I did not previously think about how age impacts the way you experience poverty, but I can see how this may be true. When we are little kids, we are still unaware of a lot of the different facets of identity that set us apart, and are more likely to be open minded to more things-- we are still very easily impressionable. As we become older though, and cliques form, there is a way clearer understanding of what may be deemed as cool or desirable for teenagers and what is embarrassing or something to be ashamed of.

Author response:

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

Editor’s summary:

This paper by Castello-Serrano et al. addresses the role of lipid rafts in trafficking in the secretory pathway. By performing carefully controlled experiments with synthetic membrane proteins derived from the transmembrane region of LAT, the authors describe, model and quantify the importance of transmembrane domains in the kinetics of trafficking of a protein through the cell. Their data suggest affinity for ordered domains influences the kinetics of exit from the Golgi. Additional microscopy data suggest that lipid-driven partitioning might segregate Golgi membranes into domains. However, the relationship between the partitioning of the synthetic membrane proteins into ordered domains visualised ex vivo in GPMVs, and the domains in the TGN, remain at best correlative. Additional experiments that relate to the existence and nature of domains at the TGN are necessary to provide a direct connection between the phase partitioning capability of the transmembrane regions of membrane proteins and the sorting potential of this phenomenon.

The authors have used the RUSH system to study the traffic of model secretory proteins containing single-pass transmembrane domains that confer defined affinities for liquid ordered (lo) phases in Giant Plasma Membrane derived Vesicles (GPMVs), out of the ER and Golgi. A native protein termed LAT partitioned into these lo-domains, unlike a synthetic model protein termed LAT-allL, which had a substituted transmembrane domain. The authors experiments provide support for the idea that ER exit relies on motifs in the cytosolic tails, but that accelerated Golgi exit is correlated with lo domain partitioning.

Additional experiments provided evidence for segregation of Golgi membranes into coexisting lipid-driven domains that potentially concentrate different proteins. Their inference is that lipid rafts play an important role in Golgi exit. While this is an attractive idea, the experiments described in this manuscript do not provide a convincing argument one way or the other. It does however revive the discussion about the relationship between the potential for phase partitioning and its influence on membrane traffic.

We thank the editors and scientific reviewers for thorough evaluation of our manuscript and for positive feedback. While we agree that our experimental findings present a correlation between trafficking rates and raft affinity, in our view, the synthetic, minimal nature of the transmembrane protein constructs in question makes a strong argument for involvement of membrane domains in their trafficking. These constructs have no known sorting determinants and are unlikely to interact directly with trafficking proteins in cells, since they contain almost no extramembrane amino acids. Yet, the LATTMD traffics through Golgi similarly to the full-length LAT protein, but quite different from mutants with lower raft phase affinity. We suggest that these observations can be best rationalized by involvement of raft domains in the trafficking fates and rates of these constructs, providing strong evidence (beyond a simple correlation) for the existence and relevance of such domains.

We have substantially revised the manuscript to address all reviewer comments, including several new experiments and analyses. These revisions have substantially improved the manuscript without changing any of the core conclusions and we are pleased to have this version considered as the “version of record” in eLife.

Below is our point-by-point response to all reviewer comments.

ER exit:

The experiments conducted to identify an ER exit motif in the C-terminal domain of LAT are straightforward and convincing. This is also consistent with available literature. The authors should comment on whether the conservation of the putative COPII association motif (detailed in Fig. 2A) is significantly higher than that of other parts of the C-terminal domain.

Thank you for this suggestion, this information has now been included as Supp Fig 2B. While there are other wellconserved residues of the LAT C-terminus, many regions have relatively low conservation. In contrast, the essential residues of the COPII association motif (P148 and A150) are completely conserved across in LAT across all species analyzed.

One cause of concern is that addition of a short cytoplasmic domain from LAT is sufficient to drive ER exit, and in its absence the synthetic constructs are all very slow. However, the argument presented that specific lo phase partitioning behaviour of the TMDs do not have a significant effect on exit from the ER is a little confusing. This is related to the choice of the allL-TMD as the 'non-lo domain' partitioning comparator. Previous data has shown that longer TMDs (23+) promote ER export (eg. Munro 91, Munro 95, Sharpe 2005). The mechanism for this is not, to my knowledge, known. One could postulate that it has something to do with the very subject of this manuscript- lipid phase partitioning. If this is the case, then a TMD length of 22 might be a poor choice of comparison. A TMD 17 Ls' long would be a more appropriate 'non-raft' cargo. It would be interesting to see a couple of experiments with a cargo like this.

The basis for the claim that raft affinity has relatively minor influence on ER exit kinetics, especially in comparison to the effect of the putative COPII interaction motif, is in Fig 1G. We do observe some differences between constructs and they may be related to raft affinity, however we considered these relatively minor compared to the nearly 4-fold increase in ER efflux induced by COPII motifs.

We have modified the wording in the manuscript to avoid the impression that we have ruled out an effect of raft affinity of ER exit.

We believe that our observations are broadly consistent with those of Munro and colleagues. In both their work and ours, long TMDs were able to exit the ER. In our experiments, this was true for several proteins with long TMDs, either as fulllength or as TMD-only versions (see Fig 1G). We intentionally did not measure shorter synthetic TMDs because these would not have been comparable with the raft-preferring variants, which all require relatively long TMDs, as demonstrated in our previous work1,2. Thus, because our manuscript does not make any claims about the influence of TMD length on trafficking, we did not feel that experiments with shorter non-raft constructs would substantively influence our conclusions.

However, to address reviewer interest, we did complete one set of experiments to test the effect of shortening the TMD on ER exit. We truncated the native LAT TMD by removing 6 residues from the C-terminal end of the TMD (LAT-TMDd6aa). This construct exited the ER similarly to all others we measured, revealing that for this set of constructs, short TMDs did not accumulate in the ER. ER exit of the truncated variant was slightly slower than the full-length LAT-TMD, but somewhat faster than the allL-TMD. These effects are consistent with our previous measurements with showed that this shortened construct has slightly lower raft phase partitioning than the LAT-TMD but higher than allL2. While these are interesting observations, a more thorough exploration of the effect of TMD length would be required to make any strong conclusion, so we did not include these data in the final manuscript.

Author response image 1.

Golgi exit:

For the LAT constructs, the kinetics of Golgi exit as shown in Fig. 3B are surprisingly slow. About half of the protein Remains in the Golgi at 1 h after biotin addition. Most secretory cargo proteins would have almost completely exited the Golgi by that time, as illustrated by VSVG in Fig. S3. There is a concern that LAT may have some tendency to linger in the Golgi, presumably due to a factor independent of the transmembrane domain, and therefore cannot be viewed as a good model protein. For kinetic modeling in particular, the existence of such an additional factor would be far from ideal. A valuable control would be to examine the Golgi exit kinetics of at least one additional secretory cargo.

We disagree that LAT is an unusual protein with respect to Golgi efflux kinetics. In our experiments, Golgi efflux of VSVG was similar to full-length LAT (t1/2 ~ 45 min), and both of these were similar to previously reported values3. Especially for the truncated (i.e. TMD) constructs, it is very unlikely that some factor independent of their TMDs affects Golgi exit, as they contain almost no amino acids outside the membrane-embedded TMD.

Practically, it has proven somewhat challenging to produce functional RUSH-Golgi constructs. We attempted the experiment suggested by the reviewer by constructing SBP-tagged versions of several model cargo proteins, but all failed to trap in the Golgi. We speculate that the Golgin84 hook is much more sensitive to the location of the SBP on the cargo, being an integral membrane protein rather than the lumenal KDEL-streptavidin hook. This limitation can likely be overcome by engineering the cargo, but we did not feel that another control cargo protein was essential for the conclusions we presented, thus we did not pursue this direction further.

Comments about the trafficking model

(1) In Figure 1E, the export of LAT-TMD from the ER is fitted to a single-exponential fit that the authors say is "well described". This is unclear and there is perhaps something more complex going on. It appears that there is an initial lag phase and then similar kinetics after that - perhaps the authors can comment on this?

This is a good observation. This effect is explainable by the mechanics of the measurement: in Figs 1 and 2, we measure not ‘fraction of protein in ER’ but ‘fraction of cells positive for ER fluorescence’. This is because the very slow ER exit of the TMD-only constructs present a major challenge for live-cell imaging, so ER exit was quantified on a population level, by fixing cells at various time points after biotin addition and quantifying the fraction of cells with observable ER localization (rather than tracking a single cell over time).

For fitting to the kinetic model (which attempts to describe ‘fraction in ER/Golgi’) we re-measured all constructs by livecell imaging (see Supp Fig 5) to directly quantify relative construct abundance in the ER or Golgi. These data did not have the plateau in Fig 1E, suggesting that this is an artifact of counting “ER positive cells” which would be expected to have a longer lag than “fraction of protein in ER”. Notably however, t1/2 measured by both methods was similar, suggesting that the population measurement agrees well with single-cell live imaging.

We have included all these explanations and caveats in the manuscript. We have also changed the wording from “well described” to “reasonably approximated”.

(2) The model for Golgi sorting is also complicated and controversial, and while the authors' intention to not overinterpreting their data in this regard must be respected, this data is in support of the two-phase Golgi export model (Patterson et al PMID:18555781).

The reviewers are correct, our observations and model are consistent with Patterson et al and it was a major oversight that a reference to this foundational work was not included. We have now added a discussion regarding the “two phase model” of Patterson and Lippincott-Schwartz.

Furthermore contrary to the statement in lines 200-202, the kinetics of VSVG exit from the Golgi (Fig. S3) are roughly linear and so are NOT consistent with the previous report by Hirschberg et al.

Regarding kinetics of VSVG, our intention was to claim that the timescale of VSVG efflux from the Golgi was similar to previously reported in Hirschberg, i.e. t1/2 roughly between 30-60 minutes. We have clarified this in the text. Minor differences in the details between our observations and Hirschberg are likely attributable to temperature, as those measurements were done at 32°C for the tsVSVG mutant.

Moreover, the kinetics of LAT export from the Golgi (Fig. 3B) appear quite different, more closely approximating exponential decay of the signal. These points should be described accurately and discussed.

Regarding linear versus exponential fits, we agree that the reality of Golgi sorting and efflux is far more complicated than accounted for by either the phenomenological curve fitting in Figs 1-3 or the modeling in Fig 4. In addition to the possibility of lateral domains within Golgi stacks, there is transport between stacks, retrograde traffic, etc. The fits in Figs 1-3 are not intended to model specifics of transport, but rather to be phenomenological descriptors that allowed us to describe efflux kinetics with one parameter (i.e. t1/2). In contrast, the more refined kinetic modeling presented in Figure 4 is designed to test a mechanistic hypothesis (i.e. coexisting membrane domains in Golgi) and describes well the key features of the trafficking data.

Relationship between membrane traffic and domain partitioning:

(1) Phase segregation in the GPMV is dictated by thermodynamics given its composition and the measurement temperature (at low temperatures 4degC). However at physiological temperatures (32-37degC) at which membrane trafficking is taking place these GPMVs are not phase separated. Hence it is difficult to argue that a sorting mechanism based solely on the partitioning of the synthetic LAT-TMD constructs into lo domains detected at low temperatures in GPMVs provide a basis (or its lack) for the differential kinetics of traffic of out of the Golgi (or ER). The mechanism in a living cell to form any lipid based sorting platforms naturally requires further elaboration, and by definition cannot resemble the lo domains generated in GPMVs at low temperatures.

We thank the reviewers for bringing up this important point. GPMVs are a useful tool because they allow direct, quantitative measurements of protein partitioning between coexisting ordered and disordered phases in complex, cell-derived membranes. However, we entirely agree, that GPMVs do not fully represent the native organization of the living cell plasma membrane and we have previously discussed some of the relevant differences4,5. Despite these caveats, many studies have supported the cellular relevance of phase separation in GPMVs and the partitioning of proteins to raft domains therein 6-9. Most notably, elegant experiments from several independent labs have shown that fluorescent lipid analogs that partition to Lo domains in GPMVs also show distinct diffusive behaviors in live cells 6,7, strongly suggesting the presence of nanoscopic Lo domains in live cells. Similarly, our recent collaborative work with the lab of Sarah Veatch showed excellent agreement between raft preference in GPMVs and protein organization in living immune cells imaged by super-resolution microscopy10. Further, several labs6,7, including ours11, have reported nice correlations between raft partitioning in GPMVs and detergent resistance, which is a classical (though controversial) assay for raft association.

Based on these points, we feel that GPMVs are a useful tool for quantifying protein preference for ordered (raft) membrane domains and that this preference is a useful proxy for the raft-associated behavior of these probes in living cells. We propose that this approach allows us to overcome a major reason for the historical controversy surrounding the raft field: nonquantitative and unreliable methodologies that prevented consistent definition of which proteins are supposed to be present in lipid rafts and why. Our work directly addresses this limitation by relating quantitative raft affinity measurements in a biological membrane with a relevant and measurable cellular outcome, specifically inter-organelle trafficking rates.

Addressing the point about phase transition temperatures in GPMVs: this is the temperature at which macroscopic domains are observed. Based on physical models of phase separation, it has been proposed that macroscopic phase separation at lower temperatures is consistent sub-microscopic, nanoscale domains at higher temperatures8,12. These smaller domains can potentially be stabilized / functionalized by protein-protein interactions in cells13 that may not be present in GPMVs (e.g. because of lack of ATP).

(2) The lipid compositions of each of these membranes - PM, ER and Golgi are drastically different. Each is likely to phase separate at different phase transition temperatures (if at all). The transition temperature is probably even lower for Golgi and the ER membranes compared to the PM. Hence, if the reported compositions of these compartments are to be taken at face value, the propensity to form phase separated domains at a physiological temperature will be very low. Are ordered domains even formed at the Golgi at physiological temperatures?

It is a good point that the membrane compositions and the resulting physical properties (including any potential phase behavior) will be very different in the PM, ER, and Golgi. Whether ordered domains are present in any of these membranes in living cells remains difficult to directly visualize, especially for non-PM membranes which are not easily accessible by probes, are nanoscopic, and have complex morphologies. However, the fact that raft-preferring probes / proteins share some trafficking characteristics, while very similar non-raft mutants behave differently argues that raft affinity plays a role in subcellular traffic.

(3) The hypothesis of 'lipid rafts' is a very specific idea, related to functional segregation, and the underlying basis for domain formation has been also hotly debated. In this article the authors conflate thermodynamic phase separation mechanisms with the potential formation of functional sorting domains, further adding to the confusion in the literature. To conclude that this segregation is indeed based on lipid environments of varying degrees of lipid order, it would probably be best to look at the heterogeneity of the various membranes directly using probes designed to measure lipid packing, and then look for colocalization of domains of different cargo with these domains.

This is a very good suggestion, and a direction we are currently following. Unfortunately, due to the dynamic nature and small size of putative lateral membrane domains, combined with the interior of a cell being filled with lipophilic environments that overlay each other, directly imaging domains in organellar membranes with lipid packing probes remains extremely difficult with current technology (or at least available to us). We argue that the TMD probes used in this manuscript are a reasonable alternative, as they are fluorescent probes with validated selectivity for membrane compartments with different physical properties.

Ultimately, the features of membrane domains suggested by a variety of techniques – i.e. nanometric, dynamic, relatively similar in composition to the surrounding membrane, potentially diverse/heterogeneous – make them inherently difficult to microscopically visualize. This is one reason why we believe studies like ours, which use a natural model system to directly quantify raft-associated behaviors and relate them to cellular effects (in our case, protein sorting), are a useful direction for this field.

We believe we have been careful in our manuscript to avoid confusing language surrounding lipid rafts, phase separation, etc. Our experiments clearly show that mammalian membranes have the capacity to phase separate, that some proteins preferentially interact with more ordered domains, and that this preference is related to the subcellular trafficking fates and rates of these proteins. We have edited the manuscript to emphasize these claims and avoid the historical controversies and confusions.

(4) In the super-resolution experiments (by SIM- where the enhancement of resolution is around two fold or less compared to optical), the authors are able to discern a segregation of the two types of Golgi-resident cargo that have different preferences for the lo-domains in GPMVs. It should be noted that TMD-allL and the LATallL end up in the late endosome after exit of the Golgi. Previous work from the Bonafacino laboratory (PMID: 28978644) has shown that proteins (such as M6PR) destined to go to the late endosome bud from a different part of the Golgi in vesicular carriers, while those that are destined for the cell surface first (including TfR) bud with tubular vesicular carriers. Thus at the resolution depicted in Fig 5, the segregation seen by the authors could be due to an alternative explanation, that these molecules are present in different areas of the Golgi for reasons different from phase partitioning. The relatively high colocalization of TfR with the GPI probe in Fig 5E is consistent with this explanation. TfR and GPI prefer different domains in the GPMV assays yet they show a high degree of colocalization and also traffic to the cell surface.

This is a good point. Even at microscopic resolutions beyond the optical diffraction limit, we cannot make any strong claims that the segregation we observe is due to lateral lipid domains and not several reasonable alternatives, including separation between cisternae (rather than within), cargo vesicles moving between cisternae, or lateral domains that are mediated by protein assemblies rather than lipids. We have explicitly included this point in the Discussion: “Our SIM imaging suggests segregation of raft from nonraft cargo in the Golgi shortly (5 min) after RUSH release (Fig 5B), but at this level of resolution, we can only report reduced colocalization, not intra-Golgi protein distributions. Moreover, segregation within a Golgi cisterna would be very difficult to distinguish from cargo moving between cisternae at different rates or exiting via Golgi-proximal vesicles.”

We have also added a similar caveat in the Results section of the manuscript: “These observations support the hypothesis that proteins can segregate in Golgi based on their affinity for distinct membrane domains; however, it is important to emphasize that this segregation does not necessarily imply lateral lipid-driven domains within a Golgi cisterna. Reasonable alternative possibilities include separation between cisternae (rather than within), cargo vesicles moving between cisternae, or lateral domains that are mediated by protein assemblies rather than lipids.”

Finally, while probes with allL TMD do eventually end up in late endosomes (consistent with the Bonifacino lab’s findings which we include), they do so while initially transiting the PM2,11.

Minor concerns:

(1) Generally, the quantitation is high quality from difficult experimental data. Although a lot appears to be manual, it appears appropriately performed and interpreted. There are some claims that are made based on this quantitation, however, where there are no statistics performed. For example, figure 1B. Any quantitation with an accompanying conclusion should be subject to a statistical test. I think the quality of the model fits- this is particularly important.

We appreciate the thoughtful feedback, the quantifications and fits were not trivial, but we believe important. We have added statistical significance to Figure 1B and others where it was missing.

(2) Modulation of lipid levels in Fig 4E shows a significant change for the trafficking rate for the LAT-TMD construct and a not so significant change for all-TMD construct. However, these data are not convincing and appear to depend on a singular data point that appears to lower the mean value. In general, the experiment with the MZA inhibitor (Fig. 4D-F) is hard to interpret because cells will likely be sick after inhibition of sphingolipid and cholesterol synthesis. Moreover, the difference in effects for LAT-TMD and allL-TMD is marginal.

We disagree with this interpretation. Fig 4E shows the average of three experiments and demonstrates clearly that the inhibitors change the Golgi efflux rate of LAT-TMD but not allL-TMD. This is summarized in the t1/2 quantifications of Fig 4F, which show a statistically significant change for LAT-TMD but not allL-TMD. This is not an effect of a singular data point, but rather the trend across the dataset.

Further, the inhibitor conditions were tuned carefully to avoid cells becoming “sick”: at higher concentrations, cells did adopt unusual morphologies and began to detach from the plates. We pursued only lower concentrations, which cells survived for at least 48 hrs and without major morphological changes.

(3) Line 173: 146-AAPSA-152 should read either 146-AAPSA-150 or 146-AAPSAPA-152, depending on what the authors intended.

Thanks for the careful reading, we intended the former and it has been fixed.

(4) What is the actual statistical significance in Fig. 3C and Fig. 3E? There is a single asterisk in each panel of the figure but two asterisks in the legend.

Apologies, a single asterisk representing p<0.05 was intended. It has been fixed.

(5) The code used to calculate the model. is not accessible. It is standard practice to host well-annotated code on Github or similar, and it would be good to have this publicly available.

We have deposited the code on a public repository (doi: 10.5281/zenodo. 10478607) and added a note to the Methods.

(1) Lorent, J. H. et al. Structural determinants and func7onal consequences of protein affinity for membrane ra=s. Nature communica/ons 8, 1219 (2017).PMC5663905

(2) Diaz-Rohrer, B. B., Levental, K. R., Simons, K. & Levental, I. Membrane ra= associa7on is a determinant of plasma membrane localiza7on. Proc Natl Acad Sci U S A 111, 8500-8505 (2014).PMC4060687

(3) Hirschberg, K. et al. Kine7c analysis of secretory protein traffic and characteriza7on of golgi to plasma membrane transport intermediates in living cells. J Cell Biol 143, 1485-1503 (1998).PMC2132993

(4) Levental, K. R. & Levental, I. Giant plasma membrane vesicles: models for understanding membrane organiza7on. Current topics in membranes 75, 25-57 (2015)

(5) Sezgin, E. et al. Elucida7ng membrane structure and protein behavior using giant plasma membrane vesicles. Nat Protoc 7, 1042-1051 (2012)

(6) Komura, N. et al. Ra=-based interac7ons of gangliosides with a GPI-anchored receptor. Nat Chem Biol 12, 402-410 (2016)

(7) Kinoshita, M. et al. Ra=-based sphingomyelin interac7ons revealed by new fluorescent sphingomyelin analogs. J Cell Biol 216, 1183-1204 (2017).PMC5379944

(8) Stone, M. B., Shelby, S. A., Nunez, M. F., Wisser, K. & Veatch, S. L. Protein sor7ng by lipid phase-like domains supports emergent signaling func7on in B lymphocyte plasma membranes. eLife 6 (2017).PMC5373823

(9) Machta, B. B. et al. Condi7ons that Stabilize Membrane Domains Also Antagonize n-Alcohol Anesthesia. Biophys J 111, 537-545 (2016)

(10) Shelby, S. A., Castello-Serrano, I., Wisser, I., Levental, I. & S., V. Membrane phase separa7on drives protein organiza7on at BCR clusters. Nat Chem Biol in press (2023)

(11) Diaz-Rohrer, B. et al. Rab3 mediates a pathway for endocy7c sor7ng and plasma membrane recycling of ordered microdomains Proc Natl Acad Sci U S A 120, e2207461120 (2023)

(12) Veatch, S. L. et al. Cri7cal fluctua7ons in plasma membrane vesicles. ACS Chem Biol 3, 287-293 (2008)

(13) Wang, H. Y. et al. Coupling of protein condensates to ordered lipid domains determines func7onal membrane organiza7on. Science advances 9, eadf6205 (2023).PMC10132753

Author response:

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

Reviewer #1 (Recommendation for the authors):

(1) On a few occasions, I found that the authors would introduce a concept, but provide evidence much later on. For example, in line 57, they introduced the idea that feedback timing modulates engagement of the hippocampus and striatum, but they provided the details much later on around line 99. There are a few instances like these, and the authors may want to go through the manuscript critically to bridge such gaps to improve the flow of reading.

First, we thank the reviewer for acknowledging the contribution of our study and the methodological choices. We acknowledge the concern raised about the flow of information in the introduction. We have critically reviewed the manuscript, especially on writing style and overall structure, to ensure a smoother transition between the introduction of concepts and the provision of supporting evidence. In the case of the concept of feedback timing and memory systems, lines 46-58 first introduce the concept enhanced with evidence regarding adults, and we then pick up the concept around line 103 again to relate it to children and their brain development to motivate our research question. To further improve readability, we have included an outline of what to expect in the introduction. Specifically, we added a sentence in line 66-68 that provides an overview of the different paragraphs: “We will introduce the key parameters in reinforcement learning and then we review the existing literature on developmental trajectories in reinforcement learning as well as on hippocampus and striatum, our two brain regions of interest.”

This should prepare the reader better when to expect more evidence regarding the concepts introduced. We included similar “road-marker” outline sentences in other occasions the reviewer commented on, to enhance consistency and readability.

(2) I am curious as to how they think the 5-second delay condition maps onto real-life examples, for example in a classroom setting feedback after 5 seconds could easily be framed as immediate feedback.

The authors may want to highlight a few illustrative examples.

Thank you for asking about the practical implications of a 5-second delay condition, which may be very relevant to the reader. We have modified the introduction example in line 39-41 towards the role of feedback timing in the classroom to point out its practical relevance early on: “For example, children must learn to raise their hand before speaking during class. The teacher may reinforce this behavior immediately or with a delay, which raises the question whether feedback timing modulates their learning”.

We have also expanded a respective discussion point in lines 720-728 to pick up the classroom example and to illustrate how we think timescale differences may apply: “In scenarios such as in the classroom, a teacher may comment on a child’s behavior immediately after the action or some moments later, in par with our experimental manipulation of 1 second versus 5 seconds. Within such short range of delay in teachers’ feedback, children’s learning ability during the first years of schooling may function equally well and depend on the striatal-dependent memory system. However, we anticipate that the reliance on the hippocampus will become even more pronounced when feedback is further delayed for longer time. Children’s capacity for learning over longer timescales relies on the hippocampal-dependent memory system, which is still under development. This knowledge could help to better structure learning according to their development.”

(3) In the methods section, there are a few instances of task description discrepancies which make things a little bit confusing, for example, line 173 reward versus punishment, or reward versus null elsewhere e.g. line 229. In the same section, line 175, there are a few instances of typos.

We appreciate your attention to detail in pointing out discrepancies in task descriptions and typos in the method section. We have revised the section, corrected typos, and now phrased the learning outcomes consistently as “reward” and “punishment”.

(4). I wasn't very clear as to why the authors did not compute choice switch probability directly from raw data but implemented this as a model that makes use of a weight parameter. Former would-be much easier and straightforward for data plotting especially for uninformed readers, i.e., people who do not have backgrounds in computational modelling.

Thank you for asking for clarification on the calculation of switching behavior. Indeed, in the behavioral results, switching behavior was directly calculated from the raw data. We now stressed this in the methods in lines 230-235, also by naming win-stay and lose-shift as “proportions” instead of as “probabilities”:“As a first step, we calculated learning outcomes diretly from the raw data, which where learning accuracy, win-stay and lose-shift behavior as well as reaction time.

Learning accuracy was defined as the proportion to choose the more rewarding option, while win-stay and lose-shift refer to the proportion of staying with the previously chosen option after a reward and switching to the alternative choice after receiving a punishment, respectively.”

In contrast to the raw data switching behavior, the computational heuristic strategy model indeed uses a weight for a relative tendency of switching behavior. We have also stressed the advantage of the computational measure and its difference to the raw data switching behavior in lines 248-252 and believe that the reader can now clearly distinguish between the raw data and the computational results: “Note that these model-based outcomes are not identical to the win-stay and lose-shift behavior that were calculated from the raw data. The use of such model-based measure offers the advantage in discerning the underlying hidden cognitive process with greather nuance, in contrast to classical approaches that directly use raw behavioral data.”

(5) I agree with the authors' assertion that both inverse temperature and outcome sensitivity parameters may lead to non-identifiability issues, but I was not 100% convinced about their modelling approach exclusively assessing a different family of models (inv temperature versus outcome sensitivity). Here, I would like to make one mid-way recommendation. They may want to redefine the inverse temperature term in terms of reaction time, i.e., B=exp^(s+g(RT-mean (RT)) where s and g are free parameters (see Webb, 2019), and keep the outcome sensitivity parameter in the model with bounds [0,2] so that the interpretation could be % increase or decrease in actual outcome. Personally, in tasks with binary outcomes i.e. [0,1: null vs reward] I do not think outcome sensitivity parameters higher than 2 are interpretable as these assign an inflated coefficient to outcomes.

We appreciate the mid-way recommendation regarding the modeling approach for inverse temperature and outcome sensitivity parameters. We have carefully revised our analysis approach by considering alternative modeling choices. Regarding the suggestion to redefine the inverse temperature in terms of reaction time by B=exp^(s+g(RT-mean (RT)), we unfortunately were not able to identify the reference Webb (2019), nor did we find references to the suggested modeling approach. Any further information that the reviewer could provide will be greatly appreciated. Regardless, we agree that including reaction times through the implementation of drift-diffusion modeling may be beneficial. However, changing the inverse temperature model in such a way would necessitate major changes in our modeling approach, which unfortunately would result in non-convergence issues in our MCMC pipeline using Rstan. Hence, this approach goes beyond the scope of the manuscript. Nonetheless, we have decided to mention the use of a drift-diffusion model, along with other methodological considerations, as future recommendation for disentangling outcome sensitivity from inverse temperature in lines 711-712: “Future studies might shed new light by examining neural activations at both task phases, by additionally modeling reaction times using a drift-diffusion approach, or by choosing a task design that allows independent manipulations of these phases and associated model parameters, e.g., by using different reward magnitudes during reinforcement learning, or by studying outcome sensitivity without decisionmaking.“

Regarding the upper bound of outcome sensitivity, we agree that traditionally, limiting the parameter values at 2 is the choice for the parameter to be best interpretable. During model fitting, we had experienced non-convergence issues and ceiling effects in the outcome sensitivity parameter when fixing the inverse temperature at 1. The non-convergence issue was not resolved when we fixed the inverse temperature at 15.47, which was the group mean of the winning inverse temperature family. Model convergence was only achieved after increasing the outcome sensitivity upper bound to 20, with inverse temperature again fixed at 1. Since this model also performed well during parameter and model recovery, we argue that the parameter is nevertheless meaningful, despite the more extreme trial-to-trial value fluctuations under higher outcome sensitivity. We described our choice for this model in the methods section in lines 282-288: “Even though outcome sensitivity is usually restricted to an upper bound of 2 to not inflate outcomes at value update, this configuration led to ceiling effects in outcome sensitivity and non-converging model results. Further, this issue was not resolved when we fixed the inverse temperature at the group mean of 15.47 of the winning inverse temperature family model. It may be that in children, individual differences in outcome sensitivity are more pronounced, leading to more extreme values. Therefore, we decided to extend the upper bound to 20, parallel to the inverse temperature, and all our models converged with Rhat < 1.1.”.

(6) I think the authors reporting optimal parameters for the model is very important (line 464), but the learning rate they report under stable contingencies is much higher than LRs reported by for example Behrens et al 2007, LRs around 0.08 for the optimal learning behaviour. The authors may want to discuss why their task design calls for higher learning rates.

Thank you for appreciating our optimal parameter analysis, and for the recommendation to discuss why optimal learning rates in our task design may call for higher learning rates compared to those reported in some other studies. As largely articulated in Zhang et al (2020; primer piece by one of our co-authors), the optimal parameter combination is determined by several factors, such as the reward schedule (e.g., 75:25, vs 80:20) and task design (e.g., no reversal, one reversal, vs multiple reversal) and number of trials (e.g., 80, vs 100, vs, 120). Notably, in these taskrelated regards, our task is different from Behrens et al. (2007), which hinders a quantitative comparison among the optimal parameters in the two tasks. We have now included more details in our discussion in lines 643-656: “However, the differences in learning rate across studies have to be interpreted with caution. The differences in the task and the analysis approach may limit their comparability. Task proporties such as the trial number per condition differed across studies. Our study included 32 trials per cue in each condition, while in adult studies, the trials per condition ranged from 28 to 100. Optimal learning rates in a stable learning environment were at around 0.25 for 10 to 30 trials, another study reported a lower optimal learning rate of around 0.08 for 120 trials. This may partly explain why in our case of 32 trials per condition and cue, optimal learning rates called for a relatively high optimal learning rate of 0.29, while in other studies, optimal learning rates may be lower. Regarding differences in the analysis approach, the hierarchical bayesian estimation approach used in our study produces more reliable results in comparison to maximum likelihood estimation, which had been used in some of the previous adult studies and may have led to biased results towards extreme values. Taken together, our study underscores the importance of using longitudinal data to examine developmental change as well as the importance of simulation-based optimal parameters to interpret the direction of developmental change.”

(7) The authors may want to report degrees of freedom in t-tests so that it would be possible to infer the final sample size for a specific analysis, for example, line 546.

We appreciate the recommendation to include degrees of freedom, which are now added in all t-test results, for example in line 579: “Episodic memory, as measured by individual corrected object recognition memory (hits - false alarms) of confident (“sure”) ratings, showed at trend better memory for items shown in the delayed feedback condition (𝛽!""#$%&’(#")%*"# = .009, SE =.005, t(df = 137) = 1.80, p = .074, see Figure 5A).”

(8) I'm not sure why reductions in lose shift behaviour are framed as an improvement between 2 assessment points, e.g. line 578. It all depends on the strength of the contingency so a discussion around this point should be expanded.

We acknowledge that a reduction in lose-shift behavior only reflect improvements under certain conditions where uncertainty is low and the learning contingencies are stable, which is the case in our task. We have added Supplementary Material 4 to illustrate the optimality of win-stay and lose-shift proportions from model simulation and to confirm that children’s longitudinal development was indeed towards more optimal switching behavior. In the manuscript, we refer to these results in lines 488-490: “We further found that the average longitudinal change in win-stay and lose-shift proportion also developed towards more optimal value-based learning (Supplementary Material 4).”

(9) If I'm not mistaken, the authors reframe a trend-level association as weak evidence. I do not think this is an accurate framing considering the association is strictly non-significant, therefore should be omitted line 585.

We thank for the point regarding the interpretation of a trend-level association as weak evidence. We changed our interpretation, corrected in lines 581-585: “The inclusion of poor learners in the complete dataset may have weakend this effect because their hippocampal function was worse and was not involved in learning (nor encoding), regardless of feedback timing. To summarize, there was inconclusive support for enhanced episodic memory during delayed compared to immediate feedback, calling for future study to test the postulation of a selective association between hippocampal volume and delayed feedback learning.” as well as lines 622-623: “Contrary to our expectations, episodic memory performance was not enhanced under delayed feedback compared to immediate feedback.”

Reviewer # 2 (Public Review):

We thank the reviewer for acknowledging the strength of our study and pointing out its weaknesses.

Weaknesses:

There were a few things that I thought would be helpful to clarify. First, what exactly are the anatomical regions included in the striatum here?

We appreciate the clarification question regarding the anatomical regions included in the striatum. The striatum included ventral and dorsal regions, i.e., accumbens, caudate and putamen. We have now specified the anatomical regions that were included in the striatum in lines 211-212: “We extracted the bilateral brain volumes for our regions of interest, which were striatum and hippocampus. The striatum regions included nucleus accumbens, caudate and putamen.”

Second, it was mentioned that for the reduced dataset, object recognition memory focused on "sure" ratings. This seems like the appropriate way to do it, but it was not clear whether this was also the case for the full analyses in the main text.

Thank you for pointing out that in the full dataset analysis, the use of “sure” ratings for object recognition memory was previously not mentioned. Including only “sure” ratings was used consistently across analyses. This detail is now described under methods in lines 332-333: “Only confident (“sure”) ratings were included in the analysis, which were 98.1 % of all given responses.”

Third, the children's fitted parameters were far from optimal; is it known whether adults would be closer to optimal on the task?

We thank for your question on whether adult learning rates in the task have been reported to be more optimal than those of the children in our study. This indeed seems to be the case, and we added this point in our discussion in line 639-643: “Adult studies that examined feedback timing during reinforcement learning reported average learning rates range from 0.12 to 0.34, which are much closer to the simulated optimal learning rates of 0.29 than children’s average learning rates of 0.02 and 0.05 at wave 1 and 2 in our study. Therefore, it is likely that individuals approach adult-like optimal learning rates later during adolescence.”

The main thing I would find helpful is to better integrate the differences between the main results reported and the many additional results reported in the supplement, for example from the reduced dataset when excluding non-learners. I found it a bit challenging to keep track of all the differences with all the analyses and parameters. It might be helpful to report some results in tables side-by-side in the two different samples. And if relevant, discuss the differences or their implication in the Discussion. For example, if the patterns change when excluding the poor learners, in particular for the associations between delayed feedback and hippocampal volume, and those participants were also those less well fit by the value-based model, is that something to be concerned about and does that affect any interpretations? What was not clear to me is whether excluding the poor learners at one extreme simply weakens the general pattern, or whether there is a more qualitative difference between learners and non-learners. The discussion points to the relevance of deficits in hippocampaldependent learning for psychopathology and understanding such a distinction may be relevant.

We appreciate the feedback that it might seem challenging to keep track of differences between the analyses of the full and the reduced dataset. We have now gathered all the analyses for the reduced dataset in Supplementary Material 6, with side-by-side tables for comparison to the full dataset results. Whenever there were differences between the results, they were pointed out in the results section, see lines 557-560: “In the results of the reduced dataset, the hippocampal association to the delayed learning score was no longer significant, suggesting a weakened pattern when excluding poor learners (Supplementary Material 6). It is likely that the exclusion reduced the group variance for hippocampal volume and delayed learning score in the model.” and lines 579-581: “Note that in the reduced dataset, delayed feedback predicted enhanced item memory significantly (Supplementary Material 6).”

The found differences were further included in our discussion in lines 737-740 in the context of deficits in hippocampal-dependent learning and psychopathology: “Interestingly, poor learners showed relatively less value-based learning in favor of stronger simple heuristic strategies, and excluding them modulated the hippocampal-dependent associations to learning and memory in our results. More studies are needed to further clarify the relationship between hippocampus and psychopathology during cognitive and brain development.”

Author response:

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

Reviewer #1 (Public Review):

(1) There appears to be a flaw in the exploration of cortical inputs. the authors never show that HFS of cortical inputs has no effect in the absence of thalamic stimulation. It appears that there is a citation showing this, but I think it would be important to show this in this study as well.

We understand that the reviewer would like us to induce an HFS protocol on cortical input and then test if there is any change in synaptic strength in thalamic input. We have done this experiment which shows that without a footshock, high-frequency stimulation (HFS) of the cortical inputs did not induce synaptic potentiation on the thalamic pathway (Extended Data Fig. 4d).

(2) t is somewhat confusing that the authors refer to the cortical input as driving heterosynaptic LTP, but this is not shown until Figure 4J, that after non-associative conditioning (unpaired shock and tone) HFS of the cortex can drive freezing and heterosynaptic LTP of thalamic inputs.

We agree with the reviewer that it is in figure 4j and figure 5,b,c which we show electrophysiological evidence for cortical input driving heterosynaptic LTP. It is only to be consistent with our terminology that initially we used behavioral evidence as the proxy for heteroLTP (figure 3c).

…, the authors are 'surprised' by this outcome, which appears to be what they predict.

We removed the phrase “To our surprise”.

(3) 'Cortex' as a stimulation site is vague. The authors have coordinates they used, it is unclear why they are not using standard anatomical nomenclature.

We replaced “cortex” with “auditory/associative cortex”.

(4) The authors' repeated use of homoLTP and heteroLTP to define the input that is being stimulated makes it challenging to understand the experimental detail. While I appreciate this is part of the goal, more descriptive words such as 'thalamic' and 'cortical' would make this much easier to understand.

We agree with the reviewer that a phrase such as “an LTP protocol on thalamic and cortical inputs” would be more descriptive. We chose the words “homoLTP” and “heteroLTP” only to clarify (for the readers) the physiological relevance of these protocols. We thought by using “thalamic” and “cortical” readers may miss this point. However, when for the first time we introduce the words “homoLTP” and “heteroLTP”, we describe which stimulated pathway each refers to.

Reviewer #2 (Public Review):

(1) …The experimental schemes in Figs. 1 and 3 (and Fig. 4e and extended data 4a,b) show that one group of animals was subjected to retrieval in the test context at 24 h, then received HFS, which was then followed by a second retrieval session. With this design, it remains unclear what the HFS impacts when it is delivered between these two 24 h memory retrieval sessions.

We understand that the reviewer has raised the concern that the increase in freezing we observed after the HFS protocol (ex. Fig. 1b, the bar labeled as Wth+24hHFSth) could be caused or modulated by the recall prior to the HFS (Fig. 1a, top branch). To address this concern, in a new group of mice, 24 hours after weak conditioning, we induced the HFS protocol, followed by testing (that is, no testing prior to the HFS protocol). We observed that homoLTP was as effective in mice that were tested prior to the induction protocol as those that were not (Fig. 1b, Extended Data Fig. 1d,e).

It would be nice to see these data parsed out in a clean experimental design for all experiments (in Figs 1, 3, and 4), that means 4 groups with different treatments that are all tested only once at 24 h, and the appropriate statistical tests (ANOVA). This would also avoid repeating data in different panels for different pairwise comparisons (Fig 1, Fig 3, Fig 4, and extended Fig 4).

While we understand the benefit of the reviewer’s suggestion, the current presentation of the data was done to match the flow of the text and the delivery of the information throughout the manuscript. We think it is unlikely that the retrieval test prior to the HFS impacts its effectiveness, as confirmed by homosynaptic HFS data (Extended Data Fig. 1d,e). It is beyond the scope of current manuscript to investigate the mechanisms and manipulations related to reconsolidation and retrieval effects.

(2) … It would be critical to know if LFPs change over 24 h in animals in which memory is not altered by HFS, and to see correlations between memory performance and LFP changes, as two animals displayed low freezing levels. … They would suggest that thalamo-LA potentiation occurs directly after learning+HFS (which could be tested) and is maintained over 24 h.

We have performed the experiment where we recorded the evoked LFP 2hrs and 24hrs following the weak conditioning protocol. We observed that a weak conditioning protocol that was not followed by an optical LTP protocol on the cortical inputs failed to produce synaptic potentiation of the thalamic inputs (tested 2hrs and 24hrs after the LTP protocol; Extended Data Fig. 5d,e).

(3) The statistical analyses need to be clarified. All statements should be supported with statistical testing (e.g. extended data 5c, pg 7 stats are missing). The specific tests should be clearly stated throughout. For ANOVAs, the post-hoc tests and their outcomes should be stated. In some cases, 2-way ANOVAs were performed, but it seems there is only one independent variable, calling for one-way ANOVA.

All the statistical analyses have been revised and the post-hoc tests performed after the ANOVAs are mentioned in the relevant figure legends.

Reviewer #2 (Recommendations For The Authors):

The wording "transient" and "persistent" used here in the context of memory seems a bit misleading, as only one timepoint was assessed for memory recall (24 h), at which the memory strength (freezing levels) seem to change.

As the reviewer mentioned, we have tested memory recall only at one time point. For this reason, throughout the text we used “transient” exclusively to refer to the experience (receiving footshock) and not to the memory. We replaced “persistence” with “stabilization” where it refers to a memory (“the induction of plasticity influences the stabilization of the memory”).

For the procedures in which the CS and US were not paired, the term "unpairing" is used (which is probably the more adequate one), but the term "non-associative conditioning" appears in the text, which seems a bit misleading, as this term may have another connotation. There is also literature that an unpairing of CS and US could lead to the formation of a safety memory to the CS, that may be disrupted by HFS stimulation.

We replaced "non-associative" with “unpaired”.

Validation of viral injection sites for all experiments: Only representative examples are shown, it would be nice to see all viral expression sites.

For this manuscript, we have used 155 mice. For this reason, including the injection sites for all the animals in the manuscript is not feasible. Except for the mice that have been excluded, (please see exclusion criteria added in the methods), the expression pattern we observed was consistent across animals and therefore the images shown are true representatives.

Extended Data 1b: Please explain what N, U, W, and S behavioral groups mean. To what groups mentioned in the text (pg 2,3) do these correspond?

The requested clarifications are implemented in the figure legend.

Please elaborate on the following aspects of your methods and approaches:

• Please explain if the protocol for HFS to manipulate behavior was the same as the one used for the LTP experiments (Fig 1d, Fig 4j) and was identical for homo/hetero inputs from thal and ctx?

We used the same HFS protocol for all the HFS inductions. We included this information in the methods section.

• Please state when the HFS was given in respect to the conditioning (what means immediately before and after?) and in which context it was given. Were animals subjected to HFS exposed to the context longer (either before or after the conditioning while receiving HFS) than the other groups? When the HFS was given in another context (for the 24 h group)- how was this controlled for?

Requested information has been added to the methods section. The control and intervention groups were treated in the same way.

• When were the footshocks given in the anesthesized recordings (Fig. 4j) and how was the temporal relationship to the HFS? Was the timing the same as for the HFS in the behavioral experiments?

Requested information has been added to the methods section.

• Please add information on how the LFP was stimulated and how the LFP- EPSP slope was determined in in vivo recordings, likewise for the whole cell recordings of EPSPs in Fig. 5d-f.

Requested information has been added to the methods section.

Here, the y-Axis in Fig. 5e should be corrected to EPSP slope rather than fEPSP slope if these are whole-cell recordings.

This has been corrected.

• Please include information if the viral injections and opto-manipulations were done bilateral or unilateral and if so in which hemisphere. Likewise, indicate where the LFP recordings were done.

Requested information has been added to the methods section.

• Were there any exclusion criteria for animals (e.g. insufficient viral targeting or placement of fibers and electrodes), other than the testing of the optical CS for adverse effects?

Requested information has been added to the methods section.

Statistics: In addition to clarifying analytical statistics, please clarify n-numbers for slice recordings (number of animals, number of slices, and number of cells if applicable).

Requested information has been added to the methods section.

It would be nice to scrutinize the results in extended data 4b. The freezing levels with U+24h HFS show a strong trend towards an increase, the effect size may be similar to immediate HFS Fig 4f and extended data 4a) if n was increased.

We agree with the reviewer. To address this point, we added “HomoLTP protocol when delivered 24hrs later, produced an increase in freezing; however, the value was not statistically significant.” To show this point, we used the same scale for freezing in Extended Data Fig. 4a and b.

In the final experiment (Fig. 5a-c), Fig. 5b seems to show results from only one animal, but behavioral results are from 4 animals (Fig 5c). It would be helpful to see the quantification of potentiation in each animal.

The results (now with error bar) include all mice.

Please spell out the abbreviation "STC".

Now, it is spelled out.

Page 8 last sentence of the discussion does not seem to fit there.

The sentence has been removed.

Reviewer #3 (Recommendations For The Authors):

(1) The authors did not determine how WTh affects Th-LA synapses, as field EPSPs were recorded only after HFS. WTh was required for the effects of HFS, as HFS alone did not produce CR in naïve and/or unpaired controls. As such the effects of the WTh protocol on synaptic strength must be investigated.

We have performed the experiment where we recorded the evoked LFP 2hrs and 24hrs following the weak conditioning protocol. We observed that a weak conditioning protocol that was not followed by an optical LTP protocol on the cortical inputs failed to produce synaptic potentiation of the thalamic inputs (tested 2hrs and 24hrs after the LTP protocol; Extended Data Fig. 5d,e).

(2) The authors provide some evidence that their dual opsin approach is feasible, particularly the use of sustained yellow light to block the effects of blue light on ChrimsonR. However, this validation was done using single pulses making it difficult to assess the effect of this protocol on Th input when HFS was used. Without strong evidence that the optogenetic methods used here are fault-proof, the main conclusions of this study are compromised. Why did the authors not use a protocol in which fibers were placed directly in the Ctx and Th while using soma-restricted opsins to avoid cross-contamination?

We understand that the reviewer raises the possibility that our dual-opsin approach, although effective with single pulses, may fail in higher frequency stimulation protocols (10Hz and 85Hz). To address this concern, in a new group of mice we applied our approach to 10Hz and 85Hz stimulation protocols. We show that our approach is effective in single-pulse as well as in 10Hz and 85Hz stimulation protocols (Fig. 2d-h).

Author response:

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

We sincerely appreciate the reviewer’s dedication to evaluating our manuscript and raising essential considerations regarding the classification of the migration behavior we described. While the reviewer suggests that this behavior aligns with the concept of itinerancy, we contend that it represents a distinct phenomenon, albeit with similarities, as both involve the non-breeding movements of birds. We acknowledge that our manuscript did not adequately address this distinction and have considered the reviewer’s feedback. In our response, we clarify the difference between the described phenomenon and itinerancy. Our revised manuscript will include a new section in the Discussion to address this issue comprehensively.

In the first part of the review, the reviewer emphasizes that the pattern we are describing is consistent with itinerancy. Regardless of the terminology used, we want to highlight the existence of two different types of migratory behavior, both of which involve movement in non-breeding areas.

The first type, called itinerancy, was first described by Moreau in 1972 in “The Palaearctic-African Bird Migration Systems.” As noted by the reviewer, this behavior involves an alternation of stopovers and movements between different short-term non-breeding residency areas. They usually occur in response to food scarcity in one part of the non-breeding range, causing birds to move to another part of the same range. These movements typically cover distances of 10 to 100 kilometers but are neither continuous nor directional. Moreau (1972) defined itinerancy as prolonged stopovers, normally lasting several months, primarily in tropical regions. He noted observations of certain species disappearing from his study areas in sub-Saharan Africa in December and others appearing, suggesting they may have multiple home ranges during the non-breeding season. Subsequent research, as mentioned by the reviewer, has confirmed itinerancy in many species, particularly among Palaearctic-African migrants in sub-Saharan Africa. In particular, the Montagu’s Harrier has been extensively studied in this regard. The reviewer rightly points out that our study does not include recent findings on this species. In our revised version, we will include references to recent studies, such as those by Trierweiler et al. (2013, Journal of Animal Ecology, 82:107-120) and Schlaich et al. (2023, Ardea, 111:321-342), which show that Montagu’s Harrier has an average of 3-4 home ranges separated by approximately 200 kilometers. These studies suggest that the species spends approximately 1.5 months at each site, with the most extended period typically observed at the last site before migrating to the breeding grounds.

In the second type, birds undertake a post-breeding migration, arrive in their non-breeding range, and then gradually move in a particular direction throughout the season. This continuous directional movement covers considerable distances and continues throughout the non-breeding period. In our study, this movement covered about 1000 km, comparable to the total migration distance of Rough-legged Buzzards of about 1500 km. As observed in our research, these movements are influenced by external factors such as snow cover. In such cases, the progression of snow cover in a south-westerly direction during winter can prevent birds from finding food, forcing them to continue migrating in the same direction. In essence, this movement represents a prolonged phase of the migration process but at a slower pace. Similar behavior has been documented in buzzards, as reported by Strandberg et al. (2009, Ibis 151:200-206). Although several transmitters in their study stopped working in mid-winter, the authors observed a phenomenon they termed ‘prolonged autumn migration.’

In the second part of the review, the reviewer questions the need to distinguish between the two behaviors we have discussed. However, we believe these behaviors differ in their structure (with the first being intermittent and often non-directional, whereas the second is continuous and directional) and in their causes (with the first being driven by seasonal food resource cycles and the second by advancing snow cover). We therefore argue that it is worth distinguishing between them. To differentiate these forms of non-breeding movement, we propose to use ‘itinerancy’ for the first type, as described initially by Moreau in 1972, and introduce a separate term for the second behavior. Although ‘slow directional itinerancy’ could be considered, we find it too cumbersome.

Moreover, ‘itinerancy’ in the literature refers not only to non-breeding movements but also to the use of different nesting sites, e.g., Lislevand et al. (2020, Journal of Avian Biology: e02595), reinforcing its association with movements between multiple sites within habitats. We, therefore, propose that the second behavior be given a distinct name. We acknowledge the reviewer’s point that we did not adequately address this distinction in the Discussion and plan to include a separate section in our paper’s revised version. In the third part of his review, the reviewer suggests an alternative title. Another reviewer, Dr Theunis Piersma, suggested the current title during the first round of reviewing, and we have chosen his version.

In the fourth part of the review, the reviewer questions whether it is appropriate to discuss the conservation aspect of this study. This type of non-breeding movement raises concerns about accurately determining non-breeding ranges and population dynamics for species that exhibit this behavior. We believe that accurate determination of range and population dynamics is critical to conservation efforts. While this may be less important for species breeding in Europe and migrating to Africa, for which monitoring breeding territories is more feasible, it’s essential for Arctic and sub-Arctic breeding species. Large-scale surveys in these regions have historically been challenging and have become even more so with the end of Arctic cooperation following Russia’s war with Ukraine (Koivurova, Shibata, 2023). For North America and Europe, non-breeding abundance is typically estimated once per season in mid-winter. In North America, these are the so-called Christmas counts (which take place once at the end of December), and in Europe, they are the IWC counts mentioned by the reviewer (as follows from their official website - “The IWC requires a single count at each site, which should be repeated each year. The exact dates vary slightly from region to region, but take place in January or February”). Because of such a single count in mid-winter, non-breeding habitats occupied in autumn and spring will be listed as ‘uncommon’ at best, while south-western habitats where birds are only present in mid-winter will be listed as ‘common.’ However, the situation will be reversed if we consider the time birds spend in these habitats.

The reviewer also highlights the introduction’s unconventional structure and information redundancy at the beginning. We have chosen this structure and provided basic explanations to improve readability for a wider audience, given eLife’s readership. At the same time, we will certainly take the reviewers’ feedback into account in the revised version. We plan to include the references to modern itinerancy research mentioned above and to add a section on itinerancy to the Discussion.

We appreciate the reviewer’s input and sincerely thank them for their time and effort in reviewing our paper. While we may not fully agree on the classification of the behavior we describe, we value the opportunity to engage in discussion and believe that presenting arguments and counterarguments to the reader is beneficial to scientific progress.

---

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

Reviewer #1 (Recommendations For The Authors):

I much enjoyed reading this manuscript, that is, once I understood what it is about. Titles like "Conserving bird populations in the Anthropocene: the significance of non-breeding movements" are a claim to so-called relevance, they have NOTHING to do with the content of the paper, so once I understood that this paper was about the "Quick quick slow: the foxtrot migration of rough-legged buzzards is a response to habitat and snow" (an alternative title), it was becoming very interesting. So the start of the abstract as well as the introduction is very tedious, as clearly much trouble is taken here to establish reputability. In my eyes this is unnecessary: eLife should be interested in publishing such a wonderful description of such a wonderful migrant in a study that comes to grips with limiting factors on a continental scale!

We sincerely appreciate your time and effort in reviewing our manuscript. Thank you for your appreciation of our study.

We agree that the focus of the article should be changed from conservation to migration patterns. We have rewritten the Introduction and Discussion as suggested. We have added the application of this pattern including conservation at the end of the Discussion by completely changing Figure 5. We have also changed the title to the suggested one.

Not sure that the first paragraph statements that seek to downplay what we know about wintering vs breeding areas are valid (although I see what purpose they serve). Migratory shorebirds have extensively been studied in the nonbreeding areas, for example, including movement aspects (see, as just one example, Verhoeven, M.A., Loonstra, A.H.J., McBride, A.D., Both, C., Senner, N.R. & Piersma, T. (2020) Migration route, stopping sites, and non breeding destinations of adult Black tailed Godwits breeding in southwest Fryslân, The Netherlands. Journal of Ornithology 162, 61-76) and there are very impressive studies on the winter biology of migrants across large scale (for example in Zwarts' Living on the Edge book on the Sahel wetlands). Think also about geese and swans and about seabirds!

We have rewritten the first paragraph and it now talks about patterns of migratory behavior. We have also rewritten the second paragraph, now it is devoted to studies of movements in the non-breeding period. We explain how our pattern differs from those already studied and give references to the papers you mentioned.

Directional movements in nonbreeding areas as a function of food (in this case locusts) have really beautifully been described by Almut Schlaich et al in JAnimEcol for Montagu's harriers.

We have added Montagu's harrier example in the second paragraph of the Introduction and the Discussion. We have added a reference to Schlaich and to Garcia and Arroyo, who suggested that Montagu's harriers have long directional migrations during the non-breeding period.

Once the paper starts talking buzzards, and the analyses of the wonderful data, all is fine. It is a very competent analysis with a description of a cool pattern.

Thank you for your appreciation of our study. We hope the revised version is better and clearer.

However, i would say that it is all a question of spatial scale. The buzzards here respond to changes in food availability, but there is not an animal that doesn't. The question is how far they have to move for an adequate response: in some birds movements of 100s of meters may be enough, and then anything to the scale of rough-legged buzzards.

In the new version of the manuscript, we emphasize that this is a large distance (about 1000 km), comparable to the distance of the fall and spring migrations (about 1400 km) in lines 70-72 of the Introduction and 379-383 of the Discussion.

And actually, several of the shorebirds I know best also do a foxtrot, such as red knots and bar-tailed godwits moulting in the Wadden Sea, then spending a few months in the UK estuaries, before returning to the Wadden Sea before the long migrations to Arctic breeding grounds. The publication of the rough-legged buzzard story may help researchers to summarize patterns such as this too. Mu problem with this paper is the framing. A story on the how and why of these continental movements in response to snow and other habitat features would be a grand contribution. Drop Anthropocene, and rethink whether foxtrot should be introduced as a hypothesis or a summary of cool descriptions. I prefer the latter, and recommend eLife to go with that too, rather than encourage "disconnected frames that seek 'respectability'" Good luck, theunis piersma

We thank the reviewer again for his valuable comments and suggestions. We have changed the framing to the suggested one and removed the Anthropocene from the article.

Reviewer #2 (Recommendations For The Authors):

We sincerely appreciate the time and effort you have taken to review our manuscript. We have carefully considered all of your comments, including both public and author comments, and provided detailed responses to each of them below. In addition, we would like to address the most important public comments.

We agree with the suggestion to shift the focus of the article from conservation to migration patterns. Accordingly, we have rewritten both the Introduction and Discussion sections to focus on migration behavior rather than conservation.

However, we respectfully disagree with the suggestion that the migration patterns we describe are synonymous with itinerancy. We acknowledge that our original presentation may have been unclear and may have hindered full understanding. In the revised version, we provide a detailed analysis of migratory behavior in the Introduction that describes how our pattern differs from itinerancy. We also revisit this distinction in the Discussion section. We have also carefully revised Figure 1 to improve clarity and avoid potential misunderstandings.

Regarding the applicability of the described migration pattern, we acknowledge that the Rough-legged Buzzard is not listed as an endangered species. However, we believe that our findings have practical implications. We have moved our discussion of this issue to the end of the Discussion section and have completely revised Figure 5. While the overall population of Rough-legged Buzzards is not declining, certain regions within its range are experiencing declines. We show that this decline does not warrant listing the species as endangered. Instead, it may represent a redistribution within the non-breeding range - a shift in range dynamics. We use the example of the Rough-legged Buzzard to illustrate this concept and emphasize the importance of considering such dynamics when assessing the conservation status of species in the future.

We also acknowledge that the hypothesis of this form of behavior has been proposed previously for Montagu's Harrier, and we have included this information in the revised manuscript. In addition, we agree that the focus on the Anthropocene is unnecessary in this context and have therefore removed it.

We believe that these revisions significantly improve the clarity and robustness of the manuscript, and we are grateful for your insightful comments and suggestions.

As a general comment, please note that including line numbers (as it is the standard in any manuscript submission) would facilitate reviewers providing more detailed comments on the text.

We apologize for this oversight and have added line numbers to our revised manuscript.

Dataset: unclear what is the frequency of GPS transmissions. Furthermore, information on relative tag mass for the tracked individuals should be reported.

We have included this information in our manuscript (L 157-163). We also refer to the study in which this dataset was first used and described in detail (L 164).

Data pre-processing: more details are needed here. What data have been removed if the bird died? The entire track of the individual? Only the data classified in the last section of the track? The section also reports on an 'iterative procedure' for annotating tracks, which is only vaguely described. A piecewise regression is mentioned, but no details are provided, not even on what is the dependent variable (I assume it should be latitude?).

Regarding the deaths. We only removed the data when the bird was already dead. We have corrected the text to make this clear (L 170).

Regarding the iterative procedure. We have added a detailed description on lines 175-188.

Data analysis: several potential issues here:

(1) Unclear why sex was not included in all mixed models. I think it should be included.

Our dataset contains 35 females and eight males. This ratio does not allow us to include sex in all models and adequately assess the influence of this factor. At the same time, because adult females disperse farther than males in some raptor species, we conducted a separate analysis of the dependence of migration distance on sex (Table S8) and found no evidence for this in our species. We have written a separate paragraph about this. This paragraph can be found on lines 356-360 of the new manuscript.

(2) Unclear what is the rationale of describing habitat use during migration; is it only to show that it is a largely unsuitable habitat for the species? But is a formal analysis required then? Wouldn't be enough to simply describe this?

Habitat use and snow cover determine the two main phases (quick and slow) of the pattern we describe. We believe that habitat analysis is appropriate in this case and that a simple description would be uninformative and would not support our conclusions.

(3) Analysis of snow cover: such a 'what if' analysis is fine but it seems to be a rather indirect assessment of the effect of snow cover on movement patterns. Can a more direct test be envisaged relating e.g. daily movement patterns to concomitant snow cover? This should be rather straightforward. The effectiveness of this method rests on among-year differences in snow cover and timing of snowfall. A further possibility would be to demonstrate habitat selection within the entire non-breeding home range of an individual in relation snow cover. Such an analysis would imply associating presence-absence of snow to every location within the non-breeding range and testing whether the proportion of locations with snow is lower than the proportion of snow of random locations within the entire non-breeding home range (95% KDE) for every individual (e.g. by setting a 1/10 ratio presence to random locations).

The proposed analysis will provide an opportunity to assess whether the Rough-legged Buzzard selects areas with the lowest snow cover, but will not provide an opportunity to follow the dynamics and will therefore give a misleading overall picture. This is especially true in the spring months. In March-April, Rough-legged Buzzards move northeast and are in an area that is not the most open to snow. At this time, areas to the southwest are more open to snow (this can be seen in Figure 4b). If we perform the proposed analysis, the control points for this period would be both to the north (where there is more snow) and to the south (where there is less snow) from the real locations, and the result would be that there is no difference in snow cover.

A step-selection analysis could be used, as we did in our previous work (Curk et al 2020 Sci Rep) with the same Rough-legged Buzzard (but during migration, not winter). But this would only give us a qualitative idea, not a quantitative one - that Rough-legged Buzzards move from snow (in the fall) and follow snowmelt progression (in the spring).

At the same time, our analysis gives a complete picture of snow cover dynamics in different parts of the non-breeding range. This allows us to see that if Rough-legged Buzzards remained at their fall migration endpoint without moving southwest, they would encounter 14.4% more snow cover (99.5% vs. 85.1%). Although this difference may seem small (14.4%), it holds significance for rodent-hunting birds, distinguishing between complete and patchy snow cover. Simultaneously, if Rough-legged Buzzards immediately flew to the southwest and stayed there throughout winter, they would experience 25.7% less snow cover (57.3% vs. 31.6%). Despite a greater difference than in the first case, it doesn't compel them to adopt this strategy, as it represents the difference between various degrees of landscape openness from snow cover.

We write about this in the new manuscript on lines 385-394.

Results: it is unclear whether the reported dispersion measures are SDs or SEs. Please provide details.

For the date and coordinates of the start and end of the different phases of migration, we specified the mean, sd, and sample size. We wrote this in line 277. For the values of the parameters of the different phases of the migration (duration, distance, speed, and direction), we used the mean, the standard error of the mean, and the confidence interval (obtained using the ‘emmeans’ package). We have indicated this in lines 302-303 and the caption of Table 1 (L 315) and Figure 2 (L 293-294). For the values of habitat and snow cover experienced by the Rough-legged Buzzards, we used the mean and the error of the mean. We reported this on lines 322 and 337 and in Figures 3 (L 332-333) and 4 (L 355-356).

Discussion: in general, it should be reshaped taking into account the comments. It is overlong, speculative and quite naive in several passages. Entire sections can be safely removed (I think it can be reduced by half without any loss of information). I provide some examples of the issues I have spotted below. For instance, the entire paragraph starting with 'Understanding....' is not clear to me. What do you mean by 'prohibited management' options? Without examples, this seems a rather general text, based on unclear premises when related to the specific of this study. Some statements are vague, derive from unsubstantiated claims, and unclear. E.g. "Despite their scarcity in these habitats, forests appear to hold significant importance for Rough-legged buzzards for nocturnal safety". I could not find any day-night analysis showing that they actually roost in forests during nighttime. Being a tundra species, it may well be possible that rough-legged buzzards perceive forests as very dangerous habitats and that they prefer instead to roost in open habitats. Analysing habitat use during day and night during the non-breeding period may be of help to clarify this. Furthermore, considering the fast migration periods, what is the flight speed during day and night above forests? Do these birds also migrate at night or do they roost during the night? Perhaps a figure visualizing day and night track segments could be of help (or an analysis of day vs. night flight speed) (there are several R packages to annotate tracks in relation to day and night). This is an example of another problematic statement: "The progression of snow cover in the wintering range of Rough-legged buzzards plays a significant role in their winter migration pattern." The manuscript does not contain any clear demonstration of this, as I wrote in my previous comments. Without such evidence, you must considerably tone down such assertions. But since providing a direct link is certainly possible, I think that additional analyses would clearly strengthen your take-home message.

The paragraph starting with "The quantification of environmental changes that could prove fatal to bird species presents yet another challenge for conservation efforts in an era of rapid global change." is quite odd. Take the following statement "For instance, the presence of small patches of woodland in the winter range might appear crucial to the survival of the Rough-legged buzzard. Elimination of these seemingly minor elements of vegetation cover through management actions could have dire consequences for the species.". It is based on the assumption that minor vegetation elements play a key role in the ecology of the species, without any evidence supporting this. Does it have any sense? I could safely say exactly the opposite and I would believe it might even be more substantiated.

We agree with these comments.

We have completely rewritten this section. As suggested, we have shortened it by removing statements that were not supported by the research. We have completely removed the statements about "prohibited management". We have also removed the statement that "forests appear to be of significant importance to Rough-legged buzzards for nocturnal safety" and everything associated with that statement, e.g. the statement about "small elements of vegetation cover", etc. We do believe that this statement is true in substance, but we also agree that it is not supported by the results and requires separate analysis. At the same time, we believe that this is a topic for a separate study and would be redundant here. Therefore, we leave it for a separate publication.

Conclusion paragraph: I believe this severely overstates the conservation importance of this study. That the results have "crucial implications for conservation efforts in the Anthropocene, where rapidly changing environmental factors can severely impact bird migration" seems completely untenable to me. What is the evidence for such crucial implications? For instance, these results may suggest that climate change, because global warming is predicted to reduce snow cover in the non-breeding areas, might well be beneficial for populations of this species, by reducing non-breeding energy expenditure and improving non-breeding survival. I think statements like these are simply not necessary, and that the study should be more focused on the actual results and evidence provided.

We have completely rewritten this section. We removed the reference to the Anthropocene and focused on migratory behavior and migration patterns.

Author response:

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

We are grateful for these balanced, nuanced evaluations of our work concerning the observed epistatic trends and our interpretations of their mechanistic origins. Overall, we think the reviewers have done an excellent job at recognizing the novel aspects of our findings while also discussing the caveats associated with our interpretations of the biophysical effects of these mutations. We believe it is important to consider both of these aspects of our work in order to appreciate these advances and what sorts of pertinent questions remain.

Notably, both reviewers are concerned that our lack of experimental approaches to compare the conformational properties of GnRHR variants weakens our claims. We would first humbly suggest that this constitutes a more general caveat that applies to nearly all investigations of the cellular misfolding of α-helical membrane proteins. Whether or not any current in vitro folding measurements report on conformational transitions that are relevant to cellular protein misfolding reactions remains an active area of debate (discussed further below). Nevertheless, while we concede that our structural and/ or computational evaluations of various mutagenic effects remain speculative, prevailing knowledge on the mechanisms of membrane protein folding suggest our mutations of interest (V276T and W107A) are highly unlikely to promote misfolding in precisely the same way. Thus, regardless of whether or not we were able experimentally compare the relevant folding energetics of GnRHR variants, we are confident that the distinct epistatic interactions formed by these mutations reflect variations in the misfolding mechanism and that they are distinct from the interactions that are observed in the context of stable proteins. In the following, we provide detailed considerations concerning these caveats in relation to the reviewers’ specific comments.

Reviewer #1 (Public Review):

The paper carries out an impressive and exhaustive non-sense mutagenesis using deep mutational scanning (DMS) of the gonadotropin-releasing hormone receptor for the WT protein and two single point mutations that I) influence TM insertion (V267T) and ii) influence protein stability (W107A), and then measures the effect of these mutants on correct plasma membrane expression (PME).

Overall, most mutations decreased mGnRHR PME levels in all three backgrounds, indicating poor mutational tolerance under these conditions. The W107A variant wasn't really recoverable with low levels of plasma membrane localisation. For the V267T variant, most additional mutations were more deleterious than WT based on correct trafficking, indicating a synergistic effect. As one might expect, there was a higher degree of positive correlation between V267T/W107A mutants and other mutants located in TM regions, confirming that improper trafficking was a likely consequence of membrane protein co-translational folding. Nevertheless, context is important, as positive synergistic mutants in the V27T could be negative in the W107A background and vice versa. Taken together, this important study highlights the complexity of membrane protein folding in dissecting the mechanism-dependent impact of disease-causing mutations related to improper trafficking.

Strengths

This is a novel and exhaustive approach to dissecting how receptor mutations under different mutational backgrounds related to co-translational folding, could influence membrane protein trafficking.

Weaknesses

The premise for the study requires an in-depth understanding of how the single-point mutations analysed affect membrane protein folding, but the single-point mutants used seem to lack proper validation.

Given our limited understanding of the structural properties of misfolded membrane proteins, it is unclear whether the relevant conformational effects of these mutations can be unambiguously validated using current biochemical and/ or biophysical folding assays. X-ray crystallography, cryo-EM, and NMR spectroscopy measurements have demonstrated that many purified GPCRs retain native-like structural ensembles within certain detergent micelles, bicelles, and/ or nanodiscs. However, helical membrane protein folding measurements typically require titration with denaturing detergents to promote the formation of a denatured state ensemble (DSE), which will invariably retain considerable secondary structure. Given that the solvation provided by mixed micelles is clearly distinct from that of native membranes, it remains unclear whether these DSEs represent a reasonable proxy for the misfolded conformations recognized by cellular quality control (QC, see https://doi.org/10.1021/acs.chemrev.8b00532). Thus, the use and interpretation of these systems for such purposes remains contentious in the membrane protein folding community. In addition to this theoretical issue, we are unaware of any instances in which GPCRs have been found to undergo reversible denaturation in vitro- a practical requirement for equilibrium folding measurements (https://doi.org/10.1146/annurev-biophys-051013-022926). We note that, while the resistance of GPCRs to aggregation, proteolysis, and/ or mechanical unfolding have also been probed in micelles, it is again unclear whether the associated thermal, kinetic, and/ or mechanical stability should necessarily correspond to their resistance to cotranslational and/ or posttranslational misfolding. Thus, even if we had attempted to validate the computational folding predictions employed herein, we suspect that any resulting correlations with cellular expression may have justifiably been viewed by many as circumstantial. Simply put, we know very little about the non-native conformations are generally involved in the cellular misfolding of α-helical membrane proteins, much less how to measure their relative abundance. From a philosophical standpoint, we prefer to let cells tell us what sorts of broken protein variants are degraded by their QC systems, then do our best to surmise what this tells us about the relevant properties of cellular DSEs.

Despite this fundamental caveat, we believe that the chosen mutations and our interpretation of their relevant conformational effects are reasonably well-informed by current modeling tools and by prevailing knowledge on the physicochemical drivers of membrane protein folding and misfolding. Specifically, the mechanistic constraints of translocon-mediated membrane integration provide an understanding of the types of mutations that are likely to disrupt cotranslational folding. Though we are still learning about the protein complexes that mediate membrane translocation (https://doi.org/10.1038/s41586-022-05336-2), it is known that this underlying process is fundamentally driven by the membrane depth-dependent amino acid transfer free energies (https://doi.org/10.1146/annurev.biophys.37.032807.125904). This energetic consideration suggests introducing polar side chains near the center of a nascent TMDs should almost invariably reduce the efficiency of topogenesis. To confirm this in the context of TMD6 specifically, we utilized a well-established biochemical reporter system to confirm that V276T attenuates its translocon-mediated membrane integration (Fig. S1)- at least in the context of a chimeric protein. We also constructed a glycosylation-based topology reporter for full-length GnRHR, but ultimately found its’ in vitro expression to be insufficient to detect changes in the nascent topological ensemble.

In contrast to V276T, the W107A mutation is predicted to preserve the native topological energetics of GnRHR due to its position within a soluble loop region. W107A is also unlike V276T in that it clearly disrupts tertiary interactions that stabilize the native structure. This mutation should preclude the formation of a structurally conserved hydrogen bonding network that has been observed in the context of at least 25 native GPCR structures (https://doi.org/10.7554/eLife.5489). However, without a relevant folding assay, the extent to which this network stabilizes the native GnRHR fold in cellular membranes remains unclear. Overall, we admit that these limitations have prevented us from measuring how much V276T alters the efficiency of GnRHR topogenesis, how much the W107A destabilizes the native fold, or vice versa. Nevertheless, given these design principles and the fact that both reduce the plasma membrane expression of GnRHR, as expected, we are highly confident that the structural defects generated by these mutations do, in fact, promote misfolding in their own ways. We also concede that the degree to which these mutagenic perturbations are indeed selective for specific folding processes is somewhat uncertain. However, it seems exceedingly unlikely that these mutations should disrupt topogenesis and/ or the folding of the native topomer to the exact same extent. From our perspective, this is the most important consideration with respect to the validity of the conclusions we have made in this manuscript.

Furthermore, plasma membrane expression has been used as a proxy for incorrect membrane protein folding, but this not necessarily be the case, as even correctly folded membrane proteins may not be trafficked correctly, at least, under heterologous expression conditions. In addition, mutations can affect trafficking and potential post-translational modifications, like glycosylation.

While the reviewer is correct that the sorting of folded proteins within the secretory pathway is generally inefficient, it is also true that the maturation of nascent proteins within the ER generally bottlenecks the plasma membrane expression of most α-helical membrane proteins. Our group and several others have demonstrated that the efficiency of ER export generally appears to scale with the propensity of membrane proteins to achieve their correct topology and/ or to achieve their native fold (see https://doi.org/10.1021/jacs.5b03743 and https://doi.org/10.1021/jacs.8b08243). Notably, these investigations all involved proteins that contain native glycosylation and various other post-translational modification sites. While we cannot rule out that certain specific combinations of mutations may alter expression through their perturbation of post-translational GnRHR modifications, we feel confident that the general trends we have observed across hundreds of variants predominantly reflect changes in folding and cellular QC. This interpretation is supported by the relationship between observed trends in variant expression and Rosetta-based stability calculations, which we identified using unbiased unsupervised machine learning approaches (compare Figs. 6B & 6D).

Reviewer #2 (Public Review):

Summary:

In this paper, Chamness and colleagues make a pioneering effort to map epistatic interactions among mutations in a membrane protein. They introduce thousands of mutations to the mouse GnRH Receptor (GnRHR), either under wild-type background or two mutant backgrounds, representing mutations that destabilize GnRHR by distinct mechanisms. The first mutant background is W107A, destabilizing the tertiary fold, and the second, V276T, perturbing the efficiency of cotranslational insertion of TM6 to the membrane, which is essential for proper folding. They then measure the surface expression of these three mutant libraries, using it as a proxy for protein stability, since misfolded proteins do not typically make it to the plasma membrane. The resulting dataset is then used to shed light on how diverse mutations interact epistatically with the two genetic background mutations. Their main conclusion is that epistatic interactions vary depending on the degree of destabilization and the mechanism through which they perturb the protein. The mutation V276T forms primarily negative (aggravating) epistatic interactions with many mutations, as is common to destabilizing mutations in soluble proteins. Surprisingly, W107A forms many positive (alleviating) epistatic interactions with other mutations. They further show that the locations of secondary mutations correlate with the types of epistatic interactions they form with the above two mutants.

Strengths:

Such a high throughput study for epistasis in membrane proteins is pioneering, and the results are indeed illuminating. Examples of interesting findings are that: (1) No single mutation can dramatically rescue the destabilization introduced by W107A. (2) Epistasis with a secondary mutation is strongly influenced by the degree of destabilization introduced by the primary mutation. (3) Misfolding caused by mis-insertion tends to be aggravated by further mutations. The discussion of how protein folding energetics affects epistasis (Fig. 7) makes a lot of sense and lays out an interesting biophysical framework for the findings.

Weaknesses:

The major weakness comes from the potential limitations in the measurements of surface expression of severely misfolded mutants. This point is discussed quite fairly in the paper, in statements like "the W107A variant already exhibits marginal surface immunostaining" and many others. It seems that only about 5% of the W107A makes it to the plasma membrane compared to wild-type (Figures 2 and 3). This might be a low starting point from which to accurately measure the effects of secondary mutations.

The reviewer raises an excellent point that we considered at length during the analysis of these data and the preparation of the manuscript. Though we remain confident in the integrity of these measurements and the corresponding analyses, we now realize this aspect of the data required further discussion and documentation which we have provided in the revised version of the manuscript as is described in the following.

Still, the authors claim that measurements of W107A double mutants "still contain cellular subpopulations with surface immunostaining intensities that are well above or below that of the W107A single mutant, which suggests that this fluorescence signal is sensitive enough to detect subtle differences in the PME of these variants". I was not entirely convinced that this was true.

We made this statement based on the simple observation that the surface immunostaining intensities across the population of recombinant cells expressing the library of W107A double mutants was consistently broader than that of recombinant cells expressing W107A GnRHR alone (see Author response image 1 for reference). Given that the recombinant cellular library represents a mix of cells expressing ~1600 individual variants that are each present at low abundance, the pronounced tails within this distribution presumably represent the composite staining of many small cellular subpopulations that express collections of variants that deviate from the expression of W107A to an extent that is significant enough to be visible on a log intensity plot.

Author response image 1.

Firstly, I think it would be important to test how much noise these measurements have and how much surface immunostaining the W107A mutant displays above the background of cells that do not express the protein at all.

For reference, the average surface immunostaining intensity of HEK293T cells transiently expressing W107A GnRHR was 2.2-fold higher than that of the IRES-eGFP negative, untransfected cells within the same sample- the WT immunostaining intensity was 9.5-fold over background by comparison. Similarly, recombinant HEK293T cells expressing the W107A double mutant library had an average surface immunostaining intensity that was 2.6-fold over background across the two DMS trials. Thus, while the surface immunostaining of this variant is certainly diminished, we were still able to reliably detect W107A at the plasma membrane even under distinct expression regimes. We have included these and other signal-to-noise metrics for each experiment in the Results section of the revised manuscript.

Beyond considerations related to intensity, we also previously noticed the relative intensity values for W107A double mutants exhibited considerable precision across our two biological replicates. If signal were too poor to detect changes in variant expression, we would have expected a plot of the intensity values across these two replicates to form a scatter. Instead, we found DMS intensity values for individual variants to be highly correlated from one replicate to the next (Pearson’s R2 = 0.95, see Author response image 2 for reference). This observation empirically demonstrates that this assay consistently differentiated between variants that exhibit slightly enhanced immunostaining from those that have even lower immunostaining than W107A GnRHR. We have included these discussion points in the Results section as well as scatter plots for replicate variant intensities within all three genetic backgrounds in Figure S3 of the revised manuscript.

Author response image 2.

But more importantly, it is not clear if under this regimen surface expression still reports on stability/protein fitness. It is unknown if the W107A retains any function or folding at all. For example, it is possible that the low amount of surface protein represents misfolded receptors that escaped the ER quality control.

While we believe that such questions are outside the scope of this work, we certainly agree that it is entirely possible that some of these variants bypass QC without achieving their native fold. This topic is quite interesting to us but is quite challenging to assess in the context of GPCRs, which have complex fitness landscapes that involve their propensity to distinguish between different ligands, engage specific components associated with divergent downstream signaling pathways, and navigate between endocytic recycling/ degradation pathways following activation. In light of the inherent complexity of GPCR function, we humbly suggest our choice of a relatively simple property of an otherwise complex protein may be viewed as a virtue rather than a shortcoming. Protein fitness is typically cast as the product of abundance and activity. Rather than measuring an oversimplified, composite fitness metric, we focused on one variable (plasma membrane expression) and its dominant effector (folding). We believe restraining the scope in this manner was key for the elucidation of clear mechanistic insights.

The differential clustering of epistatic mutations (Fig. 6) provides some interesting insights as to the rules that dictate epistasis, but these too are dominated by the magnitude of destabilization caused by one of the mutations. In this case, the secondary mutations that had the most interesting epistasis were exceedingly destabilizing. With this in mind, it is hard to interpret the results that emerge regarding the epistatic interactions of W107A. Furthermore, the most significant positive epistasis is observed when W107A is combined with additional mutations that almost completely abolish surface expression. It is likely that either mutation destabilizes the protein beyond repair. Therefore, what we can learn from the fact that such mutations have positive epistasis is not clear to me. Based on this, I am not sure that another mutation that disrupts the tertiary folding more mildly would not yield different results. With that said, I believe that the results regarding the epistasis of V276T with other mutations are strong and very interesting on their own.

We agree with the reviewer. In light of our results we believe it is virtually certain that the secondary mutations characterized herein would be likely to form distinct epistatic interactions with mutations that are only mildly destabilizing. Indeed, this insight reflects one of the key takeaway messages from this work- stability-mediated epistasis is difficult to generalize because it should depend on the extent to which each mutation changes the stability (ΔΔG) as well as initial stability of the WT/ reference sequence (ΔG, see Figure 7). Frankly, we are not so sure we would have pieced this together as clearly had we not had the fortune (or misfortune?) of including such a destructive mutation like W107A as a point of reference.

Additionally, the study draws general conclusions from the characterization of only two mutations, W107A and V276T. At this point, it is hard to know if other mutations that perturb insertion or tertiary folding would behave similarly. This should be emphasized in the text.

We agree. Our findings suggest different mutations may not behave similarly, which we believe is a key finding of this work. We have emphasized this point in the Discussion section of the revised manuscript as follows:

“These findings suggest the folding-mediated epistasis is likely to vary among different classes of destabilizing mutations in a manner that should also depend on folding efficiency and/ or the mechanism(s) of misfolding in the cell.”

Some statistical aspects of the study could be improved:

(1) It would be nice to see the level of reproducibility of the biological replicates in a plot, such as scatter or similar, with correlation values that give a sense of the noise level of the measurements. This should be done before filtering out the inconsistent data.

We thank the reviewer for this suggestion and will include scatters for each genetic background like the one shown above in Figure S3 of the revised version of the manuscript.

(2) The statements "Variants bearing mutations within the C- terminal region (ICL3-TMD6-ECL3-TMD7) fare consistently worse in the V276T background relative to WT (Fig. 4 B & E)." and "In contrast, mutations that are 210 better tolerated in the context of W107A mGnRHR are located 211 throughout the structure but are particularly abundant among residues 212 in the middle of the primary structure that form TMD4, ICL2, and ECL2 213 (Fig. 4 C & F)." are both hard to judge. Inspecting Figures 4B and C does not immediately show these trends, and importantly, a solid statistical test is missing here. In Figures 4E and F the locations of the different loops and TMs are not indicated on the structure, making these statements hard to judge.

We apologize for this oversight and thank the reviewer for pointing this out. We utilized paired Wilcoxon-Signed Rank Tests to evaluate the statistical significance of these observations and modified the description of these findings in the revised version of the results section as follows:

“Variants bearing mutations within the C-terminal regions including ICL3, TMD6, and TMD7 fare consistently worse in the V276T background relative to WT (paired Wilcoxon-Signed Rank Test p-values of 0.0001, 0.02, and 0.005, respectively) (Fig. 4 B & E). Given that V276T perturbs the cotranslational membrane integration of TMD6 (Fig. S1, Table S1), this directional bias potentially suggests that the apparent interactions between these mutations manifest during the late stages of cotranslational folding. In contrast, mutations that are better tolerated in the context of W107A mGnRHR are located throughout the structure but are particularly abundant among residues in the middle of the primary structure that form ICL2, TMD4, and ECL2 (paired Wilcoxon-Signed Rank Test p-values of 0.0005, 0.0001, and 0.004, respectively) (Fig. 4 C & F).”

(3) The following statement lacks a statistical test: "Notably, these 98 variants are enriched with TMD variants (65% TMD) relative to the overall set of 251 variants (45% TMD)." Is this enrichment significant? Further in the same paragraph, the claim that "In contrast to the sparse epistasis that is generally observed between mutations within soluble proteins, these findings suggest a relatively large proportion of random mutations form epistatic interactions in the context of unstable mGnRHR variants". Needs to be backed by relevant data and statistics, or at least a reference.

We thank the reviewer for this reasonable suggestion. In the revised manuscript, we included the results of a paired Wilcoxon-Signed Rank Test that confirms the statistical significance of this observation and modified the Results section to reflect this as follows:

“Notably, these 98 variants are enriched with TMD variants (65% TMD) relative to the overall set of 251 variants (45% TMD, Fisher’s Exact Test p = 0.0019). These findings suggest random mutations form epistatic interactions in the context of unstable mGnRHR variants in a manner that depends on the specific folding defect (V276T vs. W107A) and topological context.”

Reviewer #1 (Recommendations for the Authors):

As far as this reviewer is aware, the effect of the V267T variant on MP insertion has not been measured directly; its position corresponds to T277 in TMD6 of human GnRHR that has been measured for TM insertion, but given the clear lack of conservation (threonine vs valine) the mutation in TM6 could potentially have a different impact on the mouse homologue. Please clarify what the predicted delta TM for insertion is between human and mouse GnRHR is? Moreover, I would argue that single TM insertion by tethering to Lep is insufficient to understand MP insertion/folding, as neighbouring TM helices could help to drive TM6 insertion. Has ER microsome experiments for mouse GnRHR also been carried out in the context of neighbouring helices?

We included measurements (and predictions) of the impact of the V276T substitution on the translocon-mediated membrane integration of the mouse TMD6 in the context of a chimeric Lep protein (see Fig. S1 & Table S1). Our results reveal that this substitution decreases the efficiency of TMD6 membrane integration by ~10%. Though imperfect, this prevailing biochemical assay remains popular for a variety of theoretical and technical reasons. Importantly, extensive experimental testing of this system has shown that these measurements report apparent equilibrium constants that are well-described by two-state equilibrium partitioning models (see DOIs 10.1038/nature03216 and 10.1038/nature06387). This observation provides a reasonable rationale to interpret these measurements using energetic models as we have in this work (see Table S1). From a technical perspective, the Lep system is also advantageous due to the fact that this protein is generally well expressed in the context of in vitro translation systems containing native membranes, which generally ensures a consistent signal to noise and dynamic range for membrane integration measurements. Nevertheless, the reviewers are correct that membrane integration efficiencies are likely distinct in the context of the native mGnRHR protein. For these reasons, we attempted to develop a glycosylation-based topology reporter prior to the posting and submission of this manuscript. However, all GnRHR reporters we tested were poorly expressed in vitro and the resulting 35S-labeled proteins only generated faint smears on our phosphorimaging screens that could not be interpreted. For these reasons, we chose to rely the Lep measurements for these investigations.

The lack of a more relevant topological reporter is one of many challenges we faced in our investigations of this unstable, poorly behaved protein. We share the reviewer’s frustrations concerning the speculative aspects of this work. Nevertheless, there is increasing appreciation for the fact that our perspectives on protein biophysics have been skewed by our continuing choice to focus on the relatively small set of model proteins that are compatible with our favored methodologies (doi: 10.1016/j.tibs.2013.05.001). We humbly suggest this work represents an example of how we can gain a deeper understanding of the limits of biochemical systems when we instead choose to study the unsavory bits of cellular proteomes. But this choice requires a willingness to make some reasonable assumptions and to lean on energetic/ structural modeling from time to time. Despite this limitation, we believe there is still tremendous value in this compromise.

What is the experimental evidence the W107A variant affects the protein structure? Has its melting temperature with and without inverse agonist binding for WT vs the W107A variant been measured, for example? Even heat-FSEC of detergent-solubilised membranes would be informative to know how unstable the W107A variant is. If is very unstable in detergent, then it could be that recovery mutants are going to be unlikely as you are already starting with a poor construct showing poor folding/localisation.

We again understand the rationale for this concern, but do not believe that thermal melting measurements are likely to report the same sorts of conformational transitions involved in cellular misfolding. Heating up a protein to the point in which membranes (or micelles) are disrupted and the proteins begin to form insoluble aggregates is a distinct physical process from those that occur during co- and post-translational folding within intact ER membranes at physiological temperatures (discussed further in the Response to the Reviews). Indeed, as the reviewer points out below, there seems to be little evidence that secretion is linked to thermal stability or various other metrics that others have attempted to optimize for the sake of purification and/ or structural characterization. Thus, we believe it would be just as speculative to suggest thermal aggregation represents a relevant metric for the propensity of membrane proteins to fold in the cell. The physical interpretation of membrane protein misfolding reaction remains contentious in our field due to the key fact that the denatured states of helical membrane proteins remain highly structured in a manner that is hard to generalize beyond the fact that the denatured states retain α-helical secondary structure (doi: 10.1146/annurev-biophys-051013-022926). This is in stark contrast to soluble proteins, where random coil reference states have proven to be generally useful for energetic interpretations of protein stability. For reference, our lab is currently working to leverage epistatic measurements like this to map the prevailing physiological denatured states of an integral membrane protein. Our current findings suggest that non-native electrostatic interactions form in the context of misfolded states. We hope that more information on the structural aspects of these states will help us to develop and interpret meaningful folding measurements within the membrane.

For reference, even in cases when quantitative folding measurements can be achieved, their relevance remains actively debated. As a point of reference, the corresponding author of this work previously worked on the stability and misfolding of another human α-helical membrane protein (PMP22). Like GnRHR, PMP22 is prone to misfolding in the secretory pathway and is associated with dozens of pathogenic mutations that cause protein misfolding. To understand how the thermodynamic stability of this protein is linked to secretion, the corresponding author purified PMP22, reconstituted it into n-Dodecyl-phosphocholine (DPC) micelles, and measured its resistance to denaturation by an anionic denaturing detergent (Lauryl Sarcosine, LS). The results were initially perplexing due to the fact that equilibrium unfolding curves manifested as an exponential decay (rather than a sigmoid) and relaxation kinetics appeared to be dominated by the rate constant for unfolding (doi: 10.1021/bi301635f). Unfortunately, these data could not be fit with existing folding models due to the lack of a folded protein baseline and the absence of a folding arm in the chevron plot. We eventually found that a full sigmoidal unfolding transition and refolding kinetics could be measured upon addition of 15% (v/v) glycerol. Our measurements revealed that the free energy of unfolding in DPC micelles was 0 kcal/ mol (without glycerol). This shocking lack of WT stability made it impossible to directly measure the effects of destabilizing mutations that enhance misfolding- you can’t measure the unfolding of a protein that is already unfolded. We ultimately had to instead infer the energetic effects of such mutations from the thermodynamic coupling between cofactor binding and folding (doi: 10.1021/jacs.5b03743). Finally, after demonstrating the resulting ΔΔGs correlated with both cellular trafficking and disease phenotype, we still faced justified scrutiny about the relevance of these measurements due to the fact that they were carried out in micelles. For these reasons, we do not feel that additional biophysical measurements will add much to this work until more is understood about the nature of misfolding reactions in the membrane and how to effectively recapitulate it in vitro. We also note that PMP22 is secreted with 20% efficiency in mammalian cell lines, which is 20-fold more efficient than human GnRHR under similar conditions (doi: 10.1016/j.celrep.2021.110046). Thus, we suspect equilibrium unfolding measurements are likely out of reach using previously described measurements.

Our greatest evidence suggesting W107A destabilizes the protein has to do with the fact that it deletes a highly conserved structural contact and that this structural modification kills its secretion. The fact that this mutation clearly reduces the escape of GnRHR from ER quality control is a classic indicator of misfolding that represents the cell’s way of telling us that the mutation compromises the folding of the nascent protein in some way or another. Precisely how this mutation remodels the nascent conformational ensemble of nascent GnRHR and how this relates to the free energy difference between the native and non-native portions of its conformational ensemble under cellular conditions is a much more challenging question that lies beyond the scope of this investigation (and likely beyond the scope of what’s currently possible). Indeed, there is an entire field dedicated to understanding such. Nevertheless, the difference in the epistatic interactions formed by W107A and V276T is at the very least consistent with our speculative interpretation that these two mutations vary in their misfolding mechanism and/ or in the extent to which they destabilize the protein. For these reasons, we feel the main conclusions of this manuscript are well-justified.

Please clarify if the protein is glycosylated or not and, if it is, how would this requirement affect the conclusions of your analysis?

As we noted in the Response to the Reviewers, which also constitutes a published portion of the final manuscript, this protein is indeed glycosylated. We were well aware of this aspect of the protein since inception of this project and do not think this changes our interpretation at all. Most membrane proteins are glycosylated, and several groups have demonstrated in various ways that the secretion efficiency of glycoproteins is proportional to certain stability metrics for secreted soluble proteins and membrane proteins alike. Generally, mutations that enhance misfolding do not change the propensity of the nascent chain to undergo N-linked glycosylation, which occurs during translation before protein synthesis and/ or folding is complete. Misfolded proteins typically carry lower weight glycans, which reflects their failure to advance from the ER to the Golgi, where N-linked glycans are modified and O-linked glycans are added. From our perspective, glycosyl modifications just ensure that nascent proteins are engaged by calnexin and other lectin chaperones involved in QC. It does not decouple folding from secretion efficiency. In the case of PMP22 (described above), we found that removal of its glycosylation site allows the nascent protein to bypass the lectin chaperones in a manner that enhances its plasma membrane expression eight-fold (doi: 10.1016/j.jbc.2021.100719). Similar to WT, the expression of several misfolded PMP22 variants also significantly increases upon removal of the glycosylation site. Nevertheless, their expression is still significantly lower than the un-glycosylated WT protein, and the expression patterns of the mutants relative to WT was quite similar across this panel of un-glycosylated proteins. Thus, while glycosylation certainly impacts secretion, it does not change its dependence on folding efficiency within the ER. There are many layers of partially redundant QC within the ER, and it seems that folding imposes a key bottleneck to secretion regardless of which QC proteins are involved. For these reasons, we do not think glycosylation (or other PTMs) should factor into our interpretation of these results.

One caveat with the study is that there is a poor understanding of the factors that decide if the protein should be trafficked to the PM or not. Even secretory proteins not going through the calnexin/reticulum cycle (as they have no N-linked glycans), might still get stuck in the ER, despite the fact they are functional. Could this be a technical issue of heterologous expression overloading the Sec system?

While we agree that there is much to be learned about this topic, we disagree with the notion that our understanding of folding and secretion is insufficient to generally interpret the molecular basis of the observed trends. In collaboration with various other groups, the corresponding author of this paper has shown for several other proteins that the stability of the native topology and the native tertiary structure can constrain secretion efficiency (see dois: 10.1021/jacs.8b08243, 10.1021/jacs.5b03743, and 10.1016/j.jbc.2021.100423). Moreover, the Balch and Kelly groups demonstrated many years ago that relatively simple models for the coupling between folding and chaperone binding can recapitulate the observed effects of mutations on the secretion efficiency of various proteins (doi: 10.1016/j.cell.2007.10.025). Given a wide body of prevailing knowledge in this area, we believe it is entirely reasonable to assume that the conformational effects of these mutation have a dominant effect on plasma membrane expression.

Whether or not some of the proteins retained in the ER are folded and/ or functional is an interesting question, but is outside the scope of this work. Various lines of evidence concerning approaches to rescue misfolded membrane proteins suggest many of these variants are likely to retain residual function once they escape the ER, which may suggest there are pockets of foldable/ folded proteins within the ER. But it seems generally clear that the efficiency of folding in the ER bottlenecks secretion regardless of whether or not the ER contains some fraction of folded/ functional protein. We note that it is certainly possible, if not likely, that secretion efficiency is likely to be higher at lower expression levels (doi: 10.1074/jbc.AC120.014940). However, the mutational scanning platform used in this work was designed such that all variants are expressed from an identical promoter at the same location within the genome. Thus, for the purposes of these investigations, we believe it is entirely fair to draw “apples-to-apples” comparisons of their relative effects on plasma membrane expression.

Please see Francis Arnold's paper on this point and their mutagenesis library of the channelrhodopsin (https://www.pnas.org/doi/10.1073/pnas.1700269114), which further found that 20% of mutations improved WT trafficking. Some general comparisons to this paper might be informative.

We agree that it may be interesting to compare the results from this paper to those in our own. Indeed, we find that 20% of the point mutations characterized herein also enhance the expression of WT mGnRHR, as mentioned in the Results section. However, we think it might be a bit premature to suggest this is a more general trend in light of the fact that the channelrhodopsins engineered in those studies were not of eukaryotic origin and have likely resulted from distinct evolutionary constraints. We ultimately decided against adding more on this to our already lengthy discussion in order to maintain focus on the mechanisms of epistasis.

Chris Tate and others have shown that there is a high frequency of finding stabilising point mutations in GPCRs and this is the premise of the StAR technology used to thermostabilise GPCRs in the presence of different ligands, i.e. agonist vs inverse agonists. As far as I am aware, there is a poor correlation between expression levels and thermostability (measured by ligand binding to detergent-solubilised membranes). As such, it is possible that some of the mutants might be more stable than WT even though they have lower levels of PME.

We believe the disconnect between thermostability and expression precisely speaks to our main point about the suitability of current membrane protein folding assays for the questions we address herein. The degradative activity of ER quality control has not necessarily selected for proteins that are resistant to thermal degradation and/ or are suitable for macromolecular crystallography. For this reason, it is often not so difficult to engineer proteins with enhanced thermal stability. We do not believe this disconnect signals that quality control is insensitive to protein folding and stability, but rather that it is more likely to recognize conformational defects that are distinct from those involved in thermal degradation and/ or aggregation. Indeed, recent work from the Fluman group, which builds on a wider body of previous observations, has shown that the exposure of polar groups within the membrane is a key factor that recruits degradation machinery (doi: 0.1101/2023.12.12.571171). It is hard to imagine that these sorts of conformational defects are the same as those involved in thermal aggregation.

Reviewer #2 (Recommendations For The Authors):

(1) I believe that by focusing more on the epistasis with V276T, and less on W107A, the paper could be strengthened significantly.

We appreciate this sentiment. But we believe the comparison of these two mutants really drive home the point that destabilizing mutations are not equivalent with respect to the epistatic interactions they form.

(2) In the abstract - please define the term epistasis in a simple way, to make it accessible to a general audience. For example - negative epistasis means that... this should be explicitly explained.

We thank the reviewer for this suggestion. To meet eLife formatting, we had to cut down the abstract significantly. We simplified this as best we could in the following statement:

“Though protein stability is known to shape evolution, it is unclear how cotranslational folding constraints modulate the synergistic, epistatic interactions between mutations.”

We also define positive and negative epistasis in the results section as follows:

“Positive Ɛ values denote double mutants that have greater PME than would be expected based on the effects of single mutants. Negative Ɛ values denote double mutants that have lower PME than would be expected based on the effects of single mutants. Pairs of mutations with Ɛ values near zero have additive effects on PME.”

(3) The title is quite complex and might deter readers from outside the protein evolution field. Consider simplifying it.

We thank the reviewer for this suggestion. We have simplified the title to the following:

“Divergent Folding-Mediated Epistasis Among Unstable Membrane Protein Variants”

(4) The paper could benefit from a simple figure explaining the different stages of membrane protein folding (stages 1+2) to make it more accessible to readers from outside the membrane protein field.

This is a great suggestion. We incorporated a new schematic in the revised manuscript that outlines the nature of these processes (see Fig. 1A in the revised manuscript).

(5) For the FACS-Seq experiment - it was not clear to me if and when all cells are pulled together. For example - are the 3 libraries mixed together already at the point of transfection, or are the transfected cells pulled together at any point before sorting? This could have some implications on batch effects and should, therefore, be explicitly mentioned in the main text.

We thank the reviewer for this suggestion. We modified the description of the DNA library assembly to emphasize that the mutations were generated in the context of three mixed plasmid pools, which were then transfected into the cells and sorted independently:

“We then generated a mixed array of mutagenic oligonucleotides that collectively encode this series of substitutions (Table S3) and used nicking mutagenesis to introduce these mutations into the V276T, W107A, and WT mGnRHR cDNAs (Medina-Cucurella et al., 2019), which produced three mixed plasmid pools.”

(6) The following description in the text is quite confusing. It would be better to simplify it considerably or remove it: "scores (Ɛ) were then determined by taking the log of the double mutant fitness value divided by the difference between the single mutant fitness values (see Methods)."

We thank the reviewer for this valuable feedback and have simplified the text as follows:

“To compare epistatic trends in these libraries, we calculated epistasis scores (Ɛ) for the interactions that these 251 mutations form with V276T and W107A by comparing their relative effects on PME of the WT, V276T, and W107A variants using a previously described epistasis model (product model, see Methods) (Olson et al. 2014).”

Author response:

Reviewer #1 (Public Review):

Summary:

For many years, there has been extensive electrophysiological research investigating the relationship between local field potential patterns and individual cell spike patterns in the hippocampus. In this study, using state-of-the-art imaging techniques, they examined spike synchrony of hippocampal cells during locomotion and immobility states. In contrast to conventional understanding of the hippocampus, the authors demonstrated that hippocampal place cells exhibit prominent synchronous spikes locked to theta oscillations.

Strengths:

The voltage imaging used in this study is a highly novel method that allows recording not only suprathreshold-level spikes but also subthreshold-level activity. With its high frame rate, it offers time resolution comparable to electrophysiological recordings. Moreover, it enables the visualization of actual cell locations, allowing for the examination of spatial properties (e.g., Figure 4G).

We thank the reviewer for pointing out the technical novelty of this work.

Weaknesses:

There is a notable deviation from several observations obtained through conventional electrophysiological recordings. Particularly, as mentioned below in detail, the considerable differences in baseline firing rates and no observations of ripple-triggered firing patterns raise some concerns about potential artifacts from imaging and analysis, such as cell toxicity, abnormal excitability, and false detection of spikes. While these findings are intriguing if the validity of these methods is properly proven, accepting the current results as new insights is challenging.

We appreciate the reviewer’s insightful comments regarding the intriguing aspect of our findings. Indeed, the emergence of a novel form of CA1 population synchrony presents exciting implications for hippocampal memory research and beyond.

While we acknowledge the deviations from conventional electrophysiological recordings, we respectfully contend that these differences do not necessarily imply methodological flaws. All experiments and analyses were conducted with meticulous adherence to established standards in the field.

Regarding the observed variations in averaging firing rates, it is important to note the well-documented heterogeneity in CA1 pyramidal neuron firing rates, spanning from 0.01 to 10 Hz, with a skewed distribution toward lower frequencies (Mizuseki et al., 2013). Our exclusion criteria for neurons with low estimated firing rates may have inadvertently biased the selection towards more active neurons. Moreover, prior research has indicated that averaging firing rates tend to increase during exposure to novel environments (Karlsson et al., 2008), and among deep-layer CA1 pyramidal neurons (Mizuseki et al., 2011). Given our recording setup in a highly novel environment and the predominance of deep CA1 pyramidal neurons in our sample, the observed higher averaging firing rates could be influenced by these factors. Considering these points, our mean firing rates (3.2 Hz) are reasonable estimations compared to previously reported values obtained from electrophysiological recordings (2.1 Hz in McHugh et al., 1996 and 2.4-2.6 Hz in Buzsaki et al., 2003).

Regarding concerns about potential cell toxicity, previous studies have shown that Voltron expression and illumination do not significantly alter membrane resistance, membrane capacitance, resting membrane potentials, spike amplitudes, and spike width (see Abdelfattah 2019, Science, Supplementary Figure 11 and 12). In our recordings, imaged neurons exhibit preserved membrane and dendritic morphology during and after experiments (Author response image 1), supporting the absence of significant toxicity.

Author response image 1.

Voltron-expressing neurons exhibit preserved membrane and dendritic morphology. (A) Images of two-photon z-stack maximum intensity projection showing Voltron-expressing neurons taken after voltage image experiments in vivo. (B) Post-hoc histological images of neurons being voltage-imaged.

Regarding spike detection, we use validated algorithms (Abdelfattah et al., 2019 and 2023) to ensure robust and reliable detection of spikes. Spiking activity was first separated from slower subthreshold potentials using high-pass filtering. This way, a slow fluorescence increase will not be detected as a spike, even if its amplitude is large. We benchmarked the detection algorithm in computer simulation. The sensitivity and specificity of the algorithm exceed 98% at the level of signal-to-noise ratio of our recordings. While we acknowledge that a small number of spikes, particularly those occurring later in a burst, might be missed due to their smaller amplitudes (as illustrated in Figure 1 and 2 of the manuscript), we anticipate that any missed spikes would lead to a decrease rather than an increase in synchrony between neurons. Overall, we are confident that spike detection is performed in a rigorous and robust manner.

To further strengthen these points, we will include the following in the revision:

(1) Histological images of recorded neurons during and after experiments.

(2) Further details regarding the validation of spike detection algorithms.

(3) Analysis of publicly available electrophysiological datasets.

(4) Discussion regarding the reasons behind the novelty of some of our findings compared to previous observations.

In conclusion, we assert that our experimental and analysis approach upholds rigorous standards. We remain committed to reconciling our findings with previous observations and welcome further scrutiny and engagement from the scientific community to explore the intriguing implications of our findings.

Reviewer #2 (Public Review):

Summary:

This study employed voltage imaging in the CA1 region of the mouse hippocampus during the exploration of a novel environment. The authors report synchronous activity, involving almost half of the imaged neurons, occurred during periods of immobility. These events did not correlate with SWRs, but instead, occurred during theta oscillations and were phased-locked to the trough of theta. Moreover, pairs of neurons with high synchronization tended to display non-overlapping place fields, leading the authors to suggest these events may play a role in binding a distributed representation of the context.

We thank the reviewer for a thorough and thoughtful review of our paper.

Strengths:

Technically this is an impressive study, using an emerging approach that allows single-cell resolution voltage imaging in animals, that while head-fixed, can move through a real environment. The paper is written clearly and suggests novel observations about population-level activity in CA1.

We thank the reviewer for pointing out the technical strength and the novelty of our observations.

Weaknesses:

The evidence provided is weak, with the authors making surprising population-level claims based on a very sparse data set (5 data sets, each with less than 20 neurons simultaneously recorded) acquired with exciting, but less tested technology. Further, while the authors link these observations to the novelty of the context, both in the title and text, they do not include data from subsequent visits to support this. Detailed comments are below:

We understand the reviewer’s concerns regarding the size of the dataset. Despite this limitation, it is important to note that synchronous ensembles beyond what could be expected from chance (jittering) were detected in all examined data. In the revision, we plan to add more data, including data from subsequent visits, to further strengthen our findings.

(1) My first question for the authors, which is not addressed in the discussion, is why these events have not been observed in the countless extracellular recording experiments conducted in rodent CA1 during the exploration of novel environments. Those data sets often have 10x the neurons simultaneously recording compared to these present data, thus the highly synchronous firing should be very hard to miss. Ideally, the authors could confirm their claims via the analysis of publicly available electrophysiology data sets. Further, the claim of high extra-SWR synchrony is complicated by the observation that their recorded neurons fail to spike during the limited number of SWRs recorded during behavior- again, not agreeing with much of the previous electrophysiological recordings.

We understand the reviewer’s concern. We will examine publicly available electrophysiology datasets to gain further insights into any similarities and differences to our findings. Based on these results, we will discuss why these events have not been previously observed/reported.

(2) The authors posit that these events are linked to the novelty of the context, both in the text, as well as in the title and abstract. However, they do not include any imaging data from subsequent days to demonstrate the failure to see this synchrony in a familiar environment. If these data are available it would strengthen the proposed link to novelty if they were included.

We thank the reviewer’s constructive suggestion. We will acquire more datasets from subsequent visits to gain further insights into these synchronous events.

3) In the discussion the authors begin by speculating the theta present during these synchronous events may be slower type II or attentional theta. This can be supported by demonstrating a frequency shift in the theta recording during these events/immobility versus the theta recording during movement.

We thank the reviewer’s constructive suggestion. We did demonstrate a frequency shift to a lower frequency in the synchrony-associated theta during immobility than during locomotion (see Fig. 4B, the red vs. blue curves). We will enlarge this panel and specifically refer to it in the corresponding discussion paragraph.

(4) The authors mention in the discussion that they image deep-layer PCs in CA1, however, this is not mentioned in the text or methods. They should include data, such as imaging of a slice of a brain post-recording with immunohistochemistry for a layer-specific gene to support this.

We thank the reviewer’s constructive suggestion. We do have images of brain slices post-recordings (Author response image 2). Imaged neurons are clearly located in the deep CA1 pyramidal layer. We will add these images and quantification in the revised manuscript.

Author response image 2.

Imaged neurons are located in the deep pyramidal layer of the dorsal hippocampal CA1 region.

Reviewer #3 (Public Review):

Summary:

In the present manuscript, the authors use a few minutes of voltage imaging of CA1 pyramidal cells in head-fixed mice running on a track while local field potentials (LFPs) are recorded. The authors suggest that synchronous ensembles of neurons are differentially associated with different types of LFP patterns, theta and ripples. The experiments are flawed in that the LFP is not "local" but rather collected in the other side of the brain, and the investigation is flawed due to multiple problems with the point process analyses. The synchrony terminology refers to dozens of milliseconds as opposed to the millisecond timescale referred to in prior work, and the interpretations do not take into account theta phase locking as a simple alternative explanation.

We genuinely appreciate the reviewer’s feedback and acknowledge the concerns raised. However, we believe these concerns can be effectively addressed without undermining the validity of our conclusions. With this in mind, we respectfully disagree with the assessment that our experiments and investigation are flawed. Please allow us to address these concerns and offer additional context to support the validity of our study.

Weaknesses:

The two main messages of the manuscript indicated in the title are not supported by the data. The title gives two messages that relate to CA1 pyramidal neurons in behaving head-fixed mice: (1) synchronous ensembles are associated with theta (2) synchronous ensembles are not associated with ripples.

There are two main methodological problems with the work:

(1) Experimentally, the theta and ripple signals were recorded using electrophysiology from the opposite hemisphere to the one in which the spiking was monitored. However, both signals exhibit profound differences as a function of location: theta phase changes with the precise location along the proximo-distal and dorso-ventral axes, and importantly, even reverses with depth. And ripples are often a local phenomenon - independent ripples occur within a fraction of a millimeter within the same hemisphere, let alone different hemispheres. Ripples are very sensitive to the precise depth - 100 micrometers up or down, and only a positive deflection/sharp wave is evident.

We appreciate the reviewer’s consideration regarding the collection of LFP from the contralateral hemisphere. While we acknowledge the limitation of this design, we believe that our findings still offer valuable insights into the dynamics of synchronous ensembles. Despite potential variations in theta phases with recording locations and depth, we find that the occurrence and amplitudes of theta oscillations are generally coordinated across hemispheres (Buzsaki et al., Neurosci., 2003). Therefore, the presence of prominent contralateral LFP theta around the times of synchronous ensembles in our study (see Figure 4A of the manuscript) strongly supports our conclusion regarding their association with theta oscillations, despite the collection of LFP from the opposite hemisphere.

In addition, in our manuscript, we specifically mentioned that the “preferred phases” varied from session to session, likely due to the variability of recording locations (see Line 254-256). Therefore, we think that the reviewer’s concern regarding theta phase variability has already been addressed in the present manuscript.

Regarding ripple oscillations, while we recognize that they can sometimes occur locally, the majority of ripples occur synchronously in both hemispheres (up to 70%, see Szabo et al., Neuron, 2022; Buzsaki et al., Neurosci., 2003). Therefore, using contralateral LFP to infer ripple occurrence on the ipsilateral side has been a common practice in the field, employed by many studies published in respectable journals (Szabo et al., Neuron, 2022; Terada et al., Nature, 2021; Dudok et al., Neuron, 2021; Geiller et al., Neuron, 2020). Furthermore, our observation that 446 synchronous ensembles during immobility do not co-occur with contralateral ripples, and the remaining 313 ensembles during locomotion are not associated with ripples, as ripples rarely occur during locomotion. Therefore, our conclusion that synchronous ensembles are not associated with ripple oscillations is supported by data.

(2) The analysis of the point process data (spike trains) is entirely flawed. There are many technical issues: complex spikes ("bursts") are not accounted for; differences in spike counts between the various conditions ("locomotion" and "immobility") are not accounted for; the pooling of multiple CCGs assumes independence, whereas even conditional independence cannot be assumed; etc.

We acknowledge the reviewer’s concern regarding spike train analysis. Indeed, complex bursts or different behavioral conditions can lead to differences in spike counts that could potentially affect the detection of synchronous ensembles. However, our jittering procedure (see Line 121-132) is designed to control for the variation of spike counts. Importantly, while the jittered spike trains also contain the same spike count variations, we found 7.8-fold more synchronous events in our data compared to jitter controls (see Figure 1G of the manuscript), indicating that these factors cannot account for the observed synchrony.

