Author response:

The issue of a control without blue light illumination was raised. Clearly without the light we will not obtain any signal in the fluorescence microscopy experiments, which would not be very informative. Instead, we changed the level of blue light illumination in the fluorescence microscopy experiments (figure 4A) and the response of the bacteria scales with dosage. It is very hard to find an alternative explanation, beyond that the blue light is stressing the bacteria and modulating their membrane potentials.

One of the referees refuses to see wavefronts in our microscopy data. We struggle to understand whether it is an issue with definitions (Waigh has published a tutorial on the subject in Chapter 5 of his book ‘The physics of bacteria: from cells to biofilms’, T.A.Waigh, CUP, 2024 – figure 5.1 shows a sketch) or something subtler on diffusion in excitable systems. We stand by our claim that we observe wavefronts, similar to those observed by Prindle et al¹ and Blee et al² for B. subtilis biofilms.

The referee is questioning our use of ThT to probe the membrane potential. We believe the Pilizota and Strahl groups are treating the E. coli as unexcitable cells, leading to their problems. Instead, we believe E. coli cells are excitable (containing the voltage-gated ion channel Kch) and we now clearly state this in the manuscript. Furthermore, we include a section here discussing some of the issues with ThT.

Use of ThT as a voltage sensor in cells

ThT is now used reasonably widely in the microbiology community as a voltage sensor in both bacterial [Prindle et al]1 and fungal cells [Pena et al]12. ThT is a small cationic fluorophore that loads into the cells in proportion to their membrane potential, thus allowing the membrane potential to be measured from fluorescence microscopy measurements.

Previously ThT was widely used to quantify the growth of amyloids in molecular biology experiments (standardized protocols exist and dedicated software has been created)13 and there is a long history of its use14. ThT fluorescence is bright, stable and slow to photobleach.

Author response image 1 shows a schematic diagram of the ThT loading in E. coli in our experiments in response to illumination with blue light. Similar results were previously presented by Mancini et al15, but regimes 2 and 3 were mistakenly labelled as artefacts.

Author response image 1.

Schematic diagram of ThT loading during an experiment with E. coli cells under blue light illumination i.e. ThT fluorescence as a function of time. Three empirical regimes for the fluorescence are shown (1, 2 and 3).

The classic study of Prindle et al on bacterial biofilm electrophysiology established the use of ThT in B. subtilis biofilms by showing similar results occurred with DiSc3 which is widely used as a Nernstian voltage sensor in cellular biology1 e.g. with mitochondrial membrane potentials in eukaryotic organisms where there is a large literature. We repeated such a comparative calibration of ThT with DiSc3 in a previous publication with both B. subtilis and P. aeruginosa cells2. ThT thus functioned well in our previous publications with Gram positive and Gram negative cells.

However, to our knowledge, there are now two groups questioning the use of ThT and DiSc3 as voltage sensors with E. coli cells15-16. The first by the Pilizota group claims ThT only works as a voltage sensor in regime 1 of Author response image 1 using a method based on the rate of rotation of flagellar motors. Another slightly contradictory study by the Strahl group claims DiSc316 only acts as a voltage sensor with the addition of an ionophore for potassium which allows free movement of potassium through the E. coli membranes.

Our resolution to this contradiction is that ThT does indeed work reasonably well with E. coli. The Pilizota group’s model for rotating flagellar motors assumes the membrane voltage is not varying due to excitability of the membrane voltage (otherwise a non-linear Hodgkin Huxley type model would be needed to quantify their results) i.e. E. coli cells are unexcitable. We show clearly in our study that ThT loading in E. coli is a function of irradiation with blue light and is a stress response of the excitable cells. This is in contradiction to the Pilizota group’s model. The Pilizota group’s model also requires the awkward fiction of why cells decide to unload and then reload ThT in regimes 2 and 3 of Author response image 1 due to variable membrane partitioning of the ThT. Our simple explanation is that it is just due to the membrane voltage changing and no membrane permeability switch needs to be invoked. The Strahl group’s16 results with DiSc3 are also explained by a neglect of the excitable nature of E. coli cells that are reacting to blue light irradiation. Adding ionophores to the E. coli membranes makes the cells unexcitable, reduces their response to blue light and thus leads to simple loading of DiSc3 (the physiological control of K+ in the cells by voltage-gated ion channels has been short circuited by the addition of the ionophore).

Further evidence of our model that ThT functions as a voltage sensor with E. coli include:

1) The 3 regimes in Author response image 1 from ThT correlate well with measurements of extracellular potassium ion concentration using TMRM i.e. all 3 regimes in Author response image 1 are visible with this separate dye (figure 1d).

2) We are able to switch regime 3 in Author response image 1, off and then on again by using knock downs of the potassium ion channel Kch in the membranes of the E. coli and then reinserting the gene back into the knock downs. This cannot be explained by the Pilizota model.

We conclude that ThT works reasonably well as a sensor of membrane voltage in E. coli and the previous contradictory studies15-16 are because they neglect the excitable nature of the membrane voltage of E. coli cells in response to the light used to make the ThT fluoresce.

Three further criticisms of the Mancini et al method15 for calibrating membrane voltages include:

1) E. coli cells have clutches that are not included in their models. Otherwise the rotation of the flagella would be entirely enslaved to the membrane voltage allowing the bacteria no freedom to modulate their speed of motility.

2) Ripping off the flagella may perturb the integrity of the cell membrane and lead to different loading of the ThT in the E. coli cells.

3) Most seriously, the method ignores the activity of many other ion channels (beyond H+) on the membrane voltage that are known to exist with E. coli cells e.g. Kch for K+ ions. The Pilizota groups uses a simple Nernstian battery model developed for mitochondria in the 1960s. It is not adequate to explain our results.

An additional criticism of the Winkel et al study17 from the Strahl group is that it indiscriminately switches between discussion of mitochondria and bacteria e.g. on page 8 ‘As a consequence the membrane potential is dominated by H+’. Mitochondria are slightly alkaline intracellular organelles with external ion concentrations in the cytoplasm that are carefully controlled by the eukaryotic cells. E. coli are not i.e. they have neutral internal pHs, with widely varying extracellular ionic concentrations and have reinforced outer membranes to resist osmotic shocks (in contrast mitochondria can easily swell in response to moderate changes in osmotic pressure).

A quick calculation of the equilibrium membrane voltage of E. coli can be easily done using the Nernst equation dependent on the extracellular ion concentrations defined by the growth media (the intracellular ion concentrations in E. coli are 0.2 M K+ and 10-7 M H+ i.e. there is a factor of a million fewer H+ ions). Thus in contradiction to the claims of the groups of Pilizota15 and Strahl17, H+ is a minority determinant to the membrane voltage of E. coli. The main determinant is K+. For a textbook version of this point the authors can refer to Chapter 4 of D. White, et al’s ‘The physiology and biochemistry of prokaryotes’, OUP, 2012, 4th edition.

Even in mitochondria the assumption that H+ dominates the membrane potential and the cells are unexcitable can be questioned e.g. people have observed pulsatile depolarization phenomena with mitochondria18-19. A large number of K+ channels are now known to occur in mitochondrial membranes (not to mention Ca2+ channels; mitochondria have extensive stores of Ca2+) and they are implicated in mitochondrial membrane potentials. In this respect the seminal Nobel prize winning research of Peter Mitchell (1961) on mitochondria needs to be amended20. Furthermore, the mitochondrial work is clearly inapplicable to bacteria (the proton motive force, PMF, will instead subtly depend on non-linear Hodgkin-Huxley equations for the excitable membrane potential, similar to those presented in the current article). A much more sophisticated framework has been developed to describe electrophysiology by the mathematical biology community to describe the activity of electrically excitable cells (e.g. with neurons, sensory cells and cardiac cells), beyond Mitchell’s use of the simple stationary equilibrium thermodynamics to define the Proton Motive Force via the electrochemical potential of a proton (the use of the word ‘force’ is unfortunate, since it is a potential). The tools developed in the field of mathematical electrophysiology8 should be more extensively applied to bacteria, fungi, mitochondria and chloroplasts if real progress is to be made.

Related to the previous point, we now cite articles from the Pilizota and Strahl groups in the main text (one from each group). Unfortunately, the space constraints of eLife mean we cannot make a more detailed discussion in the main article.

In terms of modelling the ion channels, the Hodgkin-Huxley type model proposes that the Kch ion channel can be modelled as a typical voltage-gated potassium ion channel i.e. with a 𝑛⁴ term in its conductivity. The literature agrees that Kch is a voltage-gated potassium ion channel based on its primary sequence³. The protein has the typical 6 transmembrane helix motif for a voltage-gated ion channel. The agent-based model assumes little about the structure of ion channels in E. coli, other than they release potassium in response to a threshold potassium concentration in their environment. The agent based model is thus robust to the exact molecular details chosen and predicts the anomalous transport of the potassium wavefronts reasonably well (the modelling was extended in a recent Physical Review E article(⁴). Such a description of reaction-anomalous diffusion phenomena has not to our knowledge been previously achieved in the literature⁵ and in general could be used to describe other signaling molecules.

1. Prindle, A.; Liu, J.; Asally, M.; Ly, S.; Garcia-Ojalvo, J.; Sudel, G. M., Ion channels enable electrical communication in bacterial communities. Nature 2015, 527, 59.

2. Blee, J. A.; Roberts, I. S.; Waigh, T. A., Membrane potentials, oxidative stress and the dispersal response of bacterial biofilms to 405 nm light. Physical Biology 2020, 17, 036001.

3. Milkman, R., An E. col_i homologue of eukaryotic potassium channel proteins. _PNAS 1994, 91, 3510-3514.

4. Martorelli, V.; Akabuogu, E. U.; Krasovec, R.; Roberts, I. S.; Waigh, T. A., Electrical signaling in three-dimensional bacterial biofilms using an agent-based fire-diffuse-fire model. Physical Review E 2024, 109, 054402.

5. Waigh, T. A.; Korabel, N., Heterogeneous anomalous transport in cellular and molecular biology. Reports on Progress in Physics 2023, 86, 126601.

6. Hodgkin, A. L.; Huxley, A. F., A quantitative description of membrane current and its application to conduction and excitation in nerve. Journal of Physiology 1952, 117, 500.

7. Dawson, S. P.; Keizer, J.; Pearson, J. E., Fire-diffuse-fire model of dynamics of intracellular calcium waves. PNAS 1999, 96, 606.

8. Keener, J.; Sneyd, J., Mathematical Physiology. Springer: 2009.

9. Coombes, S., The effect of ion pumps on the speed of travelling waves in the fire-diffuse-fire model of Ca2+ release. Bulletin of Mathematical Biology 2001, 63, 1.

10. Blee, J. A.; Roberts, I. S.; Waigh, T. A., Spatial propagation of electrical signals in circular biofilms. Physical Review E 2019, 100, 052401.

11. Gorochowski, T. E.; Matyjaszkiewicz, A.; Todd, T.; Oak, N.; Kowalska, K., BSim: an agent-based tool for modelling bacterial populations in systems and synthetic biology. PloS One 2012, 7, 1.

12. Pena, A.; Sanchez, N. S.; Padilla-Garfias, F.; Ramiro-Cortes, Y.; Araiza-Villaneuva, M.; Calahorra, M., The use of thioflavin T for the estimation and measurement of the plasma membrane electric potential difference in different yeast strains. Journal of Fungi 2023, 9 (9), 948.

13. Xue, C.; Lin, T. Y.; Chang, D.; Guo, Z., Thioflavin T as an amyloid dye: fibril quantification, optimal concentration and effect on aggregation. Royal Society Open Science 2017, 4, 160696.

14. Meisl, G.; Kirkegaard, J. B.; Arosio, P.; Michaels, T. C. T.; Vendruscolo, M.; Dobson, C. M.; Linse, S.; Knowles, T. P. J., Molecular mechanisms of protein aggregation from global fitting of kinetic models. Nature Protocols 2016, 11 (2), 252-272.

15. Mancini, L.; Tian, T.; Guillaume, T.; Pu, Y.; Li, Y.; Lo, C. J.; Bai, F.; Pilizota, T., A general workflow for characterization of Nernstian dyes and their effects on bacterial physiology. Biophysical Journal 2020, 118 (1), 4-14.

16. Buttress, J. A.; Halte, M.; Winkel, J. D. t.; Erhardt, M.; Popp, P. F.; Strahl, H., A guide for membrane potential measurements in Gram-negative bacteria using voltage-sensitive dyes. Microbiology 2022, 168, 001227.

17. Derk te Winkel, J.; Gray, D. A.; Seistrup, K. H.; Hamoen, L. W.; Strahl, H., Analysis of antimicrobial-triggered membrane depolarization using voltage sensitive dyes. Frontiers in Cell and Developmental Biology 2016, 4, 29.

18. Schawarzlander, M.; Logan, D. C.; Johnston, I. G.; Jones, N. S.; Meyer, A. J.; Fricker, M. D.; Sweetlove, L. J., Pulsing of membrane potential in individual mitochondria. The Plant Cell 2012, 24, 1188-1201.

19. Huser, J.; Blatter, L. A., Fluctuations in mitochondrial membrane potential caused by repetitive gating of the permeability transition pore. Biochemistry Journal 1999, 343, 311-317.

20. Mitchell, P., Coupling of phosphorylation to electron and hydrogen transfer by a chemi-osmotic type of mechanism. Nature 1961, 191 (4784), 144-148.

21. Baba, T.; Ara, M.; Hasegawa, Y.; Takai, Y.; Okumura, Y.; Baba, M.; Datsenko, K. A.; Tomita, M.; Wanner, B. L.; Mori, H., Construction of Escherichia Coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Molecular Systems Biology 2006, 2, 1.

22. Schinedlin, J.; al, e., Fiji: an open-source platform for biological-image analysis. Nature Methods 2012, 9, 676.

23. Hartmann, R.; al, e., Quantitative image analysis of microbial communities with BiofilmQ. Nature Microbiology 2021, 6 (2), 151.

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

Critical synopsis of the articles cited by referee 2:

(1) ‘Generalized workflow for characterization of Nernstian dyes and their effects on bacterial physiology’, L.Mancini et al, Biophysical Journal, 2020, 118, 1, 4-14.

This is the central article used by referee 2 to argue that there are issues with the calibration of ThT for the measurement of membrane potentials. The authors use a simple Nernstian battery (SNB) model and unfortunately it is wrong when voltage-gated ion channels occur. Huge oscillations occur in the membrane potentials of E. coli that cannot be described by the SNB model. Instead a Hodgkin Huxley model is needed, as shown in our eLife manuscript and multiple other studies (see above). Arrhenius kinetics are assumed in the SNB model for pumping with no real evidence and the generalized workflow involves ripping the flagella off the bacteria! The authors construct an elaborate ‘work flow’ to insure their ThT results can be interpreted using their erroneous SNB model over a limited range of parameters.

(2) ‘Non-equivalence of membrane voltage and ion-gradient as driving forces for the bacterial flagellar motor at low load’, C.J.Lo, et al, Biophysical Journal, 2007, 93, 1, 294.

An odd de novo chimeric species is developed using an E. coli  chassis which uses Na+ instead of H+ for the motility of its flagellar motor. It is not clear the relevance to wild type E. coli, due to the massive physiological perturbations involved. A SNB model is using to fit the data over a very limited parameter range with all the concomitant errors.

(3) Single-cell bacterial electrophysiology reveals mechanisms of stress-induced damage’, E.Krasnopeeva, et al, Biophysical Journal, 2019, 116, 2390.

The abstract says ‘PMF defines the physiological state of the cell’. This statement is hyperbolic. An extremely wide range of molecules contribute to the physiological state of a cell. PMF does not even define the electrophysiology of the cell e.g. via the membrane potential. There are 0.2 M of K+ compared with 0.0000001 M of H+ in E. coli, so K+ is arguably a million times more important for the membrane potential than H+ and thus the electrophysiology!

Equation (1) in the manuscript assumes no other ions are exchanged during the experiments other than H+. This is a very bad approximation when voltage-gated potassium ion channels move the majority ion (K+) around!

In our model Figure 4A is better explained by depolarisation due to K+ channels closing than direct irreversible photodamage. Why does the THT fluorescence increase again for the second hyperpolarization event if the THT is supposed to be damaged? It does not make sense.

(4) ‘The proton motive force determines E. coli robustness to extracellular pH’, G.Terradot et al, 2024, preprint.

This article expounds the SNB model once more. It still ignores the voltage-gated ion channels. Furthermore, it ignores the effect of the dominant ion in E. coli, K+. The manuscript is incorrect as a result and I would not recommend publication.

In general, an important problem is being researched i.e. how the membrane potential of E. coli is related to motility, but there are serious flaws in the SNB approach and the experimental methodology appears tenuous.

Answers to specific questions raised by the referees

Reviewer #1 (Public Review):

Summary:

Cell-to-cell communication is essential for higher functions in bacterial biofilms. Electrical signals have proven effective in transmitting signals across biofilms. These signals are then used to coordinate cellular metabolisms or to increase antibiotic tolerance. Here, the authors have reported for the first time coordinated oscillation of membrane potential in E. coli biofilms that may have a functional role in photoprotection.

Strengths:

- The authors report original data.

- For the first time, they showed that coordinated oscillations in membrane potential occur in E. Coli biofilms.

- The authors revealed a complex two-phase dynamic involving distinct molecular response mechanisms.

- The authors developed two rigorous models inspired by 1) Hodgkin-Huxley model for the temporal dynamics of membrane potential and 2) Fire-Diffuse-Fire model for the propagation of the electric signal.

- Since its discovery by comparative genomics, the Kch ion channel has not been associated with any specific phenotype in E. coli. Here, the authors proposed a functional role for the putative K+ Kch channel : enhancing survival under photo-toxic conditions.

We thank the referee for their positive evaluations and agree with these statements.

Weaknesses:

- Since the flow of fresh medium is stopped at the beginning of the acquisition, environmental parameters such as pH and RedOx potential are likely to vary significantly during the experiment. It is therefore important to exclude the contributions of these variations to ensure that the electrical response is only induced by light stimulation. Unfortunately, no control experiments were carried out to address this issue.

The electrical responses occur almost instantaneously when the stimulation with blue light begins i.e. it is too fast to be a build of pH. We are not sure what the referee means by Redox potential since it is an attribute of all chemicals that are able to donate/receive electrons. The electrical response to stress appears to be caused by ROS, since when ROS scavengers are added the electrical response is removed i.e. pH plays a very small minority role if any.

- Furthermore, the control parameter of the experiment (light stimulation) is the same as that used to measure the electrical response, i.e. through fluorescence excitation. The use of the PROPS system could solve this problem.

>>We were enthusiastic at the start of the project to use the PROPs system in E. coli as presented by J.M.Krajl et al, ‘Electrical spiking in E. coli probed with a fluorescent voltage-indicating protein’, Science, 2011, 333, 6040, 345. However, the people we contacted in the microbiology community said that it had some technical issues and there have been no subsequent studies using PROPs in bacteria after the initial promising study. The fluorescent protein system recently presented in PNAS seems more promising, ‘Sensitive bacterial Vm sensors revealed the excitability of bacterial Vm and its role in antibiotic tolerance’, X.Jin et al, PNAS, 120, 3, e2208348120.

- Electrical signal propagation is an important aspect of the manuscript. However, a detailed quantitative analysis of the spatial dynamics within the biofilm is lacking. In addition, it is unclear if the electrical signal propagates within the biofilm during the second peak regime, which is mediated by the Kch channel. This is an important question, given that the fire-diffuse-fire model is presented with emphasis on the role of K+ ions.

We have presented a more detailed account of the electrical wavefront modelling work and it is currently under review in a physical journal, ‘Electrical signalling in three dimensional bacterial biofilms using an agent based fire-diffuse-fire model’, V.Martorelli, et al, 2024 https://www.biorxiv.org/content/10.1101/2023.11.17.567515v1

- Since deletion of the kch gene inhibits the long-term electrical response to light stimulation (regime II), the authors concluded that K+ ions play a role in the habituation response. However, Kch is a putative K+ ion channel. The use of specific drugs could help to clarify the role of K+ ions.

Our recent electrical impedance spectroscopy publication provides further evidence that Kch is associated with large changes in conductivity as expected for a voltage-gated ion channel (https://pubs.acs.org/doi/10.1021/acs.nanolett.3c04446, 'Electrical impedance spectroscopy with bacterial biofilms: neuronal-like behavior', E.Akabuogu et al, ACS Nanoletters, 2024, in print.

