise from measurements, discrete temporal sampling, automated detection thresholds, and intrinsic biological variability. Importantly, our goal is to capture essential distributional characteristics for input into the stochastic model, not to achieve perfect statistical fit across all individual datasets. The Q-Q plots confirm that these distributions provide reasonable approximations, and the qualitative agreement between model predictions and experimental observations validates this modeling approach. We have revised the Methods section to clarify this rationale.

We respectfully disagree that “new insights this provides the underlying physical/biological mechanisms are relatively limited.” Beyond confirming previous findings (e.g., scaling for gradual detachment), we believe our work provides several novel mechanistic insights. First, the Pe/Da criterion enables quantitative prediction of nutrient limitation regimes, allowing systematic decoupling of nutrient effects from other phenomena in biofilm studies. Second, we demonstrate that pressure, not shear, drives sloughing detachment events, a mechanism overlooked in previous studies where the notion of “shear-induced detachment” clearly dominates. Third, we show that sloughing-regrowth cycles occur even in single channels, establishing pressure-driven fluctuations as a signature of confined biofilm growth, independent of geometric complexity. Finally, the stochastic description of sloughing demonstrates that, while instantaneous biofilm states are irreproducible, the underlying randomness is predictable, therefore addressing a fundamental challenge in biofilm research.

Recommendations For The Authors:

Reviewer #1 (Recommendations For The Authors):

(1) In the abstract, I suggest clarifying the term "bacteria development." It is unclear if it refers to bacterial growth, biofilm formation, or biofilm detachment. The concept is expressed more clearly at the end of the Introduction.

We have modified the entire abstract to make it clearer. The abstract now explicitly establishes the key processes - growth ('nutrients necessary for growth', 'growing bacteria obstruct flow paths') and detachment ('mechanical stresses that cause detachment', 'flow-induced detachment', 'sloughing') - before using 'bacterial development' as a collective term to refer to these coupled spatiotemporal dynamics. We believe the abstract is now clear as written.

(2) Findings from Sanfilippo et al. (2019) were slightly questioned by Padron et al. (PNAS, 2023), who discovered that H2O2 transport is responsible for fro operon upregulation.

Thanks for the clarification, which is indeed significant. The new sentence now reads: Pseudomonas aeruginosa has been found to regulate the fro operon in response to flow-modulated H2O2 concentrations (Sanfilippo et al. 2019, Padron et al. 2023).

(3) Additionally, Kurz et al. (2022) account for pressure buildup as the mechanism controlling sloughing.

We respectfully disagree and note that Kurz et al. (2022) identify shear stress, not pressure buildup, as the primary mechanism controlling sloughing. Besides the title, key sentences include “opening was driven by a physical process and specifically by the shear forces associated with flow through the biofilm”, “The opening of the PFPs is driven by flow-induced shear stress, which increases as a PFP becomes narrower due to microbial growth, causing biofilm compression and rupture.” While pressure differences are measured as indicators of system state and do contribute to normal compression stresses, their mechanistic explanation emphasizes that narrowing PFPs experience increased shear rates that eventually exceed the biofilm's yield stress, triggering viscoplastic deformation and detachment. The pressure buildup is a hydraulic consequence of narrowing rather than the direct cause of sloughing. In contrast, our work demonstrates that in confined geometries, pressure differences generate tangential stresses at the biofilm-solid interface that directly drive detachment.

(4) The flow control strategy represented in Fig. 1 is not explained and should be detailed in the Methods section.

The methods section reads as follows. Inoculation and flow experiments BHI suspensions were adjusted at optical density at OD640nm= 0.2 (108 CFU/mL) and inoculated inside the microchannels from the outlet, up to approximately ¾ of the channel length in order to keep a clean inlet. The system was let at room temperature (25°C) for 3h under static conditions. Flow experiments were then performed at 0.02, 0.2, 2, 20 and 200 μL/min constant flow rates for 72h in the microchannels at room temperature. For the experiments at 0.2, 2, 20 and 200 μL/min, the fluidic system was based on a sterile culture medium reservoir pressurized by a pressure controller (Fluigent FlowEZ) and connected with a flow rate controller (Fluigent Flow unit). The flow rate was maintained constant by using a controller with a feedback loop adjusting the pressure in the liquid reservoir. The reservoir was connected to the chip using Tygon tubing (Saint Gobain Life Sciences Tygon™ ND 100-80) of 0.52 mm internal diameter and 1.52 mm external diameter, along with PEEK tubing (Cytiva Akta pure) with 0.25 mm inner diameter adapters for flow rate controller. The waste container was also pressurized by another independent pressure controller to reduce air bubble formation in the inlet part. For the experiments at 0.02 μL/min, we used an Harvard Phd2000 syringe pump for the flow.

(5) Including images of the actual biofilms formed in a portion of the channel would aid in understanding the analysis presented in Fig. 2.

Images are introduced later on (eg Figure 5). There is also supplementary material showing videos.

(6) The boundary conditions used to calculate the stress in the developed model should be discussed. The authors should specify why biofilm porosity is neglected.

We have added a detailed discussion in the supplementary (Section I.2).

(7) In the first section of the Results, the authors hypothesize that heterogeneity in biofilm development could be due to oxygen limitation. However, given the high oxygen permeability of PDMS, this hypothesis is later denied by their data. It would be prudent to avoid this hypothesis initially to streamline the presentation. Additionally, the authors should specify how oxygen levels at the inlet and outlet are measured.

We appreciate this comment and agree that streamlining would simplify the presentation. However, after careful consideration, we have chosen to retain the oxygen limitation hypothesis for the following reasons: (1) oxygen limitation is a frequently invoked mechanism in biofilm systems and deserves explicit consideration, (2) it is not immediately obvious that oxygen remains non-limiting in larger microchannels where transverse gradients could develop, and (3) systematically eliminating this plausible alternative hypothesis strengthens our mechanistic conclusion that BHI drives the observed heterogeneity. Regarding oxygen measurements: we did not directly measure dissolved oxygen concentrations. Our approach is only indirect.

(8) What is the standard deviation of the doubling time measured at different flows (page 9)?

We have indicated the standard deviation in the text. Note that the graph shows the SEM.

(9) What is the "zone of interest" in the channel mentioned on page 9?

We have added the following sentence to clarify: To further understand this effect, let us consider the mass balance of biofilm in the zone of interest -- the zone where biofilm grows in between the two UVC irradiation zones -- in the channel.

(10) Minor and major detachment events should be classified based on a defined threshold or criteria, and their frequency should be measured.

We appreciate the reviewer's concern about quantitative rigor. However, we respectfully disagree that imposing arbitrary thresholds to classify 'minor' vs. 'major' events would improve our analysis. Detachment events in our system span a continuum of magnitudes, and any threshold would be artificial and potentially misleading. Our quantitative characterization of detachment dynamics is provided through the statistical analysis of interevent times, which we show follow a gamma distribution. This stochastic framework captures the full spectrum of detachment behavior without requiring arbitrary binning. The terms 'minor' and 'major' in our manuscript are used qualitatively to illustrate the range of observed phenomena, not as formal classifications.

(11) Have the authors identified a reason for the peaks in the volume fraction in the Δpsl mutants at the highest flow rate?

The biofilm thickness following these sloughing events is below our detection limit, consistent with a residual layer of cells. However, these cells grow, leading to a time window where the fraction is measurable, before a new detachment event occurs. Our understanding is that the psl mutant forms a weaker matrix with a much lower threshold for sloughing.

(12) The fit of the probability density function for the relative density function does not match the data well. The authors should comment on this.

We have added quantile-quantile (Q-Q) plots for the Gamma distribution fits of inter-event times to the Supplementary Materials (Supplementary Figure S20). These plots demonstrate that the sample quantiles track the theoretical Gamma quantiles across all flow rates (0.2, 2, and 20 μL/min), indicating that the Gamma distribution provides a reasonable approximation of the overall distributional behavior. For detachment amplitudes, we selected the lognormal distribution based on the observed high skewness and kurtosis in the data, which are characteristic signatures of lognormal processes. Formal goodness-of-fit tests (chi-square, Kolmogorov-Smirnov) yielded mixed results across datasets, passing for some while failing for others. This variability reflects inherent noise from measurements, discrete temporal sampling, automated detection thresholds, and intrinsic biological variability. Importantly, our goal is to capture essential distributional characteristics for input into the stochastic model, not to achieve perfect statistical fit across all individual datasets. The Q-Q plots confirm that these distributions provide reasonable approximations, and the qualitative agreement between model predictions and experimental observations validates this modeling approach. We have revised the Methods section to clarify this rationale.

(13) Additionally, the simulated fraction appears very flat, with limited detachments compared to experiments. Why?

The model captures the essential dynamics of growth-detachment cycles, including the characteristic timescales and volume fraction ranges. Some event-to-event variability in the experimental data likely reflects biological stochasticity not captured by our current approach—for example, variations in local biofilm mechanical properties or matrix composition that affect the precise stress at which sloughing occurs. While incorporating such biological variability as a stochastic parameter would improve detailed agreement, it would require extensive additional characterization beyond the scope of this study. The current model successfully reproduces the key qualitative and semi-quantitative features of the system.

(14) The methods section should include a more detailed explanation of how the model was validated against experimental data.

Model validation was performed by comparing predicted biofilm volume fraction time series and sloughing event statistics against experimental observations across multiple flow rates. The model reproduces the characteristic growth-sloughing cycles, timescales, and steady-state volume fractions without additional parameter fitting beyond the experimentally measured distributions.

(15) It would be useful to include information on the reproducibility of the experiments and any variations observed between replicates.

Experiments were performed in N=3 biological replicates. Individual time series for all replicates are shown in Supplementary Figures, demonstrating consistent behavior across replicates.

(16) A discussion of the limitations of the study, particularly regarding the assumptions made in the modeling and their potential impact on the results, would strengthen the paper.

We have added a discussion on why we chose to neglect the porosity of the biofilm, and strengthened parts on the uniform biofilm layer assumption.

Reviewer #2 (Recommendations For The Authors):

Page 2: "A vast" —> "The vast"

Changed.

The text and line widths on many of the figures are far too small. I printed it out at normal size, but had to look at a PDF and magnify to actually see what the graphs are showing. Fig. 9c is particularly illegible.

Changed.

Fig. 1 caption "photonic" —> "optical"?

Changed

Can you spell out the actual mathematical definition of 𝜙 on page 5 when it is introduced? Currently it just says the "cross section volume fraction of the biofilm", but that seems potentially ambiguous. It is valid to say that this is "fraction of the cross section occupied by the biofilm"?

Changed

Bottom of page 5: can you state the physical interpretation of the assumption that M is bounded between 0 and 1. i.e. that growth is larger than detachment?

There is a comment on that in the paper. It reads “In assuming that M ∈ ]0, 1] and eliminating cases where M > 1, we have not considered situations of systematic detachment 𝜙equ = 0 for any value of the concentration, since this is not a situation that we encountered experimentally.” This comes just after presenting the expression on the only non-trivial steady-state, as it becomes easier to explain the consequences of the initial choice at this point.

Currently the choice of detachment initially used in the model is a bit confusing. You say that you are going to assume a (1-𝜙)-1 model for simplicity (bottom of page 5), but then later you find that the (1-𝜙)3/4 model is more accurate (page 16). Since the latter has already been confirmed in numerous other studies, why not start with that one from the beginning?

We thank the reviewer for this important question, which highlights an area where our presentation could be clearer. We did not find that the (1-φ)-3/4 model is "more accurate." Rather, we deliberately chose the (1-φ)-1 scaling because it captures pressure-induced detachment, which we hypothesized would dominate in confined flows where biofilms clog a large portion of the channel. The (1-φ)-3/4 scaling, widely used in previous studies, describes shear stress at the biofilm/fluid interface and was developed primarily for reactor systems where pressure effects are negligible. Our analysis on page 16 validates this choice by demonstrating that pressure stress indeed exceeds shear stress when volume fraction is large, which corresponds to late Stage I and all of Stage II precisely where our model is applied. The excellent quantitative agreement between predicted and measured φmax values across flow rates (Fig. 7f, Table 1) further supports the (1-φ)-1 scaling. We recognize that our initial presentation may have suggested the (1-φ)-1 choice was merely for "simplicity." We have revised this section to emphasize that this scaling was chosen specifically to capture pressure-driven detachment in confined geometries, with the physical justification provided by the stress analysis that follows. We have also clarified our ideas on page 16 to express clearly that (1-φ)-3/4 is never used. We could alternatively use a multi-modal detachment function combining both scalings, but the data do not require this additional complexity.

In general, the models you derived in this study could be better contrasted with that from previous works. e.g. can you compare your Eqn (4) with the steady-state solutions obtained by other previous studies? Is this consistent with previous works or different? (aside from framing the biofilm thickness in terms of 𝜙)

We are currently working on a paper dedicated to modeling biofilm development in confined flows, which will do a better job at comparing approaches.

Top of page 6 - you assume K* = 0.1 - Does this assume that cells grow at half the rate in 0.1X BHI as they do in 1X BHI? Has this been confirmed experimentally or is this just a guess?

This was estimated rather than measured directly. Model predictions were a lot more sensitive to the Damköhler number, than to the value of K.

"radial" is used widely in this paper, but you are using a square geometry. Is "transverse" a better choice?

Yes it clearly is. It’s been changed.

Fig 3. Are panels (a) and (b) showing different bioreps of the same condition? If so, please spell that out in the caption.

There was an error here in the caption of fig a. This has been changed. The correspondence is between a and c, and these are exactly the same, not bioreps.

In multiple places it noted that the change in hydraulic resistance is correlated with the "change in biofilm colonization." Why not demonstrate this directly using a cross correlation analysis? How is the latter connected to the 𝜙 parameter? (e.g. is this d(𝜙)/dt?)

We thank the reviewer for this suggestion. To clarify: φ(t) represents the volume fraction of biofilm in the channel. We measure this in two independent ways: (1) φ(t) from hydraulic resistance (black line in Fig. 3) i.e. calculated from pressure measurements using φ = 1 - √(R₀/R(t)), assuming uniform layer growth (see Methods section "Data analysis for the calculation of hydraulic resistance and volume fraction") and (2) φ(t) from fluorescence (green squares in Fig. 3) i.e. estimated from integrated GFP intensity or image segmentation of the glass/liquid interface. The reviewer is correct that we should quantify this relationship directly. We have now added correlation analysis between these two independent measurements of φ (new Supplementary Figure S21). The analysis shows strong positive correlation, with r-values ranged from 0.68 to 0.77 across all flow rates. This validates two key aspects of our approach: (1) the uniform layer assumption used to convert R(t) to φ(t) is reasonable, and (2) the pressure-based measurements accurately capture the dynamics visible in fluorescence imaging, including both growth phases and sloughing events. The strong agreement is particularly notable given that these measurements probe different aspects of the biofilm: hydraulic resistance is sensitive to the three-dimensional obstruction of flow, while fluorescence captures primarily the biofilm attached to the glass surface within our focal plane. Their correlation supports the model assumptions. We have revised the manuscript to clarify this relationship and present the correlation analysis.

Top of page 9 - a doubling time of 110 mins is reported in liquid culture - is this in shaken or static conditions? Can you provide some data on how this was calculated? (e.g. on a plate reader?) Do you think your measurements in the microfluidics could be affected by attachment/detachment of cells, rather than being solely driven by division. It is curious that your apparent growth rate varies by a factor of two across the different flow rates and there is not a monotonic dependency. Both attachment and detachment would depend on the flow rate (with some non-trivial dependencies).e.g. https://www.pnas.org/doi/10.1073/pnas.2307718120 https://doi.org/10.1016/j.bpj.2010.11.078

Given that your doubling time in the microfluidics is sole based on changes in cell number (rather than directly tracking cell divisions) it seems possible your results here are measuring the combined effect of growth, attachment and detachment, rather than just growth.

We agree with those comments regarding the doubling time measurement. We have added a description of how we performed the doubling time measurement in the Methods section.

Page 9 - you discuss the role of EPS here, but the effect of EPS is not demonstrated here and this is muddled with a discussion about the non-linearity of the putative dependency. Maybe this would be on a firmer footing if you save the discussion of EPS for the section on the Psl and Pel mutants?

Changed.

Middle of page 9: Please define what "smooth detachment" means and contrast it with catastrophic sloughing. Also, please define what you mean by "flow, seeding, and erosion" detachment are and how these three things differ from one another.

We have clearly defined each term in the revised version.

The results from wavelet scalograms seem to be underutilised and not well described. Can you clearly say what time series this analyses has been calculated on the caption? e.g. hydraulic resistance? Other than simply pointing out the "blue stripes", what can be gained from this analyses that could not be obtained with another method? It would be great if the basic features of this plot could more fully discussed (e.g. is the curved envelope at the bottom caused by edge effects?)

We have improved the text, captions and method section following the reviewer’s comment.

Fig. 5 a and b - please list the time at which each of these images were taken. Do these have the same dt between the two sets of images?

Yes the dt is the same (30 minutes). It’s been indicated in the caption.

Fig. 6: you have significant 2D variation in the biofilm width along the length of the channel. The relative contribution of pressure and shear based detachment will be different at different positions along the length. However, this variation is ignored in your model. Can you please comment on this in our manuscript and how it might affect the interpretation of your results? e.g. would the longitudinally averaged description yield the same result as one that takes the geometry into account (on average)?

Our model indeed assumes longitudinally averaged properties. A more detailed spatially resolved model would be valuable for capturing heterogeneities and will be explored in future work.

Bottom of page 11: you say standard deviations are in the range of 10-3. How does this jibe with the error bars on the middle flow rate in Fig. 7e?

This extremely low standard deviation only applies to the maximum value of 𝜙 and is a completely different measurement from the whisker boxes presented in fig7e.

Fig. 7: You are calculating the "Fraction" here. Is this "𝜙"? If so, can you put that on the y-axis instead? You calculate the volume fraction two different ways e.g. with hydraulic resistance and with imaging. Is only one of these shown in (e)? Is the same powerlaw dependence shown in (f) conserved when the other measurement of the "fraction" is used? Can you include both in Fig. 7e?

We have modified the axis and indicated 𝜙.

(e) is calculated only from hydraulic resistance. This is the most precise measurement to evaluate 𝜙 quantitatively.

Related to the previous comment: Some of the estimates of 𝜙max in Table 1 are obtained by fitting the model to integrated fluorescence data (Fig. 2b), while others are estimated from measurements of the hydraulic resistance. The former yields non-unique sets of parameters. Can the biofilm fraction instead actually be estimated directly from fluorescent imaging by segmenting biofilm and directly calculating how much of the cross section is occupied by cells on average across the length? This seems like a more direct measure of this quantity. Given there are multiple ways of estimating the same parameter, it would be better consistency checking to make sure that different methods actually yield the same result.

We have now added in Fig S21 a direct comparison of these two measurement methods. These are strongly correlated. Microscopy is more direct but only provides 2D pictures. Hydraulic resistance provides a 3D measurement, but relies on a model of biofilm distribution. Both are imperfect, but correlate well. In particular, we see that the 2D measurement does capture sloughing.

You cite a large number of supplemental figures (e.g. Fig. S21 on page 12), but the figures in your SI only go up to 11.

We have revised references to supplementary figures.

Bottom of page 11: Your data from liquid culture suggests that your psl mutant grows at half the rate of WT cells. Is that consistent with your microfluidic data (e.g. Fig. 8)? If not, might this be a sign that your growth rate analyses from the microfluidics might be affected by attachment/detachment? (see comment above) Psl cells should detach much more easily.

The approach taken to measure doubling times in the microfluidic system does not rely on the macroscopic measurements presented in figure 8, but rather on the approach presented in fig 4. These measurements require specific imaging (different magnification and time stepping) and we did not perform such experiments for the mutants.

In analyses of sloughing, you fit the times between the jumps and the relative amplitude. Are these two random variables correlated with one another? Might that influence your results? Your methods say that "jumps were identified through through the selection of local maxima" of the derivative. Do you to say "minima" here? Did you keep all local maxima/minima or did you have a threshold?

These are two random variables, not correlated with another. This is an assumption, and it would be interesting to analyze whether these are correlated. To perform this analysis, we believe that we would first need to acquire even more data and more replications to improve the statistical analysis.

Yes, it was minima (in the code we make everything positive, hence the confusion).

Yes, there is a threshold on the value of the jump itself. This value is extremely low and essentially filters out noise.

Fig. 9 - can you make it clearer in the caption what timeseries you are analysing here? I understand from the methods this that is the "volume fraction." The data/fits are difficult to see in Fig. 9 b and impossible to see in Fig. 9c because the green bars get in the way of the other two data sets. Can this visualisation be improved? It is not clear to me how good of a job the Gamma and log-normal fits are actually doing.

We have clarified that histograms are calculated from all experiments/replicates.

We have slightly modified the graph to make it clearer. This comparison is intrinsically hard, partly because it compares discrete data with continuous PDFs.

Aside from noting the results from the stochastic sloughing model are 'strikingly similar to experimental data', which seems to be based on a qualitative analysis of the lines in Fig. 7 d, e, and f. However, experimental data is not plotted in the same graph nor is the experimental data that we should be comparing this to cited in the text/caption.

We have added a note in the caption to indicate which figure it can be compared to.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

Matsen et al. describe an approach for training an antibody language model that explicitly tries to remove effects of "neutral mutation" from the language model training task, e.g. learning the codon table, which they claim results in biased functional predictions. They do so by modeling empirical sequence-derived likelihoods through a combination of a "mutation" model and a "selection" model; the mutation model is a non-neural Thrifty model previously developed by the authors, and the selection model is a small Transformer that is trained via gradient descent. The sequence likelihoods themselves are obtained from analyzing parent-child relationships in natural SHM datasets. The authors validate their method on several standard benchmark datasets and demonstrate its favorable computational cost.

They discuss how deep learning models explicitly designed to capture selection and not mutation, trained on parent-child pairs, could potentially apply to other domains such as viral evolution or protein evolution at large.

Strengths:

Overall, we think the idea behind this manuscript is really clever and shows promising empirical results. Two aspects of the study are conceptually interesting: the first is factorizing the training likelihood objective to learn properties that are not explained by simple neutral mutation rules, and the second is training not on self-supervised sequence statistics but on the differences between sequences along an antibody evolutionary trajectory. If this approach generalizes to other domains of life, it could offer a new paradigm for training sequence-to-fitness models that is less biased by phylogeny or other aspects of the underlying mutation process.

Thank you for your kind words.

Weaknesses:

Some claims made in the paper are weakly or indirectly supported by the data. In particular, the claim that learning the codon table contributes to biased functional effect predictions may be true, but requires more justification.

Thank you for this comment, which made us realize that we had not adequately explained the key insight of Figure S3. We have expanded the caption of Figure S3 to clarify:

“DASM selection factors match the pattern seen in experimental measurements, while masked language models show artifacts from the codon table.

The experimental data (left two panels) show a slight decrease in median scores for amino acids requiring multiple nucleotide mutations (“multiple”) versus single mutations (“single”).

DASM captures this pattern, showing similar distributions for both categories.

In contrast, AbLang and ESM assign radically lower scores to multinucleotide amino acid substitutions, consistent with the masked language modeling objective learning codon-level mutation probabilities as described in the main text (Figure 1a).”

This figure directly supports our claim: the experimental fitness data show similar distributions for single-mutation vs multiple-mutation amino acids, yet AbLang2 and ESM assign dramatically different scores to these groups, while DASM does not.

Additionally, the paper could benefit from additional benchmarking and comparison to enhanced versions of existing methods, such as AbLang plus a multi-hit correction.

It's an interesting idea to consider enhancing existing models. However, this approach faces some challenges. Most fundamentally, it is difficult to recast AbLang and other such models in an evolutionary framework: the masked language objective is simply not an evolutionary one. We have written a whole paper working to do this (https://doi.org/10.1371/journal.pcbi.1013758) and the results were middling despite our best efforts. Specifically regarding multihit, the effects of multihit are minor compared to the codon table effects, and those require the structure of codon-based evolutionary model.

Further descriptions of model components and validation metrics could help make the manuscript more readable.

We have clarified several aspects of the model in the revision: we now describe the Thrifty neutral model in the introduction, clarify the transformer architecture and wiggle activation function in the Methods, and explain the joint branch-length optimization procedure.

In the introduction we now describe Thrifty:

“This fixed model uses convolutions on 3-mer embeddings to deliver wide context sensitivity without needing a large number of parameters: the variant we use has around the same number of parameters as the classic S5F 5-mer model.”

In the Methods we clarify the architecture:

“We parameterize the DASM f using the standard transformer-encoder architecture: an amino-acid embedding, sinusoidal positional encodings, and PyTorch's TransformerEncoder module.

The only non-standard component to this architecture is a custom “wiggle” activation function to the output layer that prevents extreme selection factors as previously described.

This function asymptotes to zero for highly deleterious mutations and grows sub-linearly for beneficial ones.”

And the joint optimization:

“This joint optimization is performed cyclically, in which a complete cycle consists of neural network optimization followed by branch length optimization for every parent-child pair.

The parent sequence and the child sequence are pre-estimated, fixed, and used as training data.

The branch lengths are independent and so are optimized in parallel.”

Reviewer #2 (Public review):

Summary:

Endowing protein language models with the ability to predict the function of antibodies would open a world of translational possibilities. However, antibody language models have yet to achieve breakthrough success, which large language models have achieved for the understanding and generation of natural language. This paper elegantly demonstrates how training objectives imported from natural language applications lead antibody language models astray on function prediction tasks. Training models to predict masked amino acids teaches models to exploit biases of nucleotide-level mutational processes, rather than protein biophysics. Taking the underlying biology of antibody diversification and selection seriously allows for disentangling these processes through what the authors call deep amino acid selection models. These models extend previous work by the authors (Matsen MBE 2025) by providing predictions not only for the selection strength at individual sites, but also for individual amino acid substitutions. This represents a practically important advance.

Strengths:

The paper is based on a deep conceptual insight, the existence of a multitude of biological processes that affect antibody maturation trajectories. The figures and writing a very clear, which should help make the broader field aware of this important but sometimes overlooked insight. The paper adds to a growing literature proposing biology-informed tweaks for training protein language models, and should thus be of interest to a wide readership interested in the application of machine learning to protein sequence understanding and design.

Thank you for your kind words.

Weaknesses:

Proponents of the state-of-the-art protein language models might counter the claims of the paper by appealing to the ability of fine-tuning to deconvolve selection and mutation-related signatures in their high-dimensional representation spaces. Leaving the exercise of assessing this claim entirely to future work somewhat diminishes the heft of the (otherwise good!) argument.

This is an interesting idea! However, it seems to us that this approach has some fundamental limitations. Existing models operate on amino acid sequences with no nucleotide representation, so while they can be implicitly biased by the codon table, they have no signal to separate selection from effects related to the codon table and SHM rates.

We interpret this comment as proposing that we could use fine-tuning on functional data to pull out the selection components (that would only affect the functional data) versus the mutation component. That sounds like an interesting research project. We would be concerned that there are correlations between mutability and selective effects (e.g., CDRs are both more mutable and under different selection), creating identifiability problems unless separate data sources are used as we do here.

Additionally, the fine-tuning approaches we are aware of are taskspecific: they require labeled data from a specific assay (binding to antigen X, expression in system Y) that may or may not relate to the general evolutionary selection signal. Also, such approaches are limited to the specific data used and may not do a good job of guiding the model to a signal that is not present in the training data.

By structuring the model as we do, we obtain the evolutionary interpretation directly from phylogenetic signal without requiring taskspecific supervision.

In the context of predicting antibody binding affinity, the modeling strategy only allows prediction of mutations that improve affinity on average, but not those which improve binding to specific epitopes.

We agree, and this is fundamental to any general purpose model. Predictions of binding patterns for a specific target requires information about that target to be specified in the training data. We look forward to developing such task-specific models in the future.

We have added a paragraph to the Discussion clarifying this limitation:

“The current generation of DASM model does not use any antigen-labeled training data.

The signal that it leverages to infer some limited ability to predict binding comes from natural affinity maturation.

This affinity maturation comes through natural repertoires and so represents a mix of all of the antigens to which the sampled individuals have been exposed.”

Reviewer #3 (Public review):

Summary:

This work proposes DASM, a new transformer-based approach to learning the distribution of antibody sequences which outperforms current foundational models at the task of predicting mutation propensities under selected phenotypes, such as protein expression levels and target binding affinity. The key ingredient is the disentanglement, by construction, of selection-induced mutational effects and biases intrinsic to the somatic hypermutation process (which are embedded in > a pre-trained model).

Strengths:

The approach is benchmarked on a variety of available datasets and for two different phenotypes (expression and binding affinity). The biologically informed logic for model construction implemented is compelling, and the advantage, in terms of mutational effects prediction, is clearly demonstrated via comparisons to state-of-the-art models.

Thank you.

Weaknesses:

The gain in interpretability is only mentioned but not really elaborated upon or leveraged for gaining insight.

We are also excited about the ability of these models to provide interpretable predictions. We have dedicated an entire paper to this direction: “A Sitewise Model of Natural Selection on Individual Antibodies via a Transformer-Encoder" in MBE (https://doi.org/10.1093/molbev/msaf186). The interpretations offered by that paper overturn some of the oversimplified dogma about how natural selection works in antibodies (purifying in FWK and diversifying in CDR), giving a more nuanced sitewise perspective. The paper also highlights the importance of specific structural features of the antibodies.

This eLife paper, on the other hand, is focused on comparison to antibody language models and benchmarking zero-shot prediction on functional tasks.

We have better highlighted this new paper in our revision with:

“We have dedicated a companion paper to leveraging this interpretability to provide new perspectives on the operating rules of affinity maturation (Matsen et al., MBE 2025): that work provides a nuanced sitewise perspective on natural selection in antibodies that challenges classical oversimplified views of selection patterns.”

The following aspects could have been better documented: the hyperparametric search to establish the optimal model; the predictive performance of baseline approaches, to fully showcase the gain yielded by DASM.

We appreciate the concern and the desire to reveal all the factors that lead to a strong performance result. For this particular paper, we feel that this is less of a concern because we are optimizing according to an evolutionary objective function and then evaluating according to a functional one. We now describe how other than model size, hyperparameters stayed the same as in our previous paper (Matsen et al., MBE 2025).

Regarding baseline approaches, our previous paper includes comparisons to simpler models for the evolutionary objective. Here we focus on comparison to antibody language models for functional prediction. Comparing between state-of-the-art models is the standard practice for papers in this field.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

We recommend modest amounts of revision, discussed below:

Major comments:

(1) In the first section of the results, there is extensive discussion on shortcomings of existing antibody language models like AbLang2 that seems to associate all of the performance gap with the inability to separate non-synonymous mutations separated by 1 or 2+ substitutions.

In reality, some of the lower likelihoods in the 2+ substitution case could actually reflect real fitness deficits (while others could indeed be rarer occurrences in the training data). The authors should either moderate these claims or do an analysis that leverages antibody deep mutational scanning data to show that, conditioned on the fitness of the antibody (probably expression) being the same (either all high or all low), AbLang2 still artefactually considers rarer-training/less-codon-accessible variants to be less fit.

As described above, we believe that this is addressed by Figure S3, but if not please correct us.

(2) Some in the machine learning for antibody community might view the set of benchmarked datasets to be incomplete and somewhat arbitrarily selected, though we do think this is a good start, and the results are promising. A dataset commonly used in this field that is missing from this paper is from Shehata et al. (https://pubmed.ncbi.nlm.nih.gov/31553901/). A binding affinity experiment that is also commonly used in the field is from Phillips et al. (https://elifesciences.org/articles/71393) - this dataset measures combinatorial changes of framework regions on binding, which may be especially relevant here.

We're glad to have the opportunity to clarify this, thanks.

We based our evaluations on the April 2024 version of the FLAb benchmarking project (https://doi.org/10.1101/2024.01.13.575504) which preceded our work and thus was not subject to selection bias by us. We took the largest data sets in that repository. After this we became aware of the rich data sets offered by the Whitehead lab that provided binding measurements for many variants for a number of antigens, and added that to the evaluation set.

We have clarified this in the manuscript:

“We based our evaluations on the April 2024 version of the FLAb benchmarking project, which preceded our work and thus was not subject to selection bias by us.

We also benchmarked high-throughput binding data (more recent than FLAb) from the Whitehead lab that provided affinity measurements across many variants and antigens.”

The Shehata dataset is interesting but doesn't fit so much in the DASM mold: it is a survey of biophysical properties across many independent antibodies rather than a deep investigation of point mutants of a smaller collection of focal antibodies.

FLAb has grown to include the Phillips dataset. We are working full-tilt on the next version of DASM and will be including many other datasets in our paper on DASM2. Thanks for the tip!

(3) Similar to the above comment, we were also extremely curious as to why the authors did not test data from DeWitt et al. (https://pubmed.ncbi.nlm.nih.gov/40661619/). Instead, the authors only make a cryptic reference to this study on lines 201-6, but we could not even find a figure describing the results discussed on these lines. It would be great to actually include this data.

We agree, however, our model is for human rather than mouse. We would like to train a mouse model in the future but have not yet lined up the appropriate data.

(4) The authors should comment on potential data leakage if the SHM trajectories used in training have a similar sequence or antigen similarity to the benchmark expression/binding datasets.

This is a good question that we should clarify. Our model is trained only on evolutionary trajectories and not functional data. Evaluation is then done on functional data without fine-tuning. Because these evaluation data are categorically different from the training data and thus data leakage is not a problem. Recall that our model is zero-shot: it only considers evolutionary trajectories and not functional data as such. In a similar way, other self-supervised models such as MLMs do not exclude seeing an antibody in the training data when they are doing functional prediction.

We have clarified this in the manuscript with

“Because the DASM is trained exclusively on evolutionary trajectories rather than functional measurements, evaluation on expression and binding benchmarks is strictly zero-shot with no risk of data leakage.”

Relatedly, what happens if this approach is applied to completely de novo antibodies?

We direct this reviewer to the Shanehsazzadeh dataset that involves antibodies that were suggested by an AI algorithm rather than observed in nature.

If the reviewer is referring to completely synthetic antibody molecules, such as those generated by inverse folding, we have not attempted this.

(5) It makes sense that you included the multihit correction as a response to your earlier instantiation (without this correction) underestimating the probabilities of multiple mutations in a codon associated with a single amino acid substitution (lines 476-477).

However, this could potentially make for a somewhat unfair comparison to existing methods: if, say, we took AbLang (or another comparator) and also applied a multi-hit correction (even in some naive way at inference time), how would that compare to DASM? If this comparison favors DASM, it would show that models need more than just such a correction on top of existing methods to do good sequence scoring--which would only amplify the impact of the results.

Thank you for this suggestion. We believe that we have addressed it in the response to the public reviews, but please let us know if not.

Minor comments:

(1) It would be worth explicitly defining/summarizing the mutation model used in the study, e.g. giving an overview of Thrifty in the introduction or where it first appears.

Thanks, we have done this:

“Our approach separates mutation and selection processes by encoding functional effects in a Deep Amino acid Selection Model (DASM) while explicitly modeling mutation using a separate fixed model trained on neutrally evolving data.

This fixed model uses convolutions on 3-mer embeddings to deliver wide context sensitivity without needing a large number of parameters: the variant we use has around the same number of parameters as the classic S5F (Yaari et al., 2013) 5-mer model.”

(2) Paragraph starting on line 58: it sounds like you're suggesting that masked deep learning models will learn certain features of genomes in a certain order. We suggest that you weaken the language, giving examples of various things the model could learn, not implying that such models will necessarily learn the most useful features after the less useful ones.

We have fixed this by removing the "First... Second... Third... Finally" ordering:

“It could memorize the germline genes and learn about the probabilities of V(D)J recombination.

It could learn the codon table, as according to this table some aminoacid mutations are much more likely than others. It could learn rates of somatic hypermutation...

It could also learn about the impact of amino acid mutations on antibody function through natural selection in the course of affinity maturation, which is the desired signal.

However, this desired signal is confounded by the preceding factors.”

(3) Line 72: You make a strong claim that existing models conflate mutation and selection without knowing for sure that they didn't successfully learn these components separately (it seems this would require a lot of mechanistic interpretability). The language could be softened here.

We believe that we have addressed this in the response to public reviews, but please let us know if not.

(4) Line 79: Say a bit more about the separate fixed mutation model here. Why shouldn't we worry about this choice (especially the word "fixed") biasing your results? Does the empirical performance of your method suggest this doesn't really matter?

We have added to the description of the fixed mutation model, as described above.

As described in the public response, training SHM models on out-of-frame sequences is an established methodology for characterizing mutation in the absence of selection. In principle one could jointly train a model of SHM and selection, but one could have identifiability problems as there is a correlation between more mutable sites (e.g. in the CDRs) and those under relaxed selection. Using out-of-frame sequences gives a clean an independent description of the SHM process.

(5) Line 81: on what benchmarks does it outperform? State briefly.

Great suggestion. Done:

“The DASM, trained on substantially less data, outperforms AbLang2 and general protein language models including ESM2 and ProGen2-small. This outperformance holds on the largest benchmark datasets of the FLAb collection and on recent high-throughput binding assays.”

(6) Paragraph starting on line 90: The topic sentence reads a bit vague to us. Do you mean that you want to learn the extent to which models are regurgitating nucleotide similarity of AAs in determining the scores associated with AAs at masked sites?

Thank you. We have updated to

"We first sought to understand the extent to which processes such as neutral mutation rate and the codon table influence antibody language model prediction at masked sites."

(7) Paragraph starting on line 108: feels speculative and maybe better for the discussion...

We appreciate this comment, but we have decided to keep the content where it is. Although this would make sense as a Discussion item we feel like it fits well here right next to the evidence, and the structure of our Discussion doesn't really have a place for it.

(8) Paragraph starting on line 116: don't say "sequences from [12]" or "method of [15]." Explain what these are before giving the citation.

Whoops! Thanks. We have fixed these.

(9) Line 134: Consider giving a brief definition of perplexity?

Thanks. We added our favorite definition:

“Perplexity (as defined in the Methods) is the standard way of evaluating the plausibility of a sequence according to a model: it is the acrosssite geometric mean of the inverse probability of the observed amino acid.”

(10) Line 154: A citation here could be useful to support the claim that these models are learning phylogeny.

We have replaced with the more clearly established "codon table":

“We implemented a model to learn amino-acid preferences of antibodies without being influenced by germline genes, the codon table, or SHM biases.”

(11) Lines 161-162: Given that phylogenetic inference methods can be tough to scale, we're curious how you managed to get 2 million PCPs from the data? Did you construct a bunch of different phylogenies (in > parallel)?

Indeed! We now clarify in the methods section that these trees were run in parallel across clonal families:

“As in our previous work, tree inference and ancestral sequence reconstruction were performed per clonal family with the K80 substitution model...

Because these clonal families are independent these phylogenetic inferences were run in parallel.”

(12) Line 173-174: Can you say more about the joint optimization of the branch lengths? Are you conditioning on a phylogenetic tree topology only, and leaving the branch lengths unknown? Do you account for the fact that these branch lengths in the same phylogenetic tree aren't independent?

Thanks for pointing out the need to clarify these points. We have done so in the methods section and provided a pointer to the methods section in the main text.

In the main text we now say:

“We trained DASMs of several sizes (~1M, ~4M, ~7M) using joint optimization of branch length t and parameters of the DASM (see Methods for details).”

And in the Methods:

“This joint optimization is performed cyclically, in which a complete cycle consists of neural network optimization followed by branch length optimization for every parent-child pair.

The parent sequence and the child sequence are pre-estimated, fixed, and used as training data.

The branch lengths are independent and so are optimized in parallel.”

(13) Line 358: Yes, in a trivial sense, separating mutation and selection means that we know exactly how each of those two components has been learned. We would be curious if you could say anything about mechanistic interpretability within the deep learning selection model. If not, could this be a future research direction?

We believe that we have addressed this in the response to public reviews, but please let us know if not.

(14) Lines 384-386--indeed. Do you have any proposals for how a phylogeny could be constructed at this scale?

As above this is not one big phylogeny but many, which invites parallelization.

Reviewer #2 (Recommendations for the authors):

(1) I agree that a full study of fine-tuning strategies for all possible alternative models is beyond the scope of the paper. However, a little bit of fine-tuning would go a long way to demonstrate how easy (or hard) it is to extract the relevant signal from a general protein language model embedding.

As described in our response to the public reviews, we appreciate this point but have decided to focus on the core novelty of the paper and leave fine-tuning experiments to future work.

(2) The authors might want to add some discussion about what signals their models capture with regard to binding affinity (averages), and how this limitation might be addressed in future work.

As described in our response to the public reviews, we have added a paragraph to the Discussion clarifying this limitation.

Reviewer #3 (Recommendations for the authors):

(1) Introduction: I think more references have to be provided re: Antibody "foundation" language models, e.g. adding AntiBERTy and the two versions of AntiBERTa.

We have added citations to those two models, although we weren't sure what the second version of AntiBERTa was. There are very many antibody language models. If we could use number ranges we would cite a dozen or more, but I hesitate to add many of them in the eLife format, which has parenthetical citations. If there are others that you consider essential don't hesitate to suggest them.

(2) A key point of the approach is the disentanglement of “mutation” and “selection”, as mentioned in the introduction. However, the explanation of what the authors mean by mutation and selection comes only later. I would anticipate it in the introduction for clarity.

This is a great point. The revised intro has this in the second sentence:

“Natural antibodies are generated through V(D)J recombination, and refined by somatic hypermutation and affinity-based selection in germinal centers.”

and the "While the masked..." paragraph now more clearly calls out selection.

(3) Line 133: expression of what? Could the authors also explain mechanistically why expression should be impacted by a mutation? In what conditions do these data sample expression?

We have clarified that it is expression in a phage display library:

“To do so, we used the largest dataset of the FLAb collection of benchmarks, which measures the effect of single mutations on expression in a phage display library.”

(4) Line 142: Clarify that 0.49 and 0.3 are correlation coefficients. Also, what type of correlation coefficient is this?

Thanks for the catch! They are Pearson correlations as we now describe.

(5) Line 173: The hyperparametric search should have been more documented (with a description of how it was carried out and plots).

As described in our response to the public reviews, we are optimizing according to an evolutionary objective function and then evaluating according to a functional one. Other than model size, hyperparameters stayed the same as in our previous paper (Matsen et al., MBE 2025).

(6) Line 358: The authors say that 'DASMs provide direct interpretability'. However, this is not really inspected. A valuable addition would be to show how such interpretability is made possible, how it can recapitulate existing biological knowledge or provide hints for antibody engineering.

As described above, this is addressed in detail in our previous paper.

(7) Line 398: 'Inferred insertions or deletions were reversed, so that all sequences align to the naive sequence without gaps.' Could the authors comment on whether this is a limitation of the approach, why it wasn't dealt with and whether it could be the direction of future work?

Funny you should mention this! We have been planning out such an extension in detail recently. We have added a sentence in the discussion:

“We also have plans to extend the DASM framework to estimate the effect of natural selection on insertion and deletion events.”

(8) Line 430-431: Could the authors clarify 'shared' over what? Also, I believe these two lines really describe the DASM architecture. This should be spelt out more clearly and tied to the description provided in lines 173-175. A diagram of the architecture would be a valuable addition to provide a full picture of the model (this could be added to the general diagram of the modelling approach of Figure S8).

We have clarified in the text that this is indeed a description of the DASM architecture -- thanks for the catch:

“We parameterize the DASM f using the standard transformer-encoder architecture: an amino-acid embedding, sinusoidal positional encodings, and PyTorch's TransformerEncoder module.

The only non-standard component to this architecture is a custom “wiggle” activation function to the output layer that prevents extreme selection factors as previously described.”

The architecture is very “stock” - just the default torch TransformerEncoder, so I don't think that it merits a diagram. We have expanded our discussion of the simple architecture in the revision. This sits in contrast to the setup for the loss function, which is quite custom and is the subject of Figure 2 and Figure S8.

(9) Another general remark is that, to fully showcase the predictive advantage offered by DAMS with all the modelling choices entailed, one could show the performance of simpler models, like the mutation model alone (with no selection factors), or models where selection factors are just learnt independently for each site, or are learnt with a simple linear layer instead of a transformer (these are just ideas of some simpler approach that can set baselines over which DASM improvement can be shown).

This is a great suggestion. The primary focus of this paper is in comparing to alternate antibody language models in terms of functional prediction.

These simpler models could be used for comparing the evolutionary objective, which we did in our previous paper (https://doi.org/10.1093/molbev/msaf186). We note that a sitewise model with fixed sites cannot really be appropriately formulated due to sequences being of different lengths.

Additional changes

In addition to the reviewer-requested changes, we added a comparison of ESM2 model sizes (650M vs 3B parameters) on the Koenig benchmark. We found that scaling ESM2 from 650M to 3B parameters did not improve performance. Indeed, the larger model showed slightly degraded correlations, particularly for light chain predictions. This is consistent with recent observations that medium-sized protein language models can outperform larger ones on transfer learning tasks (Vieira et al., Sci. Rep. 2025). We added Table S2 documenting these results and cite this finding in the main text to justify our use of the 650M model throughout the analyses. After doing this, we realized for the Shanehsazzadeh evaluation we had accidentally used ESM2-3B instead of ESM2-650M. The corrected ESM2-650M values are slightly lower (0.191 and 0.308 for sequence lengths 119 and 120, respectively, compared to the previous values of 0.248 and 0.337). This correction does not affect our conclusions, as DASM substantially outperforms ESM2 on this benchmark before and after the change.

We also realized in the course of revision that we had been scoring AbLang2 using the masked-marginals pseudo-perplexity approach for the single-mutant Koenig dataset (Figure 1c), rather than the standard persequence pseudo-perplexity used elsewhere in the paper. For maskedmarginals, probabilities are computed using only wild-type context, whereas standard pseudo-perplexity uses each variant's own context.

The masked-marginals approach has a simple interpretation: for singlemutation variants, it is a linear transformation of the log ratio of the variant amino acid probability to the wild-type amino acid probability, both evaluated under wild-type context. This log-odds ratio directly measures how much the model prefers the mutation over the original residue.

We found that masked-marginals performed better for AbLang2 on this dataset, so we continued using it for Figure 1c. However, for the benchmarking table (Table 1), we switched to per-sequence pseudoperplexity as for the other comparisons in the paper, following the standard benchmarking protocol defined in FLAb (Chungyoun et al., 2024). We document both approaches in the Methods section:

“An alternative “masked-marginals” approach scores variants using only wild-type context.

For a wild-type sequence w, masked-marginals computes . for all amino acids a at each position i once, then uses these wild-type-derived probabilities to compute pseudoperplexity for any variant x...

For a single-mutation variant x that differs from wild-type w only at position j, all terms except position j cancel when comparing to wild-type, giving . Thus, the log-probability difference between variant and wild-type amino acids equals, up to an additive constant that depends only on the wild-type sequence, negative n times the log pseudo-perplexity of the variant.

For Figure 1c on the single-mutant Koenig dataset, we found that this approach gave a higher correlation for AbLang2 and so used it in that figure.

For benchmarking comparisons (Table 1), we followed standard practice and used per-sequence pseudo-perplexity.”

Author response:

Updated Response, March 3, 2026

In the midst of considering the thoughtful and insightful reviews of our manuscript and updating our work accordingly, we wanted to provide an interim update.

In the reviews of our paper, each of the reviewers brought up questions about the specificity and sensitivity of a new "TFD-Seq" assay for protein-DNA specificity in vivo that we had developed for this work and applied here for the first time with a complex eukaryote (Figure 4). While we remain strong proponents of developing in vivo assays for protein-DNA interaction, we took to heart the concerns that the reviewers had expressed. We have therefore, in the past few weeks, done a rather "deep dive" into both the technical aspects of the TFD-Seq data and the conceptual and statistical aspects of how TFD mutation data can be interpreted. From this analysis, we find ourselves in agreement with the concerns. In particular, our "deep dive" has suggested that conclusions from TFD data (particularly negative conclusions on the presence of binding sites) will require a better understanding of signal and noise in the kind assay used in Figure 4.

As the work is current in the submitted/preprint stage, we look forward to spending some time working (as appropriate) on both improvements to current protocols and alternative experiments to support the novel assay. An updated preprint which (for now) conveys the body of work and conclusions (which are not substantially altered), while avoiding the complexities of the TFD-seq assay is available at BioRXIV, and we will look forward to sending a version-of-record over the next few months as we have had a chance to provide robust tests for the macromolecular targets/interactors for ZNF-236 factor that was identified in this study.

We again thank the reviewers (peer review is indeed really a good thing) and look forward to updating everyone soon.

Updated bioRxiv preprint: https://www.biorxiv.org/content/10.1101/2025.10.22.683740v3

Original Response, January 5, 2026

We thank the reviewers for their insights and suggestions. We appreciate that the reviewers were engaged by both the observations and their interpretation, and consider their interest in further analysis and clarified discussion to be the best possible compliment to this work.

As noted by the reviewers, the working hypothesis of a nuclear organization role for ZNF-236 is just one model. Clarifying this model and potential alternatives will certainly add to the manuscript and this will be a key part of the revision.  Beyond this, several suggested analyses should explore extant models, while providing context for considering alternatives.  We look forward to carrying out such analyses as feasible and will report them in the revised manuscript.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

In this manuscript, Qin and colleagues aim to delineate a neural mechanism by which the internal satiety levels modulate the intake of sugar solution. They identified a three-step neuropeptidergic system that downregulates the sensitivity of sweet-sensing gustatory sensory neurons in sated flies. First, neurons that release a neuropeptide Hugin (which is an insect homolog of vertebrate Neuromedin U (NMU)) are in an active state when the concentration of glucose is high. This activation does not require synaptic inputs, suggesting that Hugin-releasing neurons sense hemolymph glucose levels directly. Next, the Hugin neuropeptides activate Allatostatin A (AstA)-releasing neurons via one of Hugin's receptors, PK2-R1. Finally, the released AstA neuropeptide suppresses sugar response in sugar-sensing Gr5a-expressing gustatory sensory neurons through AstA-R1 receptor. Suppression of sugar response in Gr5a-expressing neurons reduces the fly's sugar intake motivation (measured by proboscis extension reflex). They also found that NMU-expressing neurons in the ventromedial hypothalamus (VMH) of mice (which project to the rostral nucleus of the solitary tract (rNST)) are also activated by high concentrations of glucose, independent of synaptic transmission, and that injection of NMU reduces the glucose-induced activity in the downstream of NMU-expressing neurons in rNST. These data suggest that the function of Hugin neuropeptide in the fly is analogous to the function of NMU in the mouse.

Generally, their central conclusions are well-supported by multiple independent approaches. The parallel study in mice adds a unique comparative perspective that makes the paper interesting to a wide range of readers. It is easier said than done: the rigor of this study, which effectively combined pharmacological and genetic approaches to provide multiple lines of behavioral and physiological evidence, deserves recognition and praise.

A perceived weakness is that the behavioral effects of the manipulations of Hugin and AstA systems are modest compared to a dramatic shift of sugar solution-induced PER (the behavioral proxy of sugar sensitivity) induced by hunger, as presented in Figure 1B and E. It is true that the mutation of tyrosine hydroxylase (TH), which synthesizes dopamine, does not completely abolish the hunger-induced PER change, but the remaining effect is small. Moreover, the behavioral effect of the silencing of the Hugin/AstA system (Figure Supplement 13B, C) is difficult to interpret, leaving a possibility that this system may not be necessary for shifting PER in starved flies. These suggest that the Hugin-AstA system accounts for only a minor part of the behavioral adaptation induced by the decreased sugar levels. Their aim to "dissect out a complete neural pathway that directly senses internal energy state and modulates food-related behavioral output in the fly brain" is likely only partially achieved. While this outcome is not a shortcoming of a study per se, the depth of discussion on the mechanism of interactions between the Hugin/AstA system and the other previously characterized molecular circuit mechanisms mediating hunger-induced behavioral modulation is insufficient for readers to appreciate the novelty of this study and future challenges in the field.

We thank the reviewer for the thoughtful comment. We agree that the behavioral effects of manipulating the Hugin–AstA system alone were considerably weaker than the pronounced PER shifts induced by starvation. We have revised our Discussion to address it by positioning our findings within the broader context of energy regulation.

More specifically, we discuss that feeding behavior is controlled by two distinct, yet synergistic, types of mechanisms:

(1) Hunger-driven 'accelerators': as the reviewer notes, pathways involving dopamine and NPF are powerful drivers of sweet sensitivity. These systems are strongly activated by hunger to promote food-seeking and consumption.

(2) Satiety-driven 'brakes': our study identifies the counterpart to those systems above, aka. a satiety-driven 'brake'. The Hugin–AstA pathway acts as a direct sensor of high internal energy (glucose), which is specifically engaged during satiety to actively suppress sweet sensation and prevent overconsumption.

This framework explains the seemingly discrepancy in effect size. The dramatic PER shift seen upon starvation is a combined result of engaging the 'accelerators' (hunger pathways like TH/NPF) while simultaneously releasing the 'brake' (our Hugin–AstA pathway being inactive).

Our manipulations, which specifically target only the 'brake' system, are therefore expected to have a more modest effect than this combined physiological state. Thus, rather than being a "minor part," the Hugin–AstA pathway is a mechanistically defined, satiety-specific circuit that is essential for the precise "braking" required for energy homeostasis. We will update our Discussion to emphasize how these 'accelerator' and 'brake' circuits must work in concert to ensure precise energy regulation.

In this context, authors are encouraged to confront a limitation of the study due to the lack of subtype-level circuit characterization, despite their intriguing finding that only a subtype of Hugin- and AstA-releasing neurons are responsive to the elevated level of bath-applied glucose.

We thank the reviewer for highlighting the critical issue of subtype-level specialization within the Hugin and AstA populations.

We fully agree that the Hugin system is known for its functional heterogeneity (pleiotropy), with different Hugin neuron subclusters implicated in regulating a variety of behaviors, including feeding, aversion, and locomotion (e.g., Anna N King, Curr Biol, 2017, Andreas PLoS Biol, Sebastian et al., 2016, Nat Comm). Our finding that only a specific subcluster of Hugin neurons is responsive to glucose elevation provides a crucial first step in functionally dissecting this complexity.

we have added a dedicated paragraph to elaborate on this functional partitioning in the discussion. We propose that this subtype-level specialization allows the Hugin system to precisely link specific physiological states (like high circulating glucose) to appropriate behavioral outputs (like the suppression of sweet taste), demonstrating an elegant solution to coordinating multiple survival behaviors. Future work using high-resolution tools such as split-GAL4 and single-cell sequencing will be invaluable in fully mapping the specific functional roles corresponding to each Hugin and AstA subcluster.

Reviewer #2 (Public review):

Summary:

The question of how caloric and taste information interact and consolidate remains both active and highly relevant to human health and cognition. The authors of this work sought to understand how nutrient sensing of glucose modulates sweet sensation. They found that glucose intake activates hugin signaling to AstA neurons to suppress feeding, which contributes to our mechanistic understanding of nutrient sensation. They did this by leveraging the genetic tools of Drosophila to carry out nuanced experimental manipulations and confirmed the conservation of their main mechanism in a mammalian model. This work builds on previous studies examining sugar taste and caloric sensing, enhancing the resolution of our understanding.

Strengths:

Fully discovering neural circuits that connect body state with perception remains central to understanding homeostasis and behavior. This study expands our understanding of sugar sensing, providing mechanistic evidence for a hugin/AstA circuit that is responsive to sugar intake and suppresses feeding. In addition to effectively leveraging the genetic tools of Drosophila, this study further extends their findings into a mammalian model with the discovery that NMU neural signaling is also responsive to sugar intake.

Weaknesses:

The effect of Glut1 knockdown on PER in hugin neurons is modest, and does not show a clear difference between fed and starved flies as might be expected if this mechanism acts as a sensor of internal energy state. This could suggest that glucose intake through Glut1 may only be part of the mechanism.

We thank the reviewer for this insightful comment and agree that the modest behavioral effect of Glut1 knockdown is a critical finding that warrants further clarification. This observation strongly supports the idea that internal energy state is monitored by a sophisticated and robust network, not a single, fragile component. We believe the effect size is modest for two main reasons, which we have addressed in revised Discussion.

Firstly, the effect size is likely attenuated by technical and molecular redundancy. Specifically, the RNAi-mediated knockdown of Glut1 may be incomplete, leaving residual transporter function. Furthermore, Glut1 is likely only one part of the Hugin neuron's intrinsic sensing mechanism; other components, such as alternative glucose transporters or downstream K_ATP channel signaling, may provide molecular redundancy, meaning that the full energy-sensing function is not easily abolished by a single manipulation.

Secondly, and more importantly, the final feeding decision is an integrated output of competing circuits. While hunger-sensing pathways like the dopamine and NPF circuits act as powerful "accelerators" to drive sweet consumption, the Hugin–AstA pathway serves as a satiety-specific "brake." The modest effect of partially inhibiting just one component of this 'brake' system is the hallmark of a precisely regulated, multi-layered homeostatic system. We have clarified in the Discussion that the Hugin pathway represents one essential inhibitory circuit within this cooperative network that works together with the hunger-promoting systems to ensure precise control over energy intake.

Reviewer #3 (Public review):

Summary:

This study identifies a novel energy-sensing circuit in Drosophila and mice that directly regulates sweet taste perception. In flies, hugin+ neurons function as a glucose sensor, activated through Glut1 transport and ATP-sensitive potassium channels. Once activated, hugin neurons release hugin peptide, which stimulates downstream Allatostatin A (AstA)+ neurons via PK2-R1 receptors. AstA+ neurons then inhibit sweet-sensing Gr5a+ gustatory neurons through AstA peptide and its receptor AstA-R1, reducing sweet sensitivity after feeding. Disrupting this pathway enhances sweet taste and increases food intake, while activating the pathway suppresses feeding.

The mammalian homolog of neuromedin U (NMU) was shown to play an analogous role in mice. NMU knockout mice displayed heightened sweet preference, while NMU administration suppressed it. In addition, VMH NMU+ neurons directly sense glucose and project to rNST Calb2+ neurons, dampening sweet taste responses. The authors suggested a conserved hugin/NMU-AstA pathway that couples energy state to taste perception.

Strengths:

Interesting findings that extend from insects to mammals. Very comprehensive.

Weaknesses:

Coupling energy status to taste sensitivity is not a new story. Many pathways appear to be involved, and therefore, it raises a question as to how this hugin-AstA pathway is unique.

The reviewer is correct that several energy-sensing pathways are known. However, we now clarify that these previously established mechanisms, such as the dopaminergic and NPF pathways, primarily function as hunger-driven "accelerators." They are activated by low-energy states to promote sweet sensitivity and drive consumption.

The crucial, missing piece of the puzzle—which our study provides—is the satiety-specific "brake" mechanism. We identify the Hugin–AstA circuit as one of the “brakes”: a dedicated, central sensor that responds directly to high circulating glucose (satiety) to suppress sweet sensation and prevent overconsumption.

Thus, our work is unique because it defines the essential counterpart to the hunger pathways. In the revised Discussion, we have explained how these 'accelerator' (hunger) and 'brake' (satiety) systems work in concert to allow for the precise, bidirectional regulation of energy intake. Furthermore, by demonstrating that this Hugin/NMU 'brake' circuit is evolutionarily conserved in mice, our findings reveal a fundamental energy-sensing strategy and suggest that this pathway could represent a promising new therapeutic target for managing conditions of excessive food intake.

Recommendations for the authors:

Reviewing Editor Comments:

Considering the comments from all three reviewers, new experiments are not necessary, but the authors are welcome to provide new pieces of evidence that would strengthen their conclusions. To assist the authors with their revisions, the comments have been categorized from the highest to lowest priority based on the concerns raised by reviewers 1, 2, and 3.

High priority:

(1) Acknowledgement of partial phenotypes by the genetic manipulations, especially relative to other neuromodulators that are involved in the adjustment of sugar sensitivity after starvation (1, 2).

Please see our responses to the Public Review 1 for details.

(2) Detailed discussion on the novelty of the present work, also in light of previous studies both in flies and mammals (known Drosophila modulators, as well as NMU-rNST circuit on sugar sensation) (1, 2, 3).

Please see our responses to the Public Review 3 for details.

(3) Medium priority:

• Discussions on the subtype-specific function of hugin neurons (1).

Please see our responses to the Public Review 1 for details.

• Discussions on the pleiotropic effect of changes in the level of circulating sugar (including release of other sugar types) (2, 3).

We agree that circulating sugars represent a complex, systemic signal with broad, pleiotropic effects, and we have expanded our Discussion to address this.

We will discuss the functional distinction between key hemolymph sugars, such as trehalose (the main circulating sugar, critical for stress/flight) and glucose (the primary, rapidly mobilized energy currency). While various sugars collectively influence metabolic status, our study’s unique focus is on the direct neural link between internal energy and sweet taste modulation. We clarify that our work precisely identifies glucose as the direct, key ligand for the Hugin satiety circuit, thus providing a concrete, mechanistically defined link from systemic energy complexity to the specific regulation of sweet sensation.

• Illustration or clear explanations of sugar application methods in mouse experiments (ex. Figure 5F vs Figure 5M), as well as discussion on the concentration of sugar solutions used (3).

We have added the relevant details in the figure legends and explain the rationale for using this concentration of sugar in the results.

• Less saturated image for Figure 5K (3).

We have adjusted Figure 5K to reduce image saturation for clarity.

• Discussions on the modest effect of NMU on rNST neurons (Figure 5M) (3).

In the revised results, we have discussed that the modest suppression of rNST activity likely reflects partial peptide diffusion and the heterogeneous composition of sweet-responsive rNST neurons.

(4) Low priority:

• Systematic quantification of multiple types of sugars after starvation (3).

We agree that circulating sugars represent a complex metabolic milieu, and a fully systematic biochemical quantification of individual hemolymph sugars after starvation would be informative. While such analyses are beyond the scope of the present study, we have addressed this point at the functional level by systematically pre-feeding flies with different types of dietary sugars prior to PER assays.

We find that multiple sugars are capable of suppressing PER, indicating that satiety-related behavioral inhibition is not unique to a single carbohydrate source. Notably, sucrose produces the strongest suppression, consistent with its rapid metabolic conversion and effectiveness in elevating internal glucose levels. These results support the notion that diverse dietary sugars converge on a common satiety-signaling mechanism, while our mechanistic analyses specifically identify glucose as the key ligand engaging the Hugin satiety circuit.

We now clarify this distinction in the revised Discussion.

• Testing Gr64f neurons or mutants (3).

Our results indicate that energy sensing in the CNS suppresses sweet-sensing neuron activity (e.g., via hyperpolarization) rather than directly blocking sugar binding to receptors. Thus, sweet perception—not sugar detection—is inhibited. As evidence, in Figure supplementary4 we measured the PER to fructose and trehalose. Although Gr5a and Gr64a differ in their sensitivity to these sugars, the CNS energy state consistently suppresses sweet perception for both. As Reviewer 3 noted, Gr5a and Gr64f are co-expressed in sweet neurons; while they respond to different sugars, their labeling of the neurons is largely equivalent.

• Testing sugar preference (glucose vs. other sugars) (3)

Since our primary goal was to identify a direct satiety-sensing and sensory-modulating circuit—the "brake" mechanism—PER served as the most suitable and mechanistically specific readout. While manipulation of the Hugin–AstA circuit influences internal state, and therefore likely alters long-term sugar preference, investigating the integration of this pathway with reward and post-ingestive signaling is a critical question that lies beyond the scope of the current study.

• Cell type-specific knockout of NMU (3).

Achieving a cell type-specific knockout of NMU using the Cre approach is not feasible in the short term. While previous studies have reported the role of NMU in the VMH region in regulating feeding, our contribution lies in revealing how these neurons sense energy. We also show that these neurons project to the vicinity of Calb2 neurons and that the neuropeptide can suppress Calb2 neuronal activity. This essentially demonstrates that the hugin–Gr5a pathway in Drosophila is conserved in mice. We believe that a detailed dissection of the precise circuitry in mice is more appropriate to address in a subsequent study.

• Explanation of NMU detection in Figure 5K (3): this is GFP expressed by the Cre-dependent virus.

We have revised the Figure 5K legend to clarify that NMU⁺ neurons are labeled by GFP expression from a Cre-dependent AAV2/1-DIO-GFP, which undergoes anterograde trans-synaptic transfer. We further explain that GFP expression in rNST neurons requires local AAV-Cre injection, enabling identification of postsynaptic Calb2⁺ target neurons.

• Neuronal manipulation of NMU neurons by optogenetics or DREADD.

Please see our responses to the question “Cell type-specific knockout of NMU.”

Reviewer #1 (Recommendations for the authors):

A major concern about the study is that the effect of genetic manipulations on Hugin/AstA system appears to account for only a small part of the dramatic shift of PER probability toward smaller concentrations of sucrose solutions among starved flies. In Figure 1B and E, PER probability is significantly higher among starved flies in response to 10-200mM of sucrose solutions than fed flies. Compared to this, RNAi knockdown of glucose transporter in hugin neurons (Figure 2C), PK2-R1 pan-neuronally (Figure 3C) or in AstA-releasing neurons (Figure 3G), AstA-R1 in Gr5a neurons (Figure 4E), systemic mutation of PK-R2 (Figure Supplement 10) and AstA-R1 (Figure Supplement 12) all produce relatively minor behavioral changes. Consistent with previous works, the mutation of TH causes a robust decrease of PER across the entire range of sucrose concentration tested (Figure Supplement 1).

These discrepancies can be caused by many technical limitations that cannot be readily addressed. For instance, the large effect of TH can be confounded by the pleiotropic behavioral effect of the lack of dopamine. RNAi can suffer from incomplete elimination of targeted genes. However, the relatively small behavioral effect size of these manipulations cannot be entirely ignored in light of previous publications, which point to the importance of other neuromodulators such as dopamine, serotonin, Akh, and NPF, on sugar sensitivity (Marella et al., 2012; Inagaki et al., 2014; Yao et al., 2022), as well as other potentially parallel glucose-sensing systems, including Gr43a-expressing cells (Miyamoto et al., 2012) and sNPF-expressing CN neurons (Oh et al., 2019). While the neuropeptides initially tested (Figure 1) are not poor choices, it is a missed opportunity that so many other neuromodulators were excluded from the initial search.

We appreciate the reviewer’s detailed analysis and agree that the magnitude of behavioral effects produced by manipulating the hugin–AstA pathway is smaller than the dramatic shift in PER observed under starvation conditions. This comparison is important and highlights a central conceptual point of our study.

Starvation represents a compound physiological state that simultaneously engages multiple hunger-promoting neuromodulatory systems—most prominently dopaminergic and NPF pathways—while also releasing satiety-associated inhibitory signals. As shown previously and confirmed here (Figure supplementary 1), manipulation of dopamine synthesis produces a broad and robust reduction in PER across sucrose concentrations, consistent with its role as a powerful hunger-driven modulator.

By contrast, our genetic manipulations specifically target a satiety-associated inhibitory circuit—the hugin–AstA pathway—that is selectively engaged by high internal glucose levels. Manipulating this pathway alone therefore isolates a single “brake” component of feeding regulation, rather than recapitulating the full physiological state of starvation, which combines both accelerator activation and brake release. Accordingly, the more modest behavioral effects we observe are an expected consequence of dissecting one defined regulatory module from a larger, cooperative network.

We agree that multiple neuromodulators, including dopamine, serotonin, Akh, NPF, and others, as well as parallel glucose-sensing systems such as Gr43a-expressing cells and sNPF-expressing CN neurons, contribute to the regulation of sugar sensitivity. Rather than aiming to exhaustively screen all neuromodulators, our study was designed to identify and mechanistically define a central, glucose-responsive satiety sensor that directly links internal energy state to sweet taste modulation. In the revised discussion, we now explicitly position the hugin–AstA circuit as one essential, satiety-specific component within this broader regulatory landscape and discuss how it functionally complements previously characterized hunger-driven pathways.

I am also confused by the results of Shibirets1-mediated silencing of Hugin and AstA neurons (Figure Supplement 13B, C). It is unclear to me why a feeding assay was used instead of PER, like the activation experiments. Feeding (ingestion) and PER are qualitatively different types of behavior, which cannot be directly compared. Moreover, the definition of "fold change" is not provided either in the figure legend or in the Materials and Methods section, making it difficult to understand what the figure means.

We thank the reviewer for pointing out this important issue regarding the interpretation of the Shibire^ts1-mediated silencing experiments. We agree that proboscis extension reflex (PER) and feeding/ingestion assays reflect qualitatively different behavioral processes and should not be directly compared.

In the original submission, feeding assays were used to assess the effect of neuronal silencing, which led to ambiguity when comparing these results with PER-based activation experiments. To directly address this concern and ensure consistency across behavioral readouts, we have now performed additional PER experiments under the same Shibire^ts1-mediated silencing conditions.

These new data demonstrate that acute silencing of hugin neurons significantly enhances PER responses to sucrose (Figure supplementary 13B), indicating increased sweet sensitivity. This result is fully consistent with our activation experiments and supports the conclusion that the hugin–AstA pathway suppresses sweet taste perception under satiety conditions.

In addition, we have revised the figure legend to explicitly define the “fold change” metric used in the behavioral analysis, clarifying how the values were calculated and normalized. Together, these changes resolve the ambiguity raised by the reviewer and strengthen the behavioral consistency of our conclusions.

Of note, Marella et al. (2012) reported that silencing of Hugin-releasing neurons did not affect PER. It is therefore possible that the Hugin system is sufficient, but not necessary, for modulating PER under food deprivation.

We agree that their observation—that silencing Hugin-releasing neurons does not alter PER in starved flies—is consistent with a state-dependent role of the Hugin system in feeding regulation.

In starved animals, dopaminergic TH⁺ neurons are strongly activated and promote high PER responsiveness, while circulating glucose levels are low, placing Hugin neurons in a relatively inactive state. Under such conditions, further silencing of Hugin neurons would be expected to produce minimal additional effects on PER, which likely explains the results reported by Marella et al.

Importantly, our data show that preventing the starvation-associated reduction in Hugin neuronal activity—by thermogenetic activation of Hugin⁺ neurons (Hugin–TrpA1; Figure 1D)—significantly suppresses the hunger-induced enhancement of PER. These results indicate that dynamic downregulation of Hugin neuronal activity is a critical component of the normal behavioral shift in sweet sensitivity in response to food deprivation. Thus, while Hugin neurons may not be required to further modulate PER once animals are already in a strongly starved state, their regulated activity change is essential for mediating state-dependent modulation of sweet taste behavior. We have added discussion in the revised manuscript.

While no new experiments are requested, it is important for authors to acknowledge the limited effect size of Hugin/AstA manipulation. In the current manuscript, the authors briefly mention the previous works (lines 460-462, 472-474), which is insufficient. Discussions must include how the Hugin/AstA system may "complement these established mechanisms (line 460)" (described in the references listed above), under what situations this novel Hugin/AstA system can be relevant for controlling PER, and why the fly is equipped with seemingly redundant systems for sensing internal glucose levels and controlling feeding behavior. Without these discussions, it is difficult to recognize the novelty of the presented work. The data appears largely to be a minor and incremental progress on an already mature field.

In the revised manuscript, we have substantially expanded the Discussion to explicitly acknowledge this limited effect size and to clarify the functional role of the Hugin–AstA pathway within the broader energy-regulatory network. We now emphasize that this circuit represents a satiety-specific inhibitory branch that complements, rather than replaces, previously described hunger-promoting systems such as dopaminergic, NPF, and AKH circuits.

Importantly, we discuss the specific physiological conditions under which the Hugin–AstA system is most relevant—namely, post-feeding and high-glucose states. Unlike hunger circuits that amplify sweet sensitivity during starvation, the Hugin–AstA pathway directly senses circulating glucose and rapidly suppresses sweet taste perception when energy is sufficient, thereby acting as a brake to prevent overconsumption.

We further address the apparent redundancy among internal sugar-sensing systems. Rather than being redundant, these pathways form a coordinated and layered network with distinct sugar specificities, temporal dynamics, and functional roles. For example, Gr43a⁺ neurons primarily detect fructose, whereas hemolymph glucose represents the principal energetic currency in Drosophila. The use of multiple internal sugar sensors allows flies to fine-tune feeding decisions across different nutritional contexts and timescales.

Finally, we expand the Discussion to highlight that although the Hugin–AstA circuit constitutes only one branch of the energy-sensing network, its disruption leads to excessive energy intake (Figure supplementary 13C-E, G) and increased fat accumulation (Figure S13F), underscoring its physiological relevance. We also discuss how this pathway likely interacts with other neuromodulatory systems, including TH⁺ dopaminergic and NPF⁺ neurons, to collectively orchestrate adaptive feeding behavior and energy homeostasis.

Together, these additions clarify that our work does not simply add another neuromodulator to an already mature field, but instead identifies a distinct glucose-sensing, satiety-linked mechanism that fills a conceptual gap between internal energy state detection and sensory modulation.

Another perceived weakness is the lack of subtype-level dissection among Hugin- and AstA-releasing neurons. I make a justified request to narrow down the behaviorally relevant neuron to one (or one type), which is based on a widespread but unreasonable and dangerous assumption that every behavior must be controlled by one neuron. However, the authors present very interesting data that only a subset of Hugin- and AstA-releasing neurons responds to higher levels of sucrose (Figure 1H, Figure Supplement 7A, B), which leads to a hypothesis that a specific subtype within each peptidergic neuronal group is responsible for starvation-induced behavioral change. The authors only briefly touch upon this (lines 217-218), but this is an important hypothesis that requires further discussion.

We thank the reviewer for highlighting the importance of neuronal heterogeneity within the Hugin- and AstA-releasing populations. We fully agree that the observation that only a subset of Hugin⁺ and AstA⁺ neurons responds to elevated sucrose levels (Figure 1H; Figure Supplement 7A, B) strongly suggests functional specialization within these peptidergic groups.

In the revised Discussion, we now explicitly propose that distinct subtypes of Hugin and AstA neurons differentially contribute to energy sensing and feeding modulation. We suggest that glucose-responsive subpopulations may be specifically engaged in satiety signaling, whereas other neurons within the same genetic classes may participate in additional physiological or behavioral processes. This heterogeneity provides a plausible explanation for the partial behavioral effects observed following population-level manipulations. Although we did not perform subtype-specific perturbations in this study, our findings provide a foundation for identifying these subtypes in future work using split-GAL4 lines and connectomic datasets.

These issues are more important than the sprawling and unfocused review of various hunger and satiety-controlling systems across species in the Introduction. Lines 53-108 contain only tangential information to the main conclusion of the paper. Both the Introduction and Discussion sections must be completely restructured so that readers understand what is already known about hunger-induced changes in feeding-related behavior, what is a missing gap of knowledge in neural mechanisms controlling behavioral adaptation under starvation, and why Hugin/NMU is an interesting target in this context.

We thank the reviewer for this important structural critique. We agree that, in the original manuscript, the Introduction placed disproportionate emphasis on a broad survey of hunger- and satiety-regulating systems across species, which may have obscured the central conceptual advance of this study.

In the revised manuscript, we have substantially restructured both the Introduction and the Discussion to sharpen the narrative focus and clarify the specific knowledge gap addressed by our work.

First, the Introduction has been streamlined to focus on what is already known about hunger-induced modulation of feeding-related behaviors, particularly sweet taste sensitivity and PER in Drosophila. We now emphasize that prior studies have predominantly characterized hunger-activated, feeding-promoting pathways (e.g., dopaminergic, NPF, AKH systems) that act as accelerators of food-seeking behavior.

Second, we explicitly define the missing gap in knowledge: while hunger-driven mechanisms are well studied, it remains unclear how satiety states—specifically elevated internal glucose levels—are directly sensed by central neurons and translated into suppression of sensory gain and feeding behavior.

Third, we reposition Hugin/NMU as an attractive and conceptually distinct target because of its peptidergic nature, evolutionary conservation, and previously reported but mechanistically unresolved links to feeding regulation. This framing motivates our central question: whether Hugin/NMU neurons function as a direct internal energy sensor that actively implements a satiety-specific inhibitory control over taste perception.

In parallel, the Discussion has been reorganized to avoid an unfocused review of feeding circuits across species and instead to interpret our findings within a clear conceptual framework. We now emphasize that the Hugin–AstA (and NMU) pathway represents a satiety-driven “brake” that complements, rather than duplicates, established hunger-driven “accelerator” circuits. This restructuring clarifies both the novelty of our findings and their relevance within the existing literature.

Reviewer #2 (Recommendations for the authors):

When discussing the results of Figure 1, such as lines 203-204, "These results demonstrate that sugar intake inhibits sweet sensation, probably via increasing circulating sugar levels" it may be worth discussing the known impact of sweet sensation experience on future sweet taste responses. With the data shown here, it is difficult to conclusively separate blood glucose levels from the sweet sensation that happens during the re-feeding. The "normal diet minus sucrose" does not blunt the starved PER effect, but that could potentially be impacted by either/both sugar intake or sweet taste.

We thank the reviewer for this thoughtful and important point. We agree that sweet taste experience itself can influence subsequent sweet sensitivity, and that separating the contribution of sensory experience from nutrient-derived internal energy is non-trivial.

In the revised manuscript, we have clarified the experimental timing by explicitly stating that PER was assessed 15 minutes after refeeding. At this time point, hemolymph glucose levels have returned to baseline (Figure supplementary 5), supporting the physiological relevance of glucose-dependent activation of Hugin neurons under our experimental conditions.

We also acknowledge that sweet taste exposure can induce sensory adaptation and modulate future taste responses. To directly address this potential confound, we performed additional control experiments during revision (Figure supplementary 4B) in which starved flies were refed with sorbitol (caloric but not sweet) or arabinose (sweet but non-nutritive). We found that both manipulations partially reduced PER, but neither recapitulated the full suppressive effect of sucrose refeeding.

These results indicate that sweet taste experience and metabolic energy contribute in parallel to the regulation of sweet sensitivity. Importantly, the incomplete effects of sorbitol or arabinose alone suggest that neither sensory adaptation nor caloric value is sufficient by itself to fully account for the observed PER suppression.

Accordingly, we have revised the Discussion to clarify that the Hugin–AstA pathway likely operates within a broader, multi-layered regulatory framework, integrating internal metabolic state with sensory experience, rather than acting as a sole determinant of post-feeding sweet sensitivity. This clarification avoids over-attribution of the behavioral effect to circulating glucose alone while preserving the central conclusion that internal energy state is a key modulator of sweet perception.

Blocking cellular sugar intake or metabolism could be impacting the ability of neurons to function, distinct from any specific intracellular regulatory mechanism that glucose or its derivatives might be involved with. That may be a caveat worth mentioning in the results or discussion.

We thank the reviewer for raising this important caveat. We agree that blocking cellular sugar uptake or metabolism could, in principle, impair neuronal function in a nonspecific manner, independent of any dedicated intracellular glucose-sensing mechanism.

In the revised manuscript, we now explicitly acknowledge this possibility and clarify the scope of our interpretation. Several features of our data argue against a generalized loss of neuronal function as the primary explanation. First, the behavioral and physiological effects observed upon manipulation of glucose transport or K_ATP channel activity are rapid and reversible, consistent with state-dependent modulation rather than chronic metabolic failure. Second, these manipulations selectively affect sweet sensitivity and feeding-related behaviors, without causing gross deficits in proboscis extension or neuronal responsiveness.

Accordingly, we have revised the Results to emphasize that while intracellular glucose metabolism is required for normal neuronal activity, our findings specifically support a role for glucose-dependent modulation of neuronal excitability in satiety signaling, rather than a nonspecific energetic impairment.

Minor suggestions:

(1) Figure 2G: "Pryuvate" -> "Pyruvate."

We have corrected “Pryuvate” to “Pyruvate”

(2) "Fly" methods section: it says that flies were kept on 2% agar for 12 hours for starvation, but in the Figure 1A description, it says 24 hours.

We have corrected the description in Figure 1A.

Reviewer #3 (Recommendations for the authors):

(1) SEZ Hugin+ and AstA+ neurons were activated by glucose (Figures 1G, 1I), yet hemolymph also contains trehalose and fructose. For instance, DH44 neurons respond broadly to all hemolymph sugars (Dus et al., 2015), while Gr43a neurons specifically detect fructose (Miyamoto et al., 2012). The present study does not clarify whether Hugin+ or AstA+ neurons are similarly sugar-specific or more broadly tuned. A systematic analysis is needed to determine whether these circuits are selective for glucose.