To explicitly demonstrate that complex bursts cannot account for the observed synchrony, we have performed additional analysis to remove all latter spikes in bursts and only count the single and the first spikes of bursts. Importantly, we found that this procedure did not change the rate and size of synchronous ensembles, nor did it significantly alter the grand-average CCG (see Author response image 3). The results of this analysis explicitly rule out a significant effect of complex spikes on the analysis of synchronous ensembles.

Author response image 3.

Population synchrony remains after the removal of spikes in bursts. (A) The grand-average cross correlogram (CCG) was calculated using spike trains without latter spikes in bursts. The gray line represents the mean grand average CCG between reference cells and randomly selected cells from different sessions. (B) Pairwise comparison of the event rates of population synchrony between spike trains containing all spikes and spike trains without latter spikes in bursts. Bar heights indicate group means (n=10 segments, p=0.036, Wilcoxon signed-rank test). (C) Histogram of the ensemble sizes as percentages of cells participating in the synchronous ensembles.

Beyond those methodological issues, there are two main interpretational problems: (1) the "synchronous ensembles" may be completely consistent with phase locking to the intracellular theta (as even shown by the authors themselves in some of the supplementary figures).

We agree with the reviewer that the synchronous ensembles are indeed consistent with theta phase locking. However, it is important to note that theta phase locking alone does not necessarily imply population synchrony. In fact, theta phase locking has been shown to “reduce” population synchrony in a previous study (Mizuseki et al., 2014, Phil. Trans. R. Soc. B.). Thus, the presence of theta phase locking cannot be taken as a simple alternative explanation of the synchronous ensembles.

To directly assess the contribution of theta phase locking to synchronous ensembles, we have performed a new analysis to randomize the specific theta cycles in which neurons spike, while keeping the spike phases constant. This manipulation disrupts spike co-occurrence while preserving theta phase locking, allowing us to test whether theta phase locking alone can explain the population synchrony, or whether spike co-occurrence in specific cycles is required. The grand-average CCG shows a much smaller peak compared to the original peak (Author response image 4A). Moreover, synchronous event rates show a 4.5-fold decrease in the randomized data compared to the original event rates (Author response image 4B). Thus, the new analysis reveals theta phase locking alone cannot account for the population synchrony.

Author response image 4.

Drastic reduction of population synchrony by randomizing spikes to other theta cycles while preserving the phases. (A) The grand-average cross correlogram (CCG) was calculated using original spike trains (black) and randomized spike trains where theta phases of the spikes are kept the same but spike timings were randomly moved to other theta cycles (red). (B) Pairwise comparison of the event rates of population synchrony between the original spike trains and randomized spike trains (n=10 segments, p=0.002, Wilcoxon signed-rank test). Bar heights indicate group means. ** p<0.01

(2) The definition of "synchrony" in the present work is very loose and refers to timescales of 20-30 ms. In previous literature that relates to synchrony of point processes, the timescales discussed are 1-2 ms, and longer timescales are referred to as the "baseline" which is actually removed (using smoothing, jittering, etc.).

Regarding the timescale of synchronous ensembles, we acknowledge that it varies considerably across studies and cell types. However, it is important to note that a timescale of dozens, or even hundreds of milliseconds is common for synchrony terminology in CA1 pyramidal neurons (see Csicsvari et al., Neuron, 2000; Harris et al., Science, 2003; Malvache et al., Science, 2016; Yagi et al., Cell Reports, 2023). In fact, a timescale of 20-30 ms is considered particularly important for information transmission and storage in CA1, as it matches the membrane time constant of pyramidal neurons, the period of hippocampal gamma oscillations, and the time window for synaptic plasticity. Therefore, we believe that this timescale is relevant and in line with established practices in the field.

Author response:

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

eLife assessment

Connelly and colleagues provide convincing genetic evidence that importation from mainland Tanzania is a major source of Plasmodium falciparum lineages currently circulating in Zanzibar. This study also reveals ongoing local malaria transmission and occasional near-clonal outbreaks in Zanzibar. Overall, this research highlights the role of human movements in maintaining residual malaria transmission in an area targeted for intensive control interventions over the past decades and provides valuable information for epidemiologists and public health professionals.

Reviewer #1 (Public Review):

Zanzibar archipelago is close to achieving malaria elimination, but despite the implementation of effective control measures, there is still a low-level seasonal malaria transmission. This could be due to the frequent importation of malaria from mainland Tanzania and Kenya, reservoirs of asymptomatic infections, and competent vectors. To investigate population structure and gene flow of P. falciparum in Zanzibar and mainland Tanzania, they used 178 samples from mainland Tanzania and 213 from Zanzibar that were previously sequenced using molecular inversion probes (MIPs) panels targeting single nucleotide polymorphisms (SNPs). They performed Principal Component Analysis (PCA) and identity by descent (IBD) analysis to assess genetic relatedness between isolates. Parasites from coastal mainland Tanzania contribute to the genetic diversity in the parasite population in Zanzibar. Despite this, there is a pattern of isolation by distance and microstructure within the archipelago, and evidence of local sharing of highly related strains sustaining malaria transmission in Zanzibar that are important targets for interventions such as mass drug administration and vector control, in addition to measures against imported malaria.

Strengths:

This study presents important samples to understand population structure and gene flow between mainland Tanzania and Zanzibar, especially from the rural Bagamoyo District, where malaria transmission persists and there is a major port of entry to Zanzibar. In addition, this study includes a larger set of SNPs, providing more robustness for analyses such as PCA and IBD. Therefore, the conclusions of this paper are well supported by data.

Weaknesses:

Some points need to be clarified:

(1) SNPs in linkage disequilibrium (LD) can introduce bias in PCA and IBD analysis. Were SNPs in LD filtered out prior to these analyses?

Thank you for this point. We did not filter SNPs in LD prior to this analysis. In the PCA analysis in Figure 1, we did restrict to a single isolate among those that were clonal (high IBD values) to prevent bias in the PCA. In general, disequilibrium is minimal only over small distances <5-10kb without selective forces at play. This is much less than the average spacing of the markers in the panel. If there is minimal LD, the conclusions drawn on relative levels and connections at high IBD are unlikely to be confounded by any effects of disequilibrium.

( 2) Many IBD algorithms do not handle polyclonal infections well, despite an increasing number of algorithms that are able to handle polyclonal infections and multiallelic SNPs. How polyclonal samples were handled for IBD analysis?

Thank you for this point. We added lines 157-161 to clarify. This section now reads:

“To investigate genetic relatedness of parasites across regions, identity by descent (IBD) estimates were assessed using the within sample major alleles (coercing samples to monoclonal by calling the dominant allele at each locus) and estimated utilizing a maximum likelihood approach using the inbreeding_mle function from the MIPanalyzer package (Verity et al., 2020). This approach has previously been validated as a conservative estimate of IBD (Verity et al., 2020).”

Please see the supplement in (Verity et al., 2020) for an extensive simulation study that validates this approach.

Reviewer #1 (Recommendations For The Authors):

(3) I think Supplementary Figures 8 and 9 are more visually informative than Figure 2.

Thank you for your response. We performed the analysis in Figure 2 to show how IBD varies between different regions and is higher within a region than between.

Reviewer #2 (Public Review):

This manuscript describes P. falciparum population structure in Zanzibar and mainland Tanzania. 282 samples were typed using molecular inversion probes. The manuscript is overall well-written and shows a clear population structure. It follows a similar manuscript published earlier this year, which typed a similar number of samples collected mostly in the same sites around the same time. The current manuscript extends this work by including a large number of samples from coastal Tanzania, and by including clinical samples, allowing for a comparison with asymptomatic samples.

The two studies made overall very similar findings, including strong small-scale population structure, related infections on Zanzibar and the mainland, near-clonal expansion on Pemba, and frequency of markers of drug resistance. Despite these similarities, the previous study is mentioned a single time in the discussion (in contrast, the previous research from the authors of the current study is more thoroughly discussed). The authors missed an opportunity here to highlight the similar findings of the two studies.

Thank you for your insights. We appreciated the level of detail of your review and it strengthened our work. We have input additional sentences on lines 292-295, which now reads:

“A recent study investigating population structure in Zanzibar also found local population microstructure in Pemba (Holzschuh et al., 2023). Further, both studies found near-clonal parasites within the same district, Micheweni, and found population microstructure over Zanzibar.”

Strengths:

The overall results show a clear pattern of population structure. The finding of highly related infections detected in close proximity shows local transmission and can possibly be leveraged for targeted control.

Weaknesses:

A number of points need clarification:

(1) It is overall quite challenging to keep track of the number of samples analyzed. I believe the number of samples used to study population structure was 282 (line 141), thus this number should be included in the abstract rather than 391. It is unclear where the number 232 on line 205 comes from, I failed to deduct this number from supplementary table 1.

Thank you for this point. We have included 282 instead of 391 in the abstract. We added a statement in the results at lines 203-205 to clarify this point, which now reads:

“PCA analysis of 232 coastal Tanzanian and Zanzibari isolates, after pruning 51 samples with an IBD of greater than 0.9 to one representative sample, demonstrates little population differentiation (Figure 1A).”

(2) Also, Table 1 and Supplementary Table 1 should be swapped. It is more important for the reader to know the number of samples included in the analysis (as given in Supplementary Table 1) than the number collected. Possibly, the two tables could be combined in a clever way.

Thank you for this advice. Rather than switch to another table altogether, we appended two columns to the original table to better portray the information (see Table 1).

Methods

(3) The authors took the somewhat unusual decision to apply K-means clustering to GPS coordinates to determine how to combine their data into a cluster. There is an obvious cluster on Pemba islands and three clusters on Unguja. Based on the map, I assume that one of these three clusters is mostly urban, while the other two are more rural. It would be helpful to have a bit more information about that in the methods. See also comments on maps in Figures 1 and 2 below.

Cluster 3 is a mix of rural/urban while the clusters 2, 4 and 5 are mostly rural. This analysis was performed to see how IBD changes in relation to local context within different regions in Zanzibar, showing that there is higher IBD within locale than between locale.

(4) Following this point, in Supplemental Figure 5 I fail to see an inflection point at K=4. If there is one, it will be so weak that it is hardly informative. I think selecting 4 clusters in Zanzibar is fine, but the justification based on this figure is unclear.

The K-means clustering experiment was used to cluster a continuous space of geographic coordinates in order to compare genetic relatedness in different regions. We selected this inflection point based on the elbow plot and based the number to obtain sufficient subsections of Zanzibar to compare genetic relatedness. This point is added to the methods at lines 174-178, which now reads:

“The K-means clustering experiment was used to cluster a continuous space of geographic coordinates in order to compare genetic relatedness in different regions. We selected K = 4 as the inflection point based on the elbow plot (Supplemental Figure 5) and based the number to obtain sufficient subsections of Zanzibar to compare genetic relatedness.”

(5) For the drug resistance loci, it is stated that "we further removed SNPs with less than 0.005 population frequency." Was the denominator for this analysis the entire population, or were Zanzibar and mainland samples assessed separately? If the latter, as for all markers <200 samples were typed per site, there could not be a meaningful way of applying this threshold. Given data were available for 200-300 samples for each marker, does this simply mean that each SNP needed to be present twice?

Population frequency is calculated based on the average within sample allele frequency of each individual in the population, which is an unbiased estimator. Within sample allele frequency can range from 0 to 1. Thus, if only one sample has an allele and it is at 0.1 within sample frequency, the population allele frequency would be 0.1/100 = 0.001. This allele is removed even though this would have resulted in a prevalence of 0.01. This filtering is prior to any final summary frequency or prevalence calculations (see MIP variant Calling and Filtering section in the methods). This protects against errors occurring only at low frequency.

Discussion:

(6) I was a bit surprised to read the following statement, given Zanzibar is one of the few places that has an effective reactive case detection program in place: "Thus, directly targeting local malaria transmission, including the asymptomatic reservoir which contributes to sustained transmission (Barry et al., 2021; Sumner et al., 2021), may be an important focus for ultimately achieving malaria control in the archipelago (Björkman & Morris, 2020)." I think the current RACD program should be mentioned and referenced. A number of studies have investigated this program.

Thank you for this point. We have added additional context and clarification on lines 275-280, which now reads:

“Thus, directly targeting local malaria transmission, including the asymptomatic reservoir which contributes to sustained transmission (Barry et al., 2021; Sumner et al., 2021), may be an important focus for ultimately achieving malaria control in the archipelago (Björkman & Morris, 2020). Currently, a reactive case detection program within index case households is being implemented, but local transmission continues and further investigation into how best to control this is warranted (Mkali et al. 2023).”

(7) The discussion states that "In Zanzibar, we see this both within and between shehias, suggesting that parasite gene flow occurs over both short and long distances." I think the term 'long distances' should be better defined. Figure 4 shows that highly related infections rarely span beyond 20-30 km. In many epidemiological studies, this would still be considered short distances.

Thank you for this point. We have edited the text at lines 287-288 to indicate that highly related parasites mainly occur at the range of 20-30km, which now reads:

“In Zanzibar, highly related parasites mainly occur at the range of 20-30km.”

(8) Lines 330-331: "Polymorphisms associated with artemisinin resistance did not appear in this population." Do you refer to background mutations here? Otherwise, the sentence seems to repeat lines 324. Please clarify.

We are referring to the list of Pfk13 polymorphisms stated in the Methods from lines 146-148. We added clarifying text on lines 326-329:

“Although polymorphisms associated with artemisinin resistance did not appear in this population, continued surveillance is warranted given emergence of these mutations in East Africa and reports of rare resistance mutations on the coast consistent with spread of emerging Pfk13 mutations (Moser et al., 2021). “

(9) Line 344: The opinion paper by Bousema et al. in 2012 was followed by a field trial in Kenya (Bousema et al, 2016) that found that targeting hotspots did NOT have an impact beyond the actual hotspot. This (and other) more recent finding needs to be considered when arguing for hotspot-targeted interventions in Zanzibar.

We added a clarification on this point on lines 335-345, which now reads:

“A recent study identified “hotspot” shehias, defined as areas with comparatively higher malaria transmission than other shehias, near the port of Zanzibar town and in northern Pemba (Bisanzio et al., 2023). These regions overlapped with shehias in this study with high levels of IBD, especially in northern Pemba (Figure 4). These areas of substructure represent parasites that differentiated in relative isolation and are thus important locales to target intervention to interrupt local transmission (Bousema et al., 2012). While a field cluster-randomized control trial in Kenya targeting these hotspots did not confer much reduction of malaria outside of the hotspot (Bousema et al. 2016), if areas are isolated pockets, which genetic differentiation can help determine, targeted interventions in these areas are likely needed, potentially through both mass drug administration and vector control (Morris et al., 2018; Okell et al., 2011). Such strategies and measures preventing imported malaria could accelerate progress towards zero malaria in Zanzibar.”

Figures and Tables:

(10) Table 2: Why not enter '0' if a mutation was not detected? 'ND' is somewhat confusing, as the prevalence is indeed 0%.

Thank you for this point. We have put zero and also given CI to provide better detail.

(11) Figure 1: Panel A is very hard to read. I don't think there is a meaningful way to display a 3D-panel in 2D. Two panels showing PC1 vs. PC2 and PC1 vs. PC3 would be better. I also believe the legend 'PC2' is placed in the wrong position (along the Y-axis of panel 2).

Supplementary Figure 2B suffers from the same issue.

Thank you for your comment. A revised Figure 1 and Supplemental Figure 2 are included, where there are separate plots for PC1 vs. PC2 and PC1 vs. PC3.

(12) The maps for Figures 1 and 2 don't correspond. Assuming Kati represents cluster 4 in Figure 2, the name is put in the wrong position. If the grouping of shehias is different between the Figures, please add an explanation of why this is.

Thank you for this point. The districts with at least 5 samples present are plotted in the map in Figure 1B. In Figure 2, a totally separate analysis was performed, where all shehias were clustered into separate groups with k-means and the IBD values were compared between these clusters. These maps are not supposed to match, as they are separate analyses. Figure 1B is at the district level and Figure 2 is clustering shehias throughout Zanzibar.

The figure legend of Figure 1B on lines 410-414 now reads:

“B) A Discriminant Analysis of Principal Components (DAPC) was performed utilizing isolates with unique pseudohaplotypes, pruning highly related isolates to a single representative infection. Districts were included with at least 5 isolates remaining to have sufficient samples for the DAPC. For plotting the inset map, the district coordinates (e.g. Mainland, Kati, etc.) are calculated from the averages of the shehia centroids within each district.”

The figure legend of Figure 2 on lines 417-425 now reads:

“Figure 2. Coastal Tanzania and Zanzibari parasites have more highly related pairs within their given region than between regions. K-means clustering of shehia coordinates was performed using geographic coordinates all shehias present from the sample population to generate 5 clusters (colored boxes). All shehias were included to assay pairwise IBD between differences throughout Zanzibar. Pairwise comparisons of within cluster IBD (column 1 of IBD distribution plots) and between cluster IBD (column 2-5 of IBD distribution plots) was done for all clusters. In general, within cluster IBD had more pairwise comparisons containing high IBD identity.”

(13) Figure 2: In the main panel, please clarify what the lines indicate (median and quartiles?). It is very difficult to see anything except the outliers. I wonder whether another way of displaying these data would be clearer. Maybe a table with medians and confidence intervals would be better (or that data could be added to the plots). The current plots might be misleading as they are dominated by outliers.

Thank you for this point and it greatly improved this figure. We changed the plotting mechanisms through using a beeswarm plot, which plots all pairwise IBD values within each comparison group.

(14) In the insert, the cluster number should not only be given as a color code but also added to the map. The current version will be impossible to read for people with color vision impairment, and it is confusing for any reader as the numbers don't appear to follow any logic (e.g. north to south).

Thank you very much for these considerations. We changed the color coding to a color blind friendly palette and renamed the clusters to more informative names; Pemba, Unguja North (Unguja_N), Unguja Central (Unguja_C), Unguja South (Unguja_S) and mainland Tanzania (Mainland).

(15) The legend for Figure 3 is difficult to follow. I do not understand what the difference in binning was in panels A and B compared to C.

Thank you for this point. We have edited the legend to reflect these changes. The legend for Figure 3 on lines 427-433 now reads:

“Figure 3. Isolation by distance is shown between all Zanzibari parasites (A), only Unguja parasites (B) and only Pemba parasites (C). Samples were analyzed based on geographic location, Zanzibar (N=136) (A), Unguja (N=105) (B) or Pemba (N=31) (C) and greater circle (GC) distances between pairs of parasite isolates were calculated based on shehia centroid coordinates. These distances were binned at 4km increments out to 12 km. IBD beyond 12km is shown in Supplemental Figure 8. The maximum GC distance for all of Zanzibar was 135km, 58km on Unguja and 12km on Pemba. The mean IBD and 95% CI is plotted for each bin.”

(16) Font sizes for panel C differ, and it is not aligned with the other panels.

Thank you for pointing this out. Figure 3 and Supplemental Figure 10 are adjusted with matching formatting for each plot.

(17) Why is Kusini included in Supplemental Figure 4, but not in Figure 1?

In Supplemental Figure 4, all isolates were used in this analysis and isolates with unique pseudohaplotypes were not pruned to a single representative infection. That is why there are additional isolates in Kusini. The legend for Supplemental Figure 4 now reads:

“Supplemental Figure 4. PCA with highly related samples shows population stratification radiating from coastal Mainland to Zanzibar. PCA of 282 total samples was performed using whole sample allele frequency (A) and DAPC was performed after retaining samples with unique pseudohaplotypes in districts that had 5 or more samples present (B). As opposed to Figure 1, all isolates were used in this analysis and isolates with unique pseudohaplotypes were not pruned to a single representative infection.”

(18) Supplemental Figures 6 and 7: What does the width of the line indicate?

The sentence below was added to the figure legends of Supplemental Figures 6 and 7 and the legends of each network plot were increased in size:

“The width of each line represents higher magnitudes of IBD between pairs.”

(19) What was the motivation not to put these lines on the map, as in Figure 4A? This might make it easier to interpret the data.

Thank you for this comment. For Supplemental Figure 8 and 9, we did not put these lines that represent lower pairwise IBD to draw the reader's attention to the highly related pairs between and within shehias.

Reviewer #2 (Recommendations For The Authors):

(1) There is a rather long paragraph (lines 300-323) on COI of asymptomatic infections and their genetic structure. Given that the current study did not investigate most of the hypotheses raised there (e.g. immunity, expression of variant genes), and the overall limited number of asymptomatic samples typed, this part of the discussion feels long and often speculative.

Thank you for your perspective. The key sections highlighted in this comment, regarding immunity and expression of variant genes, were shortened. This section on lines 300-303 now reads:

“Asymptomatic parasitemia has been shown to be common in falciparum malaria around the globe and has been shown to have increasing importance in Zanzibar (Lindblade et al., 2013; Morris et al., 2015). What underlies the biology and prevalence of asymptomatic parasitemia in very low transmission settings where anti-parasite immunity is not expected to be prevalent remains unclear (Björkman & Morris, 2020).”

(2) As a detail, line 304 mentions "few previous studies" but only one is cited. Are there studies that investigated this and found opposite results?

Thank you for this comment. We added additional studies that did not find an association between clinical disease and COI. These changes are on lines 303-308, which now reads:

“Similar to a few previous studies, we found that asymptomatic infections had a higher COI than symptomatic infections across both the coastal mainland and Zanzibar parasite populations (Collins et al., 2022; Kimenyi et al., 2022; Sarah-Matio et al., 2022). Other studies have found lower COI in severe vs. mild malaria cases (Robert et al., 1996) or no significant difference between COI based on clinical status (Earland et al. 2019; Lagnika et al. 2022; Conway et al. 1991; Kun et al. 1998; Tanabe et al. 2015)”

(3) Table 2: Percentages need to be checked. To take one of several examples, for Pfk13-K189N a frequency of 0.019 for the mutant allele is given among 137 samples. 2/137 equals to 0.015, and 3/137 to 0.022. 0.019 cannot be achieved. The same is true for several other markers. Possibly, it can be explained by the presence of polyclonal infections. If so, it should be clarified what the total of clones sequenced was, and whether the prevalence is calculated with the number of samples or number of clones as the denominator.

Thank you for this point. We mistakenly reported allele frequency instead of prevalence. An updated Table 2 is now in the manuscript. The method for calculating the prevalence is now at lines 148-151:

“Prevalence was calculated separately in Zanzibar or mainland Tanzania for each polymorphism by the number of samples with alternative genotype calls for this polymorphism over the total number of samples genotyped and an exact 95% confidence interval was calculated using the Pearson-Klopper method for each prevalence.”

Author response:

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

eLife assessment

This study presents useful findings regarding the role of formin-like 2 in mouse oocyte meiosis. The submitted data are supported by incomplete analyses, and in some cases, the conclusions are overstated. If these concerns are addressed, this paper would be of interest to reproductive biologists.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The presented study focuses on the role of formin-like 2 (FMNL2) in oocyte meiosis. The authors assessed FMNL2 expression and localization in different meiotic stages and subsequently, by using siRNA, investigated the role of FMNL2 in spindle migration, polar body extrusion, and distribution of mitochondria and endoplasmic reticulum (ER) in mouse oocytes.

Strengths:

Novelty in assessing the role of formin-like 2 in oocyte meiosis.

Weaknesses:

Methods are not properly described.

Overstating presented data.

It is not clear what statistical tests were used.

My main concern is that there are missing important details of how particular experiments and analyses were done. The material and methods section are not written in the way that presented experiments could be repeated - it is missing basic information (e.g., used mouse strain, timepoints of oocytes harvest for particular experiments, used culture media, image acquisition parameters, etc.). Some of the presented data are overstated and incorrectly interpreted. It is not clear to me how the analysis of ER and mitochondria distribution was done, which is an important part of the presented data interpretation. I'm also missing important information about the timing of particular stages of assessed oocytes because the localization of both ER and mitochondria differs at different stages of oocyte meiosis. The data interpretation needs to be justified by proper analysis based on valid parameters, as there is considerable variability in the ER and mitochondria structure and localization across oocytes based on their overall quality and stage.

Thank you for your comment. We regret the oversight of omitting critical information in the manuscript. In the revised manuscript, we have included essential details such as mouse strains, culture media, stages of oocyte and statistical methods in the materials and methods section. Please find our details responses in the “Recommendations for the authors” part.

Reviewer #2 (Public Review):

Summary:

This research involves conducting experiments to determine the role of Fmnl2 during oocyte meiosis I.

Strengths:

Identifying the role of Fmnl2 during oocyte meiosis I is significant.

Weaknesses:

The quantitative analysis and the used approach to perturb FMNL2 function are currently incomplete and would benefit from more confirmatory approaches and rigorous analysis.

(1) Most of the results are expected. The new finding here is that FMNL2 regulates cytoplasmic F-actin in mouse oocytes, which is also expected given the role of FMNL2 in other cell types. Given that FMNL2 regulates cytoplasmic F-actin, it is very expected to see all the observed phenotypes. It is already established that F-actin is required for spindle migration to the oocyte cortex, extruding a small polar body and normal organelle distribution and functions.

Thank you for your comment. In the recent decade, Arp2/3 complex (Nat Cell Biol 2011), Formin2 (Nat Cell Biol 2002, Nat Commun 2020), and Spire (Curr Biol 2011) were reported to be 3 key factors to involve into this process. These factors regulate actin filaments in different ways. However, how they cross with each other for the subcellular events were still fully clear. Our current study identified that FMNL2 played a critical role in coordinating these molecules for actin assembly in oocytes. Our findings demonstrate that FMNL2 interacts with both the Arp2/3 complex and Formin2 to facilitate actin-based meiotic spindle migration. Additionally, we discovered a novel role for FMNL2 in determining the distribution and function of the endoplasmic reticulum and mitochondria, which may in turn influence meiotic spindle migration in oocytes. Our results not only uncover the novel functions of FMNL2-mediated actin for organelle distribution, but also extend our understanding of the molecular basis for the unique meiotic spindle migration in oocyte meiosis.

(2) The authors used Fmnl2 cRNA to rescue the effect of siRNA-mediated knockdown of Fmnl2. It is not clear how this works. It is expected that the siRNA will also target the exogenous cRNA construct (which should have the same sequence as endogenous Fmnl2) especially when both of them were injected at the same time. Is this construct mutated to be resistant to the siRNA?

Thank you for your question. We regret any misunderstanding that may have been caused by the inappropriate description in our manuscript. In the rescue experiments, we initially injected FMNL2 siRNA into oocytes, followed by the microinjection of FMNL2 mRNA 18-20 hours later. After conducting our previous experiments, we have verified through Western blotting that endogenous FMNL2 is effectively suppressed 18-20 hours following the microinjection of FMNL2 siRNA. Additionally, we observed a significant increase in exogenous FMNL2 protein expression 2 hours after the injection of FMNL2 mRNA. We believe that the exogenous FMNL2 could compensate the decrease by FMNL2 knockdown, and this approach was adopted in many oocyte studies.

(3) The authors used only one approach to knockdown FMNL2 which is by siRNA. Using an additional approach to inhibit FMNL2 would be beneficial to confirm that the effect of siRNA-mediated knockdown of FMNL2 is specific.

Thank you for your question. Yes, the specificity is always the concern for siRNA or morpholino microinjection due to the off-target issue. Due to the limitation we could not generate the knock out model, and there are no known inhibitors with specific targeting capabilities for FMNL2. To solve this, we performed the rescue study with exogenous mRNA to confirm the effective knock down of FMNL2. These measures provide reassurance regarding the credibility of the experimental outcomes, and this is also the general way to avoid the off-target of siRNA or morpholino.

Reviewer #3 (Public Review):

Summary:

The authors focus on the role of formin-like protein 2 in the mouse oocyte, which could play an important role in actin filament dynamics. The cytoskeleton is known to influence a number of cellular processes from transcription to cytokinesis. The results show that downregulation of FMNL2 affects spindle migration with resulting abnormalities in cytokinesis in oocyte meiosis I.

Weaknesses:

The overall description of methods and figures is overall dismissively poor. The description of the sample types and number of replicate experiments is impossible to interpret throughout, and the quantitative analysis methods are not adequately described. The number of data points presented is unconvincing and unlikely to support the conclusions. On the basis of the data presented, the conclusions appear to be preliminary, overstated, and therefore unconvincing.

Thank you for your comment. We regret the oversight of omitting critical information in the manuscript. In the revised manuscript, we have incorporated your suggestions for modification, particularly regarding the Materials and Methods section. Please see the detailed revision and responses in the “Recommendations for the authors” part.

Recommendations for the authors:

Reviewer #1 (Recommendations for The Authors):

My main concern is that there are missing important details of how particular experiments and analyses were done. The material and methods section is not written in the way that presented experiments could be repeated - it is missing basic information (e.g., used mouse strain, timepoints of oocytes harvest for particular experiments, used culture media, image acquisition parameters, etc.). Some of the presented data are overstated and incorrectly interpreted. It is not clear to me how the analysis of ER and mitochondria distribution was done, which is an important part of the presented data interpretation. I'm also missing important information about the timing of particular stages of assessed oocytes because the localization of both ER and mitochondria differs at different stages of oocyte meiosis. The data interpretation needs to be justified by proper analysis based on valid parameters, as there is considerable variability in the ER and mitochondria structure and localization across oocytes based on their overall quality and stage. My specific comments are listed below.

(1) Information about statistical tests that were used needs to be provided for all quantification experiments.

Thank you for your suggestion. Based on your suggestions, we revised the statistical analysis description in the Materials and Methods section. Additionally, we also included a description of the statistical methods in the legends of the relevant result figures.

(2) I recommend replacing the plunger plots, used in most quantification data, with alternatives allowing evaluation of the distribution of the data (dot plots, box plots, whisker plots).

Thank you for your suggestion. Following your suggestion, we replaced the plunger plots in Fig 2C, D, H, I and Fig3 B, C with dot plots.

(3) Can the authors provide information about particular time points when were individual oocyte stages (GVBD, meiosis I, and meiosis II) harvested/used for immunofluorescence protein detection, western blotting, microinjection, and ER and mitochondria staining? Were the time points always the same in all presented experiments and experimental vs control group? If not, this needs to be clarified.

Thank you for your suggestion. We used oocytes in the metaphase I (MI) stage for the statistical analysis of spindle migration, actin filament aggregation, endoplasmic reticulum localization, and mitochondrial localization. In the Western blot analysis, GV stage oocytes were utilized to evaluate the efficiency of knockdown and rescue experiments. The protein expression levels of Arp2, Formin2, INF2, Cofilin, Grp78, and Chop in different treatment groups were detected using MI-stage oocytes. In the revised version, we provided all the detailed information about the stages.

(4) Figure 1B: Can the authors comment on why there is a missing representative image of MII oocyte FMBL2-Ab? I recommend including this in the figure to have a complete view of comparing overexpressed and endogenous FMNL2 localization in oocyte meiosis.

Thank you for your suggestion. In the revised manuscript, we added immunostaining images of FMNL2 antibody in MII stage oocytes.

(5) Figure 1C: The figure legend says, "FMNL2 and actin overlapped in cortex and spindle surrounding". In MI oocytes, there is usually no accumulated actin signal around the spindle, which is also true in the presented images, so there cannot be overlapping with the FMNL2 signal. The interpretation should be changed.

We apologize for this inappropriate description that was used, and we deleted this sentence.

(6) Figure 2B: What were the parameters of the "large" and "normal" polar bodies for performing the analysis?

Thank you for your question. In order to assess the size of the polar body, we conducted a comparison between the diameter of the polar body and that of the oocyte. If the diameter of the polar body was found to be less than 1/3 of the oocyte's diameter, we categorized it as normal-sized polar body. Conversely, if the polar body's diameter exceeded 1/3 of the oocyte's diameter, we categorized it as a large polar body. We have included these details in the Results section of the manuscript.

(7) Figure 2F: Can the authors comment on what can be the second band in the rescue group?

Thank you for your question. In the rescue experiment, we microinjected exogenous FMNL2-EGFP mRNA into the oocytes. As a result, compared to endogenous FMNL2, the protein size increased due to the addition of the EGFP tag, approximately 27 kDa. Hence, in the Western blot bands of the rescue group, the upper band represents the expression of exogenous FMNL2-EGFP, while the lower band corresponds to the expression of endogenous FMNL2. We have provided annotations in the revised Figure 2F to clarify this.

(8) Can the authors comment on the variability of PBE between 2C and 2H in the FMNL2-KD groups? In panel C, the PBE in the KD group was 59.5 {plus minus} 2.82%; in panel H, the PBE in the KD group was 48.34 {plus minus} 4.2%, and in the rescue group, the PBE was 62.62 {plus minus} 3.6%. The rescue group has a similar PBE rate as the KD group in panel C. How consistent was the FMNL2 knockdown across individual replicates? Can the authors provide more details on how the rescue experiment was performed?

Thank you for your question. We believe that the difference in PBE observed in Figure 2C and 2H of the FMNL2-KD group was due to the microinjection times and the duration of in vitro arrest. The results shown in Figure 2C depict the outcome of a single injection of FMNL2 siRNA into GV stage oocytes, followed by 18 hours of in vitro arrest; the results shown in Figure 2H contain a subsequent additional injection of FMNL2-EGFP mRNA with another 2 hours of arrest. The two rounds of microinjection and the extended period of in vitro arrest both affect oocyte maturation rates.

(9). Figure 2J and K: What groups were compared together? The used statistic needs to be properly described.

Thank you for your question. The FMNL2-KD, FMNL3-KD, and FMNL2+3-KD groups were all compared to the Control group, therefore, t-test was used for analysis. We have provided explanations in the revised manuscript.

(10) Figure 4B and C: Can the authors provide representative images without oversaturated actine signal?

Thank you for your question. For the analysis of oocyte F-actin, the F-actin are divided into cortex actin and cytoplasmic actin. Due to the contrast during imaging, the strong cortex actin signals affected the detection of cytoplasmic actin, therefore, it is necessary to increase the scanning index, which will cause the overexpose the cortex actin signal. This is for the better observation of the cytoplasmic signals.

(11) Figure 4G + 5H: Can the authors comment on why they used as a housekeeping gene actin instead of tubulin, which was used in the rest of the WB experiments?

Thank you for your question. In most of the western blot experiments conducted in this study, we used tubulin as a housekeeping gene. However, due to the supply of antibodies by delivery period, we had GAPDH and actin as well for some experiments. These housekeeping genes were all valid for the study.

(12) Based on what parameters was ER considered normally or abnormally distributed, and what stages of oocytes were assessed?

Thank you for your question. In this study, we employed oocytes at the MI stage for the analysis of ER localization. In the MI stage, the ER localized around the spindle, which is regarded as the typical localization pattern. The ER displayed a dispersed distribution throughout the cytoplasm or clustered were categorized as aberrant positioning. We included relevant descriptions in the revised version of the manuscript.

(13) Figure 5H: As a housekeeping gene was used actin - the quantification is labeled as a Grp78 to tubulin ratio.

Thank you for pointing out the error. This is a label mistake and we corrected it.

(14) Information about how JC-1 staining was done needs to be provided.

Thank you for your carefully reading. We included a description of JC1 staining in the Materials and Methods section.

(15). Line 231-232: "As shown in Figure 4A" - the text doesn't correspond to the figure.

Thank you for pointing out the error. We revised this mistake in the revised manuscript by correcting "Fig3A" to "Fig4A."

(16) Line 265: there is probably a missing word "Formin2".

Thank you and we corrected the error and made the necessary changes in the revised manuscript.

Reviewer #2 (Recommendations for The Authors):

(1) Quantification and analysis:

• Fig. 3B: The rate of spindle migration should be quantified based on the distance from the spindle to the cortex. Also, the orientation of the spindle (Z-position) needs to be taken into consideration.

• Fig. 5C, D: It is unclear how the rate of ER distribution was calculated.

• Western blot: In many experiments (such as Fig. 5H), the bands are saturated which will prevent accurate intensity measurements and quantifications.

For spindle migration, we specifically focused on spindles exhibiting a distinctive spindle-like shape with clear bipolarity to eliminate any statistical discrepancies potentially caused by variations in Z-axis alignment. Our criterion for determining successful migration was based on the contact between the spindle pole and the cortical region of the oocyte. Therefore, we think that the rate is better to reflect the phenotype than the distance.

For the examination of ER localization, Reviewer 1 also raised this issue. We utilized oocytes at the MI stage in this study. The ER localized around the spindle in MI stage. The ER displayed a dispersed distribution throughout the cytoplasm or clustered were categorized as aberrant positioning. We included relevant descriptions in the revised version of the manuscript.

For the bands of the western blot results, during the experimental procedure we typically capture multiple images at different exposure levels (3-5 images). In the revised manuscript, we have replaced the inappropriate images with more suitable ones.

(2) Given that all Immunoprecipitation experiments in this manuscript were performed on the whole ovary which contains more somatic cells than oocytes, the results do not necessarily reflect meiotic oocytes. Please consider this possibility during the interpretation.

Thank you for your suggestion. Yes, we agree with you. In the revised manuscript, we made appropriate modifications to the relevant descriptions.

(3) 351-365: The conclusion that Arp2/3 compensates for the decreased formin 2 in FMNL2 knockdown oocytes is a bit unconvincing. 1- In mouse oocytes, it is already known that Arp2/3 and formin 2 regulate different pools of F-actin nucleation. 2- The authors found an increase in Arp2/3 in FMNL2 knockdown oocytes compared to control oocytes without any change in cortical F-actin. Given that Arp2/3 is primarily promoting cortical F-actin, it is expected to see an increase in cortical F-actin in FMNL2 knockdown oocytes, which was not the case.

Thank you for your question. Yes, previous studies showed that formin2 localizes to the cytoplasm of oocytes and accumulates around the spindle, which facilitate cytoplasmic actin assembly. While Arp2/3 is primarily responsible for actin assembly at the cortex region of oocytes. In invasive cells, FMNL2 is mainly localized in the leading edge of the cell, lamellipodia and filopodia tips, to improve cell migration ability by actin-based manner (Curr Biol 2012). We showed that FMNL2 localized both at spindle periphery and cortex, but depletion of FMNL2 did not affect cortex actin intensity. We think that FMNL2 and Arp2/3 both contribute to the cortex actin dynamics, when FMNL2 decreased, ARP2 increased to compensate for this, which maintained the cortex actin level. In the revised manuscript, we have made modifications to avoid excessive extrapolation from our results, ensuring that our conclusions are presented in a more objective manner.

(4) Lines 195-197: The spindle is initially formed soon after the GVBD, so there is no spindle during GVBD. Also, I can't see oocytes at anaphase I or telophase I in this figure. Please revise.

Thank you for your suggestion. We apologize for the inappropriate descriptions that were used. In the revised manuscript, we have made modifications to the respective descriptions in the Results part.

(5) Fig. 2E: It seems that the control oocyte is abnormal with mild cytokinesis defects. Please replace or delete it since this information is already included in Fig. 3A.

Thank you for your suggestion. Based on our observations, during the extrusion of the first polar body in oocytes, there is a temporary occurrence of cellular morphological fragmentation due to cortical reorganization (11h in control oocyte from Fig 2E). However, after the extrusion of the first polar body, the oocyte morphology returns to normal. Figure 2E illustrates the meiotic division process of oocytes, while Figure 3A primarily focuses on the process of oocyte spindle migration. We think that it is better to retain both to present our results.

Reviewer #3 (Recommendations for The Authors):

In the case of the observed phenotype, the stage of GV is important. The phenotypes presented also occur in meiotic or developmentally incompetent oocytes. In addition, the images of GV oocytes appear as NSN, which also show the KD phenotype in Figs. 2 and 3.

Thank you for your concern. As the oocyte grows, the proportion of SN-type oocytes gradually increases. When the oocyte diameter reaches 70-80 μm, the proportion of SN oocytes is approximately 52.7% (Mol Reprod Dev. 1995). In our study, both the control and knockdown groups collected oocytes with a diameter of around 80 μm, which is considered as fully-grown oocytes, predominantly in the SN phase. Since the collection period and size of the oocytes were consistent, we can sure that the observed differences between the control and knockdown groups in phenotype analysis could be solid and reliable.

MII is absent in Fig. 1B.

In the revised manuscript, we added immunostaining images of FMNL2 in MII stage oocytes.

The result of KD is not convincing. Also, discuss whether the heterozygous effect of Fmnl2 deletion affects reproductive fitness.

Thank you for your concern. In our investigation, limited to the setup of knock out model, we employed siRNA to knockdown FMNL2 expression, to avoid the risk of off-target, we performed rescue experiment with exogenous mRNA, which we believe that it could solve this issue. When designing siRNA sequences, we ensured their specificity for binding to FMNL2 mRNA only, and we assessed the levels of FMNL2 and FMNL3 mRNA in oocytes after injection of FMNL2 siRNA. The results showed that, compared to the control group, the expression of FMNL2 mRNA decreased by approximately 70% after 18 hours of FMNL2 siRNA injection, while the level of FMNL3 mRNA was not decreased.

Fig. 2F rescue experiment with double bands. What bands are seen here? Did the authors inject tagged or untagged FMNL2? Or does endogenous FMNL2 appear higher in the sample after KD?

Thank you for your question. In the rescue experiment, we microinjected exogenous FMNL2-EGFP mRNA into the oocytes. As a result, compared to endogenous FMNL2, the protein size increased due to the addition of the EGFP tag, approximately 27 kDa. Hence, in the Western blot bands of the rescue group, the upper band represents the expression of exogenous FMNL2-EGFP, while the lower band corresponds to the expression of endogenous FMNL2. We provided annotations in the revised Figure 2F to clarify this.

Variability in mitochondria and ER distribution patterns is also known in healthy and developing oocytes, although the authors described only a single phenotype.

Thank you for your concern. Yes, mitochondria and ER show dynamic localization in different stage of oocyte maturation. However, in this study we employed oocyte MI stage for the analysis of ER and mitochondria localization, and in MI stage, both the ER and mitochondria localize around the spindle. This pattern is considered as the normal localization. Several studies showed that dispersed or clustered localization contributed to maturation defects. We included relevant descriptions in the revised manuscript.

What exactly is meant by input in the IP experiments? Why is the target missing in the input sample?

Thank you for your question. We subjected the input samples to electrophoresis on a single channel, all the analyzed proteins demonstrated normal expression, thereby confirming the viability of the input sample. However, upon simultaneous exposure with the IP samples, we observed a lack of clear signal for certain proteins in the input group. This phenomenon is due to the excessive signal intensity resulting from protein enrichment in the IP group, which caused the low exposure of proteins in input group.

Explain the rationale for using, actin or tubulin as loading or normalization controls in the study focusing on the cytoskeleton.

Thank you for your question. Actin and tubulin are both widely used as the control due to their stable expression. For actin, there are α-actin and β-actin isoforms. Formins and Arp2/3 complex regulate the polymerization of α-actin and β-actin to form F-actin, not isoform expression. In our study F-actin (the functional type) was examined. While α-tubulin and β-tubulin are two subtypes of tubulin, and they interact with each other to form stable α/β-tubulin heterodimers. The changes of cytoskeleton dynamics could not change the expression of α/β-tubulin. Therefore, β-actin and α-tubulin could be used as normalization controls.

Fig. 6E shows only , but the legend says *.

Thank you for pointing out the error. We correct the mistake in the revised manuscript.

Spindle positioning appears to differ between control and KD. Does this affect the quantification of Fig. 6F? Adequate nomenclature should be used here.

Thank you for your question. Yes, spindle positioning was affected by FMNL2 depletion. However, central spindle or cortex spindle all belong to MI stage, and JC1 is not related with the stage difference. To avoid misunderstanding we replaced the representative images and corresponding description in Figure 6F.

The description of the methods and legends should be significantly improved.

Thank you for your suggestion. Reviewer 1 and 2 also raised the similar concern. We enriched the description of methods and legends in the revised manuscript.

Author response:

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors collected genomic information from public sources covering 423 eukaryote genomes and around 650 prokaryote genomes. Based on pre-computed CDS annotation, they estimated the frequency of alternative splicing (AS) as a single average measure for each genome and computed correlations with this measure and other genomic properties such as genome size, percentage of coding DNA, gene and intergenic span, etc. They conclude that AS frequency increases with genome complexity in a somewhat directional trend from "lower" organisms to "higher" organisms.

Strengths:

The study covers a wide range of taxonomic groups, both in prokaryotes and eukaryotes.

Weaknesses:

The study is weak both methodologically and conceptually. Current high throughput sequencing technologies, coupled with highly heterogeneous annotation methods, can observe cases of AS with great sensitivity, and one should be extremely cautious of the biases and rates of false positives associated with these methods. These issues are not addressed in the manuscript.

We are aware of the bias that may exist in annotation files. Since the source of noise can be highly variable, we have assumed that most of the data has a similar bias. However, we agree with the reviewer that we could perform some analysis to test for these biases and their association to different methodologies. Thus, we will measure the uncertainty present in the data. From one side, we will be more explicit about the data limitations and the biases it can generate in the results. On the other side, while analyzing the false positives in the data is out of our scope, we will perform a statistical test to detect possible biases regarding different methods of sequencing and annotation, and types of organisms (model or non-model organisms). If positive, we will proceed, as far as possible, to normalize the data or to estimate a confidence interval.

Here, AS measures seem to be derived directly from CDS annotations downloaded from public databases, and do not account for differing annotation methods or RNA sequencing depth and tissue sample diversity.

Beyond taking into account the differential bias that may exist in the data, we do not consider that our AS measure is problematic. The NCBI database is one of the most reliable databases that we have to date and is continuously updated from all scientific community. So, the use of this data and the corresponding procedures for deriving the AS measure are perfectly acceptable for a comparative analysis on such a huge global scale. Furthermore, the proposal of a new genome-level measure of AS that allows to compare species spanning the whole tree of life is part of the novelty of the study. We understand that small-scale studies require a high specificity about the molecular processes involved in the study. However, this is not the case, where we are dealing with a large-scale problem. On the other side, as we have previously mention, we agree with the reviewer to analyze the degree of uncertainty in the data to better interpret the results.

There is no mention of the possibility that AS could be largely caused by random splicing errors, a possibility that could very well fit with the manuscript's data. Instead, the authors adopt early on the view that AS is regulated and functional, generally citing outdated literature.

There is no question that some AS events are functional, as evidenced by strongly supported studies. However, whether all AS events are functional is questionable, and the relative fractions of functional and non-functional AS are unknown. With this in mind, the authors should be more cautious in interpreting their data.

Many studies suggest that most of the AS events observed are the result of splicing errors and are therefore neither functional nor conserved. However, we still have limited knowledge about the functionality of AS. Just because we don’t have a complete understanding of its functionality, doesn’t mean there isn’t a fundamental cause behind these events. AS is a highly dynamic process that can be associated with processes of a stochastic nature that are fundamental for phenotypic diversity and innovation. This is one of the reasons why we do not get into a discussion about the functionality of AS and consider it as a potential measure of biological innovation. Nevertheless, we agree with the reviewer’s comments, so we will add a discussion about this issue with updated literature and look at any possible misinterpretation of the results.

The "complexity" of organisms also correlates well (negatively) with effective population size. The power of selection to eliminate (slightly) deleterious mutations or errors decreases with effective population size. The correlation observed by the authors could thus easily be explained by a non-adaptive interpretation based on simple population genetics principles.

We appreciate the observation of the reviewer. We know well the M. Lynch’s theory on the role of the effective population size and its eventual correlation with genomic parameters, but we want to emphasize that our objective is not to find an adaptive or non-adaptive explanation of the evolution of AS, but rather to reveal it. Nevertheless, as the reviewer suggests, we will look at the correlation between the AS and the effective population size and discuss about a possible non-adaptive interpretation.

The manuscript contains evidence that the authors might benefit from adopting a more modern view of how evolution proceeds. Sentences such as "... suggests that only sophisticated organisms optimize alternative splicing by increasing..." (L113), or "especially in highly evolved groups such as mammals" (L130), or the repeated use of "higher" and "lower" organisms need revising.

As the reviewer suggests, we will proceed with the corresponding linguistic corrections.

Because of the lack of controls mentioned above, and because of the absence of discussion regarding an alternative non-adaptive interpretation, the analyses presented in the manuscript are of very limited use to other researchers in the field. In conclusion, the study does not present solid conclusions.

Reviewer #2 (Public Review):

Summary:

In this contribution, the authors investigate the degree of alternative splicing across the evolutionary tree and identify a trend of increasing alternative splicing as you move from the base of the tree (here, only prokaryotes are considered) towards the tips of the tree. In particular, the authors investigate how the degree of alternative splicing (roughly speaking, the number of different proteins made from a single ORF (open reading frame) via alternative splicing) relates to three genomic variables: the genome size, the gene content (meaning the fraction of the genome composed of ORFs), and finally, the coding percentage of ORFs, meaning the ratio between exons and total DNA in the ORF. When correlating the degree of alternative splicing with these three variables, they find that the different taxonomic groups have a different correlation coefficient, and identify a "progressive pattern" among metazoan groups, namely that the correlation coefficient mostly increases when moving from flowering plants to arthropods, fish, birds, and finally mammals. They conclude that therefore the amount of splicing that is performed by an organismal group could be used as a measure of its complexity.

Weaknesses:

While I find the analysis of alternative splicing interesting, I also find that it is a very imperfect measure of organismal complexity and that the manuscript as a whole is filled with unsupported statements. First, I think it is clear to anyone studying evolution over the tree of life that it is the complexity of gene regulation that is at the origin of much of organismal structural and behavioral complexity. Arguably, creating different isoforms out of a single ORF is just one example of complex gene regulation. However, the complexity of gene regulation is barely mentioned by the authors.

We disagree with the reviewer with that our measure of AS is imperfect. Just as we responded to the first reviewer, we will quantify the uncertainty in the data and correct for differential biases caused by annotation and sequencing methods. Thus, beyond correcting relevant biases in the data, we consider that our measure is adequate for a comparative analysis at a global scale. A novelty of our study is the proposal of a genome-level measure of AS that takes into account data from the entire scientific community.

We want also to emphasize that we assume from the beginning that AS may reflect some kind of biological complexity, it is not a conclusion from the results. An argument in favor of such an assumption is that AS is associated with stochastic processes that are fundamental for phenotypic diversity and innovation. Of course, we agree with the reviewer that it is not the only mechanism behind biological complexity, so we will emphasize it in the manuscript. On the other side, we will be more explicit about the assumptions and objectives, and will correct any unsupported statement.

Further, it is clear that none of their correlation coefficients actually show a simple trend (see Table 3). According to these coefficients, birds are more complex than mammals for 3 out of 4 measures.

An evolutionary trend is broadly defined as the gradual change in some characteristic of organisms as they evolve or adapt to a specific environment. Under our context, we define an evolutionary trend as the gradual change in genome composition and its association with AS across the main taxonomic groups. If we look at Figure 4 and Table 3 we can conclude that there is a progressive trend. We will be more precise about how we define an evolutionary trend and correct any possible misinterpretation of the results. On the other side, we do not assume that mammals should be more complex than birds. First, we will emphasize that our results show that birds have the highest values of such a trend. Second, after reading the reviewer’s comments, we have decided that we will perform an additional analysis to correct for differences in the taxonomic group sizes, which will allow us to have more confidence in the results.

It is also not clear why the correlation coefficient between alternative splicing ratio and genome length, gene content, and coding percentage should display such a trend, rather than the absolute value. There are only vague mechanistic arguments.

The study analyzes the relationship of AS with genomic composition for the large taxonomic groups. We assume that significant differences in these relationships are indicators of the presence of different mechanisms of genome evolution. However, we agree with the reviewer that a correlation does not imply a causal relation, so we will be more cautious when interpreting the results.

To quantify the relationships we use correlation coefficients, the slopes of such correlations, and the relation of variability. Although the absolute values of AS are also illustrated in Table 4, we consider that they are less informative than if we include how it relates to the genomic composition. For example, we observe that plants have a different genome composition and relation with AS if compared to animals, which suggest that they follow different mechanisms of genome evolution. On the other hand, we observe a trend in animals, where high values of AS are associated to a large percentage of introns and a percentage of intergenic DNA of about the 50% of genomes.

Much more troubling, however, is the statement that the data supports "lineage-specific trends" (lines 299-300). Either this is just an ambiguous formulation, or the authors claim that you can see trends within lineages.

We agree with the reviewer that this statement is not correct, so we will proceed to correct it.

The latter is clearly not the case. In fact, within each lineage, there is a tremendous amount of variation, to such an extent that many of the coefficients given in Table 3 are close to meaningless. Note that no error bars or p-values are presented for the values shown in Table 3. Figure 2 shows the actual correlation, and the coefficient for flowering plants there is given as 0.151, with a p-value of 0.193. Table 3 seems to quote r=0.174 instead. It should be clear that a correlation within a lineage or species is not a sign of a trend.

The reviewer is not understanding correctly the results in Table 3. It is precisely the variation of the genome variables what we are measuring. Given the standardization of these values by the mean values, we have proceeded to compare the variability between groups, which is the result shown in Table 3. In this case there are no error bars or p-values associated. On the other hand, we agree that a correlation is not a sign of a trend. But the relations of variability, together with the results obtained in Figure 3, are indicators of a trend. As we mentioned before, we will proceed to analyze whether the variation in the group sizes is causing a bias in the results.

There are several wrong or unsupported statements in the manuscript. Early on, the authors state that the alternative splicing ratio (a number greater or equal to one that can be roughly understood as the number of different isoforms per ORF) "quantifies the number of different isoforms that can be transcribed using the same amount of information" (lines 51-52). But in many cases, this is incorrect, because the same sequence can represent different amounts of information depending on the context. So, if a changed context gives rise to a different alternative splice, it is because the genetic sequence has a different meaning in the changed context: the information has changed.

We agree that there are not well supported statements, so we will proceed to revise them.

In line 149, the authors state that "the energetic cost of having large genomes is high". No citation is given, and while such a statement seems logical, it does not have very solid support.

We will also revise the bibliography and support our statements with updated references.

If there was indeed a strong selective force to reduce genome size, we would not see the stunning diversity of genome sizes even within lineages. This statement is repeated (without support) several times in the manuscript, apparently in support of the idea that mammals had "no choice" to increase complexity via alternative splicing because they can't increase it by having longer genomes. I don't think this reasoning can be supported.

We agree with the reviewer in this issue, so we will carefully revise the statements that indirectly (or directly) assume the action of selective forces on the genome composition.

Even more problematic is the statement that "the amount of protein-coding DNA seems to be limited to a size of about 10MB" (line 219). There is no evidence whatsoever for this statement.

In Figure 1A we observe a one-to-one relationship between the genome size and the amount of coding. However, in multicellular organisms, although the genome size increases we observe that the amount of coding does not increase by more than 10Mb, which suggest the presence of some genomic limitation. Of course, this is not an absolute or general statement, but rather a suggestion. We are only describing our results.

The reference that is cited (Choi et al 2020) suggests that there is a maximum of 150GB in total genome size due to physiological constraints. In lines 257-258, the authors write that "plants are less restricted in terms of storing DNA sequences compared to animals" (without providing evidence or a citation).

We will revise the bibliography and add updated references.

I believe this statement is made due to the observation that plants tend to have large intergenic regions. But without examining the functionality of these interagency regions (they might host long non-coding RNA stretches that are used to regulate the expression of other genes, for example) it is quite adventurous to use such a simple measure as being evidence that plants "are less restricted in terms of storing DNA sequences", whatever that even means. I do not think the authors mean that plants have better access to -80 freezers. The authors conclude that "plant's primary mechanism of genome evolution is by expanding their genome". This statement itself is empty: we know that plants are prone to whole genome duplication, but this duplication is not, as far as we understand, contributing to complexity. It is not a "primary mechanism of genome evolution".

We will revise these statements.

In lines 293-294, the authors claim that "alternative splicing is maximized in mammalian genomes". There is no evidence that this ratio cannot be increased. So, to conclude (on lines 302-303) that alternative splicing ratios are "a potential candidate to quantify organismal complexity" seems, based on this evidence, both far-fetched and weak at the same time.

Our results show the highest values of AS in mammals, but we understand that the results are limited to the availability and accuracy of data, which we will emphasize in the manuscript. As we previously mention, we will also proceed to analyze the uncertainty in data and carry out the appropriate corrections.

I am also not very comfortable with the data analysis. The authors, for example, say that they have eliminated from their analysis a number of "outlier species". They mention one: Emmer wheat because it has a genome size of 900 Mb (line 367). Since 900MB does not appear to be extreme, perhaps the authors meant to write 900 Gb. When I consulted the paper that sequenced Triticum dicoccoides, they noted that 14 chromosomes are about 10GB. Even a tetraploid species would then not be near 900Gb. But more importantly, such a study needs to state precisely which species were left out, and what the criteria are for leaving out data, lest they be accused of selecting data to fit their hypothesis.

The reviewer is right, we wanted to say 900Mb, which is approximately 7.2Gb. We had a mistake of nomenclature. This value is extreme compared to the typical values, so it generates large deviations when applying measures of central tendency and dispersion. We want to obtain mean values that are representative of the most species composing the taxonomic groups, so we find appropriate to exclude all outlier values in the study. Nevertheless, we will specify the criteria that we have used to select the data in a rigorous way.

I understand that Methods are often put at the end of a manuscript, but the measures discussed here are so fundamental to the analysis that a brief description of what the different measures are (in particular, the "alternative splicing ratio") should be in the main text, even when the mathematical definition can remain in the Methods.

We agree with the reviewer, so we will add a brief description of the genomic variables at the beginning of the Results section.

Finally, a few words on presentation. I understand that the following comments might read differently after the authors change their presentation. This manuscript was at the border of being comprehensible. In many cases, I could discern the meaning of words and sentences in contexts but sometimes even that failed (as an example above, about "species-specific trends", illustrates). The authors introduced jargon that does not have any meaning in the English language, and they do this over and over again.

Note that I completely agree with all the comments by the other reviewer, who alerted me to problems I did not catch, including the possible correlation with effective population size: a possible non-adaptive explanation for the results.

Author response:

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

We understand that this is a unique situation (or at least to us it is) to have a reviewer resign and we are now down to one reviewer. In light of this, would either of you be willing to let us know your thoughts and suggestions on our manuscript? We think this would help direct us as to how to proceed.

In light of our work being confirmed by a different laboratory as well as the numerous improvements and additions we provided adhering to the requirements of both reviewers, we were surprised and disappointed by no changes in the eLife assessment of the manuscript.

In more detail, please note and consider that:

(1) Our results have been confirmed, i.e., replicated already by a different independent team (Gonzalez-Palomares, E., Boulanger-Bertolus, J., Dupin, M. et al. Amplitude modulation pattern of rat distress vocalisations during fear conditioning. Sci Rep 13, 11173 (2023). https://doi.org/10.1038/s41598-023-38051-7); and these authors are citing our Elife preprint.

They have observed very comparable 44-kHz vocalizations during a similar fear conditioning protocol – with 10 foot shocks; 44-kHz vocalizations were described to occur during later conditioning trials, i.e., 8-10, in contrast to trials 1-3 (Fig. S4) – all of which confirms our findings. Before our Elife report, it was difficult or discouraging to report 44-kHz vocalizations since they did not follow or fit the established definitions of types/subtypes of rat vocalizations.

We are confused our findings gained no merit and reliability despite the replication.

(2) We have addressed all the concerns and comments of both reviewers, and made substantial additions. In particular, Reviewer #1 suggested fatigue as a possible cause of increased emissions of 44-kHz vocalizations. We followed this hypothesis in good faith and performed several analyses which required several months of work and manual inspection of numerous vocalizations and video recordings. In the end, these results concerning the duration and mean power of the calls, which do not support the fatigue idea were included. However, this addition was not mentioned in the reviews or the assessment.

We are not sure what to do in regards to some recommendations (e.g., concerning “behavioral response to call playback” or aversive calls which sometimes “can occur at frequencies above 22 kH”) repeated in the second review which we have already addressed.

Also, we are unsure of the basis of some of the Reviewer’s concerns, in particular: that 44-kHz calls are not a separate group but “merely represent expected variation in vocalization parameters” or a problem with referring to 44-kHz calls as “a new group” (is it because they likely have been emitted before, in different experiments – though undetected, or they are closely related to 22-kHz calls, i.e., constitute the same group).

(3) We would be happy to make additional improvements and changes (e.g., changing “Rats” to “Male rats” within the title) but it would be really helpful to have some confirmation as to which ones of these changes are crucial according to the journal Editors. For example, we are now running new analyses to show that 44-kHz calls occur in groups, i.e., they usually happen one after the other – but again we are not sure if this is the kind of analysis required at this stage of the manuscript submission; and whether it addresses the concerns of Reviewer #1.

---

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

We would like to express our gratitude to the reviewers for their suggestions and critiques as we continually strive to enhance the quality of the manuscript. We improved it, by incorporating the reviewers’ suggestions, changing the content and numbering of figures (Figs 1, 3S1 were edited; 4 figures were moved to supplemental materials), and adding several analyses suggested by the reviewers along with accompanying figures (1S2, 1S3) and tables (1 and 2). These analyses include investigating the link between freezing behavior and 44-kHz calls as well as their sound mean power and duration. Also, we have introduced detailed information regarding the experiments performed as well as expanded the description and discussion of the results section. Finally, we added the information about 44-kHz calls reported by another group – which was inspired by our findings.

Below is the point-by-point response to the reviewers’ comments.

Reviewer #1 (Public Review):

Olszyński and colleagues present data showing variability from canonical "aversive calls", typically described as long 22 kHz calls rodents emit in aversive situations. Similarly long but higher-frequency (44 kHz) calls are presented as a distinct call type, including analyses both of their acoustic properties and animals' responses to hearing playback of these calls. While this work adds an intriguing and important reminder, namely that animal behavior is often more variable and complex than perhaps we would like it to be, there is some caution warranted in the interpretation of these data. The authors also do not provide adequate justification for the use of solely male rodents. With several reported sex differences in rat vocal behaviors this means caution should be exercised when generalizing from these findings.

We fully agree that our data should be interpreted with caution and we followed the Reviewer’s suggestions along these lines (see below). Also, we appreciate the suggestion to explore the prevalence of 44-kHz calls in female subjects, which would indeed represent an important and intriguing extension of our research. However, due to present financial constraints, we can only plan such experiments. To address the comment, we have added the sentence: “Here we are showing introductory evidence that 44-kHz vocalizations are a separate and behaviorally-relevant group of rat ultrasonic calls. These results require further confirmations and additional experiments, also in form of repetition, including research on female rat subjects.”

It is important to note that the data presented in the current manuscript originates primarily from previously conducted experiments. These earlier experiments employed male subjects only; it was due to established evidence indicating that the female estrus cycle significantly influences ultrasonic vocalization (Matochik et al., 1992). Adhering to controls for the estrus cycle would require a greater number of female subjects than males, which would not only increase animal suffering but also escalate the demands of human labor and financial costs.

Firstly, the authors argue that the shift to higher-frequency aversive calls is due to an increase in arousal (caused by the animals having received multiple aversive foot shocks towards the end of the protocols). However, it cannot be ruled out that this shift would be due to factors such as the passage of time and increase in fatigue of the animals as they make vocalizations (and other responses) for extended periods of time. In fact the gradual frequency increase reported for 22 kHz calls and the drop in 44 kHz calls the next day in testing is in line with this.

Answer: We would like to point out that the “increased-arousal” hypothesis, declared in the manuscript, is only a hypothesis – as reflected by the wording used. However, we changed the beginning of the sentence in question from “It could be argued” to “We would like to propose a hypothesis” to emphasize the speculative aspect of the proposed explanation behind the increase of 44-kHz ultrasonic emissions.

Also, we do agree that other factors could contribute to the increased emission of 44kHz calls. These factors could include: heightened fear, stress/anxiety, annoyance/anger, disgust/boredom, grief/sadness, despair/helplessness, and weariness/fatigue. We are listing these potential factors in the discussion. Also, we added: “It is not possible, at this stage, to determine which factors played a decisive role. Please note that the potential contribution of these factors is not mutually exclusive”. However, we propose a list of arguments supporting the idea that 44-kHz vocalizations communicate an increased negative emotional state. Among these arguments were the conclusions drawn from additional analyses – mostly inspired by the fatigue hypothesis proposed by the Reviewer #1. In particular, we investigated changes in the sound mean power and duration of 22-kHz and 44-kHz calls. Specifically, we showed that the mean power of 44-kHz vocalizations did not change, and was higher than that of 22-kHz vocalizations (Fig. 1S2EF).

Finally, the Reviewer #1 listed “the gradual frequency increase reported for 22 kHz calls and the drop in 44 kHz calls the next day” as arguments for the fatigue hypothesis. We do not agree that the “increase” should be interpreted as a sign of fatigue [Producing and maintaining higher frequency calls require greater effort from the vocalizer, on which we elaborated in the manuscript], also we are not sure what “drop in 44 kHz calls” the Reviewer is referring to [We assume it refers to less 44-kHz calls during testing vs. training; we suppose that the levels of arousal are lower in the test due to shorter session time and lack of shocks, which additionally contributes to fear extinction].

Secondly, regarding the analysis where calls were sorted using DBSCAN based on peak frequency and duration, it is not surprising that the calls cluster based on frequency and duration, i.e. the features that are used to define the 44 kHz calls in the first place. Thus presenting this clustering as evidence of them being truly distinct call types comes across as a circular argument.

Answer: The DBSCAN sorting results were to convey that when changing the clustering ε value, the degree of cluster separation, the 44-kHz vocalizations remained distinct from the 22-kHz and various short-call clusters that merged. In other words: 44-kHz calls remained separate from long 22-kHz, short 22-kHz and 50-kHz vocalizations, which all consolidated into one common cluster. As a result, in this mathematical analysis, 44-kHz vocalizations remained distinct without applying human biases. Additionally, frequency and duration are the two most common features used to define all types of calls (Barker et al., 2010; Silkstone & Brudzynski, 2019a, 2019b; Willey & Spear, 2013). In summary, we did not expect the analysis to isolate out the 44-kHz calls, and we were surprised by this result.

The sparsity of calls in the 30-40 kHz range (shown in the individual animal panels in Figure 2C) could in theory be explained by some bioacoustics properties of rat vocal cords, without necessarily the calls below and above that range being ethologically distinct.

Answer: We respectfully disagree with the argument regarding sparsity. It is important to note that, during prolonged fear conditioning experiments, we observed an increased incidence of 44-kHz calls (Fig. 1E-G) of up to >19% (Fig. 1S2AB) of the total ultrasonic vocalizations during specific inter-trial intervals. Also, it is possible that in observed experimental circumstances almost every fifth call could be attributed to the vocal apparatus as an artifact of its functioning (assuming we are interpreting the Reviewer’s argument correctly). While we do not believe this to be the case, we acknowledge the importance of considering such a hypothesis.

The behavioral response to call playback is intriguing, although again more in line with the hypothesis that these are not a distinct type of call but merely represent expected variation in vocalization parameters. Across the board animals respond rather similarly to hearing 22 kHz calls as they do to hearing 44 kHz calls, with occasional shifts of 44 kHz call responses to an intermediate between appetitive and aversive calls. This does raise interesting questions about how, ethologically, animals may interpret such variation and integrate this interpretation in their responses. However, the categorical approach employed here does not address these questions fully.

Answer: We are unsure of the Reviewer’s critique in this paragraph and will attempt to address it to the best of our understanding. Our finding of up to >19% of long seemingly aversive, 44-kHz calls, at a frequency in the define appetitive ultrasonic range (usually >32 kHz) is unexpected rather than “expected”. We would agree that aversive call variation is expected, but not in the appetitive frequency range.

Kindly note the findings by Saito et al. (2019), which claim that frequency band plays the main role in rat ultrasonic perception. It is possible that the higher peak frequency of 44kHz calls may be a strong factor in their perception by rats, which is, however, modified by the longer duration and the lack of modulation.

Also, from our experience, it is quite challenging to demonstrate different behavioral responses of naïve rats to pre-recorded 22-kHz (aversive) vs. 50-kHz (appetitive) vocalizations. Therefore, to demonstrate a difference in response to two distinct, potentially aversive, calls, i.e., 22-kHz vs. 44-kHz calls, to be even more difficult (as to our knowledge, a comparable experiment between short vs. long 22-kHz ultrasonic vocalizations, has not been done before).

Therefore, we do not take lightly the surprising and interesting finding that “animals respond rather similarly to hearing 22 kHz calls as they do to hearing 44 kHz calls, with occasional shifts of 44 kHz call responses to an intermediate between appetitive and aversive calls”. We would rather put this description in analogous words: “the rats responded similarly to hearing 44-kHz calls as they did to hearing aversive 22-kHz calls, especially regarding heartrate change, despite the 44-kHz calls occupying the frequency band of appetitive 50-kHz vocalizations” and “other responses to 44-kHz calls were intermediate, they fell between response levels to appetitive vs. aversive playback” – which we added to the Discussion.

Finally, we acknowledge that our findings do not present a finite and complete picture of the discussed aspects of behavioral responses to the presented ultrasonic stimuli (44-kHz vocalizations). Therefore, we have incorporated the Reviewer’s suggestion in the discussion. The added sentence reads: “Overall, these initial results raise further questions about how, ethologically, animals may interpret the variation in hearing 22-kHz vs. 44-kHz calls and integrate this interpretation in their responses.”

In sum, rather than describing the 44kHz long calls as a new call type, it may be more accurate to say that sometimes aversive calls can occur at frequencies above 22 kHz. Individual and situational variability in vocalization parameters seems to be expected, much more so than all members of a species strictly adhering to extremely non-variable behavioral outputs.

Answer: The surprising fact that there are presumably aversive calls that are beyond the commonly applied thresholds, i.e. >32 kHz, while sharing some characteristics with 22-kHz calls, is the main finding of the current publication. Whether they be finally assigned as a new type, subtype, i.e. a separate category or become a supergroup of aversive calls with 22-kHz vocalizations is of secondary importance to be discussed with other researchers of the field of study.

However, we would argue – by showing a comparison – that 22-kHz calls occur at durations of <300 ms and also >300 ms, and are, usually, referred to in literature as short and long 22-kHz vocalizations, respectively (not introduced with a description that “sometimes 22kHz calls can occur at durations below 300 ms”). These are then regarded and investigated as separate groups or classes usually referred to as two different “types” (e.g., Barker et al., 2010) or “subtypes” (e.g., Brudzynski, 2015). Analogously, 44-kHz vocalizations can also be regarded as a separate type or a subtype of 22-kHz calls. The problem with the latter is that 22-kHz vocalizations are traditionally and predominantly defined by 18–32 kHz frequency bandwidth (Araya et al., 2020; Barroso et al., 2019; Browning et al., 2011; Brudzynski et al., 1993; Hinchcliffe et al., 2022; Willey & Spear, 2013).

Reviewer #2 (Public Review):

Olszyński et al. claim that they identified a "new-type" ultrasonic vocalization around 44 kHz that occurs in response to prolonged fear conditioning (using foot-shocks of relatively high intensity, i.e. 1 mA) in rats. Typically, negative 22-kHz calls and positive 50-kHz calls are distinguished in rats, commonly by using a frequency threshold of 30 or 32 kHz. Olszyński et al. now observed so-called "44-kHz" calls in a substantial number of subjects exposed to 10 tone-shock pairings, yet call emission rate was low (according to Fig. 1G around 15%, according to the result text around 7.5%).

Answer: We are thankful for praising the strengths. Please note Figure 1G referred to 10-trial Wistar rats during delay fear conditioning session in which 44-kHz constituted 14.1% of ultrasonic vocalizations. The 7.5% number in results refers to the total of vocalizations analyzed across all animal groups used in fear conditioning experiments. These values have been updated in the current version of the manuscript. Also, please note – 44-kHz calls constituted up to 19.4% of calls, on average, in one of the ITI during fear conditioning session. However, the prevalence of aversive calls and of 44-kHz vocalizations in particular varied. It varied between individual rats; we added the text: “for n = 3 rats, 44-kHz vocalizations accounted for >95% of all calls during at least one ITI (e.g., 140 of total 142, 222 of 231, and 263 of 265 tallied 44-kHz calls), and in n = 9 rats, 44-kHz vocalizations constituted >50% of calls in more than one ITI.” See also further for the description of the array of experiments analyzed and the prevalence/percentage of 44-kHz calls encountered (Tab. 1, Fig. 1S3).

Weaknesses: I see a number of major weaknesses.

While the descriptive approach applied is useful, the findings have only focused importance and scope, given the low prevalence of "44 kHz" calls and limited attempts made to systematically manipulate factors that lead to their emission. In fact, the data presented appear to be derived from reanalyses of previously conducted studies in most cases and the main claims are only partially supported. While reading the manuscript, I got the impression that the data presented here are linked to two or three previously published studies (Olszyński et al., 2020, 2021, 2023). This is important to emphasize for two reasons:

(1) It is often difficult (if not impossible) to link the reported data to the different experiments conducted before (and the individual experimental conditions therein). While reanalyzing previously collected data can lead to important insight, it is important to describe in a clear and transparent manner what data were obtained in what experiment (and more specifically, in what exact experimental condition) to allow appropriate interpretation of the data. For example, it is said that in the "trace fear conditioning experiment" both single- and grouphoused rats were included, yet I was not able to tell what data were obtained in single- versus group-housed rats. This may sound like a side aspect, however, in my view this is not a side aspect given the fact that ultrasonic vocalizations are used for communication and communication is affected by the social housing conditions.

Answer: Preparing the current manuscript, we indeed used data collected during fear conditioning experiments which were described previously (Olszyński et al., 2021; Olszyński et al., 2022). Please note, however, that vocalization behavior during the fear conditioning itself was not the main subject of these publications. Our previous publications (Olszyński et al., 2020; Olszyński et al., 2021; Olszyński et al., 2022) present primarily ultrasonic-vocalization data from playback-part of experiments whereas here we analyze recordings obtained during fear conditioning experiments, thus we are analyzing new parts, i.e., not yet analyzed, of previously published studies. Also, we have performed additional experiments.

In the first version of the current manuscript, we did not attempt to demonstrate exactly which calls were recorded in which conditions as the focus was to demonstrate that 44-kHz calls were emitted in several different fear-conditioning experiments. Also, as the experiments were not performed simultaneously and are results from different experimental situations, we would prefer to not compare these results directly.

However, in the current version of the manuscript, we have introduced an additional reference system, based on Tab. 1, to more clearly indicate which rats have been employed in each analysis, e.g. the group of “Wistar rats that undergone 10 trials of fear conditioning” are described as “Tab. 1/Exp. 1-3/#2,4,8,13; n = 46”, i.e., these are the rats listed in rows 2, 4, 8, and 13 of Tab. 1.

We have also tried to unify the analyses, in terms of rats used, as much as possible. Finally, we have also introduced Fig. 1S3 to demonstrate the prevalence of 44-kHz calls in all experiments analyzed with the note that “the experiments were not performed in parallel”.

Regarding the Reviewer’s concerns about analyzing single- and pair-housed rats together. We have examined ultrasonic vocalizations emitted and freezing behavior in these two groups.

• Ultrasonic vocalizations; when comparing the number of vocalizations, their duration, peak frequency and latency to first occurrence, equally for all types of calls and divided into types (short 22-kHz, long 22-kHz, 44-kHz, 50-kHz), the only difference was observed in peak frequency in 50-kHz vocalizations (50.7 ± 2.8 kHz for paired vs. 61.8 ± 3.1 kHz for single rats; p = 0.0280, Mann-Whitney). Since 50-kHz calls are not the subject of the current publication, we did not investigate this difference further. Also, this difference was not observed during playback experiments (Olszyński et al., 2020, Tab. 1).

• Freezing. There were no differences between single- and pair-housed groups in freezing behavior, both in the time before first shock presentation and during fear conditioning training (Mann-Whitney).

In summary, since the two groups did not differ in relevant ultrasonic features and freezing, we decided to present the results obtained from these rats together. However, we agree with the Reviewer, and it is possible that social housing conditions may in fact affect the emission of 44-kHz vocalizations, which could be a subject of another project – involving, e.g., larger experimental groups observed under hypothesis-oriented and defined conditions.

(2) In at least two of the previously published manuscripts (Olszyński et al., 2021, 2023), emission of ultrasonic vocalizations was analyzed (Figure S1 in Olszyński et al., 2021, and Fig. 1 in Olszyński et al., 2023). This includes detailed spectrographic analyses covering the frequency range between 20 and 100 kHz, i.e. including the frequency range, where the "newtype" ultrasonic vocalization, now named "44 kHz" call, occurs, as reflected in the examples provided in Fig. 1 of Olszyński et al. (2023). In the materials and methods there, it was said: "USV were assigned to one of three categories: 50-kHz (mean peak frequency, MPF >32 kHz), short 22-kHz (MPF of 18-32 kHz, <0.3 s duration), long 22-kHz (MPF of 18-32 kHz, >0.3 s duration)". Does that mean that the "44 kHz" calls were previously included in the count for 50-kHz calls? Or were 44 kHz calls (intentionally?) left out? What does that mean for the interpretation of the previously published data? What does that mean for the current data set? In my view, there is a lack of transparency here.

Answer: As mentioned above, we indeed used data collected during fear conditioning experiments which were described previously (Olszyński et al., 2021; Olszyński et al., 2022). However, in these publications, ultrasonic vocalizations emitted during playback experiments were the main subject, while the ultrasonic calls emitted during fear conditioning (performed before the playback) were only analyzed in a preliminary way. As a result, the 44-kHz vocalizations analyzed in the current manuscript were not included in the previous analyses. In particular, in Olszyński et al. (2021), we counted the overall number of ultrasonic vocalizations before fear conditioning session to determine the basal ultrasonic emissions (Fig. S1). Then, our next article (Olszyński et al., 2022), we analyzed again the number of all ultrasonic vocalizations before fear conditioning (Fig. S1) and restricted the analysis of vocalizations during fear conditioning to 22-kHz calls (Tab. S1 and S2).

Also, we re-reviewed all the data used in our previous playback publications. Overall, 44-kHz calls were extremely rare in playback parts of the experiments. There were no 44-kHz calls in the playback data used in Olszyński et al. (2022) and Olszyński et al. (2020). In Olszyński et al. (2021), one rat produced eight 44-kHz calls. These 44-kHz calls constituted 0.03% of all vocalizations analyzed in the experiment (8/24888) and were included in the total number of calls analyzed (but not in the 50-kHz group), they were not described in further detail in that publication.

Moreover, whether the newly identified call type is indeed novel is questionable, as also mentioned by the authors in their discussion section. While they wrote in the introduction that "high-pitch (>32 kHz), long and monotonous ultrasonic vocalizations have not yet been described", they wrote in the discussion that "long (or not that long (Biały et al., 2019)), frequency-stable high-pitch vocalizations have been reported before (e.g. Sales, 1979; Shimoju et al., 2020), notably as caused by intense cholinergic stimulation (Brudzynski and Bihari, 1990) or higher shock-dose fear conditioning (Wöhr et al., 2005)" (and I wish to add that to my knowledge this list provided by the authors is incomplete). Therefore, I believe, the strong claims made in abstract ("we are the first to describe a new-type..."), introduction ("have not yet been described"), and results ("new calls") are not justified.

Answer: We would argue that 44-kHz vocalizations were indeed reported but not described. As far as we are concerned, an in-depth analysis of the properties and experimental circumstance of emission of long, high-frequency calls has not yet been performed. These researchers have observed, at least to a degree, similar calls to the ones we observed – as we mentioned in the discussion section. However, since these reported 44-kHz vocalizations were not fully described, we can only guess that they may be similar to ours. We speculate that perhaps like us, these researchers unknowingly recorded 44-kHz calls in their experiments and may also be able to describe them more extensively when re-analyzing their data as we have done here.

Possibly, it was difficult to find reports on vocalizations, similar to the 44-kHz calls that we observed, because of the canonical and accepted definitions of ultrasonic vocalization types. Biały et al. (2019) allocated them as a part of 22-kHz group, perhaps because their calls were often of a step variation having both low and high components. Shimoju et al. (2020) grouped them along with 50-kHz vocalizations because they appeared during stroking rats held vertically; this procedure was compared to tickling which usually elicits appetitive calls.

The Reviewer #2 states there are other publications to complete the list. We are aware of other articles authored by the same team as Shimoju et al. (2020) with different first authors. However, they are reporting similar findings to the cited article. Otherwise, we would gladly cite a more complete list of publications showing atypical, long, monotonous highfrequency vocalizations, similar to those observed in our experiments. Therefore, we would argue that ultrasonic vocalizations which were long, flat, high in frequency, and repeatedly occurring in a defined behavioral situation, have not been reported before. However, concerning the strong claims of novelty of our finding, we toned them down where we found this was warranted.

In general, the manuscript is not well written/ not well organized, the description of the methods is insufficient, and it is often difficult (if not impossible) to link the reported data to the experiments/ experimental conditions described in the materials and methods section.

Answer: The description of the methods has been adjusted and expanded. We added the requested link to each particular experiment as a formula “Tab. 1/Exp. nos./# nos.” which shows, each time, which experiments and experimental groups were analyzed. The list of the experiments and groups is found in the Tab. 1.

For example, I miss a clear presentation of basic information: 1) How many rats emitted "44 kHz" calls (in total, per experiment, and importantly, also per experimental condition, i.e. single- versus group-housed)?

Answer: We now clearly show which experiments were performed and how many animals were tested in each condition (Tab. 1), while the prevalence of 44-kHz calls amongst experimental conditions and animal groups is shown in Fig. 1S3. Also, we included information regarding the number of animals and treatment of each group of rats when reporting results. For example, we are stating that:

(1a) “53 of all 84 conditioned Wistar rats (Tab. 1/Exp. 1-3/#2,4,6-8,13, Figs 1B, 1E, 1S1BC) displayed” 44-kHz vocalizations – as a general assessment; these numbers are different from those in the first version of the Ms, when we are mentioning Wistar rats conditioned 6 or 10 times only.

(1b) “From this group of rats (n = 46), n = 41 (89.1%) emitted long 22-kHz calls, and 32 of them (69.6%) emitted 44-kHz calls” – this time referring only to 10-times conditioned Wistar rats as the biggest group that could be analyzed together (Figs 1F, 1G, 1S2A).

(1c) “for n = 3 rats, 44-kHz vocalizations accounted for >95% of all calls during at least one ITI (e.g., 140 of total 142, 222 of 231, and 263 of 265 tallied 44-kHz calls), and in n = 9 rats, 44kHz vocalizations constituted >50% of calls in more than one ITI.”

(2) Out of the ones emitting "44 kHz" calls, what was the prevalence of "44 kHz" calls (relative to 22- and 50-kHz calls, e.g. shown as percentage)?

Answer: The prevalence of 44-kHz vocalizations in all investigated experiments and groups is shown in Fig. 1S3CD. Also, more information regarding the percentage of 44-kHz calls was demonstrated in Fig. 1S2AB where we calculated the distribution of 44-kHz calls to 22-kHz calls in Wistar rats, in 10-trial fear conditioning, across the length of the session.

Additionally, the values are listed in the sentence regarding all Wistar rats which underwent 10 trials of fear conditioning: “these vocalizations were less frequent following the first trial (1.2 ± 0.4% of all calls), and increased in subsequent trials, particularly after the 5th (8.8 ± 2.8%), through the 9th (19.4 ± 5.5%, the highest value), and the 10th (15.5 ± 4.9%) trials, where 44-kHz calls gradually replaced 22-kHz vocalizations in some rats (Fig. 1F, 1S2B, Video 1; comp Fig. 1D vs. 1E).”

(3) How did this ratio differ between experiments and experimental conditions?

Answer: The prevalence of 44-kHz vocalizations in all experimental conditions is shown in Fig. 1S3. However, the direct comparison of results obtained in different conditions was not the goal of the present work. Also, we would argue, that such direct comparisons of results of different experiments would not be allowed. These experiments were done with different groups of animals, at different times, with different timetables of experimental manipulations.

However, we are comfortable to state that:

• There were more 44-kHz vocalizations during fear conditioning training than testing in all fear-conditioned Wistar rats;

• We observed more 44-kHz vocalizations in Wistar rats compared to SHR.

(4) Was there a link to freezing? Freezing was apparently analyzed before (Olszyński et al., 2021, 2023) and it would be important to see whether there is a correlation between "44-kHz" calls and freezing. Moreover, it would be important to know what behavior the rats are displaying while such "44-kHz" calls are emitted? (Note: Even not all 22-kHz calls are synced to freezing.) All this could help to substantiate the currently highly speculative claims made in the discussion section ("frequency increases with an increase in arousal" and "it could be argued that our prolonged fear conditioning increased the arousal of the rats with no change in the valence of the aversive stimuli"). Such more detailed analyses are also important to rule out the possibility that the "new-type" ultrasonic vocalization, the so-called "44 kHz" call, is simply associated with movement/ thorax compression.

Answer: We analyzed freezing behavior and its association with ultrasonic emissions. The emission of 44-kHz vocalizations was associated with freezing. The results are now described and presented in the manuscript, i.e., Tab. 2, its legend and the description in Results: “Freezing during the bins of 22-kHz calls only (p < 0.0001, for both groups) and during 44-kHz calls only bins (p = 0.0003) was higher than during the first 5 min baseline freezing levels of the session. Also, the freezing associated with emissions of 44-kHz calls only was higher than during bins with no ultrasonic vocalizations (p = 0.0353), and it was also 9.9 percentage points higher than during time bins with only long 22-kHz vocalizations, but the difference was not significant (p = 0.1907; all Wilcoxon)” and “To further investigate this potential difference, we measured freezing during the emission of randomly selected single 44-kHz and 22-kHz vocalizations. The minimal freezing behavior detection window was reduced to compensate for the higher resolution of the measurements (3, 5, 10, or 15 video frames were used). There was no difference in freezing during the emission of 44-kHz vs. 22-kHz vocalizations for ≥150ms-long calls (3 frames, p = 0.2054) and for ≥500-ms-long calls (5 frames, p = 0.2404; 10 frames, p = 0.4498; 15 frames, p = 0.7776; all Wilcoxon, Tab. 2B).”

Please note, that the general observation that "frequency increases with an increase in arousal" is not our claim but a general rule derived from large body of observations and proposed by the others (Briefer et al., 2012); we changed the wording of this statement to: “frequency usually increases with an increase in arousal (Briefer et al., 2012)”.

The figures currently included are purely descriptive in most cases - and many of them are just examples of individual rats (e.g. majority of Fig. 1, all of Fig. 2 to my understanding, with the exception of the time course, which in case of D is only a subset of rats ("only rats that emitted 44-kHz calls in at least seven ITI are plotted" - is there any rationale for this criterion?)), or, in fact, just representative spectrograms of calls (all of Fig. 3, with the exception of G, all of Fig. 4).

Answer: Please note, the former figures 2, 4, 6, and 8 have been now moved to supplementary figures 1S1, 2S1, 3S1, and 4S1 – to better organize the presentation of data. Figures 1, 3, 5, 7 are now 1, 2, 3, 4 respectively. In regards to presenting data from individual rats, this was to show the general patterns of ultrasonic-calls distributions observed. Showing the full data set as seen in Fig. 5A (now Fig. 3A) would obscure the readability of the graph without using mathematical clustering techniques such as DBSCAN.

Concerning the Reviewer’s #2 question regarding the criterion of “minimum seven ITI”, we selected the highest vocalizers by taking animals above the 75th percentile of the number of ITI with 44-kHz calls. However, in the current version of the manuscript, we decided to omit this part of the analysis and the accompanying part of the figure, since it did not provide any additional informative value (apart from employing questionable criterion).

Moreover, the differences between Fig. 5 and Fig. 6 are not clear to me. It seems Fig. 5B is included three times - what is the benefit of including the same figure three times?

Answer: We hope that designating Fig. 6 as supplementary to Fig. 5 (now Figs 3S1 and 3, respectively) will make interpreting them more streamlined. Fig. 6A (now Fig. 3S1A) is a more detailed look on information presented in Fig. 5B (now Fig. 3B) with spectrogram images of ultrasonic vocalizations from different areas of the plot. Also, Fig. 3B (former Fig. 5B) was removed from Fig. 3S1B (former Fig. 6B).

A systematic comparison of experimental conditions is limited to Fig. 7 and Fig. 8, the figures depicting the playback results (which led to the conclusion that "the responses to 44-kHz aversive calls presented from the speaker were either similar to 22-kHz vocalizations or in between responses to 22-kHz and 50-kHz playbacks", although it remains unclear to me why differences were seen b e f o r e the experimental manipulation, i.e. the different playback types in Fig. 8B).

Answer: There were indeed instances of such before-differences. Such differences were observed in our previous studies (Olszyński et al., 2020, Tabs S9-12; Olszyński et al., 2021, Tabs S7; Olszyński et al., 2022, Tabs S4, S9, S13, S17, S18) and were most likely due to analyzing multiple comparisons. However, we think that the carry-over effect, mentioned by the Reviewer #2 (see below), also played a role.

Related to that, I miss a clear presentation of relevant methodological aspects: 1) Why were some rats single-housed but not the others?

Answer: As stated before, data were collected from our previous experiments and the observation of 44-kHz vocalizations in fear conditioning was an emergent discovery as we decided to analyze ultrasonic recordings from fear conditioning procedures. Single-housed animals were part of our experiment comparing fear conditioning and social situation on the perception of ultrasonic playback as described in Olszyński et al. (2020). Aside from this experiment, all other rats were housed in pairs.

(2) Is the experimental design of the playback study not confounded? It is said that "one group (n = 13) heard 50-kHz appetitive vocalization playback while the other (n = 16) 22-kHz and 44kHz aversive calls". How can one compare "44 kHz" calls to 22- and 50-kHz calls when "44 kHz" calls are presented together with 22-kHz calls but not 50-kHz calls? What about carry-over effects? Hearing one type of call most likely affects the response to the other type of call. It appears likely that rats are a bit more anxious after hearing aversive 22-kHz calls, for example. Therefore, it would not be very surprising to see that the response to "44 kHz" calls is more similar to 22-kHz calls than 50-kHz calls.

Of note, in case of the other playback experiment it is just said that rats "received appetitive and aversive ultrasonic vocalization playback" but it remains unclear whether "44 kHz" calls are seen as appetitive or aversive. Later it says that "rats were presented with two 10-s-long playback sets of either 22-kHz or 44-kHz calls, followed by one 50-kHz modulated call 10-s set and another two playback sets of either 44-kHz or 22-kHz calls not previously heard" (and wonder what data set was included in the figures and how - pooled?). Again, I am worried about carry-over effects here. This does not seem to be an experimental design that allows to compare the response to the three main call types in an unbiased manner.

Answer: We apologize for being confounding and brief in our original description of the playback experiments. We wanted to avoid confusion associated with including several additional playback signals (please note some are not related to the current comparisons and include different 50-kHz ultrasonic subtypes and two different subtypes of short 22-kHz calls). We lengthened the description of these playback experiments in the current version.

In general, including more than one type of ultrasonic calls as playback has a risk of a carry-over effect as well as a habituation effect (the responses become weak). However, it greatly reduces the number of required animals. Finally, regarding the first experiment, we chose 3 playbacks to compare the rats’ reactions, as this was the most conservative choice we thought of.

We would like to highlight that we wanted to compare specifically the rats’ responses to 22-kHz vs. 44-kHz playback (as well as the effects of playback of different subtypes 50-kHz calls, which is not the subject of the current work). Therefore, we would argue, that the design of both experiments is actually unbiased regarding this key comparison (responses to 22-kHz vs. 44-kHz playback). In both experiments, 22-kHz and 44-kHz playbacks were included in the same sequences of stimuli and counterbalanced regarding their order (i.e., taking into account possible carry-over effects), and presented to the same rats. We regarded the group of rats that heard 50-kHz recordings as a baseline/control, since we know from previous playback studies what reactions to expect from rats exposed to these vocalizations (and 22-kHz playback), while in the second experiment, we reduced the 50-kHz playback to one set in order to minimize possible habituation to multiple playbacks.

We agree that the design of both experiments does not allow for full comparison of the effects of aversive playbacks to 50-kHz playback. Also, we agree that some carry-over effects could play a role. It was mentioned in the discussion: ”Please factor in potential carryover effects (resulting from hearing playbacks of the same valence in a row) in the differences between responses to 50-kHz vs. 22/44-kHz playbacks, especially, those observed before the signal (Fig. 4AB).” However, we would still argue that the observed lack of difference in heartrate response (Fig. 4A) and the differences regarding the number of 50-kHz calls emitted (e.g., Fig. 4S1F) are void of the constraints raised by the Reviewer #2.

We acknowledge that our studies do not give a complete picture of 44-kHz ultrasonic perception in relation to other ultrasonic bands and, given the possibility, we would like to perform more in-depth and focused experiments to study this aspect of 44-kHz calls in the future.

Finally, regarding the second experiment, the description of the rats now includes that they “received 22-kHz, 44-kHz, and 50-kHz ultrasonic vocalization playback”, while the description of the experiment itself includes: “Responses to the pairs of playback sets were averaged”.

Of note, what exactly is meant by "control rats" in the context of fear conditioning is also not clear to me. One can think of many different controls in a fear conditioning experiment.

More concrete information is needed.

Answer: This information was included in our previous publications. However, it was now provided in the method section of the current version of the manuscript. In general, control rats were subjected to the same procedures but did not receive electric shocks.

Literature included in the answers

Araya, E. I., Baggio, D. F., Koren, L. O., Andreatini, R., Schwarting, R. K. W., Zamponi, G. W., & Chichorro, J. G. (2020). Acute orofacial pain leads to prolonged changes in behavioral and affective pain components. Pain, 161(12), 2830-2840. https://doi.org/10.1097/j.pain.0000000000001970

Barker, D. J., Root, D. H., Ma, S., Jha, S., Megehee, L., Pawlak, A. P., & West, M. O. (2010). Dose-dependent differences in short ultrasonic vocalizations emitted by rats during cocaine self-administration. Psychopharmacology (Berl), 211(4), 435-442. https://doi.org/10.1007/s00213-010-1913-9

Barroso, A. R., Araya, E. I., de Souza, C. P., Andreatini, R., & Chichorro, J. G. (2019). Characterization of rat ultrasonic vocalization in the orofacial formalin test: Influence of the social context. Eur Neuropsychopharmacol, 29(11), 1213-1226. https://doi.org/10.1016/j.euroneuro.2019.08.298

Biały, M., Podobinska, M., Barski, J., Bogacki-Rychlik, W., & Sajdel-Sulkowska, E. M. (2019). Distinct classes of low frequency ultrasonic vocalizations in rats during sexual interactions relate to different emotional states. Acta Neurobiol Exp (Wars), 79(1), 1-12. https://www.ncbi.nlm.nih.gov/pubmed/31038481

Briefer, E. F., Padilla de la Torre, M., & McElligott, A. G. (2012). Mother goats do not forget their kids' calls. Proc Biol Sci, 279(1743), 3749-3755. https://doi.org/10.1098/rspb.2012.0986

Browning, J. R., Browning, D. A., Maxwell, A. O., Dong, Y., Jansen, H. T., Panksepp, J., & Sorg, B. A. (2011). Positive affective vocalizations during cocaine and sucrose self administration: a model for spontaneous drug desire in rats. Neuropharmacology, 61(1-2), 268-275. https://doi.org/10.1016/j.neuropharm.2011.04.012

Brudzynski, S. M. (2015). Pharmacology of Ultrasonic Vocalizations in adult Rats: Significance, Call Classification and Neural Substrate. Curr Neuropharmacol, 13(2), 180-192. https://doi.org/10.2174/1570159x13999150210141444

Brudzynski, S. M., & Bihari, F. (1990). Ultrasonic vocalization in rats produced by cholinergic stimulation of the brain. Neurosci Lett, 109(1-2), 222-226. https://doi.org/10.1016/0304-3940(90)90567-s

Brudzynski, S. M., Bihari, F., Ociepa, D., & Fu, X. W. (1993). Analysis of 22 kHz ultrasonic vocalization in laboratory rats: long and short calls. Physiol Behav, 54(2), 215-221. https://doi.org/10.1016/0031-9384(93)90102-l

Hinchcliffe, J. K., Jackson, M. G., & Robinson, E. S. (2022). The use of ball pits and playpens in laboratory Lister Hooded male rats induces ultrasonic vocalisations indicating a more positive affective state and can reduce the welfare impacts of aversive procedures. Lab Anim, 56(4), 370-379. https://doi.org/10.1177/00236772211065920

Matochik, J. A., White, N. R., & Barfield, R. J. (1992). Variations in scent marking and ultrasonic vocalizations by Long-Evans rats across the estrous cycle. Physiol Behav, 51(4), 783-786. https://doi.org/10.1016/0031-9384(92)90116-j

Olszyński, K. H., Polowy, R., Małż, M., Boguszewski, P. M., & Filipkowski, R. K. (2020). Playback of Alarm and Appetitive Calls Differentially Impacts Vocal, Heart-Rate, and Motor Response in Rats. iScience, 23(10), 101577. https://doi.org/10.1016/j.isci.2020.101577

Olszyński, K. H., Polowy, R., Wardak, A. D., Grymanowska, A. W., & Filipkowski, R. K. (2021). Increased Vocalization of Rats in Response to Ultrasonic Playback as a Sign of Hypervigilance Following Fear Conditioning. Brain Sci, 11(8). https://doi.org/10.3390/brainsci11080970

Olszyński, K. H., Polowy, R., Wardak, A. D., Grymanowska, A. W., Zieliński, J., & Filipkowski, R. K. (2022). Spontaneously hypertensive rats manifest deficits in emotional response to 22-kHz and 50-kHz ultrasonic playback. Prog Neuropsychopharmacol Biol Psychiatry, 120, 110615. https://doi.org/10.1016/j.pnpbp.2022.110615

Saito, Y., Tachibana, R. O., & Okanoya, K. (2019). Acoustical cues for perception of emotional vocalizations in rats. Scientific Reports, 9(1), 10539.

Sales, G. D. (1979). Strain Differences in the Ultrasonic Behavior of Rats (Rattus norvegicus) Am Zool, 19(2), 513-527. https://www.jstor.org/stable/3882331

Shimoju, R., Shibata, H., Hori, M., & Kurosawa, M. (2020). Stroking stimulation of the skin elicits 50-kHz ultrasonic vocalizations in young adult rats. J Physiol Sci, 70(1), 41. https://doi.org/10.1186/s12576-020-00770-1

Silkstone, M., & Brudzynski, S. M. (2019a). The antagonistic relationship between aversive and appetitive emotional states in rats as studied by pharmacologically-induced ultrasonic vocalization from the nucleus accumbens and lateral septum. Pharmacology Biochemistry and Behavior, 181, 77-85. https://doi.org/10.1016/j.pbb.2019.04.009

Silkstone, M., & Brudzynski, S. M. (2019b). Intracerebral injection of R-(-)-Apomorphine into the nucleus accumbens decreased carbachol-induced 22-kHz ultrasonic vocalizations in rats. Behavioural Brain Research, 364, 264-273. https://doi.org/10.1016/j.bbr.2019.01.044

Willey, A. R., & Spear, L. P. (2013). The effects of pre-test social deprivation on a natural reward incentive test and concomitant 50 kHz ultrasonic vocalization production in adolescent and adult male Sprague-Dawley rats. Behav Brain Res, 245, 107-112. https://doi.org/10.1016/j.bbr.2013.02.020

Wöhr, M., Borta, A., & Schwarting, R. K. (2005). Overt behavior and ultrasonic vocalization in a fear conditioning paradigm: a dose-response study in the rat. Neurobiol Learn Mem, 84(3), 228-240. https://doi.org/10.1016/j.nlm.2005.07.004

Recommendations For The Authors:

Reviewer #1 (Recommendations For The Authors):

Additional considerations:

The discussion of the "perfect fifth" and the proposition that this observation could be evidence of an evolutionary mechanism underlying it is rather far-fetched, especially for being presented in the Results section (with no supporting non-anecdotal evidence).

Answer: We agree with the Reviewer #1. The text was modified, the word “evolutionary” was deleted. Instead, we expended on the possible reason for prevalence of the perfect fifth in the current version of the manuscript; we added that the prevalence of the perfect fifth: “could be explained by the observation that all physical objects capable of producing tonal sounds generate harmonic vibrations, the most prominent being the octave, perfect fifth, and major third (Christensen, 1993, discussed in Bowling and Purves, 2015).”

It is not clear why Sprague-Dawleys were used as "receivers" in the playback experiment, when presumably the calls were recorded from Wistars and SHRs. While this does not critically impact the conclusions, within the species rats should be able to respond appropriately to calls made by rats of different genetic backgrounds, it adds an unnecessary source of variance.

Answer: Sprague-Dawley rats were used to test another normotensive strain of rats. Regarding the Reviewer’s main point – we beg to differ as we think that it is worth testing playback stimuli in different strains. Diverging the stimuli between different rat strains would add unnecessary variance and it seemed logical to use the same recordings to test effects in different strains. Please note that finally, in spite of this additional variance, the results of both playback experiments are, in general, similar – which may point to a universal effect of 44-kHz playback across rat strains.

It is pertinent to note that for the trace fear conditioning experiment, the rats had previously been exposed to a vocalization playback experiment. While such a pre-exposure is unlikely to be a very strong stressor, the possibility for it to influence the vocal behaviors of these rats in later experiments cannot be ruled out. It is also not clear what the control rats in this experiment experienced (home cage only?), nor what they were used for in analyses.

Answer: In the current version of the manuscript, we have described in greater detail all the experiments performed and analyzed. We would like to emphasize that both delay and trace fear conditioning experiments with radiotelemetric transmitters were not performed specifically to elicit any particular response during fear conditioning, rather that our observation of 44-kHz vocalizations emerged as a result of re-examining the audio recordings. As a result, this work summarizes our observations of 44-kHz calls from several different experiments. It is relevant to note, that 44-kHz vocalizations were observed “in rats which were exposed to vocalization playback experiment”, in rats before the playback experiments as well as in naïve rats, without transmitters implemented, trained in fear conditioning (Tab. 1/Exp. 1-3).

Our main message is that 44-kHz vocalizations were present in several experiments, with different conditions and subjects, while we are not attempting to compare in detail the results across the different experiments. In other words, we agree that pre-exposure to playback (and even more likely – transmitters implantation) could influence, but are not necessary, for 44-kHz ultrasonic emissions by the rats. To demonstrate this, we added a prolonged fear conditioning group with naïve Wistar rats (Exp. 3) to verify the emission of 44kHz calls in the absence of those experimental factors.

We modified the methods section to clarify the circumstances under which these discoveries were made, such as including the information regarding the control rats in trace fear conditioning. In particular we mention that: “Control rats were subjected to the exact same procedures but did not receive the electric shock at the end of trace periods”.

For Figure 1A-E, only example call distributions from individual rats are shown. It would perhaps be more informative to see the full data set displayed in this manner, with color/shape codes distinguishing individuals if desired.

Answer: Please note the Fig. 1S1 shows more examples of ultrasonic call distribution. Showing all the data would make it more difficult to read and interpret. The problem is partly amended in Fig. 3A.

It is not clear what is presented in Figure 2D vs. E, i.e. panel D is shown only for "selected rats" but the legend does not clarify how and why these rats were selected. It is also not clear why the legend reports p-values for both Friedman and Wilcoxon tests; the latter is appropriate for paired data which seems to be the case when the question is whether the call peak frequency alters across time, but the Friedman assumes non-paired input data.

Answer: The question refers to the current Fig. 1S2C panel (former Fig. 2E panel) and the former Fig. 2D panel. The latter was not included in the current version of the manuscript, since both reviewers opposed the presentation of “selected rats” only (see above). The full description of the Fig. 1S2C panel is now in the results section together with p-values for Friedman and Wilcoxon test. We used the latter to investigate the difference between the first and the last ITI (selected paired data), while the Friedman to investigate the presence of change within the chain of ten ITI – since it is a suitable test for a difference between two or more paired samples.

Reviewer #2 (Recommendations For The Authors):

The weaknesses listed in the public review need to be addressed.

Answer: We have done our best to address the weaknesses.

Notes: 1) Page and line numbers would have been useful.

Answer: We are including a separate manuscript version with page and line numbers.

.(2) English language needs to be improved.

Answer: The text has been checked by two native English speakers (one with a scientific background). Both only identified minor changes to improve the text which we applied.

(3) I am a bit unsure whether the comment about the Star Wars movie (1997) and the Game of Thrones series (2011) is supposed to be a joke.

Answer: These are indeed two genuine examples of the perfect fifth in human music that we hope are easily recognizable and familiar to readers. Parts of the same examples of the perfect fifth can also heard in the rat voice files provided.

Author response:

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

Thank you very much for the careful and positive reviews of our manuscript. We have addressed each comment in the attached revised manuscript. We describe the modifications below. To avoid confusion, we've changed supplementary figure and table captions to start with "Supplement Figure" and "Supplementary Table," instead of "Figure" and "Table."

We have modified/added:

● Supplementary Table S1: AUC scores for the top 10 frequent epitope types (pathogens) in the testing set of epitope split.

● Supplementary Table S5: AUCs of TCR-epitope binding affinity prediction models with BLOSUM62 to embed epitope sequences.

● Supplementary Table S6: AUCs of TCR-epitope binding affinity prediction models trained on catELMo TCR embeddings and random-initialized epitope embeddings.

● Supplementary Table S7: AUCs of TCR-epitope binding affinity prediction models trained on catELMo and BLOSUM62 embeddings.

● Supplementary Figure 4: TCR clustering performance for the top 34 abundant epitopes representing 70.55% of TCRs in our collected databases.

● Section Discussion.

● Section 4.1 Data: TCR-epitope pairs for binding affinity prediction.

● Section 4.4.2 Epitope-specific TCR clustering.

Public Reviews:

Reviewer #1 (Public Review):

In this manuscript, the authors described a computational method catELMo for embedding TCR CDR3 sequences into numeric vectors using a deep-learning-based approach, ELMo. The authors applied catELMo to two applications: supervised TCR-epitope binding affinity prediction and unsupervised epitope-specific TCR clustering. In both applications, the authors showed that catELMo generated significantly better binding prediction and clustering performance than other established TCR embedding methods. However, there are a few major concerns that need to be addressed.

(1) There are other TCR CDR3 embedding methods in addition to TCRBert. The authors may consider incorporating a few more methods in the evaluation, such as TESSA (PMCID: PMC7799492), DeepTCR (PMCID: PMC7952906) and the embedding method in ATM-TCR (reference 10 in the manuscript). TESSA is also the embedding method in pMTnet, which is another TCR-epitope binding prediction method and is the reference 12 mentioned in this manuscript.

TESSA is designed for characterizing TCR repertoires, so we initially excluded it from the comparison. Our focus was on models developed specifically for amino acid embedding rather than TCR repertoire characterization. However, to address the reviewer's inquiry, we conducted further evaluations. Since both TESSA and DeepTCR used autoencoder-based models to embed TCR sequences, we selected one used in TESSA for evaluation in our downstream prediction task, conducting ten trials in total. It achieved an average AUC of 75.69 in TCR split and 73.3 in epitope split. Notably, catELMo significantly outperformed such performance with an AUC of 96.04 in TCR split and 94.10 in epitope split.

Regarding the embedding method in ATM-TCR, it simply uses BLOSUM as an embedding matrix which we have already compared in Section 2.1. Furthermore, we have provided the comparison results between our prediction model trained on catELMo embeddings with the state-of-the-art prediction models such as netTCR and ATM-TCR in Table 6 of the Discussion section.

(2) The TCR training data for catELMo is obtained from ImmunoSEQ platform, including SARS-CoV2, EBV, CMV, and other disease samples. Meanwhile, antigens related to these diseases and their associated TCRs are extensively annotated in databases VDJdb, IEDB and McPAS-TCR. The authors then utilized the curated TCR-epitope pairs from these databases to conduct the evaluations for eptitope binding prediction and TCR clustering. Therefore, the training data for TCR embedding may already be implicitly tuned for better representations of the TCRs used in the evaluations. This seems to be true based on Table 4, as BERT-Base-TCR outperformed TCRBert. Could catELMo be trained on PIRD as TCRBert to demonstrate catELMo's embedding for TCRs targeting unseen diseases/epitopes?

We would like to note that catELMo was trained exclusively on TCR sequences in an unsupervised manner, which means it has never been exposed to antigen information. We also ensured that the TCRs used in catELMo's training did not overlap with our downstream prediction data. Please refer to the section 4.1 Data where we explicitly stated, “We note that it includes no identical TCR sequences with the TCRs used for training the embedding models.”. Moreover, the performance gap (~1%) between BERT-Base-TCR and TCRBert, as observed in Table 4, is relatively small, especially when compared to the performance difference (>16%) between catELMo and TCRBert.

To further address this concern, we conducted experiments using the same number of TCRs, 4,173,895 in total, sourced exclusively from healthy ImmunoSeq repertoires. This alternative catELMo model demonstrated a similar prediction performance (based on 10 trials) to the one reported in our paper, with an average AUC of 96.35% in TCR split and an average AUC of 94.03% in epitope split.

We opted not to train catELMo on the PIRD dataset for several reasons. First, approximately 7.8% of the sequences in PIRD also appear in our downstream prediction data, which could be a potential source of bias. Furthermore, PIRD encompasses sequences related to diseases such as Tuberculosis, HIV, CMV, among others, which the reviewer is concerned about.

(3) In the application of TCR-epitope binding prediction, the authors mentioned that the model for embedding epitope sequences was catElMo, but how about for other methods, such as TCRBert? Do the other methods also use catELMo-embedded epitope sequences as part of the binding prediction model, or use their own model to embed the epitope sequences? Since the manuscript focuses on TCR embedding, it would be nice for other methods to be evaluated on the same epitope embedding (maybe adjusted to the same embedded vector length).

Furthermore, the authors found that catELMo requires less training data to achieve better performance. So one would think the other methods could not learn a reasonable epitope embedding with limited epitope data, and catELMo's better performance in binding prediction is mainly due to better epitope representation.

Review 1 and 3 have raised similar concerns regarding the epitope embedding approach employed in our binding affinity prediction models. We address both comments together on page 6 where we discuss the epitope embedding strategies in detail.

(4) In the epitope binding prediction evaluation, the authors generated the test data using TCR-epitope pairs from VDJdb, IEDB, McPAS, which may be dominated by epitopes from CMV. Could the authors show accuracy categorized by epitope types, i.e. the accuracy for TCR-CMV pair and accuracy for TCR-SARs-CoV2 separately?

The categorized AUC scores have been added in Supplementary Table 7. We observed significant performance boosts from catELMo compared with other embedding models.

(5) In the unsupervised TCR clustering evaluation, since GIANA and TCRdist direct outputs the clustering result, so they should not be affected by hierarchical clusters. Why did the curves of GIANA and TCRdist change in Figure 4 when relaxing the hierarchical clustering threshold?

For fair comparisons, we performed GIANA and TCRdist with hierarchical clustering instead of the nearest neighbor search. We have clarified it in the revised manuscript as follows.

“Both methods are developed on the BLOSUM62 matrix and apply nearest neighbor search to cluster TCR sequences. GIANA used the CDR3 of TCRβ chain and V gene, while TCRdist predominantly experimented with CDR1, CDR2, and CDR3 from both TCRα and TCRβ chains. For fair comparisons, we perform GIANA and TCRdist only on CDR3 β chains and with hierarchical clustering instead of the nearest neighbor search.”

(6 & 7) In the unsupervised TCR clustering evaluation, the authors examined the TCR related to the top eight epitopes. However, there are much more epitopes curated in VDJdb, IEDB and McPAS-TCR. In real application, the potential epitopes is also more complex than just eight epitopes. Could the authors evaluate the clustering result using all the TCR data from the databases? In addition to NMI, it is important to know how specific each TCR cluster is. Could the authors add the fraction of pure clusters in the results? Pure cluster means all the TCRs in the cluster are binding to the same epitope, and is a metric used in the method GIANA.

We would like to note that there is a significant disparity in TCR binding frequencies across different epitopes in current databases. For instance, the most abundant epitope (KLGGALQAK) has approximately 13k TCRs binding to it, while 836 out of 982 epitopes are associated with fewer than 100 TCRs in our dataset. Furthermore, there are 9347 TCRs having the ability to bind multiple epitopes. In order to robustly evaluate the clustering performance, we originally selected the top eight frequent epitopes from McPAS and removed TCRs binding multiple epitopes to create a more balanced dataset.