- The manuscript as such does not allow us to properly conclude on the photo-protective role of the Kch ion channel.

That Kch has a photoprotective role is our current working hypothesis. The hypothesis fits with the data, but we are not saying we have proven it beyond all possible doubt.

- The link between membrane potential dynamics and mechanosensitivity is not captured in the equation for the Q-channel opening dynamics in the Hodgkin-Huxley model (Supp Eq 2).

Our model is agnostic with respect to the mechanosensitivity of the ion channels, although we deduce that mechanosensitive ion channels contribute to ion channel Q.

- Given the large number of parameters used in the models, it is hard to distinguish between prediction and fitting.

This is always an issue with electrophysiological modelling (compared with most heart and brain modelling studies we are very conservative in the choice of parameters for the bacteria). In terms of predicting the different phenomena observed, we believe the model is very successful.

Reviewer #2 (Public Review):

Summary of what the authors were trying to achieve:

The authors thought they studied membrane potential dynamics in E.coli biofilms. They thought so because they were unaware that the dye they used to report that membrane potential in E.coli, has been previously shown not to report it. Because of this, the interpretation of the authors' results is not accurate.

We believe the Pilizota work is scientifically flawed.

Major strengths and weaknesses of the methods and results:

The strength of this work is that all the data is presented clearly, and accurately, as far as I can tell.

The major critical weakness of this paper is the use of ThT dye as a membrane potential dye in E.coli. The work is unaware of a publication from 2020 https://www.sciencedirect.com/science/article/pii/S0006349519308793 [sciencedirect.com] that demonstrates that ThT is not a membrane potential dye in E. coli. Therefore I think the results of this paper are misinterpreted. The same publication I reference above presents a protocol on how to carefully calibrate any candidate membrane potential dye in any given condition.

We are aware of this study, but believe it to be scientifically flawed. We do not cite the article because we do not think it is a particularly useful contribution to the literature.

I now go over each results section in the manuscript.

Result section 1: Blue light triggers electrical spiking in single E. coli cells

I do not think the title of the result section is correct for the following reasons. The above-referenced work demonstrates the loading profile one should expect from a Nernstian dye (Figure 1). It also demonstrates that ThT does not show that profile and explains why is this so. ThT only permeates the membrane under light exposure (Figure 5). This finding is consistent with blue light peroxidising the membrane (see also following work Figure 4 https://www.sciencedirect.com/science/article/pii/S0006349519303923 [sciencedirect.com] on light-induced damage to the electrochemical gradient of protons-I am sure there are more references for this).

The Pilizota group invokes some elaborate artefacts to explain the lack of agreement with a simple Nernstian battery model. The model is incorrect not the fluorophore.

Please note that the loading profile (only observed under light) in the current manuscript in Figure 1B as well as in the video S1 is identical to that in Figure 3 from the above-referenced paper (i.e. https://www.sciencedirect.com/science/article/pii/S0006349519308793 [sciencedirect.com]), and corresponding videos S3 and S4. This kind of profile is exactly what one would expect theoretically if the light is simultaneously lowering the membrane potential as the ThT is equilibrating, see Figure S12 of that previous work. There, it is also demonstrated by the means of monitoring the speed of bacterial flagellar motor that the electrochemical gradient of protons is being lowered by the light. The authors state that applying the blue light for different time periods and over different time scales did not change the peak profile. This is expected if the light is lowering the electrochemical gradient of protons. But, in Figure S1, it is clear that it affected the timing of the peak, which is again expected, because the light affects the timing of the decay, and thus of the decay profile of the electrochemical gradient of protons (Figure 4 https://www.sciencedirect.com/science/article/pii/S0006349519303923 [sciencedirect.com]).

We think the proton effect is a million times weaker than that due to potasium i.e. 0.2 M K+ versus 10-7 M H+. We can comfortably neglect the influx of H+ in our experiments.

If find Figure S1D interesting. There authors load TMRM, which is a membrane voltage dye that has been used extensively (as far as I am aware this is the first reference for that and it has not been cited https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1914430 [ncbi.nlm.nih.gov]/). As visible from the last TMRM reference I give, TMRM will only load the cells in Potassium Phosphate buffer with NaCl (and often we used EDTA to permeabilise the membrane). It is not fully clear (to me) whether here TMRM was prepared in rich media (it explicitly says so for ThT in Methods but not for TMRM), but it seems so. If this is the case, it likely also loads because of the damage to the membrane done with light, and therefore I am not surprised that the profiles are similar.

The vast majority of cells continue to be viable. We do not think membrane damage is dominating.

The authors then use CCCP. First, a small correction, as the authors state that it quenches membrane potential. CCCP is a protonophore (https://pubmed.ncbi.nlm.nih.gov/4962086 [pubmed.ncbi.nlm.nih.gov]/), so it collapses electrochemical gradient of protons. This means that it is possible, and this will depend on the type of pumps present in the cell, that CCCP collapses electrochemical gradient of protons, but the membrane potential is equal and opposite in sign to the DeltapH. So using CCCP does not automatically mean membrane potential will collapse (e.g. in some mammalian cells it does not need to be the case, but in E.coli it is https://www.biorxiv.org/content/10.1101/2021.11.19.469321v2 [biorxiv.org]). CCCP has also been recently found to be a substrate for TolC (https://journals.asm.org/doi/10.1128/mbio.00676-21 [journals.asm.org]), but at the concentrations the authors are using CCCP (100uM) that should not affect the results. However, the authors then state because they observed, in Figure S1E, a fast efflux of ions in all cells and no spiking dynamics this confirms that observed dynamics are membrane potential related. I do not agree that it does. First, Figure S1E, does not appear to show transients, instead, it is visible that after 50min treatment with 100uM CCCP, ThT dye shows no dynamics. The action of a Nernstian dye is defined. It is not sufficient that a charged molecule is affected in some way by electrical potential, this needs to be in a very specific way to be a Nernstian dye. Part of the profile of ThT loading observed in https://www.sciencedirect.com/science/article/pii/S0006349519308793 [sciencedirect.com] is membrane potential related, but not in a way that is characteristic of Nernstian dye.

Our understanding of the literature is CCCP poisons the whole metabolism of the bacterial cells. The ATP driven K+ channels will stop functioning and this is the dominant contributor to membrane potential.

Result section 2: Membrane potential dynamics depend on the intercellular distance

In this chapter, the authors report that the time to reach the first intensity peak during ThT loading is different when cells are in microclusters. They interpret this as electrical signalling in clusters because the peak is reached faster in microclusters (as opposed to slower because intuitively in these clusters cells could be shielded from light). However, shielding is one possibility. The other is that the membrane has changed in composition and/or the effective light power the cells can tolerate (with mechanisms to handle light-induced damage, some of which authors mention later in the paper) is lower. Given that these cells were left in a microfluidic chamber for 2h hours to attach in growth media according to Methods, there is sufficient time for that to happen. In Figure S12 C and D of that same paper from my group (https://ars.els-cdn.com/content/image/1-s2.0-S0006349519308793-mmc6.pdf [ars.els-cdn.com]) one can see the effects of peak intensity and timing of the peak on the permeability of the membrane. Therefore I do not think the distance is the explanation for what authors observe.

Shielding would provide the reverse effect, since hyperpolarization begins in the dense centres of the biofilms. For the initial 2 hours the cells receive negligible blue light. Neither of the referee’s comments thus seem tenable.

Result section 3: Emergence of synchronized global wavefronts in E. coli biofilms

In this section, the authors exposed a mature biofilm to blue light. They observe that the intensity peak is reached faster in the cells in the middle. They interpret this as the ion-channel-mediated wavefronts moved from the center of the biofilm. As above, cells in the middle can have different membrane permeability to those at the periphery, and probably even more importantly, there is no light profile shown anywhere in SI/Methods. I could be wrong, but the SI3 A profile is consistent with a potential Gaussian beam profile visible in the field of view. In Methods, I find the light source for the blue light and the type of microscope but no comments on how 'flat' the illumination is across their field of view. This is critical to assess what they are observing in this result section. I do find it interesting that the ThT intensity collapsed from the edges of the biofilms. In the publication I mentioned https://www.sciencedirect.com/science/article/pii/S0006349519308793#app2 [sciencedirect.com], the collapse of fluorescence was not understood (other than it is not membrane potential related). It was observed in Figure 5A, C, and F, that at the point of peak, electrochemical gradient of protons is already collapsed, and that at the point of peak cell expands and cytoplasmic content leaks out. This means that this part of the ThT curve is not membrane potential related. The authors see that after the first peak collapsed there is a period of time where ThT does not stain the cells and then it starts again. If after the first peak the cellular content leaks, as we have observed, then staining that occurs much later could be simply staining of cytoplasmic positively charged content, and the timing of that depends on the dynamics of cytoplasmic content leakage (we observed this to be happening over 2h in individual cells). ThT is also a non-specific amyloid dye, and in starving E. coli cells formation of protein clusters has been observed (https://pubmed.ncbi.nlm.nih.gov/30472191 [pubmed.ncbi.nlm.nih.gov]/), so such cytoplasmic staining seems possible.

>>It is very easy to see if the illumination is flat (Köhler illumination) by comparing the intensity of background pixels on the detector. It was flat in our case. Protons have little to do with our work for reasons highlighted before. Differential membrane permittivity is a speculative phenomenon not well supported by any evidence and with no clear molecular mechanism.

Finally, I note that authors observe biofilms of different shapes and sizes and state that they observe similar intensity profiles, which could mean that my comment on 'flatness' of the field of view above is not a concern. However, the scale bar in Figure 2A is not legible, so I can't compare it to the variation of sizes of the biofilms in Figure 2C (67 to 280um). Based on this, I think that the illumination profile is still a concern.

The referee now contradicts themselves and wants a scale bar to be more visible. We have changed the scale bar.

Result section 4: Voltage-gated Kch potassium channels mediate ion-channel electrical oscillations in E. coli

First I note at this point, given that I disagree that the data presented thus 'suggest that E. coli biofilms use electrical signaling to coordinate long-range responses to light stress' as the authors state, it gets harder to comment on the rest of the results.

In this result section the authors look at the effect of Kch, a putative voltage-gated potassium channel, on ThT profile in E. coli cells. And they see a difference. It is worth noting that in the publication https://www.sciencedirect.com/science/article/pii/S0006349519308793 [sciencedirect.com] it is found that ThT is also likely a substrate for TolC (Figure 4), but that scenario could not be distinguished from the one where TolC mutant has a different membrane permeability (and there is a publication that suggests the latter is happening https://onlinelibrary.wiley.com/doi/10.1111/j.1365-2958.2010.07245.x [onlinelibrary.wiley.com]). Given this, it is also possible that Kch deletion affects the membrane permeability. I do note that in video S4 I seem to see more of, what appear to be, plasmolysed cells. The authors do not see the ThT intensity with this mutant that appears long after the initial peak has disappeared, as they see in WT. It is not clear how long they waited for this, as from Figure S3C it could simply be that the dynamics of this is a lot slower, e.g. Kch deletion changes membrane permeability.

The work that TolC provides a possible passive pathway for ThT to leave cells seems slightly niche. It just demonstrates another mechanism for the cells to equilibriate the concentrations of ThT in a Nernstian manner i.e. driven by the membrane voltage.

The authors themselves state that the evidence for Kch being a voltage-gated channel is indirect (line 54). I do not think there is a need to claim function from a ThT profile of E. coli mutants (nor do I believe it's good practice), given how accurate single-channel recordings are currently. To know the exact dependency on the membrane potential, ion channel recordings on this protein are needed first.

We have good evidence form electrical impedance spectroscopy experiments that Kch increases the conductivity of biofilms  (https://pubs.acs.org/doi/10.1021/acs.nanolett.3c04446, 'Electrical impedance spectroscopy with bacterial biofilms: neuronal-like behavior', E.Akabuogu et al, ACS Nanoletters, 2024, in print.

Result section 5: Blue light influences ion-channel mediated membrane potential events in E. coli

In this chapter the authors vary the light intensity and stain the cells with PI (this dye gets into the cells when the membrane becomes very permeable), and the extracellular environment with K+ dye (I have not yet worked carefully with this dye). They find that different amounts of light influence ThT dynamics. This is in line with previous literature (both papers I have been mentioning: Figure 4 https://www.sciencedirect.com/science/article/pii/S0006349519303923 [sciencedirect.com] and https://ars.els-cdn.com/content/image/1-s2.0-S0006349519308793-mmc6.pdf [ars.els-cdn.com] especially SI12), but does not add anything new. I think the results presented here can be explained with previously published theory and do not indicate that the ion-channel mediated membrane potential dynamics is a light stress relief process.

The simple Nernstian battery model proposed by Pilizota et al is erroneous in our opinion for reasons outlined above. We believe it will prove to be a dead end for bacterial electrophysiology studies.

Result section 6: Development of a Hodgkin-Huxley model for the observed membrane potential dynamics

This results section starts with the authors stating: 'our data provide evidence that E. coli manages light stress through well-controlled modulation of its membrane potential dynamics'. As stated above, I think they are instead observing the process of ThT loading while the light is damaging the membrane and thus simultaneously collapsing the electrochemical gradient of protons. As stated above, this has been modelled before. And then, they observe a ThT staining that is independent from membrane potential.

This is an erroneous niche opinion. Protons have little say in the membrane potential since there are so few of them. The membrane potential is mostly determined by K+.

I will briefly comment on the Hodgkin Huxley (HH) based model. First, I think there is no evidence for two channels with different activation profiles as authors propose. But also, the HH model has been developed for neurons. There, the leakage and the pumping fluxes are both described by a constant representing conductivity, times the difference between the membrane potential and Nernst potential for the given ion. The conductivity in the model is given as gK*n^4 for potassium, gNa*m^3*h sodium, and gL for leakage, where gK, gNa and gL were measured experimentally for neurons. And, n, m, and h are variables that describe the experimentally observed voltage-gated mechanism of neuronal sodium and potassium channels. (Please see Hodgkin AL, Huxley AF. 1952. Currents carried by sodium and potassium ions through the membrane of the giant axon of Loligo. J. Physiol. 116:449-72 and Hodgkin AL, Huxley AF. 1952. A quantitative description of membrane current and its application to conduction and excitation in nerve. J. Physiol. 117:500-44).

In the 70 years since Hodgkin and Huxley first presented their model, a huge number of similar models have been proposed to describe cellular electrophysiology. We are not being hyperbolic when we state that the HH models for excitable cells are like the Schrödinger equation for molecules. We carefully adapted our HH model to reflect the currently understood electrophysiology of E. coli.

Thus, in applying the model to describe bacterial electrophysiology one should ensure near equilibrium requirement holds (so that (V-VQ) etc terms in authors' equation Figure 5 B hold), and potassium and other channels in a given bacterium have similar gating properties to those found in neurons. I am not aware of such measurements in any bacteria, and therefore think the pump leak model of the electrophysiology of bacteria needs to start with fluxes that are more general (for example Keener JP, Sneyd J. 2009. Mathematical physiology: I: Cellular physiology. New York: Springer or https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0000144 [journals.plos.org])

The reference is to a slightly more modern version of a simple Nernstian battery model. The model will not oscillate and thus will not help modelling membrane potentials in bacteria. We are unsure where the equilibrium requirement comes from (inadequate modelling of the dynamics?)

Result section 7: Mechanosensitive ion channels (MS) are vital for the first hyperpolarization event in E. coli.

The results that Mcs channels affect the profile of ThT dye are interesting. It is again possible that the membrane permeability of these mutants has changed and therefore the dynamics have changed, so this needs to be checked first. I also note that our results show that the peak of ThT coincides with cell expansion. For this to be understood a model is needed that also takes into account the link between maintenance of electrochemical gradients of ions in the cell and osmotic pressure.

The evidence for permeability changes in the membranes seems to be tenuous.

A side note is that the authors state that the Msc responds to stress-related voltage changes. I think this is an overstatement. Mscs respond to predominantly membrane tension and are mostly nonspecific (see how their action recovers cellular volume in this publication https://www.pnas.org/doi/full/10.1073/pnas.1522185113 [pnas.org]). Authors cite references 35-39 to support this statement. These publications still state that these channels are predominantly membrane tension-gated. Some of the references state that the presence of external ions is important for tension-related gating but sometimes they gate spontaneously in the presence of certain ions. Other publications cited don't really look at gating with respect to ions (39 is on clustering). This is why I think the statement is somewhat misleading.

We have reworded the discussion of Mscs since the literature appears to be ambiguous. We will try to run some electrical impedance spectroscopy experiments on the Msc mutants in the future to attempt to remove the ambiguity.

Result section 8: Anomalous ion-channel-mediated wavefronts propagate light stress signals in 3D E. coli biofilms.

I am not commenting on this result section, as it would only be applicable if ThT was membrane potential dye in E. coli.

Ok, but we disagree on the use of ThT.

Aims achieved/results support their conclusions:

The authors clearly present their data. I am convinced that they have accurately presented everything they observed. However, I think their interpretation of the data and conclusions is inaccurate in line with the discussion I provided above.

Likely impact of the work on the field, and the utility of the methods and data to the community:

I do not think this publication should be published in its current format. It should be revised in light of the previous literature as discussed in detail above. I believe presenting it in it's current form on eLife pages would create unnecessary confusion.

We believe many of the Pilizota group articles are scientifically flawed and are causing the confusion in the literature.

Any other comments:

I note, that while this work studies E. coli, it references papers in other bacteria using ThT. For example, in lines 35-36 authors state that bacteria (Bacillus subtilis in this case) in biofilms have been recently found to modulate membrane potential citing the relevant literature from 2015. It is worth noting that the most recent paper https://journals.asm.org/doi/10.1128/mbio.02220-23 [journals.asm.org] found that ThT binds to one or more proteins in the spore coat, suggesting that it does not act as a membrane potential in Bacillus spores. It is possible that it still reports membrane potential in Bacillus cells and the recent results are strictly spore-specific, but these should be kept in mind when using ThT with Bacillus.

>>ThT was used successfully in previous studies of normal B. subtilis cells (by our own group and A.Prindle, ‘Spatial propagation of electrical signal in circular biofilms’, J.A.Blee et al, Physical Review E, 2019, 100, 052401, J.A.Blee et al, ‘Membrane potentials, oxidative stress and the dispersal response of bacterial biofilms to 405 nm light’, Physical Biology, 2020, 17, 2, 036001, A.Prindle et al, ‘Ion channels enable electrical communication in bacterial communities’, Nature, 2015, 527, 59-63). The connection to low metabolism pore research seems speculative.

Reviewer #3 (Public Review):

It has recently been demonstrated that bacteria in biofilms show changes in membrane potential in response to changes in their environment, and that these can propagate signals through the biofilm to coordinate bacterial behavior. Akabuogu et al. contribute to this exciting research area with a study of blue light-induced membrane potential dynamics in E. coli biofilms. They demonstrate that Thioflavin-T (ThT) intensity (a proxy for membrane potential) displays multiphasic dynamics in response to blue light treatment. They additionally use genetic manipulations to implicate the potassium channel Kch in the latter part of these dynamics. Mechanosensitive ion channels may also be involved, although these channels seem to have blue light-independent effects on membrane potential as well. In addition, there are challenges to the quantitative interpretation of ThT microscopy data which require consideration. The authors then explore whether these dynamics are involved in signaling at the community level. The authors suggest that cell firing is both more coordinated when cells are clustered and happens in waves in larger, 3D biofilms; however, in both cases evidence for these claims is incomplete. The authors present two simulations to describe the ThT data. The first of these simulations, a Hodgkin-Huxley model, indicates that the data are consistent with the activity of two ion channels with different kinetics; the Kch channel mutant, which ablates a specific portion of the response curve, is consistent with this. The second model is a fire-diffuse-fire model to describe wavefront propagation of membrane potential changes in a 3D biofilm; because the wavefront data are not presented clearly, the results of this model are difficult to interpret. Finally, the authors discuss whether these membrane potential changes could be involved in generating a protective response to blue light exposure; increased death in a Kch ion channel mutant upon blue light exposure suggests that this may be the case, but a no-light control is needed to clarify this.

In a few instances, the paper is missing key control experiments that are important to the interpretation of the data. This makes it difficult to judge the meaning of some of the presented experiments.