We thank the reviewer for raising this important question regarding sugar specificity. We agree that hemolymph contains multiple sugars, including trehalose and fructose, and that distinct neural systems have been shown to differ in their tuning breadth. To address this issue, we performed additional experiments during revision in which starved wild-type flies were refed with different sugars—including sucrose, fructose, trehalose, and sorbitol—followed by PER measurements. We found that sucrose refeeding produced the strongest suppression of PER, whereas fructose, trehalose, and sorbitol induced weaker effects (Figuresupplementary 4A).

We interpret these results as suggesting a preferential sensitivity of the Hugin/AstA pathway to glucose availability rather than a broad responsiveness to all circulating sugars. One plausible explanation is that fructose, trehalose, and sorbitol require peripheral metabolic conversion before contributing to intracellular glucose levels in neurons, whereas sucrose feeding rapidly restores hemolymph glucose within the 15-minute time window used in our experiments (Figure supplementary 5).

Importantly, we now clarify in the revised Results and Discussion that our data support a functional preference for glucose under physiological conditions, rather than excluding the possibility that other sugars may influence this circuit indirectly or on longer timescales.

(2) The authors state that SEZ, but not VNC, Hugin+ neurons regulate AstA activity (lines 318-319). However, comparison of Figure Supplement 8B with the severing sample in Figure Supplement 11B shows a more pronounced reduction of sweet sensation under hug>TrpA1 activation. Although the absolute response in Figure 3F (in vivo) is higher than that in the cut-off preparation (Figure S11), comparison of Figure S11C with Figure 3F indicates that hug+ neurons drive an AstA+ calcium transient more than fourfold greater in the presence of VNC neurons. Thus, the contribution of Hugin+ VNC neurons cannot be dismissed, and the conclusion should be revised accordingly.

We thank the reviewer for this careful and quantitative comparison. We agree that our original wording overstated the exclusivity of SEZ Hugin⁺ neurons in regulating AstA activity.

Upon closer examination of the data, we now acknowledge that VNC Hugin⁺ neurons likely contribute to AstA activation. As the reviewer points out, the AstA⁺ calcium response evoked by Hugin activation is substantially larger when VNC neurons are intact (Figure supplementary11C) compared with the cut preparation (Figure 3F), indicating that descending inputs from the VNC can potentiate AstA neuronal activity.

Accordingly, we have revised the manuscript to state that SEZ Hugin⁺ neurons play a predominant role in driving AstA responses relevant to sweet sensation, while VNC Hugin⁺ neurons provide additional modulatory input that enhances the overall magnitude of Hugin signaling. These revisions have been made in the Results to more accurately reflect the contributions of distinct Hugin subpopulations.

(3) In Figure 4D, you show AstA-R1 co-localized with Gr5a-expressing cells. However, Gr5a-expressing cells also co-express Gr64f in labellum (Fuji et al., 2015, Current Biology). Are the authors sure that the sweet sensation they described is Gr5a-specific? Testing Gr64f is essential. Moreover, Fuji et al. demonstrated that Gr5a loss-of-function mutation impairs not only sucrose but also maltose, fructose, and trehalose sensation. This raises a question of whether the Hug+ and AstA+ neurons identified in the current study contribute to sensing sugars beyond sucrose. Additional experiments are required to clarify this point.

Please see our responses to the Reviewing Editor Comments (4).

(4) While nutritive sugar sensors such as Dh44 neurons have been directly implicated in sugar preference (Dus et al., 2015, Neuron), this study examines the hug+,AstA+, Gr5a neuronal circuit only in the context of PER responses. Why is sugar preference not assessed here, especially given that in mice, the comparison was made using preference tests?

We thank the reviewer for this insightful question. We agree that sugar preference assays provide important information about feeding decisions and reward-based behavior. In the present study, however, we deliberately focused on the proboscis extension reflex (PER) because it offers a direct, quantitative, and temporally precise readout of sweet sensory sensitivity at the sensory–motor level.

PER allows us to isolate changes in taste perception itself, largely independent of post-ingestive reinforcement, learning, or motivational state, all of which strongly influence preference-based assays. This distinction is particularly important given our central goal of identifying a circuit that directly links internal energy sensing to modulation of peripheral sweet-sensing neurons.

By contrast, sugar preference reflects an integrated behavioral outcome combining sensory input, internal state, and post-ingestive reward signals, including those mediated by DH44 neurons and other nutritive sensing pathways. We therefore chose PER as the most mechanistically specific assay to dissect the Hugin–AstA–Gr5a pathway. We now explicitly acknowledge in the revised Discussion that determining how this satiety-linked sensory modulation interacts with reward and post-ingestive circuits to shape long-term sugar preference will be an important direction for future studies.

Several other concerns:

(5) The intraperitoneal injection of NMU is interpreted as reflecting a brain-specific NMU effect, but such systemic delivery cannot exclude peripheral actions. In Figure 5D, the use of whole-body KO mice is insufficient; targeted manipulations (e.g., NMU-Cre-driven inactivation) are required to establish circuit-specific behavioral roles.

Please see our responses to the Reviewing Editor Comments (Low priority)

(6) In Figure 5F and 5M, neural activity is measured under different conditions: gastric glucose infusion in 5F versus glucose licking in 5M. To establish that NMU VMH neurons and Calb2 rNST neurons belong to the same circuit, this discrepancy in stimulation timing must be resolved to support the conclusions.

We thank the reviewer for pointing out this important issue regarding stimulation paradigms in Figures 5F and 5M. We agree that the difference between gastric glucose infusion and glucose licking requires explicit clarification.

In the revised manuscript, we now clearly state that these two paradigms were intentionally designed to probe complementary levels of the same NMU–Calb2 circuit. In Figure 5F, gastric glucose infusion was used to isolate the internal energy-sensing property of VMH NMU⁺ neurons, independent of oral sensory input, motor behavior, or reward expectation. This experiment establishes that NMU⁺ neurons are directly activated by elevated circulating glucose.

By contrast, Figures 5M examined how activation of this NMU pathway modulates downstream Calb2⁺ rNST neurons under physiologically relevant feeding conditions, in which sweet taste signals are naturally evoked by licking. This design allows us to test the functional consequence of NMU signaling on sweet-responsive rNST neurons during normal sensory processing.

Although the route and timing of glucose delivery differ, both paradigms converge on a unified circuit model: internal glucose elevation activates VMH NMU⁺ neurons, and NMU signaling suppresses sweet-driven activity in Calb2⁺ rNST neurons. We have revised the Results and figure legends to explicitly describe this layered experimental logic and to clarify that Figures 5F and 5M together establish distinct but connected nodes of the same circuit.

(7) Figure 5I-J. The glucose concentration used appears excessively high. In mammals, blood glucose in the sated state is ~7-8 mM. It is unclear whether the observed responses represent physiological effects or artifacts of supraphysiological stimulation. Additional experiments with lower glucose concentrations would strengthen the study.

We thank the reviewer for raising this important concern regarding the glucose concentration used in Figure 5I–J. We agree that the concentration applied in ex vivo slice experiments exceeds the typical physiological range of circulating glucose.

This higher concentration was intentionally chosen to ensure reliable neuronal activation in acute brain slices, where glucose diffusion, uptake, and metabolic access are substantially slower than in vivo. Similar approaches have been widely used in studies of glucose-sensitive hypothalamic neurons to overcome these technical limitations (e.g., Kim et al., 2025., Neuron).

Importantly, the physiological relevance of our findings is supported by in vivo fiber photometry experiments, which demonstrate that VMH NMU⁺ neurons are robustly activated following normal sugar ingestion under physiological conditions. Thus, while supraphysiological glucose was used to establish glucose responsiveness ex vivo, our in vivo data confirm that NMU⁺ neurons respond to glucose elevations within the normal physiological range.

(8) Figure 5K. The VMH images are inconsistently oriented compared with Figure 5E, lacking a 3v landmark. The NMU detection method (IHC or FISH) is not specified in the legend. The GFP-Calb2 signal is heavily saturated, making it difficult to distinguish true signals from artifacts. These issues undermine interpretability.

We thank the reviewer for pointing out these issues. In the revised manuscript, VMH images in Figure 5K have been reoriented to match Figure 5E, and the third ventricle (3v) is now indicated as an anatomical landmark. The figure legend has been revised to clarify that NMU⁺ neurons are identified by GFP expression from a Cre-dependent AAV2/1-DIO-GFP injected into NMU-Cre mice, rather than by NMU immunohistochemistry or FISH. In addition, GFP–Calb2 images have been reprocessed to clearly distinguish true signals from background and imaging artifacts.

(9) Figure 5L-M. Details of the NMU injection method are absent (route, dose, delivery parameters). The number of animals (n) is also not reported. Furthermore, AUC reduction alone is not sufficient evidence of robust inhibition. To convincingly demonstrate causality, NMU-IRES-Cre mice should be combined with DREADD or optogenetic approaches to directly inhibit NMU neurons and test whether rNST Calb2 activity is reduced.

We thank the reviewer for these helpful comments. We have revised the manuscript to include all missing methodological details. These details are now clearly described in the Methods section and figure legend.

We fully acknowledge that cell-type–specific manipulations, such as DREADD or optogenetic inhibition of NMU neurons, would provide more definitive causal evidence. However, our main goal in the mouse experiments was to demonstrate that NMU⁺ neurons can directly sense glucose and modulate sweet sensitivity, thereby supporting the evolutionary conservation of the Hugin mechanism identified in Drosophila. Detailed dissection of the downstream circuit architecture and behavioral consequences in mammals is indeed an important direction for future research, but it lies beyond the current study’s primary focus on cross-species conservation.

(10) In Drosophila, hugin neurons respond selectively to nutritive glucose (Fig. 2H), but whether NMU neurons share this property is unknown. Notably, Calb2 neurons in the rNST respond to the artificial sweetener AceK (Hao Jin et al., 2021, Cell), leaving open whether the NMU-rNST circuit is calorie-dependent or calorie-independent.

We have added a statement in the Discussion acknowledging this limitation and emphasizing that future work will be needed to test whether the NMU–Calb2 circuit is selectively engaged by metabolically active sugars or also by sweet taste signals independent of caloric value.

Minor comments

(11) All bar graphs should include individual data points.

We have added individual data points to all bar graphs.

(12) In Figures 3E, 4C, and 4D, it appears that a combination of GAL4 and LexA was used, but the information about the fly lines is missing.

We have now included the complete list of fly lines used for these experiments, including their genotypes and sources.

(13) The source for PK2-R1 KO, AstA-R1 KO fly lines and NMU-IRES-Cre, Calb2-IRES-Cre mice is missing.

We have added the complete source information for all genetic lines mentioned.

(14) Figure 5B-D, This is a sucrose preference test, so why is the y-axis labeled as glucose? Is this an error, or were the values converted to glucose equivalents?

We thank the reviewer for catching this mistake. The assay shown in Figure 5B–D measured sucrose preference, not glucose preference. The inconsistency resulted from a typographical error in the Methods description. In the revised manuscript, we have corrected this error to clearly state that sucrose was used in the preference test,

(15) Supplementary Figure 15. The NMU images are of poor quality and should be improved.

The punctate appearance of NMU signals in Supplementary Figure 15 is not due to poor image quality but rather reflects the physiological distribution of the NMU neuropeptide. As NMU is stored in secretory vesicles within neuronal terminals and somata, its immunostaining typically appears as discrete puncta rather than diffuse cytoplasmic labeling.

Editor's note:

Should you choose to revise your manuscript, if you have not already done so, please include full statistical reporting including exact p-values wherever possible alongside the summary statistics (test statistic and df) and, where appropriate, 95% confidence intervals. These should be reported for all key questions and not only when the p-value is less than 0.05 in the main manuscript.<br /> Readers would also benefit from noting that the mice were male and discussion of the exclusion of females.

In the revised manuscript, we have included full statistical reporting for all key experiments in the resource data. Regarding animal sex, we confirm that all mouse experiments were conducted using male mice. This choice was made to minimize variability caused by hormonal cycles in females, which can influence feeding behavior and glucose metabolism. We have now explicitly stated this information in the Methods section and included a brief discussion noting that sex-specific differences in NMU–Calb2 circuitry and feeding regulation represent an important question for future investigation.

Author response:

The following is the authors’ response to the previous reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

The paper uses rigorous methods to determine phase dynamics from human cortical stereotactic EEGs. It finds that the power of the phase is higher at the lowest spatial phase. The application to data illustrates the solidity of the method and their potential for discovery.

Comments on revised submission:

The authors have provided responses to the previous recommendations.

We thank the reviewer for reviewing our manuscript again, and for their positive evaluation.

Reviewer #3 (Public review):

Summary:

The authors propose a method for estimating the spatial power spectrum of cortical activity from irregularly sampled data and apply it to iEEG data from human patients during a delayed free recall task. The main findings are that the spatial spectra of cortical activity peak at low spatial frequencies and decrease with increasing spatial frequency. This is observed over a broad range of temporal frequencies (2-100 Hz).

Strenghs:

A strength of the study is the type of data that is used. As pointed out by the authors, spatial spectra of cortical activity are difficult to estimate from non-invasive measurements (EEG and MEG) and from commonly used intracranial measurements (i.e. electrocorticography or Utah arrays) due to their limited spatial extent. In contrast, iEEG measurements are easier to interpret than EEG/MEG measurements and typically have larger spatial coverage than Utah arrays. However, iEEG is irregularly sampled within the three-dimensional brain volume and this poses a methodological problem that the proposed method aims to address.

Weaknesses:

Although the proposed method is evaluated in several indirect ways, a direct evaluation is lacking. This would entail simulating cortical current source density (CSD) with known spatial spectrum and using a realistic iEEG volume-conductor model to generate iEEG signals.

Comments on revised version:

In my original review, I raised the following issue:

"The proposed method of estimating wavelength from irregularly sampled three-dimensional iEEG data involves several steps (phase-extraction, singular value-decomposition, triangle definition, dimension reduction, etc.) and it is not at all clear that the concatenation of all these steps actually yields accurate estimates. Did the authors use more realistic simulations of cortical activity (i.e. on the convoluted cortical sheet) to verify that the method indeed yields accurate estimates of phase spectra?"

And the authors' response was:

"We now included detailed surrogate testing, in which varying combinations of sEEG phase data and veridical surrogate wavelengths are added together. See our reply from the public reviewer comments. We assess that real neurophysiological data (here, sEEG plus surrogate and MEG manipulated in various ways) is a more accurate way to address these issues. In our experience, large scale TWs appear spontaneously in realistic cortical simulations, and we now cite the relevant papers in the manuscript (line 53)."

The point that I wanted to make is not that traveling waves appear in computational models of cortical activity, as the authors seem to think. My point was that the only direct way to evaluate the proposed method for estimating spatial spectra is to use simulated cortical activity with known spatial spectrum. In particular, with "realistic simulations" I refer to the iEEG volume-conductor model that describes the mapping from cortical current source density (CSD) to iEEG signals, and that incorporates the reference electrodes and the particular montage used.

Although in the revised manuscript the authors have provided indirect evidence for the soundness of the proposed estimation method, the lack of a direct evaluation using realistic simulations with ground truth as described above makes that remain sceptical about the soundness of the method.

We thank the reviewer for reviewing our manuscript again.

We have reviewed the literature again on volume conduction effects in LFP measures of cortical activity. In all publications we reviewed, the conclusion is that the range of the effect is <1cm. We now mention the range of volume conduction in the Methods section dealing with the surrogate models (lines 1054-9) as well as added emphasis in the Discussion (lines 594-9).

The highest spatial frequency we consider in the present research is 50c/m, which corresponds to a cortical distance of 2cm. This is well outside the range of volume conduction effects in LFPs. Mathematically speaking, blurring (e.g. Gaussian) acts as a low-pass filter, attenuating higher spatial frequency components. But only for components within the spatial range of the Gaussian blurring i.e. for LFPs, higher than 100c/m. There will therefore be negligible effects (mathematically speaking, zero effect) of volume conduction in the results reported by us. If the veracity of these studies on volume conduction with LFPs is accepted, then the reviewer’s requested simulation reduces to “estimating spatial spectra [using] simulated cortical activity with known spatial spectrum.” This is what we have done, in a direct and simple manner.

If the ubiquity and importance of spatio-temporal dynamics in cortex is accepted, then it is insufficient to describe “the mapping from cortical current source density (CSD) to iEEG signals”, since this presumes a model of cortical activity that does not capture the correlations in space and time that we assume are critical to cortical function. We are aware the CSD approach has a long and successful history of unravelling brain mechanisms. However, an emphasis on traveling waves (and spatio-temporal dynamics in general) is in part a challenge to this approach (and the idea of localized sources in general). CSD approaches carry similar assumptions (but at a smaller scale, <1cm) as those elaborated in Zhigalov and Jensen (2023) for extra-cranial measures. In both cases, removal of volume conduction effects emphasizes standing wave activity (localized static, oscillatory sources) over traveling wave activity. In this manner, these methods tend to confirm their starting assumptions (as does our own approach, of course). What is required is external empirical validation to break any circular confirmation of initial theoretical choice of basis. All this is a way of saying that CSD approaches are not the unproblematic, direct methods that the reviewer asserts.

We did understand the reviewer’s request to model the effects of volume conduction. Our own view of realistic cortical simulations differs from the reviewer’s, setting aside the final step in the forward modeling pipeline which would add the effects of volume conduction in the grey matter. By simulating real-time dynamics, it should be possible to untangle the effects of volume conduction from true spatio-temporal correlations. This is because the volume conduction effects are essentially instantaneous, compared to the relatively slow motion of traveling waves. So, the measurement of purely spatial phase vectors is prone to smearing artefact, but following the trajectory of a wave over one cycle can more accurately determine the range of true interactions. One could, for example, compare the usual CSD forward modelling with TWs in simulations, see which is the best predictor of future activity, and compare these to empirical measurements. Here, the CSD analysis would remove the volume conduction effects but also emphasize standing activity over motion, even where the motion was veridical in the simulation.

Even so, these tests are only relevant in <1cm range.

Another issue is ephaptic coupling, which we mention in the discussion. This means that some of the local volume conduction effects are not merely artefacts from the point of view of cortical function, but have a real causal effect. The strength of the word ‘some’ has yet to be completely resolved in the literature, and it would be technically challenging to include these effects in any simulation.

Finally, simulation should be an adjunct to empirical studies, or used when empirical studies are not possible. We do not think, in this case, they are the ‘only direct’ way to evaluate our method. We, rather, rely on the converging evidence from empirical studies of volume conduction in LFPs which show this effect is outside the range of our reported results.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

The authors present an approach that uses the transformer architecture to model epistasis in deep mutational scanning datasets. This is an original and very interesting idea. Applying the approach to 10 datasets, they quantify the contribution of higher-order epistasis, showing that it varies quite extensively.

Suggestions:

(1) The approach taken is very interesting, but it is not particularly well placed in the context of recent related work. MAVE-NN, LANTERN, and MoCHI are all approaches that different labs have developed for inferring and fitting global epistasis functions to DMS datasets. MoCHI can also be used to infer multidimensional global epistasis (for example, folding and binding energies) and also pairwise (and higher order) specific interaction terms (see 10.1186/s13059-024-03444-y and 10.1371/journal.pcbi.1012132). It doesn't distract from the current work to better introduce these recent approaches in the introduction. A comparison of the different capabilities of the methods may also be helpful. It may also be interesting to compare the contributions to variance of 1st, 2nd, and higher-order interaction terms estimated by the Epistatic transformer and MoCHI.

We thank the reviewer for the very thoughtful suggestion.

Although these methods are conceptually related to our method, none of them can be realistically used to perform the type of inference we have done in the paper on most the datasets we used, as they all require explicitly enumerating the large number of interaction terms.

We have included new text (Line 65-74) in the introduction to discuss the advantages and disadvantages of these models. We believe this has made our contribution better placed in the broader context of the field.

(2) https://doi.org/10.1371/journal.pcbi.1004771 is another useful reference that relates different metrics of epistasis, including the useful distinction between biochemical/background-relative and backgroundaveraged epistasis.

We have included this very relevant reference in the introduction. We also pointed out the limitation of these class of methods is that they typically require near combinatorically complete datasets and often have to rely on regularized regression to infer the parameters, making the inferred model parameters disconnected from their theoretical expectations. Line 49-56.

(3) Which higher-order interactions are more important? Are there any mechanistic/structural insights?

We thank the reviewer for pointing out this potential improvement. We have now included a detailed analysis of the GRB2-SH3 abundance landscape in the final section of the results. In particular, we estimated the contribution of individual amino acid sites to different orders (pairwise, 3-4th order, 4-8th order) of epistasis and discuss our finding in the context of the 3D structure of this domain. We also analyzed the sparsity of specific interactions among subsets of sites.

Please see Results section “Architecture of specific epistasis for GRB2-SH3 abundance.”

Reviewer #2 (Public review):

Summary:

This paper presents a novel transformer-based neural network model, termed the epistatic transformer, designed to isolate and quantify higher-order epistasis in protein sequence-function relationships. By modifying the multi-head attention architecture, the authors claim they can precisely control the order of specific epistatic interactions captured by the model. The approach is applied to both simulated data and ten diverse experimental deep mutational scanning (DMS) datasets, including full-length proteins. The authors argue that higher-order epistasis, although often modest in global contribution, plays critical roles in extrapolation and capturing distant genotypic effects, especially in multi-peak fitness landscapes.

Strengths:

(1) The study tackles a long-standing question in molecular evolution and protein engineering: "how significant are epistatic interactions beyond pairwise effects?" The question is relevant given the growing availability of large-scale DMS datasets and increasing reliance on machine learning in protein design.

(2) The manuscript includes both simulation and real-data experiments, as well as extrapolation tasks (e.g., predicting distant genotypes, cross-ortholog transfer). These well-rounded evaluations demonstrate robustness and applicability.

(3) The code is made available for reproducibility.

We thank the reviewer for the positive feedback.

Weaknesses:

(1) The paper mainly compares its transformer models to additive models and occasionally to linear pairwise interaction models. However, other strong baselines exist. For example, the authors should compare baseline methods such as "DANGO: Predicting higher-order genetic interactions." There are many works related to pairwise interaction detection, such as: "Detecting statistical interactions from neural network weights", "shapiq: Shapley interactions for machine learning", and "Error-controlled nonadditive interaction discovery in machine learning models."

We thank the reviewer for this very helpful comment. These references are indeed conceptually quite similar to our framework. Although they are not directly applicable to the types of analyses we performed in this paper (partitioning contribution of epistasis into different interaction orders in terms of variance components), we have included a discussion of these methods in the introduction (Line 70-74). We believe this helps better situate our method within the broader conceptual context of interpreting machine learning models for epistatic interactions.

(2) While the transformer architecture is cleverly adapted, the claim that it allows for "explicit control" and "interpretability" over interaction order may be overstated. Although the 2^M scaling with MHA layers is shown empirically, the actual biological interactions captured by the attention mechanism remain opaque. A deeper analysis of learned attention maps or embedding similarities (e.g., visualizations, site-specific interaction clusters) could substantiate claims about interpretability.

Again, we thank the reviewer for the thoughtful comment. We have addressed this comment together with a related comment by Reviewer1 by including a detailed analysis of the GRB2-SH3 landscape using a marginal epistasis framework, where we quantified the contribution of individual sites to different orders of epistasis as well as the sparsity of epistatic interactions. We also present these results in the context of the structure of this protein. Please see Results section “Architecture of specific epistasis for GRB2-SH3 abundance.”

(3) The distinction between nonspecific (global) and specific epistasis is central to the modeling framework, yet it remains conceptually underdeveloped. While a sigmoid function is used to model global effects, it's unclear to what extent this functional form suffices. The authors should justify this choice more rigorously or at least acknowledge its limitations and potential implications.

We agree that the under parameterization of the simple sigmoid function could be be potentially confounding. We did compare different choices of functional forms for modeling global epistasis. Overall, we found that there is no difference between a simple sigmoid function with four trainable parameters and the more complex version (sum of multiple sigmoid functions, used by popular methods such as MAVENN). Therefore, all results we presented in the paper were based on the model with a single scalable sigmoid function.

We have added relevant text; line 153-158. We have also included side-by-side comparisons of the model performance for the GRB-abundance and the AAV2 dataset to corroborate this claim (Supplemental Figure 1).

(4) The manuscript refers to "pairwise", "3-4-way", and ">4-way" interactions without always clearly defining the boundaries of these groupings or how exactly the order is inferred from transformer layer depth. This can be confusing to readers unfamiliar with the architecture or with statistical definitions of interaction order. The authors should clarify terminology consistently. Including a visual mapping or table linking a number of layers to the maximum modeled interaction order could be helpful.

We thank the reviewer for the thoughtful suggestion. We have rewritten the description of our metrics for measuring the importance of "pairwise", "3-4-way", and ">4-way" interactions; Line 232-239.

We have also added a table to improve clarity, as suggested; Table 2.

Reviewer #3 (Public review):

Summary:

Sethi and Zou present a new neural network to study the importance of epistatic interactions in pairs and groups of amino acids to the function of proteins. Their new model is validated on a small simulated data set and then applied to 10 empirical data sets. Results show that epistatic interactions in groups of amino acids can be important to predict the function of a protein, especially for sequences that are not very similar to the training data.

Strengths:

The manuscript relies on a novel neural network architecture that makes it easy to study specifically the contribution of interactions between 2, 3, 4, or more amino acids. The study of 10 different protein families shows that there is variation among protein families.

Weaknesses:

The manuscript is good overall, but could have gone a bit deeper by comparing the new architecture to standard transformers, and by investigating whether differences between protein families explain some of the differences in the importance of interactions between amino acids. Finally, the GitHub repository needs some more information to be usable.

We thank the reviewer for the thoughtful comments. We have listed our response below in the “Recommendations for the authors” section.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Some of the dataset labels are confusing. For example, GRB is actually the protein GRB2 and more specifically just one of the two SH3 domains from GRB2 (called GRB2-SH3 in Faure et al.).

We thank the reviewer for catching this. Our original naming of the datasets followed the designation of library number in the Faure et al paper (which constructed 3 variant libraries and performed different assays on them). To avoid confusion (and also save space in the figure titles), we have now renamed the datasets using this mapping:

Author response table 1.

Reviewer #3 (Recommendations for the authors):

(1) What is the cost of the interpretability of the model? It would be interesting to evaluate how a standard transformer, complete with its many non-linearities, performs on the simulated 13-position data, using the r2 metric. This is important as the last sentence of the discussion seems to suggest that the model proposed by the authors could be used in other contexts, where perhaps interpretability would be less important.

We thank the reviewer for this suggestion. We have run a generic transformer model on the GRBabundance and AAV2 datasets. Overall, we found minimal difference between the generic model and our interpretable model, suggesting that fitting the interpretable transformer does not incur significant cost in performance.

We have included a side-by-side comparison of the performance of the generic transformer and our three-layer model in Supplemental Figure 5 and a discussion of this finding in Line 256-259.

(2) The 10 data sets analyzed by the authors differ in their behaviour. I was wondering whether the proteins have different characteristics, beyond the number and distribution of mutants in the data sets. For instance, do high-order interactions play a bigger role in longer proteins, in proteins with more secondary structures, in more hydrophobic proteins?

We fully agree that this is a highly relevant question. Unfortunately, the paucity of datasets suitable for the type of analyses we performed in the paper limit our ability to draw general conclusions. Furthermore, the differences in genotype distribution among the 10 datasets may be the main driving factor in the behaviors of the models.

We included our thoughts on this issue in the discussion (Line 477-481).

We will definitely revisit this question if this type of high-order combinatorial DMS data becomes more available in the (hopefully) near future.

(3) Although the code appears to be available in the repository, there is no information about the content of the different folders, about what the different scripts do, or about how to reproduce the article's results. More work should be done to clarify it all.

Thank you for pointing this out. We have substantially improved our github repository and included many annotations for reproducibility.

(4) Typos and minor comments:

(a) p3 "a multi-peak fitness landscapes": landscape.

(b) p3 "Here instead of directly fitting the the regression coefficients in Eq. 2": remove 'the'.

(c) p3 "neural network architectures do not allow us to control the highest order of specific epistasis": a word is missing.

(d) p6 "up to 1,926, 3,014, and 4,102 parameters, respectively-all smaller than the size of the training dataset": it's not very clear what size of the dataset means: number of example sequences?

(e) p6 "This results confirm": This result confirms.

(f) p6 "to the convergence of of the variance components of the model landscape to the ground truth.": remove 'of'.

(g) p7 "to characterize the importance higher-order interactions": the importance of.

(h) p7 "The improvement varies across datasets and range": and ranges.

(i) p9 "over the pairwise model is due to the its ability": remove 'the'.

(j) p13 "This results suggest that pairwise": result suggests.

(k) p13 "although the role assessed by prediction for randomly sampled genotypes seems moderate": sampled. Also, I'm not sure I understand this part of the sentence: what results are used to support this claim? It's not 6b, which is only based on the mutational model.

This is in Supplemental Figure 7.

(l) p13 "potentially by modeling how the these local effects": remove the.

(m) p13 "We first note that the the higher-order models": remove the.

(n) p15 "M layers of MHA leads to a models that strictly": lead to a model.

(o) Supp Figure 1: "Solid lines shows the inverse": show.

(p) Supp p 10 "on 90% of randomly sample data": sampled.

(q) Supp p11 "Next, assume that Eq. 5 is true for m > 0. We need to show that Eq. 5 is also true for m + 1.": shouldn't it be m>=0 ? It seems important to start the recursive argument.

Good catch.

(r) Supp p11 "Since the sum in line 9 run through subsets": runs.

(s) Supp p11 "we can further simplify Eq. 11 it to": remove it.

We have fixed all these problems. We very much appreciate the reviewer’s attention.

Author response:

eLife Assessment

This study uses the yeast two-hybrid assay to identify proteins that may interact with yeast Set1 and other subunits of COMPASS/Set1C, the histone H3K4 methyltransferase, providing also some evidence for Set1 sumoylation and a role of SET1C methylating other factors in vitro. The results are valuable, and they should contribute to understanding the functions of the conserved SET1C complex, as they suggest potential functional connections with RNA biogenesis, chromatin remodeling, and non-histone methylation, whose implications would yet need to be explored. Nevertheless, apart from the fact that only a small subset of the Y2H interactions is further examined, the validating experiments are only partial or inconclusive, the strength of evidence being at this point incomplete.

We thank the reviewers for their thoughtful comments, which primarily raise three major concerns: the overinterpretation of the Y2H data, issues related to validation, and the manuscript’s structure. At the same time, the reviewers acknowledge that the dataset is extensive and that aspects of the validation work are valuable. Below, we provide point-by-point responses to the public reviews. We will prepare a revised version of the manuscript that carefully addresses the public comments and incorporates the referees’ recommendations.

Public Reviews:

Reviewer #1 (Public review):

The manuscript by Luciano et al is a collection of experiments about the yeast histone 3 lysine 4 methyltransferase, Set1, starting with 10 yeast two-hybrid screens (Y2H). Y2H screens were briefly popular 20+ years ago, but the persistently unfavourable false-to-true positive ratios limited their utility, and the conclusion emerged that Y2H is an unreliable approach for gathering protein-protein interaction data. Y2H outcomes are candidate interaction lists at best, strongly contaminated by false positives. Here, the authors employed a company (Hybridomics) to perform the Y2H screens.

The primary data is not presented, and the outcomes are summarized using the Hybridomics in-house quality scoring system in Figure 1A. It is not possible to evaluate these data, and the manuscript presents cartoon summaries that the reader must accept as valuable.

We agree that false positives contaminate the list of potential interactors. Some interactions may also be indirect through a common interactor and do not reflect a physiological interaction. Nevertheless, some positives reflect real interactions that can occur under specific physiological conditions. This is the case, for example, with the interaction between Spp1 and Mer2 (from this screen), which has led to major discoveries (Acquaviva et al. Science 2013; Sommermeyer et al. Mol Cell 2013). The publication of these 10 screens should be viewed as a valuable resource for the broader community.

Hybrigenics brings extensive experience from conducting numerous screens, enabling the team to recognize recurring false positives that commonly arise in screening assays.

(1) Based on the extensive knowledge about Set1C/COMPASS acquired from genetics and biochemistry by many labs (including the Geli lab), the results presented here from the 10 Y2H screens are notably patchy. Of the 7 subunits of this complex, only one (Spp1) was identified using Set1 as bait. Conversely, as baits, Swd2, Spp1, Shg1, captured Set1, and the Bre2-Sdc1 interaction was reciprocally identified. These interactions were scored at the highest confidence level, which lends some confidence to the screens. However, the missing interactions, even at the third confidence level, indicate that any Y2H conclusions using these data must be qualified with caution. The authors do not appear to be cautious in their lengthy evaluations of these candidate interactions, which are illustrated with cartoons in Figures 2 and 3, with some support from the literature but almost without additional evidence. Snf2 is a particularly interesting candidate, which the authors support with pull-down experiments after mixing the two proteins in vitro (Figure 4). After Y2H, this is the least convincing evidence for a protein-protein interaction, and no further, more reliable evidence is supplied.

We agree with referee 1 that more caution is needed, and we will take this into account in the revised version. We agree that Y2H interaction is an indication of potential interaction and not proof of interaction. We have therefore made a significant effort to compile elements from the literature that may support the interaction. Once again, this study can be considered a resource.

(2) Figure 5 continues the cartoon summary of extrapolations from the Y2H screens, again without supporting evidence, except that the authors state, "We have refined the interaction region between Set1, Prp8 and Prp22, showing that Prp8 and Prp22 interact strongly with Set1-F4 (n-SET). Prp22 interacts in addition with Set1-F1 (Figure S2)." However, Figure S2 does not show this evidence and is incoherent.

When we say that we have refined the interaction region between Set1, Prp8, and Prp22, we mean that we have restricted the interaction regions according to Y2H criteria. Indeed, we have not shown the spots illustrating the results. This will be corrected in the revised version.

The figure legends for Figure S2B and C (copied here in bold) do not correspond to the figure.

We agree that the legend for Figure S2 is unclear and does not accurately describe the panels shown in the figure. We will revise the legend accordingly in the updated version to ensure it accurately reflects the content of all panels.

(B) Expression of the F1-F5 fragments in yeast cells. Fusion proteins were detected with an anti-GAL4 monoclonal antibody. TOTO yeast cells (Hybrigenics) were transformed with the different pB66-Set1-F1 to F5 plasmids and subsequently with either P6, pP6-Snf2 762-968, pP6-Prp8 37-250, or pP6-Prp22 379-763 that were identified in the Y2H screens. Transformed cells were incubated 3 days at 30{degree sign}C on SD-LEU-TRP and then restreaked on SD-LEU-TRP-HIS with 3AT. Cell growth was monitored after 2 days at 30{degree sign}C.

(C) Solid and dotted arrows indicate that transformed TOTO cells transformed with pB66-Set1-F1 to F5 and the indicated prey (Snf2, Prp8, and Prp22) are growing in the presence of 20 mM and 5 mM of AT, respectively.

Figure S2D is two almost featureless dark grey panels accompanied by the figure legend D) Control experiment showing that TOTO cells transformed with p6 and pB66-Set1-F4 are not gowing (sic) in the presence of 5 mM or 20 mM AT.

Line 343. Interestingly, the two-hybrid screens reveal that Set1 1-754 interacted with Gag capsid-like proteins of Ty1 (Figure S5), raising the possibility that Set1 binding to Ty1 mRNA is linked to the interaction of Set1 1-754 with Gag.

This is another example of the primary mistake repeatedly made by the authors -Y2H interactions are candidate results and not conclusive evidence.

This statement is supported by our previous findings demonstrating that Set1 binds Ty1 mRNA independently of it dRRM and represses Ty1 mobility at a post-transcriptional stage (Luciano et al., Cell Discovery, 2017 PMID:29071121). Binding of Set1 to Ty1 mRNA could stem from the interaction between Set1 1-754 and the Gag capsid-like protein.

To further illustrate this point, the authors highlight the candidate interaction between Nis1 and 3 Set1C subunits.

While we agree that the Nis1-Set1C interaction has not been demonstrated beyond doubt, we feel that our Y2H and in vitro binding experiments provide reasonable evidence that the interactions may be relevant. It is important to consider that any interaction assay can provide negative (and false positive) results, this includes Y2H, in vitro binding and mass-spec analysis of purified complexes from cells. We feel that it is not appropriate to only trust protein interactions that are strong and stable enough to be demonstrated via purified complexes. It is clear that some protein interactions do occur in transient and weak manner and therefore are not compatible with biochemical purification approach. This indeed is the strength of alternative methods like Y2H and in vitro binding assays, that interactions can be identified and tested even if the physiological context of the interaction may be more complex.

(3) After multiple speculations based on the Y2H candidates, the authors changed to focus on sumoylation of Set1, which has previously reported to be sumoylated. Evidence identifying two sumoylation sites in Set1, in the N-SET and SET domains, is valuable and adds important progress to the role of sumoylation in the regulation of H3K4 methyltransferase, relevant for all eukaryotes. This illuminating part of the manuscript is only tenuously connected to the preceding Y2H screens and concomitant speculations.

We thank Referee 1 for their comment. While it is true that there is only a modest connection between Set1 interactors involved in direct or indirect sumoylation and the characterization of Set1 SUMOylation sites, we believe that this does not constitute a weakness of the manuscript.

(4) The manuscript then describes a red herring exercise involving Set1 methylation of Nrm1. In an already speculative and difficult manuscript, it is exasperating to read a paragraph about a failed idea. Apart from panel E, Figure 7 is a distraction, and I believe it should not be shared.

According to this comment, we will remove Fig. 7 panels A-D.

(5) However, despite the failure with Nrm1, Line 443 - The H3K4-like domain in Nrm1 raised our attention to other yeast proteins that carry such sequences.

This line of thinking is even less connected to the Y2H screens than the sumoylation work.

However, the authors present a reasonable evaluation of the yeast proteome screened for six amino acids similar to the known H3K4 motif ARTKQT (Figure 7e).

(6) However, this evaluation goes nowhere and has no connection with the next section of the manuscript, which is entirely speculation about the regulation of metabolism and stress responses based on the Y2H results and selected evidence from the literature.

We will take into account of these remarks (points 5 and 6) in the revised version.

(7) The manuscript then describes more failed experiments regarding lysine methylation of Snf2 by Set1C, which unexpectedly reports arginine methylation rather than lysine. The manuscript does not currently meet the standard expected for this type of paper - the composition is somewhat incoherent and there are no previous reports of arginine methylation by SET domain proteins.

We respectfully disagree with referee 1. We have integrated extensive in vitro reconstruction experiments with complementary in vivo studies, all conducted according to the rigorous standards expected by leading journals. These approaches have allowed us to reach the conclusions presented in this manuscript. While some of these findings are unexpected, they are supported by the data. We have carefully discussed the results and their limitations to provide a comprehensive interpretation.

The manuscript presents a very experienced grasp of the literature and a sophisticated appreciation of the forefront issues, but a surprising failure to eliminate uninformative failures and peripheral distractions. The overinterpretation of Y2H results is a dominating failure. There are some valuable parts within this manuscript, and hopefully, the authors can reformat to eliminate the defects and appropriately qualify the candidate data.

We thank Referee 1 for these insightful comments. In the revised version, we will follow the advice to remove non-informative failures and peripheral distractions. Additionally, we will exercise greater caution to avoid overinterpreting the Y2H results.

Reviewer #2 (Public review):

Summary:

This paper starts with a large-scale yeast two-hybrid (Y2H) screen using Set1 (full-length and smaller parts) and other Set1C/COMPASS subunits as bait. There are hundreds of possible interactions identified, but only a small number are given any follow-up. While it's useful to document all the possible interactions, the unfocused and preliminary nature of the results makes the paper feel scattered and incomplete.

Strengths:

The Y2H screen was very comprehensive, producing lots of interesting possible leads for further experiments.

Weaknesses:

The results are useful but incomplete because only a small subset of the Y2H interactions is further examined. Even in the case of those that were further tested, the validating experiments are only partial or inconclusive.

Referee 2’s comments align in some respects with those of Referee 1. We will follow the detailed Referee 2 suggestions to reduce the scattered nature of the manuscript.

We will follow his/her recommendations, in particular we will provide and AlphaFold model of the interaction between the Set1 N-term 1-754 with the SID domain of Kap104 that involves the proposed Set1 PY-NLS sequence.

Reviewer #3 (Public review):

The SET1C/COMPASS complex is the histone H3K4 methyltransferase in Saccharomyces cerevisiae, where it plays pivotal roles in transcriptional regulation, DNA repair, and chromatin dynamics. While its canonical function in histone methylation is well-established, its full interactome remains poorly defined. Moreover, whether SET1C methylates non-histone substrates has been an open question. In this study, Luciano et al. employ systematic yeast two-hybrid (Y2H) screening to uncover novel interactors and functions of SET1C. Their findings reveal potential functional connections to RNA biogenesis, chromatin remodeling, and non-histone methylation.

The authors performed multiple Y2H screens using Set1 (full-length, N-terminal, and C-terminal fragments) and each of its seven subunits as baits. They identified high-confidence interactors that link SET1C to diverse cellular processes, including chromatin regulation (e.g., the SWI/SNF complex via Snf2), DNA replication (e.g., Mcm2, Orc6), RNA biogenesis (e.g., spliceosome components Prp8 and Prp22; polyadenylation factors Pta1 and Ref2), tRNA processing (e.g., Trm1, Trm732), and nuclear import/export (e.g., importins Kap104 and Kap123). Some of these interactions were further validated by immunoprecipitation or in vitro assays.

Given the interaction of Set1 with Slx5 and Wss1 - proteins involved in SUMO-dependent processes - the authors investigated and convincingly demonstrated that Set1 is sumoylated. This modification may influence the function and regulation of the SET1C complex.

Finally, the authors provide evidence that SET1C methylates proteins beyond histone H3K4, notably Nrm1, a transcriptional corepressor, and Snf2, the catalytic subunit of the SWI/SNF chromatin remodeling complex. Although Nrm1 contains a domain resembling the H3K4-methylated sequence (H3K4-like domain), this region does not appear to be required for its methylation. The search for other proteins containing similar domains as potential methylation candidates (p.12, first paragraph) seems less justified, given the lack of evidence supporting the requirement for the H3K4-like domain in methylation.

This study offers valuable insights into the interactome of SET1C, suggesting potential links between the complex and a wide range of cellular processes. However, the functional implications of the Y2H interactions remain to be explored further. Additionally, the study provides intriguing information on the possible regulation of Set1 by sumoylation. The discovery of Nrm1 and Snf2 as methylation substrates could significantly expand the known targets and functions of SET1C.

The results are supported by high-quality data.

We thank referee 3 for his/her positive comments

Author response:

We sincerely appreciate the constructive comments and valuable suggestions from the editors sand reviewers. We highly value the feedback and will carefully address all concerns in our revised manuscript.

(1) We will supplement more details of the processing steps and key results in the analyses of sCCA and SVR to improve the transparency and reproducibility of our methods.

(2) According to the reviewers’ suggestions, we will adjust and present a more conventional and cautious conclusion regarding clinical specificity and neuroplasticity reserve.

(3) We will supplement the results of structural connections (termed “symptom-related network” in the manuscript) across the three subgroups to strengthen the interpretation of subgroup-specific neurobiological characteristics.

(4) All the suggestions from the reviews will be respected, and we will carefully revise our manuscript to improve its clarity, rigor, and scientific quality.

We believe these revisions will significantly improve the quality of our work.

Author response:

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

We thank the reviewers for their thoughtful comments and constructive suggestions. We describe how we have addressed each point below and are grateful for the guidance on areas where our work could be clarified or expanded. In particular, we note the following:

Selection scan summary statistics: In our revised manuscript, we have included summary statistics from the selection scans. We believe this addition will enhance transparency and provide additional context for readers.

Reporting of outliers: As highlighted by the editor, the reviewers expressed differing views on the most appropriate way to report outliers. To provide a comprehensive and balanced presentation, we now report both the empirical selection statistics and the corresponding converted p-values in either the main text or supplement, and both outputs are also provided in the full summary files. This dual approach will allow readers to fully interpret the results under both perspectives.

Expanded discussion of admixture timing and population structure: We have carefully considered the reviewers' suggestions to incorporate additional descriptions of population structure or demographic analyses, and have done so in our revisions where possible. These changes strengthen the rigor and clarity of the analyses.

Public Reviews:

Reviewer #1 (Public review):

Summary:

The paper reports an analysis of whole-genome sequence data from 40 Faroese. The authors investigate aspects of demographic history and natural selection in this population. The key findings are that the Faroese (as expected) have a small population size and are broadly of Northwest European ancestry. Accordingly, selection signatures are largely shared with other Northwest European populations, although the authors identify signals that may be specific to the Faroes. Finally, they identify a few predicted deleterious coding variants that may be enriched in the Faroes.

Strengths:

The data are appropriately quality-controlled and appear to be of high quality. Some aspects of the Faroese population history are characterized, in particular, by the relatively (compared to other European populations) high proportion of long runs of homozygosity, which may be relevant for disease mapping of recessive variants. The selection analysis is presented reasonably, although as the authors point out, many aspects, for example differences in iHS, can reflect differences in demographic history or population-specific drift and thus can't reliably be interpreted in terms of differences in the strength of selection.

Weaknesses:

The main limitations of the paper are as follows:

(1) The data are not available. I appreciate that (even de-identified) genotype data cannot be shared; however, that does substantially reduce the value of the paper. Minimally, I think the authors should share summary statistics for the selection scans, in line with the standard of the field.

We agree with the reviewer that sharing the selection scan results is important, so we have now made the selection scan summary statistics publicly available, and clearly lay out the guidelines and research questions for which the data can be accessed in our Data Availability statement.

(2) The insight into the population history of the Faroes is limited, relative to what is already known (i.e., they were settled around 1200 years ago, by people with a mixture of Scandinavian and British ancestry, have a small effective population size, and any admixture since then comes from substantially similar populations). It's obvious, for example, that the Faroese population has a smaller bottleneck than, say, GBR.

More sophisticated analyses (for example, ARG-based methods, or IBD or rare variant sharing) would be able to reveal more detailed and fine-scale information about the history of the populations that is not already known. PCA, ADMIXTURE, and HaplotNet analysis are broad summaries, but the interesting questions here would be more specific to the Faroes, for example, what are the proportions of Scandinavian vs Celtic ancestry? What is the date and extent of sex bias (as suggested by the uniparental data) in this admixture? I think that it is a bit of a missed opportunity not to address these questions.

We clarify that we did quantify the proportions of various ancestry components as estimated by HaploNet in main text Figure 5 and supplemental figures S6 and S7. To better highlight this result, we now also include the average global ancestry of the various components in the Main Text - Results - Fine-Scale Structure and Connections to Ancient Genomes.

We agree that more fine-scale demographic analyses would be informative. We now additionally provide an estimation of the admixture date in the Main Text - Results - Fine-Scale Structure and Connections to Ancient Genomes and discussion using the DATES software which is optimized for ancient genomes.

We have encountered problems with using different standard date estimation software, including DATES, which give very inconsistent and unstable results. As we note in our text, we suspect this might be due to the strong bottleneck experienced in the history of the Faroe Islands, low LD differentiation between the source populations, or multiple pulses of admixture, which may be breaking one or more of the assumptions of these methods. Assessing the limitations of these methods is beyond the scope of this current manuscript; however, we will continue working on this problem for future studies, possibly using simulations to assess where the problem might be. We recognize that our relatively small sample size places limits on the fine-scale demographic analyses that can be performed. We are addressing this in ongoing work by generating a larger cohort, which we hope will enable more detailed inference in the future.

(3) I don't really understand the rationale for looking at HLA-B allele frequencies. The authors write that "ankylosing spondylitis (AS) may be at a higher prevalence in the Faroe Islands (unpublished data), however, this has not been confirmed by follow-up epidemiological studies". So there's no evidence (certainly no published evidence) that AS is more prevalent, and hence nothing to explain with the HLA allele frequencies?

We agree that no published studies have confirmed a higher prevalence of ankylosing spondylitis (AS) in the Faroe Islands. Our recruitment data suggest that AS might be more common than in other European populations, but we understand that this is only based on limited, unpublished observations and what we are hearing from the community. We emphasized in our original manuscript that this is based on observational evidence from the FarGen project. However, as this reviewer pointed out, we can be more clear that this prevalence has not been formally studied.

In revision, we clarify in the Main Text - Results - HLA-B Allele Frequencies and Discussion that our recruitment data suggest a higher prevalence of AS may be possible, but more formal epidemiological studies are needed to confirm this observation. The reason we study HLA-B allele frequencies is to see if the genetic background of the Faroese population could help explain this possible difference, since HLA-B27 is already known to play a strong role in AS.

Reviewer #2 (Public review):

In this paper, Hamid et al present 40 genomes from the Faroe Islands. They use these data (a pilot study for an anticipated larger-scale sequencing effort) to discuss the population genetic diversity and history of the sample, and the Faroes population. I think this is an overall solid paper; it is overall well-polished and well-written. It is somewhat descriptive (as might be expected for an explorative pilot study), but does make good use of the data.

The data processing and annotation follows a state-of-the-art protocol, and at least I could not find any evidence in the results that would pinpoint towards bioinformatic issues having substantially biased some of the results, and at least preliminary results lead to the identification of some candidate disease alleles, showing that small, isolated cohorts can be an efficient way to find populations with locally common, but globally rare disease alleles.

I also enjoyed the population structure analysis in the context of ancient samples, which gives some context to the genetic ancestry of Faroese, although it would have been nice if that could have been quantified, and it is unfortunate that the sampling scheme effectively precludes within-Faroes analyses.

We note that although the ancestry proportions were not originally specified in the main text, we did quantify ancestry proportions in the modern Faroese individuals and other ancient samples, and we visualized these proportions in Figure 5 and Supplementary Figures S6 and S7. As stated in our response to Reviewer #1, in our revisions, we now more clearly state the average global ancestry of the various components in the Main Text - Results - Fine-Scale Structure and Connections to Ancient Genomes.

I am unfortunately quite critical of the selection analysis, both on a statistical level and, more importantly, I do not believe it measures what the authors think it does.

Major comments:

(1) Admixture timing/genomic scaling/localization:

As the authors lay out, the Faroes were likely colonized in the last 1,000-1,500 years, i.e., 40-60 generations ago. That means most genomic processes that have happened on the Faroese should have signatures that are on the order of ~1-2cM, whereas more local patterns likely indicate genetic history predating the colonization of the islands. Yet, the paper seems to be oblivious to this (to me) fascinating and somewhat unique premise. Maybe this thought is wrong, but I think the authors miss a chance here to explain why the reader should care beyond the fact that the small populations might have high-frequency risk alleles and the Faroes are intrinsically interesting, but more importantly, it also makes me think it leads to some misinterpretations in the selection analysis.

See response to point #3

(2) ROH:

Would the sampling scheme impact ROH? How would it deal with individuals with known parental coancestry? As an example of what I mean by my previous comment, 1MB is short enough in that I would expect most/many 1MB ROH-tracts to come from pedigree loops predating the colonization of the Faroes. (i.e, I am actually quite surprised that there isn't much more long ROH, which makes me wonder if that would be impacted by the sampling scheme).

The sampling scheme was designed to choose 40 Faroese individuals that were representative of the different regions and were minimally related. There were no pairs of third-degree relatives or closer (pi-hat > 0.125) in either the Faroese cohort or the reference populations. It is possible that this sampling scheme would reduce the amount of longer ROHs in the population, but we should still be able to see overall patterns of ROH reflective of bottlenecks in the past tens of generations. Additionally, based on this reviewer's earlier comment, 1 Mb ROHs would still be relevant to demographic events in the last 40-60 generations given that on average 1 cM corresponds to 1 Mb in humans, though we recognize that is not an exact conversion.

That said, the “sum total amount of the genome contained in long ROH” as we described in the manuscript includes all ROHs greater than 1Mb. Although we group all ROHs longer than 1Mb into one category in Main Text Figure 2, we now additionally provide the distribution in ROH lengths across all individuals for each cohort in a new Supplemental Figure S3. As this plot shows, there certainly are ROHs longer than 1Mb in the Faroese cohort, and on average there is a higher proportion of long ROH particularly in the 5-15 Mb range in the Faroese cohort relative to the other cohorts. As the reviewer points out, these longer ROHs are possibly indicative of a more recent or stronger bottleneck in the Faroes relative to the comparison cohorts. We highlight this result in Main Test - Results - Population Structure and Relatedness.

(3) Selection scan:

We are talking about a bottlenecked population that is recently admixed (Faroese), compared to a population (GBR) putatively more closely related to one of its sources. My guess would be that selection in such a scenario would be possibly very hard to detect, and even then, selection signals might not differentiate selection in Faroese vs. GBR, but rather selection/allele frequency differences between different source populations. I think it would be good to spell out why XP-EHH/iHS measures selection at the correct time scale, and how/if these statistics are expected to behave differently in an admixed population.

The reviewer brings up good points about the utility of classical selection statistics in populations that are admixed or bottlenecked, and whether the timescale at which these statistics detect selection is relevant for understanding the selective history of the Faroese population. We break down these concerns separately.

(1) Bottlenecks: Recent bottlenecks result in higher LD within a population. However, demographic events such as bottlenecks affect global genomic patterns while positive selection is expected to affect local genomic patterns. For this reason, iHS and XP-EHH statistics are standardized against the genome-wide background, to account for population-specific demographic history.

(2) Admixture: The term “admixture” has different interpretations depending on the line of inquiry and the populations being studied. Across various time and geographic scales, all human populations are admixed to some degree, as gene flow between groups is a common fixture throughout our history. For example, even the modern British population has “admixed” ancestry from North / West European sources as well, dating to at least as recently as the Medieval & Viking periods (Gretzinger et al. 2022, Leslie et al. 2015), yet we do not commonly consider it an “admixed” population, and we are not typically concerned about applying haplotype-based statistics in this population. This is due to the low divergence between the source populations. In the case of the Faroe Islands, we believe admixture likely occurred on a similar timescale or even earlier, based on the DATES estimates. We see low variance in ancestry proportions estimated by HaploNet, both from the historical Faroese individuals (dated to 260 years BP) and the modern samples. This indicates admixture predating the settlement of the Faroe Islands, where recombination has had time to break up long ancestry tracts and the global ancestry proportions have reached an equilibrium. That is, these ancestry patterns suggest that the modern Faroese are most likely descended from already admixed founders. In the original manuscript, we mentioned this as a likely possibility in the Main Text - Discussion: “This could have occurred either via a mixture of the original “West Europe” ancestry with individuals of predominantly “North Europe” ancestry, or a by replacement with individuals that were already of mixed ancestry at the time of arrival in the islands (the latter are not uncommon in Viking Age mainland Europe).” In our revisions, we further included the DATES estimations of the timing of admixture in the modern and historical Faroese samples, which pre-date the timing of settlement in both cases. We highlight these points in the Discussion. And, as with the case of the British population, the closely-related ancestral sources for the Faroese founders were likely not so diverged as to have differences in allele frequencies and long-range haplotypes that would disrupt signals of selection from iHS or XP-EHH.

(3) Time scale: It is certainly possible, and in fact likely, that iHS measures selection older than the settlement of the Faroe Islands. In our manuscript, we calculated iHS in both the Faroese and the closely related British cohort, and we highlight in the main Main Text that the top signals, with the exception of LCT, are shared between the two cohorts, indicative of selection that began prior to the population split (Discussion and Results - Signals of Positive Selection). iHS is a commonly calculated statistic, and it is often calculated in a single population without comparing to others, so we feel it is important to show our result demonstrating these shared selection signals. In our revisions, we now clarify in the Discussion the limitations and time-scale at which the iHS statistic may detect selection. As far as XP-EHH, it is a statistic designed to identify differentiated variants that are fixed or approaching fixation in one population but not others. The time-scale of selection that XP-EHH can detect would therefore be dependent on the populations used for comparison. As XP-EHH has the best power to identify alleles that are fixed or approaching fixation in one population but not others, it is less likely to detect older selection events / incomplete sweeps from the source populations. We highlight this point in the Discussion.

(4) Similarly, for the discussion of LCT, I am not convinced that the haplotypes depicted here are on the right scale to reflect processes happening on the Faroes. Given the admixture/population history, it at the very least should be discussed in the context of whether the 13910 allele frequency on the Faroes is at odds with what would be expected based on the admixture sources.

We agree that more investigation into the LCT allele frequency in the other ancient samples may provide some insight into the selection history, particularly in light of ancient admixture. Please note, we did look at the allele frequency of the LCT allele rs4988235 and stated in the main text that it was present at high frequencies in the historical (250BP) Faroese samples. The frequency of this allele in the imputed historical Faroese samples is 82% while the allele is present at ~74% frequency in modern samples. We originally did not report the exact percentage in the main text because the sample size of the historical samples (11 individuals) is small and coverage of ancient samples is low, leading to potential errors in imputation.

However, given the reviewer’s comment, we have now included the frequencies as well as these caveats in the Discussion. We additionally calculated the LCT allele frequency in other ancient samples, and assuming that we had good proxies for the sources at the time of admixture, we calculated the expected allele frequency in the admixed ancestors of the Faroese founders (Discussion), but again note the limitations in using such a calculation in this context.

(5) I am lacking information to evaluate the procedure for turning the outliers into p-values. Both iHS and XP-EHH are ratio statistics, meaning they might be heavy-tailed if one is not careful, and the central limit theorem may not apply. It would be much easier (and probably sufficient for the points being made here) to reframe this analysis in terms of empirical outliers.

Given that there are disagreements on the best approach to reporting selection scan results from the reviewers, in our revision, we have additionally supplied both the standardized iHS / XP-EHH values in Supplementary Fig. S10 as well as these values transformed to p-values in Main Text Fig. 3. Additionally, both outputs are provided in the publicly available selection scan results files. We provide the method for obtaining p-values in the subsection “Selection scan” from the Methods section - we used a method developed earlier by Fariello et al.

(6) Oldest individual predating gene flow: It seems impossible to make any statements based on a single individual. Why is it implausible that this person (or their parents), e.g., moved to the Faroes within their lifetime and died there?

We agree with the reviewer that this is a plausible explanation, and in our revisions, we have updated the Main Text - Discussion to acknowledge this possibility.

Recommendations for the authors:

Reviewing Editor Comments:

Please note that there was disagreement among the reviewers regarding the reporting of outliers.

As stated in our response to the public reviews, given the disagreement, we include both the empirical selection statistics as well as the converted p-values in the main text, supplement and selection scan files.

Reviewer #2 (Recommendations for the authors):

(1) Figure 2:

Define labels / explain why they differ from 1000k populations / make them consistent throughout the manuscript.

We apologize for the error in labels for Figure 2. These are the same populations used in other figures and analyses. We have fixed this in our revisions so that the labels are consistent with the rest of the manuscript.

(2) Figure S2 label:

"The matrix is rescaled after subsetting the individuals, so although the scales are different, the overall structure remains the same." I do not understand this sentence. The samples are different, the scale is different, the apparent pattern is different - what overall structure is supposed to be the same?

We apologize that the language was not clear in the figure label. The scales between panels A and B are different, because popkin rescales the kinship labels after subsetting so that the minimum kinship is zero. This is necessary when subsetting individuals from an already estimated kinship matrix particularly when subsetting from global populations to a single region. From the popkin documentation: “This rescaling is required when subsetting results in a more recent Most Recent Common Ancestor (MRCA) population compared to the original dataset (for example, if the original data had individuals from across the world but the subset only contains individuals from a single continent)” (https://rdrr.io/cran/popkin/man/rescale_popkin.html).

We also described this in the Methods - Population Genetics - Kinship and runs of homozygosity section: “When calculating the kinship matrix for the Faroese WGS cohort only, we used the rescale_kinship() function, which will change the most recent common ancestor and give different absolute values, but the overall relationship structure in the subpopulation remains the same.”

That is, the relative kinship within the Faroese cohort remains consistent, despite the different scale.

It is difficult to see the kinship of Faroese individuals in the larger plot with all cohorts, which is why we subset and visualize the Faroese cohort alone. We have updated the Fig. S2 label language to make this more clear.

(3) "Iron Age Wet Europe"

We have corrected this typo to “Iron Age West Europe.”

I'm confused if the ancient Faroese were part of the imputation panel: Figure 5 legend implies they are, methods imply they are not.

The ancient samples are not imputed with the modern Faroese and reference samples, but they are the imputed data downloaded from Allentoft et al. and merged with the modern Faroese cohort. We specify that we downloaded imputed ancient samples in both the Methods - Fine-scale structure estimation using ancient genomes and in the Main Text - Results - Fine-Scale Structure and Connections to Ancient Genomes. The description of the imputation panel in the Methods - Bioinformatics - Variant calling and imputation refers only to the modern samples.

(4) Kinship:

The kinship of the Faroes is useful (and nice) as a QC analysis showing the genetic data matches the expectations from the pedigree. I don't know what I should learn from the kinship of the 1000kg samples (I'd assume one could learn something about bottleneck strength from this), but it's not developed/discussed.

The global kinship matrix provides complementary information to PCA and ROH, as another way to quantify and visualize the relationships within and between populations. Additionally, as the reviewer mentioned, bottlenecks increase kinship within populations. Given that popkin estimates kinship measured from a Most Recent Common Ancestor, we can best observe this increase in kinship when comparing to other global populations. We more clearly delineate what can be observed from Fig. S2A versus Fig. S2B in the Results - Population Structure and Relatedness.

Reference

(1) Gretzinger, J. et al. The Anglo-Saxon migration and the formation of the early English gene pool. Nature 610, 112–119 (2022)

(2) Leslie, S. et al. The fine-scale genetic structure of the British population. Nature 519, 309–314 (2015).

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Henshall et al. delete the highly abundant merozoite surface protein PfMSP2 from two Plasmodium falciparum laboratory lines (3D7 and Dd2) using CRISPR-Cas9. Parasites lacking MSP2 replicate and invade red cells normally, opposing the experimental history that suggests MSP2 is essential. Unexpectedly, the knock-outs become more susceptible to several inhibitory antibodies - most strikingly those that target the apical antigen AMA1-while antibodies to other surface or secreted proteins are largely unaffected. Recombinant MSP2 added in vitro can dampen AMA1-antibody binding, supporting a "conformational masking" model. The reported data suggest that MSP2 helps shield key invasion ligands from host antibodies and may itself be a double-edged vaccine target.

Reviewer 1 did not have any comments we needed to address.

Reviewer #2 (Public review):

(1) The section describing Laverania and avian Plasmodium MSP2 comparison is a lengthy section and could be told much more concisely for clarity in delivering the key message, i.e., that conservation in distantly related Plasmodium species could indicate an important function. The identification of MSP2-like genes in avian Plasmodium species was highlighted previously in the referenced Escalante paper, so it is not entirely novel, although this paper goes into more detailed characterisation of the extent of conservation. Overall, this section takes up much more space in the manuscript than is merited by the novelty and significance of the findings.

As outlined in point (1) for Reviewer 1 (Recommendations for the authors), we have cut back through this section and focussed on the important comparisons rather than the general observation. We have also moved the elements of Table 1 to Supplementary Figures 2, 3 and 4 to streamline the manuscript. Further description of the changes is available in the Reviewer #1 (Recommendations for the authors).

(2) Characterisation of the knockout strains is generally thorough, though relatively few interactions were followed by live microscopy (Figures 3E-H). A minimum of 30 merozoites were followed in each assay (although the precise number is not specified in the figure or legend), but there are intriguing trends in the data that could potentially have become significant if n was increased.

In the Figure 3 Legend we have now indicated the number of merozoite invasions followed as per the following:

“(E-H) Key parameters of merozoite invasion were measured for both PfDd2 WT (n = 43) and PfDd2 ΔMSP2 (n = 35) parasites that had successfully invaded a RBC using live cell imaging of merozoite invasion.”

We have also removed the more general description of ‘a minimum of 30 merozoites’ from the same Figure Legend.

The number of schizont ruptures and subsequent merozoite invasions followed for each experiment is in line with previous studies that have investigated phenotypes with invasion inhibitors and gene knock-outs (e.g. Weiss et al. 2015, PLoS Pathogens). It is important to note that the data refers to merozoites that have completed invasion, and not just the number of merozoites that have been released from a schizont which is typically 2-4 times more than have invaded. This means we are comparing the kinetics of invasion across a relatively large sample size compared to other studies of inhibitory phenotypes. While it is possible that increasing the number of merozoites being filmed might lead to some statistical significance for some of the trends, we note that there is a limited growth phenotype overall in both short and long-term culture and this fits with the limited defect we are seeing. In order to better address this, as outlined in our response to point (7) for Reviewer 2 (Recommendations for the authors), we now discuss the trends seen in the data in additional detail.

(3) The comparative RNAseq data is interesting, but is not followed up to any significant degree. Multiple transcripts are up-regulated in the absence of PfMSP2, but they are largely dismissed because they are genes of unknown function, not previously linked to invasion, or lack an obvious membrane anchor. Having gone to the lengths of exploring potentially compensatory changes in gene expression, it is disappointing not to validate or explore the hits that result.

While we understand the reviewers comment, as outlined in the text we did not identify any upregulated proteins that looked like strong candidates to compensate for loss of MSP2 to explore in this manuscript. Instead, we chose to further investigate any potential loss of MSP2 phenotype that yielded the observations around improved potency of antibodies targeting some merozoite antigens with loss of MSP2. This will be explored in future studies as we try and understand the role of MSP2 in more detail and the interactions between proteins and antibodies on the merozoite surface.

(4) Given the abundance of PfMSP2 on the merozoite surface, it would have been interesting to see whether the knockout lines have any noticeable difference in surface composition, as viewed by electron microscopy, although, of course, this experiment relies on access to the appropriate facilities.

We agree with the reviewer, but this lies outside the scope of this manuscript and optimisation of the imaging platform used to gain biologically useful insights would take a considerable amount of work based on feedback from people working with these techniques.

(5) One of the key findings is that deletion of PfMSP2 increases inhibition by some antibodies/nanobodies (some anti-CSS2, some anti-AMA1) but not others (anti-EBA/RH, anti-EBA175, anti-Rh5, anti-TRAMP, some anti-CSS2, some anti-AMA1). The data supporting these changes in inhibition are solid, but the selectivity of the effect (only a few antibodies, and generally those targeting later stages in invasion) is not really discussed in any detail. Do the authors have a hypothesis for this selectivity? The authors make attempts to explore the mechanisms for this antibody-masking (Figure 7), but the data is less solid. Surface Plasmon Resonance was non-conclusive, while an ELISA approach co-incubating MSP2 and anti-AMA1 antibodies to wells coated with AMA1 lacks appropriate controls (eg, including other merozoite proteins in similar experiments).

As outlined in our response to point (7) for Reviewer 2 (Recommendations for the authors), we have repeated the ELISA based assessment of recombinant MSP2s impact on anti-AMA1 antibody binding. In addition, we have included two comparator control proteins, the intrinsically disordered MSP4 of P. falciparum and the globular domain of the neural cell adhesion molecule (NCAM, CD56, 16 kDa), and found these proteins did not impact binding of anti-AMA1 antibodies. This strengthens the data that links the presence of MSP2 to reduced activity of anti-AMA1 antibodies.

As covered in our response to point (7) for Reviewer 2 (Recommendations for the authors) we provide additional discussion of this phenotype. We note that the list of inhibitory antibodies tested is not exhaustive, and additional antibodies may be identified where loss of MSP2 could improve potency. So although we see a consistent effect with a relatively small number of antibody targets, this does not rule out additional examples that may act earlier in invasion (for example, we noticed a small, but not statistically significant, trend for mildly inhibitory antibodies targeting MSP1-19 as well) and this makes speculating on why these two initial antibody targets at this time problematic.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) If feasible, perform ex vivo assays to demonstrate that the masking effect operates with physiologically relevant antibodies.

For this manuscript, we focussed on characterising the MSP2 knock-out parasites using the best reagents available. We remain interested in understanding whether these lines can be used to investigate the activity of functional antibodies from malaria exposed human serum and this will be the subject of future studies.

Reviewer #2 (Recommendations for the authors):

(1) As noted in the Public Review, the section describing MSP2 orthologues in other Laverania and avian Plasmodium species is overly long and not the most novel section of the manuscript. It could be really radically trimmed back.

We have taken this suggestion for the reviewer on board and have significantly cut back on our descriptions of the basic similarity properties of the conserved N and C-terminal regions as well as the description of the central variable region. Effectively, we have cut back the number of words through this section from 864 across 3 paragraphs to 478 across 2 paragraphs. While we have chosen to greatly economise our description of the N and C-terminal conserved regions, we have maintained much of the description of the similarities and differences in the central variable region as we believe the observation that this variant region still maintaining repeats, though they differ in size, number and amino acid composition, across such evolutionary distances is of interest.

Taking the reviewers comment on board, we have also removed Table 1 from the manuscript (shows amino acid sequence properties of these regions) and instead have inserted the tables relevant for each alignment in Supplementary Figures 2, 3 and 4 as appropriate. This will streamline the main manuscript and better align amino acid property and alignment data in the one Figure. We thank the reviewer for this feedback and believe that this has helped focus the text on the most important observations.

(2) Figure 2C - As MSP2 has stage-specific expression, it could be informative to incorporate an antibody targeting another gene with a similar stage-specific expression pattern, such as AMA,1 into the blot. This would confirm that both protein samples were collected at a similar point during blood stage development.

We have modified Figure 2C to include both the original comparison using PfAldolase as the loading control and also the merozoite expressed PfGAP45 as a loading/stage specific control as per the Figure.

(3) Figure 2D - Magenta and red are hard to distinguish in the merge channel. Is it possible to pseudocolour one of these channels a different colour? Also, it would be simpler to keep PfMSP2 a consistent colour in both rows.

Thank you for this suggestion and we agree that the comparison could be made clearer. For this figure, we have coloured DAPI to label the nuclei (Cyan), and antibodies targeting PfMSP2 (Magenta), PfAMA1 and PfMSP1-19 (Yellow). This is also reflected in the merged image. The Figure legend now reads:

“(D) Distribution of key merozoite surface proteins in the presence or absence of PfMSP2 was visualised by immunofluorescence. PfMSP2 (magenta), the nucleus stained by DAPI (cyan) and PfAMA1 (yellow, top two rows) or PfMSP1-19 (yellow, bottom two rows), and the coloured merge of the preceding panels. Scale bar = 0.7 µm. Representative images shown from a minimum of 10 schizonts imaged per condition.”

(4) Figure 2F - Static growth relative to shaking growth is plotted in this panel; perhaps this could be more clearly described in the legend or mentioned in the text that there was not a significant alteration in growth in static or shaking conditions.

As suggested, we have clarified the result in the Figure legend text as follows:

“(E-F) Growth of Pf3D7 WT compared to Pf3D7 ΔMSP2 P. falciparum parasites, measured as fold increase in parasitaemia, over one (48 hrs) or two (96 hrs) cycles in either standard (still- (E)) or shaking (F) conditions, with no measurable difference between parasite growth rates seen between standard or shaking conditions.”

Please also describe the shaking conditions used (i.e., speed, culture size, and vessel) in the methods.

We have updated the methods to provide information on the growth conditions used in the standard versus shaking growth assays:

“The initial parasitemia of cultures was determined by flow cytometry and then measured again after the 50 mL cultures in 96 well plates were maintained under standard (still) or shaking (50 rpm) conditions for 48 hrs or 96 hrs of growth.”

(5) Figure 3G - Annotate legend for strength of deformation to describe what 1,2, or 3 refers to.

We have added the following to the Figure legend of Figure 3G:

“Deformation scores are as defined by Weiss et al (Weiss et al., 2015), with 1 = weak deformation of the RBC membrane at the point of contact, 2 = strong deformation leading to the RBC membrane extending up the sides of the merozoite and changes in RBC membrane curvature beyond the point of contact and 3 = extreme deformation indicated by the merozoite being deeply embedded in the RBC membrane and strong deformation of the RBC well beyond the point of contact.”

There is a small visible shift in the deformation event scores. Is this also not significant? Even if deformation is not significantly longer, could this small effect alter the exposure of epitopes on other proteins for antibody targeting?

We did test the deformation event scores and the differences were non-significant. We have considered this possibility raised by the reviewer, but we are cautious in over interpreting the possibility that these trends might contribute to the increased potency of certain antibodies in the absence of additional data. We note that, although deformation may happen over a slightly longer timescale and show more aggressive deformations with PfMSP2 knock-out, this also seems to translate into a weak trend for faster overall entry for those merozoites that go on to invade. Therefore, although deformation may be longer and stronger, antibodies may have less time to block invasion overall. We are not confident that we can interpret around what might be happening at the molecular scale here based on this data and have chosen not to discuss this possibility in the manuscript. However, we have added the following to the results to better explain the phenotype the phenotype we observed.

“This analysis showed that, although there was a trend for PfDd2 ΔMSP2 knock-out parasites to have a higher mean time to attach to the RBC, as well as for the length and strength of RBC deformation, these trends did not reach significance. For those merozoites that did invade the RBC, on average it took less time for PfDd2 ΔMSP2 knock-out parasites to invade then PfDd2 WT, but this again did not reach significance (Figure 3 E-H). Together these data show PfMSP2 is not essential for blood-stage replication in vitro in two P. falciparum laboratory isolates from different geographical regions and knock-out of PfMSP2 does not seem to significantly impact parasite growth or merozoite invasion in vitro.”

(6) Figure 4C - Legend refers to black lines, but on the figure, they are red? Is the horizontal red line in the correct place, or should some of the dots below it be black rather than blue if they fall outside the adjusted p-value significance cut-off? Were 4 schizont harvests performed in total, or 4 for each cell line?

We thank the reviewer for pointing this out and we have now changed the text to say red lines. We have also provided more information in the Figure legend to more clearly define what data is represented. In short, 4 harvests were performed for each cell line (8 in total across the 2 cell lines) and the data represents the distribution from one of these harvests. The blue shaded genes are those that, on average, across the 4 Pf3D7 WT and Pf3D7 ΔMSP2 paired harvests show up or down-regulated expression. This is why some of the blue shaded genes lie near or below the cut-off values represented by the red line. The Figure legend text has now been modified as follows.

“(C) Log2(fold change) for differentially expressed genes, including multigene families, between the transcriptome of Pf3D7 WT and Pf3D7 ΔMSP2 schizonts. Plot represents the results for one of four independent schizont RNA harvests for Pf3D7 WT and Pf3D7 ΔMSP2 parasites and red lines differentiate genes with a log2 (fold change) > 0.5 and < -0.5 with adjusted p-value < 0.01. Genes shaded blue represent those genes that were found to have an average log2 (fold change) > 0.5 (dark blue) or < -0.5 (light blue) across the four replicate samples compared. Significance determined as below p< 0.05 after correction for multiple testing.”

(7) Figure 7D - ELISA results don't show a convincing concentration-dependent inhibition, and repeating with another recombinant protein is essential before inferring that the effect is specific to PfMSP2

We have repeated the ELISA experiment using recombinant PfMSP2 to reduce variability across the assay and again found a dose dependent reduction of anti-PfAMA1 binding with increasing concentrations of recombinant PfMSP2. It should be noted that this is a completely new set of experiments that recapitulate the original findings. See updated Figure 7D.

We agree with the reviewer that the experiment and interpretation of the data would be strengthened by comparing any potential inhibitory impact on anti-PfAMA1 binding to a different recombinant protein. Therefore, we have completed identical experiments using the similarly intrinsically disordered PfMSP4 recombinant protein (40 kDa) and the highly structured 16 kDa immunoglobulin domain of human neural cell adhesion molecule (NCAM). We find that there is no dose dependent loss of anti-PfMAMA1 binding to recombinant PfAMA1 with addition of PfMSP4 or NCAM immunoglobulin domain recombinant protein. These controls are contained in Supplementary Figure 6, the relevant text is provided below.

‘In contrast, increasing concentrations of the intrinsically disordered MSP4 from P. falciparum 3D7 (40 kDa) and the highly structured immunoglobulin domain of neural cell adhesion molecule (NCAM, CD56, 16 kDa) recombinant proteins did not impact on binding of anti-PfAMA1 antibodies to recombinant AMA1 (Supplementary Figure 6).’

(8) Again, as noted in the public review, the target-specificity of the inhibition-masking effect is perhaps the most surprising aspect of the data - this could do with much more thorough discussion. Why only these proteins, both of which function late in invasion?

Overall, we tested several growth inhibitory and non-inhibitory antibodies shown to bind specifically to individual or some combination of nine P. falciparum merozoite surface and secreted proteins. However, we do not consider this to be an exhaustive list of potentially invasion inhibitory antibodies by any means. We mostly did not observe any non-inhibitory antibodies becoming significantly more growth inhibitory to PfMSP2 KO lines, indicating that these antibodies were not impacted by loss of PfMSP2 or had no functional inhibitory effect in these assays.

What we do demonstrate here is that we see a consistent impact with different rabbit, mouse monoclonal and i-body growth inhibitory antibodies targeting PfAMA1, indicating that it is not a spurious result from a single antibody or antibody type. We also find a second example, with nanobodies targeting the PfPCRCR complex protein PfCSS potentiated with loss of PfMSP2. This opens up the possibility that other growth inhibitory antibodies to the antigens tested here, or growth inhibitory antibodies targeting other antigens involved in merozoite invasion, may also become more potent with MSP2KO. Although both PfAMA1 and PfCSS function late in invasion, it is too early to say whether this is a functional trend or an observation that is related to the panel of antibodies tested. Therefore, further testing using lines developed in this study could yield additional examples of antibodies that become more inhibitory with MSP2 KO and provide additional information on the potential impact that MSP2 may have on their vaccine potential. In order to address this, we have added the following text to the discussion:

“Here we show consistent potency improvement with PfMSP2 knock-out for growth inhibitory rabbit, mouse monoclonal and i-body antibodies targeting PfAMA1, as well as demonstrate improved activity for and Fc-tagged nanobody targeting PfCSS, indicating that these are not outlier results from a single antibody or antibody type. However, increased antibody potency was not shared across all antibodies tested, possibly because the specific function or localisation of a target protein, the region that an antibody binds to or the functional activity (or lack thereof) of an antibody may all play a role in determining whether loss of PfMSP2 can potentiate growth inhibitory activity. Further investigation using the parasite lines developed in this study and a wider panel of antibodies that target different stages of the merozoite invasion process could shed more light on this potentially novel mechanism of vaccine derived antibody efficacy.”

(9) Typos/minor editorial points:

L111 – conserved

This text has been modified.

L235-237 - check the wording in this sentence for clarity

This text has been modified.

Figure 3E - 'attachment' on axis

This Figure has been modified.

L350 - mentions eight 'proteins' having expression increase, instead 'transcripts' should be referred to when describing RNAseq data, as transcript levels may not correspond directly with protein levels. Also, be careful when referring to transcript or protein throughout this paragraph.

This text has been modified.

Figure 4A - instead of 'transcription during schizonts', better to say 'schizont transcript abundance'

This text has been modified.

L514 - 'detectable binding to PfAMA1'

This text has been modified.

L589 - Is it a mouse Fc region or a human Fc region that is added? The human Fc region is mentioned in the results.

In the growth inhibition assays anti-AMA1 WD34 i-body with a human FC region was used and in the ELISA assays anti-AMA1 WD34 i-body with a mouse FC region (to enable detection of AMA1 binding use the same secondary anti-body for both the WD34 i-body and the 4G2 mouse monoclonal antibody) was used. The text has been been checked and modified accordingly to clearly say this.

Supplementary figure 3 - 'repeats'

This text has been modified.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

In this manuscript, the authors describe the generation of a Drosophila model of RVCL-S by disrupting the fly TREX1 ortholog cg3165 and by expressing human TREX1 transgenes (WT and the RVCL-S-associated V235Gfs variant). They evaluate organismal phenotypes using OCT-based cardiac imaging, climbing assays, and lifespan analysis. The authors show that loss of cg3165 compromises heart performance and locomotion, and that expression of human TREX1 partially rescues these phenotypes. They further report modest differences between WT and mutant hTREX1 under overexpression conditions. The study aims to establish Drosophila as an in vivo model for RVCL-S biology and future therapeutic testing.

Strengths:

(1) The manuscript addresses an understudied monogenic vascular disease where animal models are scarce.