We acknowledge that the real-world scenario is more complex than just eight epitopes. Therefore, we conducted clustering experiments using the top most abundant epitopes whose combined cognate TCRs make up at least 70% of TCRs across three databases (34 epitopes). This is illustrated in Supplementary Figure 5. Furthermore, we extended our analysis by clustering all TCRs after filtering out those that bind to multiple epitopes, resulting in 782 unique epitopes. We found that catELMo achieved the 3rd and 2nd best performance in NMI and Purity, respectively (see Table below). These are aligned with our previous observations of the eight epitopes.

Author response table 1.

Reviewer #2 (Public Review):

In the manuscript, the authors highlighted the importance of T-cell receptor (TCR) analysis and the lack of amino acid embedding methods specific to this domain. The authors proposed a novel bi-directional context-aware amino acid embedding method, catELMo, adapted from ELMo (Embeddings from Language Models), specifically designed for TCR analysis. The model is trained on TCR sequences from seven projects in the ImmunoSEQ database, instead of the generic protein sequences. They assessed the effectiveness of the proposed method in both TCR-epitope binding affinity prediction, a supervised task, and the unsupervised TCR clustering task. The results demonstrate significant performance improvements compared to existing embedding models. The authors also aimed to provide and discuss their observations on embedding model design for TCR analysis: 1) Models specifically trained on TCR sequences have better performance than models trained on general protein sequences for the TCR-related tasks; and 2) The proposed ELMo-based method outperforms TCR embedding models with BERT-based architecture. The authors also provided a comprehensive introduction and investigation of existing amino acid embedding methods. Overall, the paper is well-written and well-organized.

The work has originality and has potential prospects for immune response analysis and immunotherapy exploration. TCR-epitope pair binding plays a significant role in T cell regulation. Accurate prediction and analysis of TCR sequences are crucial for comprehending the biological foundations of binding mechanisms and advancing immunotherapy approaches. The proposed embedding method presents an efficient context-aware mathematical representation for TCR sequences, enabling the capture and analysis of their structural and functional characteristics. This method serves as a valuable tool for various downstream analyses and is essential for a wide range of applications. Thank you.

Reviewer #3 (Public Review):

Here, the authors trained catElMo, a new context-aware embedding model for TCRβ CDR3 amino acid sequences for TCR-epitope specificity and clustering tasks. This method benchmarked existing work in protein and TCR language models and investigated the role that model architecture plays in the prediction performance. The major strength of this paper is comprehensively evaluating common model architectures used, which is useful for practitioners in the field. However, some key details were missing to assess whether the benchmarking study is a fair comparison between different architectures. Major comments are as follows:

• It is not clear why epitope sequences were also embedded using catELMo for the binding prediction task. Because catELMO is trained on TCRβ CDR3 sequences, it's not clear what benefit would come from this embedding. Were the other embedding models under comparison also applied to both the TCR and epitope sequences? It may be a fairer comparison if a single method is used to encode epitope sequence for all models under comparison, so that the performance reflects the quality of the TCR embedding only.

In our study, we indeed used the same embedding model for both TCRs and epitopes in each prediction model, ensuring a consistent approach throughout.

Recognizing the importance of evaluating the impact of epitope embeddings, we conducted experiments in which we used BLOSUM62 matrix to embed epitope sequences for all models. The results (Supplementary Table 5) are well aligned with the performance reported in our paper. This suggests that epitope embedding may not play as critical a role as TCR embedding in the prediction tasks. To further validate this point, we conducted two additional experiments.

Firstly, we used catELMo to embed TCRs while employing randomly initialized embedding matrices with trainable parameters for epitope sequences. It yielded similar prediction performance as when catELMo was used for both TCR and epitope embedding (Supplementary Table 6). Secondly, we utilized BLOSUM62 to embed TCRs but employed catELMo for epitope sequence embedding, resulting in performance comparable to using BLOSUM62 for both TCRs and epitopes (Supplementary Table 4). These experiment results confirmed the limited impact of epitope embedding on downstream performance.

We conjecture that these results may be attributed to the significant disparity in data scale between TCRs (~290k) and epitopes (less than 1k). Moreover, TCRs tend to exhibit high similarity, whereas epitopes display greater distinctiveness from one another. These features of TCRs require robust embeddings to facilitate effective separation and improve downstream performance, while epitope embedding primarily serves as a categorical encoding.

We have included a detailed discussion of these findings in the revised manuscript to provide a comprehensive understanding of the role of epitope embeddings in TCR binding prediction.

• The tSNE visualization in Figure 3 is helpful. It makes sense that the last hidden layer features separate well by binding labels for the better performing models. However, it would be useful to know if positive and negative TCRs for each epitope group also separate well in the original TCR embedding space. In other words, how much separation between these groups is due to the neural network vs just the embedding?

It is important to note that we used the same downstream prediction model, a simple three-linear-layer network, for all the discussed embedding methods. We believe that the separation observed in the t-SNE visualization effectively reflects the ability of our embedding model. Also, we would like to mention that it can be hard to see a clear distinction between positive and negative TCRs in the original embedding space because embedding models were not trained on positive/negative labels. Please refer to the t-SNE of the original TCR embeddings below.

Author response image 1.

• To generate negative samples, the author randomly paired TCRs from healthy subjects to different epitopes. This could produce issues with false negatives if the epitopes used are common. Is there an estimate for how frequently there might be false negatives for those commonly occurring epitopes that most populations might also have been exposed to? Could there be a potential batch effect for the negative sampled TCR that confounds with the performance evaluation?

Thank you for bringing this valid and interesting point up. Generating negative samples is non-trivial since only a limited number of non-binding TCR-pairs are publicly available and experimentally validating non-binding pairs is costly [1]. Standard practices for generating negative pairs are (1) paring epitopes with healthy TCRs [2, 3], and (2) randomly shuffling existing TCR-epitope pairs [4,5]. We used both approaches (the former included in the main results, and the latter in the discussion). In both scenarios, catELMo embeddings consistently demonstrated superior performance.

We acknowledge the possibility of false negatives due to the finite-sized TCR database from which we randomly selected TCRs, however, we believe that the likelihood of such occurrences is low. Given the vast diversity of human TCR clonotypes, which can exceed 10^15[6], the chance of randomly selecting a TCR that specifically recognizes a target epitope is relatively small.

In order to investigate the batch effect, we generated new negative pairs using different seeds and observed consistent prediction performance across these variations. However, we agree that there could still be a potential batch effect for the negative samples due to potential data bias.

We have discussed the limitation of generative negative samples in the revised manuscript.

• Most of the models being compared were trained on general proteins rather than TCR sequences. This makes their comparison to catELMO questionable since it's not clear if the improvement is due to the training data or architecture. The authors partially addressed this with BERT-based models in section 2.4. This concern would be more fully addressed if the authors also trained the Doc2vec model (Yang et al, Figure 2) on TCR sequences as baseline models instead of using the original models trained on general protein sequences. This would make clear the strength of context-aware embeddings if the performance is worse than catElmo and BERT.

We agree it is important to distinguish between the effects of training data and architecture on model performance.

In Section 2.4, as the reviewer mentioned, we compared catELMo with BERT-based models trained on the same TCR repertoire data, demonstrating that architecture plays a significant role in improving performance. Furthermore, in Section 2.5, we compared catELMo-shallow with SeqVec, which share the same architecture but were trained on different data, highlighting the importance of data on the model performance.

To further address the reviewer's concern, we trained a Doc2Vec model on the TCR sequences that have been used for catELMo training. We observed significantly lower prediction performance compared to catELMo, with an average AUC of 50.24% in TCR split and an average AUC of 51.02% in epitope split, making the strength of context-aware embeddings clear.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) It is known that TRB CDR3, the CDR1, CDR2 on TRBV gene and the TCR alpha chain also contribute to epitope recognition, but were not modeled in catELMo. It would be nice for the authors to add this as a current limitation for catELMo in the Discussion section.

We have discussed the limitation in the revised manuscript.

“Our study focuses on modeling the TCRβ chain CDR3 region, which is known as the primary determinant of epitope binding. Other regions, such as CDR1 and CDR2 on the TRB V gene, along with the TCRα chain, may also contribute to specificity in antigen recognition. However, a limited number of available samples for those additional features can be a challenge for training embedding models. Future work may explore strategies to incorporate these regions while mitigating the challenges of working with limited samples.”

(2) I tried to follow the instructions to train a binding affinity prediction model for TCR-epitope pairs, however, the cachetools=5.3.0 seems could not be found when running "pip install -r requirements.txt" in the conda environment bap. Is this cachetools version supported after Python 3.7 so the Python 3.6.13 suggested on the GitHub repo might not work?

This has been fixed. We have updated the README.md on our github page.

Reviewer #2 (Recommendations For The Authors):

The article is well-constructed and well-written, and the analysis is comprehensive.

The comments for minor issues that I have are as follows:

(1) In the Methods section, it will be clearer if the authors interpret more on how the standard deviation is calculated in all tables. How to define the '10 trials'? Are they based on different random training and test set splits?

‘10 trials' refers to the process of splitting the dataset into training, validation, and testing sets using different seeds for each trial. Different trials have different training, validation, and testing sets. For each trial, we trained a prediction model on its training set and measured performance on its testing set. The standard deviation was calculated from the 10 measurements, estimating model performance variation across different random splits of the data.

(2) The format of AUCs and the improvement of AUCs need to be consistent, i.e., with the percent sign.

We have updated the format of AUCs.

Reviewer #3 (Recommendations For The Authors):

In addition to the recommendations in the public review, we had the following more minor questions and recommendations:

• Could you provide some more background on the data, such as overlaps between the databases, and how the training and validation split was performed between the three databases? Also summary statistics on the length of TCR and epitope sequence data would be helpful.

We have provided more details about data in our revision.

• Could you comment on the runtime to train and embed using the catELMo and BERT models?

Our training data is TCR sequences with relatively short lengths (averaging less than 20 amino acid residues). Such characteristic significantly reduces the computational resources required compared to training large-scale language models on extensive text corpora. Leveraging standard machines equipped with two GeForce RTX 2080 GPUs, we were able to complete the training tasks within a matter of days. After training, embedding one sequence can be accomplished in a matter of seconds.

• Typos and wording:

• Table 1 first row of "source": "immunoSEQ" instead of "immuneSEQ"

This has been corrected.

• L23 of abstract "negates the need of complex deep neural network architecture" is a little confusing because ELMo itself is a deep neural network architecture. Perhaps be more specific and add that the need is for downstream tasks.

We have made it more specific in our abstract.

“...negates the need for complex deep neural network architecture in downstream tasks.”

References

(1) Montemurro, Alessandro, et al. "NetTCR-2.0 enables accurate prediction of TCR-peptide binding by using paired TCRα and β sequence data." Communications biology 4.1 (2021): 1060.

(2) Jurtz, Vanessa Isabell, et al. "NetTCR: sequence-based prediction of TCR binding to peptide-MHC complexes using convolutional neural networks." BioRxiv (2018): 433706.

(3) Gielis, Sofie, et al. "Detection of enriched T cell epitope specificity in full T cell receptor sequence repertoires." Frontiers in immunology 10 (2019): 2820.

(4) Cai, Michael, et al. "ATM-TCR: TCR-epitope binding affinity prediction using a multi-head self-attention model." Frontiers in Immunology 13 (2022): 893247.

(5) Weber, Anna, et al. "TITAN: T-cell receptor specificity prediction with bimodal attention networks." Bioinformatics 37 (2021): i237-i244.

(6) Lythe, Grant, et al. "How many TCR clonotypes does a body maintain?." Journal of theoretical biology 389 (2016): 214-224.

Author response:

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

We thank the reviewers for their thoughtful comments. We were pleased that they thought our study was "well crafted and written", "important", and that it provides a "valuable resource for researchers studying color vision". They also expressed several constructive criticisms, concerning – among other things – the lack of details regarding experimental procedures and analysis, the challenge in relating retinal data to cortical recordings, and consistency of results across animals. In response to the reviewers’ comments and following their suggestions, we performed additional analyses, and substantially revised the paper:

We added a section in the Discussion about "Limitations of the stimulus paradigm". In addition, we added a new Suppl. Figure that illustrates the effect of deconvolution of calcium traces on our results and clarified in the text why we use deconvolved signals for all analyses. The new Suppl. Figure also shows an additional analysis with a more conservative threshold of neuron exclusion.

We now clarify how retinal signaling relates to our cortical results and rewrote the text to be more conservative regarding our conclusions.

In addition, we added a new Suppl. Figure showing the key analyses from Figures 2 and 4 separately for each animal. We now mention consistency across animals in the Results section and clearly state which analyses were performed an data pooled across animals.

We are positive that these additions address the issues raised by the reviewers. Please find our point-by-point replies to all comments below.

eLife assessment

Franke et al. explore and characterize the color response properties in the mouse primary visual cortex, revealing specific color opponent encoding strategies across the visual field. The data is solid; however, the evidence supporting some conclusions and details about some procedures are incomplete. In its current form, the paper makes a useful contribution to how color is coded in mouse V1. Significance would be enhanced with some additional analyses and resolution of some technical issues.

We thank the reviewers for appreciating our manuscript and their thoughtful comments.

Referee 1 (Remarks to the Author):

Summary:

In this study, Franke et al. explore and characterize the color response properties across the primary visual cortex, revealing specific color opponent encoding strategies across the visual field. The authors use awake-behaving 2P imaging to define the spectral response properties of visual interneurons in Layer 2/3. They find that opponent responses are more prominent at photopic light levels, and diversity in color opponent responses exists across the visual science, with green ON/ UV OFF responses being stronger represented in the upper visual field. This is argued to be relevant for detecting certain features that are more salient when the chromatic space is used, possibly due to noise reductions.

Strengths:

The work is well crafted and written and provides a thorough characterization that reveals an uncharacterized diversity of visual properties in V1. I find this characterization important because it reveals how strongly chromatic information can modulate the response properties in V1. In the upper visual field, 25% of the cells differentially relay chromatic information, and one may wonder how this information will be integrated and subsequently used to aid vision beyond the detection of color per see. I personally like the last paragraph of the discussion that highlights this fact.

We thank the reviewer for appreciating our manuscript.

Weaknesses: One major point highlighted in this paper is the fact that Green ON/UV OFF responses are not generated in the retina. But glancing through the literature, I saw this is not necessarily true. Fig 1. of Joesch and Meister, a paper cited, shows this can be the case. Thus, I would not emphasize that this wasn’t present in the retina. This is a minor point, but even if the retina could not generate these signals, I would be surprised if the diversity of responses would only arise through feed-forward excitation, given the intricacies of cortical connectivity. Thus, I would argue that the argument holds for most of the responses seen in V1; they need to be further processed by cortical circuitries.

We thank the reviewer for this comment. When analyzing available data from the retina using a similar center-surround color flicker stimulus (Szatko et al. 2020), we found that Green On/UV Off color opponency is very rare in the RF center of retinal ganglion cells (Suppl. Fig. 5). This suggests that center Green On/UV Off color opponency in V1 neurons is not inherited by the RF center of retinal neurons. However, we agree with the reviewer that retinal neurons might still contribute to V1 color opponency, for example by being center-surround color opponent (e.g. Joesch et al. 2016 and Szatko et al. 2020). We rephrased the text to acknowledge this fact.

This takes me to my second point, defining center and surround. The center spot is 37.5 deg of visual angle, more than 1 mm of the retinal surface. That means that all retinal cells, at least half and most likely all of their surrounds will also be activated. Although 37.5 deg is roughly the receptive field size previously determined for V1 neurons, the one-to-one comparison with retinal recording, particularly with their center/surround properties, is difficult. This should be discussed. I assume that the authors tried a similar approach with sparse or dense checker white noise stimuli. If so, it would be interesting if there were better ways of defining the properties of V1 neurons on their complex/simple receptive field properties to define how much of their responses are due to an activation of the true "center" or a coactivation of the surround. Interestingly, at least some of the cells (Fig. 1d, cells 2 and 5) don’t have a surround. Could it be that in these cases, the "center" and "surround" are being excited together? How different would the overall statistics change if one used a full-filed flicker stimulus instead of a center/surround stimulus? How stable are the results if the center/surround flicker stimulus is shifted? These results won’t change the fact that chromatic coding is present in the VC and that there are clear differences depending on their position, but it might change the interpretation. Thus, I would encourage you to test these differences and discuss them.

Thanks for this comment. We agree with the reviewer that a one-to-one comparison of retina and V1 data is challenging, due to differences in both RF and stimulus size. We rephrased the Results text to clarify this point and now also mention it in the Discussion.

To be able to record from many V1 neurons simultaneously, we used a stimulus size of 37.5 degree visual angle in diameter, which is slightly larger than center RFs of single V1 neurons. As the reviewer mentions, the disadvantage of this approach is that the stimulus is only roughly centered on the neurons’ center RFs. To reduce the impact of potential stimulus misalignment on our results, we used the following steps:

For each recording, we positioned the monitor such that the mean RF across all neurons lies within the center of the stimulus field of view.

We confirmed that this procedure results in good stimulus alignment for the large majority of recorded neurons within individual recording fields by using a sparse noise stimulus (Suppl. Fig. 1a-c). Specifically, we found that for 83% of tested neurons, more than two thirds of their center RF, determined by the sparse noise stimulus, overlapped with the center spot of the color noise stimulus.

For analysis, we excluded neurons without a significant center STA, which may be caused by misalignment of the stimulus.

Together, we believe these points strongly suggest that the center spot and the surround annulus of the noise stimulus predominantly drive center (i.e. classical RF) and surround (i.e. extraclassical RF), respectively, of the recorded V1 neurons. This is further supported by the fact that color response types identified using an automated clustering method were robust across mice (Suppl. Fig. 6c), indicating consistent stimulus centering.

Nevertheless, we cannot exclude that the stimulus was misaligned for a subset of the recorded neurons used for analysis. We agree with the reviewer that such misalignment might have contributed to cells not having surround STAs, due to simultaneous activation of antagonistic center and surround RF components by the surround stimulus. While a full-field stimulus would get rid of the misalignment problem, it would not allow to study color tuning in center and surround RF components separately. Instead, one could compare the results of our approach with an approach that centers the stimulus on individual neurons. However, we believe that performing these additional experiments is out of the scope of the current study.

To acknowledge the experimental limitations of our study and the concerns brought up by the reviewer, we now explicitly mention the steps we perform to reduce the effects of stimulus misalignment in the Results section and discuss the problem of stimulus alignment in the Discussion. We believe these changes will help the reader to interpret our results.

Referee 2 (Remarks to the Author):

Summary:

Franke et al. characterize the representation of color in the primary visual cortex of mice and how it changes across the visual field, with a particular focus on how this may influence the ability to detect aerial predators. Using calcium imaging in awake, head-fixed mice, they characterize the properties of V1 neurons (layer 2/3) using a large center-surround stimulation where green and ultra-violet were presented in random combinations. Using a clustering approach, a set of functional cell-types were identified based on their preference to different combinations of green and UV in their center and surround. These functional types were demonstrated to have varying spatial distributions in V1, including one neuronal type (Green-ON/UV-OFF) that was much more prominent in the posterior V1 (i.e. upper visual field). Modelling work suggests that these neurons likely support the detection of predator-like objects in the sky.

Strengths:

The large-scale single-cell resolution imaging used in this work allows the authors to map the responses of individual neurons across large regions of the visual cortex. Combining this large dataset with clustering analysis enabled the authors to group V1 neurons into distinct functional cell types and demonstrate their relative distribution in the upper and lower visual fields. Modelling work demonstrated the different capacity of each functional type to detect objects in the sky, providing insight into the ethological relevance of color opponent neurons in V1.

We thank the reviewer for appreciating our manuscript.

Weaknesses:

While the study presents solid evidence a few weaknesses exist, including the size of the dataset, clarity regarding details of data included in each step of the analysis and discussion of caveats of the work. The results presented here are based on recordings of 3 mice. While the number of neurons recorded is reasonably large (n > 3000) an analysis that tests for consistency across animals is missing. Related to this, it is unclear how many neurons at each stage of the analysis come from the 3 different mice (except for Suppl. Fig 4).

Thank you for this comment. We apologize that the original manuscript did not clearly indicate the consistency of our results across animals. We have revised the manuscript in the following ways:

We have added an additional Suppl. Figure, which shows the variability of the data within and across animals (Suppl. Fig. 4). Specifically, we show the distribution of color and luminance selectivity for (i) center and surround components of V1 RFs and (ii) for upper and lower visual field. This data is used for all analyses shown in Figures 2-4. The figure legend of this figure also states the number of neurons per animal.

We now clearly state in the Results section that all analyses in the main figures were performed by pooling data across animals, and refer to the Suppl. Figures for consistency across animals.

We believe these changes help the reader to interpret our results.

Finally, the paper would greatly benefit from a more in depth discussion of the caveats related to the conclusion drawn at each stage of the analysis. This is particularly relevant regarding the caveats related to using spike triggered averages to assess the response preferences of ON-OFF neurons, and the conclusions drawn about the contribution of retinal color opponency.

Thanks. We substantially revised the text to discuss caveats and limitations of the approach. For example, we added a section into the Discussion called "Limitations of the stimulus paradigm". In addition, we clarified how retinal signals relate to cortical ones and phrased our conclusions more conservatively.

The authors provide solid evidence to support an asymmetric distribution of color opponent cells in V1 and a reduced color contrast representation in lower light levels. Some statements would benefit from more direct evidence such as the integration of upstream visual signals for color opponency in V1.

Based on the comments from Reviewer 1, we have rephrased the statements regarding the integration of upstream visual signals for color opponency in V1. We think these revisions increase the clarity of the results and help the reader with interpretation.

Overall, this study will be a valuable resource for researchers studying color vision, cortical processing, and the processing of ethologically relevant information. It provides a useful basis for future work on the origin of color opponency in V1 and its ethological relevance.

Thanks! We thank the reviewer again for the helpful comments.

Referee 3 (Remarks to the Author):

This paper studies chromatic coding in mouse primary visual cortex. Calcium responses of a large collection of cells are measured in response to a simple spot stimulus. These responses are used to estimate chromatic tuning properties - specifically sensitivity to UV and green stimuli presented in a large central spot or a larger still surrounding region. Cells are divided based on their responses to these stimuli into luminance or chromatic sensitive groups. Several technical concerns limit how clearly the data support the conclusions. If these issues can be fixed, the paper would make a valuable contribution to how color is coded in mouse V1.

We thank the reviewer for the helpful comments.

Analysis: The central tool used to analyze the data is a "spike triggered average" of the responses to randomly varying stimuli. There are several steps in this analysis that are not documented, and hence evaluating how well it works is difficult. Central to this is that the paper does not measure spikes. Instead, measured calcium traces are converted to estimated spike rates, which are then used to estimate STAs. There are no raw calcium traces shown, and the approach to estimate spike rates is not described in any detail. Confirming the accuracy of these steps is essential for a reader to be able to evaluate the paper. Further, it is not clear why the linear filters connecting the recorded calcium traces and the stimulus cannot be estimated directly, without the intermediate step of estimating spike rates.

Thank you for this comment. We have used the genetically encoded calcium sensor GCaMP6s in our recordings. This sensor is a very sensitive GCaMP6 variant, but also one with slow kinetics. To remove the effect of the slow sensor kinetics from recorded calcium responses, the recorded traces are commonly deconvolved with the impulse function of the sensor to obtain the deconvolved calcium traces. We now include this reasoning in the Results section. To illustrate the effect of the deconvolution, we added a new Suppl. Figure (Suppl. Fig. 2) showing raw calcium and deconvolved traces, and the STAs estimated from both types of traces. This illustrates that the results regarding neuronal color preferences are consistent across raw and deconvolved calcium traces.

We agree with the reviewer that the term STA might be confusing. We have replaced it with the term "even-triggered-average" (ETA). In addition, we have replaced the phrase "estimated spike rate" with "deconvolved calcium trace" throughout the manuscript because the unit of the deconvolved traces is not interpretable, like spike rate would be (spikes per second). In the revised version, we now clarify in the Methods section that we estimate the ETAs based on deconvolved calcium traces, which is correlated with and an approximation for spike rate.

A further issue about the STAs is that the inclusion criterion (correlation of predicted vs measured responses of 0.25) is pretty forgiving. It would be helpful to see a distribution of those correlation values, and some control analyses to check whether the STA is providing a sufficiently accurate measure to support the results (e.g. do the central results hold for the cells with the highest correlations).

We thank the reviewer for this comment. To exclude noisy neurons from analysis, we used the following procedure:

For each of the four stimulus conditions (center and surround for green and UV stimuli), kernel quality was measured by comparing the variance of the STA with the variance of the baseline, defined as the first 500 ms of the STA. Only cells with at least 10-times more variance of the kernel compared to baseline for UV or green center STA were considered for further analysis.

We have added the distribution of quality values to a new Suppl. Figure (Suppl. Fig. 2d,e). We now also show the percentage of neurons above threshold, given different quality thresholds. Finally, we have repeated the analysis shown in Figure 2 for a much more conservative threshold, including only the top 25% of neurons (Suppl. Fig. 2e,f). We now mention this new analysis in the Methods and Results section.

Limitations of stimulus choice: The paper relies on responses to a large (37.5 degree diameter) modulated spot and surrounding region. This spot is considerably larger than the receptive fields of both V1 cells and retinal ganglion cells. As a result, the spot itself is very likely to strongly activate both center and surround mechanisms, and responses of cells are likely to depend on where the receptive fields are located within the spot (and, e.g., how much of the true neural surround samples the center spot vs the surround region). The impact of these issues on the conclusions is considered briefly at the start of the results but needs to be evaluated in considerably more detail. This is particularly true for retinal ganglion cells given the size of their receptive fields (see also next point).

We agree with the reviewer that the centering of the stimulus is critical and apologize if this point was not discussed sufficiently. To be able to record from many V1 neurons simultaneously, we used a stimulus size of 37.5 degree visual angle in diameter, which is slightly larger than center RFs of single V1 neurons. As the reviewer mentions, the disadvantage of this approach is that the stimulus is only roughly centered on the neurons’ center RFs. To reduce the impact of potential stimulus misalignment on our results, we have used different experimental and analysis steps and controls (see also second comment of Reviewer 1):

For each recording, we positioned the monitor such that the mean RF across all neurons lies within the center of the stimulus field of view.

We confirmed that this procedure results in good stimulus alignment for the large majority of recorded neurons within individual recording fields by using a sparse noise stimulus (Suppl. Fig. 1a-c). Specifically, we found that for 83% of tested neurons, more than two thirds of their center RF, determined by the sparse noise stimulus, overlapped with the center spot of the color noise stimulus.

For analysis, we excluded neurons without a significant center STA, which may be caused by misalignment of the stimulus.

We now mention those clearly in the Results section and added the limitations of our approach to the Discussion section.

Comparison with retina: A key conclusion of the paper is that the chromatic tuning in V1 is not inherited from retinal ganglion cells. This conclusion comes from comparing chromatic tuning in a previously-collected data set from retina with the present results. But the retina recordings were made using a considerably smaller spot, and hence it is not clear that the comparison made in the paper is accurate. This issue may be handled by the analysis presented in the paper, but if so it needs to be described more clearly. The paper from which the retina data is taken argues that rod-cone chromatic opponency originates largely in the outer retina. This mechanism would be expected to be shared across retinal outputs. Thus it is not clear how the Green-On/UV-Off vs Green-Off/UV-On asymmetry could originate. This should be discussed.

We agree with the reviewer that a one-to-one comparison of retina and V1 data is challenging, due to differences in both RF and stimulus size. We rephrased the Results text to clarify this point and now also mention it in the Discussion.

When analyzing available data from the retina using a similar center-surround color flicker stimulus (Szatko et al. 2020), we found that Green On/UV Off color opponency is very rare in the RF center of retinal ganglion cells (Suppl. Fig. 5). This suggests that center Green On/UV Off color opponency in V1 neurons is not inherited by the RF center of retinal neurons. However, we agree with the reviewer that retinal neurons might still contribute to V1 color opponency, for example by being center-surround color opponent (e.g. Joesch et al. 2016 and Szatko et al. 2020). We rephrased the text to acknowledge this fact.

Residual chromatic cells at low mesopic light levels The presence of chromatically tuned cells at the lowest light level probed is surprising. The authors describe these conditions as rod-dominated, in which case chromatic tuning should not be possible. This again is discussed only briefly. It either reflects the presence of an unexpected pathway that amplifies weak cone signals under low mesopic conditions such that they can create spectral opponency or something amiss in the calibrations or analysis. Data collected at still lower light levels would help resolve this.

Thank you for this comment. We call the lowest light level "low mesopic" and "rod-dominated" because the spectral contrast of V1 center responses in posterior recording fields is green-shifted for this light level (Fig. 3a). This is only expected if responses in the UV-cone dominant ventral retina are predominantly driven by rod photoreceptors. We now explain this rationale in the Results section. In addition, we mention in the Discussion that future studies are required to test whether cone signals need to be amplified for low light levels. While we agree with the reviewer that it would be exciting to use even lower light levels during recordings, we believe this is out of the scope of the current study due to the technical challenges involved in achieving scotopic stimulation.

Author response:

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

As you will see, the main changes in the revised manuscript pertain to the structure and content of the introduction. Specifically, we have tried to more clearly introduce our paradigm, the rationale behind the paradigm, why it is different from learning paradigms, and why we study “relief”.

In this rebuttal letter, we will go over the reviewers’ comments one-by-one and highlight how we have adapted our manuscript accordingly. However, because one concern was raised by all reviewers, we will start with an in-depth discussion of this concern.

The shared concern pertained to the validity of the EVA task as a model to study threat omission responses. Specifically, all reviewers questioned the effectivity of our so-called “inaccurate”, “false” or “ruse” instructions in triggering an equivalent level of shock expectancy, and relatedly, how this effectivity was affected by dynamic learning over the course of the task.

We want to thank the reviewers for raising this important issue. Indeed, it is a vital part of our design and it therefore deserves considerable attention. It is now clear to us that in the previous version of the manuscript we may have focused too little on why we moved away from a learning paradigm, and how we made sure that the instructions were successful at raising the necessary expectations; and how the instructions were affected by learning. We believe this has resulted in some misunderstandings, which consequently may have cast doubts on our results. In the following sections, we will go into these issues.

The rationale behind our instructed design

The main aim of our study was to investigate brain responses to unexpected omissions of threat in greater detail by examining their similarity to the reward prediction error axioms (Caplin & Dean, 2008), and exploring the link with subjective relief. Specifically, we hypothesized that omission-related responses should be dependent on the probability and the intensity of the expected-but-omitted aversive event (i.e., electrical stimulation), meaning that the response should be larger when the expected stimulation was stronger and more expected, and that fully predicted outcomes should not trigger a difference in responding.

To this end, we required that participants had varying levels of threat probability and intensity predictions, and that these predictions would most of the time be violated. Although we fully agree with the reviewers that fear conditioning and extinction paradigms can provide an excellent way to track the teaching properties of prediction error responses (i.e., how they are used to update expectancies on future trials), we argued that they are less suited to create the varying probability and intensity-related conditions we required (see Willems & Vervliet, 2021). Specifically, in a standard conditioning task participants generally learn fast, rendering relatively few trials on which the prediction is violated. As a result, there is generally little intraindividual variability in the prediction error responses. This precludes an in-depth analysis of the probability-related effects. Furthermore, conditioning paradigms generally only include one level of aversive outcome: the electrical stimulation is either delivered or omitted. As a result, intensity-related effects cannot be tested. Finally, because CS-US contingencies change over the course of a fear conditioning and extinction study (e.g. from acquisition to extinction), there is never complete certainty about when the US will (not) follow. This precludes a direct comparison of fully predicted outcomes.

Another added value of studying responses to the prediction error at threat omission outside a learning context is that it can offer a way to disentangle responses to the violation of threat expectancy, with those of subsequent expectancy updating.

Also note that Rutledge and colleagues (2010), who were the first to show that human fMRI responses in the Nucleus Accumbens comply to the reward prediction error axioms also did not use learning experiences to induce expectancy. In that sense, we argued it was not necessary to adopt a learning paradigm to study threat omission responses.

Adaptations in the revised manuscript: We included two new paragraphs in the introduction of the revised manuscript to elaborate on why we opted not to use a learning paradigm in the present study (lines 90-112).

“However, is a correlation with the theoretical PE over time sufficient for neural activations/relief to be classified as a PE-signal? In the context of reward, Caplin and colleagues proposed three necessary and sufficient criteria all PE-signals should comply to, independent of the exact operationalizations of expectancy and reward (the socalled axiomatic approach24,25; which has also been applied to aversive PE26–28). Specifically, the magnitude of a PE signal should: (1) be positively related to the magnitude of the reward (larger rewards trigger larger PEs); (2) be negatively related to likelihood of the reward (more probable rewards trigger smaller PEs); and (3) not differentiate between fully predicted outcomes of different magnitudes (if there is no error in prediction, there should be no difference in the PE signal).”

“It is evident that fear conditioning and extinction paradigms have been invaluable for studying the role of the threat omission PE within a learning context. However, these paradigms are not tailored to create the varying intensity and probability-related conditions that are required to evaluate the threat omission PE in the light of the PE axioms. First, conditioning paradigms generally only include one level of aversive outcome: the electrical stimulation is either delivered or omitted. As a result, the magnitude-related axiom cannot be tested. Second, in conditioning tasks people generally learn fast, rendering relatively few trials on which the prediction is violated. As a result, there is generally little intra-individual variability in the PE responses. Moreover, because of the relatively low signal to noise ratio in fMRI measures, fear extinction studies often pool across trials to compare omission-related activity between early and late extinction16, which further reduces the necessary variability to properly evaluate the probability axiom. Third, because CS-US contingencies change over the course of the task (e.g. from acquisition to extinction), there is never complete certainty about whether the US will (not) follow. This precludes a direct comparison of fully predicted outcomes. Finally, within a learning context, it remains unclear whether PErelated responses are in fact responses to the violation of expectancy itself, or whether they are the result of subsequent expectancy updating.”

Can verbal instructions be used to raise the expectancy of shock?

The most straightforward way to obtain sufficient variability in both probability and intensityrelated predictions is by directly providing participants with instructions on the probability and intensity of the electrical stimulation. In a previous behavioral study, we have shown that omission responses (self-reported relief and omission SCR) indeed varied with these instructions (Willems & Vervliet, 2021). In addition, the manipulation checks that are reported in the supplemental material provided further support that the verbal instructions were effective at raising the associated expectancy of stimulation. Specifically, participants recollected having received more stimulations after higher probability instructions (see Supplemental Figure 2). Furthermore, we found that anticipatory SCR, which we used as a proxy of fearful expectation, increased with increasing probability and intensity (see Supplemental Figure 3). This suggests that it is not necessary to have expectation based on previous experience if we want to evaluate threat omission responses in the light of the prediction error axioms.

Adaptations in the revised manuscript: We more clearly referred to the manipulation checks that are presented in the supplementary material in the results section of the main paper (lines 135-141).

“The verbal instructions were effective at raising the expectation of receiving the electrical stimulation in line with the provided probability and intensity levels. Anticipatory SCR, which we used as a proxy of fearful expectation, increased as a function of the probability and intensity instructions (see Supplementary Figure 3). Accordingly, post-experimental questions revealed that by the end of the experiment participants recollected having received more stimulations after higher probability instructions, and were willing to exert more effort to prevent stronger hypothetical stimulations (see Supplementary Figure 2).”

How did the inconsistency between the instructed and experienced probability impact our results?

All reviewers questioned how the inconsistency between the instructed and experienced probability might have impacted the probability-related results. However, judging from the way the comments were framed, it seems that part of the concern was based on a misunderstanding of the design we employed. Specifically, reviewer 1 mentions that “To ensure that the number of omissions is similar across conditions, the task employs inaccurate verbal instructions; I.e., 25% of shocks are omitted regardless of whether subjects are told that the probability is 100%, 75%, 50%, 25%, 0%.”, and reviewer 3 states that “... the fact remains that they do not get shocks outside of the 100% probability shock. So learning is occurring, at least for subjects who realize the probability cue is actually a ruse.” We want to emphasize that this was not what we did, and if it were true, we fully agree with the reviewers that it would have caused serious trust- and learning related issues, given that it would be immediately evident to participants that probability instructions were false. It is clear that under such circumstances, dynamic learning would be a big issue.

However, in our task 0% and 100% instructions were always accurate. This means that participants never received a stimulus following 0% instructions and always received the stimulation of the given intensity on the 100% instructions (see Supplemental Figure 1 for an overview of the trial types). Only for the 25%, 50% and 75% trials an equal reinforcement rate (25%) was maintained, meaning that the stimulation followed in 25% of the trials, irrespective of whether a 25%, 50% or 75% instruction was given. The reason for this was that we wanted to maximize and balance the number of omission trials across the different probability levels, while also keeping the total number of presentations per probability instruction constant. We reasoned that equating the reinforcement rate across the 25%, 50% and 75% instructions should not be detrimental, because (1) in these trials there was always the possibility that a stimulation would follow; and (2) we instructed the participants that each trial is independent of the previous ones, which should have discouraged them to actively count the number of shocks in order to predict future shocks.

Adaptations in the revised manuscript: We have tried to further clarify the design in several sections of the manuscript, including the introduction (lines 121-125), results (line 220) and methods (lines 478-484) sections:

Adaptation in the Introduction section: “Specifically, participants received trial-by-trial instructions about the probability (0%, 25%, 50%, 75% and 100%) and intensity (weak, moderate, strong) of a potentially painful upcoming electrical stimulation, time-locked by a countdown clock (see Fig.1A). While stimulations were always delivered on 100% trials and never on 0% trials, most of the other trials (25%-75%) did not contain the expected stimulation and hence provoked an omission PE.”

Adaptation in the Results section: “Indeed, the provided instructions did not map exactly onto the actually experienced probabilities, but were all followed by stimulation in 25% on the trials (except for the 0% trials and the 100% trials).”

Adaptation in the Methods section: “Since we were mainly interested in how omissions of threat are processed, we wanted to maximize and balance the number of omission trials across the different probability and intensity levels, while also keeping the total number of presentations per probability and intensity instruction constant. Therefore, we crossed all non-0% probability levels (25, 50, 75, 100) with all intensity levels (weak, moderate, strong) (12 trials). The three 100% trials were always followed by the stimulation of the instructed intensity, while stimulations were omitted in the remaining nine trials. Six additional trials were intermixed in each run: Three 0% omission trials with the information that no electrical stimulation would follow (akin to 0% Probability information, but without any Intensity information as it does not apply); and three trials from the Probability x Intensity matrix that were followed by electrical stimulation (across the four runs, each Probability x Intensity combination was paired at least once, and at most twice with the electrical stimulation).”

Could the incongruence between the instructed and experienced reinforcement rate have detrimental effects on the probability effect? We agree with reviewer 2 that it is possible that the inconsistency between instructed and experienced reinforcement rates could have rendered the exact probability information less informative to participants, which might have resulted in them paying less attention to the probability information whenever the probability was not 0% or 100%. This might to some extent explain the relatively larger difference in responding between 0% and 25% to 75% trials, but the relatively smaller differences between the 25% to 75% trials.

However, there are good reasons to believe that the relatively smaller difference between 25% to 75% trials was not caused by the “inaccurate” nature of our instructions, but is inherent to “uncertain” probabilities.

We added a description of these reasons to the supplementary materials in a supplementary note (supplementary note 4; lines 97-129 in supplementary materials), and added a reference to this note in the methods section (lines 488-490).

“Supplementary Note 4: “Accurate” probability instructions do not alter the Probability-effect

A question that was raised by the reviewers was whether the inconsistency between the probability instruction and the experienced reinforcement rate could have detrimental effects on the Probability-related results; especially because the effect of Probability was smaller when only including non-0% trials.

However, there are good reasons to believe that the relatively smaller difference between 25% to 75% trials was not caused by the “inaccurate” nature of our instructions, but that they are inherent to “uncertain” probabilities.

First, in a previously unpublished pilot study, we provided participants with “accurate” probability instructions, meaning that the instruction corresponded to the actual reinforcement rate (e.g., 75% instructions were followed by a stimulation in 75% of the trials etc.). In line with the present results and our previous behavioral study (Willems & Vervliet, 2021), the results of this pilot (N = 20) showed that the difference in the reported relief between the different probability levels was largest when comparing 0% and the rest (25%, 50% and 75%). Furthermore the overall effect size of Probability (excluding 0%) matched the one of our previous behavioral study (Willems & Vervliet, 2021): ηp2 = +/- 0.50.”

Author response image 1.

Main effect of Probability including 0% : F(1.74,31.23) = 53.94, p < .001, ηp2 = 0.75

Main effect of Probability excluding 0%: F(1.50, 28.43) = 21.03, p < .001, ηp2 = 0.53

Second, also in other published studies that used CSs with varying reinforcement rates (which either included explicit written instructions of the reinforcement rates or not) showed that the difference in expectations, anticipatory SCR or omission SCR was largest when comparing the CS0% to the other CSs of varying reinforcement rates (Grings & Sukoneck, 1971; Öhman et al., 1973; Ojala et al., 2022).

Together, this suggests that when there is a possibility of stimulation, any additional difference in probability will have a smaller effect on the omission responses, irrespective of whether the underlying reinforcement rate is accurate or not.

Adaptation to methods section: “Note that, based on previous research, we did not expect the inconsistency between the instructed and perceived reinforcement rate to have a negative effect on the Probability manipulation (see Supplementary Note 4).”

Did dynamic learning impact the believability of the instructions?

Although we tried to minimize learning in our paradigm by providing instructions that trials are independent from one another, we agree with the reviewers that this cannot preclude all learning. Any remaining learning effects should present themselves by downweighing the effect of the probability instructions over time. We controlled for this time-effect by including a “run” regressor in our analyses. Results of the Run regressor for subjective relief and omission-related SCR are presented in Supplemental Figure 5. These figures show that although there was a general drop in reported relief pleasantness and omission SCR over time, the effects of probability and intensity remained present until the last run. This indicates that even though some learning might have taken place, the main manipulations of probability and intensity were still present until the end of the task.

Adaptations in the revised manuscript: We more clearly referred to the results of the Blockregressor which were presented in the supplementary material in the results section of the main paper (lines 159-162).

Note that while there was a general drop in reported relief pleasantness and omission SCR over time, the effects of Probability and Intensity remained present until the last run (see Supplementary Figure 5). This further confirms that probability and intensity manipulations were effective until the end of the task.

In the following sections of the rebuttal letter, we will go over the rest of the comments and our responses one by one.

Reviewer #1 (Public Review):

Summary:

Willems and colleagues test whether unexpected shock omissions are associated with reward-related prediction errors by using an axiomatic approach to investigate brain activation in response to unexpected shock omission. Using an elegant design that parametrically varies shock expectancy through verbal instructions, they see a variety of responses in reward-related networks, only some of which adhere to the axioms necessary for prediction error. In addition, there were associations between omission-related responses and subjective relief. They also use machine learning to predict relief-related pleasantness, and find that none of the a priori "reward" regions were predictive of relief, which is an interesting finding that can be validated and pursued in future work.

Strengths:

The authors pre-registered their approach and the analyses are sound. In particular, the axiomatic approach tests whether a given region can truly be called a reward prediction error. Although several a priori regions of interest satisfied a subset of axioms, no ROI satisfied all three axioms, and the authors were candid about this. A second strength was their use of machine learning to identify a relief-related classifier. Interestingly, none of the ROIs that have been traditionally implicated in reward prediction error reliably predicted relief, which opens important questions for future research.

Weaknesses:

To ensure that the number of omissions is similar across conditions, the task employs inaccurate verbal instructions; i.e. 25% of shocks are omitted, regardless of whether subjects are told that the probability is 100%, 75%, 50%, 25%, or 0%. Given previous findings on interactions between verbal instruction and experiential learning (Doll et al., 2009; Li et al., 2011; Atlas et al., 2016), it seems problematic a) to treat the instructions as veridical and b) average responses over time. Based on this prior work, it seems reasonable to assume that participants would learn to downweight the instructions over time through learning (particularly in the 100% and 0% cases); this would be the purpose of prediction errors as a teaching signal. The authors do recognize this and perform a subset analysis in the 21 participants who showed parametric increases in anticipatory SCR as a function of instructed shock probability, which strengthened findings in the VTA/SN; however given that one-third of participants (n=10) did not show parametric SCR in response to instructions, it seems like some learning did occur. As prediction error is so important to such learning, a weakness of the paper is that conclusions about prediction error might differ if dynamic learning were taken into account.

We thank the reviewer for raising this important concern. We believe we replied to all the issues raised in the general reply above.

Lastly, I think that findings in threat-sensitive regions such as the anterior insula and amygdala may not be adequately captured in the title or abstract which strictly refers to the "human reward system"; more nuance would also be warranted.

We fully agree with this comment and have changed the title and abstract accordingly.

Adaptations in the revised manuscript: We adapted the title of the manuscript.

“Omissions of Threat Trigger Subjective Relief and Prediction Error-Like Signaling in the Human Reward and Salience Systems”

Adaptations in the revised manuscript: We adapted the abstract (lines 27-29).

“In line with recent animal data, we showed that the unexpected omission of (painful) electrical stimulation triggers activations within key regions of the reward and salience pathways and that these activations correlate with the pleasantness of the reported relief.”

Reviewer #2 (Public Review):

The question of whether the neural mechanisms for reward and punishment learning are similar has been a constant debate over the last two decades. Numerous studies have shown that the midbrain dopamine neurons respond to both negative and salient stimuli, some of which can't be well accounted for by the classic RL theory (Delgado et al., 2007). Other research even proposed that aversive learning can be viewed as reward learning, by treating the omission of aversive stimuli as a negative PE (Seymour et al., 2004).

Although the current study took an axiomatic approach to search for the PE encoding brain regions, which I like, I have major concerns regarding their experimental design and hence the results they obtained. My biggest concern comes from the false description of their task to the participants. To increase the number of "valid" trials for data analysis, the instructed and actual probabilities were different. Under such a circumstance, testing axiom 2 seems completely artificial. How does the experimenter know that the participants truly believe that the 75% is more probable than, say, the 25% stimulation? The potential confusion of the subjects may explain why the SCR and relief report were rather flat across the instructed probability range, and some of the canonical PE encoding regions showed a rather mixed activity pattern across different probabilities. Also for the post-hoc selection criteria, why pick the larger SCR in the 75% compared to the 25% instructions? How would the results change if other criteria were used?

We thank the reviewer for raising this important concern. We believe the general reply above covers most of the issues raised in this comment. Concerning the post-hoc selection criteria, we took 25% < 75% as criterium because this was a quite “lenient” criterium in the sense that it looked only at the effects of interest (i.e., did anticipatory SCR increase with increasing instructed probability?). However, also when the criterium was more strict (e.g., selecting participants only if their anticipatory SCR monotonically increased with each increase in instructed probability 0% < 25% < 50% < 75% < 100%, N = 11 participants), the probability effect (ωp2 = 0.08), but not the intensity effect, for the VTA/SN remained.

To test axiom 3, which was to compare the 100% stimulation to the 0% stimulation conditions, how did the actual shock delivery affect the fMRI contrast result? It would be more reasonable if this analysis could control for the shock delivery, which itself could contaminate the fMRI signal, with extra confound that subjects may engage certain behavioral strategies to "prepare for" the aversive outcome in the 100% stimulation condition. Therefore, I agree with the authors that this contrast may not be a good way to test axiom 3, not only because of the arguments made in the discussion but also the technical complexities involved in the contrast.

We thank the reviewer for addressing this additional confound. It was indeed impossible to control for the delivery of shock since the delivery of the shock was always present on the 100% trials (and thus completely overlapped with the contrast of interest). We added this limitation to our discussion in the manuscript. In addition, we have also added a suggestion for a contrast that can test the “no surprise equivalence” criterium.

Adaptations in the revised manuscript: We adapted lines 358-364.

“Thus, given that we could not control for the delivery of the stimulation in the 100% > 0% contrast (the delivery of the stimulation completely overlapped with the contrast of interest), it is impossible to disentangle responses to the salience of the stimulation from those to the predictability of the outcome. A fairer evaluation of the third axiom would require outcomes that are roughly similar in terms of salience. When evaluating threat omission PE, this implies comparing fully expected threat omissions following 0% instructions to fully expected absence of stimulation at another point in the task (e.g. during a safe intertrial interval).”

Reviewer #3 (Public Review):

We thank the reviewer for their comments. Overall, based on the reviewer’s comments, we noticed that there was an imbalance between a focus on “relief” in the introduction and the rest of the manuscript and preregistration. We believe this focus raised the expectation that all outcome measures were interpreted in terms of the relief emotion. However, this was not what we did nor what we preregistered. We therefore restructured the introduction to reduce the focus on relief.

Adaptations in the revised manuscript: We restructured the introduction of the manuscript. Specifically, after our opening sentence: “We experience a pleasurable relief when an expected threat stays away1” we only introduce the role of relief for our research in lines 79-89.

“Interestingly, unexpected omissions of threat not only trigger neural activations that resemble a reward PE, they are also accompanied by a pleasurable emotional experience: relief. Because these feelings of relief coincide with the PE at threat omission, relief has been proposed to be an emotional correlate of the threat omission PE. Indeed, emerging evidence has shown that subjective experiences of relief follow the same time-course as theoretical PE during fear extinction. Participants in fear extinction experiments report high levels of relief pleasantness during early US omissions (when the omission was unexpected and the theoretical PE was high) and decreasing relief pleasantness over later omissions (when the omission was expected and the theoretical PE was low)22,23. Accordingly, preliminary fMRI evidence has shown that the pleasantness of this relief is correlated to activations in the NAC at the time of threat omission. In that sense, studying relief may offer important insights in the mechanism driving safety learning.”

Summary:

The authors conducted a human fMRI study investigating the omission of expected electrical shocks with varying probabilities. Participants were informed of the probability of shock and shock intensity trial-by-trial. The time point corresponding to the absence of the expected shock (with varying probability) was framed as a prediction error producing the cognitive state of relief/pleasure for the participant. fMRI activity in the VTA/SN and ventral putamen corresponded to the surprising omission of a high probability shock. Participants' subjective relief at having not been shocked correlated with activity in brain regions typically associated with reward-prediction errors. The overall conclusion of the manuscript was that the absence of an expected aversive outcome in human fMRI looks like a reward-prediction error seen in other studies that use positive outcomes.

Strengths:

Overall, I found this to be a well-written human neuroimaging study investigating an often overlooked question on the role of aversive prediction errors, and how they may differ from reward-related prediction errors. The paper is well-written and the fMRI methods seem mostly rigorous and solid.

Weaknesses:

I did have some confusion over the use of the term "prediction-error" however as it is being used in this task. There is certainly an expectancy violation when participants are told there is a high probability of shock, and it doesn't occur. Yet, there is no relevant learning or updating, and participants are explicitly told that each trial is independent and the outcome (or lack thereof) does not affect the chances of getting the shock on another trial with the same instructed outcome probability. Prediction errors are primarily used in the context of a learning model (reinforcement learning, etc.), but without a need to learn, the utility of that signal is unclear.

We operationalized “prediction error” as the response to the error in prediction or the violation of expectancy at the time of threat omission. In that sense, prediction error and expectancy violation (which is more commonly used in clinical research and psychotherapy; Craske et al., 2014) are synonymous. While prediction errors (or expectancy violations) are predominantly studied in learning situations, the definition in itself does not specify how the “expectancy” or “prediction” arises: whether it was through learning based on previous experience or through mere instruction. The rationale why we moved away from a conditioning study in the present manuscript is discussed in our general reply above.

We agree with the reviewer that studying prediction errors outside a learning context limits the ecological validity of the task. However, we do believe there is also a strength to this approach. Specifically, the omission-related responses we measure are less confounded by subsequent learning (or updating of the wrongful expectation). Any difference between our results and prediction error responses in learning situation can therefore point to this exact difference in paradigm, and can thus identify responses that are specific to learning situations.

An overarching question posed by the researchers is whether relief from not receiving a shock is a reward. They take as neural evidence activity in regions usually associated with reward prediction errors, like the VTA/SN . This seems to be a strong case of reverse inference. The evidence may have been stronger had the authors compared activity to a reward prediction error, for example using a similar task but with reward outcomes. As it stands, the neural evidence that the absence of shock is actually "pleasurable" is limited-albeit there is a subjective report asking subjects if they felt relief.

We thank the reviewer for cautioning us and letting us critically reflect on our interpretation. We agree that it is important not to be overly enthusiastic when interpreting fMRI results and to attribute carelessly psychological functions to mere activations. Therefore, we will elaborate on the precautions we took not to minimize detrimental reverse inference.

First, prior to analyzing our results, we preregistered clear hypotheses that were based on previous research, in addition to clear predictions, regions of interest and a testing approach on OSF. With our study, we wanted to investigate whether unexpected omissions of threat: (1) triggered activations in the VTA/SN, putamen, NAc and vmPFC (as has previously been shown in animal and human studies); (2) represent PE signals; and (3) were related to self-reported relief, which has also been shown to follow a PE time-curve in fear extinction (Vervliet et al., 2017). Based on previous research, we selected three criteria all PE signals should comply to. This means that if omission-related activations were to represent true PE signals, they should comply to these criteria. However, we agree that it would go too far to conclude based on our research that relief is a reward, or even that the omission-related activations represent only PE signals. While we found support for most of our hypotheses, this does not preclude alternative explanations. In fact, in the discussion, we acknowledge this and also discuss alternative explanations, such as responding to the salience (lines 395-397; “One potential explanation is therefore that the deactivation resulted from a switch from default mode to salience network, triggered by the salience of the unexpected threat omission or by the salience of the experienced stimulation.”), or anticipation (line 425-426; “... we cannot conclusively dismiss the alternative interpretation that we assessed (part of) expectancy instead”).

Second, we have deliberately opted to only use descriptive labels such as omission-related activations when we are discussing fMRI results. Only when we are talking about how the activations were related to self-reported relief, we talk about relief-related activations.

I have some other comments, and I elaborate on those above comments, below:

(1) A major assumption in the paper is that the unexpected absence of danger constitutes a pleasurable event, as stated in the opening sentence of the abstract. This may sometimes be the case, but it is not universal across contexts or people. For instance, for pathological fears, any relief derived from exposure may be short-lived (the dog didn't bite me this time, but that doesn't mean it won't next time or that all dogs are safe). And even if the subjective feeling one gets is temporary relief at that moment when the expected aversive event is not delivered, I believe there is an overall conflation between the concepts of relief and pleasure throughout the manuscript. Overall, the manuscript seems to be framed on the assumption that "aversive expectations can transform neutral outcomes into pleasurable events," but this is situationally dependent and is not a common psychological construct as far as I am aware.

We thank the reviewer for their comment. We have restructured the introduction because we agree with the reviewer that the introduction might have set false expectations concerning our interpretation of the results. The statements related to relief have been toned down in the revised manuscript.

Still, we want to note that the initial opening statement “unexpected absence of danger constitutes the pleasurable emotion relief” was based on a commonly used definition of relief that states that relief refers to “the emotion that is triggered by the absence of expected or previously experienced negative stimulation ” (Deutsch, 2015). Both aspects that it is elicited by the absence of an otherwise expected aversive event and that it is pleasurable in nature has received considerable empirical support in emotion and fear conditioning research (Deutsch et al., 2015; Leknes et al., 2011; Papalini et al., 2021; Vervliet et al., 2017; Willems & Vervliet, 2021).

That said, the notion that the feeling of relief is linked to the (reward) prediction error underlying the learning of safety is included in several theoretical papers in order to explain the commonly observed dopaminergic response at the time of threat omission (both in animals and humans; Bouton et al., 2020; Kalisch et al., 2019; Pittig et al., 2020).

Together, these studies indicate that the definition of relief, and its potential role in threat omission-driven learning is – at least in our research field – established. Still, we felt that more direct research linking feelings of relief to omission-related brain responses was warranted.

One of the main reasons why we specifically focus on the “pleasantness” of the relief is to assess the hedonic impact of the threat omission, as has been done in previous studies by our lab and others (Leknes et al., 2011; Leng et al., 2022; Papalini et al., 2021; Vervliet et al., 2017; Willems & Vervliet, 2021). Nevertheless, we agree with the reviewer that the relief we measure is a short-lived emotional state that is subjected to individual differences (as are all emotions).

(2) The authors allude to this limitation, but I think it is critical. Specifically, the study takes a rather simplistic approach to prediction errors. It treats the instructed probability as the subjects' expectancy level and treats the prediction error as omission related activity to this instructed probability. There is no modeling, and any dynamic parameters affected by learning are unaccounted for in this design . That is subjects are informed that each trial is independently determined and so there is no learning "the presence/absence of stimulations on previous trials could not predict the presence/absence of stimulation on future trials." Prediction errors are central to learning. It is unclear if the "relief" subjects feel on not getting a shock on a high-probability trial is in any way analogous to a prediction error, because there is no reason to update your representation on future trials if they are all truly independent. The construct validity of the design is in question.

(3) Related to the above point, even if subjects veered away from learning by the instruction that each trial is independent, the fact remains that they do not get shocks outside of the 100% probability shock. So learning is occurring, at least for subjects who realize the probability cue is actually a ruse.

We thank the reviewer for raising these concerns. We believe that the general reply above covers the issues raised in points 2 and 3.

(4) Bouton has described very well how the absence of expected threat during extinction can create a feeling of ambiguity and uncertainty regarding the signal value of the CS. This in large part explains the contextual dependence of extinction and the "return of fear" that is so prominent even in psychologically healthy participants. The relief people feel when not receiving an expected shock would seem to have little bearing on changing the long-term value of the CS. In any event, the authors do talk about conditioning (CS-US) in the paper, but this is not a typical conditioning study, as there is no learning.

We fully agree with the reviewer that our study is no typical conditioning study. Nevertheless, because our research mostly builds on recent advances in the fear extinction domain, we felt it was necessary to introduce the fear extinction procedure and related findings. In the context of fear extinction learning, we have previously shown that relief is an emotional correlate of the prediction error driving acquisition of the novel safety memory (CSnoUS; Papalini et al., 2021; Vervliet et al., 2017). The ambiguity Bouton describes is the result of extinguished CS holding multiple meanings once the safety memory is acquired. Does it signal danger or safety? We agree with Bouton that the meaning of the CS for any new encounter will depend on the context, and the passage of time, but also on the initial strength of the safety acquisition (which is dependent on the size of the prediction error, and hence the amount of relief; Craske et al., 2014). However, it was not our objective to directly study the relation of relief to subsequent CS value, and our design is not tailored to do so post hoc.

(5) In Figure 2 A-D, the omission responses are plotted on trials with varying levels of probability. However, it seems to be missing omission responses in 0% trials in these brain regions. As depicted, it is an incomplete view of activity across the different trial types of increasing threat probability.

We thank the reviewer for pointing out this unclarity. The betas that are presented in the figures represent the ROI averages from each non-0% vs 0% contrasts (i.e., 25%>0%; 50%>0%; and 75%>0% for the weak, moderate and strong intensity levels). Any positive beta therefore indicates a stronger activation in the given region compared to a fully predicted omission. Any negative beta indicates a weaker activation.

Adaptations in the revised manuscript: We have adapted the figure captions of figures 2 and 3.

“The extracted beta-estimates in figures A-D represent the ROI averages from each non0% > 0% contrast (i.e., 25%>0%; 50%>0%; and 75%>0% for the weak, moderate and strong intensity levels). Any positive beta therefore indicates a stronger activation in the given region compared to a fully predicted omission. Any negative beta indicates a weaker activation.”

(6) If I understand Figure 2 panels E-H, these are plotting responses to the shock versus no-shock (when no-shock was expected). It is unclear why this would be especially informative, as it would just be showing activity associated with shocks versus no-shocks. If the goal was to use this as a way to compare positive and negative prediction errors, the shock would induce widespread activity that is not necessarily reflective of a prediction error. It is simply a response to a shock. Comparing activity to shocks delivered after varying levels of probability (e.g., a shock delivered at 25% expectancy, versus 75%, versus 100%) would seem to be a much better test of a prediction error signal than shock versus no-shock.

We thank the reviewer for this comment. The purpose of this preregistered contrast was to test whether fully predicted outcomes elicited equivalent activations in our ROIs (corresponding to the third prediction error axiom). Specifically, if a region represents a pure prediction error signal, the 100% (fully predicted shocks) > 0% (fully predicted shock omissions) contrast should be nonsignificant, and follow-up Bayes Factors would further provide evidence in favor of this null-hypothesis.

We agree with the reviewer that the delivery of the stimulation triggers widespread activations in our regions of interest that confounded this contrast. However, given that it was a preregistered test for the prediction error axioms, we cannot remove it from the manuscript. Instead, we have argued in the discussion that future studies who want to take an axiomatic stance should consider alternative tests to examine this axiom.

Adaptations in the revised manuscript: We adapted lines 358-364.

“Thus, given that we could not control for the delivery of the stimulation in the 100% > 0% contrast (the delivery of the stimulation completely overlapped with the contrast of interest), it is impossible to disentangle responses to the salience of the stimulation from those to the predictability of the outcome. A fairer evaluation of the third axiom would require outcomes that are roughly similar in terms of salience. When evaluating threat omission PE, this implies comparing fully expected threat omissions following 0% instructions to fully expected absence of stimulation at another point in the task (e.g. during a safe intertrial interval).”

Also note that our task did not lend itself for an in-depth analysis of aversive (worse-thanexpected) prediction error signals, given that there was only one stimulation trial for each probability x intensity level (see Supplemental Figure 1). The most informative contrast that can inform us about aversive prediction error signals contrasts all non-100% stimulation trials with all 100% stimulation trials. The results of this contrast are presented in Supplemental Figure 16 and Supplemental Table 11 for completeness.

(7) I was unclear what the results in Figure 3 E-H were showing that was unique from panels A-D, or where it was described. The images looked redundant from the images in A-D. I see that they come from different contrasts (non0% > 0%; 100% > 0%), but I was unclear why that was included.

We thank the reviewer for this comment. Our answer is related to that of the previous comment. Figure 3 presents the results of the axiomatic tests within the secondary ROIs we extracted from a wider secondary mask based on the non0%>0% contrast.

(8) As mentioned earlier, there is a tendency to imply that subjects felt relief because there was activity in "the reward pathway ."

We thank the reviewer for their comment, but we respectfully disagree. Subjective relief was explicitly probed when the instructed stimulations stayed away. In the manuscript we only talk about “relief” when discussing these subjective reports. We found that participants reported higher levels of relief-pleasantness following omissions of stronger and more probable threat. This was an observation that matches our predictions and replicates our previous behavioral study (Willems & Vervliet, 2021).

The fMRI evidence is treated separately from the “pleasantness” of the relief. Specifically, we refrain from calling the threat omission-related neural responses “relief-activity” as this would indeed imply that the activation would only be attributed to this psychological function. Instead, we talked about omission-related activity, and we assessed whether it complied to the prediction error criteria as specified by the axiomatic approach.

Only afterwards, because we hypothesized that omission-related fMRI activation and selfreported relief-pleasantness were related, and because we found a similar response pattern for both measures, we examined how relief and omission-related fMRI activations within our ROIs were related on a trial-by-trial basis. To this end, we entered relief-pleasantness ratings as a parametric modulator to the omission regressor.

By no means do we want to reduce an emotional experience (relief) to fMRI activations in isolated regions in the brain. We agree with the reviewer that this would be far too reductionist. We therefore also ran a pre-registered LASSO-PCR analysis in order to identify whether a whole-brain pattern of activations can predict subjective relief (independent from the exact instructions we gave, and independent of our a priori ROIs). This analysis used trialby-trial patterns of activation across all voxels in the brain as the predictor and self-reported relief as the outcome variable. It is therefore completely data-driven and can be seen as a preregistered exploratory analysis that is intended to inform future studies.

(9) From the methods, it wasn't entirely clear where there is jitter in the course of a trial. This centers on the question of possible collinearity in the task design between the cue and the outcome. The authors note there is "no multicollinearity between anticipation and omission regressors in the firstlevel GLMs," but how was this quantified? b The issue is of course that the activity coded as omission may be from the anticipation of the expected outcome.

We thank the reviewer for pointing out this unclarity. Jitter was introduced in all parts of the trial: i.e., the duration of the inter-trial interval (4-7s), countdown clock (3-7s), and omission window (4-8s) were all jittered (see fig. 1A and methods section, lines 499-507). We added an additional line to the method section.

Adaptations in the revised manuscript: We added an additional line of to the methods section to further clarify the jittering (lines 498-500).

“The scale remained on the screen for 8 seconds or until the participant responded, followed by an intertrial interval between 4 and 7 seconds during which only a fixation cross was shown. Note that all phases in the trial were jittered (i.e., duration countdown clock, duration outcome window, duration intertrial interval).”

Multicollinearity between the omission and anticipation regressors was assessed by calculating the variance inflation factor (VIF) of omission and anticipation regressors in the first level GLM models that were used for the parametric modulation analyses.

Adaptations in the revised manuscript: We replaced the VIF abbreviation with “variance inflation factor” (line 423-424).

“Nevertheless, there was no multicollinearity between anticipation and omission regressors in the first-level GLMs (VIFs Variance Inflation Factor, VIF < 4), making it unlikely that the omission responses purely represented anticipation.”

(10) I did not fully understand what the LASSO-PCR model using relief ratings added. This result was not discussed in much depth, and seems to show a host of clusters throughout the brain contributing positively or negatively to the model. Altogether, I would recommend highlighting what this analysis is uniquely contributing to the interpretation of the findings.

The main added value of this analyses is that it uses a different approach altogether. Where the (mass univariate) parametric modulation analysis estimated in each voxel (and each ROI) whether the activity in this voxel/ROI covaried with the reported relief, a significant activation only indicated that this voxel was related to relief. However, given that each voxel/ROI is treated independently in this analysis, it remains unclear how the activations were embedded in a wider network across the brain, and which regions contributed most to the prediction of relief. The multivariate LASSO-PCR analysis approach we took attempts to overcome this limitation by examining if a more whole-brain pattern can predict relief. Because we use the whole-brain pattern (and not only our a priori ROIs), this analysis is completely data-driven and is intended to inform future studies. In addition, the LASSO-PCR model was cross-validated using five-fold cross-validation, which is also a difference (and a strength) compared to the mass univariate GLM approach.

One interesting finding that only became evident when we combined univariate and multivariate approaches is that despite that the parametric modulation analysis showed that omission-related fMRI responses in the ROIs were modulated by the reported relief, none of these ROIs contributed significantly to the prediction of relief based on the identified signature. Instead, some of the contributing clusters fell within other valuation and errorprocessing regions (e.g. lateral OFC, mid cingulate, caudate nucleus). This suggests that other regions than our a priori ROIs may have been especially important for the subjective experience of relief, at least in this task. However, all these clusters were small and require further validation in out of sample participants. More research is necessary to test the generalizability and validity of the relief signature to new individuals and tasks, and to compare the signature with other existing signature models (e.g., signature of pain, fear, reward, pleasure). However, this was beyond the scope of the present study.

Adaptations in the revised manuscript: We altered the explanation of the LASSO-PCR approach in the results section (lines 286-295) and the discussion (lines 399-402)

Adaptations in the Results section: “The (mass univariate) parametric modulation analysis showed that omission-related fMRI activity in our primary and secondary ROIs correlated with the pleasantness of the relief. However, given that each voxel/ROI is treated independently in this analysis, it remains unclear how the activations were embedded in a wider network of activation across the brain, and which regions contributed most to the prediction of relief. To overcome these limitations, we trained a (multivariate) LASSO-PCR model (Least Absolute Shrinkage and Selection Operator-Regularized Principle Component Regression) in order to identify whether a spatially distributed pattern of brain responses can predict the perceived pleasantness of the relief (or “neural signature” of relief)31. Because we used the whole-brain pattern (and not only our a priori ROIs), this analysis is completely data driven and can thus identify which clusters contribute most to the relief prediction.”

Adaptations in the Discussion section: “In addition to examining the PE-properties of neural omission responses in our a priori ROIs, we trained a LASSO-PCR model to establish a signature pattern of relief. One interesting finding that only became evident when we compared the univariate and multivariate approach was that none of our a priori ROIs appeared to be an important contributor to the multivariate neural signature, even though all of them (except NAc) were significantly modulated by relief in the univariate analysis.”

In addition to the public peer review, the reviewers provided some recommendation on how to further improve our manuscript. We will reply to the recommendations below.

Reviewer #1 (Recommendations For The Authors):

Given that you do have trial-level estimates from the classifier analysis, it would be very informative to use learning models and examine responses trial-by-trial to test whether there are prediction errors that vary over time as a function of learning.

We thank the reviewer for the suggestion. However, based on the results of the run-regressor, we do not anticipate large learning effects in our paradigm. As we mentioned in our responses above, we controlled for time-related drops in omission-responding by including a “run” regressor in our analyses. Results of this regressor for subjective relief and omission-related SCR showed that although there was a general drop in reported relief pleasantness and omission SCR over time, the effects of probability and intensity remained present until the last run. This suggests that even though some learning might have taken place, its effect was likely small and did not abolish our manipulations of probability and intensity. In any case, we cannot use the LASSO-PCR signature model to investigate learning, as this model uses the trial-level brain pattern at the time of US omission to estimate the associated level of relief. These estimates can therefore not be used to examine learning effects.

Reviewer #2 (Recommendations For The Authors):

The LASSO-PCR model feels rather disconnected from the rest of the paper and does not add much to the main theme. I would suggest to remove this part from the paper.

We thank the reviewer for this suggestion. However, the LASSO-PCR analysis was a preregistered. We therefore cannot remove it from the manuscript. We hope to have clarified its added value in the revised version of the manuscript.

Author response:

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

We have revised the manuscript mainly in the following aspects: (1) the data of electrophysiological and behavioral responses of larvae and adults to trehalose have been added, and the related figures and texts have been modified accordingly; (2) the photos of taste organs of larvae and adults indicating the position of recorded sensilla have been added; (3) the potential off-target effects of GR knock-out on other GR expressions has been carefully explained and revised in the relevant text; (4) the abstract has been revised to present the findings more technically in a limited number of words; (5) some details of experiments in Materials and Methods and some new literatures have been added; (6) a new figure (Figure 8) summarizing the main findings of the study has been added.

In the following, we respond to the reviewers’ comments and suggestions one by one. We hope that our answers will satisfy you and the three reviewers. We are also very happy to get further valuable advices from you.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The process of taste perception is significantly more intricate and complex in Lepidopteran insects. This investigation provides valuable insights into the role of Gustatory receptors and their dynamics in the sensation of sucrose, which serves as a crucial feeding cue for insects. The article highlights the differential sensitivity of Grs to sucrose and their involvement in feeding and insect behavior.

Strengths:

To support the notion of the differential specificity of Gr to sucrose, this study employed electrophysiology, ectopic expression of Grs in Xenopus, genome editing, and behavioral studies on insects. This investigation offers a fundamental understanding of the gustation process in lepidopteran insects and its regulation of feeding and other gustation-related physiological responses. This study holds significant importance in advancing our comprehension of lepidopteran insect biology, gustation, and feeding behavior.

Thank you for your recognition of our research.

Weaknesses:

While this manuscript demonstrates technical proficiency, there exists an opportunity for additional refinement to optimize comprehensibility for the intended audience. Several crucial sugars have been overlooked in the context of electrophysiology studies and should be incorporated. Furthermore, it is imperative to consider the potential off-target effects of Gr knock-out on other Gr expressions. This investigation focuses exclusively on Gr6 and Gr10, while neglecting a comprehensive narrative regarding other Grs involved in sucrose sensation.

We accept the reviewer's suggestion. Because trehalose is a main sugar in insect blood, and it is converted by insects after feeding on plant sugars, we have added the new data on electrophysiological and behavioral responses of larvae and adults of Helicoverpa armigera to trehalose (see Figure 1-2, Figure 1-figure supplement 1, Figure 2-figure supplement 1). Now, the total eight sugars include 2 pentoses (arabinose and xylose), 4 hexoses (fructose, fucose, galactose and glucose), and 2 disaccharides (sucrose and trehalose), which were chosen because they are mainly present in host-plants of H. armigera and/or representative in the structure and source of sugars.

We fully agree to the reviewer’s opinion and have already taken the potential off-target effects of CRISPR/Cas9 knockout of Gr on other GR expressions into consideration. To predict the potential off-target sites of sgRNA of Gr6 and Gr10 establishing homozygous mutants using CRISPR/Cas9 technology, we first use online software CasOFFinder (http://www.rgenome.net/cas-offinder/) to blast the genome of the wild type cotton bollworm and set the mismatch number less than or equal to 3. We found that Gr10 sgRNA had no potential potential off-target site, and the sgRNA of Gr6 had only one potential off-target site. Therefore, we designed primers according to the sequence of potential off-target sites of Gr6 sgRNA, and conducted PCR using genomic DNA of homozygous mutant as a template， performed Sanger sequencing on the PCR products obtained, and found that the potential off-target sites of Gr6 sgRNA were no different from those of the wild type. Particularly, concerning the sgRNA of Gr6 and Gr10 may produce off-target effects on other sugar receptor genes of H. armigera, we conducted the same off-target site analysis with the designed sgRNA on each of the other eight sugar receptor genes, and found that there were no off-target sites on these receptor genes (see Line254-256).

Reviewer #2 (Public Review):

Summary:

To identify sugar receptors and assess the capacity of these genes the authors first set out to identify behavioral responses in larvae and adults as well as physiological response. They used phylogenetics and gene expression (RNAseq) to identify candidates for sugar reception. Using first an in vitro oocyte system they assess the responses to distinct sugars. A subsequent genetic analysis shows that the Gr10 and Gr6 genes provide stage specific functions in sugar perception.

Strengths:

A clear strength of the manuscript is the breadth of techniques employed allowing a comprehensive study in a non-canonical model species.

Thank you for your recognition of our research.

Weaknesses:

There are no major weaknesses in the study for the current state of knowledge in this species. Since it is much basic work to establish a broader knowledge, context with other modalities remains unknown. It might have been possible to probe certain contexts known from the fruit fly, which would have strengthened the manuscript.

Thank you so much for your suggestion. According to this suggestion, we further added some sentences probing sugar sensing and behaviors of fruit fly larvae in the Introduction and discussion sections (Line 68-71 in Introduction section, Line 395-399 in Discussion section).

Reviewer #3 (Public Review):

In this study, the authors combine electrophysiology, behavioural analyses, and genetic editing techniques on the cotton bollworm to identify the molecular basis of sugar sensing in this species.

The larval and adult forms of this species feed on different plant parts. Larvae primarily consume leaves, which have relatively lower sugar concentrations, while adults feed on nectar, rich in sugar. Through a series of experiments-spanning electrophysiological recordings from both larval and adult sensillae, qPCR expression analysis of identified GRs from these sensillae, response profiles of these GRs to various sugars via heterologous expression in Xenopus oocytes, and evaluations of CRISPR mutants based on these parameters-the authors discovered that larvae and adults employ distinct GRs for sugar sensing. While the larva uses the highly sensitive GR10, the adult uses the less sensitive and broadly tuned GR6. This differential use of GRs are in keeping with their behavioral ecology.

The data are cohesive and consistently align across the methodologies employed. They are also well presented and the manuscript is clearly written.

Recommendations for the authors:

While appreciating the quality of the work and its presentation, we have a few comments for the authors, should they wish to consider them, that would significantly improve the presentation of the work.

Title: Could the authors please revisit their title to better reflect the main finding of their work?

The title has been changed into “The larva and adult of Helicoverpa armigera use differential gustatory receptors to sense sugars”.

Text: There are a few comments related to the text, and these are listed below:

(1) Could the authors place their work in the context of what's known about sugar sensing in Drosophila larva and adult?

In the Introduction section, we added the status of research on sugar perception in Drosophila larvae, pointing out "No external sugar-sensing mechanism in Drosophila larvae has yet been characterized." (Line 70-71); in the Discussion section, the research progress of sugar sensing in Drosophila adults and larvae was also summarized (Line 397-399).

(2) For each results section, could the authors please include a sentence or two that interprets the data in the context of previously presented data?

We accept the reviewer's suggestion. In order to make it easy for readers to follow up, we included a sentence interprets the above data at the beginning of each part of the Results on the premise of avoiding duplication.

(3) Could the authors please provide details of the generation and screening of the CRISPR mutants?

We have added more details on mutant establishment and screening in the Materials and Methods section (Line 722-726, 729-732).

Figures: Could the authors please include images and schematics wherever possible? For example, a schematic depicting the position of the sense organs and one summarising the main findings of the studies.

In Figure 1 we added the photo of each taste organ, on which the recorded sensilla were indicated. We also added a new figure, Figure 8, summarizing the main findings of the study.

Choice of Sugars: Could the authors please justify their choice of sugars they have used in the analyses?

In the first paragraph of the Results section of the article, we further explain the reasons for using the sugars in the study. “We first investigated the electrophysiological responses of the lateral and medial sensilla styloconica in the larval maxillary galea to eight sugars. These sugars were chosen because they are mostly found in host-plants of H. armigera or are representative in the structure and source of sugars.”

In addition to this, there are several specific comments in the detailed reviewers comments below, which the authors could consider responding to.

Reviewer #1 (Recommendations For The Authors):

The article titled "Sucrose taste receptors exhibit dissimilarities between larval and adult stages of a moth" by Shuai-Shuai Zhang and colleagues provides an intriguing analysis. The authors have conducted a meticulously planned and executed study. However, I do have some inquiries.

(1) What precisely does the term "differ" signify in the title? It can be expounded upon in terms of differing in expression or sensitivity. The title could benefit from being more informative. The authors should appropriately specify the insect species in the title of the paper. This would make it more comprehensible to readers. Merely mentioning the term "moth" does not provide any information about the model organism. Hence, it would be preferable to mention Helicoverpa armigera instead of using the generic term "moth" in the title.

Thank you for your suggestions. We considered it better to emphasize that the receptors for sucrose are different, and we have accepted the suggestion of adding the name of the animal. The title has been changed into “The larva and adult of Helicoverpa armigera use differential gustatory receptors to sense sugars”.

(2) The abstract is written in a simple and easily understandable manner, but it overlooks important findings from a technical standpoint.

We add some key experimental techniques to illustrate some important findings in the Abstract.

(3). Almost all herbivorous insects are said to consume plants and utilize sucrose as a stimulus for feeding, as stated by the authors. Sucrose, glucose, and fructose sugar are among the commonly observed stimulants for feeding in numerous insects. It would be appropriate to incorporate not only sucrose but also glucose and fructose as feeding stimulants for almost all herbivorous insects.

Thank you for your suggestion. Sucrose is the major sugar in plants, and its concentration varies greatly from tissue to tissue, while the concentration of the hexose sugars is much lower and the concentration does not change much. In Line 48, we state that sucrose, glucose, and fructose are feeding stimuli for herbivorous insects. From the previous studies, it seems that sucrose is the strongest, followed by fructose, and finally glucose. The cotton bollworm larvae showed no electrophysiological and behavioral response to glucose.

(4) The reason why trehalose is not considered in the electrophysiology analysis is unclear. Given that trehalose is a major sugar in insects and plants, it would be intriguing to include it in the analysis.

We have accepted the reviewer's suggestion, and supplemented the electrophysiological responses of taste organs in larvae and adults of Helicoverpa armigera to trehalose (Figure 1, Figure 1-Figure Supplement 1), and also tested the behavioral responses of the larvae and adults to trehalose (Figure 2, Figure 2-Figure Supplement 1). Therefore, all the related figures have been changed.

(5) The author's intention regarding the co-receptor relationship between Gr5 and Gr6 (line 211) is unclear. If this is indeed the case, then the reason for considering Gr5 in further studies remains uncertain.

We have changed the sentence as follows: “Since Gr5 was highly expressed with Gr6 in the proboscis and tarsi (Figure 3D-3E, Figure 3—figure supplement 1), we suspected that Gr5 and Gr6 might be expressed in the same cells, and then tested the response profile of their co-expression in oocytes.”

(6) The homologous nature of Grs is emphasized by the authors. It is not specified how the author ensured that the guide RNA targeting Gr6 or Gr10 did not result in off-target effects on other Grs.

Thank you so much for your suggestion. We have rewritten the relevant paragraph (Line 238-251), detailing our tests and the results on the potential off-target effects of knocking out GRs by CRISPR/Cas9: “In order to predict the potential off-target sites of sgRNA of Gr6 and Gr10, we used online software Cas-OFFinder (http://www.rgenome.net/cas-offinder/) to blast the genome of H. armigera, and the mismatch number was set to less than or equal to 3. According to the predicted results, the Gr10 sgRNA had no potential off-target region but Gr6 sgRNA had one. Therefore, we amplified and sequenced the potential off-target region of Gr6-/- and found there was no frameshift or premature stop codon in the region compared to WT (Figure 5—figure supplement 2). It is worth mentioning that there was no potential off-target region of Gr6 and Gr10 sgRNA in other sugar receptor genes of H. armigera, Gr4, Gr5, Gr7, Gr8, Gr9, Gr11 and Gr12. We further found there was no difference in the response to xylose of the medial sensilla styloconica among WT, Gr10-/- and Gr6-/- (Figure 5—figure supplement 2). Furthermore, WT, Gr10-/- and Gr6-/- did not show differences in the larval body weight, adult lifespan, and number of eggs laid per female (Figure 5—figure supplement 2). All these results suggest that no off-target effects occurred in the study.”

(7) Is it possible that knocking out Gr10 is not compensated for by the overexpression of Gr6 or other sucrose sensing Grs? Similarly, would the vice versa scenario hold true?

In the Discussion section, we have added some sentences to discuss this issue: “From our results, knocking out Gr10 or Gr6 is unlikely to be compensated by overexpression of other sugar GRs. One of our recent studies showed that Orco knockout had no significant effect on the expression of most OR, IR and GR genes in adult antennae of H. armigera, but some genes were up- or down-regulated (Fan et al., 2022).”

(8) What was the rationale for selecting nine candidate GR genes for expression analysis?

Based on the reviewer's suggestion, we expanded the relevant paragraphs to illustrate the rationale for selecting nine candidate GR genes for expression analysis: “To reveal the molecular basis of sugar reception in the taste sensilla of H. armigera, we first analyzed the putative sugar gustatory receptor genes based on the reported gene sequences of GRs in H. armigera and their phylogenetic relationship of D. melanogaster sugar gustatory receptors (Jiang et al., 2015; Pearce et al., 2017; Xu et al., 2017). Nine putative sugar GR genes, Gr4–12 were identified, and their full-length cDNA sequences were cloned (The GenBank accession number is provided in Appendix—Table S1).” (Line 155-161)

(9) What is the potential reason for the difference between the major larval sugar receptors of Drosophila and Lepidopterans?

The difference between the major larval sugar receptors of Drosophila and Lepidopterans is probably due to differences in the food their larvae feed on. Fruit fly larvae feed on rotten fruit, the main sugar of which is fructose. The larvae of Lepidoptera mainly feed on plants, and the main sugar is sucrose. In the Discussion section, we have added a sentence “This is most likely due to fruit fly larvae feeding on rotten fruits, which contain fructose as the main sugar.” (Line 399-401)

(10) There is a disparity in GRs, specifically GR5 and GR6, between the female antenna, proboscis, and tarsi. What could be the possible justification and significance of this?

Thank you so much for this question. We have added a sentence in the Discussion section, “In this study, the expression patterns of 9 sugar GRs in three taste organs of adult H. armigera show that there is a disparity in GRs, specifically GR5 and GR6, between the female antenna, tarsi and proboscis, which may be an evolutionary adaptation reflecting subtle differentiation in the function of these taste organs in adult foraging. Antennae and tarsi play a role in the exploration of potential sugar sources, while the proboscis plays a more precise role in the final decision to feed.” (Line 433-438)

(11) I suggest that a visual representation illustrating the positioning of GSNs, particularly the lateral and medial sensilla, in both larva and adult stages would enhance the correlation with the results.

In Figure 1 we added the photo of each taste organ and the position of the recorded sensilla, and also added a new figure, Figure 8 summarizing the main findings of the studies.

(12) Further experiments can be conducted to elucidate the precise molecular mechanisms, particularly the downstream effects of GRs, in order to establish the specificity of GRs more convincingly.

Thank you so much for your suggestion. We have discussed the further experiments in the Discussion section, “To elucidate the precise molecular mechanisms of sugar reception in H. armigera is necessary to compare a series of single, double and even multiple Gr knock-out lines and investigate the downstream effects of the GRs.” (Line 363-369)

(13) Figure 6 caption: In Figure 6 (D to I), the percentage of PER is depicted. There is redundancy in the Y-axis title (Percentage of PER) and the legend. This appears to be repetitive. I suggest that it would be better to include the Y-axis title only in Figure D or in Figures D and G.

We accept the suggestion. Figure 7 (not Figure 6) has been revised accordingly.

(14) In Figures 6A and 6C, there is inconsistency in the colors used for WT, Gr6, and Gr10. This could potentially confuse the reader. I recommend using the same colors in both figures instead of using a blue color. Please specify how the authors calculated the feeding area in Figure 6.

We accept the reviewer's suggestion and have changed the color of Figure 7A, B. We have also added the detail method for calculating feeding area (Line 541-545).

(15) In Two-choice tests, why did the authors use 0.01% Tween 80? Please provide comments on this.

Use of 0.01% Tween 80 is to reduce the surface tension and increase the malleability of the solution. We have given detailed explanation in the Method section and cite the reference. (Line538-540)

(16) It would be valuable if the authors could comment on the prospects of this study, considering that GRs play a vital role in controlling behavior and developmental pathways. What are the potential consequences of blocking or disrupting these receptors in terms of behavioral and developmental phenotypic deformities? Could this potentially lead to increased insect mortality?

Thank you so much for your suggestions. In the last paragraph of the Discussion section, we have added the following perspectives, “Knockout of Gr10 or Gr6 led to a significant decrease in sugar sensitivity and food preference of the larvae and adults of H. armigera, respectively, which is bound to bring adverse consequences to survival and reproduction of the insects. Therefore, studying the molecular mechanisms underlying sugar perception in phytophagous insects may provide new insights into the behavioral ecology of this important and highly diverse group of insects, and measures blocking or disrupting sugar receptors could also have applications to control agricultural pests and improve crop yields worldwide” (Line 449-456).

Reviewer #2 (Recommendations for The Authors):

There are a few comments, that I feel would be beneficial to be addressed.

• The authors used 7 different sugars for their experimental approach. While I agree that this is a sufficiently large collection for a study, I was wondering why they specifically chose these sugars; an explanatory section might be helpful for a reader to follow the reasoning.

According to reviewer 1's suggestion, we increased trehalose to 8 sugars in experiments. Trehalose is a main sugar in insect blood. It is converted by insects after feeding on plant sugars. The 8 sugars were chosen because they are present in host-plants of H. armigera or are representative in the structure and source of sugars. They contain 2 pentoses (arabinose and xylose), 4 hexoses (fructose, fucose, galactose and glucose), and 2 disaccharides (sucrose and trehalose).

• It might be beneficial to provide some broader overview on the gustatory system in the cotton bollworm, particularly at the larval stage since this may not be common knowledge. Along these lines eg. the complexity of sensilla types, organs and overall number (or estimation) of neurons might be good to know, a graphical representation of the sense organs might be informative.

In the Introduction section, we give a more specific description on sugar sensitive GSNs in the taste system of the larva and adult of H. armigera, and cite the corresponding references.

• Concerning phylogeny of GRs, it might be relevant to know how complete the genome information is and some more general background on GR diversity in the cotton bollworm.

We agree to your opinion. According to this idea, we got the putative sugar GRs from the previously published genome (Pearce et al. 2017) and the related annotation of GRs (Jiang et al. 2015, Xu et al. 2012). We have made a more detailed explanation about this in the new version of the manuscript, “We first analyzed the putative sugar gustatory receptor genes based on the genome data of H. armigera (Pearce et al. 2017), the reported gene sequences of sugar GRs in H. armigera and their phylogenetic relationship of D. melanogaster sugar gustatory receptors (Jiang et al. 2015, Xu et al. 2012). All nine putative sugar GR genes in H. armigera, Gr4–12 were validated, and their full-length cDNA sequences were cloned (The GenBank accession number is provided in Appendix—Table S1).” (Line 155-161).

• Generation of mutants based on CRISPR is intriguing and a powerful step. While the techniques are well described in the method section, there is no information concerning efficiency or broader feasibility of the approach. I feel it would be quite interesting to learn about how feasible or laborious the approach is to generate mutants (e.g. number of initial injected eggs, the resulting F0 offspring, number of back-crosses, number of screened F1s ....).

In the Materials and Methods section, we have added specific success rates for each step in the process of building the two mutants (Line 722-726, 729-732).

Reviewer #3 (Recommendations For The Authors):

I want to congratulate the authors on this very nice study and have only minor comments for them.

(1) It would be very nice to include pictures of the larva and adult of H. armigera. It would also help to have schematics of where the sensilla they are recording from are.

We have added photos of four taste organs on which the recoded sensilla were indicated (Figure 1), and picture of the larva and adult on which the stimulating site was indicated (Figure 2).

(2) A schematic summarising their findings, including the relevance to the animal's behavioural ecology, will greatly improve interpretations for the broader audience.

A schematic summarizing the findings has been added.

(3) The manner in which PIs are represented in figure 2A, B (among others) is confusing. Can the authors please plot the PI and not the feeding area? From the PI values listed beside the plot, it actually suggests that the larvae don't really show a preference. Could the authors please comment on this?

Yes, sucrose has a significant stimulating effect on larva feeding, but the effect is not as large as the predicted based on the sensitivity of the sensillum, the main reasons are as follows: (1) there are many factors affecting larva feeding, sucrose is only one of them; (2) due to the substrate leaf discs also contain sugar, the effect of newly added sucrose may be reduced. After careful consideration, we think it is better to display the feeding area and PI together so that readers have a complete understanding of the data.

(4) The heterologous expression experiments suggest that co-expression of GR6 with either GR10 or GR5 somehow suppress the response of the GR6 alone to fucose. Am I reading the data correctly? Why would this be? Perhaps the authors could discuss this. In this context, it would help to reproduce all the GR6 data together.

Your interpretation is reasonable to a certain extent. The result of co-injection might be that Gr10 or Gr5 inhibited the response of Gr6. However, there is another possibility that the amount of Gr6 sRNA was diluted by co-injection of two GRs, resulting in a reduced response of Gr6 to fucose.

(5) In general, for each results section, it would help to have a sentence or two that interprets the data in the context of previously presented data. This would help the reader digest the data and interpret it as they read along. Currently, the authors summarise the observations and leave all the interpretation to the discussion section.

We accept the suggestion. In each part of the results, we have added a sentence to explain the above data, which will help readers to clarify the context of the research more easily.

(6) Is the GR6 data in 4C not lined up correctly?

Yes, it is right.

(7) Line 228 suggests that the mutants were validating with qPCRs - I don't see that data.

The mutants were not validating with qPCR. We used the ordinary PCR technology at the mRNA level to verify whether the related sequences were really deleted in the mutants.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors present a detailed study of a nearly complete Entomophthora muscae genome assembly and annotation, along with comparative analyses among related and non-related entomopathogenic fungi. The genome is one of the largest fungal genomes sequenced, and the authors document the proliferation and evolution of transposons and the presence/absence of related genetic machinery to explore how this may have occurred. There has also been an expansion in gene number, which appears to contain many "novel" genes unique to E. muscae. Functionally, the authors were interested in CAZymes, proteases, circadian clock related genes (due to entomopathogenicity/ host manipulation), other insect pathogenspecific genes, and secondary metabolites. There are many interesting findings including expansions in trahalases, unique insulinase, and another peptidase, and some evidence for RIP in Entomophthoralean fungi. The authors performed a separate study examining E. muscae species complex and related strains. Specifically, morphological traits were measured for strains and then compared to the 28S+ITSbased phylogeny, showing little informativeness of these morpho characters with high levels of overlap.

This work represents a big leap forward in the genomics of non-Dikarya fungi and large fungal genomes. Most of the gene homologs have been studied in species that diverged hundreds of millions of years ago, and therefore using standard comparative genomic approaches is not trivial and still relatively little is known. This paper provides many new hypotheses and potential avenues of research about fungal genome size expansion, entomopathogenesis in zygomycetes, and cellular functions like RIP and circadian mechanisms.

Strengths:

There are many strengths to this study. It represents a massive amount of work and a very thorough functional analysis of the gene content in these fungi (which are largely unsequenced and definitely understudied). Too often comparative genomic work will focus on one aspect and leave the reader wondering about all the other ways genome(s) are unique or different from others. This study really dove in and explored the relevant aspects of the E. muscae genome.

The authors used both a priori and emergent properties to shape their analyses (by searching for specific genes of interest and by analyzing genes underrepresented, expanded, or unique to their chosen taxa), enabling a detailed review of the genomic architecture and content. Specifically, I'm impressed by the analysis of missing genes (pFAMs) in E. muscae, none of which are enriched in relatives, suggesting this fungus is really different not by gene loss, but by its gene expansions.

Analyzing species-level boundaries and the data underlying those (genetic or morphological) is not something frequently presented in comparative genomic studies, however, here it is a welcome addition as the target species of the study is part of a species complex where morphology can be misleading and genetic data is infrequently collected in conjunction with the morphological data.

Thank you for your careful reading of our work. We’re glad that you identified these areas as strengths.

Weaknesses:

The conclusions of this paper are mostly well supported by data, but a few points should be clarified.

In the analysis of Orthogroups (OGs), the claim in the text is that E. muscae "has genes in multi-species OGs no more frequently than Enotomophaga maimaiga. (Fig. 3F)" I don't see that in 3F. But maybe I'm really missing something.

Thank you for catching this. You were, in fact, not missing anything at all. There was a mismatch between the data plotted in F and G and how the caption described these data. We very much apologize for the confusion that this must have caused. We have corrected these plots and also made changes to improve interpretability (see below).

Also related, based on what is written in the text of the OG section, I think portions of Figure 3G are incorrect/ duplicated. First, a general question, related to the first two portions of the graph. How do "Genes assigned to an OG" and "Genes not assigned to an OG" not equal 100% for each species? The graph as currently visualized does not show that. Then I think the bars in portion 3 "Genes in speciesspecific OG" are wrong (because in the text it says "N. thromboides had just 16.3%" species-specific OGs, but the graph clearly shows that bar at around 50%. I think portion 3 is just a duplicate of the bars in portion 4 - they look exactly the same - and in addition, as stated in the text portion 4 "Potentially speciesspecific genes" should be the simple addition of the bars in portion 2 and portion 3 for each species.

As mentioned above, we sincerely regret the error made in the plot and for the confusion that this caused. F now reflects the percentage of orthogroups (OGs) that possess at least one representative from the indicated species (left) and the percentage of OGs that are species-specific (only possess genes from one species; right). The latter is a subset of the former. G now reflects the percentage of annotated genes that were assigned an OG, per species, as well as the inverse of this - genes that were not assigned to any OG. These should, and now do, sum to 100%. The “Within species-specific OG” data summed with the “Not assigned OG” data yields the “Potentially species-specific data” in the rightmost column.

In the introduction, there is a name for the phenomenon of "clinging to or biting the tops of plants," it's called summit disease. And just for some context for the readers, summit disease is well-documented in many of these taxa in the older literature, but it is often ignored in modern studies - even though it is a fascinating effect seen in many insect hosts, caused by many, many fungi, nematodes (!), etc. This phenomenon has evolved many times. Nice discussions of this in Evans 1989 and Roy et al. 2006 (both of whom cite much of the older literature).

You’re right. We have now clarified that this behavior is called “summit disease” and referenced the suggested articles, along with a more recent review.

Reviewer #2 (Public Review):

In their study, Stajich and co-authors present a new 1.03 Gb genome assembly for an isolate of the fungal insect parasite Entomophthora muscae (Entomophthoromycota phylum, isolated from Drosophila hydei). Many species of the Entomophthoromycota phylum are specialised insect pathogens with relatively large genomes for fungi, with interesting yet largely unexplored biology. The authors compare their new E. muscae assembly to those of other species in the Entomophthorales order and also more generally to other fungi. For that, they first focus on repetitive DNA (transposons) and show that Ty3 LTRs are highly abundant in the E. muscae genome and contribute to ~40% of the species' genome, a feature that is shared by closely related species in the Entomophthorales. Next, the authors describe the major differences in protein content between species in the genus, focusing on functional domains, namely protein families (pfam), carbohydrate-active enzymes, and peptidases. They highlight several protein families that are overrepresented/underrepresented in the E. muscae genome and other

Entomophthorales genomes. The authors also highlight differences in components of the circadian rhythm, which might be relevant to the biology of these insect-infecting fungi. To gain further insights into E. muscae specificities, the authors identify orthologous proteins among four Entomophthorales species. Consistently with a larger genome and protein set in E. muscae, they find that 21% of the 17,111 orthogroups are specific to the species. To finish, the authors examine the consistency between methods for species delineation in the genus using molecular (ITS + 28S) or morphological data (# of nuclei per conidia + conidia size) and highlight major incongruences between the two.

Although most of the methods applied in the frame of this study are appropriate with the scripts made available, I believe there are some major discrepancies in the datasets that are compared which could undermine most of the results/conclusions. More precisely, most of the results are based on the comparison of protein family content between four Entomophthorales species. As the authors mention on page 5, genome (transcriptome) assembly and further annotation procedures can strongly influence gene discovery. Here, the authors re-annotated two assemblies using their own methods and recovered between 30 and 60% more genes than in the original dataset, but if I understand it correctly, they perform all downstream comparative analyses using the original annotations. Given the focus on E. muscae and the small sample size (four genomes compared), I believe performing the comparisons on the newly annotated assemblies would be more rigorous for making any claim on gene family variation.

Thank you for this comment. While we did compare gene model predictions for two of these assemblies to assess if this difference could account for discrepancies in gene counts, completely reannotating all non-E. muscae datasets was outside of the scope of this study. In our opinion, the total number of predicted genes in a genome is not a best representation of differences since splitting or fusing gene models can inflate seeming differences; the orthology and domain counts are a more accurate assessment of the content. It’s possible that annotation differences may have inflated some gene family counts, however we will note that similar domain trends were observed between the closest species to E. muscae, Entomophaga maimaiga, suggesting that these differences were not sufficient to prevent us from detecting real biological signals. We look forward to continued improvement of our genome through additional sequencing and more clarity on total gene content of E. muscae.

The authors also investigate the putative impact of repeat-induced point mutation on the architecture of the large Entomophthorales genomes (for three of the eight species in Figure 1) and report low RIP-like dinucleotide signatures despite the presence of RID1 (a gene involved in the RIP process in Neurospora crassa) and RNAi machinery. They base their analysis on the presence of specific PFAM domains across the proteome of the three Entomophthorales species. In the case of RID1, the authors searched for a DNA methyltransferase domain (PF00145), however other proteins than RID1 bear such functional domain (DNMT family) so that in the current analysis it is impossible to say if the authors are actually looking at RID1 homologs (probably not, RID1 is monophyletic to the Ascomycota I believe). Similar comments apply to the analysis of components of the RNAi machinery. A more reliable alternative to the PFAM analysis would be to work with full protein sequences in addition to the functional domains.

While we understand this concern regarding domain vs. full length protein, the advantage of the domain search is that HMM-based searches are sensitive to detecting more distantly related homologs. Entomophthoralean fungi are distantly related from the ascomycetes in which these mechanisms have been characterized, so we chose a broader search approach that may identify proteins with similar domain structure, but are not necessarily homologs. These searches are presented in the manuscript as preliminary, but worth further investigation. However, our RID-based analysis did not identify convincing homologs for RID1 in entomophthoralean fungi included in our investigation, and we reported low homology (i.e., 12-14%) among our orthogroup of interest and RID1. We have further edited this section to clarify our understanding that these candidates are not RID1 homologs. We had hoped to avoid this implication, but we felt this investigation and null result were worth reporting.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Specific points:

Results:

"1.03 Gb genome consisting of 7,810 contigs (N50 = 301.1 kb). Additional... resulted in a final contig count of 7,810 (N50 = 329.6 kb)" So you started and ended with the same contig count but a different N50? Is this a typo?

Yes, this was a typo. Thank you for bringing this to our attention.

Figure 1D.

The colors of Complete1x and Complete2x are too similar to tell them apart.

The colors have been made more distinct.

Figure 4B.

I know C. rosea has been found from insects before, but it's mostly a mycoparasite and occasionally an endophyte, and has bioactivity against a lot of things. I just saw that it's listed as an entomopathogen, and I was surprised. Anyway, leave it as is if you want to, but it's definitely better studied and better known (Google Scholar) as a mycoparasite.

Thanks for this comment. For the sake of including a more diverse representation of entomopathogenic fungi, we have opted to leave this as is.

Full references (from the public comment)

Evans, H.C., 1989. Mycopathogens of insects of epigeal and aerial habitats. Insect-fungus interactions, pp.205-238.

Roy, H.E., Steinkraus, D.C., Eilenberg, J., Hajek, A.E. and Pell, J.K., 2006. Bizarre interactions and endgames: entomopathogenic fungi and their arthropod hosts. Annu. Rev. Entomol., 51, pp.331-357.

Reviewer #2 (Recommendations For The Authors):

I believe the manuscript could largely benefit from restructuring the results section to enhance clarity. The results section reads like a lot of descriptive back and forth, so that the reader lacks a clear rationale. The absence of a consistent dataset used for the different comparisons made all along the manuscript makes it hard to follow.

Minor comments:

(No line numbers were available so I refer to page numbers).

p1

• not sure about the use of "allied" to describe other fungal species in the title and after (sister species?).

We didn’t want to use the word sister because not all of these species could be considered sister.

• Genomic defence against transposable elements rather than "anti"?

We have rephrased to genomic defense.

p3

• Extra parenthesis at Bronski et al.

This is now corrected.

• What does newly-available mean here?

We mean recent. A lot of the datasets we used were very new, and we wanted to emphasize that point.

• The back and forth between genomes and transcriptomes makes it hard to follow, would clarify from the beginning (in addition to the sequencing method - short vs long-read assemblies as in Figure 1B) or perhaps use a consistent dataset for all subsequent comparative analysis in the Entomophthorales.

We have denoted our transcriptomic datasets in Fig 1C using parentheses.

p5

• Perhaps clarify that class II DNA transposons can also "copy" (single-strand excisions can be repaired by the host machinery).

We have now included mention of “copy” as well as “jump” mechanisms of Class II transposons per your suggestion.

p6

• "beginning roughly concurrently", not clear what "began".

This is now corrected.

• "control" rather than "protect against"?

We’ve changed “protect against” to “counter”.

• I believe RIP has only been observed (experimentally) in a handful of fungal species, all from the Ascomycota phylum.

Hood et al, 2005 found signatures of RIP in anther-smut fungus and Horns et al, 2012, found evidence of hypermutability across repeat elements within several Pucciniales species.

• "RID1 contains two DNA_methylase domains", RID1 has one methyltransferase domain according to the reference Freitag et al, 2002.

Thank you for drawing this to our attention. It is true RID1 has one methyltransferase region; however, the sequence deposited by Freitag et al, 2002 (AAM27408) is predicted by HMMer to have two adjacent Pfam DNA_methylase domains (i.e., PF00145). In this exploratory analysis, we tried to leverage this characteristic to identify candidate proteins of interest. We have reworded this section to clarify this.

p8

• Here and after I would use more informative titles for each paragraph.

With the exception of the headings for Pfam, CAZy and MEROPs analyses, we believe the other headings are informative. We appreciate this comment, but opt to leave the heading titles as is.

• I believe presenting the orthology analysis before the more in-depth protein family domain search.

We leveraged the OG analysis mostly as a way to identify potentially unique genes in E. muscae, so we think the current order makes the most sense.

p10

• Figures 3F and G are confusing. The legend for Figure 3F mentions "OGs with >= 2 species" while the figure shows "multi-species OGs", and reads as redundant with the "species-specific" OGs. For the "OGs within species" do I understand it correctly that it represents the number of genes assigned to OGs for each species? If yes, the numbers are in contradiction with Figure 3G. And in Figure 3G shouldn't the sum of "genes assigned in OGs" and "genes nor assigned in OGs" add up to 100? I'm probably missing something here, but I would clarify what the different sets of orthogroups are in the figure and in the text (perhaps adopting a pangenome-like nomenclature).

Thanks for this comment. This legend, unfortunately, reflected an earlier version of the figure and was overlooked prior to submission. We have since amended this and sincerely apologize for the error on our part.

p12

• The whole first paragraph reads more like it should be part of an introduction/discussion.

We’ve moved some of this paragraph to the discussion but left the background information necessary for the reader to understand why we were looking for homologs of wc and frq.

p13

• The last paragraph reads like discussion.

We have revised this paragraph so it now reads: “Because E. muscae is an obligate insect-pathogen only living inside live flies, we investigate the presence of canonical entomopathogenic enzymes in the genome. We find that E. muscae appear to have an expanded group of acid-trehalases compared to other entomopathogenic and non-entomopathogenic Entomophthorales (Fig. 4A), which correlates with the primary sugar in insect blood (hemolymph) being trehalose (Thompson, 2003). The obligate insectpathogenic lifestyle is also evident when comparing the repertoire of lipases, subtilisin-like serine proteases, trypsins, and chitinases in our focal species versus Zoopagomycota and Ascomycota fungi that are not obligate insect pathogens (Fig. 4B). Sordariomycetes within Ascomycota contains the other major transition to insect-pathogenicity within the kingdom Fungi (Araújo and Hughes, 2016). Based on our comparison of gene numbers, Entomophthorales possess more enzymes suitable for cuticle penetration than Sordariomycetes (Fig. 4B). In contrast, insect-pathogenic fungi within Hypocreales possess a more diverse secondary metabolite biosynthesis machinery as evidenced by the absence of polyketide synthase (PKS) and indole pathways in Entomophthorales (Fig. 4C).”

p15 and 16

• This all reads as redundant with the previous protein family domain analysis. I would try to merge them.

Thank you for this comment, however we have opted to maintain the current structure.

p18

• In the first sentence, I'm not sure about what was performed here.

This has been reworded to clarify.

p20

• Regarding the assembly, do I understand it correctly that a nuclear genome can be partially haploid / diploid?

Thanks for your comment. The genome itself is, of course, some integer multiple of n, but based on BUSCO scores our assembly doesn’t appear to have completely collapsed into a haploid genome. We think it makes more sense here to say “partially haploid” than “partially diploid” so have altered this.

p21

• RIP has only been observed in a couple of Ascomycetes. RIP-like genomic signatures (GC bias) have been observed elsewhere.

Hood et al, 2005 found signatures of RIP in anther-smut fungus and Horns et al, 2012, found evidence of hypermutability across repeat elements within several Pucciniales species.

p23

• Interesting that the peptidase A2B domain is found uniquely in E. muscae genome and is associated with Ty3 activity. Does the domain often overlap with annotated Ty3 in E. muscae genome? Or how come the domain is not present in other sister species with large genomes full of Ty3 transposons? Could it relate to a new active transposon in E. muscae specifically?

Thanks for this comment. The domain-based analysis was only performed on the predicted transcriptome of the genome assembly, which does not include the repeat elements (e.g., Ty3). It could be that this peptidase reflects a new active transposon that’s specific to E. muscae, which would certainly be very interesting. We’ve now included this idea in the discussion.

p26

• In the case of fungal genomes, I would not advise masking the assembly for repeated sequences prior to gene annotation (in particular given the current focus on protein family variation).

Thank you for this comment, however we disagree with this assertion as a typical approach for genome annotation in fungi and eukaryotic genomes is to use soft masking of transposable elements before performing gene prediction to avoid over-prediction. While there could be alternative approaches that compare masked or unmasked. This is a recommended protocol for underlying tools like Augustus (10.1002/cpbi.57) and in general descriptions of genome annotation (10.1002/0471250953.bi0401s52). The false positive rate of genes predicted through TE regions is likely to be more a problem than false negatives of missed genes in our experience. Further it seems appropriate to use consistent approach to annotation throughout when including genomes from other sources (e.g., Joint Genome Institute annotated genomes) which also use a repeat masking approach first before annotation. It seems most appropriate to use consistent methods when generating datasets to be used for comparative analyses. It is outside the scope of this project to reannotate all genomes with and without repeat masking.

p27

• Interrupted sentence at "Classification of DNA and LTR .. by similarity The".

This was an unnecessary partial phrase as the information on classification of elements via RepBase was made a few sentences above this.

p28

• Enriched/depleted rather than "significantly different"?

Thank you for this comment, however we have opted to maintain the current phrasing.

Author response:

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

We thank the reviewers for a careful review of the manuscript and for their comments, which we address below.

Reviewer #1:

(1) …the authors could examine division in a population of cells with only one centrosome. Seeing some restoration of mitotic progression in the absence of SAC-dependent delays would suggest that even one centrosome with uninhibited Eg5 is sufficient to negate SAC-dependent delays, and would limit models for what exactly centrosomes contribute.

We agree that the one-centrosome question (i.e. whether cells with a single centriole, and therefore a single centrosome, have the same SAC dependence) would be interesting to address. It is known that cells with a single centriole generated through centrinone treatment also have elongated mitoses, like cells lacking centrioles (see Chinen, et. al. 2021, compare Fig 2C to Fig 2D), We have tried this experiment in RPE-1 cells with preliminary results confirming that there is a mitotic delay. It is not known whether this delay requires SAC activity, and we hope to address that in future work. In addition, we note that we show in Fig. 4b-c that cells with the normal centrosome number but with a single focus of microtubules due to Eg5 inhibition, were also sensitive to MPS1 inhibition. This suggests that centrosome presence alone cannot overcome the requirement for SAC activity, rather, the centrosomes need to be able to separate in a timely fashion.

Reviewer #2:

(1) An example is how to interpret the effect of Aurora B inhibition, which does not block acentrosomal cell division. If Aurora B is required for SAC activity, it suggests this effect of MPS1 may be a function other than SAC. Given the complexity of the SAC, it would be informative to test other SAC components. Instead, the authors conclude that the mitotic delay caused by MPS is required for acentrosomal cell division. I don't think they have ruled out, or even addressed other functions of MPS1.

We agree that it is possible that functions of the MPS1 kinase other than those involved in the SAC could be important. Although we have not directly tested other SAC components, we did “mimic” SAC activity by delaying anaphase onset using APC/C inhibition while also inhibiting MPS1 (Fig. 2b-b’’). The fact that this restored division suggests that it is the SAC function of MPS1 kinase activity that is relevant to this delay. 

(2) The authors find that when both the APC and MPS1 are inhibited, the cells eventually divide. These results are intriguing, but hard to interpret. The authors suggest that the failure to divide in MPS1-inhibited cells is because they enter anaphase, and then must back out. This is hard to understand and there is not data supporting some kind of aborted anaphase. Is the division observed with double inhibition some sort of bypass of the block caused by MPS1 inhibition alone? It is not clear why inhibition of APC causes increased cell division when MPS1 is inhibited.

As described in the response to 1), we believe that reinstating the delay to anaphase onset by APC/C inhibition provided the time needed to establish a functional bipolar spindle even in the absence of the SAC, and that cells eventually overcome the proTAME block and proceed through mitosis, as observed in control cells in our experiments. We note that we chose concentrations of proTAME specifically for each cell line (RPE-1 and U2OS) that would result only in a temporary block, following on the work of Lara-Gonzalez and Taylor (2012), who reported similar findings for HeLa cells.

(3) The authors characterize MTOC formation in these cells, which is also interesting. MTOCs are established after NEB in acentrosomal cells. Indeed, forming these MTOCs is probably a key mechanism for how these cells complete a division, like mouse oocytes.

We agree that the observed intermediates of MTOCs are interesting and likely crucial to the mechanism of cell division in acentrosomal somatic cells. We are investigating further the differences and similarities between somatic cell MTOC formation in the absence of centrosomes and the naturally-occurring form of that process in oocytes.

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

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1

Evidence, reproducibility and clarity

Seleit and colleagues set out to explore the genetics of developmental timing and tissue size by mapping natural genetic variation associated with segmentation clock period and presomitic mesoderm (PSM) size in different species of Medaka fish. They first establish the extent of variation between five different Medaka species of in terms of organismal size, segmentation rate, segment size and presomitic mesoderm size, among other traits. They find that these traits are species-specific but strongly correlated. In a massive undertaking, they then perform developmental QTL mapping for segmentation clock period and PSM size in a set of ~600 F2 fish resulting from the cross of Orizyas sakaizumii (Kaga) and Orizyas latipes (Cab). Correlation between segmentation period and segment size was lost among the F2s, indicating that distinct genetic modules control these traits. Although the researchers fail to identify causal variants driving these traits, they perform proof of concept perturbations by analyzing F0 Crispants in which candidate genes were knocked out. Overall, the study introduces a completely new methodology (QTL mapping) to the field of segmentation and developmental tempo, and therefore provides multiple valuable insights into the forces driving evolution of these traits.