(1) An additional control for the effects of autofluorescence is very important. The authors conduct an experiment where they treat cells with CCCP and see that Thioflavin-T (ThT) dynamics do not change over the course of the experiment. They suggest that this demonstrates that autofluorescence does not impact their measurements. However, cellular autofluorescence depends on the physiological state of the cell, which is impacted by CCCP treatment. A much simpler and more direct experiment would be to repeat the measurement in the absence of ThT or any other stain. This experiment should be performed both in the wild-type strain and in the ∆kch mutant.

ThT is a very bright fluorophore (much brighter than a GFP). It is clear from the images of non-stained samples that autofluorescence provides a negligible contribution to the fluorescence intensity in an image.

(2) The effects of photobleaching should be considered. Of course, the intensity varies a lot over the course of the experiment in a way that photobleaching alone cannot explain. However, photobleaching can still contribute to the kinetics observed. Photobleaching can be assessed by changing the intensity, duration, or frequency of exposure to excitation light during the experiment. Considerations about photobleaching become particularly important when considering the effect of catalase on ThT intensity. The authors find that the decrease in ThT signal after the initial "spike" is attenuated by the addition of catalase; this is what would be predicted by catalase protecting ThT from photobleaching (indeed, catalase can be used to reduce photobleaching in time lapse imaging).

Photobleaching was negligible over the course of the experiments. We employed techniques such as reducing sample exposure time and using the appropriate light intensity to minimize photobleaching.

(3) It would be helpful to have a baseline of membrane potential fluctuations in the absence of the proposed stimulus (in this case, blue light). Including traces of membrane potential recorded without light present would help support the claim that these changes in membrane potential represent a blue light-specific stress response, as the authors suggest. Of course, ThT is blue, so if the excitation light for ThT is problematic for this experiment the alternative dye tetramethylrhodamine methyl ester perchlorate (TMRM) can be used instead.

Unfortunately the fluorescent baseline is too weak to measure cleanly in this experiment. It appears the collective response of all the bacteria hyperpolarization at the same time appears to dominate the signal (measurements in the eLife article and new potentiometry measurements).

(4) The effects of ThT in combination with blue light should be more carefully considered. In mitochondria, a combination of high concentrations of blue light and ThT leads to disruption of the PMF (Skates et al. 2021 BioRXiv), and similarly, ThT treatment enhances the photodynamic effects of blue light in E. coli (Bondia et al. 2021 Chemical Communications). If present in this experiment, this effect could confound the interpretation of the PMF dynamics reported in the paper.

We think the PMF plays a minority role in determining the membrane potential in E. coli. For reasons outlined before (H+ is a minority ion in E. coli compared with K+).

(5) Figures 4D - E indicate that a ∆kch mutant has increased propidium iodide (PI) staining in the presence of blue light; this is interpreted to mean that Kch-mediated membrane potential dynamics help protect cells from blue light. However, Live/Dead staining results in these strains in the absence of blue light are not reported. This means that the possibility that the ∆kch mutant has a general decrease in survival (independent of any effects of blue light) cannot be ruled out.

>>Both strains of bacterial has similar growth curve and also engaged in membrane potential dynamics for the duration of the experiment. We were interested in bacterial cells that observed membrane potential dynamics in the presence of the stress. Bacterial cells need to be alive to engage in membrane potential  dynamics (hyperpolarize) under stress conditions. Cells that engaged in membrane potential dynamics and later stained red were only counted after the entire duration. We believe that the wildtype handles the light stress better than the ∆kch mutant as measured with the PI.

(6) Additionally in Figures 4D - E, the interpretation of this experiment can be confounded by the fact that PI uptake can sometimes be seen in bacterial cells with high membrane potential (Kirchhoff & Cypionka 2017 J Microbial Methods); the interpretation is that high membrane potential can lead to increased PI permeability. Because the membrane potential is largely higher throughout blue light treatment in the ∆kch mutant (Fig. 3AB), this complicates the interpretation of this experiment.

Kirchhoff & Cypionka 2017 J Microbial Methods, using fluorescence microscopy, suggested that changes in membrane potential dynamics can introduce experimental bias when propidium iodide is used to confirm the viability of tge bacterial strains, B subtilis (DSM-10) and Dinoroseobacter shibae, that are starved of oxygen (via N2 gassing) for 2 hours. They attempted to support their findings by using CCCP in stopping the membrane potential dynamics (but never showed any pictoral or plotted data for this confirmatory experiment). In our experiment methodology, cell death was not forced on the cells by introducing an extra burden or via anoxia. We believe that the accumulation of PI in ∆kch mutant is not due to high membrane potential dynamics but is attributed to the PI, unbiasedly showing damaged/dead cells. We think that propidium iodide is good for this experiment. Propidium iodide is a dye that is extensively used in life sciences. PI has also been used in the study of bacterial electrophysiology (https://pubmed.ncbi.nlm.nih.gov/32343961/, ) and no membrane potential related bias was reported.

Throughout the paper, many ThT intensity traces are compared, and described as "similar" or "dissimilar", without detailed discussion or a clear standard for comparison. For example, the two membrane potential curves in Fig. S1C are described as "similar" although they have very different shapes, whereas the curves in Fig. 1B and 1D are discussed in terms of their differences although they are evidently much more similar to one another. Without metrics or statistics to compare these curves, it is hard to interpret these claims. These comparative interpretations are additionally challenging because many of the figures in which average trace data are presented do not indicate standard deviation.

Comparison of small changes in the absolute intensities is problematic in such fluorescence experiments. We mean the shape of the traces is similar and they can be modelled using a HH model with similar parameters.

The differences between the TMRM and ThT curves that the authors show in Fig. S1C warrant further consideration. Some of the key features of the response in the ThT curve (on which much of the modeling work in the paper relies) are not very apparent in the TMRM data. It is not obvious to me which of these traces will be more representative of the actual underlying membrane potential dynamics.

In our experiment, TMRM was used to confirm the dynamics observed using ThT. However, ThT appear to be more photostable than TMRM (especially towars the 2nd peak). The most interesting observation is that with both dyes, all phases of the membrane potential dynamics were conspicuous (the first peak, the quiescent period and the second peak). The time periods for these three episodes were also similar.

A key claim in this paper (that dynamics of firing differ depending on whether cells are alone or in a colony) is underpinned by "time-to-first peak" analysis, but there are some challenges in interpreting these results. The authors report an average time-to-first peak of 7.34 min for the data in Figure 1B, but the average curve in Figure 1B peaks earlier than this. In Figure 1E, it appears that there are a handful of outliers in the "sparse cell" condition that likely explain this discrepancy. Either an outlier analysis should be done and the mean recomputed accordingly, or a more outlier-robust method like the median should be used instead. Then, a statistical comparison of these results will indicate whether there is a significant difference between them.

The key point is the comparison of standard errors on the standard deviation.

In two different 3D biofilm experiments, the authors report the propagation of wavefronts of membrane potential; I am unable to discern these wavefronts in the imaging data, and they are not clearly demonstrated by analysis.

The first data set is presented in Figures 2A, 2B, and Video S3. The images and video are very difficult to interpret because of how the images have been scaled: the center of the biofilm is highly saturated, and the zero value has also been set too high to consistently observe the single cells surrounding the biofilm. With the images scaled this way, it is very difficult to assess dynamics. The time stamps in Video S3 and on the panels in Figure 2A also do not correspond to one another although the same biofilm is shown (and the time course in 2B is also different from what is indicated in 2B). In either case, it appears that the center of the biofilm is consistently brighter than the edges, and the intensity of all cells in the biofilm increases in tandem; by eye, propagating wavefronts (either directed toward the edge or the center) are not evident to me. Increased brightness at the center of the biofilm could be explained by increased cell thickness there (as is typical in this type of biofilm). From the image legend, it is not clear whether the image presented is a single confocal slice or a projection. Even if this is a single confocal slice, in both Video S3 and Figure 2A there are regions of "haze" from out-of-focus light evident, suggesting that light from other focal planes is nonetheless present. This seems to me to be a simpler explanation for the fluorescence dynamics observed in this experiment: cells are all following the same trajectory that corresponds to that seen for single cells, and the center is brighter because of increased biofilm thickness.

We appreciate the reviewer for this important observation. We have made changes to the figures to address this confusion. The cell cover has no influence on the observed membrane potential dynamics. The entire biofilm was exposed to the same blue light at each time. Therefore all parts of the biofilm received equal amounts of the blue light intensity. The membrane potential dynamics was not influenced by cell density (see Fig 2C). 

The second data set is presented in Video S6B; I am similarly unable to see any wave propagation in this video. I observe only a consistent decrease in fluorescence intensity throughout the experiment that is spatially uniform (except for the bright, dynamic cells near the top; these presumably represent cells that are floating in the microfluidic and have newly arrived to the imaging region).

A visual inspection of Video S6B shows a fast rise, a decrease in fluorescence and a second rise (supplementary figure 4B). The data for the fluorescence was carefully obtained using the imaris software. We created a curved geometry on each slice of the confocal stack. We analyzed the surfaces of this curved plane along the z-axis. This was carried out in imaris.

3D imaging data can be difficult to interpret by eye, so it would perhaps be more helpful to demonstrate these propagating wavefronts by analysis; however, such analysis is not presented in a clear way. The legend in Figure 2B mentions a "wavefront trace", but there is no position information included - this trace instead seems to represent the average intensity trace of all cells. To demonstrate the propagation of a wavefront, this analysis should be shown for different subpopulations of cells at different positions from the center of the biofilm. Data is shown in Figure 8 that reflects the velocity of the wavefront as a function of biofilm position; however, because the wavefronts themselves are not evident in the data, it is difficult to interpret this analysis. The methods section additionally does not contain sufficient information about what these velocities represent and how they are calculated. Because of this, it is difficult for me to evaluate the section of the paper pertaining to wave propagation and the predicted biofilm critical size.

The analysis is considered in more detail in a more expansive modelling article, currently under peer review in a physics journal, ‘Electrical signalling in three dimensional bacterial biofilms using an agent based fire-diffuse-fire model’, V.Martorelli, et al, 2024 https://www.biorxiv.org/content/10.1101/2023.11.17.567515v1

There are some instances in the paper where claims are made that do not have data shown or are not evident in the cited data:

(1) In the first results section, "When CCCP was added, we observed a fast efflux of ions in all cells"- the data figure pertaining to this experiment is in Fig. S1E, which does not show any ion efflux. The methods section does not mention how ion efflux was measured during CCCP treatment.

We have worded this differently to properly convey our results.

(2) In the discussion of voltage-gated calcium channels, the authors refer to "spiking events", but these are not obvious in Figure S3E. Although the fluorescence intensity changes over time, it's hard to distinguish these fluctuations from measurement noise; a no-light control could help clarify this.

The calcium transients observed were not due to noise or artefacts.

(3) The authors state that the membrane potential dynamics simulated in Figure 7B are similar to those observed in 3D biofilms in Fig. S4B; however, the second peak is not clearly evident in Fig. S4B and it looks very different for the mature biofilm data reported in Fig. 2. I have some additional confusion about this data specifically: in the intensity trace shown in Fig. S4B, the intensity in the second frame is much higher than the first; this is not evident in Video S6B, in which the highest intensity is in the first frame at time 0. Similarly, the graph indicates that the intensity at 60 minutes is higher than the intensity at 4 minutes, but this is not the case in Fig. S4A or Video S6B.

The confusion stated here has now been addressed. Also it should be noted that while Fig 2.1 was obtained with LED light source, Fig S4A was obtained using a laser light source. While obtaining the confocal images (for Fig S4A ), the light intensity was controlled to further minimize photobleaching. Most importantly, there is an evidence of slow rise to the 2nd peak in Fig S4B. The first peak, quiescence and slow rise to second peak are evident.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Scientific recommendations:

- Although Fig 4A clearly shows that light stimulation has an influence on the dynamics of cell membrane potential in the biofilm, it is important to rule out the contribution of variations in environmental parameters. I understand that for technical reasons, the flow of fresh medium must be stopped during image acquisition. Therefore, I suggest performing control experiments, where the flow is stopped before image acquisition (15min, 30min, 45min, and 1h before). If there is no significant contribution from environmental variations (pH, RedOx), the dynamics of the electrical response should be superimposed whatever the delay between stopping the flow stop and switching on the light.

In this current research study, we were focused on studying how E. coli cells and biofilms react to blue light stress via their membrane potential dynamics. This involved growing the cells and biofilms, stopping the media flow and obtaining data immediately. We believe that stopping the flow not only helped us to manage data acquisition, it also helped us reduce the effect of environmental factors. In our future study we will expand the work to include how the membrane potential dynamics evolve in the presence of changing environmental factors for example such induced by stopping the flow at varied times.

- Since TMRM signal exhibits a linear increase after the first response peak (Supplementary Figure 1D), I recommend mitigating the statement at line 78.

- To improve the spatial analysis of the electrical response, I suggest plotting kymographs of the intensity profiles across the biofilm. I have plotted this kymograph for Video S3 and it appears that there is no electrical propagation for the second peak. In addition, the authors should provide technical details of how R^2(t) is measured in the first regime (Figure 7E).

See the dedicated simulation article for more details. https://www.biorxiv.org/content/10.1101/2023.11.17.567515v1

- Line 152: To assess the variability of the latency, the authors should consider measuring the variance divided by the mean instead of SD, which may depend on the average value.

We are happy with our current use of standard error on the standard deviation. It shows what we claim to be true.

- Line 154-155: To truly determine whether the amplitude of the "action potential" is independent of biofilm size, the authors should not normalise the signals.

Good point. We qualitatively compared both normalized and unnormalized data. Recent electrical impedance spectroscopy measurements (unpublished) indicate that the electrical activity is an extensive quantity i.e. it scales with the size of the biofilms.

- To precise the role of K+ in the habituation response, I suggest using valinomycin at sub-inhibitory concentrations (10µM). Besides, the high concentration of CCCP used in this study completely inhibits cell activity. Not surprisingly, no electrical response to light stimulation was observed in the presence of CCCP. Finally, the Kch complementation experiment exhibits a "drop after the first peak" on a single point. It would be more convincing to increase the temporal resolution (1min->10s) to show that there is indeed a first and a second peak.

An interesting experiment for the future.

- Line 237-238: There are only two points suggesting that the dynamics of hyperpolarization are faster at higher irradiance(Fig 4A). The authors should consider adding a third intermediate point at 17µW/mm^2 to confirm the statement made in this sentence.

Multiple repeats were performed. We are confident of the robustness of our data.

- Line 249 + Fig 4E: It seems that the data reported on Fig 4E are extracted from Fig 4D. If this is indeed the case, the data should be normalised by the total population size to compare survival probabilities under the two conditions. It would also be great to measure these probabilities (for WT and ∆kch) in the presence of ROS scavengers.

- To distinguish between model fitting and model predictions, the authors should clearly state which parameters are taken from the literature and which parameters are adjusted to fit the experimental data.

- Supplementary Figure 4A: why can't we see any wavefront in this series of images?

For the experimental data, the wavefront was analyzed by employing the imaris software. We systematically created a ROI with a curved geometry within the confocal stack (the biofilm). The fluorescence of ThT was traced along the surface of the curved geometry was analyzed along the z-axis.

- Fig 7B: Could the authors explain why the plateau is higher in the simulations than in the biofilm experiments? Could they add noise on the firing activities?

See the dedicated Martorelli modelling article. In general we would need to approach stochastic Hodgkin-Huxley modelling and the fluorescence data (and electrical impedance spectroscopy data) presented does not have extensive noise (due to collective averaging over many bacteria cells).

- Supplementary Figure 4B: Why can't we see the second peak in confocal images?

The second peak is present although not as robust as in Fig 2B. The confocal images were obtained with a laser source. Therefore we tried to create a balance between applying sufficient light stress on the bacterial cells and mitigating photobleaching.

Editing recommendations:

The editing recommendations below has been applied where appropriate

- Many important technical details are missing (e.g. R^2, curvature, and 445nm irradiance measurements). Error bars are missing from most graphs. The captions should clearly indicate if these are single-cell or biofilm experiments, strain name, illumination conditions, number of experiments, SD, or SE. Please indicate on all panels of all figures in the main text and in the supplements, which are the conditions: single cell vs. biofilm, strains, medium, centrifugal vs centripetal etc..., where relevant. Please also draw error bars everywhere.

We have now made appropriate changes. We specifically use cells when we were dealing with single cells and biofilms when we worked on biofilms. We decided to describe the strain name either on the panel or the image description.

- Line 47-51: The way the paragraph is written suggests that no coordinated electrical oscillations have been observed in Gram-negative biofilms. However, Hennes et al (referenced as 57 in this manuscript) have shown that a wave of hyperpolarized cells propagates in Neisseria gonorrhoea colony, which is a Gram-negative bacterium.

We are now aware of this work. It was not published when we first submitted our work and the authors claim the waves of activity are due to ROS diffusion NOT propagating waves of ions (coordinated electrical wavefronts).

- Line 59: "stressor" -> "stress" or "perturbation".

The correction has been made.

- Line 153: Please indicate in the Material&Methods how the size of the biofilm is measured.

The biofilm size was obtained using BiofilmQ and the step by step guide for using BiofilmQ were stated..

- Figure 2A: Please provide associated brightfield images to locate bacteria.

- Line 186: Please remove "wavefront" from the caption. Fig2B only shows the average signal as a function of time.

This correction has been implemented.

- Fig 3B,C: Please indicate single cell and biofilm on the panels and also WT and ∆kch.

- Line 289: I suggest adding "in single cell experiments" to the title of this section.

- Fig 5A: blue light is always present at regular time intervals during regime I and II. The presence of blue light only in regime I could be misleading.

- Fig 5C: The curve in Fig 5D seems to correspond to the biofilm case. The curve given by the model, should be compared with the average curve presented in Fig 1D.

- Fig 6A, B, and C: These figures could be moved to supplements.

- Line 392: Replace "turgidity" with "turgor pressure".

- Fig 7C,E: Please use a log-log scale to represent these data and indicate the line of slope 1.

- Fig 7E: The x-axis has been cropped.

- Please provide a supplementary movie for the data presented in Fig 7E.

- Line 455: E. Coli biofilms do not express ThT.

- Line 466: "\gamma is the anomalous exponent". Please remove anomalous (\gamma can equal 1 at this stage).

- Line 475: Please replace "section" with "projection".

- Line 476: Please replace "spatiotemporal" with "temporal". There is no spatial dependency in either figure.

- Line 500: Please define Eikonal approximation.

- Fig 8 could be moved to supplements.

- Line 553: "predicted" -> "predict".

- Line 593: Could the authors explain why their model offers much better quantitative agreement?

- Line 669: What does "universal" mean in that context?

- Line 671: A volume can be pipetted but not a concentration.

- Line 676: Are triplicates technical or biological replicates?

- Sup Fig1: Please use minutes instead of seconds in panel A.

- Model for membrane dynamics: "The fraction of time the Q+ channel is open" -> "The dynamics of Q+ channel activity can be written". Ditto for K+ channel...

- Model for membrane dynamics: "the term ... is a threshold-linear". This function is not linear at all. Why is it called linear? Also, please describe what \sigma is.

- ABFDF model: "releasing a given concentration" -> "releasing a local concentration" or "a given number" but it's not \sigma anymore. Besides, this \sigma is unlikely related to the previous \sigma used in the model of membrane potential dynamics in single cells. Please consider renaming one or the other. Also, ions are referred to as C+ in the text and C in equation 8. Am I missing something?

Reviewer #2 (Recommendations For The Authors):

I have included all my comments as one review. I have done so, despite the fact that some minor comments could have gone into this section, because I decided to review each Result section. I thus felt that not writing it as one review might be harder to follow. I have however highlighted which comments are minor suggestions or where I felt corrections.

However, while I am happy with all my comments being public, given their nature I think they should be shown to authors first. Perhaps the authors want to go over them and think about it before deciding if they are happy for their manuscript to be published along with these comments, or not. I will highlight this in an email to the editor. I question whether in this case, given that I am raising major issues, publishing both the manuscript and the comments is the way to go as I think it might just generate confusion among the audience.

Reviewer #3 (Recommendations For The Authors):

I was unable to find any legends for any of the supplemental videos in my review materials, and I could not open supplemental video 5.

I made some comments in the public review about the analysis and interpretation of the time-to-fire data. One of the other challenges in this data set is that the time resolution is limited- it seems that a large proportion of cells have already fired after a single acquisition frame. It would be ideal to increase the time resolution on this measurement to improve precision. This could be done by imaging more quickly, but that would perhaps necessitate more blue light exposure; an alternative is to do this experiment under lower blue light irradiance where the first spike time is increased (Figure 4A).