(2) The use of OCT imaging to quantify fly cardiac performance is technically strong and may be useful for broader applications.

(3) The authors generated both cg3165 null mutants and humanized transgenes at a defined genomic landing site.

(4) The study provided initial in vivo evidence that human TREX1 truncation variants can induce functional impairments in flies.

Weaknesses:

(1) Limited mechanistic insight.

RVCL-S pathogenesis is strongly linked to mislocalization of truncated TREX1, DNA damage accumulation, and endothelial/podocyte cellular senescence. The current manuscript does not examine any cellular, molecular, or mechanistic readouts - e.g. DNA damage markers, TREX1 subcellular localization in fly tissues, oxidative stress, apoptosis, or senescence-related pathways. As a result, the model remains largely phenotypic and descriptive.

We thank the reviewers for these suggestions. We are planning to perform experiments addressing the RVCL-S linked cellular deviations. We will examine DNA damage markers on cellular level and perform TUNEL tissue staining to visualize apoptosis, etc.

To strengthen the impact, the authors should provide at least one mechanistic assay demonstrating that the humanized TREX1 variants induce expected molecular consequences in vivo.

Yes, we are planning to demonstrate the distinct effects from TREX1 and TREX1 V235G expression on molecular level.

(2) The distinction between WT and RVCL-S TREX1 variants is modest.

In the cg3165 rescue experiments, the authors do not observe differences between hTREX1 and the V235Gfs variant (e.g., Figure 3A-B). Phenotypic differences only emerge under ubiquitous overexpression, raising two issues:

i) It is unclear whether these differences reflect disease-relevant biology or artifacts of strong Act5C-driven expression.

Thanks for pointing out this issue. We will discuss the differences between two expression models in the revised manuscript.

ii) The authors conclude that the model captures RVCL-S pathogenicity, yet the data do not robustly separate WT from mutant TREX1 under physiological expression levels.

We will provide more details related to the RVCL-S disease development and agerelated manifestations.

The authors should clarify these limitations and consider additional data or explanations to support the claim that the model distinguishes WT vs RVCL-S variants.

We will address the reviewer concerns and re-write the related manuscript sections to provide more clarity.

(3) Heart phenotypes are presented as vascular defects without sufficient justification.

RVCL-S is a small-vessel vasculopathy, but the Drosophila heart is a contractile tube without an endothelial lining. The authors refer to "vascular integrity restoration," but the Drosophila heart lacks vasculature.

We will expand the model justification section and will be more careful with our statements to avoid misunderstanding of the experimental conclusions.

The manuscript would benefit from careful wording and from a discussion of how the fly heart phenotypes relate to RVCL-S microvascular pathology.

We thank the reviewer for pointing to this issue. Justifying Drosophila usage for human disease modelling is always challenging. We will re-write the corresponding parts of the manuscript.

(4) General absence of tissue-level or cellular imaging.

No images of fly hearts, brains, eyes, or other tissues are shown. TREX1 nuclear mislocalization is a hallmark of RVCL-S, yet no localization studies are included in this manuscript. Adding one or two imaging experiments demonstrating TREX1 localization or tissue pathology would greatly enhance confidence in the model.

As suggested by the reviewers,we will add tissue imaging experiments to illustrate the pathological effects of RVCL linked TREX1 expression. We are also planning to utilize CRIMIC line CR70804 to visualize fly TREX1 tissue distribution.

Reviewer #2 (Public review):

Summary:

The authors used the Drosophila heart tube to model Retinal vasculopathy with the goal of building a model that could be used to identify druggable targets and for testing chemical compounds that might target the disease. They generated flies expressing human TREX1 as well as a line expressing the V235G mutation that causes a C-terminal truncation that has been linked to the disease. In humans, this mutation is dominant. Heart tube function was monitored using OCM; the most robust change upon overexpression of wild-type or mutant TREX1was heart tube restriction, and this effect was similar for both forms of TREX1.

Our results are consistent with the human disease nature, RVCL-S carriers and non-carriers are both healthy and asymptomatic at young age; however, the accumulation of physiological stress becomes obvious in midlife, leading to premature death in 40s and 50s. We will expand the discussion section focusing on RVCL-S manifestations in aged animals.

Lifespan and climbing assays did show differential effects between wt and mutant forms when they were strongly and ubiquitously expressed by an actin-Gal4 driver. Unfortunately, these types of assays are less useful as drug screening tools. Their conclusion that the primary effect of TREX is on neuronal function is inferential and not directly supported by the data.

We will revise this experiment discussion and plan to include additional experiments to strengthen the conclusions.

The authors do not show that CG3165 is normally expressed in the heart. Further fly heart tube function was similarly restricted in response to expression of either wild-type or mutant TREX1. The fact that expression of any form of human TREX1 had deleterious effects on heart function suggests that TREX1 serves different roles in flies compared to humans. Thus, in the case of this gene, it may not be a useful model to use to identify targets or use it as a drug screening tool.

We will examine the expression of cg3165, human TREX1 transgenes in whole organism to demonstrate tissue expression profiles, as noted above. We will also expand the relevant manuscript sections to address the systemic manifestations of RVCL.

The significant effects on lifespan and climbing that did show differential effects required ubiquitous overexpression using an actin-gal4 driver that does not allow the identification of tissue-specific effects.

We plan to carry out additional experiments to determine cg3165, and human TREX1 tissue expression profile.

Thus, their assertion that the results suggested a strong positive correlation between Drosophila neuromotor regulation and transgenic hTREX1 presence and a negative impact from hTREX1 V235G" is not supported by these data.

Thanks for pointing this out. We will revise our conclusions appropriately after we include the results from additional new experiments.

Also worrisome was the inability to identify the mutant TREX1 protein by Western blot despite the enhanced expression levels suggested by qPCR analysis. Mutant TREX1 cannot exert a dominant effect on cell function if it isn't present.

We will try to resolve this issue by technical means.

There are also some technical problems. The lifespan assays lack important controls, and the climbing assays do not appear to have been performed correctly.

We would disagree with this statement. We will re-write the method description for better clarity.

It is unclear what the WT genetic background is in Figure 1-3, so it is unclear if the appropriate controls have been used. Finally, the lack of information on the specific statistical analyses used for each graph makes it difficult to judge the significance of the data.

We will provide clearer descriptions of our controls and procedures.

Overall, the current findings establish the Retinal vasculopathy disease model platform, but with only incremental new data and without any mechanistic insights.

We will include additional experiments addressing the mechanism (see previous responses above).

Reviewing Editor Comments:

I (Hugo Bellen) also read your paper and noted that you do not document the expression pattern in the nervous system and other tissues, such as the heart. The stock https://flypush.research.bcm.edu/pscreen/crimic/info.php?CRname=CR70804 may help you do this and should allow you to compare the GAL4 induced expression of the stock you created and this stock. If compatible, you should consider reporting expression patterns.

Thank you for the suggestion. We will obtain the line and will use it for expression visualization.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

(1) The authors appear to be excluding a significant fraction of the TCRlow gamma delta T cells from their analysis in Figure 1A. Since this population is generally enriched in CD25+ gamma delta T cells, this gating strategy could significantly impact their analysis due to the exclusion of progenitor gamma delta T cell populations.

We were cautious in our gating strategy since the TCR𝛿+ CD3e+ subset is rather small and so low signal/background noise ratio can be an issue if the gates used are too broad/generous. There is some inevitable low level background staining with the TCR𝛿 that sits just above the bulk of the negative population and is CD3ε -ve. Although this background represents a tiny fraction of total cells, we were wary of gate contamination into our TCR𝛿+ CD3e⁺ subset and we wanted a gating strategy that could be applied across other organs too. We do not, however, believe this conservative strategy is impacting on measurements progenitor numbers across strains or our conclusions, since the size of this progenitor population in the various IKKΔT^CD2 and Casp8ΔT^CD2 strains was never impacted by the mutations. But to reassure the reviewer, we show our conservative gate as compared with a very broad TCR𝛿 gate and see we are not missing a substantial population of CD25+ cells just below our gate. This also helps illustrate how close the background from the CD27^int expressing αβ thymocytes (right column) comes to the TCR𝛿+ CD3+ gate and the importance of tight lineage gating.

Author response image 1.

(2) The overall phenotype of the IKKDeltaTCd2 mice is not described in any great detail. For example, it is not clear if these mice possess altered thymocyte or peripheral T cell populations beyond that of gamma delta T cells.

Given that gamma delta T cell development has been demonstrated to be influenced by gamma delta T cells (i.e, trans-conditioning), this information could have aided in the interpretation of the data.

Apologies for not being clearer on this point. We have studied conventional αβ T cell development in these strains in considerable detail, and these studies are published and discussed in some detail in the introduction in paragraph 3 on page 3-4 and in cited references Schmidt-Supprian et al 2004, SIlva et al 2014, Xing et al 2016, Webb et al 2019, Carty et al 2023. These detail how IKK expression is critical for thymic development of αβ T cells and their peripheral survival, and dissects the role of NF-κB activation and cell death regulation by IKK. However, we now add new discussion (page 11-12) that considers the potential impact of altered αβ T cell development in the strains used for this study.

We agree that trans-conditioning is also an important consideration, since CD4 TH17 T cells can enhance type 17 𝛾𝛿 T cell development (10.1038/icb.2011.50). This is of relevance to the limited conclusions we draw concerning type 17 𝛾𝛿 T cells. The REL and IKK deficient strains do lack effector populations, including type 17 αβ T cells, so it is possible that the absence of type 17 αβ T cells in these strains does contribute to the modest impact of IKK deletion in the type 17 𝛾𝛿 subset. We now highlight this information and discuss in the manuscript (page 11-12).

Related to this, it would have been helpful if the authors provided a comparison of the frequencies of each of the relevant subsets, in addition to the numbers.

We now provide both the absolute frequencies of different 𝛾𝛿 subsets and their relative frequencies to one another, as supplementary figure 2. We still believe assessing absolute numbers is the gold standard, since the differential impact of gene deletions on the αβ T cell compartments in different strains will effect whether or not αβ T cells are present, and therefore overall representation of 𝛾𝛿 T cells can vary considerably between strains. Hence, absolute numbers are more reliable measure of cell abundance.

(3) The manner in which the peripheral gamma delta T cell compartment was analyzed is somewhat unclear. The authors appear to have assessed both spleen and lymph node separately. The authors show representative data from only one of these organs (usually the lymph node) and show one analysis of peripheral gamma delta T cell numbers, where they appear to have summed up the individual spleen and lymph node gamma delta T cell counts. Since gamma deltaT17 and gamma deltaT1 are distributed somewhat differently in these compartments (lymph node is enriched in gamma deltaT17, while spleen is enriched in gamma deltaT1), combining these data does not seem warranted. The authors should have provided representative plots for both organs and calculated and analyzed the gamma delta T cell numbers for both organs separately in each of these analyses.

We did of course process and calculate numbers of different subsets in both lymph nodes and spleen. Where we saw loss of peripheral 𝛾𝛿 subsets, or rescue, this was reflected in seperate analysis of both organs and we did not see any organs specific effects in the mouse strains analysed. We therefore took the initial view that presenting aggregate data was most efficient and least repetitive representation of data. However, we very much recognise the reviewers concern, and interest to see these data, so have now included representative plots across both organs for figure 1D, and show cell numbers of lymph nodes and spleen separately, as well as together, for figures 1, 2, 4 and 7, and these plots reflect the differences observed when we combined data. We did not break down the data for all figures (e.g. figures 3 and 5) as it was more cumbersome for more complex multi-strain comparisons and so attempt to balance clarity and transparency against unnecessary repetitive data presentation.

(4) The authors make extensive use of surrogate markers in their analysis. While the markers that they choose are widely used, there is a possibility that the expression of some of these markers may be altered in some of their genetic mutants. This could skew their analysis and conclusions. A better approach would have been to employ either nuclear stains (Tbx21, RORgammaT) or intracellular cytokine staining to definitively identify functional gamma deltaT1 or gamma deltaT17 subsets.

We did share a similar concern, but think this is not an issue where subsets disappear and are almost completely absent, such as in IKK1/2 KO and Casp8 KO settings. Where we saw rescue with RIPK1^D138N in Casp8ΔT^CD2 strains, we were keen to demonstrate that the populations we saw restored did exhibit their expected function, and so confirmed this in figure 5C by intracellular cytokine staining after a short 4h restimulation in vitro. This also served to validate our gating strategy, since what we designated as Type 1 cells - CD27+CD122+CD44^int cells were the only source of IFN-gamma, while CD27–CD44^hi CD122^lo cells were the only source of IL-17. Adaptive/ naive cells made neither cytokine. So while we did not include nuclear stains, we were satisfied that the cytokine assays validated the gating strategy.

(5) The analysis and conclusion of the data in Figure 3A is not convincing. Because the data are graphed on log scale, the magnitude of the rescue by kinase dead RIPK1 appears somewhat overstated. A rough calculation suggests that in type 1 game delta T cells, there is ~ 99% decrease in gamma delta T cells in the Cre+WT strain and a ~90% decrease in the Cre+KD+ strain. Similarly, it looks as if the numbers for adaptive gamma delta T cells are a 95% decrease and an 85% decrease, respectively. Comparing these data to the data in Figure 5, which clearly show that kinase dead RIPK1 can completely rescue the Caspase 8 phenotype, the conclusion that gamma delta T cells require IKK activity to repress RIPK1-dependent pathways does not appear to be well-supported. In fact, the data seem more in line with a conclusion that IKK has a significant impact on gamma delta T cell survival in the periphery that cannot be fully explained by invoking Caspase8-dependent apoptosis or necroptosis. Indeed, while the authors seem to ultimately come to this latter conclusion in the Discussion, they clearly state in the Abstract that "IKK repression of RIPK1 is required for survival of peripheral but not thymic gamma delta T cells." Clarification of these conclusions and seeming inconsistencies would greatly strengthen the manuscript. With respect to the actual analysis in Figure 3A, it appears that the authors used a succession of non-parametric t-tests here without any correction. It may be helpful to determine if another analysis, such as ANOVA, may be more appropriate.

Yes, we completely agree with this assessment and conclusion. While kinase dead RIPK1 does provide some rescue, this appears relatively modest, and instead supports the view, validated in figure 7, that maybe the dominant function of IKK in 𝛾𝛿 T cells is to activate NF-κB dependent survival signals. Nevertheless, RIPK1^D138N does provide some significant rescue, which allows some peripheral cells to repopulate and demonstrates that IKK is repressing RIPK1 mediated cell death. It is actually not trivial to assess the relative importance of IKK-RIPK1 and IKK-NF-κB functions. In the IKKΔT^CD2 RIPK1^D138N mice, we prevent RIPK1 induced death, but still lack the NF-κB-dependent survival signal. Consistent with this, the ~1log reduction in 𝛾𝛿 numbers between WT and IKKΔT^CD2 RIPK1^D138N mice is actually similar to what we observe in the absence of REL subunits (Fig. 7) which is a smaller reduction than we observe in IKKΔT^CD2 mice. What would have been ideal is to have a scenario where IKK regulation of RIPK1 was defective but NF-κB survival signalling was intact. This would reveal the full impact of loosing IKK dependent regulation of RIPK1 alone, which we suspect would result in substantial cell death that could not be blocked by NF-κB. Unfortunately, we not have or know of suitable mouse mutants to test this. This is quite a nuanced discussion and we now clarify the scope and extent of conclusions we can draw (p. 7, 11).

(6) The conclusion that the alternative pathway is redundant for the development and persistence of the major gamma delta T cell subsets is at odds with a previous report demonstrating that Relb is required for gamma delta T17 development (Powolny-Budnicka, I., et al., Immunity 34: 364-374, 2011). This paper also reported the involvement of RelA in gamma delta T17 development. The present manuscript would be greatly improved by the inclusion of a discussion of these results.

Thank you - we include a discussion of these papers now (p12).

(7) The data in Figures 1C and 3A are somewhat confusing in that while both are from the lymph nodes of IKKdeltaTCD2 mice, the data appear to be quite different (In Figure 3A, the frequency of gamma delta T cells increases and there is a near complete loss of the CD27+ subset. In Figure 1A, the frequency of gamma delta T cells is drastically decreased, and there is only a slight loss of the CD27+ subset.)

Yes, we agree these do like quite different and could be confusing. The lymph nodes from IKKΔT^CD2 lack αβ T cells and B cells, and so the cellularity is much lower than normal. Consequently, the percentage representation of remaining cells can be more noisy, while total cellularity calculations are more consistent. This is not an issue in the other strains that all have more cells in lymph nodes. We now show plots from spleen of the same mice which appear better aligned with additional splenic data shown in Figure 1.

Reviewer #2 (Public review):

(1) All approaches used confer changes to the entire T cell compartment. Therefore, the authors are unable to resolve whether the observations are mediated by direct and/or indirect effects (e.g., disorganized lymphoid architecture impacting maintenance/survival/homing).

We address this important point in the discussion (p11-12). The impacts of gene deletions upon αβ and 𝛾𝛿 T cells operate independently of one another (as also discussed in response to reviewer 1). For instance, the phenotype of αβ T cells is identical in IKKΔT^CD2 and IKKΔT^CD4 mice - 𝛾𝛿 T cells are only targeted in IKKΔT^CD2 mice. Similarly, the phenotype of 𝛾𝛿 T cells is similar in IKKΔT^CD2 vs Casp8.IKKΔT^CD2 strains. αβ T cells are absent from IKKΔT^CD2 but present in near normal numbers in Casp8.IKKΔT^CD2 mice. Others have also noted that 𝛾𝛿 T cell development is normal in Rag deficient mice (10.1126/science.1604321). In any case, an absence of αβ T cells is expected to promote 𝛾𝛿 T cell survival in the absence of competition for common utilised cytokines such as IL-7 and IL-15, though we do not see much evidence for this in mice with and without αβ T cells such as IKKΔT^CD2 vs Casp8. IKKΔT^CD2 strains. We do now discuss the potential contribution of trans-conditioning for type 17 𝛾𝛿 T cell development (p12).

(2) Assessment of factors that impact T cell numbers in the periphery is necessary. Are there observable changes to the proliferation, survival, and migration of gd T cell subsets?

In IKKΔT^CD2 and Casp8. IKKΔT^CD2 deficient strains, we infer a defect in survival, since they lack peripheral 𝛾𝛿 T cells, despite normal thymic development. Their absence made it hard to assess proliferation and migration, though 𝛾𝛿 T cells were absent from all lymphoid organs. The conclusions that defective survival is responsible for the absence of 𝛾𝛿 T cells in the different strains is also supported by the rescue of IKKΔT^CD2 and Casp8ΔT^CD2 strains by kinase dead RIPK1D138N. Furthermore, the presence of small numbers of residual populations in lymph nodes and spleen of IKKΔT^CD2 and Casp8ΔT^CD2 strains demonstrates that migration patterns were normal. Were cells unable to recirculate, they might be expected to fail to leave the thymus, or to accumulate in the spleen. We so no evidence of either of these scenarios.

(3) TCRd chain usage, especially among type 3 gd T cells, should be assessed.

We did not unfortunately, assess chain usage, choosing rather to rely of phenotypic identity of specific subsets, which we show in figure 5C, was extremely robust. IL-17 was only secreted by CD27– CD44^hi 𝛾𝛿 T cells, while IFN-gamma was only secreted by CD27+ CD44^hi 𝛾𝛿 T cells. We argue that the production of these key effector cytokines is the most direct test of a subsets functional identity and the phenotypic designation is robust.

(4) The functional consequences of IKK signaling on gd T cells were largely unaddressed. Cytokine analyses were performed only in the RIPK1D138N Casp8∆TCD2 model, leaving open the question of how canonical NF-κB-dependent signaling impacts the long-term functionality of gd T cells.

Yes, we agree this remains an open question around the transcriptional mechanisms by which NFκB signalling promotes cell survival, and one best addressed in future studies. We did not perform cytokine staining more widely, because the cytokine assay relies on short term re-stimulation of T cells with PMA and ionomycin. PMA activates PKC which in turn activates NF-κB signalling to elicit the cytokine response measured in this assay. As such, the results of such assays would be hard to interpret. We agree it would be interesting to investigate the functional consequences of REL deficiency in future studies, although this may need a more nuanced setting where 𝛾𝛿 T cells are not lost as a result of their defective survival.

(5) The authors suggest that Caspase 8 is required for the development and maintenance of type 3 gd T cells. While the authors discussed the limitations of assessing adult mice in interpreting the data, it seems like a relatively straightforward experiment to perform.

We did attempt these experiments with collaborators by analysing type 17 𝛾𝛿 T cell development in fetal thymic organ culture (FTOC). However, the GM mice are not so easy to breed and generating the large numbers of embryos required to set up the FTOCs proved too challenging and we were unable to generate these data.

(6) While analyses of Casp8∆TCD2 RIPK1D138N mice suggest that loss of adaptive and type 1 gamma delta T cells in Casp8∆TCD2 animals is due to necroptosis, the contribution of RIPK3 kinase activity remains unexamined. RIPK3 activity determines whether cells die via necroptosis or apoptosis in RIPK1/Caspase8-dependent signaling, and inclusion of this analysis would strengthen mechanistic insights.

Given time and resources, it would have been ideal to confirm necroptotic cell death by alternative knockouts, such as RIPK3 or MLKL. However, formation of the necrosome is dependent on kinase active RIPK1, since autophosphorylation of RIPK1 changes its conformation to allow recruitment of RIPK3 and MLKL and formation of the necrosome. Therefore, the rescue of CASPASE8 deficient T cells from cell death by kinase dead RIPK1 is very solid genetic evidence of necroptosis.

(7) Canonical NF-κB signaling through cRel alone was not evaluated, leaving a gap in the understanding of transcriptional pathways required for gd T cell subsets.

This was assessed in p105/RelA knockout strain, which only express cREL. What we lacked was an assessment of what RelA/p50 dimers can support in the absence of cREL. We do however, show the impact of RelA single deficiency, and RelA/p50 deficiency.

In truth, we had many REL deficient strains and it was challenging to make all the combinations we wanted. However, we try to compensate for this by discussing what cREL:cREL dimers and cREL:P50 dimers are capable of doing by analysing 𝛾𝛿 T cell development in p105/RELA DKO and RELA KO mice - these do show that cREL:P50 can compensate in the absence of RELA, but cREL:cREL cannot.

Reviewer #3 (Public review):

Weaknesses:

The paper would benefit greatly from a graphical abstract that could summarize the key findings, making the key findings accessible to the general immunology or biochemistry reader. Ideally, this graphic would distinguish the requirements for NF-κB signals sustaining thymic γδ T cell differentiation from peripheral maintenance, taking into account the various subsets and signaling pathways required. In addition, the authors should consider adding further literature comparing the requirements for NF-κB /necroptosis pathways in regulating other non-conventional T cell populations, such as iNKT, MAIT, or FOXP3+ Treg cells. These data might help position the requirements described here for γδ T cells compared to other subsets, with respect to homeostatic cues and transcriptional states.

Thank you - we have added such discussions. We are happy to add a graphical abstract if journal constraints permit this.

Last and least, there are multiple grammatical errors throughout the manuscript, and it would benefit from further editing. Likewise, there are some minor errors in figures (e.g., Figure 3A, add percentage for plot from IKKDT.RIPK1D138N mouse; Figure 7, “Adative").

Thank you !

Author response:

Public Reviews:

Reviewer #1 (Public review):

The central pair apparatus of motile cilia consists of two singlet microtubules, termed C1 and C2, each of which is associated with a set of projections, referred to as the C1 and C2 projections. Each projection comprises multiple distinct structural domains, designated a, b, c, and so on. Biochemical studies combined with genetic analyses in Chlamydomonas identified three proteins as the major components of the C2a projection, and subsequent cryo-EM studies confirmed these findings.

In this paper, the authors aim to study the homologues of these three proteins-CCDC108/CFAP65, CFAP70, and MYCBPAP/CFAP147-using knockout mouse models. Biochemical and cell biological analyses demonstrate that, as in Chlamydomonas, these proteins are components of the C2 projection and form a complex that depends on the presence of each other. In addition, the authors use affinity purification to identify two previously uncharacterized proteins and show that they are central pair apparatus proteins that associate with the aforementioned complex. Knockout mice lacking any of the three core proteins exhibit phenotypes consistent with primary ciliary dyskinesia (PCD).

Overall, the manuscript is clearly written, and the data are convincing and support the authors' conclusions. However, given the previous findings in Chlamydomonas, this work provides limited conceptual advances to the field. Nonetheless, it represents a useful and well-documented resource for understanding the conserved organization of the central pair apparatus in motile cilia. It will be of interest to cell and developmental biologists, biochemists, and clinicians studying and treating human ciliopathies.

We thank the reviewer for their positive comments on our work.

Reviewer #2 (Public review):

Summary:

This manuscript investigates the protein composition and functional role of the C2a projection of the central apparatus (CA) in vertebrate motile cilia. Using three knockout mouse models (Ccdc108, Mycbpap, and Cfap70), the authors demonstrate that these genes - homologs of Chlamydomonas FAP65, FAP147, and FAP70 - are required for normal motile cilia function in ependymal and tracheal multiciliated cells. Specifically, the authors show that:

(1) Knockout mice for each gene exhibit primary ciliary dyskinesia phenotypes (hydrocephalus and sinusitis), accompanied by abnormal ciliary motion and reduced ciliary beat frequency. 

(2) CCDC108, MYCBPAP, and CFAP70 physically interact and localize to the axonemal central lumen, consistent with the C2a projection. 

(3) Loss of any one of these proteins destabilizes the others and disrupts CA integrity in a tissue-specific manner. 

(4) ARMC3 and MYCBP are C2a-associated proteins. 

Strengths:

(1) Clarity: the results are presented in a coherent sequence that facilitates understanding of both the rationale and conclusions. 

(2) Genetic rigor: three independent knockout mouse lines that exhibit consistent motile cilia phenotypes provide in vivo support for the proposed role of these proteins. 

(3) Integration of structural and functional analyses: combination of ultrastructural (TEM) and immunofluorescence data with CBF measurements provides convincing correlation between structural defects and impaired ciliary function. 

(4) Mutual dependency model: reciprocal destabilization of CCDC108, MYCBPAP, and CFAP70 supports their interdependence in the C2a assembly. 

(5) Expansion of the vertebrate C2a proteome: the identification of ARMC3 and MYCBP as C2a-associated proteins provides a foundation for future mechanistic studies. 

We appreciate our reviewer's positive comments.

Weaknesses:

(1) Mechanistic depth: the data show a convincing correlation between C2a and ciliary function, but the cell type-specificity of CCDC108, MYCBPAP, and CFAP70 knockout effects is underdeveloped. This is an interesting observation that raises mechanistic/structural questions not addressed in the study, such as what is the role of C2a in CP nucleation, maintenance, or mechanical stabilization? Is C2a composition different in different cell types? 

We agree with our reviewer and value their insightful comments. Indeed, CP-MT defects, including the loss of one or both CP-MTs, were only observed in a subset of mouse ependymal cells (mEPCs) at day 10 post-serum starvation, and were rare in tracheal multiciliated cells, although the C2a projections were severely damaged in these tracheal cells. Based on these observations, we hypothesize that the loss of CP-MTs is probably a secondary effect caused by mechanical stress during ciliary movement. To investigate the role of C2a in CP-MT nucleation, maintenance, or mechanical stabilization, we plan to examine the axoneme structures of mEPCs at day 5 post-serum starvation using TEM. By comparing axoneme defects in these cells at days 5 and 10, we hope to gain insights into this question. Based on our findings and previous findings in Chlamydomonas, we speculate that the core components (CCDC108/FAP65, MYCBPAP/FAP147, and CFAP70/FAP70) of the C2a projection are highly conserved across species, but the peripheral associated C2a proteins may vary among different cell types. Therefore, we will perform co-immunoprecipitation using mEPCs and mouse tracheal epithelial cells to investigate potential cell-type-specific differences and expand the related discussion.

(2) Cell model choice: co-immunoprecipitation was performed using mouse testis lysates. While this is a reasonable source of CA proteins from flagellated cells, the functional analyses in this study focus on ependymal and tracheal multiciliated cells. It would therefore be helpful for the authors to clarify the extent to which these interactions are expected to be conserved across ciliated cell types, and to discuss potential tissue-specific differences in CA assembly.

We appreciate our reviewer's insightful comments. We will follow their suggestion and perform co-immunoprecipitation using mEPCs and mouse tracheal epithelial cells to investigate potential cell-type-specific differences and expand the related discussion.

(3) Statistical analysis: the manuscript states "Statistical significance was defined as P < 0.5", which is likely a typo, but should be P < 0.05. In general, the statistical methods require more clarification. In several figures (e.g., 2B, 2D, 5J, 5K), multiple knockout genotypes are compared with WT, yet unpaired t-tests are reported. When more than two groups are analyzed, multiple pairwise t-tests inflate Type I error unless appropriately corrected; a one-way ANOVA with post hoc comparisons (e.g., Dunnett's test for WT-referenced comparisons) would be more appropriate. Furthermore, the analysis of ciliary movement modes (Figure 2D) involves categorical data, for which a t-test is not statistically appropriate. These comparisons could instead be evaluated using chi-square or Fisher's exact tests. Addressing these issues is important to ensure accurate statistical inference.

We thank our reviewer for pointing out these errors. We will double-check our statistical results and perform new analyses following their suggestion.

(4) Methods section: does not sufficiently describe how image-based quantifications were performed. For example, the criteria used to define cilia number, basal body number, and rotational beating are not specified, nor is how CBF measurements were analyzed. The authors should also provide details regarding analysis software and imaging parameters used (and whether they were kept constant across genotypes). 

We apologize for overlooking these method details. We will expand the relevant method section to include this information.

Author response:

We thank the reviewer for the thoughtful and constructive evaluation of our work and for recognizing its potential interest to researchers working on cardiac development and regeneration. We are planning to address the specific concerns as noted by the reviewers in the following way:

Public Reviews:

Reviewer #1 (Public review):

Summary:

This manuscript addresses an important question in cardiac biology: whether distinct cardiomyocyte (CM) subpopulations play specialized roles during heart development and regeneration. Using single-cell RNA sequencing and newly generated genetic tools, the authors identify phlda2 as a specific marker of primordial cardiomyocytes in the adult zebrafish heart. They further show that these primordial CMs function are essential for myocardial morphogenesis and coronary vascularization but are dispensable for myocardial regeneration or revascularization after injury. These findings indicate that heart regeneration doesn't simply recapitulate developmental processes.

Strengths:

A major strength of the study is the generation of a phlda2 BAC reporter, which provides a specific and reliable marker for primordial cardiomyocytes. The lack of genetic tools has previously limited functional analysis of this CM population. By using phlda2 regulatory elements to generate reporter and NTR-based ablation lines, the authors can visualize and selectively manipulate primordial CMs in vivo. This enables a direct functional interrogation rather than relying on lineage tracing or correlative evidence. Through genetic ablation, the authors convincingly demonstrate that primordial CMs are essential for myocardial morphogenesis and coronary vascular organization during development but are not necessary for heart regeneration.

Weaknesses:

(1) The manuscript would benefit from clarifying whether the primordial cardiomyocytes ablation affects epicardial cell behaviors during heart development, given that the well-established role of the epicardium in supporting coronary vessel growth, it is possible that the vascular phenotypes observed after primordial CM ablation may be affected, at least in part, by altered epicardial cells.

We thank the reviewer for this thoughtful comment and agree that primordial cardiomyocyte ablation may indirectly affect coronary vessel growth through changes in epicardial cell behavior. Therefore, we will perform additional analyses to examine epicardial cell behaviors, including epicardial coverage and migration following primordial cardiomyocyte ablation using the established epicardial reporter line tcf21:nucEGFP during heart development.

(2) Because primordial cardiomyocytes form a dense, single-cell-thick layer covering the ventricular surface, it would be informative to determine whether their loss alters the spatial distribution or inward migration of coronary endothelial cells or epicardial cells.

We thank the reviewer for this important comment. We will analyze the spatial distribution and inward migration of coronary endothelial and epicardial cells after primordial cardiomyocyte ablation using high-resolution imaging and quantitative analysis

(3) The manuscript carefully examines the relationship between primordial CMs and gata4⁺ cardiomyocytes during regeneration. However, their relationship during heart development should be more fully addressed.

We appreciate the suggestion and will carefully investigate the relationship between primordial cardiomyocytes and gata4⁺ cardiomyocytes during heart development.

(4) As loss of cardiomyocytes is known to induce gata4:GFP activation during regeneration, it would be important to determine whether ablation of primordial cardiomyocytes alone triggers gata4:GFP expression in neighboring cardiomyocytes. This analysis would further support the conclusion that primordial cardiomyocytes are not required for regenerative responses.

We acknowledge the reviewer’s comments and will test whether primordial cardiomyocyte ablation induces gata4:GFP activation in neighboring cardiomyocytes in the adult heart.

Reviewer #2 (Public review):

Summary:

In the manuscript "Primordial Cardiomyocytes orchestrate myocardial morphogenesis and vascularization but are dispensable for regeneration", Sun et al. identify a novel marker of primordial cardiomyocytes and use it to visualize and ablate the population during development and regeneration. The role of the primordial layer has not been investigated because the tools to manipulate this population have not existed. The manuscript is straightforward, easy to understand, and addresses an important question that has not been explored.

While the manuscript provides important insights into the role of primordial CMs, backed by a convincing methodology, the authors should clarify their requirements for heart development and maturation. Specifically, is the primordial layer required for the fish to survive?

We thank the reviewer for this important question. We will examine the survival of fish following primordial cardiomyocyte ablation during development.

Do primordial CMs regenerate when ablated during development, and do the defects observed (in trabecular and compact CMs and coronary vessels) resolve after 10 days post-treatment when they were detected?

We thank the reviewer for this valuable comment. We will perform additional analyses to determine whether primordial cardiomyocytes regenerate after ablation during development and to assess the extent and dynamics of their recovery. We will also evaluate whether the defects in trabecular and compact myocardium and coronary vasculature persist or resolve in adult hearts following primordial cardiomyocyte ablation during development.

Reviewer #3 (Public review):

Summary:

The authors performed single-cell RNA sequencing of adult zebrafish hearts and identified markers for distinct cardiomyocyte subpopulations. One marker, phlda2, marks primordial cardiomyocytes. They generated transgenic reporter lines to characterize phlda2 expression patterns and a phlda2-NTR ablation line to determine the functional requirement of primordial cardiomyocytes during heart regeneration. They found that phlda2+ primordial cardiomyocytes are essential for myocardial morphogenesis and coronary vessel development. Interestingly, when phlda2+ primordial cardiomyocytes are ablated during heart regeneration, gata4+ cortical cardiomyocytes, coronary vessel revascularization, and scar tissue formation are not affected.

Strengths:

The authors identified a new primordial cardiomyocyte marker, phlda2. They further demonstrated that primordial cardiomyocytes are important for heart morphogenesis but dispensable for heart regeneration. Their findings reveal a potential difference between heart development and regeneration programs.

Weakness:

Despite the interesting findings, the authors did not provide supplemental data for their scRNAseq to demonstrate the data quality and support their conclusions, and some results are not well described.

We appreciate the reviewer’s comment. We will include supplemental data to demonstrate the quality of our single-cell RNA sequencing. Additionally, we will provide more detailed descriptions of the key results in the main text and figure legends to clearly support our conclusions regarding primordial cardiomyocytes and their roles in heart morphogenesis and regeneration.

Author response:

Public reviews:

Reviewer #1 (Public review):

In the manuscript entitled "Flexible and high-throughput simultaneous profiling of gene expression and chromatin accessibility in single cells," Soltys and colleagues present easySHARE-seq, a method described as an improvement upon SHARE-seq for the simultaneous measurement of RNA transcripts and chromatin accessibility.

The authors demonstrate the utility of easySHARE-seq by profiling approximately 20,000 nuclei from the murine liver, successfully annotating cell types and linking cis-regulatory elements to target genes. The authors claim that easySHARE-seq supports longer read lengths potentially enabling better variant discovery or allele-specific signal assessment, though they do not provide direct evidence to support these specific claims.

A key strength of the protocol is enhanced sequencing efficiency, achieved by shortening the Index 1 read from 99 to 17 nucleotides. This reduction does not come at a significant cost to barcode diversity, retaining approximately 3.5 million combinations. Additionally, the approach allows for the sequencing of a sub-library to assess quality prior to final barcoding and sequencing which seems quite clever.

While the increase in RNA transcript recovery is substantial, it appears to come at a cost: there is a notable decrease in ATAC fragments per cell compared to the original SHARE-seq (and other platforms). Likely as a result, the dimensionality reduction (UMAP) shows good resolution for RNA profiles but relatively poor resolution for accessibility profiles. Furthermore, the presented data suggests potential ambient RNA contamination; specifically, the detection of Albumin in HSCs and B cells is likely an artifact of the protocol rather than a biological signal.

Overall, the study is well-presented and represents a promising advance. However, there are significant shortcomings that should be addressed, particularly regarding "leaky" transcript recovery and reduced ATAC performance.

Recommendations:

(1) To provide a comprehensive view of the current field, the authors should include Scale Biosciences (Scale Bio) in their discussion of available commercial platforms.

(2) A head-to-head comparison with the 10x Genomics Multiome platform would be of significant interest to the single-cell genomics community and would better contextualize the performance of easySHARE-seq.

(3) Optimizing ATAC Performance: I strongly suggest exploring methods to improve ATAC sensitivity. As the authors note, the improvement in RNA recovery may result from fewer processing steps and stronger fixation. It would be valuable to test if decreasing fixation back to 2% (as in the original SHARE-seq) recovers ATAC data quality, and to determine if the fixation level or the number of steps is the key variable in preserving transcripts.

(4) The authors allude to the possibility of scaling this assay using a barcoded poly(T). Explicit inclusion or demonstration of this capability would dramatically increase interest in this protocol. Perhaps ATAC could be scaled using a barcoded Tn5?

(5) The number of HSCs and B cells expressing Albumin is problematic and suggests significant ambient RNA issues that need to be addressed or computationally corrected.

We thank reviewer #1 for his comments and critique. We will include a direct comparison of easySHARE-seq with the 10x Multiome platform by adding this comparison to Fig. 1 E&F and more directly point to Table 1 as a comparison of overall assay possibilities. We will also more explicitly state and describe the possibilities and limitations of how to scale this assay up. We also thank the reviewer for raising the possible issue of ambient RNA contamination. We aim to quantify ambient RNA contamination and explore its impact as well as possibilities to correct for it if needed. Unfortunately, external circumstances make it difficult to perform further wetlab experiments in order to optimize ATAC-seq performance. We will thus update our discussion to include possibilities on how to improve ATAC-seq data quality.

Reviewer #2 (Public review):

Aims:

The authors sought to optimize SHARE-seq, a multimodal single-cell method, to improve the simultaneous profiling of gene expression and chromatin accessibility. Their goal was to enhance barcode design for better sequencing efficiency and cost savings, while improving overall data quality. They then applied their optimized method, easySHARE-seq, to study liver sinusoidal endothelial cells (LSECs) to demonstrate its utility in examining gene regulation and spatial zonation.

Strengths:

The improved barcode design is an advance, increasing the proportion of sequencing reads dedicated to biological information rather than barcode identification. This modification offers practical benefits in terms of sequencing costs and read length, potentially reducing alignment errors. The method also demonstrates improved RNA detection compared to the original SHARE-seq protocol. The biological applications showcase how simultaneous measurement of both modalities enables analyses that would be practically impossible with single-modality approaches, particularly in examining how chromatin states change along developmental or spatial trajectories.

Weaknesses:

There is a notable reduction in chromatin accessibility detection compared to the original SHARE-seq method, likely limiting the broad use of the method. While the authors are transparent about this tradeoff, additional discussion would be helpful regarding how this affects data interpretation. Comparisons showing consistency between easySHARE-seq and SHARE-seq chromatin accessibility patterns at the single-cell level would strengthen confidence in the method.

We thank reviewer #2 for his comments and great suggestions for further analyses. We will emphasize ATAC-seq data quality issues further in our discussions and more explicitly discuss the resulting implications and shortcomings. We agree with reviewer #2 that this dataset allows exploration of enhancer logic. We aim to incorporate the suggested analyses regarding RNA-ATAC correlations, expand our exploration of enhancer biology and include these results in our revisions. We will also improve clarity of our zonation analysis procedure.

Overall:

The authors achieve their aim of creating an optimized protocol with improved barcode design and enhanced RNA detection. The method represents a useful advance for specific experimental contexts where the tradeoffs are appropriate.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

In the ecological interactions between wild plants and specialized herbivorous insects, structural innovation-based diversification of secondary metabolites often occurs. In this study, Agrawal et al. utilized two milkweed species (Asclepias curassavica and Asclepias incarnata) and the specialist Monarch butterfly (Danaus plexippus) as a model system to investigate the effects of two N,S-cardenolides - formed through structural diversification and innovation in A. curassavica-on the growth, feeding, and chemical sequestration of D. plexippus, compared to other conventional cardenolides. Additionally, the study examined how cardenolide diversification resulting from the formation of N,S-cardenolides influences the growth and sequestration of D. plexippus. On this basis, the research elucidates the ecophysiological impact of toxin diversity in wild plants on the detoxification and transport mechanisms of highly adapted herbivores.

Strengths:

The study is characterized by the use of milkweed plants and the specialist Monarch butterfly, which represent a well-established model in chemical ecology research. On one hand, these two organisms have undergone extensive co-evolutionary interactions; on the other hand, the butterfly has developed a remarkable capacity for toxin sequestration. The authors, building upon their substantial prior research in this field and earlier observations of structural evolutionary innovation in cardenolides in A. curassavica, proposed two novel ecological hypotheses. While experimentally validating these hypotheses, they introduced the intriguing concept of a "non-additive diversity effect" of trace plant secondary metabolites when mixed, contrasting with traditional synergistic perspectives, in their impact on herbivores.

Weaknesses:

The manuscript has two main weaknesses. First, as a study reliant on the control of compound concentrations, the authors did not provide sufficient or persuasive justification for their selection of the natural proportions (and concentrations) of cardenolides. The ratios of these compounds likely vary significantly across different environmental conditions, developmental stages, pre- and post-herbivory, and different plant tissues. The ecological relevance of the "natural proportions" emphasized by the authors remains questionable. Furthermore, the same compound may even exert different effects on herbivorous insects at different concentrations. The authors should address this issue in detail within the Introduction, Methods, or Discussion sections.

Second, the study was conducted using leaf discs in an in vitro setting, which may not accurately reflect the responses of Monarch butterflies on living plants. This limitation undermines the foundation for the novel ecological theory proposed by the authors. If the observed phenomena could be validated using specifically engineered plant lines-such as those created through gene editing, knockdown, or overexpression of key enzymes involved in the synthesis of specific N,S-cardenolides - the findings would be substantially more compelling.

Reviewer #2 (Public review):

This study examined the effects of several cardenolides, including N,S-ring containing variants, on sequestration and performance metrics in monarch larvae. The authors confirm that some cardenolides, which are toxic to non-adapted herbivores, are sequestered by monarchs and enhance performance. Interestingly, N,S-ring-containing cardenolides did not have the same effects and were poorly sequestered, with minimal recovery in frass, suggesting an alternate detoxification or metabolic strategy. These N,S-containing compounds are also known to be less potent defences against non-adapted herbivores. The authors further report that mixtures of cardenolides reduce herbivore performance and sequestration compared to single compounds, highlighting the important role of phytochemical diversity in shaping plant-herbivore interactions.

Overall, this study is clearly written, well-conducted and has the potential to make a valuable contribution to the field. However, I have one major concern regarding the interpretations of the mixture results. From what I understand of the methods, all tested mixtures contain all five compounds. As such, it is not possible to determine whether reduced performance and sequestration result from the complete mixture or from the presence of a single compound, such as voruscharin for performance and uscharin for sequestration. For instance, if all compounds except voruscharin (or uscharin) were combined, would the same pattern emerge? I suspect not, since the effects of the individual N,S-containing compounds alone are generally similar to those of the full mixture (Figure S3). By taking the average of all single compounds, the individual effects of the N,S-containing ones are being inflated by the non-N,S-containing ones (in the main text, Figure 4). In the mix, of course, they are not being 'diluted', as they are always present. This interpretation is further supported by the fact that in the equimolar mix, the relative proportion of voruscharin decreases (from 50% in the 'real mix'), and the target measurements of performance and sequestration tend to increase in the equimolar mix compared to the real mix.

Despite this issue, the discussion of mixtures in the context of plant defence against both adapted and non-adapted herbivores is fascinating and convincing. The rationale that mixtures may serve as a chemical tool-kit that targets different sets of herbivores is compelling. The non-N,S cardenolides are effective against non-adapted herbivores and the N,S-containing cardenolides are effective against adapted herbivores. However, the current experiments focus exclusively on an adapted species. It would be especially interesting to test whether such mixtures reduce overall herbivory when both adapted and non-adapted species are present.

It remains possible that mixtures, even in the absence of voruscharin or uscharin, genuinely reduce sequestration or performance; however, this would need to be tested directly to address the abovementioned concern.

Thanks for these insightful reviews and your summary assessment. We certainly agree that ours was a laboratory study with a single specialized insect, and both mixtures types had all five compounds (controlling for total toxin concentration). Thus, our conclusion that combined effects of naturally occurring toxins (within the cardenolide class) have non-additive effects for the specialized sequestering monarch are constrained by our experimental conditions. In our assay we used two mixture types, equimolar and “natural” proportions. We acknowledge that the natural proportions will vary with plant age, damage history, etc. of the host plant, Asclepias curassavica. Our proportions were based on growing the plants a few different times under variable conditions. Although we did not conduct these experiments on non-adapted insects, we discuss a related experiment that was conducted with wild-type and genetically engineered Drosophila (Lopez-Goldar et al. 2024, PNAS). In sum, we appreciate the reviewers’ comments.

Recommendations for the authors:

Reviewing Editor Comments:

(i) More convincingly justify the choice and ecological relevance of the "natural" cardenolide ratios, (ii) Clarify the interpretation of mixture effects, and (iii) more explicitly discuss the limitations of leaf-disc assays and the absence of non-adapted herbivores in light of the broader coevolutionary claims.

Thank you for these suggestions. We have added several sentences of text to the Discussion section to make these points.

Reviewer #1 (Recommendations for the authors):

(1) Statistical analysis is missing from Figure 3 and Figure S3, making it difficult to assess the significance of the data.

Much of the data in Fig. 3 is meant for descriptive presentation, with the main statistical analysis (contrast between N,S and non-N,S cardenolides given in the main text of the results. We have added treatment differences between the sequestration efficiencies to the figure as well.

(2) To help readers intuitively understand how certain results (such as ECD and sequestration efficiency) were calculated, the authors can provide the equations used for these computations.

Thank you, this was given in the methods and we have added it to the Result on first mention as well.

(3) For Figure 4, we suggest presenting the results of the equal mixture treatment and the realistic mixture treatment separately, rather than averaging the results from these two types of treatments.

We understand and appreciate this comment – all of the treatment means are given in Fig. S3. For this particular figure we have opted to stick with the binary comparison (singles vs. mixed) to maximize replication for statistical tests (typically n = 25 vs. 10).

Reviewer #2 (Recommendations for the authors):

Given the interpretations and discussion generally, I feel the manuscript would benefit from either additional experiments (mixtures w/o N-S compounds), inclusion of non-adapted herbivore performance, or reframing of the explicit interpretations from your findings.

We have added some caveats to the text but not added any additional experiments.

Also, for all treatments/mixtures are concentrations above the IC50? Perhaps this could be calculated from the information presented, but it may be best to explicitly mention this.

This is an interesting question. IC50’s are estimated from in vitro assays (with the enzyme and toxins in microplate wells) and so are not translatable to foliar concentrations. As indicated in the text, we chose cardenolide levels based on foliar concentrations to match A. curassavica.

Some minor points:

(1) Although the intact N,S-ring-containing compounds are recovered in low amounts in frass (and not sequestered), is there evidence of N,S-ring components being otherwise traceable in the frass? For example, can excess S or N be detected in frass? This could provide insight into differential detoxification or reincorporation of these elements, potentially explaining variation between voruscharin and uscharin.

Great question! We have not been able to detect breakdown projects. In other experiments we have conducted mass spectrometric analysis of bodies and frass, but have not been able to find the features representing breakdown products. Nonetheless, as mentioned below, the main conversion products are evident and measurable, as in this study.

(2) As a point of curiosity, is there evidence of interconversion between such compounds? For instance, if monarchs are fed only voruscharin, can other cardenolides be detected in their tissues?

Yes, we have tried to make this more clear in the text. Both uscharin and voruscharin are converted to calotropin and calactin.

Author response:

General Statements

Our study provides important mechanistic insights into how the perinuclear actomyosin network PANEM facilitates the interaction of unfavorably positioned chromosomes, i.e. peripheral and polar chromosomes, with the mitotic spindle in early mitosis to ensure their correct segregation in subsequent anaphase. All reviewers agree that our study makes important contribution to the field of mitosis and chromosome segregation. They make positive comments on our manuscript, for example, ‘The work highlights the PANEM as a key spatial and temporal element of chromosome congression’, ‘The work is an excellent addition to the field’, and ‘the concept of PANEM could be integrated into textbooks and models of chromosome congression’. All three reviewers also acknowledge the high quality of the data, rigorous and accurate analyses, and convincing quantification in our study. Reviewers 1 and 3 give several comments and suggestions for revision of our manuscript. Please find our point-by-point revision plan of the manuscript from page 3.

Description of the planned revisions

Reviewer #1 (Evidence, reproducibility and clarity):

Summary

Sheidaei and colleagues report a novel and potentially important role for an early mitotic actomyosin-based mechanism, PANEM contraction, in promoting timely congression of chromosomes located at the nuclear periphery, particularly those in polar positions. The manuscript will interest researchers studying cell division, cytoskeletal dynamics, and motor proteins. Although some data overlap with the group's prior work, the authors extend those findings by optimizing key perturbations and performing more detailed analyses of chromosome movements, which together provide a clearer mechanistic explanation. The study also builds naturally on recent ideas from other groups about how chromosome positioning influences both early and later mitotic movements.

In its current form, however, the manuscript is not acceptable for publication. It suffers from major organizational problems, an overcrowded and confusing Results section and figures, and a lack of essential experimental controls and contextual discussion. These deficiencies make it difficult to evaluate the data and the authors' conclusions. A substantial structural revision is required to improve clarity and persuasiveness. In addition, several key control experiments and more conceptual context are needed to establish the specificity and relevance of PANEM relative to other microtubule- and actin-based mitotic mechanisms. Testing PANEM in additional cell lines or contexts would also strengthen the claim. I therefore recommend Major Revision, addressing the structural, conceptual, and experimental issues detailed below.

Major Comments

A. Structural overhaul and figure reorganization

The Results section is overly dense, lacks clear structure, and includes descriptive content that belongs in the Methods. Many figure panels should be moved to Supplementary Materials. A substantial reorganization is required to transform the manuscript into a focused, "Reports"-type article.

Figure 4I: This panel is currently unclear and should be drastically simplified.

We will follow this suggestion and simplify this figure. For example, we plan to remove the column of “Start” because it is obvious and does not provide much new information.

I recommend to reorganize figures as follows:

Figure I: Keep as single figure but simplify. Figure 1D and 1E could be combined, move unnormalized SCV to supplementary materials. Same goes for 1F.

We will follow this suggestion and reorganize Figure 1 accordingly.

New Figure 4: Combine Figures 7A, 7B, 7D, 7E, 7F, expanded Supplementary Figure S7, and new data to demonstrate that PANEM actively pushes peripheral chromosomes inward which is important for efficient chromosome congression in diverse cellular contexts.

As suggested, we will conduct new experiments to demonstrate the role of PANEM in diverse cellular contexts, as detailed below. We will then combine the new results with Figure S7 to make the new Figure 8.

On the other hand, in our view, combining Figure 7A-E and the extended Figure S7 would be confusing because the two parts address different topics. Although we respect this suggestion from the reviewer, we would like to keep Figure 7 and the extended Figure S7 (i.e. Figure 8) separate.

C. Expansion of PANEM functional analysis

To strengthen the conclusions and broaden the study beyond the group's previous work, PANEM function should be tested in additional contexts (some may be considered optional but important for broader impact): [underlined by authors]

Test PANEM function in at least one additional cell line that displays PANEM to rule out cellline-specific effects.

As suggested, we will study the effect of PANEM contraction in one or two additional cell lines that form PANEM during prophase. For example, we plan to inhibit the PANEM contraction and study the outcome, focusing on the generation of polar chromosomes, which is the major defect after the inhibition of PANEM contraction in U2OS cells.

Evaluate PANEM contraction role in unsynchronized U2OS cells, where centrosome separation can occur before NEBD in a subset of cells (Koprivec et al., 2025), and in other cell types with variable spindle elongation timing.

As suggested, we will investigate the outcome (e.g. generation of polar chromosomes) of reduced PANEM contraction in unsynchronized U2OS cells, and address whether the two subsets of cells, where centrosomes’ separation occurs before and after NEBD, show any difference in the outcome.

D. Conceptual integration in Introduction and Discussion

The manuscript should better situate its findings within the context of early mitotic chromosome movements:

Clearly state in the Introduction and elaborate in the Discussion that initiation of congression is coupled to biorientation (Vukušić & Tolić, 2025). This provides essential context for how PANEM-mediated nuclear volume reduction supports efficient congression of polar chromosomes.

To explain the new interpretation of our results more clearly, we plan to add a new diagram to a supplemental figure in the revised manuscript.

Minor Comments

Sixth subheading (currently in Discussion): Move the final paragraph of the Discussion into the Results and expand it with preliminary analyses linking PANEM contraction to congression efficiency across untreated cell types or under mild nocodazole treatment.

As suggested, we will move the final paragraph of the Discussion to make a new final section in the Results. Moreover, as suggested, we will study the outcome of inhibiting PANEM contraction in cell lines other than U2OS, and add the results to the new final section in the Results.

Significance

Advance

This study's main strength is its novel and potentially important demonstration that contraction of PANEM, a peripheral actomyosin network that operates contracts early mitosis, contributes to the timely initiation of chromosome congression, especially for polar chromosomes. While PANEM itself was previously described by this group, this manuscript provides new mechanistic evidence, improved perturbations, and detailed chromosome tracking. To my knowledge, no prior studies have mechanistically connected this contraction to polar chromosome congression in this level of detail. The work complements dominant microtubule-centric models of chromosome congression and introduces actomyosin-based forces as a cooperating system during very early mitosis. However, the impact of the study is currently limited by major organizational issues, insufficient controls, and incomplete contextualization within existing literature. Addressing these issues will substantially improve clarity and credibility. [underlined by authors]

We have addressed or will address the underlined criticisms as detailed above.

Audience

Primary audience of this study will be researchers working in cell division, mitosis, cytoskeleton dynamics, and motor proteins. The findings may interest also the wider cell biology community, particularly those studying chromosome segregation fidelity, spindle mechanics, and cytoskeletal crosstalk. If validated and clarified, the concept of PANEM could be integrated into textbooks and models of chromosome congression and could inform studies on mitotic errors and cancer cell mechanics.

Expertise

My expertise lies in kinetochore-microtubule interactions, spindle mechanics, chromosome congression, and mitotic signaling pathways.

Reviewer #2 (Evidence, reproducibility and clarity):

In this manuscript, Sheidaei et al. reported on their study of chromosome congression during the early stages of mitotic spindle assembly. Building on their previous study (ref. #15, Booth et al., Elife, 2019), they focused on the exact role of the actin-myosin-based contraction of the nuclear envelope. First, they addressed a technical issue from their previous study, finding a way to specifically impair the actomyosin contraction of the nuclear membrane without affecting the contraction of the plasma membrane. This allowed them to study the former more specifically. They then tracked individual kinetochores to reveal which were affected by nuclear membrane contraction and at what stage of displacement towards the metaphase plate. The investigation is rigorous, with all the necessary controls performed. The images are of high quality. The analyses are accurate and supported by convincing quantifications. In summary, they found that peripheral chromosomes, which are close to the nuclear membrane, are more influenced by nuclear membrane contraction than internal chromosomes. They discovered that nuclear membrane contraction primarily contributes to the initial displacement of peripheral chromosomes by moving them towards the microtubules. The microtubules then become the sole contributors to their motion towards the pole and subsequently the midplane. This step is particularly critical for the outermost chromosomes, which are located behind the spindle pole and are most likely to be missegregated.

Significance

While the conclusions are somewhat intuitive and could be considered incremental with regard to previous works, they are solid and improve our understanding of mitotic fidelity. The authors had already reported the overall role of nuclear membrane contraction in reducing chromosome missegregation in their previous study, as mentioned fairly and transparently in the text. However, the reason for this is now described in more detail with solid quantification. Overall, this is good-quality work which does not drastically change our understanding of chromosome congression, but contributes to improving it. Personally, I am surprised by the impact of such a small contraction (of around one micron) on the proper capture of chromosomes and wonder whether the signalling associated with the contraction has a local impact on microtubule dynamics. However, investigating this point is clearly beyond the scope of this study, which can be published as it is. [underlined by authors]

The suggested topic (underlined) is intriguing. However, we agree with the reviewer that it is beyond the scope of this paper. The reviewer recommends publication of our manuscript as it is. So, we do not plan a revision based on this reviewer’s comments.

Reviewer #3:

Sheidaei et al., report how chromosomes are brought to positions that facilitate kinetochoremicrotubule interactions during mitosis. The study focusses on an important early step of the highly orchestrated chromosome segregation process. Studying kinetochore capture during early prophase is extremely difficult due to kinetochore crowding but the team has taken up the challenge by classifying the types of kinetochore movements, carefully marking kinetochore positions in early mitosis and linking these to map their fate/next-positions over time. The work is an excellent addition to the field as most of the literature has thus far focussed on tracking kinetochore in slightly later stages of mitosis. The authors show that the PANEM facilitates chromosome positioning towards the interior of the newly forming spindle, which in turn facilitates chromosome congression - in the absence of PANEM chromosomes end up in unfavourable locations, and they fail to form proper kinetochore-microtubule interactions. The work highlights the perinuclear actomyosin network in early mitosis (PANEM) as a key spatial and temporal element of chromosome congression which precedes the segregation process.

Major points

(4) The work has high quality manual tracking of objects in early mitosis- if this would be made available to the field, it can help build AI models for tracking. The authors could consider depositing the tracking data and increasing the impact of their work.

As suggested, we will include kinetochore tracking data as supplemental data in the revised manuscript.

Minor points

(2) Discussion point: If cells had not separated their centrosomes before NEBD, would PANEM still be effective? Perhaps the cancer cell lines or examples as shown in Figure 6A have some clues here.

The same question has been raised by Reviewer #1’s major point. We will undergo new experiments to directly address this question in a revised manuscript. If we do not obtain interpretable results, we will discuss this issue further in the Discussion, as suggested.

(3) Figure 7 cartoon shows misalignment leading to missegregation. It may be useful to consider this in the context of the centrosome directed kinetochore movements via pivoting microtubules. Is this process blocked in azBB-treated cells?

This issue is closely relevant to point 2 above. As discussed above, we will first address this issue experimentally. If we do not obtain interpretable results, we will discuss this issue further in the Discussion.

Description of the revisions that have already been incorporated in the transferred manuscript

Reviewer #1 (Evidence, reproducibility and clarity):

Summary

Sheidaei and colleagues report a novel and potentially important role for an early mitotic actomyosin-based mechanism, PANEM contraction, in promoting timely congression of chromosomes located at the nuclear periphery, particularly those in polar positions. The manuscript will interest researchers studying cell division, cytoskeletal dynamics, and motor proteins. Although some data overlap with the group's prior work, the authors extend those findings by optimizing key perturbations and performing more detailed analyses of chromosome movements, which together provide a clearer mechanistic explanation. The study also builds naturally on recent ideas from other groups about how chromosome positioning influences both early and later mitotic movements.

In its current form, however, the manuscript is not acceptable for publication. It suffers from major organizational problems, an overcrowded and confusing Results section and figures, and a lack of essential experimental controls and contextual discussion. These deficiencies make it difficult to evaluate the data and the authors' conclusions. A substantial structural revision is required to improve clarity and persuasiveness. In addition, several key control experiments and more conceptual context are needed to establish the specificity and relevance of PANEM relative to other microtubule- and actin-based mitotic mechanisms. Testing PANEM in additional cell lines or contexts would also strengthen the claim. I therefore recommend Major Revision, addressing the structural, conceptual, and experimental issues detailed below.

Major Comments

A. Structural overhaul and figure reorganization

The Results section is overly dense, lacks clear structure, and includes descriptive content that belongs in the Methods. Many figure panels should be moved to Supplementary Materials. A substantial reorganization is required to transform the manuscript into a focused, "Reports"-type article.

Remove repetitive statements that simply restate that later phenotypes arise as consequences of delayed Phase 1 (applicable to subheadings 3 onward).

As suggested, we have removed the statement for the delayed start of Phase 2 for peripheral kinetochores in azBB-treated cells (Page 9, second paragraph). We have also simplified the statement for the delayed start of Phase 3 and Phase 4 to avoid repetition (Page 9, third paragraph; Page 10, second paragraph).

B. Specificity and redundancy of actin perturbation

To establish the specificity and relevance of PANEM, the authors should include or discuss appropriate controls:

Apply global actin inhibitors (e.g., cytochalasin D, latrunculin A) to disrupt the entire actin cytoskeleton. These perturbations strongly affect mitotic rounding and cytokinesis but only modestly influence early chromosome movements, as reported previously (Lancaster et al., 2013; Dewey et al., 2017; Koprivec et al., 2025). The minimal effect of global inhibition must be addressed when proposing a localized actomyosin mechanism. Comment if the apparent differences in this approach and one that the authors were using arises due to different cell types.

We did experiments along this line, using a dominant-negative LINC construct, in our previous study (Booth et al eLife 2019). LINC-DN should more specifically remove/reduce PANEM than the global actin inhibitors mentioned above. LINC-DN attenuated the reduction of CSV soon after NEBD and increased the number of polar chromosomes (Booth et al eLife 2019); i.e. in this regard, the outcome was similar to azBB treatment in the current study. One can expect that global actin inhibitors would also inhibit the PANEM formation and show effects similar to LINC-DN. By contrast, the indicated references reported that global actin inhibitors strongly affect mitotic rounding and cytokinesis but only modestly influence early chromosome movements, as pointed out by the reviewer. Such a difference may have arisen due to different cell types (e.g. some cells form the PANEM and others do not: Figure S7), a different extent in the inhibition of PANEM formation, and/or the inhibition of cell rounding and cytokinesis (e.g. if cytokinesis is more sensitive to inhibitors than is the PANEM formation, we may not observe the possible effects on early chromosome movements due to PANEM inhibition while cytokinesis is still affected). As suggested, we discussed this topic in the Discussion (page 15, second paragraph). 

Clarify why spindle-associated actin, especially near centrosomes, as reported in prior studies using human cultured cells (Kita et al., 2019; Plessner et al., 2019; Aquino-Perez et al., 2024), was not observed in this study. The Myosin-10 and actin were also observed close to centrosomes during mitosis in X.laevis mitotic spindles (Woolner et al., 2008). Possible explanations include differences in fixation, probe selection, imaging methods, or cell type. Note that some actin probes (e.g., phalloidin) poorly penetrate internal actin, and certain antibodies require harsh extraction protocols. Comment on possibility that interference with a pool of Myo10 at the centrosomes is important for effects on congression.

As the reviewer implies, we cannot rule out that we could not detect actin associated with the spindle or centrosomes because of the difference in methods or cell lines between the current study and the literature mentioned by the reviewer. We have therefore moderated our claim in the Discussion that ‘we did not detect any actin network inside the nucleus, on the spindle or between chromosomes’ by adding ‘at least, using the method and the cell line in the current study’ to this statement (Page 13, second paragraph). We have also cited the three references mentioned by the reviewer in the Discussion (Page 13, second paragraph). Regarding Myosin10, azBB (blebbistatin variant) should have negligible effects on class-X myosin, including Myosin-10 (Limouze et al 2004 [PMID 15548862]). It is therefore unlikely that the effects of azBB that we observed in the current study are due to the inhibition of Myosin-10. We have cited Woolner et al 2008 and another paper and discussed this topic in the Discussion (Page 13, second paragraph).

C. Expansion of PANEM functional analysis

Quantify not only the percentage of affected cells after azBB but also the number of chromosomes per cell with congression defects in the current and future experiments.

It is tricky to count the number of chromosomes because they frequently overlap. Counting kinetochores is more feasible, but kinetochore signals show some non-specific background (e.g. those outside of the nucleus in prophase). We therefore quantified the chromosome volume at polar regions in azBB-treated cells (Figure 6C).

D. Conceptual integration in Introduction and Discussion

The manuscript should better situate its findings within the context of early mitotic chromosome movements:

Clearly state in the Introduction and elaborate in the Discussion that initiation of congression is coupled to biorientation (Vukušić & Tolić, 2025). This provides essential context for how PANEM-mediated nuclear volume reduction supports efficient congression of polar chromosomes.

It has been a widely accepted view in the field that chromosome congression precedes biorientation, since the publication in 2006 (Kapoor et al Science 2006). Very recently, this view has been challenged by the new publication (Vukušić & Tolić, Nat comm 2025), as indicated by this reviewer. We have mentioned this new model and discussed the new interpretation of our results based on this new model, in the Discussion (page 14; ‘It has been a widely accepted view…’).

To explain the new interpretation of our results more clearly, we plan to add a new diagram to a supplemental figure in the revised manuscript.

Explain that PANEM is most critical for polar chromosomes because their peripheral positions are unfavorable for rapid biorientation (Barišić et al., 2014; Vukušić & Tolić, 2025).

We have included such a statement in the Discussion, as a part of the new interpretation of our results based on the new model that chromosome biorientation precedes congression (see above). We have also cited the indicated two papers.

Discuss how cell lines lacking PANEM (e.g., HeLa and others) nonetheless achieve efficient congression, and what alternative mechanisms compensate in the absence of PANEM. For example, it is well established that cells congress chromosomes after monastrol or nocodazole washout, which essentially bypasses the contribution of PANEM contraction.

Following this suggestion, we discussed three possible mechanisms that could compensate for a lack of PANEM and facilitate kinetochore-MT interaction and chromosome congression, based on previous literature (Page 16): 1) the enhanced assembly rate of spindle MTs may facilitate kinetochore-MT interactions in N-CIN+ cancer cells, 2) chromosome biorientation may precede congression more frequently to promote the congression towards the spindle midplane, and 3) the balance between CENP-E, Dynein and chromokinesin’s activities may incline to greater chromosome-arm ejection forces towards the spindle midplane.

Minor Comments

These issues are more easily addressable but will significantly improve clarity and presentation.

Introduction

Remove the reference to Figure 1A in the Introduction. The portion of Figure 1 and related text that recapitulates the authors' previous work should be incorporated into the Introduction, not the Results.

As suggested in the second sentence of this comment, we have moved most of the second paragraph of the first section of Results to Introduction (Page 4) and cited Figure 1A and 1B in Introduction. We would like to keep the reference to Figure 1A in the Introduction, because showing the PANEM images at the beginning of the manuscript would help readers’ understanding of our study. In addition, citing Figure 1A in the Introduction is more consistent with the suggestion in the second sentence of this comment.

Results (by subheading)

First subheading: When introducing the ~8-minute early mitotic interval, cite additional studies that have characterized this period: Magidson et al., 2011 (Cell); Renda et al., 2022 (Cell Reports); Koprivec et al., 2025 (bioRxiv); Vukušić & Tolić, 2025 (Nat Commun); Barišić et al., 2013 (Nat Cell Biol).

As suggested, we cited these references at the indicated part of the first section of the Results (page 5).

Second subheading: Cite key reviews and foundational research on kinetochore architecture and sequential chromosome movement during early mitosis: Mussachio & Desai, 2017

(Biology); Itoh et al., 2018 (Sci Rep); Magidson et al., 2011 (Cell); Vukušić & Tolić, 2025 (Nat Commun); Koprivec et al., 2025 (bioRxiv); Rieder & Alexander, 1990 (J Cell Biol); Skibbens et al., 1993 (J Cell Biol); Kapoor et al., 2006 (Science); Armond et al., 2015 (PLoS Comput Biol); Jaqaman et al., 2010 (J Cell Biol).

Rieder & Alexander, 1990 (J Cell Biol) and Kapoor et al., 2006 (Science) have already been cited in the second section of the Results in the original manuscript. We agree that all other references should be cited in this manuscript, and they are now cited in the Introduction and/or Discussion where they fit best (e.g. Mussachio & Desai 2017 reviews the kinetochore in general and is therefore best cited in the Introduction).

Third subheading: Clarify why some kinetochores on Figure 3A appear outside the white boundaries if these boundaries are intended to represent the nuclear envelope.

We interpret that these are background signals in the cytoplasm, which do not come from kinetochores, because 1) before NEBD, they were outside of the nucleus, and 2) after NEBD, they did not show any characteristic kinetochore motions such as those towards a spindle pole (Phase 2) and the spindle mid-plane (Phase 4). We have commented on these background signals in the legend for Figure 3A.

Fifth subheading: Cite studies on polar chromosome movements: Klaasen et al., 2022 (Nature); Koprivec et al., 2025 (bioRxiv). Clarify that Figure 5F displays only those kinetochores that initiated directed congression movements.

These two references have already been cited and discussed in this Result section of our original manuscript. However, considering this suggestion, we have discussed more about polar chromosome movements reported by Koprivec et al (page 11). Meanwhile, the reviewer is correct about Figure 5F, and we have clarified this point in the Figure 5F legend.

Discussion

When discussing cortical actin, cite key reviews on its presence and function during mitosis:

Kunda & Baum, 2009 (Trends Cell Biol); Pollard & O'Shaughnessy, 2019 (Annu Rev Biochem); Di Pietro et al., 2016 (EMBO Rep).

As suggested, we have cited all these review papers in the Discussion (page 15), and mentioned the role of the cortical actin on the spindle orientation and positioning (Kunda & Baum, 2009; Di Pietro et al., 2016), as well as the function of the actomyosin ring on cytokinesis (Pollard & O'Shaughnessy, 2019).

Significance

Advance

This study's main strength is its novel and potentially important demonstration that contraction of PANEM, a peripheral actomyosin network that operates contracts early mitosis, contributes to the timely initiation of chromosome congression, especially for polar chromosomes. While PANEM itself was previously described by this group, this manuscript provides new mechanistic evidence, improved perturbations, and detailed chromosome tracking. To my knowledge, no prior studies have mechanistically connected this contraction to polar chromosome congression in this level of detail. The work complements dominant microtubule-centric models of chromosome congression and introduces actomyosin-based forces as a cooperating system during very early mitosis. However, the impact of the study is currently limited by major organizational issues, insufficient controls, and incomplete contextualization within existing literature. Addressing these issues will substantially improve clarity and credibility. [underlined by authors]

We have addressed or will address the underlined criticisms as detailed above.

Audience

Primary audience of this study will be researchers working in cell division, mitosis, cytoskeleton dynamics, and motor proteins. The findings may interest also the wider cell biology community, particularly those studying chromosome segregation fidelity, spindle mechanics, and cytoskeletal crosstalk. If validated and clarified, the concept of PANEM could be integrated into textbooks and models of chromosome congression and could inform studies on mitotic errors and cancer cell mechanics.

Expertise

My expertise lies in kinetochore-microtubule interactions, spindle mechanics, chromosome congression, and mitotic signaling pathways.

Reviewer #2 (Evidence, reproducibility and clarity):

In this manuscript, Sheidaei et al. reported on their study of chromosome congression during the early stages of mitotic spindle assembly. Building on their previous study (ref. #15, Booth et al., Elife, 2019), they focused on the exact role of the actin-myosin-based contraction of the nuclear envelope. First, they addressed a technical issue from their previous study, finding a way to specifically impair the actomyosin contraction of the nuclear membrane without affecting the contraction of the plasma membrane. This allowed them to study the former more specifically. They then tracked individual kinetochores to reveal which were affected by nuclear membrane contraction and at what stage of displacement towards the metaphase plate. The investigation is rigorous, with all the necessary controls performed. The images are of high quality. The analyses are accurate and supported by convincing quantifications. In summary, they found that peripheral chromosomes, which are close to the nuclear membrane, are more influenced by nuclear membrane contraction than internal chromosomes. They discovered that nuclear membrane contraction primarily contributes to the initial displacement of peripheral chromosomes by moving them towards the microtubules. The microtubules then become the sole contributors to their motion towards the pole and subsequently the midplane. This step is particularly critical for the outermost chromosomes, which are located behind the spindle pole and are most likely to be missegregated.

Significance

While the conclusions are somewhat intuitive and could be considered incremental with regard to previous works, they are solid and improve our understanding of mitotic fidelity. The authors had already reported the overall role of nuclear membrane contraction in reducing chromosome missegregation in their previous study, as mentioned fairly and transparently in the text. However, the reason for this is now described in more detail with solid quantification. Overall, this is good-quality work which does not drastically change our understanding of chromosome congression, but contributes to improving it. Personally, I am surprised by the impact of such a small contraction (of around one micron) on the proper capture of chromosomes and wonder whether the signalling associated with the contraction has a local impact on microtubule dynamics. However, investigating this point is clearly beyond the scope of this study, which can be published as it is. [underlined by authors]

The suggested topic (underlined) is intriguing. However, we agree with the reviewer that it is beyond the scope of this paper. The reviewer recommends publication of our manuscript as it is. So, we do not plan a revision based on this reviewer’s comments.

Reviewer #3:

Sheidaei et al., report how chromosomes are brought to positions that facilitate kinetochoremicrotubule interactions during mitosis. The study focusses on an important early step of the highly orchestrated chromosome segregation process. Studying kinetochore capture during early prophase is extremely difficult due to kinetochore crowding but the team has taken up the challenge by classifying the types of kinetochore movements, carefully marking kinetochore positions in early mitosis and linking these to map their fate/next-positions over time. The work is an excellent addition to the field as most of the literature has thus far focussed on tracking kinetochore in slightly later stages of mitosis. The authors show that the PANEM facilitates chromosome positioning towards the interior of the newly forming spindle, which in turn facilitates chromosome congression - in the absence of PANEM chromosomes end up in unfavourable locations, and they fail to form proper kinetochore-microtubule interactions. The work highlights the perinuclear actomyosin network in early mitosis (PANEM) as a key spatial and temporal element of chromosome congression which precedes the segregation process.

Major points

(1) The complexity of tracking has been managed by classifying kinetochore movements into 4 categories, considering motions towards or away from the spindle mid-plane. While this is a very creative solution in most cases, there may be some difficult phases that involve movement in both directions or no dominant direction (eg Phase3-like). It is unclear if all kinetochores go through phase1, 2, 3 and 4 in a sequential or a few deviate from this pattern. A comment on this would be helpful. Also, it may be interesting to compare those that deviate from the sequence, and ask how they recover in the presence and absence of azBB.

To respond to this comment, we would like to first clarify how we selected kinetochores for our analysis. We selected kinetochores that can be individually tracked. If kinetochore tracking was difficult (before the start of Phase 4 in control and azBB-treated cells or before observing the extended Phase 3 in azBB-treated cells) because of kinetochore crowding, we did not choose such kinetochores. We also did not include kinetochores close to spindle poles (within 4 µm) at NEBD in our analysis for the following two reasons: First, these kinetochores often did not show clear and rapid movements towards a spindle pole, which we used to define Phase 2. Second, although we referred to kinetochore co-localization with a microtubule signal for the start of Phase 2, this was difficult for kinetochores close to spindle poles because of a high density of microtubules. As requested, we have added this comment to the Method section (page 23).

With the above selection, all selected kinetochores without azBB treatment (control) showed the poleward motion (Phase 2) and congression (Phase 4) in this order, though their extents were varied among kinetochores. All selected kinetochores with azBB treatment also showed the poleward motion (Phase 2), and some of them showed congression (Phase 4) after Phase 2. Then, Phase 1 and Phase 3 were defined as intervals between NEBD and Phase 2 and between Phase 2 and Phase 4, respectively. If no Phase 4 was observed with azBB, we judged that Phase 3 continued till the end of tracking. We have added this comment to the Method section (page 23-24).

(2) Would peripheral kinetochore close to poles behave differently compared to peripheral kinetochore close to the midplane (figure S4)? In figure 3D, are they separated? If not, would it look different?

Since we did not include kinetochores close to spindle poles (at NEBD), for which it was difficult to define Phase 2 (see our response to the above major point 1), in our analysis, the suggested comparison is not feasible.

(3) Uncongressed polar chromosomes (eg., CENPE inhibited cells) are known to promote tumbling of the spindle. In figure 5B with polar chromosomes, it will be helpful to indicate how the authors decouple spindle pole movements from individual kinetochore movements.

In contrast to CENPE-inhibited cells, azBB-treated cells did not show much tumbling of the spindle, though both cells showed uncongressed polar chromosomes. The reason for this difference may be fewer uncongressed polar chromosomes in azBB-treated cells. There were still modest spindle motions in azBB-treated cells. However, because kinetochore motions were assessed relative to a spindle pole (and other reference points on the spindle) in our study (Figure 2A, C), the modest spindle motions were offset in our analyses of kinetochore motions. We have clarified the underlined part in the Method section (page 22).

Minor points

(1) It will be helpful for readers to see how many kinetochores/cell were considered in the tracking studies. Figure legends show kinetochore numbers but not cell numbers.

As suggested, we have now mentioned the number of cells, where the kinetochore motions were analyzed, in the legends for Figures 3, 4, 5, S4 and S5.

(4) Are all the N-CIN- lines with PANEM highly sensitive to azBB? In other words, is PANEM essential for normal congression in some of these lines.

We checked the sensitivity of cell lines in Figure S7B to blebbistatin (the original form of azBB) on DepMap. There was no plausible difference between PANEM+ and PANEM- cell lines, although the blebbistatin sensitivity data were available only for 4 cell lines (HCT116, MCF7, U2OS and HT29) in Figure S7B. Nonetheless, because blebbistatin could kill cells by inhibiting cytokinesis, the blebbistatin sensitivity may not necessarily reflect how essential the PANEM contraction is for chromosome congression.

(5) Are congression times delayed in lines that naturally lack PANEM?

For example, it takes 10-20 min for HeLa cells (lacking PANEM) to complete chromosome congression after the NEBD (Bancroft et al 2025: https://doi.org/10.1242/jcs.163659). This is not significantly different from the time (8-18 min) for chromosome congression we observed in U2OS cells (forming PANEM). We assume that cells lacking PANEM have developed a compensatory mechanism for efficient chromosome congression – we have newly discussed possible compensatory mechanisms in the last paragraph of the Discussion (page 16).

(6) Page 23 "we first identified the end of congression" how does this relate to kinetochore oscillations that move kinetochores away from the metaphase plate?

The start of kinetochore oscillation was defined as the end of Phase 4 if we could track the kinetochore until that point. In some cases where the kinetochore became close to the midplane (< 2.5 µm), it was not possible to track it further due to kinetochore crowding around the spindle mid-plane – in such cases, the end of Phase 4 was assigned as the end of tracking. In the original manuscript, it was not clear that the end of Phase 4 was defined in the same way for both non-polar and polar kinetochores, while the start of Phase 4 was defined differently for the two groups. This was confusing in the original manuscript. We have now clarified these points in the Method section (page 23).

(7) Are spindle pole distances (spindle sizes) different in early and late mitotic cells (4min vs 6min after NEBD) in control vs azBB-treated cells? Please comment on Figure S2E (mean distance) in the context of when phase 4 is completed. Does spindle size return to normal after congression?

In Figure S2E, we did not observe a significant difference in the spindle-pole distance (the spindle size) between control and azBB-treated cells at any individual time points. The smallest p-value was 0.094 at 6.0 min. As suggested, we have explained this in the legend for Figure S2E.

Significance:

The current work builds upon their previous work, in which the authors demonstrated that an actomyosin network forms on the cytoplasmic side of the nuclear envelope during prophase. This work explains how the network facilitates chromosome capture and congression by tracking motions of individual kinetochores during early mitosis. The findings can be broadly useful for cell division and the cytoskeletal fields.