Major comments: - The first sentence in the abstract reads "How the timing of development is linked to organismal size is a longstanding question". It is therefore disappointing that organismal size is not reported for the F2 hybrids. Was larval length measured in the F2s? If so, it should be reported. It is critical to understand whether the correlation between larval size and segmentation clock period is preserved in F2s or not, therefore determining if they represent a single or separate developmental modules. If larval length data were not collected, the authors need to be more careful with their wording.

The question the reviewer raises here is indeed a very relevant one, and a question that we also were curious about ourselves. While it was not possible (logistically) to grow the 600 F2 fish to adulthood, we did measure larval length in a subset of F2 hatchling (n=72) to ask precisely the question the reviewer raises here. Our results (new Supplementary Figure 5) show that the correlation between larval length and segmentation timing (which we report across the Oryzias species) is absent in the F2s. This indeed argues that the traits represent separate developmental modules.

In the current version of the paper, organismal size is often incorrectly equated to tissue size (e.g. PSM size, segment size). For example, in page 3 lines 33-34, the authors state that faster segmentation occurred in embryos of smaller size (Fig. 1D). However, Fig. 1D shows correlation between segmentation rate and unsegmented PSM area. The appropriate data to show would be segmentation rate vs. larval or adult length.

The reviewer is correct. We have now linked the data more clearly to data we show in Supplementary Figure 1, which shows that adult length and adult mass are strongly correlated (S1A) and that adult mass is in turn strongly correlated with segmentation rate in the different Oryzias species (S1B). Additionally main Figure 1B shows that larval length is correlated with PSM length. We have corrected the main text to reflect these relationships more clearly.

• Is my understanding correct in that the her7-venus reporter is carried by the Cab F0 but not the Kaga F0? Presumably only F2s which carried the reporter were selected for phenotyping. I would expect the location of the reporter in the genome to be obvious in Figure 3J as a region that is only Cab or het but never Kaga. Can the authors please point to the location of the reporter?

The reviewer is correct. Indeed the location of our her7-venus KI is on chromosome 16 and the recombination patterns on this chromosome overwhelmingly show either Hom Cab (green) or Het Cab/Kaga (Black). This is expected as we selected fish carrying the her7-venus KI for phenotyping.

• devQTL mapping in this study seems like a wasted opportunity. The authors perform mapping only to then hand pick their targets based on GO annotations. This biases the study towards genes known to be involved in PSM development, when part of the appeal of QTL mapping is precisely its unbiased nature and the potential to discover new functionally relevant genes. The authors need to better justify their rationale for candidate prioritization from devQTL peaks. The GO analysis should be shown as supplemental data. What criteria were used to select genes based on GO annotations?

We have now commented on these valid points and outlined our rationale in more detail in the text (page 4, lines 20-30). Our rationale now also includes selection of differentially expressed genes (n=5 genes) that fall within segmentation timing devQTL hits (for more details see below). Essentially, while we indeed finally focused on the proof of principle using known genes, these genes were previously not known to play a role in either setting the timing of segmentation or controlling the size of the PSM. Hence, we do think our strategy demonstrates the "the potential to discover new functionally relevant genes", even though the genes themselves had been involved overall in somitogenesis. We added the GO analysis as supplemental data as requested (new Supplementary Figure 7E).

• Analysis of the predicted functional consequence of divergent SNPs (Fig. S6B, F) is superficial. Among missense variants, which genes harbor the most deleterious mutations? Which missense variants are located in highly conserved residues? Which genes carry variants in splice donors/acceptors? Carefully assessing the predicted effect of SNPs in coding regions would provide an alternative, less biased approach to prioritize candidate genes.

We now included our analysis of SNPs based on the Variant effect predictor (VEP) tool from ensembl. This analysis does rank the predicted severity of the SNP on protein structure and function (Impact: low, moderate, high) and does annotate which variants can affect splice donors/acceptors. The VEP analysis for both phenotypes is now added to the manuscript as supplemental data (new Supplementary Data S2, S5).

• Another potential way to prioritize candidate genes within devQTL peaks would be to use the RNA seq data. The authors should perform differential expression analysis between Kaga and Cab RNA-seq datasets. Do any of the differentially expressed genes fall within the devQTL peaks?

As suggested we have performed this additional experiment and report the RNAseq differential analysis in new Supplement Figure 7C-D. The analysis revealed 2606 differentially expressed genes in the PSM between Kaga and Cab, five of which were candidate genes from the devQTL analysis. We now tested all of these (5 in total, 4 new and 1 previously targeted adgrg1) for segmentation timing by CRISPR/Cas9 KO in the her7-venus background, none of which showed a timing phenotype (new Supplementary Figure 7F-F'). We provide the complete set of results in new Supplementary Figure 7 , Supplementary Data file 3 (DE-genes), all data were deposited on publicly available repository Biostudies under accession number: E-MTAB-13927.

• The use of crispants to functionally test candidate genes is inappropriate. Crispants do not mimic the effect of divergent SNPs and therefore completely fail to prove causality. While it is completely understandable that Medaka fish are not amenable to the creation of multiple knock-in lines where divergent SNPs are interconverted between species, better justification is needed. For instance, is there enough data to suggest that the divergent alleles for the candidate genes tested are loss of function? Why was a knockout approach chosen as opposed to overexpression?

We agree with the reviewer that we do not address the causality of SNPs with the CRISPR/Cas9 KO approach we followed. And medaka does offer the genome editing capabilities to create tailored sequence modifications. So in principle, this can be done. In practice, however, we reasoned that any given SNP will contribute only partially to the observed phenotypes and combinatorial sequence edits are simply very laborious given the current state of the art in genome editing technologies. We therefore opted for an alternative proof of principle approach that aims to "to discover new functionally relevant genes", not SNPs.

-Along the same line, now that two candidate genes have been shown to modulate the clock period in crispants (mespb and pcdh10b), the authors should at least attempt to knock in the respective divergent SNPs for one of the genes. This is of course optional because it would imply several months of work, but it would significantly increase the impact of the study.

As above, this is in principle the correct rationale to follow though very time, cost and labour intensive. It is for the later practical consideration that we decided not to follow this option.

Minor Comments - It would be highly beneficial to describe the ecological differences between the two Medaka species. For example, do the northern O. sakaizumii inhabit a colder climate than the southern O. latipes? Is food more abundant or easily accessible for one species compared to the other? What, if anything, has been described about each species' ecology?

There are indeed differences in the ecology of both species, with the northern O.sakaizumii inhabiting a colder climate than the southern O. latipes. In addition, it is known that the breeding season is shorter in the north than the south, and also there is the fact that northern species have been shown to have a faster juvenile growth rate than southern species. While it would be premature to link those ecological factors to the timing differences we observe, we can certainly speculate. A line to this effect has been added to the main text (Page 5, line 28-30).

• The authors describe two different methods for quantifying segmentation clock period (mean vs. intercept). It is still unclear what is the difference between Figs. 3A (clock period), S4A (mean period) and S4B (intercept period). Is clock period just mean period? Are the data then shown twice? How do Fig. 3A and S4A differ?

The clock period shown in all the main figures is the intercept period, which was also used for the devQTL analysis. Both measurements (mean and intercept) are indeed highly correlated and we include both in supplement for completeness.

• devQTL as shorthand for developmental QTL should be defined in page 4 line 1 (where the term first appears), not later in line 12 of the same page.

Noted and corrected, we thank the reviewer for spotting this error.

• Python code for period quantification should be uploaded to Github and shared with reviewers.

All period quantification code that was used in this study was obtained from the publicly available tool Pyboat (https://www.biorxiv.org/content/10.1101/2020.04.29.067744v3). All code that is used in PyBoat is available from the Github page of the creator of the tool (https://github.com/tensionhead/pyBOAT). Both are linked in the references and materials and methods sections.

• RNA-seq data should be uploaded to a publicly accessible repository and the reviewer token shared with reviewers.

We have uploaded all RNA-sequencing Data to public repository BioStudies under accession numbers : E-MTAB-13927, E-MTAB-13928. This information is now also added to material and methods in the manuscript text.

Why are the maintenance (27-28C) vs. imaging (30C) temperatures different?

Medaka fish have a wide range of temperatures they can physiologically tolerate, i.e. 17-33. The temperature 30C was chosen for practical reasons, i.e. a slightly faster developmental rate enables higher sample throughput in overnight real-time imaging experiments.

• For Crispants, control injections should have included a non-targeting sgRNA control instead of simply omitting the sgRNA.

We agree a non-targeting sgRNA control can be included, though we choose a different approach. For clarity, we now also include a control targeting Oca2, a gene involved in the pigmentation of the eye to probe for any injection related effect on timing and PSM size. As expected, 3 sgRNAs + Cas9 against Oca2 had no impact on timing or PSM size. This data is now shown in new Supplementary Figure 9 F-G'.

It is difficult to keep track of the species and strains. It would be most helpful if Fig. S1 appeared instead in main figure 1.

We agree and included an overview of the phylogenetic relationship of all species and their geographical locales in new Figure 1 A-B.

Significance

• The study introduces a new way of thinking about segmentation timing and size scaling by considering natural variation in the context of selection. This new framing will have an important impact on the field.
• Perhaps the most significant finding is that the correlation between segment timing and size in wild populations is driven not by developmental constraints but rather selection pressure, whereas segment size scaling does form a single developmental module. This finding should be of interest to a broad audience and will influence how researchers in the field approach future studies.
• It would be helpful to add to the conclusion the author's opinion on whether segmentation timing is a quantitative trait based on the number of QTL peaks identified.
• The authors should be careful not to assign any causality to the candidate genes that they test in crispants.
• The data and results are generally well-presented, and the research is highly rigorous.
• Please note I do have the expertise to evaluate the statistical/bioinformatic methods used for devQTL mapping.

Reviewer #2

Evidence, reproducibility and clarity

Seleit et al. investigate the correlation between segment size, presomitic mesoderm and the rhythm of periodic oscilations in the segmentation clock of developing medaka fish. Specifically, they aim to identify the genetic determinants for said traits. To do so, they employ a common garden approach and measure such traits in separate strains (F0) and in interbreedings across two generations (F1 and F2). They find that whereas presomitic mesoderm and segment size are genetically coupled, the tempo of her7 oscilations it is not. Genetic mapping of the F0 and F2 progeny allows them to identify regions associated to said traits. They go on an perturb 7 loci associated to the segmentation clock and X related to segment size. They show that 2/7 have a tempo defect, and 2/ affect size.

Major comments: The conclusions are convincing and well supported by the data. I think the work could be published as is in its current state, and no additional experiments that I can think of are needed to support the claims in the paper.

Minor comments: - The authors could provide a more detailed characterization of the identified SNPs associated to the clock and to PSM size. For the segmentation clock, the authors identify 46872 SNPs, most of which correspond to non-coding regions and are associated to 57 genes. They narrow down their approach to those expressed in the PSM of Cab Kaga. Was the RNA selected from F1 hybrids? I wonder if this would impact the analysis for tempo and or size in any way, as F2 are derived from these, and they show broader variability in the clock period than the F0 and F1 fishes.

The RNA was obtained from the pure F0 strains and we have now extended this analysis by deep bulk-RNA sequencing and differential gene expression analysis. As indicated also to reviewer 1, this revealed 2606 differentially expressed genes in the unsegmented tails of Kaga and Cab embryos, some of which occurred in devQTL peaks. Based on this information we expanded our list of CRISPR/Cas9 KOs by targeting all differentially expressed genes (5 in total, 4 new and 1 previously targeted) for segmentation timing, none of which showed a timing phenotype (new Supplementary figure 7C-D). We provide the complete set of results in new Supplementary Figure 7, Supplementary Data file 3 (DE-genes). All data were deposited on publicly available repository Biostudies under accession number: E-MTAB-13927.

It would be good if the authors could discuss if there were any associated categories or overall functional relationships between the SNPs/genes associated to size. And what about in the case of timing?

In the case of PSM size there were no clear GO terms or functional relationships between the genes that passed the significance threshold on chromosome 3.

For the 35 genes related to segmentation timing, there were a number of GO enrichment terms directly related to somitogenesis. We have included the GO analysis in the new Supplementary Figure 7E.

• Have any of the candidate genes or regulatory loci been associated to clock defects (57) or segment size (204) previously in the literature?

To the best of our knowledge none of the genes have been associated with clock or PSM size defects so far. It might be worthwhile using our results to probe their function in other systems enabling higher throughput functional analysis, such as newly developed organoid models.

• When the authors narrow down the candidate list, it is not clear if the genes selected as expressed in the PSM are tissue specific. If they are, I wonder if genes with ubiquitous expression would be more informative to investigate tempo of development more broadly. It would be good if the authors could specifically discuss this point in the manuscript.

We have not addressed the spatial expression pattern of the 35 identified PSM genes in this study, so we cannot speculate further. But the reviewer raises an important point, how timing of individual processes (body axis segmentation) are linked at organismal scale is indeed a fundamental, additional, question that will be addressed in future studies, indeed the in-vivo context we follow here would be ideal for such investigations.

Can the authors speculate mechanistically why mespb or pchd10b accelerates the period of her7 oscillations?

While we do not have a mechanistic explanation yet, an additional experiment we performed, i.e. bulk-RNAsequencing on WT and mespb mutant tails, provided additional insight, we now added this data to the manuscript . This analysis revealed 808 differentially expressed genes between wt and mespb mutants. Interestingly, many of these affected genes are known to be expressed outside of the mespb domain, i.e. in the most posterior PSM (i.e. tbxt, foxb1,msgn1, axin2, fgf8, amongst others). This indicates that the effect of mespb downregulation is widespread and possibly occurs at an earlier developmental stage. This requires more follow up studies. This data is now shown in new Supplementary figure 9A, Supplementary Data file S4. We now comment on this point in the revised manuscript.

• Are there any size difference associated to the functionally validated clock mutants?

We addressed this point directly and added this analysis as supplementary Figure 9H-H'. While pcdh10b mutants do not show any detectable difference in PSM size, we find a small, statistically significant reduction in PSM size (area but not length) in mespb mutants. All this data is now included in the revised manuscript.

-Ref 27 shows a lack of correlation between body size and the segmentation period in various species of mammals. The work supports their findings, and it would be good to see this discussed in the text.

We are not certain how best to compare our in-vivo results in externally developing fish embryos to in-vitro mammalian 2-D cell cultures. In our view, the correlation of embryo size, larval and adult size that we find in Oryzias might not necessarily hold in mammalian species, which would make a comparison more difficult. We do cite the work mentioned so the reader is pointed towards this interesting, complementary literature.

Significance

The work is quite remarkable in terms of the multigenerational genetic analysis performed. The authors have analysed >600 embryos from three separate generations to obtain quantitative data to answer their question (herculean task!). Moreover, they have associated this characterization to specific SNPs. Then, to go beyond the association, they have generated mutant lines and identified specific genes associated to the traits they set out to decipher.

To my knowledge, this is the first project that aims to identify the genetic determinants for developmental timing. Recent work on developmental timing in mammals has focused on interspecies comparisons and does not provide genetic evidence or insight into how tempo is regulated in the genome. As for vertebrates, recent work from zebrafish has profiled temperature effects on cell proportions and developmental timing. However, the genetic approach of this work is quite elegant and neat.

Conceptually, it is quite important and unexpected that overall size and tempo are not related. Body size, lifespan, basal metabolic rates and gestational period correlate positively and we tend to think that mechanistically they would all be connected to one another. This paper and Lazaro et al. 2023 (ref 27) are one of the first in which this preconception is challenged in a very methodical and conclusive manner. I believe the work is a breakthrough for the field and this work would be interesting for the field of biological timing, for the segmentation clock community and more broadly for all developmental biologists.

My field is quantitative stem cell biology and I work on developmental timing myself, so I acknowledge that I am biased in the enthusiasm for the work. It should be noted that as an expert on the field, I have identified instances where other work hasn't been as insightful or well developed in comparison to this piece. It is also worth noting that I am not an expert in fish development, phylogenetic studies or GWAS analyses, so I am not capable to asses any pitfalls in that respect.

__Reviewer #3 (Evidence, reproducibility and clarity (Required)): __

__Summary: __

This manuscript explores the temporal and spatial regulation of vertebrate body axis development and patterning. In the early stages of vertebrate embryo development, the axial mesoderm (presomitic mesoderm - PSM) undergoes segmentation, forming structures known as somites. The exact genetic regulation governing somite and PSM size, and their relationship to the periodicity of somite formation remains unclear.

To address this, the authors used two evolutionarily closely related Medaka species, Oryzias sakaizumii and Oryzias latipes, which, although having distinct characteristics, can produce viable offspring. Through analysis spanning parental (generation F0) and offspring (generations F1 and F2) generations, the authors observed a correlation between PSM and somite size. However, they found that size scaling does not correlate with the timing of somitogenesis.

Furthermore, employing developmental quantitative trait loci (devQTL) mapping, the authors identified several new candidate loci that may play a role during somitogenesis, influencing timing of segment formation or segment size. The significance of these loci was confirmed through an innovative CRISPR-Cas9 gene editing approach.

This study highlights that the spatial and temporal aspects of vertebrate segmentation are independently controlled by distinct genetic modular mechanisms.

__Major comments: __

1) In the main text page 3, lines 11 and 12, the authors state that the periodicity of the embryo clock of the F1 generation is the intermediate between the parental F0 lineages. However, the authors look only at the periodicity of the Cab strain (Oryzias latipes) segmentation clock. The authors should have a reporter fish line for the Kaga strain (Oryzias sakaizumii) to compare the segmentation clock of both parental strains and their offspring. Since it could be time consuming and laborious, I advise to alternatively rephrase the text of the manuscript.

We agree a careful distinction between segment forming rate (measured based on morphology) and clock period (measured using the novel reporter we generated) is essential. We show that both measures correlate very well in Cab, in both F0 and F1 and F2 carrying the Cab allele. For Kaga F0, we indeed can only provide the rate of somite formation, which nevertheless allows comparison due to the strong correlation to the clock period we have found. We have rephrased the text accordingly.

2) It is evident that only a few F0 and F1 animals were analyzed in comparison with the F2 generation. Could the authors kindly explain whether and how this could bias or skew the observed results?

We provide statistical evidence through the F-test of equality that the variances between the F0, F1 and F2 samples are equal. Additionally if we sub-sample and separate the F2 data into groups of 100 embryos (instead of all 638) we get the same distribution of the F2s. We therefore believe that this is sufficient evidence against a bias or skew in the results.

3) It would be interesting to create fish lines with the validated CRISPR-Cas9 gene manipulations in different genetic contexts (Cab or Kaga) to analyze the true impact on the segmentation clock and/or PSM & somite sizes.

We agree with the reviewer this would in principle be of interest indeed, please see our response to reviewer 1 earlier.

4) Please add the results of the Go Analysis as supplementary material.

We have added the GO analysis in new Supplementary Figure 7E.

__Minor comments: __

1) In the main text, page 2, line 29, Supplementary Figure 1D should be referenced.

We have added a clearer phylogeny and geographical location of the different species in new Figure 1 A-B. And reference it at the requested location.

2) In the main text, page 2, line 32, the authors refer to Figure 1B, but it should be 1C.

We have corrected the information.

3) Regarding the topic "Correlation of segmentation timing and size in the Oryzias genus" the authors should also give information on the total time of development of the different Oryzias species, as well as the total number of formed somites.

We follow this recommendation and have added this information in new Supplementary Figure 5. We also now include segment number measured in F2 embryos. We indeed view segmentation rate as a proxy for developmental rate, which however needs to be distinguished from total developmental time. The latter can be measured for instance by quantifying hatching time, which we did. These measurements show that Kaga, Cab and O.hubbsi embryos kept at constant 28 degrees started hatching on the same day while O.minutillus and O.mekongensis embryos started hatching one day earlier. We have not included this data in the manuscript because we think a distinction should be made between rate of development and total development time.

4) In Figures 3A and B, please add info on the F1 lines for comparison.

The information on F1 lines is provided in Supplementary Figure 3

5) Supplementary Figures 2F shows that the generation F1 PSM is similar to Cab F0, and not an intermediate between Kaga F0 and Cab F0. This is interesting and should be discussed.

We show that the F1 PSM is indeed closer to the PSM of Cab than it is to the Kaga PSM. This is indeed intriguing and we have now commented on this point directly in the text.

6) Supplementary Figures 6C to H are not mentioned either in the main text or in the extended information. Please add/mention accordingly.

We have added references to both in the text

7) The order of Supplementary Figure 8 E to H and A to D appears to be not correct and not following the flow of the text. Please update/correct accordingly.

We have updated the text accordingly.

8) The authors should choose between "Fig.", "Fig", "fig.", "fig" or "Figure". All 'variants' can be found in the text.

Noted, and updated. Fig. is used for main figures and fig. is used for supplementary figures.

9) The color scheme of several figures (graphs with colored dots) should be revised. Several appear to be difficult to discern and analyze.

We have enhanced the colours and increased the font on the figure panels. The colour panel was chosen to be colour-blind friendly.

10) Please address/discuss following questions: What are the known somitogenesis regulating genes in Medaka? How do they correlate with the new candidates?

The candidates we found and tested had not been implicated in regulating the tempo of segmentation or PSM size, while for some a role in somite formation had been previously established, hence the enrichment in GO analysis Somitogenesis.

Reviewer #3 (Significance (Required)):

General assessment:

This interesting manuscript describes a novel approach to study and find new players relevant to the regulation of vertebrate segmentation. By employing this innovative methodology, the authors could elegantly demonstrate that the segmentation clock periodicity is independent from the sizes of the PSM and forming somites. The authors were further able to find new genes that may be involved in the regulation of the segmentation clock periodicity and/or the size of the PSM & somites. A limitation of this study is the fact that the results mainly rely on differences between the two species. The integration of additional Medaka species would be beneficial and may help uncover relevant genes and genetic contexts.

Advance:

To my best knowledge this is the first time that such a methodology was employed to study the segmentation clock and axial development. Although the topic has been extensively studied in several model organisms, such as mice, chicken, and zebrafish, none of them correlated the size of the embryonic tissues and the periodicity of the embryo clock. This study brings novel technological and functional advances to the study of vertebrate axial development.

Audience:

This work is particularly interesting to basic researchers, especially in the field of developmental biology and represents a fresh new approach to study a core developmental process. This study further opens the exciting possibility of using a similar methodology to investigate other aspects of vertebrate development. It is a timely and important manuscript which could be of interest to a wider scientific audience and readership.

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

Learn more at Review Commons

---

Reply to the reviewers

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

Summary:* In this paper the authors explore the function of Syndecan in Drosophila stem cells focussing primarily on the intestinal stem cells. They use RNAi knockdown to conclude that Syndecan is required for long term stem cell maintenance as its knockdown results in apoptosis. They suggest that this effect is independent of LINC complex proteins but is associated with changes to nuclear morphology and DNA damage. They go on to show that a similar impact on nuclear shape can be seen in larval neuroblasts but not in stem cells of the female germline. *

Major Comments: *The key conclusion that underpins the paper is that reduced Syndecan causes loss of stem cells. This is based entirely on evidence from cell-type specific RNAi using 3 independent RNAi lines. Overexpression has no phenotype and there is no analysis of loss of function mutants. SdcRNAi3 gives strong phenotypes that are statistically significant and is used throughout the paper. SdcRNAi2 gives comparatively moderate phenotypes which trend in the same direction but it is not clear if these are statistically significant (Fig S1). SdcRNAi line 1 appears to have very little effect (and if anything trends in the opposite direction in S1A). In addition, the knockdown efficiency of the three lines has not been assessed. Another possible concern given the dependence on RNAi3 is that the RNAi control line used is not an ideal match for the VDRC GD RNAi lines as it is in a different genetic background. In order to robustly draw conclusions: the phenotypes with RNAi lines 1 and 2 should be tested for significance; the extent of knockdown in each should be quantified either by qPCR in whole tissue knockdown, or by staining for protein levels if possible, to assess whether the variation in phenotypes is due to different knockdown levels. The use of a loss of function mutant in clones or tissue specific CRISPR-Cas9 KO or KD would also significantly increase confidence in the findings. *

• Our qPCR data indicate that SdcRNAi3 produces the most efficient knockdown, whilst SdcRNAi1 generates the weakest knockdown. The new manuscript version will incorporate this data in figure S1. Knockdown efficacy of SdcRNAi 3 has also been previously reported (Eveland et al., 2016).

• We apologise for omitting to add the statistical tests on phenotypic categories in figure S1A, this will be revised. We confirm that all Sdc RNAi phenotypic distributions are significantly different to that seen for age-matched controls (p- It should also be noted that despite weaker knockdowns with SdcRNAi1 and 2, we still observed statistically significant ISC depletion after 28 days of RNAi expression - we will add this data in figure S1. Overall, we are confident about Sdc’s role in maintaining intestinal stem cells.

*Similarly, the evidence for a lack of LINC protein role in the phenotype relies on single RNAi lines without validation of knockdowns. The authors should ideally validate these lines in this system or reference other studies that have validated the lines in this or other contexts. *

• The klarsicht RNAi line (BDSC 36721) and klaroid RNAi line (BDSC 40924) used in this study have been validated and used in other studies. (Falo-Sanjuan & Bray, 2022; Collins et al., 2017)

• For Msp300 RNAi knockdown we have used two independent RNAi lines which gave similar results. We will amend the text to clarify these points. In addition, the line reported in the manuscript was previously validated (Dondi et al., 2021; Frost et al., 2016).

Minor Comments: *The figures are generally very clear but some of the IF image panels are very small and require significant on-screen enlargement to be legible. In particular in Figure 1B the cross section views make it difficult to assess expression in the different cell types (and don't show very many cells), could this be shown in wholemount or as separated channels in a supplementary figure? In addition, it would strengthen the argument to include counterstains for markers of the different cell types (particularly to distinguish ISC/EB from EE). This could include esg-lacZ to mark ISC/EBs or prospero for EEs. However, if a broader view of these panels makes it clearer that all epithelial cells are expressing Syndecan this may not be essential. *

• We are happy to incorporate larger fields of view, and co-immunostaining with different cell type markers.

*Syndecan is referred to throughout as a stem cell regulator. This implies that in certain contexts or in response to certain stimuli its expression may be altered to elicit a stem cell response but no examples of this are shown. Moreover, only knockdown and not overexpression gives phenotypes suggesting its role may be as a required protein than a regulator. Either examples of its expression being modulated in homeostasis or in response to a challenge could be included or the wording could be amended. *

• We agree with the reviewer and will amend the wording.

*Expression of Syndecan in neuroblasts is described as data not shown, it would be better to include this for completeness. *

• We will add this data in figure 4.

*In addition to the intestinal validation of the Syndecan RNAi lines, validation of knockdown in the germline would be valuable to support the conclusions of Fig S4 given differences of knockdown in the germline with some RNAi lines (although inclusion of Dicer in the driver line should have overcome this). *

• Sdc expression is very low in the germline, compared to the surrounding somatic cells, therefore we are not confident that we can detect differences in expression level after knockdown. We suggest adding a panel in figure S4 to show the low expression and adding a comment in the text. Reviewer #1 (Significance (Required)): *The study describes a potentially very interesting, novel link between Syndecan, nuclear shape and apoptosis in cycling cells that could have broad relevance. If fully validated this could have implications for other stem cell populations, including those in mammals and disease relevance in the context of cancer. The paper is fundamentally descriptive in nature and so the level of significance hinges on the strength of evidence and how interesting the phenotype itself is. At this stage the audience will be primarily in the areas of fundamental research in biology of the nucleus and cytoskeleton. Defining the mechanistic link between Syndecan and nuclear morphology will be a critical next step and while not essential for this study would significantly increase the likely interest in the paper. *

• We thank the reviewer for these constructive comments. We agree that discovering the mechanistic links between Syndecan and nuclear morphology in future studies, in this and other model systems, will be relevant to many areas of biological research.

*In terms of significance in stem cell biology the distinction between a regulator and a requirement to prevent stem cell apoptosis is important and the lack of evidence for a context in which Syndecan plays a regulatory role somewhat detracts from the breadth of impact. My field of expertise is in epithelial stem cell biology. *

• We agree and will amend our wording.

Reviewer #2 *(Evidence, reproducibility and clarity (Required)): ** Summary: Stem cell (SC) maintenance and proliferation are necessary for tissue morphogenesis and homeostasis. The basement membrane (BM) has been shown to play a key role in regulating stem cell behavior. In this work, the authors unravel a new connection between the receptor for BM components Syndecan (Sdc) and SC behavior, using Drosophila as model system. They show that Sdc is required for intestine stem cell (ISC) maintenance, as Sdc depletion results in their progressive loss. At a cellular level, they also find that Sdc depletion in ISCs affects cell survival, cell and nuclear shape, nuclear lamina and DNA damage. In addition, they show that the defects in shape are not related to cell death. They also find that Sdc depletion in neural stem cells also results in nuclear envelope remodeling during cell division. This is in contrast to what happens in female germline stem cells where Sdc does not seem to be required for their survival or maintenance. In general, I believe that this work unravels a connection between Sdc and stem cell behavior. However, I think the study is still at a preliminary stage, as how Sdc regulates different facets of stem cell behavior remains unclear.

Major comments: 1. To clearly show that the cellular changes produced by loss of Sdc are not due to cell death, one should quantify the ISC area and shape of Sdc-depleted ISCs expressing DIAP1 and compare it to that of Sdc-depleted ISCs. As DIAP1 overexpression only partially rescues ISC loss due to Sdc depletion, one should show that the Sdc-depleted ISCs expressing DIAP1 that still show cellular changes are not dying, as overexpression of Diap1 might not be sufficient to completely rescue cell death in all Sdc-depleted ISCs. In fact, apoptosis in Sdc depleted guts and the ability of Diap1 overexpression to rescue cell death should be analyzed using markers of caspase activity, this will provide a better idea of the contribution of apoptosis to the phenotypes associated to Sdc depletion. *

• We can, as suggested by the reviewer, quantify the area and shape of Sdc-depleted ISCs expressing DIAP1 and compare it to that of Sdc-depleted ISCs. However, our immunostainings with anti-Caspase 3 or Drice do not pick up apoptotic cells in the fly gut. This is not entirely unexpected, as apoptosis is unfortunately not easily detected in this tissue. In the absence of a positive readout of apoptosis, we will not be able to discriminate between apoptotic and non-apoptotic stem cells when quantifying area and shape and will only have global quantifications.

• The authors show that ISC loss is associated with reduced cell density, suggesting that this is most likely due to failure in new cell production. What do they mean with cell production? Is this related to a problem in regulating cell division or to the fact that as some ISCs are lost by apoptosis there is progressively less ISCs or to a combination of both? I think that cell division should be monitored throughout time as well as cell death in ISCs.*

• Based on esgF/O experiments (fig. 1D-F and S1C) where we can trace the production of new cells with GFP, we know that Sdc RNAi expression (i) impairs the appearance of newly differentiated cells in the tissue and (ii) results in the disappearance of progenitor cells (fig. S1C). Supporting these points, (i) we have observed PH3+ mitotic stem cells upon Sdc RNAi, so we are confident the cells are able to initiate cell division (see also fig. 2G), and (ii) we have occasionally noted in fixed samples stem cells looking like they were in the process of delaminating. Overall, the failure of cell production is likely related to problems with both completion of cell division and progressive stem cell loss. High resolution live imaging will in future give us a better insight into stem cell division dynamics/behaviour, however, the technical improvements required are beyond the scope of this project. In the meantime, we propose to clarify our statement in the text.

• The authors report that in contrast to what happens when Sdc is eliminated from ISCs, its elimination from EEs results in an increase in the number of these cells. An explanation for this result is missing.*

• Based on known roles of Syndecan in other Drosophila tissues (Johnson et al., 2004; Steigemann et al., 2004; Chanana et al., 2009; Schulz et al., 2011), we speculate that Syndecan may contribute to robo/slit signalling, which is an important regulator of EE activity in the Drosophila gut (Biteau & Jasper 2014; Zeng et al., 2015). We propose to amend the text to express this hypothesis.

• The authors suggest that "Sdc function is unlikely to be fully accounted for by individual LINC complex proteins, although these proteins might act redundantly". Checking redundancy seems a straight forward experiment, which only requires the simultaneous expression of RNAis against several of these proteins. This would help to settle the implication of LINC complex proteins on Sdc function.*

• To check redundancy, we propose to combine Klaroid RNAi with Msp300 or Klarsicht RNAis, and express two RNAis at a time in ISCs. We will then measure stem cell proportions and the proportion of ISCs with DNA damage.

• Although quantification of DNA damage, by immunolabelling with gH2Av, reveals that knockdown of individual LINC complex components did not recapitulate the damage observed upon Sdc depletion (Fig.3G), the image shown in Fig.3F reflects much higher levels of gH2Av in Msp300 RNAi cells compared to Sdc RNAi cells. Authors should clarify this. *

• Like the reviewer, we are intrigued by the higher levels of H2Av staining in the tissue, despite Msp300 knockdown in stem cells only (fig. 3F). It is worth noting that we observed this with two independent RNAi lines (we showed only one RNAi in the manuscript, but we will amend the text to indicate this). In fig. 3F, we will indicate with an arrow the only ISC that is H2Av positive, and mention in the text that the majority of DNA damage signal observed in the Msp300 RNAi condition is in enterocytes, not ISCs. We currently do not have an explanation for why loss of Msp300 in ISCs should cause DNA damage in neighboring cells.

*In addition, the consequences of the simultaneous elimination of more than one component of the LINC complex on DNA damage should be analyzed. *

• We agree, and as we check for redundancy (as in point 4), we will also immunostain the tissues for H2Av.

• The authors claim that the fact that "DNA damage was found more frequently in Sdc-depleted ISCs with lamina invaginations compared to those without (Figure 3H), supports a model whereby the development of nuclear lamina invaginations precedes the acquisition of DNA damage". However, to me, these results show that there is a relation between these two phenotypes, but not that one precedes the other. In order to show which one is the possible cause and which the consequence, the authors should perform a time course of the appearance of each of these phenotypes.*

• We agree with the reviewer that we should rephrase our statement to indicate a relationship between lamina invaginations and DNA damage, rather than a causality (as stated in fig. 3H).

(In terms of performing a time course analysis, the difficulty is that after 3 days of Sdc RNAi expression, the apparent DNA damage (fig. 3G) corresponds to a very small proportion of stem cells, meaning that an exceptionally large sample size would be required to achieve robust statistical analysis.)

• When studying the role of Sdc in neural stem cells, the authors show that elimination of Sdc in neuroblasts also affect nuclear envelope and shape. Furthermore, in this case, they also show that Sdc elimination affects cell division. To look for a more conserved role of Sdc in stem cell behavior, I believe the authors should also analyze whether Sdc elimination in neural stem cells results in an increase in DNA damage, as it is the case in ISCs.*

• We will stain larval brains for H2Av to see if DNA damage is also observed following Sdc knockdown in neuroblasts.

• When analyzing a possible role of Sdc in fGSCs, quantification of germline stem cells and gH2Av levels in control nosGal4 and nos>Sdc RNAi germaria should be done. In addition, it is not clear to me whether Sdc is in fact expressed in fGSCs.*

• *

• As mentioned in comments to reviewer 1, we will add a panel in figure S4 to show the low Sdc expression in fGSCs. We will also clarify in the text that we do not see any H2Av staining in the fGSCs (thus, there is nothing to quantify in this case).

* The authors should show presence of Sdc in neuroblasts.*

• Yes, we agree, as also mentioned in comments to reviewer 1.

Reviewer #2 (Significance (Required)): *In general, although this work reveals that elimination of Sdc affects different aspects of intestinal and neural stem cell behavior, including cell survival, cell production, nuclear shape, nuclear lamina or DNA damage, their contribution to stem cell loss and interactions between them have not been analyzed in detail. The role of the basement membrane in stem cell behavior has been extensively studied. In particular, the role of syndecan in stem cell regulation has been primarily confined to cancer, muscle, neural and hematopoietic stem cells. Thus, the study here presented could extend the role of Sdc to intestinal stem cells and could potentially reveals a conserved role for Sdc in neural stem cell behavior. However, the problem with the data mentioned above, hinders the assessment of the significance of this work. *

• We thank the reviewer for their assessment and are glad that they also find that our study provides novel connections between Syndecan and the regulation of intestinal and neural stem cell behaviors. To strengthen our conclusions, we will include additional experiments or amend the text, as indicated above.

Reviewer #3* (Evidence, reproducibility and clarity (Required)): ** Peer-review: The transmembrane protein Syndecan regulates stem cell nuclear properties and cell maintenance.

In this work, the authors investigate the role of the transmembrane protein Syndecan (Sdc) in nuclear organisation and stem cell maintenance. Theys show that Sdc knockdown in intestinal stem cells (ISCs) results in a reduction of the ISC pool as well as of their progeny. They hypothesise that these ISCs might get eliminated via cell death, however, expression of the apoptotic inhibitor DIAP1 only rescued ISC loss by 50%. Hence, they suggest that apoptosis can not account for the total decrease in ISCs observed upon Sdc loss. ISCs depleted from Sdc exhibited abnormal cytoplasmic and nuclear morphologies. As Sdc has previously been implicated in the abscission machinery in mammalian cultured cells, they tested if Sdc could be playing a similar role in the abscission of ISCs. However, ISCs were capable of undergoing cytokinesis. Next, they tested if Sdc depletion could be altering the linkage between the plasma membrane and the nucleus mediated by the Linker of Nucleoskeleton and Cytoskeleton (LINC) complex. However, individual knockdowns of the different components of the complex did not disrupt the nuclear morphology to the same extent as Sdc knockdown, suggesting that Sdc function may be independent of the LINC complex. Finally, they observed that Sdc-depleted ISCs exhibited DNA damage, suggesting that Sdc may play a role in DNA protection. The authors next tested if Sdc played similar roles in other stem cell types such as the female germline stem cells (fGSCs) and larval neural stem cells (NSCs). While Sdc depletion appeared dispensable for fGSC maintenance, it prolonged NSC divisions and altered the nuclear morphology of NSCs. Upon further investigations, they observed that the NSC's nuclear envelope was disrupted upon division, hence causing defects in the nuclear size ratio of NSC and their progeny. This study provides with interesting findings in the field and proves a new role for Sdc in the regulation of intestinal and neural stem cell maintenance. I would recommend this manuscript to be accepted if the authors address the following comments.

__Major comments: __ 1. In Figure 2 A-B, Sdc RNAi should ideally have a UAS control transgene to match the number of UAS being expressed to that of Sdc RNAi, DIAP1. Otherwise, it is plausible that reduced RNAi expression of Sdc RNAi, DIAP1 animals is the cause of the partial rescue. Staining against cell death markers such as Dcp-1 or TUNEL might also quantify the number of cells undergoing cell death in each of the genotypes. *

• As mentioned in comments to reviewer 2 (point 1), it is difficult to label apoptotic cells in the fly gut. However, we could set up an additional control to test that the partial rescue observed upon DIAP1 expression is not a result of Gal4 dilution.

• " These phenotypes were observed both with and without DIAP1 expression (Figure 2C), indicating that these cell shapes are not caused by apoptosis."Misleading, as DIAP overexpression in Sdc knockdown background only rescued apoptosis by 50%. Hence, it is possible that those cells undergoing morphological defects, protrusions and blebbing might still undergo death - also considering those morphological changes are typically observed in apoptotic cells...Therefore, to rule apoptosis out, these cells should be shown to be negative for cell death markers. *

• We agree, however, it is difficult to label apoptotic cells. We think that the quantification of shape and area (as suggested by reviewer 2, point 1) will clearly show that the cell shapes resulting from Sdc depletion are not caused by apoptosis.

• Show if Sdc is expressed in fGSCs - the lack of phenotype caused by Sdc knockdown might be due to lack of expression of Sdc.*

• As mentioned in comments to reviewers 1&2, we will add a panel in figure S4 to show the low sdc expression in fGSCs.

• "After confirming the presence of Sdc in neuroblasts (data not shown)."Data should be shown. It would be of great interest for researchers if you showed a staining of different brain cell types (NBs, glia, neurons) and the Sdc expression patterns.*

• As mentioned in comments to reviewers 1&2, we will add a panel in figure 4 to show sdc expression in NBs and the overall expression pattern.

• You show how Slc-depleted NBs have disrupted nuclear morphologies. However, does Slc KD in NB lineages affect their ability to self-renew and generate differentiated progeny? Is the number of NBs and of their progeny cells altered as it is for ISCs?*

• We propose to knockdown Sdc in NBs and quantify brain size in 3rd instar larvae to test if the ability to generate progeny is affected.

• Does protection against DNA damage in an Slc knockdown background prevent the defects observed with the single knockdown and ISC elimination?*

• This is a good question, and we should emphasize this point in the discussion. However, because of the multiple routes of DNA damage response, and the multiple lines needed to explore this connection, we feel that investigating this question is beyond this project.

• Any idea the similarities between ISC and NBs that can account for why Sdc knockdown has effects in those systems, while no effect was observed in the germ cells?*

• Besides the differences in expression level, we speculate that GSCs may have a different nuclear / lamina architecture which might reflect differences in how GSCs control the physical integrity of their nuclei. It is also possible that the differences observed between tissues reflect the way stem cells connect to their microenvironment. Notably, fGSCs rely extensively on E-Cadherin mediated adhesion with neighbouring cells, and it is possible that contact with the extracellular matrix is dispensable. We will consider these possibilities in the discussion.

Minor comments:* ** 8. Lamina invaginations, for example in Figure 3 A, could be indicated with an arrow for easier detection. *

• Thanks for this suggestion, we will amend the figure.

Specify the type and location of NB imaged during live cell experiments.

• The NBs were imaged in the brain lobes, and we did not distinguish between type I and II NBs. We will add a sentence in the method section to clarify.

*Reviewer #3 (Significance (Required)): Expertise: Drosophila stem cells *

• Many thanks for the constructive comments.

Author response:

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

Reviewer #1 (Public Review):

(1) The data strongly suggest that iron depletion in urine leads to conditional essentiality of some genes. It would be informative to test the single gene deletions (Figure 3G) for growth in urine supplemented with iron, to determine how many of those genes support growth in urine due to iron limitation.

We appreciate this suggestion. We have now included this suggested experiment as a new panel (Figure 5G).

(2) Line 641. The authors raise the intriguing possibility that some mutants can "cheat" by benefitting from the surrounding cells that are phenotypically wild-type. Growing a fepA deletion strain in urine, either alone or mixed with wild-type cells, would address this question. Given that other mutants may be similarly "masked", it is important to know whether this phenomenon occurs.

We thank the reviewer for this suggestion but believe that this would be very difficult to ascertain in K. pneumoniae as several redundant iron uptake systems exist. This would require significantly more time to construct sequential/combinatorial iron-uptake mutants to exactly determine this “cheating” and “masking” phenomenon and such work is beyond the scope of the current study.

(3) In cases where there are disparities between studies, e.g., for genes inferred to be essential for serum resistance, it would be informative to test individual deletions for genes described as essential in only one study.

We thank the reviewer for this suggestion, and we agree that deleting conditionally essential genes (i.e. serum resistance) could help identify discrepancies in methodology with other studies but this is beyond the scope of this study. Furthermore, we do not have these other strains readily available to us and importing these strains into Australia is challenging due to the strict import/quarantine laws.

Reviewer #1 (Recommendations For The Authors)

(4) Line 529. Why was 50 chosen as the read count threshold?

This was chosen as the minimum threshold needed to exclude essential genes from the comparative analysis, as these can contribute false positive results where a change from, for example, 2 to 5 reads between conditions is considered a >2-fold change. We have updated the manuscript text to highlight this: “were removed from downstream analysis to exclude confounding essential genes and minimize the effect of stochastic mutant loss” (line 539

(5) The titles for Figure 5 and Figure 6 appear to be switched.

Thank you, we have now corrected this error.

(6) Line 381. "Forty-six of these regions contain potential open reading frames that could encode proteins". How is a potential ORF defined?

This was based on submitting the selected 145bp regions to BLASTx using default parameters and listing the top hit (if one was found). We have now edited the manuscript text to make this clearer. (Line 394)

(7) Two previous TnSeq studies looking at Escherichia coli and Vibrio cholerae suggest that H-NS can prevent transposon insertion, leading to false positive essentiality calls. Is there any evidence of this phenomenon here? A/T content could be used as a proxy for H-NS occupancy.

We thank the reviewer for this point and also agree that H-NS or other DNA-binding proteins could indeed lead to false-positive essentiality calls using TraDIS. Based on this, we have now included a sentence in the conclusion section mentioning this methodological caveat (Line 631). We believe that A/T content could potentially be used as a proxy for H-NS occupancy,

Reviewer #2 (Recommendations For The Authors):

(1) The authors may wish to reformat the manuscript by decanting a number of panels and figures as supplementary material. These include the panels related to the description of TraDIS (for example Fig 1D, 1E, 1F. 1G, Fig 2A, Fig 3C, 3D, 3E, 3F, Fig 5C, Fig 6D). This is a well-established method.

We thank the reviewer for this suggestion but believe that these panels allow the methodology and resulting insertion plots to be more followable and allow other researchers, of varying expertise, to better understand this functional genetic screen technique.

(2) The authors need to indicate how relevant the strain they have probed is. Is it a good reference strain of the KpI group?

This is a great suggestion and we have now included a new figure illustrating the genetic context and relatedness of K. pneumoniae ECL8 within the KpI phylogroup (New Figure 3).

(3) The authors need to provide an extensive comparison between the data obtained and those reported testing other Klebsiella strains. A Table identifying the common and different genes, as well as a figure, may suffice. I would encourage authors to compare also their data against E. coli and Salmonella. For example, igaA seems to be not essential in Kebsiella although data indicates it is in Salmonella.

We thank the reviewer for their comment and appreciate that our data could be extended and compared to other relevant Enterobacteriaceae members. However, we believe this is beyond the scope of this study as the focus is more on K. pneumoniae.

(4) None of the mutants tested further are complemented. Without these experiments, it cannot be rigorously claimed that these loci play any role in the phenotypes investigated.

We agree that complementation is an important tenet for validation of mutant gene phenotypes to specific gene loci, in this case wbbY has already been complemented and believe complementation for an already known molecular mechanism would be redundant. Please refer to our response in point 6.

We complemented isolated transposon mutants hns7::Tn5 and hns18::Tn5 with a mid-copy IPTG inducible . We observed a slight increase in serum susceptibility but not full rescue of the WT phenotype (i.e. serum susceptibility). We suspect that the imperfect rescue of the serum-resistance phenotype observed could be due to the expression levels and copy number of the complement hns plasmid used. As hns is a known global regulator its possible pleiotropic role is complex as many aspects of stress response, metabolism or capsule could be affected in Klebsiella (doi.org/10.1186/1471-2180-6-72, doi.org/10.3389/fcimb.2016.00013). We have now included in the text our efforts in complementation and have included a new supplementary figure (Figure S11).

(5) The contribution of siderophores to survival in urine is not conclusively established. Authors may wish to test the transcription of relevant genes, and to assess whether the expression is fur dependent in urine. Also, authors may wish to identify the main siderophore needed for survival in urine by probing a number of mutants; this will allow us to assess whether there is a degree of selection and redundancy.

We thank the reviewer for their comment and agree siderophore uptake is important. We have now included an additional panel (Figure 5G) interrogating the importance of iron-uptake genes grown in urine which is iron limited. We do appreciate that further experiments looking into the Fur regulon and siderophore biosynthesis would be interesting but believe this is outside the scope of this study.

(6) The role of wbbY is intriguing, pointing towards the importance of high molecular weight O-polysaccharide. In this mutant background, the authors need to assess whether the expression of the capsule, and ECA is affected. Authors need also to complement the mutant. Which is the mechanism conferring resistance?

We thank the reviewer for their comment and would like to mention that wbbY has already been shown to play a role in LPS profile/biosynthesis and serum-resistance (10.3389/fmicb.2014.00608 ). Furthermore, blast analysis shows that the wbbY gene between the NTUH-K2044 (strain used in aforementioned study) and ECL8 shares 100% sequence identity and also shares lps operon structure. Hence, we do not find it pertinent to complement this mutant as we believe its molecular mechanism has already been established. We have now in the text more prominently highlighted the results of this study and how our screen was robust enough to also identify this gene for serum resistance.

(7) hns and gnd mutants most likely will have their capsule affected. The authors need to assess whether this is the case. Which is the mechanism conferring resistance?

As mentioned in point 6, we believe that the serum resistance phenotype is attributable to the LPS phenotype. Previous studies have listed hns and gnd mutants would likely have differences in capsule but due to hns being pleiotropic and gnd being intercalated/adjacent to the LPS/O-antigen biosynthesis it would be difficult to exactly delineate which cellular surface structure is involved.

(8) The conclusion section can be shortened significantly as much of the text is a repetition of the results/discussion section.

We thank the reviewer for their suggestion and have made edits to limit repetition in the conclusion section.

Reviewer #3 (Public Review):

Below I include several comments regarding potential weaknesses in the methodology used:

• The study was done with biological duplicates. In vitro studies usually require 3 samples for performing statistical robust analysis. Thus, are two duplicates enough to reach reproducible results? This is important because many genes are analyzed which could lead to false positives. That said, I acknowledge that genes that were confirmed through targeted mutagenesis led to similar phenotypic results. However, what about all those genes with higher p and q values that were not confirmed? Will those differences be real or represent false positives? Could this explain the differences obtained between this and other studies?

We thank the reviewer for their comment and apologize for the confusion, data were only pooled for the statistical analysis of gene essentiality. Here, two technical replicates of the input library were sequenced and the number of insertions per gene quantified (insertion index scores). These replicates had a correlation coefficient of r2 = 0.955, and the insertions per gene data were pooled to give total insertions index scores to predict gene essentiality. For conditional analyses (growth in urine or serum), replicate data were not combined. As mentioned previously, differences between this and other studies could also be attributed to inherent genomic differences or due to differences in experimental methodology, computational approaches, or the stringency of analysis used to categorize these genes.

• Two approaches are performed to investigate genes required for K. pneumoniae resistance to serum. In the first approach, the resistance to complement in serum is investigated. And here a total of 356 genes were identified to be relevant. In contrast, when genes required for overall resistance to serum are studied, only 52 genes seem to be involved. In principle, one would expect to see more genes required for overall resistance to serum and within them identify the genes required for resistance to complement. So this result is unexpected. In addition, it seems unlikely that 356 genes are involved in resistance to complement. Thus, is it possible false positives account for some of the results obtained?

We thank the reviewer for their comment and do believe false positives may account for some of the identified genes. Specifically, to the large contrast in genes, we believe this is due to the methodology as alluded to in our conclusion section. For overall resistance to serum, we used a longer time point (180 min exposure) where fewer surviving mutants are recovered hence fewer overall genes will be identified, whereas strains with short killing windows will have more (i.e. complement-mediated killing, 90 minute exposure).

Reviewer #3 (Recommendations For The Authors):

• In Figure 4 it is shown that genes important for growth in urine include several that are required for enterobactin uptake. Moreover, an in vitro experiment shows that the complementation of urine with iron increases K. pneumoniae growth. It would have been informative to do a competition experiment between the WT and Fep mutants in urine supplemented with iron. This could demonstrate that the genes identified are only necessary for conditions in which iron is in limiting concentrations and confirm that the defect of the mutants is not due to other characteristics of urine.

We appreciate this suggestion. We have now included a new panel (Figure 5G) addressing the supplementation of iron in urine for these select mutants.

• Considering the results section, the title for Figure 6 seems to be more appropriate for Figure 5.

Thank you, this has now been corrected.

Other points:

• Line 44: treat instead of treating

Thank you, this has now been corrected.

• Line 63: found that only 3 genes played a role instead of "found only 3 genes played a role"

Thank you, this has now been corrected.

• Line 105: is there any reason for only using males? Since UTIs are frequent in women? Why not use urine from women volunteers?

Due to accessibility of willing volunteers and human ethic application processes, only male samples were available. We are currently undertaking further studies to understand how male and female urine influences growth of uropathogens.

• Line 105: since the urine was filter-sterilized, maybe the authors can comment that another point that is missing in urine - and that it may be important to study - will be the presence of the urine microbiome and how this affects growth of K. pneumoniae.

We again thank the reviewer for this comment and have now edited the manuscript discussing how the absence of urine microbiome could affect growth (Line 659). As an aside, future studies in our lab are interested in looking at the role of commensal/microbiome co-interactions for essentiality/pathogenesis using TraDIS.

• Line 116: I understand that the 8 healthy volunteers combined males and females

Thank you, we have now edited this methods line to make this clearer.

• Line 120: incubate in serum 90 min and 180 RPM shaking: any reasons for using these conditions, any reference supporting these conditions?

Thank you for pointing this out, we were mirroring a previous K. pneumoniae serum-resistance study (doi.org/10.1128/iai.00043-).

• Line 156: space after the dot.

Thank you, we have now corrected this in the manuscript.

• Line 164: resulting reads were mapped to the K. pneumoniae: what are the parameters used for mapping (e.g. % of identity...)?

Thank you for bringing this to our attention, we have now included in our manuscript that we used the default parameters of BWA-MEM for mapping for minimum seed length (default -k =20bp exact match)

• Line 180: it will be good to upload to a repository the In-house scripts used or indicate the link beside the reference for those scripts.

Our scripts are derived from the pioneering TraDIS study (doi: 10.1101/gr.097097.109). We are currently still optimizing our scripts and intend to upload these to be publicly available. However, in the meantime we are more than happy to share them with other parties upon request.

• Line 191: why were genes classified as 12 times more likely to be situated in the left mode? Any particular reason for using this threshold?

We opted for a more-stringent threshold for classifying essential genes, in keeping with previous and comparable studies (doi.org/10.1371/journal.pgen.1003834).

• Line 209: do you mean Q-value of <0.05 instead of >0.05 ? How is this Q value is calculated, and which specific tests are applied?

Thank you for pointing out this Q value error, we have now corrected this in the manuscript. These values were generated using the biotradis tradis_comparison.R script which uses the EdgeR package. For further reading please see DOI: 10.1093/bioinformatics/btp616. The Q-values are from P values corrected for multiple testing by the Benjamini-Hochberg method.

• Line 212: again, which type of test is used? What about the urine growth analysis? The same type of tests were applied?

Thank you for bringing this to our attention, we have now indicated in the referenced method section the use of which package for which datasets (i.e. or serum). Line 212 refers to our use of the AlbaTraDIS package, which builds on the biotradis toolkit, to identify gene commonalities/differences in the selected growth conditions again using multiple testing by the Benjamini-Hochberg methods. For further reading, please refer to DOI: 10.1371/journal.pcbi.1007980

• Line 226: do the authors mean Sanger sequencing instead of SangerSanger sequencing?

Thank you, we have now corrected this in the manuscript.

• Line 239: does the WT strain contain another marker for differentiating this strain from the mutant? Or is the calculation of the number of WT CFUs done by subtracting the number of CFUs in media with antibiotics from the total number of CFUs in media without antibiotics? The former will be a more accurate method.

The calculation was based on the latter assumption, “number of WT CFUs done by subtracting the number of CFUs in media with antibiotics from the total number of CFUs in media without antibiotics”. We have now updated the methods section to make this clearer.

• Line 266: can you indicate approximately how many CFUs you have in this OD?

Thank you, we have now also indicated an approximate CFU for this mentioned OD600 (OD600 1 = 7 × 108 cells).

• Line 309: besides indicating Figure 1D please indicate here Dataset S1 (the table where one can see the list of essential and non-essential genes). This table is shown afterwards but I think it will be more appropriate to show it at the begging of the section.

Thank you, we have now taken on this recommendation and have now edited the manuscript to also indicate Dataset S1 earlier.

• Table 3. regarding the comparison of essential genes between different strains. I think it will be more clear if a Venn diagram was drawn including only genes that have homologs in all the studied strains (i.e. defining the core genome essentially).

We would like to thank the reviewer for suggesting a venn diagram and have now removed Table 3 which has been replaced with a new Figure 3.

• Line 461: replicates were combined for downstream analyses? But are replicates combined for doing the statistical analysis? If so, how is the statistical analysis performed? How is it taken into account the potential variability in the abundance in each library? An r of 0.9 is high but not perfect.

Technical replicates of the sequenced input library were combined following identification of a correlation coefficient of r2 = 0.955, for the calculation of insertion index scores used in gene essentiality analysis. While r2 = 0.955 is not perfect, discrepancies here can be attributed to higher variance in insertion index scores when sampling small genes, as these are represented by fewer insertions and the stochastic absence of a single insertion event has a greater effect on the overall IIS. Replicate data were not pooled for statistical analysis of mutant fitness (growth in urine and serum).

• Line 487: is there any control strain containing the kanamycin gene in a part of the genome that does not affect the growth of K. pneumoniae? This could be used to show that having the kanamycin gene does not provide any defect in urine growth.

We thank the reviewer for this suggestion but argue that introduction of the kanamycin gene into each unique loci may result in various levels of gene fitness that would be incomparable to a single control strain. Instead, we culture the ECL8 mutant library in urine and ensure that its kinetics are comparable to the wildtype. As the library contains thousands of kanamycin cassettes uniquely positioned across most of the genome with no observable growth defect, we do not anticipate the presence or expression of the cassette to have an appreciable impact.

• Line 569: in the methodology it was indicated that control cells were incubated in PBS for the same amount of time. I think this is an important control that is not cited in the results section. Please can you indicate?

We apologise for this misunderstanding due to how the methodology was written. The experiment did not sequence the PBS incubated samples as this was solely used a check for viability of the used K. pneumoniae ECL8 stock solution.

• Line 597: "Mutants in igaA are enriched in our experiments". Can you show this data?

We have now included this as a supplementary (Figure S11A)

• Line 615: when doing this calculation, I guess the authors take into account only genes that are also present in the other strains.

That is correct, we were aiming to highlight the high conservation of “essential genes” among all the selected strains.

• Line 627: why surprisingly? Because is too low. Then indicate.

Thank you, we have now edited this sentence to indicate that.

• Figure 4: please, for clarity, can you indicate the meaning of the colors in the figure itself besides indicating it in the figure legend?

Thank you, we have now included a color legend in these figure panels for clarity.

Author response:

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

eLife assessment

The authors build upon prior data implicating the secreted peptidoglycan hydrolase SagA produced by Enterococcus faecium in immunotherapy. Leveraging new strains with sagA deletion/complementation constructs, the investigators reveal that sagA is non-essential, with sagA deletion leading to a marked growth defect due to impaired cell division, and sagA being necessary for the immunogenic and anti-tumor effects of E. faecium. In aggregate, the study utilizes compelling methods to provide both fundamental new insights into E. faecium biology and host interactions and a proof-of-concept for identifying the bacterial effectors of immunotherapy response.

We thank the Reviewers for their positive feedback on our manuscript. We also appreciate their helpful comments/critiques and have revised the manuscript as indicated below.

Public Reviews:

Reviewer #1 (Public Review):

Klupt, Fam, Zhang, Hang, and colleagues present a novel study examining the function of sagA in E. faecium, including impacts on growth, peptidoglycan cleavage, cell separation, antibiotic sensitivity, NOD2 activation, and modulation of cancer immunotherapy. This manuscript represents a substantial advance over their prior work, where they found that sagA-expressing strains (including naturally-expressing strains and versions of non-expressing strains forced to overexpress sagA) were superior in activating NOD2 and improving cancer immunotherapy. Prior to the current study, an examination of sagA mutant E. faecium was not possible and sagA was thought to be an essential gene.

The study is overall very carefully performed with appropriate controls and experimental checks, including confirmation of similar densities of ΔsagA throughout. Results are overall interpreted cautiously and appropriately.

I have only two comments that I think addressing would strengthen what is already an excellent manuscript.

In the experiments depicted in Figure 3, the authors should clarify the quantification of peptidoglycans from cellular material vs supernatants. It should also be clarified whether the sagA need to be expressed endogenously within E. faecium, and whether ambient endopeptidases (perhaps expressed by other nearby bacteria or recombinant enzymes added) can enzymatically work on ΔsagA cell wall products to produce NOD2 ligands?

We mentioned in the main text that peptidoglycan was isolated from bacterial sacculi and digested with mutanolysin for LC-MS analysis. We have now also included “mutanolysin-digested” sacculi in the Figure 3 legend as well.

We have added the following text “We next evaluated live bacterial cultures with mammalian cells to determine their ability to activate the peptidoglycan pattern recognition receptor NOD2” and “our analysis of these bacterial strains” to indicate live cultures were evaluated for NOD2 activation.

We have also added the following text “Our results also demonstrated that while many enzymes are required for the biosynthesis and remodeling of peptidoglycan in E. faecium, SagA is essential for generating NOD2 activating muropeptides ex vivo.”

In the murine experiments depicted in Figure 4, because the bacterial intervention is being performed continuously in the drinking water, the investigators have not distinguished between colonization vs continuous oral dosing of the mice peptidoglycans. While I do not think additional experimentation is required to distinguish the individual contributions of these 2 components in their therapeutic intervention, I do think the interpretation of their results should include this perspective.

We have added the following text “We note that by continuous oral administration in the drinking water, live E. faecium and soluble muropeptides that are released into the media during bacterial growth may both contribute to NOD2 activation in vivo.” and revised the following text “Nonetheless, these results demonstrate SagA is not essential for E. faecium colonization, but required for promoting the ICI antitumor activity through NOD2 in vivo.

Reviewer #2 (Public Review):

Summary:

The gut microbiome contributes to variation in the efficacy of immune checkpoint blockade in cancer therapy; however, the mechanisms responsible remain unclear. Klupt et al. build upon prior data implicating the secreted peptidoglycan hydrolase SagA produced by Enterococcus faecium in immunotherapy, leveraging novel strains with sagA deleted and complemented. They find that sagA is non-essential, but sagA deletion leads to a marked growth defect due to impaired cell division. Furthermore, sagA is necessary for the immunogenic and anti-tumor effects of E. faecium. Together, this study utilizes compelling methods to provide fundamental new insights into E. faecium biology and host interactions, and a proof-of-concept for identifying the bacterial effectors of immunotherapy response.

Strengths:

Klupt et al. provide a well-written manuscript with clear and compelling main and supplemental figures. The methods used are state-of-the-art, including various imaging modalities, bacterial genetics, mass spectrometry, sequencing, flow cytometry, and mouse models of immunotherapy response. Overall, the data supports the conclusions, which are a valuable addition to the literature.

Weaknesses:

Only minor revision recommendations were noted.

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):

General comments - the number/type of replicates and statistics are missing from some of the figure panels. Please be sure to add these throughout - all main figure panels should have replicates. I've also noted some specific cases below.

Abstract - sagA is non-essential, need to edit text at "essential functions".

This change has been made.

"small number of mutations" - specify how many in the text.

We revised the text. “Small number” is changed to “11”.

"under control of its native promoter" - what was the plasmid copy number? It looks clearly overexpressed in Figure 1d despite using a native promoter, although it's a bit hard to know for sure without a loading control.

pAM401 has p15A origin of replication, therefore the plasmid copy number ~20-30 copies (Lutz R. et al Nucleic Acids Res. 1997). Total protein was visualized by Stain-Free™ imaging technology (BioRad) and serves as protein loading control and has been relabeled accordingly.

"decrease levels of small muropeptides" - the asterisks are missing from Figure 3a.

Green asterisks for peaks 2, 3, 7 and purple asterisks for peaks 13, 14 were added.

The use of "Com 15 WT" in the figures is confusing - just replace it with "wt" and specify the strain in the text. Presumably, all of the strains are on the Com 15 background.

“Com15 WT” was replaced to “WT” in figures and main text.

Change 1d to 1b so that the panels are in order (reading left to right and then top to bottom).

Figure 1 legend is missing a number of replicates and statistics for 1a.

Number of replicates were added.

Figure 1b - it's unclear to me what to look at here, could add arrows indicating the feature or interest and expand the relevant text.

Arrows pointing to cell clusters were added.

Figure 1d - what is "stain free"? It would be preferable to show a loading control using an antibody against a constitutive protein to allow for normalization of the loading control.

Stain-Free Imaging technology (BioRad) utilizes gel-containing trihalo compound to make proteins fluorescent directly in the gel with a short photoactivation, allowing the immediate visualization of proteins at any point during electrophoresis and western blotting. Stain-Free total protein measurement serves as a reliable loading control comparable to Coomassie Blue Staining. This has been relabeled a “Total protein” in the Figure and Stain-free imaging technology is noted in the legend.

ED Figure 1 - representative of how many biological replicates?

Legends are updated.

ED Figure 2a - I would replace this with a table, it's not necessary to show the strip images. Also, please specify the number of replicates per group.

Additional Extended Data Table 2 was added.

ED Figure 2b - This data was not that convincing since the sagA KO has a marked growth defect and the time points are cut off too soon to know if growth would occur later. The MIC definition is potentially misleading. Should specific a % growth cutoff (i.e. <10% of vehicle control) and the metric used (carrying capacity or AUC). Then assign MIC to the tested concentration, not a range. The empty vector also seems to impact MIC, which is concerning and complicates the interpretation. Specify the number of replicates and add statistics. Given these various concerns, I might suggest removing this figure, as it doesn't really add much to the story.

We appreciate this comment from the Reviewer, but believe this data is helpful for paper and have included longer time points for the growth data. The definition of MIC for ED Fig. 2b has been included in the legend.

Figure 2 - specify the type of replicate. Number of cells? Number of slices? Number of independent cultures?

For Cryo-ET experiments single bacterial cultures were prepared. Number of cells and slices for analysis are indicated in the legend. Legends are updated.

Figure 4e - missing the water group, was it measured?

Water (αPD-L1) group was not included in immune profiling of tumor infiltrating lymphocytes (TILs) experiment, as we have previously demonstrated limited impact on ICI anti-tumor activity and T cell activation in this setting (Griffin M et al Science 2021).

Figure 4d - is this media specific to your strains? If not, qPCR may be a better method using strain-specific primers.

Yes, HiCrome™ Enterococcus faecium agar plates (HIMEDIA 1580) are selective for Enterococcus species, moreover the agar is chromogenic allowing to identify E. faecium as yellow colonies among other Enterococcus species.

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

Learn more at Review Commons

---

Reply to the reviewers

RC-2023-02105R: Brunetta et al.,

IF1 is a cold-regulated switch of ATP synthase to support thermogenesis in brown fat

We are happy to submit our revised manuscript after considering the suggestions made by reviewers. The comments were overall positive, and the changes requested were mostly editorial. We have, nevertheless, added new experiments as quality controls. These experiments did not affect the main conclusions of our work. In addition, we also included two in vivo experimental models of gain and loss-of-function, to further address the physiological relevance of IF1 in BAT thermogenesis. We believe with these additional experiments, quality controls as well as in vivo models, our study has improved considerably. We hope our efforts will be appreciated by the reviewers and we make ourselves available to answer any further questions.

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

Summary: In the present manuscript, the authors present data in support of their primary discovery that "IF1 controls UCP1-dependent mitochondrial bioenergetics in brown adipocytes". The opening figure convincingly demonstrates that IF1 expression is cold-exposure dependent. They then go on to show that loss of IF1 has functional consequences that would be predicted based on IF1's know role as a regulator of ATP hydrolysis by CV. They go on to make a few additional claims, succinctly detailed in the Discussion section. Specific claims include the following: 1) IF1 is downregulated in cold-adapted BAT, allowing greater hydrolytic activity of ATP synthase by operating in the reverse mode; 2) when IF1 is upregulated in brown adipocytes in vitro mitochondria unable to sustain the MMP upon adrenergic stimulation, 3) IF1 ablation in brown adipocytes phenocopies the metabolic adaptation of BAT to cold, and 4) IF1 overexpression blunts mitochondrial respiration without any apparent compensator response in glycolytic activity. The claims described above are well supported by the evidence. The manuscript is very well written, figures are clear and succinct. Overall, the quality of the work is very high. Given that IF1 is implicated across many fields of study, the novel discovery of IF1 as a regulator of brown adipose mitochondrial bioenergetics will be of significance across several fields. That said, a few areas of concern were apparent. Concerns are detailed in the "Major" and "Minor" comments section below. Additional experiments do not appear to be required, assuming the authors adequately acknowledge the limitations of the study and either remove or qualify speculative claims.

Major Comments:

1. The authors convincingly demonstrate that IF1 expression is specifically down-regulated in BAT upon cold-exposure. These data strongly implicate a role for IF1 in BAT bioenergetics, a major claim of the authors and a novel finding herein. Additional major strengths of the paper, which provide excellent scientific rigor include the use of both loss of function and gain of function approaches for IF1. In addition, the mutant IF1 experiments are excellent, as they convincingly show that the effects of IF1 are dependent on its ability to bind CV. RESPONSE: We thank the reviewer for the positive feedback on our work.

Regarding Figure 1 - Did the content of ATP synthase change? In figure 1A-B, the authors show that ATPase activity of CV is higher in cold-adapted mice. While this result could be due to a loss of IF1, it could also be due to a higher expression of CV. To control for this, the authors should consider blotting for CV, which would allow for ATPase activity to be normalized to expression.

RESPONSE: Thank you for this suggestion. We have now determined complex V subunit A in our experimental protocol. We found that cold exposure does not impact complex V protein levels. Given the importance of this control, we have now included it in Figure 1 (Please, see the revised version) alongside the IF1/complex V ratio. In addition, we have now performed WBs in the BAT from mice exposed for 3 and 7 days to thermoneutrality (~28°C). We found that IF1 is not reduced following whitening of BAT by this approach whilst UCP1 and other mitochondrial proteins are reduced. This set of data is now included in Figure 1I,K,L.

Regarding MMP generated specifically by ATP hydrolysis at CV, the reversal potential for ANT occurs at a more negative MMP than that of CV (PMID: 21486564). Because reverse transport of ATP (cytosol to matrix) via ANT will also generate a MMP, it is speculative to state that the MMP in the assay is driven by ATP hydrolysis at CV. It is possible and maybe even likely that the majority of the MMP is driven by ANT flux, which in turn limits the amount of ATP hydrolyzed by CV. Admittedly, it is very challenging to different MMP from ANT vs that from CV, thus the authors simply need to acknowledge that the specific contribution of ATP hydrolysis to MMP remains to be fully determined. That said, the fact that ATP-dependent MMP tracks with IF1 expression does certainly implicate a role for ATP hydrolysis in the process. The authors should consider including a discussion of the ambiguity of the assay to avoid confusion. A role for ANT likely should be incorporated in the Fig. 1J cartoon.

RESPONSE: Thank you for bringing the ANT contribution to MMP to our attention. The effects of ATP in the real-time MMP measurements were totally abolished by the addition of oligomycin in BAT-derived isolated mitochondria, thus suggesting dependency of complex V in this process. However, the assessment of MMP in intact cells is much more challenging given cytosolic vs. mitochondrial contribution to ATP pool, and ATP synthase vs. ANT reversal capacity depending on MMP. Nevertheless, we have addressed these points in the discussion section as well as added to our schematic cartoon in Figure 1m.

Regarding the lack of effect of IF1 silencing on MMP, it is possible that IF1 total protein levels are simply lower in cultured brown fat cells relative to tissue? The authors could consider testing this by blotting for IF1 and CV in BAT and brown fat cells. The ratio of IF1/ATP5A1 in tissue versus cells may provide some amount of mechanistic evidence as to their findings.

RESPONSE: We have now blotted for complex V and IF1 in both differentiated primary brown adipocytes and BAT homogenates derived from mice kept at room temperature (~22°C). We found the levels of complex V in primary brown adipocytes are higher than BAT homogenates. Therefore, IF1/complex V ratio is different between these two systems. This has indeed the potential to influence our gain and loss-of-function experiments. We have added these results alongside their interpretation in the revised manuscript.

The calculation of ATP synthesis from respiration sensitive to oligomycin has many conceptual flaws. Unlike glycolysis, where ATP is produced via substrate level phosphorylation, during OXPHOS, the stoichiometry of ATP produced per 2e transfer is not known in intact brown adipose cells. This is a major limitation of this "calculated ATP synthesis" approach that is beginning to become common. Such claims are speculative and thus likely do more harm than good. In addition to ANT and CV, there are many proton-consuming reactions driven by the proton motive force (e.g., metabolite transport, Ca2+ cycling, NADPH synthesis). Although it remains unclear how much proton conductance is diverted to non-ATP synthesis dependent processes, it seems highly likely that these processes contribute to respiratory demand inside living cells. Moreover, just as occurs with UCP1 in response to adrenergic stimuli, proton conductance across the various proton-dependent processes likely changes depending on the cellular context, which is another reason why using a fixed stoichiometry to calculate how much ATP is produced from oxygen consumption is so highly flawed. Maximal P/O values that are often used for NAD/FAD linked flux are generated using experimental conditions that favor near complete flux through the ATP synthesis system (supraphysiological substrate and ADP levels). The true P/O value inside living cells is likely to be lower.

RESPONSE: We agree with the reviewer regarding the limitations on calculating ATP production in intact cells based on respiration and proton flux. However, this was only one experiment on which we based our conclusions, as these were also supported by i.e. ATP/ADP ratio measurements and oxygen consumption using different substrates. Therefore, we do not rely exclusively on the ATP production estimative, rather we use this experiment to support complementary methodologies. Nevertheless, we have now better detailed our experimental protocol as well as acknowledged the limitations of the method, so the reader is aware of our procedure and its limitations. We hope the reviewer understands our motivation to perform these experiments and the contribution to our study.

Why are the results in Figure 3K expressed as a % of basal? Could the authors please normalize the OCR data to protein and/or provide a justification for why different normalization strategies were used between 3K and 3M?

RESPONSE: We apologize for the lack of consistency. We have now updated Figure 3 to show all the data in absolute values divided by protein content. This change does not affect the overall interpretation of the findings.

The authors claim that IF1 overexpression lowers ATP production via OXPHOS. However, given the major limitations of this assay (ass discussed above), these claims should be viewed as speculation. This needs to be addressed by the authors as a major limitation. The fact that the ATP/ADP levels did not change do not support of reduction in ATP production, as claimed in the title of Figure 4.

RESPONSE: The reduction in ATP levels and mitochondrial respiration (independent of the substrate offered) suggests a reduction in ATP production rather than an increase in ATP consumption. Moreover, the maintenance of ATP/ADP ratio suggests the existence of a compensatory mechanism to avoid cellular energy crises, which we interpreted as reduced metabolic activity of the cells. Nevertheless, we have now reworded our statements to address the limitations of the methods and our interpretation of the data.

In the discussion, the authors state "However, considering that IF1 inhibits F1-ATP synthase in a 1:1 stoichiometric ratio, the relatively higher expression of IF1 in BAT at room temperature could represent an additional inhibitory factor for ATP synthesis in this tissue." This does not appear to be correct. Although IF1 has been suggested to partially lower maximal rates of ATP synthesis rates, most of this evidence comes from over-expression experiments. According to the current understanding of IF1-CV interaction, the protein is expelled from the complex during rotation in favor of ATP synthesis (PMID: 37002198). It is far more likely that ATP synthesis is low in BAT mitochondria due to the low CV expression. Relative to heart and when normalized to mitochondrial content, CV expression in BAT mitochondria is about 10% that of heart (PMID: 33077793).

RESPONSE: We agree with the reviewer and removed this sentence.

The last sentence of the manuscript states, "Given the importance of IF1 to control brown adipocyte energy metabolism, lowering IF1 levels therapeutically might enhance approaches to enhance NST for improving cardiometabolic health in humans." This sentence seems at odds with the evidence that IF1 levels go up, not down, in human BAT upon cold exposure.

RESPONSE: In light of our new experiments, we have now updated our conclusions.

Minor Comments:

The term "anaerobic glycolysis" is used throughout. All experiments were performed under normoxic conditions, thus the correct term is "aerobic glycolysis.

RESPONSE: Thank you for this comment and we have replaced this term as suggested.

Only male mice were used in the study, could the authors please provide a justification for this?

RESPONSE: Given we devoted most of our efforts to the manipulation of IF1 in vitro, we have used the mouse model as a proof-of-principle on the impact of IF1 in adrenergic-induced thermogenesis. We have now included IF1 KO male and female mice to address the role of IF1 in adrenergic-induced thermogenesis. However, due to the limitation of material, we could only perform AAV in vivo gain-of-function in male mice, therefore, our results cannot be immediately transferred to both sexes, unfortunately.

Reviewer #1 (Significance (Required)):

Overall, the quality of the work is very high. Given that IF1 is implicated across many fields of study, the novel discovery of IF1 as a regulator of brown adipose mitochondrial bioenergetics will be of significance across several fields.

My expertise is in mitochondrial thermodynamics; thus, I do not feel there are any parts of the paper that I do not have sufficient expertise to evaluate.

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

Summary

The manuscript by Brunetta and colleagues conveys the message that the ATPase inhibitory factor 1 (IF1) protein, a physiological inhibitor of mitochondrial ATP synthase, is expressed in BAT of C57BL/6J mice. Moreover, upon cold-adaption of mice they report that the content of IF1 in BAT is downregulated to sustain the mitochondrial membrane potential (MMP) as a result of reverse functioning of the enzyme. In experiments of loss and gain of function of IF1 in cultured brown adipocytes and WT cells they further stress that IF1 silencing promotes metabolic reprogramming to an enhanced glycolysis and lipid oxidation, whereas IF1 overexpression blunts ATP production rendering a quiescent cellular state of the adipocytes.

RESPONSE: We appreciate the time the reviewer invested in our work. Please, see our responses below in a point-by-point manner.

Reviewer #2 (Significance (Required)):

Claims and conclusions:

I have been surprised by the claim that IF1 protein is expressed in BAT under basal conditions and that its expression is downregulated in the cold-adapted tissue. In a previously published work by Forner et al., (2009) Cell Metab 10, 324-335 (reference 43), using a quantitative proteomic approach, it is reported that the mitochondrial proteome of mouse BAT under basal conditions contains a low content of IF1 (at level comparable to the background of the analysis). Remarkably, in the same study they show that there is roughly a 2-fold increase in the content of IF1 protein in mitochondria of BAT at 4d and 24d of cold-adaptation of mice. In other words, just the opposite of what is being reported in the Brunetta study.

RESPONSE: We are aware of the inconsistencies between our findings and Forner et al. (2009). We would like to point out that we have determined IF1 levels in BAT in two separate cohorts with the same findings, and in a third cohort, we observed IF1 mRNA levels to be downregulated in a much shorter timeframe. Our functional analysis is line with this pattern of regulation. A closer look at the supplementary table provided by Forner et al. (2009), shows that the increase in IF1 content following cold exposure is not supported and since we do not have further insight into the methods and analysis employed by the Forner et al. group, we believe a direct comparison should be avoided at the moment. Regarding the baseline levels of IF1 in BAT, the relatively high abundance of IF1 in BAT was also found by another independent group (https://doi.org/10.1101/2020.09.24.311076).

Importantly, the last paragraph of the discussion needs to be amended when mentioning the work of Forner et al. (ref.43). The mentioned reference studied changes in the mouse mitochondrial proteome not in human mitochondria, as it is stated in the alluded paragraph.

RESPONSE: We apologize for this overlook; we have now reworded our statement.

More puzzling are the western blots in Figures 1E, 1H, Supp. Fig. 1C, D were IF1 (ATP5IF1) is identified by a 17kDa band. However, in other Figures (Fig. 2, Fig. 3, Fig. 4, Supp Fig. 2) IF1 is identified by its well-known 12kDa band. What is the reason for this change in labeling of the IF1 band? The reactivity of the anti-IF1 antibody used? It has been previously documented that liver of C57BL/6J and FVB mouse strains do not express IF1 to a significant level when compared to heart IF1 levels (Esparza-Molto (2019) FASEB J. 33, 1836-1851). However, in Fig. 1E they show opposite findings, much higher levels of IF1 in liver than in heart as reveal by the 17kDa band. Moreover, in Fig. 1H they show the vanishing of the 17 kDa band under cold adaptation, which is not the migration of IF1 in gels as shown in their own figures (see Fig. 2, Fig. 3, Fig. 4, Supp Fig. 2). I am certainly reluctant to accept that the 17kDa band shown in Figures 1E, 1H, Supp. Fig. 1C, D is indeed IF1. Most likely it represents a non-specific protein recognized by the antibody in the tissue extracts analyzed. Cellular overexpression experiments of IF1 in WT1 cells (Fig. 2E) and primary brown adipocytes (Fig. 4B) also support this argument. Overall, I do not support publication of this study for the reasons stated above.

RESPONSE: We understand the concerns raised by the reviewer and apologize for the lack of details in our experimental procedures. While we used the same antibody in the study (Cell Sig. cat. Num. 8528, 1:500), we used two different types of gels. The difference in the molecular weight appearance of IF1 is likely through the migration of the protein in the agarose gel. By using custom-made gels, we observe the protein ~17kDa (Fig. 1 and 5), whereas by using commercial gels (Fig. 2, 3, and 4), we observe the protein closer to the predicted molecular weight (i.e. ~12kDa). Of note, gain and loss-of-function experiments, both in vivo as well as in vitro confirm this statement and the specificity of the antibody (Fig. 2, 3, 4, 5, Fig. EV2). In addition, when we ran a custom-made gel with primary BAT cells, we observed again the ~17kDa band (see Figure for the reviewer below). These experiments alongside the absence of other bands in the gels (see uncropped membranes in Supplementary Figure 1) make us conclude that the band we observe is indeed IF1. Nevertheless, we have now updated our methods section, so the reader is aware of our approaches. We hope the reviewer is satisfied with our additional experiments and editions throughout the manuscript.

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

Summary:

In this manuscript, Brunneta et al describe the role of IF1 in brown adipose tissue activation using in vivo and in vitro experimental models. They observed that cold adaptation promotes a reduction in IF1 expression and an increase in the reverse activity of mitochondrial ATPase or Complex V. Based on these results, the authors explore the contribution of IF1 in this metabolic pathway by modeling the thermogenic process in differentiated primary brown adipocytes. They silenced and overexpressed IF1 in culture and studied their adrenergic stimulation under norepinephrine.

Major comments:

The experiments are well explained and the manuscript flows very well. There are several comments that should be addressed.

RESPONSE: We thank the reviewer for the kind words regarding our work.

1. The authors measure ATP hydrolysis in isolated mitochondria from BAT in Figure 1. They observed that IF1 is decreased upon cold exposure and that ATP hydrolysis is increased. They assess protein levels of different OXPHOS proteins, including IF1 but not other proteins of Complex V (ATP5A) as they do in Figures 3 and 4. It is important to see that cold exposure only affects IF1 levels but not other proteins from Complex V. Does IF1/Complex V ratio change? RESPONSE: We thank the reviewer for this suggestion which was also raised by Reviewer #1. We have now measured complex V subunit A in our experimental protocol. We found that cold exposure does not impact complex V protein levels. Given the importance of this information, we have now included it in Figure 1 (Please, see the revised version) alongside the IF1/complex V ratio. In addition, we have now performed WBs in the BAT exposed for 3 and 7 days to thermoneutrality (~28°C) where we found that IF1 is not reduced following whitening of BAT by this approach whilst UCP1 and other mitochondrial proteins are reduced.

This set of data is now included in Figure 1I,K,L.

In Figure 2J, the drop in MMP is lower upon adrenergic stimulation than in Figure 2E. The same observation applies to other results when the reduction in MMP after NE addition is minimal. Why do the authors remove TMRM for the measurements of membrane potential? TMRM imaging is normally done in the presence of the dye in non-quenching mode. Treatments should be done prior to the addition of the dye and then TMRM should be added and left during the imaging analysis and measure in non-quenching mode. This might explain some of the above-mentioned points regarding the MMP data. Alternatively, if the dye is removed before the measurements, they should let the cells to adapt and so the dye equilibrates between mitochondria and cytosol. A more elegant method to measure membrane potential could be live-cell imaging. In addition, authors propose that mitochondrial membrane potential upon NE stimulation is maintained by reversal of ATP synthase. If this is the case, one would expect that addition of oligomycin in NE treated adipocytes would cause depolarization. However, in FigS2A this is not the case. Authors should comment on this in addition to considering more elegant approach to measure MMP.

RESPONSE: We apologize for the lack of details in the methods. All treatments (i.e., transfection and norepinephrine stimulation) were performed before the addition of TMRM. Indeed, this approach does not have the resolution compared to safranine in isolated mitochondria (Fig. 1D), which limits our interpretation regarding the dynamic role of IF1 on MMP in brown adipocytes. We have taken care to state the limitations of our method throughout the entire paper to avoid overinterpretation of our data. Regarding the removal of the dye before the measurements, our internal controls indicate that this procedure does not change the ability of our method to detect fluctuations in MMP (i.e., oligomycin and FCCP as internal controls). Nevertheless, as suggested by the reviewer, to test the time effect of the probe equilibrium (i.e., mitochondria versus cytosol) in our method, we loaded cells with TMRM 20 nM for 30 min and measured the fluorescence right after the removal of the probe/washing steps for another 10 min. We were not able to detect differences in the fluorescence in a time-dependent manner (see below). Therefore, we conclude the removal of TMRM does not influence the fluorescence of the probe in differentiated brown adipocytes.

+NE

-NE

In addition, we performed a similar experiment using TMRM in the quenching mode (200 nM), however, after the removal of TMRM, we added FCCP (1 mM) to the cells for 10 min under constant agitations at 37°C. This approach aimed to expel all TMRM that accumulated within the mitochondria in an MMP-dependent manner. Therefore, excluding the dynamic Brownian movement that we could have caused by the removal of the dye before the measurement mentioned by the reviewer. By doing this, we found the same effect of IF1 overexpression in the reduction of MMP in the presence of norepinephrine.

Protocol:

• Transfection (24h) on day 4 of differentiation + 24h just normal media

• 30 min norepinephrine 10 µM

• 200 nM TMRM on top of NE

• Washing step

• Add FCCP 1 µM for 10 min, and read (The aim here was to release all TMRM accumulated inside of mitochondria in a MMP-dependent manner)

In summary, the data suggests the removal of the dye from the cells does not influence the fluorescence of TMRM, therefore, enabling us to make conclusions regarding the biological effects of IF1 manipulation in the MMP of brown adipocytes. Regarding the reverse mode of ATP synthase and the absence of effects with oligomycin, given oligomycin inhibits both rotation of ATP synthase and even uncoupled brown adipocytes respond to oligomycin (i.e. reduction in O2 consumption), the prediction of lowering MMP in the presence of oligomycin due to inhibition of the reserve mode of ATP synthase is more complicated than anticipated. Nevertheless, we have now addressed this topic in the discussion section. Lastly, we generally observe a reduction in MMP around 10-25% in differentiated adipocytes upon NE treatment (30 minutes, 10mM). However, due to the differentiation state of the cells, MMP response from norepinephrine fluctuated from experiment to experiment. Therefore, we did not compare experiments performed on different days or batches, but only within the same differentiation batch to reduce variability.

In Figure 2, in the model of siIF1, there is baseline more phosphorylation of AMPK than in the scramble control (pAMPK). However, this is not the case of p-p38MAPK. Do the authors have any explanation for those differences in baseline activation of the stress kinases when IF1 is silenced? In the same experimental group, addition of NE seems to have more effect in the scrambled than in siIF1, but the plotted data does not reflect these differences. In contrast, increase in pAMPK upon NE is higher in IF1 overexpressing cells compared to EV (Figure 2H), but again this is not reflected in western blot quantification (Figure 2I).

RESPONSE: Although some differences in pAMPK in the treatments were observed as gathered by the representative blots, these changes were not confirmed later in different biological replicates, therefore, the overall effect of IF1 manipulation in pAMPK does not change. Given we used this approach as quality control for our experiments to guarantee norepinephrine treatment works, we removed the pAMPK data from the study and kept p38 as a marker of adrenergic signaling activation (please see revised Fig. 2 in the main file).

Does NE promote decrease of IF1 expression in control (siScramble and EV) adipocytes? The authors should test it and see whether it goes in the same direction as the observations derived from the experiments in cold exposed mice. This is very important point, as it could explain the lack of an additional effect of IF1 silencing on NE-induced depolarization (Figure 2E).

RESPONSE: We thank the reviewer for this suggestion. In line, with the in vivo data, acute NE treatment in differentiated brown adipocytes does not change IF1 mRNA and protein levels. We have now added this information and the corresponding interpretation to the updated manuscript.

Does NE promote decrease of IF1 expression in the scramble and EV adipocytes? The authors should test it and see whether it goes in the same direction as the observations derived from the experiments in cold exposed mice.

RESPONSE: As this question is the same as #4, we believe the reviewer may have erroneously pasted this here.

For MMP data in Fig2, they should include significance between non treated and NE-treated groups. They say: "While UCP1 ablation did not cause any effect on MMP upon adrenergic stimulation...", but NE caused (probably significant) depolarization in siUCP1, which seems even stronger than depolarization in EV. This is opposite to what you would expect. They also didn't confirm UCP1 silencing with western blot.

RESPONSE: We thank the reviewer for this suggestion. We have now included the expected statistical main effect of NE upon MMP. Although the effects of IF1 overexpression were blunted when Ucp1 was silenced, we indeed still observed the same degree of reduction in MMP in brown adipocytes. This finding has two possible explanations, one is the effectiveness of the silencing protocol, therefore, residual Ucp1 expression may still play a role in this experiment; second, other ATP-consuming processes are able to lower MMP in a UCP1-independent manner. We have added this information to the updated manuscript to make the reader aware of our findings as well as the limitations of the method. Unfortunately, we were not able to detect UCP1 protein levels due to technical issues. Given the effects of IF1 overexpression were blunted when Ucp1 was silenced, we believe this functional outcome is sufficient, alongside mRNA levels, to demonstrate the effectiveness of our silencing protocol.

It has been established that decreased expression of IF1 promotes increase in the reverse activity of Complex V, ATP hydrolytic activity. Increase in ATP hydrolysis also affects ECAR. The authors should consider this when calculating the contribution of ATP glycolysis versus ATP OXPHOS since the ATP hydrolysis is also playing a role in the ECAR increase. The data should be reinterpreted. ATP hydrolysis should be measured in the situation where IF1 is silenced and overexpressed. These measurements can be done in cells using the seahorse.

RESPONSE: The only differences we observed in MMP are in the presence of norepinephrine (i.e. UCP-1-dependent proton conductance), which is not present during the estimation of ATP production by Seahorse analysis. Nevertheless, we have now improved the description of our experimental protocol and limitations to estimate ATP production to make it as clear as possible to the reader. Lastly, given the addition of in vivo gain-of-function experiments, we have now determined the ATP hydrolytic activity in this model, which offers a better understanding of the in vivo modulation of IF1 levels affecting ATP synthase activity (reverse mode). We hope the reviewer understands our motivation to focus on the in vivo model of gain-of-function regarding ATP synthase activity.

The authors use GAPDH as loading control in western blots. They should use another protein since GAPDH is part of the intermediary metabolism and plays a role in glycolysis.

RESPONSE: We understand the concern of the reviewer regarding the use of GAPDH as a loading control for the studies of metabolism. However, as can be observed by the western blot images, GAPDH levels do not change in our experimental models, therefore, we feel confident that our loading is homogeneous throughout our gels.

The authors show that reduction of IF1 involves more lipid utilization. They should include more experiments showing the connection of the metabolic adaptation in the absence of IF1 and some lipid imaging.

RESPONSE: We appreciate this suggestion. We have now performed Oil Red O staining in differentiated adipocytes following ablation of IF1. However, we did not observe any effect on lipid accumulation in primary brown adipocytes following IF1 knockdown. Therefore, the effects of IF1 ablation on lipid mobilization are not due to lipid content or reflected in lipid accumulation. We have now added this new information to the manuscript (please, see the revised form Fig. EV3).

In the text, "Despite this adjustment of experimental conditions, we did not detect any effect of IF1 ablation on mitochondrial oxygen consumption (Supplementary Fig. 3A,B)", this is true for baseline, NE-driven and ATP-linked respiration, but what about maximal respiration? There is a huge increase in IF1 knockdown... They should explain these results.

RESPONSE: We perform this experiment to address the question of whether the lipid mobilization induced by norepinephrine would uncouple mitochondria in a UCP1-independent manner. Given the absence of effect between scrambled and IF1 ablated cells in mitochondrial respiration in the presence of norepinephrine and following the addition of oligomycin, we concluded no effect of lipolysis-induced UCP1-independent uncoupling. However, as observed by the reviewer and consistent with other data within the study, the interaction between lipid metabolism and IF1 knockdown seems to affect maximal electron transport chain activity, which although interesting, was not the focus of the present study. Nevertheless, we have now acknowledged these findings and a possible explanation for them in the revised manuscript.

In Figure 3K they present OCR as % of baseline, but in a similar experiment in Figire 4G it is OCR/protein, they should make the Y axis consistent across experiments.

RESPONSE: We apologize for this overlook. We have now edited all the axes and labels for consistency.

The graphical abstract is confusing. In BAT there are two populations of mitochondria, the cytosolic and the mitochondria attached to the lipid droplet, peridroplet mitochondria (PDM). Upon adrenergic stimulation, PDM leave the lipid droplet and lipolysis takes place. The authors propose that upon adrenergic stimulation, IF1 is reduced and there is lipid mobilization. The part of the scheme where it says "fully recruited" should be removed or rewritten, since adrenergic stimulation is not compatible with mitochondria recruitment around the lipid droplet.

RESPONSE: Thank you for this input. Given the addition of new experiments and interpretation, we have now redrawn the graphical abstract and addressed this topic in the discussion section.

The title should be rewritten to better reflect the research presented in the manuscript.

RESPONSE: Thank you for this input. Given the addition of new experiments, we have now rewritten the title accordingly.

Minor comments:

Some of the Y axis should be corrected. For example, in Figure 2J, L and M should say % of EV untreated, Similarly, in Figure 2E, it should say % of scramble untreated. In Figure 3N, the Y axis is misspelled. All the Y axis referring to percentages should have the same scale for comparison purposes.

RESPONSE: Thank you for the proofreading. We have now edited the scales and labels to keep consistency.

The authors should describe better the results corresponding to Figure 2. There is a lot of information and they should improve the description pertaining the connection between the different pieces of data relating the different signaling pathways that are shown. For westerns in this Figure, they should provide some rationale (one to two sentences in the results section) as to why they are checking the expression of pAMPK and p38-MAPK.

RESPONSE: We have now edited the description of our results to make them as clear as possible.

Here are some comments referring to the methods section:

For Complex V hydrolytic activity, the reaction buffer contains 10mM Na-azide. I guess this is to inhibit respiration, but wouldn't azide also inhibit complex V at this concentration?

RESPONSE: We thank the reviewer for this question. To test that, we performed complex V activity in buffers containing or not 10 mM sodium azide. As demonstrated below, the presence of sodium azide in the buffer does not influence complex V activity in two different tissues with low and high complex V activity (BAT and heart, respectively).

Table 1. ATP synthase hydrolytic activity in the presence or absence of Na-azide.

BAT

Heart

+Na-azide

100 ± 43.01

100 ± 39.36

-Na-azide

82.6 ± 4.33

111.3 ± 43.32

+Na-azide + oligomycin

15.3 ± 4.32*

13.8 ± 14.01*

-Na-azide + oligomycin

14.2 ± 3.53*

11.9 ± 2.88*

Data presented as % of control (i.e. presence of Na-azide and absence of oligomycin) for both tissues independently. N = 2-3/condition. Statistical test: two-way ANOVA. * main effect of oligomycin (p In the mitochondrial isolation protocol, they say "mitochondria were centrifuged at 800g for 10min..." Will this speed pellet the mitochondria? I think this is a mistake in writing.

RESPONSE: We apologize for the lack of clarity. What was centrifuged at 800 g was the whole-tissue homogenate to discard cellular debris, before pelleting mitochondria at 5000 g. We have now corrected this mistake in the methods section.

For the safranin-O experiment, they don't mention mitochondrial substrate used, probably it's in the reference that they provide, but I think it should be included in the text.

RESPONSE: We did not use any substrate because our goal was to test the contribution of ATP synthase to mitochondrial membrane potential. For that, we inhibited proton movement within the ETC with antimycin A and through UCP1 with GDP (see Methods). We have now edited our Method’s description to make sure the reader is aware of our approach.

Reviewer #3 (Significance (Required)):

The manuscript is well written, and it flows well when reading. However, there are some additional experiments that need to be performed to reach the conclusions the authors claim.

RESPONSE: We thank the reviewer for the positive commentaries regarding our work and hope to have answered the open questions with the edits and new experiments.

The role of ATP hydrolysis in BAT thermogenesis is novel and interesting as it can sed some light onto potential approaches to promotes BAT activation.

Reviewer #4 (Evidence, reproducibility and clarity (Required)):

This is an interesting investigation into the activity of IF1 in brown adipocytes. The findings are innovative and the conclusion is well-supported by the data. The conclusion is in line with previous reports on IF1 activities in other cell types, particularly in terms of its regulation of FoF1-ATPase. The authors have executed an exceptional job in designing the study, preparing the figures, and writing the manuscript. Overall, this study significantly contributes to the understanding of IF1 activity in brown adipocytes and its role in thermogenesis.

RESPONSE: We thank the reviewer for the kind words. Please, find below our answers in a point-by-point manner.

Reviewer #4 (Significance (Required)):

The study demonstrates involvement of IF1 in regulating thermogenesis in brown adipocytes, which is a unique aspect not covered in existing literature. Advantage of the study is well-designed cellular studies. The major weakness is lack of proof of conclusion in vivo. There are a few minor concerns that should be addressed to further enhance quality of the manuscript.

RESPONSE: We have now included two in vivo models, whole-body IF1 KO mice and BAT-injected IF1 overexpression to test the role of IF1 in BAT biology. The whole dataset is included in the main manuscript, where we conclude the BAT IF1 overexpression partially suppresses b3-adrenergic induction of thermogenesis alongside a reduction (overall and UCP1 dependent) in mitochondrial oxygen consumption. Also, similar to our in vitro experiments, IF1 KO mice did not present any difference in adrenergic-stimulated oxygen consumption.

1. Current discussion does not mention the regulation of IF1 protein by the cAMP/PKA pathway. This point should be included to provide a comprehensive understanding of the regulatory mechanisms of IF1 protein. RESPONSE: Thank you for this suggestion. We have now added this topic to the discussion.

It has been reported that IF1 also influences the structure of mitochondrial crista. Considering the observed changes with IF1 knockdown, it would be valuable to discuss this activity in relation to the findings of the study.

RESPONSE: We discussed the implications of IF1 modulation in mitochondrial morphology in the revised manuscript.

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

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2023-02218R

Corresponding author(s): Steven, McMahon

1. General Statements [optional]

*We were pleased to receive the encouraging critiques and very much appreciate the Reviewer's specific comments and suggestions. In this revised version of our manuscript, we have made a number of substantive additions and modifications in response to these comments/suggestions. We hope you agree that the study is now improved to the point where it is suitable for publication. *

2. Point-by-point description of the revisions

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

Summary This study describes efforts to characterize differences in the roles of the two related human decapping factors Dcp1a and Dcp1b by assessing mRNA decay and protein associations in knockdown and knockout cell lines. The authors conclude that these proteins are non-redundant based on the observations that loss of DCp1a versus Dcp1b impacts the decapping complex (interactome) and the transcriptome differentially.

Major comments • While the experiments appear to be well designed and executed and the data of generally high quality, the conclusions are drawn without sufficient consideration for the fact that these two proteins form a heterotrimeric complex. The authors assume that there are distinct homotrimeric complexes rather than a single complex with both proteins in. Homotrimers may have new/different functions not normally seen when both proteins are expressed. Thus while it is acceptable to infer that the functions of these two proteins within the decapping complex are distinct, it is not clear that they act separately, or that complexes naturally exist without one or the other. A careful evaluation of the relative ratios of Dcp1a and b overall and in decapping complexes would be informative if the authors want to make stronger statements about the roles of these two factors.

RESPONSE: Thank you for this valuable comment. We have substantially edited the manuscript to incorporate these points. Examples include a detailed analysis of iBAQ values for the DDX6, DCP1a, and DCP1b interactomes (which now allows us to estimate the ratios of DCP1a and DCP1b in these complexes) and cellular fractionation to interrogate complex integrity (using Superose 6).

• The concept of buffering is not adequately introduced and the interpretation of observations that RNAs with increased half life do not show increased protein abundance - that Dcp1a/b are involved in transcript buffering is nebulous. In order to support this interpretation, the mRNA abundances (NOT protein abundances) should be assessed, and even then, there is no way to rule out indirect effects. RESPONSE: Thank you for this comment. In the revised version of the manuscript, we introduced the concept of transcript buffering at an earlier stage as one of the potential explanations for our findings. We were also able to use a new algorithm (grandR) to estimate half-lives and synthesis rates from our data. These new data add strength to the argument that DCP1a and DCP1b are linked to transcript buffering pathways.

• It might be interesting to see what happens when both factors are depleted to get an idea of the overall importance of each one.

RESPONSE: In our work we tried to emphasize the differences between the two paralogs. We believe that doing double knockout or knockdown would mask the distinct impacts of the paralogs. In data not included in this study, we have shown that cells lacking both DCP1a and DCP1b are viable. We did check PARP cleavage in the CRISPR generated cell pools of DCP1a KO, DCP1b KO, and the double KO. The WB measuring the PARP cleavage is shown in the supplemental material (Supplementary Material: Replicates)

• The algorithms etc used for data analysis should be included at the time of publication. Version number and settings used for SMART to define protein domains, and webgestalt should be indicated

RESPONSE: We apologize for this oversight. Version number and settings used for the webtools (SMART, Webgestalt) are now included. The analysis pipeline for half-lives and synthesis rates estimation as well as all the files and the code needed to generate the figures in the paper are available on zenodo (https://zenodo.org/records/10725429).

• Statistical analysis is not provided for the IP experiments, the number of replicates performed is not indicated and quantification of KD efficiency are not provided.

RESPONSE: The number of replicates performed in each experiment is now clearly indicated and quantifications of knockdown efficiency are provided (Supplemental Figure 3A and 3B, Figure 3A, Figure 3B).

• The possibility that the IP Antibody interferes with protein-protein interactions is not mentioned.

RESPONSE: Thank you for this comment. The revised manuscript includes a discussion of the antibody epitope location and the potential for impact on protein-protein interactions.

Minor comments • P4 - "This translational repression of mRNA associated with decapping can be reversed, providing another point at which gene expression can be regulated (21)" - implies that decapping can be reversed or that decapped RNAs are translated. I don't think this is technically true.

RESPONSE: There have been several studies that document the reversal of decapping. These findings are summarized in the following reviews.

Schoenberg, D. R., & Maquat, L. E. (2009). Re-capping the message. Trends in biochemical sciences, 34(9), 435-442.

Trotman, J. B., & Schoenberg, D. R. (2019). A recap of RNA recapping. Wiley Interdisciplinary Reviews: RNA, 10(1), e1504.

• P11 - how common is it for higher eukaryotes to have 2 DCP genes? *RESPONSE: Metazoans have 2 DCP1 genes. *

• Fig S1 - says "mammalian tissues" in the text but the data is all human. The statement that "expression analyses revealed that DCP1a and DCP1b have concordant rather than reciprocal expression patterns across different mammalian tissues (Supplemental Figure 1)" is a bit misleading as no evidence for correlation or anti-correlation is provided. Also co-expression is not strong support for the idea that these genes have non-redundant functions. Both genes are just expressed in all tissues - there's no evidence provided that they are concordantly expressed. In bone marrow it may be worth noting that one is high and the other low - i.e. reciprocal. *RESPONSE: We appreciate this comment. We have corrected the interpretation of the aforementioned dataset. We have also incorporated a more detailed discussion in the text of the paper. As the Reviewer pointed out, there are a subset of tissues where their expression appears to be reciprocal. *

• Fig 1A - it is not clear what the different colors mean. Does Sc DCP1 have 1 larger EVH or 2 distinct ones. Are the low complexity regions in Sc DCP2 the SLiMs. *RESPONSE: Thank you for this comment. We have corrected this ambiguity to reflect that Sc DCP1 has one EVH1 domain that is interconnected by a flexible hinge. The low-complexity regions typically contain short linear motifs (SLIMs), however, not all low-complexity regions have been verified to contain them. In the figure, only low-complexity regions are shown. The text of the paper refers only to verified SLIMs . *

• P11 - why were HCT116 cells selected? RESPONSE: HCT116 cells are an easily transfectable human cell line and have been widely used in biochemical and molecular studies, including studies of mRNA decapping (see references below). Since decapping is impacted by viral proteins we avoided the use of other commonly used cell models such as HEK293T or HeLa.

https://pubmed.ncbi.nlm.nih.gov/?term=decapping+hct116&sort=date&size=200

• Fig 1B - what are the asterisks by the RNA names? Might be worth noting that over-expression of DCP1b reduced IP of DCP1a. There's no quantification and no indication of the number of times this experiment was repeated. Data from replicates and quantification of the knockdown efficiency in each replicate would be nice to see. *RESPONSE: Thank you for this comment. Asterisks indicate that those bands were from a second gel, as DCP1a and DCP1b run at approximately the same molecular weight. We have now included a note in our figure legend to indicate this. The knockdown efficiency is provided (Figure 3 and Supplemental Figure 3). We also noted the number of replicas for each IP in figure 1. The replicas are provided as supplementary material (Supplementary Materials: Replicates). *

• Fig 1C/1D - why are there 3 bands in the DCP1a blot? Quantification of the IP bands is necessary to say whether there is an effect or not of over-expression/KO. RESPONSE: The additional bands in DCP1a blots are background. When we stained the whole blot for DCP1a, in cells which with complete DCP1a KO cells (clone A3), these bands still appear (Supplementary Material: Validation of the KO clones). Quantifications of the bands in the overexpression experiments is now provided.

• Fig 3 - is it possible that differences are due to epitope positions for the antibodies used for IP? RESPONSE: We do not believe so. DCP1a antibody binds roughly 300-400 residues on DCP1a, and DCP1b antibody binds around Val202. Antibodies therefore do not bind DCP1a or DCP1b low-complexity regions (which are largely responsible for interacting with the decapping complex interactome). Antibodies don't bind the EVH1 domains or the trimerization domain, which are needed for their interaction with DCP2 and each other.

• Fig 5A - the legend doesn't match the colors in the figure. It is not clear how the pRESPONSE: Thank you for this comment. We have corrected this issue in the revised version of the paper. High-confidence proteins are those with pRESPONSE: Thank you for this comment. We have corrected this issue in the revised version of the paper.*

• There are a few more recent studies on buffering that should be cited and more discussion of this in the introduction is necessary if conclusions are going to be drawn about buffering. *RESPONSE: We have included a discussion of transcript buffering in the introduction. *

• The heatmaps in figure 2 are hard to interpret. RESPONSE: To clarify the heatmaps, we included a more detailed description in the figure legends, have enlarged the heatmaps themselves, and have added more extensive labeling.

Reviewer #1 (Significance (Required)):

• Strengths: The experiments appear to be done well and the datasets should be useful for the field. • Limitations: The results are overinterpreted - different genes are affected by knocking down one or other of these two similar proteins but this does not really tell us all that much about how the two proteins are functioning in a cell where both are expressed. • Audience: This study will appeal most to a specialized audience consisting of those interested in the basic mechanisms of mRNA decay. Others may find the dataset useful. • This study might complement and/or be informed by another recent study in BioRXiv - https://doi.org/10.1101/2023.09.04.556219 • My field of expertise is mRNA decay - I am qualified to evaluate the findings within the context of this field. I do not have much experience of LC-MS-MS and therefore cannot evaluate the methods/analysis of this part of the study.

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

The authors provide evidence that Dcp1a and Dcp1b - two paralogous proteins of the mRNA decapping complex - may have divergent functions in a cancer cell line. In the first part, the authors show that interaction of Dcp2 with EDC4 is diminished upon depletion of Dcp1a but not affected by depletion of Dcp1b. The results have been controlled by overexpression of Dcp1b as it may be limiting factor (i.e. expression levels too low to compensate for depletion of Dcp1a reduced interaction with EDC3/4 while depletion of Dcp1b lead to opposite and increase interactions). They then defined the protein interactome of DDX6 in parental and Dcp1a or Dcp1b depleted cells. Here, the authors show some differential association with EDC4 again, which is along results shown in the first part. The authors further performed SLAM-seq and identified subsets of mRNA whose decay rates are common but also different upon depletion with Dcp1a and Dcp1b. Interestingly, it seems that Dcp1a preferentially targets mRNAs for proteins regulating lymphocyte differentiation. To further test whether changes in RNA decay rates are also reflected at the protein levels, they finally performed an MS analysis with Dcp1a/b depleted cells. However no significant overlap with mRNAs showing altered stability could be observed; and the authors suggested that the lack of congruence reflects translational repression.

Major comments: 1. While functional difference between Dcp1a and Dcp1b are interesting and likely true, there are overinterpretations that need correction or further evidence for support. Sentences like "DCP1a regulates RNA cap binding proteins association with the decapping complex and DCP1b controls translational initiation factors interactions (Figure 2E)" sound misleading. While differential association with proteins has been recognised with MS-data, it does not necessary implement an active process of control/regulation. To make the claim on 'control/regulation', and inducible system or introduction of mutants would be required.

RESPONSE: This set of comments were particularly useful in helping us refine the presentation of our findings. We have edited our manuscript to be more specific about the limits of our data.

1. The MS analysis is not clearly described in the text and it is unclear how authors selected high-confident proteins. The reader needs to consider the supplemental tables to find out what controls were used. Furthermore, the authors should show correlation plots of MS data between replicates. For instance, there seems to be limited correlation among some of the replicates (e.g. Dcp1b_ko3 sample, Fig. 2c). Any explanation in this variance?

*RESPONSE: We have now included a clear description of how all high-confidence proteins were selected in the Methods and Results sections. The revised manuscript also includes a more thorough description of the controls used and the number of replicates for individual experiments. The PCA plots have now been included where appropriate. The variance in this sample is likely technical. *

1. GO analysis for the proteome analysis should consider the proteome and not the genome as the background. The authors should also indicate the corrected P-values (multiple testing) FDRs.

*RESPONSE: Webgestalt uses a reference set of IDs to recognize the input IDs, and it does not use it for the background analysis in the classical sense. We repeated a subset of our proteome analyses using the 'genome-protein coding' as background and obtained the same result as in our original analysis. All ontology analyses now include raw p-values and/or FDRs when appropriate. *

1. Fig 2E. The figures display GO enrichments needs better explanation and additional data can be added. The enrichment ratio is not explained (is this normalised?) and p-values and FDRs, number of proteins in respective GO category should be added. *RESPONSE: More thorough explanations of the GO enrichments are now included. The supplemental data contains all p-values (raw and adjusted), as well as the number of proteins in each GO category. The Enrichment ratio is normalized and contains information about the number of proteins that are redundant in multiple groups. GO Ontology analyses are now displayed with p-values and/or FDR values, and in this case the enrichment ratio contains information regarding the number of proteins found in our input set and the number of expected proteins in the GO group. The network analysis shows the FDR values and the number of proteins found in the groups compared. *

Minor: 5. These studies were performed in a colorectal carcinoma cell line (HCT116). The authors should justify the choice of this specialised cell line. Furthermore, one wonders whether similar conclusions can be drawn with other cell lines or whether findings are specific to this cancer line.

RESPONSE: The study that is currently in pre-print in BioRxiv (https://doi.org/10.1101/2023.09.04.556219*) utilized HEK293Ts and found similar results to ours when examining the various relationships between the core decapping core members. *

1. Fig. 1B. It is unclear what DCP1b* refers to? There are bands of different size that are not mentioned by the authors - are those protein isoforms or what are those referring to? A molecular marker should be added to each Blots. Uncropped Western images and markers should be provided in the Supplement. *RESPONSE: The asterisk indicates that these images came from a second western blot gel (DCP1a and DCP1b have a similar molecular weight and cannot be probed on the same membrane). Uncropped western blot images and markers (as available) are provided in the supplement. *

2. MS data submitted to public repository with access. No. indicated in the manuscript.

RESPONSE: MS data is submitted as supplementary datasets to the paper. It contains the analyzed data as well as the LCMSMS output. We are in the process of submitting the raw LSMSMS data to a public repository.