In the public review, I mentioned the possible impact of high membrane potential on PI permeability. To address this, the experiment could be repeated with other stains, or the viability of blue light-treated cells could be addressed more directly by outgrowth or colony-forming unit assays.

In the public review, I mentioned the possible combined toxicity of ThT and blue light. Live/dead experiments after blue light exposure with and without ThT could be used to test for such effects, and/or the growth curve experiment in Figure 1F could be repeated with blue light exposure at a comparable irradiance used in the experiment.

Throughout the paper and figure legends, it would help to have more methodological details in the main text, especially those that are critical for the interpretation of the experiment. The experimental details in the methods section are nicely described, but the data analysis section should be expanded significantly.

At the end of the results section, the authors suggest a critical biofilm size of only 4 µm for wavefront propagation (not much larger than a single cell!). The authors show responses for various biofilm sizes in Fig. 2C, but these are all substantially larger. Are there data for cell clusters above and below this size that could support this claim more directly?

The authors mention image registration as part of their analysis pipeline, but the 3D data sets in Video S6B and Fig. S4A do not appear to be registered- were these registered prior to the velocity analysis reported in Fig. 8?

One of the most challenging claims to demonstrate in this paper is that these membrane potential wavefronts are involved in coordinating a large, biofilm-scale response to blue light. One possible way to test this might be to repeat the Live/Dead experiment in planktonic culture or the single-cell condition. If the protection from blue light specifically emerges due to coordinated activity of the biofilm, the Kch mutant would not be expected to show a change in Live/Dead staining in non-biofilm conditions.

Line 140: How is "mature biofilm" defined? Also on this same line, what does "spontaneous" mean here?

Line 151: "much smaller": Given that the reported time for 3D biofilms is 2.73 {plus minus} 0.85 min and in microclusters is 3.27 {plus minus} 1.77 min, this seems overly strong.

Line 155: How is "biofilm density" characterized? Additionally, the data in Figure 2C are presented in distance units (µm), but the text refers to "areal coverage"- please define the meaning of these distance units in the legend and/or here in the text (is this the average radius?).

Lines 161-162: These claims seem strong given the data presented before, and the logic is not very explicit. For example, in the second sentence, the idea that this signaling is used to "coordinate long-range responses to light stress" does not seem strongly evidenced at this point in the paper. What is meant by a long-range response to light stress- are there processes to respond to light that occur at long-length scales (rather than on the single-cell scale)? If so, is there evidence that these membrane potential changes could induce these responses? Please clarify the logic behind these conclusions.

Lines 235-236: In the lower irradiance conditions, the responses are slower overall, and it looks like the ThT intensity is beginning to rise at the end of the measurement. Could a more prominent second peak be observed in these cases if the measurement time was extended?

Line 242-243: The overall trajectories of extracellular potassium are indeed similar, but the kinetics of the second peak of potassium are different than those observed by ThT (it rises some minutes earlier)- is this consistent with the idea that Kch is responsible for that peak? Additionally, the potassium dynamics also reflect the first peak- is this surprising given that the Kch channel has no effect on this peak?

Line 255-256: Again, this seems like a very strong claim. There are several possible interpretations of the catalase experiment (which should be discussed); this experiment perhaps suggests that ROS impacts membrane potential, but does not obviously indicate that these membrane potential fluctuations mitigate ROS levels or help the cells respond to ROS stress. The loss of viability in the ∆kch mutant might indicate a link between these membrane potential experiments and viability, but it is hard to interpret without the no-light control I mention in the public review.

Lines 313-315: "The model predicts... the external light stress". Please clarify this section. Where this prediction arises from in the modeling work? Second, I am not sure what is meant by "modulates the light stress" or "keeps the cell dynamics robust to the intensity of external light stress" (especially since the dynamics clearly vary with irradiance, as seen in Figure 4A).

Line 322: I am not sure what "handles the ROS by adjusting the profile of the membrane potential dynamics" means. What is meant by "handling" ROS? Is the hypothesis that membrane potential dynamics themselves are protective against ROS, or that they induce a ROS-protective response downstream, or something else? Later in lines 327-8 the authors write that changes in the response to ROS in the model agree with the hypothesis, but just showing that ROS impacts the membrane potential does not seem to demonstrate that this has a protective effect against ROS.

Line 365-366: This section title seems confusing- mechanosensitive ion channels totally ablate membrane potential dynamics, they don't have a specific effect on the first hyperpolarization event. The claim that mechanonsensitive ion channels are specifically involved in the first event also appears in the abstract.

Also, the apparent membrane potential is much lower even at the start of the experiment in these mutants- is this expected? This seems to imply that these ion channels also have a blue light independent effect.

Lines 368, 371: Should be VGCCs rather than VGGCs.

Line 477: I believe the figure reference here should be to Figure 7B, not 6B.

Line 567-568: "The initial spike is key to registering the presence of the light stress." What is the evidence for this claim?

Line 592-594: "We have presented much better quantitative agreement..." This is a strong claim; it is not immediately evident to me that the agreement between model and prediction is "much better" in this work than in the cited work. The model in Figure 4 of reference 57 seems to capture the key features of their data. Clarification is needed about this claim.

Line 613: "...strains did not have any additional mutations." This seems to imply that whole genome sequencing was performed- is this the case?

Line 627: I believe this should refer to Figure S2A-B rather than S1.

Line 719: What percentage of cells did not hyperpolarize in these experiments?

Lines 751-754: As I mentioned above, significant detail is missing here about how these measurements were made. How is "radius" defined in 3D biofilms like the one shown in Video S6B, which looks very flat? What is meant by the distance from the substrate to the core, since usually in this biofilm geometry, the core is directly on the substrate? Most importantly, this only describes the process of sectioning the data- how were these sections used to compute the velocity of ThT signal propagation?

I also have some comments specifically on the figure presentation:

Normalization from 0 to 1 has been done in some of the ThT traces in the paper, but not all. The claims in the paper would be easiest to evaluate if the non-normalized data were shown- this is important for the interpretation of some of the claims.

Some indication of standard deviation (error bars or shading) should be added to all figures where mean traces are plotted.

Throughout the paper, I am a bit confused by the time axis; the data consistently starts at 1 minute. This is not intuitive to me, because it seems that the blue light being applied to the cells is also the excitation laser for ThT- in that case, shouldn't the first imaging frame be at time 0 (when the blue light is first applied)? Or is there an additional exposure of blue light 1 minute before imaging starts? This is consequential because it impacts the measured time to the first spike. (Additionally, all of the video time stamps start at 0).

Please increase the size of the scale bars and bar labels throughout, especially in Figure 2A and S4A.

In Figure 1B and D, it would help to decrease the opacity on the individual traces so that more of them can be discerned. It would also improve clarity to have data from the different experiments shown with different colored lines, so that variability between experiments can be clearly visualized.

Results in Figure 1E would be easier to interpret if the frequency were normalized to total N. It is hard to tell from this graph whether the edges and bin widths are the same between the data sets, but if not, they should be. Also, it would help to reduce the opacity of the sparse cell data set so that the full microcluster data set can be seen as well.

Biofilm images are shown in Figures 2A, S3A, and Video S3- these are all of the same biofilm. Why not take the opportunity to show different experimental replicates in these different figures? The same goes for Figure S4A and Video S6B, which again are of the same biofilm.

Figure 2C would be much easier to read if the curves were colored in order of their size; the same is true for Figure 4A and irradiance.

The complementation data in Figure S3D should be moved to the main text figure 3 alongside the data about the corresponding knockout to make it easier to compare the curves.

Fig.ure S3E: Is the Y-axis in this graph mislabeled? It is labeled as ThT fluorescence, but it seems that it is reporting fluorescence from the calcium indicator?

Video S6B is very confusing - why does the video play first forwards and then backwards? Unless I am looking very carefully at the time stamps it is easy to misinterpret this as a rise in the intensity at the end of the experiment. Without a video legend, it's hard to understand this, but I think it would be much more straightforward to interpret if it only played forward. (Also, why is this video labeled 6B when there is no video 6A?)

Author response:

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

The points raised let us critically rethink our approach, our results, and our conclusions. Furthermore, it gave us the chance to elaborate on some critical aspects that were mentioned. With the help of the reviewers, we made some clarifications in the point-by-point responses and implemented them in the manuscript. Furthermore, we modified the figures as suggested:

- The colors in Figure 1C, D, G and H have been adapted as suggested

- We added a Figure2-figure supplement 1, which strengthens our conclusion in Figure 2

- As asked by reviewer #1 (weaknesses #3), we added the data about neutrophil numbers in the different organs (Figure 6-figure supplement 3C).

Reviewer #1 (Public Review):

Summary:

- Extracellular ATP represents a danger-associated molecular pattern associated to tissue damage and can act also in an autocrine fashion in macrophages to promote proinflammatory responses, as observed in a previous paper by the authors in abdominal sepsis. The present study addresses an important aspect possibly conditioning the outcome of sepsis that is the release of ATP by bacteria. The authors show that sepsis-associated bacteria do in fact release ATP in a growth dependent and strain-specific manner. However, whether this bacterial derived ATP play a role in the pathogenesis of abdominal sepsis has not been determined. To address this question, a number of mutant strains of E. coli has been used first to correlate bacterial ATP release with growth and then, with outer membrane integrity and bacterial death. By using E. coli transformants expressing the ATP-degrading enzyme apyrase in the periplasmic space, the paper nicely shows that abdominal sepsis by these transformants results in significantly improved survival. This effect was associated with a reduction of peritoneal macrophages and CX3CR1+ monocytes, and an increase in neutrophils. To extrapolate the function of bacterial ATP from the systemic response to microorganisms, the authors exploited bacterial OMVs either loaded or not with ATP to investigate the systemic effects devoid of living microorganisms. This approach showed that ATP-loaded OMVs induced degranulation of neutrophils after lysosomal uptake, suggesting that this mechanism could contribute to sepsis severity.

Strengths:

- A strong part of the study is the analysis of E. coli mutants to address different aspects of bacterial release of ATP that could be relevant during systemic dissemination of bacteria in the host.

We want to thank the reviewer for recognizing this important aspect of our experimental approach.

Weaknesses:

- As pointed out in the limitations of the study whether ATP-loaded OMVs provide a mechanistic proof of the pathogenetic role of bacteria-derived ATP independently of live microorganisms in sepsis is interesting but not definitively convincing. It could be useful to see whether degranulation of neutrophils is differentially induced by apyrase-expressing vs control E. coli transformants.

We thank the reviewer for raising several important points. In our study, we assessed local and systemic effects of released bacterial ATP. The consequences of local bacterial ATP release were assessed using an apyrase-expressing E. coli transformant. Locally, bacterial ATP resulted in a decrease in neutrophil numbers and we hypothesize that directly released bacterial ATP either leads to neutrophil death (e.g. via P2X7 receptor (Proietti et al., 2019)) or interferes with the recruitment of neutrophils (e.g. via P2Y receptors (Junger, 2011)).

The systemic consequences were assessed using ATP-loaded and empty OMV. We have shown that degranulation is induced by OMV-derived bacterial ATP. ATP-containing OMV are engulfed by neutrophils, reach its endolysosomal compartment and might activate purinergic receptors, which then lead to aberrant degranulation. This concept, that needs to be explored in future studies, is fundamentally different from classical purinergic signaling via directly released bacterial ATP into the extracellular space.

It is possible that neutrophil degranulation is also modulated by directly released bacterial ATP. We agree that this should be assessed in future studies. Also, the role of OMV-derived bacterial ATP should be assessed locally as well as the importance of directly released vs. OMV-mediated bacterial ATP dissected locally. Based on our measurements (Figure 4-figure supplement 1A and Figure 5C), we estimate that the effect of OMV-derived bacterial ATP might be much smaller than the effects of directly released bacterial ATP. Thus, direct ATP release might predominate locally. However, we fully agree that this has to be investigated in a future study to reconcile the different aspects of bacterial ATP signaling. A paragraph will be added to the manuscript, in which we discuss this particular issue.

- Also, the increase of neutrophils in bacterial ATP-depleted abdominal sepsis, which has better outcomes than "ATP-proficient" sepsis, seems difficult to correlate to the hypothesized tissue damage induced by ATP delivered via non-infectious OMVs.

We fully acknowledge the mentioned discrepancy. What we propose is that bacterial ATP exhibits different functions that are dependent on the release mechanism (see above). Locally, in the peritoneal cavity, neutrophil numbers are decreased by directly released bacterial ATP. Remotely, ATP is delivered via OMV and impacts on neutrophil function. We agree that, in particular, in the peritoneal cavity, both effects may play a role. However, the impact of directly released bacterial ATP seems to be dominant (see above).

We propose that neutrophils are decreased locally because of directly released bacterial ATP, which prevents efficient infection control and, therefore, impairs sepsis survival. In addition, these fewer neutrophils might even be dysregulated by the engulfment of bacterial ATP delivered via OMV, which leads to an upregulated and possibly aberrant degranulation process worsening local and remote tissue damage. We agree that in addition to neutrophil numbers, the function of local neutrophils should be assessed with and without the influence of OMV-delivered bacterial ATP. This could be done by RNA sequencing of primary neutrophils from the peritoneal cavity or neutrophil cell lines as well as degranulation assays.

- Are the neutrophils counts affected by ATP delivered via OMVs?

This is difficult to show in the peritoneal cavity where we have both, directly released bacterial ATP and OMV-derived bacterial ATP. We assessed such putative difference, however, for the systemic organs and the blood, where we did not find any differences in neutrophil numbers.

Author response image 1.

- A comparison of cytokine profiles in the abdominal fluids of E. coli and OMV treated animals could be helpful in defining the different responses induced by OMV-delivered vs bacterial-released ATP. The analyses performed on OMV treated versus E. coli infected mice are not closely related and difficult to combine when trying to draw a hypothesis for bacterial ATP in sepsis.

We fully agree that there are several open questions that remain to be elucidated, in particular, to differentiate the local role of directly released versus OMV-delivered bacterial ATP. In this study, we laid the foundation for future in vivo research to examine the specific role of bacterial ATP in sepsis. Such future research avenues might be to investigate the local effects of OMV-delivered bacterial ATP, and how neutrophil migration, apoptosis and degranulation are altered. We agree that exploration of the local secretory immune response and cytokine profiles are relevant to understand the different mechanisms of how bacterial ATP alters sepsis. However, such experiments should be ideally performed in systems where the source and the delivery of ATP can be modulated locally.

- Also it was not clear why lung neutrophils were used for the RNAseq data generation and analysis.

Thank you for this remark. We have chosen primary lung neutrophils for four reasons:

(1) Isolation of primary lung neutrophils allowed us to assess an in vivo response that would not have been possible with cell lines.

(2) The lung and the respiratory system are among the clinically most important organs affected during sepsis resulting in a significant cause of mortality.

(3) We show in Figure 6C that specifically in the lung, OMV are engulfed by neutrophils, which shows the relevance of the lung also in our study context.

(4) And finally, lung neutrophils were chosen to examine specifically distant and not local effects.

Reviewer #2 (Public Review):

Summary:

- In their manuscript "Released Bacterial ATP Shapes Local and Systemic Inflammation during Abdominal Sepsis", Daniel Spari et al. explored the dual role of ATP in exacerbating sepsis, revealing that ATP from both host and bacteria significantly impacts immune responses and disease progression.

Strengths:

- The study meticulously examines the complex relationship between ATP release and bacterial growth, membrane integrity, and how bacterial ATP potentially dampens inflammatory responses, thereby impairing survival in sepsis models. Additionally, this compelling paper implies a concept that bacterial OMVs act as vehicles for the systemic distribution of ATP, influencing neutrophil activity and exacerbating sepsis severity.

We thank the reviewer for mentioning these key points and supporting the relevance of our study.

Weaknesses:

(1) The researchers extracted and cultivated abdominal fluid on LB agar plates, then randomly picked 25 colonies for analysis. However, they did not conduct 16S rRNA gene amplicon sequencing on the fluid itself. It is worth noting that the bacterial species present may vary depending on the individual patients. It would be beneficial if the authors could specify whether they've verified the existence of unculturable species capable of secreting high levels of Extracellular ATP.

Most septic complications are caused by a limited spectrum of bacteria, belonging mainly either to the Firmicutes or the Proteobacteria phyla, including E. coli, K. pneumoniae, S. aureus or E. faecalis (Diekema et al., 2019; Mureșan et al., 2018). We validated this well documented existing evidence by randomly assessing 25 colonies. For the planned experiments, it was crucial to work with culturable bacteria; otherwise, ATP measurements, the modulation of ATP generation or loading of OMV would not have been possible. Using such culturable bacteria allowed us to describe mechanisms of ATP release.

We fully agree that hard-to-culture or unculturable bacteria might contribute significantly to septic complications. This, however, would need to be explored in future studies using extensive culturing methods (Cheng et al., 2022).

(2) Do mice lacking commensal bacteria show a lack of extracellular ATP following cecal ligation puncture?

ATP is typically secreted by many cells of the host in active and passive manners in the case of any injury, including cecal ligation and puncture (Burnstock, 2016; Dosch et al., 2018; Eltzschig et al., 2012; Idzko et al., 2014). We hypothesize that bacterial ATP is a potential priming agent at early stages of sepsis, and indeed, at such early time points, a comparison of peritoneal ATP levels between germfree and colonized mice could support our hypothesis. Future studies addressing this question must, however, correct for the different immune responses between germ-free and colonized mice. This is of utmost importance, especially for the cecal ligation and puncture model, since the cecum of germ-free mice is extremely large, making such experiments hard to control.

(3) The authors isolated various bacteria from abdominal fluid, encompassing both Gram-negative and Gram-positive types. Nevertheless, their emphasis appeared to be primarily on the Gram-negative E. coli. It would be beneficial to ascertain whether the mechanisms of Extracellular ATP release differ between Gram-positive and Gram-negative bacteria. This is particularly relevant given that the Gram-positive bacterium E. faecalis, also isolated from the abdominal fluid, is recognized for its propensity to release substantial amounts of Extracellular ATP.

We fully agree with this comment. In this paper, we used E. coli as our model organism to determine the principles of sepsis-associated bacterial ATP release and therefore focused on gram-negative bacteria. In addition to the direct, growth-dependent release, we found a relevant impact of OMV-delivered bacterial ATP. For this latter purpose, a gram-negative strain, in which OMV generation has been well described (Schwechheimer & Kuehn, 2015), was chosen. Recently, gram-positive bacteria have been shown to secrete ATP and OMV as well (Briaud & Carroll, 2020; Hironaka et al., 2013; Iwase et al., 2010). Given the fundamental differences in the structure of the cell wall of gram-positive bacteria and the mechanisms of OMV generation and release, future studies are required to assess the relevance of directly released and OMV-delivered ATP in gram-positive bacteria.

(4) The authors observed changes in the levels of LPM, SPM, and neutrophils in vivo. However, it remains uncertain whether the proliferation or migration of these cells is modulated or inhibited by ATP receptors like P2Y receptors. This aspect requires further investigation to establish a convincing connection.

We fully agree with this comment. The decrease in LPM and the consequential predomination of SPM have been well described after inflammatory stimuli in the context of the macrophage disappearance reaction (Ghosn et al., 2010). Also, it has been shown that purinergic signaling modulates infiltration of neutrophils and can lead to cell death as a consequence of  P2Y and P2X receptor activation (Junger, 2011; Proietti et al., 2019). In our study, we propose that intracellular purinergic receptors contribute to neutrophil function during sepsis. After introducing the general principles and fundaments of bacterial ATP with our studies, we fully agree that additional experiments need to address downstream purinergic receptor activation. That, however, would go beyond the scope of our study.

(5) Additionally, is it possible that the observed in vivo changes could be triggered by bacterial components other than Extracellular ATP? In this research field, a comprehensive collection of inhibitors is available, so it is desirable to utilize them to demonstrate clearer results.

This question is of utmost importance and defined the choice of our model and experimental approach. When we started the project, we used two different E. coli mutants that release low (ompC) and high (eaeH) amounts of ATP. However, the limitation of this approach is that these are different bacteria, which may also differ in the components they secrete or the surface proteins they express. We, therefore, decided against that approach. With the approach we finally used (same bacterium, just with and without ATP), we aimed to minimize the influence of non-ATP bacterial components.

(6) Have the authors considered the role of host-derived Extracellular ATP in the context of inflammation?