Description of analyses that authors prefer not to carry out

Reviewer #1 (Evidence, reproducibility and clarity):

Summary

Sheidaei and colleagues report a novel and potentially important role for an early mitotic actomyosin-based mechanism, PANEM contraction, in promoting timely congression of chromosomes located at the nuclear periphery, particularly those in polar positions. The manuscript will interest researchers studying cell division, cytoskeletal dynamics, and motor proteins. Although some data overlap with the group's prior work, the authors extend those findings by optimizing key perturbations and performing more detailed analyses of chromosome movements, which together provide a clearer mechanistic explanation. The study also builds naturally on recent ideas from other groups about how chromosome positioning influences both early and later mitotic movements.

In its current form, however, the manuscript is not acceptable for publication. It suffers from major organizational problems, an overcrowded and confusing Results section and figures, and a lack of essential experimental controls and contextual discussion. These deficiencies make it difficult to evaluate the data and the authors' conclusions. A substantial structural revision is required to improve clarity and persuasiveness. In addition, several key control experiments and more conceptual context are needed to establish the specificity and relevance of PANEM relative to other microtubule- and actin-based mitotic mechanisms. Testing PANEM in additional cell lines or contexts would also strengthen the claim. I therefore recommend Major Revision, addressing the structural, conceptual, and experimental issues detailed below.

Major Comments

A. Structural overhaul and figure reorganization

The Results section is overly dense, lacks clear structure, and includes descriptive content that belongs in the Methods. Many figure panels should be moved to Supplementary Materials. A substantial reorganization is required to transform the manuscript into a focused, "Reports"-type article.

Move methodological and descriptive details (e.g., especially from the second Results subheading and Figure 2) to the Methods or Supplementary Materials.

In these parts, we define four phases of kinetochore motion in early mitosis. Without such a description in the main text, readers would be confused about subsequent analyses. Figure 2 is also important to show examples of how the four phases develop. Although we respect this suggestion from the reviewer, we would like to keep these parts in the main text and main figure.

New Figure 2: Combine current Figures 2A, 3A, 3C, 3D, 4C, 4F, and 4H to illustrate how PANEM contraction facilitates initial interactions of peripheral chromosomes with spindle microtubules which increases speed of congression initiation.

If we were to follow this suggestion, we would lose Figure 2B, D, Figure 3B and Figure 4A, where examples of kinetochore motions are shown in images and 3D diagrams. The new Figure would mostly consist of only graphs. Without examples of images and 3D diagrams, readers would have difficulty understanding the study. Although we respect this suggestion from the reviewer, we would like to keep Figures 2, 3 and 4, as they are (except for making Figure 4I simpler; see above).

New Figure 3: Combine current Figures 5A, 5C, 5D, 5F, 6B, 6C, and lower panels of 4H to show how PANEM contraction repositions polar chromosomes and reduces chromosome volume in early mitosis to enable rapid initiation of congression.

If we were to follow this suggestion, we would lose Figure 5B and Figure 6A, where examples of kinetochore/chromosome dynamics are shown in images and 3D diagrams. For the same reason as above, we would like to keep Figure 5 and 6 as they are, although we respect this suggestion from the reviewer.

New Figure 4: Combine Figures 7A, 7B, 7D, 7E, 7F, expanded Supplementary Figure S7, and new data to demonstrate that PANEM actively pushes peripheral chromosomes inward which is important for efficient chromosome congression in diverse cellular contexts.

As suggested, we will conduct new experiments to demonstrate the role of PANEM in diverse cellular contexts, as detailed below. We will then combine the new results with Figure S7 to make the new Figure 8.

On the other hand, in our view, combining Figure 7A-E and the extended Figure S7 would be confusing because the two parts address different topics. Although we respect this suggestion from the reviewer, we would like to keep Figure 7 and the extended Figure S7 (i.e. Figure 8) separate.

B. Specificity and redundancy of actin perturbation

To establish the specificity and relevance of PANEM, the authors should include or discuss appropriate controls:

Examine higher-ploidy or binucleated cells to determine whether multiple PANEM contractions are coordinated and if PANEM contraction contributes more in cells of higher ploidies or specific nuclear morphologies.

This is an interesting suggestion, but it takes lots of time to conduct such a study, and it goes beyond the scope of this paper.

Investigate dependency on nuclear shape or lamina stiffness; test whether PANEM force transmission requires a rigid nuclear remnant.

This is an interesting suggestion, but it takes lots of time to conduct such a study, and it goes beyond the scope of this paper.

Analyze PANEM's contribution under mild microtubule perturbations that are known to induce congression problems (e.g., low-dose nocodazole).

In the current study, we found that PANEM contraction affects chromosome motions in Phase 1 and Phase 3 but not Phase 2 or Phase 4. Mild microtubule perturbation itself could affect chromosome motions in all four Phases. We do not think it would be so informative to study what additional effects the reduced PANEM contraction shows when combined with mild microtubule perturbation.

D. Conceptual integration in Introduction and Discussion

The manuscript should better situate its findings within the context of early mitotic chromosome movements:

Minor Comments

These issues are more easily addressable but will significantly improve clarity and presentation.

Results (by subheading)

Fourth subheading: Note that congression speed is lower for centrally located kinetochores because they achieve biorientation more rapidly (Barišić et al., 2013, Nat Cell Biol; Vukušić & Tolić, 2025, Nat Commun).

We respect this comment. However, if biorientation were established more rapidly for centrally located kinetochores, it would advance the initiation of congression, but would not necessarily change congression speed.

Author response:

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

We greatly appreciate the reviewers’ constructive comments and have followed their recommendations to improve our manuscript. These improvements include additional experiments, new analyses, and a rewriting of the text. We believe these changes significantly improved the paper and hope the editor and the reviewers agree. The following is a summary of the major changes made and our point-by-point response to reviewers’ comments.

Summary of major changes:

(1) Expanded labeling options: We generated a new nMAGIC vector containing miRFP680 as an infrared fluorescent protein (IFP) marker. We used gRNA-40D2(IFP) to demonstrate clones labeled by this marker in the wing imaginal disc (Figure 1M). This vector is available via Addgene for the generation of new gRNA-markers with our recommended or customer-designed gRNA target sequences.

(2) Validated Gal80 potency: We provide new data in Figure 1E demonstrating complete suppression of pxn-Gal4>CD4-tdTom by tub-GAL80-DE-SV40. The exact transgenes used in the comparisons are clarified in the figure and figure legend.

(3) Verified clone fitness: We compared the sizes of nMAGIC twin spots in wing discs and found no intrinsic growth or viability bias between marker/marker and WT/WT clones (Figure 1O).

(4) Methodological Schematics: We added supplemental figures to Figure 1 to illustrate the principle of MAGIC, the difference between pMAGIC and nMAGIC, and an example of pMAGIC crossing scheme.

(5) Inducible induction: We provide new data (Figure 3J-K’) showing the induction of sparse neuronal clones in the adult brain by heat shock (hs)-Cas9.

(6) We revised texts to incorporate all other recommendations suggested by the reviewers. We also made other small changes to the manuscript to improve its readability.

Public Reviews:

Reviewer #1 (Public review):

Summary:

In this manuscript, Shen et al. have improved upon the mitotic clone analysis tool MAGIC that their lab previously developed. MAGIC uses CRISPR/Cas9-mediated double-stranded breaks to induce mitotic recombination. The authors have replaced the sgRNA scaffold with a more effective scaffold to increase clone frequency. They also introduced modifications to positive and negative clonal markers to improve signal-to-noise and mark the cytoplasm of the cells instead of the nuclei. The changes result in increase in clonal frequencies and marker brightness. The authors also generated the MAGIC transgenics to target all chromosome arms and tested the clone induction efficacy.

Strengths:

MAGIC is a mitotic clone generation tool that works without prior recombination to special chromosomes (e.g., FRT). It can also generate mutant clones for genes for which the existing FRT lines could not be used (e.g., the genes that are between the FRT transgene and the centromere).

This manuscript does a thorough job in describing the method and provides compelling data that support improvement over the existing method.

Weaknesses:

It would be beneficial to have a greater variety of clonal markers for nMAGIC. Currently, the only marker is BFP, which may clash with other genetic tools (e.g., some FRET probes) depending on the application. It would be nice to have far-red clonal markers.

We thank the reviewer for the positive comments about our study. We agree with the reviewer that adding a far-red option for nMAGIC increases the flexibility of this method. We replaced the BFP coding sequence in the nMAGIC cloning vector pAC-U63-QtgRNA2.1-tubBFP(HA) with that of miRFP680-T2A-HO1. We then used the resulting cloning vector to make a gRNA-40D2(IFP) transgene and tested it in the wing disc. Result showing clones in the wing disc are now in Figure 1M. The new cloning vector, along with others reported in our study, are available from Addgene.

Reviewer #2 (Public review):

Summary:

In this study, the authors present the latest improvement of their previously published methods, pMAGIC and nMAGIC, which can be used to engineer mosaic gene expression in wild-type animals and in a tissue-specific manner. They address the main limitation of MAGIC, the lack of gRNA-marker transgenes, which has hampered the broader adoption of MAGIC in the fly community. To do so, they create an entire toolkit of gRNA markers for every Drosophila chromosome and test them across a range of different tissues and in the context of making Drosophila species hybrid mosaic animals. The study provides a significant and broadly useful improvement compared to earlier versions, as it broadens the use-cases for transgenic manipulation with MAGIC to virtually any subfield of Drosophila cell biology.

Strengths:

Major improvements to MAGIC were made in terms of clone induction efficiency and usability across the Drosophila model system, including wild-type genotypes and the use in non-melanogaster species.

Notably, mosaic mutants can now be created for genes residing on the 4th chromosome, which is exciting and possibly long-awaited by 4th chromosome gene enthusiasts.

Selection of the standard set of gRNA markers was done thoughtfully, using non-repetitive conserved and unique sequences.

The authors demonstrate that MAGIC can be used easily in the context of interspecific hybrids. I believe this is a great advancement for the Drosophila community, especially for evolutionary biologists, because this may allow for easy access to mechanistic, tissue-specific insight into the process of a range of hybrid incompatibilities, an important speciation process that is normally difficult to study at the level of molecular and cell biology.

In the same way, because it is not limited to usage in any particular genetic background, genome-wide MAGIC can be potentially used in wild-type genotypes relatively easily. This is exciting, especially because natural genetic diversity is rarely investigated more mechanistically and at the scale/resolution of cells or specific tissues. Now, one can ask how a particular naturally occurring allele influences cell physiology compared to another (control) while keeping the global physiological context of the particular genetic background largely intact.

Weaknesses:

It is not entirely clear how functionally non-critical regions were evaluated, besides that they are selected based on conservation of sequence between species. It may be useful to directly test the difference in viability or other functionally relevant phenotype for flies carrying different markers. Similarly, the frequency of off-targets could be investigated or documented in a bit more detail, especially if one of the major use-cases is meant for naturally derived, diverse genetic backgrounds. It is, at the moment, unclear how consistently the clones are induced for each new gRNA marker across different WT genetic backgrounds, for example, a set of DGRP genotypes, which could be highly useful information for future users.

We thank the reviewer for the positive comments about our study. The reviewer raises an excellent point regarding the consistency of clone induction and potential background effects in diverse genetic backgrounds. As a standard step in building the MAGIC kit, we tested all gRNA-marker transgenes with the Cas9-LEThAL assay (Poe et al., Genetics, 2019), in which the gRNA-marker transgene was crossed to lig4 Act5C-Cas9 homozygotes. All crosses led to viable and apparently healthy female progeny, suggesting that ubiquitously mutating the chosen gRNA targeting sites does not cause obvious defects.

For standard mutant analysis, we recommend researchers to use a well-characterized wildtype chromosome as a negative control. For studies utilizing diverse wildtype backgrounds where a standard control chromosome is inapplicable (e.g., DGRP screens), we recommend an internal validation strategy: researchers should confirm their key phenotypic findings by inducing clones with a second, independent gRNA-marker located on the same chromosomal arm (e.g., comparing clones induced by gRNA-40D2 vs. gRNA-40D4 ). This ensures that any observed phenotypes or variations in clone induction are linked to the selected genetic background rather than an off-target artifact or target-site specific effect.

We admit that the above approach may not resolve concerns about off-targets. Performing deep sequencing to map empirical off-targets for all 34 gRNA pairs across multiple genetic backgrounds is experimentally prohibitive for a toolkit resource. However, our in silico selection pipeline strictly required target sequences to be unique within the D. melanogaster genome to mathematically minimize off-target probability. In addition, our requirement that target sequences be conserved in closely related Drosophila species acts as a stringent filter against intraspecies variation. Sequences conserved across species are subject to purifying selection, substantially reducing the likelihood that SNPs within the DGRP lines will disrupt the PAM or seed sequences required for Cas9 induction.

Reviewer #3 (Public review):

Summary:

In the manuscript by Shen, Yeung, and colleagues, the authors generate an improved and expanded Mosaic analysis by gRNA-induced crossing-over (MAGIC) toolkit for use in making mosaic clones in Drosophila. This is a clever method by which mitotic clones can be induced in dividing cells by using CRISPR/Cas9 to generate double-strand breaks at specific locations that induce crossing over at those locations. This is conceptually similar to previous mosaic methods in flies that utilized FRT sites that had been inserted near centromeres along with heat-shock inducible FLPase. The advantage of the MAGIC system is that it can be used along with chromosomes lacking FRT sites already introduced, such as those found in many deficiency collections or in EMS mutant lines. It may also be simpler to implement than FRT-based mosaic systems. There are two flavors of the MAGIC system: nMAGIC and pMAGIC. In nMAGIC, the main constituents are a transgene insertion that contains gRNAs that target DNA near the centromere, along with a fluorescent marker. In pMAGIC, the main constituents are a transgenic insertion that contains gRNAs that target DNA near the centromere, along with ubiquitous expression of GAL80. As such, nMAGIC can be used to generate clones that are not labelled, whereas pMAGIC (along with a GAL4 line and UAS-marker) can be used much like MARCM to positively label a clone of cells. This manuscript introduces MAGIC transgenic reagents that allow all 4 chromosomes to be targeted. They demonstrate its use in a variety of tissues, including with mutants not compatible with current FLP/FRT methods, and also show it works well in tissues that prove challenging for FLP/FRT mosaic analyses (such as motor neurons). They further demonstrate that it can be used to generate mosaic clones in non-melanogaster hybrid tissues. Overall, this work represents a valuable improvement to the MAGIC method that should promote even more widespread adoption of this powerful genetic technique.

Strengths:

(1) Improves the design of the gRNA-marker by updating the gRNA backbone and also the markers used. GAL80 now includes a DE region that reduces the perdurance of the protein and thus better labeling of pMAGIC clones. The data presented to demonstrate these improvements is rigorous and of high quality.

(2) Introduces a toolkit that now covers all chromosome arms in Drosophila. In addition, the efficiency of 3 target different sites is characterized for each chromosome arm (e.g., 3 different gRNA-Marker combinations), which demonstrate differences in efficiency. This could be useful to titrate how many clones an experimenter might want (e.g., lower efficiency combinations might prove advantageous).

(3) The manuscript is well written and easy to follow. The authors achieved their aims of creating and demonstrating MAGIC reagents suitable for mosaic analysis of any Drosophila chromosome arm.

(4) The MAGIC method is a valuable addition to the Drosophila genetics toolkit, and the new reagents described in this manuscript should allow it to become more widely adopted.

Weaknesses:

(1) The MAGIC method might not be well known to most readers, and the manuscript could have benefited from schematics introducing the technique.

We thank the reviewer for the positive evaluation of our study and for making this kind suggestion. We have added diagrams that explain the principle of MAGIC and the difference between pMAGIC and nMAGIC in Figure 1 - Figure Supplement 1.

(2) Traditional mosaic analyses using the FLP/FRT system have strongly utilized heat-shock FLPase for inducible temporal control over mitotic clones, as well as a way to titrate how many clones are induced (e.g., shorter heat shocks will induce fewer clones). This has proven highly valuable, especially for developmental studies. A heat-shock Cas9 is available, and it would have been beneficial to determine the efficiency of inducing MAGIC clones using this Cas9 source.

We thank the reviewer for suggesting this experiment. We agree that demonstrating inducible clone induction in the adult brain is an effective way for people to compare MAGIC with the MARCM method they are probably more familiar with. We used a heat shock Cas9 developed by the Tzumin Lee group (Chen et al., Development, 2020) to experiment with clone induction, and the results are shown in the new Figure 3 (K and J). We show that, with a pan-neuronal Gal4, heat shock during the wandering 3rd instar larval stage induced more clones than during the pupal stage, and the later heat shock readily produced sparsely labeled neurons whose single-cell morphology can be easily visualized.

Recommendations for the authors:

Reviewing Editor Comments:

The following are some consolidated review remarks after discussions amongst all three reviewers:

The reviewers feel the evidence level could be raised from 'convincing' to 'compelling' if the following key (and partially shared) suggestions by the reviewers are followed adequately:

(1) Expand labeling options for nMAGIC, which is currently just a BFP marker. This would increase the utility of the method. A far-red marker would be very helpful. Could the authors just do this for one chromosome arm and make the reagent available for others to generate other chromosome arms?

We agree with the editor and reviewers that adding a far-red option for nMAGIC increases the flexibility of this method. We replaced the BFP coding sequence in the nMAGIC cloning vector pAC-U63-QtgRNA2.1-tubBFP(HA) with that of miRFP680-T2A-HO1. We then used the resulting cloning vector to make a gRNA-40D2(IFP) transgene and tested it in the wing disc. Result showing clones in the wing disc are now in Figure 1M. The new cloning vector, along with others reported in our study, will be available from Addgene.

(2) Verify that destabilized GAL80 is potent enough to suppress GAL4. Repeat Figure 1C-E with tub-GAL80-DE-SV40.

We replaced the experiment using gRNA-42A4-tDES, which successfully achieved complete suppression of pxn>CD4-tdTom (Figure 1E).

(3) Concern about the health of the induced mitotic clones. This is an important consideration, but the reviewers were not sure what the necessary experiments would be. To gauge twin-spot clone sizes? Please address.

We agree that clone fitness is an important consideration for MAGIC experiments. To test it, we generated WT clones in the wing imaginal disc using nMAGIC and quantified the sizes of the twin spots (BFP/BFP and WT/WT clones). Our results show that there is no statistical difference between these two types of clones. Thus, there is no intrinsic growth disadvantage to either type of mitotic clones generated by MAGIC.

(4) Include a schematic of the MAGIC method as Figure 1 or add it to Figure 1. Many may not be familiar with the method, so to promote its adoption, the authors should clearly introduce the MAGIC method in this paper (and not rely on readers to go to previous publications). For this paper to become a MAGIC reference paper, it should be self-contained.

We thank the reviewers for this suggestion. We have added diagrams that explain the principle of MAGIC and the difference between pMAGIC and nMAGIC in Figure 1 - Figure Supplement 1.

(5) Determine the utility of using a hs-Cas9 line for temporal induction of MAGIC clones. This is a traditional method for mitotic clone induction (with hsFLP/FRTs), and its use with the MAGIC system (especially pMAGIC) could also make it more attractive, especially to label small populations of neurons born at known times. To this point, the authors could generate pMAGIC clones using hs-Cas9 for commonly used adult target neurons, such as projection neurons, central complex neurons, or mushroom body neurons. The method to label small numbers of these adult neurons is well worked out with known GAL4 lines, and demonstrating that pMAGIC could have similar results would capture the attention of many not familiar with the pMAGIC method.

We agree that demonstrating inducible clone induction in the adult brain is an effective way for people to compare MAGIC with the MARCM method they are probably more familiar with. We used a heat shock Cas9 developed by the Tzumin Lee group (GarciaMarques, Espinosa-Medina et al. 2020) to experiment with clone induction, and the results are shown in the new Figure 3 (J-K’). We show that, with a pan-neuronal Gal4, heat shock during wandering 3rd instar larval stage induced more clones than during the pupal stage, and the later heat shock readily produced sparsely labeled neurons whose single-cell morphology can be easily visualized.

Reviewer #1 (Recommendations for the authors):

This is a marked improvement over the existing methods that the authors' lab has previously generated. It will be a nice addition to the Drosophila genetic tool kit after minor revisions.

We appreciate the reviewer’s recognition of the new tools we developed.

Minor issues:

(1) In the data in Figures 1G and H, it is not ideal to compare the effect of different modifications on two different transgenes. uH and uDEH are compared in gRNA-40D2, whereas uDEH, tDEH, and tDES are compared in gRNA-42A4. If the transgenics are already available, it would be better to compare the uH, uDEH, tDEH, and tDES on either gRNA-40D2 or gRNA-42A4.

We appreciate the reviewer’s concern. These transgenes were developed during different phases of this project. We first adopted the uDEH design during improvement of gRNA40D2, which solved both the leaky activity of pxn-Gal4 and dim epidermal clones. However, when we tried to expand this design to 2R (such as 42A4), we found that the clones were still too dim (probably due to positional effects). Thus, we next used uDEH in gRNA-42A4 as a base for further improvements. We did not make a uH version for gRNA-42A4 because we already knew that it is inferior to uDEH. Because of this history, we did not have the full set for gRNA42A4.

Despite the lack of uH for gRNA-42A4, we believe our comparisons of different designs are still valid, given that uH and uDEH were compared with identical sequences elsewhere in the transgenic vector (including the gRNA target sequence) and in the identical insertion site.

(2) It is not clear whether the authors tested destabilized Gal80 is potent to suppress Gal4 (e.g., in suppressing pxn>CD4-tdTom in hemocytes). The results in Figure 1C-E should be repeated with tub-Gal80-DE-SV40.

We apologize for omitting the transgene identities in these experiments. We have redone the experiment using gRNA-42A4-tDES and updated the figures to clearly indicate which transgenes were used.

(3) The difference in sgRNA scaffolds can be better explained in the text. The explanation here is very bare bones and reads like jargon. (i.e., changing F+E gRNA scaffold with gRNA2.1 scaffold is not a sufficient explanation).

We have added more explanations to the differences between the scaffolds as suggested.

(4) The stocks should be sent to Bloomington Stock Center to ensure widespread adoption of the method. This includes the Cas9 lines that are generated and used.

It is our plan to freely share the reagents developed in this study with the community. Most of the fly lines are already available at Bloomington (https://bdsc.indiana.edu/stocks/misc/magic.html and https://bdsc.indiana.edu/stocks/genome_editing/crispr_cas9.html). We are in the process of depositing the remaining ones to BDSC.

In conclusion, this is a nicely written manuscript that improves currently available tools and should be of interest to the readership of this journal.

Reviewer #2 (Recommendations for the authors):

Typos spotted:

Line 163 issues -> tissues

Line 613 significance -> significant

We thank the reviewer for catching these typos. We have corrected them.

Reviewer #3 (Recommendations for the authors):

This is a welcome update to the MAGIC system, which is a brilliant method that has not been as widely adopted as it should be. The authors validate and introduce updates to this system to increase clonal efficiency and more robust labeling (for both pMAGIC and nMAGIC). The data presented are robust and convincing.

We appreciate the reviewer’s positive comments about our study.

Suggestions to improve the presentation and adoption of this work:

(1) The MAGIC system might not be well known, and the manuscript would have benefited from an introductory schematic of how the system works. I realize this was already done in the PLoS Biology paper, but the authors should not assume readers will know that paper, or be willing to look it up. So a standalone schematic, as Figure 1, or something added to Figure 1, would greatly aid in understanding how this system works and what the new updates are doing.

We thank the reviewer for this kind suggestion. We have added diagrams that explain the principle of MAGIC and the difference between pMAGIC and nMAGIC in Figure 1 - figure supplement 1.

(2) There were many instances where abbreviations were not clearly defined, especially in the Figures and Figure legends. The main text is well-written, and while the information is in there, it is beneficial when the Figures and Figure legends can stand alone. For example:

(a) Figure 1. DE, not defined in the Figure or Figure legend.

(b) Figure 1. 'p' and 'n' not defined in the Figure legend.

(c) The different Cas9 lines or GAL4 lines used-a brief description of their expression patterns might be helpful in the legend. E.g., zk-Cas9, vas-Cas9, gcm-Cas9, R38F11-GAL4, RabX4Gal4.

We apologize for omitting the details mentioned. They have been added to the figures and figure legends.

(3) "Traditional" mosaic analyses took advantage of hsFLP for inducible induction and to control the number of mitotic clones that were induced. A hs-Cas9 line does exist (as correctly pointed out by the authors), and it would be a valuable addition if the authors tested the utility of this reagent with the MAGIC system. Many possible adopters may not like the idea that an alwayson Cas9 line is used, which could result in too many clones, especially if one wanted to label very few cells. Granted, one could use a 'worse' gRNA-Marker line as mentioned in the manuscript, but this might still be hard to titrate, as well as an inducible system that uses a heatshock promoter. A hs promoter is especially useful for birthdating cells during development.

We thank the reviewer for suggesting this experiment. We agree that demonstrating inducible clone induction in the adult brain is an effective way for people to compare MAGIC with the MARCM method they are probably more familiar with. We used a heat shock Cas9 developed by the Tzumin Lee group (Chen et al., Development, 2020) to experiment with clone induction, and the results are shown in the new Figure 3 (K and J). We show that, with a panneuronal Gal4, heat shock during wandering 3rd instar larval stage induced more clones than during the pupal stage, and the later heat shock readily produced sparsely labeled neurons whose single-cell morphology can be easily visualized.

(4) Lines 61-63. "However, most of these mutant chromosomes cannot be analyzed by traditional mosaic techniques due to the lack of FRT sites or incompatibility with the FRT/Flp system." It might also be worth mentioning that recombining existing reagents (e.g., mutants, etc) onto an FRT chromosome can be labor and time-intensive. A brilliant advantage of MAGIC is that it can be used with any existing stock, such as from classical EMS mutant screens, Df screens (as pointed out), etc. So the more the authors can emphasize a new way of thinking (e.g, you don't need to recombine your mutant of interest onto an FRT stock before you can get started), the better!

We thank the reviewer for this kind suggestion. As suggested, we have expanded our introduction and discussion to emphasize the advantages of the MAGIC system over traditional mosaic techniques.

(5) One incredible advantage of the MAGIC system is that it can direct where recombination occurs. So if one had two mutations on a chromosome arm, it could be possible to make the most distal homozygous mutant while the other remains heterozygous. This is not possible with current FRT-based methods. It's not necessary to demonstrate this, but perhaps the authors could mention it as a possible next step? This was somewhat implied by lines 66-67 "In comparison, MAGIC can potentially be used to study these genes because the crossover site in MAGIC can be flexibly defined by users".

Again, we thank the reviewer for this nice suggestion. We have added this point to the discussion.

(6) How stable are the MAGIC lines? If gRNA (with Cas9 expressed) induced a germline mutation of the target site, the MAGIC line would break down. How often is this observed? Some mention of this would be appreciated, especially to end users, if caution is necessary and gRNA-marker stocks should not be maintained in the same flies as an x-Cas9 line.

The reviewer made a very important point. Keeping gRNA and Cas9 in the same strain will risk mutating the target sequence in the germline, if the Cas9 has any activity in the germline. Thus, it is not recommended to keep gRNA and Cas9 in the same flies over multiple generations. For MAGIC experiments, this concern is lessened because by crossing gRNA + Cas9 flies to another strain containing the chromosome of interest, clones can still be induced (possibly with less efficiency) because the chromosome of interest is still cuttable by Cas9. Nevertheless, to address this concern, we have recently developed anti-CRISPR tools to suppress Cas9 activity in such strains. These tools will be reported in a separate study.

In the revised manuscript, we added this point in Discussion to caution users.

(7) Line 157, "identify efficient gRNAs for every chromosomal arm.". What is considered "efficient"? Is this quantifiable? Eg., >= 10 clones.

Thanks for pointing this out! “Efficient” is an arbitrary evaluation, as different experiments may require different efficiencies. But operationally, we consider any gRNA that can generate >= 10 neuronal clones per larva as being efficient. We have clarified it in the text.

(8) Line 163, "highly packed _issues_ such as the brain"; spelling, should be "tissues"

Thanks for catching this typo. It has been corrected.

(9) The authors use ey-Cas9 for their demonstration of adult brain labeling. Additional adult brain examples would increase exposure of this method and attract wider attention- targeting structures that have been well characterized, such as projection neurons (GH146-GAL4), central complex, mushroom bodies, etc. Especially if hs-Cas9 could be utilized to mimic previous MARCM clones (for example).

We thank the reviewer for suggesting heat shock-induced clones in the adult brain. We have conducted the experiment as explained above and shown in Figure 3J-3K’. We showed a single neuronal clone that resembles lateral horn Leucokinin neurons.

(10) Line 216, "Despite these advances, existing mutations on FRT-lacking 4th chromosomes still cannot be analyzed by the FRT/Flp system." For context, it might be worth pointing out that meiotic recombination is exceedingly rare on the 4th chromosome, which means it is practically impossible to recombine existing 4th chromosome mutations onto an FRT chromosome.

We thank the reviewer for this kind suggestion. We have added a note about the difficulty of recombining FRT onto the 4th chromosome.

(11) Figure 2 legend. What is the full genotype for D and E? eg, what is RabX4>MApHS?

We apologize for being brief with the details. RabX4-Gal4 is a pan-neuronal driver. UAS-MApHS is a membrane fluorescent marker (UAS-pHluorin-CD4-tdTom). The genotypes have been added to the figure legend.

(12) It would be good to include the Bloomington Stock numbers for the MAGIC toolkit, especially in Table 1. And include an HTML reference to their MAGIC page at Bloomington

(https://bdsc.indiana.edu/stocks/misc/magic.html).

Thank you for this suggestion! We have done as suggested.

(13) Similarly, the key plasmids to create the improved gRNA-marker insertions should be deposited to Addgene (or similar repository) and their ID numbers included in the resources table.

The plasmids have been deposited to Addgene and are currently being validated.

(14) The authors might consider including (perhaps as supplementary to Figure 1 or Figure 2) a crossing scheme for one of their MAGIC experiments. This will make it even clearer how a MAGIC experiment could be set up using existing fly reagents.

This is a good suggestion! We have added an example crossing scheme in Figure 1 – figure supplement 1C.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

The manuscript "Synaptotagmin 1 and Synaptotagmin 7 promote MR1-mediated presentation of Mycobacterium tuberculosis antigens", authored by Kim et al., showed that the calcium-sensing trafficking proteins Synaptotagmin (Syt) 1 and Syt7 specifically promote (are critical for) MAIT cell activation in response to Mtb-infected bronchial epithelial cell line BEAS-2B (Fig. 1) and monocyte-like cell line THP-1 (Figure 3) . This work also showed co-localization of Syt1 and Syt7 with Rab7a and Lamp1, but not with Rab5a (Figure 5). Loss of Syt1 and Syt7 resulted in a larger area of MR1 vesicles (Figure 6f) and an increased number of MR1 vesicles in close proximity to an Auxotrophic Mtb-containing vacuoles during infection (Figure 7ab). Moreover, flow organellometry was used to separate phagosomes from other subcellular fractions and identify enrichment of auxotrophic Mtb-containing vacuoles in fractions 42-50, which were enriched with Lamp1+ vacuoles or phagosomes (Figures 7e-f).

Strengths:

This work nicely associated Syt1 and Syt7 with late endocytic compartments and Mtb+ vacuoles. Gene editing of Syt1 and Syt7 loci of bronchial epithelial and monocyte-like cells supported Syt1 and Syt7 facilitated maintaining a normal level of antigen presentation for MAIT cell activation in Mtb infection. Imaging analyses further supported that Syt1 and Syt7 mutants enhanced the overlaps of MR1 with Mtb fluorescence, and the MR1 proximity with Mtb-infected vacuoles, suggesting that Syt1 and Syt7 proteins help antigen presentation in Mtb infection for MAIT activation.

Weaknesses:

Additional data are needed to support the conclusion, "identify a novel pathway in which Syt1 and Syt7 facilitate the translocation of MR1 from Mtb-containing vacuoles" and some pieces of other evidence may be seen by some to contradict this conclusion.

We thank the reviewer for their positive and constructive comments. Because MR1 presents small molecule metabolites, specifically identifying MR1 molecules loaded with antigens derived from intracellular Mtb infection remains a significant technical challenge. Therefore, we agree that some of our approaches measure antigen-loaded MR1 indirectly. For example, IFN-γ release from a MAIT cell clone serves as a sensitive surrogate readout for the presence of antigen-loaded MR1 at the cell surface. This has been demonstrated in previous work showing that IFN-γ release from MAIT cells correlated with loaded MR1 molecules measured using flow cytometry and a TCR based tetramer (Kulicke et al., 2024). In this context, Syt1 and Syt7 represent the first endosomal trafficking proteins we have identified that play a specific role in MR1-mediated presentation of Mtb-derived metabolites. Syt1 and Syt7 do not contribute to the presentation of an exogenously delivered MR1 ligands, such as Ac-6-FP loaded in the ER or M. smegmatis supernatant. In Syt1 and Syt7 knockout cells expressing MR1-GFP, larger MR1 vesicles are observed, but MR1 continues to co-localize with LAMP1 similar to wildtype cells. Furthermore, Syt1 and Syt7 knockout cells exhibit an increased number of MR1 vesicles near the Mtb-containing vacuoles compared to wildtype cells. To increase the statistical power of our microscopy analyses, we have analyzed additional cells. Although the absolute magnitude of the observed effects is modest, T cell activation is highly sensitive to the number of loaded antigen presenting molecules at the cell surface. Also, a complementary approach using flow organellometry confirmed increased MR1 expression within Mtb⁺LAMP1⁺ vesicles in Syt7 knockout cells. Thus, these findings suggest a mechanism whereby Syt1 and Syt7 facilitate the trafficking of loaded MR1 molecules from the Mtb-containing vacuoles to the plasma membrane. This specialized mechanism may be analogous to the previously described role of Syt7 in MHC class II trafficking (Becker et al., 2009). In our model, we observed increased accumulation and expression of MR1 within Mtb-containing vacuoles in Syt7 knockout cells.

Reviewer #2 (Public review):

Summary:

The study demonstrates that calcium-sensing trafficking proteins Synaptotagmin (Syt) 1 and Syt7 are involved in the efficient presentation of mycobacterial antigens by MR1 during M. tuberculosis infection. This is achieved by creating antigen-presenting cells in which the Syt1 and Syt7 genes are knocked out. These mutated cell lines show significantly reduced stimulation of MAIT cells, while their stimulation of HLA class I-restricted T cells remains unchanged. Syt1 and Syt7 co-localize in a late endo-lysosomal compartment where MR1 molecules are also located, near M. tuberculosis-containing vacuoles.

Strengths:

This work uncovers a new aspect of how mycobacterial antigens generated during infection are presented. The finding that Syt1 and Syt7 are relevant for final MR1 surface expression and presentation to MR1-restricted T cells is novel and adds valuable information to this process. The experiments include all necessary controls and convincingly validate the role of Syt1 and Syt7. Another key point is that these proteins are essential during infection, but they are not significant when an exogenous synthetic antigen is used in the experiments. This emphasizes the importance of studying infection as a physiological context for antigen presentation to MAIT cells. An additional relevant aspect is that the study reveals the existence of different MR1 antigen presentation pathways, which differ from the endoplasmic reticulum or endosomal pathways that are typical for MHC-presented peptides.

Weaknesses:

The reduced MAIT cell response observed with Syt1 and Syt7-deficient cell lines is statistically significant but not completely abolished. This may suggest that only some MR1-loaded molecules depend on these two Syt proteins. Further research is needed to determine whether, during persistent M. tuberculosis infection, enough MR1-loaded molecules are produced and transported to the plasma membrane to sufficiently stimulate MAIT cells. The study proposes that other Syt proteins might also play a role, as outlined by the authors. However, exploring potential redundant mechanisms that facilitate MR1 loading with antigens remains a challenging task.

We appreciate the reviewer’s comments and feedback. Syt1 and Syt7 knockout cells do not completely abolish MR1-mediated presentation of Mtb-derived metabolites. We agree that the likely explanation is that there are redundancies within the antigen presentation pathways. Whether these redundancies are due to other endosomal trafficking proteins or other intracellular compartments where MR1 loading can occur remains unknown. Moreover, Mtb-derived antigens can access the ER, where Syt1 and Syt7 are not involved, thereby enabling an ER-mediated pathway for MR1 antigen presentation. It is also important to note that relatively few (<10) loaded MHC class I molecules are sufficient to trigger T cell activation (Brower et al., 1994; Sykulev et al., 1995; Sykulev et al., 1996). A major challenge in exploring these mechanisms is due to the inability to directly track small molecule Mtb-derived antigens as they are loaded onto MR1 and presented at the cell surface. These hurdles are briefly outlined in the discussion as future directions. Nonetheless, Syt1 and Syt7 are the first endosomal trafficking proteins identified to have a specific effect on MR1-mediated presentation of Mtb-derived antigens.

Reviewer #3 (Public review):

Summary:

In the submitted manuscript, the authors investigate the role of Synaptotagmins (Syt1) and (Syt7) in MR1 presentation of MtB.

Strengths:

In the first series of experiments, the authors determined that knocking down Syt1 and Sy7 in antigenpresenting cells decreases IFN-γ production following cellular infection with Mtb. These experiments are well performed and controlled.

Weaknesses:

Next, they aim to mechanistically investigate how Syt1 and Syt7 affect MtB presentation. In particular, they focus on MR1, a non-classical MHC-I molecule known to present endogenous and exogenous metabolites, including MtB metabolites. Results from these next series of experiments are less clear. Firstly, they show that knocking down Syt1 and Sy7 does not change MtB phagocytosis as well as MR1 ER-plasma membrane translocation. Based on this, they suggest that Syt1 and Syt7 may affect MR1 trafficking in endosomal compartments. However, neither subcellular compartment analysis nor flow organelleometry clearly establishes the role of Syt1 and Syt7 in MtB trafficking. Altogether, the notion that Synaptotagmins facilitate MR1 interaction with Mtb-containing compartments and its vesicular transport was already known. As such, the manuscript should add additional insight on where/how the interaction occurs. The reviewer is left with the notion that Syt1 and Sy7 may affect MR1 presentation, facilitating the trafficking of MR1 vesicles from endosomal compartments to either the cell surface or other endosomal compartments. The analysis is observational and additional data or discussion could address what the insight gained beyond what is already known from the literature.

We thank Reviewer 3 for their comments. Our hypothesis is that Syt1 and Syt7 mediate MR1 trafficking rather than Mtb trafficking. While Syt7 has previously been implicated in MHC class II trafficking and vesicular transport, this study is the first to explore in detail the roles of Syt1 and Syt7 in MR1-mediated presentation of Mtb-derived metabolites. Since current technologies do not allow direct tracking of Mtbderived antigens loaded onto MR1, we relied on complementary approaches including IFN-γ release from MAIT cells, flow cytometry, fluorescence microscopy, and flow organelleometry. Both flow organelleometry and fluorescence microscopy show increased MR1 expression at Mtb-containing vacuoles in Syt7 knockout cells. Since total MR1 expression measured by flow cytometry and the overall number of MR1 vesicles remain unchanged, these data support a mechanism in which Syt7 facilitates the trafficking of antigen-loaded MR1 from Mtb-containing vacuoles to the cell surface, consistent with the observed reduction in MAIT cell IFN-γ release.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Concern 1, the data in the current manuscript have not been sufficient to "identify a novel pathway in which Syt1 and Syt7 facilitate the translocation of MR1 from Mtb-containing vacuoles, potentially to the cell surface for antigen presentation" (Last part of Abstract). To conclude this, additional pieces of data are needed: (a) Mtb-containing vacuoles associate with MR1 protein expression; (b) MR1+ vesicles traffic from one subcellular location to another; (c) Syt1 or Syt7 KO reduces MR1 vesicles at a downstream subcellular location, e.g., the cell surface. Important evidence supporting the "facilitation of translocation" is missing on whether Syt1 or Syt7 KO reduces MR1 vesicle traffic from one location to another.

We thank the reviewer for their detailed suggestions to improve our proposed model. We would like to clarify that Figure 7g demonstrates increased MR1 protein expression in Syt7 knockout cells, as assessed by flow organellometry. This approach allowed us to specifically distinguish AuxMtb⁺LAMP1⁺ compartments (Mtb-containing vacuoles) and to quantify MR1 expression using geometric mean fluorescence intensity. Moreover, in both Syt1 and Syt7 knockout cells, MR1+ vesicles are retained within lysosomal compartments, characterized by vesicle enlargement and accumulation. Therefore, we did not observe trafficking of MR1+ vesicles to other subcellular locations or to the plasma membrane. A key limitation, however, is the lack of current technologies that allow direct measurement of MR1 surface expression specifically during intracellular Mtb infection via flow cytometry. Given this limitation, IFN-γ ELISpot is a sensitive surrogate and supports the conclusion that loss of Syt1 and Syt7 results in decreased MR1 presentation of Mtb-derived antigens at the plasma membrane.

The results "a significant increase in the number of MR1 vesicles within 1 μm of AuxMtb for Syt1 (1.13 {plus minus} 0.46) and Syt7 KO (1.31 {plus minus} 0.46) cells compared to WT cells (Fig.7b)." and "the surface of MR1 vesicles in Syt1 and Syt7 KO cells showed a 3-fold increase in overlap area with Mtb surfaces (Fig.7d)." may need to be further elaborated on whether MR1+vacuoles and Mtb+ vacuoles are overlapped or are adjacent. Figure 7b shows several groups of vacuoles with the same distance. This needs a larger sample size to randomize this distance measurement, for example, calculating 50~100 Mtb+ vacuoles.

We appreciate the reviewer’s critical comments and suggestions. To quantify distance and surface overlap, the microscopy images were acquired from a single optical plane rather than full z-stacks. As a result, it is not possible to definitively determine whether MR1+ vesicles and Mtb-containing vacuoles are directly overlapping or adjacent. In response to the reviewer’s suggestion, we increased the sample size for both distance (n=51-53) and surface overlap analyses (n=51-53). Using the larger sample size, we observed a significant increase in the number of MR1 vesicles located within 1μm of AuxMtb in both Syt1 (1.23±0.21) and Syt7 knockout (1.28±0.22) cells. Also, there was an approximately 4-fold increase in MR1-Mtb surface overlap area compared to wildtype cells.

Results from "performed flow organellometry to separate phagosomes from other subcellular fractions and identified enrichment of Mtb-containing vacuoles in fractions 42-50 (Fig.7e-f)" could not distinguish the difference between WT and Syt1/Syt7 KO, or further support the role of Syt1/Syt7 in endocytic trafficking. More specifically, authors claimed that "enhanced MR1 expression in Mtb+LAMP1+ compartments via flow organellometry in Syt1 and Syt7 KO cells.", may not be supported by Figure 7f, which does not show a difference in MR1 expression between Syt1 KO or Syt7 KO and WT.

We appreciate the reviewer’s concerns and would like to clarify the interpretation of Figures 7f and 7g. Figure 7f demonstrates: (a) enrichment Mtb-containing vacuoles within fractions 42-50, (b) coenrichment of LAMP1+ vesicles within these Mtb-containing fractions, and (c) comparable subcellular fractionation profiles across wildtype, Syt1 knockout, and Syt7 knockout cells, indicating no major differences in fraction distribution. Differences in MR1 expression are shown in Figure 7g, which compares MR1 expression as the geometric mean fluorescence intensity within the fraction exhibiting the highest percentage of AuxMtb⁺LAMP1⁺ across all fractions. We observed significant increase in MR1 expression in Syt7 knockout cells compared to wildtype cells.

Concern 2, in abstract, "Loss of Syt1 and Syt7 results in enlarged MR1 vesicles and an increased number of MR1 vesicles in close proximity to Mtb-containing vacuoles during infection.". Although numbers of MR1 vesicles within 1um of Mtb increase (Figure 7b) and areas of MR1+ vacuoles for WT and KO cells enhance (Figure 6f), but numbers of MR1 vesicles/cells are not different between WT and Syt1 and Sy7 KO (Fig. 7c). These imaging analyses, including other figure panels, need more explicit presentation of (most if not all) random images for calculation, annotation of MR1-vacuoles for calculation, and raw statistical data for mean and p value calculation. These raw data can be presented in supplemental figure panels.

We thank the reviewer for these suggestions. We have included more details on randomization, technical procedures, and statistical analyses in methods section for “Fluorescence Microscopy,” “Image Analysis,” and “Statistical Analysis.” Raw data collection and statistical data are presented in the supplemental data.

Concern 3, additional evidence that does not support the conclusion "This study identifies a novel pathway in which Syt1 and Syt7 facilitate the translocation of MR1 from Mtb-containing vacuoles" (the last part of Abstract). This additional unsupportive evidence includes: (a) MR1 expression on the cell surface is not impacted or not different among WT, Syt1 KO, and Syt7 KO of BEAS-2B cells (Fig. 6d). (b) "Live-cell imaging showed no differences in MR1 cellular distribution in the presence or absence of Ac-6FP between WT, Syt1, and Syt7 KO BEAS-2B:TET-MR1GFP cells as MR1 translocated from the ER and vesicles to the cell surface as expected (Figure 6c).

We thank the reviewer for this comment and would like to clarify our use of Ac-6-FP. Figures 6c and 6d examine MR1 cellular distribution and surface expression in the presence or absence of Ac-6-FP. Ac-6-FP is a small MR1 ligand that is loaded in the ER and promotes MR1 surface stabilization and trafficking to the cell membrane. In contrast, Mtb primarily resides within membrane-bound phagosomes. MR1 presentations of soluble/exogenously delivered ligands versus intracellular Mtb-derived antigens have shown to involve distinct pathways and endosomal trafficking proteins (Harriff et al., 2016; Karamooz et al., 2019; Karamooz et al., 2025). Findings from Figures 6c and 6d show that Syt1 and Syt7 do not contribute to the presentation of small soluble and ER-loaded ligands such as Ac-6-FP. Instead, they specifically contribute in MR1 presentation of Mtb-derived metabolites by translocating MR1 from Mtbcontaining vacuoles in the context of intracellular Mtb infection

Other concerns:

(1) Figure 1a uses Ct value to measure Syt1 and Syt7 expression levels, but a comparison with GAPDH Ct cycle numbers in different cell types will be helpful for understanding.

We appreciate the reviewer’s suggestion of including GADPH Ct cycle numbers. We have revised Figure 1a to show Ct values for Syt1, Syt7, and GAPDH in both BEAS-2B and THP-1 cells.

(2) Figure 1b indel, shown with an ICE method, should be confirmed with protein expression levels to interpret functional results.

We thank the reviewer for raising this concern. We attempted to assess protein levels by western blot using multiple antibodies from both Abcam and Synaptic Systems. However, we were unable to identify a suitable antibody that reliably detected endogenous Syt1 or Syt7 protein levels.

(3) Figure 1c. HLA-B45-restricted T cell clones also show some marginal reduction of IFN-γ spot responses and are more different in Figure 6b. Please discuss this conflicting data. Also, need a reference to support whether the exogenous CFP peptide antigen is presented via surface or endocytic antigen loading.

We agree with the reviewer that there are some marginal reductions of IFN-γ responses for HLA-B45restricted T cell clones. Since T cell clones are used from frozen, there can be differences in maximal responses between T cell clones and expansions of the same T cell clone. However, the comparisons include a control arm and pool data from multiple experiments to reach statistical power and validity. In addition, Figure 6b shows Syt1 and Syt7 KO cells in the background of BEAS-2B MR1KO:tetMR1-GFP clone D4 cells, which overexpresses MR1 that may contribute to variability and potentially account for the observed differences. With respect to exogenous CFP peptide loading, earlier studies on peptides and antigen presenting cells demonstrated that peptides can be loaded onto fixed cells and subsequently presented to T cells (Shimonkevitz et al., 1983; Watts et al., 1985). Based on these findings, it is reasonable to assume that substantial peptide exchange occurs at the cell surface when exogenous peptides are added to antigen presenting cells.

(4) Figure 2e: Delta CT values of Syt1, Syt7 in WT, KO cells can be shown together with Ct values of GAPDH or B2m house-keeping genes to help readers determine the efficiency of Syt1 and 7 mutation at the gene expression level. Also, in Figure 4a, the baseline of Ct values for GAPDH can be plotted together.

As suggested by the reviewer, we have revised Figure 2e and 4a to include CT values for the genes of interest as well as housekeeping gene GAPDH.

(5) Figure 3c and Figure 1d: M.smeg infection can be shown to be more comparable with Mtb infection.

We thank the reviewer for this thoughtful comment. Although M. smegmatis infection could serve as a comparable control, M. smegmatis secretes large amounts of MR1 ligands derived from riboflavin metabolism. This makes it difficult to distinguish between extracellular and intracellular antigens, and to directly compare with Mtb infection, which is specifically an intracellular infection model.

(6) Figure 4e: It appears Esyt2 Knockdown shows strong inhibition of MAIT activation mediated by BEAS2B cells with Mtb infection and M.smeg supernatant stimulation. Please add other relevant data, such as MR1 cell surface expression and colocalization, and discuss these results with Syt proteins.

We appreciate the reviewer’s suggestion to include relevant data for Esyt2 knockdown. We performed flow cytometry analysis of Esyt2 knockdown cells and found surface MR1 expression under basal conditions. Treatment with Ac-6-FP resulted in increased MR1 surface stabilization, but MR1 surface level was significantly lower than those observed in missense control cells. Therefore, Esyt2 is not specific to MR1 presentation of Mtb-derived metabolites and instead may play a broader role in overall MR1 antigen presentation, including intracellular Mtb-derived antigens, exogenous antigens, and ER-loaded Ac-6-FP.

(7) Figure 5 colocalization computational analyses can be more explicitly presented regarding randomization, technical procedures, and statistical analyses, as stated in Concern 2.

As suggested, we have included more details in methods section and added the supplemental data.

(8) Figure 6a: Syt1 and Syt7 protein expressions are also suggested to confirm the mutation, similar to the confirmation for Figures 1 and 3.

We thank the reviewer for raising this concern. As discussed previously, we have not identified a suitable antibody for human Syt1 and Syt7. We have tested multiple antibodies from Abcam and Synaptic Systems.

(9) For statistical analyses, "non-linear regression analysis comparing best-fit values of top and EC50 were used to calculate p-values by extra sum-of-squares F test" (Figure 6b) and "non-linear regression analysis of pairwise comparison to WT on best-fit values of top and EC50 were used to calculate p-values by extra sum-of-squares F test." (Figure 3bc), readers may need more specific demonstration in supplemental figures on how statistical analyses have been performed.

We appreciate the reviewer’s suggestion to include more detailed information regarding the statistical analyses. For clarification, data presented in Figures 6b and 3bc were analyzed using the same statistical analysis in Prism 10. Specifically, nonlinear regression (curve fit) was performed using the [Agonist] vs. response model with three parameters. Best-fit values for the top and EC₅₀ parameters were compared using an extra sum-of-squares F test.No constraints were applied to the bottom and top parameters, and the EC₅₀ parameter was constrained to be greater than 0 for p-value calculation. We have revised the Statistical Analysis section of the Methods to more clearly describe this approach.

(10) In discussion, the background section for Syt1 and Syt7 is more appropriate to be in the introduction. This will allow readers to better understand the association of Syt proteins with MR1 and the necessity to study the impact of Syt on MR1 trafficking.

We thank the reviewer for this suggestion. We believe that the basic background and relevance of Syt1 and Syt7 in MR1 trafficking are covered in the introduction; however, we have added details to help readers understand their impact.

Reviewer #2 (Recommendations for the authors):

This reviewer has no requests for implementation and congratulates the authors on this nice piece of work.

We thank the reviewer for the positive comments.

Reviewer #3 (Recommendations for the authors):

Complete trafficking experiments to pinpoint the trafficking relationship between Syt 1 and 7 and MR1 in MtB infection.

We appreciate the reviewer’s insightful comment. As this study represents the first detailed investigation into the roles of Syt1 and Syt7 in MR1-mediated presentation of Mtb-derived metabolites, we agree that a fully resolved trafficking mechanism has not yet been established. A major limitation is the inability to directly track Mtb-derived antigens as they are loaded onto MR1 and trafficked to the cell surface. Therefore, we relied on complementary functional and microscopy-based approaches, including IFN-γ ELISpot assays, flow cytometry, fluorescence microscopy, and flow organellometry, to infer the trafficking relationships between Syt1, Syt7, and MR1 during intracellular Mtb infection. Our data support a model that Syt1 and Syt7 facilitates the trafficking of MR1 from Mtb-containing vacuoles to the plasma membrane. This interpretation is supported with the increased accumulation of MR1 in Mtb-containing vacuoles and reduction in MAIT cell IFN-γ release observed in Syt1 and Syt7 knockout cells.

References

(1) Becker, S. M., Delamarre, L., Mellman, I., & Andrews, N. W. (2009). Differential role of the Ca(2+) sensor synaptotagmin VII in macrophages and dendritic cells. Immunobiology, 214(7), 495–505.

(2) Brower, R. C., England, R., Takeshita, T., Kozlowski, S., Margulies, D. H., Berzofsky, J. A., & Delisi, C. (1994). Minimal requirements for peptide-mediated activation of CD8+ CTL. Molecular immunology, 31(16), 1285–1293.

(3) Harriff, M. J., Karamooz, E., Burr, A., Grant, W. F., Canfield, E. T., Sorensen, M. L., Moita, L. F., & Lewinsohn, D. M. (2016). Endosomal MR1 Trafficking Plays a Key Role in Presentation of Mycobacterium tuberculosis Ligands to MAIT Cells. PLoS pathogens, 12(3), e1005524.

(4) Karamooz, E., Harriff, M. J., Narayanan, G. A., Worley, A., & Lewinsohn, D. M. (2019). MR1 recycling and blockade of endosomal trafficking reveal distinguishable antigen presentation pathways between Mycobacterium tuberculosis infection and exogenously delivered antigens. Scientific reports, 9(1), 4797.

(5) Karamooz, E., Kim, S. J., Peterson, J. C., Tammen, A. E., Soma, S., Soll, A. C. R., Meermeier, E. W., Khuzwayo, S., & Lewinsohn, D. M. (2025). Two-pore channels in MR1-dependent presentation of Mycobacterium tuberculosis infection. PLoS pathogens, 21(8), e1013342.

(6) Kulicke, C. A., Swarbrick, G. M., Ladd, N. A., Cansler, M., Null, M., Worley, A., Lemon, C., Ahmed, T., Bennett, J., Lust, T. N., Heisler, C. M., Huber, M. E., Krawic, J. R., Ankley, L. M., McBride, S. K., Tafesse, F. G., Olive, A. J., Hildebrand, W. H., Lewinsohn, D. A., Adams, E. J., … Harriff, M. J. (2024). Delivery of loaded MR1 monomer results in efficient ligand exchange to host MR1 and subsequent MR1T cell activation. Communications biology, 7(1), 228.

(7) Shimonkevitz, R., Kappler, J., Marrack, P., & Grey, H. (1983). Antigen recognition by H-2restricted T cells. I. Cell-free antigen processing. The Journal of Experimental Medicine, 158(2), 303–316.

(8) Sykulev, Y., Cohen, R. J., & Eisen, H. N. (1995). The law of mass action governs antigen-stimulated cytolytic activity of CD8+ cytotoxic T lymphocytes. Proceedings of the National Academy of Sciences of the United States of America, 92(26), 11990–11992.

(9) Sykulev, Y., Joo, M., Vturina, I., Tsomides, T. J., & Eisen, H. N. (1996). Evidence that a single peptide-MHC complex on a target cell can elicit a cytolytic T cell response. Immunity, 4(6), 565– 571.

(10) Watts, T. H., Gariépy, J., Schoolnik, G. K., & McConnell, H. M. (1985). T-cell activation by peptide antigen: effect of peptide sequence and method of antigen presentation. Proceedings of the National Academy of Sciences of the United States of America, 82(16), 5480–5484.

Author response:

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

eLife Assessment

This valuable work investigates the role of protein N-glycosylation in regulating T-cell activation and function and suggests that B4GALT1 is a potential target for tumor immunotherapy. The strength of evidence is solid, and further mechanistic validation could be provided.

We sincerely thank the editor and reviewers for their time and constructive feedback. Your recognition of our work is much appreciated. We clarify our mechanistic studies as stated below.

Public Reviews:

Reviewer #1 (Public review):

Summary:

The study by Yu et al investigated the role of protein N-glycosylation in regulating T-cell activation and functions is an interesting work. By using genome-wide CRISPR/Cas9 screenings, the authors found that B4GALT1 deficiency could activate expression of PD-1 and enhance functions of CD8+ T cells both in vitro and in vivo, suggesting the important roles of protein N-glycosylation in regulating functions of CD8+ T cells, which indicates that B4GALT1 is a potential target for tumor immunotherapy.

Strengths:

The strengths of this study are the findings of novel function of B4GALT1 deficiency in CD8 T cells.

Weaknesses:

However, authors did not directly demonstrate that B4GALT1 deficiency regulates the interaction between TCR and CD8, as well as functional outcomes of this interaction, such as TCR signaling enhancements.

We are very sorry that we did not highlight our results in Fig. 5f-h enough. In those figures, we demonstrated the interaction between TCR and CD8 increased significantly in B4GALT1 deficient T-cells, by FRET assays. To confirm the important role of TCR-CD8 interaction in mediating the functions of B4GALT1 in regulating T-cell functions, such as in vitro killing of target cells, we artificially tethered TCR and CD8 by a CD8β-CD3ε fusion protein and tested its functions in both WT and B4GALT1 knockout CD8⁺ T-cell. Our results demonstrate that such fusion protein could bypass the effect of B4GALT1 knockout in CD8⁺ T-cells (Fig. 5g-h). Together with the results that B4GALT1 directly regulates the galactosylation of TCR and CD8, those results strongly support the model that B4GALT1 modulates T-cell functions mainly by galactosylations of TCR and CD8 that interfere their interaction.

Reviewer #2 (Public review):

Summary:

In this study, the authors identify the N-glycosylation factor B4GALT1 as an important regulator of CD8 T-cell function.

Strengths:

(1) The use of complementary ex vivo and in vivo CRISPR screens is commendable and provides a useful dataset for future studies of CD8 T-cell biology.

(2) The authors perform multiple untargeted analyses (RNAseq, glycoproteomics) to hone their model on how B4GALT1 functions in CD8 T-cell activation.

(3) B4GALT1 is shown to be important in both in vitro T-cell killing assays and a mouse model of tumor control, reinforcing the authors' claims.

Weaknesses:

(1) The authors did not verify the efficiency of knockout in their single-gene KO lines.

Thank reviewer for reminding. We verified the efficiency of some gRNAs by T7E1 assay. We will add those data in supplementary results in revised version later.

(2) As B4GALT1 is a general N-glycosylation factor, the phenotypes the authors observe could formally be attributable to indirect effects on glycosylation of other proteins.

Please see response to reviewer #1.

(3) The specific N-glycosylation sites of TCR and CD8 are not identified, and would be helpful for site-specific mutational analysis to further the authors' model.

Thank reviewer for suggestion! Unfortunately, there are multiple-sites of TCR and CD8 involved in N-glycosylation (https://glycosmos.org/glycomeatlas). We worry that mutations of all these sites may not only affect glycosylation of TCR and CD8 but also other essential functions of those proteins.

(4) The study could benefit from further in vivo experiments testing the role of B4GALT1 in other physiological contexts relevant to CD8 T cells, for example, autoimmune disease or infectious disease.

Thank reviewer for this great suggestion to expand the roles of B4GALT1 in autoimmune and infection diseases. However, since in current manuscript we are mainly focusing on tumor immunology, we think we should leave these studies for future works.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The study by Yu et al investigated the role of protein N-glycosylation in regulating T-cell activation and functions is an interesting work. By using genome-wide CRISPR/Cas9 screenings, the authors found that B4GALT1 deficiency could activate expression of PD-1 and enhance functions of CD8+ T cells both in vitro and in vivo, suggesting the important roles of protein N-glycosylation in regulating functions of CD8+ T cells, which indicates that B4GALT1 is a potential target for tumor immunotherapy. However, authors need to directly demonstrate that B4GALT1 deficiency regulates the interaction between TCR and CD8, as well as functional outcomes of this interaction, such as TCR signaling enhancements. In addition, blocking PD1 has been shown to enhance antitumor effect, whereas the presented data in this study suggest that the activation of PD1 expression in the condition of B4GALT1 deficiency in T cells enhanced antitumor effect. How to reconcile this discrepancy? Finally, several minor questions need to be addressed to strengthen the conclusions in this manuscript.

(1) We used a FRET (Fluorescence Resonance Energy Transfer) assay to measure interaction between TCR and CD8. FRET signals of TCR-CD8 increased significantly in B4GALT1 deficient T-cells, compared with control cells (Fig. 5f). For functional outcomes of this interaction, we observed enhanced T-cell killing activities in B4GALT1 deficient CD8⁺ T-cells (Fig. 3f and Fig. 5h).

To confirm whether reduced TCR-CD8 interaction is the major cause of TCR activation phenotypes in B4GALT1 knockout CD8⁺ T-cells, we generated a construct in which we fused the CD8b ectodomain (ECD) with CD3e to artificially tether TCR with CD8 (Fig.5g). Overexpression of such CD8β-CD3ε fusion led to enhanced in vitro killing activities in control wild-type CD8⁺ T-cells. On the other hand, in B4GALT1 deficient CD8⁺T-cells, such enhanced T-cell killing activities by fusion construct was significantly diminished (Fig.5h), suggesting it bypassed the regulation by B4GALT1.

(2) PD-1 is both an early T-cell activation marker upon TCR activation and a T-exhausted marker under consecutive or repeated stimulations. In our screenings, PD-1 was used as an early activation marker for T-cells.

We have clarified this in new Discussion section.

(1) The present data relies on statistical graphs (e.g., bar and line charts) for all data, excluding the bioinformatics analysis. Including data such as flow cytometry plots, photomicrographs, or immunohistochemistry staining images will provide more direct support for the conclusions.

Thank the reviewer for valuable suggestions! We added original flow cytometry gating strategies for Cas9 screening sorting (Fig. S1a), TIL analysis (Fig.S5), and FRET assay (Fig. S8) in revised version to provide more direct support for our conclusions.

(2) To further validate the enhanced tumor infiltration phenotype resulting from B4GALT1 knockout, the following data would strengthen the manuscript:

(a) Flow cytometric analysis of TILs or immunofluorescence data from tumor sections.

Thank the reviewer for valuable suggestion! We added original flow cytometry gating strategies for TILs in Fig. S5 in revised version.

(b) Assessment of in vivo T cell proliferation, for example, by tracking changes in the proportion of CD8+ T cells in the peripheral blood over time.

We analyzed in vivo T-cell proliferation within tumor by CFSE (carboxyfluorescein succinimidyl ester) analysis. As shown in Fig. S6b, 6 days after infusion, B4GALT1 knockout OT-I T-cell showed increased proliferation within tumors, comparing with wild type control OT-I cells.

(c) Evaluation of the proliferation and activation status of OT-1 CD8+ T cells specifically in the draining lymph nodes of the mouse model.

Thank the reviewer for valuable suggestion! We plan to perform this experiment in the future.

(3) The authors provide evidence that B4GALT1 knockout enhances CD8+ T cell function in both mouse models and human TCR-T cells (in vitro). Definitive support for the translational potential of this strategy would come from showing that B4GALT1-knockout human TCR-T cells also mediate potent in vivo function (NSG tumor-bearing model may be a better choice).

Thank the reviewer for valuable suggestion! We are going to perform those experiments in the future. However, we do not expect that in vitro and in vivo (NSG mice) experiments will show much different results, which may also not add too much for current manuscript.

(4) It would be preferable to include data on T cell activation and effector function (e.g., flow cytometry for IL-2, TNF-α, and IFN-γ, or ELISPOT) following stimulation with an OVA-specific peptide or co-culturing of OVA-expressing tumor cells with B4GALT1-knockout OT-1 CD8 T cells, especially the changes in the TILs compared with the non-targeting control group.

Following co-culturing of B16-OVA tumor cells with B4GALT1-knockout or wild-type OT-I CD8⁺ T-cells, the RNA levels and secretion levels of TNFα and IFNγ were detected by RT-qPCR and ELISA, respectively (Fig. 3c). B4GALT1-deficient OT-I T-cells showed increased expression of T-cell activation and cytotoxic markers such as IFNγ and TNFα.

(5) What is the correlation between the expression of B4GALT1, PD-1, and TCR activation markers at various time points during a long-term T cell co-culture with tumor cells?

Thanks for the reviewer for valuable suggestion! We don’t have this data now. While we agree that exploring this might be interesting, we think it falls outside the scope of the current study.

(6) In line 136: Regarding the genetic targeting of B4GALT1 in T cells, it is unclear whether single or multiple gRNAs were used and if potential off-target effects were assessed. To fully validate the model, it would be important to clarify these strategies, and it is essential to include data on the knockout efficiency at both the protein (e.g., Western blot) and mRNA levels.

We are sorry about the unclear statements for gene knockout strategy. In current study, single sgRNAs were used in all experiments for gene knockout. B4galt1 sg2 was used in Fig. 3a. Both B4galt1 sg1 and sg2 were used in Fig. S1d. We clarified this in each figure legend in revised version.

The phenotypes of B4galt1 knockout T-cells could be rescued by overexpression of either a short or long isoform of mouse B4galt1 cDNA (Fig. 3b), indicating that potential off-target effects could be excluded.

The sgRNA knockout efficiencies were confirmed by T7E1 assay in revised version (Fig. S2). Regrettably, anti-mouse B4galt1 antibody didn’t work in western blot.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

(1) Rationale for excluding clades G and H and clarification of clade definitions

We appreciate this important request for clarification. In the revised manuscript, we now explicitly state (Methods, Tree generation) that the phylogenetic framework used in this study follows the clade definitions established by Techtmann et al. (Front. Microbiol. 2012, 3, 132), which classify [NiFe]-CODHs into clades based on high supporting values in nodes (bootstrap >75). We deem Techtmann et al.’s work as best lead, since their approach with two different types of trees (ML vs. Bayesian) gives solid support to this classification of clades. We ourselves did not perform Bayesian statistics, instead we used the known clades from literature to assign ours.

Clades G and H were not deliberately excluded from downstream genomic-context and operon analyses. They were excluded by our pipeline, because their data did not fulfil our initial quality assessments, such as: host classified down to species level (https://github.com/boehmax/protein-per-organism), and protein exists in the IPG database of NCBI (https://github.com/boehmax/protein-to-genome).

Clade G and H are both represented by only a very small number of sequences, most of which derive from fragmented or poorly annotated genomes, preventing reliable assessment of operon organization and gene neighbourhood conservation. As a result, inclusion of these clades would not allow statistically meaningful or biologically interpretable comparisons with clades A–F.

To improve transparency, we have added a brief explanation of these limitations in the Results (Results, Neighbor analysis).

(2) Presentation and interpretation of co-occurrence data

We agree that the presentation of the co-occurrence data required improvement. In the revised supplementary material, we now include a table in the long format that might be easier to interpret than a matrix representation as seen in Fig. 3B.

We have also revised the Results text to more precisely reflect the numerical trends. Specifically, we clarify that clade D shows co-occurrence with clades A, E and F, while clade C only displays co-occurrence with clade E. The statement that clades C and D “more often co-occur” has been removed and rephrased to avoid overgeneralization and to better align with the quantitative data shown in Figure 3B and the supplementary table (Results, Co-occurrence and Correlation).

(3) Rationale for operon-level rather than organism-level analysis

We thank the reviewer for highlighting this conceptual point. In the revised manuscript, we now explicitly state that our analysis was conducted at the operon level because individual genomes frequently encode multiple CODH operons that are phylogenetically and functionally distinct. Treating each operon as an independent functional unit allows us to capture this intra-genomic diversity and to associate specific gene neighbourhoods with individual CODH clades. We furthermore discuss in the introduction explicitly technical reasons why we decided to limit this study to the operon level for more transparency.

Nevertheless, we acknowledge that this approach may overlook higher-level regulatory or physiological interactions among multiple CODHs encoded within the same genome. This limitation is now discussed explicitly, and we acknowledge that operon-level analysis should be a complementary, not exhaustive, framework for functional inference.

Reviewer #2 (Public review):

We thank Reviewer #2 for their positive assessment of the conceptual clarity and methodological utility of our approach, as well as for their thoughtful discussion of its limitations.

Regarding incomplete genome assemblies, limited representation of class II HCPs, and potential omission of distal pathway components, we agree fully. We stress that our conclusions are probabilistic and hypothesis-generating rather than definitive functional assignments.

In response to the concern about reproducibility of the visual filtering step, we have added a more explicit description (Methods, Data collection and refinement) of the criteria used to exclude non-CODH homologs (e.g., absence of conserved active-site motifs, unknown folds predicted with AlphaFold3, extremely long tree branches). This clarification improves transparency and facilitates replication of the analysis.

Finally, we concur that extrapolating enzymatic activity or inactivity from a limited number of characterized representatives should be done cautiously. We have revised the wording throughout the manuscript to further temper such generalizations and to frame our interpretations explicitly as predictions that require experimental validation.

Once again, we thank both reviewers for their constructive feedback, which has significantly improved the clarity, rigor, and transparency of the manuscript. We believe that the revisions address all concerns raised and strengthen the overall contribution of this work.

Recommendation from authors:

Reviewer #1 (Recommendations for the authors):

All suggested editorial and stylistic corrections were implemented. These include refinements to the wording in the Abstract, grammatical corrections, streamlined phrasing, standardized figure callouts and supplementary file references, corrected abbreviations, and consistent formatting of references and author names. The only exception concerns the suggested change from MetCODH to MtCODH. We have retained MetCODH, as this abbreviation is well established in the literature for the Methanothermobacter thermophila CODH and is commonly used in prior studies (e.g., https://doi.org/10.1073/pnas.2410995121 ). MtCODH has historically been referring to CODH from Neomoorella thermoacetica (previously Moorella thermoacetica, hence the abbreviation Mt). We chose to rename that to NtCODH but to avoid confusion, keep MetCODH for Methanothermobacter thermophila.

Reviewer #2 (Recommendations for the authors):

We likewise addressed the majority of recommendations. We now report the versions of all software tools and databases used, standardized capitalization and naming of software and platforms (e.g., GitHub, eggNOG), clarified the BLAST implementation and database employed, and added direct repository links for custom scripts in both the Methods section and the bibliography. Overall grammatical consistency and formatting were improved throughout the manuscript. In addition, the criteria and procedure used for visual inspection to remove non-CODH sequences are now described more explicitly to enhance reproducibility, and several methodological sections were streamlined as suggested. Minor textual redundancies were removed, and phrasing was simplified where appropriate.

Figure legends and formatting were revised to improve clarity and consistency. Adjustments to color usage and font consistency were made where feasible to enhance readability. The color scheme in Figure 1 was adjusted as suggested, and darker shades were chosen for clade H and G. This change was also implemented in the Supplementary File 9_Tree5. Figure 3A was retained, as it provides information on the frequency of multiple CODHs from the same clade within genomes, which cannot be inferred from the probability matrix shown in Figure 3B; together, these panels offer complementary insights. We adjusted the figure caption to make this clearer. We increased the visibility of data points in Figure 4B. To allow inclusion of the full dataset we did not collapse the x-axis as suggested. Figure 4C was retained in its original format to emphasize the characteristic operon “fingerprints” of each CODH clade, which is a central focus of this work. A table is supplied in Supplementary File 2, which allows data exploration with the preferred focus of the reader.

A small number of suggestions were therefore not implemented exactly as proposed, primarily where alternative revisions were judged to better preserve clarity or analytical intent. These decisions are minor and do not affect the conclusions or reproducibility of the study.

Overall, we believe that these revisions have substantially improved the manuscript’s readability, transparency, and technical rigor, and we thank the reviewers again for their careful and constructive feedback.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

This study compares four models - VALOR (dynamic visual-text alignment), CLIP (static visual-text alignment), AlexNet (vision-only), and WordNet (text-only) - in their ability to predict human brain responses using voxel-wise encoding modeling. The results show that VALOR not only achieves the highest accuracy in predicting neural responses but also generalizes more effectively to novel datasets. In addition, VALOR captures meaningful semantic dimensions across the cortical surface and demonstrates impressive predictive power for brain responses elicited by future events.

Strengths:

The study leverages a multimodal machine learning model to investigate how the human brain aligns visual and textual information. Overall, the manuscript is logically organized, clearly written, and easy to follow. The results well support the main conclusions of the paper.

(1) My primary concern is that the performance difference between VALOR and CLIP is not sufficiently explained. Both models are trained using contrastive learning on visual and textual inputs, yet CLIP performs significantly worse. The authors suggest that this may be due to VALOR being trained on dynamic movie data while CLIP is trained on static images. However, this explanation remains speculative. More in-depth discussion is needed on the architectural and inductive biases of the two models, and how these may contribute to their differences in modeling brain responses.

Thank you for this thoughtful comment. We agree that attributing VALOR’s advantage over CLIP solely to ‘dynamic (video) versus static (image) pretraining’ would be incomplete, and that the architectural and inductive biases of the two models are central to understanding the observed performance gap.

Both VALOR and CLIP use contrastive learning to align visual and textual representations, but they differ in several key inductive biases that are particularly relevant for modeling brain responses during continuous movie viewing. First, VALOR is trained to align temporally extended video segments with text, introducing an explicit temporal integration window that aggregates information across consecutive frames. This encourages representations that maintain context, stabilize semantics across time, and encode event-level structure. Second, VALOR’s alignment operates at the level of multi-second narrative units, rather than isolated visual snapshots, biasing the model toward representations that are sensitive to unfolding events and cross-frame consistency.

In contrast, CLIP processes frames independently and aligns single static images with text. As a result, it lacks an intrinsic mechanism for temporal binding, context accumulation, or event-level representation. While CLIP can capture rich visual–semantic associations at the image level, it is less well suited to represent higher-order temporal structure, which is known to strongly drive responses in association cortex during naturalistic narrative perception.

We therefore interpret VALOR’s superior encoding performance as reflecting not only exposure to dynamic audiovisual data, but also inductive biases—temporal integration and event-level alignment—that more closely match how the brain integrates information over time during movie watching. We have revised the Discussion (p. 16) to articulate these architectural and representational differences explicitly, rather than attributing the effect solely to training data modality.

(On page 16) “Additionally, VALOR exceeds the performance of CLIP, a leading static multimodal model, as its training objective aligns multi-second video–text units, enforcing a temporal integration window and event-level semantics that maintain cross-frame consistency and narrative context, whereas CLIP’s image-level alignment provides no intrinsic mechanism for such temporal continuity.”

(2) The methods section lacks clarity regarding which layers of VALOR and CLIP were used to extract features for voxel-wise encoding modeling. A more detailed methodological description is necessary to ensure reproducibility and interpretability. Furthermore, discussion of the inductive biases inherent in these models-and their implications for brain alignment - is crucial.

Thank you for this comment. We agree that reproducibility and interpretability require precise specification of which model representations were used for voxel-wise encoding, as well as clearer discussion of the inductive biases inherent in these models and their implications for brain alignment.

In the revised Methods, we now explicitly specify the feature sources for both models. For CLIP (ViT-B/32), we use the final pooled image embedding after projection into the shared image–text space, extracted frame-by-frame; one representative frame is sampled per TR, and its projected embedding serves as the regressor. For VALOR, we use the final joint video–text projection head, yielding a 512-dimensional embedding computed at the segment/TR level that integrates information across consecutive frames and aligns each multi-second video segment with its associated text. These procedures are now described step-by-step in the Methods (p. 21).

In addition, we expanded the Discussion (p. 16) to explicitly articulate the models’ inductive biases and their relevance for brain alignment. In particular, we contrast CLIP’s image-level, framewise alignment—which lacks intrinsic temporal integration—with VALOR’s event-level, temporally extended video–text alignment, which biases representations toward context maintenance and narrative continuity. This distinction helps explain why the two models differ in their ability to predict neural responses during continuous movie viewing.

(Methods, On page 21)

“(1) Video–text alignment features (VALOR): To extract video-based multimodal features, we used VALOR (VALOR-large checkpoint), an open-source pretrained video–text alignment model24. VALOR combines visual encoders (CLIP and Video Swin Transformer) for extracting visual features and a text encoder (BERT) for extracting textual features 23,51,52. These representations are aligned in a shared embedding space through contrastive learning. We segmented each movie at the TR level and, for each segment, extracted VALOR’s projected video–text embedding from the final projection head of the alignment module to obtain a 512-dimensional feature vector. These embeddings were then time-aligned to the corresponding BOLD responses.

(2) CLIP features: To compare with static image-based multimodal models, we utilized CLIP (ViT-B/32), which aligns visual and textual representations through contrastive learning but processes individual frames independently without capturing temporal context. One video frame was sampled per TR, and the pooled image embedding after CLIP’s projection into the shared image–text space was extracted to obtain a 512-dimensional feature vector. These TR-aligned vectors were used directly as regressors in the voxel-wise encoding models.”

(Discussion, On page 16)

“Additionally, VALOR exceeds the performance of CLIP, a leading static multimodal model, as its training objective aligns multi-second video–text units, enforcing a temporal integration window and event-level semantics that maintain cross-frame consistency and narrative context, whereas CLIP’s image-level alignment provides no intrinsic mechanism for such temporal continuity. More broadly, this difference reflects distinct inductive biases in how the two models represent visual–linguistic information. CLIP is optimized for framewise image–text correspondence, encouraging representations that emphasize instantaneous visual semantics but remain agnostic to temporal structure. In contrast, VALOR is explicitly biased toward aggregating information over multiple consecutive frames and aligning representations at the level of temporally extended events. These inductive biases favor context maintenance, semantic stabilization, and narrative coherence over time, which are known to be critical for driving responses in association cortex during continuous movie perception.”

(3) A broader question remains insufficiently addressed: what is the purpose of visual-text alignment in the human brain? One hypothesis is that it supports the formation of abstract semantic representations that rely on no specific input modality. While VALOR performs well in voxel-wise encoding, it is unclear whether this necessarily indicates the emergence of such abstract semantics. The authors are encouraged to discuss how the computational architecture of VALOR may reflect this alignment mechanism and what implications it has for understanding brain function.

Thank you for this important conceptual question. We agree that improved voxel-wise encoding performance does not, by itself, imply the emergence of fully amodal or modality-independent semantic representations in the brain. In the revision, we therefore avoid framing our findings as evidence for abstract amodal semantics and instead clarify a more constrained interpretation.

Specifically, we suggest that visual–text alignment may support the stabilization and coordination of scene-level meaning across modalities and over time, rather than the formation of modality-free semantic codes. From this perspective, VALOR’s advantage reflects inductive biases that promote (i) integration of visual information over multi-second windows and (ii) alignment of temporally extended visual events with linguistic descriptions, yielding representations that are more temporally stable, context-sensitive, and constrained by language.

We therefore interpret VALOR’s superior encoding performance as identifying cortical regions whose responses are better captured by temporally stabilized, cross-modal representations, rather than as evidence that these regions encode fully abstract semantics independent of input modality. We have expanded the Discussion (p. 16) to articulate this interpretation and to clarify the implications of video–text alignment for understanding how the brain integrates perception and language during naturalistic cognition.

(On page 16) “Together, the relative gains over AlexNet (purely visual), WordNet (manual semantic annotation), and CLIP (static image–text alignment) indicate cortical systems whose responses are best captured by multi-second, multimodal integration, and highlight regions that accumulate and stabilize narrative context over time. At the same time, these findings do not imply that visual–text alignment in the brain gives rise to fully amodal, modality-independent semantic representations. Instead, we suggest that alignment between visual and linguistic signals may serve to stabilize and coordinate scene-level meaning across modalities and over time. From this perspective, VALOR’s architecture—by integrating visual information over multi-second windows and aligning temporally extended video segments with language—provides a computational proxy for how the brain may use linguistic constraints to organize, disambiguate, and maintain coherent representations of unfolding events. The observed encoding gains therefore highlight regions engaged in temporally stabilized, cross-modal integration during naturalistic perception, rather than providing evidence for abstract semantic codes divorced from sensory input.”

(4) The current methods section does not provide enough details about the network architectures, parameter settings, or whether pretrained models were used. If so, please provide links to the pretrained models to facilitate reproducible science.

We appreciate this comment and agree that our original description of model sources and implementation details was not sufficiently explicit. These details are essential for both reproducibility and interpretability. We have now made these specifications explicit in the revised Methods.