Fig 3. A Venn Diagram displaying the overlap of identified proteins should be added. GO analysis should be done considering the proteome as background (as mentioned above).

*RESPONSE: A Venn diagram showing the overlap among the proteins identified is now included in the revised version. *

Reviewer #2 (Significance (Required)):

Overall, this is a large-scale integrative -omics study that suggest functional difference between Dcp1 paralogues. While it seems clear that both paralogous have some different functions and impact, there are overinterpretations in place and further evidence would to be provided to substantiate conclusions made in the paper. For instance, while the interactions with Dcp2/Ddx6 in the absence of Dcp1a,b with EDC4/3 may be altered (Fig. 1, 2), the functional implications of this changed associations remains unresolved and not further discussed. As such, it remains somehow disconnected with the following experiments and compromises the flow of the study. The observed differences in decay-rates for distinct functionally related sets of mRNAs is interesting; however, it remains unclear whether those are direct or rather indirect effects. This is further obscured by the absence of any correlation to changes in protein levels, which the authors interpreted as 'transcriptional buffering'. In this regard, it is puzzling how the authors can make a statement about transcriptional buffering? While this may be an interesting aspect and concept of the discussion, there is no primary data showing such a functional impact.

As such, the study is interesting as it claims functional differences between DCP1a/b paralogous in a cancer cell line. Nevertheless, I am not sure how trustful the MS analysis and decay measurements are as there is not further validation. It woudl be interesting if the authors could go a bit further and draw some hypothesis how the selectivty could be achieved i.e interaction with RNA-binding proteins that may add some specificity towards the target RNAs for differential decay. As such, the study remains unfortunately rather descriptive without further functional insight.

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

Review on "Non-redundant roles for the human mRNA decapping cofactor paralogs DCP1a and DCP1b" by Steven McMahon and co-workers mRNA decay is a critical step in the regulation of gene expression. In eukaryotes, mRNA turnover typically begins with the removal of the poly(A) tail, followed by either removal of the 5' cap structure or exonucleolytic 3'-5' decay catalyzed by the exosome. The decapping enzyme DCP2 forms a complex with its co-activator DCP1, which enhances decapping activity. Mammals are equipped with two DCP1 paralogs, namely DCP1a and DCP1b. Metazoans' decapping complexes feature additional components, such as enhancer of decapping 4 (EDC4), which supports the interaction between DCP1 and DCP2, thereby amplifying the efficiency of decapping. This work focuses on DCP1a and DCP1b and investigates their distinct functions. Using DCP1a- and DCP1a-specific knockdowns as well as K.O. cell lines, the authors find surprising differences between the DCP1 paralogs. While DCP1a is essential for the assembly of EDC4-containig decapping complexes and interactions with mRNA cap binding proteins, DCP1b mediates interactions with the translational machinery. Furthermore, DCP1a and DCP1b target different mRNAs for degradation, indicating that they execute non-overlapping functions. The findings reported here expand our understanding of mRNA decapping in human cells, shedding light on the unique contributions of DCP1a and DCP1b to mRNA metabolism. The manuscript tackles an interesting subject. Historically, the emphasis has been on studying DCP1a, while DCP1b has been deemed a functionally redundant homolog of DCP1a. Therefore, it is commendable that the authors have taken on this topic and, with the help of knockout cell lines, aimed to dissect the function of DCP1a and DCP1b. Despite recognizing the significance of the subject and approach, the manuscript falls short of persuading me. Following a promising start in Figure 1 (which still has room for improvement), there is a distinct decline in overall quality, with only relatively standard analyses being conducted. However, I do not want to give the authors a detailed advice on maximizing the potential of their data and presenting it convincingly. So, here are just a few key points for improvement: Figure 1C: Upon closer examination, a faint band is still visible at the size of DCP1a in the DCP1a knockout cells. Could this be leaky expression of DCP1a? The authors should provide an in-depth characterization of their cells (possibly as supplementary material), including identification of genomic changes (e.g. by sequencing of the locus) and Western blots with longer exposure, etc.

*RESPONSE: Thank you for this comment. The in-depth characterization of our cells is now included in the Supplementary Material. DCP1a KO cells and DCP1b KO cells indicated as single cell clones have been confirmed to have no DCP1a or DCP1b expression. In Figure 1D and Figure 3, polyclonal pool cells were used as indicated (only for DCP1a KO). *

Figure 2: It is great to see that the effects of the KOs are also visible in the DDX6 immunoprecipitation. However, I wonder if the IP clearly confirms that the KO cells indeed do not express DCP1a or DCP1b. In the heatmap in Figure 2B, it appears as if the proteins are only reduced by a log2-fold change of approximately 1.5? Additionally, Figure 2 shows a problem that persists in the subsequent figures. The visual presentation is not particularly appealing, and essential details, such as the scale of the heatmap in 2B (is it log2 fold?), are lacking.

*RESPONSE: The in-depth characterization of our cells is included in the Supplementary Materials and confirms the presence of single-cell clones where indicated. As noted above, only Figure 1D and Figure 3 used DCP1a KO pooled cells. The heatmap in Figure 2B is scaled by row using the pheatfunction in R studio. The actual data for the heatmap comes from protein intensities from the LC-MS/MS analysis. We have improved the visual presentation in the revised manuscript. *

Figure 3: I wonder why there are no primary data shown here, only processed GO analyses. Wouldn't one expect that DCP2 interacts mainly with DCP1a, but less with DCP1b? Is this visible in the data? Moreover, such analyses are rather uninformative (as reflected in the GO terms themselves, for instance, "oxoglutarate dehydrogenase complex" doesn't provide much meaningful insight). The authors should rather try to derive functional and mechanistic insights from their data.

RESPONSE: We have now revised this Figure to include primary data as well as the IP of DCP1a in DCP1b KO cells (single cell clones) and the IP of DCP1b in DCP1a KO cells (pooled cells). We identified EDC3 in the high-confidence protein pool. The EDC3:DCP1a interaction is enhanced in DCP1b KO cells. We also found that the EDC3:DCP1b interaction is less abundant in DCP1a KO cells. This is consistent with our data in Figures 1 and 2. DCP2 was not identified in the interactomes of either DCP1a or DCP1b. This is not unusual as DCP2 is highly flexible and the association between DCP1s with DCP2 is transient and facilitated by other proteins.

In Fig. 4 the potential of the approach is not fully exploited. Firstly, I would advocate for omitting the GO analyses, as, in my opinion, they offer little insight. Again, crucial information is missing to assess the results. While 75 nt reads are mentioned in the methods, the sequencing depth remains unspecified. Figure 4b should be included in the supplements. Furthermore, I strongly recommend concentrating on insights into the mechanisms of DCP1a and DCP1b-containing complexes. E.g. what characteristics distinguish DCP1a and DCP1b-dependent mRNAs? Are these targets inherently unstable? Why are they degraded? Are they known decapping substrates?

*RESPONSE: Thank you for this comment. We have now revised this figure and have included information about sequencing depth and other pertinent information. We have been able to use a newly available algorithm (grandR) and were able to estimate half-lives and synthesis rates. This is a significant addition to the paper. We were also able to compare significantly impacted mRNAs (by DCP1a or DCP1b loss) to the established DCP2 target list. *

In general, I suggest the authors revise the manuscript with a focus on the potential readers. Reduce Gene Ontology (GO) analyses and heatmaps, and instead, incorporate more analyses regarding the molecular processes associated with the different decapping complexes.

*RESPONSE: We removed selected GO analyses and heatmaps from the main body of the manuscript (included as Supplementary Figures instead). For our LC-MS/MS datasets, we added iBAQ analyses of the DDX6 IP, DCP1a IP, and DCP1b IP in the control conditions. Cellular fractionation studies (using Superose 6 chromatography) were also added to the paper and allow us to interrogate decapping complex composition in more detail. The revised version of the manuscript includes a new 4SU labeling experiment (pulse-chase) as well as estimation of half-lives and synthesis rates in our conditions. Also included is relevant information about DCP1b transcriptional regulation. *

Reviewer #3 (Significance (Required)):

The manuscript in its current form could benefit from substantial revisions for it to be considered impactful for researchers in the field.

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

Learn more at Review Commons

---

Reply to the reviewers

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

I have trialled the package on my lab's data and it works as advertised. It was straightforward to use and did not require any special training. I am confident this is a tool that will be approachable even to users with limited computational experience. The use of artificial data to validate the approach - and to provide clear limits on applicability - is particularly helpful.

The main limitation of the tool is that it requires the user to manually select regions. This somewhat limits the generalisability and is also more subjective - users can easily choose "nice" regions that better match with their hypothesis, rather than quantifying the data in an unbiased manner. However, given the inherent challenges in quantifying biological data, such problems are not easily circumventable.

*

* I have some comments to clarify the manuscript:

1. A "straightforward installation" is mentioned. Given this is a Method paper, the means of installation should be clearly laid out.*

__This sentence is now modified. In the revised manuscript we now describe how to install the toolset and we give the link to the toolset website if further information is needed. __On this website, we provide a full video tutorial and a user manual. The user manual is provided as a supplementary material of the manuscript.

* It would be helpful if there was an option to generate an output with the regions analysed (i.e., a JPG image with the data and the drawn line(s) on top). There are two reasons for this: i) A major problem with user-driven quantification is accidental double counting of regions (e.g., a user quantifies a part of an image and then later quantifies the same region). ii) Allows other users to independently verify measurements at a later time.*

We agree that it is helpful to save the analyzed regions. To answer this comment and the other two reviewers' comments pointing at a similar feature, we have now included an automatic saving of the regions of interest. The user will be able to reopen saved regions of interest using a new function we included in the new version of PatternJ.

* 3. Related to the above point, it is highlighted that each time point would need to be analysed separately (line 361-362). It seems like it should be relatively straightforward to allow a function where the analysis line can be mapped onto the next time point. The user could then adjust slightly for changes in position, but still be starting from near the previous timepoint. Given how prevalent timelapse imaging is, this seems like (or something similar) a clear benefit to add to the software.*

We agree that the analysis of time series images can be a useful addition. We have added the analysis of time-lapse series in the new version of PatternJ. The principles behind the analysis of time-lapse series and an example of such analysis are provided in Figure 1 - figure supplement 3 and Figure 5, with accompanying text lines 140-153 and 360-372. The analysis includes a semi-automated selection of regions of interest, which will make the analysis of such sequences more straightforward than having to draw a selection on each image of the series. The user is required to draw at least two regions of interest in two different frames, and the algorithm will automatically generate regions of interest in frames in which selections were not drawn. The algorithm generates the analysis immediately after selections are drawn by the user, which includes the tracking of the reference channel.

* Line 134-135. The level of accuracy of the searching should be clarified here. This is discussed later in the manuscript, but it would be helpful to give readers an idea at this point what level of tolerance the software has to noise and aperiodicity.

*

We agree with the reviewer that a clarification of this part of the algorithm will help the user better understand the manuscript.__ We have modified the sentence to clarify the range of search used and the resulting limits in aperiodicity (now lines 176-181). __Regarding the tolerance to noise, it is difficult to estimate it a priori from the choice made at the algorithm stage, so we prefer to leave it to the validation part of the manuscript. We hope this solution satisfies the reviewer and future users.

*

**Referees cross-commenting**

I think the other reviewer comments are very pertinent. The authors have a fair bit to do, but they are reasonable requests. So, they should be encouraged to do the revisions fully so that the final software tool is as useful as possible.

Reviewer #1 (Significance (Required)):

Developing software tools for quantifying biological data that are approachable for a wide range of users remains a longstanding challenge. This challenge is due to: (1) the inherent problem of variability in biological systems; (2) the complexity of defining clearly quantifiable measurables; and (3) the broad spread of computational skills amongst likely users of such software.

In this work, Blin et al., develop a simple plugin for ImageJ designed to quickly and easily quantify regular repeating units within biological systems - e.g., muscle fibre structure. They clearly and fairly discuss existing tools, with their pros and cons. The motivation for PatternJ is properly justified (which is sadly not always the case with such software tools).

Overall, the paper is well written and accessible. The tool has limitations but it is clearly useful and easy to use. Therefore, this work is publishable with only minor corrections.

*We thank the reviewer for the positive evaluation of PatternJ and for pointing out its accessibility to the users.

*

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

# Summary

The authors present an ImageJ Macro GUI tool set for the quantification of one-dimensional repeated patterns that are commonly occurring in microscopy images of muscles.

# Major comments

In our view the article and also software could be improved in terms of defining the scope of its applicability and user-ship. In many parts the article and software suggest that general biological patterns can be analysed, but then in other parts very specific muscle actin wordings are used. We are pointing this out in the "Minor comments" sections below. We feel that the authors could improve their work by making a clear choice here. One option would be to clearly limit the scope of the tool to the analysis of actin structures in muscles. In this case we would recommend to also rename the tool, e.g. MusclePatternJ. The other option would be to make the tool about the generic analysis of one-dimensional patterns, maybe calling the tool LinePatternJ. In the latter case we would recommend to remove all actin specific wordings from the macro tool set and also the article should be in parts slightly re-written.

*

We agree with the reviewer that our initial manuscript used a mix of general and muscle-oriented vocabulary, which could make the use of PatternJ confusing especially outside of the muscle field. To make PatternJ useful for the largest community, we corrected the manuscript and the PatternJ toolset to provide the general vocabulary needed to make it understandable for every biologist. We modified the manuscript accordingly.

* # Minor/detailed comments

# Software

We recommend considering the following suggestions for improving the software.

## File and folder selection dialogs

In general, clicking on many of the buttons just opens up a file-browser dialog without any further information. For novel users it is not clear what the tool expects one to select here. It would be very good if the software could be rewritten such that there are always clear instructions displayed about which file or folder one should open for the different buttons.*

We experienced with the current version of macOS that the file-browser dialog does not display any message; we suspect this is the issue raised by the reviewer. This is a known issue of Fiji on Mac and all applications on Mac since 2016. We provided guidelines in the user manual and on the tutorial video to correct this issue by changing a parameter in Fiji. Given the issues the reviewer had accessing the material on the PatternJ website, which we apologize for, we understand the issue raised. We added an extra warning on the PatternJ website to point at this problem and its solution. Additionally, we have limited the file-browser dialog appearance to what we thought was strictly necessary. Thus, the user will experience fewer prompts, speeding up the analysis.

*

## Extract button

The tool asks one to specify things like whether selections are drawn "M-line-to-M-line"; for users that are not experts in muscle morphology this is not understandable. It would be great to find more generally applicable formulations. *

We agree that this muscle-oriented vocabulary can make the use of PatternJ confusing. We have now corrected the user interface to provide both general and muscle-specific vocabulary ("center-to-center or edge-to-edge (M-line-to-M-line or Z-disc-to-Z-disc)").*

## Manual selection accuracy

The 1st step of the analysis is always to start from a user hand-drawn profile across intensity patterns in the image. However, this step can cause inaccuracy that varies with the shape and curve of the line profile drawn. If not strictly perpendicular to for example the M line patterns, the distance between intensity peaks will be different. This will be more problematic when dealing with non-straight and parallelly poised features in the image. If the structure is bended with a curve, the line drawn over it also needs to reproduce this curve, to precisely capture the intensity pattern. I found this limits the reproducibility and easy-usability of the software.*

We understand the concern of the reviewer. On curved selections this will be an issue that is difficult to solve, especially on "S" curved or more complex selections. The user will have to be very careful in these situations. On non-curved samples, the issue may be concerning at first sight, but the errors go with the inverse of cosine and are therefore rather low. For example, if the user creates a selection off by 5 degrees, which is visually obvious, lengths will be affected by an increase of only 0.38%. The point raised by the reviewer is important to discuss, and we therefore added a paragraph to comment on the choice of selection (lines 94-98) and a supplementary figure to help make it clear (Figure 1 - figure supplement 1).*

### Reproducibility

Since the line profile drawn on the image is the first step and very essential to the entire process, it should be considered to save together with the analysis result. For example, as ImageJ ROI or ROIset files that can be re-imported, correctly positioned, and visualized in the measured images. This would greatly improve the reproducibility of the proposed workflow. In the manuscript, only the extracted features are being saved (because the save button is also just asking for a folder containing images, so I cannot verify its functionality). *

We agree that this is a very useful and important feature. We have added ROI automatic saving. Additionally, we now provide a simplified import function of all ROIs generated with PatternJ and the automated extraction and analysis of the list of ROIs. This can be done from ROIs generated previously in PatternJ or with ROIs generated from other ImageJ/Fiji algorithms. These new features are described in the manuscript in lines 120-121 and 130-132.

*

## ? button

It would be great if that button would open up some usage instructions.

*

We agree with the reviewer that the "?" button can be used in a better way. We have replaced this button with a Help menu, including a simple tutorial showing a series of images detailing the steps to follow by the user, a link to the user website, and a link to our video tutorial.

* ## Easy improvement of workflow

I would suggest a reasonable expansion of the current workflow, by fitting and displaying 2D lines to the band or line structure in the image, that form the "patterns" the author aims to address. Thus, it extracts geometry models from the image, and the inter-line distance, and even the curve formed by these sets of lines can be further analyzed and studied. These fitted 2D lines can be also well integrated into ImageJ as Line ROI, and thus be saved, imported back, and checked or being further modified. I think this can largely increase the usefulness and reproducibility of the software.

*

We hope that we understood this comment correctly. We had sent a clarification request to the editor, but unfortunately did not receive an answer within the requested 4 weeks of this revision. We understood the following: instead of using our 1D approach, in which we extract positions from a profile, the reviewer suggests extracting the positions of features not as a single point, but as a series of coordinates defining its shape. If this is the case, this is a major modification of the tool that is beyond the scope of PatternJ. We believe that keeping our tool simple, makes it robust. This is the major strength of PatternJ. Local fitting will not use line average for instance, which would make the tool less reliable.

* # Manuscript

We recommend considering the following suggestions for improving the manuscript. Abstract: The abstract suggests that general patterns can be quantified, however the actual tool quantifies specific subtypes of one-dimensional patterns. We recommend adapting the abstract accordingly.

*

We modified the abstract to make this point clearer.

* Line 58: Gray-level co-occurrence matrix (GLCM) based feature extraction and analysis approach is not mentioned nor compared. At least there's a relatively recent study on Sarcomeres structure based on GLCM feature extraction: https://github.com/steinjm/SotaTool with publication: *https://doi.org/10.1002/cpz1.462

• *

We thank the reviewer for making us aware of this publication. We cite it now and have added it to our comparison of available approaches.

* Line 75: "...these simple geometrical features will address most quantitative needs..." We feel that this may be an overstatement, e.g. we can imagine that there should be many relevant two-dimensional patterns in biology?!*

We have modified this sentence to avoid potential confusion (lines 76-77).

• *

• Line 83: "After a straightforward installation by the user, ...". We think it would be convenient to add the installation steps at this place into the manuscript. *

__This sentence is now modified. We now mention how to install the toolset and we provide the link to the toolset website, if further information is needed (lines 86-88). __On the website, we provide a full video tutorial and a user manual.

* Line 87: "Multicolor images will give a graph with one profile per color." The 'Multicolor images' here should be more precisely stated as "multi-channel" images. Multi-color images could be confused with RGB images which will be treated as 8-bit gray value (type conversion first) images by profile plot in ImageJ. *

We agree with the reviewer that this could create some confusion. We modified "multicolor" to "multi-channel".

* Line 92: "...such as individual bands, blocks, or sarcomeric actin...". While bands and blocks are generic pattern terms, the biological term "sarcomeric actin" does not seem to fit in this list. Could a more generic wording be found, such as "block with spike"? *

We agree with the reviewer that "sarcomeric actin" alone will not be clear to all readers. We modified the text to "block with a central band, as often observed in the muscle field for sarcomeric actin" (lines 103-104). The toolset was modified accordingly.

* Line 95: "the algorithm defines one pattern by having the features of highest intensity in its centre". Could this be rephrased? We did not understand what that exactly means.*

We agree with the reviewer that this was not clear. We rewrote this paragraph (lines 101-114) and provided a supplementary figure to illustrate these definitions (Figure 1 - figure supplement 2).

* Line 124 - 147: This part the only description of the algorithm behind the feature extraction and analysis, but not clearly stated. Many details are missing or assumed known by the reader. For example, how it achieved sub-pixel resolution results is not clear. One can only assume that by fitting Gaussian to the band, the center position (peak) thus can be calculated from continuous curves other than pixels. *

Note that the two sentences introducing this description are "Automated feature extraction is the core of the tool. The algorithm takes multiple steps to achieve this (Fig. S2):". We were hoping this statement was clear, but the reviewer may refer to something else. We agree that the description of some of the details of the steps was too quick. We have now expanded the description where needed.

* Line 407: We think the availability of both the tool and the code could be improved. For Fiji tools it is common practice to create an Update Site and to make the code available on GitHub. In addition, downloading the example file (https://drive.google.com/file/d/1eMazyQJlisWPwmozvyb8VPVbfAgaH7Hz/view?usp=drive_link) required a Google login and access request, which is not very convenient; in fact, we asked for access but it was denied. It would be important for the download to be easier, e.g. from GitHub or Zenodo.

*

We are sorry for issues encountered when downloading the tool and additional material. We thank the reviewer for pointing out these issues that limited the accessibility of our tool. We simplified the downloading procedure on the website, which does not go through the google drive interface nor requires a google account. Additionally, for the coder community the code, user manual and examples are now available from GitHub at github.com/PierreMangeol/PatternJ, and are provided as supplementary material with the manuscript. To our knowledge, update sites work for plugins but not for macro toolsets. Having experience sharing our codes with non-specialists, a classical website with a tutorial video is more accessible than more coder-oriented websites, which deter many users.

* Reviewer #2 (Significance (Required)):

The strength of this study is that a tool for the analysis of one-dimensional repeated patterns occurring in muscle fibres is made available in the accessible open-source platform ImageJ/Fiji. In the introduction to the article the authors provide an extensive review of comparable existing tools. Their new tool fills a gap in terms of providing an easy-to-use software for users without computational skills that enables the analysis of muscle sarcomere patterns. We feel that if the below mentioned limitations could be addressed the tool could indeed be valuable to life scientists interested in muscle patterning without computational skills.

In our view there are a few limitations, including the accessibility of example data and tutorials at sites.google.com/view/patternj, which we had trouble to access. In addition, we think that the workflow in Fiji, which currently requires pressing several buttons in the correct order, could be further simplified and streamlined by adopting some "wizard" approach, where the user is guided through the steps.

*As answered above, the links on the PatternJ website are now corrected. Regarding the workflow, we now provide a Help menu with:

1. __a basic set of instructions to use the tool, __
2. a direct link to the tutorial video in the PatternJ toolset
3. a direct link to the website on which both the tutorial video and a detailed user manual can be found. We hope this addresses the issues raised by this reviewer.

*Another limitation is the reproducibility of the analysis; here we recommend enabling IJ Macro recording as well as saving of the drawn line ROIs. For more detailed suggestions for improvements please see the above sections of our review. *

We agree that saving ROIs is very useful. It is now implemented in PatternJ.

We are not sure what this reviewer means by "enabling IJ Macro recording". The ImageJ Macro Recorder is indeed very useful, but to our knowledge, it is limited to built-in functions. Our code is open and we hope this will be sufficient for advanced users to modify the code and make it fit their needs.*

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

Summary In this manuscript, the authors present a new toolset for the analysis of repetitive patterns in biological images named PatternJ. One of the main advantages of this new tool over existing ones is that it is simple to install and run and does not require any coding skills whatsoever, since it runs on the ImageJ GUI. Another advantage is that it does not only provide the mean length of the pattern unit but also the subpixel localization of each unit and the distributions of lengths and that it does not require GPU processing to run, unlike other existing tools. The major disadvantage of the PatternJ is that it requires heavy, although very simple, user input in both the selection of the region to be analyzed and in the analysis steps. Another limitation is that, at least in its current version, PatternJ is not suitable for time-lapse imaging. The authors clearly explain the algorithm used by the tool to find the localization of pattern features and they thoroughly test the limits of their tool in conditions of varying SNR, periodicity and band intensity. Finally, they also show the performance of PatternJ across several biological models such as different kinds of muscle cells, neurons and fish embryonic somites, as well as different imaging modalities such as brightfield, fluorescence confocal microscopy, STORM and even electron microscopy.

This manuscript is clearly written, and both the section and the figures are well organized and tell a cohesive story. By testing PatternJ, I can attest to its ease of installation and use. Overall, I consider that PatternJ is a useful tool for the analysis of patterned microscopy images and this article is fit for publication. However, i do have some minor suggestions and questions that I would like the authors to address, as I consider they could improve this manuscript and the tool:

*We are grateful to this reviewer for this very positive assessment of PatternJ and of our manuscript.

* Minor Suggestions: In the methodology section is missing a more detailed description about how the metric plotted was obtained: as normalized intensity or precision in pixels. *

We agree with the reviewer that a more detailed description of the metric plotted was missing. We added this information in the method part and added information in the Figure captions where more details could help to clarify the value displayed.

* The validation is based mostly on the SNR and patterns. They should include a dataset of real data to validate the algorithm in three of the standard patterns tested. *

We validated our tool using computer-generated images, in which we know with certainty the localization of patterns. This allowed us to automatically analyze 30 000 images, and with varying settings, we sometimes analyzed 10 times the same image, leading to about 150 000 selections analyzed. From these analyses, we can provide with confidence an unbiased assessment of the tool precision and the tool capacity to extract patterns. We already provided examples of various biological data images in Figures 4-6, showing all possible features that can be extracted with PatternJ. In these examples, we can claim by eye that PatternJ extracts patterns efficiently, but we cannot know how precise these extractions are because of the nature of biological data: "real" positions of features are unknown in biological data. Such validation will be limited to assessing whether a pattern was found or not, which we believe we already provided with the examples in Figures 4-6.

* The video tutorial available in the PatternJ website is very useful, maybe it would be worth it to include it as supplemental material for this manuscript, if the journal allows it. *

As the video tutorial may have been missed by other reviewers, we agree it is important to make it more prominent to users. We have now added a Help menu in the toolset that opens the tutorial video. Having the video as supplementary material could indeed be a useful addition if the size of the video is compatible with the journal limits.

* An example image is provided to test the macro. However, it would be useful to provide further example images for each of the three possible standard patterns suggested: Block, actin sarcomere or individual band.*

We agree this can help users. We now provide another multi-channel example image on the PatternJ website including blocks and a pattern made of a linear intensity gradient that can be extracted with our simpler "single pattern" algorithm, which were missing in the first example. Additionally, we provide an example to be used with our new time-lapse analysis.

* Access to both the manual and the sample images in the PatternJ website should be made publicly available. Right now they both sit in a private Drive account. *

As mentioned above, we apologize for access issues that occurred during the review process. These files can now be downloaded directly on the website without any sort of authentication. Additionally, these files are now also available on GitHub.

* Some common errors are not properly handled by the macro and could be confusing for the user: When there is no selection and one tries to run a Check or Extraction: "Selection required in line 307 (called from line 14). profile=getProfile( ;". A simple "a line selection is required" message would be useful there. When "band" or "block" is selected for a channel in the "Set parameters" window, yet a 0 value is entered into the corresponding "Number of bands or blocks" section, one gets this error when trying to Extract: "Empty array in line 842 (called from line 113). if ( ( subloc . length == 1 ) & ( subloc [ 0 == 0) ) {". This error is not too rare, since the "Number of bands or blocks" section is populated with a 0 after choosing "sarcomeric actin" (after accepting the settings) and stays that way when one changes back to "blocks" or "bands".*

We thank the reviewer for pointing out these bugs. These bugs are now corrected in the revised version.

* The fact that every time one clicks on the most used buttons, the getDirectory window appears is not only quite annoying but also, ultimately a waste of time. Isn't it possible to choose the directory in which to store the files only once, from the "Set parameters" window?*

We have now found a solution to avoid this step. The user is only prompted to provide the image folder when pressing the "Set parameter" button. We kept the prompt for directory only when the user selects the time-lapse analysis or the analysis of multiple ROIs. The main reason is that it is very easy for the analysis to end up in the wrong folder otherwise.

* The authors state that the outputs of the workflow are "user friendly text files". However, some of them lack descriptive headers (like the localisations and profiles) or even file names (like colors.txt). If there is something lacking in the manuscript, it is a brief description of all the output files generated during the workflow.*

PatternJ generates multiple files, several of which are internal to the toolset. They are needed to keep track of which analyses were done, and which colors were used in the images, amongst others. From the user part, only the files obtained after the analysis All_localizations.channel_X.txt and sarcomere_lengths.txt are useful. To improve the user experience, we now moved all internal files to a folder named "internal", which we think will clarify which outputs are useful for further analysis, and which ones are not. We thank the reviewer for raising this point and we now mention it in our Tutorial.

I don't really see the point in saving the localizations from the "Extraction" step, they are even named "temp".

We thank the reviewer for this comment, this was indeed not necessary. We modified PatternJ to delete these files after they are used.

* In the same line, I DO see the point of saving the profiles and localizations from the "Extract & Save" step, but I think they should be deleted during the "Analysis" step, since all their information is then grouped in a single file, with descriptive headers. This deleting could be optional and set in the "Set parameters" window.*

We understand the point raised by the reviewer. However, the analysis depends on the reference channel picked, which is asked for when starting an analysis, and can be augmented with additional selections. If a user chooses to modify the reference channel or to add a new profile to the analysis, deleting all these files would mean that the user will have to start over again, which we believe will create frustration. An optional deletion at the analysis step is simple to implement, but it could create problems for users who do not understand what it means practically.

* Moreover, I think it would be useful to also save the linear roi used for the "Extract & Save" step, and eventually combine them during the "Analysis step" into a single roi set file so that future re-analysis could be made on the same regions. This could be an optional feature set from the "Set parameters" window. *

We agree with the reviewer that saving ROIs is very useful. ROIs are now saved into a single file each time the user extracts and saves positions from a selection. Additionally, the user can re-use previous ROIs and analyze an image or image series in a single step.

* In the "PatternJ workflow" section of the manuscript, the authors state that after the "Extract & Save" step "(...) steps 1, 2, 4, and 5 can be repeated on other selections (...)". However, technically, only steps 1 and 5 are really necessary (alternatively 1, 4 and 5 if the user is unsure of the quality of the patterning). If a user follows this to the letter, I think it can lead to wasted time.

*

We agree with the reviewer and have corrected the manuscript accordingly (line 119-120).

• *

*I believe that the "Version Information" button, although important, has potential to be more useful if used as a "Help" button for the toolset. There could be links to useful sources like the manuscript or the PatternJ website but also some tips like "whenever possible, use a higher linewidth for your line selection" *

We agree with the reviewer as pointed out in our previous answers to the other reviewers. This button is now replaced by a Help menu, including a simple tutorial in a series of images detailing the steps to follow, a link to the user website, and a link to our video tutorial.

* It would be interesting to mention to what extent does the orientation of the line selection in relation to the patterned structure (i.e. perfectly parallel vs more diagonal) affect pattern length variability?*

As answered to reviewer 1, we understand this concern, which needs to be clarified for readers. The issue may be concerning at first sight, but the errors grow only with the inverse of cosine and are therefore rather low. For example, if the user creates a selection off by 3 degrees, which is visually obvious, lengths will be affected by an increase of only 0.14%. The point raised by the reviewer is important to discuss, and we therefore have added a comment on the choice of selection (lines 94-98) as well as a supplementary figure (Figure 1 - figure supplement 1).

* When "the algorithm uses the peak of highest intensity as a starting point and then searches for peak intensity values one spatial period away on each side of this starting point" (line 133-135), does that search have a range? If so, what is the range? *

We agree that this information is useful to share with the reader. The range is one pattern size. We have modified the sentence to clarify the range of search used and the resulting limits in aperiodicity (now lines 176-181).

* Line 144 states that the parameters of the fit are saved and given to the user, yet I could not find such information in the outputs. *

The parameters of the fits are saved for blocks. We have now clarified this point by modifying the manuscript (lines 186-198) and modifying Figure 1 - figure supplement 5. We realized we made an error in the description of how edges of "block with middle band" are extracted. This is now corrected.

* In line 286, authors finish by saying "More complex patterns from electron microscopy images may also be used with PatternJ.". Since this statement is not backed by evidence in the manuscript, I suggest deleting it (or at the very least, providing some examples of what more complex patterns the authors refer to). *

This sentence is now deleted.

* In the TEM image of the fly wing muscle in fig. 4 there is a subtle but clearly visible white stripe pattern in the original image. Since that pattern consists of 'dips', rather than 'peaks' in the profile of the inverted image, they do not get analyzed. I think it is worth mentioning that if the image of interest contains both "bright" and "dark" patterns, then the analysis should be performed in both the original and the inverted images because the nature of the algorithm does not allow it to detect "dark" patterns. *

We agree with the reviewer's comment. We now mention this point in lines 337-339.

* In line 283, the authors mention using background correction. They should explicit what method of background correction they used. If they used ImageJ's "subtract background' tool, then specify the radius.*

We now describe this step in the method section.

*

Reviewer #3 (Significance (Required)):

• Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. Being a software paper, the advance proposed by the authors is technical in nature. The novelty and significance of this tool is that it offers quick and simple pattern analysis at the single unit level to a broad audience, since it runs on the ImageJ GUI and does not require any programming knowledge. Moreover, all the modules and steps are well described in the paper, which allows easy going through the analysis.
• Place the work in the context of the existing literature (provide references, where appropriate). The authors themselves provide a good and thorough comparison of their tool with other existing ones, both in terms of ease of use and on the type of information extracted by each method. While PatternJ is not necessarily superior in all aspects, it succeeds at providing precise single pattern unit measurements in a user-friendly manner.
• State what audience might be interested in and influenced by the reported findings. Most researchers working with microscopy images of muscle cells or fibers or any other patterned sample and interested in analyzing changes in that pattern in response to perturbations, time, development, etc. could use this tool to obtain useful, and otherwise laborious, information. *

We thank the reviewer for these enthusiastic comments about how straightforward for biologists it is to use PatternJ and its broad applicability in the bio community.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary

The authors present an ImageJ Macro GUI tool set for the quantification of one-dimensional repeated patterns that are commonly occurring in microscopy images of muscles.

Major comments

In our view the article and also software could be improved in terms of defining the scope of its applicability and user-ship. In many parts the article and software suggest that general biological patterns can be analysed, but then in other parts very specific muscle actin wordings are used. We are pointing this out in the "Minor comments" sections below. We feel that the authors could improve their work by making a clear choice here. One option would be to clearly limit the scope of the tool to the analysis of actin structures in muscles. In this case we would recommend to also rename the tool, e.g. MusclePatternJ. The other option would be to make the tool about the generic analysis of one-dimensional patterns, maybe calling the tool LinePatternJ. In the latter case we would recommend to remove all actin specific wordings from the macro tool set and also the article should be in parts slightly re-written.

Minor/detailed comments

Software

We recommend considering the following suggestions for improving the software.

File and folder selection dialogs

In general, clicking on many of the buttons just opens up a file-browser dialog without any further information. For novel users it is not clear what the tool expects one to select here. It would be very good if the software could be rewritten such that there are always clear instructions displayed about which file or folder one should open for the different buttons.

Extract button

The tool asks one to specify things like whether selections are drawn "M-line-to-M-line"; for users that are not experts in muscle morphology this is not understandable. It would be great to find more generally applicable formulations.

Manual selection accuracy

The 1st step of the analysis is always to start from a user hand-drawn profile across intensity patterns in the image. However, this step can cause inaccuracy that varies with the shape and curve of the line profile drawn. If not strictly perpendicular to for example the M line patterns, the distance between intensity peaks will be different. This will be more problematic when dealing with non-straight and parallelly poised features in the image. If the structure is bended with a curve, the line drawn over it also needs to reproduce this curve, to precisely capture the intensity pattern. I found this limits the reproducibility and easy-usability of the software.

Reproducibility

Since the line profile drawn on the image is the first step and very essential to the entire process, it should be considered to save together with the analysis result. For example, as ImageJ ROI or ROIset files that can be re-imported, correctly positioned, and visualized in the measured images. This would greatly improve the reproducibility of the proposed workflow. In the manuscript, only the extracted features are being saved (because the save button is also just asking for a folder containing images, so I cannot verify its functionality).

? button

It would be great if that button would open up some usage instructions.

Easy improvement of workflow

I would suggest a reasonable expansion of the current workflow, by fitting and displaying 2D lines to the band or line structure in the image, that form the "patterns" the author aims to address. Thus, it extracts geometry models from the image, and the inter-line distance, and even the curve formed by these sets of lines can be further analyzed and studied. These fitted 2D lines can be also well integrated into ImageJ as Line ROI, and thus be saved, imported back, and checked or being further modified. I think this can largely increase the usefulness and reproducibility of the software.

Manuscript

We recommend considering the following suggestions for improving the manuscript. Abstract: The abstract suggests that general patterns can be quantified, however the actual tool quantifies specific subtypes of one-dimensional patterns. We recommend adapting the abstract accordingly.

Line 58: Gray-level co-occurrence matrix (GLCM) based feature extraction and analysis approach is not mentioned nor compared. At least there's a relatively recent study on Sarcomeres structure based on GLCM feature extraction: https://github.com/steinjm/SotaTool with publication: https://doi.org/10.1002/cpz1.462

Line 75: "...these simple geometrical features will address most quantitative needs..." We feel that this may be an overstatement, e.g. we can imagine that there should be many relevant two-dimensional patterns in biology?!

Line 83: "After a straightforward installation by the user, ...". We think it would be convenient to add the installation steps at this place into the manuscript.

Line 87: "Multicolor images will give a graph with one profile per color." The 'Multicolor images' here should be more precisely stated as "multi-channel" images. Multi-color images could be confused with RGB images which will be treated as 8-bit gray value (type conversion first) images by profile plot in ImageJ.

Line 92: "...such as individual bands, blocks, or sarcomeric actin...". While bands and blocks are generic pattern terms, the biological term "sarcomeric actin" does not seem to fit in this list. Could a more generic wording be found, such as "block with spike"?

Line 95: "the algorithm defines one pattern by having the features of highest intensity in its centre". Could this be rephrased? We did not understand what that exactly means.

Line 124 - 147: This part the only description of the algorithm behind the feature extraction and analysis, but not clearly stated. Many details are missing or assumed known by the reader. For example, how it achieved sub-pixel resolution results is not clear. One can only assume that by fitting Gaussian to the band, the center position (peak) thus can be calculated from continuous curves other than pixels.

Line 407: We think the availability of both the tool and the code could be improved. For Fiji tools it is common practice to create an Update Site and to make the code available on GitHub. In addition, downloading the example file (https://drive.google.com/file/d/1eMazyQJlisWPwmozvyb8VPVbfAgaH7Hz/view?usp=drive_link) required a Google login and access request, which is not very convenient; in fact, we asked for access but it was denied. It would be important for the download to be easier, e.g. from GitHub or Zenodo.

Significance

The strength of this study is that a tool for the analysis of one-dimensional repeated patterns occurring in muscle fibres is made available in the accessible open-source platform ImageJ/Fiji. In the introduction to the article the authors provide an extensive review of comparable existing tools. Their new tool fills a gap in terms of providing an easy-to-use software for users without computational skills that enables the analysis of muscle sarcomere patterns. We feel that if the below mentioned limitations could be addressed the tool could indeed be valuable to life scientists interested in muscle patterning without computational skills.

In our view there are a few limitations, including the accessibility of example data and tutorials at sites.google.com/view/patternj, which we had trouble to access. In addition, we think that the workflow in Fiji, which currently requires pressing several buttons in the correct order, could be further simplified and streamlined by adopting some "wizard" approach, where the user is guided through the steps. Another limitation is the reproducibility of the analysis; here we recommend enabling IJ Macro recording as well as saving of the drawn line ROIs. For more detailed suggestions for improvements please see the above sections of our review.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, Bell et al. provide an exhaustive and clear description of the diversity of a new class of predicted type IV restriction systems that the authors denote as CoCoNuTs, for their characteristic presence of coiled-coil segments and nuclease tandems. Along with a comprehensive analysis that includes phylogenetics, protein structure prediction, extensive protein domain annotations, and an in-depth investigation of encoding genomic contexts, they also provide detailed hypotheses about the biological activity and molecular functions of the members of this class of predicted systems. This work is highly relevant, it underscores the wide diversity of defence systems that are used by prokaryotes and demonstrates that there are still many systems to be discovered. The work is sound and backed-up by a clear and reasonable bioinformatics approach. I do not have any major issues with the manuscript, but only some minor comments.

Strengths:

The analysis provided by the authors is extensive and covers the three most important aspects that can be covered computationally when analysing a new family/superfamily: phylogenetics, genomic context analysis, and protein-structure-based domain content annotation. With this, one can directly have an idea about the superfamily of the predicted system and infer their biological role. The bioinformatics approach is sound and makes use of the most current advances in the fields of protein evolution and structural bioinformatics.

Weaknesses:

It is not clear how coiled-coil segments were assigned if only based on AF2-predicted models or also backed by sequence analysis, as no description is provided in the methods. The structure prediction quality assessment is based solely on the average pLDDT of the obtained models (with a threshold of 80 or better). However, this is not enough, particularly when multimeric models are used. The PAE matrix should be used to evaluate relative orientations, particularly in the case where there is a prediction that parts from 2 proteins are interacting. In the case of multimers, interface quality scores, such as the ipTM or pDockQ, should also be considered and, at minimum, reported.

A description of the coiled-coil predictions has been added to the Methods. For multimeric models, PAE matrices and ipTM+pTM scores have been included in Supplementary Data File S1.

Reviewer #2 (Public Review):

Summary:

In this work, using in-depth computational analysis, Bell et al. explore the diverse repertoire of type IV McrBC modification-dependent restriction systems. The prototypical two-component McrBC system has been structurally and functionally characterised and is known to act as a defence by restricting phage and foreign DNA containing methylated cytosines. Here, the authors find previously unanticipated complexity and versatility of these systems and focus on detailed analysis and classification of a distinct branch, the so-called CoCoNut, named after its composition of coiled-coil structures and tandem nucleases. These CoCoNut systems are predicted to target RNA as well as DNA and to utilise defence mechanisms with some similarity to type III CRISPR-Cas systems.

Strengths:

This work is enriched with a plethora of ideas and a myriad of compelling hypotheses that now await experimental verification. The study comes from the group that was amongst the first to describe, characterize, and classify CRISPR-Cas systems. By analogy, the findings described here can similarly promote ingenious experimental and conceptual research that could further drive technological advances. It could also instigate vigorous scientific debates that will ultimately benefit the community.

Weaknesses:

The multi-component systems described here function in the context of large oligomeric complexes. Some of the single chain AF2 predictions shown in this work are not compatible, for example, with homohexameric complex formation due to incompatible orientation of domains. The recent advances in protein structure prediction, in particular AlphaFold2 (AF2) multimer, now allow us to confidently probe potential protein-protein interactions and protein complex formation. This predictive power could be exploited here to produce a better glimpse of these multimeric protein systems. It can also provide a more sound explanation for some of the observed differences amongst different McrBC types.

Hexameric CnuB complexes with CnuC stimulatory monomers for Type I-A, I-B, I-C, II, and III-A CoCoNuT systems have been modeled with AF2 and included in Supplementary Data File S1, albeit without the domains fused to the GTPase N-terminus (with the exception of Type I-B, which lacks the long coiled-coil domain fused to the GTPase and was modeled with its entire sequence). Attempts to model the other full-length CnuB hexamers did not lead to convincing results.

Recommendations for the authors:

Reviewing Editor:

The detailed recommendations by the two reviewers will help the authors to further strengthen the manuscript, but two points seem particularly worth considering: 1. The methods are barely sketched in the manuscript, but it could be useful to detail them more closely. Particularly regarding the coiled-coil segments, which are currently just statists, useful mainly for the name of the family, more detail on their prediction, structural properties, and purpose would be very helpful. 2. Due to its encyclopedic nature, the wealth of material presented in the paper makes it hard to penetrate in one go. Any effort to make it more accessible would be very welcome. Reviewer 1 in particular has made a number of suggestions regarding the figures, which would make them provide more support for the findings described in the text.

A description of the techniques used to identify coiled-coil segments has been added to the Methods. Our predictions ranged from near certainty in the coiled-coils detected in CnuB homologs, to shorter helices at the limit of detection in other factors. We chose to report all probable coiled-coils, as the extensive coiled-coils fused to CnuB, which are often the only domain present other than the GTPase, imply involvement in mediating complex formation by interacting with coiled-coils in other factors, particularly the other CoCoNuT factors. The suggestions made by Reviewer 1 were thoughtful and we made an effort to incorporate them.

Reviewer #1 (Recommendations For The Authors):

I do not have any major issues with the manuscript. I have however some minor comments, as described below.

• The last sentence of the abstract at first reads as a fact and not a hypothesis resulting from the work described in the manuscript. After the second read, I noticed the nuances in the sentence. I would suggest a rephrasing to emphasize that the activity described is a theoretical hypothesis not backed-up by experiments.

This sentence has been rephrased to make explicit the hypothetical nature of the statement.

• In line 64, the authors rename DUF3578 as ADAM because indeed its function is not unknown. Did the authors consider reaching out to InterPro to add this designation to this DUF? A search in interpro with DUF3578 results in "MrcB-like, N-terminal domain" and if a name is suggested, it may be worthwhile to take it to the IntrePro team.

We will suggest this nomenclature to InterPro.

• I find Figure 1E hard to analyse and think it occupies too much space for the information it provides. The color scheme, the large amount of small slices, and the lack of numbers make its information content very small. I would suggest moving this to the supplementary and making it instead a bar plot. If removed from Figure 1, more space is made available for the other panels, particularly the structural superpositions, which in my opinion are much more important.

We have removed Figure 1E from the paper as it adds little information beyond the abundance and phyletic distribution of sequenced prokaryotes, in which McrBC systems are plentiful.

• In Figure 2, it is not clear due to the presence of many colorful "operon schemes" that the tree is for a single gene and not for the full operon segment. Highlighting the target gene in the operons or signalling it somehow would make the figure easy to understand even in the absence of the text and legend. The same applies to Supplementary Figure 1.

The legend has been modified to show more clearly that this is a tree of McrB-like GTPases.

• In line 146, the authors write "AlphaFold-predicted endonucelase fold" to say that a protein contains a region that AF2 predicts to fold like an endonuclease. This is a weird way of writing it and can be confusing to non-expert readers. I would suggest rephrasing for increased clarity.

This sentence has been rephrased for greater clarity.

• In line 167, there is a [47]. I believe this is probably due to a previous reference formatting.

Indeed, this was a reference formatting error and has been fixed.

• In most figures, the color palette and the use of very similar color palettes for taxonomy pie charts, genomic context composition schemes, and domain composition diagrams make it really hard to have a good understanding of the image at first. Legends are often close to each other, and it is not obvious at first which belong to what. I would suggest changing the layouts and maybe some color schemes to make it easier to extract the information that these figures want to convey.

It seemed that Figure 4 was the most glaring example of these issues, and it has been rearranged for easier comprehension.

• In the paragraph that starts at line 199, the authors mention an Ig-like domain that is often found at the N-terminus of Type I CoCoNuTs. Are they all related to each other? How conserved are these domains?

These domains are all predicted to adopt a similar beta-sandwich fold and are found at the N-terminus of most CoCoNuT CnuC homologs, suggesting they are part of the same family, but we did not undertake a more detailed sequenced-based analysis of these regions.

We also find comparable domains in the CnuC/McrC-like partners of the abundant McrB-like NxD motif GTPases that are not part of CoCoNuT systems, and given the similarity of some of their predicted structures to Rho GDP-dissociation inhibitor 1, we suspect that they have coevolved as regulators of the non-canonical NxD motif GTPase type. Our CnuBC multimer models showing consistent proximity between these domains in CnuC and CnuB GTPase domains suggest this could indeed be the case. We plan to explore these findings further in a forthcoming publication.

• In line 210, the authors write "suggesting a role in overcrowding-induced stress response". Why so? In >all other cases, the authors justify their hypothesis, which I really appreciated, but not here.

A supplementary note justifying this hypothesis has been added to Supplementary Data File S1.

• At the end of the paragraph that starts in line 264, the authors mention that they constructed AF2 multimeric models to predict if 2 proteins would interact. However, no quality scores were provided, particularly the PAE matrix. This would allow for a better judgement of this prediction, and I would suggest adding the PAE matrix as another panel in the figure where the 3D model of the complex is displayed.

The PAE matrix and ipTM+pTM scores for this and other multimer models have been added to Supplementary Data File S1. For this model in particular, the surface charge distribution of the model has been presented to support the role of the domains that have a higher PAE in RNA binding.

• In line 306, "(supplementary data)" refers to what part of the file?

This file has been renamed Supplementary Table S3 and referenced as such.

• In line 464, the authors suggest that ShdA could interact with CoCoNuTs. Why not model the complex as done for other cases? what would co-folding suggest?

As we were not able to convincingly model full-length CnuB hexamers with N-terminal coiled-coils, we did not attempt modeling of this hypothetical complex with another protein with a long coiled-coil, but it remains an interesting possibility.

• In line 528, why and how were some genes additionally analyzed with HHPred?

Justification for this analysis has been added to the Methods, but briefly, these genes were additionally analyzed if there were no BLAST hits or to confirm the hits that were obtained.

• In the first section of the methods, the first and second (particularly the second) paragraphs are extremely long. I would suggest breaking them to facilitate reading.

This change has been made.

• In line 545, what do the authors mean by "the alignment (...) were analyzed with HHPred"?

A more detailed description of this step has been added to the Methods.

• The authors provide the models they produced as well as extensive supplementary tables that make their data reusable, but they do not provide the code for the automated steps, as to excise target sequence sections out of multiple sequence alignments, for example.

The code used for these steps has been in use in our group at the NCBI for many years. It will be difficult to utilize outside of the NCBI software environment, but for full disclosure, we have included a zipped repository with the scripts and custom-code dependencies, although there are external dependencies as well such as FastTree and BLAST. In brief, it involves PSI-BLAST detection of regions with the most significant homology to one of a set of provided alignments (seals-2-master/bin/wrappers/cog_psicognitor). In this case, the reference alignments of McrB-like GTPases and DUF2357 were generated manually using HHpred to analyze alignments of clustered PSI-BLAST results. This step provided an output of coordinates defining domain footprints in each query sequence, which were then combined and/or extended using scripts based on manual analysis of many examples with HHpred (footprint_finders/get_GTPase_frags.py and footprint_finders/get_DUF2357_frags.py), then these coordinates were used to excise such regions from the query amino acid sequence with a final script (seals-2-master/bin/misc/fa2frag).

Reviewer #2 (Recommendations For The Authors):

(1) Page 4, line 77 - 'PUA superfamily domains' could be more appropriate to use instead of "EVE superfamily".

While this statement could perhaps be applied to PUA superfamily domains, our previous work we refer to, which strongly supports the assertion, was restricted to the EVE-like domains and we prefer to retain the original language.

(2) Page 5. lines 128-130 - AF2 multimer prediction model could provide a more sound explanation for these differences.

Our AF2 multimer predictions added in this revision indeed show that the NxD motif McrB-like CoCoNuT GTPases interact with their respective McrC-like partners such that an immunoglobulin-like beta-sandwich domain, fused to the N-termini of the McrC homologs and similar to Rho GDP-dissociation inhibitor 1, has the potential to physically interact with the GTPase variants. However, we did not probe this in greater detail, as it is beyond the scope of this already highly complex article, but we plan to study it in the future.

(3) Page 8, line 252 - The surface charge distribution of CnuH OB fold domain looks very different from SmpB (pdb3iyr). In fact, the regions that are in contact with RNA in SmpB are highly acidic in CoCoNut CnuH. Although it looks likely that this domain is involved in RNA binding, the mode of interaction should be very different.

We did not detect a strong similarity between the CnuH SmpB-like SPB domain and PDB 3IYR, but when we compare the surface charge distribution of PDB 1WJX and the SPB domain, while there is a significant area that is positively charged in 1WJX that is negatively charged in SPB, there is much that overlaps with the same charge in both domains.

The similarity between SmpB and the SPB domain is significant, but definitely not exact. An important question for future studies is: If the domains are indeed related due to an ancient fusion of SmpB to an ancestor of CnuH, would this degree of divergence be expected?

In other words, can we say anything about how the function of a stand-alone tmRNA-binding protein could evolve after being fused to a complex predicted RNA helicase with other predicted RNA binding domains already present? Experimental validation will ultimately be necessary to resolve these kinds of questions, but for now, it may be safe to say that the presence of this domain, especially in conjunction with the neighboring RelE-like RTL domain and UPF1-like helicase domain, signals a likely interaction with the A-site of the ribosome, and perhaps restriction of aberrant/viral mRNA.

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

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1

Evidence, reproducibility and clarity

The manuscript by Barba-Aliaga and colleagues describe a potential function of eIF5A for the control of TIM50 translation. The authors showed that in temperature-sensitive mutants of eIF5A several mitochondrial proteins are decreased including OXPHOS subunits, proteins of the TCA cycle and some components of protein translocases. Some precursor proteins appear to localize into the cytosol. As consequent of mitochondrial dysfunction, the expression of some stress components is induced. The idea is that eIF5A ribosome-stalling of the proline-rich Tim50 of the TIM23 complex and thereby controls mitochondrial protein set-up.

The findings are potentially interesting. However, some control experiments are required to substantiate the findings.

1. To support their conclusion the authors should show whether Tim50 levels are affected in the eIF5A-ts mutants used. Tim50 protein half-life is approximately 9.6 h (Christiano et al, 2014), which makes difficult to measure large differences in new protein synthesis upon eIF5A depletion. However, we used different approaches to show that reduction in eIF5A provokes a reduction in Tim50 protein levels and synthesis. 1) The steady-state levels of Tim50 protein (genomic HA-tagged version) are shown by western blotting analysis in Fig. S4B and confirm a significant drop of approximately 20% in the tif51A-1 mutant at restrictive temperature. 2) The use of a construct in which Tim50 is fused to a nanoluciferase reporter under the control of a tetO7 inducible promoter shows a significant 3-fold reduction in Tim50 protein synthesis in the tif51A-1 mutant compared to wild-type (Fig. 4C). In addition, the protein synthesis time is calculated and indicates that it takes the double time for the tif51A-1 strain to synthesize Tim50 protein than the wild-type (Fig. 4E). 3) The expression of a FLAG-TIM50-GFP version under a GAL inducible system also shows a significant reduction in Tim50 protein synthesis in the two eIF5A temperature-sensitive strains (Fig. S4C). 4) The proteomic analysis performed at 41ºC showed a 20% reduction in Tim50 protein levels in the two eIF5A temperature-sensitive strains, although not being statistically significant (Table S1). Furthermore, TIM50 mRNA levels were determined by RT-qPCR across all the experiments mentioned to confirm that the low levels of Tim50 protein were not due to decreased transcription or increased mRNA degradation. 5) An additional experiment of polysome profiling has been included in Fig. R1 (Figure for Reviewers) showing a higher TIM50 mRNA abundance at low polysomal fractions and a lower mRNA abundance at heavy polysomal fractions upon eIF5A depletion. This indicates that the TIM50 mRNA abundance is significantly shifted to earlier fractions and translation of Tim50 is reduced in the tif51A-1 mutant at restrictive temperature but not at permissive temperature. Altoghether, all these experiments confirm a significant reduction of Tim50 protein levels upon eIF5A depletion and conclusions are supported on these results.

How are the levels of TOM and TIM23 subunits?

Response: Our proteomic analysis shows that the protein levels of Tom70 and Tom20 receptor subunits of the TOM complex are significantly decreased in the two eIF5A temperature-sensitive strains (Table S1). These results are in agreement with the polysome profiling results, where it is seen a significant reduction of TOM70 and TOM20 mRNAs in the heavy polysomal fractions while a significant increase of these mRNAs is observed in the light fractions of eIF5A-depleted cells (Fig. 2C and Fig. S2D). Apart from Tim50, no other proteins of the Tim23 translocase complex were detected in the proteomic analysis.

Furthermore, how are the levels of the Tim50 variant that lack the proline residues? Is the stability or function of Tim50 affected by these mutations?

Although we did not specifically analysed the Tim50ΔPro protein levels, a quantification of the Tim50ΔPro fluorescent signal has been performed to address this matter and is shown in Fig. R2 and mentioned in the corresponding Results section. Results indicate that the Tim50 variant lacking the proline residues has similar protein levels to the wild-type version and therefore, it is tempting to say that its stability should also be similar. However, if Reviewers consider this to be essential for publishing, additional experiments using cycloheximide could be conducted in order to better assess the stability and half-life of this Tim50 version.

Additionally, functional levels of Tim50ΔPro protein is shown by the fact that wild-type cells carrying this Tim50 protein version as the only copy of Tim50 grew well in glycerol media, where Tim50 is essential for the mitochondrial function (Fig. 5A). However, we suspect that Tim50ΔPro is a bit less efficient protein since a double mutant tif51A-1 Tim50ΔPro shows even reduced growth than the single tif51A-1 mutant (Fig. 5A). This information also responds to the comments made by Reviewer #2.

How specific is the effect of eIF5A on Tim50? Is there any other mitochondrial substrate of eIF5A? It is not so clear to the reviewer why the authors focused on Tim50.

Response: eIF5A has been shown to be necessary for the translation of mRNA codons encoding for consecutive prolines and, consequently, lack of eIF5A causes ribosome stalling in these polyproline motifs (Gutierrez et al., 2013; Pelechano and Alepuz, 2017; Schuller et al., 2017). In our manuscript we showed: 1) using an artificial tetO7-TIM50-nanoLuc genomic construct we demonstrate that the synthesis of Tim50 protein (measured as appearance of luciferase activity upon induction of tetO promoter) is significantly reduced by 3-fold under eIF5A depletion only when Tim50 contains the stretch of 7 consecutive prolines (Fig. 4A-D); 2) genomic Tim50-HA and plasmid FLAG-TIM50-GFP protein levels are significantly reduced upon eIF5A depletion (Fig. S4B); 3) calculation of the time for translation elongation of Tim50 mRNA shows that this time is double in cells with eIF5A depletion than in cells containing normal eIF5A levels (Fig. 4E); and 4) analysis of published ribosome profiling data shows a precipitous drop-off in ribosome density exactly where the stretch of polyprolines is located in Tim50 (540-561bp) upon eIF5A depletion but not in the control strain (Fig. 4F). This result is indicative of ribosome stalling at Tim50 polyproline motif upon eIF5A depletion. Altogether, our results strongly support a direct and specific role of eIF5A in Tim50 protein synthesis. However, as we discuss in relation to Fig. 5 and in the Discussion section, Tim50 does not seem to be the only mitochondrial substrate of eIF5A, since recovery of Tim50 protein synthesis does not rescue the growth of eIF5A mutants under respiratory conditions. In this line, we have added further data pointing to ribosome stalling for other co-translationally inserted mitoproteins which are potential substrates of eIF5A (Table S6). Accordingly, this has also been included in the Discussion section. This information also responds to the comments made by Reviewer #4.

Our focus on Tim50 in this manuscript resides in that we found a global downregulation of mitochondrial protein synthesis (Fig.1 and 2) in parallel to the accumulation of mitochondrial precursor proteins in the cytoplasm and induction of the mitoCPR response (Fig.3). All these data were pointing to a mitochondrial protein import defect. Since Tim50 is an essential component of the Tim23 translocase complex, its protein levels are reduced in eIF5A mutants and Tim50 contains a polyproline motif, all these data were pointing towards a Tim50-dependent effect in mitochondrial protein import upon eIF5A depletion, which we addressed in the manuscript.

Figure 1A: Which tif51A strain was used?

Response: The proteomic analysis was performed with tif51A-1 and tif51A-3 temperature-sensitive strains (see Table S1) and Fig.1A shows the average of the values obtained for the two mutants (proteins detected as down-regulated in these two samples and from 3 different biological replicates). This is now clarified in the Figure 1A legend. A similar approach was also followed in Pelechano and Alepuz, 2017. Additionally, the ratios between the protein level in the temperature-sensitive mutant respect wild-type for each protein and for each eIF5A mutant are also shown in Table S1. This information also responds to the comments made by Reviewer #2.

Figure 1C: The authors should show the steady state levels of some OXPHOS/TCA components to confirm the findings of Figure 1A.

Response: Proteomic findings have been confirmed for several proteins. The steady state levels of Por1 and Hsp60 proteins were investigated by western blotting (Figs. 1C,D) and results show a significant down-regulation on the two eIF5A temperature-sensitive strains at 41ºC, which confirms the findings of Fig. 1A. Additionally, we have included the same experiment performed at 37ºC (Fig. S1E), which also confirms the same conclusion.

Furthermore, the steady-state levels of Tim50 protein were also investigated by western blotting (Fig. S4B), and results also showed a significant down-regulation in the tif51A-1 mutant at restrictive temperature (37ºC), compared to wild-type. This result also confirms the findings of Fig. 1A.

However, if Reviewers consider that additional confirmation for OXPHOS/TCA proteins to be essential for publishing, additional experiments could be conducted to assess the protein levels of other OXPHOS/TCA proteins.

The manuscript contains several quantifications. However, central information like number of repeats or whether a standard deviation or S.E.M. is depicted are missing.

Response: Clear information on the number of repeats, type of graphical representation and statistical analysis is now included for all figures in the corresponding figure legends and also detailed in the Materials and Methods section. This information also responds to the comments made by Reviewer #2.

Figure 3: The authors propose that precursor form aggregates outside mitochondria. To assess the data, a quantification should address in how many cells are protein aggregates.

Response: The quantification of cytoplasmic Yta12 aggregates is now included in Fig.3E, which shows significant differences between the tif51A-1 mutant and the wild-type strain. In addition, quantification of cytosolic Tim50 aggregates was already included in Fig. 4H, which also shows significant differences between the tif51A-1 mutant and the wild-type strain. These two figures include the individual values from three biological replicates (at least 150 cells were analyzed), mean, standard deviation and statistical analysis.

Do the observed aggregated proteins interact with Hsp104? recycled?

Response: Yes, the cytoplasmic mitochondrial precursor aggregated proteins co-localize with Hsp104 as shown in Fig. 3I for Cyc1 and in Fig. 4J for Tim50. The quantification of Cyc1 and Tim50 co-localization with Hsp104 is shown in Fig S5D.

---

Significance

See above

---

Reviewer #2

Evidence, reproducibility and clarity

The authors report here novel findings concerning the role of eIF5A in mediating protein import to mitochondria in the model eukaryote Saccharomyces cerevisiae. It was previously known from structural and other studies that the translation factor eIF5A binds to the E-site of stalled ribosomes to help promote peptide bond formation. It was inferred by ribosome footprinting and reporter studies assessing the impact of eIF5A depletion that eIF5A is particularly needed to translate several specific amino acid motifs including polyproline stretches. However additional target sequences are known.

Here a proteomics approach reveals clear evidence that mitochondrially targeted proteins are impacted by temperature sensitive mutations in eIF5A that deplete the factor, including those without polyprolines. The authors then use a range of molecular and cell biology to focus on the role of mitochondrial signal sequences/mitochondrial protein import and the mitochondrial stress response, before highlighting a role for poly-prolines in Tim50, a major mitochondrial protein import factor. Consistent with the ribosome footprinting done previously it is shown that a stretch of 7 prolines limit its translation when eIF5A is depleted and studies shown here are consistent with the idea that this has wider consequences for mitochondrial protein import and hence translation/stability of other proteins. However improved Tim50 translation alone, by eliminating the poly-proline motif, is not sufficient to overcome all consequences of eIF5A depletion for mitochondrial protein import and for viability, suggesting a wider role.

In general the text flows nicely, this could be a study that explains why a large number of mitochondrially targeted proteins are impacted by depletion of eIF5A in yeast. As the poly Pro sequence in Tim50 is not conserved in higher eukaryotes it is unclear how this observation will scale to other systems, but it provides an example of how studies in a relatively simple system can trace wide-spread impact of the loss of one component of a central pathway-here protein synthesis to altered translation of a key component of another process-mitochondrial protein import. Given that eIF5A and its hypusine modifying enzymes are mutated in rare human disorders, it is likely there will be interest in this study.

However, while the conclusions may be justified, there are significant deficiencies in how the experiments have been analysed and presented in this version of the manuscript that impact every figure shown, coupled with deficiencies in the methods section that all need to be addressed. Thus, we have here the basis of what should be a very interesting paper here, but there is a lot of work to do to remedy perceived weaknesses. It may be that the overall conclusions are entirely sound and appropriate, but I suspect that performing the statistics in less biased ways may change some of the significant differences claimed. Some explanations concerning how data analyses were conducted and the reasons for specific analysis decisions being made would also improve the narrative. These points are expanded on below.

All the edits suggested here are aimed at improving the rigor of reporting in this study. Depending on how they are answered some may become major issues, or they could all be minor.

1 Figure 1 shows proteomic data for response to heat shock at 41{degree sign}C. In the text it is made clear that two different temperature sensitive missense alleles the 51A-1 and 51A-3 were analysed, but the single volcano plot in Figure 1A does not say whether it is reporting one of these experiments compared to WT (which one) or some other analysis (ie have data from the 2 mutants been amalgamated somehow?). I would assume only one, but which one, and why only one plot? How different is the other experiment? Why does the Figure title say the experiment is an eIF5A deletion when it is not this?

Response: The data shown in Figure 1A corresponds to the average values obtained in the proteomic analysis for the two temperature-sensitive mutants tif51A-1 and tif51A-3 (with data for each mutant obtained from 3 different biological replicates). Highly reproducible proteomic results and similar between the two mutants were obtained (see in Fig. S1A the MDS-plot showing all replicates for each strain and condition studied in the proteomic analysis). In addition, the proteomic data showing the protein 41°C/25°C ratio for each eIF5A temperature-sensitive mutant with respect to wild-type is shown in the Table S1. This is now clarified in the Figure 1A legend. A similar approach using the mean values of the two mutants was followed in the analysis of ribosome footprintings made in Pelechano and Alepuz, 2017. Additionally, the ratios between the protein level in the temperature-sensitive mutant respect wild-type for each protein and for each eIF5A mutant are also shown in Table S1. This information also responds to the comments made by Reviewer #1.

Reviewer #2 is right with his/her comment and there was a mistake in the Fig.1 title. Now it is corrected and written “depletion” instead of the wrong “deletion”.

2 Why were the experiments shown in Figure 1 done at 41{degree sign}C when all other experiments are done at 37{degree sign}C? This experimental difference is ignored in the text and no comparison of the impact of 37 vs 41 is made anywhere in the manuscript. For example it would be straightforward to perform a comparison of eIF5A depletion (by western blot), polyribosome profiles, strain growth/inhibition at both temperatures.

Response: Our aim carrying out a proteomic experiment after 4 hours of incubation of the temperature-sensitive strains at 41°C was to get a more profound depletion of the eIF5A protein, which is very abundant and stable at normal conditions, in order to get clear proteomic results. The proteomic results were pointing to a reduction in the levels of many mitochondrial proteins, corroborating previous results obtained in murine embryonic fibroblasts upon depletion of active eIF5A conditions (https://doi.org/10.1016/j.cmet.2019.05.003). From this starting point we tried to find out the molecular mechanism involved and all the rest of experiments are done with temperature sensitive eIF5A mutants under restrictive temperature of 37°C that is the most common conditions used in yeast by us and others, and in which wild-type yeast cells still grow vigorously.

In our previous manuscript version, the depletion of eIF5A after growing the cells at 41ºC for 4 h was shown in Fig. 1C. These data has been expanded and we have now included in Fig. S1E a western blotting analysis that shows the depletion of eIF5A after incubating the cells at 37ºC and 41 ºC for 4 h (Fig. S1E). The steady state level of the mitochondrial Por1 protein was investigated by western blotting (Figs. 1C,D) and results show a significant down-regulation in the two eIF5A temperature-sensitive strains at 41ºC. We have now included the same experiment performed at 37ºC (Fig. S1E), which also confirms the same conclusion. In addition, following Reviewer #2 suggestions, growth of the wild-type and tif51A-1 strains was tested by serial drop assays conducted at 25ºC, 37ºC and 41ºC and results confirm that both 37ºC and 41ºC temperatures impair the growth of the tif51A-1 strain but not the wild-type (Fig.S1B). The new information included in Figure S1 is now explained in the Results section. This information also responds to the comments made by Reviewer #4.

3 Western blot quantification. In Figure 1D and E the authors present western blot quantification. However they have chosen to normalise every panel to the signal in lane 1. This means that there is no variation at all in that sample as every replicate is =1. This completely skews the statistical assumptions made (because there will be variation in that sample) and effectively invalidates all the statistics shown. An appropriate approach to use is to normalise the signal in each lane to the mean signal across all lanes in a single blot. That way if all are identical they remain at 1, but importantly variation across all samples is captured. This should be done to the loading controls as well before working out ratios or performing any statistical analyses.

Response: Following Reviewer #2 suggestions we have changed the normalization methodology for the Western blots and we have now normalized the signal in each lane to the mean signal across all lanes in each single blot, and do so also for the loading controls. We have conducted this analysis in every western blotting experiment shown in the manuscript (Figs. 1D, S4B and S4C) and statistical analyses have been performed again to capture variation across all samples. In addition, this is also included in the Materials and Methods section (“Western blotting” subsection). Results obtain are similar to previous ones but we agree that this new approach improves the data presentation.

For this type of experiment it is more appropriate to use Anova than a T-test. This advice applies to every western data analysis figure in the whole manuscript and so all associated statistics need to be done again from the original quantification values. If T-test is justified then a correction for multiple hypothesis testing should be applied.

Response: After reviewing a large number of publications analysing similar data, and also following the recommendations of our statistical department, we have retained the statistics used in our previous version (with the new data normalisation as explained above, following the recommendations of Reviewer #2). This is because for each western blot figure shown, we have performed experiments with two different biological samples, wild-type cells and eIF5A mutant cells, and compared results for a single variable (Por1 protein level; eIF5A protein level or Hsp60 protein level) using three or more biological replicates. In this context, we compare the mean of the protein levels obtained from the biological replicate for two groups: wild-type and eIF5A mutant. Therefore, we believe that the statistical T-test is more appropriate. However, we could repeat the statistic if it is finally considered more appropriate.

In all bar chart figures in addition to showing the mean and SD, each replicate value should be shown (eg as done in Fig 2C). Graphpad allows individual points to be plotted easily.

Response: All Figures along the manuscript now include individual values from each replicate, in addition to showing the mean, SD and statistical analysis. All figure legends have been corrected accordingly.

5 Figure 2. Polysome profiles. The impact of translation elongation stalls on global polysome profiles is complex, but a global run off is highly unlikely. Stalls later in the coding region would be anticipated to cause an increase in ribosome density as more ribosomes accumulate (like cars queueing held at a red light). However where a stall is early in a longer ORF, for example at a signal sequence, then there is less opportunity for ribosomes to join and so for those mRNAs moving to lighter points in the gradient may be observed. This may also cause knock on effects on AUG clearance and initiation which the authors appear to see as there may be an increased 60S peak in the traces shown. Are there differences in overall -low vs high polysomes, the traces shown suggest there may be? Discussion of these points is merited in the results section given the subsequent qPCR experiment.

Response: The comments made by the Reviewer #2 are very interesting and we have made changes accordingly. First, we now show in Fig. 2A,B and Fig.S2B,C the quantification of polysomal and monosomal fractions in wild-type and tif51A-1 mutants at permissive and restrictive temperatures. It can be appreciated that there is no impact on global polysomal and monosomal fractions under eIF5A depletion. This result does not support a global stall at 3’ region of the ORF, because then an increase in polysomal fractions should be detected; nor a global stall at the 5’ region of the ORF, because then a decrease in polysomal fractions should be detected. However, with respect to individual mRNAs, our data show a significant reduction in the heavier polysomal fractions and a significant increase in lighter polysomal fractions for mRNAs encoding mitochondrial proteins, while no significant changes were observed for mRNAs encoding cytoplasmic proteins (Fig. 2C and Fig. S2D-I). These results could be interpreted as a result of ribosome stalls in the 5’ ORF regions, for example at the signal sequence, according to Reviewer #2 comments.

We have now introduced this comment in the Results and Discussion sections.

Figure 2 qPCR. Using qPCR to analyse RNA levels across polysome gradients is tricky for multiple reasons including that the total RNA level varies across fractions that can impact recovery efficiencies following precipitation of gradient fractions. Often investigators use a spike in control to act as a normalising factor. Here it is completely unclear what analysis was done because details are not stated anywhere. How were primers optimized, was amplification efficiency determined? Or are they assumed to be 100%, which they will not be? A detailed description or reference to a study where that is written is needed.

Response: The RNA extraction and analyses by RT-qPCR of the mRNA levels in the polysomal gradients was done as in previous studies of our lab (Romero et al. Sci Rep. 2020;10(1):233. doi: 10.1038/s41598-019-57132-0; Ramos-Alonso et al. PLoS Genet. 2018;14(6):e1007476. doi: 10.1371/journal.pgen.1007476; van Wijlick et al. PLoS Genet. 2016;12(10):e1006395. doi: 10.1371/journal.pgen.1006395; Garre et al., 2012 Mol Biol Cell. ;23(1):137-50. doi: 10.1091/mbc.E11-05-0419.). Three independent replicates were analyzed and results were reproducible and statistically significant, as shown in Fig. S2. Total RNA was extracted from each fraction using the SpeedTools Total RNA Extraction kit (Biotools B&M Labs). In the first replicate a spike in RNA control (Phenylalanine) was added and tested that no significant differences in the results were obtained when using or not the spike in control (see below Figure R3 for referees). mRNA relative values are always obtained from qPCR using a calibrating efficiency standard curve for each pair of oligos, after the initial set up of the qPCR for this specific pair of oligos. Therefore, slight differences in amplification efficiencies for each oligo pair are taken into account. More details about qPCR are now included in the Materials and Methods section (“Polyribosome profile analysis” subsection) and one additional reference is also included for the processing of polysomal gradient fractions.

It would be helpful to state how long CDS are for these mRNAs and where 2-3/2-8 cut off made is what for determining what is 'short' vs 'long' and the scientific basis for selecting 2-3 vs 2-8, why 8? Were M fractions also used in qPCR, they appear to be ignored in the analysis as currently presented?

Response: The CDS lengths of the mRNAs analyzed by polysome profiling and other important features are now included in new Table S5. We decided to classify as short length mRNAs those with a length below 600 bp, while mRNAs with lengths above 600 bp were classified as long length mRNAs. This classification was made on the basis of specific mRNA profiles obtained by qPCR analysis. mRNAs with short lengths behaved similarly and we selected 2n-3n fractions since the main polysomal peak under normal conditions appeared among 4n-5n fractions. In this line, long length mRNAs also behaved similarly between them, and we selected 2n to 8n fractions since the main polysomal peak under normal conditions appeared right after the 8n fraction. This information is now included in the Results and Materials and Methods sections.

Regarding the use of the Monosomal fractions, yes, they were used as it can be seen in Fig. S2 which includes the distribution in Monosomal (M), lighter (2n-3n/2n-8n) or heavier (n>3/n>8, P) polysomal fractions. In the polysomal profiles we can be see that depletion of eIF5A causes a reduction in the amount of mitochondrial mRNAs in the heavier fractions and a corresponding increase in the amount of mRNAs in the lighter polysomal fractions, while no significant changes are found in the monosomal fractions. Therefore, the statistically significant change in the heavier/lighter polysomal fraction ratio is indicative of the translation down-regulation and these ratios are shown in Fig. 2C. As the Reviewer #2 commented in point 5, the change in mRNA distribution to lighter polysomal fractions may be indicative or ribosome stalling at the 5’ ORF region, compatible with a stall at the mitochondrial target signal (MTS), and this discussion is now included in the Results and Discussion section.

Which transcripts studied here encode proteins with signal sequences? As Signal sequence pauses early in translation should impact ribosome loading this is potentially important here as discussed above.

Response: Yes, we agree with Reviewer #2 that this information may be relevant according to the hypothesis of ribosome stall at the MTS. Therefore, a score value of probability of harbouring an MTS presequence (Fukasawa et al., 2015) is now included in Table S5 for each of the mRNAS analyzed by polysome profiling. The discussion of this point has also been included in the Results and Discussion sections.

While it has been shown that SRP recognition is able to slow and even arrest translation of ER signal recognition peptides, there is currently no known direct SRP like correlate for mitochondrial signal sequences. We are therefore unaware of literature showing that mitochondrial signal sequences pause translation in a manner similar to ER signal sequences. We have previously found that downstream translational slowing is important for mitochondrial mRNA targeting (Tsuboi et al 2020, Arceo et al 2022), but we believe that to be distinct to what the Reviewer #2 is addressing.

Figures 3-5. Microscopy. The false green color images in Figure 3B do not show up well. They may be better shown in grayscale, with only the multiple overlays in color.

Response: False color for fluorescent microscopy images are widely used because they help to visualize the results to the readers and also facilitate the interpretation of multiple overlays. The use of false color is also suggested by Reviewer #4.

Figure 3C should show the data spread for all 150 cells and normalise differently as discussed above for westerns. I do not believe that all 150 WT cells have exactly the same GFP intensity, which is what the present plot claims.

Response: As answered to point 3 made by this Reviewer, now all figures, including Fig. 3C, are made with Graphpad and scatter plot with all individual points plotted, additionally to showing the mean, SD and statistical analysis. Results correspond to three independent experiments and show a statistically significant difference in Pdr5-GFP intensity signal between wild-type and tif51A-1 mutant. Figure legend has been corrected accordingly.

For panels 3D-F image quantification should be shown so that the variation across a population is clear. Eg in violin plots, or showing every point. It should be clear what proportion of cells have GFP aggregates and what the variation in number of granules is.

Response: The quantification of cytoplasmic Yta12 aggregates is now included in Fig.3E, which shows significant differences between the tif51A-1 mutant and the wild-type strain. Results show the individual values from three independent experiments with a minimum of 150 cells counted. We used a bar graph in which the values (% of cells with 0, 1, 2 or 3 aggregates) for each independent experiment are shown together with the mean, SD and statistical analysis. Figure legend has been corrected accordingly. This information also responds to the comments made by Reviewer #1.

Figure 4H has no error bars.

Response: New Fig.4H now shows the individual values of each of the three independent replicates, mean and error bars (SD). Figure legend has been corrected accordingly.

Figure 5C normalises 2 WTs to 1 as in Figure 3C. Both would be better as violin plots.

Response: Results in Fig. 5C are now shown using Graphpad and scatter plot in which all individual values are plotted (not normalized wild-type to 1), and also mean, SD and statistical significance. Results correspond to three independent replicates with the fluorescence intensity measured in more than 150 cells.

Figure 5D/E shows 37{degree sign}C data only. Do tif51A-1 cells have aggregates at 25{degree sign}C?There are no error bars in Figure 5E or any indication of how many cells/replicates were quantified.

Response: Figures 5D and 5E only show data at 37ºC since there are no Tim50-GFP aggregates, nor aggregates of other mitochondrial proteins, in tif51A-1 mutants at 25ºC, as shown in Fig. S3C-F and Fig. S5C.

New Fig. 5E shows individual values from each of the three independent experiments, mean, SD and statistical significance. Results correspond to the measurement of Tim50 protein aggregates in more than 150 cells. Figure legend has been corrected accordingly.

There are no sizing bars on any of the micrographs.

Response: Now, all sets of microscopy figures contain a size bar and this is indicated in the corresponding Figure legend.

The methods states that all quantification was done using ImageJ, but there is no detail given about how this was done. There are lots of ways to use ImageJ.

Response: A detailed description of the quantifications made using ImageJ is now included in the Materials and Methods section (“Fluorescent microscopy and analysis” subsection).

Figure 4. Luciferase assay. It is clear that there are differences in Tim50 vs Tim50∆7pro signal over time from the primary plots. It is not clear why the quantification plots on the right are from 2 selected time points. It is more typical to calculate the rate of increase in RLU per min in the linear portion of the plot, for these examples it would be approximately 30-40 mins.

Response: As luciferase mRNA level is also increasing with time, the total amount of luciferase protein will increase exponentially. At some point mRNA levels will reach a steady state and for a brief period there could be a linear portion of RLU increase, but that will be different for each condition and reporter as ribosome quality control can have a direct impact on mRNA half-life. We have instead chosen two time points to show that statistical differences in Tim50 protein expression upon eIF5A depletion are not dependent on the time point chosen. We have also included the full data plots for readers to view the raw data.

Figure 4F. The text on p6 states Fig 4F is evidence of RQC induction. This is an overstatement. There are no data presented relating to RQC.

Response: Ribosome-associated quality control (RQC) is a mechanism by which elongation-stalled ribosomes are sensed in the cell, and then removed from the stall site by ribosomal subunit dissociation. This is the definition of RQC. With high levels of RQC this will cause a drop in ribosome density downstream of the stall site because of ribosome removal. While we would agree that most studies do not show actual buildup of ribosomes at ribosome stalls, and removal after the stall, we do. Our ribosome profiling analysis shows in vivo distribution of ribosome density across the TIM50 mRNA in wild-type and upon eIF5A depletion. We show that in the eIF5A depletion the ribosome density is similar to wild-type for the first ~200 bp, then there is a buildup of ribosomes for ~300 bps up to the stretch of polyproline residues, indicative of slowed ribosome movement. This slowed ribosome movement is further supported by our translation duration measurements in Fig. 4E. Then the transcript is almost completely devoid of ribosomes after the stretch of proline residues, indicating the ribosomes are removed at the proline stretch. This combination of ribosome stalling (Fig. 4E,4F) and subsequent ribosome removal (Fig 4F) is the textbook definition of RQC, so we indicate this as evidence for RQC.

Figure 5G. It is not clear to this reviewer why the CYC1 reporter is impacted by Tim50∆pro at 25{degree sign}C. Can the authors comment?

Response: This is also not clear to us, however, no differences are seen with and without eIF5A depletion, supporting the interpretation that Cyc1 translation is not affected by eIF5A depletion when Tim50 protein levels are restored in the Tim50∆pro strain. However, in order to clarify this point, we propose, if it is considered necessary, to remake the Tim50∆pro CYC1 reporter strain.

Does ∆pro impact Tim50 function or is there possibly some other off target impact of integrating the reporter in this strain?

Response: As answered to Reviewer #1 in her/his point 1, the functionality of Tim50ΔPro is shown by the fact that wild-type cells carrying this Tim50 protein version as the only copy of Tim50 grew well in glycerol media, where Tim50 is essential for the mitochondrial function (Fig. 5A). However, we suspect that Tim50ΔPro is a bit less efficient protein since a double mutant tif51A-1 Tim50ΔPro shows even reduced growth than the single tif51A-1 mutant (Fig. 5A). We do not expect off target impact in this Tim50ΔPro strains, although we cannot exclude this 100%, as in any other yeast strain obtained by transformation.

Significance

Strengths and Limitations:

Strengths are that the study uses a wide range of molecular approaches to address the questions and that the results present a clear story.

Limitations are that the poly-proline residues identified in yeast Tim50 are not conserved through to humans, so the direct relevance to higher organisms is unclear. However there are many more poly-proline proteins in human genes than in yeast and there are rare genetic conditions affecting eIF5A and its hypusination

Advance. provides a clear link between dysregulation of eIF5A, Tim50 expression and wider impact on mitochondria.

Audience. Scientists interested in protein synthesis, mitochondrial biology and clinicians investigating rare human disorders of eIF5A and hypusination.

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

eIF5A is required to mediate efficient translation elongation of some amino-acid sequences like polyproline motifs, and eIF5A depletion was reported to impair mitochondrial respiration functions, decreasing mitochondrial protein levels. In this study, Barba-Aliaga et al. showed that eIF5A is important for the translation of the Pro-repeat containing protein, Tim50, an essential subunit of the TIM23 complex, the presequence translocase in the mitochondrial inner membrane. eIF5A ts mutants caused ribosome stalling of Tim50 mRNA on the mitochondrial surface at non-permissive temperature, and the removal of the Pro-repeat from Tim50 (Tim50-delta7Pro mutant) made its translation independent of eIF5A. However, the replacement of endogenous Tim50 with Tim50-delta7Pro did not recover the cell growth defects of eIF5A ts mutant on respiration medium at semi-permissive temperature, suggesting that Tim50 is not the only reason for the global mitochondrial defects caused by defective eIF5A.

(1) I am wondering why the authors mainly used the eIF5A ts mutant strains instead of the eIF5A degron strain since, for example, the decrease in the level of Tim50 was only marginal (Fig. EV4A).

Response: eIF5A is a very abundant protein and with high stability (SGD data: 273594 molecules/cell in YPD and 9.1 h protein half-life). We have used temperature-sensitive strains, tif51A-1, instead of eIF5A-degron because eIF5A is depleted much quicker in the first than the second system. As it can be seen in Schuller et al., Mol Cell. 2017;66(2):194-205.e5. doi: 10.1016/j.molcel.2017.03.003, with the eIF5A-degron system the addition of auxin was made in parallel to a transcriptional shut off using GAL promoter to express eIF5A-degron, changing the media from galactose to glucose and incubating the cells for 10 hours. With our approach using temperature-sensitive proteins, almost full depletion (without affecting viability, see Li et al., Genetics 2014; 197(4):1191-200 doi: 10.1534/genetics.114.166926) can be done after 4-6 h incubation at 37ºC or 4 h incubation at 41ºC (Fig. 1C and Fig. S1E, almost no signal is detected by western blotting). Therefore, we chose to use eIF5A depletion with temperature-sensitive yeast strains to achieve stronger protein depletion with shorter times and avoid secondary effects. In addition, the two eIF5A temperature-sensitive strains used in this study have been widely used by us and others (Pelechano and Alepuz, 2017; Zanelli and Valentini, 2005; Zanelli et al., 2006; Dias et al., 2008; Muñoz-Soriano et al., 2017; Rossi et al., 2014; Li et al., 2014; Xiao et al., 2024).

(2) To show that the compromised translation of Tim50 in the absence of functional eIF5A causes defects in the mitochondrial protein import by clogging the import channels, the authors should directly observe the accumulation of the precursor forms of several matrix-targeting proteins by immunoblotting. In this sense, the results in Fig. 1C for Hsp60 do not fit the interpretation of import channel clogging.

Response: We did not see precursor mitochondrial proteins by Western blot upon eIF5A depletion possibly because: 1) the mature protein form is more abundant and stable; 2) the precursor mito-protein appears in cytoplasmic aggregates and this may not be easily extracted during preparation of proteins for Western blot analysis. In the work by Weidberg and Amon, 2018, who described the mitoCPR response; Krämer et al., 2023, who described mitostores; and others (Wrobel et al., 2015; Boos et al., 2019) the authors use extreme over-expression of mitoproteins or mutations in essential proteins for mitochondrial biogenesis to induce clogging of translocases and accumulation of precursors in the cytosol. However, we are using and detecting proteins at their physiological levels, expressed under their native promoters, what may explain why we do not detect precursor mito-proteins. We are using what we believe to be a much more physiologically relevant system, where we use endogenous expression of mitochondrially imported proteins. Yet we see similar transcriptional induction of mitoCPR targets (CIS1, PDR5, PDR15) and mislocalization of mitochondrial proteins to Hsp104 marked aggregates (MitoStores).

(3) The authors speculated in the Discussion section that import defects caused by compromised translation of Tim50 could cause down-regulation of translation through prolonged mitochondrial stress. However, this lacks experimental evidence.

Response: We do see that depletion of eIF5A causes import defects through Tim50 and correlates with the down-regulation of translation of mRNAs encoding mitoproteins as shown in Fig. 2C and Fig. S2. In these figures it can be seen that mito-mRNAs move from heavier to lighter polysomal fractions upon eIF5A depletion, indicating that less ribosomes are bound to these mRNAs. Importantly, synthesis of Cyc1 and Cox5A mitochondrial proteins is recovered when TIM50 gene is replaced by an eIF5A-translation independent TIM50ΔPro gene, arguing in favor of a translation defect caused by eIF5A depletion through the collapse of import systems produced by the ribosome stalling in TIM50 mRNA.

As discussed by Reviewer #2 and in our answers to his/her points 5 and 6, the reduction in the number of ribosomes bound to mito-mRNAs upon eIF5A depletion may be a consequence of the stall of ribosomes after the mRNA 5’ coding region encoding the MTS. This discussion has now been introduced in the Discussion section. This information also responds to the comments made by Reviewer #2.

(4) The authors stated that human Tim50 does not have Pro-repeat motif, but how about other organisms (like other fungi species)? Is the present observation specific only to S. cerevisiae?

Response: We have now included a sequence alignment of the Tim50 protein sequences of different yeast species (Saccharomyces cerevisiae, Candida albicans, Candida glabrata, Candida lipolytica, Schizzosaccharomyces pombe, Schizzosaccharomyces jamonicus), mouse and human (Fig. S4A). The resulting alignment shows that S. cerevisiae is the only organism presenting the seven consecutive proline residues. Still, C. albicans and C. glabrata conserve five consecutive prolines while C. lipolytica conserves five non-consecutive prolines. Furthermore, S. pombe and S. jamonicus, and mouse and human, conserve three and four non-consecutive prolines respectively. This means that the observations presented in this manuscript could be extended to other fungi species as well since most of the proline residues are conserved and are predicted to behave as eIF5A-dependent motifs for translation. Moreover, the described eIF5A-dependent tripeptide motif PDP is found in humans, mice and S. pombe at the Tim50 region where we found the PPP motif inducing ribosome stalling in S. cerevisiae (Fig S4A). This may confer eIF5A-dependent ribosome stall since as we showed in our previous ribosome footprinting (Pelechano et al., 2017), this PDP motif causes a similar high ribosome stall as the PPP motif. This discussion has now been introduced in the Results and Discussion sections.

(5) Two references in the text are marked with "?", which should be corrected.

Response: We thank you the Reviewer #3 for noticing this, references have been corrected in the text.

__Reviewer #3 (Significance (Required)): __

The essence of this work, the role of eIF5A in the efficient translation of Pro-repeat containing Tim50 (Figs. 4 and 5), is important and worth publication. However, the results of the effects of defective eIF5A on the levels and localization of mitochondrial proteins (Figs.1-3) can be even deleted to make clear the point of this work.

Reviewer #4 (Evidence, reproducibility and clarity (Required)):

The manuscript submitted by Barba-Aliaga et al. aims to understand on the molecular level how eIF5A influences mitochondrial function. elF5A promotes translation elongation at stretches prone to translational stalling like e.g. polyproline sequence. The finding that eIF5a influences mitochondrial function has been previously reported by the same group and by others. In this context, it was suggested that eIF5a promotes translation of N-terminal mitochondrial targeting signals. Here, the authors propose an alternative mechanism and suggest that "eIF5a directly controls mitochondrial protein import through alleviation of ribosome stalling along TIM50 mRNA." Using luciferase reporter assay, the authors indeed convincingly show that the speed of Tim50 translation is dependent on the presence of functional TIF51A, the major eIF5a in yeast, and that this dependence comes from the presence of the polyproline stretch in Tim50. The rest of the manuscript is unfortunately less clear and it is very hard, if not impossible, to sort out direct from secondary effects and compensations. The authors use proteomics, biochemical methods, RNAseq and fluorescence microscopy to analyze the temperature sensitive tif51A mutant but the conditions used in the manuscript are non-consistent between various experiments presented, in respect to the medium, temperature, preculture condition and the length of treatment used.

Response: We do not agree with this Reviewer #4 appreciation. We used different molecular approaches to investigate different questions. Indeed, this is one of the Strengths that is highlighted by Reviewer 2 as it reads above: “Strengths are that the study uses a wide range of molecular approaches to address the questions and that the results present a clear story.” All the experiments presented in the manuscript, apart from proteomics analysis (Fig. 1), have been performed in the same conditions respect to the medium (SGal), temperature (25ºC/37ºC), preculture condition (SGal, 25ºC) and length of treatment used (4 h of depletion at 37ºC). This is already clearly specified in every Figure legend along the whole manuscript and also in the Materials and Methods section. In addition, individual values from each replicate, mean, standard deviation and statistical tests are shown for every Figure in the manuscript. Therefore, we do believe that conditions are consistent between experiments and conclusions are made based on different experiments and different scientific approaches.

We agree with Reviewer #4 in that depletion of eIF5A protein in the temperature sensitive tif51A-1 mutant was done in the proteomic at 41°C for 4 h, whereas in the rest of experiments depletion is made at 37°C for 4 h. As answered to Reviewer #2 (see answer to point 2), stronger depletion conditions were used to get clear proteomic results, and in order to compare both temperatures we have added now some controls showing eIF5A depletion and growth of tif51A-1 mutant at 41°C and 37°C; importantly, we also show the reduction in mito-protein levels upon eIF5A depletion at 37°C (Fig. S1B and E).

In some cases, the genetic background of the yeast strains and plasmids used are also unclear (e.g. pYES2-pGAL-FLAG-TIM50-GFP-URA3 - based on the provided description, TIM50 was inserted between FLAG and GFP tags; if so, mitochondrial targeting signal of Tim50 would be masked making its import into mitochondria impossible).

Response: We do not agree with this appreciation. The genetic background of the yeast strains is always the same along the whole manuscript (BY4741 background) and is clearly specified in Table S2. In this line, all the information regarding the plasmids used can be found at Table S3 and plasmids construction is extensively detailed in the Materials and Methods section (“Yeast strains, plasmids, and growth conditions” subsection).

Regarding the pYES2-pGAL-FLAG-TIM50-GFP-URA3 plasmid and as already mentioned in the text, we only used this plasmid to analyze by western blotting the protein synthesis of Tim50 independently of its subcellular localization. Our results (Fig. S4C) confirm that the synthesis defect of this Tim50 version upon eIF5A depletion is only due to the presence of the polyproline region. Importantly, we did not make any conclusion regarding import defects or protein localization based on these results.

I have no doubt that upon exposure of tif51A cells to 41{degree sign}C for 4h cells initiate a number of cellular responses including mitoCPR and formation of MitoStores, however, I don´t think that the authors convincingly show that these are initiated by reduced levels of Tim50 - on the contrary, the authors show that levels of Tim50 are actually not significantly changed. This can hardly be reconciled with the model proposed. In addition, should the effect of Tif51A on mitochondria primarily be due to its effect on Tim50, then Tim50deltaPro should rescue the phenotype of tif51a mutant but it didn´t; if anything, it made it worse (see Fig 5A - the double mutant grows worse than the single ones). Furthermore, expression of Cyc1 luciferase reporter is reduced in Tim50deltaPro strain even at permissive temperature, Figure 5G. Since cytochrome c is not a substrate of the presequence pathway this again points to the secondary effects that are being observed.

Response: We believe that our main results, summarized next and all performed at 37°C, do show that translation defects in TIM50 mRNA are the cause of the mitoCPR induction and formation of MitoStores. First, Tim50 protein levels are significantly reduced upon eIF5A depletion, as shown in Fig. S4A and S4B. Although being statistically significant, we agree that the reduction in Tim50 protein level is quantitatively low. This can be explained by the high stability of Tim50 protein, with a half-life of approximately 9.6 h (Christiano et al, 2014), which makes it more difficult to measure large differences in new protein synthesis. This is why we additionally used an accurate and quantitative test for showing the eIF5A-dependency for TIM50 mRNA translation: the fusion of the TIM50 DNA sequence to a TetO7-inducible nLuc reporter, which allows to monitor the appearance of new Tim50 protein and to estimate the translation elongation rate (Fig.4C-E). The ribosome stalling at TIM50 mRNA provoked by eIF5A depletion, where this mRNA is located at the mitochondrial surface to promote the import of nascent Tim50 protein during translation (Fig. S5B), may cause by itself the clogging of the protein import system even though yields only a slight reduction in total Tim50 cellular protein. Second, as Reviewer #4 pointed with our model, Tim50deltaPro should rescue the phenotype of tif51A-1 mutant and it does it: no mitoCPR induction and no mito-protein cytoplasmic aggregation are observed (Fig. 5D-F). Moreover, no differences in Cyc1- and Cox5a-nanoLuc synthesis are observed in the tif51A-1 Tim50ΔPro strain between depletion and not depletion conditions (Fig. 5E). These results strongly suggest that the mitochondrial protein import defects (and consequently the mitoCPR induction and mito-protein cytoplasmic aggregation) caused by eIF5A depletion are a consequence of ribosome stalling during TIM50 mRNA translation. However, Reviewer #4 is right in that mitochondrial respiration and growth in glycerol are not restored in the tif51A-1 Tim50ΔPro strain, even though Tim50 protein levels have been restored under eIF5A depletion conditions. As we discuss in the manuscript, we expect that there are additional mitochondrial proteins as targets of eIF5A, such as Yta12 and/or others. We have added further data pointing to ribosome stalling and RQC for other cotranslationally inserted mitochondrial proteins (Table S6). Accordingly, this has also been included in the Discussion section. However, the identification and study of these other mitochondrial targets goes beyond the aim of our current study.

Minor comments

1. Page 1, mitochondrial proteins cross do not the intermembrane space through Tom40 but rather the outer membrane Response: We think the Reviewer #4 misunderstood the sentence because we are saying exactly what he/she states: mitoproteins cross the outer membrane to the intermembrane space through Tom40. Thus our sentence is:

“Usually, mitoproteins contact the central receptor Tom20 and cross to the intermembrane space (IMS) through Tom40, the β-barrel pore-forming subunit.”

Therefore, we kept the sentence.

Page 4, ATP1 is present in the matrix and not the inner membrane

Response: This has been corrected. We thank the Reviewer for pointing this.

The citations are missing at several places - they are left as "?"

Response: References have been corrected in the text.

It would be nice if microscopy images were colored in magenta and cyan, rather than red and green, to make them accessible to a wider audience.

Response: Green and red colors for fluorescent microscopy images are widely used in high-impact journals, especially when showing mitochondrial proteins and mitochondrial marker Su9-mCherry (Hughes et al., 2016, eLife, doi: 10.7554/eLife.13943; Kakimoto et al., 2018, Scientific Reports, doi: 10.1038/s41598-018-24466-0; Kreimendahl et al, 2020, BMC Biology, doi: 10.1186/s12915-020-00888-z). However, if the Reviewers think this is essential for publication, microscopy images can be colored in magenta and cyan instead.

Formally speaking, Tim50 is not per se a translocase, it is either a component of the translocase or, more precisely, a receptor of the translocase. Similarly, Tom20 and Tom70 are not membrane transporters but rather receptors of the TOM complex.

Response: We have changed the title and text to be more precise in the description of the components of the mitochondrial import systems as suggested by Reviewer #4.

Reviewer #4 (Significance (Required)):

---

This is a potentially interesting story, however, the conditions used for the analysis of the temperature sensitive mutants were either too harsh or these mutants are in general impossible to control, making the manuscript, in my opinion, unfortunately too premature for publication.

Response: We do not agree with the Reviewer #4 opinion, all experiments were done at 37ºC except the proteomic analysis that it is also confirmed further for Tim50 and Por1 proteins at 37ºC. We want to stress that we show all experiments with at least three biological replicates, individual values for each measurement are included now in the graphics as recommended by Reviewer #2, and the mean, SD and statistical tests are included. We make conclusions based in statistical significant differences along the manuscript. The temperature-sensitive yeast mutants used show reproducible analysis, they behave as expected in the controlled conditions used and they have been widely used in our lab and others (Pelechano and Alepuz, 2017; Zanelli and Valentini, 2005; Zanelli et al., 2006; Dias et al., 2008; Muñoz-Soriano et al., 2017; Rossi et al., 2014; Li et al., 2014; Xiao et al., 2024).

Author Response

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

eLife assessment

This work provides a valuable contribution and assessment of what it means to replicate a null study finding, and what are the appropriate methods for doing so (apart from a rote p-value assessment). Through a convincing re-analysis of results from the Reproducibility Project: Cancer Biology using frequentist equivalence testing and Bayes factors, the authors demonstrate that even when reducing 'replicability success' to a single criterion, how precisely replication is measured may yield differing results. Less focus is directed to appropriate replication of non-null findings.

Reviewer #1 (Public Review):

Summary:

The goal of Pawel et al. is to provide a more rigorous and quantitative approach for judging whether or not an initial null finding (conventionally with p ≥ 0.05) has been replicated by a second similarly null finding. They discuss important objections to relying on the qualitative significant/non-significant dichotomy to make this judgment. They present two complementary methods (one frequentist and the other Bayesian) which provide a superior quantitative framework for assessing the replicability of null findings.

Strengths:

Clear presentation; illuminating examples drawn from the well-known Reproducibility Project: Cancer Biology data set; R-code that implements suggested analyses. Using both methods as suggested provides a superior procedure for judging the replicability of null findings.

Weaknesses:

The proposed frequentist and the Bayesian methods both rely on binary assessments of an original finding and its replication. I'm not sure if this is a weakness or is inherent to making binary decisions based on continuous data.

For the frequentist method, a null finding is considered replicated if the original and replication 90% confidence intervals for the effects both fall within the equivalence range. According to this approach, a null finding would be considered replicated if p-values of both equivalences tests (original and replication) were, say, 0.049, whereas would not be considered replicated if, for example, the equivalence test of the original study had a p-value of 0.051 and the replication had a p-value of 0.001. Intuitively, the evidence for replication would seem to be stronger in the second instance. The recommended Bayesian approach similarly relies on a dichotomy (e.g., Bayes factor > 1).

Thanks for the suggestions, we now emphasize more strongly in the “Methods for assessing replicability of null results” and “Conclusions” sections that both TOST p-values and Bayes factors are quantitative measures of evidence that do not require dichotomization into “success” or “failure”.

Reviewer #2 (Public Review):

Summary:

The study demonstrates how inconclusive replications of studies initially with p > 0.05 can be and employs equivalence tests and Bayesian factor approaches to illustrate this concept. Interestingly, the study reveals that achieving a success rate of 11 out of 15, or 73%, as was accomplished with the non-significance criterion from the RPCB (Reproducibility Project: Cancer Biology), requires unrealistic margins of Δ > 2 for equivalence testing.

Strengths:

The study uses reliable and shareable/open data to demonstrate its findings, sharing as well the code for statistical analysis. The study provides sensitivity analysis for different scenarios of equivalence margin and alfa level, as well as for different scenarios of standard deviations for the prior of Bayes factors and different thresholds to consider. All analysis and code of the work is open and can be replicated. As well, the study demonstrates on a case-by-case basis how the different criteria can diverge, regarding one sample of a field of science: preclinical cancer biology. It also explains clearly what Bayes factors and equivalence tests are.

Weaknesses:

It would be interesting to investigate whether using Bayes factors and equivalence tests in addition to p-values results in a clearer scenario when applied to replication data from other fields. As mentioned by the authors, the Reproducibility Project: Experimental Philosophy (RPEP) and the Reproducibility Project: Psychology (RPP) have data attempting to replicate some original studies with null results. While the RPCB analysis yielded a similar picture when using both criteria, it is worth exploring whether this holds true for RPP and RPEP. Considerations for further research in this direction are suggested. Even if the original null results were excluded in the calculation of an overall replicability rate based on significance, sensitivity analyses considering them could have been conducted. The present authors can demonstrate replication success using the significance criteria in these two projects with initially p < 0.05 studies, both positive and non-positive.

Other comments:

• Introduction: The study demonstrates how inconclusive replications of studies initially with p > 0.05 can be and employs equivalence tests and Bayesian factor approaches to illustrate this concept. Interestingly, the study reveals that achieving a success rate of 11 out of 15, or 73%, as was accomplished with the non-significance criterion from the RPCB (Reproducibility Project: Cancer Biology), requires unrealistic margins of Δ > 2 for equivalence testing.

• Overall picture vs. case-by-case scenario: An interesting finding is that the authors observe that in most cases, there is no substantial evidence for either the absence or the presence of an effect, as evidenced by the equivalence tests. Thus, using both suggested criteria results in a picture similar to the one initially raised by the paper itself. The work done by the authors highlights additional criteria that can be used to further analyze replication success on a case-by-case basis, and I believe that this is where the paper's main contributions lie. Despite not changing the overall picture much, I agree that the p-value criterion by itself does not distinguish between (1) a situation where the original study had low statistical power, resulting in a highly inconclusive non-significant result that does not provide evidence for the absence of an effect and (2) a scenario where the original study was adequately powered, and a non-significant result may indeed provide some evidence for the absence of an effect when analyzed with appropriate methods. Equivalence testing and Bayesian factor approaches are valuable tools in both cases.

Regarding the 0.05 threshold, the choice of the prior distribution for the SMD under the alternative H1 is debatable, and this also applies to the equivalence margin. Sensitivity analyses, as highlighted by the authors, are helpful in these scenarios.

Thank you for the thorough review and constructive feedback. We have added an additional “Appendix C: Null results from the RPP and EPRP” that shows equivalence testing and Bayes factor analyses for the RPP and EPRP null results.

Reviewer #3 (Public Review):

Summary:

The paper points out that non-significance in both the original study and a replication does not ensure that the studies provide evidence for the absence of an effect. Also, it can not be considered a "replication success". The main point of the paper is rather obvious. It may be that both studies are underpowered, in which case their non-significance does not prove anything. The absence of evidence is not evidence of absence! On the other hand, statistical significance is a confusing concept for many, so some extra clarification is always welcome.

One might wonder if the problem that the paper addresses is really a big issue. The authors point to the "Reproducibility Project: Cancer Biology" (RPCB, Errington et al., 2021). They criticize Errington et al. because they "explicitly defined null results in both the original and the replication study as a criterion for replication success." This is true in a literal sense, but it is also a little bit uncharitable. Errington et al. assessed replication success of "null results" with respect to 5 criteria, just one of which was statistical (non-)significance.

It is very hard to decide if a replication was "successful" or not. After all, the original significant result could have been a false positive, and the original null-result a false negative. In light of these difficulties, I found the paper of Errington et al. quite balanced and thoughtful. Replication has been called "the cornerstone of science" but it turns out that it's actually very difficult to define "replication success". I find the paper of Pawel, Heyard, Micheloud, and Held to be a useful addition to the discussion.

Strengths:

This is a clearly written paper that is a useful addition to the important discussion of what constitutes a successful replication.

Weaknesses:

To me, it seems rather obvious that non-significance in both the original study and a replication does not ensure that the studies provide evidence for the absence of an effect. I'm not sure how often this mistake is made.

Thanks for the feedback. We do not have systematic data on how often the mistake of confusing absence of evidence with evidence of absence has been made in the replication context, but we do know that it has been made in at least three prominent large-scale replication projects (the RPP, RPEP, RPCB). We therefore believe that there is a need for our article.

Moreover, we agree that the RPCB provided a nuanced assessment of replication success using five different criteria for the original null results. We emphasize this now more in the “Introduction” section. However, we do not consider our article as “a little bit uncharitable” to the RPCB, as we discuss all other criteria used in the RPCB and note that our intent is not to diminish the important contributions of the RPCB, but rather to build on their work and provide constructive recommendations for future researchers. Furthermore, in response to comments made by Reviewer #2, we have added an additional “Appendix B: Null results from the RPP and EPRP” that shows equivalence testing and Bayes factor analyses for null results from two other replication projects, where the same issue arises.

Reviewer #1 (Recommendations For The Authors):

The authors may wish to address the dichotomy issue I raise above, either in the analysis or in the discussion.

Thank you, we now emphasize that Bayes factors and TOST p-values do not need to be dichotomized but can be interpreted as quantitative measures of evidence, both in the “Methods for assessing replicability of null results” and the “Conclusions” sections.

Reviewer #2 (Recommendations For The Authors):

Given that, here follow additional suggestions that the authors should consider in light of the manuscript's word count limit, to avoid confusing the paper's main idea:

2) Referencing: Could you reference the three interesting cases among the 15 RPCB null results (specifically, the three effects from the original paper #48) where the Bayes factor differs qualitatively from the equivalence test?

We now explicitly cite the original and replication study from paper #48.

3) Equivalence testing: As the authors state, only 4 out of the 15 study pairs are able to establish replication success at the 5% level, in the sense that both the original and the replication 90% confidence intervals fall within the equivalence range. Among these 4, two (Paper #48, Exp #2, Effect #5 and Paper #48, Exp #2, Effect #6) were initially positive with very low p-values, one (Paper #48, Exp #2, Effect #4) had an initial p of 0.06 and was very precisely estimated, and the only one in which equivalence testing provides a clearer picture of replication success is Paper #41, Exp #2, Effect #1, which had an initial p-value of 0.54 and a replication p-value of 0.05. In this latter case (or in all these ones), one might question whether the "liberal" equivalence range of Δ = 0.74 is the most appropriate. As the authors state, "The post-hoc specification of equivalence margins is controversial."

We agree that the post hoc choice of equivalence ranges is a controversial issue. The margins define an equivalence region where effect sizes are considered practically negligible, and we agree that in many contexts SMD = 0.74 is a large effect size that is not practically negligible. We therefore present sensitivity analyses for a wide range of margins. However, we do not think that the choice of this margin is more controversial for the mentioned studies with low p-values than for other studies with greater p-values, since the question of whether a margin plausibly encodes practically negligible effect sizes is not related to the observed p-value of a study. Nevertheless, for the new analyses of the RPP and EPRP data in Appendix B, we have added additional sensitivity analyses showing how the individual TOST p-values and Bayes factors vary as a function of the margin and the prior standard deviation. We think that these analyses provide readers with an even more transparent picture regarding the implications of the choice of these parameters than the “project-wise” sensitivity analyses in Appendix A.

4) Bayes factor suggestions: For the Bayes factor approach, it would be interesting to discuss examples where the BF differs slightly. This is likely to occur in scenarios where sample sizes differ significantly between the original study and replication. For example, in Paper #48, Exp #2 and Effect #4, the initial p is 0.06, but the BF is 8.1. In the replication, the BF dramatically drops to < 1/1000, as does the p-value. The initial evidence of 8.1 indicates some evidence for the absence of an effect, but not strong evidence ("strong evidence for H0"), whereas a p-value of 0.06 does not lead to such a conclusion; instead, it favors H1. It would be interesting if the authors discussed other similar cases in the paper. It's worth noting that in Paper #5, Exp #1, Effect #3, the replication p-value is 0.99, while the BF01 is 2.4, almost indicating "moderate" evidence for H0, even though the p-value is inconclusive.

We agree that some of the examples nicely illustrate conceptual differences between p-values and Bayes factors, e.g., how they take into account sample size and effect size. As methodologists, we find these aspects interesting ourselves, but we think that emphasizing them is beyond the scope of the paper and would distract eLife readers from the main messages.

Concerning the conceptual differences between Bayes factors and TOST p-values, we already discuss a case where there are qualitative differences in more detail (original paper #48). We added another discussion of this phenomenon in the Appendix C as it also occurs for the replication of Ranganath and Nosek (2008) that was part of the RPP.

5) p-values, magnitude and precision: It's noteworthy to emphasize, if the authors decide to discuss this, that the p-value is influenced by both the effect's magnitude and its precision, so in Paper #9, Exp #2, Effect #6, BF01 = 4.1 has a higher p-value than a BF01 = 2.3 in its replication. However, there are cases where both p-values and BF agree. For example, in Paper #15, Exp #2, Effect #2, both the original and replication studies have similar sample sizes, and as the p-value decreases from p = 0.95 to p = 0.23, BF01 decreases from 5.1 ("moderate evidence for H0") to 1.3 (region of "Absence of evidence"), moving away from H0 in both cases. This also occurs in Paper #24, Exp #3, Effect #6.

We appreciate the suggestions but, as explained before, think that the message of our paper is better understood without additional discussion of more general differences between p-values and Bayes factors.

6) The grey zone: Given the above topic, it is important to highlight that in the "Absence of evidence grey zone" for the null hypothesis, for example, in Paper #5, Exp #1, Effect #3 with a p = 0.99 and a BF01 = 2.4 in the replication, BF and p-values reach similar conclusions. It's interesting to note, as the authors emphasize, that Dawson et al. (2011), Exp #2, Effect #2 is an interesting example, as the p-value decreases, favoring H1, likely due to the effect's magnitude, even with a small sample size (n = 3 in both original and replications). Bayes factors are very close to one due to the small sample sizes, as discussed by the authors.

We appreciate the constructive comments. We think that the two examples from Dawson et al. (2011) and Goetz et al. (2011) already nicely illustrate absence of evidence and evidence of absence, respectively, and therefore decided not to discuss additional examples in detail, to avoid redundancy.