Yes, the role of host-derived extracellular ATP in inflammation and sepsis is well-established with contradictory results (Csóka et al., 2015; Ledderose et al., 2016). This conflicting data was the rationale to test the relevance of bacterial ATP. We suggest that bacterial ATP is essential in the early phase of sepsis when bacteria invade the sterile compartment and before efficient host response, including the eukaryotic release of ATP, is established.

(7) The authors mention that Extracellular ATP is rapidly hydrolyzed by ectonucleotases in vivo. Are the changes of immune cells within the peritoneal cavity caused by Extracellular ATP released from bacterial death or by OMVs?

This is a relevant question that was also asked by reviewer #1, and we answered it in detail above (weaknesses comment #1 and #2). From our ATP measurements (Figure 4-figure supplement 1A and Figure 5C), we conclude that locally, the role of directly released bacterial ATP (extracellular) predominates over OMV-derived bacterial ATP. Furthermore, the mechanisms between directly released and OMV-derived bacterial ATP (within OMV, engulfed and transported to the endolysosomal compartment) are different, and especially extracellular ATP has been described to lead to apoptosis via P2X7 signaling.

(8) In the manuscript, the sample size (n) for the data consistently remains at 2. I would suggest expanding the sample size to enhance the robustness and rigor of the results.

Two biological replicates (independent cultures) were only used for the bacteria cultures in Figure 1, Figure 2, and Figure 3, which achieved similar results and the standard deviation remained very small, indicating its robustness. In the in vitro experiments in Figure 5 we used a sample size of 6 (three biological replicates measured in technical duplicates), since we saw bigger deviations in our measurements. For the in vivo experiments, we always used 5 or more animals in at least two independent experiments.

Reviewer #2 (Recommendations For The Authors):

(9). Line 37: 11 million sepsis-related deaths were reported "in" 2017.

The passage has been corrected as suggested.

(10) By the way, the similar colors used in Figure 1C and G are too chaotic, making it difficult to distinguish.

We agree, the colors have been adapted.

Author response image 2.

(11). All "in vivo" and "in vitro" should be italicized.

We italicized all of them.

(12). The title of Figure 4 is confusing: "Impairs sepsis outcome in vivo?" Could you make it more specific?

We agree, the title has been rephrased:

“Bacterial ATP reduces neutrophil counts and reduces survival in a mouse model of abdominal sepsis.”

(13) Line 314-316: The sentence "Potentially, despite the lack of a transporter, ATP may similarly to eukaryotic cells leak (Yegutkin et al., 2006) across the inner membrane into the periplasmic space that lacks the enzymes for ATP generation." sounds odd.

This passage was reformulated in the manuscript.

“Despite the lack of a transporter, ATP may leak across the inner membrane into the periplasmic space. Such leakage may be similar to baseline leakage in eukaryotic cells (Yegutkin et al., 2006).”

(14) The numerical notation in the paper is odd: sometimes it uses a prime symbol as a superscript (such as line 504), and sometimes it does not (such as line 421). Should it be standardized to "3,200" and "150,000"?

Thank you for this remark. The numbers have been standardized throughout the manuscript.

(15) Line "0.4 mm EP cuvettes" should be "0.4 cm EP cuvettes"

The specified passage has been corrected as suggested.

References

Briaud, P., & Carroll, R. K. (2020). Extracellular Vesicle Biogenesis and Functions in Gram-Positive Bacteria. Infection and Immunity, 88(12), 10.1128/iai.00433-20. https://doi.org/10.1128/iai.00433-20

Burnstock, G. (2016). P2X ion channel receptors and inflammation. Purinergic Signalling, 12(1), 59–67. https://doi.org/10.1007/s11302-015-9493-0

Cheng, A. G., Ho, P.-Y., Aranda-Díaz, A., Jain, S., Yu, F. B., Meng, X., Wang, M., Iakiviak, M., Nagashima, K., Zhao, A., Murugkar, P., Patil, A., Atabakhsh, K., Weakley, A., Yan, J., Brumbaugh, A. R., Higginbottom, S., Dimas, A., Shiver, A. L., … Fischbach, M. A. (2022). Design, construction, and in vivo augmentation of a complex gut microbiome. Cell, 185(19), 3617-3636.e19. https://doi.org/10.1016/j.cell.2022.08.003

Csóka, B., Németh, Z. H., Törő, G., Idzko, M., Zech, A., Koscsó, B., Spolarics, Z., Antonioli, L., Cseri, K., Erdélyi, K., Pacher, P., & Haskó, G. (2015). Extracellular ATP protects against sepsis through macrophage P2X7 purinergic receptors by enhancing intracellular bacterial killing. The FASEB Journal, 29(9), 3626–3637. https://doi.org/10.1096/fj.15-272450

Diekema, D. J., Hsueh, P.-R., Mendes, R. E., Pfaller, M. A., Rolston, K. V., Sader, H. S., & Jones, R. N. (2019). The Microbiology of Bloodstream Infection: 20-Year Trends from the SENTRY Antimicrobial Surveillance Program. Antimicrobial Agents and Chemotherapy, 63(7), e00355-19. https://doi.org/10.1128/AAC.00355-19

Dosch, M., Gerber, J., Jebbawi, F., & Beldi, G. (2018). Mechanisms of ATP Release by Inflammatory Cells. International Journal of Molecular Sciences, 19(4), 1222. https://doi.org/10.3390/ijms19041222

Eltzschig, H. K., Sitkovsky, M. V., & Robson, S. C. (2012). Purinergic Signaling during Inflammation. New England Journal of Medicine, 367(24), 2322–2333. https://doi.org/10.1056/NEJMra1205750

Ghosn, E. E. B., Cassado, A. A., Govoni, G. R., Fukuhara, T., Yang, Y., Monack, D. M., Bortoluci, K. R., Almeida, S. R., Herzenberg, L. A., & Herzenberg, L. A. (2010). Two physically, functionally, and developmentally distinct peritoneal macrophage subsets. Proceedings of the National Academy of Sciences, 107(6), 2568–2573. https://doi.org/10.1073/pnas.0915000107

Hironaka, I., Iwase, T., Sugimoto, S., Okuda, K., Tajima, A., Yanaga, K., & Mizunoe, Y. (2013). Glucose Triggers ATP Secretion from Bacteria in a Growth-Phase-Dependent Manner. Applied and Environmental Microbiology, 79(7), 2328–2335. https://doi.org/10.1128/AEM.03871-12

Idzko, M., Ferrari, D., & Eltzschig, H. K. (2014). Nucleotide signalling during inflammation. Nature, 509(7500), 310–317. https://doi.org/10.1038/nature13085

Iwase, T., Shinji, H., Tajima, A., Sato, F., Tamura, T., Iwamoto, T., Yoneda, M., & Mizunoe, Y. (2010). Isolation and Identification of ATP-Secreting Bacteria from Mice and Humans. Journal of Clinical Microbiology, 48(5), 1949–1951. https://doi.org/10.1128/JCM.01941-09

Junger, W. G. (2011). Immune cell regulation by autocrine purinergic signalling. Nature Reviews Immunology, 11(3), 201–212. https://doi.org/10.1038/nri2938

Ledderose, C., Bao, Y., Kondo, Y., Fakhari, M., Slubowski, C., Zhang, J., & Junger, W. G. (2016). Purinergic Signaling and the Immune Response in Sepsis: A Review. Clinical Therapeutics, 38(5), 1054–1065. https://doi.org/10.1016/j.clinthera.2016.04.002

Mureșan, M. G., Balmoș, I. A., Badea, I., & Santini, A. (2018). Abdominal Sepsis: An Update. The Journal of Critical Care Medicine, 4(4), 120–125. https://doi.org/10.2478/jccm-2018-0023

Proietti, M., Perruzza, L., Scribano, D., Pellegrini, G., D’Antuono, R., Strati, F., Raffaelli, M., Gonzalez, S. F., Thelen, M., Hardt, W.-D., Slack, E., Nicoletti, M., & Grassi, F. (2019). ATP released by intestinal bacteria limits the generation of protective IgA against enteropathogens. Nature Communications, 10(1), Article 1. https://doi.org/10.1038/s41467-018-08156-z

Schwechheimer, C., & Kuehn, M. J. (2015). Outer-membrane vesicles from Gram-negative bacteria: Biogenesis and functions. Nature Reviews Microbiology, 13(10), 605–619. https://doi.org/10.1038/nrmicro3525

Author response:

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

Reviewer #1 (Recommendations For The Authors):

Major concerns:

(1) It is not clear about the biological significance of the inhibitory effects of human Abeta42 on gammasecretase activity. As the authors mentioned in the Discussion, it is plausible that Abeta42 may concentrate up to microM level in endosomes. However, subsets of FAD mutations in APP and presenilin 1 and 2 increase Abeta42/Abeta40 ratio and lead to Abeta42 deposition in brain. APP knock-in mice NLF and NLGF also develop Abeta42 deposition in age-dependent manner, although they produce more human Abeta42 than human Abeta40. 

If the production of Abeta42 is attenuated, which results in less Abeta42 deposition in brain. So, it is unlikely that human Abeta42 interferes gamma-secretase activity in physiological conditions. This reviewer has an impression that inhibition of gamma-secretase by human Abeta42 is an interesting artifact in high Abeta42 concentration. If the authors disagree with this reviewer's comment, this manuscript needs more discussion in this point of view. 

We thank the Reviewer for raising this key conceptual point, we acknowledge that it was insufficiently discussed in the original manuscript. In response to this point, we introduced the following paragraph in the discussion section of the revised manuscript:

“From a mechanistic standpoint, the competitive nature of the Aβ42-mediated inhibition implies

that it is partial, reversible, and regulated by the relative concentrations of the Aβ42 peptide (inhibitor) and the endogenous substrates (Figure 10C and 10D). The model that we put forward is that cellular uptake, as well as endosomal production of Aβ, result in increased intracellular concentration of Aβ42, facilitating γ-secretase inhibition and leading to the buildup of APP-CTFs (and γ-secretase substrates in general). As Aβ42 levels fall, the augmented concentration of substrates shifts the equilibrium towards their processing and subsequent Aβ production. As Aβ42 levels rise again, the equilibrium is shifted back towards inhibition. This cyclic inhibitory mechanism will translate into pulses of (partial) γsecretase inhibition, which will alter γ-secretase mediated-signaling (arising from increased CTF levels at the membrane or decreased release of soluble intracellular domains from substrates). These alterations may affect the dynamics of systems oscillating in the brain, such as NOTCH signaling, implicated in memory formation, and potentially others (related to e.g. cadherins, p75 or neuregulins). It is worth noting that oscillations in γ-secretase activity induced by treatment with a γ-secretase inhibitor semagacestat have been proposed to have contributed to the cognitive alterations observed in semagacestat treated patients in the failed Phase-3 IDENTITY clinical trial (7) and that semagacestat, like Aβ42, acts as a high affinity competitor of substrates (85).

The convergence of Aβ42 and tau at the synapse has been proposed to underlie synaptic dysfunction in AD (86-89), and recent assessment of APP-CTF levels in synaptosome-enriched fractions from healthy control, SAD and FAD brains (temporal cortices) has shown that APP fragments concentrate at higher levels in the synapse in AD-affected than in control individuals (90).  Our analysis adds that endogenous Aβ42 concentrates in synaptosomes derived from end-stage AD brains to reach ~10 nM, a concentration that in CM from human neurons inhibits γ-secretase in PC12 cells (Figure 7). Furthermore, the restricted localization of Aβ in endolysosomal vesicles, within synaptosomes, likely increases the local peptide concentration to the levels that inhibit γ-secretase-mediated processing of substrates in this compartment. In addition, we argue that the deposition of Aβ42 in plaques may be preceded a critical increase in the levels of Aβ present in endosomes and the cyclical inhibition of γsecretase activity that we propose. Under this view, reductions in γ-secretase activity may be a (transient) downstream consequence of increases in Aβ due to failed clearance, as represented by plaque deposition, contributing to AD pathogenesis.“

We have also added figures 10C and 10D, presented here for convenience.

Author response image 1.

(2) It is not clear whether the FRET-based assay in living cells really reflects gamma-secretase activity.

This reviewer thinks that the authors need at least biochemical data, such as levels of Abeta. 

We have established a novel, HiBiT tag based assay reporting on the global γ-secretase activity in cells, using as a proxy the total levels of secreted HiBiT-tagged Aβ peptides. The assay and findings are presented in the revised manuscript as follows:

In the result section, in the “Aβ42 treatment leads to the accumulation of APP C-terminal fragments in neuronal cell lines and human neuron” subsection:

“The increments in the APP-CTF/FL ratio suggested that Aβ42 (partially) inhibits the global γ-

secretase activity. To further investigate this, we measured the direct products of the γ-secretase mediated proteolysis of APP. Since the detection of the endogenous Aβ products via standard ELISA methods was precluded by the presence of exogenous human Aβ42 (treatment), we used an N-terminally tagged version of APPC99 and quantified the amount of total secreted Aβ, which is a proxy for the global γsecretase activity. Briefly, we overexpressed human APPC99 N-terminally tagged with a short 11 amino acid long HiBiT tag in human embryonic kidney (HEK) cells, treated these cultures with human Aβ42 or p3 17-42 peptides at 1 μM or DAPT (GSI) at 10 µM, and determined total HiBiT-Aβ levels in conditioned media (CM). DAPT was considered to result in full γ-secretase inhibition, and hence the values recorded in DAPT treated conditions were used for the background subtraction. We found a ~50% reduction in luminescence signal, directly linked to HiBiT-Aβ levels, in CM of cells treated with human Aβ42 and no effect of p3 peptide treatment, relative to the DMSO control (Figure 3D). The observed reduction in the total Aβ products is consistent with the partial inhibition of γ -secretase by Aβ42.”

In Methods:

“Analysis of γ-secretase substrate proteolysis in cultured cells using secreted HiBiT-Aβ or -Aβ-like peptide levels as a proxy for the global γ-secretase endopeptidase activity

HEK293 stably expressing APP-CTF (C99) or a NOTCH1-based substrate (similar in size as

APP- C99) both N-terminally tagged with the HiBiT tag were plated at the density of 10000 cells per 96-well, and 24h after plating treated with Aβ or p3 peptides diluted in OPTIMEM (Thermo Fisher Scientific) supplemented with 5% FBS (Gibco). Conditioned media was collected and subjected to analysis using Nano-Glo® HiBiT Extracellular Detection System (Promega). Briefly, 50 µl of the medium was mixed with 50 µl of the reaction mixture containing LgBiT Protein (1:100) and Nano-Glo HiBiT Extracellular Substrate (1:50) in Nano-Glo HiBiT Extracellular Buffer, and the reaction was incubated for 10 minutes at room temperature. Luminescence signal corresponding to the amount of the extracellular HiBiT-Aβ or -Aβ-like peptides was measured using victor plate reader with default luminescence measurement settings.”

As the direct substrate of γ -secretase was used in this analysis, the observed reduction (~50%) in the levels of N-terminally-tagged (HiBiT) Aβ peptides in the presence of 1 µM Aβ42, relative to control conditions, demonstrates a selective inhibition of γ-secretase by Aβ42 (not by the p3). These data complement the FRET-based findings presented in Figure 5.

(3) Processing of APP-CTF in living cells is not only the cleavage by gamma-secretase. This reviewer thinks that the authors need at least biochemical data, such as levels of Abeta in Figures 4, 5 and 7.

We tried to measure the levels of Aβ peptides secreted by cells into the culture medium directly by ELISA (using different protocols) or MS (using established methods, as reported in Koch et al, 2023), but exogenous Aβ42 (treatment) present at relatively high levels interfered with the readout and rendered the analysis inconclusive. 

However, we were successful in the determination of total secreted (HiBiT-tagged) Aβ peptides from the HiBiT tagged APP-C99 substrate, as indicated in the previous point. The quantification of the levels of these peptides showed that Aβ42 treatment resulted in ~50% reduction in the γ -secretase mediated processing of the tagged substrate.    

In addition, we would like to highlight that our analysis of the contribution of other APP-CTF degradation pathways, using cycloheximide-based assays in the constant presence of γ-secretase inhibitor, failed to reveal significant differences between Aβ42 treated cells and controls (Figure 6B & C). The lack of a significant impact of Aβ42 on the half-life of APP-CTFs under the conditions of γsecretase inhibition maintained by inhibitor treatment is consistent with the proposed Aβ42-mediated inhibitory mechanism.

(4) Similar to comment #3. Processing of Pancad-CTF and p75 in living cells may be not only the cleavage by gamma-secretase. This reviewer thinks that the authors need at least biochemical data, such as levels of ICDs in Figures 6C and E. 

To address this comment we have now performed additional experiments where we measured Nterminal Aβ-like peptides derived from NOTCH1-based substrate using the HiBiT-based assay. These experiments showed a reduction in the aforementioned peptides in the cells treated with Aβ42 relative to the vehicle control, and hence further confirmed the inhibitory action of Aβ42. These new data have been included as Figure 8D in the revised manuscript and described as follow:

Finally, we measured the direct N-terminal products generated by γ-secretase proteolysis from a HiBiT-tagged NOTCH1-based substrate, an estimate of the global γ-secretase activity. We quantified the Aβ-like peptides secreted by HEK 293 cells stably expressing this HiBiT-tagged substrate upon treatment with 1 µM Aβ1-42,  p3 17-42 peptide or  DAPT (GSI) (Figure 8D). DAPT treatment was considered to result in a complete γ-secretase inhibition, and hence the values recorded in the DAPT condition were used for background subtraction. A ~20% significant reduction in the amount of secreted

N-terminal HiBiT-tagged peptides derived from the NOTCH1-based substrates in cells treated with Aβ1-

42 supports the inhibitory action of Aβ1-42 on γ-secretase mediated proteolysis.

Minor concerns:

(1) Murine Abeta42 may be converted to murine Abeta38 easily, compared to human Abeta42. This may be a reason why murine Abeta42 exhibits no inhibitory effect on gamma-secretase activity. 

In order to address this question, we performed additional experiments where we assessed the processing of murine Aβ42 into Aβ38. Analogous to human Aβ42, the murine Aβ42 peptide was not processed to Aβ38 in the assay conditions. These new data have been integrated in the manuscript and added as a Supplementary figure 1B.

(2) It is curious to know the levels of C99 and C83 in cells in supplementary figure 3.  

The conditions used in these assays were analogous to the conditions used in the figure 3 (i.e. treatment with Aβ peptides at 1 µM concentrations). Such conditions were associated with profound and consistent APP-CTF accumulation in this model system.

Reviewer #2 (Recommendations For The Authors):

In the current study, the authors show that Aβs with low affinity for γ-secretase, but when present at relatively high concentrations, can compete with the longer, higher affinity APPC99 substrate for binding and processing. They also performed kinetic analyses and demonstrate that human Aβ1-42 inhibits γ-secretase-mediated processing of APP C99 and other substrates. Interestingly, neither murine Aβ1-42 nor human p3 (17-42 amino acids in Aβ) peptides exerted inhibition under similar conditions. The authors also show that human Aβ1-42-mediated inhibition of γ-secretase activity results in the accumulation of unprocessed, which leads to p75-dependent activation of caspase 3 in basal forebrain cholinergic neurons (BFCNs) and PC12 cells. 

These analyses demonstrate that, as seen for γ-secretase inhibitors, Aβ1-42 potentiates this marker of apoptosis. However, these are no any in vivo data to support the physiological significance of the current finding. The author should show in APP KO mice whether gamma-secretase enzymatic activity is elevated or not, and putting back Aβ42 peptide will abolish these in vivo effects. 

The findings presented in this manuscript form the basis for further in vitro and in vivo research to investigate the mechanisms of inhibition and its contribution to brain pathophysiology. Here, we used well-controlled model systems to investigate a novel mechanism of Aβ42 toxicity. Multiple mechanisms regulate the local concentration of Aβ42 in vivo, making the dissection of the biochemical mechanisms of the inhibition more complex. Nevertheless, beyond the scope of this report, we consider these very reasonable comments as a motivation for further research activities. 

The experimental concentrations for Aβ42 peptide in the assay are too high, which are far beyond the physiological concentrations or pathological levels. The artificial observations are not supported by any in vivo experimental evidence.

It is correct that in the majority of the experiments we used low μM concentrations of Aβ42. However, we would like to note that we have also performed experiments where conditioned medium collected from human APP.Swe expressing neurons was used as a source of Aβ. In these experiments total Aβ concentration was in low nM range (0.5-1 nM) (Figure 7). Treatment with this conditioned medium  led to the increase APP-CTF levels, supporting  that low nM concentrations of Aβ are sufficient for partial inhibition of  γ-secretase. 