In particular, we now state for each model:

VALOR. We use the publicly released pretrained VALOR-large checkpoint. For each movie segment, we extract the joint video–text projection head output (512-D) that encodes the aligned segment-level audiovisual semantics. We report the checkpoint source, the segment duration (in frames/seconds), and how these segment-level embeddings are temporally aligned to TRs for voxel-wise encoding.

CLIP (ViT-B/32). We use the standard pretrained CLIP weights. For each video frame, we extract the final pooled image representation after projection into CLIP’s shared image–text embedding space (512-D). We also clarify that one representative frame is sampled and aligned to each TR, and that these projected embeddings are used as regressors in the encoding model.

AlexNet. We use the ImageNet-pretrained AlexNet. We take activations from conv5, and then apply PCA to reduce them to 512 dimensions before mapping them to the fMRI time series.

For each model, the revised Methods now specify: the pretrained source/checkpoint, the layer or head from which features were taken, output dimensionality, any preprocessing or dimensionality reduction, and the temporal alignment procedure used to generate TR-level regressors. These revisions appear in the updated Methods (page 21).

(On page 21) “(1) Video–text alignment features (VALOR): To extract video-based multimodal features, we used VALOR (VALOR-large checkpoint), an open-source pretrained video–text alignment model24. VALOR combines visual encoders (CLIP and Video Swin Transformer) for extracting visual features and a text encoder (BERT) for extracting textual features 23,51,52. These representations are aligned in a shared embedding space through contrastive learning. We segmented each movie at the TR level and, for each segment, extracted VALOR’s projected video–text embedding from the final projection head of the alignment module to obtain a 512-dimensional feature vector. These embeddings were then time-aligned to the corresponding BOLD responses.

(2) P features: To compare with static image-based multimodal models, we utilized CLIP (ViT-B/32), which aligns visual and textual representations through contrastive learning but processes individual frames independently without capturing temporal context. One video frame was sampled per TR, and the pooled image embedding after CLIP’s projection into the shared image–text space was extracted to obtain a 512-dimensional feature vector. These TR-aligned vectors were used directly as regressors in the voxel-wise encoding models.

(3) AlexNet features: Visual features were extracted by sampling frames at the TR level and processing them with AlexNet, an eight-layer convolutional neural network comprising five convolutional layers followed by three fully connected layers. Features from all five convolutional layers were evaluated in preliminary analyses; the fifth convolutional layer showed the best performance and was used in subsequent analyses. Intra-image z-score normalization was applied to reduce amplitude effects. Principal component analysis (PCA) was used to reduce dimensionality, retaining the top 512 components to match the dimensionality of multimodal feature spaces. This pipeline was implemented using the DNNBrain toolkit 53.

(4) WordNet features: Semantic features were obtained from publicly available WordNet annotations provided with the HCP dataset (7T_movie_resources/WordNetFeatures.hdf5), following the procedure of Huth et al. (2012). Each second of the movie clips was manually annotated with WordNet categories according to predefined guidelines: (a) identifying clear objects and actions in the scene; (b) labeling categories that dominated for more than half of the segment duration; and (c) using specific category labels rather than general ones. A semantic feature matrix was constructed with rows corresponding to time points and columns to semantic categories, with category presence coded as binary values. More specific categories from the WordNet hierarchy were added to each labeled category, yielding a total of 859 semantic features. These features were used directly as regressors. We also evaluated a PCA-reduced 512-dimensional variant (fit within each training fold to avoid leakage); because this version performed slightly worse, we report results from the full 859-dimensional representation in the main text. For the generalization analysis in Study 2, annotations for the SFM dataset were aligned to the same WordNet category space to ensure consistency.”

Reviewer #2 (Public review):

Fu and colleagues have shown that VALOR, a model of multimodal and dynamic stimulus features, better predicts brain responses compared to unimodal or static models such as AlexNet, WordNet, or CLIP. The authors demonstrated the robustness of their findings by generalizing encoding results to an external dataset. They demonstrated the models' practical benefit by showing that semantic mappings were comparable to another model that required labor-intensive manual annotation. Finally, the authors showed that the model reveals predictive coding mechanisms of the brain, which held a meaningful relationship with individuals' fluid intelligence measures.

Strengths:

Recent advances in neural network models that extract visual, linguistic, and semantic features from real-world stimuli have enabled neuroscientists to build encoding models that predict brain responses from these features. Higher prediction accuracy indicates greater explained variance in neural activity, and therefore a better model of brain function. Commonly used models include AlexNet for visual features, WordNet for audio-semantic features, and CLIP for visuo-semantic features; these served as comparison models in the study. Building on this line of work, the authors developed an encoding model using VALOR, which captures the multimodal and dynamic nature of real-world stimuli. VALOR outperformed the comparison models in predicting brain responses. It also recapitulated known semantic mappings and revealed evidence of predictive processing in the brain. These findings support VALOR as a strong candidate model of brain function.

(1) The authors argue that this modeling contributes to a better understanding of how the brain works. However, upon reading, I am less convinced about how VALOR's superior performance over other models tells us more about the brain. VALOR is a better model of the audiovisual stimulus because it processes multimodal and dynamic stimuli compared to other unimodal or static models. If the model better captures real-world stimuli, then I almost feel that it has to better capture brain responses, assuming that the brain is a system that is optimized to process multimodal and dynamic inputs from the real world. The authors could strengthen the manuscript if the significance of their encoding model findings were better explained.

We thank the reviewer for this thoughtful comment and agree with the premise that a model preserving multimodal and temporal structure might a priori be expected to better predict brain responses to naturalistic stimuli. Our intent is not to claim that higher accuracy alone explains brain function, but rather that where and how VALOR improves prediction provides diagnostic insight into cortical processing. We have revised the Discussion to make this distinction explicit.

Specifically, we clarify three ways in which VALOR’s gains are scientifically informative rather than merely unsurprising:

(1) Anatomical specificity of improvement. VALOR’s advantage is not uniform across the cortex; gains are largest in regions implicated in multi-second, cross-modal integration. This spatial pattern constrains where the brain accumulates information over time and stabilizes visual representations using linguistic context.

(2) Model as a computational probe. Beyond prediction accuracy, VALOR’s feature space recovers large-scale semantic organization without manual annotation and enables targeted tests of predictive processing. Features reflecting upcoming content selectively improve fits in specific regions, consistent with anticipatory coding during continuous narrative perception.

(3) Link to individual differences. Individuals whose neural responses are better captured by anticipatory features show higher fluid intelligence, suggesting that VALOR indexes meaningful variability in forward-looking representations rather than merely tracking stimulus complexity.

Accordingly, we have revised the Discussion (p. 16) to frame VALOR as a tool for mapping cortical integration profiles, probing semantic and predictive structure, and linking representational dynamics to cognition, rather than asserting that higher encoding accuracy alone explains brain function.

(On page 16) “Together, the relative gains over AlexNet (purely visual), WordNet (manual semantic annotation), and CLIP (static image–text alignment) indicate cortical systems whose responses are best captured by multi-second, multimodal integration, and highlight regions that accumulate and stabilize narrative context over time.”

(2) In Study 3, the authors show high alignment between WordNet and VALOR feature PCs. Upon reading the method together with Figure 3, I suspect that the alignment almost has to be high, given that the authors projected VALOR features to the Huth et al.'s PC space. Could the authors conduct non-parametric permutation tests, such as shuffling the VALOR features prior to mapping onto Huth et al.'s PC space, and then calculating the Jaccard scores? I imagine that the null distribution would be positively shifted. Still, I would be convinced if the alignment is higher than this shifted null distribution for each PC. If my understanding of this is incorrect, I suggest editing the relevant Method section (line 508) because this analysis was not easy to understand.

Thank you for this helpful comment and for pointing out a potential source of confusion. We apologize that the original Methods description was not sufficiently clear. Importantly, VALOR features were never projected into the Huth et al. PC space, and no optimization or rotation toward the WordNet basis occurred at any stage.

The analysis proceeded as follows:

(1) VALOR PCs. We first fit voxel-wise encoding models using VALOR features on the Huth et al. dataset. We then applied PCA to the resulting cortical weight maps, yielding spatial components (‘VALOR PCs’) that summarize shared patterns of VALOR feature weights across the cortex.

(2) WordNet PCs. We used the semantic principal components reported by Huth et al. (2012) directly as published, with no refitting, projection, or modification using VALOR.

(3) Correspondence analysis. Only after obtaining these two independent sets of cortical maps did we threshold each to their top-loading vertices and compute Jaccard overlap between VALOR PCs and WordNet PCs.

Although a permutation that shuffles VALOR features prior to projection addresses a scenario that does not apply here, we agree that the Methods description should more clearly convey the independence of the two decompositions. We have therefore revised the Methods (p. 24) to describe the procedure step-by-step and explicitly state that no projection, refitting, or optimization toward the WordNet basis was performed.

(On page 24) “We first fit voxel-wise encoding models using VALOR features for each of the five participants in the Huth et al. dataset. For each participant, this yielded a weight map linking each VALOR feature to each voxel. We then stacked these weight maps across participants to form a single voxel-by-feature weight matrix and applied principal component analysis (PCA). The top four principal components from this analysis (“VALOR PCs”) captured shared spatial patterns of VALOR feature weights across cortex. To interpret these components, we projected VALOR feature vectors from >20,000 video segments in the VALOR training set onto each VALOR PC, which revealed dominant semantic axes (e.g., mobility, sociality, civilization). For comparison, we used the semantic principal components reported by Huth et al. (2012) from their WordNet-based encoding model; these “WordNet PCs” were taken directly from the published work and were not refit or reweighted using VALOR.”

(3) In Study 4, the authors show that individuals whose superior parietal gyrus (SPG) exhibited high prediction distance had high fluid cognitive scores (Figure 4C). I had a hard time believing that this was a hypothesis-driven analysis. The authors motivate the analysis that "SPG and PCu have been strongly linked to fluid intelligence (line 304)". Did the authors conduct two analyses only-SPG-fluid intelligence and PCu-fluid intelligence-without relating other brain regions to other individual differences measures? Even if so, the authors should have reported the same r-value and p-value for PCu-fluid intelligence. If SPG-fluid intelligence indeed holds specificity in terms of statistical significance compared to all possible scenarios that were tested, is this rationally an expected result, and could the authors explain the specificity? Also, the authors should explain why they considered fluid intelligence to be the proxy of one's ability to anticipate upcoming scenes during movie watching. I would have understood the rationale better if the authors had at least aggregated predictive scores for all brain regions that held significance into one summary statistic and found a significant correlation with the fluid intelligence measure.

We thank the reviewer for this careful and constructive comment and agree that greater transparency about analytic intent, specificity, and rationale is needed. We have revised the manuscript accordingly.

(1) Analytic scope and a priori restriction. The analysis in Fig. 4C was hypothesis-driven and restricted a priori to two regions — superior parietal gyrus (SPG) and precuneus (PCu) — based on convergent evidence linking frontoparietal and medial parietal systems to fluid reasoning, relational integration, and domain-general cognitive control. Importantly, we did not conduct a whole-brain search across regions or behaviors to identify the strongest correlation post hoc.

(2) Specificity and reporting. In response to the reviewer’s request, we now report the full results for both hypothesized regions. Prediction horizon in SPG showed a statistically reliable association with fluid intelligence, whereas PCu showed a positive but weaker trend that did not survive correction. Reporting both results makes the regional specificity explicit rather than implicit.

(3) Why SPG over PCu? Although both regions are implicated in fluid cognition, SPG has been more consistently linked to active maintenance and manipulation of relational structure and top-down attentional control, whereas PCu is more often associated with internally oriented and mnemonic processes. We therefore interpret the stronger SPG association as consistent with a role for sustained, externally driven predictive processing during continuous perception, rather than as evidence of exclusivity.

(4) Why fluid intelligence? We do not equate fluid intelligence with “anticipation” per se. Rather, we used gF as an a priori proxy for domain-general capacities — maintaining and updating relational context over multi-second windows, integrating multiple constraints, and exerting flexible control — that are plausibly recruited when anticipating upcoming events during naturalistic narratives. The reported relationship is associative and hypothesis-consistent, not causal.

(5) Why not aggregate across regions? We agree that aggregation could reveal more global relationships; however, our goal in this analysis was to test whether predictive timescales in theoretically motivated control regions relate to individual differences, rather than to maximize correlation by pooling heterogeneous regions. We now clarify this rationale in the Results.

These clarifications and additional statistics have been incorporated in the revised Results section (p. 14).

(On page 14) “Finally, we examined whether prediction horizons were linked to individual differences in cognition. We focused on fluid intelligence (gF) because gF is widely taken to index domain-general capacities such as maintaining and updating relational context over several seconds, integrating multiple constraints, and exerting flexible top-down control — functions that should support anticipating what will happen next in a continuous narrative. We targeted two parietal regions, the SPG and the PCu, which have both been repeatedly linked to gF and high-level cognitive control in the individual-differences literature 36,37. For each participant, we correlated fluid cognition scores with that participant’s average prediction horizon in each region. As shown in Fig. 4c, individuals with longer prediction horizons in SPG showed higher fluid cognition scores (SPG: r = 0.172, FDR-corrected p = 0.047). PCu showed a similar positive trend (PCu: r = 0.111, FDR-corrected p = 0.146) but did not reach significance. These associations suggest that the ability to sustain a longer predictive timescale during naturalistic perception co-varies with broader fluid cognitive capacity. No additional brain regions or behavioral measures were examined in this analysis.”

Reviewer #3 (Public review):

In this work, the authors aim to improve neural encoding models for naturalistic video stimuli by integrating temporally aligned multimodal features derived from a deep learning model (VALOR) to predict fMRI responses during movie viewing.

Strengths:

The major strength of the study lies in its systematic comparison across unimodal and multimodal models using large-scale, high-resolution fMRI datasets. The VALOR model demonstrates improved predictive accuracy and cross-dataset generalization. The model also reveals inherent semantic dimensions of cortical organization and can be used to evaluate the integration timescale of predictive coding.

This study demonstrates the utility of modern multimodal pretrained models for improving brain encoding in naturalistic contexts. While not conceptually novel, the application is technically sound, and the data and modeling pipeline may serve as a valuable benchmark for future studies.

(1) Lines 95-96: The authors claim that "cortical areas share a common space," citing references [22-24]. However, these references primarily support the notion that different modalities or representations can be aligned in a common embedding space from a modeling perspective, rather than providing direct evidence that cortical areas themselves are aligned in a shared neural representational space.

We thank the reviewer for this important clarification. We agree that the cited works do not provide direct evidence that cortical areas themselves are aligned in a single neural representational space. Rather, they demonstrate that representations derived from different modalities can be mapped into a shared embedding space from a modeling and computational perspective.

We have therefore revised the text to avoid overstatement and to more precisely reflect what these studies support. In the revised manuscript (p. 4), we now frame the claim in terms of a shared representational framework or feature space used for modeling, rather than implying that cortical areas themselves intrinsically share a unified neural space. This clarification aligns the conceptual claim with the scope of the cited literature.

(On page 4) “As a result, researchers are turning to multimodal deep learning, which learns from visual, linguistic, and auditory streams to model complex brain functions. This trend is supported by neuroscience evidence that cortical responses across regions can be jointly modeled within a common representational space.”

(2) The authors discuss semantic annotation as if it is still a critical component of encoding models. However, recent advances in AI-based encoding methods rely on features derived from large-scale pretrained models (e.g., CLIP, GPT), which automatically capture semantic structure without requiring explicit annotation. While the manuscript does not systematically address this transition, it is important to clarify that the use of such pretrained models is now standard in the field and should not be positioned as an innovation of the present work. Additionally, the citation of Huth et al. (2012, Neuron) to justify the use of WordNet-based annotation omits the important methodological shift in Huth et al. (2016, Nature), which moved away from manual semantic labeling altogether. Since the 2012 dataset is used primarily to enable comparison in study 3, the emphasis should not be placed on reiterating the disadvantages of semantic annotation, which have already been addressed in prior work. Instead, the manuscript's strength lies in its direct comparison between data-driven feature representations and semantic annotation based on WordNet categories. The authors should place greater emphasis on analyzing and discussing the differences revealed by these two approaches, rather than focusing mainly on the general advantage of automated semantic mapping.

Thank you for this thoughtful and constructive comment. We agree with the reviewer that the field has largely transitioned away from manual semantic annotation toward features derived from large-scale pretrained models (e.g., CLIP, GPT-style architectures), and that this shift is now standard rather than a novelty of the present work.

We have revised the manuscript to clarify this positioning. Our goal is not to claim automated semantic extraction as an innovation, but rather to demonstrate how a multimodal, temporally informed video–text model can be used as a direct feature space for voxel-wise encoding of naturalistic movie fMRI data. VALOR is used as a representative example of this broader class of pretrained models, and our emphasis is on the general modeling approach rather than on promoting a specific architecture.

We also agree that our original discussion underemphasized the important methodological shift introduced in Huth et al. (2016, Nature), which moved away from manual semantic labeling in the context of continuous spoken narratives. We now explicitly acknowledge this work and clarify that our use of WordNet-based annotations from Huth et al. (2012) serves a different purpose: it provides an interpretable, historically grounded benchmark for comparison in Study 3, rather than a claim that semantic annotation remains necessary or state-of-the-art.

In response to the reviewer’s suggestion, we have revised the Results (p.10) and Discussion (p.18) to place greater emphasis on what is revealed by directly comparing data-driven multimodal features with category-based semantic annotation under matched conditions. Specifically, we focus on how these two approaches converge at the level of large-scale semantic organization while differing in their flexibility, temporal resolution, and dependence on human-defined categories. These revisions better reflect the current state of the field and sharpen the manuscript’s central contribution as a principled comparison between modeling approaches, rather than a general argument for automated semantic mapping.

(On page 10) “Study 3: Comparing data-driven multimodal representations with category-based semantic annotation

A central question in naturalistic encoding is how data-driven feature representations derived from pretrained models relate to more interpretable, category-based semantic annotations that have historically been used to study cortical semantic organization. Although recent work has shown that pretrained language and vision–language models can capture semantic structure without explicit annotation, category-based approaches such as WordNet remain valuable as interpretable reference frameworks. Here, we leverage the WordNet-based semantic components reported by Huth et al. (2012) 5 not as a state-of-the-art alternative, but as a historically grounded benchmark, allowing a controlled comparison between data-driven multimodal representations and manually defined semantic categories under matched naturalistic movie stimuli.”

(On page 18) “Study 3 demonstrates the utility of video–text alignment models for probing higher-order semantic representations during naturalistic perception. Our comparison between VALOR-derived representations and WordNet-based semantic components highlights an important distinction between data-driven and category-based approaches to modeling meaning in the brain. While multimodal pretrained models offer flexible, high-dimensional representations that capture semantic structure without explicit annotation, category-based frameworks provide interpretability and theoretical anchoring 4,48. Using WordNet-based labeling from prior work as an interpretable reference point, we show that VALOR automatically extracts semantic dimensions—including mobility, sociality, and civilization—that closely mirror those identified using manual semantic categories (Fig. 3). The observed alignment between VALOR PCs and WordNet semantic components suggests that large-scale semantic organization emerges consistently across these approaches, even though they differ in how semantic structure is defined and learned. This convergence supports the use of pretrained multimodal models as practical encoding tools for naturalistic stimuli, while also underscoring the continued value of interpretable semantic benchmarks for understanding which aspects of meaning are represented across cortex. We do not argue that semantic annotation is required for modern encoding models; rather, WordNet-based features serve here as a historically grounded and interpretable reference for contextualizing data-driven multimodal representations.”

(3) The authors use subject-specific encoding models trained on the HCP dataset to predict group-level mean responses in an independent in-house dataset. While this analysis is framed as testing model generalization, it is important to clarify that it is not assessing traditional out-of-distribution (OOD) generalization, where the same subject is tested on novel stimuli, but rather evaluating which encoding model's feature space contains more stimulus-specific and cross-subject-consistent information that can transfer across datasets.

We thank the reviewer for this helpful clarification and agree that the type of generalization tested here should be described more precisely. Our analysis does not assess classical within-subject out-of-distribution (OOD) generalization, in which the same individual is tested on novel stimuli.

Instead, for each HCP participant we train a subject-specific encoding model and transfer it to predict group-mean responses in an independent in-house dataset collected at a different site, with different participants, different movies, and different acquisition conditions. This design evaluates which encoding model’s feature space contains stimulus-locked representations that are consistent across individuals and robust to changes in dataset and experimental context, rather than within-subject stimulus novelty per se.

We have revised the Results (p. 10) and Discussion section (p. 17) to explicitly describe this analysis as a test of cross-subject and cross-dataset transferability of stimulus representations, and to clarify the distinction from traditional OOD generalization.

(On Page 10) “Although this analysis is not a classical within-subject out-of-distribution generalization test, it evaluates the extent to which different feature spaces capture stimulus-locked representations that are consistent across subjects and transferable across datasets, stimuli, and acquisition environments.”

(On Page 17) “By contrast, VALOR exhibited stronger generalization in a cross-cohort, cross-stimulus, and cross-site transfer evaluation.”

(4) Within this setup, the finding that VALOR outperforms CLIP, AlexNet, and WordNet is somewhat expected. VALOR encodes rich spatiotemporal information from videos, making it more aligned with movie-based neural responses. CLIP and AlexNet are static image-based models and thus lack temporal context, while WordNet only provides coarse categorical labels with no stimulus-specific detail. Therefore, the results primarily reflect the advantage of temporally-aware features in capturing shared neural dynamics, rather than revealing surprising model generalization. A direct comparison to pure video-based models, such as Video Swin Transformers or other more recent video models, would help strengthen the argument.

We thank the reviewer for this baseline-focused comment and agree that, in naturalistic movie paradigms, a temporally structured audiovisual model would be expected to outperform static or unimodal feature spaces. Our intent in this comparison is therefore not to claim a surprising advantage, but to isolate which inductive biases matter for cross-dataset transfer of movie-evoked neural responses.

The baseline models were chosen deliberately to span feature spaces that are widely used and interpretable in cognitive neuroscience: AlexNet (vision-only, frame-based), WordNet (human-defined semantic categories without learned visual features), and CLIP (static image–text alignment without temporal context). Comparing VALOR against these established baselines under matched preprocessing, TR alignment, and dimensionality control allows us to attribute performance differences specifically to temporal integration and audiovisual alignment, rather than to generic model capacity.

We agree that a direct comparison with purely visual spatiotemporal encoders (e.g., Video Swin or TimeSformer-style models) would further dissociate the contribution of temporal visual processing from cross-modal video–text alignment. We now explicitly note this as an important direction for future work and frame VALOR as one representative of a broader class of multimodal video models, rather than as a uniquely optimal solution (Discussion, p. 16).

(On page 16) “Second, we did not directly compare VALOR to state-of-the-art video-only spatiotemporal models (e.g., Video Swin Transformer, VideoMAE, and related architectures) that are designed to capture temporal visual structure without language grounding; such comparisons will be important for isolating the specific contributions of temporal visual processing versus cross-modal video–text alignment in naturalistic neural responses.”

(5) Moreover, while WordNet-based encoding models perform reasonably well within-subject in the HCP dataset, their generalization to group-level responses in the Short Fun Movies (SFM) dataset is markedly poorer. This could indicate that these models capture a considerable amount of subject-specific variance, which fails to translate to consistent group-level activity. This observation highlights the importance of distinguishing between encoding models that capture stimulus-driven representations and those that overfit to individual heterogeneities.

Thank you for this thoughtful observation. We agree with the reviewer’s interpretation. In our analyses, WordNet-based models perform reasonably well when fit and evaluated within individual HCP participants, but their performance degrades substantially when transferred to predict group-averaged responses in the independent SFM dataset. This dissociation suggests that, while WordNet annotations capture meaningful variance at the individual level, a larger fraction of that variance may be subject-specific or idiosyncratic, and therefore does not translate into consistent, stimulus-locked responses at the group level.

One motivation for our cross-dataset, cross-subject evaluation is precisely to distinguish encoding models that primarily capture shared stimulus-driven structure from those whose apparent performance depends more strongly on individual heterogeneity. In this context, the reduced transferability of WordNet-based models highlights a potential limitation of category-based semantic features for capturing population-consistent neural dynamics during naturalistic viewing.

We note that this effect likely reflects multiple factors rather than a single failure mode, including differences in annotation schemes, labeling granularity, and semantic coverage across datasets. By contrast, video–text models provide time-aligned linguistic features directly from the stimulus itself, reducing reliance on dataset-specific human annotation and exhibiting stronger transfer across cohorts. We have clarified this interpretation in the revised Discussion (p. 17).

(Page 17) “Together, these findings underscore the importance of distinguishing encoding models that primarily capture shared, stimulus-driven neural structure from those whose performance relies more heavily on subject-specific heterogeneity, particularly when evaluating generalization across participants and datasets.”

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) In the Methods section, please clarify which specific layer of VALOR the 512-dimensional feature vector was extracted from.

Thank you for this suggestion. We have revised the Methods to state explicitly that the 512-dimensional feature vector is extracted from VALOR’s joint video–text projection head, i.e., the final projection layer of the contrastive alignment module that maps video and text representations into a shared embedding space. We also clarify that these 512-D embeddings are computed at the segment/TR level and then time-aligned to the BOLD signal (Methods, p. 21).

(On page 21) “We segmented each movie at the TR level and, for each segment, extracted VALOR’s projected video–text embedding from the final projection head of the alignment module to obtain a 512-dimensional feature vector. These embeddings were then time-aligned to the corresponding BOLD responses.”

(2) It would be helpful to include more detailed descriptions of the network architectures and parameters for all models used.

Thank you for the suggestion. We have revised the Methods to include model-specific subsections for all feature spaces used (VALOR, CLIP, AlexNet, and WordNet). For each model, we now explicitly report (i) the backbone architecture and training objective, (ii) the exact feature source (layer or projection head) and output dimensionality, and (iii) how features were temporally aligned to the BOLD signal. All models were used with their publicly released pretrained parameters, without additional fine-tuning. These additions are intended to improve transparency and reproducibility (Methods, p. 21).

(On page 21) “Movie Feature Extraction

(1) Video–text alignment features (VALOR): To extract video-based multimodal features, we used VALOR (VALOR-large checkpoint), an open-source pretrained video–text alignment model24. VALOR combines visual encoders (CLIP and Video Swin Transformer) for extracting visual features and a text encoder (BERT) for extracting textual features 23,51,52. These representations are aligned in a shared embedding space through contrastive learning. We segmented each movie at the TR level and, for each segment, extracted VALOR’s projected video–text embedding from the final projection head of the alignment module to obtain a 512-dimensional feature vector. These embeddings were then time-aligned to the corresponding BOLD responses.

(2) CLIP features: To compare with static image-based multimodal models, we utilized CLIP (ViT-B/32), which aligns visual and textual representations through contrastive learning but processes individual frames independently without capturing temporal context. One video frame was sampled per TR, and the pooled image embedding after CLIP’s projection into the shared image–text space was extracted to obtain a 512-dimensional feature vector. These TR-aligned vectors were used directly as regressors in the voxel-wise encoding models.

(3) AlexNet features: Visual features were extracted by sampling frames at the TR level and processing them with AlexNet, an eight-layer convolutional neural network comprising five convolutional layers followed by three fully connected layers. Features from all five convolutional layers were evaluated in preliminary analyses; the fifth convolutional layer showed the best performance and was used in subsequent analyses. Intra-image z-score normalization was applied to reduce amplitude effects. Principal component analysis (PCA) was used to reduce dimensionality, retaining the top 512 components to match the dimensionality of multimodal feature spaces. This pipeline was implemented using the DNNBrain toolkit 53.

(4) WordNet features: Semantic features were obtained from publicly available WordNet annotations provided with the HCP dataset (7T_movie_resources/WordNetFeatures.hdf5), following the procedure of Huth et al. (2012). Throughout this manuscript, we use the term “semantic features” to refer to such human-annotated, category-based representations of scene content, and we reserve the term “linguistic features” for continuous language embeddings derived automatically from pretrained language or vision–language models. Each second of the movie clips was manually annotated with WordNet categories according to predefined guidelines: (a) identifying clear objects and actions in the scene; (b) labeling categories that dominated for more than half of the segment duration; and (c) using specific category labels rather than general ones. A semantic feature matrix was constructed with rows corresponding to time points and columns to semantic categories, with category presence coded as binary values. More specific categories from the WordNet hierarchy were added to each labeled category, yielding a total of 859 semantic features. These features were used directly as regressors. We also evaluated a PCA-reduced 512-dimensional variant (fit within each training fold to avoid leakage); because this version performed slightly worse, we report results from the full 859-dimensional representation in the main text. For the generalization analysis in Study 2, annotations for the SFM dataset were aligned to the same WordNet category space to ensure consistency.”

(3) In Figure 3, consider following Huth et al.'s approach by using 3-4 distinct colors to visualize semantic representations across the cortical surface more clearly.

Thank you for this excellent suggestion. We have generated an alternative visualization using a discrete 3–4 color scheme following Huth et al. to display the semantic components on the cortical surface. This version makes the spatial correspondence between components and the boundaries between cortical territories easier to see. We now include this visualization in the Supplement (Fig. S3)

(4) In Figure 2, the brain renderings are too small. Please consider creating a separate, enlarged figure with clearer delineation of relevant ROIs.

We appreciate this suggestion and agree that clear delineation of ROIs is important. We evaluated larger brain renderings; however, within the multi-panel layout of Fig. 2, enlarging them compressed accompanying plots/legends and introduced visual crowding, which reduced overall readability. To preserve a balanced layout and consistent typography across panels, we have kept the current rendering size in the main text and added Fig. S4 with enlarged brain renderings showing clearer ROI boundaries for the same ROIs.

Reviewer #2 (Recommendations for the authors):

(1) From the introduction, I feel like naïve readers would have a hard time understanding what semantic models (e.g., WordNet) are, which the authors write are based on "labor-intensive and subjective manual annotation of semantic content". It would be straightforward to explain the process-how scientists have written descriptions or denoted categories of what's happening within a TR and transformed these into embedding vectors based on language models. This description would explain what the authors mean by "labor-intensive, time-consuming, and subjective". Related to this point, the authors seem to be using the words "semantic model/feature" and "linguistic model/feature" interchangeably, which may exacerbate the confusion.

Thank you for this helpful suggestion. We agree that naïve readers would benefit from a clearer explanation of how “semantic” models such as WordNet are constructed and from a more precise distinction between semantic and linguistic features.

In response, we expanded the Introduction (p. 3) to explicitly describe the process by which semantic features are generated via dense human annotation (i.e., raters label objects, actions, and events within each TR and map these labels onto a predefined ontology to form feature vectors), clarifying why this approach is labor-intensive, time-consuming, and subject to rater variability.

To avoid disrupting the conceptual flow of the Introduction, we placed the explicit terminology clarification in the Methods section (p. 22), where feature extraction is described. There, we now define semantic features as human-annotated, category-based representations of scene content, and linguistic features as continuous language embeddings derived automatically from pretrained language or vision–language models. These revisions are intended to improve clarity and consistency for both expert and non-expert readers.

(On page 3) “Critically, semantic models often rely on dense human annotation. In early naturalistic encoding studies, trained raters watched the stimulus and labeled what was happening within each TR or short time window—for example, identifying objects, actions, or events present in the scene. These labels were then mapped onto a predefined semantic ontology (such as WordNet), yielding high-dimensional categorical feature vectors that served as regressors in encoding models. While this approach provides interpretable semantic features, it is labor-intensive, time-consuming, and inherently subjective, as annotations depend on rater judgment, labeling guidelines, and dataset-specific conventions, limiting scalability and reproducibility.”

(On page 22) “Throughout this manuscript, we use the term “semantic features” to refer to such human-annotated, category-based representations of scene content, and we reserve the term “linguistic features” for continuous language embeddings derived automatically from pretrained language or vision–language models.”

(2) Figure 1A does not look like an accurate schematic of the encoding method. For example, shouldn't the "Train" give rise to weight matrices, and Movies come from moments at Test? I would appreciate it if this schematic figure would explain what the encoding model is to naïve readers.

(3) Figure 1B emphasizes that VALOR is utilizing multimodal features, but does not emphasize that the model is trained on dynamic video. The current figure looks like the model extracted visual and linguistic features from a screenshot of the video, much like the CLIP model.

Thank you for this helpful comment. We agree that the original Fig. 1A did not sufficiently clarify what is learned during training versus what is applied during testing, and that this distinction is particularly important for naïve readers unfamiliar with encoding models. We also agree that the original Fig. 1B did not sufficiently emphasize that VALOR is trained on dynamic video segments, and that the schematic could be misinterpreted as aligning a single video frame with text, similar to CLIP-style image–text models.

We have revised Fig. 1A (p. 6) to make the encoding procedure explicit and pedagogical. Specifically, we now clearly depict that, during the training phase (HCP dataset), voxel-wise encoding models learn feature-to-voxel weight matrices from stimulus features and BOLD responses. These learned weights are explicitly labeled as voxel-wise weight matrices and visually associated with the training stage. In the testing/generalization phase (SFM dataset), we now indicate that these learned weights are held fixed and applied to features extracted from novel movies to generate predicted BOLD responses. Additional labels were added to distinguish “Training (learn weights)” from “Testing/Transfer (apply fixed weights)” and to clarify that the encoding model implements a linear mapping from stimulus features to voxel responses. We have also rewritten the Fig. 1 legend (p. 6) to explicitly explain the encoding workflow in words, including (i) the learning of voxel-specific weights during training, (ii) their reuse during cross-dataset transfer, and (iii) how generalization performance is evaluated. These changes are intended to ensure that Fig. 1A accurately reflects the encoding methodology and is understandable to readers without prior experience with encoding models.

We have revised Fig. 1B (p. 6) to explicitly highlight the temporal nature of the video input used by VALOR. In the updated schematic, the visual stream is depicted as a sequence of consecutive frames spanning multiple seconds, grouped into a video segment, rather than as a single static image. Additional labels indicate that VALOR encodes temporally extended video clips and aligns them with corresponding textual descriptions in a shared embedding space via contrastive learning. We have also updated the figure legend (p. 6) to clarify that VALOR operates on multi-frame video segments and explicitly models temporal structure, distinguishing it from static image–text models such as CLIP. These changes are intended to make clear that VALOR’s advantage derives not only from multimodality, but also from learning representations over time.

(4) Regarding Figure 2, why were paired t-tests conducted in one-sided comparisons? Shouldn't this be two-sided, given that there is no reason to assume one is higher or lower than another?

Thank you for raising this point. We agree that, in the absence of a preregistered directional hypothesis, paired comparisons should be evaluated using two-sided statistical tests.

In response, we have re-run all paired comparisons reported in Figure 2 (p. 9) using two-sided paired t-tests, recomputed the corresponding p-values and false discovery rate (FDR) corrections, and updated the significance markers in the figure and captions accordingly. Importantly, this change does not alter the qualitative pattern of results or the main conclusions reported in the manuscript.

(5) Regarding Study 4, I am curious whether the results are specific to forward-looking representations (predictive coding) or whether the results broadly reveal regions that are sensitive to contexts. For example, if the authors were to incorporate nearby past scenes in the analysis rather than the nearby future scenes, would different brain regions light up?

Thank you for this thoughtful question. We agree that it is important to distinguish forward-looking (predictive) representations from more general sensitivity to temporal context. In Study 4, we deliberately operationalized prediction using future-aligned features, such that only information from upcoming scenes was incorporated into the encoding model. Accordingly, the reported effects should be interpreted as reflecting forward-oriented representations rather than generic context sensitivity.

To make this interpretive scope explicit, we have added a clarifying sentence at the beginning of the Study 4 paragraph in the Discussion (p.18), noting that our analysis incorporates only future-aligned features and that directly contrasting past- and future-aligned features will be an important direction for future work. This clarification is intended to clearly bound our claims while addressing the reviewer’s conceptual distinction..

(On page 18) “In Study 4, we used a video-text alignment model to investigate predictive coding mechanisms. Because our analysis incorporates only future-aligned features, the reported effects should be interpreted as reflecting forward-oriented representations rather than generic sensitivity to temporal context; directly contrasting past- and future-aligned features will be an important direction for future work.”

(6) In the paragraph starting in line 447, were WordNet feature time series also reduced to 512 dimensions like the rest of the model features?

Thank you for the question. In the main analyses, WordNet feature time series were not reduced to 512 dimensions and were instead used at their full dimensionality (859 features).

For comparability with the other feature spaces, we additionally conducted a control analysis in which WordNet features were reduced to 512 dimensions using PCA. The PCA was fit within each training fold to avoid information leakage, and the resulting 512-D features were evaluated using the same encoding pipeline. This PCA-reduced version performed slightly worse than the full 859-D WordNet representation. Accordingly, we report results from the full 859-D WordNet features in the main text. We have clarified this point in the Methods section (p. 22).

(On page 22) “We also evaluated a PCA-reduced 512-dimensional variant (fit within each training fold to avoid leakage); because this version performed slightly worse, we report results from the full 859-dimensional representation in the main text.”

(7) I don't think authors have written what VALOR stands for.

Thank you for the reminder. We now define the VALOR acronym at its first mention in the Abstract and Introduction and use the abbreviation thereafter.

(On page 2) “Using a state-of-the-art deep learning model (VALOR; Vision-Audio-Language Omni-peRception)”

(On page 5) “To answer this, we apply a video-text alignment encoding framework, using VALOR (Vision-Audio-Language Omni-peRception)—a high-performing, open-source model that aligns visual and linguistic features over time—to predict brain responses during movie watching.”

(8) When calculating equation (3), please make sure that the correlation values are Fisher's r-to-z transformed.

Thank you for this reminder. We confirm that all correlation coefficients used in Equation (3) are now Fisher r-to-z transformed prior to any averaging, contrasts, or statistical testing, and this procedure is now explicitly stated in the Methods. We have also updated Fig. 4a (p. 15) to reflect this transformation. Importantly, applying the r-to-z transformation does not change the qualitative pattern of results or their statistical significance.

(9) I wasn't able to check the OSF data/codes because it required permission.

Thank you for flagging this, and we apologize for the inconvenience. We have removed the permission restriction and set the OSF repository to public read-only access, which should resolve the issue.

Reviewer #3 (Recommendations for the authors):

(1) The current approach extracts features from a single "best" layer of each model, which may be suboptimal for predicting neural responses. Prior work has shown that combining features across multiple layers through optimized fusion strategies (e.g., St-Yves et al., 2023) or using model ensembles (e.g., Li et al., 2024) can substantially improve encoding performance. The authors may consider these more comprehensive approaches either as additional baselines or as alternative directions to enhance model accuracy.

Thank you for this constructive suggestion. We agree that combining features across multiple layers or using optimized fusion and ensemble strategies, as demonstrated in recent work (e.g., St-Yves et al., 2023; Li et al., 2024), can substantially improve absolute encoding performance.

In the present study, however, we intentionally evaluated each model using its single best-performing layer within a matched encoding pipeline. This design choice was made to maintain model-agnostic comparability and interpretability, and to ensure that performance differences could be attributed primarily to the type of representation (e.g., temporally informed video–text features versus static or unimodal features), rather than to differences in model complexity, parameter count, or fusion strategy. Importantly, this constraint was applied uniformly across all models and therefore does not favor VALOR over the baselines.

We now explicitly note in the Discussion (p. 19) that multilayer fusion and ensemble approaches represent a natural and promising extension of our framework and are likely to further improve absolute prediction accuracy. Our goal in the current work was to establish the practical utility and generalizability of temporally aligned video–text features for naturalistic movie fMRI under a controlled and comparable evaluation setting..

(On page 19) “Third, for comparability across models we evaluated each model using its single best-performing layer within a matched encoding pipeline rather than using multilayer fusion or ensembling, which allowed us to attribute performance differences to representational format but likely underestimates the absolute performance ceiling.”

(2) Given the naturalistic video-based task, the manuscript would benefit from including state-of-the-art video-only models (e.g., Video Swin Transformer, VideoMAE, and other more recent architectures) as explicit baselines. These models are designed to capture spatiotemporal structure without relying on language input and would provide a more targeted comparison to assess the specific contribution of temporal visual processing.

Thank you for this thoughtful suggestion. We agree that state-of-the-art video-only spatiotemporal models (e.g., Video Swin Transformer, VideoMAE) are highly relevant baselines for naturalistic movie paradigms and would provide a more targeted comparison for isolating the contribution of temporal visual processing independent of language input.

In the present study, our primary goal was not to exhaustively benchmark all possible video architectures, but to evaluate whether temporally informed video–text features can serve as a practical and general-purpose encoding framework that improves upon the models most commonly used in cognitive neuroscience for naturalistic fMRI (e.g., AlexNet for vision, WordNet for semantic annotation, and CLIP for static multimodal alignment). Using these established baselines allowed us to place our results in direct continuity with prior neuroimaging work and to attribute performance differences to representational format under a controlled encoding pipeline.

We agree that incorporating modern video-only spatiotemporal encoders is an important next step, particularly for disentangling the relative contributions of temporal visual structure and cross-modal video–text alignment. We now explicitly note this point in the Discussion (p.19) as a limitation and future direction, and view such comparisons as a natural extension of the current framework within the same TR-aligned encoding setup.

(On page 19) “Second, we did not directly compare VALOR to state-of-the-art video-only spatiotemporal models (e.g., Video Swin Transformer, VideoMAE, and related architectures) that are designed to capture temporal visual structure without language grounding; such comparisons will be important for isolating the specific contributions of temporal visual processing versus cross-modal video–text alignment in naturalistic neural responses.”

(3) An additional consideration is the scale of the AI models used for feature extraction. Previous studies (e.g., Matsuyama et al., 2023) have indicated that model size - particularly the number of parameters - can influence neural prediction performance, independently of architecture. A discussion or analysis of how model size contributes to the observed encoding gains would help clarify whether improvements are due to the representational quality of the model or simply its scale

Thank you for this important point. We agree that model scale—particularly parameter count—can influence neural prediction performance independently of architecture, as noted in prior work (e.g., Matsuyama et al., 2023).

In the present study, our primary goal was to evaluate whether temporally informed video–text representations provide practical advantages over unimodal and static multimodal baselines that are widely used in cognitive neuroscience for naturalistic movie fMRI, under a matched encoding pipeline. We did not perform a systematic scale-controlled analysis in this revision because doing so would require training or evaluating multiple size-matched variants across video-only and video–text architectures, which is beyond the scope of the current work.

We therefore agree that part of the observed performance gains may reflect model capacity in addition to representational format, and we caution against attributing all improvements solely to cross-modal alignment or temporal structure. We now explicitly acknowledge this limitation in the Discussion and note that comparing size-matched video-only and video–text models within the same pipeline is an important next step for disentangling model scale from representational content.

(On page 19) “Finally, part of VALOR’s advantage may reflect model capacity: larger pretrained models often yield higher encoding accuracy, so repeating these analyses with size-matched image-only and image–text models will be critical for disentangling model scale from representational content.”

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

In the current study, Huang et al. examined ACC response during a novel discrimination-avoid task. The authors concluded that ACC neurons primarily encode post-action variables over extended periods, reflecting the animal's preceding actions rather than the outcomes or values of those actions. Specifically, they identified two subgroups of ACC neurons that responded to different aspects of the actions. This work represents admirable efforts to investigate the role of ACC in task-performing mice. However, in my opinion, alternative explanations of the data were not sufficiently explored, and some key findings were not well supported.

Strengths:

The development of the new discrimination-avoid task is applauded. Single-unit electrophysiology in task-performing animals represents admirable efforts and the datasets are valuable. The identification of different groups of encoding neurons in ACC can be potentially important.

Weaknesses:

One major conclusion is that ACC primarily encodes the so-called post-action variables (specifically shuttle crossing). However, only a single example session was included in Figure 2, while in Supplementary Figure 2 a considerable fraction of ACC neurons appears to respond to either the onset of movement or ramp up their activity prior to movement onset. How did the authors reach the conclusion that ACC preferentially respond to shuttle crossing?

We now include more example sessions and the main results from individual animals (Fig. 3; Figs. S2–S3; Fig. 8). Overall, the results are consistent across recording sessions and animals.

While shuttle crossings were the primary reference for most analysis, using shuttle initiation as a reference led to similar conclusions (Fig.4). Namely, we found that most ACC neurons exhibit either robust (22%; Types 1a & 2a) or moderate (51%; Types 1b & 2b) post-shuttle activity changes (Fig.4), while only a subset exhibits ramping pre-shuttle activity (16%; Types 3b & 3c). Therefore, our conclusion was intended to highlight the role of post-shuttle activity in learning. While we do not exclude the possibility that pre-shuttle ACC activity contributes to learning, its involvement is likely more limited.

In Figure 4, it was concluded that ACC neurons respond to action independent of outcome. Since these neurons are active on both correct and incorrect shuttle but not stay trials, they seem to primarily respond to overt movement. If so, the rationale for linking ACC activity and adaptive behavior/ associative learning is not very clear to me. Further analyses are needed to test whether their firing rates correlated with locomotion speed or acceleration/deceleration. On a similar note, to what extent are the action state neurons actually responding to locomotion-related signals? And can ACC activity actually differentiate correct vs. incorrect stays?

In this study, we highlight two distinct groups of ACC neurons: action-state and action-content neurons. Both groups of neurons tend to show sustained activity even when the animals remain immobile after completing shuttle behaviors, suggesting that their activity is not directly driven by locomotion. Furthermore, action-content neurons are selectively engaged in only one of the two shuttle categories, either rooms A→B or B→A shuttles. Therefore, differences in neuronal activity are unlikely to reflect locomotor differences, given that both shuttle types involve similar movement patterns. Finally, we analyzed ACC neuronal activity in relation to locomotion speed. Our results indicate that only a small fraction of neurons (<15%) show speed-correlated activity (Fig.5), suggesting that most ACC neurons do not encode movement-related information. Taken together, these findings support the distinction between ACC activity and locomotion encoding.

As for the small subset of speed-related neurons, it remains unclear whether these speed-related neurons represent a distinct subpopulation within the ACC or reflect recordings from the nearby motor cortex. Postmortem examination of the recording sites suggests that most neurons were recorded from the ACC, while a small subset may be located at the border between the ACC and motor cortex (Fig. S2). Therefore, it is possible that the small fraction of speed-related neurons originated from the motor cortex.

Lastly, given that the ACC neurons display no or limited activity during stay trials, their activity generally does not differentiate correct vs. incorrect stays (Fig.S7). However, ACC activity does show moderate differentiation between room-A vs. room-B stays (Fig.S7).

Given that a considerable amount of ACC neurons encode 'action content', it is not surprising that by including all neurons the model is able to make accurate predictions in Figure 6. How would the model performance change by removing the content neurons?

We thank the reviewer for this thoughtful analysis idea. Excluding action-content neurons drastically reduces decoding accuracy (Fig.8), suggesting that they are the main drivers for differentiating rooms A→B vs. B→A shuttles.

Moving on to Figure 7. Since Figure 4 showed that ACC neurons respond to movement regardless of outcome, it is somewhat puzzling how ACC activity can be linked to future performance.

As discussed earlier (point #2), ACC activity does not simply reflect locomotion itself. We interpret the post-shuttle ACC activity as encoding both the preceding shuttle state (shuttle or stay) and shuttle content (rooms A→B or B→A). Regardless of the outcome (safety or shock), such encoding is essential for cue–action–outcome associative learning, because both positive and negative feedback can drive learning. The level of post-shuttle ACC activity may reflect task engagement, with greater engagement facilitating learning and improving future performance.

Two mice contributed about 50% of all the recorded cells. How robust are the results when analyzing mouse by mouse?

We have added further analysis of highlighting the results of each mouse. Although the total number of recorded neurons varied across mice, the major findings were consistent. In every mouse, we observed sustained post-shuttle ACC activity (Fig.S2), and population-level ACC activity reliably decoded shuttle contents (rooms A→B vs. B→A; Fig.8).

Lastly, the development of the new discrimination-avoid task is applauded. However, a major missing piece here is to show the importance of ACC in this task and what aspects of this behavior require ACC.

We appreciate this feedback. We are currently conducting additional experiments to determine whether inhibiting ACC activity during distinct time windows disrupts task learning. We hope to publish a follow-up paper on these findings in the near future.

Reviewer #2 (Public review):

Summary:

The current dataset utilized a 2x2 factorial shuttle-escape task in combination with extracellular single-unit recording in the anterior cingulate cortex (ACC) of mice to determine ACC action coding. The contributions of neocortical signaling to action-outcome learning as assessed by behavioral tasks outside of the prototypical reward versus non-reward or punished vs non-punished is an important and relevant research topic, given that ACC plays a clear role in several human neurological and psychiatric conditions. The authors present useful findings regarding the role of ACC in action monitoring and learning. The core methods themselves - electrophysiology and behavior - are adequate; however, the analyses are incomplete since ruling out alternative explanations for neural activity, such as movement itself, requires substantial control analyses, and details on statistical methods are not clear.

Strengths:

(1) The factorial design nicely controls for sensory coding and value coding, since the same stimulus can signal different actions and values.

(2) The figures are mostly well-presented, labeled, and easy to read.

(3) Additional analyses, such as the 2.5/7.5s windows and place-field analysis, are nice to see and indicate that the authors were careful in their neural analyses.

(4) The n-trial + 1 analysis where ACC activity was higher on trials that preceded correct responses is a nice addition, since it shows that ACC activity predicts future behavior, well before it happens.

(5) The authors identified ACC neurons that fire to shuttle crossings in one direction or to crossings in both directions. This is very clear in the spike rasters and population-scaled color images. While other factors such as place fields, sensory input, and their integration can account for this activity, the authors discuss this and provide additional supplemental analyses.

Weaknesses:

(1) The behavioral data could use slightly more characterization, such as separating stay versus shuttle trials.

We appreciate this feedback. In the revised manuscript, we present data separating stay versus shuttle trials (Fig.1). Additionally, we provide new data from extended training sessions (Fig.S2).

(2) Some of the neural analyses could use the necessary and sufficient comparisons to strengthen the authors' claims.

We have now used the necessary and sufficient comparisons where applicable. In the SVM decoding analysis, we show that population ACC activity is sufficient to decode A→B or B→A shuttles. We also show that excluding action-content, but not other ACC neurons, drastically reduces decoding accuracy, suggesting that these neurons are necessary for the decoding (Fig.8).

(3) Many of the neural analyses seem to utilize long time windows, not leveraging the very real strength of recording spike times. Specifics on the exact neural activity binning/averaging, tests, classifier validation, and methods for quantification are difficult to find.

We chose to perform our neural analyses on a longer time scale, given the sustained activity we see in the data. To further justify that decision, we now provide additional results highlighting the sustained activity of ACC neurons in our task (Fig.2; Fig.S2). Additionally, we now provide more specifics of the neural analyses in Methods section.

(4) The neural analyses seem to suggest that ACC neurons encode one variable or the other, but are there any that multiplex? Given the overwhelming evidence of multiplexing in the ACC a bit more discussion of its presence or absence is warranted.

This is an interesting point of discussion, and we thank the reviewer for pointing this out. Overall, our results suggest that individual ACC neurons preferentially engage in only one of the proposed functions, rather than multiplexing across them. For example, action-state and action-content ACC neurons primarily engage in action monitoring, but not in decision-making, planning, or outcome tracking. Nevertheless, we cannot rule out the possibility that other ACC neurons, through their distinct connectivity or location in different ACC subregions, engage in other proposed functions. Thus, when considering the ACC as a whole, its function may still be multiplexed.

Another possible reason we do not see clear multiplexing of neurons may be due to the dynamic nature of our task. Unlike established tasks that often assign fixed positive or negative values to cues, the cues in our task are not inherently associated with valence. Instead, their meaning is dynamically determined by the animal’s location (context) at the time of cue presentation. Since values are not fixed and change based on context, value-related responses may not be reflected in the ACC in our tasks.

We have now incorporated the above discussions into our revised manuscript.

Reviewer #3 (Public review):

Summary:

The authors record from the ACC during a task in which animals must switch contexts to avoid shock as instructed by a cue. As expected, they find neurons that encode context, with some encoding of actions prior to the context, and encoding of neurons post-action. The primary novelty of the task seems to be dynamically encoding action-outcome in a discrimination-avoidance domain, while this is traditionally done using operant methods. While I'm not sure that this task is all that novel, I can't recall this being applied to the frontal cortex before, and this extends the well-known action/context/post-context encoding of ACC to the discrimination-avoidance domain.

While the analysis is well done, there are several points that I believe should be elaborated upon. First, I had questions about several details (see point 3 below). Second, I wonder why the authors downplayed the clear action coding of ACC ensembles. Third, I wonder if the purported 'novelty' of the task (which I'm not sure of) and pseudo-debate on ACC's role undermines the real novelty - action/context/outcome encoding of ACC in discrimination-avoidance and early learning.

Strengths:

Recording frontal cortical ensembles during this task is particularly novel, and the analyses are sophisticated. The task has the potential to generate elegant comparisons of action and outcome, and the analyses are sophisticated.

Weaknesses:

I had some questions that might help me understand this work better.

(1) I wonder if the field would agree that there is a true 'debate' and 'controversy' about the ACC and conflict monitoring, or if this is a pseudodebate (Line 34). They cite 2 very old papers to support this point. I might reframe this in terms of the frontal cortex studying action-outcome associations in discrimination-avoidance, as the bulk of evidence in rodents comes from overtrained operant behavior, and in humans comes from high-level tasks, and humans are unlikely to get aversive stimuli such as shocks.

We appreciate this feedback. We have revised the Introduction and Discussion.

(2) Does the purported novelty of the task undermine the argument? While I don't have an exhaustive knowledge of this behavior, the novelty involves applying this ACC. There are many paradigms where a shock triggers some action that could be antecedents to this task.

We argue our newly designed discrimination–avoidance task is unique for several reasons. First, it requires animals to discriminate both sensory cues and environment contexts. Unlike established tasks that often assign fixed positive or negative values to cues, the cues in our task are not inherently associated with valence. Instead, their meaning is dynamically determined by the animal’s location (context) at the time of cue presentation, which reflects a conceptual advance over previous techniques. Furthermore, by removing valence from the cues, this design helps disentangle the ACC’s potential role in value encoding from other cognitive functions.

Second, this task involves robust, ethologically relevant actions (i.e., shuttles), unlike many established paradigms that rely on less naturalistic behaviors such as saccades or lever presses. We view this as a key distinction from prior approaches, as even previous paradigms that utilize shutting responses or other naturalistic responses, fail to incorporate dynamic integration of cues and contexts.

Finally, the clear temporal separation between actions and outcomes further helps disentangle the ACC’s roles in action monitoring vs. outcome tracking.

(3) The lack of details was confusing to me:

(a) How many total mice? Are the same mice in all analyses? Are the same neurons? Which training day? Is it 4 mice in Figure 3? Five mice in line 382? An accounting of mice should be in the methods. All data points and figures should have the number of neurons and mice clearly indicated, along with a table. Without these details, it is challenging to interpret the findings.

We are sorry for the confusion. We now provide additional details and clear N numbers for each analysis to improve clarity.

(b) How many neurons are from which stage of training? In some figures, I see 325, in some ~350, and in S5/S2B, 370. The number of neurons should be clearly indicated in each figure, and perhaps a table.

All data were obtained from well-trained mice. For some analyses, the N is smaller because certain task sessions contained very few incorrect trials (≤3), which prevented us from examining ACC activity during those trials. We have modified figure legend so that neuron count is clear.

(c) Were the tetrodes driven deeper each day? The depth should be used as a regressor in all analyses?

Yes, the tetrodes were driven slightly deeper across task sessions (~80 µm per step; 2–4 depths per mouse). Given limited depth changes, preliminary analyses indicate no clear differences in ACC activity across these recording depths. However, we cannot rule out potential dorsal–ventral subregion differences if recordings were to span larger depth ranges.

(d) Was is really ACC (Figure 2A)? Some shanks are in M2? All electrodes from all mice need to be plotted as a main figure with the drive length indicated.

We have now included a supplementary figure showing all recording sites (Fig.S2). It is likely that a small subset of neurons was recorded at the ACC/M2 border area. Unfortunately, we are unable to separate them out due to blind recording design of our tetrode arrays.

(e) It's not clear which sessions and how many go into which analysis

We have now specified the number of task sessions for each analysis (see Methods).

(f) How many correct and incorrect trials (<7?) are there per session?

We have now specified the number of correct and incorrect trials per session (see Methods).

(g) Why 'up to 10 shocks' on line 358? What amplitudes were tried? What does scrambled mean?

We decided to use up to 10 mild shocks per trial because mice do not necessarily shuttle to the safe room after one or even a few shocks during the early stages of training. This design allows mice to efficiently learn the concept of the task (i.e., one room is safe while the other delivers shocks). Each shock was specified in the Methods section as 0.5 mA, 0.1 s. A “scrambled shock” refers to an electric shock delivered through multiple floor bars in a randomized pattern, effectively preventing the animal from avoiding the stimulus.

(4) Why do the authors downplay pre-action encoding? It is clearly evident in the PETHs, and the classifiers are above chance. It's not surprising that post-shuttle classification is so high because the behavior has occurred. This is most evident in Figure S2B, which likely should be a main figure.

We did not intend to downplay pre-action encoding. Our analysis shows that most ACC neurons exhibit either robust (22%; Types 1a & 2a) or moderate (51%;Types 1b & 2b) post-shuttle activity changes (Fig.4). Although a subset of ACC neurons exhibits ramping pre-shuttle activity, they represent a much smaller fraction (16%; Types 3b & 3c). Therefore, our conclusion was intended to highlight the role of post-shuttle activity in learning. While we do not exclude the possibility that pre-shuttle ACC activity contributes to learning, its involvement is likely more limited

(5) The statistics seem inappropriate. A linear mixed effects model accounting for between-mouse variance seems most appropriate. Statistical power or effect size is needed to interpret these results. This is important in analyses like Figure 7C or 6B.

We appreciate this feedback. We now use appropriate statistics and report effect size.

(6) Better behavioral details might help readers understand the task. These can be pulled from Figures S2 and S5. This is particularly important in a 'novel' task.

We now provide more details to help better understand the task and have added new figures (Fig.1; Figs. S1&S2).

(7) Can the authors put post-action encoding on the same classification accuracy axes as Figure 6B? It'd be useful to compare.

We appreciate the comment, but we are unsure what clarification is being requested.

(8) What limitations are there? I can think of several - number of animals, lack of causal manipulations, ACC in rodents and humans.

We now include discussions on limitation of our study. One caveat of our study is that the discrimination–avoidance task requires weeks of training in mice. By the time they master the task, ACC activity may reflect modified neural circuits. Investigating ACC activity during early phase of learning, such as by introducing a new pair of cues or contexts, could provide further insights into ACC’s role in learning and cognitive processes. Additionally, a limitation of the current study is the lack of evidence for the causal role of post-action ACC activity in complex associative learning. Future investigations using closed-loop strategies to selectively disrupt ACC activity during the post-action phase could help address this question.

Minor:

(1) Each PCA analysis needs a scree plot to understand the variance explained.

We have added a scree plot for each PCA analysis.

(2) Figure 4C - y and x-axes have the same label?

We have corrected the y-axis label.

(3) What bin size do the authors use for machine learning (Not clear from line 416)?

The bin sizes used were 2.5, 5, 7.5, or 10 sec which have now been discussed in the Methods section.

(4) Why not just use PCA instead of 'dimension reduction' (of which there are many?)

We have adjusted the phrasing where appropriate.

(5) Would a video enhance understanding of the behavior?

We appreciate this feedback. We now include a few videos to accompany our paper.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Is Figure 1C sufficiently powered?

We have now included data from additional mice and updated the figure accordingly.

(2) Task performance was not plateaued after 10 sessions in Figure 1B. How variable is task performance in the datasets with ephys recordings (session to session, mouse to mouse).

We have now included additional data from extended training (15 sessions; Fig.S2). Moderate variations across both sessions and mice are observed. Specifically, the total number of correct/incorrect shuttles used for ephys analysis are 19/5, 19/4, 21/5, 20/4 (mouse #1; 4 sessions); 20/7, 23/7, 20/7 (mouse #2; 3 sessions); 19/4, 16/2 (mouse #3; 2 sessions); 26/4, 23/4, 17/6, 25/5 (mouse #4; 4 sessions); 20/5, and 17/4 (mouse #5; 2 sessions), respectively.

(3) Please quantify the results in Figure 3, for both within individual mice and across mice.

We have calculated maximum trajectory length within the 3-D space (Fig. 3C).

(4) What is the effect size in Figure 7C?

We now report the effect size.

(5) Please provide more details for spike sorting.

We have now included more details in the Methods section.

(6) More detailed cell type or correlation analysis in Figures 4 and 5 may be helpful. For example, if putative regular and fast-spiking neurons were simultaneously recorded, did the FS directly inhibit the RS to give rise to the apparent encoding properties?

We recorded a small number of putative interneurons (n = 13) from only three mice, which precludes drawing meaningful conclusions, particularly given their heterogeneous responses during discrimination–avoidance tasks. Accordingly, we include only an example interneuron demonstrating discrimination between A→B vs. B→A shuttles (Fig. S5). Nevertheless, it is evident there are reciprocal monosynaptic connections between putative interneurons and certain pyramidal neurons, as indicated by short-latency (~2 ms) excitatory or inhibitory interactions (Fig. S5). That said, follow up studies with greater Ns are needed to parse out these details

Reviewer #2 (Recommendations for the authors):

(1) While I appreciate displaying the success rate for the sake of simplifying behavioral data in Figure 1B, it would be nice to also see these data broken out as correct vs incorrect for stay vs shuttle trials, since it is difficult to determine whether the performance increases are primarily driven by mice improving at stay vs shuttle responses

We appreciate this feedback. In the revised manuscript, we present data separating stay versus shuttle trials (Fig.1; Fig.S2).

(2) In Figure 2 the comparison between shuttle and stay is not particularly convincing, since the comparison is also essentially movement vs no movement and place1-->place2 vs place1-->place1. A more appropriate comparison might be action state neurons vs action content neurons during A-->B, B-->A, or both crossings. If it is true that these populations contain this information, then action state neurons should traverse a large component space in both directions, action content neurons only one direction, and so on.

We agree that the comparison is not ideal due to differences in locomotion. However, it provides valuable information suggesting that the ACC plays a limited role during stay trials, despite these trials involve mental and cognitive processes comparable to shuttle trials. While we appreciate the reviewer’s suggestion, the proposed analysis is not particularly reliable given the relatively small number of simultaneously recorded action-state or action-content neurons.

(3) I would say the above point applies to Figure 3 as well. I would also note that this reviewer greatly appreciates the rigor of showing ensemble activity in each subject.

We appreciate this comment. See our response above.

(4) In Figure 5 do these neurons show the same A-->B vs B-->A firing patterns during correct vs incorrect shuttles? The text describing the data in Figure 4 suggests this should be the case but even from a quick glance it sort of seems like the population dynamics during correct vs incorrect shuttles are not the same. My concern is that averaging neural activity over 5s windows washes out all these dynamics

Preliminary analysis suggests that these firing patterns apply to both correct and incorrect shuttles. However, the main reason we did not compare correct and incorrect trials is the limited amount of data. In many sessions, there are only a few (≤5) incorrect shuttles, which include both A→B or B→A shuttles (Fig.1C; Fig.S2), thus lacking the statistical power for a meaningful comparison.

(5) Some information on classifier validation is required - was this leave-out validation and if so how many trials were left-out vs tested? K-fold, and if so, how many folds? Was the trial order shuffled for each simulation? Classifiers will pick up within-session temporal information. In addition to this classifier accuracy during the different time points should be compared by a non-parametric test, and compared to the 95th percentile of the label-shuffled distribution.

Yes, we use standard 10-fold cross-validation. We appreciate the suggestion on trial-order shuffling, and implementing this procedure does not change our original conclusion. Additionally, we have applied a non-parametric test.

(6) How exactly were neurons classified as content vs state? Was it the average activity during the 5s following the shuttle? If this is stated I could not really find it easily so I might suggest clarifying.

We now use a new method for classification of the two neuron types (Fig.7). We have included detailed methods in the revised manuscript.

(7) Movement drives cortical neuron activity more than anything else I have ever seen. Really, more than anything else, it would be nice to demonstrate that it is not movement alone or movement multiplexed with place/sensory information/direction driving these responses.

We have analyzed ACC neuronal activity in relation to locomotion speed. Our results indicate that only a small fraction of ACC neurons (<15%) show speed-correlated activity (Fig.5). It remains unclear whether these speed-related neurons represent a distinct subpopulation within the ACC or reflect recordings from nearby motor cortex. Postmortem examination of the recording sites suggests that most neurons were recorded from the ACC, while a small subset may be located at the border between the ACC and motor cortex. Therefore, it is possible that the small fraction of speed-related neurons originated from the motor cortex.

Furthermore, we identify two distinct groups of ACC neurons: <iaction-state and action-content neurons, both of which tend to show sustained activity even when the animals remain immobile after completing shuttle behaviors. This prolonged activation in the absence of movement suggests that their activity is not directly driven by locomotion. Moreover, action-content neurons are selectively engaged in only one of the two shuttle categories, either rooms A→B or B→A shuttles. Therefore, differences in neuronal activity are unlikely to reflect locomotor differences, given that both shuttle types involve similar movement patterns.

(8) In addition to the above, the place-field analysis in Supplemental Figure 5 only shows 4 neurons. Was the whole population analyzed? Is it possible to decode place from the population during the ITI? The data in this figure sort of look exactly like place fields - many cortical neurons and also some hippocampal neurons have more than 1 place field

We have now provided additional place-field analysis. A comparison with hippocampal CA1 neurons (recorded during the same task) suggests that ACC neurons encode limited spatial information.

(9) "a simple Pavlovian association strategy is unlikely to be sufficient for learning the task" ... is Pavlovian occasion setting not a simple association? Tones and contexts both readily act as Pavlovian occasion setters. Similarly positive/negative patterning might also explain how the task is learned.

We appreciate this comment and have revised the sentence accordingly. It is possible that animals use multiple strategies to learn and perform the task effectively. In the early stages, animals may rely more heavily on sensory–spatial integration, whereas in later stages, sensory- or location-related Pavlovian associative strategies may contribute to performance, particularly when animals begin to show place preferences during inter-trial intervals.

(10) I might suggest softening this language and others like it. For example, 2x2 factorial designs are not really novel.

We have revised the language used to describe the task.

(11) Some of the color-scale bars and figures do not have labels. For example, Supplementary Figure 3, Supplementary Figure 5. Please add labels.

We have added the missing labels to all color bars.

Reviewer #3 (Recommendations for the authors):

(1) Some relevant papers that should be cited:

https://doi.org/10.1523/JNEUROSCI.4450-08.2008

10.1016/j.neuron.2018.11.016

https://doi.org/10.1016/j.jphysparis.2014.12.001

We appreciate these suggestions.

(2) Where can we download the data and code?

We will upload the essential data and MATLAB code to GitHub to accompany the publication of the final version of this paper.

Author response:

Thank you for the reviews of our article “PKMζ-PKCι/λ double-knockout demonstrates atypical PKC is crucial for the persistence of hippocampus LTP and spatial memory.” We will address all of the reviewers’ issues point-by-point in a revised version.

Author response:

We thank the reviewers for their insightful comments on our work.

We agree with reviewer #1 that further experiments would be needed to figure out how the observations done on lab strains can apply to yeast in various ecological conditions and particularly in the wild. We here provide a proof of principle that multicellularity selection can arise as a side-effect. It obviously does not prove that it took place during yeast evolution, but we would like to emphasize that resource fluctuations are very common in ecological conditions, making it highly likely that the environmental conditions necessary for the selection of the side effects described have arisen.

We agree with reviewer #2 that our work on yeast strains is “somewhat artificial” as often the case with model organisms under laboratory conditions. Importantly though, we showed that the effect found with the cln3 knock-out mutation can be phenocopied by overexpression of WHI5 (encoding the yeast equivalent of Rb). We propose that variations in the levels of cell cycle regulators during evolution may have played a role in multicellularity selection as a side effect. We agree that this is merely a hypothesis to explain the selection of multicellularity (just like predator escape) and that there is no direct evidence that this occurred in the history of the lineage. Nevertheless, our work provides a first evidence that such a selection of multicellularity as a side effect could be possible, and gives a framework to understand how multicellularity can persist in the wild, even when it is not the primary target of selection.

We are currently working on the text and figure revisions suggested by the reviewers.

Author response:

Public Reviews:

Reviewer #1 (Public review):

In this paper, the authors use a doxycycline-inducible DLD1 cell line expressing a Clover-tagged RNA-binding-defective TDP-43 2KQ mutant that forms nuclear "anisosomes" (TDP-43 shell with HSP70 core) to carry out a small-molecule screen using the LOPAC 1280 library to identify compounds that reduce anisosome number or shift their morphology and dynamics. They also conducted a genome-wide siRNA screen to identify genetic modifiers of anisosome formation and dynamics. From these screens, the authors identify pathways in RNA splicing, translation, proteostasis (proteasome and HSP90), and nuclear transport, including XPO1. They then focus on XPO1 as their primary hit. Pharmacological inhibition of XPO1 using KPT-276, Verdinexor, and Leptomycin B reduces anisosome number while enlarging remaining condensates, which retain liquid-like behavior by FRAP and fusion assays. XPO1 overexpression causes fewer, enlarged TDP-43 puncta, including cytoplasmic puncta, with little or no FRAP recovery, interpreted as gel or solid-like aggregates. Anisosome induction reduces detectable nucleoplasmic XPO1 staining. Finally, the authors examine a homozygous TDP-43 K181E iPSC-derived forebrain organoid model, showing increased cytosolic pTDP-43 in K181E/K181E organoids compared to wild-type controls. Chronic low-dose KPT-276 reduces cytoplasmic pTDP-43 without changing total TDP-43 levels. Bulk RNA-seq shows only a modest fraction of dysregulated genes in K181E/K181E organoids are rescued by KPT-276. They conclude that nuclear export, via XPO1, is a key regulator of TDP-43 liquid-to-solid phase transitions and that cytoplasmic aggregation per se may contribute only modestly to TDP-43 proteinopathy, with RNA-processing defects being dominant.

We thank the reviewer for carefully summarizing our study.

The study presents well-executed chemical and genome-wide siRNA screens in a DLD1 TDP-43 2KQ anisosome model and follows up on nuclear transport, particularly XPO1, as a modulator of TDP-43 phase behavior and cytoplasmic aggregation. The screens are impressive in scale, and the microscopy and fluorescence recovery after photobleaching (FRAP) work is technically strong. However, the central mechanistic and disease-relevance claims are not yet sufficiently supported. There are major concerns about the heavy reliance on non-physiological, RNA-binding-defective, and acetylation-mimetic TDP-43 (2KQ) and a homozygous TDP-43 K181E organoid model. An underdeveloped and partly contradictory mechanistic link exists between XPO1 and TDP-43 phase transitions in the context of prior work showing TDP-43 is not a canonical XPO1 cargo. The paper also appears to overinterpret organoid data to conclude that cytoplasmic TDP-43 aggregation plays only a minor role in pathology, based largely on pTDP-43 antibody staining with limited sensitivity and relatively modest rescue readouts. A deeper mechanistic analysis and additional, more physiological validation are needed for this to reach the level of rigor and impact implied by the title and abstract. The work feels screen-rich but conceptually underdeveloped, with key claims outpacing the data. A major revision with substantial new data and tempering of conclusions is warranted. I outline several problematic areas below:

(1) The central mechanistic discoveries are derived almost entirely from a DLD1 colon cancer cell line overexpressing an RNA-binding-defective, acetylation-mimetic TDP-43 2KQ mutant and homozygous TDP-43 K181E iPSC-derived organoids. Both systems are far from physiological. The 2KQ mutation is a synthetic double lysine-to-glutamine mutant originally designed to mimic acetylation and disrupt RNA binding. In this study, essentially all cell-based mechanistic data on phase behavior, screens, and XPO1 effects rely on 2KQ. Yet there is no quantification of how much endogenous TDP-43 is acetylated in degenerating human neurons, nor whether a 2KQ-like acetylation state is ever achieved in vivo. It is not established that the phase behavior of 2KQ recapitulates the physiological or pathological phase behavior of wild-type TDP-43 or genuine disease-linked mutants, which may retain partial RNA binding and different post-translational modification patterns. As a result, it is difficult to know whether the modifiers identified here regulate a highly artificial 2KQ condensate or physiologically relevant TDP-43 condensates. To address this concern, the paper would benefit from quantifying endogenous TDP-43 acetylation at the relevant lysines in control and ALS/FTD patient tissue or more disease-proximal models such as heterozygous TARDBP mutant iPSC neurons, which would justify the focus on an acetyl-mimetic mutant. Key phenomena, including XPO1 dependence of phase behavior, effects of proteasome and HSP90 inhibition, and effects of splicing and translation inhibitors, should be tested for wild-type TDP-43 expressed at near-physiological levels and for one or more bona fide ALS/FTD-linked TARDBP mutants that are not acetyl mimetics. At a minimum, the authors should show that endogenous TDP-43 in neuronally differentiated cells exhibits qualitatively similar responses to XPO1 modulation, rather than exclusively relying on DLD1 2KQ overexpression.

Acetylation of endogenous TDP-43 was reported by several studies. Although it occurs at low levels under normal conditions, TDP-43 acetylation is upregulated under stress conditions (e.g. oxidative stress and proteotoxic stress) (PMID: 25556531; PMID: 28724966). Importantly, Cohen et al. reported the identification of acetylated TDP-43 in ALS patient spinal cord (PMID: 25556531), while Yu et al. showed that endogenous wildtype TDP-43 undergoes demixing when neurons were treated with either a deacetylase inhibitor or proteasome inhibitors (PMID: 33335017). These studies also show that acetylated TDP-43 is defective in RNA binding and more prone to aggregation. Furthermore, ectopic expression of acetylated TDP-43 mimetics in cells and mice induces cellular defects similar to those observed in disease models (PMID: 28724966). Thus, our findings, based on previously established TDP-43 mimetics, should provide valuable information regarding the regulation of TDP-43 phase behavior. We agree with the reviewers that the model used in this study has its limitations, and we will be happy to revise the manuscript to tone down some conclusions, and include more background information to justify the use of TDP-43 acetylation mimetics.

(2) The organoid model is based on a homozygous K181E knock-in line. However, in patients, TARDBP mutations are overwhelmingly heterozygous. Homozygosity is thus a severe, arguably non-physiological sensitized background that may exaggerate nuclear RNA mis-splicing and phase defects and alter the relative contribution of cytoplasmic aggregation versus nuclear loss-of-function. In addition, it is not fully clear from this manuscript whether the structures in K181E organoids are bona fide anisosomes as defined in Yu et al. 2021, characterized by HSP70-enriched central liquid cores with TDP-43 shells and similar FRAP and fusion behavior to anisosomes in the DLD1 model. At present, the organoid section is framed as validation of "anisosome-bearing organoids," but the figures in this manuscript mainly show pTDP-43 puncta and total TDP-43 immunostaining, without detailed structural or biophysical characterization. The authors should explicitly compare heterozygous K181E/+ organoids or another heterozygous TARDBP mutant line with homozygous K181E/K181E organoids to assess whether XPO1 inhibition has similar effects in a genotype that more closely resembles patient genetics. They should provide direct evidence that the K181E condensates in organoids are anisosomes through HSP70 core immunostaining, three-dimensional reconstruction, and FRAP measurements, and clarify whether KPT-276 is acting on anisosome-like structures or more generic cytoplasmic aggregates or puncta. Without this, the leap from a DLD1 2KQ cancer cell model to human ALS/FTD-relevant neurons is not convincingly supported.

The reviewer is correct that the use of homozygous K181E organoids generates a homogenous background that is more sensitive for detecting phosphor-TDP43. The goal of the experiment was to test whether XPO1 inhibition mitigates the aggregation of a TDP-43 disease mutant. For this purpose, we believe that our experimental setup is suitable. We agree that we should not extrapolate the result to overemphasize on its disease connections. We will revise the paper to tone down this part.

Regarding the immunostained signals in K181E organoids, we did not report them as anisosomes. As widely documented in the literature, p-TPD-43 is widely used as a marker of pathological TDP-43 aggregation. P-TDP-43 is enriched in pathological aggregates in human ALS and FTLD patients, colocalized with other aggregation signatures such as ubiquitin and other aggregation prone proteins (PMID: 36008843), and is being used as a diagnostic marker for neurodegeneration (PMID: 31661037). Figure 7A showed that inhibiting nuclear export mitigates the accumulation of p-TDP-43 in mutant tissues. We will revise the subheading and the corresponding text to avoid the confusion.

(3) The title and framing assert that "nuclear export governs TDP-43 phase transitions." However, prior studies such as Pinarbasi et al. 2018 and Duan et al. 2022 indicate that TDP-43 is not a canonical XPO1 cargo and that its export is largely passive, with active nuclear import being the dominant determinant of nuclear localization. The authors cite these studies but still position XPO1 as a central, quasi-direct regulator. The data presented are largely correlative or based on pharmacologic manipulation and overexpression in an overexpression mutant background, with no direct evidence that XPO1 engages TDP-43 in a specific, regulated manner. Even if XPO1 does not engage WT TDP-43, it could still engage the 2KQ variant, which needs to be tested.

We did not conclude or imply the regulation of TDP-43 by XPO1 is direct. In fact, we explicatively mentioned on page 8 that the regulation is likely indirect and mediated by other factors. The sentence reads as “Since XPO1 does not bind TDP-43 directly (Pinarbasi et al., 2018), additional factors likely facilitate XPO1-mediated TDP-43 nuclear egression under this condition.” We can revise the part to make it clearer. We will also revise the title and change the framing accordingly. 

(4) The XPO1 perturbations yield somewhat confusing phenotypes. XPO1 inhibition using Leptomycin B, KPT-276, and Verdinexor reduces anisosome number and enlarges remaining anisosomes, which remain liquid-like by FRAP recovery and fusion assays and stay nuclear. XPO1 overexpression causes fewer, enlarged puncta, but these are FRAP-impaired (gel-like) and redistribute to the cytoplasm. Thus, both decreased and increased XPO1 activity reduce anisosome number and enlarge puncta, but with opposite phase behaviors and subcellular localizations. The model presented in Figure 5L is relatively qualitative and does not resolve these issues. Moreover, XPO1 inhibition globally impairs nuclear export of many cargos and profoundly alters the nuclear environment, transcription, RNA processing, and chromatin. It is therefore difficult to conclude that the observed effects are specific to TDP-43 phase regulation as opposed to secondary consequences of broad nuclear export blockade.

The reviewer correctly summarizes our data and interpretation: XPO1 loss-of-function and gain-of-function generate opposite phenotypes regarding TDP-43 phase behavior. We agree that additional studies are needed to elucidate the underlying mechanism (e.g. direct or indirect), but we feel that belong to a separate study. We plan to re-test the effect of nuclear export inhibition on the subcellular distribution of WT TDP-43 and the acetylation mimetics. We will also add more discussions about the potential indirect effect of XPO-1 inhibition on TDP-43 phase behavior.

(5) The authors show that anisosome induction depletes nucleoplasmic XPO1 signal and that mCherry-XPO1 can be seen in some TDP-43 puncta. However, antibody penetration into anisosomes is limited, so XPO1 depletion from nucleoplasm could reflect sequestration in the anisosome shell or core, but this is not demonstrated. There is no demonstration of physical interaction, even indirect interaction, between XPO1 and TDP-43 or a defined adaptor, nor identification of a specific mutant of XPO1 that selectively disrupts this putative interaction while preserving other functions. The known TDP-43 NES has been shown to be weak and not a functional XPO1-dependent NES in multiple studies. If XPO1 is acting through an adaptor that recognizes 2KQ or K181E specifically, that by itself would bring into question the generality of the mechanism for wild-type TDP-43.

We agree that our observation does not demonstrate an interaction between XPO1 and TDP-43. As mentioned above, we did discuss that the regulation of TDP-43 by XPO1 is likely indirect. We will revise our paper further to separate any speculative statements from the data and narrow our mechanistic claim.