7) Using meta-analytical results (?): For papers from RPCB, comparing the initial study with the meta-analytical results using Bayes factor and equivalence testing approaches (thus, increasing the sample size of the analysis, but creating dependency of results since the initial study would affect the meta-analytical one) could change the conclusions. This would be interesting to explore in initial studies that are replicated by much larger ones, such as: Paper #9, Exp #2, Effect #6; Goetz et al. (2011), Exp #1, Effect #1; Paper #28, Exp #3, Effect #3; Paper #41, Exp #2, Effect #1; and Paper #47, Exp #1, Effect #5).

Thank you for the suggestion. We considered adding meta-analytic TOST p-values and Bayes factors before, but decided that Figure 3 and the results section are already quite technical, so adding more analyses may confuse more than help. Nevertheless, these meta-analytic approaches are discussed in the “Conclusions” section.

8) Other samples of fields of science: It would be interesting to investigate whether using Bayes factors and equivalence tests in addition to p-values results in a clearer scenario when applied to replication data from other fields. As mentioned by the authors, the Reproducibility Project: Experimental Philosophy (RPEP) and the Reproducibility Project: Psychology (RPP) have data attempting to replicate some original studies with null results. While the RPCB analysis yielded a similar picture when using both criteria, it is worth exploring whether this holds true for RPP and RPEP. Considerations for further research in this direction are suggested. Even if the original null results were excluded in the calculation of an overall replicability rate based on significance, sensitivity analyses considering them could have been conducted. The present authors can demonstrate replication success using the significance criteria in these two projects with initially p < 0.05 studies, both positive and non-positive.

Thank you for the excellent suggestion. We added an Appendix B where the null results from the RPP and EPRP are analyzed with our proposed approaches. The results are also discussed in the “Results” and “Conclusions” sections.

9) Other approaches: I am curious about the potential impact of using an approach based on equivalence testing (as described in https://arxiv.org/abs/2308.09112). It would be valuable if the authors could run such analyses or reference the mentioned work.

Thank you. We were unaware of this preprint. It seems related to the framework proposed by Stahel W. A. (2021) New relevance and significance measures to replace p-values. PLoS ONE 16(6): e0252991. https://doi.org/10.1371/journal.pone.0252991

We now cite both papers in the discussion.

10) Additional evidence: There is another study in which replications of initially p > 0.05 studies with p > 0.05 replications were also considered as replication successes. You can find it here: https://www.medrxiv.org/content/10.1101/2022.05.31.22275810v2. Although it involves a small sample of initially p > 0.05 studies with already large sample sizes, the work is currently under consideration for publication in PLOS ONE, and all data and materials can be accessed through OSF (links provided in the work).

Thank you for sharing this interesting study with us. We feel that it is beyond the scope of the paper to include further analyses as there are already analyses of the RPCB, RPP, and EPRP null results. However, we will keep this study in mind for future analysis, especially since all data are openly available.

11) Additional evidence 02: Ongoing replication projects, such as the Brazilian Reproducibility Initiative (BRI) and The Sports Replication Centre (https://ssreplicationcentre.com/), continue to generate valuable data. BRI is nearing completion of its results, and it promises interesting data for analyzing replication success using p-values, equivalence regions, and Bayes factor approaches.

We now cite these two initiatives as examples of ongoing replication projects in the introduction. Similarly as for your last point, we think that it is beyond the scope of the paper to include further analyses as there are already analyses of the RPCB, RPP, and EPRP null results.

Reviewer #3 (Recommendations For The Authors):

I have no specific recommendations for the authors.

Thank you for the constructive review.

Reviewing Editor (Recommendations For the Authors):

I recognize that it was suggested to the authors by the previous Reviewing Editor to reduce the amount of statistical material to be made more suitable for a non-statistical audience, and so what I am about to say contradicts advice you were given before. But, with this revised version, I actually found it difficult to understand the particulars of the construction of the Bayes Factors and would have appreciated a few more sentences on the underlying models that fed into the calculations. In my opinion, the provided citations (e.g., Dienes Z. 2014. Using Bayes to get the most out of non-significant results) did not provide sufficient background to warrant a lack of more technical presentation here.

Thank you for the feedback. We added a new “Appendix C: Technical details on Bayes factors” that provides technical details on the models, priors, and calculations underlying the Bayes factors.

Author response:

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

Reviewer #1 (Public Review):

Summary:

Bendzunas, Byrne et al. explore two highly topical areas of protein kinase regulation in this manuscript. Firstly, the idea that Cys modification could regulate kinase activity. The senior authors have published some standout papers exploring this idea of late, and the current work adds to the picture of how active site Cys might have been favoured in evolution to serve critical regulatory functions. Second, BRSK1/2 are understudied kinases listed as part of the "dark kinome" so any knowledge of their underlying regulation is of critical importance to advancing the field.

Strengths:

In this study, the author pinpoints highly-conserved, but BRSK-specific, Cys residues as key players in kinase regulation. There is a delicate balance between equating what happens in vitro with recombinant proteins relative to what the functional consequence of Cys mutation might be in cells or organisms, but the authors are very clear with the caveats relating to these connections in their descriptions and discussion. Accordingly, by extension, they present a very sound biochemical case for how Cys modification might influence kinase activity in cellular environs.

Weaknesses:

I have very few critiques for this study, and my major points are barely major.

Major points

(1) My sense is that the influence of Cys mutation on dimerization is going to be one of the first queries readers consider as they read the work. It would be, in my opinion, useful to bring forward the dimer section in the manuscript.

We agree that the influence of Cys on BRSK dimerization is a topic of significant interest. Our primary focus was to explore oxidative regulation of the understudied BRSK kinases as they contain a conserved T-loop Cys, and we have previously demonstrated that equivalent residues at this position in related kinases were critical drivers of oxidative modulation of catalytic activity. We have demonstrated here that BRSK1 & 2 are similarly regulated by redox and this is due to oxidative modification of the T+2 Cys, in addition to Cys residues that are conserved amongst related ARKs as well as BRSK-specific Cys. Although we also provide evidence for limited redox-sensitive higher order BRSK species (dimers) in our in vitro analysis, these represent a small population of the total BRSK protein pool (this was validated by SEC-MALs analysis). As such, we do not have strong evidence to suggest that these limited dimers significantly contribute to the pronounced inhibition of BRSK1 & 2 in the presence of oxidizing agents, and instead believe that other biochemical mechanisms likely drive this response. This may result from oxidized Cys altering the conformation of the activation loop. Indeed, the formation of an intramolecular disulfide within the T-loop of BRSK1 & 2, which we detected by MS, is one such regulatory modification. It is noteworthy, that intramolecular disulfide bonds within the T-loop of AKT and MELK have already been shown to induce an inactive state in the kinase, and we posit a similar mechanism for BRSKs.

While we recognize the potential importance of dimerization in this context, our current data from in vitro and cell-based assays do not provide substantial evidence to assert dimerization as a primary regulatory mechanism. Hence, we maintained a more conservative stance in our manuscript, discussing dimerization in later sections where it naturally followed from the initial findings. That being said, we acknowledge the potential significance of dimerization in the regulation of the BRSK T-loop cysteine. We believe this aspect merits further investigation and could indeed be the focus of a follow-up study.

(2) Relatedly, the effect of Cys mutation on the dimerization properties of preparations of recombinant protein is not very clear as it stands. Some SEC traces would be helpful; these could be included in the supplement.

In order to determine whether our recombinant BRSK proteins (and T-loop mutants) existed as monomers or dimers, we performed SDS-PAGE under reducing and non-reducing conditions (Fig 7). This unambiguously revealed that a monomer was the prominent species, with little evidence of dimers under these experimental conditions (even in the presence of oxidizing agents). Although we cannot discount a regulatory role for BRSK dimers in other physiological contexts, we could not produce sufficient evidence to suggest that multimerization played a substantial role in modifying BRSK kinase activity in our assays. We note that our in vitro analysis was performed using truncated forms of the protein, and as such it is entirely possible that regions of the protein that flank the kinase domain may serve additional regulatory functions that may include higher order BRSK conformations. In this regard, although we have not included SEC traces of our recombinant proteins, we have included analytical SEC-MALS of the truncated proteins (Supplementary Figure 6) which we believe to be more informative. We have also now included additional SEC-MALS data for BRSK2 C176A and C183A (Supplementary Figure 6d and e), which supports our findings in Fig 7, demonstrating the presence of limited dimer species under non-reducing conditions.

(3) Is there any knowledge of Cys mutants in disease for BRSK1/2?

We have conducted an extensive search across several databases: COSMIC (Catalogue of Somatic Mutations in Cancer), ProKinO (Protein Kinase Ontology), and TCGA (The Cancer Genome Atlas). These databases are well-regarded for their comprehensive and detailed records of mutations related to cancer and protein kinases. Our analysis using the COSMIC and TCGA databases focused on identifying any reported instances of Cys mutations in BRSK1/2 that are implicated in cancer. Additionally, we utilized the ProKinO database to explore the broader landscape of protein kinase mutations, including any potential disease associations of Cys mutations in BRSK1/2. However, we found no evidence to indicate the presence of Cys mutations in BRSK1/2 that are associated with cancer or disease. This lack of association in the current literature and database records suggests that, as of our latest search, Cys mutations in BRSK1/2 have not been reported as significant contributors to pathogenesis.

(4) In bar charts, I'd recommend plotting data points. Plus, it is crucial to report in the legend what error measure is shown, the number of replicates, and the statistical method used in any tests.

We have added the data points to the bar charts and included statistical methods in figure legends.

(5) In Figure 5b, the GAPDH loading control doesn't look quite right.

The blot has been repeated and updated.

(6) In Figure 7 there is no indication of what mode of detection was used for these gels.

We have updated the figure legend to confirm that the detection method was western blot.

(7) Recombinant proteins - more detail should be included on how they were prepared. Was there a reducing agent present during purification? Where did they elute off SEC... consistent with a monomer of higher order species?

We have added ‘produced in the absence of reducing agents unless stated otherwise’ in the methods section to improve clarity. Although we have not added additional sentences to describe the elution profile of the BRSK proteins by SEC during purification, we believe that the inclusion of analytical SEC-MALS data is sufficient evidence that the proteins are largely monomeric under non-reducing conditions.

Reviewer #2 (Public Review):

Summary:

In this study by Bendzunas et al, the authors show that the formation of intra-molecular disulfide bonds involving a pair of Cys residues near the catalytic HRD motif and a highly conserved T-Loop Cys with a BRSK-specific Cys at an unusual CPE motif at the end of the activation segment function as repressive regulatory mechanisms in BSK1 and 2. They observed that mutation of the CPE-Cys only, contrary to the double mutation of the pair, increases catalytic activity in vitro and drives phosphorylation of the BRSK substrate Tau in cells. Molecular modeling and molecular dynamics simulations indicate that oxidation of the CPE-Cys destabilizes a conserved salt bridge network critical for allosteric activation. The occurrence of spatially proximal Cys amino acids in diverse Ser/Thr protein kinase families suggests that disulfide-mediated control of catalytic activity may be a prevalent mechanism for regulation within the broader AMPK family. Understanding the molecular mechanisms underlying kinase regulation by redox-active Cys residues is fundamental as it appears to be widespread in signaling proteins and provides new opportunities to develop specific covalent compounds for the targeted modulation of protein kinases.

The authors demonstrate that intramolecular cysteine disulfide bonding between conserved cysteines can function as a repressing mechanism as indicated by the effect of DTT and the consequent increase in activity by BSK-1 and -2 (WT). The cause-effect relationship of why mutation of the CPE-Cys only increases catalytic activity in vitro and drives phosphorylation of the BRSK substrate Tau in cells is not clear to me. The explanation given by the authors based on molecular modeling and molecular dynamics simulations is that oxidation of the CPE-Cys (that will favor disulfide bonding) destabilizes a conserved salt bridge network critical for allosteric activation. However, no functional evidence of the impact of the salt-bridge network is provided. If you mutated the two main Cys-pairs (aE-CHRD and A-loop T+2-CPE) you lose the effect of DTT, as the disulfide pairs cannot be formed, hence no repression mechanisms take place, however when looking at individual residues I do not understand why mutating the CPE only results in the opposite effect unless it is independent of its connection with the T+2residue on the A-loop.

Strengths:

This is an important and interesting study providing new knowledge in the protein kinase field with important therapeutic implications for the rationale design and development of next-generation inhibitors.

Weaknesses:

There are several issues with the figures that this reviewer considers should be addressed.

Reviewer #1 (Recommendations for The Authors):

Major points

Page 26 - the discussion could be more concise. There's an element of recapping the results, which should be avoided.

Regarding the conciseness of the discussion section, we have thoroughly revised it to ensure a more succinct presentation, deliberately avoiding the recapitulation of results. The revised discussion now focuses on interpreting the findings and their implications, steering clear of redundancy with the results section.

Figure 1b seems to be mislabeled/annotated. I recommend checking whether the figure legends match more broadly. Figure 1 appears to be incorrectly cited throughout the results.

Thank you for pointing out the discrepancies in the labeling and citation of Figure 1b. We have carefully reviewed and corrected these issues to ensure that all figure labels, legends, and citations accurately reflect the corresponding data and illustrations. We appreciate your attention to detail and the opportunity to improve the clarity and accuracy of our presentation.

Figure 6 - please include a color-coding key in the figure. Further support for these simulations could be provided by supplementary movies or plots of the interaction. Figure 4 colour palette should be adjusted for the spheres in the Richardson diagrams to have greater distinction.

As suggested, we have amended the colour palette in Figure 4 to improve conformity throughout the figure.

Minor points

Figure 2 - it'd be helpful to know what the percentage coverage of peptides is.

We have updated the figure legend to include peptide coverage for both proteins

Some typos - Supp 2 legend "Domians".

Fixed

Figure 6 legend - analyzed by needs a space;

Fixed

Fig 8 legend schematic misspelled.

Fixed

Broadly, if you Google T-loop you get a pot pourri of enzyme answers. Why not just use Activation loop?

The choice of "T-loop" over "Activation loop" in our manuscript was made to maintain consistency with other literature in the field, and in particular our previous paper “Aurora A regulation by reversible cysteine oxidation reveals evolutionarily conserved redox control of Ser/Thr protein kinase activity” where we refer to the activation loop cysteine as T-loop + 2. We acknowledge the varied enzyme contexts in which "T-loop" is used and agree on the importance of clarity. To address this, we made an explicit note in the manuscript that the "T-loop" is also referred to as the "Activation loop", ensuring readers are aware of the interchangeable use of these terms. Additionally, this nomenclature facilitates a more straightforward designation of cysteine residues within the loop (T+2 Cysteine). We believe this approach balances adherence to established conventions with the need for clarity and precision in our descriptions.

Methods - what is LR cloning. Requires some definition. Some manufacturer detail is missing in methods, and referring to prior work is not sufficient to empower readers to replicate.

We agree, and have added the following to the methods section:

“BRSK1 and 2 were sub-cloned into pDest vectors (to encode the expression of N-terminal Flag or HA tagged proteins) using the Gateway LR Clonase II system (Invitrogen) according to the manufacturer’s instructions. pENtR BRSK1/2 clones were obtained in the form of Gateway-compatible donor vectors from Dr Ben Major (Washington University in St. Louis). The Gateway LR Clonase II enzyme mix mediates recombination between the attL sites on the Entry clone and the attR sites on the destination vector. All cloned BRSK1/2 genes were fully sequenced prior to use.”

Page 7 - optimal settings should be reported. How were pTau signals quantified and normalised?

We have added the following to the methods section:

“Two-color Western blot detection method employing infrared fluorescence was used to measure the ratio of Tau phospho serine 262 to total Tau. Total GFP Tau was detected using a mouse anti GFP antibody and visualized at 680 nm using goat anti mouse IRdye 680 while phospho-tau was detected using a Tau phospho serine 262 specific antibody and visualized at 800 nm using goat anti rabbit IRdye 800. Imaging was performed using a Licor Odessey Clx with scan control settings set to 169 μm, medium quality, and 0.0 mm distance. Quantification was performed using Licor image studio on the raw image files. Total Tau to phospho Tau ratio was determined by measuring the ratio of the fluorescence intensities measured at 800 nm (pTau) to those at 680 nm (total tau).”

In the Figure 6g-j legend, the salt bridge is incorrectly annotated as E185-R248 rather than 258.

Fixed

Lines 393-395 provides a repeat statement on BRSKs phosphorylating Tau (from 388-389).

We have removed the repetition and reworded the opening lines of the results section to improve the overall flow of the manuscript.

Supp. Figure 1 is difficult to view - would it be possible to increase the size of the phylogenetic analysis?

We thank the reviewer for this observation. We have rotated (90°) and expanded the figure so that it can be more clearly viewed

Supp. Figure 2 - BRSK1/2 incorrectly spelled.

Fixed

Please check the alignment of labels in Supp. Figure 3e.

Fixed

Reviewer #2 (Recommendations For The Authors):

(1) In Figure 1, current panel b is not mentioned/described in the figure legend and as a consequence, the rest of the panels in the legends do not fit the content of the figure.

Reviewer 1 also noted this error, and we have amended the manuscript accordingly.

What is the rationale for using the HEK293T cells as the main experimental/cellular system? Are there cell lines that express both proteins endogenously so that the authors can recapitulate the results obtained from ectopic overexpression?

The selection of HEK-293T cells was driven by their well-established utility in overexpression studies, which make them ideal for the investigation of protein interactions and redox regulation. This cell line's robust transfection efficiency and well-characterized biology provide a reliable platform for dissecting the molecular mechanisms underlying the redox regulation of proteins. Furthermore, the use of HEK-293T cells aligns with the broader scientific practice, serving as a common ground for comparability with existing literature in the field of BRSK1/2 signaling, protein regulation and interaction studies.

The application of HEK-293T cells as a model system in our study serves as a foundational step towards eventually elucidating the functions of BRSK1/2 in neuronal cells, where these kinases are predominantly expressed and play critical roles. Given the fact that BRSKs are classed as ‘understudied’ kinases, the choice of a HEK-293T co-overexpression system allowed us to analyze the direct effects of BRSK kinase activity (using phosphorylation of Tau as a readout) in a cellular context and in more controlled manner. This approach not only aids in the establishment of a baseline understanding of the redox regulation of BRSK1/2, but also sets the stage for subsequent investigations in more physiologically relevant neuronal models

In current panel d, could the authors recapitulate the same experimental conditions as in current panel c?

Figure 1 panel c shows that both BRSK1 and 2 are reversibly inhibited by oxidizing agents such as H2O2, whilst panels d and e show the concentration dependent activation and inhibition of the BRSKs with increasing concentrations of DTT and H2O2 respectively. The experimental conditions were identical, other than changing amounts of reducing and oxidizing agents, and used the same peptide coupled assays. Data for all experiments were originally collected in ‘real time’ as depicted in Fig 1c (increase in substrate phosphorylation over time). However, to aid interpretation of the data, we elected to present the latter two panels as dose response curves by calculating the change in the rate of enzyme activity (shown as pmol phosphate incorporated into the peptide substrate per min) for each condition. To aid the reader, we now include an additional supplementary figure (new supplementary figure 2) depicting BRSK1 and 2 dependent phosphorylation of the peptide substrate in the presence of different concentrations of DTT and H2O2 in a real time (kinetic) assay. The new data shown is a subset of the unprocessed data that was used to calculate the rates of BRSK activity in Fig 1d & e.

Why did the authors use full-length constructs in these experiments and did not in e.g. Figure 2 where they used KD constructs instead?

In the initial experiments, illustrated in Figure 1, we employed full-length protein constructs to establish a proof of concept, demonstrating the overall behavior and interactions of the proteins in their full-length form. This confirmed that BRSK1 & 2, which both contain a conserved T + 2 Cys residue that is frequently prognostic for redox sensitivity in related kinases, displayed a near-obligate requirement for reducing agents to promote kinase activity.  

Subsequently, in Figure 2, our focus shifted towards delineating the specific regions within the proteins that are critical for redox regulation. By using constructs that encompass only the kinase domain, we aimed to demonstrate that the redox-sensitive regulation of these proteins is predominantly mediated by specific cysteine residues located within the kinase domain itself. This strategic use of the kinase domain of the protein allowed for a more targeted investigation. Furthermore, in our hands these truncated forms of the protein were more stable at higher concentrations, enabling more detailed characterization of the proteins by DSF and SEC-MALS. We predict that the flanking disordered regions of the full-length protein (as predicted by AlphaFold) contribute to this effect.

(2) In Figure 2, Did the authors try to do LC/MS-MS in the same experimental conditions as in Figure 1 (e.g. buffer minus/plus DTT, H2O2, H2O2 + DTT)?

We would like to clarify that the mass spectrometry experiments were conducted exclusively on proteins purified under native (non-reducing) conditions. We did not extend the LC/MS-MS analyses to include proteins treated with various buffer conditions such as minus/plus DTT, H2O2, or H2O2 + DTT as used in the experiments depicted in Figure 1. Given that we could readily detect disulfides in the absence of oxidizing agents, we did not see the benefit of additional treatment conditions as peroxide treatment of protein samples can frequently complicate interpretation of MS data. However, it should be noted that prior to MS analysis, tryptic peptides were subjected to a 50:50 split, with one half alkylated in the presence of DTT (as described in the methods section) to eliminate disulfides and other transiently oxidized Cys forms. Comparative analysis between reduced and non-reduced tryptic peptides improved our confidence when assigning disulfide bonds (which were eliminated in identical peptides in the presence of DTT).

On panel b, why did the authors show alphafold predictions and not empiric structural information (e.g. X-ray, EM,..)?

The AlphaFold models were primarily utilized to map the general locations of redox-sensitive cysteine pairs within the proteins of interest. Although we have access to the crystal structure of mouse BRSK2, they do not fully capture the active conformation seen in the Alphafold model of the human version. The use of AlphaFold models for human proteins in this study aids in consistently tracking residue numbering across the manuscript, offering a useful framework for understanding the spatial arrangement of these critical cysteine pairs in their potentially active-like states. This approach facilitates our analysis and discussion by providing a reference for the structural context of these residues in the human proteins.

What was the rationale for using the KD construct and not the FL as in Figure 1?

The rationale to use the kinase domain was primarily based on the significantly lower confidence in the structural predictions for regions outside the kinase domain (KD). Our experimental focus was to investigate the role of conserved cysteine residues within the kinase domain, which are critical for the protein's function and regulation. This targeted approach allowed us to concentrate our analyses on the most functionally relevant and structurally defined portion of the protein, thereby enhancing the precision and relevance of our findings. As is frequently the case, truncated forms of the protein, consisting only of the kinase domain, are much more stable than their full length counterparts and are therefore more amenable to in vitro biochemical analysis. In our hands this was true for both BRSK1 and 2, and as such much of the data collected here was generated using kinase-domain (KD) constructs. Simulations using the KD structures are therefore much more representative of our original experimental setup.

The BSK1 KD construct appears to be rather inactive and not responsive to DTT treatment. Could the authors comment on the differences observed with the FL construct of Figure 1

It is important to note that BRSK1, in general, exhibits lower intrinsic activity compared to BRSK2. This reduced activity could be attributed to a range of factors, including the need for activation by upstream kinases such as LKB1, as well as potential post-translational modifications (PTMs) that may be absent in the bacterially expressed KD construct. The full-length forms of the protein were purified from Sf21 cells, and as such may have additional modifications that are lacking in the bacterially derived KD counterparts. We also cannot discount additional regulatory roles of the regions that flank the KD, and these may contribute in part to the modest discrepancy observed between constructs.  Despite these differences, it is crucial to emphasize that both the KD and FL constructs of BRSK1 are regulated by DTT, indicating a conserved redox-dependent activation for both of the related BRSK proteins.  

(3) In Figure 4, on panel A wouldn´t the authors expect that mutating on the pairs e.g. C198A in BSK1 would have the same effect as mutating the C191 from the T+2 site? Did they try mutating individual sites of the aE/CHRD pair? The same will apply to BSK2

We appreciate the insightful comment. It's important to clarify that the redox regulation of these proteins is influenced not solely by the formation of disulfide bonds but also by the oxidation state of individual cysteine residues, particularly the T+2 Cys. This nuanced mechanism of regulation allows for a diverse range of functional outcomes based on the specific cysteine involved and its state of oxidation. This aspect forms a key finding of our paper, highlighting the complexity of redox regulation beyond mere disulfide bond formation. For example, AURA kinase activity is regulated by oxidation of a single T+2 Cys (Cys290, equivalent to Cys191 and Cys176 of BRSK1 and 2 respectively), but this regulation can be supplemented through artificial incorporation of a secondary Cys at the DFG+2 position (Byrne et al., 2020). This targeted genetic modification or AURA mirrors equivalent regulatory disulfide-forming Cys pairs that naturally occur in kinases such as AKT and MELK, and which provide an extra layer of regulatory fine tuning (and a possible protective role to prevent deleterious over oxidation) to the T+2 Cys. We surmise that the CPE Cys is also an accessory regulatory element to the T+2 Cys in BRSK1 +2, which is the dominant driver of BRSK redox sensitivity (as judged by the fact that CPE Cys mutants are still potently regulated by redox [Fig 4]), by locking it in an inactive disulfide configuration.

In our preliminary analysis of BRSK1, we observed that mutations of individual sites within the aE/CHRD pair was similarly detrimental to kinase activity as a tandem mutation (see reviewer figure 1). As discussed in the manuscript, we think that these Cys may serve important structural regulatory functions and opted to focus on co-mutations of the aE/CHRD pair for the remainder of our investigation.

Author response image 1.

In vitro kinase assays showing rates of in vitro peptide phosphorylation by WT and Cys-to-Ala (aE/CHRD residues) variants of BRSK1 after activation by LKB1.

In panels C and D, the same experimental conditions should have been measured as in A and B.

Panels A and B were designed to demonstrate the enzymatic activity and the response to DTT treatment to establish the baseline redox regulation of the kinase and a panel of Cys-to-Ala mutant variants. In contrast, panels C and D were specifically focused on rescue experiments with mutants that showed a significant effect under the conditions tested in A and B. These panels were intended to further explore the role of redox regulation in modulating the activity of these mutants, particularly those that retained some level of activity or exhibited a notable response to redox changes.

The rationale for this experimental design was to prioritize the investigation of mutants, such as those at the T+2 and CPE cysteine sites, which provided the most insight into the redox-dependent modulation of kinase activity. Other mutants, which resulted in inactivation, were deprioritized in this context as they offered limited additional information regarding the redox regulation mechanism. This focused approach allowed us to delve deeper into understanding how specific cysteine residues contribute to the redox-sensitive control of kinase function, aligning with the overall objective of elucidating the nuanced roles of redox regulation in kinase activity.

(4) In figure 5: Why did the authors use reduced Glutathione instead of DTT? The authors should have recapitulated the same experimental conditions as in Figure 4 and not focused only on the T+2 or the CPE single mutants but using the double and the aE/CHRD mutants as well, as internal controls and validation of the enzymatic assays using the modified peptide

Regarding the use of reduced glutathione (GSH) instead of DTT in Figure 5, we chose GSH for its well characterized biological relevance as an antioxidant in cellular responses to oxidative stress. Furthermore, while DTT has been widely used in experimental setups, it is also potentially cytotoxic at high concentrations.

Addressing the point on experimental consistency with Figure 4, we appreciate the suggestion and indeed had already conducted such experiments (Previously Supp Fig 3, now changed to current Supp Fig 4). These experiments include analyses of BRSK mutant activity in a HEK-293T model. However, we chose not to focus on inactivating mutants (such as the aE/CHRD mutants which had depleted expression levels possibly as a consequence of compromised structural integrity) or pursue the generation of double mutant CMV plasmids, as these were deemed unlikely to add significant insights into the core narrative of our study. Our focus remained on the mutants that yielded the most informative results regarding the redox regulation mechanisms in the in vitro setting, ensuring a clear and impactful presentation of our findings.

A time course evaluation of the reducing or oxidizing reagents should have been performed. Would we expect that in WT samples, and in the presence of GSH, and also in the case of the CPE mutant, an increment in the levels of Tau phosphorylation as a readout of BSK1-2 activity?

We acknowledge the importance of such analyses in understanding the dynamic nature of redox regulation on kinase activity and have included a time course (Supp Fig 2 e-g). These results confirm a depletion of Tau phosphorylation over time in response to peroxide generated by the enzyme glucose oxidase.

(5) In Figure 6, did the authors look at the functional impact of the residues with which interact the T+2 and the CPE motifs e.g. T174 and the E185-R258 tether?

Our primary focus was on the salt bridges, as this is a key regulatory structural feature that is conserved across many kinases. Regarding the additional interactions mentioned, we have thoroughly evaluated their roles and dynamics through molecular dynamics (MD) simulations but did not find any results of significant relevance to warrant inclusion.

(6) In Figure 7: Did the author look at the oligomerization state of the BSK1-2 multimers under non-reducing conditions? Were they also observed in the case of the FL constructs? What was the stoichiometry?

Our current work indicates that the kinase domain of BRSK1-2 primarily exists in a monomeric state, with some evidence of dimerization or multimer formation under specific conditions. Our SEC-MALS (Supp Fig 6) and SDS-PAGE analysis (Figure 7) clearly demonstrates that monomers are overwhelmingly the dominant species under non-reducing conditions (>90 %). We also conclude that these limited oligomeric species can be removed by inclusion of reducing agents such as DTT (Figure 7), which may suggest a role for a Cys residue(s). Notably, removal of the T+2 Cys was insufficient to prevent multimerization.

We were unable to obtain reliable SEC-MALS data for the full-length forms of the protein, likely due to the presence of disordered regions that flank the kinase domain which results in a highly heterodispersed and unstable preparation (at the concentrations required for SEC-MALS). Although we are therefore unable to comment on the stoichiometry of FL BRSK dimers, we can detect BRSK1 and 2 hetero- and homo-complexes in HEK-293T cells by IP, which supports the existence of limited BRSK1 & 2 dimers (Supp Fig 6a). However, we were unable to detect intermolecular disulfide bonds by MS, although this does not necessarily preclude their existence. The physiological role of BRSK multimerization (if any) and establishing specifically which Cys residues drive this phenomenon is of significant interest to our future investigations.

Author response:

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

Response to Reviewer #1:

Thank you for the careful reading and the positive evaluation of our manuscript. As you mentioned, the present study tried to address the question of how the lost genomic functions could be compensated by evolutionary adaptation, indicating the potential mechanism of "constructive" rather than "destructive" evolution. Thank you for the instructive comments that helped us to improve the manuscript. We sincerely hope the revised manuscript and the following point-to-point response meet your concerns.

• Line 80 "Growth Fitness" is this growth rate?

Yes. The sentence was revised as follows.

(L87-88) “The results demonstrated that most evolved populations (Evos) showed improved growth rates, in which eight out of nine Evos were highly significant (Fig. 1B, upper).”

• Line 94 a more nuanced understanding of r/K selection theory, allows for trade-ups between R and K, as well as trade-offs. This may explain why you did not see a trade-off between growth and carrying capacity in this study. See this paper https://doi.org/10.1038/s41396-023-01543-5. Overall, your evos lineages evolved higher growth rates and lower carrying capacity (Figures 1B, C, E). If selection was driving the evolution of higher growth rates, it may have been that there was no selective pressure to maintain high carrying capacity. This means that the evolutionary change you observed in carrying capacity may have been neutral "drift" of the carrying capacity trait, during selection for growth rate, not because of a trade-off between R and K. This is especially likely since carrying capacity declined during evolution. Unless the authors have convincing evidence for a tradeoff, I suggest they remove this claim.

• Line 96 the authors introduce a previous result where they use colony size to measure growth rate, this finding needs to be properly introduced and explained so that we can understand the context of the conclusion.

• Line 97 This sentence "the collapse of the trade-off law likely resulted from genome reduction." I am not sure how the authors can draw this conclusion, what is the evidence supporting that the genome size reduction causes the breakdown of the tradeoff between R and K (if there was a tradeoff)?

Thank you for the reference information and the thoughtful comments. The recommended paper was newly cited, and the description of the trade-off collapse was deleted. Accordingly, the corresponding paragraph was rewritten as follows.

(L100-115) “Intriguingly, a positive correlation was observed between the growth fitness and the carrying capacity of the Evos (Fig. 1D). It was somehow consistent with the positive correlations between the colony growth rate and the colony size of a genome-reduced strain 11 and between the growth rates and the saturated population size of an assortment of genome reduced strains 13. Nevertheless, the negative correlation between growth rate and carrying capacity, known as the r/K selection30,31 was often observed as the trade-off relationship between r and K in the evolution and ecology studies 32 33,34. As the r/K trade-off was proposed to balance the cellular metabolism that resulted from the cost of enzymes involved 34, the deleted genes might play a role in maintaining the metabolism balance for the r/K correlation. On the other hand, the experimental evolution (i.e., serial transfer) was strictly performed within the exponential growth phase; thus, the evolutionary selection was supposed to be driven by the growth rate without selective pressure to maintain the carrying capacity. The declined carrying capacity might have been its neutral "drift" but not a trade-off to the growth rate. Independent and parallel experimental evolution of the reduced genomes selecting either r or K is required to clarify the actual mechanisms.”

• Line 103 Genome mutations. The authors claim that there are no mutations in parallel but I see that there is a 1199 base pair deletion in eight of the nine evo strains (Table S3). I would like the author to mention this and I'm actually curious about why the authors don't consider this parallel evolution.

Thank you for your careful reading. According to your comment, we added a brief description of the 1199-bp deletion detected in the Evos as follows.

(L119-122) “The number of mutations largely varied among the nine Evos, from two to 13, and no common mutation was detected in all nine Evos (Table S3). A 1,199-bp deletion of insH was frequently found in the Evos (Table S3, highlighted), which well agreed with its function as a transposable sequence.”

• Line 297 Please describe the media in full here - this is an important detail for the evolution experiment. Very frustrating to go to reference 13 and find another reference, but no details of the method. Looked online for the M63 growth media and the carbon source is not specified. This is critical for working out what selection pressures might have driven the genetic and transcriptional changes that you have measured. For example, the parallel genetic change in 8/9 populations is a deletion of insH and tdcD (according to Table S3). This is acetate kinase, essential for the final step in the overflow metabolism of glucose into acetate. If you have a very low glucose concentration, then it could be that there was selection to avoid fermentation and devote all the pyruvate that results from glycolysis into the TCA cycle (which is more efficient than fermentation in terms of ATP produced per pyruvate).

Sorry for the missing information on the medium composition, which was additionally described in the Materials and Methods. The glucose concentration in M63 was 22 mM, which was supposed to be enough for bacterial growth. Thank you for your intriguing thinking about linking the medium component to the genome mutation-mediated metabolic changes. As there was no experimental result regarding the biological function of gene mutation in the present study, please allow us to address this issue in our future work.

(L334-337) “In brief, the medium contains 62 mM dipotassium hydrogen phosphate, 39 mM potassium dihydrogen phosphate, 15 mM ammonium sulfate, 15 μM thiamine hydrochloride, 1.8 μM Iron (II) sulfate, 0.2 mM magnesium sulfate, and 22 mM glucose.”

• Line 115. I do not understand this argument "They seemed highly related to essentiality, as 11 out of 49 mutated genes were essential (Table S3)." Is this a significant enrichment compared to the expectation, i.e. the number of essential genes in the genome? This enrichment needs to be tested with a Hypergeometric test or something similar.

• Also, "As the essential genes were known to be more conserved than nonessential ones, the high frequency of the mutations fixed in the essential genes suggested the mutation in essentiality for fitness increase was the evolutionary strategy for reduced genome." I do not think that there is enough evidence to support this claim, and it should be removed.

Sorry for the unclear description. Yes, the mutations were significantly enriched in the essential genes (11 out of 45 genes) compared to the essential genes in the whole genome (286 out of 3290 genes). The improper description linking the mutation in essential genes to the fitness increase was removed, and an additional explanation on the ratio of essential genes was newly supplied as follows.

(L139-143) “The ratio of essential genes in the mutated genes was significantly higher than in the total genes (286 out of 3290 genes, Chi-square test p=0.008). As the essential genes were determined according to the growth35 and were known to be more conserved than nonessential ones 36,37, the high frequency of the mutations fixed in the essential genes was highly intriguing and reasonable.”

• Line 124 Regarding the mutation simulations, I do not understand how the observed data were compared to the simulated data, and how conclusions were drawn. Can the authors please explain the motivation for carrying out this analysis, and clearly explain the conclusions?

Random simulation was additionally explained in the Materials and Methods and the conclusion of the random simulation was revised in the Results, as follows.

(L392-401) “The mutation simulation was performed with Python in the following steps. A total of 65 mutations were randomly generated on the reduced genome, and the distances from the mutated genomic locations to the nearest genomic scars caused by genome reduction were calculated. Subsequently, Welch's t-test was performed to evaluate whether the distances calculated from the random mutations were significantly longer or shorter than those calculated from the mutations that occurred in Evos. The random simulation, distance calculation, and statistic test were performed 1,000 times, which resulted in 1,000 p values. Finally, the mean of p values (μp) was calculated, and a 95% reliable region was applied. It was used to evaluate whether the 65 mutations in the Evos were significantly close to the genomic scars, i.e., the locational bias.”

(L148-157) “Random simulation was performed to verify whether there was any bias or hotspot in the genomic location for mutation accumulation due to the genome reduction. A total of 65 mutations were randomly generated on the reduced genome (Fig. 2B), and the genomic distances from the mutations to the nearest genome reduction-mediated scars were calculated. Welch's t-test was performed to evaluate whether the genomic distances calculated from random mutations significantly differed from those from the mutations accumulated in the Evos. As the mean of p values (1,000 times of random simulations) was insignificant (Fig. 2C, μp > 0.05), the mutations fixed on the reduced genome were either closer or farther to the genomic scars, indicating there was no locational bias for mutation accumulation caused by genome reduction.”

• Line 140 The authors should give some background here - explain the idea underlying chromosomal periodicity of the transcriptome, to help the reader understand this analysis.

• Line 142 Here and elsewhere, when referring to a method, do not just give the citation, but also refer to the methods section or relevant supplementary material.

The analytical process (references and methods) was described in the Materials and Methods, and the reason we performed the chromosomal periodicity was added in the Results as follows.

(L165-172) “As the E. coli chromosome was structured, whether the genome reduction caused the changes in its architecture, which led to the differentiated transcriptome reorganization in the Evos, was investigated. The chromosomal periodicity of gene expression was analyzed to determine the structural feature of genome-wide pattern, as previously described 28,38. The analytical results showed that the transcriptomes of all Evos presented a common six-period with statistical significance, equivalent to those of the wild-type and ancestral reduced genomes (Fig. 3A, Table S4).”

• Line 151 "The expression levels of the mutated genes were higher than those of the remaining genes (Figure 3B)"- did this depend on the type of mutation? There were quite a few early stops in genes, were these also more likely to be expressed? And how about the transcriptional regulators, can you see evidence of their downstream impact?

Sorry, we didn't investigate the detailed regulatory mechanisms of 49 mutated genes, which was supposed to be out of the scope of the present study. Fig. 3B was the statistical comparison between 3225 and 49 genes. It didn't mean that all mutated genes expressed higher than the others. The following sentences were added to address your concern.

(L181-185) “As the regulatory mechanisms or the gene functions were supposed to be disturbed by the mutations, the expression levels of individual genes might have been either up- or down-regulated. Nevertheless, the overall expression levels of all mutated genes tended to be increased. One of the reasons was assumed to be the mutation essentiality, which remained to be experimentally verified.”

• Line 199 onward. The authors used WGCNA to analyze the gene expression data of evolved organisms. They identified distinct gene modules in the reduced genome, and through further analysis, they found that specific modules were strongly associated with key biological traits like growth fitness, gene expression changes, and mutation rates. Did the authors expect that there was variation in mutation rate across their populations? Is variation from 3-16 mutations that they observed beyond the expectation for the wt mutation rate? The genetic causes of mutation rate variation are well understood, but I could not see any dinB, mutT,Y, rad, or pol genes among the discovered mutations. I would like the authors to justify the claim that there was mutation rate variation in the evolved populations.

Thank you for the intriguing thinking. We don't think the mutation rates were significantly varied across the nine populations, as no mutation occurred in the MMR genes, as you noticed. Our previous study showed that the spontaneous mutation rate of the reduced genome was higher than that of the wild-type genome (Nishimura et al., 2017, mBio). As nonsynonymous mutations were not detected in all nine Evos, the spontaneous mutation rate couldn't be calculated (because it should be evaluated according to the ratio of nonsynonymous and synonymous single-nucleotide substitutions in molecular evolution). Therefore, discussing the mutation rate in the present study was unavailable. The following sentence was added for a better understanding of the gene modules.

(L242-245) “These modules M2, M10 and M16 might be considered as the hotspots for the genes responsible for growth fitness, transcriptional reorganization, and mutation accumulation of the reduced genome in evolution, respectively.”

• Line 254 I get the idea of all roads leading to Rome, which is very fitting. However, describing the various evolutionary strategies and homeostatic and variable consequence does not sound correct - although I am not sure exactly what is meant here. Looking at Figure 7, I will call strategy I "parallel evolution", that is following the same or similar genetic pathways to adaptation and strategy ii I would call divergent evolution. I am not sure what strategy iii is. I don't want the authors to use the terms parallel and divergent if that's not what they mean. My request here would be that the authors clearly describe these strategies, but then show how their results fit in with the results, and if possible, fit with the naming conventions, of evolutionary biology.

Thank you for your kind consideration and excellent suggestion. It's our pleasure to adopt your idea in tour study. The evolutionary strategies were renamed according to your recommendation. Both the main text and Fig. 7 were revised as follows.

(L285-293) “Common mutations22,44 or identical genetic functions45 were reported in the experimental evolution with different reduced genomes, commonly known as parallel evolution (Fig. 7, i). In addition, as not all mutations contribute to the evolved fitness 22,45, another strategy for varied phenotypes was known as divergent evolution (Fig. 7, ii). The present study accentuated the variety of mutations fixed during evolution. Considering the high essentiality of the mutated genes (Table S3), most or all mutations were assumed to benefit the fitness increase, partially demonstrated previously 20. Nevertheless, the evolved transcriptomes presented a homeostatic architecture, revealing the divergent to convergent evolutionary strategy (Fig. 7, iii).”

Author response image 1.

• Line 327 Growth rates/fitness. I don't think this should be called growth fitness- a rate is being calculated. I would like the authors to explain how the times were chosen - do the three points have to be during the log phase? Can you also explain what you mean by choosing three ri that have the largest mean and minor variance?

Sorry for the confusing term usage. The fitness assay was changed to the growth assay. Choosing three ri that have the largest mean and minor variance was to avoid the occasional large values (blue circle), as shown in the following figure. In addition, the details of the growth analysis can be found at https://doi.org/10.3791/56197 (ref. 59), where the video of experimental manipulation, protocol, and data analysis is deposited. The following sentence was added in accordance.

Author response image 2.

(L369-371) “The growth rate was determined as the average of three consecutive ri, showing the largest mean and minor variance to avoid the unreliable calculation caused by the occasionally occurring values. The details of the experimental and analytical processes can be found at https://doi.org/10.3791/56197.”

• Line 403 Chromosomal periodicity analysis. The windows chosen for smoothing (100kb) seem big. Large windows make sense for some things - for example looking at how transcription relates to DNA replication timing, which is a whole-genome scale trend. However, here the authors are looking for the differences after evolution, which will be local trends dependent on specific genes and transcription factors. 100kb of the genome would carry on the order of one hundred genes and might be too coarse-grained to see differences between evos lineages.

Thank you for the advice. We agree that the present analysis focused on the global trend of gene expression. Varying the sizes may lead to different patterns. Additional analysis was performed according to your comment. The results showed that changes in window size (1, 10, 50, 100, and 200 kb) didn't alter the periodicity of the reduced genome, which agreed with the previous study on a different reduced genome MDS42 of a conserved periodicity (Ying et al., 2013, BMC Genomics). The following sentence was added in the Materials and Methods.

(L460-461) “Note that altering the moving average did not change the max peak.”

• Figures - the figures look great. Figure 7 needs a legend.

Thank you. The following legend was added.

(L774-777) “Three evolutionary strategies are proposed. Pink and blue arrowed lines indicate experimental evolution and genome reduction, respectively. The size of the open cycles represents the genome size. Black and grey indicate the ancestor and evolved genomes, respectively.”

Response to Reviewer #2:

Thank you for reviewing our manuscript and for your fruitful comments. We agree that our study leaned towards elaborating observed findings rather than explaining the detailed biological mechanisms. We focused on the genome-wide biological features rather than the specific biological functions. The underlying mechanisms indeed remained unknown, leaving the questions as you commented. We didn't perform the fitness assay on reconstituted (single and combinatorial) mutants because the research purpose was not to clarify the regulatory or metabolic mechanisms. It's why the RNA-Seq analysis provided the findings on genome-wide patterns and chromosomal view, which were supposed to be biologically valuable. We did understand your comments and complaints that the conclusions were biologically meaningless, as ALE studies that found the specific gene regulation or improved pathway was the preferred story in common, which was not the flow of the present study.

For this reason, our revision may not address all these concerns. Considering your comments, we tried our best to revise the manuscript. The changes made were highlighted. We sincerely hope the revision and the following point-to-point response are acceptable.

Major remarks:

(1) The authors outlined the significance of ALE in genome-reduced organisms and important findings from published literature throughout the Introduction section. The description in L65-69, which I believe pertains to the motivation of this study, seems vague and insufficient to convey the novelty or necessity of this study i.e. it is difficult to grasp what aspects of genome-reduced biology that this manuscript intends to focus/find/address.

Sorry for the unclear writing. The sentences were rewritten for clarity as follows.

(L64-70) “Although the reduced growth rate caused by genome reduction could be recovered by experimental evolution, it remains unclear whether such an evolutionary improvement in growth fitness was a general feature of the reduced genome and how the genome-wide changes occurred to match the growth fitness increase. In the present study, we performed the experimental evolution with a reduced genome in multiple lineages and analyzed the evolutionary changes of the genome and transcriptome.”

(2) What is the rationale behind the lineage selection described in Figure S1 legend "Only one of the four overnight cultures in the exponential growth phase (OD600 = 0.01~0.1) was chosen for the following serial transfer, highlighted in red."?

The four wells (cultures of different initial cell concentrations) were measured every day, and only the well that showed OD600=0.01~0.1 (red) was transferred with four different dilution rates (e.g., 10, 100, 1000, and 10000 dilution rates). It resulted in four wells of different initial cell concentrations. Multiple dilutions promised that at least one of the wells would show the OD600 within the range of 0.01 to 0.1 after the overnight culture. They were then used for the next serial transfer. Fig. S1 provides the details of the experimental records. The experimental evolution was strictly controlled within the exponential phase, quite different from the commonly conducted ALE that transferred a single culture in a fixed dilution rate. Serial transfer with multiple dilution rates was previously applied in our evolution experiments and well described in Nishimura et al., 2017, mBio; Lu et al., 2022, Comm Biol; Kurokawa et al., 2022, Front Microbiol, etc. The following sentence was added in the Materials and Methods.

(L344-345) “Multiple dilutions changing in order promised at least one of the wells within the exponential growth phase after the overnight culture.”

(3) The measured growth rate of the end-point 'F2 lineage' shown in Figure S2 seemed comparable to the rest of the lineages (A1 to H2), but the growth rate of 'F2' illustrated in Figure 1B indicates otherwise (L83-84). What is the reason for the incongruence between the two datasets?

Sorry for the unclear description. The growth rates shown in Fig. S2 were obtained during the evolution experiment using the daily transfer's initial and final OD600 values. The growth rates shown in Fig. 1B were obtained from the final population (Evos) growth assay and calculated from the growth curves (biological replication, N=4). Fig. 1B shows the precisely evaluated growth rates, and Fig. S2 shows the evolutionary changes in growth rates. Accordingly, the following sentence was added to the Results.

(L84-87) “As the growth increases were calculated according to the initial and final records, the exponential growth rates of the ancestor and evolved populations were obtained according to the growth curves for a precise evaluation of the evolutionary changes in growth.”

(4) Are the differences in growth rate statistically significant in Figure 1B?

Eight out of nine Evos were significant, except F2. The sentences were rewritten and associated with the revised Fig. 1B, indicating significance.

(L87-90) “The results demonstrated that most evolved populations (Evos) showed improved growth rates, in which eight out of nine Evos were highly significant (Fig. 1B, upper). However, the magnitudes of growth improvement were considerably varied, and the evolutionary dynamics of the nine lineages were somehow divergent (Fig. S2).”

(5) The evolved lineages showed a decrease in their maximal optical densities (OD600) compared to the ancestral strain (L85-86). ALE could accompany changes in cell size and morphologies, (doi: 10.1038/s41586-023-06288-x; 10.1128/AEM.01120-17), which may render OD600 relatively inaccurate for cell density comparison. I suggest using CFU/mL metrics for the sake of a fair comparison between Anc and Evo.

The methods evaluating the carrying capacity (i.e., cell density, population size, etc.) do not change the results. Even using CFU is unfair for the living cells that can not form colonies and unfair if the cell size changes. Optical density (OD600) provides us with the temporal changes of cell growth in a 15-minute interval, which results in an exact evaluation of the growth rate in the exponential phase. CFU is poor at recording the temporal changes of population changes, which tend to result in an inappropriate growth rate. Taken together, we believe that our method was reasonable and reliable. We hope you can accept the different way of study.

(6) Please provide evidence in support of the statement in L115-119. i.e. statistical analysis supporting that the observed ratio of essential genes in the mutant pool is not random.

The statistic test was performed, and the following sentence was added.

(L139-141) “The ratio of essential genes in the mutated genes was significantly higher than in the total genes (286 out of 3290 genes, Chi-square test p=0.008).”

(7) The assumption that "mutation abundance would correlate to fitness improvement" described in L120-122: "The large variety in genome mutations and no correlation of mutation abundance to fitness improvement strongly suggested that no mutations were specifically responsible or crucially essential for recovering the growth rate of the reduced genome" is not easy to digest, in the sense that (i) the effect of multiple beneficial mutations are not necessarily summative, but are riddled with various epistatic interactions (doi: 10.1016/j.mec.2023.e00227); (ii) neutral hitchhikers are of common presence (you could easily find reference on this one); (iii) hypermutators that accumulate greater number of mutations in a given time are not always the eventual winners in competition games (doi: 10.1126/science.1056421). In this sense, the notion that "mutation abundance correlates to fitness improvement" in L120-122 seems flawed (for your perusal, doi: 10.1186/gb-2009-10-10-r118).

Sorry for the improper description and confusing writing, and thank you for the fruitful knowledge on molecular evolution. The sentence was deleted, and the following one was added.

(L145-146) “Nevertheless, it was unclear whether and how these mutations were explicitly responsible for recovering the growth rate of the reduced genome.”

(8) Could it be possible that the large variation in genome mutations in independent lineages results from a highly rugged fitness landscape characterized by multiple fitness optima (doi: 10.1073/pnas.1507916112)? If this is the case, I disagree with the notion in L121-122 "that no mutations were specifically responsible or crucially essential" It does seem to me that, for example, the mutations in evo A2 are specifically responsible and essential for the fitness improvement of evo A2 in the evolutionary condition (M63 medium). Fitness assessment of individual (or combinatorial) mutants reconstituted in the Ancestral background would be a bonus.

Thank you for the intriguing thinking. The sentence was deleted. Please allow us to adapt your comment to the manuscript as follows.

(L143-145) “The large variety of genome mutations fixed in the independent lineages might result from a highly rugged fitness landscape 38.”

(9) L121-122: "...no mutations were specifically responsible or crucially essential for recovering the growth rate of the reduced genome". Strictly speaking, the authors should provide a reference case of wild-type E. coli ALE in order to reach definitive conclusions that the observed mutation events are exclusive to the genome-reduced strain. It is strongly recommended that the authors perform comparative analysis with an ALEed non-genome-reduced control for a more definitive characterization of the evolutionary biology in a genome-reduced organism, as it was done for "JCVI-syn3.0B vs non-minimal M. mycoides" (doi: 10.1038/s41586-023-06288-x) and "E. coli eMS57 vs MG1655" (doi: 10.1038/s41467-019-08888-6).

The improper description was deleted in response to comments 7 and 8. The mentioned references were cited in the manuscript (refs 21 and 23). Thank you for the experimental advice. We are sorry that the comparison of wild-type and reduced genomes was not in the scope of the present study and will probably be reported soon in our future work.

(10) L146-148: "The homeostatic periodicity was consistent with our previous findings that the chromosomal periodicity of the transcriptome was independent of genomic or environmental variation" A Previous study also suggested that the amplitudes of the periodic transcriptomes were significantly correlated with the growth rates (doi: 10.1093/dnares/dsaa018). Growth rates of 8/9 Evos were higher compared to Anc, while that of Evo F2 remained similar. Please comment on the changes in amplitudes of the periodic transcriptomes between Anc and each Evo.

Thank you for the suggestion. The correlation between the growth rates and the amplitudes of chromosomal periodicity was statistically insignificant (p>0.05). It might be a result of the limited data points. Compared with the only nine data points in the present study, the previous study analyzed hundreds of transcriptomes associated with the corresponding growth rates, which are suitable for statistical evaluation. In addition, the changes in growth rates were more significant in the previous study than in the present study, which might influence the significance. It's why we did not discuss the periodic amplitude.

(11) Please elaborate on L159-161: "It strongly suggested the essentiality mutation for homeostatic transcriptome architecture happened in the reduced genome.".

Sorry for the improper description. The sentence was rewritten as follows.

(L191-193) “The essentiality of the mutations might have participated in maintaining the homeostatic transcriptome architecture of the reduced genome.”

(12) Is FPKM a valid metric for between-sample comparison? The growing consensus in the community adopts Transcripts Per Kilobase Million (TPM) for comparing gene expression levels between different samples (Figure 3B; L372-379).

Sorry for the unclear description. The FPKM indicated here was globally normalized, statistically equivalent to TPM. The following sentence was added to the Materials and Methods.

(L421-422) “The resulting normalized FPKM values were statistically equivalent to TPM.”

(13) Please provide % mapped frequency of mutations in Table S3.

They were all 100%. The partially fixed mutations were excluded in the present study. The following sentence was added to the caption of Table S3.

(Supplementary file, p 9) “Note that the entire population held the mutations, i.e., 100% frequency in DNA sequencing.”

(14) To my knowledge, M63 medium contains glucose and glycerol as carbon sources. The manuscript would benefit from discussing the elements that impose selection pressure in the M63 culture condition.

Sorry for the missing information on M63, which contains 22 mM glucose as the only carbon source. The medium composition was added in the Materials and Methods, as follows.

(L334-337) “In brief, the medium contains 62 mM dipotassium hydrogen phosphate, 39 mM potassium dihydrogen phosphate, 15 mM ammonium sulfate, 15 μM thiamine hydrochloride, 1.8 μM Iron (II) sulfate, 0.2 mM magnesium sulfate, and 22 mM glucose.”

(15) The RNA-Seq datasets for Evo strains seemed equally heterogenous, just as their mutation profiles. However, the missing element in their analysis is the directionality of gene expression changes. I wonder what sort of biological significance can be derived from grouping expression changes based solely on DEGs, without considering the magnitude and the direction (up- and down-regulation) of changes? RNA-seq analysis in its current form seems superficial to derive biologically meaningful interpretations.

We agree that most studies often discuss the direction of transcriptional changes. The present study aimed to capture a global view of the magnitude of transcriptome reorganization. Thus, the analyses focused on the overall features, such as the abundance of DEGs, instead of the details of the changes, e.g., the up- and down-regulation of DEGs. The biological meaning of the DEGs' overview was how significantly the genome-wide gene expression fluctuated, which might be short of an in-depth view of individual gene expression. The following sentence was added to indicate the limitation of the present analysis.

(L199-202) “Instead of an in-depth survey on the directional changes of the DEGs, the abundance and functional enrichment of DEGs were investigated to achieve an overview of how significant the genome-wide fluctuation in gene expression, which ignored the details of individual genes.”

Minor remarks

(1) L41: brackets italicized "(E. coli)".

It was fixed as follows.

(L40) “… Escherichia coli (E. coli) cells …”

(2) Figure S1. It is suggested that the x-axis of ALE monitor be set to 'generations' or 'cumulative generations', rather than 'days'.

Thank you for the suggestion. Fig. S1 describes the experimental procedure, so the" day" was used. Fig. S2 presents the evolutionary process, so the "generation" was used, as you recommended here.

(3) I found it difficult to digest through L61-64. Although it is not within the job scope of reviewers to comment on the language style, I must point out that the manuscript would benefit from professional language editing services.

Sorry for the unclear writing. The sentences were revised as follows.

(L60-64) “Previous studies have identified conserved features in transcriptome reorganization, despite significant disruption to gene expression patterns resulting from either genome reduction or experimental evolution 27-29. The findings indicated that experimental evolution might reinstate growth rates that have been disrupted by genome reduction to maintain homeostasis in growing cells.”

(4) Duplicate references (No. 21, 42).

Sorry for the mistake. It was fixed (leaving ref. 21).

(5) Inconsistency in L105-106: "from two to 13".

"From two to 13" was adopted from the language editing. It was changed as follows.

(L119) “… from 2 to 13, …”

Response to Reviewer #3:

Thank you for reviewing our manuscript and for the helpful comments, which improved the strength of the manuscript. The recommended statistical analyses essentially supported the statement in the manuscript were performed, and those supposed to be the new results in the scope of further studies remained unconducted. The changes made in the revision were highlighted. We sincerely hope the revised manuscript and the following point-to-point response meet your concerns. You will find all your suggested statistic tests in our future work that report an extensive study on the experimental evolution of an assortment of reduced genomes.

(1) Line 106 - "As 36 out of 45 SNPs were nonsynonymous, the mutated genes might benefit the fitness increase." This argument can be strengthened. For example, the null expectation of nonsynonymous SNPs should be discussed. Is the number of observed nonsynonymous SNPs significantly higher than the expected one?

(2) Line 107 - "In addition, the abundance of mutations was unlikely to be related to the magnitude of fitness increase." Instead of just listing examples, a regression analysis can be added.

Yes, it's significant. Random mutations lead to ~33% of nonsynonymous SNP in a rough estimation. Additionally, the regression is unreliable because there's no statistical significance between the number of mutations and the magnitude of fitness increase. Accordingly, the corresponding sentences were revised with additional statistical tests.

(L123-129) “As 36 out of 45 SNPs were nonsynonymous, which was highly significant compared to random mutations (p < 0.01), the mutated genes might benefit fitness increase. In addition, the abundance of mutations was unlikely to be related to the magnitude of fitness increase. There was no significant correlation between the number of mutations and the growth rate in a statistical view (p > 0.1). Even from an individual close-up viewpoint, the abundance of mutations poorly explained the fitness increase.”

(3) Line 114 - "They seemed highly related to essentiality, as 11 out of 49 mutated genes were essential (Table S3)." Here, the information mentioned in line 153 ("the ratio of essential to all genes (302 out of 3,290) in the reduced genome.") can be used. Then a statistical test for a contingency table can be used.

(4) Line 117 - "the high frequency of the mutations fixed in the essential genes suggested the mutation in essentiality for fitness increase was the evolutionary strategy for reduced genome." What is the expected number of fixed mutations in essential genes vs non-essential genes? Is the observed number statistically significantly higher?

Sorry for the improper and insufficient information on the essential genes. Yes, it's significant. The statistical test was additionally performed. The corresponding part was revised as follows.

(L134-146) “They seemed highly related to essentiality7 (https://shigen.nig.ac.jp/ecoli/pec/genes.jsp), as 11 out of 49 mutated genes were essential (Table S3). Although the essentiality of genes might differ between the wild-type and reduced genomes, the experimentally determined 302 essential genes in the wild-type E. coli strain were used for the analysis, of which 286 were annotated in the reduced genome. The ratio of essential genes in the mutated genes was significantly higher than in the total genes (286 out of 3290 genes, Chi-square test p=0.008). As the essential genes were determined according to the growth35 and were known to be more conserved than nonessential ones 36,37, the high frequency of the mutations fixed in the essential genes was highly intriguing and reasonable. The large variety of genome mutations fixed in the independent lineages might result from a highly rugged fitness landscape 38. Nevertheless, it was unclear whether and how these mutations were explicitly responsible for recovering the growth rate of the reduced genome.”

(5) The authors mentioned no overlapping in the single mutation level. Is that statistically significant? The authors can bring up what the no-overlap probability is given that there are in total x number of fixed mutations observed (either theory or simulation is good).

Sorry, we feel confused about this comment. It's unclear to us why it needs to be statistically simulated. Firstly, the mutations were experimentally observed. The result that no overlapped mutated genes were detected was an Experimental Fact but not a Computational Prediction. We feel sorry that you may over-interpret our finding as an evolutionary rule, which always requires testing its reliability statistically. We didn't conclude that the evolution had no overlapped mutations. Secondly, considering 65 times random mutations happened to a ~3.9 Mb sequence, the statistical test was meaningful only if the experimental results found the overlapped mutations. It is interesting how often the random mutations cause the overlapped mutations in parallel evolutionary lineages while increasing the evolutionary lineages, which seems to be out of the scope of the present study. We are happy to include the analysis in our ongoing study on the experimental evolution of reduced genomes.

(6) The authors mentioned no overlapping in the single mutation level. How about at the genetic level? Some fixed mutations occur in the same coding gene. Is there any gene with a significantly enriched number of mutations?

No mutations were fixed in the same gene of biological function, as shown in Table S3. If we say the coding region, the only exception is the IS sequences, well known as the transposable sequences without genetic function. The following description was added.

(L119-122) “The number of mutations largely varied among the nine Evos, from 2 to 13, and no common mutation was detected in all nine Evos (Table S3). A 1,199-bp deletion of insH was frequently found in the Evos (Table S3, highlighted), which well agreed with its function as a transposable sequence.”

(7) Line 151-156- It seems like the authors argue that the expression level differences can be just explained by the percentage of essential genes that get fixed mutations. One further step for the argument could be to compare the expression level of essential genes with vs without fixed mutations. Also, the authors can compare the expression level of non-essential genes with vs without fixed mutations. And the authors can report whether the differences in expression level became insignificant after the control of the essentiality.

It's our pleasure that the essentiality intrigued you. Thank you for the analytical suggestion, which is exciting and valuable for our studies. As only 11 essential genes were detected here and "Mutation in essentiality" was an indication but not the conclusion of the present study, we would like to apply the recommended analysis to the datasets of our ongoing study to demonstrate this statement. Thank you again for your fruitful analytical advice.

(8) Line 169- "The number of DEGs partially overlapped among the Evos declined significantly along with the increased lineages of Evos (Figure 4B). " There is a lack of statistical significance here while the word "significantly" is used. One statistical test that can be done is to use re-sampling/simulation to generate a null expectation of the overlapping numbers given the DEGs for each Evo line and the total number of genes in the genome. The observed number can then be compared to the distribution of the simulated numbers.

Sorry for the inappropriate usage of the term. Whether it's statistically significant didn't matter here. The word "significant" was deleted as follows.

(L205--206) “The number of DEGs partially overlapped among the Evos declined along with the increased lineages of Evos (Fig. 4B).”

(9) Line 177-179- "In comparison,1,226 DEGs were induced by genome reduction. The common DEGs 177 of genome reduction and evolution varied from 168 to 540, fewer than half of the DEGs 178 responsible for genome reduction in all Evos" Is the overlapping number significantly lower than the expectation? The hypergeometric test can be used for testing the overlap between two gene sets.

There's no expectation for how many DEGs were reasonable. Not all numbers experimentally obtained are required to be statistically meaningful, which is commonly essential in computational and data science.

(10) The authors should give more information about the ancestral line used at the beginning of experimental evolution. I guess it is one of the KHK collection lines, but I can not find more details. There are many genome-reduced lines. Why is this certain one picked?

Sorry for the insufficient information on the reduced genome used for the experimental evolution. The following descriptions were added in the Results and the Materials and Methods, respectively.

(L75-79) “The E. coli strain carrying a reduced genome, derived from the wild-type genome W3110, showed a significant decline in its growth rate in the minimal medium compared to the wild-type strain 13. To improve the genome reduction-mediated decreased growth rate, the serial transfer of the genome-reduced strain was performed with multiple dilution rates to keep the bacterial growth within the exponential phase (Fig. S1), as described 17,20.”

(L331-334) “The reduced genome has been constructed by multiple deletions of large genomic fragments 58, which led to an approximately 21% smaller size than its parent wild-type genome W3110.”

(11) How was the saturated density in Figure 1 actually determined? In particular, the fitness assay of growth curves is 48h. But it seems like the experimental evolution is done for ~24 h cycles. If the Evos never experienced a situation like a stationary phase between 24-48h, and if the author reported the saturated density 48 h in Figure 1, the explanation of the lower saturated density can be just relaxation from selection and may have nothing to do with the increase of growth rate.

Sorry for the unclear description. Yes, you are right. The evolution was performed within the exponential growth phase (keeping cell division constant), which means the Evos never experienced the stationary phase (saturation). The final evolved populations were subjected to the growth assay to obtain the entire growth curves for calculating the growth rate and the saturated density. Whether the decreased saturated density and the increased growth rate were in a trade-off relationship remained unclear. The corresponding paragraph was revised as follows.

(L100-115) “Intriguingly, a positive correlation was observed between the growth fitness and the carrying capacity of the Evos (Fig. 1D). It was somehow consistent with the positive correlations between the colony growth rate and the colony size of a genome-reduced strain 11 and between the growth rates and the saturated population size of an assortment of genome reduced strains 13. Nevertheless, the negative correlation between growth rate and carrying capacity, known as the r/K selection30,31 was often observed as the trade-off relationship between r and K in the evolution and ecology studies 32 33,34. As the r/K trade-off was proposed to balance the cellular metabolism that resulted from the cost of enzymes involved 34, the deleted genes might play a role in maintaining the metabolism balance for the r/K correlation. On the other hand, the experimental evolution (i.e., serial transfer) was strictly performed within the exponential growth phase; thus, the evolutionary selection was supposed to be driven by the growth rate without selective pressure to maintain the carrying capacity. The declined carrying capacity might have been its neutral "drift" but not a trade-off to the growth rate. Independent and parallel experimental evolution of the reduced genomes selecting either r or K is required to clarify the actual mechanisms.”

(12) What annotation of essentiality was used in this paper? In particular, the essentiality can be different in the reduced genome background compared to the WT background.

Sorry for the unclear definition of the essential genes. They are strictly limited to the 302 essential genes experimentally determined in the wild-type E coli strain. Detailed information can be found at the following website: https://shigen.nig.ac.jp/ecoli/pec/genes.jsp. We agree that the essentiality could differ between the WT and reduced genomes. Identifying the essential genes in the reduced genome will be an exhaustedly vast work. The information on the essential genes defined in the present study was added as follows.

(L134-139) “They seemed highly related to essentiality7 (https://shigen.nig.ac.jp/ecoli/pec/genes.jsp), as 11 out of 49 mutated genes were essential (Table S3). Although the essentiality of genes might differ between the wild-type and reduced genomes, the experimentally determined 302 essential genes in the wild-type E. coli strain were used for the analysis, of which 286 were annotated in the reduced genome.”

(13) The fixed mutations in essential genes are probably not rarely observed in experimental evolution. For example, fixed mutations related to RNA polymerase can be frequently seen when evolving to stressful environments. I think the author can discuss this more and elaborate more on whether they think these mutations in essential genes are important in adaptation or not.

Thank you for your careful reading and the suggestion. As you mentioned, we noticed that the mutations in RNA polymerases (rpoA, rpoB, and rpoD) were identified in three Evos. As they were not shared across all Evos, we didn't discuss the contribution of these mutations to evolution. Instead of the individual functions of the mutated essential gene functions, we focused on the enriched gene functions related to the transcriptome reorganization because they were the common feature observed across all Evos and linked to the whole metabolic or regulatory pathways, which are supposed to be more biologically reasonable and interpretable. The following sentence was added to clarify our thinking.

(L268-273) “In particular, mutations in the essential genes, such as RNA polymerases (rpoA, rpoB, rpoD) identified in three Evos (Table S3), were supposed to participate in the global regulation for improved growth. Nevertheless, the considerable variation in the fixed mutations without overlaps among the nine Evos (Table 1) implied no common mutagenetic strategy for the evolutionary improvement of growth fitness.”

(14) In experimental evolution to new environments, several previous literature also show that long-term experimental evolution in transcriptome is not consistent or even reverts the short-term response; short-term responses were just rather considered as an emergency plan. They seem to echo what the authors found in this manuscript. I think the author can refer to some of those studies more and make a more throughput discussion on short-term vs long-term responses in evolution.

Thank you for the advice. It's unclear to us what the short-term and long-term responses referred to mentioned in this comment. The "Response" is usually used as the phenotypic or transcriptional changes within a few hours after environmental fluctuation, generally non-genetic (no mutation). In comparison, long-term or short-term experimental "Evolution" is associated with genetic changes (mutations). Concerning the Evolution (not the Response), the long-term experimental evolution (>10,000 generations) was performed only with the wild-type genome, and the short-term experimental evolution (500~2,000 generations) was more often conducted with both wild-type and reduced genomes, to our knowledge. Previous landmark studies have intensively discussed comparing the wild-type and reduced genomes. Our study was restricted to the reduced genome, which was constructed differently from those reduced genomes used in the reported studies. The experimental evolution of the reduced genomes has been performed in the presence of additional additives, e.g., antibiotics, alternative carbon sources, etc. That is, neither the genomic backgrounds nor the evolutionary conditions were comparable. Comparison of nothing common seems to be unproductive. We sincerely hope the recommended topics can be applied in our future work.

Some minor suggestions

• Figures S3 & Table S2 need an explanation of the abbreviations of gene categories.

Sorry for the missing information. Figure S3 and Table S3 were revised to include the names of gene categories. The figure was pasted followingly for a quick reference.

Author response image 3.

• I hope the authors can re-consider the title; "Diversity for commonality" does not make much sense to me. For example, it can be simply just "Diversity and commonality."

Thank you for the suggestion. The title was simplified as follows.

(L1) “Experimental evolution for the recovery of growth loss due to genome reduction.”

• It is not easy for me to locate and distinguish the RNA-seq vs DNA-seq files in DRA013662 at DDBJ. Could you make some notes on what RNA-seq actually are, vs what DNA-seq files actually are?

Sorry for the mistakes in the DRA number of DNA-seq. DNA-seq and RNA-seq were deposited separately with the accession IDs of DRA013661 and DRA013662, respectively. The following correction was made in the revision.

(L382-383) “The raw datasets of DNA-seq were deposited in the DDBJ Sequence Read Archive under the accession number DRA013661.”

Author response:

eLife assessment

In this valuable study, Kumar et al., provide evidence suggesting that the p130Cas drives the formation of condensates that sprout from focal adhesions to cytoplasm and suppress translation. Pending further substantiation, this study was found to be likely to provide previously unappreciated insights into the mechanisms linking focal adhesions to the regulation of protein synthesis and was thus considered to be of broad general interest. However, the evidence supporting the proposed model was incomplete; additional evidence is warranted to substantiate the relationship between p130Cas condensates and mRNA translation and establish corresponding functional consequences.

We thank the Elife editorial team for their positive assessment of the broad significance of our manuscript. We fully agree that the functional consequences need to be explored in more detail. We feel that many of the criticisms are valid points that are not easily addressed via available tools, thus, should be considered limitations of present approaches. We hope that readers appreciate that identification of a new class of liquid-liquid phase separations calls for much more work to fully explore their characteristics, regulation and function, which will likely advance many areas of cell biology and perhaps even medicine.

Reviewer #1 (Public Review):

Summary:

The authors demonstrated the phenomenon of p130Cas, a protein primarily localized at focal adhesions, and its formation of condensates. They identified the constituents within the condensates, which include other focal adhesion proteins, paxillin, and RNAs. Furthermore, they proposed a link between p130Cas condensates and translation.

Strengths:

Adhesion components undergo rapid exchange with the cytoplasm for some unclear biological functions. Given that p130Cas is recognized as a prominent mechanical focal adhesion component, investigating its role in condensate formation, particularly its impact on the translation process, is intriguing and significant.

We thank the reviewer for recognizing the functional significance of the work.

Weaknesses:

The authors identified the disordered region of p130Cas and investigated the formation of p130Cas condensate. They attempted to demonstrate that p130Cas condensates inhibit translation, but the results did not fully support this assertion. There are several comments below:

(1) Despite isolating p130Cas-GFP protein using GFP-trap beads, the authors cannot conclusively eliminate the possibility of isolating p130Cas from focal adhesions. While the characterization of the GFP-tagged pulls can reveal the proteins and RNAs associated with p130Cas, they need to clarify their intramolecular mechanism of localization within p130Cas droplets. Whether the protein condensates retain their liquid phase or these GFP-p130Cas pulls represent protein aggregate remains uncertain.

We agree, the isolation from cell lysates does not distinguish between focal adhesions and cytoplasmic LLPS. We note that p130Cas in focal adhesions also appears to be in LLPS. But there are no methods available to isolate them separately. We acknowledge this is a limitation of the study.

(2) The authors utilized hexanediol and ammonium acetate to highlight the phenomenon of p130Cas condensates. Although hexanediol is an inhibitor for hydrophobic interactions and ammonium acetate is a salt, a more thorough explanation of the intramolecular mechanisms underlying p130Cas protein-protein interaction is required. Additionally, given that the size of p130Cas condensates can exceed >100um2, classification is needed to differentiate between p130Cas condensates and protein aggregation.

Ammonium acetate, which works by promoting hydrophobic interactions and weak Van der Waals forces, has been widely used in phase separation studies to change ionic strength without altering intracellular pH. Conversely, hexanediol weakens hydrophobic/ Van der Walls interactions that commonly mediate phase separation of IDRs. In the case of p130Cas, the multiple tyrosines and within the scaffolding domain are obvious targets. If the reviewer is asking us to resolve the detailed hydrophobic interactions within the scaffolding domain, this is far beyond the scope of the current paper.

Protein aggregates are defined by their characteristics (e.g irreversibility, departure from spherical) not by size. Older, larger droplets remain circular and show slower but still measurable rates of exchange. Moreover, droplets are essentially absent after trypsinizing and replating cells. All these results argue against aggregates.

(3) The connection between p130Cas condensates and translation inhibition appears tenuous. The data only suggests a correlation between p130Cas expression and translation inhibition. Further evidence is required to bolster this hypothesis.

The optogenetic experiment shows that triggering LLPS by dimerizing p130Cas results in inhibition of translation. This is a causal not a correlative experiment. The reviewer may be thinking that dimerizing p130Cas could stimulate focal adhesion signaling, activating FAK or a src family kinase or other signals. However, none of these signals has been linked to inhibition of cell growth or migration. Thus, we agree that this is a limitation but consider it a low probability mechanism.

Reviewer #2 (Public Review):

Summary:

In this article, Kumar et al., report on a previously unappreciated mechanism of translational regulation whereby p130Cas induces LLPS condensates that then traffic out from focal adhesion into the cytoplasm to modulate mRNA translation. Specifically, the authors employed EGFP-tagged p130Cas constructs, endogenous p130Cas, and p130Cas knockouts and mutants in cell-based systems. These experiments in conjunction with various imaging techniques revealed that p130Cas drives assembly of LLPS condensates in a manner that is largely independent of tyrosine phosphorylation. This was followed by in vitro EGFP-tagged p130Cas-dependent induction of LLPS condensates and determination of their composition by mass spectrometry, which revealed enrichment of proteins involved in RNA metabolism in the condensates. The authors excluded the plausibility that p130Cas-containing condensates co-localize with stress granules or p-bodies. Next, the authors determined mRNA compendium of p130Cas-containing condensates which revealed that they are enriched in transcripts encoding proteins implicated in cell cycle progression, survival, and cell-cell communication. These findings were followed by the authors demonstrating that p130Cas-containing condensates may be implicated in the suppression of protein synthesis using puromycylation assay. Altogether, it was found that this study significantly advances the knowledge pertinent to the understanding of molecular underpinnings of the role of p130Cas and more broadly focal adhesions on cellular function, and to this end, it is likely that this report will be of interest to a broad range of scientists from a wide spectrum of biomedical disciplines including cell, molecular, developmental and cancer biologists.

Strengths:

Altogether, this study was found to be of potentially broad interest inasmuch as it delineates a hitherto unappreciated link between p130Cas, LLPS, and regulation of mRNA translation. More broadly, this report provides unique molecular insights into the previously unappreciated mechanisms of the role of focal adhesions in regulating protein synthesis. Overall, it was thought that the provided data sufficiently supported most of the authors' conclusions. It was also thought that this study incorporates an appropriate balance of imaging, cell and molecular biology, and biochemical techniques, whereby the methodology was found to be largely appropriate.

We thank reviewer for this positive assessment.

Weaknesses:

Two major weaknesses of the study were noted. The first issue is related to the experiments establishing the role of p130Cas-driven condensates in translational suppression, whereby it remained unclear whether these effects are affecting global mRNA translation or are specific to the mRNAs contained in the condensates. Moreover, some of the results in this section (e.g., experiments using cycloheximide) may be open to alternative interpretation. The second issue is the apparent lack of functional studies, and although the authors speculate that the described mechanism is likely to mediate the effects of focal adhesions on e.g., quiescence, experimental testing of this tenet was lacking.

We appreciate the reviewer’s insights. Assessing translational inhibition for specific genes rather than global measurement of translation is an important direction for future work.

Regarding the cycloheximide experiments, we are unsure what the reviewer means. We used it as a control for puromycin labeling but this is a very standard approach. It seems more likely that the question concerns Fig 5G, where we used it to sequester mRNAs on ribosomes to deplete from other pools. In this case, p130cas condensates decrease after 2 minutes. The reviewer may be suggesting that this effect could be due to blocked translation per se and loss of short-lived proteins. We acknowledge that this is possible but given the very rapid effect (2 min), we think it unlikely.

Lastly, we agree with the reviewer that further functional studies in quiescence or senescence are warranted; however, these are extensive, open-ended studies and we will not be able to include them as part of the current paper.

Author response:

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

eLife assessment

In this valuable study, the authors investigate the transcriptional landscape of tuberculous meningitis, revealing important molecular differences contributed by HIV co-infection. Whilst some of the evidence presented is compelling, the bioinformatics analysis is limited to a descriptive narrative of gene-level functional annotations, which are somewhat basic and fail to define aspects of biology very precisely. Whilst the work will be of broad interest to the infectious disease community, validation of the data is critical for future utility.

We appreciate with eLife’s positive assessment, although we challenge the conclusion that we ‘fail to define aspects of biology very precisely’. Our stated objective was to use bioinformatics tools to identify the biological pathways and hub genes associated with TBM pathogenesis and the eLife assessment affirms we have investigated ‘the transcriptional landscape of tuberculous meningitis’. To more precisely define aspects of the biology will require another study with different design and methods.

Reviewer #1 (Public Review):

Summary:

Tuberculous meningitis (TBM) is one of the most severe forms of extrapulmonary TB. TBM is especially prevalent in people who are immunocompromised (e.g. HIV-positive). Delays in diagnosis and treatment could lead to severe disease or mortality. In this study, the authors performed the largest-ever host whole blood transcriptomics analysis on a cohort of 606 Vietnamese participants. The results indicated that TBM mortality is associated with increased neutrophil activation and decreased T and B cell activation pathways. Furthermore, increased angiogenesis was also observed in HIV-positive patients who died from TBM, whereas activated TNF signaling and down-regulated extracellular matrix organisation were seen in the HIV-negative group. Despite similarities in transcriptional profiles between PTB and TBM compared to healthy controls, inflammatory genes were more active in HIV-positive TBM. Finally, 4 hub genes (MCEMP1, NELL2, ZNF354C, and CD4) were identified as strong predictors of death from TBM.

Strengths:

This is a really impressive piece of work, both in terms of the size of the cohort which took years of effort to recruit, sample, and analyse, and also the meticulous bioinformatics performed. The biggest advantage of obtaining a whole blood signature is that it allows an easier translational development into a test that can be used in the clinical with a minimally invasive sample. Furthermore, the data from this study has also revealed important insights into the mechanisms associated with mortality and the differences in pathogenesis between HIV-positive and HIV-negative patients, which would have diagnostic and therapeutic implications.

Weaknesses:

The data on blood neutrophil count is really intriguing and seems to provide a very powerful yet easy-to-measure method to differentiate survival vs. death in TBM patients. It would be quite useful in this case to perform predictive analysis to see if neutrophil count alone, or in combination with gene signature, can predict (or better predict) mortality, as it would be far easier for clinical implementation than the RNA-based method. Moreover, genes associated with increased neutrophil activation and decreased T cell activation both have significantly higher enrichment scores in TBM (Figure 9) and in morality (Figure 8). While I understand the basis of selecting hub genes in the significant modules, they often do not represent these biological pathways (at least not directly associated in most cases). If genes were selected based on these biologically relevant pathways, would they have better predictive values?

We conducted a sensitivity analysis including blood neutrophil as a potential predictor in the multivariate Cox elastic-net regression model for important predictor selection (Table S14). In this analysis, all six selected important predictors (genes and clinical risk factors) identified in the original analysis (Table S13) were also selected, together with blood neutrophil number. Additionally, we evaluated the predictive value of blood neutrophil alone, which demonstrated poor performance, with an optimism-corrected AUC of 0.63 for all TBM, 0.67 for HIV-negative TBM, and 0.70 for HIV-positive TBM. Even when combined with identified gene signatures, blood neutrophil did not improve the overall performance of predictive model (optimism-corrected AUC of 0.79 for all TBM, 0.76 for HIV-negative TBM, and 0.80 for HIV-positive). These results indicate that identified hub genes exhibit better predictive values compared to blood neutrophil alone or in combination. These findings have been incorporated into our manuscript results.

To test whether pathway representative genes have better predictive values than hub genes, we included all these genes in the analysis for important predictor selection. Pathway representative genes comprised ANXA3 and CXCR2 representing neutrophil activation and IL1b representing acute inflammatory response. We observed that all hub genes (MCEMP1, NELL2, ZNF354C, and CD4) consistently emerged as the most important genes with the highest selection in the models, compared to the rest, in both the HIV-negative TBM and HIV-positive TBM cohorts. Additionally, these identified hub genes were still selected when testing together with other hub genes representing relevant biological pathways associated with TBM mortality, such as CYSTM1 involved in neutrophil activation, TRAF5 involved in NF-kappa B signaling pathway, CD28 and TESPA1 involved in T cell receptor signaling. These results show that selected genes based on known biologically relevant pathways did not give better predictive values than the identified hub genes in the significant modules.

Reviewer #2 (Public Review):

Summary:

This manuscript describes the analysis of blood transcriptomic data from patients with TB meningitis, with and without HIV infection, with some comparison to those of patients with pulmonary tuberculosis and healthy volunteers. The objectives were to describe the comparative biological differences represented by the blood transcriptome in TBM associated with HIV co-infection or survival/mortality outcomes and to identify a blood transcriptional signature to predict these outcomes. The authors report an association between mortality and increased levels of acute inflammation and neutrophil activation, but decreased levels of adaptive immunity and T/B cell activation. They propose a 4-gene prognostic signature to predict mortality.

Strengths:

Biological evaluations of blood transcriptomes in TB meningitis and their relationship to outcomes have not been extensively reported previously.

The size of the data set is a major strength and is likely to be used extensively for secondary analyses in this field of research.

Weaknesses:

The bioinformatic analysis is limited to a descriptive narrative of gene-level functional annotations curated in GO and KEGG databases. This analysis cannot be used to make causal inferences. In addition, the functional annotations are limited to 'high-level' terms that fail to define biology very precisely. At best, they require independent validation for a given context. As a result, the conclusions are not adequately substantiated. The identification of a prognostic blood transcriptomic signature uses an unusual discovery approach that leverages weighted gene network analysis that underpins the bioinformatic analyses. However, the main problem is that authors seem to use all the data for discovery and do not undertake any true external validation of their gene signature. As a result, the proposed gene signature is likely to be overfitted to these data and not generalisable. Even this does not achieve significantly better prognostic discrimination than the existing clinical scoring.

As explained in response to the eLife assessment, our objective was to use bioinformatics tools to identify the biological pathways and hub genes associated with TBM pathogenesis. We agree that ‘This analysis cannot be used to make causal inferences’: that would require different study design and approaches. The proposed gene signature has higher AUC values than the existing clinical model alone or in combination with clinical risk factors (Table 4). We agree that independent validation of the gene signature will be a crucial next step for future utility. We have performed qPCR in another sample set, and have added these results in the revision (Table 4 and supplementary figure S8)

Reviewer #1 (Recommendations For The Authors):

I have a few additional comments most of which are relatively minor:

(1) Can the authors please clarify if all the PTB cases are also HIV-negative?

This has been added to the methods section.

(2) For Table 1, can the authors please list the total number of patients with microbiologically confirmed TB regardless of the methods used? And for the two TBM groups, was the positive microbiology based on CSF findings?

The total number of patients with microbiologically confirmed TB was presented in Table 2 in definite TBM group, which was microbiologically confirmed TB diagnosed using microscopy, culture, and Xpert testing in cerebrospinal fluid (CSF) samples. We have updated the note in Table 2 to provide clarity on the definition.

(3) How was the discovery and validation set selected? Was it based on randomisation?

We randomly split TBM data into two datasets, a discovery cohort (n=142) and a validation cohort (n=139) with a purpose to ensure reproducibility of data analysis. We described this in the methods section.

(4) Line 107 can be better clarified by stating that the overall 3-month mortality rate is 21.7% for TBM regardless of HIV status.

Thank you, we have restated this sentence in the results section.

(5) The authors stated that samples were collected at enrolment when patients would have received less than 6 days of anti-tubercular treatment. Is there information on the median and IQR on the number of days that the patients would have received Rx, especially between the groups? Did the authors control for this variable when analysing for DEGs?

One of criteria to enroll participants in LAST-ACT and ACT-HIV trials is that they must receive less than 6 consecutive days of two or more drugs active against M. tuberculosis. However, the information of the days that the patients would have received Rx was not recorded and we could not control this variable when performing differential expression analysis for DEGs. This has been clarified further in the methods section: ‘The samples were taken at enrollment, when patients could not have received more than 6 consecutive days of two or more drugs active against M. tuberculosis.’

(6) I am a little bit concerned with the reads mapping accuracy (57%) to the human genome, which is fairly low. Did the authors investigate the reasons behind this low accuracy?

Thank you. It was indeed a typo. We have corrected it in the results section.

(7) On Tables S2-S4, can the authors please clarify what the last column (labelled as "B") shows?

Tables S2-S4 now have been changed to S3-S5. We have updated the legend of these tables to provide clarification regarding the meaning of the last column.

Reviewer #2 (Recommendations For The Authors):

If the authors wish to revise their manuscript, I suggest the following amendments:

(1) Provide a consort diagram for the selection of samples included in the present analysis (from parent study cohorts), allocation to test and validation splits for bioinformatics analysis, and outcomes.

We have provided our consort diagram in supplementary Figure S10.

(2) Provide details of inclusion criteria for pulmonary TB cohort, and how samples from this cohort were selected for inclusion in the present analysis. Please clarify whether this cohort excluded HIV-positive participants by design or by chance.

The inclusion criteria for the pulmonary TB cohort were described in the methods section. Due to the very low prevalence of HIV in this prospective observational study, HIV-positive participants were excluded. We have clarified in the amended manuscript that the pulmonary TB cohort only included HIV-negative participants.

(3) Baseline characteristics of HIV-positive participants (Table 1) should include CD4 count, HIV viral load, and whether anti-retroviral therapy was naïve or experienced.

We have included pre-treatment CD4 cell count, information on anti-retroviral therapy, and HIV viral load data in Table 1, as well as described these information in the results section.

(4) I note that the TBM samples were derived from RCTs of adjunctive steroid therapy, but not stratified in the present analysis by treatment arm allocation. Clearly, this may affect the survival/mortality outcomes that are the central focus of this manuscript. Therefore, they should be included in the models for differential gene expression analysis and prognostic signature discovery. To do so, the authors may need to wait until they are able to unblind the trial metadata.

With permission from the trial investigators, we were able to adjust the analyses for treatment with corticosteroids. The investigators remained blind to the allocation and we have not reported any direct effects of corticosteroids on outcome – such an analysis could only be done once the LAST-ACT trial has been reported (which won’t be until the end of 2024). Treatment outcome and effect were blinded by extracting only the fold change difference between survival and death in the linear regression model, in which gene expression was outcome and survival and treatment were covariates.

(5) I understood from the methods (lines 460-461) that batch correction of the RNAseq data was necessary. However, it is not clear how the samples were batched. PCA of the transcriptomes before and after batch correction with batch and study group labels should be provided. I would also advocate for a sensitivity analysis to check the robustness of the main findings without batch correction. I assume Fig2A represents batch-corrected data, but this is not clear.

We have now added information about the RNA sequencing batch and the batch correction approach, analyses and data visualizations utilized batch-corrected data in the methods section. We have also updated results related to batch correction in Fig. 2A and Supplementary Figure S9.

(6) I would encourage the authors to include a differential gene expression analysis to directly compare the transcriptome of TBM to that of pulmonary TB. I think it would add additional value to their focus on describing the transcriptome in TBM.

We thank for reviewer’s suggestion. Conducting differential gene expression analysis to compare the transcriptome of TBM with that of PTB is beyond the scope of this manuscript and we will examine this question separately.

(7) I don't really understand the purpose of splitting their data set into test and validation for the purposes of showing that WGCNA analysis is mostly reproduced in the two halves of the data. I would advocate that they scrap this approach to maximise the statistical power of their analysis in the descriptive work.

As mentioned in response to reviewer #1 in question #3, the purpose of splitting data is to ensure the reproducibility of the data analysis as suggested by Langfelder et al. (PMID: 21283776). This approach served two purposes: (i) to affirm the existence of functional modules in an independent cohort and (ii) to validate the association of interested modules or their hub genes with survival outcomes.

(8) The authors should soften the confidence in their interpretation of the GO/KEGG annotations of WGCNA modules. At least, they should include a paragraph that explicitly details the limitations of their analyses, including (i) the accuracy GO/KEGG annotations are not validated in this context (if at all), (ii) that none of the data can be used to make causal inferences and (iii) that peripheral blood assessments that are obviously impacted by changes in cellular composition of peripheral blood do not necessarily reflect immunopathogenesis at the site of disease - in fact if circulating cells are being recruited to the site of disease or other immune compartments, then quite the opposite interpretations may be true.

We appreciate the reviewer's comment. (i) In our analysis, we initially confirmed the existence of Weighted Gene Co-expression Network Analysis (WGCNA) modules in discovery cohort and validated the association of these modules with mortality outcomes in validation cohort. We then applied GO/KEGG annotations to define the biological functions involved in WGCNA modules. Finally, we performed Qusage analysis to directly test the association of top-hit pathways of each WGCNA module with mortality outcomes (see supplementary S6). This analysis approach helped to identify and validate modules and biological pathways associated with TBM mortality in this context, avoiding potential false positives in GO/KEGG annotations of WGCNA modules. (ii) We agree with the assessment that 'This analysis cannot be used to make causal inferences,' as that would require a different study design and approach. (iii) The focus of this study is to investigate the pathogenesis of TBM in the systemic immune system. We have highlighted this focus in the title and the aim of the manuscript.

(9) For the prognostic signature discovery and validation, I strongly recommend the authors include more robust validation. For example, to undertake an 80:20 split for sequential discovery (for feature selection and derivation of a prognostic model), followed by validation of a 'locked' model in data that made no contribution to discovery. In two separate sensitivity analyses. I also suggest they split their dataset (i) by treatment allocation in the RCT and (ii) by HIV status. In addition, their method for feature selection has to be clearer- precisely how they select hub genes from their WGCNA analysis as candidate predictors is not explained. Since this is such a prominent output of their manuscript, the results of this analysis should really be included in the main manuscript, and all performance metrics for discrimination should include confidence intervals.

Employing an 80:20 split for training and testing models is a good approach for an internal validation. However, we addressed the issue of overestimating the performance of a prognostic model by bootstrapping sampling approach proposed by Steyerberg et al. (PMID: 11470385). This approach has been proven to provide stable estimates with low bias. The overall model performance for discrimination, reported in our manuscript, was corrected for “optimism” to ensure internal validity. This adjustment was achieved through a 1000-times bootstrapping approach, which effectively accounted for estimation uncertainty. As such, there is no need to present confidence intervals for these metrics.

Moreover, in our revision, to confirm prognostic signatures independently, we have evaluated the predictive value of identified gene signatures using qPCR in another set of samples. The results have been added in Table 4, supplementary Figure S8 and the results section.

For the reasons given above (comment 4), we are unable to split our dataset by treatment allocation in this analysis. But as described, we have adjusted the analysis for corticosteroid treatment. Once the primary results of the LAST ACT trial have been published, we will examine the impact of corticosteroids on TBM pathophysiology and outcomes, seeking to better understand the mechanisms by which steroids have their therapeutic effects.

Given the difference in pathogenesis and immune response by HIV-coinfection, we stratified our analysis by HIV status. As the reviewer’s suggestion, we have provided additional details in the methods section regarding the selection of hub genes from associated WGCNA modules and the feature selection process for predictive modeling.

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

Learn more at Review Commons

---

Reply to the reviewers

__Below is our point-by-point reply to the reviewer's comments __

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

PNKP is one of critical end-processing enzymes for DNA damage repair, mainly base excision & single strand break repair, and double strand break repair to a certain extent. This protein has dual enzyme function: 3' phosphatase and 5' kinase to make DNA ends proper for ligation. It has been demonstrated that PTM of PNKP (e.g., S114, S126), particularly phosphorylation by either ATM or DNAPK, is important for PNKP function in DNA damage repair. The authors found a new phosphorylation site, T118, of PNKP which might be modified by CDK1 or 2 during S phase. This modification of phosphorylation is involved in maintenance and stability of the lagging strand, particularly Okazaki fragments. Loss of this phosphorylation could result in increased single strand gaps, accelerated speed of fork progression, and eventually genomic instability. And for this process, PNKP enzyme activity is not that important. And the authors concluded that PNKP T118 phosphorylation is important for lagging strand stability and DNA damage repair.

Major comments

In general, enzymes have protein interactions with its/their substrates. If PNKP is phosphorylated by either/both CDK1/2, the protein interaction between these would be expected. However, the authors did not provide any protein interactions in PNKP and CDKs. *Thank you for your suggestion. We will perform GFP-pulldown assays using cell extracts from HEK293 cells expressing GFP-WT-PNKP, GFP-T118A-PNKP. And then to confirm the interaction of PNKP and CDK1/2, we will blot with CDK1 and CDK2 antibodies. *

It is not clear how T118 phosphorylation is involved in DNA damage repair itself as the authors suggested. The data presenting the involvement of T118 phosphorylation in this mechanism are limited. This claim opens more questions than answers. CDK1/2 still phosphorylates T118 in this DNA damage repair process? What would happen to DNA damage repair in which PNKP involves outside of S phase in terms of T118 phosphorylation?

Thank you for your comment. We have investigated how T118 phosphorylation is important in DNA damage repair by several experiments. In figure S8, we tested SSB and DSB repair abilities of PNKP KO cells expressing PNKP T118A mutant, in which PNKP T118 phosphorylation has critical roles in both SSB and DSB repair pathways. Interestingly, the result of SSB repair assay (figure S8A & B) may indirectly indicate that T118 phosphorylation is important for SSB repair throughout cell cycle as these SSBs are instantly induced by IR exposure and recovered only for 30 mins that is presumably not enough time for cells to go through cell cycle. Along with the repair abilities, we also analyzed a recruitment kinetics/ability to DNA damage in PNKP T118A and T118D mutants using laser micro-irradiation assay in figure S9. This result indicates that the phosphorylation of PNKP at T118 is controlling its recruitment to at least laser-induced DNA damage sites. Moreover, we have analyzed recruitment of PNKP to a single-strand DNA gap structure, which mimics intermediates of some DNA repair pathways and incomplete Okazaki fragment maturation, using cell extracts from PNKP KO cells expressing PNKP T118A and T118D mutants and biochemical assay in figure 4H. This assay is much cleaner and shows that loss of T118 phosphorylation impairs PNKP recruitment to the ssDNA gap structure. We believe that these data sufficiently support our model that the phosphorylation of T118 on PNKP is involved in DNA repair in general. However, we agree with that we have not yet directly tested DNA repair ability of PNKP T118A in outside of S-phase. Therefore, in addition to these data, we will perform H2O2-induced SSB and IR-induced DSB repair assay using EdU (S phase) pulse labelling in PNKP KO cells expressing PNKP T118A mutant, then we will measure the ADP-ribose intensity and pH2AX foci in EdU negative cells (outside of S phase as the reviewer suggested).

Along the same line with #1/2 comments, the recruitment of PNKP to the damage sites is XRCC1 dependent. Is not clear whether PNKP recruitment to gaps on the lagging strand is XRCC1 independent or dependent. It might be interesting to examine (OPTIONAL)

*Thank you for an important suggestion. XRCC1 acts as a scaffold of PNKP and is required for recruitment of PNKP for canonical SSB repair, although we propose that PNKP is involved in two pathways in DNA replication: PARP1-XRCC1-dependent ssDNA gap filling pathway and Okazaki fragment maturation pathway working with FEN1. It is still important to address how XRCC1 is required for PNKP recruitment to the single-strand gaps on nascent DNA. Therefore, we will perform iPOND analysis in XRCC1 knock down + GFP-WT-PNKP expressed HEK293 cells. *

Minor comments

In results: 'Generation of PNKP knock out U2OS cell line' - In figure S2A; There are no data regarding diminishing the phosphorylation of g-H2AX.

Thank you for your suggestion. We will add pH2AX blot data in fig S2A (all reviewers requested).

• By showing data in figure S2B/C/D/E, the authors describe 'PNKP KO cells impaired the SSBs repair activity'. However, as the authors mentioned in this manuscript, PNKP could bind to either XRCC1 or XRCC4. Also for this experiment, IR had been applied, which induces DNA double strand breaks. Therefore, it is not certain that the authors' description is fully supported by these data presented. Perhaps, SSB inducing reagents should be used instead of IR.

In figure S2B/C/D/E, we used gamma-ray as IR source, which classified as low energy transfer irradiation. which mainly act as indirect effect to the DNA. It is estimated gamma-ray induce DNA damage as 60-80% SSBs and 20-40 % DSBs. We believe our results are reasonable. In addition to these results, we will perform poly-ADP-ribose assay with H2O2 treatment to more specifically assess SSBs repair activity.

• Is there any FACS analysis data to support the description of the last sentence 'especially the phosphorylation of PNKP T118, is required for S phase progression and proper cell proliferation'?

Thank you for your suggestion. We will add the FACS analysis data of cell cycle profiles in PNKP KO cells expressing GFP, GFP-PNKP WT, T118A.

In results: 'CDKs phosphorylate T118 of PNKP ~~~ replication forks'

• In figure 3A, Is there any change in total PNKP (both GFP-tagged & endogenous) level?

*Thank you for your suggestion. We agree with your comment. We will add the PNKP expression analysis in different cell cycle population in asynchronized and synchronized cells (G1, S, G2/M samples). *

In results: 'Phosphorylation of PNKP at T118 ~~~ between Okazaki fragments'

• In figure 4D, What happens in the ADP-ribose level, when T118D PNKP is expressed?

*Thank you for your suggestion. This is interesting question. We will perform ADP-ribosylation assay in PNKP KO cells and PNKP KO cells expressing PNKP WT and T118D, and add data of ADP-ribose levels in those cells. *

In results: 'PNKP is involved in postreplicative single-strand DNA gap-filling pathway'

• The description regarding data presented in figure 6 is not clear enough. These data might suggest that wildtype U2OS does not have SSB which is a substrate for S1 nuclease (except under FEN1i and PARPi treatment), whereas PNKP KO has SSB during both IdU and CIdU incorporation, so that S1 nuclease treatment dramatically reduces the speed of fork formation in PNKP KO cells. Also In figure 6B/C/D, adding an experimental group of PNKP KO with S1 nuclease + PARPi might help to understand the role of PNKP during replication better. Also these additional data could support the description in discussion 'Furthermore, PNKP is required for the PARP1-dependent single-strand gap-filling pathway ~~~ DNA gap structure'.

• *

*We agree with reviewer's comment and suggestion. Since this point is also raised by reviewer 3, we will add the rationale of the experiment and more detailed description about the results, which would substantially improve this manuscript. We will also revise our representation in text followed by the comment. In addition to revising the text, we will add experiment groups of PNKP KO with S1 nuclease with/without PARPi as the reviewer suggested. *

In results: 'Phosphorylation of PNKP at T118 is essential for genome stability'

• In figure S8C, Did you measure g-H2AX foci disappearance for later time point, such as 24 hrs after DNA damage? Is not clear whether non-phosphorylated PNKP at T118 inhibit DNA damage repair or make it slower? How does T114A-PNKP behave in this experimental condition? T114 is well known target of ATM/DNAPK for DDR & DSB repair.

Thank you for your suggestion. We agree with your point. It is very important to analyze whether T118A mutant shows delayed or total loss of DSB repair ability. We will add the measurement of pH2AX foci at 24 hrs after IR in PNKP KO cells expressing GFP, WT-PNKP, T118A-PNKP. Although the analysis of pS114 PNKP is previously reported (Segal-Raz et al., EMBO reports, 2011 and Zolner et al., Nucleic Acids Research, 2011), we will also perform pH2AX assay in PNKP KO cells expressing S114A-PNKP as a control.

The result shown in figure S9 should be described in the result section, not in the discussion section.

Thank you for your suggestion. This is a point also raised by Reviewer 3. Since we are going to re-consider the layout of the manuscript upon the planned revision (as reviewer 3 suggested), we will move these points to the appropriate result section from the discussion.

**Referees cross-commenting**

I could see a similar degree of positive tendency toward the manuscript. I agree with the comments and suggestions in additional experiments made by reviewers 2 and 3. Those suggestions will improve an impact of the manuscript in the DNA damage repair field.

Reviewer #1 (Significance (Required)):

Significance

The authors discovered new phosphorylation site (T118) of PNKP which is an important DNA repair protein. This modification seems to play a role in maintenance of the lagging strand stability in S phase. This discovery is something positive in DNA repair field to expand the canonical and non-canonical functions of DNA repair factors.

The data presented to support PNKP functions and T118 phosphorylation in S phase seem solid in general, yet it is not sure how much PNKP is critical in the Okazaki fragment maturation process which is known that several end processing enzymes (like FEN1, EXO1, DNA2 etc which leave clean DNA ends.) are involved.

These finding might draw good attentions from researchers interested broadly in cell cycle, DNA damage repair, replication, and possibly new tumor treatment.

My field and research interest: DNA damage response (including cell cycle arrest and programmed cell death), DNA damage repair (including BER, SSBR, DSBR)

Thank you very much for your positive comment. As you mentioned, there are several other end processing enzymes that seem to be involved in Okazaki fragment maturation, however, none of those enzymes is reported as a protein involved in the gap-filling pathway as well. Therefore, the role(s) of PNKP in DNA replication are very outstanding as PNKP could be involved in two separate pathways, Okazaki fragment maturation and a back-up gap-filling repair process. As you suggested, we will add several experiments such as iPOND experiments using XRCC1-depleted cells, analysis of DNA repair ability of PNKP T118A mutant throughout cell cycle and S1 nuclease DNA fiber assays in PNKP KO cells with/without PARP inhibitor treatment, to reveal how much PNKP is critical in the Okazaki fragment maturation. We believe that performing those experiments makes the conclusion and this manuscript more solid and convincing.

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

Polynucleotide kinase phosphatase (PNPK) participates in multiple DNA repair processes, where it acts on DNA breaks to generate 5'-phosphate and 3'-OH ends, facilitating the downstream activities of DNA ligases or polymerases.

This manuscript identifies a CDK-dependent phosphorylation site on threonine 118 in PNKP's linker region. The authors provide some convincing evidence that this modification is important to direct the activity of PNPK towards ssDNA gaps between Okazaki fragments during DNA replication. The authors monitored protein expression levels, enzymatic activity, the growth rate and replication fork speed, as well as the presence of ssDNA damage to make a comprehensive overview of the features of PNKP necessary for its function.

Overall, the conclusions are sufficiently supported by the results and this manuscript is relevant and of general interest to the DNA repair and genome stability fields. Some level of revision to the experimental data and text would help strengthen its message and conclusions.

Major points:

In an iPOND experiment the authors detect the wt PNKP and the T118 phosphorylated form at the forks and conclude that this phosphorylation promotes interaction with nascent DNA (Figure 3E). An informative sample to include here would have been the T118A mutant. Based on the model proposed, the prediction would be that it would not be associated with the forks, or at least, associated at reduced levels compared to the wt. *Thank you for your suggestion. We agree with your comment. We will add the iPOND analysis in PNKP KO cells expressing T118A mutant to confirm that pT118 is important for recruitment of PNKP at nascent DNA. *

The quality of the gels showing the phosphatase and kinase assays in Figure 5 could be improved to facilitate quantification of the results. The gel showing the phosphatase activity has a deformed band corresponding to K378A mutant. The gel showing the kinase activity seems to be hitting the detection limits, and the overall high background might influence the quantification of D171A mutant in the area of interest. The authors should provide a better quality of these gels, focusing on better separation (running them longer, eventually with a slightly increased electric current) and higher signal of the analyzed bands (longer incubation phosphatase/kinase prior to quenching or loading higher amount of DNA).

We agree with your suggestion. This phosphatase and kinase assay could be improved. We will perform this assay again followed by reviewer's suggestions.

The authors sometimes make statements like: "a slight increase, slightly increased, relatively high" without an evaluation of the statistical significance for the presented data. An example of such a statement is: "T118A mutant-expressing cells exhibited a marked delay in cell growth, which was not observed for S114A, although T122A, S126A, and S143A were slightly delayed," based on the figure 2E. A similar comment applies also to figures 4A, 5A, 5E. Whenever possible, the authors should include also an evaluation of the statistical significance in the statement.

Thank you for your suggestion. We will check manuscript and revise representation as reviewer's suggestion.

Minor revisions:

I could not find a gH2AX blot for figure S2A.

Thank you for your suggestion. We will add pH2AX blot data in fig S2A.

The authors established two PNKP-/- clones and supported it with sequencing and several functional observations However, the C-terminal antibody appears to detect lower-intensity bands (Figure 1A). Can authors comment on those bands?

Thank you for your comment. One possibility of this band is artificially recognized bands. To improve this problem, we will try electrophoresis for longer time to separate this band.

Why the S1 nuclease data on DNA fibers do not show the same level of epistasis with the Fen1i, as do those on ADP-ribosylation?

Because FEN1 dependent Okazaki fragment maturation and PARP1-XRCC1 dependent gap-filling pathway are different pathways, FEN1i and PARPi treatment resulted in an additive effect in S1 nuclease data in PNKP WT cells. To facilitate better understanding, we will add graphical scheme in figure 6 (a similar problem was raised by Reviewer 3 below) and revise the description of the result.

**Referees cross-commenting**

I agree with all the comments from the reviewers 1 and 3.

Reviewer #2 (Significance (Required)):

Significance:

The manuscript identifies a CDK phosphorylation site in a relevant DNA repair protein. The experiments on this part are elegant and convincing. It seems that this phosphorylation is important during DNA replication and there is some supporting evidence in this point, although not as robust, meaning that it is not clear whether this phosphorylation is controlling specifically the recruitment to Okazaki fragments, or a general role in DNA repair. Maybe if they see a reduced recruitment of the T118A mutant to the forks (iPOND experiment) this would further increase the impact.

This work will be relevant to the basic research, especially in the fields of DNA repair and DNA replication.

My expertise: DNA replication, genome stability, telomere biology.

Thank you very much for your positive comment. As you suggested, we will perform an iPOND assay using PNKP T118A mutant. In addition of the T118A iPOND assay, we will also analyze the DNA repair function of PNKP T118A mutant throughout cell cycle as reviewer 1 suggested. We believe that results of these experiments will pin down whether the phosphorylation of PNKP on T118 is controlling its recruitment to Okazaki fragments specifically or single-strand DNA gaps in general, and solidify the conclusion of the manuscript.

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

Tsukada and colleagues studied the role of PNKP phosphorylation in processing single-strand DNA gaps and its link to fork progression and processing of Okazaki fragments.

They generated two PNKP KO human clonal cell lines and described defects in cell growth, accumulation in S-phase, and faster fork progression. With some elegant experiments, they complement the KO cell lines with deletion and point mutants for PNKP, identifying a critical phosphorylation site (T118) in the linker regions, which is important for cell growth and DNA replication.

They show that phosphorylation of PNKP peaks in the mid-S phase. CDK1 and CDK2/ with Cyclin A2 are the two main CDK complexes responsible for this modification. With the IPOND experiment, the author shows that PNKP is recruited at nascent DNA during replication.

They described increased parylation activity in PNKP KO cells, and by using HU and emetin, they concluded that this increased activity depends on replication and synthesis of Okazaki fragments.

Interfering with Okazaki fragment maturation by FEN1 inhibition is epistatic with PNKP KO (and T118A) in influencing parylation activity in the S phase and fork progression. The authors try to understand by mutant complementation which of the two functions (Phosphatase vs Kinase) is important in processing OF, and they propose a primary role for the phosphatase activity of PNKP. They also show that T118 is important in controlling genome stability following different genotoxic stress. Finally, by coupling the measurement of fork progression with PARP/FEN1 inhibitors and S1 treatment, they propose a role of PNKP in the post-replicative repair of single-strand gaps due to unligated OF.

Here are my major points:

The authors use a poly ADP ribose deposition measurement to estimate SSB nick/gap formation. Even if PARP activity is strictly linked to SSB repair, ADP ribosylation does not directly estimate SSB/nick gap formation. In addition, in Figs S2A, B, and C, the authors use IR and PARG inhibition to measure poly-ADP ribosylation in WT and PNKP KO cells. IR produces both SSB and DSB. A better and cleaner experiment would be to directly measure SSB formation (with alkaline comet assay, for example) in combination with treatments that are known to mainly cause SSB (H2O2, or low doses of bleomycin). Thank you for your suggestion. The main purpose of this manuscript is to clarify the potential role of PNKP in DNA replication. Therefore, we generated PNKP KO human cells and figure S2 showed confirmation of function of established role of PNKP in SSBs and DSBs repair. In addition, previous our report published in EMBO Journal (Shimada et al., 2015), we showed SSBs and DSBs repair defect in PNKP KO MEF with comet assay (both alkaline and neutral) after IR and H2O2 treatment. In addition to those observations, we will also perform BrdU incorporation assay in PNKP WT and KO cells treated with H2O2. BrdU staining under an undenatured condition has now been commonly used and is a more direct method to detect ssDNA nick/gap formation. We believe that the importance of PNKP in SSB repair is sufficiently supported by all data such as previous comet assays in PNKP KO MEF cells and two SSB repair assays in human cells using ADP-ribose staining or BrdU incorporation, which will be provided in the revised manuscript.

The manuscript would benefit from substantially restructuring the figures' order and panels. Before starting the T118 part, the authors could create several figures to explain the main consequences of the loss of PNKP. A figure could be focused on DSB-driven genome instability (fig1 + fig S8 and S9). Then, a figure for the single-strand break and link to the S-phase. For example, by using data from Figure 6 and showing only WT vs PNKP KO +- Nuclease S1 (without FEN1 or PARP inhibitors), the authors could easily convince the readers that loss of PNKP leads to the accumulation of single-strand gaps. Only in the second part of the manuscript could they introduce all the T118 parts. Thank you for your suggestion. The layout of this manuscript makes reviewers feeling confusing. After performing all planned experiments, we will carefully re-consider the total layout of the revised manuscript.

I understand the use of a FEN1 inhibitor to link the PNKP KO phenotype to OF processing, but this drug does not either rescue or exacerbate any of the phenotypes described by the authors. It seems to have just an epistatic effect everywhere. So, what other conclusion can we have if not that PNKO has a similar effect to FEN1? I think that the presence of this inhibitor in many plots complicates the digestion of several figures a little bit. Maybe clustering the data in a different way (DMSO on one side FEN1i on the other) would help. Thank you for your suggestion. We agree that this data set is complicate. To facilitate better understanding, we will change organization of the data according to your suggestion and add graphical scheme in figure 6.