In addition, we highlight that analyses of the brains of the AD affected individuals have shown that APPCTFs accumulate in both sporadic and genetic forms of the disease (Pera et al. 2013, Vaillant-Beuchot et al. 2021); and recently, Ferrer-Raventós et al. 2023 have revealed a correlation between APP-CTFs and Aβ levels at the synapse (Ferrer-Raventós et al. 2023). We therefore assessed the concentration of Aβ42 in synaptosomes derived from frontal cortices of post-mortem AD and age-matched non-demented (ND) control individuals. Our findings and conclusions are included in the revised version as follows: 

In the results section:

“We next investigated the levels of Aβ42 in synaptosomes derived from frontal cortices of post-mortem AD and age-matched non-demented (ND) control individuals (Figure 10B). Towards this, we prepared synaptosomes from frozen brain tissues using Percoll gradient procedure (62, 63). Intact synaptosomes were spun to obtain a pellet which was resuspended in minimum amount of PBS, allowing us to estimate the volume containing the resuspended synaptosome sample. This is likely an overestimate of the actual synaptosome volume. Finally, synaptosomes were lysed in RIPA buffer and Aβ peptide concentrations measured using ELISA (MSD). We observed that the concentration of Aβ42 in the synaptosomes from (end-stage) AD tissues was significantly higher (10.7 nM)  than those isolated from non-demented tissues (0.7 nM), p<0.0005***. These data provide evidence for accumulation at nM concentrations of endogenous Aβ42 in synaptosomes in end-stage AD brains. Given that we measured Aβ42 concentration in synaptosomes, we speculate that even higher concentrations of this peptide may be present in the endolysosome vesicle system, and therein inhibit the endogenous processing of APP-CTF at the synapse. Of note treatment of PC12 cells with conditioned medium containing even lower amounts of Aβ (low nanomolar range (0.5-1 nM)) resulted in the accumulation of APP-CTFs.” 

In the discussion: 

“The convergence of Aβ42 and tau at the synapse has been proposed to underlie synaptic dysfunction in AD (86-89), and recent assessment of APP-CTF levels in synaptosome-enriched fractions from healthy control, SAD and FAD brains (temporal cortices) has shown that APP fragments concentrate at higher levels in the synapse in AD-affected than in control individuals (90).  Our analysis adds that endogenous Aβ42 concentrates in synaptosomes derived from end-stage AD brains to reach ~10 nM, a concentration that in CM from human neurons inhibits γ-secretase in PC12 cells (Figure 7). Furthermore, the restricted localization of Aβ in endolysosomal vesicles, within synaptosomes, likely increases the local peptide concentration to the levels that inhibit γ-secretase-mediated processing of substrates in this compartment. In addition, we argue that the deposition of Aβ42 in plaques may be preceded by a critical increase in the levels of Aβ present in endosomes and the cyclical inhibition of γ-secretase activity that we propose. Under this view, reductions in γ-secretase activity may be a (transient) downstream consequence of increases in Aβ due to failed clearance, as represented by plaque deposition, contributing to AD pathogenesis. ”

Author Response

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

Reviewer #1 (Public Review):

In 2019, Wilkinson and colleagues (PMID: 31142833) managed to break the veil in a 20-year open question on how to properly culture and expand Hematopoietic Stem Cells (HSCs). Although this study is revolutionizing the HSC biology field, several questions regarding the mechanisms of expansion remain open. Leveraging on this gap, Zhang et al.; embarked on a much-needed investigation regarding HSC self-renewal in this particular culturing setting.

The authors firstly tacked the known caveat that some HSC membrane markers are altered during in vitro cultures by functionally establishing EPCR (CD201) as a reliable and stable HSC marker (Figure 1), demonstrating that this compartment is also responsible for long-term hematopoietic reconstitution (Figure 3). Next in Figure 2, the authors performed single-cell omics to shed light on the potential mechanisms involved in HSC maintenance, and interestingly it was shown that several hematopoietic populations like monocytes and neutrophils are also present in this culture conditions, which has not been reported. The study goes on to functionally characterize these cultured HSCs (cHSC). The authors elegantly demonstrate using state-of-the-art barcoding strategies that these culturing conditions provoke heterogeneity in the expanding HSC pool (Figure 4). In the last experiment (Figure 5), it was demonstrated that cHSC not only retain their high EPCR expression levels but upon transplantation, these cells remain more quiescent than freshly-isolated controls.

Taken together, this study independently validates that the proposed culturing system works and provides new insights into the mechanisms whereby HSC expansion takes place.

Most of the conclusions of this study are well supported by the present manuscript, some aspects regarding experimental design and especially the data analysis should be clarified and possibly extended.

1) The first major point regards the single-cell (sc) omics performed on whole cultured cells (Figure 2):

a. The authors claim that both RNA and ATAC were performed and indeed some ATAC-seq data is shown in Figure 2B, but this collected data seems to be highly underused.

We appreciate the opportunity to clarify our analytical approach and the rationale behind it. In our study, we employed a novel deep learning framework, SAILERX, for our analysis. This framework is specifically designed to integrate multimodal data, such as RNAseq and ATACseq. The advantage of SAILERX lies in its ability to correct for technical noise inherent in sequencing processes and to align information from different modalities. Unlike methods that force a hard alignment of modalities into a shared latent space, SAILERX allows for a more refined integration. It achieves this by encouraging the local structures of the two modalities, as measured by pairwise similarities.

To put it more simply, SAILERX combines RNAseq and ATACseq data, ensuring that the unique characteristics of each data type are respected and used to enhance the overall biological picture, rather than forcing them into a uniform framework.

While it is indeed possible to analyze the ATAC-seq and RNA-seq modalities separately, and we acknowledge the potential value in such an approach, our primary objective in this study was to highlight the relatively low content of HSCs in cultures. This finding is a key point of our work, and the multiome data support this from a molecular point of view.

The Seurat object we provide was created to facilitate further analysis by interested researchers. This object simplifies the exploration of both the ATAC-seq and RNA-seq data, allowing for additional investigations that may be of interest to the scientific community. We hope this explanation clarifies our methodology and its implications.

b. It's not entirely clear to this reviewer the nature of the so-called "HSC signatures"(SF2C) and why exactly these genes were selected. There are genes such as Mpl and Angpt1 which are used for Mk-biased HSCs. Maybe relying on other HSC molecular signatures (PMID: 12228721, for example) would not only bring this study more into the current field context but would also have a more favorable analysis outcome. Moreover reclustering based on a different signature can also clarify the emergence of relevant HSC clusters.

In our study, the selection of the HSC signature in our work was based on well-referenced datasets on well-defined HSPCs, as detailed in the "v. HSC signature" section of our methods. This signature was projected also to another single-cell RNA sequencing dataset generated from ex vivo expanded HSC culture (PMID: 35971894, see Author response image 1 below), demonstrating again an association primarily to the most primitive cells (at least based on gene expression).

Author response image 1.

Projection of "our" HSC signature on scRNAseq data from independent work.

In further response to the suggestion here, we have also examined the molecular signature of HSCs referenced in PMID: 12228721 but also of another HSC signature from PMID: 26004780 in our data (Author response image 2). While these signatures do indeed enrich for cells that fall in the cluster of molecularly defined HSCs, our analysis indicates that neither of them significantly improves the identification of HSCs in our dataset compared to the signature we originally used. This finding reinforces our confidence in the appropriateness of our chosen HSC signature for this study.

Author response image 2.

Projection of alternative HSC signatures onto the SAILERX UMAP.

Regarding the specific genes Mpl and Angpt1, we respectfully oppose the view that these genes are exclusively associated with MK-biased HSCs. There is substantial evidence supporting the broader role of Mpl in regulating HSCs, regardless of any particular "lineage bias". Similarly, while Angpt1 has been less extensively studied, its role in HSCs, as examined in PMID: 25821987, suggests a more general association with HSCs rather than a specific impact on MKs. Therefore, we maintain that it is more accurate to consider these genes as HSC-associated rather than restricted to MK-biased HSCs.

Finally, addressing the comment on reclustering based on different signatures, we would like to clarify that the clustering process is independent of the projection of signatures. The clustering aims to identify cell populations based on their overall molecular profiles, and while signatures can aid in characterizing these populations, they do not influence the clustering process itself.

c. The authors took the hard road to perform experiments with the elegant HSC-specific Fgd5-reporter, and they claim in lines 170-171 that it "failed to clearly demarcate in our single-cell multimodal data". This seems like a rather vague statement and leads to the idea that the scRNA-seq experiment is not reliable. It would be interesting to show a UMAP with this gene expression regardless and also potentially some other HSC markers.

We understand the concerns raised about our statement on the performance of the Fgd5-reporter in our multimodal data analysis. Our aim was not to suggest that single-cell molecular data are unreliable. Instead, we intended to point out specific challenges associated with scRNA sequencing, notably the high rates of dropout. Regarding the specific example of Fgd5, it appears this transcript is not efficiently captured by 10x technology. Our previous 10x scRNA-seq experiments on cells from the Fgd5 reporter strain (Säwén et al., eLife 2018; Konturek-Ciesla et al., Cell Rep. 2023) support this observation. Despite cells being sorted as Fgd5-reporter positive, many showed no detectable transcripts.

We consider it pertinent to note that our study integrates ATAC-seq data in conjunction with single-cell molecular data. We believe that this integration, coupled with the analytical methods we have employed, potentially offers a way to address some of the limitations typically associated with scRNA sequencing. However, in assessing frequencies, we observe that the number of candidate HSCs identified via single-cell molecular data is substantially higher compared to those identified through flow cytometry, the latter which we demonstrate correlate functionally with genuine long-term repopulating activity.

With respect to Fgd5, as depicted in our analysis below, there appears to be an enrichment of cells in the cluster identified as HSCs, as well as a significant representation in the cycling cell cluster (Author response image 3). Regarding the projection of other individual genes, the Seurat object we have provided allows for such projections to be readily performed. This offers an opportunity for further exploration and validation of our findings by interested researchers.

Author response image 3.

Feature plot depicting Fgd5 expression in the SAILERX UMAP.

2) During the discussion and in Figure 4, the authors ponder and demonstrate that this culturing system can provoke divert HSC close expansion, having also functional consequences. This a known caveat from the original system, but in more recent publications from the original group (PMID: 36809781 and PMID: 37385251) small alterations into the protocol seem to alleviate clone selection. It's intriguing why the authors have not included these parameters at least in some experiments to show reproducibility or why these studies are not mentioned during the discussion section.

Thank you for pointing out the recent publications (PMID: 36809781 and PMID: 37385251) that discuss modifications to the HSC culturing system. We appreciate the opportunity to address why these were not included in our discussion or experiments.

Firstly, it is important to note that these papers were published after the submission of our manuscript. In fact, one of the studies (PMID: 36809781) references the preprint version of our work on Biorxiv. This timing meant that we were unable to consider these studies in our initial manuscript or incorporate any of their findings into our experimental designs.

Furthermore, as strong advocates for the peer-review system, we prioritize references that have undergone this rigorous process. Preprints, while valuable for early dissemination of research findings, do not offer the same level of scrutiny and validation as peer-reviewed publications. Our approach was to rely on the most relevant and rigorously reviewed literature available to us at the time of submission. This included, most notably, the original and ground-breaking work by Wilkinson et al., which provided a foundational basis for our research.

We acknowledge that the field of HSC research is rapidly evolving, and new findings, such as those mentioned, are continually emerging. These new studies undoubtedly contribute valuable insights into HSC culturing systems and their optimization. However, given the timing of their publication relative to our study, we were not able to include them in our analysis or discussion.

3) In this reviewer's opinion, the finding that transplanted cHSC are more quiescent than freshly isolated controls is the most remarkable aspect of this manuscript. There is a point of concern and an intriguing thought that sprouts from this experiment. It is empirical that for this experiment the same HSC dose is transplanted between both groups. This however is technically difficult since the membrane markers from both groups are different. Although after 8 weeks chimerism levels seem to be the same (SF5D) for both groups, it would strengthen the evidence if the author could demonstrate that the same number of HSCs were transplanted in both groups, likely by limiting dose experiments. Finally, it's interesting that even though EE100 cells underwent multiple replication rounds (adding to their replicative aging), these cells remained more quiescent once they were in an in vivo setting. Since the last author of this manuscript has also expertise in HSC aging, it would be interesting to explore whether these cells have "aged" during the expansion process by assessing whether they display an aged phenotype (myeloid-skewed output in serial transplantations and/or assisting their transcriptional age).

We thank the reviewer for the insightful observations regarding the quiescence of transplanted cultured HSCs. We appreciate the opportunity to clarify the experimental design and its implications, particularly in the context of HSC aging.

The primary aim of comparing cKit-enriched bone BM cells with cultured cells was to investigate if ex vivo activated HSCs exhibit a similar proliferation pattern to in vivo quiescent HSCs post-transplantation. This comparison was crucial for evaluating the similarity between in vitro cultured and "unmanipulated" HSC behavior. While we acknowledge the technical challenge of transplanting equivalent HSC doses between groups due to differing membrane markers, our study design focused on assessing stem cell activity post-culture. This was quantitatively evaluated by calculating the repopulating units (detailed in Table 1 and Fig S4G), rather than through a limiting dilution assay. There exists a plethora of literature demonstrating the correlation between these assays, although of course the limiting dilution assay is designed to provide a more exact output.

Regarding the intriguing aspect of HSC aging in the context of ex vivo expansion, our observations indicate that both the subfraction of ex vivo expanded cells (Fig 3 and Fig S3) and the entire cultured population (Fig 4B, Fig 5B, Fig S4A, and Fig S5B) maintain long-term multilineage reconstitution capacity post-transplantation. This suggests that the PVA-culture system does not lead to apparent signs of "HSC aging," despite the cells undergoing active self-renewal in vitro. This is further supported by our serial transplantation experiments, where cultured cells continued to demonstrate multilineage capacity rather than any evident myeloid-biased reconstitution 16 weeks post-second transplantation (see Author response image 4 below).

Author response image 4.

Serial transplantation behavior of ex vivo expanded HSCs. 5 million whole BM cells from primary transplantation were transplanted together with 5 million competitor whole BM cells. The control group was transplanted with 100 cHSCs freshly isolated from BM for the primary transplantation. Mann-Whitney test was applied and the asterisks indicate significant differences. , p < 0.05; , p < 0.01; ***, p < 0.0001. Error bars denote SEM.

However, we recognize the complexity of defining HSC aging and the potential for the culture system to influence certain aspects of this process. The association of aging signature genes with HSC primitiveness and young signature genes with differentiation presents an interesting dichotomy. Our analysis of a native dataset on young mice and the projection of aged signatures onto our multiome data (as shown below for a set of genes known to be induced at higher levels in aged HSCs (f.i. Wahlestedt et al., Nature Comm 2017), aging scRNAseq data from PMID: 36581635) does not directly indicate that the culture system promotes HSC aging compared to aged Lin-Sca+Kit+ cells. Yet, we do not rule out the possibility that culturing may influence other facets of the HSC aging process.

In conclusion, while our current data do not provide direct evidence of induced HSC aging through the culture system, this remains a compelling area for future research. The potential impact of ex vivo culture on aspects of the HSC aging process warrants further exploration, and we appreciate your suggestion in this regard.

Author response image 5.

No evident signs of "molecular aging" following ex vivo expansion of HSCs. Young and aged scRNAseq data from PMID: 36581635 were integrated and explored from the perspective of known genes associating to HSC aging. The top row depicts contribution to UMAPs from young and aged cells (two left plots), cell cycle scores of the cells, and the expression of EPCR and CD48 as examples markers for primitive and more differentiated cells, respectively. The expression of the HSC aging-associated genes Wwtr1, Cavin2, Ghr, Clu and Aldh1a1 was then assessed in the data as well as in the SAILERX UMAP of cultured HSCs (bottom row).

Reviewer #2 (Public Review):

Summary:

In this study, Zhang and colleagues characterise the behaviour of mouse hematopoietic stem cells when cultured in PVA conditions, a recently published method for HSC expansion (Wilkinson et al., Nature, 2019), using multiome analysis (scRNA-seq and scATACseq in the same single cell) and extensive transplantation experiments. The latter are performed in several settings including barcoding and avoiding recipient conditioning. Collectively the authors identify several interesting properties of these cultures namely: 1) only very few cells within these cultures have long-term repopulation capacity, many others, however, have progenitor properties that can rescue mice from lethal myeloablation; 2) single-cell characterisation by combined scRNAseq and scATACseq is not sufficient to identify cells with repopulation capacity; 3) expanded HSCs can be engrafted in unconditioned host and return to quiescence.

The authors also confirm previous studies that EPCRhigh HSCs have better reconstitution capability than EPCRlow HSCs when transplanted.

Strengths:

The major strength of this manuscript is that it describes how functional HSCs are expanded in PVA cultures to a deeper extent than what has been done in the original publication. The authors are also mindful of considering the complexities of interpreting transplantation data. As these PVA cultures become more widely used by the HSC community, this manuscript is valuable as it provides a better understanding of the model and its limitations.

Novelty aspects include:

• The authors determined that small numbers of expanded HSCs enable transplantation into non-conditioned syngeneic recipients.

• This is to my knowledge the first report characterising the output of PVA cultures by multiome. This could be a very useful resource for the field.

• They are also the first to my knowledge to use barcoding to quantify HSC repopulation capacity at the clonal level after PVA culture.

• It is also useful to report that HSCs isolated from fetal livers do expand less than their adult counterparts in these PVA cultures.

Weaknesses:

• The analysis of the multiome experiment is limited. The authors do not discuss what cell types, other than functional or phenotypic HSCs are present in these cultures (are they mostly progenitors or bona fide mature cells?) and no quantifications are provided.

The primary objective of our manuscript was to characterize the features of HSCs expanded from ex vivo culture. In this context, our analysis of the single cell multiome sequencing data was predominantly centered on elucidating the heterogeneity of cultures, along with subsequent in vivo functional analysis. This focus is reflected in our comparisons between the molecular features of ex vivo cultured candidate HSCs (cHSCs) and "fresh/unmanipulated" HSCs, as illustrated in Figures 2D-E of our manuscript.

Our findings provide substantial evidence that ex vivo expanded cells share significant similarities with HSCs isolated from the BM in terms of molecular features, differentiation potential, heterogeneity, and in vivo stem cell activity/function. This suggests that the ex vivo culture system closely mimics several aspects of the in vivo environment, thereby broadening the potential applications of this system for HSC research.

Regarding the presence of other cell types in the cultures, it is important to note that most cells did not express mature lineage markers, suggesting their immature status. However, we acknowledge the presence of some mature lineage marker-positive cells within the cultures. These cells are represented by the endpoints in our SAILERX UMAP, indicating a progression from immature to more differentiated states within the culture system.

While the main emphasis of our study was on HSCs, we understand the importance of acknowledging and briefly discussing the presence and characteristics of other cell types in the cultures. This aspect provides a more comprehensive understanding of the culture system and its impact on cellular heterogeneity, although it was for the most part beyond the scope of our studies.

• Barcoding experiments are technically elegant but do not bring particularly novel insights. We respectfully disagree with the view that our barcoding experiments do not offer novel insights. We believe that the application of barcoding technology in our study represents a significant advancement over previous methods, both in terms of quantitative rigor and ethical considerations.

In the foundational work by Wilkinson et al., clonal assessments were indeed performed, but these were limited in scope and largely served as proof of concept. Our use of barcoding technology, on the other hand, allowed for a comprehensive quantitative assessment of the expansion potential of HSC clones. This technology enabled us to rigorously quantify the number of HSC clones capable of undergoing at least three self-renewing divisions (e.g. those clones present in 5 separate animals), while also revealing the heterogeneity in their expansion potential.

One alternative approach could have been to culture single HSCs and distribute the progeny among multiple mice for analysis. However, when considering the sheer number of mice that would be required for such an experiment for quantitative assessments, it becomes evident that viral barcoding is a far superior method. Not only does it provide a more efficient and scalable approach to assessing clonal expansion, but it also significantly reduces the number of animals required for the study, aligning with the principles of ethical research and animal welfare.

In conclusion, we assert that the barcoding experiments conducted in our study are not only technically robust but also yield novel quantitative insights into the dynamics of HSC clones within expansion cultures. These insights have value not only for current research but also hold potential implications for future applications.

• The number of mice analysed in certain experiments is fairly low (Figures 1 and 5).