(6) To support a mechanistic claim that nuclear export governs TDP-43 phase transitions, more targeted evidence is needed. The authors should test whether siRNA knockdown or CRISPR interference of XPO1 in the DLD1 2KQ model reproduces the effects seen with Leptomycin B and KPT-276, including FRAP and fusion phenotypes, and verify on-target effects by rescue with an siRNA-resistant XPO1 construct. They should demonstrate that canonical XPO1 cargos behave as expected under the inhibitor conditions used, as a positive control, and that the concentrations used are not grossly toxic. They should attempt to identify or at least constrain candidate adaptors that might enable XPO1-dependent export of TDP-43 through proteomic analysis of XPO1 co-purifying with 2KQ condensates or loss-of-function studies of candidate adaptors from the siRNA screen. Finally, they should test whether a TDP-43 mutant that cannot bind the proposed adaptor still responds to XPO1 manipulation.

The anisosome enlargement phenotype upon XPO1 depletion was seen in our siRNA screend, which was identified by machine-based image analyses using 6 distinct siRNAs. This, together with the chemical inhibition experiments, convinced us that the phenotype is specifically caused by XPO1 inactivation.

When characterizing the effect of XPO1 inhibition on anisosome dynamics, we preferred chemical inhibitor because the effect is acute, and is therefore, less likely to be caused by secondary effects.

Regarding the inhibitor concentration, a literature survey suggested that 50-200nM of Leptomycin B was commonly used. We chose 200nm to ensure a quick and complete inhibition of XPO1-mediated nuclear export (see Figure 3 in PMID: 9628873). This dose is also well tolerated by our cells, at least during the chosen time window.

We did not propose any specific adaptor that mediates XPO1 interaction with TDP-43. The identification of such adaptor is out of the scope of this study. We will revise our paper to avoid this confusion.

(7) Even with these data, what is currently shown is that global modulation of nuclear export capacity can alter the phase behavior and localization of a highly overexpressed RNA-binding-defective TDP-43 mutant and of K181E in organoids. This is important, but it is weaker than asserting that XPO1 directly governs TDP-43 phase transitions in physiological contexts. The title, abstract, and Discussion should be tempered to reflect that nuclear export is one of several pathways, alongside RNA splicing, translation, and proteostasis, that influence TDP-43 phase states in this model, and that the specific mechanism and cargo relationship between XPO1 and TDP-43 remain unresolved and may be indirect.

We will revise the title, abstract, and discussion to temper the conclusion.

(8) The authors conclude that cytoplasmic TDP-43 aggregation plays only a modest role in TDP-43 proteinopathies because in homozygous K181E organoids, chronic KPT-276 treatment almost abolishes cytoplasmic pTDP-43 puncta, yet bulk RNA-seq shows only a relatively small fraction of dysregulated genes are rescued. There are several issues with this inference. Relying primarily on pTDP-43 antibody staining to define cytoplasmic TDP-43 aggregation is limiting. pTDP-43 antibodies label only phosphorylated species and may miss non-phosphorylated, oligomeric, or amorphous TDP-43 species that could still be toxic. Different pTDP-43 antibodies vary in epitope accessibility depending on aggregate conformation and subcellular location. More sensitive approaches, such as high-affinity TDP-43 RNA aptamer probes developed by Gregory and colleagues, biochemical fractionation for SDS-insoluble and urea-soluble TDP-43, and filter-trap assays, would provide a more quantitative assessment of cytoplasmic aggregation and its reduction by KPT-276. Without these, it is not safe to assume that cytoplasmic aggregation has been eliminated, as opposed to one antigenic subclass.

We agree with the reviewer that p-TDP-43 may not represent all aggregate species. However, p-TDP-43 antibodies detect the pathologically validated species most tightly associated with TDP-43 proteinopatheis. In human ALS and FTLD-TDP tissues, cytoplasmic inclusions are strongly immunoreactive for phosphorylated TDP-43 (typically S409/410, as used here). Additionally, p-TDP-43 immunohistochemistry is a routine diagnostic criterion in neuropathology. For these reasons, we believe that the observation that inhibition of XPO1 significantly reduces p-TDP-43 is a very significant finding, as it suggests that an improvement in TDP-43 proteinopathy can be achieved by the inhibition of nuclear transport. We plan to revise the text to better explain the significance of p-TDP-43 staining.

(9) The treatment window, spanning from day 87 to 122 with 20 nanomolar KPT-276, may be too late or too mild to reverse entrenched nuclear RNA-processing defects, even if cytoplasmic inclusions are cleared. Once widespread cryptic exon inclusion and alternative polyadenylation misregulation are established, many downstream changes may become self-sustaining or only partially reversible. Moreover, XPO1 inhibition will massively rewire nucleocytoplasmic transport of many transcription factors, splicing factors, and RNA-binding proteins. Thus, the lack of full transcriptomic rescue cannot be cleanly interpreted as evidence that cytoplasmic aggregates are only modest contributors. It may instead reflect that nuclear dysfunction is primary and XPO1 inhibition does not correct, and may even exacerbate, certain nuclear defects.

We agree with the reviewer that the lack of rescue may be caused by technical issues. We will remove the RNAseq data and related texts since it is not essential for our main conclusion.

(10) To support a causal statement about the modest contribution of cytoplasmic aggregates, one would want more direct measures of neuronal health and function, such as cell death, neurite complexity, synaptic markers, and electrophysiology before and after KPT-276, not only transcriptomics. A way to selectively reduce cytoplasmic aggregation without globally inhibiting nuclear export would allow comparison of outcomes.

We will remove the discussion regarding the role of cytoplasmic aggregates in disease.

(11) Given these caveats, the concluding statements that cytoplasmic TDP-43 aggregation is only a modest contributor should be substantially softened. A more defensible interpretation is that in this homozygous K181E organoid model, chronic global XPO1 inhibition reduces pTDP-43-positive cytoplasmic puncta but only partially normalizes the steady-state transcriptome, suggesting that persistent nuclear RNA-processing defects and other pathways continue to drive pathology.

We agree with the review and will revise this part accordingly.

(12) The screens are a major strength but need more rigorous validation for key hits, especially nuclear transport factors. For the siRNA screen, hits are filtered by anisosome number per nucleus, but there is no direct demonstration in the main text that XPO1 or CSE1L knockdown is efficient at the messenger RNA or protein level. For the highlighted genes, Western blot or quantitative polymerase chain reaction validation and phenotypic rescue would strengthen confidence. For small-molecule hits, it is not systematically shown that anisosome modulation is independent of changes in total TDP-43 2KQ expression or gross toxicity. Translation inhibitors are tested for this, but for many other hits, including proteasome, HSP90, and kinase inhibitors, expression and general nuclear structure should be monitored. Given the reliance on anisosome count as a readout, secondary screens that specifically distinguish changes in TDP-43 expression levels, changes in nuclear morphology or cell cycle, and specific changes in anisosome phase behavior, including FRAP and fusion for top hits, would greatly increase interpretability.

For the siRNA screen, each positive hit was confirmed by two rounds of screen with 6 independent siRNAs in total. Although we did not validate the knockdown efficiency due to the large number of hits, we routinely include a positive siRNA control in our study (siRNAdeath), which targets an essential gene. Transfection efficiency was controlled by measuring cell viability after knocking down this essential gene. In addition, the identification of XPO1 as a positive regulator of TDP-43 phase behavior was independently validated by our chemical genetic screens. We feel confident that XPO1 is a key modulator of TDP-43 phase behavior. For chemical treatment experiments, the anisosome fusion phenotypes could be detected as early as 5 h post treatment. Given the short treatment, we do not expect a significant change in protein level or toxicity.

(13) The classification of condensates as liquid versus gel-like or solid is based almost entirely on FRAP recovery or lack thereof. While FRAP is appropriate, interpretations could be made more robust by including half-region-of-interest bleach controls and assessing mobile fractions and recovery kinetics more quantitatively across conditions. Complementing FRAP with other phase-behavior assays such as sensitivity to 1,6-hexanediol, shape relaxation after deformation, and coarsening behavior over longer timescales would strengthen the analysis. At present, some assignments, such as that XPO1 overexpression drives a gel-like transition, are reasonable but somewhat qualitative.

In this study, we described two types of condensates formed by TDP-43 2KQ, one characterized previously as nuclear anisosome and the other as cytosolic puncta in XPO1 over-expressing cells. The two can be clearly distinguished by several features including the subcellular localization, shape, and mobility. We feel that our FRAP data clearly segregate these puncta into two distinctive types of assemblies. The difference in fluorescence recovery rate is huge. The proposed half-region-of-interest bleach is technically challenging for small anisosomes under normal conditions. When they were enlarged by Leptomycin B treatment, we did perform both whole anisosome bleach and partial bleach (Figure 5D, I). Both assays demonstrate that TDP-43 in these enlarged anisosomes is highly mobile.

(14) For the Leptomycin B and KPT-276 experiments in cells and organoids, it would be important to confirm that canonical XPO1 cargo proteins accumulate in the nucleus and that the concentrations used are within a range that is not overtly toxic over the experimental timeframe. Assessing nuclear morphology, chromatin condensation, and general transcriptional activity through global RNA synthesis or key reporter genes would ensure that observed effects are not secondary to severe global nuclear export collapse.

In Leptomycin B treatment experiments, we carefully chose a dose that was previously validated (see Figure 3 in PMID: 9628873). Based on our DAPI staining, the nuclear morphology appears normal (Figure 5A). Additionally, in cell line-based experiment, the effect of Leptomycin B on anisosomes was detected 6-8 hours post treatment. The change in global protein synthesis should be relatively minor at this time point. In the organoid experiment, the drug dose was determined by a pre-experiment in which the morphology of organoids was evaluated after prolonged treatment with different doses of the inhibitors.

(15) In the organoid section, it is not clear how many independent iPSC clones and organoid batches were used per condition, nor whether batch effects were assessed in the bulk RNA-seq analysis. This should be fully specified and ideally controlled with isogenic wild-type and K181E clones. For transcriptional rescue, it is important to know whether the changes in wild-type organoids treated with KPT-276 are negligible. A direct wild-type comparison with or without KPT-276 is important to disentangle general drug effects from K181E-specific rescue. More detailed quantification of total TDP-43 and pTDP-43 in both nuclear and cytoplasmic fractions, including biochemical fractionation if possible, would strengthen the assertion that KPT-276 specifically reduces cytosolic pTDP-43 aggregates while sparing nuclear TDP-43.

The organoid experiment was performed with two batches per condition. This is to reduce the effect of batch variation. The wildtype cells and K181E mutant are derived from the same genetic background. We will revise the text to clarify these issues. Given the cost of this experiment, we did not include drug-treated wild-type as a control. Given the criticisms by review 1 and 2 on the RNAseq data, we will remove this non-essential data from our revision.

(16) Beyond the core issues above, several additions could greatly enhance the impact. The manuscript currently emphasizes XPO1, but the genetic and chemical data clearly implicate RNA splicing, translation, and proteostasis as equally strong or stronger regulators of TDP-43 phase states. A more integrated model that explains how these pathways intersect, for example, how splicing factor availability, ribosome loading, and proteasome capacity co-govern anisosome nucleation, growth, and hardening, would be valuable.

We agree with the reviewer that these are important directions for future studies. We will include some discussions on a possible model that integrate these factors.

(17) A key unresolved question is whether XPO1 is acting directly on TDP-43, or instead primarily regulates anisosomes by exporting other factors that more proximally control TDP-43 phase behavior. Given that TDP-43 is not a canonical XPO1 cargo and prior work indicates that its nuclear export is largely passive, it seems at least as plausible that XPO1 inhibition alters the nuclear concentration or localization of splicing factors, RNA-binding proteins, chaperones, or other modifiers identified in the screens, and that changes in these proteins secondarily reshape anisosome dynamics. In other words, XPO1 may be exporting a more direct regulator of anisome formation and hardening, rather than exporting TDP-43 itself in a specific, regulated way. The current data do not distinguish between these possibilities. Systematic identification of XPO1-dependent cargos that colocalize with or biochemically associate with anisosomes, combined with targeted perturbation of their nuclear export, would be needed to determine whether the relevant XPO1 substrate in this system is actually TDP-43 or an upstream modulator of its phase behavior.

The reviewer raises an important point. We did include some discussions along this line in our paper. We can add more to further clarify this issue. Again, as mentioned in the original draft, we did not conclude there is an interaction between TDP-43 and XPO1.

(18) Testing whether identified modifiers converge on nuclear TDP-43 concentration would be informative. Since phase separation is concentration-dependent, measuring nuclear versus cytoplasmic TDP-43 levels across key perturbations, including splicing inhibition, translation inhibition, proteasome inhibition, HSP90 inhibition, and XPO1 modulation, would help determine whether modifiers mainly work by changing nuclear TDP-43 concentration or by altering interaction networks and the material properties of condensates.

We will measure the nuclear TDP-43 concentration in our imaging experiments and add the data to a revised version.

(19) Examining other ALS-relevant RNA-binding proteins would be valuable. Given the role of XPO1 and other hits, it would be informative to briefly test whether similar principles apply to FUS, hnRNPA1, or other ALS-relevant RNA-binding proteins in the same cellular context, to argue for generality versus TDP-43-specific idiosyncrasies of the 2KQ system.

We agree that this is an important issue but we feel the proposed experiments are beyond the scope of the study.

(20) The Introduction sometimes implies that anisosomes are common and well-established intermediates en route to pathology. It would be helpful to more clearly state that, to date, anisosomes are primarily observed in overexpression and mutant systems and have not yet been unequivocally demonstrated in human patient tissue. The link between PDGFRβ, PAK4, GSK-3β, and YAP and TDP-43 phase dynamics is intriguing but only briefly mentioned. The authors should either expand on this or tone down the emphasis in the Results section.

We will revise the introduction accordingly.

(21) In the organoid methods, the authors should consider clarifying whether doxycycline is continuously used, which might alter TDP-43 expression and nuclear transport in a non-negligible way.

The organoid model does not involve protein overexpression or doxycycline treatment. We measured endogenous p-TDP-43. We will revise to paper to avoid the confusion.

(22) For statistical methods, it would be beneficial to indicate whether multiple-comparison corrections were applied for the many FRAP, anisosome count, and size comparisons beyond DESeq2 internal corrections for RNA-seq.

We will add this information to the figure legends during revision.

(23) Some figure legends could more clearly indicate whether the images shown are single z-planes or maximum intensity projections and how the thresholding for anisosome detection was performed.

We will revise the figure legends to include this information. As for anisosome detection, because they are so obvious, standard thresholding was sufficient to identify them.

(24) In its current form, the manuscript contains an impressive set of screens and some nicely executed imaging of TDP-43 condensates, highlighting nuclear export among other pathways as a modulator of TDP-43 phase behavior. However, the physiological relevance is undercut by heavy reliance on an acetylation-mimetic, RNA-binding-defective TDP-43 mutant and a homozygous K181E organoid model. The mechanistic link between XPO1 and TDP-43 remains largely inferential and partly at odds with prior work. The conclusion that cytoplasmic TDP-43 aggregation is only a modest contributor to disease is not firmly supported by the available data.

We agree with the reviewer that the strength of the study is our unbiased approach that identify pathways capable of modulating TDP-43 phase separation behavior. We will revise our paper to carefully discuss the potential physiological relevance of our study and tone down some mechanistic conclusions, as suggested by the reviewer.

(25) With substantial additional mechanistic work, particularly around XPO1, rigorous validation in more physiological TDP-43 contexts, more sensitive detection of cytoplasmic TDP-43 aggregates, and a tempering of the central claims, this study could make a meaningful contribution to understanding how nucleocytoplasmic transport and other cellular pathways influence TDP-43 phase transitions and aggregation. The work should be reframed as an important screening study that identifies nuclear export as one among several cellular processes that modulate TDP-43 phase behavior in a model system, rather than as a definitive demonstration that nuclear export governs pathological TDP-43 aggregation in disease.

We will reframe the study as an important screening study that identifies nuclear export among several other pathways as modulators of TDP-43 phase behavior.

Reviewer #2 (Public review):

Summary:

This manuscript addresses an important and timely question in TDP-43 biology by systematically identifying regulators of TDP-43 anisosome formation, with a particular focus on nuclear export via XPO1. Using a combination of unbiased chemical screening, genetic perturbation, and advanced imaging approaches, the authors propose that inhibition of nuclear export modulates the abundance and biophysical properties of TDP-43 anisosomes. The study is conceptually innovative and has potential relevance for neurodegenerative diseases characterized by TDP-43 pathology. However, significant concerns regarding experimental controls, reporting transparency, and model translatability currently limit the strength of the conclusions and the interpretability of several key findings.

We thank the reviewer for acknowledging the significance and innovation of our study.

Strengths:

(1) The study employs an unbiased, hypothesis-free compound screen to identify regulators of TDP-43 anisosome formation, which is a major strength and reduces confirmation bias.

(2) The authors combine chemical and genetic screening approaches, providing orthogonal validation of key pathways and increasing confidence in the biological relevance of top hits.

(3) The focus on biophysical properties of TDP-43 assemblies, assessed through imaging and FRAP, moves beyond simple presence/absence of aggregates and provides mechanistic insight into the biophysical states of TDP-43.

(4) The use of multiple experimental modalities, including live-cell imaging, FRAP, pharmacological perturbation, and transcriptomic analysis, reflects a technically sophisticated and ambitious study design.

(5) The authors attempt to extend findings beyond immortalized cancer cell lines by incorporating organoid models, demonstrating awareness of disease relevance and translational importance.

Overall, the manuscript is clearly written and logically structured, making complex experimental workflows accessible and the central hypotheses easy to follow.

Weaknesses:

Despite its strengths, the manuscript has several major limitations that affect data interpretation and confidence in the conclusions.

(1) Lack of appropriate controls for overexpression experiments:

A central concern is the absence of proper controls for TDP-43 and XPO1 overexpression. Prior studies (including those cited by the authors, Archbold et al.2018) show that overexpression of WT TDP-43 alone is toxic to neurons. Thus, the experimental system itself may induce anisosome formation independently of the mechanisms under study. Similarly, XPO1 overexpression lacks a suitable control (e.g., mCherry alone or mCherry fused to a protein known to be independent of TDP-43). The near-complete colocalization of XPO1 with TDP-43 anisosomes upon overexpression raises the possibility that these structures reflect non-physiological protein accumulation rather than regulated assemblies.

As mentioned in our response to reviewer 1, point 1, we will add more discussion regarding the use of acetylation mimetics in our study. We agree with the reviewer that these large puncta (both anisosomes and gel-like structures) likely resulted from TDP-43 overexpression. Nevertheless, in a titration experiment done by Yu et al. 2020 (PMID: 33335017), they showed that ectopic TDP-43 undergo demixing even at concentrations lower than endogenous TDP-43, although the demixed puncta were very small. Their result suggested that overexpression per se does not change TDP-43 phase behavior, only enlarging the demixed TDP-43 structures. This is necessary for our screen and imaging-based characterization. We will revise the text to clarify this point.

For XPO1, we did include mCherry alone control in the study but due to space limit in Figure 5, we did not include it. We can put the data in a Supplementary Figure during revision.

(2) Insufficient experimental and analytical transparency:

The manuscript frequently lacks clear reporting of experimental details. In multiple figures, the stated number of independent experiments does not match the number of data points shown, making it difficult to assess statistical validity. Concentrations used in the compound screen are not clearly defined, nor is it stated whether multiple concentrations were tested. It is unclear how many wells, cells, or independent cultures were analyzed. The criteria used to reduce 1,533 screening hits to 211 candidates via STRING analysis are not explained. Knockdown and overexpression efficiencies are not reported.

We apologize for these omissions. We will add more experimental details to the figure legends and the method part. For the imaging experiments, data points reflect randomly selected individual cells imaged in 2-3 independent biological repeats. For chemical screens, we screened against NCATS libraries first at top concentration (10 mM) to ensure inhibitory efficacy for all compounds. In the follow-up study, we validated the top hits using a series of concentrations, as shown in Figure 1B.

We will explain the STRING analysis in more detail. We did not check XPO1 knockdown efficiency in high through-put screens (HTS) for several reasons. Firstly, the large number of positive hits makes it impossible to check knockdown efficiency for all these hits. Secondly, the effect of XPO1 knockdown on anisosomes was seen with 6 different siRNAs in two rounds of screens. Thirdly, in the HTS protocol, we routinely included a transfection control (siRNAdeath) to indicate high transfection efficiency. We would only process the data if siRNAdeath control killed > 90% of the cells.

(3) RNA-seq concerns:

The RNA-seq experiments are particularly problematic. The number of biological replicates per condition is not stated, and heatmaps suggest that only one sample per group may have been used, which would preclude statistical analysis. No baseline comparison between WT and mutant TDP-43 is shown. Given that TDP-43 is an RNA-binding protein, splicing analyses would be far more informative than gene expression alone, yet no splicing data are presented. Moreover, nuclear retention of TDP-43 does not preclude nuclear aggregation, which may still impair its splicing function.

We apologize for the lack of clarity regarding the RNA-seq design. For each condition, organoids of two independently differentiated batches were treated in triplicate. We pooled the organoids of the same treatment from the two batches to reduce the impact of batch variation.

Given the criticisms from both reviewer 1 and 2 on the limitation of the RNAseq study, we plan to remove this data from the revised manuscript.

(4) Limited translatability to neuronal biology:

All anisosome analyses are performed in a cancer cell line, raising concerns about relevance to post-mitotic neurons. While organoids are used as a secondary model, the assays performed do not overlap with those used in cancer cells, making it difficult to assess whether anisosome-related mechanisms are conserved. Neuronal toxicity, a critical outcome given known TDP-43 biology, is not assessed. Prior work has shown that WT TDP-43 overexpression alone is toxic to neurons, yet this is not addressed.

We agree with the reviewer that the model used in this study is not directly relevant to neurodegeneration. However, as pointed out by the reviewer, neurons are much more sensitive to TDP-43-associated toxicity. By contrast, the cell line used in this study can tolerate TDP-43 overexpression with no detectable cytotoxicity. This feature makes it feasible to evaluate how different cellular processes modulate TDP-43 phase behavior without the confounding effect from toxicity. The fact that TDP-43 expression was induced for a short period of time also help minimize the impact of toxicity. Notably, the processes identified by our screens are all house-keeping pathways that is present in neurons. Thus, we believe that the reported findings are likely applicable to neurons, though we will revise our paper to make sure that we don’t overstate the clinical relevance of our work.

(5) Conceptual and interpretational gaps:

The authors quantify anisosome number but also report conditions in which anisosome number decreases while size increases. The biological interpretation of larger anisosomes is not discussed, and whether this reflects improvement or worsening of pathology is unclear. Compounds targeting the same mechanism (e.g., nuclear export inhibition) are inconsistently used across experiments (KPT compounds, verdinexor, leptomycin B), raising concerns about reproducibility. In organoids, the experimental paradigm shifts to long-term treatment (35 days vs. 16 hours), further complicating interpretation.

As pointed out by the reviewer 1 in point 4 above, we do not have evidence to establish a convincing correlation between the size of anisosomes and clinical phenotypes. Regarding the use of different drugs for different experiments, the initial screen identified KPT and Verdinexor because Leptomycin B was not in our library. In the follow-up studies, we switched to Leptomycin B because 1) it is commercially available; 2) it is highly potent and specific; 3) it was more commonly used as inhibitors of XPO1 according to the literature. However, for the organoid study, we had to switch back to KPT because of the toxicity issue associated with long-term application of Leptomycin B.

(6) Overinterpretation of rescue effects:

Although the authors state that they aim to test whether nuclear export inhibition rescues neuronal defects, no functional neuronal readouts are provided (e.g., viability, morphology, axon outgrowth, or electrophysiological measures). RNA-seq alone is insufficient to support claims of rescue.

Our interpretation of the RNA-seq data was that the rescue effect by nuclear export inhibition was limited and likely insignificant. Given that this negative data is not conclusive, we will remove it from the revised manuscript.

(7) Finally, the model does not appear to exhibit cytosolic TDP-43 aggregation at baseline. It remains unclear whether longer induction would produce cytosolic gel-like assemblies and whether these would be prevented by nuclear export inhibition. Long-term data are shown only in organoids, yet anisosome formation is not assessed there.

The expression system used in the study reaches a steady state after 48 h of induction. At this point, we did not observe any gel-like structures. We can clarify this point during revision.

Reviewer #3 (Public review):

Summary:

TDP-43 proteinopathy is broadly found in neurodegenerative diseases. This manuscript investigates how nuclear export influences the biophysical properties of TDP-43. The authors use a combination of chemical screening and genome-wide siRNA screening to identify pathways that modulate TDP-43 liquid-to-solid transitions. Overall, the study employs a broad array of approaches and addresses an important question in TDP-43 pathobiology. The identification of nuclear export as a central regulator is compelling and conceptually aligns with the emerging view that TDP-43 nucleocytoplasmic trafficking is a major defect in neurodegeneration.

Strengths:

This work integrates chemical and genetic screening to identify novel modifiers. The candidates were validated in both reporter cell lines and iPS-differentiated organoids. The findings support the nucleocytoplasmic transport is important for the biophysical properties of TDP-43.

We thank the reviewer for acknowledging the significance and strength of our study.

Weaknesses:

The mechanisms underlying the connection between nuclear export and phase transition need further clarification. Broader consequences of XPO1 inhibition are not addressed.

We agree that our study does not address how nuclear export inhibition affect TDP-43 phase behavior. As discussed in the paper, we proposed that the effect of nuclear export inhibition on TDP-43 phase separation is likely indirect. The most likely scenario is that inhibition of nuclear export changes the nuclear environment over time, which affects TDP-43 phase separation. We have tried to isolate nuclear extracts from control and LMB-treated cells and used mass spec to identify proteins that are differentially present in the nucleus. However, knockdown of the identified top candidates did not abolish LMB-induced phase alteration. Considering our observation that RNA splicing is another modulator of TDP-43 phase behavior, it is possible that it is the combined change of RNA and protein composition in the nucleus that alters TDP-43 phase behavior. However, defining the mechanism would require substantial work that is beyond the scope of the current study.

Author response:

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

Joint Public review:

Weaknesses:

(1) Controls for the genetic background are incomplete, leaving open the possibility that the observed oviposition timing defects may be due to targeted knockdown of the period (per) gene but from the GAL4, Gal80, and UAS transgenes themselves. To resolve this issue the authors should determine the egg-laying rhythms of the relevant controls (GAL4/+, UAS-RNAi/+, etc); this only needs to be done for those genotypes that produced an arrhythmic egg-laying rhythm.

(2) Reliance on a single genetic tool to generate targeted disruption of clock function leaves the study vulnerable to associated false positive and false negative effects: a) The per RNAi transgene used may only cause partial knockdown of gene function, as suggested by the persistent rhythmicity observed when per RNAi was targeted to all clock neurons. This could indicate that the results in Fig 2C-H underestimate the phenotypes of targeted disruption of clock function. b) Use of a single per RNAi transgene makes it difficult to rule out that off-target effects contributed significantly to the observed phenotypes. We suggest that the authors repeat the critical experiments using a separate UAS-RNAi line (for period or for a different clock gene), or, better yet, use the dominant negative UAS-cycle transgene produced by the Hardin lab (https://doi.org/10.1038/22566).

We have followed the referee advice,repeating the experiments with the dominant negative UAS-cyc^DN. They nicely confirm our conclusions: the abolition of the cellular clock in LNd neurons rule out the rhythmicity of oviposition. The results are presented in Fig. 3 of the new manuscript, panels H to N. We thank the reviewer for this suggestion that has definitely improved our paper, since it allows us to confirm our result using both a different driver and a different UAS sequence. In addition, we included the required GAL4 controls, which can be found in Panels E, L of the figure as well as average egglaying profiles for all genotypes involved (Panels B, D, F, I, K and M). Regarding the MB122Bsplit-Gal4>UAS-per^RNAi experiment, we moved it to a supplementary figure (Figure 3S1). The paragraph where the new Figure 3 is discussed has been modified accordingly.

(3) The egg-laying profiles obtained show clear damping/decaying trends which necessitates careful trend removal from the data to make any sense of the rhythm. Further, the detrending approach used by the authors is not tested for artifacts introduced by the 24h moving average used.

The method used for the assessment of rhythmicity is now more fully explained and tested in the supplementary material. In particular, the issue of trend removal is treated in the second section of the SM, and the absence of "artifacts" (interpreted as the possibility of deciding that a signal is rhythmic when it is not, or vice versa) shown in figs. S3 to S5.

(4) According to the authors the oviposition device cannot sample at a resolution finer than 4 hours, which will compel any experimenter to record egg laying for longer durations to have a suitably long time series which could be useful for circadian analyses.

The choice of sampling every 4 hours is not due to a limitation imposed by the device used. In fact the device can be programmed to move at whatever times are desired. As mentioned in the Material and Methods section, "more frequent sampling gives rise to less consistent rhythmic patterns", because the number of eggs sampled at each time slot become too small. In particular, we have tested sampling at intervals of 2 hours, and we have observed that this doubles the work performed by the experimenter but does not lead to an improvement in the assessment of rhythmicity.

(5) Despite reducing the interference caused by manually measuring egg-laying, the rhythm does not improve the signal quality such that enough individual rhythmic flies could be included in the analysis methods used. The authors devise a workaround by combining both strongly and weakly rhythmic (LSpower > 0.2 but less than LSpower at p < 0.05) data series into an averaged time series, which is then tested for the presence of a 16-32h "circadian" rhythm. This approach loses valuable information about the phase and period present in the individual mated females, and instead assumes that all flies have a similar period and phase in their "signal" component while the distribution of the "noise" component varies amongst them. This assumption has not yet been tested rigorously and the evidence suggests a lot more variability in the inter-fly period for the egg-laying rhythm.

As stressed in the paper, and in the new Supplementary Material, the individual egg records are very noisy, which in general precludes the extraction of any information about the underlying period and phase. The workaround we (and others, e.g. Howlader et al. 2006) have used is analyzing average egg records for each genotype. Even though this implies assuming the same period and phase for all individuals, we have observed, using experiments with synthetic data, that small variations in individual periods (of the same amount as those present in real experiments where the period of some flies can be assessed individually) still allow us to use our method to decide if the genotype is rhythmic or not. This issue is discussed at length in the new Supplementary Material. There we also discuss an experiment with real flies, showing the individual records, and the corresponding periodograms, for each fly, for a rhythmic (Fig. S14) and an arrhythmic genotype (Fig. S17).

(6) This variability could also depend on the genotype being tested, as the authors themselves observe between their Canton-S and YW wild-type controls for which their egg-laying profiles show clearly different dynamics. Interestingly, the averaged records for these genotypes are not distinguishable but are reflected in the different proportions of rhythmic flies observed. Unfortunately, the authors also do not provide further data on these averaged profiles, as they did for the wild-type controls in Figure 1, when they discuss their clock circuit manipulations using perRNAi. These profiles could have been included in Supplementary figures, where they would have helped the reader decide for themselves what might have been the reason for the loss of power in the LS periodogram for some of these experimental lines.

We have added the individual periodograms of the arrhythmic lines to the Supplementary material (Figs. 3S2, 3S5 and panel G of Fig. 3S1), where they can be compared with their respective controls (Figs 3S3, 3S4, 3S6, 3S7 and panel F of Fig. 3S1).

(7) By selecting 'the best egg layers' for inclusion in the oviposition analyses an inadvertent bias may be introduced and the results of the assays may not be representative of the whole population.

We agree that the results may be biased for 'the best egg layers'. We remark however, that the flies that have been left out lay very few eggs, some of them even laying no eggs on a whole day. For these flies it is difficult to understand how one can even speak of egg laying rhythmicity (let alone how one can experimentally assess it). Thus, we think it might be misleading to speak of results as "representative of the whole population". Furthermore, it is even possible that the very concept of egg laying rhythmicity makes little sense if flies do not lay enough eggs.

(8) An approach that measures rhythmicity for groups of individual records rather than separate individual records is vulnerable to outliers in the data, such as the inclusion of a single anomalous individual record. Additionally, the number of individual records that are included in a group may become a somewhat arbitrary determinant for the observed level of rhythmicity. Therefore, the experimental data used to map the clock neurons responsible for oviposition rhythms would be more convincing if presented alongside individual fly statistics, in the same format as used for Figure 1.

In general, we have checked that there are no "outliers", in the sense of flies that lay many more eggs than the others in the experiment. But maybe the reviewer is referring to the possibility that a few rhythmic flies make the average rhythmic. This issue is addressed in the supplementary material, at the end of section "Example of rhythmicity assessment for a synthetic experiment". In short, we found that eliminating some of the most rhythmic flies from a rhythmic population makes the average a bit less rhythmic, but still significantly so. Conversely, if these flies are transferred to an arrhythmic population, the average is still non rhythmic.

Regarding "the number of individual records that are included in a group may become a somewhat arbitrary determinant for the observed level of rhythmicity", we stress that we have not performed a selection of flies for the averages. All of the flies tested are included in the average, independently of their individual rhythmicity, provided only that they lay enough eggs.

(9) The features in the experimental periodogram data in Figures 3B and D are consistent with weakened complex rhythmicity rather than arrhythmicity. The inclusion of more individual records in the groups might have provided the added statistical power to demonstrate this. Graphs similar to those in 1G and 1I, might have better illustrated qualitative and quantitative aspects of the oviposition rhythms upon per knockdown via MB122B and Mai179; Pdf-Gal80.

We are aware that in the studies of the rhythmicity of locomotor activity the presence of two significant peaks is usually interpreted as a “complex rhythm”, i.e. as evidence of the existence of two different mechanisms producing two different rhythms in the same individual. In our case, since the periodograms we show assess the rhythmicity of the average time series of several individuals, the two non-significant peaks could also correspond to the periods of two different subpopulations of individuals. However, a close examination of the individual periodograms, now provided as Supplementary Figures 3S2 to 3S9, does not show any convincing evidence of any of these two possibilities.

Another possibility could be that such peaks are simply an artifact of the method in the analysis of time series that consist of very few cycles and also few points per cycle. In the supplemenatry material we show that this can indeed happen. Consider, for example, periodograms 2 and 4 in Fig. S12 of the SM. Even though both of them display two non significant peaks, these periodograms correspond to two synthetic time series that are completely arrhythmic.

We have added to the manuscript a paragraph discussing the issue of possible bimodality (next to last paragraph in subsection "The molecular clock in Cry+ LNd neurons is necessary for rhythmic egg-laying").

Wider context:

The study of the neural basis of oviposition rhythms in Drosophila melanogaster can serve as a model for the analogous mechanisms in other animals. In particular, research in this area can have wider implications for the management of insects with societal impact such as pests, disease vectors, and pollinators. One key aspect of D. melanogaster oviposition that is not addressed here is its strong social modulation (see Bailly et al.. Curr Biol 33:2865-2877.e4. doi:10.1016/j.cub.2023.05.074). It is plausible that most natural oviposition events do not involve isolated individuals, but rather groups of flies. As oviposition is encouraged by aggregation pheromones (e.g., Dumenil et al., J Chem Ecol 2016 https://link.springer.com/article/10.1007/s10886-016-0681-3) its propensity changes upon the pre-conditioning of the oviposition substrates, which is a complication in assays of oviposition rhythms that periodically move the flies to fresh substrate.

We agree that social modulation can be important for oviposition, as has been shown in the paper cited by the reviewer. But we think that, in order to understand the contribution of social modulation to oviposition, it is important to know, as a reference for comparisons, what the flies do when they are isolated. Our aim in this work has been to provide such a reference.

Recommendations for the authors:

(1) The weaknesses identified in the Public review could be addressed as follows: etc.

We have followed the suggestions of the editor and addressed each of the weaknesses mentioned (see details above).

(2) Could the authors comment on their choice of using individual flies for their assay rather than (small) groups of flies? Is it possible that their assay would produce less noisy results with the latter?

First we want to emphasize that our aim here was to assess the presence of individual rhythmicity, free from any external influences, whether arising from environmental external cues (such as light or temperature changes) or by social interactions (with other females or males). However, we were also curious about the behavior when males were put in the same chamber with each female. We performed a few tests and the results were very similar to what we obtained with single females.

(3) Minor points:

(a) Line 57-58 - "around 24 h and a peak near night onset (Manjunatha et al., 2008). Egglaying rhythmicity is temperature-compensated and remains invariant despite the nutritional state": Rephrase to something simpler like temperature and nutrition compensated.

Corrected.

(b) Line 56-57 - "The circadian nature of this behavior was revealed by its persistence under DD with a period around 24 h and a peak near night onset (Manjunatha et al., 2008)." A better reference here would be to Sheeba et al, 2001 for preliminary investigations into the egg-laying rhythms of individual flies and McCabe and Birley, 1998 for groups of flies under LD12:12 and DD.

Suggestion accepted.

(c) Line 65-67 - "We determined..... molecular clock in the entire clock network reduced the LNv did not." This suggests that it was unknown until now that LNv does not have a role, whereas Howlader et al 2006 already suggested that. The reader becomes aware of this at a later part of the manuscript. Please revise.

This has been revised, and the citation to Howlader et al 2006 added to the new sentence.

(d) Line 67 - "impairing the molecular clock in the entire clock network reduced the circadian rhythm of.."; saying "Reduced the power of the circadian rhythm" might be better phrasing."

Suggestion accepted.

(e) Line 72 - using the Janelia hemibrain dataset.

Corrected

(f) Line 72 typo "ussing", should be 'using'.

Corrected.

(g) Line 94: why is the periodic signal the same for all on the first day of DD?

It is well known that in LD conditions activity is driven by the environmental light-dark cycle, which entrains the endogenous circadian clock of all flies. Even after the transition to DD, the effects of this entrainment persist for a few days, allowing the individual rhythmic patterns set by the light-dark cycle to remain synchronized for at least a few cycles. We are assuming that the same happens with oviposition. A sentence has been added explaining this (beginning of third paragraph of subsection "Egg-laying is rhythmic when registered with a semiautomated egg collection device").

(h) Figure 1A-D, Were all flies included or only rhythmic flies? Please make this clear. How do you distinguish rhythmic and arrhythmic flies in Figure 1E? Their representative individual plots of egg number graphs are required. Why was the number of flies under DD decreased from 20 to 18?

Throughout the paper, the analysis of average rhythmicity has been performed including all flies, since we postulate that even flies that individually can be classified as non rhythmic have a rhythm that is corrupted by noise, and that this noise can be partially subtracted by performing an average. The explanation of the characterization of rhythmic and arrhythmic individuals is in the Methods section, under the Data Analysis subsection. This is now fully developed in the Supplementary material, where the individual plots for some of the genotypes are included.

Regarding the question of the number of flies having "decreased from 20 to 18?", there is a misunderstanding here. The results depicted in Figure 1, and in particular in panel E, correspond to two different experiments: one performed only in LD (7 days, n=20), and a second one performed for 5 days in DD, with one previous day in LD (n=18).

(i) Figure E and K, Are n=20, 18, and n=30, 22 the total numbers of flies including both rhythmic and nonrhythmic? If so, it would be better to put them in the column, not in the rhythmic column.

The figure has been corrected.

(j) Line 107-108, please provide a citation for this statement.

We have added two references: Shindey et al. 2016, and Deppisch et al. 2022.

(k) Figure 1, 2, etc., please write a peak value inside the periodogram graph. This makes comparison easier.

The peak values have been added in all Figures.

(l) Line 184-185, Figure 2F, tau appears shorter in Clk4.1>perRNAi flies than in control, which suggests that DNp1 may play a role?

As explained in the Supplementary Material, the particularities of oviposition records (discrete values, noise, few samples per period, etc.) preclude an accurate determination of the period if the record is considered as rhythmic. In particular, Fig. S4 shows that differences of 1 hour between the real and the estimated periods are not unusual.

(m) Figure 4. Why are 2 controls shown? Please explain. Are they the same strains?

The two controls shown are the UAS control and the GAL4 control. This information has now been added to the figure.

(n) Line 314 'that' should be 'than'?

Corrected.

(o) Line 73-74 - Phrasing is not clear in: "LNds and oviposition neurons, consisting with, the essential role of LNds neurons in the control of this behavior.""

Corrected.

(p) Line 81-84 - "the experiments particularly demanding and labor-intensive. In this approach, eggs are typically collected every 4 hours (sometimes also every 2 hours), which usually implies transferring the fly to a new vial or extracting the food with the eggs and replacing it with fresh food in the same vial (McCabe and Birley, 1998; Menon et al., 2014)." McCabe and Birley had an automated egg collection device designed for groups of flies, which sampled eggs laid every hour for 6 days. Please remove this reference in this context

Reference removed.

(q) Line 91-92 - "The assessment of oviposition rhythmicity is challenging because the decision of laying an egg relies on many different internal and external factors making this behavior very noisy." This sentence makes it appear that 'assessment' is the limitation. Even locomotor activity is governed by many internal and external factors, yet we can obtain very robust rhythms. The sentence that follows is also not easy to digest. Can the authors frame the idea better?

We have rewritten the corresponding paragraph in order to make it more clear (second paragraph of the Results section). Additionally, the Supplementary Material contains now a more detailed explanation and analysis of the method used.

(r) Line 104-107 - rhythmic (with a period close to 24 h, Figure 1F) although the average egg record is strongly rhythmic with a period around 24 h (Figure 1B). Under DD condition, individual rhythmicity percentages are the same as in LD (Figure 1E) and their average record is also very rhythmic with a period of 24 h (Figure 1D). 'Strongly rhythmic' and 'very rhythmic' are less indicative of what is happening with the oviposition rhythm and can be phrased as robust instead, with a focus on their power measured.

We have accepted the suggestion.

(s) Line 108-110 - "Thus, egg-laying displays a much larger variability than locomotor activity, compounding the difficulty of observing the influence of the circadian clock on this behavior." The section discussed here does not illustrate the variability in egg-laying as much as the lack of robustness of the rhythm. The variation in rhythmicity going from CS flies (~70% rhythmic) to yw flies (~50% rhythmic) showcases the variability in this rhythm and how it is difficult to observe when compared to locomotor rhythms, which are usually consistently >90% rhythmic across multiple genotypes. These lines can be placed after the discussion about yw and perS flies. Moreover, previous studies using individual flies have reported that egg-laying rhythm is more variable than others Figure 1, Sheeba et al 2001.

We have accepted the suggestion, replacing "Thus, egg-laying displays a much larger variability than locomotor activity..." by "This shows that, at the individual level, egg-laying is much less robust than locomotor activity ..."

(t) Figure 1. Genotype notation within the figure panels is not consistent with the accepted / conventional notation or with the main text or legend notations throughout the manuscript.

We are sorry for this mistake. We have corrected the genotype names in Figures and text in order to make notation consistent across the paper.

(u) Supplementary Figure 1 Legend. Error in upper right corner? Not left corner? The photo does not clearly show the apparatus. The authors may wish to consider clearer images and more details about the apparatus including details of the 3D printing of the device and perhaps even include a short video where the motor moves the flies to a new chamber (This is only a suggestion to advertise the apparatus, not related to the review of the manuscript). They could also provide information about what fraction of females survived till the end of each trial when 21 flies were examined with 4-hour sampling across 4-5 cycles.

In general, more than 80% of the females are alive at the end of a one week oviposition experiment. We have added this information in the Methods section at the end of the corresponding subsection ("Automated egg collection device"). Regarding the eggcollection device, we have replaced the photographs in what is now Supplementary Figure 1S1, and a short supplementary movie showing its operation.

(v) The results depicted in Figure 2B are that of averaged time series. Hence the reader does not know 'the fact' that knocked-down animals are not completely rhythmic. Is the "not completely arrhythmic" in reference to flies with a power > 0.2 (weakly rhythmic) in their egg-laying rhythm or to the presence of ~40% of male flies (Supplementary Table 1) with a locomotor rhythm after perRNAi silencing of most of their clock neurons? This is confusing because no intermediate category of flies is discussed in Figure 2. Please edit for clarity.

We were referring to the rhythmicity of the genotype, not of the individuals. We have rewritten the corresponding paragraph in order to make it clearer (last paragraph of the first subsection of the Results section).

(w) Line 173 - ablation or electrically silencing all PDF+ neurons (Howlader et al., 2006). There were no experiments carried out using electrical silencing of PDF+ neurons in the referenced paper.

We are sorry for this mistake. This has been corrected (we have deleted the mention to electrical silencing).

(x) Line 173 - Shortening of period by nearly 3 hours cannot be considered minor.

We agree, and we have deleted the word "minor".

(y) Line 332-333 - "We also disrupted the molecular clock (or electrically silenced) in PDFexpressing neurons as well as in the DN1p group with no apparent effect on egg-laying rhythms". There was period shortening observed for pdf GAL4 > perRNAi manipulation so there was an effect on the egg-laying rhythm. Additionally, perRNAi based silencing does not electrically silence PDF neurons as the kir 2.1 was expressed only using Clk4.1 GAL4 in the Dn1ps. This line should be rewritten.

We have rewritten the paragraph mentioned (third paragraph of the Discussion) in order to make it more accurate.

(4) Page 22 - Data Analysis

Since the number of eggs laid by a mated female tend to show a downward trend, we proceeded as follows, in order to detrend the data (see the Supplementary Material for further details). First, a moving average of the data is performed, with a 6 point window, and a new time series T is obtained. In principle, T is a good approximation to the trend of the data. Then, a new, detrended, time series D is generated by pointwise dividing the two series (i.e. D(i)=E(i)/T(i), where i indexes the points of each series)." Can the authors provide a reference for this method of detrending? Smoothing can frequently introduce artifacts in the data and give incorrect period estimates. Additionally, the trend visible in the data, especially in Figure 1, suggests a linear decay that can be easily subtracted. Also, there is no discussion of detrending in the Supplementary material attached.

We are sorry for the confusion with the Supplementary materials. The method used for subtracting both noise and trend from the data is now fully explained in the new Supplementary Material. All the issues raised by the reviewer in this comment have been addressed there.

(5) Figure by figure

Page - Type (Figure or text) - Comment

(a) Page 6 Figure 1C There is remarkable phase coherence seen in the average egg laying time series for CS flies 5 days into DD and as the authors note in Lines 94-95 in the text "Under light-dark (LD) conditions, or in the first days of DD, it can be that the periodic signal is the same for all flies". Since this observation is crucial to constructing the figures seen later in the paper, a note should be made about why this rhythm could persist across flies, so deep into DD.

As mentioned above, we have added a couple of lines explaining why we think that the assumption of a synchronized periodic signal is reasonable, at least during the first cycles (second paragraph of the first subsection of section Results).

(b) Figure 1 G The effect of period/phase decoherence seems to be showing up here in the average profile for yw flies as they seem to completely dampen out after 2 days in DD and yet have a 24-hour rhythm in the averaged periodogram. The authors should make a note here if the LS periodogram is over-representing the periodicity of the first few days in DD or if comparing the first 3 vs. the last 3 days in DD gives different results.

The dampening observed in average oviposition records is a product of the dampening of the oviposition records, which is well known phenomenon, probably caused by the depletion of sperm in the female spermatheque. One of the aims of the method used in the paper was to avoid the bias introduced by this dampening, by means of a detrending procedure. This is explained in the Materials an Methods, and now full details are given in the new Supplementary Materials.

(c) Figure 1E, K Is this data pooled across 2-3 experiments, as discussed in lines 500-01 under 'Statistical Analysis'? Also, what test is being performed to check for differences between proportions here, seeing as there are no error bars to denote error around a mean value and no other viable tests mentioned in Statistical Analysis?

We are sorry for this omission. For the comparison of proportions we used the 'N-1' Chisquared test. We have added a sentence detailing this at the end of the Statistical analysis section.

(d) Figure 1 F, L Can the total number of weakly and strongly rhythmic values be indicated in the scatter plot?

Corrected.

(e) Figure 1F, L (legend) Is the Chi-squared test being performed on the proportion values of Figure 1(E, K) or for Figure 1(F, L)?"

The chi-squared test mentioned was used for Fig1 F-L. As explained above, for the comparison of proportions we used 'N-1' Chi-squared test. This has now been added to the legend of the figure

(f) Page 8 Figure 2B Seeing as individual flies with a LS periodogram power < 0.2 are considered weakly rhythmic in Figure 1 F, L can Clk856 > perRNAi flies on average also be considered weakly rhythmic, as the peak in the periodogram is above 0.3?

We prefer to use the weakly rhythmic class only for individual flies. Nevertheless, we agree that this periodogram shows that the genotype analyzed is not completely arrhythmic, and that this might be due to some remaining individual rhythmicity. As mentioned above, we have rewritten the last paragraph of the first subsection of section Results in order to discuss this.

(g) Figure 2D Can the authors comment on why there is a shorter period rhythm when PDF neurons have a dysfunctional clock, whereas previous evidence (Howlader et al., 2004) suggested that these neurons play no role in egg-laying rhythm? They should also refer to McCabe and Birley, 1998 to see if their results (where they observed a shorter period of ~19h with groups of per0 flies), might be of interest in their interpretations.

We have added a line commenting this in the corresponding subsection ("LNv and DN1 neurons are not necessary for egg-laying rhythmicity") of the Results, as well as a discussion of this in the third paragraph of the Discussion. In a nutshell, even though Howlader et al did not find a shortening when PDF neurons are ablated, they did find it in pdf01 flies.

(h) Figure 2 F, H As the authors mention in their Discussion on Page 16, lines 340-45, the manipulation of DN1p neurons might abolish the circadian rhythm in oogenesis as reported by Zhang et al, which is why they looked at this circuit driven by Clk4.1 neurons and comment that "The persistence of the rhythm of oviposition implies that it is not based on the availability of eggs but is instead an intrinsic property of the motor program". However, no change in fecundity is reported for either kir2.1 or perRNAi-based manipulations of these neurons, to help the reader understand if egg availability (at the level of egg formation) is playing any role in the downstream (and seemingly independent) act of egg laying. The authors should report if they see any change in total fecundity for either set of flies w.r.t their respective controls. Also, is the reduction in power seen with electrical silencing vs perRNAi expression of any relevance? Does the percentage of rhythmic flies change between these two manipulations?

In the line mentioned by the reviewer what we meant is that our results show that the rhythm of oviposition does not seem to be based in the rhythmic production of oocytes, which is not necessarily connected with the total number of eggs produced. We have modified the corresponding line in the paper, in order to avoid this misunderstanding. Regarding the "reduction in power" mentioned, it must be stressed that, in general, the height of the peak is correlated with the fraction of rhythmic individuals. The problem is that this fraction is a much more noisy output, and that is the reason why we have chosen to work with periodograms of averages.

(i) Figure 2 E and G, a loss of rhythmicity could also be due to a decrease in fecundity in the experimental lines. Since the number of eggs laid for each genotype is already known, can the authors show statistically relevant comparisons between the experimental lines and their respective controls? In this vein, can the averaged time series profiles also be provided for all the genotypes tested (as seen previously in Figure 1 A, C, G, I), perhaps in the supplementary?

We did not focus on fecundity in the present work. However, our observations do not seem to show any definite relationship with rhythmicity. We plan to address the issue of fecundity more systematically in a future work. The averaged time series profiles have now been added to the figure.

(j) Scatter plots showing the average period and SEM as seen in Figure 1 (F, L) would help in understanding if these manipulations have any effect on variation in the period of the egg-laying rhythm across flies. Particularly for pdf GAL4 > perRNAi flies which have a net shorter period, (but this might vary across the 34 flies tested).

We have added a Supplementary Figure (2S1) that shows that the shortening of oviposition period can be also observed at the individual level. We have also added a line commenting this in the corresponding subsection ("LNv and DN1 neurons are not necessary for egg-laying rhythmicity") of the Results, as well as a discussion of this in the third paragraph of the Discussion.

(k) Page 11 Figure 3B Does the presence of two peaks in the LS periodogram at a power > 0.2 indicate the presence of weakly rhythmic flies with both a short(20h) and a long(~27h) period component or either one? The short-period peak is nearly at p < 0.05 level of significance. So then, do most of the flies in MB122B GAL4 > perRNAi line show a weakly rhythmic shorter period?

(l) Figure 3D A similar peak is observed again at 20h (LS power > 0.2 and nearly at p < 0.05 significance level again) and a different longer one at (~30h) though this one is almost near 0.2 on the power scale. Given the consistency of this feature in both LNd manipulations, the authors should comment on whether this is driven by variation in periods detected or the presence of complex rhythms (splitting or change in period) in the oviposition time series for these lines.

(m) Figure 3 General scatter plots showing average period {plus minus} SEM could help explain the bimodality seen in the periodograms. Additionally indicating just how many flies are weakly rhythmic vs. strongly rhythmic can also help to illustrate how important the CRY+ LnDs are to the oviposition rhythm's stability.

For these three comments (k, l and m), we note that the issue of bimodality has been addressed above, in our response to Weakness 9.

(o) Figure 4B Same as comments under Figure 1, what is the statistical test done to compare the proportions for these three genotypes?

As mentioned above, for the comparison of proportions we used the 'N-1' Chi-squared test. We have added a sentence detailing this at the end of the Statistical analysis section.

(p) Figure 4C Are all flies significantly rhythmic? The authors should also provide an averaged LS periodogram measure for each genotype, to help illustrate the difference in power between activity-rest and egg-laying rhythms.

Yes, the points represent periods of (significantly) rhythmic flies. This has been added to the caption, to avoid misunderstandings. The differences that arise when assessing rhythmicity in activity records vs. egg-laying records is addressed at length in the Supplementary Material (see e.g. Fig S1).

(q) Page 15 Figure 5 - general As the authors discuss the possible contribution of DN1ps to evening activity and control over oogenesis rhythm, investigating the connections of the few that are characterized in the connectome (or lack thereof) with the Oviposition neurons, can help illustrate the distinct role they play in the female Drosophila's reproductive rhythm.

This information was in the text and the Supplementary Tables. Lines 273-275 of the old manuscript read: "The full results are displayed in Supplementary Tables 2 and Table 3, but in short, we found that whereas there are no connections between LNv or DN1 neurons and oviposition neurons..."

(r) Minor: The dark shading of the circles depicting some of the clusters makes it difficult to read. Consider changing the colors or moving the names outside the circles.

Figure corrected.

(s) Line 38: The estimated number of clock neurons has been revised recently (https://www.biorxiv.org/content/10.1101/2023.09.11.557222v2.article-info).

Thank you for the reference. We have corrected the number of clock neurons in the Introduction of the new manuscript.

Author response:

Public Reviews: 

Reviewer #1 (Public review): 

Summary: 

In this study, Li et al. used genetically engineered murine intestinal organoids to investigate how the temporal order of oncogenic mutations influences cell state and tumourigenicity of colorectal epithelial cells. By sequentially introducing Apc and Trp53 loss-of-function mutations in alternate orders within a Kras^G12D background, the authors generated isogenic organoid lines for both in vitro and in vivo characterisation. Bulk RNA-seq reveals expected transcriptional changes with relatively modest differences between the two triple-mutant configurations (KAT vs KTA). The key finding emerges from transplantation assays: while KAT and KTA organoids show equivalent tumourigenic potential in immunodeficient mice, only KAT organoids form tumours in immunocompetent hosts (5/10 vs 0/10), suggesting that mutation order shapes susceptibility to immune-mediated clearance. The experiments are well-executed, and the conclusions are generally supported by the data. 

Strengths: 

The experimental system is well-designed for the question. By combining a Kras^G12D transgenic background with sequential CRISPR-mediated knockout of Apc and Trp53 in alternate orders, the authors generated truly isogenic organoid lines that differ only in mutational sequence. This is technically non-trivial and provides a clean platform for dissecting order effects, a question otherwise difficult to address experimentally. 

The authors performed comprehensive baseline characterisation of these organoids, including morphological and histological assessment, quantification of organoid-forming efficiency and proliferation, and bulk RNA-seq profiling. While these analyses revealed no major differences between KAT and KTA organoids, and the observed enhancement of epithelial stemness upon Apc loss and proliferative advantage conferred by Trp53 loss are largely expected, the systematic nature of this characterisation establishes a useful methodological template for future organoid-based studies. 

The authors further investigated the functional impact of mutational order using subcutaneous transplantation assays. By comparing tumour formation in immunodeficient versus immunocompetent hosts, the authors uncover a genuinely unexpected finding: KAT and KTA organoids behave equivalently in the absence of adaptive immunity, but diverge dramatically when immune pressure is applied (KAT: 5/10; KTA: 0/10). This observation is arguably the most compelling aspect of the study and opens an interesting line of inquiry. 

We greatly appreciate your positive comments on our study.

Weaknesses: 

The authors acknowledge that initiating with Kras^G12D does not reflect the typical human sporadic CRC trajectory, where APC loss is usually the first event. While this design choice was pragmatic, it means the observed order effects are contextualised within an artificial starting point. It remains unclear whether the Apc/Trp53 order would matter in a Kras-wild-type background, or whether the Kras-driven cellular state is a prerequisite for these phenotypes to emerge. 

We agree with the reviewer that initiating tumorigenesis with Kras^G12D does not fully recapitulate the most common trajectory of sporadic human CRC, where APC loss typically occurs first. We had noted this point in the original Discussion and will further clarify it more explicitly in the Introduction part of the revised manuscript.

Our experimental design was intended to establish a controlled and genetically tractable system to interrogate the principle of mutation order effects. In this context, Kras^G12D activation provides a stable oncogenic baseline that facilitates sequential genome engineering and comparison of isogenic lines.

Although APC loss is frequently the initiation event, a recent study has suggested that Kras^G12D priming can reshape the selective landscape for subsequent driver events, including Apc alterations (PMID: 41339549). Consistent with this notion, our data indicate that Kras^G12D activation induces a permissive oncogenic cellular state that may influence the phenotypic consequences of later mutations. We therefore speculate that the Kras^G12D-primed context may contribute to the observed order-dependent effects.

We agree that testing Apc/Trp53 order in a Kras-wild-type background would be an important future direction, and we will point this out explicitly in the revised Discussion.

Subcutaneous implantation provides a tractable readout of tumourigenicity, but the cutaneous immune microenvironment differs substantially from that of the intestinal mucosa. Given that the central claim concerns immune-mediated selection, orthotopic transplantation would more directly test whether the observed order effects hold in a physiologically relevant context. 

In the present study, we employed subcutaneous transplantation, which is a widely used platform to assess tumorigenic potential under controlled immune conditions. This approach offers high reproducibility, straightforward tumor monitoring, and has been broadly applied in organoid-based cancer studies in both immunodeficient (PMID: 23273993, 23776211, 32209571, 33055221) and immunocompetent (PMID: 32209571, 33055221, 41672595) settings.

Importantly, our primary goal was to determine whether mutation order influences susceptibility to immune-mediated clearance, rather than to model the full complexity of the intestinal niche. The clear divergence between KAT and KTA specifically in immunocompetent hosts supports the existence of intrinsic mutation order-dependent immune vulnerability.

Nevertheless, we fully agree with the reviewer that orthotopic transplantation would provide a more physiologically relevant immune microenvironment and represents also an important direction for future investigation. We will explicitly discuss this limitation and highlight orthotopic validation as an important future direction in the revised Discussion.

The ssGSEA comparison involves only 14 ATK tumours, and the key comparisons (Figure 6E) yield borderline significance (p=0.052). More fundamentally, since mutation order cannot be inferred from the clinical samples, the authors are correlating organoid-derived IFN signatures with tumour immunophenotypes without direct evidence that these patients' tumours followed a KAT-like trajectory. The reasoning becomes circular: KAT organoids define the signature used to identify KAT-like clinical tumours. 

We thank the reviewer for raising this important point. We would like to clarify that our intention was not to infer the actual mutation order in clinical samples, which indeed cannot be reliably reconstructed from bulk tumor RNA-seq data.

Instead, our goal was to determine whether the transcriptional programs distinguishing KAT and KTA organoids could be observed in human CRC cohorts. In this context, the organoid-derived IFN-related signature was used as a molecular reference to assess potential clinical relevance, rather than to classify tumors by evolutionary trajectory.

We agree that the statistical significance in Figure 6E is modest (p = 0.052), and we would like to revise the text to present this analysis more cautiously as a suggestive trend rather than definitive evidence. We will also clarify this limitation explicitly in the revised manuscript to avoid overinterpretation.

Furthermore, the most striking finding of the study, that KTA organoids fail to form tumours in immunocompetent hosts while KAT organoids can, lacks a mechanistic follow-up. The transcriptomic differences between KAT and KTA are modest when cultured as monocultures, yet their in vivo fates diverge dramatically. The authors do not address why these subtle intrinsic differences translate into such divergent immune susceptibility, nor do they characterise the immune response adequately (beyond limited CD4/CD8 IHC at tumour peripheries). 

We thank the reviewer for this important point. We agree that the mechanistic basis underlying the differential immune susceptibility between KAT and KTA remains incompletely resolved.

A practical limitation of the current study is that KTA grafts failed to establish tumors in immunocompetent hosts, which precluded downstream histological and immune profiling of established lesions. As a result, our in vivo immune characterization of KTA grafts is nearly impossible.

Nevertheless, our transcriptomic analyses indicate that KAT and KTA organoids differ in interferon-response and immune-related programs prior to transplantation, and those differentially expressed genes were consistently preserved in tumor cells derived from immunodeficient hosts. These results suggest the presence of intrinsic tumor-cell-autonomous differences may influence immune recognition and clearance.

We will expand the Discussion to outline several non-mutually exclusive mechanisms that could account for this phenotype, including altered interferon responsiveness, differential antigen presentation capacity, and changes in tumor cell-intrinsic immune visibility programs. These hypotheses are consistent with the transcriptional differences observed prior to transplantation and provide a framework for future mechanistic investigation. We agree that deeper immune profiling (e.g., immune infiltrate composition, antigen presentation status, and functional immune assays) will be important to fully elucidate the mechanism and represents a key direction for future work.

Reviewer #2 (Public review): 

Summary: 

This study addresses an important and timely question in colorectal cancer biology by systematically examining the effects of the common driver mutations APC, KRAS G12D, and TP53 in murine colorectal organoids, with particular emphasis on how the order of APC and TP53 acquisition influences tumor phenotype. These mutations are well known to be frequent, truncal, and often co-occurring in colorectal cancer. While it is increasingly appreciated that mutational order can shape tumor behavior, studies directly comparing the phenotypic consequences of alternative APC-TP53 mutation orders remain rare. This work, therefore, addresses a relevant and timely question. 

Strengths: 

A major strength of the study is its focus on previously unexplored biology, combined with the generation of multiple isogenic murine organoid models with controlled mutational sequences. The authors employ careful and robust quality control of the CRISPR-mediated alterations, and the inclusion of both in vitro and in vivo experiments strengthens the relevance of the work.

We greatly appreciate your positive comments on our study.

Weaknesses: 

There are, however, several limitations that should be considered when interpreting the findings. First, KRAS G12D activation is used as the initiating alteration, whereas APC loss is generally believed to be the initiating event in most human colorectal cancers.

We sincerely thank the reviewer for their insightful comments regarding the initiation of tumorigenesis with a Kras mutation rather than the more canonical Apc loss, which was also raised by the reviewer #1. We fully agree that the Apc-first represents the most prevalent sequence in human colorectal cancer (CRC), We will more clearly explain the rationale for our experimental design in the revised Introduction part as outlined in our response to reviewer #1.

Second, the analysis is restricted to comparing only two mutation orders (KAT versus KTA), which limits the breadth of conclusions that can be drawn about mutation ordering more generally.

We thank the reviewer for pointing this limitation out. However, as a proof-of-concept, study of Apc and Trp53 loss, two major oncogenic events in CRC, serves as a biologically meaningful starting point for dissecting order-dependent effects. Although it is of great significance to compare all six possible mutation orders of these three driver genes, generating and thoroughly characterizing all genotypes represents a substantial undertaking beyond the scope of this initial study.

Finally, key RNA-sequencing and in vivo experiments rely on a single isogenic line, which substantially constrains interpretability. 

The aim of the study was to systematically investigate how mutation accumulation and order influence colorectal cancer initiation. While the data suggest that the relative timing of APC and TP53 loss may be particularly important for tumor initiation, the absence of biological replication makes it difficult to draw robust conclusions. Engraftment efficiency and tumor behavior can be influenced by many factors for a single clone, including additional passenger mutations acquired during culturing, as well as epigenetic differences that are independent of the engineered mutations.

We thank the reviewer for raising his/her concern. We apologize that we have not made a clear presentation of our data source. Indeed, for all major in vitro and in vivo assays of double and triple mutants (KA/KT/KAT/KTA), we analyzed at least two independently derived clones per genotype. These independent clones harbor distinct mutations in target genes and were treated as biological replicates throughout the study.

To improve clarity and transparency, we will revise the relevant figures and figure legends to explicitly indicate the clonal origin of each data point.

Author response:

The following is the authors’ response to the original reviews

We would like to sincerely thank the editor and reviewers for their thoughtful and constructive feedback on our manuscript. We are grateful not only for the close reading and insightful suggestions, but also for the open and generous way in which the reviewers engaged with our work. In revising the manuscript, we have clarified how our contribution is situated within the existing literature, conducted additional analyses to examine individual differences in exploration strategies, expanded and refined our description of the DDM analyses, and added correlations between strategies and other behavioral measures. We have also clarified methodological points, such as the estimation of thresholds, and provided new supplementary figures and analyses where appropriate. In several places, we have modified and qualified our interpretations in line with the reviewers’ comments. We believe these changes have significantly strengthened the manuscript, and we are grateful for the scientific dialogue with the reviewers.

Review 1 (Public review):

This manuscript reports on the behavior of participants playing a game to measure exploration. Specifically, participants completed a task with blocks of exploratory choices (choosing between two 'tables', and within each table, two 'card decks', each of which had a specific probability of showing cards with one color versus another) and test choices, where participants were asked to choose which of the two decks per table had a higher likelihood of one color. Blocks differed on how long (how many trials) the exploration phase lasted. Participants' choices were fit to increasingly complex models of next-trial exploration. Participants' choices were best fit by an intermediate model where the difference in uncertainty between tables influenced the choice. Next, the authors investigated factors affecting whether participants sought out or avoided uncertainty, their choice reaction times, and the relationship of these measures with performance during the test phase of each block. Participants were uncertainty-seeking (exploratory) under most levels of overall uncertainty but became less uncertainty-seeking at high levels of total uncertainty. Participants with a stronger tendency to approach uncertainty at lower levels of total uncertainty were more accurate in the test phase, while the tendency to avoid uncertainty when total uncertainty was high was also weakly positively related to test accuracy. In terms of reaction times, participants whose reaction times were more related to the level of uncertainty, and who deliberated longer, performed better. The individual tendency to repeat choices was related to avoidance of uncertainty under high total uncertainty and better test performance. Lastly, choices made after a longer lag were less affected by these measures.

The authors note that their paradigm, which does not provide immediate rewarding feedback, is novel. However, the resulting behavior appears similar to other exploratory learning tasks, so it's unclear what this task design adds - besides perhaps showing that exploratory behavior is similar across types of reward environments. Several papers have shown that cognitive constraints modulate exploration (PMIDs: 30667262, 24664860, 35917612, 35260717); although this paper provides novel insights, it does not situate its findings in the context of this prior literature. As a result, what it adds to the literature is difficult to discern.

We are grateful for your thoughtful reading of our paper and for pointing us to these relevant references. We appreciate the need to clarify how our work is situated within the existing literature. In brief, the novelty of our paper lies in measuring exploration in contexts where it is not in direct competition with the need to exploit knowledge for reward. This approach enables us to include orders of magnitude more exploration trials. With this increased power, we were able— for the first time— to distinguish between competing algorithms for addressing uncertainty, and we identified a novel tendency to avoid uncertainty when overall uncertainty is high. We now state this more clearly in the discussion section and cite the suggested papers.

“While the literature on exploration is expansive, the paradigm presented here extends it in important ways. Researchers of reinforcement learning have previously examined exploration in the context of reward-seeking decisions. Using such paradigms as the bandit task Schulz and Gershman (2019), it was demonstrated that humans don't always choose the option they believe will yield the most reward, but also make random and directed choices with the aim of exploring other uncertain options (Schulz and Gershman, 2019; Wilson et al., 2014). Recently, studies using the bandit task have lent empirical support to the notion that exploration is difficult, as participants explore less under time pressure or cognitive load (Brown et al., 2022; Otto et al., 2014; Cogliati Dezza et al., 2019; Wu et al., 2022). Crucially, this literature has focused on cases where reward can be gained on each trial (Brown et al., 2022; Cohen et al., 2007; Daw et al., 2006; Schulz and Gershman, 2019; Song et al., 2019; Tversky and Edwards, 1966; Wilson et al., 2014; Wu et al., 2022). In such tasks, the motivation to exploit current knowledge predominates exploration, rendering it rare and difficult to measure (Findling et al., 2019). In contrast, our task was designed to remove the impetus to immediately exploit current knowledge , and as a result we were able to observe many exploratory choices. With this increased experimental power, we were able to compare different algorithms approximating the goal of approaching uncertainty, and describe how and when humans avoid uncertainty instead of approaching it.”

Reviewer #1 (Recommendations For The Authors):

Are all participants best fit by the delta uncertainty model? Since other parts of the paper focus on individual differences, it would be useful to examine if people differ in the computational complexity of their exploration strategies and if this difference relates to other behavior.

We thank you for this helpful suggestion, which prompted us to conduct additional analyses. To address your question, we summarized point-wise predictive accuracy for each participant and compared it across the three models. The results are presented in the new Supplements 2 and 3 to Figure 6.

These analyses show that, for the vast majority of participants, uncertainty was favored over exposure as a choice strategy, and for a sizable majority, it was also favored over EIG. As detailed in Figure 6 and its supplements, 125 participants were best described by uncertainty relative to EIG, 58 by EIG, and 11 showed inconclusive results. Similarly, 96 participants were better fit by uncertainty than exposure, while an additional 72 had negative exposure coefficients (consistent with uncertainty-based choice). Exposure was supported for 26 participants.

We also examined how these strategies relate to other behavioral measures. Exposure was not strongly linked to test performance. EIG, by contrast, showed a positive association with test performance, perhaps because it is more closely correlated with uncertainty. Importantly, however, across posterior predictive checks in the main text and supplements, approaching uncertainty continues to provide the best overall description of participants’ strategies.

The authors construct a hierarchy of exploratory strategies. Perseveration/switching is also an explore/exploit strategy that would lie above random exploration in the authors' hierarchy.

We chose not to place perseveration within the hierarchy, as from a normative perspective it is not, strictly speaking, an exploration strategy. At its extreme, perseveration would lead a participant to repeatedly sample only one option, leaving the others entirely unexplored. Switching is represented in the hierachy by the equating exposure strategy – they are very similar.

For the analyses examining uncertainty seeking vs. aversion by total uncertainty, how was the cut point determined? Did this differ across people?

Thank you for highlighting the need for greater clarity on this point. The threshold was indeed fitted to the data and varied significantly across participants (see Table 6 in Appendix 3). For each participant, the threshold marks the point at which behavior shifts from approaching to avoiding uncertainty. This threshold is a key factor underlying individual differences in the tendency to avoid uncertainty when overall uncertainty is high, as illustrated in the analyses of Figure 6 and related results. We now make this point clearer in the methods section:

“To quantify how the influence of Δ-uncertainty on choice varied with overall uncertainty, we fit a multilevel piecewise logistic regression model. This model estimated a threshold in overall uncertainty, treated as a free parameter, and allowed the slope of Δ-uncertainty on choice to differ below and above this threshold. Below the threshold, a positive slope reflects a tendency to approach uncertainty; above the threshold, a negative interaction captures the tendency to avoid Δ-uncertainty with higher values of overall uncertainty.”

More details on the DDM analyses are needed - it's not clear how the outputs of the DDM correspond to what is stated in the text in the results.

We agree that the section detailing the DDM analyses could be clarified. We analyzed two key parameters of the DDM: the drift rate, which we interpret as reflecting the efficacy of deliberation over uncertainty, and the bound separation, which corresponds to the tendency to deliberate rather than respond quickly. Our results show that good learners exhibit both higher drift rates and higher bounds. When participants repeat a previous choice, both the drift rate and bounds are lower. We changed the way we report the results:

“We found that RTs indeed varied in relation to the absolute value of Δ-uncertainty as expected b=0.69, 95\% PI=[0.58,0.78]. Crucially, a stronger dependence of RT on the absolute value of Δ-uncertainty predicted better performance at test (drift-rate and test performance association b=0.81, 95% PI=[0.58,1.07]). We further found that participants who tended to deliberate longer for the sake of accuracy also tended to perform better at test (bound height and test perfromance association b=1.46, 95% PI=[0.58,2.34]; Figure8c). In summary, participants who were better at deliberating about uncertainty during exploration, and who deliberated for longer, performed better at test. Thus, making good exploratory choices that lead to efficient learning involves prolonged deliberation.”

We also provide a detailed explanation of this correspondence in the Methods section:

“The DDM explains RTs as the culmination of three interpretable terms. The first is the efficacy of a participant’s thought process in furnishing relevant evidence for the decision - in our case the efficacy of choosing according to Δ-uncertainty (the drift rate in DDM parlance). The second term governs the participant’s speed-accuracy tradeoff by determining how much evidence they require to commit to a decision. This can also be thought of as how long a participant is willing to deliberate when a decision is difficult (bound height). Finally, the portion of the RT not linked to the deliberation process is captured by a third term (non-decision time).”

The authors note that "the three choice strategies prescribe different table choices on most trials" but (from what I can see) only provide a representative participant's plot in Figure 2. What was the overall correlation of predicted choices from the three models?

Thank you for pointing out this oversight. The correlations are now shown in the supplement to Figure 2. In brief, correlations between exposure and the other two strategies are low, while the correlation between EIG and uncertainty is moderate. These dependencies motivated our decision to fit a separate logistic regression model for each strategy and to compare strategies using formal model comparison and posterior predictive checks, rather than including them all in a single regression model.

It appears that the models are all constructed to predict table choices and not card deck choices. Can the authors clarify this? If so, what role do the card deck choices have?

Indeed, the manuscript focuses on table choices, as these are the choices of primary interest from an exploration perspective. It is most straightforward to define the three exploration strategies with respect to table choices, whereas for deck choices it is not clear how to define EIG in respect to the perforamnce at test. The hierarchical structure of the task was originally chosen to increase complexity, with the goal of creating a rich task that engages cognitive resources. We have not formally tested this assumption, and do not expect that the patterns we observe should be absent in a flat version of the task.

Reviewer 2 (Public review):

Summary:

This paper focuses on an interesting question that has puzzled psychologists for decades, that is, why do people demonstrate a mix of uncertainty approach and avoidance behavior, given the fact that reducing uncertainty could always gain information and seems beneficial? This paper designed a novel task to demonstrate behavioral signatures of uncertainty approaching and avoidance during the exploration phase within the same task at both a within-subject and betweensubject level. On the algorithmic level, this paper compared four different implementations of uncertainty-guided exploration and found that the model sensitive to relative uncertainty provides the best fit for human behavior compared to its counterparts using expected information gain or past exposure. This paper then links people's uncertainty attitude with accuracy and finds that uncertainty avoidance during exploration does not impair task performance, implying that uncertainty avoidance may be the output of a resource-rational decision-making process. To examine this account, this paper uses reaction time as an independent proxy of costly deliberation and shows that people deliberate shorter when engaging in repetitive choice, which presumably saves cognitive resources. Finally, the paper shows that people's tendency to engage in repetitive choice correlates with their tendency to avoid uncertainty, which supports the argument that avoiding uncertainty could be a strategy developed under the constraint of limited cognitive resources.

Strengths:

One of the highlights of this paper, as mentioned in the previous paragraph, is that the authors can establish the existence of the uncertainty approach and avoidance behavior within the same task whereas previous work usually focuses on one of them. This dissociation allows the authors to examine what situational factor is related to the emergence of the act of avoiding uncertainty, and extract parameters describing participants' attitude towards uncertainty during baseline as well as during situations where uncertainty avoidance is more common. Besides documenting the existence of uncertainty avoidance behavior, this paper also tried to explain this behavior by proposing under the resource rational framework and has carefully quantified different aspects (e.g., accuracy; choice speed) of participants' behavior as well as examined their relationships. Though more experiments are needed to fully understand human uncertainty avoidance behavior, this paper has provided both empirical and theoretical contributions toward a mechanistic understanding of how people balance approaching and avoiding uncertainty.

Weaknesses:

I have a couple of concerns related to this paper. First, there seems to exist an anticorrelation between total uncertainty and absolute relative uncertainty (Figure 5 panel C, \delta uncertainty is restricted to a small range when total uncertainty is high). It seems to be a natural product of the exploration process since the high total uncertainty phase is usually the period where the participant knows little about either option, leading to a less distinguishable relative uncertainty. However, it remains unknown whether the documented uncertainty avoidance still applies when extrapolating to larger absolute relative uncertainty.

We sincerely thank you for your close reading of our manuscript and for highlighting its strengths. In the paradigm we study, overall and relative uncertainty are not anticorrelated. While the two are related—as in any finite-information exploration task, where the value of overall uncertainty constrains the possible range of relative uncertainty—they are not correlated and can therefore be used as predictors in a single regression model. We agree that strategies could differ substantially in a (near) infinite-information setting, such as when people seek semantic knowledge. The advantage of a finite-information task is its tractability, which enables the computational analyses we conducted. That said, the inherently greater intractability of an infinite-information task would likely alter human strategies, as it poses challenges both to participants and to researchers.

It would be great if the experiment allows for a manipulation of uncertainty in the middle of the experiment (e.g., introducing a new deck/informing that one deck has been updated)

We agree, and look forward to probing this question in the future. We’ve added the point to our discussion section:

“Our theoretical analysis and experiments leave several open questions. One concerns the relationship between overall uncertainty and time on task: in our paradigm, overall uncertainty was correlated with the number of cards observed. Although our findings remain robust when trial number is included as a covariate in the regression models, future work could more directly disentangle these factors by orthogonalizing overall uncertainty and elapsed time. This might be achieved, for instance, by manipulating overall uncertainty within a game—such as by introducing new tables or altering outcome probabilities mid-round.”

Relatedly, the current 'threshold' of uncertainty avoidance behavior, if I understand correctly, is found by empirically fitting participants' data. This brings the question: can we predict when people will demonstrate uncertainty avoidance behavior before collecting any data? Or, is it possible that by measuring some metrics related to cognitive cost sensitivity, we could predict the proportion of choices that participants will show uncertainty-avoidant behavior?

Thank you again for probing our thinking further. The threshold of uncertainty is indeed fitted on an individual basis using a hierarchical model. We believe there should be ways to predict it. In the current data, we find that it is correlated with the baseline tendency to approach uncertainty: in other words, participants who perform better show a slightly stronger tendency to avoid uncertainty when overall uncertainty is high. This underscores the complexity of identifying correlates of a coping strategy, as it is intricately linked to the difficulty being coped with. We speculate that working memory capacity may play an important role in this strategy, as well as the interplay between working memory–based learning and slower incremental learning mechanisms. Beyond speculation, however, we currently have no data to test these ideas.

Finally, regarding the analysis of different behavior patterns in the game, it seems that the authors try to link repetitive behavior, uncertainty attitude, and accuracy together by testing the correlation between the two of them. I wonder whether other multivariate statistical methods e.g., mediation analysis, will be better suited for this purpose.

This was a very insightful comment. We revisited the data and fitted test performance using a multiple regression model, predicting performance from the three exploration-phase strategies simultaneously: baseline tendency to approach uncertainty, tendency to avoid uncertainty when overall uncertainty is high, and tendency to repeat previous choices. When adjusting for the baseline tendency to approach, we find that the tendency to avoid uncertainty is indeed associated with a slight decrement in test performance. However, in our sample, the better learners—who are more effective at approaching uncertainty—also tend to avoid it when overall uncertainty is high. This nuance highlights the point discussed earlier. We find similar results when fitting the data with a mediation model, but we favour the multiple regression approach, since have no strong convictions about which exploration strategy causes another. We have detailed this analysis in the main text and have accordingly modified and qualified our interpretation of this finding:

“In contrast, the relationship between the tendency to avoid uncertainty and test performance was more nuanced. In both samples, participants who were more inclined to approach uncertainty also tended to avoid it when overall uncertainty was high r=0.43, p=5.42 x 10^-10. Accordingly, avoidance was positively correlated with test performance at the population level b=1.18, 95% PI=[0.80, 1.58] Figure 7b; see Methods for parameter estimation). However, once we adjusted for the tendency to approach, avoidance was reliably associated with worse test performance b=-0.83, 95% PI=[-1.28,-0.40].”

Reviewer #2 (Recommendations For The Authors):

Could the authors elaborate more on why the negative relationship between exposure and choice (Figure 4a) is a natural phenomenon under the relative uncertainty model?