In terms of the other conclusion we can have from those experiments, the other conclusion is that PNKP might plays two important roles in DNA replication: Okazaki fragment maturation, which seems an epistatic effect with FEN1, and PARP1-XRCC1 dependent single-strand gap filling pathway, which is required for repairing single-strand gaps between Okazaki fragments when Okazaki fragment maturation pathway does not work properly (e.g., loss of FEN1 or PNKP). In figure 6D, we show that a double treatment of FEN1i and PARPi in PNKP WT cells with S1 nuclease treatment shows extensive amount of digested DNA fibers, although a single treatment of either FEN1i or PARPi in PNKP WT cells with S1 nuclease treatment leads to only limited amount of digested DNA fibers, which indicates that two pathways regulated by FEN1 or PARP are coordinately required for preventing eruption of ssDNA gaps in DNA replication. On the other hand, PNKP KO cells with S1 nuclease treatment cause extensive amount of digested DNA fibers even without FEN1i and PARP1i treatments, also it is not further increased by FEN1i and PARPi treatment. Those results indicate that PNKP itself is involved in two pathways mentioned above. Therefore, loss of PNKP has a similar phenotype with loss of FEN1 in terms of Okazaki fragment maturation, but also there is an additional effect in repairing ssDNA nicks/gaps, which is created in FEN1 loss condition, but FEN1 seems not dealing with it.

Fig S9 should be removed from the discussion. Additionally, the authors should consider whether they want to keep that piece of data in a manuscript that is already pretty dense. Why should we focus on additional linker residues and microirradiation data at the end of this manuscript? *Thank you for your suggestion. This is a point also raised by Reviewer 1. Since we are going to re-consider the layout of the manuscript upon the planned revision, we will move these points to the appropriate result section from the discussion. *

I suggest using a free AI writing assistant. I think this manuscript would substantially benefit from one. As a non-native English speaker, I personally use one of them and find it extremely useful. Thank you for your suggestion. Our manuscript was revised by a native speaker from an English correction company. However, for revised manuscript, we will discuss with native speakers as well as use a free AI writing assistant to improve the quality of the manuscript.

Minor points:

In Figure S1A, the author refers to P-H2AX, but I do not see this marker in the western blot. Thank you for your suggestion. We will add pH2AX blot data in fig S2A.

**Referees cross-commenting**

I agree with all comments from reviewer 1 and 2.

Reviewer #3 (Significance (Required)):

This is an interesting paper with generally solid data and proper statistical analysis. The figures are pretty straightforward. Unfortunately, the manuscript is dry, and the reader needs help to follow the logical order and the rationale of the experiments proposed. This is also complicated by the enormous amount of data the authors have generated. The authors should improve their narrative, explaining better why they are performing the experiment and not simply referring to a previous citation. Reordering panels and figures would help in this regard. Overall, with some new experiments, tone-downs over strong claims and a better explanation of the rationale behind experiments the authors could create a fascinating paper.

Thank you very much for your positive comment about the data/analysis and the logic behind the experiments provided in the manuscript. We agree with that a manner and a structure of the manuscript could be improved by reordering figures, cutting down some redundant experiments, adding better explanation of the rationale behind experiments, and toning-down some claims. With rewriting the manuscript as stated above and performing several additional experiments suggested by the reviewers, we believe that the revised manuscript will be more convincing and fascinating.

1. Description of the revisions that have already been incorporated in the transferred manuscript

Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. If no revisions have been carried out yet, please leave this section empty.

• *

• *

Reviewer #1:

Minor comments

• Is there any difference (except for PARGi exposure time?!) between figure S2B/C and S2D/E? Both data show increased ADP ribose after IR. It seems redundancy. Also it is hard to imagine that there is absolutely no sign of ADP ribose after IR w/o PARGi treatment (figure S2D).

Figure S2B/C show spontaneous single strand DNA breaks (SSBs) in PNKP KO cells, on the other hand, figure 2S/E show ectopic SSBs induced by IR exposure in PNKP KO cells. We believe these data help for readers to understand the effect of endo or exo damage in PNKP KO cells. Poly-ADP ribosylations are immediately removed from SSB sites after repair as demonstrated previously (Tsukada, et al., PLoS One 2019, Kalasova et al., Nucleic Acids Research, 2020), although not zero (low level), it is very difficult to detect without PARGi treatment.

• *

Legend for figure S3 - typo!

Thank you for your suggestion about typo. The legend for figure S3 is corrected as "Protein expression of PNKP mutants in U2OS cells".

• *

• In figure S3A/B, it is quite interesting that the PNKP antibody used for this analysis can detect all truncated and alanine substituted PNKP proteins. It might be helpful to indicate for other researchers which antibody used (Novus; epitope - 57aa to 189 aa or Abcam; epitope not revealed).

In S3A/B, Novus PNKP antibody was used for all blots. We indicated this in the figure legend as "PNKP antibody (Novus: NBP1-87257) was used for comparing expression levels of endogenous and exogenous PNKP".

• *

In results: 'PNKP phosphorylation, especially of T118 ~~~ proliferation'

• In the fork progression experiment (figure 2C), is there any statistical difference between D2 and D3/4 expressing cells?

*Thank you for your suggestion. We performed statistical analysis as the reviewer suggested. Statistical analysis shows that there are no significant differences between D2 and D3/D4. Meanwhile, there are significant differences between WT and D3(P- What is the basis of the description 'Since the linker region of PNKP is considered to be involved in fork progression'? Any reference?

This sentence was considered based on the above sentences "Furthermore, D2 mutant-expressing cells also showed an increased speed of the replication fork compared to WT and D1 mutant-expressing cells, although D3 and D4 showed mildly high-speed fork progression.". The D2 mutant lacks a whole linker region, which shows increased speed of DNA fiber in figure 2C. Therefore, we originally explained as the sentence above. We have revised the sentence to "Since these results may indicate the linker region of PNKP is involved in proper fork progression".

• *

• In figure 3B: pS114-PNKP (also pS15-p53) is DNA damage inducible. In this experiment, was DNA damage introduced? Roscovitine could hinder DNA repair process, but not inducing DNA damage itself.

Thank you for your suggestion. DNA damage induction was not applied in this experiment. We agree that this panel makes confusing. We think that endogenously S114-PNKP (also S15-p53) might be phosphorylated slightly but not significant, although this is not the scope of this manuscript. This result showing that phosphorylated-T118 is reduced by Roscovitine treatment maybe redundant as we also have a result of in vitro phosphorylation assay using several combinations of CDKs and Cyclin proteins, which is a cleaner experiment to prove which CDK/Cyclin complex is directly controlling the T118 phosphorylation. Since the manuscript already contains enough amount of data to support the conclusion (as reviewer 3 also stated), we removed those blots result from the panel to avoid complicating the conclusion.

• *

In results: 'Phosphatase activity of PNKP is ~~~ of Okazaki fragments'

• In figure 5C, any statistical analysis between WT-PNKP KO vs D171A-PNKP KO or K378A-PNKP KO has been done?

Thank you for your comment. Statistical analysis shows P *

In discussion, 'In contrast, the T118A mutants showed the absence of both SSBs and DSBs repair (Fig. S7) : figure S7 does not indicate what the authors describe.

Thank you for pointing out this. This should refer to figure S8 instead of figure S7. We have corrected this error.

In addition, the same sentence in discussion: No evidence demonstrate that 'the absence of both SSBs and DSBs repair', and the following sentence is not clear.

*This is same point with above. We have corrected this mis-referencing and revised the sentence to "In contrast, the T118A mutants showed the impaired abilities of both SSBs and DSBs repair (Fig. S8).". We also revised the following sentence to "However, residual SSBs due to impaired SSB repair ability (e.g., in PARPi-treated cells and T118A cells) sometimes cause DNA replication-coupled DSBs formation in S phase, and the phenotype in DSB repair assay of the T118A mutant may be caused by an accumulated formation of DNA replication-coupled DSBs. Future works will be needed to distinguish whether the T118 phosphorylation directly regulate PNKP recruitment to DSBs as well as SSBs." for better explanation of the result. *

• *

In discussion, 'Because both CDK1/cyclin A2 and CDK2/cyclin A2 are involved in PNKP phosphorylation, cyclin A2 is likely important for these activities': It is not clear what this description intends? Is 'cyclin A2' important in what stance?

This description is coming from Fig3C observation. Since both CDK1 and CDK2 activities are cyclin A2 dependent, we speculated cyclin A2 is important for CDK1/CDK2 dependent PNKP T118 phosphorylation. We revised the description to "Since both CDK1/Cyclin A2 and CDK2/Cyclin A2 phosphorylate T118 of PNKP, we speculated that PNKP T118 is phosphorylated in S phase to G2 phase in CDK1/Cyclin A2- and CDK2/Cyclin A2-dependent manner (Fig. 3B and C)".

• *

In discussion, 'This may be explained by the fact that mutations in the phosphorylated residue in the linker region are embryonic lethal': any reference to support this embryonic lethality?

Thank you for your suggestion. We agree with that this sentence is overwriting. We revise the sentence to "This observation may indicate that mutations in the phosphorylated residue (T118) in the linker region are potentially embryonic lethal due to the importance of T118 in DNA replication, which is revealed in the present study.".

• *

• *

Reviewer #2:

Minor comments

Sometimes there are incorrect references to the figures in the discussion (e.g. FigS9A, B, and C, are called out instead of E, F and G), a similar issue is found 4 lines below in the same page.

Thank you for pointing out these errors. We checked the references in the discussion and corrected to the appropriate references.

Based on the data in Figure 3A the authors suggest that pT118-PNKP follows Cyclin A2 levels, but this does not appear very clearly in the gel, especially for the last point. Even though the results are convincing, the authors should rephrase the conclusions of Figure 3A to reflect better the results.

Thank you for your suggestion. We agree that this phrase is overwriting. We revised the conclusion to "pT118-PNKP was detected in asynchronized cells but increased particularly in the S phase, similar to Cyclin A2 expression levels, although the reduction of pT118, possibly dephosphorylation of T118, seems not as robust as the reduction of the Cyclin A2 expression level at the 12 hours time point. However, this effect was very weak during mitosis, suggesting that T118 phosphorylation plays a specific role in the S phase.".

I did not find a reference to what seems to be a relevant work in this topic: PMID: 22171004

Thank you for your suggestion. We have added the ref (Coquelle et al., PNAS, 2011) in Introduction section.

---

Reviewer #3:

Major comments

The authors should consider and discuss the potential role of PNKP KO outside of the S-phase. In Figure 4C, while it is clear that poly ADP ribosylation is higher in S-phase, the effects of PNKP KO and complementation by WT or T118A are equally present. This would be more immediate if comparison, fold change, and statistical significance calculation were done within the same cell cycle phase instead of between cell stages. This is also clear by IF in Figure 4B. How do the authors explain this? Thank you for your suggestion. We agree with reviewer's suggestion. We compared intensities of ADP-ribose between cell lines in same cell cycle rather than between different cell cycles in a same cell line and added the respective statistics in figure 4C. Also, we agree with that poly ADP-ribose intensity is changed outside of S phase between WT and T118A PNKP expressing PNKP KO cells. As shown in figure S8, PNKP pT118 is also involved in DNA repair. These results might reflect of PNKP function outside of S phase. We have added the sentence "Of note, PNKP−/−*cells and PNKP T118A cells showed markedly higher ADP-ribose intensity in outside the S phase as well, which indicate that PNKP and T118 may have an endogenous role to prevent SSBs formation in outside the S phase. Since FEN1 has been reported to function in R-loop processing, PNKP could also be involved in this process. Future studies of a role of PNKP in different cell cycle will be able to address this question." to discuss about the function of PNKP outside the S phase. We have added the ref (Cristini et al., Cell Reports, 2019, and Laverde et al., Genes, 2022). *

• *

• *

In connection with the previous point, can the author provide the same quantification in Figure 4E also for G2/M and not only the S phase? This should give an estimate of the activity of FEN1 outside the S-phase. This is important because FEN1 has other functions apart from OF maturation, such as R loop processing (Cristini 2019; Laverde 2023) Thank you for your suggestion. Here attached is the data of ADP-ribose intensity in cells outside the S phase as you suggested. FEN1i treatment still induces increased ADP-ribose intensity in outside the S phase as well, although the difference between with/without FEN1i treatment is much smaller than that in S phase, indicating that FEN1 has other functions outside the S phase. This finding is very interesting. However, the function of FEN1 in outside the S phase is outside the scope of this manuscript. Therefore, we would like to not put this data in the manuscript to avoid complicating the conclusion (as reviewer 3 also suggested).

• *

Why does FEN1 inhibition induce a faster fork progression in Fig4 but not in Fig5 and Fig6? Yes, it does in figure 4 and figure 5. In PNKP WT cells, FEN1i-treated fibers (CldU) show an increased speed of forks compared to non-treated fibers (IdU). However, loss of PNKP and T118 phosphorylation themselves cause a faster fork progression even if without FEN1i treatment, therefore the difference of speeds of forks before/after FEN1i treatment in PNKP KO and T118A cells is disappeared as both fibers grow faster than intact fibers in normal cells. In regard to figure 6, as you mentioned in a latter comment about figure 6, the title of vertical axis of the graph showing CldU length should not be speeds of replication forks as those DNA fibers are potentially digested by S1 nuclease, which is modified in the revised manuscript. Even so, DNA fibers from FEN1i-treated cells (CldU) with S1 nuclease shows similar length with fibers from untreated cells with S1 nuclease, whereas FEN1 inhibitor treatment accelerates a speed of forks in general (figure 4 and figure 5, assays without S1 nuclease), indicating that FEN1i treatment induces remaining of some ssDNA nicks/gaps which are substrates of S1 nuclease.

• *

How do the authors explain the impaired DNA gap binding activity of the phospho-mimetic T118D? Thank you for your suggestion. We think that the appropriate timing of phosphorylation of PNKP T118 is important, while the phosphor-mimetic mutant T118D mimics consecutively phosphorylated situation that may result in incomplete complementation of PNKP function.

• *

I would like to see a representative fiber image from Fig 6. Additionally, in Figure 6, the author should not label the y-axis as CldU-fork speed. Nuclease S1 treatment destroys single-strand gaps (in vitro) and does not affect the fork speed (in vivo) Thank you for your suggestion. We have added a representative fiber image. We also agree with that CldU fork speed is not a right label of y-axis as CldU fibers are potentially digested by S1 nuclease. We changed the y-axis label to "CldU tract length [kb/min]" in figure 6.

• *

Figure 5E: both mutants (kinase vs phosphatase) increase polyADP ribose intensity, while the title of this figure only emphasizes the phosphatase activity. We agree with your comment. We have changed this subtitle to "Enzymatic activities of PNKP is important for the end-processing of Okazaki fragments".

• *

• *

Minor comments

• *

The authors refer to Hoch Nature 2017 when referring to polyADP ribose IF + PARG inhibition. Should they not refer to Hanzlikova Mol Cell 2018?

Thank you for your suggestion. We have added the ref (Hanzlikova et al., Mol Cell 2018).

Statistical analysis should be performed on the cell cycle profile in Figure 1B * *

We performed statistical analysis to check whether there are significant differences of S phase population between WT and PNKP KO cells. There were significant differences between WT vs PNKP KO C1 (PThe authors should not refer to fork degradation or protection as a given fact without assessing it in these conditions. Thank you for your suggestion. We assume that this comment refers to the result section of figure 1 and figure 4. We have added a sentence "although future studies will be needed to investigate whether PNKP−/− cells has the fork protection phenotype" in the result section of figure 1. We have changed representation in the section according to the reviewer's suggestion in the result section of figure 4.*

• *

• *

Reviewer #1 (Public Review):

Summary:

The authors want to explore how much two known minibinder protein domains against the Spike protein of SARS-CoV-2 can function as a binding domain of 2 sets of synthetic receptors (SNIPR and CAR); the authors also want to know how some modifications of the linkers of these new receptors affect their activation profile.

Major strengths and weaknesses of the methods and results:

- Strengths include: analysis of synthetic receptor function for 2 classes of synthetic receptors, with robust and appropriate assays for both kinds of receptors. The modifications of the linkers are also interesting and the types of modifications that are often used in the field.

- Weaknesses include: none of the data analysis provides statistical interpretation of the results (that I could find). One dataset is confusing: Figures 5A and C, are said to be the same assay with the same constructs, but the results are 30% in A, and 70% in C.

An appraisal of whether the authors achieved their aims, and whether the results support their conclusions:

Given the open-ended nature of the goal (implicit in it being an exploration), it is hard to say if the authors have reached their aims; they have done an exploration for sure; is it big enough an exploration? This reviewer is not sure.

The results are extremely clearly presented, both in the figures and in the text, both for the methods and the results. The claims put forward (with limited exceptions see below) are very solidly supported by the presented data.

A discussion of the likely impact of the work on the field, and the utility of the methods and data to the community:

The work may stimulate others to consider minibinders as potential binding domains for synthetic receptors. The modifications that are presented although not novel, do provide a starting point for larger-scale analysis.

It is not clear how much this is generalizable to other binders (the authors don't make such claims though). The claims are very focused on the tested modifications, and the 2 receptors and minibinder used, a scope that I would define as narrow; the take-home message if one wants to try it with other minibinders or other receptors seems to be: test a few things, and your results may surprise you.

Any additional context you think would help readers interpret or understand the significance of the work:

We are at the infancy stage of synthetic receptors optimization and next-generation derivation; there is a dearth of systematic studies, as most focus is on developing a few ones that work. This work is an interesting attempt to catalyze more research with these new minibinders. Will it be picked up based on this? Not sure.

I agree with this statement and feel deeply concerned. When we use most social media platforms, they usually select user preference content for us when registering an account in order to push content to users that they are more interested in. And this is actually a way to obtain information and find ways to attract the user's attention. Moreover, as we use the software, we are also using different algorithms to infer how our interests have changed. At the same time, we may also cooperate with shopping software to directly push the items of interest we just mentioned in the video or forum social media so that we can purchase them. I think it’s a bit scary how much big data knows about us.

Author response:

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

Reviewer #1 (Public Review):

In this manuscript, Huang and colleagues explored the role of iron in bacterial therapy for cancer. Using proteomics, they revealed the upregulation of bacterial genes that uptake iron, and reasoned that such regulation is an adaptation to the iron-deficient tumor microenvironment. Logically, they engineered E. Coli strains with enhanced iron-uptake efficiency, and showed that these strains, together with iron scavengers, suppress tumor growth in a mouse model. Lastly, they reported the tumor suppression by IroA-E. Coli provides immunological memory via CD8+ T cells. In general, I find the findings in the manuscript novel and the evidence convincing.

(1) Although the genetic and proteomic data are convincing, would it be possible to directly quantify the iron concentration in (1) E. Coli in different growth environments, and (2) tumor microenvironment? This will provide the functional consequences of upregulating genes that import iron into the bacteria.

We appreciate the reviewer’s comment regarding the precise quantification of iron concentrations. In our study, we attempted various experimental approaches, including Immunohistochemistry utilizing an a Fe3+ probe, iron assay kit (ab83366), and Inductively Coupled Plasma Mass Spectrometry (ICP-MS). Despite these attempts, the quantification of oxidized Fe3+ concentrations proved challenging due to the inherently low levels of Fe ions and difficulty to distinguish Fe2+ and Fe3+. We observed measurements below the detection threshold of even the sensitive ICP-MS technique. To circumvent this limitation, we designed an experiment wherein bacteria were cultured in a medium supplemented with Chrome Azurol S (CAS) reagent, which colormetrically detects siderophore activity. We compared WT bacteria and IroA-expressing bacteria at varying levels of Lcn2 proteins. The outcome, as depicted in the updated Fig. 3b, reveals an enhanced iron acquisition capability in IroA-E. coli under the presence of Lcn2 proteins, in comparison to the wild-type E. coli strains. In addition to the Lcn2 study, the proteomic study in Figure 4 highlights the competitive landscape between cancer cells and bacteria. We observed that IroA-E. coli showed reduced stress responses and exerted elevated iron-associated stress to cancer cells, thus further supporting the IroA-E. coli’s iron-scavenging capability against nutritional immunity.

(2) Related to 1, the experiment to study the synergistic effect of CDG and VLX600 (lines 139-175) is very nice and promising, but one flaw here is a lack of the measurement of iron concentration. Therefore, a possible explanation could be that CDG acts in another manner, unrelated to iron uptake, that synergizes with VLX600's function to deplete iron from cancer cells. Here, a direct measurement of iron concentration will show the effect of CDG on iron uptake, thus complementing the missing link.

We appreciate the reviewer’s comment and would like to point the reviewer to our results in Figure S3, which shows that the expression of CDG enhances bacteria survival in the presence of LCN2 proteins, which reflects the competitive relationship between CDG and enterobactin for LCN2 proteins as previously shown by Li et al. [Nat Commun 6:8330, 2015]. We regret to inform the reviewer that direct measurement of iron concentration was attempted to no avail due to the limited sensitivity of iron detecting assays. We do acknowledge that CDG may exert different effects in addition to enhancing iron uptake, particularly the potentiation of the STING pathway. We pointed out such effect in Fig 2c that shows enhanced macrophage stimulation by the CDG-expressing bacteria. We would like to accentuate, however, that a primary objective of the experiment is to show that the manipulation of nutritional immunity for promoting anticancer bacterial therapy can be achieved by combining bacteria with iron chelator VLX600. The multifaceted effects of CDG prompted us to focus on IroA-E. coli in subsequent experiments to examine the role of nutritional immunity on bacterial therapy. We have updated the associated text to better convey our experimental design principle.

Lines 250-268: Although statistically significant, I would recommend the authors characterize the CD8+ T cells a little more, as the mechanism now seems quite elusive. What signals or memories do CD8+ T cells acquire after IroA-E. Coli treatment to confer their long-term immunogenicity?

We apologize for the overinterpretation of the immune memory response in our previous manuscript and appreciate the reviewer’s recommendation to further characterize CD8+ T cells post-IroA-E. coli treatment. Our findings, which show robust tumor inhibition in rechallenge studies, indicate establishment of anticancer adaptive immune responses. As the scope of the present work is aimed at demonstrating the value of engineered bacteria for overcoming nutritional immunity, expounding on the memory phenotypes of the resulting cellular immunity is beyond the scope of the study. We do acknowledge that our initial writing overextended our claims and have revised the manuscript accordingly. The revised manuscript highlights induction of anticancer adaptive immunity, attributable to CD8+ T cells, following the bacterial therapy.

(3) Perhaps this goes beyond the scope of the current manuscript, but how broadly applicable is the observed iron-transport phenomenon in other tumor models? I would recommend the authors to either experimentally test it in another model or at least discuss this question.

We highly appreciate the reviewer’s suggestion regarding the generalizability of the iron-transport phenomenon in diverse tumor models. To address this, we extended our investigations beyond the initial model, employing B16-F10 melanoma and E0771 breast cancer in mouse subcutaneous models. The results, as depicted in Figures 3g to 3j and Figure S5, demonstrate the superiority of IroA-E. coli over WT bacteria in tumor inhibition. These findings support the broad implication of nutritional immunity as well as the potential of iron-scavenging bacteria for different solid tumor treatments.

Reviewer #2 (Public Review):

Summary:

The authors provide strong evidence that bacteria, such as E. coli, compete with tumor cells for iron resources and consequently reduce tumor growth. When sequestration between LCN2 and bacterobactin is blocked by upregulating CDG(DGC-E. coli) or salmochelin(IroA-E.coli), E. coli increase iron uptake from the tumor microenvironment (TME) and restrict iron availability for tumor cells. Long-term remission in IroA-E.coli treated mice is associated with enhanced CD8+ T cell activity. Additionally, systemic delivery of IroA-E.coli shows a synergistic effect with chemotherapy reagent oxaliplatin to reduce tumor growth.

Strengths:

It is important to identify the iron-related crosstalk between E. coli and TME. Blocking lcn2-bacterobactin sequestration by different strategies consistently reduces tumor growth.

Weaknesses:

As engineered E.coli upregulate their function to uptake iron, they may increase the likelihood of escaping from nutritional immunity (LCN2 becomes insensitive to sequester iron from the bacteria). Would this raise the chance of developing sepsis? Do authors think that it is safe to administrate these engineered bacteria in mice or humans?

We appreciate the reviewer’s comment on the safety evaluation of the iron-scavenging bacteria. To address the concern, we assessed the potential risk of sepsis development by measuring the bacterial burden and performing whole blood cell analyses following intravenous injection of the engineered bacteria. As illustrated in Figures 3k and 3l, our findings indicate that the administration of these engineered bacteria does not elevate the risk of sepsis. The blood cell analysis suggests that mice treated with the bacteria eventually return to baseline levels comparable to untreated mice, supporting the safety of this approach in our experimental models.

Reviewer #3 (Public Review):

Summary:

Based on their observation that tumor has an iron-deficient microenvironment, and the assumption that nutritional immunity is important in bacteria-mediated tumor modulation, the authors postulate that manipulation of iron homeostasis can affect tumor growth. They show that iron chelation and engineered DGC-E. coli have synergistic effects on tumor growth suppression. Using engineered IroA-E. coli that presumably have more resistance to LCN2, they show improved tumor suppression and survival rate. They also conclude that the IroA-E. coli treated mice develop immunological memory, as they are resistant to repeat tumor injections, and these effects are mediated by CD8+ T cells. Finally, they show synergistic effects of IroA-E. coli and oxaliplatin in tumor suppression, which may have important clinical implications.

Strengths:

This paper uses straightforward in vitro and in vivo techniques to examine a specific and important question of nutritional immunity in bacteria-mediated tumor therapy. They are successful in showing that manipulation of iron regulation during nutritional immunity does affect the virulence of the bacteria, and in turn the tumor. These findings open future avenues of investigation, including the use of different bacteria, different delivery systems for therapeutics, and different tumor types.

Weaknesses:

• There is no discussion of the cancer type and why this cancer type was chosen. Colon cancer is not one of the more prominently studied cancer types for LCN2 activity. While this is a proof-of-concept paper, there should be some recognition of the potential different effects on different tumor types. For example, this model is dependent on significant LCN production, and different tumors have variable levels of LCN expression. Would the response of the tumor depend on the role of iron in that cancer type? For example, breast cancer aggressiveness has been shown to be influenced by FPN levels and labile iron pools.

We highly appreciate the reviewer’s insightful comment on the varying LCN2 activities across different tumor types. In light of the reviewer’s suggestion, we extended our investigations beyond the initial colon cancer model, employing B16-F10 melanoma and E0771 breast cancer in mouse subcutaneous models. The results, as depicted in Figures 3g to 3j and Figure S5, demonstrate that IroA-E. coli consistently outperforms WT bacteria in tumor inhibition. We acknowledge the reviewer’s comment regarding LCN2 being more prominently examined in breast cancer and have highlighted this aspect in the revised manuscript. For colon and melanoma cancers, several reports have pointed out the correlation of LCN2 expression and the aggressiveness of these cancers [Int J Cancer. 2021 Oct 1;149(7):1495-1511][Nat Cancer. 2023 Mar;4(3):401-418], albeit to a lesser extent. These findings support the broad implication of nutritional immunity as well as the potential of iron-scavenging bacteria for different solid tumor treatments. The manuscript has been revised to reflect the reviewer’s insightful comment.

• Are the effects on tumor suppression assumed to be from E. coli virulence, i.e. Does the higher number of bacteria result in increased immune-mediated tumor suppression? Or are the effects partially from iron status in the tumor cells and the TME?

We appreciate the reviewer’s question regarding the therapeutic mechanism of IroA-E. coli. Bacterial therapy exerts its anticancer action through several different mechanisms, including bacterial virulence, nutrient and ecological competition, and immune stimulation. Decoupling one mechanism from another would be technically challenging and beyond the scope of the present work. With the objective of demonstrating that an iron-scavenging bacteria can elevate anticancer activity by circumventing nutritional immunity, we highlight our data in Fig. S6, which shows that IroA-E. coli administration resulted in higher bacterial colonization within solid tumors compared to WT-E. coli on Day 15. This increased bacterial presence supports our iron-scavenging bacteria design, and we highlight a few anticancer mechanisms mediated by the engineered bacteria. Firstly, as shown in Fig. 4d, IroA-E. coli is shown to induce an elevated iron stress response in tumor cells as the treated tumor cells show increased expression of transferrin receptors. Secondly, our experiments involving CD8+ T cell depletion indicates that the IroA-E. coli establishes a more robust anticancer CD8+ T cell response than WT bacteria. Both immune-mediated responses and alterations in iron status within the tumor microenvironment are demonstrated to contribute to the enhanced anticancer activity of IroA-E. coli in the present study.

• If the effects are iron-related, could the authors provide some quantification of iron status in tumor cells and/or the TME? Could the proteomic data be queried for this data?

We appreciate the reviewer’s query regarding the quantification of iron concentrations. In our study, we attempted various experimental approaches, including Immunohistochemistry utilizing an a Fe3+ probe, iron assay kit (ab83366), and Inductively Coupled Plasma Mass Spectrometry (ICP-MS). Despite these attempts, the quantification of oxidized Fe3+ concentrations proved challenging due to the inherently low levels of Fe ions and difficulty to distinguish Fe2+ and Fe3+. We observed measurements below the detection threshold of even the sensitive ICP-MS technique. Consequently, to circumvent this limitation, we designed an experiment wherein bacteria were cultured in a medium supplemented with Chrome Azurol S (CAS) reagent, which colormetrically detects siderophore activity. We compared WT bacteria and IroA-expressing bacteria at varying levels of Lcn2 proteins. The outcome, as depicted in the updated Fig. 3b, reveals an enhanced iron acquisition capability in IroA-E. coli under the presence of Lcn2 proteins, in comparison to the wild-type E. coli strains. In addition to the Lcn2 study, the proteomic study in Figure 4 highlights the competitive landscape between cancer cells and bacteria. We observed that IroA-E. coli showed reduced stress responses and exerted elevated iron-associated stress to cancer cells, thus further supporting the IroA-E. coli’s iron-scavenging capability against nutritional immunity.

Reviewing Editor:

The authors provide compelling technically sound evidence that bacteria, such as E. coli, can be engineered to sequester iron to potentially compete with tumor cells for iron resources and consequently reduce tumor growth. Long-term remission in IroA-E.coli treated mice is associated with enhanced CD8+ T cell activity and a synergistic effect with chemotherapy reagent oxaliplatin is observed to reduce tumor growth. The following additional assessments are needed to fully evaluate the current work for completeness; please see individual reviews for further details.

We appreciate the editor’s positive comment.

(1) The premise is one of translation yet the authors have not demonstrated that manipulating bacteria to sequester iron does not provide a potential for sepsis or other evidence that this does not increase the competitiveness of bacteria relative to the host. Only tumor volume was provided rather than animal survival and cause of death, but bacterial virulence is enhanced including the possibility of septic demise. Alternatively, postulated by the authors, that tumor volume is decreased due to iron sequestration but they do not directly quantify the iron concentration in (1) E. Coli in different growth environments, and (2) tumor microenvironment. These important endpoints will provide the functional consequences of upregulating genes that import iron into the bacteria.

We appreciate the editor’s comment and have added substantial data to support the translational potential of the iron-scavenging bacteria. In particular, we added evidence that the iron-scavenging bacteria does not increase the risk of sepsis (Fig. 3k, l), evidence of increased bacteria competitiveness and survival in tumor (Fig. S6), and iron-scavenging bacteria’s superior anticancer ability and survival benefit across 3 different tumor models (Fig. 3e-j; Fig. S5). While direct measurement of iron concentration in the tumor environment is technically difficult due to the challenge in differentiating Fe2+ and Fe3+ by available techniques, we added a colormetric CAS assay to demonstrate the iron-scavenging bacteria can more effectively utility Fe than WT bacteria in the presence of LCN2 (Fig. 3b). These results substantiate the translational relevance of the engineered bacteria.

(2) There is no discussion of the cancer type and why this cancer type was chosen. If the current tumor modulation system is dependent on LCN2 activity, there would need to be some recognition that different tumors have variable levels of LCN expression. Would the response of the tumor depend on the role of iron in that cancer type?

We appreciate the comment and added relevant text and citations describing clinical relevance of LCN2 expression associated with the tumor types used in the study (breast cancer, melanoma, and colon cancer). Elevated LCN2 has been associated with higher aggressiveness for all three cancer types.

(3) To demonstrate long-term anti-cancer memory was established through enhancement of CD8+ T cell activity (Fig 5c), the "2nd seeding tumor cells" experiment may need to be done in CD8 antibody-treated IronA mice since CD8+ T cells may play a role in tumor suppression regardless of whether or not iron regulation is being manipulated. It appears that the control group for this experiment is naive mice (and not WT-E. coli treated mice), in which case the immunologic memory could be from having had tumor/E. coli rather than the effect of IroA-E. coli.

We acknowledge that our prior writing may have overstated our claim on immunological memory. Our intention is to show that upon treatment and tumor eradication by iron-scavenging bacteria, adaptive immunity mediated by CD8 T cells can be elicited. We also did not consider a WT-E. coli control as no WT-E. coli treated group achieved complete tumor regression. We have modified our text to reflect our intended message.

Reviewer #1 (Recommendations For The Authors):

All the figures seem to be in low resolution and pixelated. Please upload high-resolution ones.

We have updated figures to high-resolution ones.

Reviewer #2 (Recommendations For The Authors):

Some specific comments towards experiments:

(1) For Fig 2 f/ Fig 3f/ Fig 5d/Fig6c, the survival rate is based on the tumor volume (the mouse was considered dead when the tumor volume exceeded 1,500 mm3). Did the mice die from the experiment (how many from each group)? If it only reflects the tumor size, do these figures deliver the same information as the tumor growth figure?

We appreciate the reviewer’s comment. The survival rate is indeed based on tumor volume, and we used a cutoff of 1500 mm3. No death event was observed prior to the tumors reaching 1500 mm3. Although the survival figures cover some of the information conveyed by the tumor volume tracking, the figures offer additional temporal resolution of tumor progression with the survival figures. Having both tumor volume and survival tracking are commonly adopted to depict tumor progression. We have the protocol regarding survival monitoring to the materials and method section.

(2) Fig 3a, not sure if entE is a good negative control for this experiment. Neg. Ctrl should maintain its CFU/ml at a certain level regardless of Lcn2 conc. However, entE conc. is at 100 CUF/ml throughout the experiment suggesting there is no entE in media or if it is supersensitive to Lcn2 that bacteria die at the dose of 0.1nM?

We appreciate the reviewer’s comment. The △entE-E. coli was indeed observed to be highly sensitive to LCN2. We included the control to highlight the competitive relationship between entE and LCN2 for iron chelation, which is previously reported in literature [Biometals 32, 453–467 (2019)].

(3) Fig 4, the authors harvested bacteria from the tumor by centrifuging homogenized samples at different speeds. Internal controls confirming sample purity (positive for bacteria and negative for cells for panels a,b,c; or vice versa for panel d) may be necessary. This comment may also apply to samples from Fig 1.

We acknowledge the reviewer’s concern and would like to point out that the proteomic analysis was performed using a highly cited protocol that provides reference and normalization standards for E. coli proteins [Mol Cell Proteomics. 2014 Sep; 13(9): 2513–2526]. The reference is cited in the Materials and Method section associated with the proteomic analysis.

(4) To demonstrate long-term anti-caner memory was established through enhancement of CD8+ T cell activity, the "2nd seeding tumor cells" experiment may need to be done in CD8 antibody-treated IronA mice.

We have modified our claims to highlight that the tumor eradication by iron scavenging bacteria can establish adaptive anticancer immunity through the elicitation of CD8 T cells. We apologize for overstating our claim in the previous manuscript draft.

Minor suggestions:

(1) Please include the tumor re-challenge experiment in the method section.

The re-challenge experiment has been added to the method section as instructed.

(2) Please cite others' and your previous work. E.g. line 281, 282, line 306-307.

We have added the citations as instructed.

(3) Line 448, BL21 is bacteria, not cells.

We have made the correction accordingly.

Reviewer #3 (Recommendations For The Authors):

• The authors postulate that IroA-E. coli is more potent than DGC-E. coli in resisting LCN2 activity, and that this potency is the cause of the increased tumor suppression of this engineered strain. If so, Fig 3a should include DGC-E. coli for direct comparison.

We appreciate the reviewer for the comment and would like to clarify that we intended construct IroA-E. coli as a more specific iron-scavenging strategy, which can aide the discussion of nutritional immunity and minimize compounding factors from the immune-stimulatory effect of CDG. We have modified our text to clarify our stance.

• The data refers to the effects of WT bacteria-mediated tumor suppression, e.g. Figure 3e shows that even WT bacteria have a significant suppressive effect on tumor growth. Could the authors provide background on what is known about the mechanism of this tumor suppression, outside of tumor targeting and engineerability? They only reference "immune system stimulation."

We appreciate the reviewer’s comment and would like to refer the reviewer to our recently published article [Lim et al., EMBO Molecular Medicine 2024; DOI: 10.1038/s44321-023-00022-w], which shows that in addition to immune system stimulation, WT bacteria can also be perceived as an invading species in the tumor that can exert differential selective pressure against cancer cells. Competition for nutrient is highlighted as a major contribution to contain tumor growth. In fact, the nutrient competition that we observed in the prior article inspired the design of the iron scavenging bacteria towards overcoming nutritional immunity. We have cited this recently published article to the revised manuscript to enrich the background.

• The authors claim that there is immunologic memory because of tumor resistance in re-challenged mice after IroA-E. coli treatment (Fig 5c). It appears that the control group for this experiment is naive mice (and not WT-E. coli treated mice), in which case the immunologic memory could be from having had tumor/E. coli rather than the effect of IroA-E. coli.

We have modified our claims to highlight that the tumor eradication by iron scavenging bacteria can establish adaptive anticancer immunity through the elicitation of CD8 T cells. We did not intend to highlight that the adaptive immunity stemmed from IroA-E. coli only, and we intend to build upon current literature that has reported CD8+ T cell elicitation by bacterial therapy. The IroA-E.coli is shown to enhance adaptive immunity. We also did not consider a WT-E. coli control as no WT-E. coli treated group achieved complete tumor regression.

• The authors claim that CD8+ T cells are mechanistically important in the effects of iron status manipulation in E. coli-mediated tumor suppression (Fig 5). In order to show this, it seems that Fig 5c should include WT-E. coli and WT-E. coli+CD8 ab groups, as it may be that CD8+ T cells play a role in tumor suppression regardless of whether or not iron regulation is being manipulated.

We apologize for the confusion from our prior writing. We have modified our claims to highlight that the tumor eradication by iron scavenging bacteria can establish adaptive anticancer immunity through the elicitation of CD8 T cells. We did not intend to convey that CD8+ T cells are mechanistically important in the effects of iron status manipulation.

Author response:

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

eLife assessment

This useful manuscript challenges the utility of current paradigms for estimating brain-age with magnetic resonance imaging measures, but presents inadequate evidence to support the suggestion that an alternative approach focused on predicting cognition is more useful. The paper would benefit from a clearer explication of the methods and a more critical evaluation of the conceptual basis of the different models. This work will be of interest to researchers working on brain-age and related models.

Thank you so much for providing high-quality reviews on our manuscript. We revised the manuscript to address all of the reviewers’ comments and provided full responses to each of the comments below. Importantly, in this revision, we clarified that we did not intend to use Brain Cognition as an alternative approach. This is because, by design, the variation in fluid cognition explained by Brain Cognition should be higher or equal to that explained by Brain Age. Here we made this point more explicit and further stated that the relationship between Brain Cognition and fluid cognition indicates the upper limit of Brain Age’s capability in capturing fluid cognition. By examining what was captured by Brain Cognition, over and above Brain Age and chronological age via the unique effects of Brain Cognition, we were able to quantify the amount of co-variation between brain MRI and fluid cognition that was missed by Brain Age. And such quantification is the third aim of this study.

Public Reviews:

Reviewer 1 (Public Review):

In this paper, the authors evaluate the utility of brain-age-derived metrics for predicting cognitive decline by performing a 'commonality' analysis in a downstream regression that enables the different contribution of different predictors to be assessed. The main conclusion is that brain-age-derived metrics do not explain much additional variation in cognition over and above what is already explained by age. The authors propose to use a regression model trained to predict cognition ("brain-cognition") as an alternative suited to applications of cognitive decline. While this is less accurate overall than brain age, it explains more unique variance in the downstream regression.

(1) I thank the authors for addressing many of my concerns with this revision. However, I do not feel they have addressed them all. In particular I think the authors could do more to address the concern I raised about the instability of the regression coefficients and about providing enough detail to determine that the stacked regression models do not overfit.

Thank you Reviewer 1 for the comment. We addressed them in our response to Reviewer 1 Recommendations For The Authors #1 and #2 (see below).

(2) In considering my responses to the authors revision, I also must say that I agree with Reviewer 3 about the limitations of the brain age and brain cognition methods conceptually. In particular that the regression model used to predict fluid cognition will by construction explain more variance in cognition than a brain age model that is trained to predict age. To be fair, these conceptual problems are more widespread than this paper alone, so I do not believe the authors should be penalised for that. However, I would recommend to make these concerns more explicit in the manuscript

Thank you Reviewer 1 for the comment. We addressed them in our response to Reviewer 1 Recommendations For The Authors #3 (see below).

Reviewer 2 (Public Review):

In this study, the authors aimed to evaluate the contribution of brain-age indices in capturing variance in cognitive decline and proposed an alternative index, brain-cognition, for consideration.

The study employs suitable methods and data to address the research questions, and the methods and results sections are generally clear and easy to follow.

I appreciate the authors' efforts in significantly improving the paper, including some considerable changes, from the original submission. While not all reviewer points were tackled, the majority of them were adequately addressed. These include additional analyses, more clarity in the methods and a much richer and nuanced discussion. While recognising the merits of the revised paper, I have a few additional comments.

(1) Perhaps it would help the reader to note that it might be expected for brain-cognition to account for a significantly larger variance (11%) in fluid cognition, in contrast to brain-age. This stems from the fact that the authors specifically trained brain-cognition to predict fluid cognition, the very variable under consideration. In line with this, the authors later recommend that researchers considering the use of brain-age should evaluate its utility using a regression approach. The latter involves including a brain index (e.g. brain-cognition) previously trained to predict the regression's target variable (e.g. fluid cognition) alongside a brain-age index (e.g., corrected brain-age gap). If the target-trained brain index outperforms the brain-age metric, it suggests that relying solely on brain-age might not be the optimal choice. Although not necessarily the case, is it surprising for the target-trained brain index to demonstrate better performance than brain-age? This harks back to the broader point raised in the initial review: while brain-age may prove useful (though sometimes with modest effect sizes) across diverse outcomes as a generally applicable metric, a brain index tailored for predicting a specific outcome, such as brain-cognition in this case, might capture a considerably larger share of variance in that specific context but could lack broader applicability. The latter aspect needs to be empirically assessed.

Thank you so much for raising this point. Reviewer 1 (Public Review #2/Recommendations For The Authors #3) and Reviewer 3 (Recommendations for the Authors #1) made a similar observation. We now made changes to the introduction and discussion to address this concern (please see our responses to Reviewer 1 Recommendations For The Authors #3 below).

Briefly, as in our 2nd revision, we did not intend to compare Brain Age with Brain Cognition since, by design, the variation in fluid cognition explained by Brain Cognition should be higher or equal to that explained by Brain Age. Here we made this point more explicit and further stated that the relationship between Brain Cognition and fluid cognition indicates the upper limit of Brain Age’s capability in capturing fluid cognition. By examining what was captured by Brain Cognition, over and above Brain Age and chronological age via the unique effects of Brain Cognition, we were able to quantify the amount of co-variation between brain MRI and fluid cognition that was missed by Brain Age. And such quantification is the third aim of this study.

(2) Furthermore, the discussion pertaining to training brain-age models on healthy populations for subsequent testing on individuals with neurological or psychological disorders seems somewhat one-sided within the broader debate. This one-sidedness might potentially confuse readers. It is worth noting that the choice to employ healthy participants in the training model is likely deliberate, serving as a norm against which atypical populations are compared. To provide a more comprehensive understanding, referencing Tim Hans's counterargument to Bashyam's perspective could offer a more complete view (https://academic.oup.com/brain/article/144/3/e31/6214475?login=false).

Thank you Reviewer 2 for bringing up this issue. We have now revised the paragraph in question and added nuances on the usage of Brain Age for normative vs. case-control studies. We also cited Tim Hahn’s article that explained the conceptual foundation of the use of Brain Age in case-control studies. Please see below. Additionally, we also made a statement about our study not being able to address issues about the case-control studies directly in the newly written conclusion (see Reviewer 3 Recommendations for the Authors #3).

Discussion:

“There is a notable difference between studies investigating the utility of Brain Age in explaining cognitive functioning, including ours and others (e.g., Butler et al., 2021; Cole, 2020, 2020; Jirsaraie et al., 2023) and those explaining neurological/psychological disorders (e.g., Bashyam et al., 2020; Rokicki et al., 2021). We consider the former as a normative type of study and the latter as a case-control type of study (Insel et al., 2010; Marquand et al., 2016). Those case-control Brain Age studies focusing on neurological/psychological disorders often build age-prediction models from MRI data of largely healthy participants (e.g., controls in a case-control design or large samples in a population-based design), apply the built age-prediction models to participants without vs. with neurological/psychological disorders and compare Brain Age indices between the two groups. On the one hand, this means that case-control studies treat Brain Age as a method to detect anomalies in the neurological/psychological group (Hahn et al., 2021). On the other hand, this also means that case-control studies have to ignore under-fitted models when applied prediction models built from largely healthy participants to participants with neurological/psychological disorders (i.e., Brain Age may predict chronological age well for the controls, but not for those with a disorder). On the contrary, our study and other normative studies focusing on cognitive functioning often build age-prediction models from MRI data of largely healthy participants and apply the built age-prediction models to participants who are also largely healthy. Accordingly, the age-prediction models for explaining cognitive functioning in normative studies, while not allowing us to detect group-level anomalies, do not suffer from being under-fitted. This unfortunately might limit the generalisability of our study into just the normative type of study. Future work is still needed to test the utility of brain age in the case-control case.”

(3) Overall, this paper makes a significant contribution to the field of brain-age and related brain indices and their utility.

Thank you for the encouragement.

Reviewer 3 (Public Review):

The main question of this article is as follows: "To what extent does having information on brain-age improve our ability to capture declines in fluid cognition beyond knowing a person's chronological age?" This question is worthwhile, considering that there is considerable confusion in the field about the nature of brain-age.

(1) Thank you to the authors for addressing so many of my concerns with this revision. There are a few points that I feel still need addressing/clarifying related to 1) calculating brain cognition, 2) the inevitability of their results, and 3) their continued recommendation to use brain-age metrics.

Thank you Reviewer 3 for the comment. We addressed them in our response to Reviewer 3 Recommendations For The Authors #1-3 (see below).

Recommendations for the authors:

Reviewer 1 (Recommendations For The Authors):

(1) I do not feel the authors have fully addressed the concern I raised about the stacked regression models. Despite the new figure, it is still not entirely clear what the authors are using as the training set in the final step. To be clear, the problem occurs because of the parameters, not the hyperparameters (which the authors now state that they are optimising via nested grid search). in other words, given a regression model y = X*beta, if the X are taken to be predictions from a lower level regression model, then they contain information that is derived from both the training set at the test set for the model that this was trained on. If the split is the same (i.e. the predictions are derived on the same test set as is being used at the second level), then this can lead to overfitting. It is not clear to me whether the authors have done this or not. Please provide additional detail to clarify this point.

Thank you for allowing us an opportunity to clarify our stacked model. We wanted to confirm that we did not use test sets to build a stacked model in both lower and higher levels of the Elastic Net models. Test sets were there just for testing the performance of the models. We made additional clarification to make this clearer (see below). Let us explain what we did and provide the rationales below.

From Methods:

“We used nested cross-validation (CV) to build these prediction models (see Figure 7). We first split the data into five outer folds, leaving each outer fold with around 100 participants. This number of participants in each fold is to ensure the stability of the test performance across folds. In each outer-fold CV loop, one of the outer folds was treated as an outer-fold test set, and the rest was treated as an outer-fold training set. Ultimately, looping through the nested CV resulted in a) prediction models from each of the 18 sets of features as well as b) prediction models that drew information across different combinations of the 18 separate sets, known as “stacked models.” We specified eight stacked models: “All” (i.e., including all 18 sets of features), “All excluding Task FC”, “All excluding Task Contrast”, “Non-Task” (i.e., including only Rest FC and sMRI), “Resting and Task FC”, “Task Contrast and FC”, “Task Contrast” and “Task FC”. Accordingly, there were 26 prediction models in total for both Brain Age and Brain Cognition.

To create these 26 prediction models, we applied three steps for each outer-fold loop. The first step aimed at tuning prediction models for each of 18 sets of features. This step only involved the outer-fold training set and did not involve the outer-fold test set. Here, we divided the outer-fold training set into five inner folds and applied inner-fold CV to tune hyperparameters with grid search. Specifically, in each inner-fold CV, one of the inner folds was treated as an inner-fold validation set, and the rest was treated as an inner-fold training set. Within each inner-fold CV loop, we used the inner-fold training set to estimate parameters of the prediction model with a particular set of hyperparameters and applied the estimated model to the inner-fold validation set. After looping through the inner-fold CV, we, then, chose the prediction models that led to the highest performance, reflected by coefficient of determination (R2), on average across the inner-fold validation sets. This led to 18 tuned models, one for each of the 18 sets of features, for each outer fold.

The second step aimed at tuning stacked models. Same as the first step, the second step only involved the outer-fold training set and did not involve the outer-fold test set. Here, using the same outer-fold training set as the first step, we applied tuned models, created from the first step, one from each of the 18 sets of features, resulting in 18 predicted values for each participant. We, then, re-divided this outer-fold training set into new five inner folds. In each inner fold, we treated different combinations of the 18 predicted values from separate sets of features as features to predict the targets in separate “stacked” models. Same as the first step, in each inner-fold CV loop, we treated one out of five inner folds as an inner-fold validation set, and the rest as an inner-fold training set. Also as in the first step, we used the inner-fold training set to estimate parameters of the prediction model with a particular set of hyperparameters from our grid. We tuned the hyperparameters of stacked models using grid search by selecting the models with the highest R2 on average across the inner-fold validation sets. This led to eight tuned stacked models.

The third step aimed at testing the predictive performance of the 18 tuned prediction models from each of the set of features, built from the first step, and eight tuned stacked models, built from the second step. Unlike the first two steps, here we applied the already tuned models to the outer-fold test set. We started by applying the 18 tuned prediction models from each of the sets of features to each observation in the outer-fold test set, resulting in 18 predicted values. We then applied the tuned stacked models to these predicted values from separate sets of features, resulting in eight predicted values.

To demonstrate the predictive performance, we assessed the similarity between the observed values and the predicted values of each model across outer-fold test sets, using Pearson’s r, coefficient of determination (R2) and mean absolute error (MAE). Note that for R2, we used the sum of squares definition (i.e., R2 = 1 – (sum of squares residuals/total sum of squares)) per a previous recommendation (Poldrack et al., 2020). We considered the predicted values from the outer-fold test sets of models predicting age or fluid cognition, as Brain Age and Brain Cognition, respectively.”

Author response image 1.

Diagram of the nested cross-validation used for creating predictions for models of each set of features as well as predictions for stacked models.

Note some previous research, including ours (Tetereva et al., 2022), splits the observations in the outer-fold training set into layer 1 and layer 2 and applies the first and second steps to layers 1 and 2, respectively. Here we decided against this approach and used the same outer-fold training set for both first and second steps in order to avoid potential bias toward the stacked models. This is because, when the data are split into two layers, predictive models built for each separate set of features only use the data from layer 1, while the stacked models use the data from both layers 1 and 2. In practice with large enough data, these two approaches might not differ much, as we demonstrated previously (Tetereva et al., 2022).

(2) I also do not feel the authors have fully addressed the concern I raised about stability of the regression coefficients over splits of the data. I wanted to see the regression coefficients, not the predictions. The predictions can be stable when the coefficients are not.

The focus of this article is on the predictions. Still, as pointed out by reviewer 1, it is informative for readers to understand how stable the feature importance (i.e., Elastic Net coefficients) is. To demonstrate the stability of feature importance, we now examined the rank stability of feature importance using Spearman’s ρ (see Figure 4). Specifically, we correlated the feature importance between two prediction models of the same features, used in two different outer-fold test sets. Given that there were five outer-fold test sets, we computed 10 Spearman’s ρ for each prediction model of the same features. We found Spearman’s ρ to be varied dramatically in both age-prediction (range=.31-.94) and fluid cognition-prediction (range=.16-.84) models. This means that some prediction models were much more stable in their feature importance than others. This is probably due to various factors such as a) the collinearity of features in the model, b) the number of features (e.g., 71,631 features in functional connectivity, which were further reduced to 75 PCAs, as compared to 19 features in subcortical volume based on the ASEG atlas), c) the penalisation of coefficients either with ‘Ridge’ or ‘Lasso’ methods, which resulted in reduction as a group of features or selection of a feature among correlated features, respectively, and d) the predictive performance of the models. Understanding the stability of feature importance is beyond the scope of the current article. As mentioned by Reviewer 1, “The predictions can be stable when the coefficients are not,” and we chose to focus on the prediction in the current article.

Author response image 2.

Stability of feature importance (i.e., Elastic Net Coefficients) of prediction models. Each dot represents rank stability (reflected by Spearman’s ρ) in the feature importance between two prediction models of the same features, used in two different outer-fold test sets. Given that there were five outer-fold test sets, there were 10 Spearman’s ρs for each prediction model. The numbers to the right of the plots indicate the mean of Spearman’s ρ for each prediction model.

(3) I also must say that I agree with Reviewer 3 about the limitations of the brain-age and brain-cognition methods conceptually. In particular that the regression model used to predict fluid cognition will by construction explain more variance in cognition than a brain-age model that is trained to predict age. This suffers from the same problem the authors raise with brain-age and I agree that this would probably disappear if the authors had a separate measure of cognition against which to validate and were then to regress this out as they do for age correction. I am aware that these conceptual problems are more widespread than this paper alone (in fact throughout the brain-age literature), so I do not believe the authors should be penalised for that. However, I do think they can make these concerns more explicit and further tone down the comments they make about the utility of brain-cognition.

Thank you so much for raising this point. Reviewer 2 (Public Review #1) and Reviewer 3 (Recommendations for the Authors #1) made a similar observation. We now made changes to the introduction and discussion to address this concern (see below).

Briefly, we made it explicit that, by design, the variation in fluid cognition explained by Brain Cognition should be higher or equal to that explained by Brain Age. That is, the relationship between Brain Cognition and fluid cognition indicates the upper limit of Brain Age’s capability in capturing fluid cognition. More importantly, by examining what was captured by Brain Cognition, over and above Brain Age and chronological age via the unique effects of Brain Cognition, we were able to quantify the amount of co-variation between brain MRI and fluid cognition that was missed by Brain Age. And this is the third goal of this present study.

From Introduction:

“Third and finally, certain variation in fluid cognition is related to brain MRI, but to what extent does Brain Age not capture this variation? To estimate the variation in fluid cognition that is related to the brain MRI, we could build prediction models that directly predict fluid cognition (i.e., as opposed to chronological age) from brain MRI data. Previous studies found reasonable predictive performances of these cognition-prediction models, built from certain MRI modalities (Dubois et al., 2018; Pat et al., 2022; Rasero et al., 2021; Sripada et al., 2020; Tetereva et al., 2022; for review, see Vieira et al., 2022). Analogous to Brain Age, we called the predicted values from these cognition-prediction models, Brain Cognition. The strength of an out-of-sample relationship between Brain Cognition and fluid cognition reflects variation in fluid cognition that is related to the brain MRI and, therefore, indicates the upper limit of Brain Age’s capability in capturing fluid cognition. This is, by design, the variation in fluid cognition explained by Brain Cognition should be higher or equal to that explained by Brain Age. Consequently, if we included Brain Cognition, Brain Age and chronological age in the same model to explain fluid cognition, we would be able to examine the unique effects of Brain Cognition that explain fluid cognition beyond Brain Age and chronological age. These unique effects of Brain Cognition, in turn, would indicate the amount of co-variation between brain MRI and fluid cognition that is missed by Brain Age.”

From Discussion:

“Third, by introducing Brain Cognition, we showed the extent to which Brain Age indices were not able to capture the variation in fluid cognition that is related to brain MRI. More specifically, using Brain Cognition allowed us to gauge the variation in fluid cognition that is related to the brain MRI, and thereby, to estimate the upper limit of what Brain Age can do. Moreover, by examining what was captured by Brain Cognition, over and above Brain Age and chronological age via the unique effects of Brain Cognition, we were able to quantify the amount of co-variation between brain MRI and fluid cognition that was missed by Brain Age.

From our results, Brain Cognition, especially from certain cognition-prediction models such as the stacked models, has relatively good predictive performance, consistent with previous studies (Dubois et al., 2018; Pat et al., 2022; Rasero et al., 2021; Sripada et al., 2020; Tetereva et al., 2022; for review, see Vieira et al., 2022). We then examined Brain Cognition using commonality analyses (Nimon et al., 2008) in multiple regression models having a Brain Age index, chronological age and Brain Cognition as regressors to explain fluid cognition. Similar to Brain Age indices, Brain Cognition exhibited large common effects with chronological age. But more importantly, unlike Brain Age indices, Brain Cognition showed large unique effects, up to around 11%. As explained above, the unique effects of Brain Cognition indicated the amount of co-variation between brain MRI and fluid cognition that was missed by a Brain Age index and chronological age. This missing amount was relatively high, considering that Brain Age and chronological age together explained around 32% of the total variation in fluid cognition. Accordingly, if a Brain Age index was used as a biomarker along with chronological age, we would have missed an opportunity to improve the performance of the model by around one-third of the variation explained.”

Reviewer #3 (Recommendations For The Authors):

Thank you to the authors for addressing so many of my concerns with this revision. There are a few points that I feel still need addressing/clarifying related to: 1) calculating brain cognition, 2) the inevitability of their results, and 3) their continued recommendation to use brain age metrics.

(1) I understand your point here. I think the distinction is that it is fine to build predictive models, but then there is no need to go through this intermediate step of "brain-cognition". Just say that brain features can predict cognition XX well, and brain-age (or some related metric) can predict cognition YY well. It creates a confusing framework for the reader that can lead them to believe that "brain-cognition" is not just a predicted value of fluid cognition from a model using brain features to predict cognition. While you clearly state that that is in fact what it is in the text, which is a huge improvement, I do not see what is added by going through brain-cognition instead of simply just obtaining a change in R2 where the first model uses brain features alone to predict cognition, and the second adds on brain-age (or related metrics), or visa versa, depending on the question. Please do this analysis, and either compare and contrast it with going through "brain-cognition" in your paper, or switch to this analysis, as it more directly addresses the question of the incremental predictive utility of brain-age above and beyond brain features.

Thank you so much for raising this point. Reviewer 1 (Public Review #2/Recommendations For The Authors #3) and Reviewer 2 (Public Review #1) made a similar observation. We now made changes to the introduction and discussion to address this concern (see our responses to Reviewer 1 Recommendations For The Authors #3 above).

Briefly, as in our 2nd revision, we made it explicitly clear that we did not intend to compare Brain Age with Brain Cognition since, by design, the variation in fluid cognition explained by Brain Cognition should be higher or equal to that explained by Brain Age. And, by examining what was captured by Brain Cognition, over and above Brain Age and chronological age via the unique effects of Brain Cognition, we were able to quantify the amount of co-variation between brain MRI and fluid cognition that was missed by Brain Age.

We have thought about changing the name Brain Cognition into something along the lines of “predicted values of prediction models predicting fluid cognition based on brain MRI.” However, this made the manuscript hard to follow, especially with the commonality analyses. For instance, the sentence, “Here, we tested Brain Cognition’s unique effects in multiple regression models with a Brain Age index, chronological age and Brain Cognition as regressors to explain fluid cognition” would become “Here, we tested predicted values of prediction models predicting fluid cognition based on brain MRI unique effects in multiple regression models with a Brain Age index, chronological age and predicted values of prediction models predicting fluid cognition based on brain MRI as regressors to explain fluid cognition.” We believe, given our additional explanation (see our responses to Reviewer 1 Recommendations For The Authors #3 above), readers should understand what Brain Cognition is, and that we did not intend to compare Brain Age and Brain Cognition directly.

As for the suggested analysis, “obtaining a change in R2 where the first model uses brain features alone to predict cognition, and the second adds on brain-age (or related metrics), or visa versa,” we have already done this in the form of commonality analysis (Nimon et al., 2008) (see Figure 7 below). That is, to obtain unique and common effects of the regressors, we need to look at all of the possible changes in R2 when all possible subsets of regressors were excluded or included, see equations 12 and 13 below.

From Methods:

“Similar to the above multiple regression model, we had chronological age, each Brain Age index and Brain Cognition as the regressors for fluid cognition:

Fluid Cognitioni = β0 + β1 Chronological Agei + β2 Brain Age Indexi,j + β3 Brain Cognitioni + εi, (12)

Applying the commonality analysis here allowed us, first, to investigate the addictive, unique effects of Brain Cognition, over and above chronological age and Brain Age indices. More importantly, the commonality analysis also enabled us to test the common, shared effects that Brain Cognition had with chronological age and Brain Age indices in explaining fluid cognition. We calculated the commonality analysis as follows (Nimon et al., 2017):

Unique Effectchronological age = ΔR2chronological age = R2chronological age, Brain Age index, Brain Cognition – R2 Brain Age index, Brain Cognition

Unique EffectBrain Age index = ΔR2Brain Age index = R2chronological age, Brain Age index, Brain Cognition – R2 chronological age, Brain Cognition

Unique EffectBrain Cognition = ΔR2Brain Cognition = R2chronological age, Brain Age index, Brain Cognition – R2 chronological age, Brain Age Index

Common Effectchronological age, Brain Age index = R2chronological age, Brain Cognition + R2 Brain Age index, Brain Cognition – R2 Brain Cognition – R2chronological age, Brain Age index, Brain Cognition

Common Effectchronological age, Brain Cognition = R2chronological age, Brain Age Index + R2 Brain Age index, Brain Cognition – R2 Brain Age Index – R2chronological age, Brain Age index, Brain Cognition

Common Effect Brain Age index, Brain Cognition = R2chronological age, Brain Age Index + R2 chronological age, Brain Cognition – R2 chronological age – R2chronological age, Brain Age index, Brain Cognition

Common Effect chronological age, Brain Age index, Brain Cognition = R2 chronological age + R2 Brain Age Index + R2 Brain Cognition – R2chronological age, Brain Age Index – R2 chronological age, Brain Cognition – R2 Brain Age Index, Brain Cognition – R2chronological age, Brain Age index, Brain Cognition , (13)”

(2) I agree that the solution is not to exclude age as a covariate, and that there is a big difference between inevitable and obvious. I simply think a further discussion of the inevitability of the results would be clarifying for the readers. There is a big opportunity in the brain-age literature to be as direct as possible about why you are finding what you are finding. People need to know not only what you found, but why you found what you found.

Thank you. We agreed that we need to make this point more explicit and direct. In the revised manuscript, we had the statements in both Introduction and Discussion (see below) about the tight relationship between Brain Age and chronological age by design, making the small unique effects of Brain Age inevitable.

Introduction:

“Accordingly, by design, Brain Age is tightly close to chronological age. Because chronological age usually has a strong relationship with fluid cognition, to begin with, it is unclear how much Brain Age adds to what is already captured by chronological age.“

Discussion:

“First, Brain Age itself did not add much more information to help us capture fluid cognition than what we had already known from a person’s chronological age. This can clearly be seen from the small unique effects of Brain Age indices in the multiple regression models having Brain Age and chronological age as the regressors. While the unique effects of some Brain Age indices from certain age-prediction models were statistically significant, there were all relatively small. Without Brain Age indices, chronological age by itself already explained around 32% of the variation in fluid cognition. Including Brain Age indices only added around 1.6% at best. We believe the small unique effects of Brain Age were inevitable because, by design, Brain Age is tightly close to chronological age. Therefore, chronological age and Brain Age captured mostly a similar variation in fluid cognition.

Investigating the simple regression models and the commonality analysis between each Brain Age index and chronological age provided additional insights….”

(3) I believe it is very important to critically examine the use of brain-age and related metrics. As part of this process, I think we should be asking ourselves the following questions (among others): Why go through age prediction? Wouldn't the predictions of cognition (or another variable) using the same set of brain features always be as good or better? You still have not justified the use of brain-age. As I said before, if you are going to continue to recommend the use of brain-age, you need a very strong argument for why you are recommending this. What does it truly add? Otherwise, temper your statements to indicate possible better paths forward.

Thank you Reviewer 3 for making an argument against the use of Brain Age. We largely agree with you. However, our work only focuses on one phenotype, fluid cognition, and on the normative situation (i.e., not having a case vs control group). As Reviewer 2 pointed out, Brain Age might still have utility in other cases, not studied here. Still, future studies that focus on other phenotypes may consider using our approach as a template to test the utility of Brain Age in other situations. We added the conclusion statement to reflect this.

From Discussion:

“Altogether, we examined the utility of Brain Age as a biomarker for fluid cognition. Here are the three conclusions. First, Brain Age failed to add substantially more information over and above chronological age. Second, a higher ability to predict chronological age did not correspond to a higher utility to capture fluid cognition. Third, Brain Age missed up to around one-third of the variation in fluid cognition that could have been explained by brain MRI. Yet, given our focus on fluid cognition, future empirical research is needed to test the utility of Brain Age on other phenotypes, especially when Brain Age is used for anomaly detection in case-control studies (e.g., Bashyam et al., 2020; Rokicki et al., 2021). We hope that future studies may consider applying our approach (i.e., using the commonality analysis that includes predicted values from a model that directly predicts the phenotype of interest) to test the utility of Brain Age as a biomarker for other phenotypes.”

References

Bashyam, V. M., Erus, G., Doshi, J., Habes, M., Nasrallah, I. M., Truelove-Hill, M., Srinivasan, D., Mamourian, L., Pomponio, R., Fan, Y., Launer, L. J., Masters, C. L., Maruff, P., Zhuo, C., Völzke, H., Johnson, S. C., Fripp, J., Koutsouleris, N., Satterthwaite, T. D., … on behalf of the ISTAGING Consortium, the P. A. disease C., ADNI, and CARDIA studies. (2020). MRI signatures of brain age and disease over the lifespan based on a deep brain network and 14 468 individuals worldwide. Brain, 143(7), 2312–2324. https://doi.org/10.1093/brain/awaa160

Butler, E. R., Chen, A., Ramadan, R., Le, T. T., Ruparel, K., Moore, T. M., Satterthwaite, T. D., Zhang, F., Shou, H., Gur, R. C., Nichols, T. E., & Shinohara, R. T. (2021). Pitfalls in brain age analyses. Human Brain Mapping, 42(13), 4092–4101. https://doi.org/10.1002/hbm.25533

Cole, J. H. (2020). Multimodality neuroimaging brain-age in UK biobank: Relationship to biomedical, lifestyle, and cognitive factors. Neurobiology of Aging, 92, 34–42. https://doi.org/10.1016/j.neurobiolaging.2020.03.014

Dubois, J., Galdi, P., Paul, L. K., & Adolphs, R. (2018). A distributed brain network predicts general intelligence from resting-state human neuroimaging data. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1756), 20170284. https://doi.org/10.1098/rstb.2017.0284

Hahn, T., Fisch, L., Ernsting, J., Winter, N. R., Leenings, R., Sarink, K., Emden, D., Kircher, T., Berger, K., & Dannlowski, U. (2021). From ‘loose fitting’ to high-performance, uncertainty-aware brain-age modelling. Brain, 144(3), e31–e31. https://doi.org/10.1093/brain/awaa454

Insel, T., Cuthbert, B., Garvey, M., Heinssen, R., Pine, D. S., Quinn, K., Sanislow, C., & Wang, P. (2010). Research Domain Criteria (RDoC): Toward a New Classification Framework for Research on Mental Disorders. American Journal of Psychiatry, 167(7), 748–751. https://doi.org/10.1176/appi.ajp.2010.09091379

Jirsaraie, R. J., Kaufmann, T., Bashyam, V., Erus, G., Luby, J. L., Westlye, L. T., Davatzikos, C., Barch, D. M., & Sotiras, A. (2023). Benchmarking the generalizability of brain age models: Challenges posed by scanner variance and prediction bias. Human Brain Mapping, 44(3), 1118–1128. https://doi.org/10.1002/hbm.26144

Marquand, A. F., Rezek, I., Buitelaar, J., & Beckmann, C. F. (2016). Understanding Heterogeneity in Clinical Cohorts Using Normative Models: Beyond Case-Control Studies. Biological Psychiatry, 80(7), 552–561. https://doi.org/10.1016/j.biopsych.2015.12.023

Nimon, K., Lewis, M., Kane, R., & Haynes, R. M. (2008). An R package to compute commonality coefficients in the multiple regression case: An introduction to the package and a practical example. Behavior Research Methods, 40(2), 457–466. https://doi.org/10.3758/BRM.40.2.457

Pat, N., Wang, Y., Anney, R., Riglin, L., Thapar, A., & Stringaris, A. (2022). Longitudinally stable, brain‐based predictive models mediate the relationships between childhood cognition and socio‐demographic, psychological and genetic factors. Human Brain Mapping, hbm.26027. https://doi.org/10.1002/hbm.26027

Poldrack, R. A., Huckins, G., & Varoquaux, G. (2020). Establishment of Best Practices for Evidence for Prediction: A Review. JAMA Psychiatry, 77(5), 534–540. https://doi.org/10.1001/jamapsychiatry.2019.3671

Rasero, J., Sentis, A. I., Yeh, F.-C., & Verstynen, T. (2021). Integrating across neuroimaging modalities boosts prediction accuracy of cognitive ability. PLOS Computational Biology, 17(3), e1008347. https://doi.org/10.1371/journal.pcbi.1008347

Rokicki, J., Wolfers, T., Nordhøy, W., Tesli, N., Quintana, D. S., Alnæs, D., Richard, G., de Lange, A.-M. G., Lund, M. J., Norbom, L., Agartz, I., Melle, I., Nærland, T., Selbæk, G., Persson, K., Nordvik, J. E., Schwarz, E., Andreassen, O. A., Kaufmann, T., & Westlye, L. T. (2021). Multimodal imaging improves brain age prediction and reveals distinct abnormalities in patients with psychiatric and neurological disorders. Human Brain Mapping, 42(6), 1714–1726. https://doi.org/10.1002/hbm.25323

Sripada, C., Angstadt, M., Rutherford, S., Taxali, A., & Shedden, K. (2020). Toward a “treadmill test” for cognition: Improved prediction of general cognitive ability from the task activated brain. Human Brain Mapping, 41(12), 3186–3197. https://doi.org/10.1002/hbm.25007

Tetereva, A., Li, J., Deng, J. D., Stringaris, A., & Pat, N. (2022). Capturing brain‐cognition relationship: Integrating task‐based fMRI across tasks markedly boosts prediction and test‐retest reliability. NeuroImage, 263, 119588. https://doi.org/10.1016/j.neuroimage.2022.119588

Vieira, B. H., Pamplona, G. S. P., Fachinello, K., Silva, A. K., Foss, M. P., & Salmon, C. E. G. (2022). On the prediction of human intelligence from neuroimaging: A systematic review of methods and reporting. Intelligence, 93, 101654. https://doi.org/10.1016/j.intell.2022.101654

Author response:

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

The reviewers praised multiple aspects of our study. Reviewer 1 noted that “the work aligns well with current research trends and will greatly interest researchers in the field.” Reviewer 2 highlighted the unique capability of our imaging approach, which “allows for investigation of the heterogeneity of response across individual dopamine axons, unlike other common approaches such as fiber photometry.” Reviewer 3 commented that “the experiments are beautifully executed” and “are revealing novel information about how aversive and rewarding stimuli is encoded at the level of individual axons, in a way that has not been done before.”

In addition to the positive feedback, the reviewers also provided useful criticisms and suggestions, some of which may not be fully addressed in a single study. For instance, questions regarding whether dopamine axons encode the valence or specific identity of the stimuli, or the most salient aspects of the environment, remain open. At the same time, as all the reviewers agreed, our report on the diversity of dopamine axonal responses using a novel imaging design introduces significant new insights to the neuroscience community. Following the reviewers’ recommendations, we have refrained from making interpretations that could be perceived as overinterpretation, such as concluding that “dopamine axons are involved in aversive processing.” This has necessitated extensive revisions, including modifying the title of our manuscript to make clear that the novelty of our work is revealing ‘functional diversity’ using our new imaging approach.

Below, we respond to the reviewers’ comments point by point.

eLife assessment

This valuable study shows that distinct midbrain dopaminergic axons in the medial prefrontal cortex respond to aversive and rewarding stimuli and suggest that they are biased toward aversive processing. The use of innovative microprism based two-photon calcium imaging to study single axon heterogeneity is solid, although the experimental design could be optimized to distinguish aversive valence from stimulus salience and identity in this dopamine projection. This work will be of interest to neuroscientists working on neuromodulatory systems, cortical function and decision making.

Reviewer #1

Summary:

In this manuscript, Abe and colleagues employ in vivo 2-photon calcium imaging of dopaminergic axons in the mPFC. The study reveals that these axons primarily respond to unconditioned aversive stimuli (US) and enhance their responses to initially-neutral stimuli after classical association learning. The manuscript is well-structured and presents results clearly. The utilization of a refined prism-based imaging technique, though not entirely novel, is well-implemented. The study's significance lies in its contribution to the existing literature by offering single-axon resolution functional insights, supplementing prior bulk measurements of calcium or dopamine release. Given the current focus on neuromodulator neuron heterogeneity, the work aligns well with current research trends and will greatly interest researchers in the field.

However, I would like to highlight that the authors could further enhance their manuscript by addressing study limitations more comprehensively and by providing essential details to ensure the reproducibility of their research. In light of this, I have a number of comments and suggestions that, if incorporated, would significantly contribute to the manuscript's value to the field.

Strengths:

• Descriptive.

• Utilization of a well-optimized prism-based imaging method.

• Provides valuable single-axon resolution functional observations, filling a gap in existing literature.

• Timely contribution to the study of neuromodulator neuron heterogeneity.

We thank the reviewer for this positive assessment.

Weaknesses:

(1) It's important to fully discuss the fact that the measurements were carried out only on superficial layers (30-100um), while major dopamine projections target deep layers of the mPFC as discussed in the cited literature (Vander Weele et al., 2018) and as illustrated in FigS1B,C. This limitation should be explicitly acknowledged and discussed in the manuscript, especially given the potential functional heterogeneity among dopamine neurons in different layers. This potential across-layer heterogeneity could also be the cause of discrepancy among past recording studies with different measurement modalities. Also, mentioning technical limitations would be informative. For example: how deep the authors can perform 2p-imaging through the prism? was the "30-100um" maximum depth the authors could get?

Thank you for pointing out this important issue about layer differences.

It is possible that the mesocortial pathway has layer-specific channels, with some neurons targeting supra granular layers and others targeting infragranular ones. Alternatively, it is also plausible that the axons of the same neurons branch into both superficial and deep layers. This is a critical issue that has not been investigated in anatomical studies and will require single-cell labeling of dopamine neurons (Matsuda et al 2009 and Aransay et al 2015). We now discuss this issue in the Discussion.

As for the imaging depth of 30–100 m, we were unable to visualize deeper axons in a live view mode. Our imaging system has already been optimized to detect weak signals (e.g., we have employed an excitation wavelength of 980 nm, dispersion compensation, and a hybrid photodetector). It is possible that future studies using improved imaging approaches may be able to visualize deeper layers. Importantly, sparse axons in the supragranular layers are advantageous in detecting weak signals; dense labeling of axons would increase the background fluorescence relative to signals. We now reference this layer issue in the Results and Discussion sections.

(2) In the introduction, it seems that the authors intended to refer to Poulin et al. 2018 regarding molecular/anatomical heterogeneity of dopamine neurons, but they inadvertently cited Poulin et al. 2016 (a general review on scRNAseq). Additionally, the statement that "dopamine neurons that project to the PFC show unique genetic profiles (line 85)" requires clarification, as Poulin et al. 2018 did not specifically establish this point. Instead, they found at least the Vglut2/Cck+ population projects into mPFC, and they did not reject the possibility of other subclasses projecting to mPFC. Rather, they observed denser innervation with DAT-cre, suggesting that non-Vglut2/Cck populations would also project to mPFC. Discuss the potential molecular heterogeneity among mPFC dopamine axons in light of the sampling limitation mentioned earlier.

We thank the reviewer for pointing this out. Genetic profiles of PFC-projecting DA neurons are still being investigated, so describing them as “unique” was misleading. We have edited the Introduction accordingly, and now discuss this issue in detail in the Discussion.

(3) I find the data presented in Figure 2 to be odd. Firstly, the latency of shock responses in the representative axons (right panels of G, H) is consistently very long - nearly 500ms. It raises a query whether this is a biological phenomenon or if it stems from a potential technical artifact, possibly arising from an issue in synchronization between the 2-photon imaging and stimulus presentation. My reservations are compounded by the notable absence of comprehensive information concerning the synchronization of the experimental system in the method section.

The synchronization of the stimulus and data acquisition is accomplished at a sub-millisecond resolution. We use a custom-made MATLAB program that sends TTL commands to standard imaging software (ThorImage or ScanImage) and a stimulator for electrical shocks. All events are recorded as analogue inputs to a different DAQ to ensure synchronization. We have provided additional details regarding the configuration in the Methods section.

We consider that the long latency of shock response is biological. For instance, a similar long latency was found after electrical shock in a photometry imaging study (Kim, …, Deisseroth, 2016).

Secondly, there appear to be irregularities in Panel J. While the authors indicate that "Significant axons were classified as either reward-preferring (cyan) or aversive-preferring (magenta), based on whether the axons are above or below the unity line of the reward/aversive scatter plot (Line 566)," a cyan dot slightly but clearly deviates above the unity line (around coordinates (x, y) = (20, 21)). This needs clarification. Lastly, when categorizing axons for analysis of conditioning data in Fig3 (not Fig2), the authors stated "The color-coded classification (cyan/magenta) was based on k-means clustering, using the responses before classical conditioning (Figure 2J)". I do not understand why the authors used different classification methods for two almost identical datasets.

We thank the reviewer for pointing out these insufficient descriptions. We classified the axons using k-means clustering, and the separation of the two clusters happened to roughly coincide with the unity line of the reward/aversive scatter plot in Fig 2J. In other words, we did not use the unity line to classify the data points (which is why the color separation of the histogram is not at 45 degrees). We have clarified this point in the Methods section.

(4) In connection with Point 3, conducting separate statistical analyses for aversive and rewarding stimuli would offer a fairer approach. This could potentially reveal a subset of axons that display responses to both aversive and appetitive stimuli, aligning more accurately with the true underlying dynamics. Moreover, the characterization of Figure 2J as a bimodal distribution while disregarding the presence of axons responsive to both aversive and appetitive cues seems somewhat arbitrary and circular logic. A more inclusive consideration of this dual-responsive population could contribute to a more comprehensive interpretation.

We also attempted k-means clustering with additional dimensions (e.g., temporal domains as shown in Fig. 3I, J), but no additional clusters were evident. We note that the lack of other clusters does not exclude the possibility of their existence, which may only become apparent with a substantial increase in the number of samples. In the current report, we present the clusters that were the easiest/simplest for us to identify.

Additionally, we have revised our manuscript to reflect that many axons respond to both reward and aversive stimuli, and that aversive-preferring axons do not exclusively respond to the aversive stimulus.

(5) The contrast in initialization to novel cues between aversive and appetitive axons mirrors findings in other areas, such as the tail-of-striatum (TS) and ventral striatum (VS) projecting dopamine neurons (Menegas et al., 2017, not 2018). You might consider citing this very relevant study and discussing potential collateral projections between mPFC and TS or VS.

Thank you for pointing this out. We have now included Menegas et al., 2017, and also discuss the possibility of collaterals to these areas. In addition, we also referred to Azcorra et al., 2023 - this was published after our initial submission.

(6) The use of correlation values (here >0.65) to group ROIs into axons is common but should be justified based on axon density in the FOV and imaging quality. It's important to present the distribution of correlation values and demonstrate the consistency of results with varying cut-off values. Also, provide insights into the reliability of aversive/appetitive classifications for individual ROIs with high correlations. Importantly, if you do the statistical testing and aversive/appetitive classifications for individual ROIs with above-threshold high correlation (to be grouped into the same axon), do they always fall into the same category? How many false positives/false negatives are observed?


"Our results remained similar for different correlation threshold values (Line 556)" (data not shown) is obsolete.

We have conducted additional analysis using correlation values 0.5 and 0.3 that resulted in a smaller number of axon terminals. In essence, the relationship between reward responses and aversive responses remained very similar to Fig. 2J, K.

Author response image 1.

Reviewer #2 (Public Review):

Summary:

This study aims to address existing differences in the literature regarding the extent of reward versus aversive dopamine signaling in the prefrontal cortex. To do so, the authors chose to present mice with both a reward and an aversive stimulus during different trials each day. The authors used high spatial resolution two-photon calcium imaging of individual dopaminergic axons in the medial PFC to characterize the response of these axons to determine the selectivity of responses in unique axons. They also paired the reward (water) and an aversive stimulus (tail shock) with auditory tones and recorded across 12 days of associative learning.

The authors find that some axons respond to both reward and aversive unconditioned stimuli, but overall, there is a strong preference to respond to aversive stimuli consistent with expectations from prior studies that used other recording methods. The authors find that both of their two auditory stimuli initially drive responses in axons, but that with training axons develop more selective responses for the shock associated tone indicating that associative learning led to changes in these axon's responses. Finally, the authors use anticipatory behaviors during the conditioned stimuli and facial expressions to determine stimulus discrimination and relate dopamine axons signals with this behavioral evidence of discrimination. This study takes advantage of cutting-edge imaging approaches to resolve the extent to which dopamine axons in PFC respond appetitive or aversive stimuli. They conclude that there is a strong bias to respond to the aversive tail shock in most axons and weaker more sparse representation of water reward.

Strengths:

The strength of this study is the imaging approach that allows for investigation of the heterogeneity of response across individual dopamine axons, unlike other common approaches such as fiber photometry which provide a measure of the average population activity. The use of appetitive and aversive stimuli to probe responses across individual axons is another strength.

We thank the reviewer for this positive assessment.

Weaknesses:

A weakness of this study is the design of the associative conditioning paradigm. The use of only a single reward and single aversive stimulus makes it difficult to know whether these results are specific to the valence of the stimuli versus the specific identity of the stimuli. Further, the reward presentations are more numerous than the aversive trials making it unclear how much novelty and habituation account for results. Moreover, the training seems somewhat limited by the low number of trials and did not result in strong associative conditioning. The lack of omission responses reported may reflect weak associative conditioning. Finally, the study provides a small advance in our understanding of dopamine signaling in the PFC and lacks evidence for if and what might be the consequence of these axonal responses on PFC dopamine concentrations and PFC neuron activity.

We thank the reviewer for the suggestions.

We agree that interpreting the response change during classical conditioning is not straightforward. Although the reward and aversive stimuli we employed are commonly used in the field, future studies with more sophisticated paradigms will be necessary to address whether dopamine axons encode the valence of the stimuli, the specific identity of the stimuli, or novelty and habituation. In our current manuscript, we refrain from making a conclusion that distinct groups of neurons encode different valances. In fact, many axons respond to both stimuli, at different ratios. We have removed descriptions that may suggest exclusive coding of reward or aversive processing. Additionally, we have extensively discussed possible interpretations.

In terms of the strength of the conditioning association, behavioral results indicated that the learning plateaued – anticipatory behaviors did not increase during the last two phases when the conditioned span was divided into six phases (Figure 3–figure supplement 1).

Our goal in the current manuscript is to provide new insight into the functional diversity of dopamine axons in the mPFC. Investigating the impact of dopamine axons on local dopamine concentration and neural activity in the mPFC is important but falls beyond the scope of our current study. In particular, given the functional diversity of dopamine axons, interpreting bulk optogenetic or chemogenetic axonal manipulation experiments would not be straightforward. As suggested, measuring the dopamine concentration through two-photon imaging of dopamine sensors and monitoring the activity of dopamine recipient neurons (e.g., D1R- or D2R-expressing neurons) is a promising approach that we plan to undertake in the near future.

Reviewer #3 (Public Review):

Summary:

The authors image dopamine axons in medial prefrontal cortex (mPFC) using microprism-mediated two-photon calcium imaging. They image these axons as mice learn that two auditory cues predict two distinct outcomes, tailshock or water delivery. They find that some axons show a preference for encoding of the shock and some show a preference for encoding of water. The authors report a greater number of dopamine axons in mPFC that respond to shock. Across time, the shock-preferring axons begin to respond preferentially to the cue predicting shock, while there is a less pronounced increase in the water-responsive axons that acquire a response to the water-predictive cue (these axons also increase non-significantly to the shock-predictive cue). These data lead the authors to argue that dopamine axons in mPFC preferentially encode aversive stimuli.

Strengths:

The experiments are beautifully executed and the authors have mastered an impressively complex technique. Specifically, they are able to image and track individual dopamine axons in mPFC across days of learning. This technique is used the way it should be: the authors isolate distinct dopamine axons in mPFC and characterize their encoding preferences and how this evolves across learning of cue-shock and cue-water contingencies. Thus, these experiments are revealing novel information about how aversive and rewarding stimuli is encoded at the level of individual axons, in a way that has not been done before. This is timely and important.

We thank the reviewer for this positive assessment.

Weaknesses:

The overarching conclusion of the paper is that dopamine axons preferentially encode aversive stimuli. This is prevalent in the title, abstract, and throughout the manuscript. This is fundamentally confounded. As the authors point out themselves, the axonal response to stimuli is sensitive to outcome magnitude (Supp Fig 3). That is, if you increase the magnitude of water or shock that is delivered, you increase the change in fluorescence that is seen in the axons. Unsurprisingly, the change in fluorescence that is seen to shock is considerably higher than water reward.

We agree that the interpretation of our results is not straightforward. Our current manuscript now focuses on our strength, which is reporting the functional diversity of dopamine axons. Therefore, we avoid using the word ‘encode’ when describing the response.

We believe that our results could reconcile the apparent discrepancy as to why some previous studies reported only aversive responses while others reported reward responses. In particular, if the reward volume were very small, the reward response could go undetected.

Further, when the mice are first given unexpected water delivery and have not yet experienced the aversive stimuli, over 40% of the axons respond [yet just a few lines below the authors write: "Previous studies have demonstrated that the overall dopamine release at the mPFC or the summed activity of mPFC dopamine axons exhibits a strong response to aversive stimuli (e.g., tail shock), but little to rewards", which seems inconsistent with their own data].

We always recorded the reward and aversive response together, which might have confused the reviewer. Therefore, there is no inconsistency in our data. We have clarified our methods and reasoning accordingly.

Given these aspects of the data, it could be the case that the dopamine axons in mPFC encodes different types of information and delegates preferential processing to the most salient outcome across time.

This is certainly an exciting interpretation, so we have included it in our discussion. Meanwhile, ‘the most salient outcome’ alone cannot fully capture the diverse response patterns of the dopaminergic axons, particularly reward-preferring axons. We discuss our findings in more detail in the revised manuscript.

The use of two similar sounding tones (9Khz and 12KHz) for the reward and aversive predicting cues are likely to enhance this as it requires a fine-grained distinction between the two cues in order to learn effectively. There is considerable literature on mPFC function across species that would support such a view. Specifically, theories of mPFC function (in particular prelimbic cortex, which is where the axon images are mostly taken) generally center around resolution of conflict in what to respond, learn about, and attend to. That is, mPFC is important for devoting the most resources (learning, behavior) to the most relevant outcomes in the environment. This data then, provides a mechanism for this to occur in mPFC. That is, dopamine axons signal to the mPFC the most salient aspects of the environment, which should be preferentially learned about and responded towards. This is also consistent with the absence of a negative prediction error during omission: the dopamine axons show increases in responses during receipt of unexpected outcomes, but do not encode negative errors. This supports a role for this projection in helping to allocate resources to the most salient outcomes and their predictors, and not learning per se. Below are a just few references from the rich literature on mPFC function (some consider rodent mPFC analogous to DLPFC, some mPFC), which advocate for a role in this region in allocating attention and cognitive resources to most relevant stimuli, and do not indicate preferential processing of aversive stimuli.

Distinguishing between 9 kHz and 12 kHz sound tones may not be that difficult, considering anticipatory licking and running are differentially manifested. In addition, previous studies have shown that mice can distinguish between two sound tones when they are separated by 7% (de Hoz and Nelken 2014). Nonetheless, we agree with the attractive interpretation that “the mPFC devotes the most resources (learning, behavior) to the most relevant outcomes in the environment” and that dopamine is a mechanism for this. Therefore, we discuss this interpretation in the revised text.

References:

(1) Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Annual review of neuroscience, 24(1), 167-202.

(2) Bissonette, G. B., Powell, E. M., & Roesch, M. R. (2013). Neural structures underlying set-shifting: roles of medial prefrontal cortex and anterior cingulate cortex. Behavioural brain research, 250, 91101.

(3) Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention. Annual review of neuroscience, 18(1), 193-222.

(4) Sharpe, M. J., Stalnaker, T., Schuck, N. W., Killcross, S., Schoenbaum, G., & Niv, Y. (2019). An integrated model of action selection: distinct modes of cortical control of striatal decision making. Annual review of psychology, 70, 53-76.

(5) Ridderinkhof, K. R., Ullsperger, M., Crone, E. A., & Nieuwenhuis, S. (2004). The role of the medial frontal cortex in cognitive control. science, 306(5695), 443-447.

(6) Nee, D. E., Kastner, S., & Brown, J. W. (2011). Functional heterogeneity of conflict, error, taskswitching, and unexpectedness effects within medial prefrontal cortex. Neuroimage, 54(1), 528-540.

(7) Isoda, M., & Hikosaka, O. (2007). Switching from automatic to controlled action by monkey medial frontal cortex. Nature neuroscience, 10(2), 240-248.

Reviewer #1 (Recommendations For The Authors):

Specific Suggestions and Questions on the Methods Section:

In general, the methods part is not well documented and sometimes confusing. Thus, as it stands, it hinders reproducible research. Specific suggestions/questions are listed in the following section.

(1) Broussard et al. 2018 introduced axon-GCaMP6 instead of axon-jGCaMP8m. The authors should provide details about the source of this material. If it was custom-made, a description of the subcloning process would be appreciated. Additionally, consider depositing sequence information or preferably the plasmid itself. Furthermore, the introduction of the jGCaMP8 series by Zhang, Rozsa, et al. 2023 should be acknowledged and referenced in your manuscript.

We thank the reviewer for pointing this out. We have now included details on how we prepared the axon-jGCaMP8m, which was based on plasmids available at Addgene. Additionally, we have deposited our construct to Addgene ( https://www.addgene.org/216533/ ). We have also cited Janelia’s report on jGCaMP8, Zhang et al.

(2) The authors elaborate on the approach taken for experimental synchronization. Specifically, how was the alignment achieved between 2-photon imaging, treadmill recordings, aversive/appetitive stimuli, and videography? It would be important to document the details of the software and hardware components employed for generating TTLs that trigger the pump, stimulator, cameras, etc.

We have now included a more detailed explanation about the timing control. We utilize a custommade MATLAB program that sends TTL square waves and analogue waves via a single National Instruments board (USB-6229) to control two-photon image acquisition, behavior camera image acquisition, water syringe movement, current flow from a stimulator, and sound presentation. We also continuously recorded at 30 kHz via a separate National Instrument board (PCIe-6363) the frame timing of two-photon imaging, the frame timing of a behavior camera, copies of command waves (sent to the syringe pump, the stimulator, and the speaker), and signals from the treadmill corresponding to running speed.

(3) The information regarding the cameras utilized in the study presents some confusion. In one instance, you mention, "To monitor licking behavior, the face of each mouse was filmed with a camera at 60 Hz (CM3-U3-13Y3M-CS, FLIR)" (Line 488). However, there's also a reference to filming facial expressions using an infrared web camera (Line 613). Could you clarify whether the FLIR camera (which is an industrial CMOS not a webcam) is referred to as a webcam? Alternatively, if it's a different camera being discussed, please provide product details, including pixel numbers and frame rate for clarity.

We thank the reviewer for pointing this out. This was a mistake on our end. The camera used in the current project was a CM3-U3-13Y3M-CS, not a web camera. We have now corrected this.

(4) Please provide more information about the methodology employed for lick detection. Specifically, did the authors solely rely on videography for this purpose? If so, why was an electrical (or capacitive) detector not used? It would provide greater accuracy in detecting licking.

Lick detection was performed offline based on videography, using DeepLabCut. As licking occurs at a frequency of ~6.5 Hz (Xu, …, O’Connor Nature Neurosci, 2022), the movement can be detected at a frame rate of 60 Hz. Initially, we used both a lick sensor and videography. However, we favored videography because it could potentially provide non-binary information.

Other Minor Points:

(5) Ensure consistency in the citation format; both Vander Weele et al. 2018 and Weele et al. 2019, share the same first author.

Thank you for pointing this out. Endnote processes the first author’s name differently depending on the journal. We fixed the error manually. The first paper (2018) is an original research paper, and the second one (2019) is a review about how dopamine modulates aversive processing in the mPFC. We cited the second one in three instances where we mentioned review papers.

(6) The distinction between "dashed vs dotted lines" in Figure 3K and 3M appears to be very confusing. Please consider providing a clearer visualization/labeling to mitigate this confusion.

We have now changed the line styles.

(7) Additionally plotting mean polar angles of aversive/appetitive axons as vectors in the Cartesian scatter plots (2J, 3I,J) would make interpretation easier.

We have now made this change to Figures 2, 3, 4.

(8) Data and codes should be shared in a public database. This is important for reproducible research and we believe that "available from the corresponding author upon reasonable request" is outdated language.

We have uploaded the data to GitHub, https://github.com/pharmedku/2024-elife-da-axon.

Reviewer #2 (Recommendations For The Authors):

(1) Authors don't show which mouse each axon data comes from making it hard to know if differences arise from inter-mouse differences vs differences in axons. The best way to address this point is to show similar plots as Figure 2J & K but broken down by mouse to shows whether each mouse had evidence of these two clusters.

We have now made this change to Figure 2-figure supplement 3.

(2) Line 166: Should this sentence point to panels 2F, G, H rather than 2I which doesn't show a shock response?

We thank the reviewer for pointing this out. We have fixed the incorrect labels.

Line 195: The population level bias to aversive stimuli was shown previously using photometry so it is not justified to say "for the first time" regarding this statement.

We have adjusted this sentences so the claim of ”for the first time” is not associated with the population-level bias.

(4) The paper lacks a discussion of the potential role that novelty plays in the amplitude of the responses given that tail shocks occur less often that rewards. Is the amplitude of the first reward of the day larger than subsequent rewards? Would tail shock responses decay if they occurred in sequential trials?

Following the reviewer's suggestion, we conducted a comparison of individual axonal responses to both conditioned and unconditioned stimuli across the first trial and subsequent trials. Our findings reveal a notable trend: aversive-preferring axons exhibited attenuation in response to CSreward, yet enhancement in response to CSaversive. Conversely, the response of these axons to USreward was attenuated, with no significant change observed for USaversive. In contrast, reward-preferring axons displayed an invariable activity pattern from the initial trial, highlighting the functional diversity present within dopamine axons. This analysis has been integrated into Figure 3-figure supplement 4 and is elaborated upon in the Discussion section.

(5) Fix typo in Figure 1 - supplement 1. Shift

We have now corrected this. Thank you.

(6) The methods section needs information about trial numbers. Please indicate how many trials were presented to each mouse per day.

We have now added the information about trial numbers to the Methods section.

Reviewer #3 (Recommendations For The Authors):

In line with the public review, my recommendation is for the authors to remain as objective about their data as possible. There are many points in the manuscript where the authors seem to directly contradict their own data. For example, they first detail that dopamine axons respond to unexpected water rewards. Indeed, they find that there are 40% of dopamine axons that respond in this way. Then, a few paragraphs later they state: "Previous studies have demonstrated that the overall dopamine release at the mPFC or the summed activity of mPFC dopamine axons exhibits a strong response to aversive stimuli (e.g., tail shock), but little to rewards". As detailed above, I do not think these data support an idea that dopamine axons in mPFC preferentially encode aversive outcomes. If the authors wanted to examine a role for mPFC in preferential encoding of aversive stimuli, you would first have to equate the outcomes by magnitude and then compare how the axons acquire preferences across time. Alternatively, a prediction of a more general process that I detail above would predict that you could give mice two rewards that differ in magnitude (e.g., lots of food vs. small water) and you would see the same results that the authors have seen here (i.e., a preference for the food, which is the larger and more salient outcome). Without other tests of how dopamine axons in mPFC respond to situations like this, I don't think any conclusion around mPFC in favoring aversive stimuli can be made.

As suggested, we have made the current manuscript as objective as possible, removing interpretation aspects regarding what dopamine axons encode and emphasizing their functional diversity. In particular, we remove the word ‘encode’ when describing the response of dopamine axons.

Although it may have appeared unclear, there was no contradiction within our data regarding the response to reward and aversive stimuli. We have now improved the readability of the Results and Methods sections. Concerning the interpretation of what exactly the mPFC dopamine axons encode, we have rewritten the discussion to be as objective about our data as possible, as suggested. We also have edited our title and abstract accordingly. Meanwhile, we wish to emphasize that our reward and aversive stimuli are standard paradigms commonly used in the field. We believe, and all the reviewers agreed, that reporting the diversity of dopamine axonal responses with a novel imaging design constitutes new insight for the neuroscience community. Therefore, we have decided to leave the introduction of new behavioral tasks for future studies and instead expanded our discussion.

As mentioned, I think the experiments are executed really well and the technological aspects of the authors' methods are impressive. However, there are also some aspects of the data presentation that would be improved. Some of the graphs took a considerable amount of effort to unpack. For example, Figure 4 is hard going. Is there a way to better illustrate the main points that this figure wants to convey? Some of this might be helped by a more complete description in the figure captions about what the data are showing. It would also be great to see how the response of dopamine axons changes across trial within a session to the shock and water-predictive cues. Supp Figure 1 should be in the main text with standard error and analyses across time. Clarifying these aspects of the data would make the paper more relevant and accessible to the field.

We thank the reviewer for pointing out that the legend of Figure 4 was incomplete. We have fixed it, along with improving the presentation of the figure. We have also prepared a new figure (Figure 3– figure supplement 4) to compare CSaversive and CSreward signals for the first and rest of the trials within daily sessions, revealing further functional diversity in dopamine axons. We have decided to keep Figure 1–figure supplement 2 as a figure supplement with an additional analysis, as another reviewer pointed out that the design is not completely new. Furthermore, as eLife readers can easily access figure supplements, we believe it is appropriate to maintain it in this way.

Minor points:

(1) What is the control period for the omission test? Was omission conducted for the shock?

The control period for reward omission is a 2-second period just before the CS onset. We did not include shock omission, because a sufficient number of trials (> 6 trials) for the rare omission condition could not be achieved within a single day.

(2) The authors should mention how similar the tones were that predicted water and shock.

According to de Hoz and Nelken (2014), a frequency difference of 4–7% is enough for mice to discriminate between tones. In addition, anticipatory licking and running confirmed that the mice could discriminate between the frequencies. We have now included this information in the Discussion.

(3) I realize the viral approach used in the current studies may not allow for an idea of where in VTA dopamine neurons are that project to mPFC- is there data in the literature that speak to this? Particularly important as we now know that there is considerable heterogeneity in dopamine neuronal responses, which is often captured by differences in medial/lateral position within VTA.

Some studies have suggested that mesocortical dopamine neurons are located in the medial posterior VTA (e.g., Lammel et al., 2008). However, in mouse anterograde tracing, it is not possible to spatially confine the injection of conventional viruses/tracers. We now refer to Lammel et al., 2008 in the Introduction.

Author response:

eLife assessment

This study provides valuable information on the mechanism of PepT2 through enhanced-sampling molecular dynamics, backed by cell-based assays, highlighting the importance of protonation of selected residues for the function of a proton-coupled oligopeptide transporter (hsPepT2). The molecular dynamics approaches are convincing, but with limitations that could be addressed in the manuscript, including lack of incorporation of a protonation coordinate in the free energy landscape, possibility of protonation of the substrate, errors with the chosen constant pH MD method for membrane proteins, dismissal of hysteresis emerging from the MEMENTO method, and the likelihood of other residues being affected by peptide binding. Some changes to the presentation could be considered, including a better description of pKa calculations and the inclusion of error bars in all PMFs. Overall, the findings will appeal to structural biologists, biochemists, and biophysicists studying membrane transporters.

We would like to express our gratitude to the reviewers for providing their feedback on our manuscript, and also for recognising the variety of computational methods employed, the amount of sampling collected and the experimental validation undertaken. Following the individual reviewer comments, as addressed point-by-point below, we will shortly prepare a revised version of this paper. Intended changes to the revised manuscript are marked up in bold font in the detailed responses below, but before that we address some of the comments made above in the general assessment:

• “lack of incorporation of a protonation coordinate in the free energy landscape”. We acknowledge that of course it would be highly desirable to treat protonation state changes explicitly and fully coupled to conformational changes. However, at this point in time, evaluating such a free energy landscape is not computationally feasible (especially considering that the non-reactive approach taken here already amounts to almost 1ms of total sampling time). Previous reports in the literature tend to focus on either simpler systems or a reduced subset of a larger problem. As we were trying to obtain information on the whole transport cycle, we decided to focus here on non-reactive methods.

• “possibility of protonation of the substrate”. The reviewers are correct in pointing out this possibility, which we had not discussed explicitly in our manuscript. Briefly, while we describe a mechanism in which protonation of only protein residues (with an unprotonated ligand) can account for driving all the necessary conformational changes of the transport cycle, there is some evidence for a further intermediate protonation site in our data (as we commented on in the first version of the manuscript as well), which may or may not be the substrate itself. A future explicit treatment of the proton movements through the transporter, when it will become computationally tractable to do so, will have to include the substrate as a possible protonation site; for the present moment, we will amend our discussion to alert the reader to the possibility that the substrate could be an intermediate to proton transport. This has repercussions for our study of the E56 pKa value, where – if protons reside with a significant population at the substrate C-terminus – our calculated shift in pKa upon substrate binding could be an overestimate, although we would qualitatively expect the direction of shift to be unaffected. However, we also anticipate that treating this potential coupling explicitly would make convergence of any CpHMD calculation impractical to achieve and thus it may be the case that for now only a semi-quantitative conclusion is all that can be obtained.

• “errors with the chosen constant pH MD method for membrane proteins”. We acknowledge that – as reviewer #1 has reminded us – the AMBER implementation of hybrid-solvent CpHMD is not rigorous for membrane proteins, and as such we will add a cautionary note to our paper. We will also explain how the use of the ABFE thermodynamic cycle calculations helps to validate the CpHMD results in a completely orthogonal manner (we will promote this validation which was in the supplementary figures into the main text in the revised version). We therefore remain reasonably confident in the results presented with regards to the reported pKa shift of E56 upon substrate binding, and suggest that if the impact of neglecting the membrane in the implicit-solvent stage of CpHMD is significant, then there is likely an error cancellation when considering shifts induced by the incoming substrate.

• “dismissal of hysteresis emerging from the MEMENTO method”. We have shown in our method design paper how the use of the MEMENTO method drastically reduces hysteresis compared to steered MD and metadynamics for path generation, and find this improvement again for PepT2 in this study. We will address reviewer #3’s concern about our presentation on this point by revising our introduction of the MEMENTO method, as detailed in the response below.

• “the likelihood of other residues being affected by peptide binding”. In this study, we have investigated in detail the involvement of several residues in proton-coupled di-peptide transport by PepT2. Short of the potential intermediate protonation site mentioned above, the set of residues we investigate form a minimal set of sorts within which the important driving forces of alternating access can be rationalised. We have not investigated in substantial detail here the residues involved in holding the peptide in the binding site, as they are well studied in the literature and ligand promiscuity is not the problem of interest here. It remains entirely possible that further processes contribute to the mechanism of driving conformational changes by involving other residues not considered in this paper. We will make our speculation that an ensemble of different processes may be contributing simultaneously more explicit in our revision, but do not believe any of our conclusions would be affected by this.

As for the additional suggested changes in presentation, we will provide the requested details on the CpHMD analysis. Furthermore, we will use the convergence data presented separately in figures S12 and S16 to include error bars on our 1D-reprojections of the 2D-PMFs in figures 3, 4 and 5. (Note that we will opt to not do so in figures S10 and S15 which collate all 1D PMF reprojections for the OCC ↔ OF and OCC ↔ IF transitions in single reference plots, respectively, to avoid overcrowding those necessarily busy figures). We are also changing the colours schemes of these plots in our revision to improve accessibility.

Reviewer #1 (Public Review):

The authors have performed all-atom MD simulations to study the working mechanism of hsPepT2. It is widely accepted that conformational transitions of proton-coupled oligopeptide transporters (POTs) are linked with gating hydrogen bonds and salt bridges involving protonatable residues, whose protonation triggers gate openings. Through unbiased MD simulations, the authors identified extra-cellular (H87 and D342) and intra-cellular (E53 and E622) triggers. The authors then validated these triggers using free energy calculations (FECs) and assessed the engagement of the substrate (Ala-Phe dipeptide). The linkage of substrate release with the protonation of the ExxER motif (E53 and E56) was confirmed using constant-pH molecular dynamics (CpHMD) simulations and cellbased transport assays. An alternating-access mechanism was proposed. The study was largely conducted properly, and the paper was well-organized. However, I have a couple of concerns for the authors to consider addressing.

We would like to note here that it may be slightly misleading to the reader to state that “The linkage of substrate release with the protonation of the ExxER motif (E53 and E56) was confirmed using constant-pH molecular dynamics (CpHMD) simulations and cell-based transport assays.” The cellbased transport assays confirmed the importance of the extracellular gating trigger residues H87, S321 and D342 (as mentioned in the preceding sentence), not of the substrate-protonation link as this line might be understood to suggest.

(1) As a proton-coupled membrane protein, the conformational dynamics of hsPepT2 are closely coupled to protonation events of gating residues. Instead of using semi-reactive methods like CpHMD or reactive methods such as reactive MD, where the coupling is accounted for, the authors opted for extensive non-reactive regular MD simulations to explore this coupling. Note that I am not criticizing the choice of methods, and I think those regular MD simulations were well-designed and conducted. But I do have two concerns.

a) Ideally, proton-coupled conformational transitions should be modelled using a free energy landscape with two or more reaction coordinates (or CVs), with one describing the protonation event and the other describing the conformational transitions. The minimum free energy path then illustrates the reaction progress, such as OCC/H87D342- → OCC/H87HD342H → OF/H87HD342H as displayed in Figure 3.

We concur with the reviewer that the ideal way of describing the processes studied in our paper would be as a higher-dimensional free energy landscapes obtained from a simulation method that can explicitly model proton-transfer processes. Indeed, it would have been particularly interesting and potentially informative with regards to the movement of protons down into the transporter in the OF → OCC → IF sequence of transitions. As we note in our discussion on the H87→E56 proton transfer:

“This could be investigated using reactive MD or QM/MM simulations (both approaches have been employed for other protonation steps of prokaryotic peptide transporters, see Parker et al. (2017) and Li et al. (2022)). However, the putative path is very long (≈ 1.7 nm between H87 and E56) and may or may not involve a large number of intermediate protonatable residues, in addition to binding site water. While such an investigation is possible in principle, it is beyond the scope of the present study.”

Where even sampling the proton transfer step itself in an essentially static protein conformation would be pushing the boundaries of what has been achieved in the field, we believe that considering the current state-of-the-art, a fully coupled investigation of large-scale conformational changes and proton-transfer reaction is not yet feasible in a realistic/practical time frame. We also note this limitation already when we say that:

“The question of whether proton binding happens in OCC or OF warrants further investigation, and indeed the co-existence of several mechanisms may be plausible here”.

Nonetheless, we are actively exploring approaches to treat uptake and movement of protons explicitly for future work.

In our revision, we will expand on our discussion of the reasoning behind employing a nonreactive approach and the limitations that imposes on what questions can be answered in this study.

Without including the protonation as a CV, the authors tried to model the free energy changes from multiple FECs using different charge states of H87 and D342. This is a practical workaround, and the conclusion drawn (the OCC→ OF transition is downhill with protonated H87 and D342) seems valid. However, I don't think the OF states with different charge states (OF/H87D342-, OF/H87HD342-, OF/H87D342H, and OF/H87HD342H) are equally stable, as plotted in Figure 3b. The concern extends to other cases like Figures 4b, S7, S10, S12, S15, and S16. While it may be appropriate to match all four OF states in the free energy plot for comparison purposes, the authors should clarify this to ensure readers are not misled.

The reviewer is correct in their assessment that the aligning of PMFs in these figures is arbitrary; no relative free energies of the PMFs to each other can be estimated without explicit free energy calculations at least of protonation events at the end state basins. The PMFs in our figures are merely superimposed for illustrating the differences in shape between the obtained profiles in each condition, as discussed in the text, and we will make this clear in the appropriate figure captions in our revision.

b) Regarding the substrate impact, it appears that the authors assumed fixed protonation states. I am afraid this is not necessarily the case. Variations in PepT2 stoichiometry suggest that substrates likely participate in proton transport, like the Phe-Ala (2:1) and Phe-Gln (1:1) dipeptides mentioned in the introduction. And it is not rigorous to assume that the N- and C-termini of a peptide do not protonate/deprotonate when transported. I think the authors should explicitly state that the current work and the proposed mechanism (Figure 8) are based on the assumption that the substrates do not uptake/release proton(s).

This is indeed an assumption inherent in the current work. While we do “speculate that the proton movement processes may happen as an ensemble of different mechanisms, and potentially occur contemporaneously with the conformational change” we do not in the current version indicate explicitly that this may involve the substrate. We will make clear the assumption and this possibility in the revised version of our paper. Indeed, as we discuss, there is some evidence in our PMFs of an additional protonation site not considered thus far, which may or may not be the substrate. We will make note of this point in the revised manuscript.

As for what information can be drawn from the given experimental stoichiometries, we note in our paper that “a 2:1 stoichiometry was reported for the neutral di-peptide D-Phe-L-Ala and 3:1 for anionic D-Phe-L-Glu. (Chen et al., 1999) Alternatively, Fei et al. (1999) have found 1:1 stoichiometries for either of D-Phe-L-Gln (neutral), D-Phe-L-Glu (anionic), and D-Phe-L-Lys (cationic).”

We do not assume that it is our place to arbit among the apparent discrepancies in the experimental data here, although we believe that our assumed 2:1 stoichiometry is additionally “motivated also by our computational results that indicate distinct and additive roles played by two protons in the conformational cycle mechanism”.

(2) I have more serious concerns about the CpHMD employed in the study.

a) The CpHMD in AMBER is not rigorous for membrane simulations. The underlying generalized Born model fails to consider the membrane environment when updating charge states. In other words, the CpHMD places a membrane protein in a water environment to judge if changes in charge states are energetically favorable. While this might not be a big issue for peripheral residues of membrane proteins, it is likely unphysical for internal residues like the ExxER motif. As I recall, the developers have never used the method to study membrane proteins themselves. The only CpHMD variant suitable for membrane proteins is the membrane-enabled hybrid-solvent CpHMD in CHARMM. While I do not expect the authors to redo their CpHMD simulations, I do hope the authors recognize the limitations of their method.

We will discuss the limitations of the AMBER CpHMD implementation in the revised version. However, despite that, we believe we have in fact provided sufficient grounds for our conclusion that substrate binding affects ExxER motif protonation in the following way:

In addition to CpHMD simulations, we establish the same effect via ABFE calculations, where the substrate affinity is different at the E56 deprotonated vs protonated protein. This is currently figure S20, though in the revised version we will move this piece of validation into a new panel of figure 6 in the main text, since it becomes more important with the CpHMD membrane problem in mind. Since the ABFE calculations are conducted with an all-atom representation of the lipids and the thermodynamic cycle closes well, it would appear that if the chosen CpHMD method has a systematic error of significant magnitude for this particular membrane protein system, there may be the benefit of error cancellation. While the calculated absolute pKa values may not be reliable, the difference made by substrate binding appears to be so, as judged by the orthogonal ABFE technique.

Although the reviewer does “not expect the authors to redo their CpHMD simulations”, we consider that it may be helpful to the reader to share in this response some results from trials using the continuous, all-atom constant pH implementation that has recently become available in GROMACS (Aho et al 2022, https://pubs.acs.org/doi/10.1021/acs.jctc.2c00516) and can be used rigorously with membrane proteins, given its all-atom lipid representation.