We would like to clarify our approach in the context of the 3R (replacement, refinement, and reduction) policy, which guides ethical considerations in animal research.

In alignment with the 3R principles, our study was designed to minimize the use of experimental animals wherever possible. For most experiments, including those presented in Figures 1 and 5, we adopted a standard of using five mice per group. Based on the effect sizes we observed, we concluded that this sample size was appropriate for most parts of our study.

Specifically for Figure 5, we used two animals per time point, totaling seven animals per treatment group. It is important to note that we did not monitor the same animals over time but used different animals at each time point, as mice had to be sacrificed for the type of analyses conducted. Despite the seemingly small sample size, the results we obtained were remarkably consistent across groups. This consistency provided strong evidence that ex vivo activated HSCs return to a more quiescent state after being transplanted into unconditioned recipients. Given the clear and consistent nature of these results, we determined that including more animals for the purpose of additional statistical analysis was not necessary.

Our approach reflects a balance between adhering to ethical standards in animal research and ensuring the scientific validity and reliability of our findings. We believe that the sample sizes chosen for our experiments are justified by the consistent and significant results we obtained, which contribute meaningfully to our understanding of HSC behavior post-transplantation.

• The manuscript remains largely descriptive. While the data can be used to make useful recommendations to future users working with PVA cultures and in general with HSCs, those recommendations could be more clearly spelled out in the discussion.

We fully agree that many aspects of our study are indeed descriptive, which is reflective of the exploratory and foundational nature of this type of research.

We have strived to provide clear and direct recommendations for researchers interested in utilizing the PVA culture system, which we believe are evident throughout our manuscript:

1) Utility of Viral Delivery in HSC Research: Our research, particularly through the use of barcoding experiments, underscores the effectiveness of viral delivery methods in HSC studies. While barcoding itself is a significant tool, it is the underlying process of viral delivery that truly exemplifies the potential of this approach. Our work shows that the culture system is highly conducive to maintaining HSC activity, which is critical for genetic manipulation. This is evident not only in our current study but also in our previous work that included for transient delivery methods (Eldeeb et al., Cell Reports 2023).

2) Non-conditioned transplantation: Our findings suggest that non-conditioned transplantation can be a valuable method in studying both normal and malignant hematopoiesis. This approach can complement genetic lineage tracing models, providing a more native and physiological context for hematopoietic research. We state this explicitly in our discussion.

3) Integration with recent technical advances: The combination of the PVA culture system with recent developments in transplantation biology, genome engineering, and single-cell technologies holds significant promise. This integration is likely to yield exciting discoveries with relevance to both basic and clinically oriented hematopoietic research. This is the end statement of our discussion.

While our manuscript is in a way tailored to those with experience in HSC research, we have made a concerted effort to ensure that the content is accessible and informative to a broader audience, including those less familiar with this area of study. Our intention is to provide a resource that is both informative for experts in the field and approachable for newcomers.

• The authors should also provide a discussion of the other publications that have used these methods to date.

We would like to clarify that the scope of literature on the specific methods we employed, particularly in the context of our research objectives, is not extensive. Most of the existing references on these methods come from a relatively narrow range of research groups. In preparing our manuscript, we tried to be comprehensive yet selective in our citations to maintain focus and relevance. Our referencing strategy was guided by the aim to include literature that was most directly pertinent to our study's methodologies and findings.

Overall, the authors succeeded in providing a useful set of experiments to better interpret what type of HSCs are expanded in PVA cultures. More in-depth mining of their bioinformatic data (by the authors or other groups) is likely to highlight other interesting/relevant aspects of HSC biology in relation to this expansion methodology.

We are grateful for the overall positive assessment of our work and the recognition of its contributions to understanding HSC expansion in PVA cultures.

We agree that every study, including ours, has its limitations, particularly regarding the scope and depth of exploration. It is challenging to cover every aspect comprehensively in a single study. Our research aimed to provide a foundational understanding of HSCs in PVA cultures, and we are pleased that this goal appears to have been met.

We also concur with your point on the potential for further in-depth mining of our bioinformatic data. Our hope is that this data can serve as a resource (or at least a starting point) for other investigators.

In conclusion, we hope that our responses have adequately addressed your queries and clarified any concerns. We are committed to contributing to the growth of knowledge in HSC research and look forward to the advancements that our study might enable, both within our team and the wider scientific community.

Reviewer #1 (Recommendations For The Authors):

1) In Line 150, the R packages can/should be mentioned just in the method section;

We have moved this text to the methods section.

2) In Figure F3C adding a legend next to the plot would assist the reader in identifying which populations are referred to, as the same color pellet is used for other panels;

We have now adjusted the figure legend position to make it more clear for the reader.

3) In Figure 4D, for the pre-culture experiments 1000 cHSCs were used and then in the post-culture 1200 cHSCs were used. Can the authors justify the different numbers?

The decision to use 1000 cHSCs in the pre-culture experiments and 1200 cHSCs in the post-culture experiments was not based on a specific rationale favoring one cell number over the other. In our Method section, we have detailed our experimental design, which was structured to provide robust and reliable readouts of HSC behavior and characteristics in different conditions.

We consider the two cell numbers – 1000 and 1200 – to be quite similar in the context of our experimental aims. Since the readouts here are based on clonal assessments, this slight difference in cell numbers is unlikely to significantly impact the overall conclusions drawn from these experiments. The primary focus of our study was on qualitative aspects of HSC behavior and function, rather than on quantitative differences that might arise from small variations in initial cell numbers.

4) In SF5F it would help readers if a line plot (per group) was also shown together with the dot plots. Moreover, applying statistics to the trend lines (Wilcoxon, for example) would strengthen the argument that cHSCs divide less than control cells.

We would like to clarify that the data presented in SF5F were derived from different animals at each respective time point. As such, the data points at each time point represent independent measurements from separate animals, rather than a continuous measurement from the same set of animals over time. Therefore, creating a line plot that connects each time point within a group would inadvertently convey a misleading impression of a longitudinal study on the same animals, which is not reflective of the actual experimental design. Instead, the dot plot format was chosen as it more accurately depicts the independent and discrete nature of the measurements at each time point. Our current data presentation method was selected to provide the most accurate and transparent representation of our findings.

Reviewer #2 (Recommendations For The Authors):

Listed below are recommendations to further improve this manuscript:

Major Comments

1) Fig 1: the authors showed that EPCRhigh HSCs have better reconstitution capability than EPCRlow HSCs via bone marrow transplantation. Additionally, mice receiving cultured EPCRhigh SLAM LSK cells were more efficiently radioprotected than those receiving PVA expanded EPCRlow SLAM LSK.

a. In addition to Fig.1F, authors should show the lineage distributions and chimerism of mice receiving cultured EPCRhigh and EPCRlow SLAM LSK respectively.

We have indeed analyzed the lineage distribution in these experiments, and our findings indicate no statistically significant differences between the groups (see graph in Author response image 6). This suggests that the cultured EPCRhigh and EPCRlow SLAM LSK cells do not preferentially differentiate into specific lineages in a way that would impact the overall interpretation of our results.

Author response image 6.

Regarding the chimerism in peripheral blood (PB) lineages, Fig. 1F in our manuscript currently shows the PB myeloid chimerism. We chose to focus on this parameter as it most directly relates to our study's objectives. We did here not transplant with competitor cells, and in most cases, the chimerism levels reached 100% for lineages other than T cells (T cells being more radioresistant). Based on our analysis, including data on chimerism in other PB lineages would not significantly enhance the understanding of the functional capacity of the transplanted cells, as the myeloid chimerism data already provides a robust indicator of their engraftment and functional potential.

We believe that our current presentation of data in Fig. 1F, along with the additional analyses provided in the results section, offers a comprehensive understanding of the behavior and potential of the cultured EPCRhigh and EPCRlow SLAM LSK cells.

b. Fig1F: only 5 mice were used in each group. Could this result occur by chance? Testing with Fisher's exact test with the data provided results in p=0.16. The authors should consider adding more animals or adding the p-value above (or from another relevant test) for readers' consideration.

We acknowledge the point that only five mice were used in each group and understand the concern regarding the robustness of our findings.

As correctly noted, applying Fisher's exact test to the data in Fig. 1F results in a p-value which does not reach the conventional threshold for statistical significance. However, one might also consider the analysis of the KM survival curve, which associated with a p-value of 0.0528 (Fig. 1F, left graph below; Gehan-Breslow-Wilcoxon test). A similar test on the single-cell culture transplantation experiment (Fig. 1E, right graph below) also demonstrated statistical significance (p-value = 0.0485).

While these p-values meet (or are very close to) the conventional criteria for statistical significance (p<0.05), we have chosen to place greater emphasis on effect sizes rather than strictly on p-values. This decision is based on our belief that effect sizes provide a more direct and meaningful measure of the biological impact observed in our experiments. We find that the effect sizes observed are compelling and consistent with the overall narrative of our study.

Author response image 5.

2) The characterisation of the multiome experiment is highly underdeveloped.

a. From an experimental point of view, it is not clear how the PVA culture for this experiment was started. Are there technical/biological replicates? Have several PVA cultures been pooled together?

We have included these details in the revised text to ensure a comprehensive understanding of our experimental setup.

b. Fig2B: The authors should present more data as to how each of the clusters was annotated (bubble plot of marker genes used for annotation?) and importantly the percentage of cells in each of the clusters. It is particularly relevant to note what % is the cluster annotated as HSCs and compare that to the % of phenotypic HSCs and the % repopulating HSCs calculated in the transplantation experiments.

In our study, the annotation of clusters was primarily based on reference genes for cell types from prior works in the field, such as from our recent work (Konturek-Ciesla et al., Cell Reports 2023). Additionally, we employed transcription factor (TF) motifs to assign identities to these clusters. This approach is relatively standard in the field, and we believe it provides a robust framework for our analysis. We included information on some of the key TF motifs used to guide our annotations.

Regarding the assignment of a percentage to cells within the HSC cluster, we initially had reservations about the utility of this measure. This is because the transcriptional identity of HSCs might not align precisely with their identity based on candidate HSC protein markers. There are complexities related to transcriptional continuums that could influence the interpretation of such data. However, acknowledging your request for this information, we have now included the percentage of cells in the HSC cluster in Fig. 2B for reference.

We also wish to highlight that when isolating EPCR+ cells, which encompasses a range of CD48 expression, clustering becomes much less distinct, as shown in Fig. 2E. Most of these cells do not demonstrate long-term functional HSC activity in a transplantation setting (as presented in Figure 3). This observation underscores the challenges in deducing HSC identity based solely on molecular data and reinforces the importance of functional validation.

c. Are there any mature cells in these PVA cultures? The annotations presented in the table under the UMAP are vague: Are cluster 4 monocytes or monocytes progenitors? Same for clusters 0,1 and 7 - are these progenitors or more mature cells? How were HPCs (cluster 3) distinguished from cHSCs (cluster 5)?

We agree with your observation that the annotations for certain clusters, such as clusters 4, 0, 1, and 7, as well as the distinction between HPCs (cluster 3) and cHSCs (cluster 5), appear vague. This vagueness to some extent stems from the challenges inherent in comparing cultured cells to their counterparts isolated directly from animals. Most reference data defining cell types are derived from cells in their native state, and less is known about how these definitions translate to the progeny of HSPCs cultured in vitro.

In our study, we used the expression of reference genes and enriched transcription factor motifs to annotate clusters. This method, while useful, has its limitations in precisely defining the maturation stage of cells in culture. The enrichment of lineage-defining factors at the ends of the UMAP suggests the presence of more mature cells, whereas the lack of lineage marker expression in the majority of cells implies a general lack of terminal differentiation.

This issue is not necessarily unique to the culture situation, as similar challenges in cell type annotation are encountered in other contexts, such as the analysis of granulocyte-macrophage progenitors in bone marrow, where a vast range of cell types and clusters are identified (e.g., PMID: 26627738). To try to address these challenges, we employed an approach detailed in the methods section under the header "iv. ATAC processing and cluster annotation." We assessed marker genes for clusters using Enrichr for cell types, relying on databases designed to provide gene expression identities to defined cell types. This methodology informed our references to the clusters.

In summary, while our annotations provide a general overview of the cell types present in the cultures, we acknowledge the complexities and limitations in precisely defining these types, particularly in distinguishing between progenitors and more mature cells. We hope this explanation clarifies our approach and the considerations behind our cluster annotations, but at the same time feel that the alternative approaches have their own drawbacks.

d. What is the meaning of the trajectories presented in Figure 2C? In the absence of a comparison to i) what is observed either when HSCs are cultured in control/non-expanding conditions ii) an in vivo landscape of differentiation in mouse bone marrow; this analysis does not bring any relevant piece of information.

We understand the perspective on comparisons to control conditions and in vivo differentiation landscapes. However, we respectfully disagree with the viewpoint that the analysis that we have performed does not bring relevant information.

The trajectory analysis in Figure 2C is intended to provide insights into the cell types generated in our PVA cultures and the potential differentiation pathways they may follow. This kind of analysis is particularly valuable in the context of understanding how in vitro cultures can support HSC maintenance and differentiation, which is a topic of significant interest in the field. For instance, studies like PMID: 31974159 have highlighted the importance of combining in vitro HSC cultures with molecular investigations.

While we acknowledge that our analysis would benefit from a direct comparison to control or non-expanding conditions, as well as to an in vivo differentiation landscape, we believe that the information provided by our current analysis still holds substantial value. It offers a glimpse into the possible cellular dynamics and differentiation routes within our culture system, which can be a valuable reference point for other investigators working with similar systems.

Regarding the confidence in computed differentiation trajectories, we recognize that this is an area where caution is warranted. Computational approaches to define cell differentiation pathways have inherent limitations and should be interpreted within the context of their assumptions and the data available. This challenge is not unique to our work but is a broader issue in the field of computational biology.

In conclusion, while we agree that additional comparative analyses could further enrich our findings, we maintain that the trajectory analysis presented in Figure 2C contributes meaningful insights into cell differentiation in our PVA culture system. We believe these insights are of interest and value to researchers exploring the complex interplay of HSC maintenance and differentiation in vitro.

3) The addition of barcoding experiments is appreciated. However, it is already known that upon transplantation clonal output is highly heteroegeneous, with a small number of clones predominating over others. This is particularly the case after myeloablation conditioning.

a. The "pre-culture" experimental design makes sense. The "post-culture" one is however ambiguous in terms of result interpretation. The authors observe fewer clones contributing to a large proportion of the graft (>5%) than in the "pre-culture" setting. Their interpretation is that expanded HSCs are functionally more homogeneous than the input HSCs. However, in the pre-culture experiment, there are 19 days of expansion during which there will be selection pressures over culture plus ongoing differentiation. In the post-culture experiment, there is no time for such pressures to be exerted. Therefore the conclusion drawn by the authors is not the only conclusion. I would encourage the authors to compare the "pre-culture" experiment to an experiment in which cHSCs are in culture for 48h, then barcoded, and then transplanted. This would be much more informative and would allow a proper comparison of expanded HSCs vs input HSCs.

We understand the perspective that a shorter culture period would reduce the influence of selection pressures and differentiation, potentially allowing for a more direct comparison between expanded HSCs and input HSCs. However, we would like to point out that similar experiments have been conducted in the past, as referenced in our work (PMID: 28224997) and others (PMID: 21964413). These studies have demonstrated a significant heterogeneity in the reconstituting clones when barcoding is done early and cells are transplanted directly.

In light of previous research, we are confident that our methodology — tracking the fates of candidate HSC clones throughout the culture period and assessing the outcomes of individual cells from these expanding clones — yields significant and pertinent insights. We want to highlight the significance of barcoding cells late in the culture, a strategy that allows us to barcode cells that have already been subjected to potential selection pressures within the culture environment. Our primary objective is to investigate the effects of these selection pressures on the subsequent in vivo behavior of the cells that emerge from this process. By focusing on this aspect, we aim to deepen the understanding of how in vitro culture conditions influence the functional characteristics and heterogeneity of HSCs after expansion. We believe this approach provides a unique perspective on the adaptive changes HSCs undergo during culture and their implications for transplantation efficacy and HSC biology. Our study thus addresses a critical question in the field: how do the conditions and selection pressures inherent to in vitro culture impact the quality and behavior of HSCs upon their return to an in vivo environment?

b. Another experiment the authors may consider is barcoding in unconditioned recipients as there the bottleneck of selecting specific clones should be lower. In addition, this could nicely complement the return to quiescence observed in Figure 5 (see point below)

We agree that this experiment could provide valuable insights, particularly in understanding how different selection pressures might affect HSC clones in various transplantation contexts. It would indeed be a worthwhile complement to our observations in Figure 5 regarding the return to quiescence of HSCs post-transplantation.

However, we would like to point out that our study already includes a substantial amount of data and analyses aimed at addressing specific research questions within this defined scope. The addition of an experiment with barcoding in unconditioned recipients, while undoubtedly relevant and interesting, would extend beyond the boundaries we set for this particular study.

4) Figure 5D-F, only 2 animals per condition were tested, so the experiment is underpowered for any statistics. How about cell viability of cHSC after in vitro culture? The authors have also not tested whether there is a difference in cell viability post-transplant between EE100 and control. In addition, comparing cell cycle profiles of donor EPCR+ HSCs in these transplanted mice would provide additional evidence to support the conclusion.

Regarding the sample size, we acknowledge that only two animals per condition were used in these experiments, which limits the statistical power for robust quantitative analysis. This decision was guided by ethical considerations to minimize animal use, in line with the 3Rs principle (Replacement, Reduction, Refinement). Despite the small sample size, we believe that the strong trends observed in these experiments are indicative and consistent with our broader findings, although we recognize the limitations in terms of statistical generalization. At the same time, as we have written in the public response: "Specifically for Figure 5, we used two animals per time point, totaling seven animals per treatment group. It is important to note that we did not monitor the same animals over time but used different animals at each time point, as mice had to be sacrificed for the type of analyses conducted."

In the context of post-transplant analysis, conducting separate viability assessments on transplanted cells is not typically informative. This is because non-viable cells would naturally be eliminated through biological processes such as phagocytosis soon after transplantation. Therefore, any post-transplant viability analysis would not provide meaningful insights into the engraftment potential or behavior of the transplanted cells.

However, it is important to note that in all our cell isolation and analysis protocols, we routinely include viability markers. This practice ensures that the cell populations we study and report on are indeed viable. Including these markers is a standard part of our methodology and contributes to the accuracy and reliability of our data.

Regarding the comparison of cell cycle profiles, we chose to focus on the cell trace assay as a means to monitor and track cell division history, which directly addresses the central theme here - informing on the proliferation and quiescence dynamics of transplanted HSCs. While comparing cell cycle profiles could perhaps offer an additional layer of information, we did not deem it essential for our core objectives.

5) Several publications have used these PVA cultures and made comments on their strengths and limitations. They do not overlap with this study but should be discussed here for completeness (for example Che et al, Cell Reports, 2022; Becker et al., Cell Stem Cell, 2023; Igarashi, Blood Advances, 2023).

See comments to reviewer 1.

Minor Comments

Figure 1C: should add in the legend that this is in peripheral blood.

Figure 2C: typo in the title.

Figure 3A: typo in "equivalent". We thank the reviewer for catching these errors, which we have now corrected.

Figure 3B and 3C: symbol colours of EPCRhighCD48+ and EPCR- are too similar to distinguish the 2 groups easily. We highly recommend using contrasting colours.

For easier visualization, we have changed the symbol types and colors in our revised version.

Fig3B and S3A-B: authors should show statistical significance in comparing the 4 fractions. We have now added this information.

In the discussion, the authors rightly point out a paper that described EPCR+ HSCs. There are other papers that also looked at EPCR intensity (high vs low), for example, Umemoto et al., EMBO J, 2022.

While we acknowledge the relevance of the paper you mentioned, we faced constraints in the number of references we could include. Therefore, we prioritized citing the original demonstration of EPCR as an HSC marker, particularly focusing on the work by the Mulligan laboratory, which established that cells expressing the highest levels of EPCR exhibit the most potent HSC activity. We believe this reference most directly supports the core focus of our study and provides the necessary context for our findings.

Author Response

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

eLife assessment

This valuable study reports on the potential of neural networks to emulate simulations of human ventricular cardiomyocyte action potentials for various ion channel parameters with the advantage of saving simulation time in certain conditions. The evidence supporting the claims of the authors is solid, although the inclusion of open analysis of drop-off accuracy and validation of the neural network emulators against experimental data would have strengthened the study. The work will be of interest to scientists working in cardiac simulation and quantitative pharmacology.