Indeed, we believe this is a natural phenomenon under the uncertainty model. When simulating an uncertainty-driven agent, the negative relationship arises naturally. We interpret this as the agent repeatedly pursuing tables that are more difficult to learn—those with smaller probability differences. The agent is drawn to these tables precisely because they are harder to master. By contrast, an EIG-driven agent would not repeatedly return to tables that are too difficult to learn. We have revised the Results section to make this point clearer:

“The simulations demonstrate that the surprising negative correlation between choice and Δ-exposure is an epiphenomenon of uncertainty-driven exploration: agents repeatedly return to harder-to-learn tables, gaining more exposure to them precisely because they remain more uncertain about these tables.”

It would be great if the authors could provide the correlation between different uncertainty estimates to help the readers have a better sense of how different these estimates are.

We’ve added this information in the supplement to Figure 2. In brief, correlations between exposure and the other two strategies are low, while the correlation between EIG and uncertainty is moderate. These dependencies motivated our decision to fit a separate logistic regression model for each strategy and to compare strategies using formal model comparison and posterior predictive checks, rather than including them all in a single regression model.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

Pierre Despas et al. studied the role of Salmonella typhimurium LppB in outer membrane tethering. Using E. coli ∆lpp mutant the authors showed that Salmonella LppB is covalently attached to PG throug K58 and that these crosslinks are formed by the L,Dtranspeptidase LdtB, primarily. Additionally, authors demonstrate that LppB forms homodimers via a disulfide bond through C57, but when Lpp is present it can also form heterotrimers with it. Thus, suggesting a regulatory role in Lpp-PG crosslinking.

Strengths:

In my view, this is a nice piece of work that expands our understanding of the role of lpp homologs. The experiments were well-designed and executed, the manuscript is wellwritten and the figures are well-presented.

Weaknesses:

I have some suggestions to give a clearer message, because I think a few images don't reflect much of what the authors wrote.

We thank Reviewer #1 for this important comment. We agree that several figures could more directly illustrate the points made in the text. In a revised version, we intend to revise the relevant figure panels and legends to better align the visual message with the conclusions, and we will adjust the corresponding text to explicitly state what each figure demonstrates and how the data support our interpretation. We anticipate that these changes will improve clarity and strengthen the alignment between figures and text.

It'd be helpful for readers to see the phylogenetic tree of the rest of the organisms that harbor LppB homologs and Lpp.

We thank Reviewer #1 for this suggestion. We examined the distribution of Lpp-family proteins across closely related Enterobacteriaceae. While species such as Escherichia fergusonii, Shigella flexneri and Shigella dysenteriae encode Lpp and as well as a paralogous small lipoprotein (YqhH, see Fig.S7), we find that LppB-like orthologs (equivalent to lppB from Salmonella) appear to be restricted to Salmonella species to our knowledge. Because LppB shows this lineage-specific distribution, inclusion of a broader phylogenetic tree would primarily highlight its restricted presence rather that provide additional evolutionary insight. We will clarify this point in the revised manuscript.

Increased expression of LppB under low pH is subtle. This result would benefit from quantifying the blots (Fig. S1) and performing statistical analysis.

We thank Reviewer #1 for this observation. We agree that the increase in LppB levels at acidic pH appears modest. We will carefully reassess this result across independent experiments and, where technically appropriate, provide quantitative information to better document the magnitude of the effect. Additionally, we will revise the text to more accurately described the observed difference.

Similarly, the SDS-EDTA sensitivity result (Fig. S2) is not convincing; the image doesn't seem to show isolated colonies at low pH (Fig. S2B). Please measure CFU/mL and report endpoint growth graphs instead. Statistical analysis should also be presented.

We thank Reviewer #1 for this suggestion. We agree that the SDS-EDTA sensitivity assay presented in Fig. S2 could benefit from a more quantitative assessment. We will perform CFU/mL measurements from independent biological replicates to better quantify the observed differences and include statistical analysis when appropriate. In addition, we will revise the corresponding text to more accurately reflect the magnitude of the phenotype.

The reduction to PG crosslinking of the C57R mutant is unclear (Fig 4B lane 22). The authors state: "suggesting that additional features of the LppB C-terminal region underlie its reduced efficiency." Does this mean additional amino acids play a role? Did the authors try to substitute Cys with other amino acid residues like Ala or Ser and quantify protein levels to find a mutant with similar expression levels? Do these have less crosslinking too?

We thank Reviewer #1 for this important comment. As correctly noted, the reduced abundance of the LppB_C57R variant likely contributes to its reduced level of peptidoglycancrosslinked species. Therefore, we cannot formally distinguish whether the reduced peptidoglycan crosslinking reflects decreased intrinsic crosslinking efficiency or simply reduced protein abundance and stability. We will revise the text to clarify this point and explicitly acknowledge this limitation. The C57R substitution was chosen because arginine is present at the equivalent position in the Salmonella LppA homolog, allowing us to assess the functional consequences of a naturally occurring sequence variation between Lpp-family members. While substitutions such as C57A or C57S could further dissect the specific contribution of the cysteine residue, our use of the C57R substitution provides direct insight into the functional implications of this naturally occurring difference between Lpp homologs.

Reviewer #2 (Public review):

Summary:

The manuscript by Pierre Despas and co-workers, reports the biochemical characterization of LppB a peculiar Lpp (Braun's lipoprotein) homolog found in Salmonella enterica. S. enterica encodes two Lpp homologs LppA and LppB: while LppA and Lpp function similarly, the role of LppB is less clear. LppB shares with Lpp the Cterminal Lys needed for covalent attachment to peptidoglycan (PG) but diverges in residues that precede the terminal Lys featuring a Cys residue at the penultimate position. By using E. coli as a surrogate model, the authors show that LppB can be covalently linked to PG via the terminal Lys residues and that the penultimate Cys residue can be used to form homodimer species when expressed alone and heterotrimeric complexes when co-expressed with Lpp. Interestingly, LppB expressed in E. coli seems to be stabilized at acidic pH a condition Salmonella encounters in macrophage phagosomes. Finally, based on decreased intensity of LppB-PG crosslinked bands as LppB expression increases the authors suggest that LppB is able to negatively modulate the outer membrane-peptidoglycan connectivity.

Strengths:

The manuscript is interesting, describes a novel strategy employed by bacteria to fine tuning outer membrane-PG attachment and provides new insights into how envelope remodeling processes can contribute to bacterial fitness and pathogenicity.

Weaknesses:

The analysis and quantification of muropeptides formed in E. coli strains overexpressing LppB would strengthen the main conclusion of the manuscript.

We thank Reviewer #2 for this insightful comment. We agree that quantitative analysis of muropeptides in E. coli strains expressing LppB would strengthen the main conclusion. This point was also raised in the editorial assessment and by Reviewer #3, underscoring its importance. In a revised version, we plan to perform muropeptide profiling by HPLC, coupled where appropriate to mass spectrometry, to quantitatively assess peptidoglycan composition in the relevant strains.

Reviewer #3 (Public review):

Summary:

The manuscript is interesting, and it is clearly written. While the experiments are well executed, a general flaw is that the LppA/B analyses are done in the E. coli K12 host as surrogate for Salmonella enterica. For the mechanistic and molecular analyses of LppB a surrogate host is certainly adequate, yet it limits extrapolation of the physiological implications of LppB in the natural context. 

Strengths:

The work convincingly demonstrates that LppB forms disulfide-based dimers and that it is crosslinked to PG via LdtB in E. coli. Moreover, dimerization is required for LppB abundance in E. coli and LppB can inhibit crosslinking of Lpp/A to PG in E. coli. 

Weaknesses:

Regarding the key conclusion of the work: while it is shown that LppB is oxidized in E. coli, whether envelope integrity (or OMV production) changes arise from switches in oxidation of the LppB cysteines remains to be shown, for E. coli let alone in the native host Salmonella. Does expression of LppB influence Lpp/A activity or OM tethering in E. coli? Since the inhibition of the Lpp/A linking to PG is not affected by the oxidation state of LppB, the abstract/title implies redox-control of envelope integrity which is a bit misleading and an overstatement. Both are features of LppB: i.e. it dimerizes through disulfide bond formation and it reduces PG binding of Lpp/A through trimerization. However, no link between the two is shown.

We thank Reviewer #3 for this important comment and for highlighting the need to clarify the relationship between LppB oxidation, oligomerization, and its effect on peptidoglycan crosslinking. We agree that while our data demonstrate that LppB forms disulfide-linked oligomers and that LppB expression reduces Lpp/A attachment to peptidoglycan, our current results do not establish a direct causal link between the oxidation state of LppB and its ability to modulate outer membrane–peptidoglycan tethering. Therefore, we will revise the manuscript to avoid implying redox-dependent control of envelope integrity and to more clearly present these as distinct but potentially related properties of LppB.

Author response:

We thank the reviewers for their constructive feedback and careful evaluation of our manuscript. We are encouraged that the study was viewed as well designed and clearly presented, that its computational modeling approach was recognized as a strength, and that the key findings were appreciated. We agree that some claims would benefit from additional support and clarification. Below, we outline the main revisions we will undertake to strengthen the manuscript and address the points raised in the reviews. These revisions are intended to strengthen the evidential support for our conclusions and clarify aspects of the results and modeling.

(1) Statistical support.

Some claims were judged to lack sufficient statistical support [Reviewer 1]. In the revised manuscript, we will carefully review all inferential claims and ensure that they are supported by appropriate statistical analyses. Where necessary, we will implement additional statistical tests and expand statistical reporting to ensure that differences between conditions, models, or behavioral measures are formally evaluated and that key aspects of the data are appropriately described.

(2) Modeling clarification.

Some aspects of the modeling were considered insufficiently clear, particularly regarding how the models were implemented [Reviewers 1 and 2]. We will expand the Methods section to provide a clearer and more complete description of the Bayesian models and their implementation. In particular, we will clarify that full probability distributions were computed (without reduced approximations such as those used in simplified Bayesian variants), and that the only approximation concerns numerical discretization of continuous state spaces at fine resolution. We will clarify that variance is part of the joint multidimensional state space and is inferred jointly with the mean. We will also explicitly state that apparent learning rates are derived from predicted paddle responses in the same way as for participants, and are not directly computed within the Bayesian inference process.

(3) Model fitting.

The absence of direct model fitting to individual participants was identified as a limitation [Reviewers 1 and 3]. In response, we will implement individual-level model fitting (to the extent feasible in practice) and conduct formal model comparison based on the fitted models. We will further validate the fitted models by examining whether they reproduce the main behavioral signatures observed in the data.

(4) Normative interpretation and control analyses.

The interpretation of the models as normative was questioned in light of the response-probability mechanism [Reviewer 2]. In the revision, we will clarify the distinction between the normative inference component of the model and the response-level mechanism. We will revise the framing of the results accordingly and ensure that normative claims are restricted to the inference component. We will also expand the discussion to integrate relevant literature on perseveration and satisficing, and clarify how normative and non-normative mechanisms may jointly shape behavior. In addition, following the reviewer’s suggestion, we will include control analyses using standard Rescorla–Wagner models, with and without the response-probability mechanism, to evaluate whether the observed signatures can be accounted for by simpler learning rules.

(5) Additional points.

We will also address the additional points raised in the reviews. Specifically, we will include supplementary histograms of apparent learning rates [Reviewer 2]. We will provide additional clarification and analyses regarding the effects of stochasticity on learning [Reviewer 1]. Finally, we will explore hybrid or mixture models and strategies and expand the discussion of this possibility [Reviewer 3].

We believe that these revisions will substantially strengthen the support for our claims and address the concerns raised in the current assessment. We are grateful for the reviewers’ engagement with our work and for their comments, which will allow us to significantly improve the clarity and strength of the manuscript.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

This work presents a GUI with SEM images of 8 Utah arrays (8 of which were explanted, and 4 of which were used for creating cortical lesions).

Strengths:

Visual comparison of electrode tips with SEM images, showing that electrolytic lesioning did not appear to cause extra damage to electrodes.

Weaknesses:

Given that the analysis was conducted on explanted arrays, and no functional or behavioural in vivo data or histological data are provided, any damage to the arrays may have occurred after explantation. This makes the results limited and inconclusive (firstly, that there was no significant relationship between degree of electrode damage and use of electrolytic lesioning, and secondly, that electrodes closer to the edge of the arrays showed more damage than those in the center).

We agree insofar as we could not fully control the circumstances of each array during explantation. However, array explantation is potentially damaging, but not universally damaging, as demonstrated by some largely intact arrays in this paper. If electrolytic lesions were damaging to the array, they would be observed. All arrays examined in this paper were carefully stored as described in the paper. All analyses of this type require an explant surgery [?????]. Our conclusions remain as strong as any of the results of these analyses.

Overall, these results do not add new insight to the field, although they do add more data and reference images.

We respectfully disagree, as there is no extant SEM analysis on electrode arrays used for lesioning.

Reviewer #2 (Public review):

In this study, the authors used scanning electron microscopy (SEM) to image and analyze eleven Utah multielectrode arrays (including eight chronically implanted in four macaques). Four of the eight arrays had previously been used to deliver electrolytic lesions. Each intact electrode was scored in five damage categories. They found that damage disproportionately occurred to the outer edges of arrays. Importantly, the authors conclude that their electrolytic Lesioning protocol does not significantly increase material degradation compared to normal chronic use without lesion. Additionally, the authors have released a substantial public dataset of single-electrode SEM images of explanted Utah arrays. The paper is well-written and addresses an important stability issue for long-term chronically implanted array recordings and electrolytic lesioning, which is relevant to both basic science and translational research. By comparing lesioning and non-lesioning electrodes on the same array and within the same animal, the study effectively controls for confounds related to the animal and surgical procedures. The shared dataset, accessible via interactive plots, enhances transparency and serves as a valuable reference for future investigations. Below, we outline some major and minor concerns that could help improve the work.

Major concerns:

(1) Electrode impedance is a critical measurement to evaluate the performance of recording electrodes. It would be helpful if the authors could provide pre-explant and post-explant impedance values for each electrode alongside the five SEM damage scores. This would allow the readers to assess how well the morphological scores align with functional degradation.

We agree, electrode impedance is very important in determining electrode performance. However, due to the multi-year, multi-subject nature of this work, we unfortunately do not have this data.

(2) The lesion parameters differ across experiments and electrodes. It would be helpful if the authors could evaluate whether damage scores (and/or impedance changes) correlate with total charge, current amplitude, duration, or frequency.

Thank you for this recommendation. We have included additional analyses in Supplementary Materials.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) ‘Both in vitro and in vivo testing of electrode arrays revealed environmental damage to these materials, such as cracking, textural defects, and degradation in response to the brain’s temperature and salinity [32]. The immune response of the brain also damages the electrodes due to effects like glial scarring (gliosis) and inflammation [33, 34]. This damage may be exacerbated by the surgical techniques used during implantation, which include pushing the electrode array into cortex and tethering the implant to the skull [33, 35, 36].’

In the above text, several relevant references have been left out, e.g.:

Barrese et al., 2013

Patel et al., 2023

Woeppel et al, 2021

Chen et al., 2023

Bjanes et al., 2025

Thank you for this recommendation. This section has been updated.

(2) ‘Aggressive electrical stimulation is known to dissolve platinum-based electrodes [37, 38]. Other studies have shown iridium oxide to be more resistant to stimulation-related damage, but not completely insusceptible [39, 40].’ Reference number 25 is relevant here.

Thank you for this recommendation. This section has been updated.

(3) ‘F’s and C’s PMd arrays were used for electrolytic lesioning experiments Monkey U was implanted with three 96-channel arrays; two in M1 and one in PMd.’ There seems to be a punctuation mark missing.

Thank you for this recommendation. This section has been updated.

(4) Methods: How much charge was injected via the electrodes that were used for lesioning? What current amplitudes, voltages, durations, and number of pulses were used? If more than 1 pulse was applied, what were the frequencies? Was the pulse cathode-only/ anode/only? What were the electrode impedance values at the time of stimulation? How many electrodes were used for lesioning at any given moment? How long after lesioning did the arrays remain in the tissue?

Thank you for your questions. An additional supplemental table (Supplemental Table 6) detailing specific NHP lesions parameters has been added. A summary of the lesion procedure (DC, bipolar, two electrodes at a time) has also been included in Methods. All arrays remained in the subject until explant, which ranged between hours (same-day lesion and explant) to several years. Further details on the lesioning procedure are available in citation [?]. Explant dates are available in Supplemental Table 1. Unfortunately, we do not have the impedance values at time of lesioning as this is not a measure we record frequently after implant, though we agree the data would be useful to have.

(5) Caption for Figure 1: ‘All array images are displayed with the wire bundle to the right side.’ I recommend adding this text from Figure 2 to the caption of Figure 1: ’electrode tips facing viewer’.

Thank you for this recommendation. This section has been updated.

(6) ‘Electrodes used for electrolytic lesioning are denoted with blue dots.’ Was stimulation carried out across all these electrodes simultaneously?

No, stimulation was not carried out across all electrode simultaneously. Pairs of electrodes were stimulated at the same time to create lesions. Lesions were performed on different days. We have updated our methods section to reflect this. See the Methods section and citation [?] for more details.

(7) For the control array, in Figure 1: ‘Click each column to view a close-up of the 5th row (from top to bottom) of electrodes:’ . It would be clearer to state: ’Click each column to view a close-up of a single electrode in the 5th row (from top to bottom):’.

Thank you for this recommendation. This section has been updated.

(8) Figure 2 caption: ‘Blank electrodes and electrodes with shank fractures are ignored and displayed in black, as they are not scored.’. What is a ‘blank’ electrode?

A ‘blank’ electrode is an electrode on the array that physically exists but is not wire bonded at time of manufacture to produce recordings. The corner electrodes of the Utah array are all blank electrodes. We have updated this wording to ‘unwired’ for clarity.

(9) I recommend incorporating Supplementary Figure 1 into Figure 2, so that the reader can immediately see where the rings are, without referring to the Supplementary Materials.

Thank you for this recommendation. We have chosen to keep these figures separate for stylistic reasons.

(10) Supplementary Figures: The figures should have the word ’Supplementary’ in the title, i.e., ‘Supplementary Figure X,’ not just ‘Figure X.’

Thank you for this recommendation. These captions have been updated.

(11) Throughout the results, the text is overly focused on the type of statistical test used and the p-values, e.g.: ‘When comparing lesioning and non-lesioning electrodes within the same array, each of the two nonparametric statistical tests (Mann-Whitney U-test, Levene Test) returned insignificant p-values for each category of damage as well as for total damage scores for all four arrays used in lesioning experiments.’.

To make the findings more digestible for the reader, the text should be rephrased in terms of whether the metrics being compared were significantly different or not. E.g.: ‘For each category of damage, as well as for the total damage score, no significant difference was found between electrodes that were or were not used for lesioning (either the mean or the variance of the scores).’.

Thank you for this recommendation. We have rephrased the text to reflect this note.

(12) ‘In Monkey H, the Mann-Whitney U test resulted in an insignificant p-value for coating cracks and parylene C delamination scores, while the Levene test resulted in an insignificant p-value for abnormal debris, coating cracks, and parylene C cracking scores. In Monkey F, the Mann-Whitney U test resulted in an insignificant p-value for parylene C delamination scores, while the Levene test resulted in an insignificant p-value for coating cracks, parylene C delamination, and parylene C cracking scores. In Monkey U, the Mann-Whitney U test resulted in significant p-values for all scores, while the Levene test resulted in an insignificant p-value for abnormal debris, tip breakage, and coating cracks scores. Finally, in Monkey C, the Mann-Whitney U test resulted in an insignificant p-value for parylene C delamination and parylene C cracking scores, while the Levene test resulted in an insignificant p-value for abnormal debris, parylene C delamination, and parylene C cracking scores.’

To point out another example, this chunk of text is highly repetitive and is unnecessary, as the reader can simply refer to Supplementary Table 4. It should be completely rephrased and summarized, to deliver the key message, i.e. briefly describe what kinds of damage occurred for which arrays. Also, what is the point of the two statistical tests? What are the authors trying to conclude?

Thank you for this recommendation. We have rephrased and pared down the text to reflect this note.

(13) Discussion: ‘Similarly, other work did not show significant differences in SEM-visible degradation between both platinum and iridium oxide coated electrodes used for stimulation [24, 25].’ What differences are being referred to here? Differences in degradation between stimulated Pt versus stimulated IrOx electrodes? Or between stimulated Pt and unstimulated PT electrodes? Stimulated IrOx and unstimulated IrOx? Or something else?

Thank you for your questions. We are comparing platinum against iridium oxide in this sentence. The wording of our original text has been updated to clarify our intention.

(14) Supplementary Tables: P-values lower than .05, .01, and .001 should simply be replaced with ¡.05, ¡.01, and ¡.001. The alpha value after a Bonferroni correction should be stated somewhere in each table or table caption.

Thank you for this recommendation. We have edited the tables to reflect this note.

(15) Title: ‘Material Damage to Multielectrode Arrays after Electrolytic Lesioning is in the Noise’ I don’t understand what the title means. What is in the noise? And what is ‘the noise’?

“In the noise” is a colloquialism referring to how background information (“noise”) may obscure or distract from other features. This title conveys how material damage to multielectrode arrays due to electrolytic lesioning is largely obscured by the general damage observed on multielectrode arrays after implant and explant.

(16) This reference has been left out altogether: Chen et al., 2014. The effect of chronic intracortical microstimulation on the electrode-tissue interface.

Thank you, this reference is now included.

Reviewer #2 (Recommendations for the authors):

(1) The number of lesion electrodes is low, especially since there are only 2-10 lesion electrodes on three of the four arrays, yielding limited statistical power.

We agree that the low number of lesioned electrodes limits statistical power. However, due to ethical considerations, it is unlikely for arrays to contain much more than this number of lesion electrodes.

(2) The dataset includes both platinum and iridium oxide-coated electrodes. A direct comparison of their damage profiles would be informative.

Thank you for this recommendation. We have included this additional analysis in Supplementary Materials.

(3) It is unclear what “is in the Noise” in the title means without reading the manuscript. It is helpful to improve the clarity of the title.

Thank you for this recommendation.

(4) Please spell out “PMd” and “M1” at first mention to facilitate reading.

Thank you for this note. The text has been updated to reflect this recommendation.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

Using single-unit recording in 4 regions of non-human primate brains, the authors tested whether these regions encode computational variables related to model-based and model-free reinforcement learning strategies. While some of the variables seem to be encoded by all regions, there is clear evidence for stronger encoding of model-based information in the anterior cingulate cortex and caudate.

Strengths:

The analyses are thorough, the writing is clear, and the work is well-motivated by prior theory and empirical studies.

Weaknesses:

My comments here are quite minor.

The correlation between transition and reward coefficients is interesting, but I'm a little worried that this might be an artifact. I suspect that reward probability is higher after common transitions, due to the fact that animals are choosing actions they think will lead to higher reward. This suggests that the coefficients might be inevitably correlated by virtue of the task design and the fact that all regions are sensitive to reward. Can the authors rule out this possibility (e.g., by simulation)?

We fully agree with the reviewer that the task design has in-built correlations between transition and reward, and thus the correlation between neural selectivity for feedback and transition (Figure 3E) may be due to the different reward expectation after common or rare transitions. We did try to make this point in the manuscript:

This suggests that the brain treats being diverted away from your current objective equivalent to losing reward, which is sensible as the subject would normally expect lower rewards on rare trials if their reward-seeking behaviour was efficient.

We’ve now updated the wording of this statement to try and better make this point and avoid confusion that any non-reward-related encoding is involved:

“As the reward expectation will be higher on common compared to rare trials, this demonstrates that the brain encodes being diverted to an area with a lower reward expectation equivalent to actually receiving a low reward (and vice versa).”

We have also adjusted the significance test of this correlation to use a circular permutation test that accounts for correlations between the regressors. This test still found there to be significant correlation in all areas.

We have described this new permutation test in Methods:

“For comparing correlations between weights for different features (i.e., between transition and reward coding, Figure 3E), the null distribution of correlations observed in circularly shifted data was compared to the correlation seen in the actual data. This accounts for any correlations between features that existed in the task by preserving the structure of the design matrices.”

And updated the text in Results accordingly:

“All regions, but particularly ACC, encoded a common transition (at the time of transition) similar to a high reward (at the time of feedback), as there was a positive correlation between the coefficients for reward and transition (the transition parameter was signed such that common and rare transitions were equivalent to high and low rewards, respectively) (ACC r=0.4963, DLPFC r=0.3273, caudate r=0.4712, putamen, r=0.5052; all p<0.002 except DLPFC where p=0.006, circular permutation test; Figure 3E, S5).”

The explore/exploit section seems somewhat randomly tacked on. Is this really relevant? If yes, then I think it needs to be integrated more coherently.

We thank the reviewer for this comment. We agree that the motivation for the explore/exploit analysis was not sufficiently clear in the original version.

Our aim was not to introduce this as a separate or tangential effect, but rather to highlight how the task’s reward structure (with outcome levels stable for 5–9 trials) naturally created alternating periods favoring exploitation of a known high-value option versus exploration when outcomes changed. This feature of the task is tightly linked to MB-RL computations, as it requires integration of state-transition knowledge and updating across trials.

Importantly, we show previously in the manuscript that ACC encoded state-transition structure (i.e., common versus rare transition) and MB-value estimates (at choice epoch). However, here we aimed to highlight that the same region also modulated choice encoding as a function of whether the subject was in an exploratory or exploitative regime – by knowing another feature of the task that relies on state-transition and outcome. We have revised this section to better integrate it into the main logic of the paper:

“In our task, the outcome level (high, medium, low) of each second-stage stimulus remained the same for 5-9 trials before potentially changing. This design naturally created periods where subjects could ‘exploit’ the same Choice 1 to maximize reward for several trials; and other periods where they had to ‘explore’ different second-stage stimuli to optimize reward (as contingencies shifted). In classical MB-RL, the transition between reward states can be learned by keeping counts of observed transitions from a current state-action pair to a subsequent state, yielding a maximum-likelihood estimate of the environment’s dynamics [42]. In fact, knowledge about the reward contingency schedule could support decision-making in both exploitation – by enabling efficient choice when rewards are stable; and exploration – by guiding alternative behaviour most likely to yield improved outcomes (this is different from MF learning, where exploration is more random since the agent lacks explicit state-transition knowledge).

We thus repeated our decoding analysis of choice 1 stimulus identity, but this time limited trials to those where they had not received a high reward for the previous two trials (‘explore’ trials), and those where the previous two rewards had been the highest level (‘exploit’ trials). All regions encoded choice 1 for some duration of the choice epoch for both explore (p<0.002 in all cases, permutation test; Figure 7A) and exploit (p<0.002 in all cases; Figure 7B) conditions, but decoding accuracy was strongest in ACC. Choice 1 was less strongly decoded – particularly in ACC – in the former condition compared to the latter (p<0.002 for at least 140 ms in all cases, permutation test on differences observed; Figure 7C); and, also during exploitation, the ACC encoded choice 1 before the choice was even presented to the subject (Figure S8). This pre-choice ACC encoding in exploit trials may reflect the need to allocate cognitive (or attentive) resources to features – i.e., choice 1 stimulus identity – that are most certain predictors of important outcomes. As a control, we also decoded the direction of the Choice 1 (where choice was indicated via joystick movement), which was randomised each trial and therefore orthogonal to the stimulus that was chosen. Again, all four regions encoded its direction in both explore (p<0.002 in all cases; Figure 7D) and exploit (p<0.002 in all cases; Figure 7E). However, there were minimal differences in the strength of the representation between explore and exploit conditions (ACC, p=0.088, cluster-based permutation test; DLPFC p=0.016; caudate p=0.32; putamen p=1; Figure 7F). Therefore, exploit behaviour specifically upregulated relevant task parameters that were worth remembering across trials.”

Reviewer #2 (Public review):

Summary:

The authors investigate single-neuron activity in rhesus macaques during model-based (MB) and model-free (MF) reinforcement learning (RL). Using a well-established two-step choice task, they analyze neural correlates of MB and MF learning across four brain regions: the anterior cingulate cortex (ACC), dorsolateral PFC (DLPFC), caudate, and putamen. The study provides strong evidence that these regions encode distinct RL-related signals, with ACC playing a dominant role in MB learning and caudate updating value representations after rare transitions. The authors apply rigorous statistical analyses to characterize neural encoding at both population and single-neuron levels.

Strengths:

(1) The research fills a gap in the literature, which has been limited in directly dissociating MB vs. MF learning at the single unit level and across brain areas known to be involved in reinforcement learning. This study advances our understanding of how different brain regions are involved in RL computations.

(2) The study used a two-step choice task Miranda et al., (2020), which was previously established for distinguishing MB and MF reinforcement learning strategies.

(3) The use of multiple brain regions (ACC, DLPFC, caudate, and putamen) in the study enabled comparisons across cortical and subcortical structures.

(4) The study used multiple GLMs, population-level encoding analyses, and decoding approaches. With each analysis, they conducted the appropriate controls for multiple comparisons and described their methods clearly.

(5) They implemented control regressors to account for neural drift and temporal autocorrelation.

(6) The authors showed evidence for three main findings:

(a) ACC as the strongest encoder of MB variables from the four areas, which emphasizes its role in tracking transition structures and reward-based learning. The ACC also showed sustained representation of feedback that went into the next trial. b) ACC was the only area to represent both MB and MF value representations.

(c) The caudate selectively updates value representations when rare transitions occur, supporting its role in MB updating.

(7) The findings support the idea that MB and MF reinforcement learning operate in parallel rather than strictly competing.

(8) The paper also discusses how MB computations could be an extension of sophisticated MF strategies.

Weaknesses:

(1) There is limited evidence for a causal relationship between neural activity and behavior. The authors cite previous lesion studies, but causality between neural encoding in ACC, caudate, and putamen and behavioral reliance on MB or MF learning is not established.

We agree with the reviewer that the present study does not establish causal relationships, and we do not claim otherwise in the manuscript. Our work was designed as a comprehensive characterization of neural activity across ACC, DLPFC, caudate, and putamen during reward-seeking decision-making. By systematically comparing MB- and MF- RL signals across these regions, we provide new insights into the division of labor and cooperative interactions within cortico-striatal networks.

While causal manipulations (e.g., lesions, inactivations, stimulation) are indeed required to directly establish necessity or sufficiency, correlational studies such as ours play a crucial role in identifying where and how computationally relevant signals are represented. Importantly, our findings align with and extend prior causal work, for example showing that ACC and striatal lesions disrupt MB control. Thus, our study contributes a detailed functional mapping of MB and MF RL encoding across multiple nodes of this circuit, which serves as an important foundation for future causal investigations (e.g., using transcranial ultrasound stimulation).

(2) There is a heavy emphasis on ACC versus other areas, but it is unclear how much of this signal drives behavior relative to the caudate.

We appreciate the reviewer's observation regarding this matter. Our intention was not to place a heavy emphasis on ACC, rather this came naturally from the data. The ACC demonstrated considerably more robust and enduring neural activity compared to other brain regions – for instance, reward-related signals in the ACC continued well beyond individual trials (Fig. 2A-B), and encoding of state transitions remained active from the initial transition through to the feedback phase (Fig. 3A-B). By comparison, distinctions among other regions were less pronounced, which naturally resulted in the ACC receiving greater attention in our analytical findings.

We acknowledge that the caudate plays an essential and complementary role in driving behavior, and we believe that this is emphasized in the two key subsections of our “Results”. First, caudate neurons encoded model-based choice values (Fig. 4A, 4C) and uniquely remapped these values following rare transitions (Fig. 5), reflecting flexible adjustment of action values. Second, decoding analyses showed that both ACC and caudate populations predicted first-stage choices (Fig. 6C), linking their activity directly to behavioral decisions. In the Discussion section, we also highlight that “the distinctive caudate signal of updating (flipping) the value estimates of the currently experienced option on rare trials” goes beyond a “general temporal-difference RPE” and rather supports “the role of caudate in MB valuation”.

(3) The role of the putamen is somewhat underexplored here.

Our analyses were conducted in an identical manner across all four recorded regions (ACC, DLPFC, caudate, and putamen), and we consistently reported the results for putamen alongside the others. For example, in the Results section we describe how “both caudate and putamen encoded the reward from the previous trial negatively during the feedback period of the current trial” (Fig. 2F-G), and that “all regions had a significant population of neurons that encoded MB-, but not MF-, derived value” including putamen (Fig. 4F). Similarly, we show that putamen, like caudate, encoded a dopamine-like RPE signal at feedback (“both caudate and putamen neurons clearly responded at feedback with the parametric features of a dopamine-like RPE”; Discussion). These findings align with previous work linking the putamen to MF learning and are discussed explicitly in the context of MF-MB dissociations. We therefore believe that the putamen was not underexplored, but rather that its contribution was more circumscribed relative to ACC and caudate because the signals observed were quantitatively weaker and less distinctive for MB computations.

(4) The authors mention the monkeys were overtrained before recording, which might have led to a bias in the MB versus MF strategy.

We agree that extensive training can influence the balance between MB and MF in choice behaviour and neuronal responses.

In a previous comprehensive behavioral analysis of the same dataset (Miranda et al., 2020, PLoS Computational Biology - ref. 36, Figure S6B) we showed that both MB and MF strategies contributed to behavior, with MB dominance stable across weeks of testing – supporting that overtraining did not eliminate MF influences (but rather stabilized a mixed strategy with robust MB contributions).

In the same manuscript, we have also: i) cautioned the readers when comparing our results to data from the original human studies; ii) acknowledged that our extensive training cannot address earlier phases of learning in which sensitivity to the task structure is first acquired; and iii) also provided task-related reasons for such MB dominance – as training made the transition structure well learned (making MB computationally less costly and faster to implement) and the non-stationary outcomes favored the flexibility of MB strategies.

In the present manuscript, we also have acknowledged that overtraining may have shifted neural signals toward stronger MB representations, or alternatively enabled more sophisticated task representations:

“On the other hand, MF-based estimates were neither as striking nor as specific to striatal regions as expected and observed in previous studies [18]. The monkeys were extensively trained on the task before recordings commenced, which may have caused a shift towards both MB behaviour and MB value representation within the striatum. Alternatively, this training may have allowed more sophisticated representations to occur, such as using latent states to expand the task space [54].”

Importantly, we strongly believe that this possibility does not detract from our main finding that both MB and MF signals were present across regions, with ACC showing the strongest multiplexing of the two.

(5) The GLM3 model combines MB and MF value estimates but does not clearly mention how hyperparameters were optimized to prevent overfitting. While the hybrid model explains behavior well, it does not clarify whether MB/MF weighting changes dynamically over time.

We appreciate this comment and would like to note that, for completeness, we have on several occasions directed the reader to our prior behavioural analysis of the same dataset (Miranda et al., 2020, PLoS Computational Biology, ref 36). In that work, we provide a full and detailed description of both the task and the computational modeling approach (see particularly the “Model fitting procedures” section). Furthermore, our model-fitting was grounded in the MF/MB RL framework used in the original human two-step study (Daw et al., 2011); and the fitting procedures also followed previous studies (Huys et al., 2011).

Hyperparameters – including the MB/MF weighting parameter (ω) - were estimated using maximum likelihood under two complementary approaches and with priors providing regularization across sessions. First, we performed a fixed-effects analysis, in which parameters were estimated independently for each session by maximizing the likelihood separately; secondly, we conducted a mixed-effects analysis, treating parameters as random effects across sessions within each subject. The effect of the prior procedure reduces the risk of overfitting by constraining parameters based on their empirical distributions, rather than allowing unconstrained session-by-session estimates. Finally, all model fitting procedures were verified on surrogate generated data.

With regard to dynamic weighting, our approach – consistent with most two-step studies – assumed ω to be constant across trials within each session. This was a deliberate choice, both for comparability with prior work and because our subjects were extensively trained, making session-level stability of strategy weights a reasonable assumption. Indeed, our analyses showed no systematic drift in ω across sessions, suggesting that MB/MF balance was stable over sessions. While approaches that allow dynamic ω estimation are possible, we believe such extensions would likely have minimal impact in the current dataset.

(6) It was unclear from the task description whether the images used changed periodically or how the transition effect (e.g., in Figure 3) could be disambiguated from a visual response to the pair of cues.

All images were kept constant across sessions. Common/Rare transitions themselves were not explicitly cued, but rather each second-stage state was associated with a specific background colour, followed ~1s later by the presentation of two specific second-stage choice cues (Figure 1B). Hence the subject could infer whether they were transitioned down a Rare or Common path by the background colour, which can be disambiguated in time from the visual responses to the second-stage cues. We’ve updated the Results text to make this clearer:

“Tracking the state-transition structure of the task is imperative for solving the task as a MB-learner. All four regions encoded whether the current trial’s first-stage choice transitioned to the common or rare second-stage state (which could be inferred by a change in background colour immediately after choice indicating which second stage state they had just entered, Figure 1A).”

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Figure 7 appears to be missing.

We thank the reviewer for pointing this out. Figure 7 was inadvertently omitted in the previous version and has now been included in the revised manuscript.

(2) No stats reported in the section on explore/exploit.

We apologise for this oversight. This section now also reports the relevant statistics:

“We thus repeated our decoding analysis of choice 1 stimulus identity, but this time limited trials to those where they had not received a high reward for the previous two trials (‘explore’ trials), and those where the previous two rewards had been the highest level (‘exploit’ trials). All regions encoded choice 1 for some duration of the choice epoch for both explore (p<0.002 in all cases, permutation test; Figure 7A) and exploit (p<0.002 in all cases; Figure 7B) conditions, but decoding accuracy was strongest in ACC. Choice 1 was less strongly decoded – particularly in ACC – in the former condition compared to the latter (p<0.002 for at least 140 ms in all cases, permutation test on differences observed; Figure 7C); and, also during exploitation, the ACC encoded choice 1 before the choice was even presented to the subject (Figure S8). This pre-choice ACC encoding in exploit trials may reflect the need to allocate cognitive (or attentive) resources to features – i.e., choice 1 stimulus identity – that are most certain predictors of important outcomes. As a control, we also decoded the direction of the Choice 1 (where choice was indicated via joystick movement), which was randomised each trial and therefore orthogonal to the stimulus that was chosen. Again, all four regions encoded its direction in both explore (p<0.002 in all cases; Figure 7D) and exploit (p<0.002 in all cases; Figure 7E). However, there were minimal differences in the strength of the representation between explore and exploit conditions (ACC, p=0.088, cluster-based permutation test; DLPFC p=0.016; caudate p=0.32; putamen p=1; Figure 7F).”

(3) Make sure that error bars are explained in all figure captions where appropriate.

We apologise that this information was absent. Error bars always represent the standard error of the mean. This has now been added to all relevant figure legends.

Reviewer #2 (Recommendations for the authors):

Overall, I think this is a great manuscript and was presented clearly and succinctly. I have some minor suggestions:

(1) Typo: Abstract "ACC, DLPFC, caudate and striatum" I think should be "caudate and putamen".

We have amended this incorrect reference in the introduction:

“One such task that does enable the dissociation of MB and MF computations is Daw et al. (2011)’s ‘two-step’ task [18]. It contains a probabilistic transition between task states to uncouple MF learners (who would assign credit to which state was rewarded regardless of the transition) from MB learners (who would appropriately assign credit based on the reward and transition that occurred). Rodents [19], monkeys [36], and humans [18] all use MB-like behaviour to solve the task. Evidence in rodents suggests dorsal anterior cingulate cortex (ACC) tracks rewards, states, and the probabilistic transition structure, and that ACC is essential in implementing a MB-strategy [37]. Here, we compare primate single neuron activity of 4 different subregions implicated in reward-based learning and choice (ACC, dorsolateral PFC (DLPFC), caudate, and putamen) during performance of the classic two-step task, and demonstrate signatures of MB-RL primarily in ACC, and MF-RL signatures most notably in putamen.”

(2) Could the authors provide a rationale for why they did the single-level encoding the way they did, instead of running an ANOVA?

We thank the reviewer for this point. We are not entirely certain which specific ANOVA approach is being suggested, but our rationale for using a GLM-based encoding analysis is that such approach allows us to model continuous, trial-by-trial variables (e.g., value signals, prediction errors, transitions) while simultaneously controlling for multiple correlated predictors. This approach is widely used in systems neuroscience (particularly in decision-making research) offering analytical flexibility and comparability with prior approaches.

(3) How were the 20 iterations for decoding decided? That seems low.

We do not agree that 20 repetitions of 5-fold cross validation is low. The error bars in panels 6C-E demonstrate what low variance occurred across these 20 repetitions. It is the average of these low variance repetitions against which we performed statistics by performing a permutation test where these 20 repetitions were repeated a further 500 times.

(4) It was unclear to me how the authors reached the conclusion "Thus, caudate activity appeared to represent the value of the state the subject was currently in." when the state value wasn't computed directly. I don't see how encoding the chosen and unchosen option is the same as the state the animal is in, which should also incorporate where the animal is in a block of trials or session, and the knowledge regarding the chosen and unchosen option.

We agree with this point and have tempered this statement:

“Thus, caudate’s encoding of an option’s value also reflected the availability of the option.”

(5) Figures 1C, D, and E were not legible to me even at 200% zoom.

We apologise for this oversight. We’ve now updated panels 1C-E to a more readable size:

(6) There is a Figure 2H in the figure legend, but the panel appears to be missing from Figure 2.

This text has been removed.

(7) Figure 2: It would've been nice to see F and G for all areas.

We have now added this data as additional panels in Figure 2.

(8) Figure 3: How is the transition disambiguated from a visual response to the set of images?

This was indicated by the background changing colour to that of the learned second stage state before the actual choices were presented. We’ve updated the Results text to make this clearer:

“Tracking the state-transition structure of the task is imperative for solving the task as a MB-learner. All four regions encoded whether the current trial’s first-stage choice transitioned to the common or rare second-stage state (which was indicated by a change in background colour before the second stage choices were presented, Figure 1A).”

(9) Figure 4F: Is this collapsed across time points? So neurons that were significant at any time? I'm confused how Figure 4A relates to 4F, as 4A shows much lower percentages of significant neurons.

Figure 4F counts the total number of neurons that had a significant period of encoding at any timepoint over the epoch (as assessed with a length-based permutation test). Whereas, 4A shows the amount of significant encoding neurons at any one time point. Investigating this further, we found that the encoding was dynamic with different neurons encoding different parts of the epoch. We have now added a new supplementary figure to highlight this and refer to it in Results:

“Examination of the strongest signal observed, ACC’s encoding of MB Q-values, showed a dynamic pattern with different neurons encoding the signal at different parts of the epoch (Figure S6). When aggregating the number of significant coders throughout the epoch, and examining the specificity of MB versus MF coding, we found that all regions had a significant population of neurons that encoded MB-, but not MF-, derived value (30, 18.72, 23 and 24% of neurons in ACC, DLPFC, caudate and putamen respectively; all p<0.0014 binomial test against 10% (as the strongest response to either of the two options was used); Figure 4F).“

(10) Data/ code could be made publicly available instead of upon request.

All data and code to reproduce figures are now available at https://github.com/jamesbutler01/TwoStepExperiment. The manuscript has been updated to reflect this:

Data and materials availability:

All data and code to reproduce figures are available at https://github.com/jamesbutler01/TwoStepExperiment.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

The authors' goal was to advance the understanding of metabolic flux in the bradyzoite cyst form of the parasite T. gondii, since this is a major form of transmission of this ubiquitous parasite, but very little is understood about cyst metabolism and growth. Nonetheless, this is an important advance in understanding and targeting bradyzoite growth.

Strengths:

The study used a newly developed technique for growing T. gondii cystic parasites in a human muscle-cell myotube format, which enables culturing and analysis of cysts. This enabled the screening of a set of anti-parasitic compounds to identify those that inhibit growth in both vegetative (tachyzoite) forms and bradyzoites (cysts). Three of these compounds were used for comparative Metabolomic profiling to demonstrate differences in metabolism between the two cellular forms.

One of the compounds yielded a pattern consistent with targeting the mitochondrial bc1 complex and suggests a role for this complex in metabolism in the bradyzoite form, an important advance in understanding this life stage.

Weaknesses:

Studies such as these provide important insights into the overall metabolic differences between different life stages, and they also underscore the challenge of interpreting individual patterns caused by metabolic inhibitors due to the systemic level of some of the targets, so that some observed effects are indirect consequences of the inhibitor action. While the authors make a compelling argument for focusing on the role of the bc1 complex, there are some inconsistencies in the patterns that underscore the complexity of metabolic systems.

We agree with reviewer #1 that metabolic fingerprints are complex to interpret and we did try to approach this problem by including mock treatment and non-metabolic inhibitors as controls. We address specific concerns below.

Reviewer #2 ( Public review):

Summary:

A particular challenge in treating infections caused by the parasite Toxoplasma gondii is to target (and ultimately clear) the tissue cysts that persist for the lifetime of an infected individual. The study by Maus and colleagues leverages the development of a powerful in vitro culture system for the cyst-forming bradyzoite stage of Toxoplasma parasites to screen a compound library for candidate inhibitors of parasite proliferation and survival. They identify numerous inhibitors capable of inhibiting both the disease-causing tachyzoite and the cyst-forming bradyzoite stages of the parasite. To characterize the potential targets of some of these inhibitors, they undertake metabolomic analyses. The metabolic signatures from these analyses lead them to identify one compound (MMV1028806) that interferes with aspects of parasite mitochondrial metabolism. The authors claim that MV1028806 targets the bc1 complex of the mitochondrial electron transport chain of the parasite, although the evidence for this is indirect and speculative. Nevertheless, the study presents an exciting approach for identifying and characterizing much-needed inhibitors for targeting tissue cysts in these parasites.

Strengths:

The study presents convincing proof-of-principle evidence that the myotube-based in vitro culture system for T. gondii bradyzoites can be used to screen compound libraries, enabling the identification of compounds that target the proliferation and/or survival of this stage of the parasite. The study also utilizes metabolomic approaches to characterize metabolic 'signatures' that provide clues to the potential targets of candidate inhibitors, although falls short of identifying the actual targets.

Weaknesses:

(1) The authors claim to have identified a compound in their screen (MMV1028806) that targets the bc1 complex of the mitochondrial electron transport chain (ETC). The evidence they present for this claim is indirect (metabolomic signatures and changes in mitochondrial membrane potential) and could be explained by the compound targeting other components of the ETC or affecting mitochondrial biology or metabolism in other ways. In order to make the conclusion that MMV1028806 targets the bc1 complex, the authors should test specifically whether MMV1028806 inhibits bc1-complex activity (i.e. in a direct enzymatic assay for bc1 complex activity). Testing the activity of MMV1028806 against other mitochondrial dehydrogenases (e.g. dihydroorotate dehydrogenase) that feed electrons into the ETC might also provide valuable insights. The experiments the authors perform also do not directly measure whether MMV1028806 impairs ETC activity, and the authors could also test whether this compound inhibits mitochondrial O2 consumption (as would be expected for a bc1 inhibitor).

We thank the reviewer for highlighting this important aspect. To further investigate the effect of MMV1028806 on the mETC, we adapted a commercial oxygen consumption assay and demonstrated that MMV1028806, like Atovaquone and Buparvaquone, inhibits the ETC, leading to reduced oxygen consumption similar to Antimycin A, which inhibits the bc1-complex. These results are now included in the revised manuscript (Methods, lines 210–233; Results, lines 460–468).

(2) The authors claim that compounds targeting bradyzoites have greater lipophilicity than other compounds in the library (and imply that these compounds also have greater gastrointestinal absorbability and permeability across the blood-brain barrier). While it is an attractive idea that lipophilicity influences drug targeting against bradyzoites, the effect seems pretty small and is complicated by the fact that the comparison is being made to compounds that are not active against parasites. If the authors are correct in their assertion that lipophilicity is a major determinant of bradyzoicidal compounds compared to compounds that target tachyzoites alone, you would expect that compounds that target tachyzoites alone would have lower lipophilicity than those that target bradyzoites. It would therefore make more sense to (statistically) compare the bradyzoicidal and dual-acting compounds to those that are only active in tachyzoites (visually the differences seem small in Figure S2B). This hypothesis would be better tested through a structure-activity relationship study of select compounds (which is beyond the scope of the study). Overall, the evidence the authors present that high lipophilicity is a determinant of bradyzoite targeting is not very convincing, and the authors should present their conclusions in a more cautious manner.

Thank you for raising this excellent point. We performed a statistical test of tachyzoidal and both bradyzoidal and dually active compounds and find indeed no significant difference (P = 0.06). We altered the results text line 367-368 and the figure S2B caption to explicitly mention this.

(3) Page 11 and Figure 7. The authors claim that their data indicate that ATP is produced by the mitochondria of bradyzoites "independently of exogenous glucose and HDQ-target enzymes." The authors cite their previous study (Christiansen et al, 2022) as evidence that HDQ can enter bradyzoites, since HDQ causes a decrease in mitochondrial membrane potential. Membrane potential is linked to the synthesis of ATP via oxidative phosphorylation. If HDQ is really causing a depletion of membrane potential, is it surprising that the authors observe no decrease in ATP levels in these parasites? Testing the importance of HDQ-target enzymes using genetic approaches (e.g. gene knockout approaches) would provide better insights than the ATP measurements presented in the manuscript, although would require considerable extra work that may be beyond the scope of the study. Given that the authors' assay can't distinguish between ATP synthesized in the mitochondrion vs glycolysis, they may wish to interpret their data with greater caution.

We thank the reviewer for addressing this important point. The enzymatic assay used in our study cannot distinguish whether ATP is produced via glycolysis or mitochondrial respiration. However, we minimized glycolytic ATP production in bradyzoites by starving them for one week without glucose. After this period, amylopectin stores are depleted, forcing the parasites to utilize glutamine via the GABA shunt to fuel the TCA cycle and generate ATP predominantly through respiration. While minor ATP production via gluconeogenic fluxes cannot be excluded, the main ATP supply under these conditions is expected to originate from the mitochondrial electron transport chain. Indeed, ATP levels are lower in HDQ-treated bradyzoites, which we attribute to the compound’s impact on electron-supplying enzymes upstream of the bc1 complex, although this inhibition is not sufficient to fully abolish ATP production as observed with Atovaquone treatment.

Reviewer #3 (Public review):

Summary:

The authors describe an exciting 400-drug screening using a MMV pathogen box to select compounds that effectively affect the medically important Toxoplasma parasite bradyzoite stage. This work utilises a bradyzoites culture technique that was published recently by the same group. They focused on compounds that affected directly the mitochondria electron transport chain (mETC) bc1-complex and compared them with other bc1 inhibitors described in the literature such as atovaquone and HDQs. They further provide metabolomics analysis of inhibited parasites which serves to provide support for the target and to characterise the outcome of the different inhibitors.

Strengths:

This work is important as, until now, there are no effective drugs that clear cysts during T. gondii infection. So, the discovery of new inhibitors that are effective against this parasite stage in culture and thus have the potential to battle chronic infection is needed. The further metabolic characterization provides indirect target validation and highlights different metabolic outcomes for different inhibitors. The latter forms the basis for new studies in the field to understand the mode of inhibition and mechanism of bc1-complex function in detail.

The authors focused on the function of one compound, MMV1028806, that is demonstrated to have a similar metabolic outcome to burvaquone. Furthermore, the authors evaluated the importance of ATP production in tachyzoite and bradyzoites stages and under atovaquone/HDQs drugs.

Weaknesses:

Although the authors did experiments to identify the metabolomic profile of the compounds and suggested bc-1 complex as the main target of MMV1028806, they did not provide experimental validation for that.

In our updated manuscript we performed additional experiments such as oxygen consumption assay to further qualify the bc1 complex as the target. We also toned down some of our statements to make sure that no false claims are made.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Introduction: It would be helpful to briefly describe what the pathogen Box is, what compounds are in it, and the rationale for using a drug screen to better understand mitochondrial function in cysts.

Thank you for this suggestion, we added an introduction of the MMV pathogen box and outlined our rationale for our experimental approach in lines 90 to 99.

Please explain why dual-active drugs were useful for understanding differences, rather than just seeking drugs that might target bradyzoites alone.

We focused on dually active compounds for two reasons. First, these are the most promising and potent targets to develop drugs against. Both stages might occur simultaneously and these dually active drugs may eliminate the need for treatment with a drug combination. Second, we speculated that monitoring the responses to inhibition of the same process in both parasite stages would reveal its functional consequences. Dually active compounds enable this direct comparison. Bradyzoite-specific compounds may be interesting from a developmental perspective but may require a reverse genetic follow-up to compare differences between stages. The lack of a well-established inducible expression system in bradyzoites that allows short term and synchronized knock-down makes metabolomic approaches difficult. We added these two points in brief to the results section (line 378 – 381).

Figure 4: this is a very important figure in understanding the significance of the work, but it is not well described in the legend. Even if these graphics have been used in other manuscripts, it would be helpful to provide better annotation in the figure legend.

Thank you for pointing this out. We expanded the figure legend to explain the isotopologues data in more detail. Line 793 to 802.

B,D: Explain what the three columns for each drug category represent.

Addressed

C,E: Explain what isotopologues are, what the M+ notation means, and what the pie charts represent. Other main figures have suitable legends.

Addressed

Discussion: there are several places where the reasoning is a bit hard to follow, and rearrangement to provide a clear logical flow would be helpful. In particular, the reasoning for why HDQ impairs active but non-essential processes could be laid out more clearly.

We added additional clarifications to the discussion section and re-wrote the HDQ paragraph. We hope that our reasoning is now easier to follow.

Abbreviations: A list of abbreviations for the entire manuscript would be helpful.

This is a good idea and we now provide an abbreviations list.

Minor typos:

P12, 2d paragraph: sentence beginning with: Consistent with this hypothesis... "cysts" is used twice

Corrected

P15, top of the second paragraph: "nano" and "molar" should be one word

Corrected

Reviewer #2 (Recommendations for the authors):

Major comments (not already covered in the weaknesses section of the public review)

(1) Figure 2 and the related description of these experiments in the methods section (page 3). The approach for calculating IC50 values for the compounds against tachyzoites is unclear. How did the authors determine the time point for calculating IC50 vacuoles? Was this when the DMSO control wells reached maximum fluorescence? This could be described in a clearer manner. A concern with calculating IC50 values on different days is that parasites will have undergone more lytic cycles after 7 days compared to 4 days, which means that the IC50 values for fast- vs slow-acting compounds might be quite different between these days. As a more minor comment on these experiments, the methods section does not describe whether the test compound was removed after 7 days, as the experimental scheme in Figure S1A seems to imply. Please clarify in the methods section.

This is a very good point and we clarified this in the methods section, line 157–160. In brief, we choose the latest time point when exponential growth could be observed in the fastest growing cultures, generally this was in mock treated cultures and at day 4 post infection. We also clarified that we changed media and removed treatment after 7 days.

Minor Comments

(2) Page 2. "we employed a recently developed human myotube-based culture system to generate mature T. gondii drug-tolerant bradyzoites". What makes these bradyzoites 'drug-tolerant' or to which drugs are they tolerant? This isn't clear from the description.

We added these details in the introduction (line 94 to 96) and state that these cysts develop resistance against anti-folates, bumped kinase inhibitors and HDQ, a Co-enzyme Q analog.

(3) Figure 1E. The number of compounds in this pie chart adds up to 384, whereas the methods describe that 371 compounds were tested. What explains this discrepancy in numbers?

We understand the confusion. We now updated the pie chart to reflect only compounds that were included in the primary screen (371) as reflected in Supplementary Table S1. We separately analysed 29 compounds that were previously tested against tachyzoites by Spalenka et al., and found an additional 13 compound, that were originally included in the pie chart. In a secondary test the activity of 10 of these 13 compounds could be confirmed. All in all we found the 16 compounds shown in Fig. 2 E-G.

(4) Page 3. The resazurin assays for measuring host cell viability could be explained in a clearer manner. What host cells were used? Were the host cells confluent when the drug was added (and the assay conducted) or was the drug added when the host cells were first seeded? How long were the host cells cultured in the candidate inhibitors before the assays were performed? What concentration (or concentration range) were the compounds tested? The host inhibition data are not easily accessible to the reader - the authors might consider including these data as part of Table S2D.

The necessary information was added to the methods section (line 145 to 153). We tested for host toxicity in both HFF and KD3 myotubes during the primary screen at 10 µM in triplicates. The colorimetric assay was performed after tachyzoite growth assays in HFFs 7 days post infection and after completion of the 4 week re-growth phase of bradyzoites in myotubes. The resulting data is already part of Supplementary File 1. In addition, we performed concentration dependent resazurin assays after secondary concentration dependent growth inhibition assays and also included data in Supplementary File 1. For the bradyzoite growth assay we performed visual inspection after drug exposure for one week and before tachyzoite re-growth to detect missing or damaged monolayer. Also, this data is included in the Supplementary File 1. We also included the cytotoxicity data as suggested into Table S2D.

(5) Page 7. "Except for four compounds (MMV021013, MMV022478, MMV658988, MMV659004), minimal lethal concentrations were higher in bradyzoites". The variation in these data seems quite large to be making this claim. Consider a statistical analysis of these data to compare potencies in tachyzoites vs bradyzoites.

With this sentence we aimed to describe the results and not to make a statement. We toned down the sentence to “… minimal lethal concentrations appear generally higher in bradyzoites… “ line 344 to 347. We also added a line 1 µM in the charts to facilitate easier comparison of compound efficacies.

(6) It would be helpful to readers to include the structures of hit compounds in the figures (perhaps as part of Figure 3).

This is a good idea and would improve the manuscript. To not overburden figure 3 we added structures to Fig S3.

(7) Page 8. "Infected monolayers were treated for three hours with a 3-fold of respective IC50 concentrations". 3-fold higher than IC50 concentrations? This isn't clear.

Thank you for noticing this: We clarified the sentence and also corrected the concentration, corresponding to five times their IC50s as stated in the methods section: “Infected monolayers were treated for three hours with compound concentrations five times their respective IC₅₀ values or the solvent DMSO.” Line 374 - 376

(8) Page 9. "buparvaquone, which we found to be dually active against T. gondii tachyzoites and bradyzoites, targets the bc1-complex in Theileria annulata (McHardy et al. 1985) and Neospora caninum (Müller et al. 2015) and was recently found active against T. gondii tachyzoites (Hayward et al. 2023)." The latter paper showed that buparvaquone targets the bc1 complex in T. gondii tachyzoites as well.

Yes, it was found to inhibit O2 consumption rate in tachyzoites. We changed the sentence accordingly. Line 407 to 411.

(9) Page 9. "Anaplerotic substrates were also affected by all three treatments, most notably a strong accumulation of aspartic acid." It is interesting that the M+3 isotopologue of aspartate (presumably synthesised from pyruvate) is the predominant form (rather than the M+2 and M+4 isotopologues that would derive from the TCA cycle, and as the diagram in Figure 4A seems to suggest). Given that aspartate is a precursor of pyrimidine biosynthesis that is upstream of the DHODH reaction, it is conceivable that its accumulation is related to the depletion of pyrimidine biosynthesis (so would tie into the point about the accumulation of DHO and CarbAsp noted earlier in the paragraph).

Yes, we assume the same. We altered the text and summarized the changes in Asp as a result of DHOD inhibition, as we also already do in the next paragraph using ¹⁵N-glutamine labelling. Line: 416 - 418

(10) Figure 6 and Page 10. Regarding the metabolomic experiments that show increased levels of acyl-carnitines. The authors note that "Since [beta-oxidation] is thought to be absent in T. gondii, we attribute these changes to inhibition of host mitochondria". This is conceivable, although the T. gondii genome does encode homologs of the proteins necessary for beta-oxidation (e.g. see PMID 35298557). If the carnitine is coming from host mitochondria, is host contamination a concern for interpreting the metabolomic data? Or do the authors think that parasites are scavenging carnitine from host cells? It is curious that the carnitine accumulation is observed in parasites treated with buparvaquone (and MMV1028806) but not atovaquone, even though buparvaquone and atovaquone (and possibly MMV1028806) target the same enzyme. Do the authors have any thoughts on why that might be the case?

Yes, thank you for raising this point. We changed the discussion elaborating on this and included the debated presence of beta-oxidation: line 640: “We also detect elevated levels of acyl-carnitines in BPQ and MMV1028806 treated bradyzoites. These molecules act as shuttles for the mitochondrial import of fatty acids for β-oxidation. However, this pathway has not been shown to be active and is deemed absent in T. gondii (35298557, 18775675). The presence of acyl-carnitines in bradyzoites might reflect import from the host. It is conceivable that their elevation in response to buparvaquone and MMV1028806 indicates compromised functionality of the host bc1-complex and subsequently accumulating β-oxidation substrates. Indeed, BPQ has a very broad activity across Apicomplexa (Hudson et al. 1985) and kinetoplastids (Croft et al. 1992).“ Regarding the existence of beta-oxidation: some potential enzymes might be conserved, but those could in part take part in branched chain amino acid degradation pathways. On a separate note: we looked extensively on beta-oxidation using stable isotope labelling and became convinced that any activity occurred in the host cell only but not in the parasite (unpublished).

(11) Page 11. "the mitochondrial [electron] transport chain in bradyzoites".

Corrected.

(12) Figure S6B. Were these optimization experiments performed in tachyzoites or bradyzoites? If the former, and given that bradyzoites have apparently smaller amounts of ATP per parasite (Figure 7C), are these values in the linear range for 10^5 bradyzoites?

Yes, we do think that the assay remains linear for these lower concentrations. Tachyzoites give a linear response starting from 10^3 parasites per sample. In the actual experiment we used 10^5 parasites, both tachyzoites and bradyzoites. Under the tested conditions bradyzoites maintain 10% of the ATP pools of tachyzoites, which should be well within the linear range of the assay. Also in Atovaquone-treated bradyzoites ATP concentration could be lower to 10% and still remain in the linear range of the assay. For practical reasons, we simply acknowledge this limitation and consider it acceptable within the scope of this study.

Reviewer #3 (Recommendations for the authors):

Major comments

(1) The authors should provide a negative control for the experiment on Figure 5. I would suggest doing the same experiment with an inhibitor that has no effect on mitochondrial potential.

We addressed this criticism by repeating the assay on tachyzoites and additionally including inhibitors that do not have the mitochondrial electron transport chain as their primary target (Pyrimethamine, Clindamycin, 6-Diazo-5-oxo-L-norleucin). The results are summarized in the supplementary Fig S5, line 445 – 449) and show that there is no effect of these inhibitors on the mitochondrial membrane potential. This supports the specificity of the assay and suggests that MMV1028806 and BPQ indeed target a mitochondrial process in this stage. Also, in this repetition ATQ, BPQ and MMV1028806 did significantly deplete the Mitotracker signal.

(2) Figure 5 - Did the authors perform this experiment in 3 biological replicates? This requires clarification of the figure legend.

No, we did not perform the experiment in 3 biological replicates. After establishing the assay thoroughly, we performed it once on tachyzoites and bradyzoites. The sampling was done on every vacuole we encountered during microscopy going through the slide from left to right. That is the reason the sample size varies from treatment to treatment. The sample size is mentioned in the caption of figure 5. However, we repeated the experiment with additional controls (see Fig. S5), which showed that the Mitotracker signals were significantly depleted in a very similar manner in ATQ, BPQ and MMV1028806 treated parasites.

(3) The authors identify that MMV1028806 has bc1-complex as the main target. I suggest that they should perform a complex III activity assay to affirm this. Also, it would be good to test if other mETC complexes are affected by this compound to prove its specificity. There is only one paper showing complex III activity in tachyzoites (PMID:37471441) and no papers in bradyzoites. So if the authors cannot do this assay, I suggest that they should change the text indicating that bc-1 complex could be the main target of the compound but more experimental validation is needed.

We hope to have satisfied the reviewer’s request by performing an oxygen consumption assay on tachyzoites. Together with metabolic profiling and labelling data, this shows that both upstream and downstream processes are impacted by MMV1028806 and strongly suggest the bc1-complex as a target (Fig 5E).

(4) Figure S5 - Are the differences shown in the EM experiment statistically supported?

We analyzed 28 images and measured the areas in 12 to 26 images. We substituted the table of means in Fig S6B by a graph showing individual values. These areas are indeed statistically different between DMSO and ATQ / MMV treated parasites. We changed the wording in the results section accordingly “Analysis by thin section electron microscopy revealed a largely unaffected sub-mitochondrial ultrastructure but the areas of mitochondrial profiles were changed in comparison to control after exposure with ATQ and MMV1028806 but not with BPQ (Fig. S6)“. The description of Fig S6B was changed to “(B) Measured areas of mitochondrial profiles from 21, 12, 15 and 26 images showing DMSO, ATQ, BPQ and MMV1028806 treated parasites (* denotes p < 0.05 in Mann-Whitney tests)”.

Minor comments:

(1) What was the criteria to choose the example compounds in Figure 1B and 1D? The authors should clarify this in the text.

These graphs are shown for illustrative purposes and were chosen based on their display of different drug efficacies. We considered this helpful for interpreting the screening data.

(2) Figure 2G - add statistical analysis.

We added Mann-Whitney tests and updated the figure legend and results text accordingly in line 344 – 347.

(3) The authors should provide more insights in the discussion about why this new compound is the next step in drug discovery compared to atovaquone or burvaquone - for example, do you expect better availability in the brain, etc.

We used MMV1028806 and the other hits ATQ and BPQ to make the point that the bc1-complex is a good target in bradyzoites that allows curative treatment. We do not suggest that the compound itself is a good starting point. We point to other actively developed candidates such as ELQ series in the discussion, line 719.

(4) Scale bars in Figure 5 should be aligned and have equal thickness.

We re-formatted the scale bars and aligned them when not obscuring parasites.

(5) The authors should be consistent with font sizes and styles in all the figures.

We adjusted the font styles to match each other.

Author response:

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

Both reviewers indicated broad approval of the revised work, for which we are grateful.

Reviewer #1 requested no further changes.

Reviewer #2’s Public review states:

The authors indicate that the adaptors of inflammatory signalosomes act as energy reservoirs for signal amplification. This is not demonstrated, but it is assumed that the energy stored in the supersaturated state is released upon polymerization.

The “assumed” link between supersaturation and energy release is in fact a thermodynamic necessity. Supersaturation is, by definition, a high free energy state. Our data shows that triggering nucleation via optogenetics results in an immediate avalanche of polymerization and cell death. This is not an assumption; it is a direct observation of work performed by the system when the kinetic barrier is removed.

Reviewer #2 recommended:

Ideally, signal amplification could be tested by determining the levels of the final product, e.g., cytokines, activated caspases...

We did measure CASP3/7 activation, demonstrating a correlation with supersaturation of upstream adaptors. We do agree however that measuring the levels of other signaling products, including for each of the supersaturated pathways, would strengthen our claims. This will be the subject of future work.

The authors indicate a significant anticorrelation between the saturating concentrations and the transcript abundances (Figure 2B), reporting an R = -0.285.

This is correct… no change appears to be requested or warranted.

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

This is a high-quality and extensive study that reveals differences in the self-assembly properties of the full set of 109 human death fold domains (DFDs). Distributed amphifluoric FRET (DAmFRET) is a powerful tool that reveals the self-assembly behaviour of the DFDs, in non-seeded and seeded contexts, and allows comparison of the nature and extent of self-assembly. The nature of the barriers to nucleation is revealed in the transition from low to high AmFRET. Alongside analysis of the saturation concentration and protein concentration in the absence of seed, the subset of proteins that exhibited discontinuous transitions to higher-order assemblies was observed to have higher concentrations than DFDs that exhibited continuous transitions. The experiments probing the ~20% of DFDs that exhibit discontinuous transition to polymeric form suggest that they populate a metastable, supersaturated form in the absence of cognate signal. This is suggestive of a high intrinsic barrier to nucleation.

Strengths:

The differences in self-assembly behaviour are significant and likely identify mechanistic differences across this large family of signalling adapter domains. The work is of high quality, and the evidence for a range of behaviours is strong. This is an important and useful starting point since the different assembly mechanisms point towards specific cellular roles. However, understanding the molecular basis for these differences will require further analysis.

An impressive optogenetic approach was engineered and applied to initiate self-assembly of CASP1 and CASP9 DFDs, as a model for apoptosome initiation in these two DFDs with differing continuous or discontinuous assembly properties. This comparison revealed clear differences in the stability and reversibility of the assemblies, supporting the hypothesis that supersaturation-mediated DFD assembly underlies signal amplification in at least some of the DFDs.

The study reveals interesting correlations between supersaturation of DFD adapters in short- and long-lived cells, suggestive of a relationship between the mechanism of assembly and cellular context. Additionally, the comprehensive nature of the study provides strong evidence that the interactions are almost all homomeric or limited to members of the same DFD subfamily or interaction network. Similar approaches with bacterial proteins from innate immunity operons suggest that their polymerisation may be driven by similar mechanisms.

Weaknesses:

Only a limited investigation of assembly morphology was conducted by microscopy. There was a tendency for discontinuous structures to form fibrillar structures and continuous to populate diffuse or punctate structures, but there was overlap across all categories, which is not fully explored.

We agree that an in-depth exploration of aggregate morphology would be interesting, but we feel it has limited relevance to the central findings of the manuscript. Our analysis established a relationship between discontinuous transitions and ordering based on the assumption that ordered assembly by DFDs involves polymerization, for which there is much precedent in the literature. Nevertheless, polymers of similar structure can form with different kinetics and hence, polymerization does not by itself imply an ability to supersaturate. We see this empirically in the “fibrillar” column in Fig. 1B. We have now elaborated this important point more fully in the relevant results section and in the discussion. Only five of the 108 DFDs in Fig. 1B warrant additional explanation. CASP4^CARD and IFIH1^tCARD lacked AmFRET but formed puncta; this could result from interactions with endogenous structures or condensates. DAPK1^DD and UNC5A^DD were classified as continuous (low) and fibrillar, but their AmFRET values are in fact higher than monomer control revealing that the fibrils simply comprise a small fraction of the protein. The puncta of UNC5A^DD additionally do not resemble the fibrillar puncta of other DFDs; we suspect it may be a false-positive resulting from localization to mitochondrial or other intracellular membranes. Finally, CASP2^CARD was inadvertently classified as punctate; this turns out to have been a technical artifact that has now been corrected (the fibrils wrapped around the cell perimeter to form ring-like puncta with anomalously low aspect ratios). We have now updated the methods section describing manual validation of our automated classification procedure, including which samples required reclassification. We have also now included all microscopy data in the public repository accompanying this manuscript.

The methodology used to probe oligomeric assembly and stability (SDD-AGE) does not justify the conclusions drawn regarding stability and native structure within the assemblies.

The reviewer is correct that SDD-AGE does not provide evidence against non-amyloid misfolding. It merely provides evidence that the DFDs are not forming amyloid (which are characteristically sarkosyl resistant). We have revised the sentence and further clarified that the distinction with amyloid specifically is important because amyloid is the only known form of ordered assembly (other than DFD polymers) with a nucleation barrier large enough to support deep supersaturation. Together with the series of interfacial mutants tested (and shown to impede assembly in all cases), the lack of sarkosyl-resistance provides evidence that the discontinuous DFDs are assembling through canonical DFD subunit interfaces.

The work identifies important differences between DFDs and clearly different patterns of association. However, most of the detailed analysis is of the DFDs that exhibit a discontinuous transition, and important questions remain about the majority of other DFDs and why some assemblies should be reversible and others not, and about the nature of signalling arising from a continuous transition to polymeric form.

We focused on discontinuous DFDs because this property allows for executive control over their respective pathways. They make signaling switch-like, which we argue is essential for innate immune responses. By contrast, and as illustrated in Figure 6D, supersaturation is required for a DFD to drive its own polymerization -- hence activation for a continuous DFD must be stoichiometrically coupled either with D/PAMP binding or positive feedback from downstream or orthogonal processes. We consider the principles underlying such regulation of signaling to be better established and understood than supersaturation, and hence built our narrative for this manuscript around the latter. Our original text addresses the fact that only a small fraction of DFDs are discontinuous. Specifically, this is expected in light of the fact that a) only one supersaturated DFD is needed to make a signaling pathway switch-like, and b) every supersaturated DFD renders the cell susceptible to spontaneous death. Evolution should therefore limit supersaturation to only the highly connected DFDs (i.e. adaptors), which is what is seen. In this view, the many nonsupersaturable DFDs have evolved to accessorize the central supersaturable DFDs with various sensor and effector modules. Our revised text attempts to further clarify this perspective.

Some key examples of well-studied DFDs, such as MyD88 and RIPK,1 deserve more discussion, since they display somewhat surprising results. More detailed exploration of these candidates, where much is known about their structures and the nature of the assemblies from other work, could substantiate the conclusions here and transform some of the conclusions from speculative to convincing.

We were likewise initially surprised about the inability of MyD88 and RIPK1 to supersaturate. We have now elaborated in the Discussion how our findings can be rationalized by the apparent supersaturability of other adaptors in MyD88 and RIPK1 signaling pathways. We additionally discuss prior evidence that MyD88 may indeed be supersaturable, and how our experimental system could have led to a false positive in the unique case of MyD88.

The study concludes with general statements about the relationship between stochastic nucleation and mortality, which provide food for thought and discussion but which, as they concede, are highly speculative. The analogies that are drawn with batteries and privatisation will likely not be clearly understood by all readers. The authors do not discuss limitations of the study or elaborate on further experiments that could interrogate the model.

We have now added to the discussion a section on the limitations of our study. We appreciate that our use of “privatisation” was confusing and have omitted it. However, we consider the battery analogy to accurately convey the newfound function of DFDs and anticipate that this analogy will ultimately prove valuable for biologists. To facilitate comprehension, we have now broadened our description of phase change batteries in the introduction.

Reviewer #2 (Public review):

Summary:

The manuscript from Rodriguez Gama et al. proposes several interesting conclusions based on different oligomerization properties of Death-Fold Domains (DFDs) in cells, their natural abundance, and supersaturation properties. These ideas are:

(1) DFDs broadly store the cell's energy by remaining in a supersaturated state;

(2) Cells are constantly in a vulnerable state that could lead to cell death;

(3) The cell's lifespan depends on the supersaturation levels of certain DFDs.

Overall, the evidence supporting these claims is not completely solid. Some concerns were noted.

Strengths:

Systematic analysis of DFD self-assembly and its relationship with protein abundance, supersaturation, cell longevity, and evolution.

Weaknesses:

(1) On page 2, it is stated, "Nucleation barriers increase with the entropic cost of assembly. Assemblies with large barriers, therefore, tend to be more ordered than those without. Ordered assembly often manifests as long filaments in cells," as a way to explain the observed results that DFDs assemblies that transitioned discontinuously form fibrils, whereas those that transitioned continuously (low-to-high) formed spherical or amorphous puncta. It is unlikely to be able to differentiate between amorphous and structured puncta by conventional confocal microscopy. Some DFDs self-assemble into structured puncta formed by intertwined fibrils. Such fibril nets are more structured and thus should be associated with a higher entropic cost. Therefore, the results in Figure 1B do not seem to agree with the reasoning described.

The formation of microscopically visible elongated structures necessitates ordering on the length scale of 100s of nanometers. Otherwise surface tension would favor rounded aggregates. Conventional confocal microscopy is in fact well-suited and widely used to distinguish ordered from disordered assemblies in cells based on this principle.1,2 We are unaware of any examples of isolated DFDs forming regular polymers that manifest as round puncta or nets. The reviewer may be referring to full-length ASC, which forms a roughly spherical mesh of filaments because it has two DFDs joined by a flexible linker. This is not applicable to our analysis with single DFDs. Single DFDs polymerize in effectively one dimension; hence a spherical punctum formed by a single DFD can only happen through noncanonical interactions or clustering of small filaments, both of which reduce order relative to long filaments.

(2) Errors for the data shown in Figure 1B would have been very useful to determine whether the population differences between diffuse, punctate, and fibrillar for the continuous (low-to-high) transition are meaningful.

We have now performed two statistical analyses to address this. First, using Fisher’s exact test, we observe a highly significant association between the DAmFRET and morphology classifications (p-value: 0.0001). Second, to specifically address whether the continuous (low to high) category has a preferred morphology, we applied an Exact Multinomial Test using the total frequencies of each morphology. This test revealed that all categories are significantly enriched for particular morphologies, as now indicated in the figure and legend.

(3) A main concern in the data shown in Figure 1B and F is that the number of counts for discontinuous compared to continuous is small. Thus, the significance of the results is difficult to evaluate in the context of the broad function of DFDs as batteries, as stated at the beginning of the manuscript.

Fig. 1B simply reports the numerical intersections between fluorescence distribution classifications and DAmFRET classifications. In Fig. 1F, our use of the chi-square test is justified by a sufficiently large sample size. Nevertheless, we obtain similar results with Fisher's exact test that accounts for smaller sample size (Odds Ratio: 75.0, P-value: < 0.0001). See also our response to the related critique by Reviewer 1 regarding the small number of discontinuous DFDs.

(4) The proteins or domains that are self-seeded (Figure 1F) should be listed such that the reader has a better understanding of whether domains or full-length proteins are considered, whether other domains have an effect on self-seeding (which is not discussed), and whether there is repetition.

We define and consistently use “DFDs” to refer to domains, and “FL” or “DFD-containing protein” to refer to FL proteins. The Figure 1 title and corresponding section title both indicate the data refer to “DFDs”. The text callout for Figure 1F also directs readers to Table S1 where we believe the self-seeding results and details of constructs are clearly presented. There is no repetition. We have modified the legend to clarify that “Each DFD was co-expressed with an orthogonally fluorescent μNS-fused version of the same DFD.” We did not systematically evaluate seeding of FL proteins. We did however previously test self-seeding on seven representative FL proteins, and have now included those data in a new supplemental figure (S5). In short, only FL proteins with discontinuous distributions are self-seedable. These are limited to adaptors that had discontinuous seedable DFDs, revealing no adverse effect of FL protein context on seedability of adaptors (unlike receptors and effectors).

(5) The authors indicate an anticorrelation between transcript abundance and Csat based on the data shown in Figure 2B; however, the data are scattered. It is not clear why an anticorrelation is inferred.

An anticorrelation is indicated by the clearly placed negative R value at the top of the graph and the figure legend describing the statistical analysis.

(6) It would be useful to indicate the expected range of degree centrality. The differences observed are very small. This is specifically the case for the BC values. The lack of context and the small differences cast doubts on their significance. It would be beneficial to describe these data in the context of the centrality values of other proteins.

The possible range of centrality scores is 0 - 1, where 1 represents a protein interacting with every other protein in the network (degree centrality) or is on the shortest path between every other pair of proteins in the network (betweenness centrality). The expected range is difficult to address, as centrality values strongly depend on the size and function of the network. We considered that the SAM domain network could provide the most relevant comparison to the DFD network, as SAM domains resemble DFDs in size and structure, function heavily in signaling, are comparably numerous (76 in humans), and many of them form homopolymers (but importantly of a geometry that does not support nucleation barriers). We found that SAM domains have much lower betweenness centrality in their physical interaction network as compared to discontinuous DFDs (p = 0. 0003) while their degree centrality is not significantly different (Figure S3F). Nevertheless, we stress that what matters for our conclusion is that the continuous and discontinuous values are significantly different among DFDs. Since there is a large overlap in the distributions of centrality scores between the two classes of DFDs, we performed a more robust permutation test with the Mann Whitney U statistic and n = 10000. These tests reiterated that continuous and discontinuous DFDs have significantly different centrality scores (Degree centrality p = 0.008; Betweenness centrality p = 0.028) (Figure S3E).

(7) Page 3 section title: "Nucleation barriers are a characteristic feature of inflammatory signalosome adaptors." This title seems to contradict the results shown in Figure 2D, where full-length CARD9 and CARD11 are classified as sensors, but it has been reported that they are adaptor proteins with key roles in the inflammatory response. Please see the following references as examples: The adaptor protein CARD9 is essential for the activation of myeloid cells through ITAM-associated and Toll-like receptors. Nat Immunol 8, 619-629 (2007), and Mechanisms of Regulated and Dysregulated CARD11 Signaling in Adaptive Immunity and Disease. Front Immunol. 2018 Sep 19;9:2105. However, both CARD9 and CARD11 show discontinuous to continuous behavior for the individual DFDs versus full-length proteins, respectively, in contrast to the results obtained for ASC, FADD, etc.