Unfortunately, when trying to titrate E56 in this CpHMD implementation, we found few protonationstate transitions taking place, and the system often got stuck in protonation state–local conformation coupled minima (which need to interconvert through rearrangements of the salt bridge network involving slow side-chain dihedral rotations in E53, E56 and R57). Author response image 1 shows this for the apo OF state, Author response image 2 shows how noisy attempts at pKa estimation from this data turn out to be, necessitating the use of a hybrid-solvent method.

Author response image 1.

All-atom CpHMD simulations of apo-OF PepT2. Red indicates protonated E56, blue is deprotonated.

Author response image 2.

Difficulty in calculating the E56 pKa value from the noisy all-atom CpHMD data shown in Author response image 1

b) It appears that the authors did not make the substrate (Ala-Phe dipeptide) protonatable in holosimulations. This oversight prevents a complete representation of ligand-induced protonation events, particularly given that the substrate ion pairs with hsPepT2 through its N- & C-termini. I believe it would be valuable for the authors to acknowledge this potential limitation.

In this study, we implicitly assumed from the outset that the substrate does not get protonated, which – as by way of response to the comment above – we will acknowledge explicitly in revision. This potential limitation for the available mechanisms for proton transfer also applies to our investigation of the ExxER protonation states. In particular, a semi-grand canonical ensemble that takes into account the possibility of substrate C-terminus protonation may also sample states in which the substrate is protonated and oriented away from R57, thus leaving the ExxER salt bridge network in an apo-like state. The consequence would be that while the direction of shift in E56 pKa value will be the same, our CpHMD may overestimate its magnitude. It would thus be interesting to make the C-terminus protonatable for obtaining better quantitative estimates of the E56 pKa shift (as is indeed true in general for any other protein protonatable residue, though the effects are usually assumed to be negligible). We do note, however, that convergence of the CpHMD simulations would be much harder if the slow degree of freedom of substrate reorientation (which in our experience takes 10s to 100s of ns in this binding pocket) needs to be implicitly equilibrated upon protonation state transitions. We will discuss such considerations in the revision.

Reviewer #2 (Public Review):

This is an interesting manuscript that describes a series of molecular dynamics studies on the peptide transporter PepT2 (SLC15A2). They examine, in particular, the effect on the transport cycle of protonation of various charged amino acids within the protein. They then validate their conclusions by mutating two of the residues that they predict to be critical for transport in cell-based transport assays. The study suggests a series of protonation steps that are necessary for transport to occur in Petp2. Comparison with bacterial proteins from the same family shows that while the overall architecture of the proteins and likely mechanism are similar, the residues involved in the mechanism may differ.

Strengths:

This is an interesting and rigorous study that uses various state-of-the-art molecular dynamics techniques to dissect the transport cycle of PepT2 with nearly 1ms of sampling. It gives insight into the transport mechanism, investigating how the protonation of selected residues can alter the energetic barriers between various states of the transport cycle. The authors have, in general, been very careful in their interpretation of the data.

Weaknesses:

Interestingly, they suggest that there is an additional protonation event that may take place as the protein goes from occluded to inward-facing but they have not identified this residue.

We have indeed suggested that there may be an additional protonation site involved in the conformational cycle that we have not been able to capture, which – as we discuss in our paper – might be indicated by the shapes of the OCC ↔ IF PMFs given in Figure S15. One possibility is for this to be the substrate itself (see the response to reviewer #1 above) though within the scope of this study the precise pathway by which protons move down the transporter and the exact ordering of conformational change and proton transfer reactions remains a (partially) open question. We acknowledge this and denote it with question marks in the mechanistic overview we give in Figure 8, and also “speculate that the proton movement processes may happen as an ensemble of different mechanisms, and potentially occur contemporaneously with the conformational change”.

Some things are a little unclear. For instance, where does the state that they have defined as occluded sit on the diagram in Figure 1a? - is it truly the occluded state as shown on the diagram or does it tend to inward- or outward-facing?

Figure 1a is a simple schematic overview intended to show which structures of PepT2 homologues are available to use in simulations. This was not meant to be a quantitative classification of states. Nonetheless, we can note that the OCC state we derived has extra- and intracellular gate opening distances (as measured by the simple CVs defined in the methods and illustrated in Figure 2a) that indicate full gate closure at both sides. In particular, although it was derived from the IF state via biased sampling, the intracellular gate opening distance in the OCC state used for our conformational change enhanced sampling was comparable to that of the OF state (ie, full closure of the gate), see Figure S2b and the grey bars therein. Therefore, we would schematically classify the OCC state to lie at the center of the diagram in Figure 1a. Furthermore, it is largely stable over triplicates of 1 μslong unbiased MD, where in 2/3 replicates the gates remain stable, and the remaining replicate there is partial opening of the intracellular gate (as shown in Figure 2 b/c under the “apo standard” condition). We comment on this in the main text by saying that “The intracellular gate, by contrast, is more flexible than the extracellular gate even in the apo, standard protonation state”, and link it to the lower barrier for transition to IF than to OF. We did this by saying that “As for the OCC↔OF transitions, these results explain the behaviour we had previously observed in the unbiased MD of Figure 2c.” We acknowledge this was not sufficiently clear and will add details to the latter sentence in revision to help clarify better the nature of the occluded state.

The pKa calculations and their interpretation are a bit unclear. Firstly, it is unclear whether they are using all the data in the calculations of the histograms, or just selected data and if so on what basis was this selection done. Secondly, they dismiss the pKa calculations of E53 in the outward-facing form as not being affected by peptide binding but say that E56 is when there seems to be a similar change in profile in the histograms.

In our manuscript, we have provided two distinct analyses of the raw CpHMD data. Firstly, we analysed the data by the replicates in which our simulations were conducted (Figure 6, shown as bar plots with mean from triplicates +/- standard deviation), where we found that only the effect on E56 protonation was distinct as lying beyond the combined error bars. This analysis uses the full amount of sampling conducted for each replicate. However, since we found that the range of pKa values estimated from 10ns/window chunks was larger than the error bars obtained from the replicate analysis (Figures S17 and S18), we sought to verify our conclusion by pooling all chunk estimates and plotting histograms (Figure S19). We recover from those the effect of substrate binding on the E56 protonation state on both the OF and OCC states. However, as the reviewer has pointed out (something we did not discuss in our original manuscript), there is a shift in the pKa of E53 of the OF state only. In fact, the trend is also apparent in the replicate-based analysis of Figure 6, though here the larger error bars overlap. In our revision, we will add more details of these analyses for clarity (including more detailed figure captions regarding the data used in Figure 6) as well as a discussion of the partial effect on the E53 pKa value.

We do not believe, however, that our key conclusions are negatively affected. If anything, a further effect on the E53 pKa which we had not previously commented on (since we saw the evidence as weaker, pertaining to only one conformational state) would strengthen the case for an involvement of the ExxER motif in ligand coupling.

Reviewer #3 (Public Review):

Summary:

Lichtinger et al. have used an extensive set of molecular dynamics (MD) simulations to study the conformational dynamics and transport cycle of an important member of the proton-coupled oligopeptide transporters (POTs), namely SLC15A2 or PepT2. This protein is one of the most wellstudied mammalian POT transporters that provides a good model with enough insight and structural information to be studied computationally using advanced enhanced sampling methods employed in this work. The authors have used microsecond-level MD simulations, constant-PH MD, and alchemical binding free energy calculations along with cell-based transport assay measurements; however, the most important part of this work is the use of enhanced sampling techniques to study the conformational dynamics of PepT2 under different conditions.

The study attempts to identify links between conformational dynamics and chemical events such as proton binding, ligand-protein interactions, and intramolecular interactions. The ultimate goal is of course to understand the proton-coupled peptide and drug transport by PepT2 and homologous transporters in the solute carrier family.

Some of the key results include:

(1) Protonation of H87 and D342 initiate the occluded (Occ) to the outward-facing (OF) state transition.

(2) In the OF state, through engaging R57, substrate entry increases the pKa value of E56 and thermodynamically facilitates the movement of protons further down.

(3) E622 is not only essential for peptide recognition but also its protonation facilitates substrate release and contributes to the intracellular gate opening. In addition, cell-based transport assays show that mutation of residues such as H87 and D342 significantly decreases transport activity as expected from simulations.

Strengths:

(1) This is an extensive MD-based study of PepT2, which is beyond the typical MD studies both in terms of the sheer volume of simulations as well as the advanced methodology used. The authors have not limited themselves to one approach and have appropriately combined equilibrium MD with alchemical free energy calculations, constant-pH MD, and geometry-based free energy calculations. Each of these 4 methods provides a unique insight regarding the transport mechanism of PepT2.

(2) The authors have not limited themselves to computational work and have performed experiments as well. The cell-based transport assays clearly establish the importance of the residues that have been identified as significant contributors to the transport mechanism using simulations.

(3) The conclusions made based on the simulations are mostly convincing and provide useful information regarding the proton pathway and the role of important residues in proton binding, protein-ligand interaction, and conformational changes.

Weaknesses:

(1) Some of the statements made in the manuscript are not convincing and do not abide by the standards that are mostly followed in the manuscript. For instance, on page 4, it is stated that "the K64-D317 interaction is formed in only ≈ 70% of MD frames and therefore is unlikely to contribute much to extracellular gate stability." I do not agree that 70% is negligible. Particularly, Figure S3 does not include the time series so it is not clear whether the 30% of the time where the salt bridge is broken is in the beginning or the end of simulations. For instance, it is likely that the salt bridge is not initially present and then it forms very strongly. Of course, this is just one possible scenario but the point is that Figure S3 does not rule out the possibility of a significant role for the K64-D317 salt bridge.

The reviewer is right to point out that the statement and Figure S3 as they stand do not adequately support our decision to exclude the K64-D317 salt-bridge in our further investigations. The violin plot shown in Figure S3, visualised as pooled data from unbiased 1 μs triplicates, does indeed not rule out a scenario where the salt bridge only formed late in our simulations (or only in some replicates), but then is stable. Therefore, in our revision, we will include the appropriate time-series of the salt bridge distances, showing how K64-D317 is initially stable but then falls apart in replicate 1, and is transiently formed and disengaged across the trajectories in replicates 2 and 3. We will also remake the data for this plot as we discovered a bug in the relevant analysis script that meant the D170-K642 distance was not calculated accurately. The results are however almost identical, and our conclusions remain.

(2) Similarly, on page 4, it is stated that "whether by protonation or mutation - the extracellular gate only opens spontaneously when both the H87 interaction network and D342-R206 are perturbed (Figure S5)." I do not agree with this assessment. The authors need to be aware of the limitations of this approach. Consider "WT H87-prot" and "D342A H87-prot": when D342 residue is mutated, in one out of 3 simulations, we see the opening of the gate within 1 us. When D342 residue is not mutated we do not see the opening in any of the 3 simulations within 1 us. It is quite likely that if rather than 3 we have 10 simulations or rather than 1 us we have 10 us simulations, the 0/3 to 1/3 changes significantly. I do not find this argument and conclusion compelling at all.

If the conclusions were based on that alone, then we would agree. However, this section of work covers merely the observations of the initial unbiased simulations which we go on to test/explore with enhanced sampling in the rest of the paper, and which then lead us to the eventual conclusions.

Figure S5 shows the results from triplicate 1 μs-long trajectories as violin-plot histograms of the extracellular gate opening distance, also indicating the first and final frames of the trajectories as connected by an arrow for orientation – a format we chose for intuitively comparing 48 trajectories in one plot. The reviewer reads the plot correctly when they analyse the “WT H87-prot” vs “D342A H87-prot” conditions. In the former case, no spontaneous opening in unbiased MD is taking place, whereas when D342 is mutated to alanine in addition to H87 protonation, we see spontaneous transition in 1 out of 3 replicates. However, the reviewer does not seem to interpret the statement in question in our paper (“the extracellular gate only opens spontaneously when both the H87 interaction network and D342-R206 are perturbed”) in the way we intended it to be understood. We merely want to note here a correlation in the unbiased dataset we collected at this stage, and indeed the one spontaneous opening in the case comparison picked out by the reviewer is in the condition where both the H87 interaction network and D342-R206 are perturbed. In noting this we do not intend to make statistically significant statements from the limited dataset. Instead, we write that “these simulations show a large amount of stochasticity and drawing clean conclusions from the data is difficult”. We do however stand by our assessment that from this limited data we can “already appreciate a possible mechanism where protons move down the transporter pore” – a hypothesis we investigate more rigorously with enhanced sampling in the rest of the paper. We will revise the section in question to make clearer that the unbiased MD is only meant to give an initial hypothesis here to be investigated in more detail in the following sections. In doing so, we will also incorporate, as we had not done before, the case (not picked out by the reviewer here but concerning the same figure) of S321A & H87 prot. In the third replicate, this shows partial gate opening towards the end of the unbiased trajectory (despite D342 not being affected), highlighting further the stochastic nature that makes even clear correlative conclusions difficult to draw.

(3) While the MEMENTO methodology is novel and interesting, the method is presented as flawless in the manuscript, which is not true at all. It is stated on Page 5 with regards to the path generated by MEMENTO that "These paths are then by definition non-hysteretic." I think this is too big of a claim to say the paths generated by MEMENTO are non-hysteretic by definition. This claim is not even mentioned in the original MEMENTO paper. What is mentioned is that linear interpolation generates a hysteresis-free path by definition. There are two important problems here: (a) MEMENTO uses the linear interpolation as an initial step but modifies the intermediates significantly later so they are no longer linearly interpolated structures and thus the path is no longer hysteresisfree; (b) a more serious problem is the attribution of by-definition hysteresis-free features to the linearly interpolated states. This is based on conflating the hysteresis-free and unique concepts. The hysteresis in MD-based enhanced sampling is related to the presence of barriers in orthogonal space. For instance, one may use a non-linear interpolation of any type and get a unique pathway, which could be substantially different from the one coming from the linear interpolation. None of these paths will be hysteresis-free necessarily once subjected to MD-based enhanced sampling techniques.

We certainly do not intend to claim that the MEMENTO method is flawless. The concern the reviewer raises around the statement "These paths are then by definition non-hysteretic" is perhaps best addressed by a clarification of the language used and considering how MEMENTO is applied in this work.

Hysteresis in the most general sense denotes the dependence of a system on its history, or – more specifically – the lagging behind of the system state with regards to some physical driver (for example the external field in magnetism, whence the term originates). In the context of biased MD and enhanced sampling, hysteresis commonly denotes the phenomenon where a path created by a biased dynamics method along a certain collective variable lags behind in phase space in slow orthogonal degrees of freedom (see Figure 1 in Lichtinger and Biggin 2023, https://doi.org/10.1021/acs.jctc.3c00140). When used to generate free energy profiles, this can manifest as starting state bias, where the conformational state that was used to seed the biased dynamics appears lower in free energy than alternative states. Figure S6 shows this effect on the PepT2 system for both steered MD (heavy atom RMSD CV) + umbrella sampling (tip CV) and metadynamics (tip CV). There is, in essence, a coupled problem: without an appropriate CV (which we did not have to start with here), path generation that is required for enhanced sampling displays hysteresis, but the refinement of CVs is only feasible when paths connecting the true phase space basins of the two conformations are available. MEMENTO helps solve this issue by reconstructing protein conformations along morphing paths which perform much better than steered MD paths with respect to giving consistent free energy profiles (see Figure S7 and the validation cases in the MEMENTO paper), even if the same CV is used in umbrella sampling.

There are still differences between replicates in those PMFs, indicating slow conformational flexibility propagated from end-state sampling through MEMENTO. We use this to refine the CVs further with dimensionality reduction (see the Method section and Figure S8), before moving to 2D-umbrella sampling (figure 3). Here, we think, the reviewer’s point seems to bear. The MEMENTO paths are ‘non-hysteretic by definition’ with respect to given end states in the sense that they connect (by definition) the correct conformations at both end-states (unlike steered MD), which in enhanced sampling manifests as the absence of the strong starting-state bias we had previously observed (Figure S7 vs S6). They are not, however, hysteresis-free with regards to how representative of the end-state conformational flexibility the structures given to MEMENTO really were, which is where the iterative CV design and combination of several MEMENTO paths in 2D-PMFs comes in.

We also cannot make a direct claim about whether in the transition region the MEMENTO paths might be separated from the true (lower free energy) transition paths by slow orthogonal degrees of freedom, which may conceivably result in overestimated barrier heights separating two free energy basins. We cannot guarantee that this is not the case, but neither in our MEMENTO validation examples nor in this work have we encountered any indications of a problem here.

We hope that the reviewer will be satisfied by our revision, where we will replace the wording in question by a statement that the MEMENTO paths do not suffer from hysteresis that is otherwise incurred as a consequence of not reaching the correct target state in the biased run (in some orthogonal degrees of freedom).

Reviewer #3 (Public Review):

Summary:

Lichtinger et al. have used an extensive set of molecular dynamics (MD) simulations to study the conformational dynamics and transport cycle of an important member of the proton-coupled oligopeptide transporters (POTs), namely SLC15A2 or PepT2. This protein is one of the most well-studied mammalian POT transporters that provides a good model with enough insight and structural information to be studied computationally using advanced enhanced sampling methods employed in this work. The authors have used microsecond-level MD simulations, constant-PH MD, and alchemical binding free energy calculations along with cell-based transport assay measurements; however, the most important part of this work is the use of enhanced sampling techniques to study the conformational dynamics of PepT2 under different conditions.

The study attempts to identify links between conformational dynamics and chemical events such as proton binding, ligand-protein interactions, and intramolecular interactions. The ultimate goal is of course to understand the proton-coupled peptide and drug transport by PepT2 and homologous transporters in the solute carrier family.

Some of the key results include<br /> (1) Protonation of H87 and D342 initiate the occluded (Occ) to the outward-facing (OF) state transition.

(2) In the OF state, through engaging R57, substrate entry increases the pKa value of E56 and thermodynamically facilitates the movement of protons further down.

(3) E622 is not only essential for peptide recognition but also its protonation facilitates substrate release and contributes to the intracellular gate opening. In addition, cell-based transport assays show that mutation of residues such as H87 and D342 significantly decreases transport activity as expected from simulations.

Strengths:

(1) This is an extensive MD-based study of PepT2, which is beyond the typical MD studies both in terms of the sheer volume of simulations as well as the advanced methodology used. The authors have not limited themselves to one approach and have appropriately combined equilibrium MD with alchemical free energy calculations, constant-pH MD, and geometry-based free energy calculations. Each of these 4 methods provides a unique insight regarding the transport mechanism of PepT2.

(2) The authors have not limited themselves to computational work and have performed experiments as well. The cell-based transport assays clearly establish the importance of the residues that have been identified as significant contributors to the transport mechanism using simulations.

(3) The conclusions made based on the simulations are mostly convincing and provide useful information regarding the proton pathway and the role of important residues in proton binding, protein-ligand interaction, and conformational changes.

Weaknesses:

(1) Some of the statements made in the manuscript are not convincing and do not abide by the standards that are mostly followed in the manuscript. For instance, on page 4, it is stated that "the K64-D317 interaction is formed in only ≈ 70% of MD frames and therefore is unlikely to contribute much to extracellular gate stability." I do not agree that 70% is negligible. Particularly, Figure S3 does not include the time series so it is not clear whether the 30% of the time where the salt bridge is broken is in the beginning or the end of simulations. For instance, it is likely that the salt bridge is not initially present and then it forms very strongly. Of course, this is just one possible scenario but the point is that Figure S3 does not rule out the possibility of a significant role for the K64-D317 salt bridge.

(2) Similarly, on page 4, it is stated that "whether by protonation or mutation - the extracellular gate only opens spontaneously when both the H87 interaction network and D342-R206 are perturbed (Figure S5)." I do not agree with this assessment. The authors need to be aware of the limitations of this approach. Consider "WT H87-prot" and "D342A H87-prot": when D342 residue is mutated, in one out of 3 simulations, we see the opening of the gate within 1 us. When D342 residue is not mutated we do not see the opening in any of the 3 simulations within 1 us. It is quite likely that if rather than 3 we have 10 simulations or rather than 1 us we have 10 us simulations, the 0/3 to 1/3 changes significantly. I do not find this argument and conclusion compelling at all.

(3) While the MEMENTO methodology is novel and interesting, the method is presented as flawless in the manuscript, which is not true at all. It is stated on Page 5 with regards to the path generated by MEMENTO that "These paths are then by definition non-hysteretic." I think this is too big of a claim to say the paths generated by MEMENTO are non-hysteretic by definition. This claim is not even mentioned in the original MEMENTO paper. What is mentioned is that linear interpolation generates a hysteresis-free path by definition. There are two important problems here: (a) MEMENTO uses the linear interpolation as an initial step but modifies the intermediates significantly later so they are no longer linearly interpolated structures and thus the path is no longer hysteresis-free; (b) a more serious problem is the attribution of by-definition hysteresis-free features to the linearly interpolated states. This is based on conflating the hysteresis-free and unique concepts. The hysteresis in MD-based enhanced sampling is related to the presence of barriers in orthogonal space. For instance, one may use a non-linear interpolation of any type and get a unique pathway, which could be substantially different from the one coming from the linear interpolation. None of these paths will be hysteresis-free necessarily once subjected to MD-based enhanced sampling techniques.

Author response:

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

Reviewer 1

(1) Given the low trial numbers, and the point of sequential vs clustered reactivation mentioned in the public review, it would be reassuring to see an additional sanity check demonstrating that future items that are currently not on-screen can be decoded with confidence, and if so, when in time the peak reactivation occurs. For example, the authors could show separately the decoding accuracy for near and far items in Fig. 5A, instead of plotting only the difference between them.

We have now added the requested analysis showing the raw decoded probabilities for near and distant items separately in Figure 5A. We have also chosen to replace Figure 5B with the new figure as we think it provides more information than the previous Figure 5B. Instead, we have moved Figure 5B to the supplement. The median peak decoded accuracy for near and distant items is equivalent. We have added the following description to the figure:

“Decoded raw probabilities for off-screen items, that were up to two steps ahead of the current stimulus cue (‘near’,) vs. distant items that were more than two steps away on the graph, on trials with correct answers. The median peak decoded probability for near and distant items was at the same time point for both probability categories. Note that displayed lines reflect the average probability while, to eliminate influence of outliers, the peak displays the median.”

(2) The non-sequential reactivation analyses often use a time window of peak decodability, and it was not entirely clear to me what data this time window is determined on, e.g., was it determined based on all future reactivations irrespective of graph distance? This should be clarified in the methods.

Thank you for raising this. We now clarify this in the relevant section to read: “First, we calculated a time point of interest by computing the peak probability estimate of decoders across all trials, i.e., the average probability for each timepoint of all trials (except previous onscreen items) of all distances, which is equivalent to the peak of the differential reactivation analysis”

(3) Fig 4 shows evidence for forward and backward sequential reactivation, suggesting that both forward and backward replay peak at a lag of 40-50msec. It would be helpful if this counterintuitive finding could be picked up in the discussion, explaining how plausible it is, physiologically, to find forward and backward replay at the same lag, and whether this could be an artifact of the TDLM method.

This is an important point and we agree that it appears counterintuitive. However, we would highlight this exact time range has been reported in previous studies, though t never for both forward and backward replay. We now include a discussion of this finding. The section now reads:

“[… ] Even though we primarily focused on the mean sequenceness scores across time lags, there appears s to be a (non-significant) peak at 40-60 milliseconds. While simultaneous forward and backward replay is theoretically possible, we acknowledge that it is somewhat surprising and, given our paradigm, could relate to other factors such as autocorrelations (Liu, Dolan, et al., 2021).”

(4) It is reported that participants with below 30% decoding accuracy are excluded from the main analyses. It would be helpful if the manuscript included very specific information about this exclusion, e.g., was the criterion established based on the localizer cross-validated data, the temporal generalisation to the cued item (Fig. 2), or only based on peak decodability of the future sequence items? If the latter, is it applied based on near or far reactivations, or both?

We now clarify this point to include more specific information, which reads:

“[…] Therefore, we decided a priori that participants with a peak decoding accuracy of below 30% would be excluded from the analysis (nine participants in all) as obtained from the cross-validation of localizer trials”

(5) Regarding the low amount of data for the reactivation analysis, the manuscript should be explicit about the number of trials available for each participant. For example, Supplemental Fig. 1 could provide this information directly, rather than the proportion of excluded trials.

We have adapted the plot in the supplement to show the absolute number of rejected epochs per participant, in addition to the ratio.

(6) More generally, the supplements could include more detailed information in the legends.

We agree and have added more extensive explanation of the plots in the supplement legends.

(7) The choice of comparing the 2 nearest with all other future items in the clustered reactivation analysis should be better motivated, e.g., was this based on the Wimmer et al. (2020) study?

We have added our motivation for taking the two nearest items and contrasting them with the items further away. The paragraph reads:

“[…] We chose to combine the following two items for two reasons: First, this doubled the number of included trials; secondly, using this approach the number of trials for each category (“near” and “distant”) was more balanced. […]”

Reviewer 2

(1) Focus exclusively on retrieval data (and here just on the current image trials).

If I understand correctly, you focus all your analyses (behavioural as well as MEG analyses) on retrieval data only and here just on the current image trials. I am surprised by that since I see some shortcomings due to that. These shortcomings can likely be addressed by including the learning data (and predecessor image trials) in your analyses.

a) Number of trials: During each block, you presented each of the twelve edges once. During retrieval, participants then did one "single testing session block". Does that mean that all your results are based on max. 12 trials? Given that participants remembered, on average, 80% this means even fewer trials, i.e., 9-10 trials?

This is correct and a limitation of the paper. However, while we used only correct trials for the reactivation analysis, the sequential analysis was conducted using all trials disregarding the response behaviour. To retain comparability with previous studies we mainly focused on data from after a consolidation phase. Nevertheless, despite the trial limitation we consider the results are robust and worth reporting. Additionally, based on the suggestion of the referee, we now include results from learning blocks (see below).

b) Extend the behavioural and replay/reactivation analysis to predecessor images.

Why do you restrict your analyses to the current image trials? Especially given that you have such a low trial number for your analyses, I was wondering why you did not include the predecessor trials (except the non-deterministic trials, like the zebra and the foot according to Figure 2B) as well.

We agree it would be great to increase power by adding the predecessor images to the current image cue analysis, excluding the ambiguous trials, we did not do so as we considered the underlying retrieval processes of these trial types are not the same, i.e. cannot be simply combined. Nevertheless, we have performed the suggested analysis to check if it increases our power. We found, that the reactivation effect is robust and significant at the same time point of 220-230 ms. However, the effect size actually decreased: While before, peak differential reactivation was at 0.13, it is now at 0.07. This in fact makes conceptual sense. We suspect that the two processes that are elicited by showing a single cue and by showing a second, related, cue are distinct insofar as the predecessor image acts as a primer for the current image, potentially changing the time course/speed of retrieval. Given our concerns that the two processes are not actually the same we consider it important to avoid mixing these data.

We have added a statement to the manuscript discussing this point. The section reads:

“Note that we only included data from the current image cue, and not from the predecessor image cue, as we assume the retrieval processes differ and should not be concatenated.”

c) Extend the behavioural and replay/reactivation analysis to learning trials.

Similar to point 1b, why did you not include learning trials in your analyses?

The advantage of including (correct and incorrect) learning trials has the advantage that you do not have to exclude 7 participants due to ceiling performance (100%).

Further, you could actually test the hypothesis that you outline in your discussion: "This implies that there may be a switch from sequential replay to clustered reactivation corresponding to when learned material can be accessed simultaneously without interference." Accordingly, you would expect to see more replay (and less "clustered" reactivation) in the first learning blocks compared to retrieval (after the rest period).

To track reactivation and replay over the course of learning is a great idea. We have given a lot of thought as to how to integrate these findings but have not found a satisfying solution. Thus, analysis of the learning data turned out to be quite tricky: We decided that each participant should perform as many blocks as necessary to reach at least 80% (with a limit of six and lower bound of two, see Supplement figure 4). Indeed, some participant learned 100% of the sequence after one block (these were mostly medical students, learning things by hard is their daily task). With the benefit of hindsight, we realise our design means that different blocks are not directly comparable between participants. In theory, we would expect that replay emerges in parallel with learning and then gradually changes to clustered reactivation as memory traces become consolidated/stronger. However, it is unclear when replay should emerge and when precisely a switch to clustered reactivation would happen. For this reason, we initially decided not to include the learning trials into the paper.

Nevertheless, to provide some insight into the learning process, and to see how consolidation impacts differential reactivation and replay, we have split our data into pre and post resting state, aggregating all learning trials of each participant. While this does not allow us to track processes on a block basis, it does offer potential (albeit limited) insight into the hypothesis we outline in the discussion.

For reactivation, we see emergence of a clear increase, further strengthening the outlined hypothesis, however, for replay the evidence is less clear, as we do not know over how many learning blocks replay is expected.

We calculated individual trajectories of how reactivation and replay changes from learning to retrieval and related these to performance. Indeed, we see an increase of reactivation is nominally associated with higher learning performance, while an increase in replay strength is associated with lower performance (both non-significant). However, due to the above-mentioned reasons we think it would premature to add this weak evidence to the paper.

To mitigate problems of experiment design in relation to this question we are currently implementing a follow-study, where we aim to normalize the learning process across participants and index how replay/reactivation changes over the course of learning and after consolidation.

We have added plots showing clustered reactivation sequential replay measures during learning (Figure 5D and Supplement 8)

The added section(s) now read:

“To provide greater detail on how the 8-minute consolidation period affected reactivation we, post-hoc, looked at relevant measures across learning trials in contrast to retrieval trials. For all learning trials, for each participant, we calculated differential reactivation for the same time point we found significant in the previous analysis (220-260 milliseconds). On average, differential reactivation probability increased from pre to post resting state (Figure 5D). […]

Nevertheless, even though our results show a nominal increase in reactivation from learning to retrieval (see Figure 5D), due to experimental design features our data do not enable us to test for an hypothesized switch for sequential replay (see also “limitations” and Supplement 8).”

d) Introduction (last paragraph): "We examined the relationship of graph learning to reactivation and replay in a task where participants learned a ..." If all your behavioural analyses are based on retrieval performance, I think that you do not investigate graph learning (since you exclusively focus the analyses on retrieving the graph structure). However, relating the graph learning performance and replay/reactivation activity during learning trials (i.e., during graph learning) to retrieval trials might be interesting but beyond the scope of this paper.

We agree. We have changed the wording to be more accurate. Indeed, we do not examine graph learning but instead examine retrieval from a graph, after graph learning. The mentioned sentence now read

“[…] relationship of retrieval from a learned graph structure to reactivation [...]”

e) It is sometimes difficult to follow what phase of the experiment you refer to since you use the terms retrieval and test synonymously. Not a huge problem at all but maybe you want to stick to one term throughout the whole paper.

Thank you for pointing this out. We have now adapted the manuscript to exclusively refer to “retrieval” and not to “test”.

(2) Is your reactivation clustered?

In Figure 5A, you compare the reactivation strength of the two items following the cue image (i.e., current image trials) with items further away on the graph. I do not completely understand why your results are evidence for clustered reactivation in contrast to replay.

First, it would be interesting to see the reactivation of near vs. distant items before taking the difference (time course of item probabilities).

(copied answer from response to Reviewer 1, as the same remark was raised)

We have added the requested analysis showing the raw decoded probabilities for near and distant items separately in Figure 5A. We have chosen to replace Figure 5B with the new figure as we think that it offers more information than the previous Figure 5B. Instead, we have moved Figure 5B to the supplement. The median peak decoded accuracy for near and distant items is equivalent. We have added the following description to the figure:

“Decoded raw probabilities for off-screen items, that were up to two steps ahead of the current stimulus cue (‘near’,) vs. distant items that were more than two steps away on the graph, on trials with correct answers. The median peak decoded probability for near and distant items was at the same time point for both probability categories. Note that displayed lines reflect the average probability while, to eliminate influence of outliers, the peak displays the median. .”

Second, could it still be that the first item is reactivated before the second item? By averaging across both items, it becomes not apparent what the temporal courses of probabilities of both items look like (and whether they follow a sequential pattern). Additionally, the Gaussian smoothing kernel across the time dimension might diminish sequential reactivation and favour clustered reactivation. (In the manuscript, what does a Gaussian smoothing kernel of  = 1 refer to?). Could you please explain in more detail why you assume non-sequential clustered reactivation here and substantiate this with additional analyses?

We apologise for the unclear description. Note the Gaussian kernel is in fact only used for the reactivation analysis and not the replay analysis, so any small temporal successions would have been picked up by the sequential analysis. We now clarify this in the respective section of the sequential analysis and also explain the parameter of delta= 1 in the reactivation analysis section. The paragraph now reads

“[…] As input for the sequential analysis, we used the raw probabilities of the ten classifiers corresponding to the stimuli. [...]

[…] Therefore, to address this we applied a Gaussian smoothing kernel (using scipy.ndimage.gaussian_filter with the default parameter of σ=1 which corresponds approximately to taking the surrounding timesteps in both direction with the following weighting: current time step: 40%, ±1 step: 25%, ±2 step: 5%, ±3 step: 0.5%) [...]”

(3) Replay and/or clustered reactivation?

The relationship between the sequential forward replay, differential reactivation, and graph reactivation analysis is not really apparent. Wimmer et al. demonstrated that high performers show clustered reactivation rather than sequential reactivation. However, you did not differentiate in your differential reactivation analysis between high vs. low performers. (You point out in the discussion that this is due to a low number of low performers.)

We agree that a split into high vs low performers would have been preferably for our analysis. However, there is one major obstacle that made us opt for a correlational analysis instead: We employed criteria learning, rendering a categorical grouping conceptually biased. Even though not all participants reached the criteria of 80%, our sample did not naturally split between high and low performers but was biased towards higher performance, leaving the groups uneven. The median performance was 83% (mean ~81%), with six of our subjects (~1/4th of included participant) having this exact performance. This makes a median or mean split difficult, as either binning assignment choice would strongly affect the results. We have added a limitations section in which we extensively discuss this shortcoming and reasoning for not performing a median split as in Wimmer et al (2020). The section now reads:

“There are some limitations to our study, most of which originate from a suboptimal study design. [...], as we performed criteria learning, a sub-group analysis as in Wimmer et al., (2020) was not feasible, as median performance in our sample would have been 83% (mean 81%), with six participants exactly at that threshold. [...]”

It might be worth trying to bring the analysis together, for example by comparing sequential forward replay and differential reactivation at the beginning of graph learning (when performance is low) vs. retrieval (when performance is high).

Thank you for the suggestion to include the learning segments, which we think improves the paper quite substantially. However, analysis of the learning data turned out to be quite tricky> We had decided that each participant should perform as many blocks as necessary to reach at least 80% accuracy (with a limit of six and lower bound of two, see Supplement figure 4). Some participants learned 100% of the sequence after one block (these were mostly medical students, learning things by hard is their daily task). This in hindsight is an unfortunate design feature in relation to learning as it means different blocks are not directly comparable between participants.

In theory, we would expect that replay emerges in parallel with learning and then gradually change to clustered reactivation, as memory traces get consolidated/stronger. However, it is unclear when replay would emerge and when the switch to reactivation would happen. For this reason, we initially decided not to include the learning trials into the paper at all.

Nevertheless, to give some insight into the learning process and to see how consolidation effects differential reactivation and replay, we have split our data into pre and post resting state, aggregating all learning trials of each participant. While this does not allow us to track measures of interest on a block basis, it gives some (albeit limited) insight into the hypothesis outlined in our discussion.

For reactivation, we see a clear increase, further strengthening the outlined hypothesis, However, for replay the evidence is less obvious, potentially due to that fact that we do not know across how many learning blocks replay is to be expected.

The added section(s) now read:

“To examine how the 8-minute consolidation period affected reactivation we, post-hoc, looked at relevant measures during learning trials in contrast to retrieval trials. For all learning trial, for each participant, we calculated differential reactivation for the time point we found significant during the previous analysis (220-260 milliseconds). On average, differential reactivation probability increased from pre to post resting state (Figure 5D).

[…]

Nevertheless, even though our results show a nominal increase in reactivation from learning to retrieval (see Figure 5D), our data does not enable us to show an hypothesized switch for sequential replay (see also “limitations” and Supplement 8).”

Additionally, the main research question is not that clear to me. Based on the introduction, I thought the focus was on replay vs. clustered reactivation and high vs. low performance (which I think is really interesting). However, the title is more about reactivation strength and graph distance within cognitive maps. Are these two research questions related? And if so, how?

We agree we need to be clearer on this point. We have added two sentences to the introduction, which should address this point. The section now reads:

“[…] In particular, the question remains how the brain keeps track of graph distances for successful recall and whether the previously found difference between high and low performers also holds true within a more complex graph learning context.”

(4) Learning the graph structure.

I was wondering whether you have any behavioural measures to show that participants actually learn the graph structure (instead of just pairs or triplets of objects). For example, do you see that participants chose the distractor image that was closer to the target more frequently than the distractor image that was further away (close vs. distal target comparison)? It should be random at the beginning of learning but might become more biased towards the close target.

Thanks, this is an excellent suggestion. Our analysis indeed shows that people take the near lure more often than the far lure in later blocks, while it is random in the first block.

Nevertheless, we have decided to put these data into the supplement and reference it in the text. This is because analysis of the learning blocks is challenging and biased in general. Each participant had a different number of learning blocks based on their learning rate, and this makes it difficult to compare learning across participants. We have tried our best to accommodate and explain these difficulties in the figure legend. Nevertheless, we thank the referee for guidance here and this analysis indeed provides further evidence that participants learned the actual graph structure.

The added section reads

“Additionally, we have included an analysis showing how wrong answers participants provided were random in the first block and biased towards closer graph nodes in later blocks. This is consistent with participants actually learning the underlying graph structure as opposed to independent triplets (see figure and legend of Supplement 6 for details).”

(5) Minor comments

a) "Replay analysis relies on a successive detection of stimuli where the chance of detection exponentially decreases with each step (e.g., detecting two successive stimuli with a chance of 30% leaves a 9% chance of detecting the replay event). " Could you explain in more detail why 30% is a good threshold then?

Thank you. We have further clarified the section. As we are working mainly with probabilities, it is useful to keep in mind that accuracy is a class metric that only provides a rough estimate of classifier ability. Alternatively, something like a Top-3-Accuracy would be preferable, but also slightly silly in the context of 10 classes.

Nevertheless, subtle changes in probability estimates are present and can be picked up by the methods we employ. Therefore, the 30% is a rough lower bound and decided based on pilot data that showed that clean MEG data from attentive participants can usually reach this threshold. The section now reads:

“(e.g., detecting two successive stimuli with a chance of 30% leaves a 9% chance of detecting a replay event). However, one needs to bear in mind that accuracy is a “winnertakes-all” metric indicating whether the top choice also has the highest probability, disregarding subtle, relative changes in assigned probability. As the methods used in this analysis are performed on probability estimates and not class labels, one can expect that the 30% are a rough lower bound and that the actual sensitivity within the analysis will be higher. Additionally, based on pilot data, we found that attentive participants were able to reach 30% decodability, allowing us to use decodability as a data quality check. “

b) Could you make explicit how your decoders were designed? Especially given that you added null data, did you train individual decoders for one class vs. all other classes (n = 9 + null data) or one class vs. null data?

We added detail to the decoder training. The section now reads

“Decoders were trained using a one-vs-all approach, which means that for each class, a separate classifier was trained using positive examples (target class) and negative examples (all other classes) plus null examples (data from before stimulus presentation, see below). In detail, null data was.”

c) Why did you choose a ratio of 1:2 for your null data?

Our choice for using a higher ratio was based upon previous publications reporting better sensitivity of TDLM using higher ratios, as spatial sensor correlations are decreasing. Nevertheless, this choice was not well investigated beforehand. We have added more information to this to the manuscript

d) You could think about putting the questionnaire results into the supplement if they are sanity checks.

We have added the questionnaire results. However, due to the size of the tables, we have decided to add them as excel files into the supplementary files of the code repository. We have mentioned the existence file in the publication.

e) Figure 2. There is a typo in D: It says "Precessor Image" instead of "Predecessor Image".

Fixed typo in figure.

f) You write "Trials for the localizer task were created from -0.1 to 0.5 seconds relative to visual stimulus onset to train the decoders and for the retrieval task, from 0 to 1.5 seconds after onset of the second visual cue image." But the Figure legend 3D starts at -0.1 seconds for the retrieval test.

We have now clarified this. For the classifier cross-validation and transfer sanity check and clustered analysis we used trials from -0.1 to 0.5s, whereas for the sequenceness analysis of the retrieval, we used trials from 0 to 1.5 seconds

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

Learn more at Review Commons

---

Reply to the reviewers

We thank all reviewers for their thorough assessment and constructive comments. We are glad that the reviewers appreciate that our findings are of interest to the nuclear transport field and that our extension of the use of the RITE methodology can be a valuable tool for the further characterization of NPCs that differ in composition and potentially function. In response to the reviewers’ comments, we have revised the text to incorporate their suggestions and improve overall readability and clarity. Furthermore, we propose to perform a set of additional experiments to address the reviewers’ most important critiques. Below we list our response with the reviewer comments reprinted in dark grey and our response in blue for easier orientation. We have added numbering of the comments for easier orientation.

Many of the comments made by the reviewers have already been implemented, additional points will be addressed in a revised version of the manuscript as detailed below.

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

The authors extended the existing recombination-induced tag exchange (RITE) technology to show that they can image a subset of NPCs, improving signal-to-noise ratios for live cell imaging in yeast, and to track the stability or dynamics of specific nuclear pore proteins across multiple cell divisions. Further, the authors use this technology to show that the nuclear basket proteins Mlp1, Mlp2 and Pml39 are stably associated with "old NPCs" through multiple cell cycles. The authors show that the presence of Mlp1 in these "old NPCs" correlates with exclusion of Mlp1-positive NPCs from the nucleolar territory. A surprising result is that basket-less NPCs can be excluded from the non-nucleolar region, an observation that correlates with the presence of Nup2 on the NPC regardless of maturation state of the NPC. In support of the proposal that retention of NPCs via Mlp1 and Nup2 in non-nucleolar regions, simulation data is presented to suggest that basket-less NPCs diffuse faster in the plane of the nuclear envelope.

However, there are some points that do need addressing:

Major Points 1. Taking into account that the Nup2 result in Figure 4B forms the basis for one half of the proposed model in Figure 6 regarding the exclusion of NPCs from the nucleolar region of the NE, there is a relatively small amount of data in support of this finding and this proposed model. For example, the only data for Nup2 in the manuscript is a column chart in Figure 4B with no supporting fluorescence microscopy examples for any Nup2 deletion. Further, the Nup60 deletion mutant will have zero basket-containing NPCs, whereas the Nup2 deletion will be a mixture of basket-containing and basket-less NPCs. The only support for the localization of basket-containing NPCs in the Nup2 deletion mutant is through a reference "Since Mlp1-positive NPCs remain excluded from the nucleolar territory in nup2Δ cells (Galy et al., 2004), the homogenous distribution observed in this mutant must be caused predominantly by the redistribution of Mlp-negative NPCs into the nucleolar territory."

We have already added fluorescent images of the nup2d strain to figure 4A in the preliminary revision.

In addition, we will repeat the experiment from Galy et al. 2004 to test whether Mlp-positive NPCs are excluded from nucleoli in our hands as well.

Furthermore, we propose to carry out more experiments to pinpoint which domains of Nup2 contribute to nucleolar exclusion, which will provide more insight into the mechanism behind this effect. We propose to do this by analyzing NPC localization in mutants expressing truncations of Nup2 with deletions for individual domains as their only copy of Nup2. Regardless of whether we find a single domain of Nup2 responsible of a combinatorial action, this experiment will indicate a potential molecular mechanism for nucleolar exclusion.

1. The authors could consider utilizing this opportunity to discuss their technological innovations in the context of the prior work of Onischenko et al., 2020. This work is referenced for the statement "RITE can be used to distinguish between old and new NPCs" Page 2, Line 43. However, it is not referenced for the statement "We constructed a RITE-cassette that allows the switch from a GFP-labelled protein to a new protein that is not fluorescently labelled (RITE(GFP-to-dark))" despite Onischenko et al., 2020 having already constructed a RITE-cassette for the GFP-to-dark transition. The authors could consider taking this opportunity to instead focus on their innovative approach to apply this technology to decrease the number of fluorescently-tagged NPCs by dilution across multiple cell divisions and to interpret this finding as a measure of the stability of nuclear pore proteins within the broader NPC.

We apologize for this imprecise citation. We have modified the text to indicate that our RITE cassette was previously used in two publications. It now reads: “We used a RITE-cassette that allows the switch from a GFP-labelled protein to a new protein that is not fluorescently labelled (RITE(GFP-to-dark)) (Onischenko et al., 2020, Kralt et al., 2022). “

1. The authors could also consider taking this opportunity to discuss their results in the context of the Saccharomyces cerevisiae nuclear pore complex structures published e.g. in Kim et al., 2018, Akey et al., 2022, Akey et al., 2023 in which the arrangement of proteins in the nuclear basket is presented, and also work from the Kohler lab (Mészáros et al., 2015) on how the basket proteins are anchored to the NPC. There is additional literature that also might help provide some perspective to the findings in the current manuscript, such as the observation that a lesser amount of Mlp2 to Mlp1 observed is consistent with prior work (e.g. Kim et al., 2018) and that intranuclear Mlp1 foci are also formed after Mlp1 overexpression (Strambio-de-Castillia et al., 1999).

Following the reviewer’s suggestion, we extended our discussion of basket Nup stoichiometry and organization in the discussion section including several of the citations mentioned. At this point, we did not see a good way to incorporate discussion about the nuclear Mlp1 foci formed after Mlp1 overexpression. However, this observation is in line with the foci formed in cells lacking Nup60, suggesting that Mlp1 that cannot be incorporated into NPCs forms nuclear foci.

Minor Points 1. What is the "lag time" of the doRITE switching? Do the authors believe that it is comparable to the approximate 1-hour timeframe following beta-estradiol induction as shown previously in Chen et al. Nucleic Acids Research, Volume 28, Issue 24, 15 December 2000, Page e108, https://doi.org/10.1093/nar/28.24.e108

Our data (e.g. newRITE, Figure S3B) suggest that the switch occurs on a similar timeframe at

1. The authors could consider a brief explanation of radial position (um) for the benefit of the reader, in Figures 1E (right panel) and 2B (right panel), perhaps using a diagram to make it easier to understand the X-axis (um).

To address this, we have now included a diagram and refer to it in the figure legend.

1. In Figure 1G, would the authors consider changing the vertical axis title and the figure legend wording from "mean number of NPCs per cell" to "mean labeled NPC # per cell" to reflect that what is being characterized are the remaining GFP-bearing NPCs over time?

Thank you for spotting this inaccuracy. We have changed the label to “mean # of labeled NPCs per cell”.

1. In Figure 2C, the magenta-labeled protein in the micrographs is not described in the figure or the legend.

As requested, a description has been added in figure and legend.

1. In Figure S2A, there is an arrow indicating a Nup159 focus, but this is not described in the figure legend, as is done in Figure 2C.

A description has been added to the legend.

1. In Figure S3C, the figure legend does not match the figure. Was this supposed to be designed like Figure 3C and is missing part of the figure? Or is the legend a typographical error?

We apologize for this error and thank the reviewer for spotting it. The legend has been corrected.

1. In Figure S4B, the spontaneously recombined RITE (GFP-to-dark) Nup133-V5 appears in the western blot as equally abundant to pre-recombined Nup133-V5-GFP. In the figure legend, this is explained as cells grown in synthetic media without selection to eliminate cells that have lost their resistance marker from the population. In Cheng et al. Nucleic Acids Res. 2000 Dec 15; 28(24): e108, Cre-EBD was not active in the absence of B-estradiol, despite galactose-induced Cre-EBD overexpression. Would the authors be able to comment further on the Cre-Lox RITE system in the manuscript?

We note that also in the cited publication, cells are grown in the presence of selection to select (as stated in this publication) “against pre-excision events that occur because of low but measurable basal expression of the recombinase”. Although the authors report that spontaneous recombination is reduced with the b-estradiol inducible system (compared to pGAL expression control of the recombinase only), they show negligible spontaneous recombination only within a two-hour time window. Indeed, we also observe low levels of uninduced recombination on a short timeframe, but occasional events can become significant in longer incubation times (e.g. overnight growth) in the absence of selection. It should be noted that in our system, Cre expression is continuously high (TDH3-promoter) and not controlled by an inducible GAL promoter. We have added the information about the promoter controlling Cre-expression in the methods section.

1. In Figure 6, the authors may want to consider inverting the flow of the cartoon model to start from the wild type condition and apply the deletion mutations at each step to "arrive" at the mutant conditions, rather than starting with mutant conditions and "adding back" proteins.

Following the suggestions of the reviewer, we have modified our model to more clearly represent the contributions of the different basket components.

Reviewer #1 (Significance (Required)):

Recent work has drawn attention to the fact that not all NPCs are structurally or functionally the same, even within a single cell. In this light, the work here from Zsok et al. is an important demonstration of the kind of methodologies that can shed light on the stability and functions of different subpopulations of NPCs. Altogether, these data are used to support an interesting and topical model for Nup2 and nuclear-basket driven retention of NPCs in non-nucleolar regions of the nuclear envelope.

We thank the reviewer for this positive assessment of our work.

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

In this study, Zsok et al. develop innovative methods to examine the dynamics of individual nuclear pore complexes (NPCs) at the nuclear envelope of budding yeast. The underlying premise is that with the emergence of biochemically distinct NPCs that co-exist in the same cell, there is a need to develop tools to functionally isolate and study them. For example, there is a pool of NPCs that lack the nuclear basket over the nucleolus. Although the nature of this exclusion has been investigated in the past, the authors take advantage of a modification of recombination induced tag exchange (RITE), the slow turnover of scaffold nups, the closed mitosis of budding yeast, and extensive high quality time lapse microscopy to ultimately monitor the dynamics of individual NPCs over the nucleolus. By leveraging genetic knockout approaches and auxin-induced degradation with sophisticated quantitative and rigorous analyses, the authors conclude that there may be two mechanisms dependent on nuclear basket proteins that impact nucleolar exclusion. They also incorporate some computational simulations to help support their conclusions. Overall, the data are of the highest quality and are rigorously quantified, the manuscript is well written, accessible, and scholarly - the conclusions are thus on solid footing.

We thank the reviewer for this assessment.

Reviewer #2 (Significance (Required)):

I have no concerns about the data or the conclusions in this manuscript. However, the significance is not overly clear as there is no major conceptual advance put forward, nor is there any new function suggested for the NPCs over nucleoli. As NPCs are immobile in metazoans, the significance may also be limited to a specialized audience.

We respectfully disagree with this assessment. It is becoming increasingly clear that NPC variants are also present in other model systems. We characterize the interaction between conserved nuclear components, the NPC, the nucleolus and chromatin. While the specific architecture of the nucleus varies between species, many of these interactions are conserved. For example, Nup50, the homologue of Nup2, interacts with chromatin also in other systems including mammalian cells and thus may contribute to regulating the interplay between the nuclear basket and adjoining chromatin. Furthermore, our work demonstrates the use of a novel approach in the application of RITE that can be useful for other researchers in the field of NPC biology and beyond. For example, doRITE could be applied to study the properties of aged NPCs in the context of young cells. In the revised manuscript, we attempt to better highlight and discuss the conceptual advances of our manuscript.

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

The manuscript of Zsok et al. describes the role of nuclear basket proteins in the distribution and mobility of nuclear pore complexes in budding yeast. In particular, the authors showed that the doRITE approach can be used for the analysis of stable and dynamically associated NUPs. Moreover, it can distinguish individual NUPs and follow the inheritance of individual NPCs from mother to daughter cells. The author's findings highlight that Mlp1, Mlp2, and Pml39 are stably associated with the nuclear pore; deletion of Mlp1-Mlp2 and Nup60 leads to the higher NPC density in the nucleolar territory; and NPCs exhibit increased mobility in the absence of the nuclear basket components.

The manuscript contains most figures supporting the data, and data supports the conclusions. However, authors need to include better explanations for figures in the text and figure legends. Lack of detailed explanation can pose challenges for non-experts. In addition, the authors jump over figures and shuffle them through the manuscript, which disrupts the flow and coherence of the manuscript.

We thank the reviewer for pointing this out. We have modified the figure legends throughout the manuscript in an attempt to make them more accessible to the reader. In addition, we will revise the figure order and text as suggested to improve the flow of the manuscript.

Major comments: 1) The nuclear basket contains Nup1, Nup2, Nup60, Mlp1, and Mlp2 in yeast. Nup60 works as a seed for Mlp1/Mlp2 and Nup2 recruitment and plays a key role in the assembly of nuclear pore basket scaffold (PMID: 35148185). Logically, the authors focused primarily on Nup60 in the current manuscript. However, NUP153 has another ortholog of yeast - Nup1, which has not been studied in this work. I recommend adjusting the title of the manuscript to: Nup60 and Mlp1/Mlp2 regulate the distribution and mobility of nuclear pore complexes in budding yeast. I also suggest discussing why work on Nup1 was not included/performed in the manuscript.

We have changed the title to “Nuclear basket proteins regulate the distribution and mobility of nuclear pore complexes in budding yeast”. We think that this better captures the essence of our manuscript than listing all four proteins (Mlp1/2, Nup60 and Nup2) in the title.

We initially focused on the network that is involved in Mlp1/2 interaction at the NPC. However, we agree that it would be interesting to test, whether Nup1 plays a role in the analyzed processes as well. Since Nup1 is essential in our yeast background, we will use auxin-inducible degradation of Nup1 to test its involvement in NPC distribution.

2) Figure 2B: I suggest choosing a more representative image for Pml39. It looks not like a stable component but rather dynamic as NUP60 or Gle1 based on figure showed in Figure 2B.

Due to its lower copy number, Pml39 is much more difficult to visualize than the other Nups. To guide the reader, we have now added arrow heads to point to remaining Pml39 foci at the 14 hour timepoint. The 11 hour time point most clearly show that Pml39 is less dynamic than other Nups such as Nup116, Nup60 or Gle1. At this time point, clear dots for Pml39 can be detected, while e.g. Nup116 in the same figure exhibits a more distributed signal and the signal for Nup60 and Gle1 is no longer visible. We will describe this more clearly in our revised manuscript as well.

3) Depletion of AID-tagged proteins needs to be supported by Western blot analysis with protein-specific antibodies, and PCR results should be included in supplementary data to demonstrate the homozygosity of the strains.

The correct genomic tagging of the depleted proteins by AID was confirmed by PCR. We will include this PCR analysis in the supplemental data. Please note that we are working with haploid yeast cells. Therefore, all strains only carry a single copy of the genes. Unfortunately, we do not have protein-specific antibodies against the depleted proteins. However, the Mlp1-mislocalization phenotype demonstrates that depletion of Nup60 is successful and the depletion strain for PolII depletion was used and characterized previously (PMID: 31753862, PMID: 36220102).

4) Figure 5B: Snapshots of images from the movie are required. There are no images, only quantifications.

We have replaced the supplemental movie with a movie showing the detection by Trackmate as well as overlaid tracks. As requested, a snapshot of this movie was inserted in figure 5B. We have also moved the example tracks from the supplement to the main figure. Furthermore, we will deposit the tracking dataset in the ETH Research Collection to make it available to the community.

5) Description of figure legends is more technical than supporting/explaining the figure. For example, below my suggestions for Figure 1D. Please, consider more detailed explanation for other figures. (D) Left: Schematic of the RITE cassette. NUP of interest is tagged with V5 tag and eGFP fluorescent protein where LoxP sites flank eGFP. Before the beta-estradiol-induced recombination, the old NPCs are marked with eGFP signal, whereas new NPCs lack an eGFP signal after the recombination. ORF: open reading frame; V5: V5-tag; loxP: loxP recombination site; eGFP: enhanced green fluorescent protein. Right: doRITE assay schematic of stable or dynamic Nup behavior over cell divisions in yeast after the recombination.

We have modified the figure legends throughout the manuscript to make them more explanatory and helpful for the reader.

In addition, I recommend highlighting the result in the title of the figures. Please, re-consider titles for Figure S3.

We have revised the title for Figure S3 to state a result. It now reads: “Mlp1 truncations localize preferentially to non-nucleolar NPCs.”

Minor: i) P.1 Line 31. Extra period symbol before the "(Figure 1A)".

Fixed

ii) P.2 Line 10. Inconsistent writing of PML39 and MLP1. Both genes are capitalized. The same for P.4 Line 16. In some cases all letters are capitalized in other only the first one.

We are following the official yeast gene nomenclature by spelling gene names in italicized capitals and protein names with only the first letter capitalized. We are sorry that this can be confusing for readers more familiar with other model systems but we adhere to the accepted yeast nomenclature standards.

iii) P.2 Line 18-22. The sentence is too long and hard to read. I recommend splitting it into two sentences.

We agree and have fixed this.

iv) P.2-3 Line 46-47. The sentence is unclear. Suggestion: We expected that successive cell divisions would dilute the signal of labelled and stably associated with the NPC nucleoporins. By contrast, ...

We have modified the sentence to read: “When tagging a Nup that stably associates with the NPC, we expected that successive cell divisions would dilute labelled NPCs by inheritance to both mother and daughter cells leading to a low density of labelled NPCs. By contrast,…”

v) P.4 Line 17-21. Please, consider adding extra information and clarifying lines 19-21. For example, in Line 19 Figure 2B you can add that the reader needs to compare row 1 and row 4.

Thank you, we have fixed this as suggested.

vi) P. 5 Line 15. When a number begins a sentence, that number should always be spelled out. You can pe-phrase the sentence to avoid it. Also, I recommend adding an explanation/hypothesis of why new NPCs are less frequently detected in nucleolar territory.

We have formatted the text. Interestingly, new NPCs are more frequently detected in the nucleolar territory. We have reformulated this section to make it clearer, also in response to the next comment.

vii) P.5 Line 17-22. I recommend re-phrasing these two sentences. Logically, it is clear that Mlp1/Mlp2 loss mimics "old NPCs" to look more like "new NPCs", and for that reason, they are more frequently included in the nucleolar territory, but it is not clear when you read these two sentences from the first time.

We have reformulated this section to make it clearer.

viii) P6. Line 16. No figure supporting data on graph (Figure 3B).

We have added fluorescent images of the nup2d strain to figure 4A.

ix) P.7 Line 10-13. The sentence is unclear.

We have shortened the sentence and moved part of the content to the discussion in the next paragraph.

x) P.13,14 etc. If 0h timepoint has been used for normalization, why is it present on the graph?

The 0h timepoint is shown for comparison and to illustrate the standard deviation in the data.

xi) P.15. Line 32-33. There is no image here. Potentially wrong description of the figure.

Thank you for spotting this. This was fixed.

xii) Figures: - Inconsistent labeling of figures. For example, Fig.1, Fig.1S, Figure 2 etc.

Thank you, this has been corrected.

• Inconsistent labeling of figures. For example, Fig.1 G "mean number of NPCs per cell" - no capitalization of the first letter. Fig.1S "Fraction in population" is capitalize d. In general, titles of axis should be capitalized.

Thank you for spotting this. This was fixed.

Suggestions for Figure 1D and Figure 6 are attached as a separate file.

We thank the reviewer for their suggestions to improve these figures. We have taken their recommendation and revised the figures accordingly (see also response to reviewer 1, minor point 8).

Reviewer #3 (Significance (Required)):

Zsok et al. used the recombination-induced tag exchange (RITE) approach, which is an interesting and powerful method to follow individual NUPs over time with respect to their localization and abundance. This approach has been used before in PMID: 36515990 to distinguish pre-existing and newly synthesized Nup2 populations and has been extended to other basket NUPs in this work. Using this method, the authors support the earlier data on basket nucleoporins and highlight new insights on how basket nucleoporins regulate NPCs distribution and mobility. Overall, the manuscript provides new details on the stability of nucleoporins in yeast and how these data align with the mass spectrometry and FRAP data performed earlier in other studies. The limitation of this study is the absence of data on Nup1. It was unclear why these data were not present. Additional data can be included on the dynamics of Pml39, for example, using the FRAP method. The dynamic of Pml39 at the pore was shown only using the doRITE method.

As suggested, we propose to test whether Nup1 influences NPC organization (see also above). Unfortunately, we are not able to provide orthologous data for the dynamics of Pml39. As we have discussed in the manuscript, FRAP is not suitable for the analysis of the dynamics of most nucleoporins in yeast due to the high lateral mobility of NPCs in the nuclear envelope and has previously generated misleading results for Mlp1. Furthermore, the low expression levels of Pml39 will make it difficult to obtain reliable FRAP curves for this protein. We therefore do not think that adding FRAP experiments with Pml39 will provide valuable insight.

However, in addition to the Pml39 doRITE result itself, our observation that the Pml39-dependent pool of Mlp1 exhibits stable association with the NPC supports the interpretation of Pml39 as a stable protein as well.

In general, this study represents a unique research study of basic research on nuclear pore proteins that will be of general interest to the nuclear transport field.

Field of expertise: nuclear-cytoplasmic transport, nuclear pore, inducible protein degradation. I do not have sufficient expertise in ExTrack.

Author response:

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

eLife assessment

This important study advances our understanding of how past and future information is jointly considered in visual working memory by studying gaze biases in a memory task that dissociates the locations during encoding and memory tests. The evidence supporting the conclusions is convincing, with state-of-the-art gaze analyses that build on a recent series of experiments introduced by the authors. This work, with further improvements incorporating the existing literature, will be of broad interest to vision scientists interested in the interplay of vision, eye movements, and memory.

We thank the Editors and the Reviewers for their enthusiasm and appreciation of our task, our findings, and our article. We also wish to thank the Reviewers for their constructive comments that we have embraced to improve our article. Please find below our point-by-point responses to this valuable feedback, where we also state relevant revisions that we have made to our article.

In addition, please note that we have now also made our data and code publicly available.

Reviewer 1, Comments:

In this study, the authors offer a fresh perspective on how visual working memory operates. They delve into the link between anticipating future events and retaining previous visual information in memory. To achieve this, the authors build upon their recent series of experiments that investigated the interplay between gaze biases and visual working memory. In this study, they introduce an innovative twist to their fundamental task. Specifically, they disentangle the location where information is initially stored from the location where it will be tested in the future. Participants are tasked with learning a novel rule that dictates how the initial storage location relates to the eventual test location. The authors leverage participants' gaze patterns as an indicator of memory selection. Intriguingly, they observe that microsaccades are directed toward both the past encoding location and the anticipated future test location. This observation is noteworthy for several reasons. Firstly, participants' gaze is biased towards the past encoding location, even though that location lacks relevance to the memory test. Secondly, there's a simultaneous occurrence of an increased gaze bias towards both the past and future locations. To explore this temporal aspect further, the authors conduct a compelling analysis that reveals the joint consideration of past and future locations during memory maintenance. Notably, microsaccades biased towards the future test location also exhibit a bias towards the past encoding location. In summary, the authors present an innovative perspective on the adaptable nature of visual working memory. They illustrate how information relevant to the future is integrated with past information to guide behavior.

Thank you for your enthusiasm for our article and findings as well as for your constructive suggestions for additional analyses that we respond to in detail below.

This short manuscript presents one experiment with straightforward analyses, clear visualizations, and a convincing interpretation. For their analysis, the authors focus on a single time window in the experimental trial (i.e., 0-1000 ms after retro cue onset). While this time window is most straightforward for the purpose of their study, other time windows are similarly interesting for characterizing the joint consideration of past and future information in memory. First, assessing the gaze biases in the delay period following the cue offset would allow the authors to determine whether the gaze bias towards the future location is sustained throughout the entire interval before the memory test onset. Presumably, the gaze bias towards the past location may not resurface during this delay period, but it is unclear how the bias towards the future location develops in that time window. Also, the disappearance of the retro cue constitutes a visual transient that may leave traces on the gaze biases which speaks again for assessing gaze biases also in the delay period following the cue offset.

Thank you for raising this important point. We initially focused on the time window during the cue given that our central focus was on gaze-biases associated with mnemonic item selection. By zooming in on this window, we could best visualize our main effects of interest: the joint selection (in time) of past and future memory attributes.

At the same time, we fully agree that examining the gaze biases over a more extended time window yields a more comprehensive view of our data. To this end, we have now also extended our analysis to include a wider time range that includes the period between cue offset (1000 ms after cue onset) and test onset (1500 ms after cue onset). We present these data below. Because we believe our future readers are likely to be interested in this as well, we have now added this complementary visualization as Supplementary Figure 4 (while preserving the focus in our main figure on the critical mnemonic selection period of interest).

Author response image 1.

Supplementary Figure 4. Gaze biases in extended time window as a complement to Figure 1 and Supplementary Figure 2. This extended analysis reveals that while the gaze bias towards the past location disappears around 600 ms after cue onset, the gaze bias towards the future location persists (panel a) and that while the early (joint) future bias occurs predominantly in the microsaccade range below 1 degree visual angle, the later bias to the future location incorporates larger eye movement that likely involve preparing for optimally perceiving the anticipated test stimulus (panel b).

This extended analysis reveals that while the gaze bias towards the past location disappears around 600 ms after cue onset (consistent with our prior reports of this bias), the gaze bias towards the future location persists. Moreover, as revealed by the data in panel b above, while the early (joint) future bias occurs predominantly in the microsaccade range below 1 degree visual angle, the later bias to the future location incorporates larger eye movement that likely involve preparing for optimally perceiving the anticipated test stimulus.

We now also call out these additional findings and figure in our article:

Page 2 (Results): “Gaze biases in both axes were driven predominantly by microsaccades (Supplementary Fig. 2) and occurred similarly in horizontal-to-vertical and vertical-tohorizontal trials (Supplementary Fig. 3). Moreover, while the past bias was relatively transient, the future bias continued to increase in anticipation of the of the test stimulus and increasingly incorporated eye-movements beyond the microsaccade range (see Supplementary Fig. 4 for a more extended time range)”.

Moreover, assessing the gaze bias before retro-cue onset allows the authors to further characterize the observed gaze biases in their study. More specifically, the authors could determine whether the future location is considered already during memory encoding and the subsequent delay period (i.e., before the onset of the retro cue). In a trial, participants encode two oriented gratings presented at opposite locations. The future rule indicates the test locations relative to the encoding locations. In their example (Figure 1a), the test locations are shifted clockwise relative to the encoding location. Thus, there are two pairs of relevant locations (each pair consists of one stimulus location and one potential test location) facing each other at opposite locations and therefore forming an axis (in the illustration the axis would go from bottom left to top right). As the future rule is already known to the participants before trial onset it is possible that participants use that information already during encoding. This could be tested by assessing whether more microsaccades are directed along the relevant axis as compared to the orthogonal axis. The authors should assess whether such a gaze bias exists already before retro cue onset and discuss the theoretical consequences for their main conclusions (e.g., is the future location only jointly used if the test location is implicitly revealed by the retro cue).

Thank you – this is another interesting point. We fully agree that additional analysis looking at the period prior to retrocue onset may also prove informative. In accordance with the suggested analysis, we have therefore now also analysed the distribution of saccade directions (including in the period from encoding to retrocue) as a function of the future rule (presented below, and now also included as Supplementary Fig. 5). Complementary recent work from our lab has shown how microsaccade directions can align to the axis of memory contents during retention (see de Vries & van Ede, eNeuro, 2024). Based on this finding, one may predict that if participants retain the items in a remapped fashion, their microsaccades may align with the axis of the future rule, and this could potentially already happen prior to cue onset.

These complementary analyses show that saccade directions are predominantly influenced by the encoding locations rather than the test locations, as seen most clearly by the saccade distribution plots in the middle row of the figure below. To obtain time-courses, we categorized saccades as occurring along the axis of the future rule or along the orthogonal axis (bottom row of the figure below). Like the distribution plots, these time course plots also did not reveal any sign of a bias along the axis of the future rule itself.

Importantly, note how this does not argue against our main findings of joint selection of past and future memory attributes, as for that central analysis we focused on saccade biases that were specific to the selected memory item, whereas the analyses we present below focus on biases in the axes in which both memory items are defined; not only the cued/selected memory item.

Author response image 2.

Supplementary Figure 5. Distribution of saccade directions relative to the future rule from encoding onset. (Top panel) The spatial layouts in the four future rules. (Middle panel) Polar distributions of saccades during 0 to 1500 ms after encoding onset (i.e., the period between encoding onset and cue onset). The purple quadrants represent the axis of the future rule and the grey quadrants the orthogonal axis. (Bottom panel) Time courses of saccades along the above two axes. We did not observe any sign of a bias along the axis of the future rule itself.

We agree that these additional results are important to bring forward when we interpret our findings. Accordingly, we now mention these findings at the relevant section in our Discussion:

Page 5 (Discussion): “First, memory contents could have directly been remapped (cf. 4,24–26) to their future-relevant location. However, in this case, one may have expected to exclusively find a future-directed gaze bias, unlike what we observed. Moreover, using a complementary analysis of saccade directions along the axis of the future rule (cf. 24), we found no direct evidence for remapping in the period between encoding and cue (Supplementary Fig. 5)”.

Reviewer 2, Comments:

The manuscript by Liu et al. reports a task that is designed to examine the extent to which "past" and "future" information is encoded in working memory that combines a retro cue with rules that indicate the location of an upcoming test probe. An analysis of microsaccades on a fine temporal scale shows the extent to which shifts of attention track the location of the location of the encoded item (past) and the location of the future item (test probe). The location of the encoded grating of the test probe was always on orthogonal axes (horizontal, vertical) so that biases in microsaccades could be used to track shifts of attention to one or the other axis (or mixtures of the two). The overall goal here was then to (1) create a methodology that could tease apart memory for the past and future, respectively, (2) to look at the time-course attention to past/future, and (3) to test the extent to which microsaccades might jointly encode past and future memoranda. Finally, some remarks are made about the plausibility of various accounts of working memory encoding/maintenance based on the examination of these time courses.

Strengths:

This research has several notable strengths. It has a clear statement of its aims, is lucidly presented, and uses a clever experimental design that neatly orthogonalizes "past" and "future" as operationalized by the authors. Figure 1b-d shows fairly clearly that saccade directions have an early peak (around 300ms) for the past and a "ramping" up of saccades moving in the forward direction. This seems to be a nice demonstration the method can measure shifts of attention at a fine temporal resolution and differentiate past from future-oriented saccades due to the orthogonal cue approach. The second analysis shown in Figure 2, reveals a dependency in saccade direction such that saccades toward the probe future were more likely also to be toward the encoded location than away from the encoded direction. This suggests saccades are jointly biased by both locations "in memory".

Thank you for your overall appreciation of our work and for highlighting the above strengths. We also thank you for your constructive comments and call for clarifications that we respond to below.

Weaknesses:

(1) The "central contribution" (as the authors characterize it) is that "the brain simultaneously retains the copy of both past and future-relevant locations in working memory, and (re)activates each during mnemonic selection", and that: "... while it is not surprising that the future location is considered, it is far less trivial that both past and future attributes would be retained and (re)activated together. This is our central contribution." However, to succeed at the task, participants must retain the content (grating orientation, past) and probe location (future) in working memory during the delay period. It is true that the location of the grating is functionally irrelevant once the cue is shown, but if we assume that features of a visual object are bound in memory, it is not surprising that location information of the encoded object would bias processing as indicated by microsaccades. Here the authors claim that joint representation of past and future is "far less trivial", this needs to be evaluaed from the standpoint of prior empirical data on memory decay in such circumstances, or some reference to the time-course of the "unbinding" of features in an encoded object.

Thank you. We agree that our participants have to use the future rule – as otherwise they do not know to which test stimulus they should respond. This was a deliberate decision when designing the task. Critically, however, this does not require (nor imply) that participants have to incorporate and apply the rule to both memory items already prior to the selection cue. It is at least as conceivable that participants would initially retain the two items at their encoded (past) locations, then wait for the cue to select the target memory item, and only then consider the future location associated with the target memory item. After all, in every trial, there is only 1 relevant future location: the one associated with the cued memory item. The time-resolved nature of our gaze markers argues against such a scenario, by virtue of our observation of the joint (simultaneous) consideration of past and future memory attributes (as opposed to selection of past-before-future). These temporal dynamics are central to the insights provided by our study.

In our view, it is thus not obvious that the rule would be applied at encoding. In this sense, we do not assume that the future location is part of both memory objects from encoding, but rather ask whether this is the case – and, if so, whether the future location takes over the role of the past location, or whether past and future locations are retained jointly.

Our statements regarding what is “trivial” and what is “less trivial” regard exactly this point: it is trivial that the future is considered (after all, our task demanded it). However, it is less trivial that (1) the future location was already available at the time of initial item selection (as reflected in the simultaneous engagement of past and future locations), and (2) that in presence of the future location, the past location was still also present in the observed gaze biases.

Having said that, we agree that an interesting possibility is that participants remap both memory items to their future-relevant locations ahead of the cue, but that the past location is not yet fully “unbound” by the time of the cue. This may trigger a gaze bias not only to the new future location but also to the “sticky” (unbound) past location. We now acknowledge this possibility in our discussion (also in response to comment 3 below) where we also suggest how future work may be able to tap into this:

Page 6 (Discussion): “In our study, the past location of the memory items was technically irrelevant for the task and could thus, in principle, be dropped after encoding. One possibility is that participants remapped the two memory items to their future locations soon after encoding, and had started – but not finished – dropping the past location by the time the cue arrived. In such a scenario, the past signal is merely a residual trace of the memory items that serves no purpose but still pulls gaze. Alternatively, however, the past locations may be utilised by the brain to help individuate/separate the two memory items. Moreover, by storing items with regard to multiple spatial frames (cf. 37) – here with regard to both past and future visual locations – it is conceivable that memories may become more robust to decay and/or interference. Also, while in our task past locations were never probed, in everyday life it may be useful to remember where you last saw something before it disappeared behind an occluder. In future work, it will prove interesting to systematically vary to the delay between encoding and cue to assess whether the reliance on the past location gradually dissipates with time (consistent with dropping an irrelevant feature), or whether the past trace remains preserved despite longer delays (consistent with preserving utility for working memory).”

(2) The authors refer to "future" and "past" information in working memory and this makes sense at a surface level. However, once the retrocue is revealed, the "rule" is retrieved from long-term memory, and the feature (e.g. right/left, top/bottom) is maintained in memory like any other item representation. Consider the classic test of digit span. The digits are presented and then recalled. Are the digits of the past or future? The authors might say that one cannot know, because past and future are perfectly confounded. An alternative view is that some information in working memory is relevant and some is irrelevant. In the digit span task, all the digits are relevant. Relevant information is relevant precisely because it is thought be necessary in the future. Irrelevant information is irrelevant precisely because it is not thought to be needed in the immediate future. In the current study, the orientation of the grating is relevant, but its location is irrelevant; and the location of the test probe is also relevant.

Thank you for this stimulating reflection. We agree that in our set-up, past location is technically “task-irrelevant” while future location is certainly “task-relevant”. At the same time, the engagement of the past location suggests to us that the brain uses past location for the selection – presumably because the brain uses spatial location to help individuate/separate the items, even if encoded locations are never asked about. Therefore, whether something is relevant or irrelevant ultimately depends on how one defines relevance (past location may be relevant/useful for the brain even if technically irrelevant from the perspective of the task). In comparison, the use of “past” and “future” may be less ambiguous.

It is also worth noting how we interpret our findings in relation to demands on visual working memory, inspired by dynamic situations whereby visual stimuli may be last seen at one location but expected to re-appear at another, such as a bird disappearing behind a building (the example in our introduction). Thus, past for us does not refer to the memory item perse (like in the digit span analogue) but, rather, quite specifically to the past location of a dynamic visual stimulus in memory (which, in our experiment, was operationalised by the future rule, for convenience).

(3) It is not clear how the authors interpret the "joint representation" of past and future. Put aside "future" and "past" for a moment. If there are two elements in memory, both of which are associated with spatial bindings, the attentional focus might be a spatial average of the associated spatial indices. One might also view this as an interference effect, such that the location of the encoded location attracts spatial attention since it has not been fully deleted/removed from working memory. Again, for the impact of the encoded location to be exactly zero after the retrieval cue, requires zero interference or instantaneous decay of the bound location information. It would be helpful for the authors to expand their discussion to further explain how the results fit within a broader theoretical framework and how it fits with empirical data on how quickly an irrelevant feature of an object can be deleted from working memory.

Thank you also for this point (that is related to the two points above). As we stated in our reply to comment 1 above, we agree that one possibility is that the past location is merely “sticky” and pulls the task-relevant future bias toward the past location. If so, our time courses suggest that such “pulling” occurs only until approximately 600 ms after cue onset, as the past bias is only transient. An alternative interpretation is that the past location may not be merely a residual irrelevant trace, but actually be useful and used by the brain.

For example, the encoded (past) item locations provide a coordinate system in which to individuate/separate the two memory items. While the future locations also provide such a coordinate system, the brain may benefit from holding onto both coordinate systems at the same time, rendering our observation of joint selection in both frames. Indeed, in a recent VR experiment in which we had participants (rather than the items) rotate, we also found evidence for the joint use of two spatial frames, even if neither was technically required for the upcoming task (see Draschkow, Nobre, van Ede, Nature Human Behaviour, 2022). Though highly speculative at this stage, such reliance on multiple spatial frames may make our memories more robust to decay and/or interference. Moreover, while past location was never explicitly probed in our task, in daily life the past location may sometimes (unexpectedly) become relevant, hence it may be useful to hold onto it, just in case. Thus, considering the past location merely as an “irrelevant feature” (that takes time to delete) may not do sufficient justice to the potential roles of retaining past locations of dynamic visual objects held in working memory.

As also stated in response to comment 1 above, we now added these relevant considerations to our Discussion:

Page 5 (Discussion): “In our study, the past location of the memory items was technically irrelevant for the task and could thus, in principle, be dropped after encoding. One possibility is that participants remapped the two memory items to their future locations soon after encoding, and had started – but not finished – dropping the past location by the time the cue arrived. In such a scenario, the past signal is merely a residual trace of the memory items that serves no purpose but still pulls gaze. Alternatively, however, the past locations may be utilised by the brain to help individuate/separate the two memory items. Moreover, by storing items with regard to multiple spatial frames (cf. 37) – here with regard to both past and future visual locations – it is conceivable that memories may become more robust to decay and/or interference. Also, while in our task past locations were never probed, in everyday life it may be useful to remember where you last saw something before it disappeared behind an occluder. In future work, it will prove interesting to systematically vary to the delay between encoding and cue to assess whether the reliance on the past location gradually dissipates with time (consistent with dropping an irrelevant feature), or whether the past trace remains preserved despite longer delays (consistent with preserving utility for working memory).”

Reviewer 3, Comments:

This study utilizes saccade metrics to explore, what the authors term the "past and future" of working memory. The study features an original design: in each trial, two pairs of stimuli are presented, first a vertical pair and then a horizontal one. Between these two pairs comes the cue that points the participant to one target of the first pair and another of the second pair. The task is to compare the two cued targets. The design is novel and original but it can be split into two known tasks - the first is a classic working memory task (a post-cue informs participants which of two memorized items is the target), which the authors have used before; and the second is a classic spatial attention task (a pre-cue signal that attention should be oriented left or right), which was used by numerous other studies in the past. The combination of these two tasks in one design is novel and important, as it enables the examination of the dynamics and overlapping processes of these tasks, and this has a lot of merit. However, each task separately is not new. There are quite a few studies on working memory and microsaccades and many on spatial attention and microsaccades. I am concerned that the interpretation of "past vs. future" could mislead readers to think that this is a new field of research, when in fact it is the (nice) extension of an existing one. Since there are so many studies that examined pre-cues and post-cues relative to microsaccades, I expected the interpretation here to rely more heavily on the existing knowledge base in this field. I believe this would have provided a better context of these findings, which are not only on "past" vs. "future" but also on "working memory" vs. "spatial attention".

Thank you for considering our findings novel and important, while at the same time reminding us of the parallels to prior tasks studying spatial attention in perception and working memory. We fully agree that our task likely engages both attention to the (past) memory item as well as spatial attention to the upcoming (future) test stimulus. At the same time, there is a critical difference in spatial attention for the future in our task compared with ample prior tasks engaging spatial cueing of attention for perception. In our task, the cue never directly cues the future location. Rather, it exclusively cues the relevant memory item. It is the memory item that is associated with the relevant future location, according to the future rule. This integration of the rule-based future location into the memory representation is distinct from classical spatial-attention tasks in which attention is cued directly to a specific location via, for example, a spatial cue such as an arrow.

Thus, if we wish to think about our task as engaging cueing of spatial attention for perception, we have to at least also invoke the process of cueing the relevant location via the appropriate memory item. We feel it is more parsimonious to think of this as attending to both the past and future location of a dynamic visual object in working memory.

If we return to our opening example, when we see a bird disappear behind a building, we can keep in working memory where we last saw it, while anticipating where it will re-appear to guide our external spatial attention. Here too, spatial attention is fully dependent on working-memory content (the bird itself) – mirroring the dynamic semng in our study. Thus, we believe our findings contribute a fresh perspective, while of course also extending established fields. We now contextualize our finding within the literature and clarify our unique contribution in our revised manuscript:

Page 5 (Discussion): “Building on the above, at face value, our task may appear like a study that simply combines two established tasks: tasks using retro-cues to study attention in working memory (e.g.,2,31-33) and tasks using pre-cues to study orienting of spatial attention to an upcoming external stimulus (e.g., 31,32,34–36). A critical difference with common pre-cue studies, however, is that the cue in our task never directly informed the relevant future location. Rather, as also stressed above, the future location was a feature of the cued memory item (according to the future rule), and not of the cue itself. Note how this type of scenario may not be uncommon in everyday life, such as in our opening example of a bird flying behind a building. Here too, the future relevant location is determined by the bird – i.e. the memory content – itself.”

Reviewer 2, Recommendations:

It would be helpful to set up predictions based on existing working memory models. Otherwise, the claim that the joint coding of past/future is "not trivial" is simply asserted, rather than contradicting an existing model or prior empirical results. If the non-trivial aspect is simply the ability to demonstrate the joint coding empirical through a good experimental design, make it clear that this is the contribution. For example, it may be that prevailing models predict exactly this finding, but nobody has been able to demonstrate it cleanly, as the authors do here. So the non-triviality is not that the result contradicts working memory models, but rather relates to the methodological difficulty of revealing such an effect.

Thank you for your recommendation. First, please see our point-by-point responses to the individual comments above, where we also state relevant changes that we have made to our article, and where we clarify what we meant with “non trivial”. As we currently also state in our introduction, our work took as a starting point the framework that working memory is inherently about the past while being for the future (cf. van Ede & Nobre, Annual Review of Psychology, 2023). By virtue of our unique task design, we were able to empirically demonstrate that visual contents in working memory are selected via both their past and their future-relevant locations – with past and future memory attributes being engaged together in time. With “not trivial” we merely intend to make clear that there are viable alternatives than the findings we observed. For example, past could have been replaced by the future, or it could have been that item selection (through its past location) was required before its future-relevant location could be considered (i.e. past-before-future, rather than joint selection as we reported). We outline these alternatives in the second paragraph of our Discussion:

Page 5 (Discussion): “Our finding of joint utilisation of past and future memory attributes emerged from at least two alternative scenarios of how the brain may deal with dynamic everyday working memory demands in which memory content is encoded at one location but needed at another.

First, [….]”

Our work was not motivated from a particular theoretical debate and did not aim to challenge ongoing debates in the working-memory literature, such as: slot vs. resource, active vs. silent coding, decay vs. interference, and so on. To our knowledge, none of these debates makes specific claims about the retention and selection of past and future visual memory attributes – despite this being an important question for understanding working memory in dynamics everyday semngs, as we hoped to make clear by our opening example.

Reviewer 3, Recommendations:

I recommend that the present findings be more clearly interpreted in the context of previous findings on working memory and attention. The task design includes two components - the first (post-cue) is a classic working memory task and the second (the pre-cue) is a classic spatial attention design. Both components were thoroughly studied in the past and this previous knowledge should be better integrated into the present conclusions. I specifically feel uncomfortable with the interpretation of past vs. future. I find this framework to be misleading because it reads like this paper is on a topic that is completely new and never studied before, when in fact this is a study on the interaction between working memory and spatial attention. I recommend the authors minimize this past-future framing or be more explicit in explaining how this new framework relates to the more common terminology in the field and make sure that the findings are not presented in a vacuum, as another contribution to the vibrant field that they are part of.

Thank you for these recommendations. Please also see our point-by-point responses to the individual comments above. Here, we explained our logic behind using the terminology of past vs. future (in addition, see also our response to point 2 or reviewer 2). Here, we also stated relevant changes that we have made to our manuscript to explain how our findings complement – but are also distinct from – prior tasks that used pre-cues to direct spatial attention to an upcoming stimulus. As we explained above, in our task, the cue itself never contained information about the upcoming test location. Rather, the upcoming test location was a property of the memory item (given the future rule). Hence, we referred to this as a “future attribute” of the cued memory item, rather than as the “cued location” for external spatial attention. Still, we agree the future bias likely (also) reflects spatial allocation to the upcoming test array, and we explicitly acknowledge this in our discussion. For example:

Page 5 (Discussion): “This signal may reflect either of two situations: the selection of a future-copy of the cued memory content or anticipatory attention to its the anticipated location of its associated test-stimulus. Either way, by the nature of our experimental design, this future signal should be considered a content-specific memory attribute for two reasons. First, the two memory contents were always associated with opposite testing locations, hence the observed bias to the relevant future location must be attributed specifically to the cued memory content. Second, we cued which memory item would become tested based on its colour, but the to-be-tested location was dependent on the item’s encoding location, regardless of its colour. Hence, consideration of the item’s future-relevant location must have been mediated by selecting the memory item itself, as it could not have proceeded via cue colour directly.”

Page 6 (Discussion): “Building on the above, at face value, our task may appear like a study that simply combines two established tasks: tasks using retro-cues to study attention in working memory (e.g.,2,31-33) and tasks using pre-cues to study orienting of spatial attention to an upcoming external stimulus (e.g., 31,32,34–36). A critical difference with common pre-cue studies, however, is that the cue in our task never directly informed the relevant future location. Rather, as also stressed above, the future location was a feature of the cued memory item (according to the future rule), and not of the cue itself. Note how this type of scenario may not be uncommon in everyday life, such as in our opening example of a bird flying behind a building. Here too, the future relevant location is determined by the bird – i.e. the memory content – itself.”

Author response:

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

Reviewer #1:

We are grateful for the overall positive feedback from the reviewer.

We agree with the reviewer that our data showing cellular co-localization between PRC1 and BIN1 requires further investigation in future studies, however, we are confident that in the current form, our manuscript already presents multiple evidences for the role of BIN1 in mitotic processes. We would like to emphasize that PRC1 is not the sole BIN1 partner that connects it to mitotic processes, but it is only one out of more than a dozen that we identified in our study. Furthermore, the mitotic connection with BIN1 is not absolutely novel as BIN1 levels are mildly fluctuating during the cell cycle, similar to other proteins involved in the regulation of the cell cycle (Santos et al., 2015) and because DNM2 is also a well-accepted actor during mitosis (Thompson et al., 2002).

The less marked co-localization between BIN1 and PRC1 compared to the strong co-localization between BIN1 and DNM2 can be a consequence of their weaker affinity and their partial binding. Yet, this does not necessarily imply that stronger interactions have more biological significance. For example, weaker affinities can be compensated by local concentrations to achieve an even higher degree of cellular complexes than of strongly binding interactions that are separated within the cell. Furthermore, even the degree of complex formation cannot be used intuitively to estimate the biological significance of a complex because complexes can trigger very important biological processes even at very low abundances, e.g. by catalyzing enzymatic reactions. Deciding what is and what is not “biologically significant” among the identified interactions remains to be answered in the future, once we are able to overview complex biological processes in a holistic manner.

In the revised version, we implemented minor changes to further clarify the raised points.

Reviewer #2:

We thank the reviewer for the careful assessment and we are pleased to see the positive enthusiasm regarding our affinity interactomic strategy.

The reviewer points out that affinities were only measured with a single technique, which is relatively unproven. While it is true that our work uses two techniques building on the same holdup concept, we rather believe that this approach is well-proven. The original holdup method was described almost 20 years ago and since then, it has been used in more than 10 publications for quantitative interactomics. Over the years, at least five distinct generations of the assay were developed, all building on the expertise of the preceding one. In the past, we extensively proved that the resulting affinities show excellent agreement with affinities measured with other methods, such as fluorescence polarization, isothermal titration calorimetry, or surface plasmon resonance (for example in Vincentelli et al. Nat. Meth. 2015; Gogl et al. 2020 Structure; Gogl et al. 2022 Nat.Com.). However, it is true that the most recent variation of this method family, called native holdup, is a fairly new approach published just a bit more than a year ago and this is only the third work that utilizes this method. Yet, in our original work describing the method, we demonstrated good agreement with the results of previous holdup experiments, as well as with orthogonal affinity measurements (Zambo et al. 2022).

Importantly, the reviewer raises concerns regarding the number of replicates used in our study, as well as the reliability of our methodology. We are glad for such a comment as it allows us to explain our motives behind experimental design which is most often left out from scientific works to save space and keep focus on results. The reason why we use technical replicates instead of the typical biological replicates lies in the nature of the holdup assay. In a typical interactomic assay, such as immunoprecipitation, a lot of variables can perturb the outcome of the measurement, such as bait immobilization, or captured prey leakage during washing steps. The output of such an experiment is a list of statistically significant partners and to minimize these variabilities, biological replicates are used. In the case of a native holdup approach, a panel of an equal amount of resins, all saturated with different baits or controls, is mixed with an equal amount of cell extract, taken from a single tube, and after a brief incubation, the supernatant of this mixture is analyzed. The output of such an experiment is a list of relative concentrations of prey and to maximize its accuracy, we use technical replicates. Using an ideal analytical method, such as fluorescence, it is not necessary to use technical replicates to reach accurate results. For example, the general accuracy of a holdup experiment coupled with a robust analytical approach can be seen clearly in our fragmentomic holdup data shown in Figure 7C where mutant domains that do not have any impact on the interactome show extreme agreement in affinities. Unfortunately, mass spectrometry is less accurate as an analytical method, hence we use technical triplicates to compensate for this. Finally, in the case of BIN1, an independent nHU measurement was also performed using a less capable mass spectrometer. Not counting the 117 detected partners of BIN1 that were only detected in only one of these proteomic measurements, 29 partners were identified as common significant partners in both of these measurements showing nearly identical affinities with a mean standard deviation between measured pKapp values of 0.18, meaning that the obtained dissociation constants are within a <2.5-fold range with >95% probability. There were also 61 BIN1 partners that were detected in both proteomic measurements but were only identified as a significant interaction partner in one of these experiments. Yet many of them show binding in both assays, albeit were found to be not significant in one of these assays. For example, CDC20 shows 66% depletion in one assay (significant binding) while it shows 54% depletion in the other (not significant binding), or CKAP2 shows 58% depletion in one assay (significant binding) while it shows 41% depletion in the other (not significant binding). We hope that these examples show that statistical significance in nHU experiments rather signifies how certain we are in a particular affinity measurement and not the accuracy of the affinity measurement itself. While there are true discrepancies between some of the affinity measurements between these experiments, that would be possible to clarify with more experimental replicates, the raw data presented in our work clearly demonstrate the strength and robustness of a fully quantitative interactomic assay.

In the revised version, we clarified the number of replicates in the text, in the figure legends, and included some of this discussion in the method section.

The reviewer had some very useful comments regarding affinity differences between short fragments and full-length proteins. In his comment, he possibly made a typo as we find that fulllength proteins typically interact with higher affinities compared to short PxxP motif fragments in isolation and not weaker. The reviewer also comments that we explain this difference with cooperativity. In a previous preprint version, which the reviewer may have seen, this was indeed the case, but since we realized that we did not have sufficient evidence supporting this model, therefore we did not discuss this in detail in the last version submitted to eLife. To clarify this, we included more discussion about the observed differences in the affinities between fragments and full-length proteins, but since we have limited data to make solid conclusions, we do not go into details about underlying models.

Instead of cooperativity, the reviewer suggests that the observed differences may originate from additional residues that were not included in our peptides. Indeed, many similar experiments fail because of suboptimal peptide library design. Our peptide library was constructed as 15-mer, xxxxxxPxxPxxxxx motifs and we do not see a strong contribution of residues at the far end of these peptides. Specificity logo reconstructions are expected to identify all key residues that participate in SH3 domain binding, and based on this, all key residues of the identified motifs can be included in shorter 10-mer, xxxPxxPxxx motifs. Therefore, it is unlikely that residues outside our peptide regions will greatly contribute to the site-specific interactions of SH3 domains. It is however possible that other sites, that are sequentially far away from the studied PxxP motifs, are also capable of binding to SH3 through a different surface, but in light of the small size of an isolated SH3 domain, we believe it is very unlikely. It is also possible that BIN1 could also interact with other types of SH3 binding motifs that were not included in our peptide library. We think a more likely explanation is some sort of cooperativity. Cooperativity, or rather synergism between different sites can be easily explained in typical situations, such as in the case of a bimolecular interaction that is mediated by two independent sites. In such an event, once one site is bound, the second binding event will likely also occur because of the high effective local concentration of the binding sites. However, cooperativity can also form in atypical conditions and a molecular explanation for these events is rather elusive. As BIN1 contains a single SH3 domain, its binding to targets containing more binding sites can be challenging to interpret. If these sites are part of a greater Pro-rich region, such as in the case of DNM2, it is possible that the entire region adopts a fuzzy, malleable, yet PPII-like helical conformation. Once the SH3 domain is recruited to this helical region, it can freely trans-locate within this region via lateral diffusion and it will pause on optimal PxxP motifs. As an alternative to this sliding mechanism, a diffusion-limited cooperative binding can also occur. If the two motifs are not part of the same Pro-rich region, but are relatively close in space, such as in the case of ITCH or PRC1, once a BIN1 molecule dissociates from one site, it has a higher chance to rebind to the second site due to higher local concentrations. Such an event can more likely occur if a transient, but relatively stable encounter complex exists between the two molecules, from which complex formation can occur at both sites (A+B↔AB; AB↔ABsite1; AB*↔ABsite2). However, this large effective local concentration in this encounter complex is only temporary because diffusion rapidly diminishes it, although weak electrostatic interactions can increase the lifetime of such encounter complexes. In contrast, the large effective local concentration in conventional multivalent binding is time-independent and only determined by the geometry of the complex. Finally, it may also occur that our empirical bait concentration estimation for immobilized biotinylated proteins is less accurate than the concentration estimation of peptide baits because we approximate this value based on peptide baits. For this technical reason, which was discussed in detail in the original paper describing the nHU approach, we are carefully using apparent affinities for nHU experiments. Nevertheless, even without accurate bait concentrations, our nHU experiment provides precise relative affinities and, thus partner ranking. Either of the mechanisms underlying the interactions we study would be difficult to further explore experimentally, especially at the proteomic level.

Author response:

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

We greatly appreciate the comments from the editor and the reviewers, based on which we have made the revisions. We have responded to all the questions and summarized the revisions below. The changes are also highlighted in the manuscript.

Additionally, we’ve noticed a few typos in the manuscript presented on the eLife website, which were not there in our originally submitted file.

(1) In both the “Full text” presented on the eLife website and the pdf file generated after clicking “Download”: the last FC1000 in the second paragraph of the “Extensive induction curves fitting of TetR mutants” section should be FC1000WT .

(2) In the pdf file generated after clicking “Download”: the brackets are all incorrectly formatted in the captions of Figure 4 and Figure 3—figure supplement 6.

eLife assessment

The fundamental study presents a two-domain thermodynamic model for TetR which accurately predicts in vivo phenotype changes brought about as a result of various mutations. The evidence provided is solid and features the first innovative observations with a computational model that captures the structural behavior, much more than the current single-domain models.

We appreciate the supportive comments by the editor and reviewers.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors’ earlier deep mutational scanning work observed that allosteric mutations in TetR (the tetracycline repressor) and its homologous transcriptional factors are distributed across the structure instead of along the presumed allosteric pathways as commonly expected. Especially, in addition, the loss of the allosteric communications promoted by those mutations, was rescued by additional distributed mutations. Now the authors develop a two-domain thermodynamic model for TetR that explains these compelling data. The model is consistent with the in vivo phenotypes of the mutants with changes in parameters, which permits quantification. Taken together their work connects intra- and inter-domain allosteric regulation that correlate with structural features. This leads the authors to suggest broader applicability to other multidomain allosteric proteins. Here the authors follow their first innovative observations with a computational model that captures the structural behavior, aiming to make it broadly applicable to multidomain proteins. Altogether, an innovative and potentially useful contribution.

We thank the reviewer for the supportive comments.

Weaknesses:

None that I see, except that I hope that in the future, if possible, the authors would follow with additional proteins to further substantiate the model and show its broad applicability. I realize however the extensive work that this would entail.

We thank the reviewer for the supportive comments and the suggestion to extend the model to other proteins, which we indeed plan to pursue in future studies.

Reviewer #2 (Public Review):

Summary:

This combined experimental-theoretical paper introduces a novel two-domain statistical thermodynamic model (primarily Equation 1) to study allostery in generic systems but focusing here on the tetracycline repressor (TetR) family of transcription factors. This model, building on a function-centric approach, accurately captures induction data, maps mutants with precision, and reveals insights into epistasis between mutations.

Strengths:

The study contributes innovative modeling, successful data fitting, and valuable insights into the interconnectivity of allosteric networks, establishing a flexible and detailed framework for investigating TetR allostery. The manuscript is generally well-structured and communicates key findings effectively.

We thank the reviewer for the supportive comments.

Weaknesses:

The only minor weakness I found was that I still don’t have a better sense into (a) intuition and (b) mathematical derivation of Equation 1, which is so central to the work. I would recommend that the authors provide this early on in the main text.

We thank the reviewer for the suggestion. The full mathematical derivation of Equation 1 is given in the first section of the supplementary file. Given the length of the derivation, we think it’s better to keep it in the supplementary file rather than the main text. In the main text, the first subsection (overview of the two-domain thermodynamic model of allostery) of the Results section and the paragraph right before Equation 1 are meant for providing intuitive understandings of the two-domain model and the derivation of Equation 1, respectively.

We would also like to point the reviewer to Figure 2-figure supplement 2 and Equations (12) to (18) in the supplementary file for an alternative derivation. They show that the equilibria among all molecular species containing the operator are dictated by the binding free energies, the ligand concentration, and the allosteric parameters. The probability of an unbound operator (proportional to the probability that the promoter is bound by a RNA polymerase, or the gene expression level) can thus be calculated using Equation (12), which then leads to main text Equation 1 following the derivation given there.

Additionally, we’ve added a paragraph to the main text (line 248-260) to aid an intuitive understanding of Equation 1.

“The distinctive roles of the three biophysical parameter on the induction curve as stipulated in Equation 1 could be understood in an intuitive manner as well. First, the value of εD controls the intrinsic strength of binding of TetR to the operator, or the intrinsic difficulty for ligand to induce their separation. Therefore, it controls how tightly the downstream gene is regulated by TetR without ligands (reflected in leakiness) and affects the performance limit of ligands (reflected in saturation). Second, the value of εL controls how favorable ligand binding is in free energy. When εL increases, the binding of ligand at low concentrations become unfavorable, where the ligands cannot effectively bind to TetR to induce its separation from the operator. Therefore, the fold-change as a function of ligand concentration only starts to noticeably increase at higher ligand concentrations, resulting in larger EC50. Third, as discussed above, γ controls the level of anti-cooperativity between the ligand and operator binding of TetR, which is the basis of its allosteric regulation. In other words, γ controls how strongly ligand binding is incompatible with operator binding for TetR, hence it controls the performance limit of ligand (reflected in saturation).”

We hope that the reviewer will find this explanation helpful.

Reviewer #3 (Public Review):

Summary:

Allosteric regulations are complicated in multi-domain proteins and many large-scale mutational data cannot be explained by current theoretical models, especially for those that are neither in the functional/allosteric sites nor on the allosteric pathways. This work provides a statistical thermodynamic model for a two-domain protein, in which one domain contains an effector binding site and the other domain contains a functional site. The authors build the model to explain the mutational experimental data of TetR, a transcriptional repress protein that contains a ligand and a DNA-binding domain. They incorporate three basic parameters, the energy change of the ligand and DNA binding domains before and after binding, and the coupling between the two domains to explain the free energy landscape of TetR’s conformational and binding states. They go further to quantitatively explain the in vivo expression level of the TetR-regulated gene by fitting into the induction curves of TetR mutants. The effects of most of the mutants studied could be well explained by the model. This approach can be extended to understand the allosteric regulation of other two-domain proteins, especially to explain the effects of widespread mutants not on the allosteric pathways. Strengths: The effects of mutations that are neither in the functional or allosteric sites nor in the allosteric pathways are difficult to explain and quantify. This work develops a statistical thermodynamic model to explain these complicated effects. For simple two-domain proteins, the model is quite clean and theoretically solid. For the real TetR protein that forms a dimeric structure containing two chains with each of them composed of two domains, the model can explain many of the experimental observations. The model separates intra and inter-domain influences that provide a novel angle to analyse allosteric effects in multi-domain proteins.

We thank the reviewer for the supportive comments.

Weaknesses:

As mentioned above, the TetR protein is not a simple two-main protein, but forms a dimeric structure in which the DNA binding domain in each chain forms contacts with the ligand-binding domain in the other chain. In addition, the two ligand-binding domains have strong interactions. Without considering these interactions, especially those mutants that are on these interfaces, the model may be oversimplified for TetR.

We thank the reviewer for this valid concern and acknowledge that TetR is a homodimer. However, we’ve deliberately chosen to simplify this complexity in our model for the following reasons.

(1) In this work, we aim to build a minimalist model for two-domain allostery withonly the most essential parameters for capturing experimental data. The simplicity of the model helps promote its mechanistic clarity and potential transferability to other allosteric systems.

(2) Fewer parameters are needed in a simpler model. Our two-domain modelcurrently uses only three biophysical parameters, which are all demonstrated to have distinct influences on the induction curve (see the main text section “System-level ramifications of the two-domain model”). This enables the inference of parameters with high precision for the mutants, and the quantification of the most essential mechanistic effects of their mutations, provided that the model is shown to accurately recapitulate the comprehensive dataset. Thus, we found it was unnecessary to add another parameter for explicitly describing inter-chain coupling, which would likely incur uncertainty in the inference of parameters due to the redundancy of their effects on induction data, and prevent the model from making faithful predictions.

(3) From a more biological point of view, TetR is an obligate dimer, meaning thatthe two chains must synchronize for function, supporting the two-domain simplification of TetR for binding concerns.

Additionally, as shown in the subsection “Inclusion of single-ligand-bound state of repressor” of section 1 of the supplementary file, incorporating the dimeric nature of TetR in our model by allowing partial ligand binding does not change the functional form of main text equation 1 in any practical sense. Therefore, considering all the factors stated above, we think that increasing the complexity of the two-domain model will only be necessary if additional data emerge to suggest the limitation of our model.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

This is an excellent work. I have only one suggestion for the authors. Interestingly, the authors also note that the epistatic interactions that they obtain are consistent with the structural features of the protein, which is not surprising. Within this framework, have the authors considered rescue mutations? Please see for example PMID: 18195360 and PMID: 15683227. If I understand right, this might further extend the applicability of their model. If so, the authors may want to add a comment to that effect.

We thank the reviewer for the supportive comments and for pointing us to the useful references. We have added some comments to the main text regarding this point in line 332-336: “The diverse mechanistic origins of the rescuing mutations revealed here provide a rational basis for the broad distributions of such mutations. Integrating such thermodynamic analysis with structural and dynamic assessment of allosteric proteins for efficient and quantitative rescuing mutation design could present an interesting avenue for future research, particularly in the context of biomedical applications (PMID: 18195360, PMID: 15683227).”

Reviewer #3 (Recommendations For The Authors):

The authors should try to build a more realistic dimeric model for TetR to see if it could better explain experimental data. If it were too complicated for a revision, more discussions on the weakness of the current model should be given.

We thank the reviewer for this valid concern and for the suggestion. The reasons for refraining from increasing the complexity of the model are fully discussed in our response to the reviewer’s public review given above. Primarily, we think that the value of a simple physical model is two-fold (e.g., the paradigm Ising model in statistical physics and the classic MWC model), first, its mechanistic clarity and potential transferability makes it a useful conceptual framework for understanding complex systems and establishing universal rules by comparing seemingly unrelated phenomena; second, it provides useful insights and design principles of specific systems if it can quantitatively capture the corresponding experimental data. Thus, given the current experimental data set, we believe it is justified to keep the two-domain model in its current form, while additional experimental data could necessitate a more complex model for TetR allostery in the future. Relevant discussions are added to the main text (line 443-446) and section 8 of the supplementary file.

“It’s noted that the homodimeric nature of TetR is ignored in the current two-domain model to minimize the number of parameters, and additional experimental data could necessitate a more complex model for TetR allostery in the future (see supplementary file section 8 for more discussions).”

Minor issues:

(1) There is an error in Figure 3A, the 13th and 14th subgraphs are the same and should be corrected.

We thank the reviewer for capturing this error, which has been corrected in the revised manuscript.

(2) The criteria for the selection of mutants for analysis should be clearly given. Apart from deleting mutants that are in direct contact with the ligand of DNA, how many mutants are left, and how far are they are from the two sites? In line 257, what are the criteria for selecting these 15 mutants? Similarly, in line 332, what are the criteria for selecting these 8 mutants?

We thank the reviewer for this comment. The data selection criteria are now added in section 7 of the supplementary file. The distances to the DNA operator and ligand of the 21 residues under mutational study are now added in Table 1 (Figure 3-figure supplement 9). The added materials are referenced in the main text where relevant.

“7. Mutation selection for two-domain model analysis

In this work, there are 24 mutants studied in total including the WT, and they contain mutations at 21 WT residues. We did not perform model parameter inference for the mutant G102D because of its flat induction curve (see the second subsection of section 2 and main text Figure 2—figure Supplement 3). Therefore, there are 23 mutants analyzed in main text Figure 5.

Measuring the induction curve of a mutant involves a significant amount of experimental effort, which therefore is hard to be extended to a large number of mutants. Nonetheless, we aim to compose a set of comprehensive induction data here for validating our two-domain model for TetR allostery. To this end, we picked 15 individual mutants in the first round of induction curve measurements, which contains mutations spanning different regions in the sequence and structure of TetR (main text Figure 3—figure Supplement 1). Such broad distribution of mutations across LBD, DBD and the domain interface could potentially lead to diverse induction curve shapes and mutant phenotypes for validating the two-domain model. Indeed, as discussed in the main text section "Extensive induction curves fitting of TetR mutants", the diverse effects on induction curve from mutations perturbing different allosteric parameters predicted by the model, are successfully observed in these 15 experimental induction curves. Additionally, 5 of the 15 mutants contain a dead-rescue mutation pair, which helps us validate the model prediction that a dead mutation could be rescued by rescuing mutations that perturb the allosteric parameters in various ways.

Eight mutation combinations were chosen for the second round of induction curve measurement for studying epistasis, where we paired up C203V and Y132A with mutations from different regions of the TetR structure. Such choice is largely based on two considerations. 1. As both C203V and Y132A greatly enhance the allosteric response of TetR, we want to probe why they cannot rescue a range of dead mutations as observed previously (PMID: 32999067). 2. C203V and Y132A are the only two mutants that show enhanced allosteric response in the first round of analysis. Combining detrimental mutations of allostery in a combined mutant could potentially lead to near flat induction curve, which is less useful for inference (see the second subsection of section 2).”

Since the number of hotspots identified by DMS is not very large, why not analyze them all?

We thank the reviewer for this comment. There are 41 hotspot residues in TetR (PMID: 36226916), which have 41*19=779 possible single mutations. It’s unfeasible to perform induction curve measurements for all of these 779 mutants in our current experiment. However, we agree that it would be helpful if we can obtain such a dataset in an efficient way.

In line 257, there are 15 mutants mentioned, while in Figure 5, there are 23 mutants mentioned, in Figure 3-figure supplement 1, there are 21 mutants mentioned, and in line 226 of the supplementary file, there are 24 mutants mentioned, which is very confusing. Therefore, the data selection criteria used in this article should be given.

We thank the reviewer for this comment. The data selection criteria are now given in section 7 of the supplementary file, which should clarify this confusion.

(3) In Figure 4 of the Exploring epistasis between mutations section, the 6 weights of the additive models corresponding to each mutation combination are different. On one hand, it seems that there are no universal laws in these experimental data. On the other hand, unique parameters of a single mutation combination were not validated in other mutation combinations, which somewhat weakened the conclusions about the potential physical significance of these additive weights.

We thank the reviewer for this comment. We admit that a quantitative universal law for tuning the 6 weights of the additive model does not manifest in our data, which indicates the mutation-specific nature of epistatic interactions in TetR as hinted in the different rescuing mutation distributions of different dead mutations (PMCID: PMC7568325). However, clear common trends in the weight tuning of combined mutants that contain common mutations do emerge, which comply with the structural features of the protein and provide explanations as to why C203V and Y132A don’t rescue a range of dead mutations (main text section “Exploring epistasis between mutations”). Additionally, the lack of a quantitative universal rule for tuning the 6 weights in our simple model doesn’t exclude the possibility of the existence of universal law for epistasis in TetR in another functional form, a point that could be explored in the future with more extensive joint experimental and computational investigations.

In Eq. (27) of the supplementary file, the prior distribution of inter-domain coupling γ is given as a Gaussian distribution centered at 5 kBT. Since the absolute value of γ is important, can the authors explain why the prior distribution of γ is set to this value and what happens if other values are used?

We thank the reviewer for the question. As explained in the corresponding discussions of Eq. (27) in the supplementary file, the prior of γ is chosen to serve as a soft constraint on its possible values based on the consideration that 1. inter-domain energetics for a TetR-like protein should be on the order of a few kBT; and 2. the prior distribution should reflect the experimental observation in the literature that γ has a small probability of adopting negative values upon mutations. Given our thorough validation of the statistical model and computational algorithm (see section 3 of the supplementary file), and the high precision in the parameter fitting results using experimental data (Figure 3 and Figure 4-figure supplement 2), we conclude that 1. the physical range of parameters encoded in their chosen prior distributions agrees well with the value reflected in the experimental data; 2. the inference results are predominantly informed by the data. Thus, changing the mean of the prior distribution of γ should not affect the inference results significantly given that it remains in the physical range.

This point is explicitly shown in the added Table 2 (Figure 3-figure supplement 10), where we compare the current Bayesian inference results with those obtained after increasing the standard deviation of the Gaussian prior of γ from 2.5 to 5 kBT. As shown in the table, most inference results stay virtually unchanged at the use of this less informative prior, which confirms that they are predominantly informed by the data. The only exceptions are the slight increase of the inferred γ values for C203V, C203V-Y132A and C203V-G102D-L146A, reflecting the intrinsic difficulty of precise inference of large γ values with our model, as is already discussed in the second subsection of section 3 of the supplementary file. However, such observations comply with the common trend of epistatic interactions involving C203V presented in the main text and don’t compromise the ability of our model to accurately capture the induction curves of mutants. Relevant discussions are now added to the second subsection of section 3 of the supplementary file (line 368-385).

“In our experimental dataset, such inference difficulty is only observed in the case of C203V, Y132A-C203V and C203V-G102D-L146A due to their large γ and γ + εL values (see main text Figure 3, Figure 3—figure Supplement 10 and Figure 4). As shown in main text Figure 3—figure Supplement 10, the inference results for the other 20 mutants stay highly precise and virtually unchanged after increasing the standard deviation of the Gaussian prior of γ (gstdγ ) from 2.5 to 5 kBT. This demonstrates that the inference results for these mutants are strongly informed by the induction data and there is no difficulty in the precise inference of the parameter values. On the other hand, the inferred γ values (especially the upper bound of the 95% credible region) for C203V, Y132A-C203V and C203V-G102D-L146A increased with gstdγ . This is because the induction curves in these cases are not sensitive to the value of γ given that it’s large enough as discussed above. Hence, when unphysically large γ values are permitted by the prior distribution, they could enter the posterior distribution as well. Such difficulty in the precise inference of γ values for these three mutants however, doesn’t compromise the ability of our model in accurately capturing the comprehensive set of induction data (see part iv below). Additionally, the increase of the inferred γ value of C203V at the use of larger gstdγ complies with the results presented in main text Figure 4, which show that the effect of C203V on γ tends to be compromised when combined with mutations closer to the domain interface."