We rigorously counter the inconsistent usage of the term “adaptor” in the signalosome literature by quantifying the centrality of each protein in the physical interaction network of DFD proteins. Such analysis shows that BCL10, which is also described as an adaptor, is the more central member of the CARD9 and CARD11 (CBM signalosome) pathways, and is therefore more “adaptor-like”. We have now elaborated this view in the text.

FADD plays a key role in apoptosis but shows the same behavior as BCL10 and ASC. However, the manuscript indicates that this behavior is characteristic of inflammatory signalosomes. What is the explanation for adaptor proteins behaving in different ways? This casts doubts about the possibility of deriving general conclusions on the significance of these observations, or the subtitles in the results section seem to be oversimplifications.

We agree that our initial presentation of these results and brief description of each protein’s function was insufficient to fully justify our conclusions. We have now elaborated that while FADD was historically considered an adaptor of extrinsic apoptosis, it is now appreciated as a pleiotropic molecule with both anti- and pro-inflammatory signaling functions. FADD’s pro-inflammatory roles include inflammasome activation and activating NF-kB through the FADDosome. We have now revised our section headings to avoid oversimplification.

(8) IFI16-PYD displays discontinuous behavior according to Figure S1H; however, it is not included in Figure 2D, but AIM 2 is.

We only tested a subset of FL proteins spanning different functions within diverse signalosomes. IFI16 was not included. Hence it could not be meaningfully included in Fig. 2D.

(9) To demonstrate that "Nucleation barriers facilitate signal amplification in human cells," constructs using APAF1 CARD, NLRC4 CARD, caspase-9 CARD, and a chimera of the latter are used to create what the authors refer to as apoptsomes. Even though puncta are observed, referring to these assemblies as apoptosomes seems somewhat misleading. In addition, it is not clear why the activity of caspase-9 was not measured directly, instead of that of capsae-3 and 7, which could be activated by other means.

We agree that describing our chimeric assemblies as “apoptosomes” could be misleading, and have now refrained from doing so. We measured caspase-3/7 instead of caspase-9 for purely technical reasons -- we were unable to find any reliable caspase-9 activity assays that were also compatible with our optogenetic and imaging wavelengths. In any case, our data with the widely used caspase3/7 reporter dyes confirm comparably effective signal propagation from the CASP9 versions to their relevant endogenous substrate for apoptotic signaling (pro-caspase-3/7). The subsequent differences in cell death efficiency between the two versions of CASP9 (Fig. 3E) cannot be attributed to indirect effects of blue light stimulation, because both versions received the same treatment. Note our stated justification for using these DFDs in the HEK293T background is that these cells lack NLCR4 and CASP1 proteins and therefore the activity we measure is due to the direct optogenetic activation.

The polymerization of caspase-1 CARD with NLRC4 CARD, leading to irreversible puncta, could just mean that the polymers are more stable. In fact, not all DFDs form equally stable or identical complexes, which does not necessarily imply that a nucleation barrier facilitates signal amplification. Could this conclusion be an overstatement?

Figure 3C shows that the polymers don’t simply persist following the transient stimulus -- they continue to grow. That is, the soluble protein continues to join the polymers for a net increase even though there is no longer a stimulus directing them to do so. This means the drive to polymerize is independent of the stimulus, i.e. the protein is supersaturated. In the absence of supersaturation, a difference in stability would simply change the rates at which the polymers shrink. That we see continued growth instead of shrinkage therefore cannot be explained just by a difference in stability. Nevertheless, the reviewer’s critique caused us to realize that increased persistence of the CASP1CARD polymers could contribute to signal amplification independently of supersaturation if they act catalytically (i.e. where each polymerized CASP9 subunit sequentially activates multiple CASP3/7 molecules), and we had not adequately considered this. Unfortunately, the relevant experimentalist has now moved on from the lab leaving us unable to conduct the necessary experiments to resolve these two effects in a timely fashion. Consequently, we have now tempered our interpretation of these data. 

(10) To demonstrate that "Innate immune adaptors are endogenously supersaturated," it is stated on page 5 that ASC clusters continue to grow for the full duration of the time course and that AIM2-PYD stops growing after 5 min. The data shown in Figure 4F indicate that AIM2-PYD grows after 5 mins, although slowly, and ASC starts to slow down at ~ 13 min. Because ASC has two DFDs, assemblies can grow faster and become bigger. How is this related to supersaturation?

That AIM2-PYD assemblies appear to grow somewhat (although not significantly statistically) would be consistent with AIM2-PYD’s sequestration into the growing ASC clusters. All that matters for our conclusion regarding ASC is that ASC assemblies grow following cessation of the stimulus, which we now describe quantitatively. Supersaturation is defined as the ratio of total concentration to saturating concentration, which is an equilibrium property. For a given protein concentration, the presence of two DFDs, each contributing their own interactions to overall stability of the assembly, will increase supersaturation relative to the individual DFDs. Importantly, growth will not occur if the protein concentration lies below its C_sat, no matter how many DFDs it has.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

It isn't clear what is implied by the final sentence of the Abstract. Some of the conclusions have a speculative tone and would be better described in less certain terms. The final sentence of the abstract should be omitted.

We have revised the abstract to add appropriate nuance but consider the final sentence to be both justified by our data and important to convey our findings to a broad audience.

How does the size and nature of the seed influence the outcome of these DFD interactions? Although some non-seeded experiments are described, the majority of the results are derived from seeded experiments. Further details about the seeds should be included. How is the size of the nucleus controlled, and will seeds of smaller or larger size generate the same pattern of results?

This is a very important question! The seeds comprised genetic fusions of each DFD to a condensate-forming domain, as described. While this system is insufficient to explore the size-dependence of nucleation, we are developing tools to do exactly that, for example our recently published multivalent nanobody against mEos3,[3] wherein we piloted its use to compare the size-dependence of ASC versus amyloid nucleation. Much further work will be needed to fully utilize this approach for the question of interest, and that is the subject of ongoing but open-ended work in the lab.

What is the implication of the observation that only ~20% of the DFDs exhibited a discontinuous transition from no to high AmFRET signal? Further discussion of the DFDs that exhibit a continuous transition would enrich the manuscript.

We consider the relationship to mortality important for understanding this observation. In the discussion we now explain that each supersaturated protein in a death-inducing pathway imposes a risk of unintentional death. We speculate that evolution therefore minimizes the number of supersaturated DFDs by restricting them to central nodes in the network. That way, a small number of supersaturable DFDs can be continuously “repurposed” with new receptor proteins for each D/PAMP. Additionally, as stated in our response to the related critique, we felt it was important to focus this manuscript on the novel concept of functional supersaturation necessarily at the expense of signaling regulation through better understood mechanisms.

Were the initial experiments with DFDs unseeded (Figure S1, F-G)? Clarify this in the text. The morphologies of all the subcellular assemblies appear similar. It is not possible to distinguish between long filaments and spherical or amorphous puncta (Figure S1F-G). Higher magnification images that allow evaluation and comparison of morphology should be provided.

The initial experiments were unseeded, as now clarified in the legend. We believe there was a misinterpretation resulting from both panels (S1F and G) showing fibrillar examples. To clarify, we have now added panel S1H showing representative DFDs classified as “punctate”, which we hope the reviewer agrees are clearly distinct from fibrillar.

The ASC and CARD14 assemblies in Figure S1G show very distinct fibrillar structures emerging from the mNS-DFD seeds. Please provide further explanation of the nature of these. Do these resemble ASC and CARD assemblies generated as a result of native stimuli rather than mNS-DFD seeds?

The μNS-DFD puncta contain numerous seeding competent sites, which presumably causes multiple fibrils to initiate and emanate from them. This and potential bundling of these fibrils produces the star-like shape. We have no reason to believe the internal structure of these fibers differs from native signalosome assemblies. For example, point mutations at native subunit interfaces that were previously shown to disrupt fibrilization and signaling likewise disrupt assembly in our DAmFRET experiments (Figure S2A). To our knowledge there exist no examples of high-resolution DFD fibril structures that were induced by native stimuli. However, recent work using super-resolution imaging confirmed that nigericin-triggered endogenous ASC specks comprise a network of filaments that superficially resembles our star-like assemblies.[4]

Figure S2B is presented as evidence that assembly is mediated by native-like interfaces rather than amyloid-like misfolding. These SDD-Age gels cannot be used to infer a native-like structure for the protein within the assemblies, only that the assemblies are (mostly) solubilised by incubation with sarkosyl. Many misfolding but non-amyloid-structure assemblies could be consistent with these results. Additionally, several of the samples appear to show insoluble aggregates within the wells, which could also be consistent with amyloid-type structures. What is the nature of these aggregates? Why is the NLRP3PYD sample so much more intense than the others? Why was FL-ZBP1 included when it does not contain a DFD? Why were no sarkosyl-resistant assemblies observed with RIPK3-RHIM when this is known to be highly amyloidogenic?

ZBP1 and RIPK3^RHIM were one of multiple proteins inadvertently included on the complete gel shown in the original figure that is not relevant to the manuscript; we have now spliced out these unnecessary lanes (indicated with dashed lines) to avoid confusion. We have found that the specific fragment of RIPK3^RHIM used in this experiment -- residues 446-464 -- does not allow for robust amyloid formation. We believe this is a steric artifact due to its small size (19 residues) relative to the fused mEos3, because a longer fragment (446-518) forms amyloid robustly. However the latter construct was not available at the time this experiment was done. Nevertheless, another known amyloid protein, RIPK1^RHIM, does show the expected smears on this gel and suffices for the positive control for amyloid. We do not understand why the NLRP3^PYD sample is more intense than the others. However, this anomaly does not impact our conclusion that DFDs do not form sarkosyl-resistant smears that would be indicative of amyloid.

Expand on the concept of autoinhibited oligomerisation. Is this due to structural features? What might be the advantage of autoinhibited oligomerisation for these DFDs?

We have elaborated on this section in the results.

End of page 3, which "former set of adaptors" are referred to here? This is ambiguous.

We have replaced “former” with “innate immune”.

Page 5, the authors state that a kinetic barrier governs the activity of inflammatory signalosomes. While under the circumstances generated in this particular system, there is a kinetic barrier to the formation of large fibrillar complexes, can the same be said to be true in cells that respond to signals? They experience a specific triggering event. This should be redrafted to distinguish between the specific trigger in cells (downstream of a binding-driven event) and the kinetic barrier to self-association observed in this model system.

Yes, our findings establish that a kinetic barrier governs signalosome activation. By engineering a triggering event that is more specific than natural triggering events (see Figure 3), we exclude the possibility that the cell first responds to the signal to create conditions that stabilize inflammasome formation. This means that regardless of what may happen with a natural trigger, the driving force for assembly clearly pre-exists and is therefore held in check by a kinetic barrier.

On page 6, the statement "...lifespan may be limited by the thermodynamic drive for inflammatory signal amplification" is not clear. While this is strictly true following the initial triggering event, isn't lifespan limited by the stochastic activation? These very general statements stray beyond what can be substantiated on the basis of the data presented here.

We believe the source of confusion here was our misuse of the term “lifespan”. We have now replaced it with “life expectancy”, which we believe is substantiated by our statements as written.

Overall, the work presents a compelling, comprehensive analysis of the seeded self-assembly of DFDs. It identifies distinct properties for assembly of these domains that may underlie their particular physiological roles. However, some of the statements are quite general and not substantiated.

Page 6. Is "end cell fate" the intended phrase?

We have revised the phrase.

The data regarding conservation of DFD-like modules and activity is interesting and probably deserves inclusion. However, without substantial evidence of expression levels (i.e., results) and a more complete understanding of these other systems, the statement "These results suggest that the function of DFDs as energy reservoirs preceded the evolution of animals" appears as an over-reach.

We demonstrated that sequence-encoded nucleation barriers of DFDs are shared across animal signalosomes (human, zebrafish, sponge). This is not trivial as such nucleation barriers are uncommon even among targeted screens of prion-like proteins.5 Therefore, they appear to have existed in the basal animal. We have now omitted the data concerning bacterial DFDs as these systems are indeed much less understood, and the concerned pathways lack the tripartite architecture of animal signalosomes. We therefore revised the sentence in question by replacing “evolution” with “radiation”.

Only a small number of DFDs exhibit this behaviour, so why is the conclusion drawn that energy storage for on-demand signalling may be the principal ancestral function of DFDs?

The totality of the data supports this conclusion. Briefly (but elaborated in the text), 1) intrinsic nucleation barriers are unusual even among self-associating proteins, the vast majority of which (e.g. condensates) would suffice for the only other major function ascribed to DFDs -- bringing effectors close enough for proximity-dependent activation (which has been repeatedly demonstrated in DFD-replacement experiments), 2) nucleation barriers are nevertheless conserved in innate immune signaling pathway, 3) that they are limited to approximately one DFD in each pathway is consistent with evolutionary selection to minimize accidental death.

Are there any other adapters like MyD88 that are inconsistent with this hypothesis? Are any others known to be controlled by oligomer formation? How strong is the evidence for hexameric oligomers? If there is a threshold size for oligomers, how does this differ from a stable seed/nucleus that triggers assembly, as in the discontinuous transition?

These are all good questions related to critiques that we have now addressed.

The use of the term "privatisation" is likely not consistently understood across the community and should be explained. Is it simply meant to imply independent operation? How is it actually different from other forms of deployment of DFDs that exhibit continuous assembly? Are they not also independent? What is implied by the opposite of privatisation here? The term may introduce ambiguity in this context.

We have now omitted this term.

Is there strong evidence that well-validated physiologically relevant LLPS systems exhibit supersaturation at concentrations that are very different from those of the DFDs examined in this study?

No, and this is a major point. As discussed in the text (with references), LLPS is incompatible with cell-wide supersaturation to a comparable magnitude as crystalline transitions, which precludes them from driving signal amplification. This helps to explain why the active state of DFD assemblies is ordered, when it has been repeatedly demonstrated that signal propagation itself does not require ordering.

The paragraph discussing TIR domains and functional amyloids would be enhanced with a comparison of amyloid systems where seeded nucleation results in assembly of a polymer with significant conformational change in the constituent monomers.

We do not yet understand how DFDs (and TIR domains) in some cases exhibit amyloid-like nucleation barriers without overt conformational differences between monomers and polymers. Work is underway in the lab to test specific hypotheses, but such discussion would be too speculative for the present paper.

The statement "High specificity also insulates pathways from each other" should be elaborated to discuss the issue of highly similar monomers that apparently assemble into filamentous forms with minimal structural rearrangement. How is the specificity generated?

We have elaborated the paragraph.

The final paragraph is speculative and utilises language that detracts from the quality and rigour of the study. While important principles have been revealed, more discussion of the limitations of the work would allow readers to evaluate the significance of the study and could be used to effectively stimulate further efforts to study the multiple different mechanisms that underpin critical signalling pathways in innate immunity and control cell fate.

We have now revised the final paragraph and included an extensive discussion of the limitations of the work.

Reviewer #2 (Recommendations for the authors):

(1) For clarity, it would be useful to include the names of the proteins in the bottom table of STable1, and such information at the top and bottom tables can be connected.

We are unable to determine what is meant by this suggestion. Table S1 does not have a “top” and “bottom table”. Every entry in Table S1 and S2 contains the protein name, its most frequently used alias in the literature (when not the official name), and the corresponding Uniprot protein ID.

(2) The language used in the abstract makes analogies between scientific and mundane terms, which compromises clarity. For example, what is meant by the terms shown below?

(a) "......specifically templated by other DFDs....."

We have revised this phrase.

(b) "...function like batteries, storing and converting energy for life-or-death decisions."

Batteries convert chemical energy into electrical energy or thermal energy. What is the electrical energy produced by DFDs? Is there any evidence that DFDs change the temperature of the cells or transfer heat?

We have now included a familiar example of a thermal battery that operates analogously to the manner we show for DFDs. As now elaborated extensively, such batteries operate via a physical rather than chemical process -- a change in the state of matter (solute to crystalline) of a supersaturated “phase change material” (this is an established term). This is exactly what we show is happening for DFDs. While it would be illustrative to measure the heat released upon DFD polymerization in cells, the much faster rate of heat transfer relative to molecular diffusion makes that impossible with present methods. Nevertheless, such measurements are unnecessary because disorder-to-order phase transitions are fundamentally exothermic.

(c) "....privatizing..."

We now avoid this term.

Using appropriate scientific terms to explain the scientific results presented in this manuscript will increase clarity. Analogously, it is difficult to understand what the title of the manuscript means, "Protein phase change batteries..."

We appreciate this critique and have removed “batteries” from the title to make the work more accessible to biologists. However, we reject the implication that such terminology is inappropriate. We presume the reviewer meant “unfamiliar” instead of “inappropriate”. The well-reasoned application of terms from other fields is standard practice and arguably essential to convey new concepts in biology. The modern biology lexicon is built on this. For example, Robert Hooke co-opted “cell” from the architecture of monasteries. More recently cell biologists appropriated “condensates” from soft matter physics. In both cases, the term while initially foreign to biologists usefully introduced a concept that lacked recognized precedent in biology. Similarly, “phase change battery” provides an accurate analogy for the central finding of our work, and we have now elaborated this analogy in the text.

Bibliography

(1) Garcia-Seisdedos, H., Empereur-Mot, C., Elad, N. & Levy, E. D. Proteins evolve on the edge of supramolecular self-assembly. Nature 548, 244–247 (2017).

(2) Alberti, S., Halfmann, R., King, O., Kapila, A. & Lindquist, S. A systematic survey identifies prions and illuminates sequence features of prionogenic proteins. Cell 137, 146–158 (2009).

(3) Kimbrough, H. et al. A tool to dissect heterotypic determinants of homotypic protein phase behavior. Protein Sci. 34, e70194 (2025).

(4) Glück, I. M. et al. Nanoscale organization of the endogenous ASC speck. iScience 26, 108382 (2023).

(5) Posey, A. E. et al. Mechanistic inferences from analysis of measurements of protein phase transitions in live cells. J. Mol. Biol. 433, 166848 (2021).

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

Here the authors attempted to test whether the function of Mettl5 in sleep regulation was conserved in drosophila, and if so, by which molecular mechanisms. To do so they performed sleep analysis, as well as RNA-seq and ribo-seq in order to identify the downstream targets. They found that the loss of one copy of Mettl5 affects sleep and that its catalytic activity is important for this function. Transcriptional and proteomic analyses show that multiple pathways were altered, including the clock signaling pathway and the proteasome. Based on these changes the authors propose that Mettl5 modulate sleep through regulation of the clock genes, both at the level of their production and degradation.

Strengths:

The phenotypical consequence of the loss of one copy of Mettl5 on sleep function is clear and well-documented.

Weaknesses:

The imaging and molecular parts are less convincing.

- The colocalization of Mettl5 with glial and neuronal cells is not very clear

We truly appreciate your suggestion. We repeated the staining experiments. To ensure better results, we tried another antibody of ELAV (mouse) and optimized the experimental conditions. This result has been included in the Figure S1 of the revised version.

- The section on gene ontology analysis is long and confusing

The session is revised for clarity. To get a better flow of logic, we deleted the paragraph which describing the details of Figure S6.

- Among all the pathways affected the focus on proteosome sounds like cherry picking. And there is no experiment demonstrating its impact in the Mettl5 phenotype

Thank you for the comments. The changes of period oppositely at transcriptional versus translational levels puzzled us a while until we found the ubiquitin pathway components changes. The regulation of Period protein degradation by ubiquitin-proteasome pathway has been well documented (Grima et al., 2002; Ko et al., 2002; Chiu et al., 2008). In addition, previous reports indicated that N6 methyladenosine (m6A) regulates ubiquitin proteasome pathway in skeletal muscle physiology (Sun et al., 2023). This information has been included in the revised manuscript in the last paragraph under the title: Mettl5 regulates the clock gene regulatory loop.

Indeed, we haven’t found a proper way to manipulate proteasome levels in genetic tests. Proteasome is a large protein complex which is composed of many subunits. Enhancing the its activity by overexpressing its components was not applicable. Moreover, proteasome has important function during many biological processed. Disrupting its function by simply MG132 treatment which we tried results in lots of side effects.

In this study, we also noticed the codon usage alteration caused by mettl5 mutant. Please refer to the answers to the following question for details. Previous reports also found the regulation of mettl5 on translation in other systems (Rong et al, 2020; Peng et al., 2022). Based on these analyses, it is possible that both the regulation on translation and protein degradation contributed the period protein upregulation found in mettl5 mutant. This idea has been included in the Discussion session of the revised manuscript.

References

Sun J, Zhou H, Chen Z, et al. Altered m6A RNA methylation governs denervation-induced muscle atrophy by regulating ubiquitin proteasome pathway. J Transl Med. 2023;21(1):845. Published 2023 Nov 23. doi:10.1186/s12967-023-04694-3

Grima, B. et al. The F-box protein slimb controls the levels of clock proteins period and timeless. Nature 420, 178–182 (2002).

Ko, H. W., Jiang, J. & Edery, I. Role for Slimb in the degradation of Drosophila period protein phosphorylated by doubletime. Nature 420, 673–678 (2002).

Chiu, J. C., Vanselow, J. T., Kramer, A. & Edery, I. The phosphooccupancy of an atypical SLIMB-binding site on PERIOD that is phosphorylated by DOUBLETIME controls the pace of the clock. Genes Dev. 22, 1758–1772 (2008).

- The ribo seq shows some changes at the level of translation efficiency but there is no connection with the Mettl5 phenotypes. In other words, how the increased usage of some codons impact clock signalling. Are the genes enriched for these codons?

Thank you for raising this point. In our analysis, we observed an increased usage of the codons for Asp in the Mettl5 mutant. Prior work has reported a possible connection between codon usage and per protein activity. In the report, a per version with optimized codon cannot rescue circadian rhythmicity caused by per mutant, in contrast to WT version (Fu J et al. 2016). Further study indicated that dPER protein levels were also elevated in the mutant flies, suggesting a role for codon optimization in enhancing dPER expression (Figure 2B in Fu J et al. 2016). Consistent with this, we analyzed the region of codon optimization in Fu J et al. 2016. The result indicated that that GAC has a relatively high usage rate in these regions (indicated in the following two Author response image charts by the red arrow), suggesting that the Mettl5 mutation may influence per protein accumulation through altered GAC usage. Further experiments are needed to confirm this possibility. We included these details in the second last paragraph of the Discussion session.

Author response image 1.

15-21

SDSAYSN

Author response image 2.

43-316

SSGSSGYGGKPSTQASSSDMIIKRNKEKSRKKKKPKCIALATATTVSLEGTEESPLPANGGCEKVLQELQDTQQLGEPLVVTETQLSEQLLETEQNEDQNKSEQLAQFPLPTPIVTTLSPGIGPGHDCVGGASGGAVAGGCSVVGAGTDKTSELIPGKLESAGTKPSQERPKEESFCCVISMHDGIVLYTTPSISDVLGFPRDMWLGRSFIDFVHHKDRATFASQITTGIPIAESRGCMPKDARSTFCVMLRRYRGLNSGGFGVIGRAVNYEPF

Fu J, Murphy KA, Zhou M, Li YH, Lam VH, Tabuloc CA, Chiu JC, Liu Y. Codon usage affects the structure and function of the Drosophila circadian clock protein PERIOD. Genes Dev. 2016 Aug 1;30(15):1761-75.

- A few papers already demonstrated the role of Mettl5 in translation, even at the structural level (Rong et al, Cell reports 2020) and this was not commented by the authors. In Peng et al, 2022 the authors show that the m6A bridges the 18S rRNA with RPL24. Is this conserved in Drosophila?

Thanks for the reminder. We discussed and cited these papers in the revised version.

Rong B, Zhang Q, Wan J, et al. Ribosome 18S m⁶A Methyltransferase METTL5 Promotes Translation Initiation and Breast Cancer Cell Growth. Cell Rep. 2020;33(12):108544. doi:10.1016/j.celrep.2020.108544

Peng H, Chen B, Wei W, et al. N⁶-methyladenosine (m⁶A) in 18S rRNA promotes fatty acid metabolism and oncogenic transformation. Nat Metab. 2022;4(8):1041-1054. doi:10.1038/s42255-022-00622-9

- The text will require strong editing and the authors should check and review extensively for improvements to the use of English.

Thanks. The text of the paper are thoroughly revised.

Conclusion

Despite the effort to identify the underlying molecular defects following the loss of Mettl5 the authors felt short in doing so. Some of the results are over-interpreted and more experiments will be needed to understand how Mettl5 controls the translation of its targets. References to previous works was poorly commented.

Thanks for your suggestion. We have incorporated the references mentioned above. However, our efforts have thus far fallen short of elucidating a precise picture of METTL5's functional mechanism. To address this, the limitations of the current study have been discussed more thoroughly in the revised main text.

Reviewer #2 (Public review):

Summary:

The authors define the m6A methyltransferase Mettl5 as a novel sleep-regulatory gene that contributes to specific aspects of Drosophila sleep behaviors (i.e., sleep drive and arousal at early night; sleep homeostasis) and propose the possible implication of Mettl5-dependent clocks in this process. The model was primarily based on the assessment of sleep changes upon genetic/transgenic manipulations of Mettl5 expression (including CRISPR-deletion allele); differentially expressed genes between wild-type vs. Mettl5 mutant; and interaction effects of Mettl5 and clock genes on sleep. These findings exemplify how a subclass of m6A modifications (i.e., Mettl5-dependent m6A) and possible epi-transcriptomic control of gene expression could impact animal behaviors.

Strengths:

Comprehensive DEG analyses between control and Mettl5 mutant flies reveal the landscape of Mettl5-dependent gene regulation at both transcriptome and translatome levels. The molecular/genetic features underlying Mettl5-dependent gene expression may provide important clues to molecular substrates for circadian clocks, sleep, and other physiology relevant to Mettl5 function in Drosophila.

Weaknesses:

While these findings indicate the potential implication of Mettl5-dependent gene regulation in circadian clocks and sleep, several key data require substantial improvement and rigor of experimental design and data interpretation for fair conclusions. Weaknesses of this study and possible complications in the original observations include but are not limited to:

(1) Genetic backgrounds in Mettl5 mutants: the heterozygosity of Mettl5 deletion causes sleep suppression at early night and long-period rhythms in circadian behaviors. The transgenic rescue using Gal4/UAS may support the specificity of the Mettl5 effects on sleep. However, it does not necessarily exclude the possibility that the Mettl5 deletion stocks somehow acquired long-period mutation allelic to other clock genes. Additional genetic/transgenic models of Mettl5 (e.g., homozygous or trans-heterozygous mutants of independent Mettl5 alleles; Mettl5 RNAi etc.) can address the background issue and determine 1) whether sleep suppression tightly correlates with long-period rhythms in Mettl5 mutants; and 2) whether Mettl5 effects are actually mapped to circadian pacemaker neurons (e.g., PDF- or tim-positive neurons) to affect circadian behaviors, clock gene expression, and synaptic plasticity in a cell-autonomous manner and thereby regulate sleep. Unfortunately, most experiments in the current study rely on a single genetic model (i.e., Mettl5 heterozygous mutant).

We believe that the multiple rescue experiments presented in Figure 1H-L and Figure 2H-L have effectively addressed the background concern. To further confirm this, we have subsequently repeated sleep and circadian rhythm assays using RNAi lines, aiming to further eliminate any remaining concerns in this regard. It appears to replicate the reduced sleep phenotype seen at night. This result has been included in the Figure S1. It is true that we have not specifically addressed whether the effects of Mettl5 are mapped to circadian pacemaker neurons in this study. We acknowledge this as a limitation and appreciate the importance of this question. Further investigations focusing on circadian pacemaker neurons, such as PDF- or tim-positive neurons, would be necessary to clarify the precise role of Mettl5 in regulating circadian behaviors and related molecular mechanisms.

(2) Gene expression and synaptic plasticity: gene expression profiles and the synaptic plasticity should be assessed by multiple time-point analyses since 1) they display high-amplitude oscillations over the 24-h window and 2) any phase-delaying mutation (e.g., Mettl5 deletion) could significantly affect their circadian changes. The current study performed a single time-point assessment of circadian clock/synaptic gene expression, misleading the conclusion for Mettl5 effects. Considering long-period rhythms in Mettl5 mutant clocks, transcriptome/translatome profiles in Mettl5 cannot distinguish between direct vs. indirect targets of Mettl5 (i.e., gene regulation by the loss of Mettl5-dependent m6A vs. by the delayed circadian phase in Mettl5 mutants).

In the revised version, we provided data collected at multiple time points. Specifically, we reexamined the per expression at both transcriptional and translational levels at different timepoints. The corresponding results were incorporated in Figure 4 D-F. We also dissected fly brains from UAS-DenMark, UAS-syt.eGFP/+; pdf-GAL4/+ and UAS-DenMark, UAS-syt.eGFP/+; pdf-GAL4/Mettl5^1bp at these four time points to quantify the synaptic structures of PDF neurons. The result has been included in revised Figure 6.

(3) The text description for gene expression profiling and Mettl5-dependent gene regulation was very detailed, yet there is a huge gap between gene expression profiling and sleep/behavioral analyses. The model in Figure 5 should be better addressed and validated.

Thank you for your suggestion. We added data to better confirm the expression changes of PER protein at different time points. Indeed, what you mention is the weak point of this paper. We did analysis thoroughly during the revision process.

The opposing changes in Period at the transcriptional versus translational levels puzzled us for some time until we identified alterations in the ubiquitin pathway components. The regulation of Period protein degradation by the ubiquitin-proteasome pathway is well-documented (Grima et al., 2002; Ko et al., 2002; Chiu et al., 2008). Additionally, previous studies have shown that N6-methyladenosine (m6A) modulates the ubiquitin-proteasome pathway in skeletal muscle physiology (Sun et al., 2023). We have incorporated this information into the revised manuscript in the last paragraph under the section titled: Clock gene regulatory loop regulating circadian rhythm was affected by Mettl5^1bp

Indeed, we have not yet identified an effective method to manipulate proteasome levels in genetic tests. The proteasome is a large protein complex composed of numerous subunits, making it impractical to enhance its activity simply by overexpressing individual components. Furthermore, the proteasome plays a critical role in many biological processes. Disrupting its function—such as through MG132 treatment, which we attempted—leads to significant off-target effects.

Sun J, Zhou H, Chen Z, et al. Altered m6A RNA methylation governs denervation-induced muscle atrophy by regulating ubiquitin proteasome pathway. J Transl Med. 2023;21(1):845. Published 2023 Nov 23. doi:10.1186/s12967-023-04694-3

Grima, B. et al. The F-box protein slimb controls the levels of clock proteins period and timeless. Nature 420, 178–182 (2002).

Ko, H. W., Jiang, J. & Edery, I. Role for Slimb in the degradation of Drosophila period protein phosphorylated by doubletime. Nature 420, 673–678 (2002).

Chiu, J. C., Vanselow, J. T., Kramer, A. & Edery, I. The phosphooccupancy of an atypical SLIMB-binding site on PERIOD that is phosphorylated by DOUBLETIME controls the pace of the clock. Genes Dev. 22, 1758–1772 (2008).

Reviewer #3 (Public review):

Xiaoyu Wu and colleagues examined the potential role in sleep of a Drosophila ribosomal RNA methyltransferase, mettl5. Based on sleep defects reported in CRISPR generated mutants, the authors performed both RNA-seq and Ribo-seq analyses of head tissue from mutants and compared to control animals collected at the same time point. While these data were subjected to a thorough analysis, it was difficult to understand the relative direction of differential expression between the two genotypes. In any case, a major conclusion was that the mutant showed altered expression of circadian clock genes, and that the altered expression of the period gene in particular accounted for the sleep defect reported in the mettl5 mutant. As noted above, a strength of this work is its relevance to a human developmental disorder as well as the transcriptomic and ribosomal profiling of the mutant. However, there are numerous weaknesses in the manuscript, most of which stem from misinterpretation of the findings, some methodological approaches, and also a lack of method detail provided. The authors seemed to have missed a major phenotype associated with the mettl5 mutant, which is that it caused a significant increase in period length, which was apparent even in a light: dark cycle. Thus the effect of the mutant on clock gene expression more likely contributed to this phenotype than any associated with changes in sleep behavior.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Some of the questions that the authors should address are the following ones:

How does Mettl5 control the translation of the clock genes ? Why the level of some genes are specifically increased or decreased? What is the relation with the effect on uORF and dORF, overlapping and non overlapping ones? The observation of these defects is interesting but how they occurs and how they impact clock signaling is missing.

Thank you for your suggestion. This is the weak point of this paper. We did analysis thoroughly during the revision process.

The opposing changes in Period at the transcriptional versus translational levels puzzled us for some time until we identified alterations in the ubiquitin pathway components. The regulation of Period protein degradation by the ubiquitin-proteasome pathway is well-documented (Grima et al., 2002; Ko et al., 2002; Chiu et al., 2008). Additionally, previous studies have shown that N6-methyladenosine (m6A) modulates the ubiquitin-proteasome pathway in skeletal muscle physiology (Sun et al., 2023). We have incorporated this information into the revised manuscript in the last paragraph under the section titled: Clock gene regulatory loop regulating circadian rhythm was affected by Mettl5^1bp.

Indeed, we have not yet identified an effective method to manipulate proteasome levels in genetic tests. The proteasome is a large protein complex composed of numerous subunits, making it impractical to enhance its activity simply by overexpressing individual components. Furthermore, the proteasome plays a critical role in many biological processes. Disrupting its function—such as through MG132 treatment, which we attempted—leads to significant off-target effects.

In this study, we also observed codon usage alterations caused by the mettl5 mutant. For details, please refer to our responses to 4th question of the weakness session above. Previous studies have reported mettl5's role in translational regulation in other systems (Rong et al., 2020; Peng et al., 2022). Based on these findings, we propose that both translational regulation and protein degradation may contribute to the upregulation of Period protein in the mettl5 mutant. This hypothesis has been included in the Discussion section of the revised manuscript.

“The mechanism by which METTL5 regulates translation warrants further investigation. Previous studies have demonstrated that METTL5 influences translation (Rong et al., 2020; Peng et al., 2022), but whether the mechanisms identified here are conserved across other systems remains an intriguing question. In our analysis, we observed increased usage of aspartate (Asp) codons in Mettl5 mutants. Notably, prior work has linked codon usage to PER protein function—specifically, a codon-optimized version of PER failed to rescue circadian rhythmicity in per mutant flies, unlike the wild-type version (Fu et al., 2016). Further analysis revealed that PER protein levels were elevated in these mutants, suggesting that codon optimization enhances PER expression (Figure 2B in Fu et al., 2016). Strikingly, when we examined the codon-optimized region from Fu et al. (2016), we found that GAC (Asp) was highly enriched, raising the possibility that Mettl5 mutation affects PER protein accumulation by altering GAC codon usage. Additional experiments will be needed to validate this hypothesis. Furthermore, we detected changes in upstream open reading frames (uORFs) in Mettl5 mutants, but their relationship to translational regulation requires further exploration.”

References

Sun J, Zhou H, Chen Z, et al. Altered m6A RNA methylation governs denervation-induced muscle atrophy by regulating ubiquitin proteasome pathway. J Transl Med. 2023;21(1):845. Published 2023 Nov 23. doi:10.1186/s12967-023-04694-3

Grima, B. et al. The F-box protein slimb controls the levels of clock proteins period and timeless. Nature 420, 178–182 (2002).

Ko, H. W., Jiang, J. & Edery, I. Role for Slimb in the degradation of Drosophila period protein phosphorylated by doubletime. Nature 420, 673–678 (2002).

Chiu, J. C., Vanselow, J. T., Kramer, A. & Edery, I. The phosphooccupancy of an atypical SLIMB-binding site on PERIOD that is phosphorylated by DOUBLETIME controls the pace of the clock. Genes Dev. 22, 1758–1772 (2008).

Rong B, Zhang Q, Wan J, et al. Ribosome 18S m⁶A Methyltransferase METTL5 Promotes Translation Initiation and Breast Cancer Cell Growth. Cell Rep. 2020;33(12):108544. doi:10.1016/j.celrep.2020.108544

Peng H, Chen B, Wei W, et al. N⁶-methyladenosine (m⁶A) in 18S rRNA promotes fatty acid metabolism and oncogenic transformation. Nat Metab. 2022;4(8):1041-1054. doi:10.1038/s42255-022-00622-9

Fu J, Murphy KA, Zhou M, Li YH, Lam VH, Tabuloc CA, Chiu JC, Liu Y. Codon usage affects the structure and function of the Drosophila circadian clock protein PERIOD. Genes Dev. 2016 Aug 1;30(15):1761-75.

Reviewer #2 (Recommendations for the authors):

Please find my comments to improve the quality of your manuscript.

Major comments

(1) The quality of text writing in English needs to be at publishable levels. It is not a trivial problem, but it literally impairs the readability of your work. So please have professionals edit your manuscript text appropriately.

We have carefully revised the language throughout the manuscript during the revision process.

(2) Fig 1O: please include the total sleep profile and other analyses for rebound sleep phenotypes in control vs. Mettl5 to better validate that both genotypes were comparably sleep-deprived, but the latter shows less sleep rebound.

Thank you for your suggestion, The other reviewer also suggested to reanalyze the sleep rebound data. We did the analysis according to the following reference. We included data sleep profiles of both genotypes in original Fig 1O. Total sleep profile and other analyses for rebound sleep phenotypes are included in the revised panel. As shown in this revised panel (now Figure 1K, L), both genotypes were comparably sleep-deprived.

Cirelli C, Bushey D, Hill S, Huber R, Kreber R, Ganetzky B, Tononi G. 2005. Reduced sleep in Drosophila Shaker mutants. Nature 434:1087-92.

(3) Line 90: the authors did not actually address this critical question. Additional Gal4 mapping (e.g., Mettl5 rescue or Mettl5 RNAi) will determine which cells/neural circuits are important for Mettl5-dependent sleep.

This sentence has been revised into “The observed expression pattern of Mettl5 further supports its sleep regulatory function.”

(4) Fig 1H-L; Fig 2H-L: the authors should check if overexpression of wild-type or mutant Mettl5 in control backgrounds could affect nighttime sleep to better define the transgenic effects among overexpression, rescue, and dominant-negative.

Thank you for the comment. We added the overexpression phenotypes in the revised version.

(5) Lines 225-226. Fig S11: The neural projections from PDF-expressing neurons should be better imaged and quantified. Current images can visualize PDF projections onto the optic lobe but not others (e.g., dorsal, POT), so the conclusion is not validated.

Thank you for the suggestion. We acknowledge the limitation in the current images of PDF-expressing neuronal projections. We included new, higher-resolution images to better visualize and quantify the neural projections, including the dorsal and POT regions, to ensure the conclusion is well-supported.

(6) Lines 230-232: per RNA/PER protein expression oscillates daily, so the authors should perform time-point experiments to conclude Mettl5 effects on clock gene expression, including per.

Thank you for the insightful comment. We performed experiments in the Mettl5 mutant background at four time points to analyze PER protein expression using both RT-PCR and Western blot (anti-PER). The updated results have been included in Figure 4D-F.

(7) Lines 235-238: the authors should note that Mettl5 effects on sleep in Clk or per mutant backgrounds are actually opposite to those in w1118/control one. Mettl5 deletion promotes daytime or nighttime sleep in Clk or per mutants, respectively. Any explanation? 

We are trying to use epistasis analysis to determine which gene is upstream here. Epistasis (or epistatic effect) in genetics refers to the interaction between different genes where the expression of one gene (the epistatic gene) masks or modifies the expression of another gene (the hypostatic gene). The epistatic gene (masking gene) usually functions downstream in the pathway because its effect overrides the output of the hypostatic gene. The double mutant showed the similar phenotype as downstream genes. Thus, Clk or per functions downstream of Mettl5.

(8) Fig 6: The dorsal PDF projections actually show time-dependent plasticity. Results from the single time-point are not conclusive.

Thank you for the insightful comment. we further dissected fly brains from UAS-DenMark, UAS-syt.eGFP/+; pdf-GAL4/+ and UAS-DenMark, UAS-syt.eGFP/+; pdf-GAL4/Mettl5^1bp at these four time points to analyze the morphology of PDF neurons. The results have been included in figure 6.

Minor comments

(1) Please avoid simple bar graphs in the data presentation-include individual data points or use a different graph showing the distribution of raw data (e.g., violin plot, box plot, etc.).

Thank you for the suggestion. In the revised version of the manuscript, we have included individual data points, violin plots, and box plots to present the data, effectively showing both the distribution and differences in the raw data.

(2) Line 19: "Clock" indicates the gene name or general terminology such as "circadian clock". Please clarify it and revise the font accordingly.

This has been revised into“clock”

(3) The overall flow in the Abstract/Summary is somewhat challenging for a general audience to follow.

We have revised the text, especially the overall flow in the Abstract/Summary.

(4) Fonts for the names of genes and gene products (i.e., mRNA, protein) should be appropriately corrected throughout the manuscript.

We have checked the text and made changes where necessary.

(5) Methods: the authors should provide detailed information on the methods. For instance, there is little description of how they generate Mettl5 deletions (e.g., sgRNA/target sequence). Also, they should clarify whether they test heterozygous vs. homozygous mutants of Mettl5 deletions in each experiment since the genotype description in the figure appears mixed-up (e.g., Fig 1B vs. Fig 1I-L).

Thank you for pointing this out. In the updated version, we provided detailed information about the strains used, including the sgRNA/target sequences for generating Mettl5 deletions. Regarding the genotypes, Figure 1B represents homozygous mutants, while Figures 1I-L represent heterozygous mutants. This distinction has been clarified in the figure legends, and the genotype notation for Figures 1I-L will be revised for consistency and clarity.

(6) Fig 1: the figure panels should be re-arranged based on the order of their text description (i.e., Fig 1H-L should go after Fig 1M-O).

Thank you for the suggestion. In the revised version, we rearranged the figure panels so that Figures 1H-L appear after Figures 1M-O, following the order of their description in the text.

(7) Sleep education in Trmt112 RNAi looks different from that in Mettl5 mutant het. Any explanation?

The functional divergence between Trmt112 and Mettl5 may also contribute to the observed sleep phenotype. While Trmt112 and Mettl5 share some downstream targets, they each regulate many unique genes, some of which could influence sleep. Sleep is a highly sensitive trait that can be modulated by numerous genetic factors. Previous studies have also suggested that sleep behaves more like a quantitative trait, reflecting the combined effects of multiple genes (Mackay and Huang, 2018).

Mackay TFC, Huang W. Charting the genotype-phenotype map: lessons from the Drosophila melanogaster Genetic Reference Panel. Wiley Interdiscip Rev Dev Biol. 2018;7(1):10.1002/wdev.289. doi:10.1002/wdev.289

Reviewer #3 (Recommendations for the authors):

A detailed critique is provided below. Generally, the authors can greatly improve this manuscript if they focus more rigorously on the circadian phenotype associated with the Mettl5 mutant, which could be the basis for the apparent sleep phenotype.

(1) Please provide more information as to how each of the mettl5 mutants were generated. This information should include, specifically, the gRNA sequences, plasmids generated for the 5' and 3' arms, and anything related to the CRISPR approach for generating the mutants. Was any sequencing done to verify the CRISPR alleles, or was this limited to the analysis of mettl5 expression and behavior? Please indicate where the qPCR primers (used in Fig 1B) are located relative to the mutant loci. The figure legend is also incomplete in that there is no reference to the boxed area in Fig 1A.

In the updated version, we have provided detailed information about the how each of the mettl5 mutants were generated. The sequence was verified by sequencing following PCR. The following references to the boxed area were added in the revised version.

Reference

Iyer LM, Zhang D, Aravind L. Adenine methylation in eukaryotes: Apprehending the complex evolutionary history and functional potential of an epigenetic modification. Bioessays. 2016 Jan;38(1):27-40. doi: 10.1002/bies.201500104.

(2) As noted, I am not in agreement with the interpretation of findings for the sleep defect reported in the mettl5[1b]/+ mutants. There is a clear increase in morning sleep in the mutants that may not have reached significance by lumping the data in 12h increments (Fig1C-E). Were the overall 24h sleep values between the mutants and controls the same? The sleep profile appears to be shifted, such that nighttime sleep onset in the mutants occurs much later than wild type, and daytime waking is also much later, all pointing to a long period phenotype, which is very strongly supported by the data in Table 1, as well as the RNA- and ribo-seq data. The implications for this leading to sleep disturbances in humans is very exciting. An additional suggestion to the authors here is to report the nighttime sleep latency values (time to onset of the first sleep bout after lights off).

We appreciate your insightful observation. As shown in Table 1, the Mettl51bp/+ mutant exhibits a robust long-period phenotype, with circadian rhythms significantly extended to 28.3 ± 0.4 hours compared to the wild-type's 23.9 ± 0.05 hours. This prolonged period perfectly aligns with the observed behavioral phenotypes, including delayed nighttime sleep onset, later daytime waking, and the overall shift in sleep profile. This is indeed quite similar to previous report on Period3 variant (Zhang et al., 2016). We agree that the prolonged circadian period contributes to the observed sleep phenotype. However, since total sleep time was significantly reduced in the mutant, we cannot attribute the phenotype solely to period lengthening. Furthermore, our 24-hour PER expression analysis in mettl5 mutants revealed elevated PER protein levels at ZT1 and ZT18, while ZT6 and ZT12 showed no significant changes, with no apparent phase shift. These findings collectively suggest that the phenotype primarily results from PER protein stabilization and accumulation.

Importantly, genetic rescue experiments restoring wild-type Mettl5 function (UAS-Mettl5/Mettl5-Gal4; Figure 1 and Table 1) completely normalized the circadian period to 24 ± 0.02 hours, providing compelling evidence that these phenotypes specifically result from loss of Mettl5 function. Together with the sleep architecture data, these findings establish Mettl5 as a crucial regulator of circadian rhythms, with important implications for understanding human sleep disorders. To further substantiate these observations, we have now included quantitative nighttime sleep latency measurements in the revised manuscript to better document the delayed sleep onset in mutants (Figure S1G).

We have discussed this in the third paragraph of the Discussion session and included the reference in the revised manuscript.

Zhang L, Hirano A, Hsu PK, et al. A PERIOD3 variant causes a circadian phenotype and is associated with a seasonal mood trait. Proc Natl Acad Sci U S A. 2016;113(11):E1536-E1544. doi:10.1073/pnas.1600039113.

(3) The description for how circadian behavior was measured and analyzed (Table 1) is missing from the methods section.

We have included a detailed description of the methods used to measure and analyze circadian behavior, as presented in Table 1, in the revised methods “Sleep behavior assays” section.

(4) Please explain what the "awake %" values reported in Figs 1G, 1L, Fig 2G, and 2L, Fig 4G and 4M are. Is this simply the number of flies that are awake at a given time point? This does not provide useful information beyond what is already reported for the sleep profiling in other parts of these figures. If it is an arousal threshold assay, as shown in supplementary Fig 1H, please indicate this. The description for "sleep arousal" in the methods (lines 368-371) is also concerning. If most of the mutant flies are already awake at ZT 14, then I would expect that this assay would not work at this time of day. A more suitable time point would be ZT 19, or later, when the mutants are falling asleep. Moreover, calculating the number of flies awakened as long as 5 minutes after a stimulus pulse cannot be distinguished from a spontaneous awakening, and so is not really a metric of arousal threshold. The number of sleeping flies awakened by the stimulus should be calculated within, at most, one minute afterward.

Thank you for your suggestion. Regarding the 'awake %' metric, it indicates that at specific time points (e.g., ZT14), the percentage of awake fruit fly population at that moment. In the revised version, we further clarify the definition and significance of 'awake %'. Additionally, we have reevaluated the time points for the arousal threshold assay, selecting a more appropriate time (e.g., ZT19) to better reflect the sleep state of the mutants. Based on your suggestion, we calculate the number of flies awakened within one minute after the stimulus to ensure a more accurate measurement of arousal threshold. This has been included in the revised Figure 1M.

(5) Fig1M-O is problematic. First, is it possible that expression of Mettl5 mRNA fluctuates with time-of-day and is not affected by sleep loss? There are no undisturbed controls collected at equivalent time points. The method used for quantifying sleep rebound in Fig 1O (lines 365-367) does not make sense, as negative values would be expected. Moreover, since the Mettl5 mutants show high sleep amounts in the morning and very low sleep amounts from ZT 12-18, this analysis would be severely confounded. Also, the sleep deprivation applied would not produce equivalent amounts of sleep loss as compared to wild type controls, so this also needs to be corrected. The authors should consider consulting Cirelli et al (2005, DOI: 10.1038/nature03486 ) as an approach for quantifying sleep homeostasis in a short-sleeping mutant. Please also show the sleep profiling in the mutants for these experiments.

Thank you for your valuable suggestions. Regarding the possibility that Mettl5 mRNA expression fluctuates with circadian rhythms rather than being affected by sleep deprivation, we acknowledge that collecting undisturbed control samples at equivalent time points would provide critical insights. In the revised version, we included undisturbed controls to distinguish between circadian-driven fluctuations and the effects of sleep deprivation on Mettl5 expression.

For the quantification of sleep rebound in Figure 1O, we agree that the current method may not fully capture the dynamics of sleep recovery, especially in Mettl5 mutants, where sleep patterns differ significantly from wild-type. We have referred to the method proposed by Cirelli et al. paper for quantifying sleep homeostasis in short-sleeping mutants, ensuring a more accurate evaluation of sleep rebound. The results have been included in Figure 1K-L of the revised version.

(6) Fig 3B and C (minor) - while the volcano plots are clear, it is not clear whether "down" or "up" means for the mutant relative to wild type or the other way around? Please clarify. In Fig 3P, the legend indicates a depiction of the "top 5 pathway associated genes", but it seems there are 10 pathways depicted. Which of these are the "top 5"?

In the volcano plots (Fig. 3B and 3C), “up” and “down” refer to genes that the mutant relative to the wild-type strain. In Fig. 3P, the legend was mislabeled as “top 5” pathway-associated genes. In fact, we displayed the top 10 pathway-associated genes. We apologize for the confusion and will correct both the figure legend and the corresponding text in our revised manuscript.

(7) Fig 4 D-E, and F,G do not have sufficient information to draw the conclusion that Per mRNA/protein expression is increased in the Mettl5 mutant. Since both mRNA protein of this gene oscillates significantly throughout the day, it is still possible that the single time point shown in this figure might indicate a disruption in cycling rather than overall expression level. Please first indicate what time of day the tissue was collected, second, consider adding more time points to both assays. For the first part of this figure, A and B, per and Clock gene expression are expected to be in different phases, and so this aspect is not unexpected. However, it is notable that it is reversed in the mutant vs wild type. Again, an alternate interpretation of this finding that the authors have not considered is a change in period duration of gene cycling.

Thank you for your suggestion. For the PER WB experiments, we have included multiple time points in the revised version to more comprehensively evaluate PER expression in the Mettl5 mutant and better understand its circadian rhythm changes. We appreciate your observation regarding the potential changes in the period duration of gene cycling. This has been discussed in the 3^rd paragraph of the Discussion session of the revised version.

(8) The data shown in Figs 4H-M does not support the conclusion that "Clock and Per genes were downstream of Mettl5" (line 236-237). The daytime sleep phenotype, in particular, appears additive between both circadian genes and mutant because the morning sleep of the double mutant is much higher than either mutant by itself. Statistical comparisons between the double mutant and each clock mutant are also noticeably missing. These data are difficult to interpret. One potential explanation is that Mettl5 alters gene expression of non-circadian genes, and that the phenotypes become additive when both clock and Mettl5 genes are missing. A full molecular analysis of clock gene cycling in the Mettl5 mutant may help improve understanding of the relationship between the circadian clock Mettl5 gene expression. It may also be worthwhile checking whether Mettl5 gene expression itself shows a daily oscillation.

Thank you for your suggestion. In the revised version, we have included four additional time points to analyze the oscillatory expression of Per and Clock in the Mettl5 mutant, providing a more comprehensive understanding of their circadian rhythm changes. In Figs 4H-M, we are trying to use epistasis analysis to determine which gene is upstream here. Epistasis (or epistatic effect) in genetics refers to the interaction between different genes where the expression of one gene (the epistatic gene) masks or modifies the expression of another gene (the hypostatic gene). The epistatic gene (masking gene) usually functions downstream in the pathway because its effect overrides the output of the hypostatic gene. The double mutant showed the similar phenotype as downstream genes. Thus, Clk or per functions downstream of Mettl5. Statistical comparisons between the double mutant and each clock mutant are added.

(9) In Fig 6, what time of day were the flies collected? PDF terminal morphology is known to change throughout the day; this is another piece of data that could indicate a defect in circadian function rather than a chronic change in synaptic morphology.

The flies were collected around ZT14. We included additional dissection time points in future experiments. Differences between the control and Mettl5 mutants are observed consistently across multiple time points, suggesting that Mettl5 has an impact on synaptic plasticity.

Minor:

There are letter indicators, presumably for statistical comparisons, depicted in Figs 1 and 2 (panels I-L), but no explanation as to what these mean in the figure legends.

We have added notes in the revised version.

What is the purpose of the boxed regions shown in Fig S1A-F? There is no explanation of these in the figure legend nor in the text.

The boxed regions highlight the significant co-localization of two proteins. We have included this explanation in the figure legend in the revised version.

The statement (lines 310-311) that per and clock genes "exhibit more pronounced sleep rebound after sleep deprivation" is inaccurate. The article cited for this (Shaw et al 2002) showed that it was female mutants of the cycle gene which showed prolonged sleep rebound; other clock mutants were normal.

Thank you for pointing out this. We revised the statement accordingly.

Overall, the manuscript may benefit from editing or writing assistance to improve the language. There were many incomplete sentences, grammatical errors, etc.

We have carefully refined the language throughout the manuscript during the revision process.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

The authors report intracranial EEG findings from 12 epilepsy patients performing an associative recognition memory task under the influence of scopolamine. They show that scopolamine administered before encoding disrupts hippocampal theta phenomena and reduces memory performance, and that scopolamine administered after encoding but before retrieval impairs hippocampal theta phenomena (theta power, theta phase reset) and neural reinstatement but does not impair memory performance. This is an important study with exciting, novel results and translational implications. The manuscript is well-written, the analyses are thorough and comprehensive, and the results seem robust.

Strengths:

(1) Very rare experimental design (intracranial neural recordings in humans coupled with pharmacological intervention).

(2) Extensive analysis of different theta phenomena.

(3) Well-established task with different conditions for familiarity versus recollection.

(4) Clear presentation of findings and excellent figures.

(5) Translational implications for diseases with cholinergic dysfunction (e.g., AD).

(6) Findings challenge existing memory models, and the discussion presents interesting novel ideas.

Weaknesses:

(1) One of the most important results is the lack of memory impairment when scopolamine is administered after encoding but before retrieval (scopolamine block 2). The effect goes in the same direction as for scopolamine during encoding (p = 0.15). Could it be that this null effect is simply due to reduced statistical power (12 subjects with only one block per subject, while there are two blocks per subject for the condition with scopolamine during encoding), which may become significant with more patients? Is there actually an interaction effect indicating that memory impairment is significantly stronger when scopolamine is applied before encoding (Figure 1d)? Similar questions apply to familiarity versus recollection (lines 78-80). This is a very critical point that could alter major conclusions from this study, so more discussion/analysis of these aspects is needed. If there are no interaction effects, then the statements in lines 84-86 (and elsewhere) should be toned down.

The reviewer highlights important concerns regarding the statistical power of the behavioral effects. We address these concerns in the revised manuscript in two ways: (1) we provide a supplemental analysis using a matched number of blocks between the placebo and scopolamine conditions to avoid statistical bias related to differing trial counts, and (2) we include a supplemental figure illustrating paired comparisons between blocks.

(2) Further, could it simply be that scopolamine hadn't reached its major impact during retrieval after administration in block 2? Figure 2e speaks in favor of this possibility. I believe this is a critical limitation of the experimental design that should be discussed.

The reviewer raises an important methodological concern regarding the time required for scopolamine's effect to manifest and the subsequent impact on the study outcomes. Previous studies report that the average time to maximum serum concentration after intravenous (IV) scopolamine administration is approximately 5 minutes (Renner et al., 2005), with the corresponding clinical onset estimated at 10 minutes. In our study, the retrieval period in Block 2 commenced at 15 ± 0.2 post-injection across all subjects. Given this timing, there is sufficient reason to conclude that scopolamine had reached its major impact during the Block 2 retrieval phase. Furthermore, the observation of significant disruptions to theta oscillations during this same retrieval phase provides strong evidence that the drug was in full effect at that time.

(3) It is not totally clear to me why slow theta was excluded from the reinstatement analysis. For example, despite an overall reduction in theta power, relative patterns may have been retained between encoding and recall. What are the results when using 1-128 Hz as input frequencies?

Slow theta (2–4 Hz) was excluded from the reinstatement analysis to avoid potential confounding effects. Given the observed disruption to slow theta power following scopolamine administration, any subsequent changes in slow theta reinstatement would be causally ambiguous, potentially arising directly from the power effects. Therefore, we would be unable to determine whether changes in slow theta reinstatement were genuinely independent of changes in power.

(4) In what way are the results affected by epileptic artifacts occurring during the task (in particular, IEDs)?

To exclude abnormal events and interictal activity, a kurtosis threshold of 4 was applied to each trial, effectively filtering out segments exhibiting significant epileptic artifacts.

Reviewer #2 (Public review):

Summary:

In this study, performed in human patients, the authors aimed at dissecting out the role of cholinergic modulation in different types of memory (recollection-based vs familiarity and novelty-based) and during different memory phases (encoding and retrieval). Moreover, their goal was to obtain the electrophysiological signature of cholinergic modulation on network activity of the hippocampus and the entorhinal cortex.

Strengths:

The authors combined cognitive tasks and intracranial EEG recordings in neurosurgical epilepsy patients. The study confirms previous evidence regarding the deleterious effects of scopolamine, a muscarinic acetylcholine receptor antagonist, on memory performance when administered prior to the encoding phase of the task. During both encoding and retrieval phases, scopolamine disrupts the power of theta oscillations in terms of amplitude and phase synchronization. These results raise the question of the role of theta oscillations during retrieval and the meaning of scopolamine's effect on retrieval-associated theta rhythm without cognitive changes. The authors clearly discussed this issue in the discussion session. A major point is the finding that the scopolamine-mediated effect is selective for recollection-based memory and not for familiarity- and novelty-based memory.

The methodology used is powerful, and the data underwent a detailed and rigorous analysis.

Weaknesses:

A limited cohort of patients; the age of the patients is not specified in the table.

To comply with human subject privacy protection policies, age was not reported; however, we did not find any significant effects of age on the behavioral or neural measures.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Regarding dosage, did you take the patients' body weight into account? Do the effects hold when controlling for it?

We controlled for participant weight, yet the observed effects were more strongly correlated with the absolute scopolamine dosage, irrespective of weight. This outcome indicates that scopolamine likely rapidly crosses the blood-brain barrier, producing swift effects that are not initially influenced by metabolic variability.

(2) Line 96: Corrected for what kind of multiple comparisons?

We apologize for this confusion. The statistical analysis presented in this line does not require multiple-comparison correction, and we will therefore remove the annotation.

(3) Line 165: These are very interesting results. How do they relate to Rizzuto et al., NeuroImage, 2006?

Our findings show that successful retrieval is tied to an encoding-retrieval phase match, which is a refinement and application of the Rizzuto et al. (2006) work. Rizzuto et al. showed that memory events are phase-locked; we show that maintaining a specific, matched phase relationship between encoding and retrieval events is critical for memory success, and that this process is dependent on the cholinergic system.

Reviewer #2 (Recommendations for the authors):

Figure 1b: It would be useful for clarity to have the cartoon of the treatment paradigm for the encoding phase (blocks 3 and 4).

The treatment paradigm only involved a single intravenous (IV) injection of scopolamine (or saline, for the placebo condition). The injections were administered by the participant's attending nurse, with a board-certified anesthesiologist present at the time of injection and available throughout the experiment. These details are fully documented in the Methods section.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

An interesting manuscript from the Carrington lab is presented investigating the behavior of single vs double GPI-anchored nutrient receptors in bloodstream form (BSF) T. brucei. These include the transferrin receptor (TfR), the HpHb receptor (HpHbR), and the factor H receptor (FHR). The central question is why these critical proteins are not targeted by host-acquired immunity. It has generally been thought that they are sequestered in the flagellar pocket (FP), where they are subject to rapid endocytosis - any Ab:receptor complexes would be rapidly removed from the cell surface. This manuscript challenges that assumption by showing that these receptors can be found all over the outer cell body and flagella surfaces, if one looks in an appropriate manner (rapid direct fixation in culture media).

The main part of the manuscript focuses on TfR, typically a GPI1 heterodimer of very similar E6 (GPI anchored) and E7 (truncated, no GPI) subunits. These are expressed coordinately from 15 telomeric expression sites (BES), of which only one can be transcribed at a time. The authors identify a native E6:E7 pair in BES7 in which E7 is not truncated and therefore forms a GPI2 heterodimer. By in situ genetic manipulation, they generate two different sets of GPI1:GPI2 TfR combinations expressed from two different BESs (BES1 and BES7). Comparative analyses of these receptors form the bulk of the data.

The main findings are:

(1) Both GPI1 and GPI2 TfR can be found on the cell body/flagellar surface.

(2) Both are functional for Tf binding and uptake.

(3) GPI2 TfR is expressed at ~1.5x relative to GPI1 TfR

(4) Ultimate TfR expression level (protein) is dependent on the BES from which it is expressed.

Most of these results are quite reasonably explained in light of the hydrodynamic flow model of the Engstler lab and the GPI valence model of the Bangs lab. Additional experiments, again by rapid fixation, with HpHbR and FHR, show that these GPI1 receptors can also be seen on the cell surface, in contrast to published localizations.

It is quite interesting that the authors have identified a native GPI2 TfR. However, essentially all of the data with GPI2 TfR are confirmatory for the prior, more detailed studies of Tiengwe et al. (2017). That said, the suggestion that GPI2 was the ancestral state makes good evolutionary sense, and begs the question of why trypanosomes prefer GPI1 TfR in 14 of 15 ESs (i.e., what is the selection pressure?)

Strengths and weaknesses:

(1) BES7 TfR subunit genes (BES7_Tb427v10): There are actually three (in order 5'3'): E7gpi, E6.1 and E6.2. E6.1 and E6.2 have a single nucleotide difference. This raises the issue of coordinate expression. If overall levels of E6 (2 genes) are not down-regulated to match E7 (1 gene), this will result in a 2x excess of E6 subunits. The most likely fate of these is the formation of non-functional GPI2 homodimers on the cell surface, as shown in Tiengwe et al. (2017), which will contribute to the elevated TfR expression seen in BES7.

We would like to thank the reviewer for pointing out that there are two ESAG6 genes in BES7, we had relied on the publicly available annotation and should have known better.

For transferrin expression levels, see the discussion in response to reviewer 1 point 3 below

(2) Surface binding studies: This is the most puzzling aspect of the entire manuscript. That surface GPI2 TfR should be functional for Tf binding and uptake is not surprising, as this has already been shown by Tiengwe et al. (2017), but the methodology for this assay raises important questions. First, labeled Tf is added at 500 nM to live cells in complete media containing 2.5 uM unlabeled Tf - a 5x excess. It is difficult to see how significant binding of labeled TfR could occur in as little as 15 seconds under these conditions.

The k_on for transferrin is very rapid (BES1 TfR / bovine transferrin at pH7.4 = 4.5 x 10⁵ M^-1s^-1 (Trevor et al., 2019) and binding would occur to unoccupied receptors within 15 sec. The k_off is also fast (BES1 TfR / bovine transferrin at pH7.4 = 3.6 x 10^-2 s^-1 (Trevor et al., 2019) and there would be exchange of transferrin within the time taken for endocytosis. These values are in vitro with purified proteins, the in vivo values may be affected by the VSG coat.

The failure to bind canine transferrin (Supp. Figure 4B) acts as a control for specificity of the interaction.

We have now performed a competition experiment as an additional control; cells in culture were supplemented with: A, 0.5 µM labelled transferrin; B, 0.5 µM labelled and 2.5 µM unlabelled transferrin; C, 0.5 µM labelled and 5 µM unlabelled transferrin, fixed after 60 s and visualised by fluorescence microscopy (Figure S4C). There was effective competition and greatly reduced binding of transferrin was seen in the presence of a 10-fold excess of unlabelled. We would like to thank the reviewer for suggesting this experiment.

Second, Tiengwe et al. (2017) found that trypanosomes taken directly from culture could not bind labeled Tf in direct surface labelling experiments. To achieve binding, it was necessary to first culture cells in serum-free media for a sufficient time to allow new unligated TfR to be synthesized and transported to the surface. This result suggests that essentially all surface TfR is normally ligated and unavailable to the added probe.

As part of the preliminary experiments for this paper we found that centrifugation followed by resuspension in either complete or serum free (but 1% BSA) medium resulted in a reduction is total cellular TfR and determined by western blotting. We have now included this experiment (Figure S4D). The inference from this experiment is that centrifugation and subsequently incubation will have an effect on receptor detection and endocytosis rates for a discreet time period.

The amount of binding of labelled transferrin to cells in culture will depend on the specific activity of the labelled transferrin. This reasoning was behind the use of 0.5 µM labelled transferrin when roughly 1 in 6 molecules in the culture medium are labelled and there was only a small effect on the overall concentration of transferrin.

Third, the authors have themselves argued previously, based on binding affinities, that all surface-exposed TfR is likely ligated in a natural setting (DOI:10.1002/bies.202400053). Could the observed binding actually be non-specific due to the high levels of fixative used?

The absence of binding/uptake of canine transferrin argues against a non-specific interaction. In our previous publication, we did not pay enough attention to the on and off rates which allow for a degree of exchange and, here, TfR newly appearing on the cell surface has a 1 in 6 chance of binding a labelled transferrin.

(3) Variable TfR expression in different BESs: It appears that native TfR is expressed at higher levels from BES7 compared to BES1, and even more so when compared to BES3. This raises the possibility that the anti-TfR used in these experiments has differential reactivity with the three sets of TfRs. The authors discount this possibility due to the overall high sequence similarities of E6s and E7s from the various ESs. However, their own analyses show that the BES1, BES3, and BES7 TfRs are relatively distal to each other in the phylogenetic trees, and this Reviewer strongly suspects that the apparent difference in expression is due to differential reactivity with the anti-TfR used in this work. In the grand scheme, this is a minor issue that does not impact the other major conclusions concerning TfR localization and function, nor the behavior of HpHbR and FHR. However, the authors make very strong conclusions about the role of BESs in TfR expression levels, even claiming that it is the 'dominant determinant' (line 189).

This point is valid but exceptionally difficult to address at the protein level. As an orthogonal approach, we performed RNAseq analysis of the ‘wild type’ BES1, BES3, and BES7 cell lines to determine whether differences in receptor mRNA levels were consistent with the proposed difference in protein levels (Table S1). The analysis showed total ESAG6/7 mRNA levels to vary in a similar manner to the protein estimates with BES3 < BES1 < BES7 providing support for the differences in protein levels.

The strongest evidence for the expression site determining the TfR level is the comparison of the cell lines in which the VSG were exchanged. This had no effect on TfR levels and so there is no evidence that the identity of the VSG alters TfR expression.

(4) Surface immuno-localization of receptors: These experiments are compelling and useful to the field. To explain the difference with essentially all prior studies, the authors suggest that typical fixation procedures allow for clearance of receptor:ligand complexes by hydrodynamic flow due to extended manipulation prior to fixation (washing steps). Despite the fact that these protocols typically involve ice-cold physiological buffers that minimize membrane mobility, this is a reasonable possibility. Have the authors challenged their hypothesis by testing more typical protocols themselves? Other contributing factors that could play a role are the use of deconvolution, which tends to minimize weak signals, and also the fact that investigators tend to discount weak surface signals as background relative to stronger internal signals.

We have added preliminary experiments that compared fixation protocols in two parts. First the effect on TfR levels of washing and resuspending cells discussed above (Figure S4D), and second how different fixation protocols alter apparent TfR immunolocalisation (Supp Figure S5A-B). The comparison shows that both the absence of glutaraldeyde and the use of washing alters the outcome.

(5) Shedding: A central aspect of the GPI valence model (Schwartz et al., 2005, Tiengwe et al., 2017) is that GPI1 reporters that reach the cell body surface are shed into the media because a single dimyristoylglycerol-containing GPI anchor does not stably associate with biological membranes. As the authors point out, this is a major factor contributing to higher steady-state levels of cell-associated GPI2 TfR relative to GPI1 TfR. Those studies also found that the size/complexity of the attached protein correlated inversely with shedding, suggesting exit from the flagellar pocket as a restricting factor in cell body surface localization. The amount of newly synthesized TfR shed into the media was ~5%, indicating that very little actually exits the FP to the outer surface. In this regard, is it possible to know the overall ratio of cell surface:FP:endosomal localized receptors? Could these data not be 'harvested' from the 3D structural illumination imaging?

A ratio could be determined but we did not do this as it would only be valid if the antibody has equal access to the internal TfR in a diluted VSG environment and the external VSG embedded in a densely packed and cross-linked VSG layer As such, we would have no confidence in the accuracy of any estimate.

Reviewer #2 (Public review):

The work has significant implications for understanding immune evasion and nutrient uptake mechanisms in trypanosomes.

While the experimental rigor is commendable, revisions are needed to clarify methodological limitations and to broaden the discussion of functional consequences.

The authors argue that prior studies missed surface-localized TfR due to harsh washing/fixation (e.g., methanol). While this is plausible, additional evidence would strengthen the claim.

Preliminary experiments that compared fixation protocols are now included to show that method affects outcome.

It remains unclear how centrifugation steps of various lengths (as in previous publications) can equally and quantitatively redistribute TfR into the flagellar pocket. If this were the case, it should be straightforward for the authors to test this experimentally.

Not aware of previous studies that demonstrate equal and quantitative redistribution to the flagellar pocket. In previous reports, there is variation in cell surface/flagellar pocket localisation depending on expression levels, for example (Mussmann et al., 2003) (Mussmann et al., 2004), it’s worth noting that the increase in TfR expression in these papers is similar to the difference in the cell lines used here. In addition, most report the presence of TfR in endosomal compartments. In the experiments here, there are cells where the majority of signal from labelled transferrin is present in the flagellar pocket and the argument is that this is a stage of a continuous process in which the receptor picks up a transferrin on the cell surface and is swept towards the pocket.

If TfR is distributed over the cell surface, live-cell imaging with fluorescent transferrin should be performed as a control. Modern detection limits now reach the singlemolecule level, and transient immobilization of live trypanosomes has been established, which would exclude hydrodynamic surface clearance as a confounding factor.

This is non-trivial and is a longer-term aim. The immobilisation involves significant manipulation of the cells prior to restraining.

In most images, TfR is not evenly distributed on the surface but rather appears punctate. Could this reflect localization to membrane domains? Immuno-EM with high-pressure frozen parasites could resolve this question and is relatively straightforward.

There is a non-uniform appearance in the super-resolution images for both TfR and FHR. We cannot distinguish whether this represents random variation in receptor density over the cell surface or results from a biological phenomenon. Whatever the cause, the experiments showed unambiguous cell surface localisation.

The authors might consider discussing whether differences in parasite life cycle stages (procyclic versus bloodstream forms) or culture conditions (e.g., cell density) affect localization. The developmentally regulated retention of GPI-anchored procyclin in the flagellar pocket might be worth mentioning.

The aim of this paper was to determine the localisation of receptors in proliferating bloodstream form trypanosomes in culture. TfR and HpHbR are not expressed in insect stages in culture. FHR is expressed in insect stages and is present all over the cell surface (Macleod et al., 2020). A procyclin-based reporter was distributed over the whole cell surface in one report (Schwartz et al. 2005). In other reports, the retention of procyclin in the flagellar pocket of proliferating bloodstream forms is probably dependent on structure/sequence as other single GPI-anchored proteins, such as FHR (Macleod et al., 2020) and GPI-anchored sfGFP (Martos-Esteban et al., 2022) can access the surface.

References:

MacGregor, P., Gonzalez-Munoz, A. L., Jobe, F., Taylor, M. C., Rust, S., Sandercock, A. M., Macleod, O. J. S., Van Bocxlaer, K., Francisco, A. F., D’Hooge, F., Tiberghien, A., Barry, C. S., Howard, P., Higgins, M. K., Vaughan, T. J., Minter, R., & Carrington, M. (2019). A single dose of antibody-drug conjugate cures a stage 1 model of African trypanosomiasis. PLoS Neglected Tropical Diseases, 13(5), e0007373. https://doi.org/10.1371/journal.pntd.0007373

Macleod, O. J. S., Bart, J.-M., MacGregor, P., Peacock, L., Savill, N. J., Hester, S., Ravel, S., Sunter, J. D., Trevor, C., Rust, S., Vaughan, T. J., Minter, R., Mohammed, S., Gibson, W., Taylor, M. C., Higgins, M. K., & Carrington, M. (2020). A receptor for the complement regulator factor H increases transmission of trypanosomes to tsetse flies. Nature Communications, 11(1), 1326. https://doi.org/10.1038/s41467-020-15125-y

Martos-Esteban, A., Macleod, O. J. S., Maudlin, I., Kalogeropoulos, K., Jürgensen, J. A., Carrington, M., & Laustsen, A. H. (2022). Black-necked spitting cobra (Naja nigricollis) phospholipases A2 may cause Trypanosoma brucei death by blocking endocytosis through the flagellar pocket. Scientific Reports, 12(1), 6394. https://doi.org/10.1038/s41598-02210091-5

Mussmann, R., Engstler, M., Gerrits, H., Kieft, R., Toaldo, C. B., Onderwater, J., Koerten, H., van Luenen, H. G. A. M., & Borst, P. (2004). Factors affecting the level and localization of the transferrin receptor in Trypanosoma brucei. The Journal of Biological Chemistry, 279(39), 40690–40698. https://doi.org/10.1074/jbc.M404697200

Mussmann, R., Janssen, H., Calafat, J., Engstler, M., Ansorge, I., Clayton, C., & Borst, P. (2003). The expression level determines the surface distribution of the transferrin receptor in Trypanosoma brucei. Molecular Microbiology, 47(1), 23–35. https://doi.org/10.1046/j.13652958.2003.03245.x

Schwartz, K. J., Peck, R. F., Tazeh, N. N., & Bangs, J. D. (2005). GPI valence and the fate of secretory membrane proteins in African trypanosomes. Journal of Cell Science, 118(Pt 23), 5499–5511. https://doi.org/10.1242/jcs.02667

Trevor, C. E., Gonzalez-Munoz, A. L., Macleod, O. J. S., Woodcock, P. G., Rust, S., Vaughan, T. J., Garman, E. F., Minter, R., Carrington, M., & Higgins, M. K. (2019). Structure of the trypanosome transferrin receptor reveals mechanisms of ligand recognition and immune evasion. Nature Microbiology, 4(12), 2074–2081. https://doi.org/10.1038/s41564-019-0589-0

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Major Recommendations:

(1) 2 E6 gene in BES7s: This does not affect the overall conclusions, but the text should be modified to reflect the existence of the second gene, and to discuss the ramifications.

This has been corrected

(2) Surface binding studies: To clarify this issue, two experimental approaches are strongly recommended. First: additional excess unlabelled Tf should be added. If binding is truly receptor-mediated, it must by definition be saturable at some experimentally achievable level. Second: TfR expression should be abrogated by RNAi silencing to show that binding is TfR-dependent. Without some validation of specific binding by one or both of these approaches, these counter-intuitive results must be questioned.

The excess unlabelled transferrin experiment is now included (we would like to thank the reviewer for this suggestion). The absence of binding of canine transferrin provides strong evidence for the specificity.

(3) Variable TfR expression in different BESs: To make such claims, quantitative RTPCR should be performed with conserved primers to assess the actual relative expression at the transcriptional level. Absent this, the claims should be eliminated, or at the very least greatly tempered.

This has been done using an RNAseq analysis.

(4) Surface immuno-localization of receptors: An example of discounting weak signals as background can be seen in Figure 8 of Duncan et al. (2024). It has also been shown that at least one other GPI1 reporter (procyclin) is readily detected on the outer cell surface under ectopic expression in BSF trypanosomes (Schwartz et al., 2005) using typical fixation procedures. This could be cited, and the authors could discuss the fact that procyclin is not a receptor and may not be susceptible to hydrodynamic drag.

Yes

Minor issues:

(1) Fully appreciating the data presented requires an understanding of the hydrodynamic flow and GPI valence models of the Engstler and Bangs labs, respectively. For the uninitiated,d it might perhaps be useful to include brief summaries of each in the Introduction.

Added to the introduction

(2) Lines 110-112: ISG65 and ISG75 both have strong localizations in endosomal compartments. This should be noted with citation of any of the work from the Field lab.

Added

(3) Lines 121-132: This passage presents the role of GPI anchors (1 vs 2) in a rather digital manner (in or out). Schwartz et al (2005) present a much more nuanced view of what is likely taking place. This is one reason summaries of hydrodynamic flow and GPI valence would be helpful.

Modified

(4) Lines 182-184: The increased size of GPI-anchored E7 is in part due to the presence of the GPI itself, as the authors state, but there are also 24 additional amino acid residues in this protein that contribute.

Modified

(5) Lines 212-214: Do p>0.95 and p>0.99 indicate statistical significance? This must be a typo.

Thank you, corrected

(6) Lines 218-219: The better references documenting GPI number in regard to turnover/shedding are Schwartz et al. 2005 and Tiengwe et al. 2017.

Changed

(7) Line 241 and Figures 3, 4, and 6: The transverse sections add little to the presentation. That there is signal variation in all dimensions is readily apparent from the images themselves, and similar profiles would be obtained regardless of the transect. Was there some process/rationale in the selection of the individual transects intended to make a broader point? If so, a description of the process should be provided.

The point was to show that the signal had a pattern consistent with plasma membrane (two distal peaks) as opposed to cytoplasm (single central peak). As such, we think it is important.

(8) Lines 582-596: Methodology for quantitation of cellular fluorescent signals should be provided.

Has been expanded

Reviewer #2 (Recommendations for the authors):

(1) As a less critical but still useful control, antibody accessibility assays on live versus fixed parasites could test whether VSG coats limit detection.

This could only be quantified by using a range of monoclonal antibodies which are not available.

(2) The rapid transferrin uptake (15-60 seconds) could reflect fast endocytic recycling rather than stable surface residency. A pulse-chase experiment tracking receptor movement would clarify this (though I acknowledge that this is technically challenging).

We agree that endocytic recycling is probably the main source of unoccupied TfR on the cell surface. It is hard to see how the pulse chase experiment could be performed without centrifugation which will affect the outcome – see above.

(3) Statistical and quantitative reporting

Added as Table S2- S4

(4) Report confidence intervals (e.g., for fluorescence intensity comparisons in Figure 3B) to contextualize claims of "no significant difference."

We do not claim ‘no significant difference’ and the SD overlap due to a high level of variation in the population

(5) Specify the number of biological replicates and cells analyzed per condition in the figure legends.

Added

(6) The study notes that surface-exposed receptors avoid antibody detection, but does not explore how.

We don’t claim that receptors avoid detection and have published evidence to the contrary. The cell has evolved mechanisms to reduce/minimise the effect of antibody binding.

(7) Comparing antibody binding to TfR in VSG221 versus VSG224 coats.

This is already present in Figure 3D

(8) Testing whether receptor shedding or conformational masking contributes to immune evasion.

A lifetime’s work

(9) Evolutionary trade-offs: Discuss why T. brucei maintains ~15 TfR variants if the GPI-anchor number has minimal impact on function (Figure 3).

The possible reason for the evolution of ~15 TfR variants was discussed in a previous publication.

(10) How do their findings align with recent studies on ISG75 surface exposure?

If this refers to the finding that ISG75 is an Ig Fc receptor, this has been included

(11) Add scale bars to 3D reconstructions (Figure 5).

Added

(12) Include a schematic summarizing key findings in the main text.

Chosen not to do

(13) Explicitly state where raw microscopy images, flow cytometry data, and analysis scripts are deposited.

Microscope Images have deposited in Bioimage Archive repository at EMBL/EBI No flow cytometry used

(14) Correct inconsistent GPI-anchor terminology (e.g., "glycosylphosphoinositol" to "glycosylphosphatidylinositol").

Our typo, corrected

(15) Clarify ambiguous phrases (e.g., "subtle mechanisms" in the Discussion).

Corrected