Thank you for the kind assessment. It is important for us to point out that, while limited, experimental validation was performed in this study and is thoroughly described in the work.

Reviewer 1 - Comments

This manuscript describes a method to solve the inverse problem of finding the initial cardiac activations to produce a desired ECG. This is an important question. The techniques presented are novel and clearly demonstrate that they work in the given situation. The paper is well-organized and logical.

Strengths:

This is a well-designed study, which explores an area that many in the cardiac simulation community will be interested in. The article is well written and I particularly commend the authors on transparency of methods description, code sharing, etc. - it feels rather exemplary in this regard and I only wish more authors of cardiac simulation studies took such an approach. The training speed of the network is encouraging and the technique is accessible to anyone with a reasonably strong GPU, not needing specialized equipment.

Weaknesses:

Below are several points that I consider to be weaknesses and/or uncertainties of the work:

C I-(a) I am not convinced by the authors’ premise that there is a great need for further acceleration of cellular cardiac simulations - it is easy to simulate tens of thousands of cells per day on a workstation computer, using simulation conditions similar to those of the authors. I do not really see an unsolved task in the field that would require further speedup of single-cell simulations. At the same time, simulations offer multiple advantages, such as the possibility to dissect mechanisms of the model behaviour, and the capability to test its behaviour in a wide array of protocols - whereas a NN is trained for a single purpose/protocol, and does not enable a deep investigation of mechanisms. Therefore, I am not sure the cost/benefit ratio is that strong for single-cell emulation currently.

An area that is definitely in need of acceleration is simulations of whole ventricles or hearts, but it is not clear how much potential for speedup the presented technology would bring there. I can imagine interesting applications of rapid emulation in such a setting, some of which could be hybrid in nature (e.g. using simulation for the region around the wavefront of propagating electrical waves, while emulating the rest of the tissue, which is behaving more regularly/predictable, and is likely to be emulated well), but this is definitely beyond of the scope of this article.

Thank you for this point of view. Simulating a population of few thousand cells is completely feasible on single desktop machines and for fixed, known parameters, emulation may not fill ones need. Yet we still foresee a great untapped potential for rapid evaluations of ionic models, such as for the gradient-based inverse problem, presented in the paper. Such inverse optimization requires several thousand evaluations per cell and thus finding maximum conductances for the presented experimental data set (13 cell pairs control/drug → 26 APs) purely through simulations would require roughly a day of simulation time even in a very conservative estimation (3.5 seconds per simulation, 1000 simulations per optimization). Additionally, the emulator provides local sensitivity information between the AP and maximum conductances in the form of the gradient, which enables a whole new array of efficient optimization algorithms [Beck, 2017]. To further emphasize these points, we added the number of emulations and runtime of each conducted experiment in the specific section and a paragraph in the discussion that addresses this point:

"Cardiomyocyte EP models are already very quick to evaluate in the scale of seconds (see Section 2.3.1), but the achieved runtime of emulations allows to solve time consuming simulation protocols markedly more efficient. One such scenario is the presented inverse maximum conductance estimation problem (see Section 3.1.2 and Section 3.1.3), where for estimating maximum conductances of a single AP, we need to emulate the steady state AP at least several hundred times as part of an optimization procedure. Further applications include the probabilistic use of cardiomyocyte EP models with uncertainty quantification [Chang et al., 2017, Johnstone et al., 2016] where thousands of samples of parameters are potentially necessary to compute a distribution of the steady-state properties of subsequent APs, and the creation of cell populations [Muszkiewicz et al., 2016, Gemmell et al., 2016, Britton et al., 2013]." (Section 4.2)

We believe that rapid emulations are valuable for several use-cases, where thousands of evaluations are necessary. These include the shown inverse problem, but similarly arise in uncertainty quantification, or cardiomyocyte population creation. Similarly, new use-cases may arise as such efficient tools become available. Additionally, we provided the number of evaluations along with the runtimes for each of the conducted experiments, showing how essential these speedups are to realize these experiments in reasonable timeframes. Utilizing these emulations in organ-level electrophysiological models is a possibility, but the potential problems in such scenarios are much more varied and depend on a number of factors, making it hard to pin-point the achievable speed-up using ionic emulations.

C I-(b) The authors run a cell simulation for 1000 beats, training the NN emulator to mimic the last beat. It is reported that the simulation of a single cell takes 293 seconds, while emulation takes only milliseconds, implying a massive speedup. However, I consider the claimed speedup achieved by emulation to be highly context-dependent, and somewhat too flattering to the presented method of emulation. Two specific points below:

First, it appears that a not overly efficient (fixed-step) numerical solver scheme is used for the simulation. On my (comparable, also a Threadripper) CPU, using the same model (”ToR-ORd-dyncl”), but a variable step solver ode15s in Matlab, a simulation of a cell for 1000 beats takes ca. 50 seconds, rather than 293 of the authors. This can be further sped up by parallelization when more cells than available cores are simulated: on 32 cores, this translates into ca. 2 seconds amortized time per cell simulation (I suspect that the NN-based approach cannot be parallelized in a similar way?). By amortization, I mean that if 32 models can be simulated at once, a simulation of X cells will not take X50 seconds, but (X/32)50. (with only minor overhead, as this task scales well across cores).

Second, and this is perhaps more important - the reported speed-up critically depends on the number of beats in the simulation - if I am reading the article correctly, the runtime compares a simulation of 1000 beats versus the emulation of a single beat. If I run a simulation of a single beat across multiple simulated cells (on a 32-core machine), the amortized runtime is around 20 ms per cell, which is only marginally slower than the NN emulation. On the other hand, if the model was simulated for aeons, comparing this to a fixed runtime of the NN, one can get an arbitrarily high speedup.

Therefore, I’d probably emphasize the concrete speedup less in an abstract and I’d provide some background on the speedup calculation such as above, so that the readers understand the context-dependence. That said, I do think that a simulation for anywhere between 250 and 1000 beats is among the most reasonable points of comparison (long enough for reasonable stability, but not too long to beat an already stable horse; pun with stables was actually completely unintended, but here it is...). I.e., the speedup observed is still valuable and valid, albeit in (I believe) a somewhat limited sense.

We agree that the speedup comparison only focused on a very specific case and needs to be more thoroughly discussed and benchmarked. One of the main strengths of the emulator is to cut the time of prepacing to steady state, which is known to be a potential bottleneck for the speed of the single-cell simulations. The time it takes to reach the steady state in the simulator is heavily dependant on the actual maximum conductance configuration and the speed-up is thus heavily reliant on a per-case basis. The differences in architecture of the simulator and emulator further makes direct comparisons very difficult. In the revised version we now go into more detail regarding the runtime calculations and also compare it to an adaptive time stepping simulation (Myokit [Clerx et al., 2016]) in a new subsection:

"The simulation of a single AP (see Section 2.1) sampled at a resolution of 20kHz took 293s on one core of a AMD Ryzen Threadripper 2990WX (clock rate: 3.0GHz) in CARPentry. Adaptive timestep solver of variable order, such as implemented in Myokit [Clerx et al., 2016], can significantly lower the simulation time (30s for our setup) by using small step sizes close to the depolarization (phase 0) and increasing the time step in all other phases. The emulation of a steady state AP sampled at a resolution of 20kHz for t ∈ [−10, 1000]ms took 18.7ms on a AMD Ryzen 7 3800X (clock rate: 3.9GHz) and 1.2ms on a Nvidia A100 (Nvidia Corporation, USA), including synchronization and data copy overhead between CPU and GPU.

"The amount of required beats to reach the steady state of the cell in the simulator has a major impact on the runtime and is not known a-priori. On the other hand, both simulator and emulator runtime linearly depends on the time resolution, but since the output of the emulator is learned, the time resolution can be chosen at arbitrarily without affecting the AP at the sampled times. This makes direct performance comparisons between the two methodologies difficult. To still be able to quantify the speed-up, we ran Myokit using 100 beats to reach steady state, taking 3.2s of simulation time. In this scenario, we witnessed a speed-up of 171 and 2 · 103 of our emulator on CPU and GPU respectively (again including synchronization and data copy overhead between CPU and GPU in the latter case). Note that both methods are similarly expected to have a linear parallelization speedup across multiple cells.

For the inverse problem, we parallelized the problem for multiple cells and keep the problem on the GPU to minimize the overhead, achieving emulations (including backpropagation) that run in 120µs per AP at an average temporal resolution of 2kHz. We consider this the peak performance which will be necessary for the inverse problem in Section 3.1.2." (Section 2.3.1)

Note that the mentioned parallelization across multiple machines/hardware applies equally to the emulator and simulator (linear speed-up), though the utilization for single cells is most likely different (single vs. multi-cell parallelization).

C I-(c) It appears that the accuracy of emulation drops off relatively sharply with increasing real-world applicability/relevance of the tasks it is applied to. That said, the authors are to be commended on declaring this transparently, rather than withholding such analyses. I particularly enjoyed the discussion of the not-always amazing results of the inverse problem on the experimental data. The point on low parameter identifiability is an important one and serves as a warning against overconfidence in our ability to infer cellular parameters from action potentials alone. On the other hand, I’m not that sure the difference between small tissue preps and single cells which authors propose as another source of the discrepancy will be that vast beyond the AP peak potential (probably much of the tissue prep is affected by the pacing electrode?), but that is a subjective view only. The influence of coupling could be checked if the simulated data were generated from 2D tissue samples/fibres, e.g. using the Myokit software.

Given the points above (particularly the uncertain need for further speedup compared to running single-cell simulations), I am not sure that the technology generated will be that broadly adopted in the near future.

However, this does not make the study uninteresting in the slightest - on the contrary, it explores something that many of us are thinking about, and it is likely to stimulate further development in the direction of computationally efficient emulation of relatively complex simulations.

We agree that the parameter identifiability is an important point of discussion. While the provided experimental data gave us great insights already, we still believe that given the differences in the setup, we can not draw conclusions about the source of inaccuracies with absolute certainty. The suggested experiment to test the influence of coupling is of interest for future works and has been integrated into the discussion. Further details are given in the response to the recommendation R III- (t)

Reviewer 2 - Comments

Summary:

This study provided a neural network emulator of the human ventricular cardiomyocyte action potential. The inputs are the corresponding maximum conductances and the output is the action potential (AP). It used the forward and inverse problems to evaluate the model. The forward problem was solved for synthetic data, while the inverse problem was solved for both synthetic and experimental data. The NN emulator tool enables the acceleration of simulations, maintains high accuracy in modeling APs, effectively handles experimental data, and enhances the overall efficiency of pharmacological studies. This, in turn, has the potential to advance drug development and safety assessment in the field of cardiac electrophysiology.

Strengths:

1) Low computational cost: The NN emulator demonstrated a massive speed-up of more than 10,000 times compared to the simulator. This substantial increase in computational speed has the potential to expedite research and drug development processes

2) High accuracy in the forward problem: The NN emulator exhibited high accuracy in solving the forward problem when tested with synthetic data. It accurately predicted normal APs and, to a large extent, abnormal APs with early afterdepolarizations (EADs). High accuracy is a notable advantage over existing emulation methods, as it ensures reliable modeling and prediction of AP behavior

C II-(a) Input space constraints: The emulator relies on maximum conductances as inputs, which explain a significant portion of the AP variability between cardiomyocytes. Expanding the input space to include channel kinetics parameters might be challenging when solving the inverse problem with only AP data available.

Thank you for this comment. We consider this limitation a major drawback, as discussed in Section 4.3. Identifiability is already an issue when only considering the most important maximum conductances. Further extending the problem to include kinetics will most likely only increase the difficulty of the inverse problem. For the forward problem though, it might be of interest to people studying ionic models to further analyze the effects of channel kinetics.

C II-(b) Simplified drug-target interaction: In reality, drug interactions can be time-, voltage-, and channel statedependent, requiring more complex models with multiple parameters compared to the oversimplified model that represents the drug-target interactions by scaling the maximum conductance at control. The complex model could also pose challenges when solving the inverse problem using only AP data.

Thank you pointing out this limitation. We slightly adapted Section 4.3 to further highlight some of these limitations. Note however that the experimental drugs used have been shown to be influenced by this drug interaction in varying degrees [Li et al., 2017] (e.g. dofetilide vs. cisapride). However, the discrepancy in identifiability was mostly channel-based (0%-100%), whereas the variation in identifiability between drugs was much lower (39%-66%).

C II-(c) Limited data variety: The inverse problem was solved using AP data obtained from a single stimulation protocol, potentially limiting the accuracy of parameter estimates. Including AP data from various stimulation protocols and incorporating pacing cycle length as an additional input could improve parameter identifiability and the accuracy of predictions.

The proposed emulator architecture currently only considers the discussed maximum conductances as input and thus can only compensate when using different stimulation protocols. However, the architecture itself does not prohibit including any of these as parameters for future variants of the emulator. We potentially foresee future works extending on the architecture with modified datasets to include other parameters of importance, such as channel kinetics, stimulation protocols and pacing cycle lengths. These will however vary between the actual use-cases one is interested in.

C II-(d) Larger inaccuracies in the inverse problem using experimental data: The reasons for this result are not quite clear. Hypotheses suggest that it may be attributed to the low parameter identifiability or the training data set were collected in small tissue preparation.

The low parameter identifiability on some channels (e.g. GK1) poses a problem, for which we state multiple potential reasons. As of yet, no final conclusion can be drawn, warranting further research in this area.

Reviewer 3 - Comments

Summary:

Grandits and colleagues were trying to develop a new tool to accelerate pharmacological studies by using neural networks to emulate the human ventricular cardiomyocyte action potential (AP). The AP is a complex electrical signal that governs the heartbeat, and it is important to accurately model the effects of drugs on the AP to assess their safety and efficacy. Traditional biophysical simulations of the AP are computationally expensive and time-consuming. The authors hypothesized that neural network emulators could be trained to predict the AP with high accuracy and that these emulators could also be used to quickly and accurately predict the effects of drugs on the AP.

Strengths:

One of the study’s major strengths is that the authors use a large and high-quality dataset to train their neural network emulator. The dataset includes a wide range of APs, including normal and abnormal APs exhibiting EADs. This ensures that the emulator is robust and can be used to predict the AP for a variety of different conditions.

Another major strength of the study is that the authors demonstrate that their neural network emulator can be used to accelerate pharmacological studies. For example, they use the emulator to predict the effects of a set of known arrhythmogenic drugs on the AP. The emulator is able to predict the effects of these drugs, even though it had not been trained on these drugs specifically.

C III-(a) One weakness of the study is that it is important to validate neural network emulators against experimental data to ensure that they are accurate and reliable. The authors do this to some extent, but further validation would be beneficial. In particular for the inverse problem, where the estimation of pharmacological parameters was very challenging and led to particularly large inaccuracies.

Thank you for this recommendation. Further experimental validation of the emulator in the context of the inverse problem would be definitely beneficial. Still, an important observation is that the identifiability varies greatly between channels. While the inverse problem is an essential reason for utilizing the emulator, it is also empirically validated for the pure forward problem and synthetic inverse problem, together with the (limited) experimental validation. The sources of problems arising in estimating the maximum conductances of the experimental tissue preparations are important to discuss in future works, as we now further emphasize in the discussion. See also the response to the recommendations R III-(t).

Reviewer 1 - Recommendations

R I-(a) Could further detail on the software used for the emulation be provided? E.g. based on section 2.2.2, it sounds like a CPU, as well as GPU-based emulation, is possible, which is neat.

Indeed as suspected, the emulator can run on both CPUs and GPUs and features automatic parallelization (per-cell, but also multi-cell), which is enabled by the engineering feats of PyTorch [Paszke et al., 2019]. This is now outlined in a bit more detail in Sec. 2 and 5.

"The trained emulator is provided as a Python package, heavily utilizing PyTorch [Paszke et al., 2019] for the neural network execution, allowing it to be executed on both CPUs and NVidia GPUs." (Section 5)

R I-(b) I believe that a potential use of NN emulation could be also in helping save time on prepacing models to stability - using the NN for ”rough” prepacing (e.g. 1000 beats), and then running a simulation from that point for a smaller amount of time (e.g. 50 beats). One could monitor the stability of states, so if the prepacing was inaccurate, one could quickly tell that these models develop their state vector substantially, and they should be simulated for longer for full accuracy - but if the model was stable within the 50 simulated beats, it could be kept as it is. In this way, the speedup of the NN and accuracy and insightfulness of the simulation could be combined. However, as I mentioned in the public review, I’m not sure there is a great need for further speedup of single-cell simulations. Such a hybrid scheme as described above might be perhaps used to accelerate genetic algorithms used to develop new models, where it’s true that hundreds of thousands to millions of cells are eventually simulated, and a speedup there could be practical. However one would have to have a separate NN trained for each protocol in the fitness function that is to be accelerated, and this would have to be retrained for each explored model architecture. I’m not sure if the extra effort would be worth it - but maybe yes to some people.

Thank you for this valuable suggestion. As pointed out in C I-(a), one goal of this study was to reduce the timeconsuming task of prepacing. Still, in its current form the emulator could not be utilized for prepacing simulators, as only the AP is computed by the emulator. For initializing a simulation at the N-th beat, one would additionally need all computed channel state variables. However, a simple adaptation of the emulator architecture would allow to also output the mentioned state variables.

R I-(c) Re: ”Several emulator architectures were tried on the training and validation data sets and the final choice was hand-picked as a good trade-off between high accuracy and low computational cost” - is it that the emulator architecture was chosen early in the development, and the analyses presented in the paper were all done with one previously selected architecture? Or is it that the analyses were attempted with all considered architectures, and the well-performing one was chosen? In the latter case, this could flatter the performance artificially and a test set evaluation would be worth carrying out.

We apologize for the unclear description of the architectural validation. The validation was in fact carried out with 20% of the training data (data set #1), which is however completely disjoint with the test set (#2, #3, #4, formerly data set #1 and #2) on which the evaluation was presented. To further clarify the four different data sets used in the study, we now dedicated an additional section to describing each set and where it was used (see also our response below R I-(d)), and summarize them in Table 1, which we also added at R II-(a). The cited statement was slightly reworked.

"Several emulator architectures were tried on the training and validation data sets and the final choice was hand-picked as a good trade-off between high accuracy on the validation set (#1) and low computational runtime cost." (Section 2.2.2)

R I-(d) When using synthetic data for the forward and inverse problem, with the various simulated drugs, is it that split of the data into training/validation test set was done by the drug simulated (i.e., putting 80 drugs and the underlying models in the training set, and 20 into test set)? Or were the data all mixed together, and 20% (including drugs in the test set) were used for validation? I’m slightly concerned by the potential of ”soft” data leaks between training/validation sets if the latter holds. Presumably, the real-world use case, especially for the inverse problem, will be to test drugs that were not seen in any form in the training process. I’m also not sure whether it’s okay to reuse cell models (sets of max conductances) between training and validation tests - wouldn’t it be better if these were also entirely distinct? Could you please comment on this?

We completely agree with the main points of apprehension that training, validation and test sets all serve a distinct purpose and should not be arbitrarily mixed. However, this is only a result of the sub-optimal description of our datasets, which we heavily revised in Section 2.2.1 (Data, formerly 2.3.1). We now present the data using four distinct numbers: The initial training/validation data, now called data set #1 (formerly no number), is split 80%/20% into training and validation sets (for architectural choices) respectively. The presented evaluations in Section 2.3 (Evaluation) are purely performed on data set #2 (normal APs, formerly #1), #3 (EADs, formerly #2) and #4 (experimental).

R I-(e) For the forward problem on EADs, I’m not sure if the 72% accuracy is that great (although I do agree that the traces in Fig 12-left also typically show substantial ICaL reactivation, but this definitely should be present, given the IKr and ICaL changes). I would suggest that you also consider the following design for the EAD investigation: include models with less severe upregulation of ICaL and downregulation of IKr, getting a population of models where a part manifests EADs and a part does not. Then you could run the emulator on the input data of this population and be able to quantify true, falsexpositive, negative detections. I think this is closer to a real-world use case where we have drug parameters and a cell population, and we want to quickly assess the arrhythmic risk, with some drugs being likely entirely nonrisky, some entirely risky, and some between (although I still am not convinced it’s that much of an issue to just simulate this in a couple of thousands of cells).

Thank you for pointing out this alternative to address the EAD identification task. Even though the values chosen in Table 2 seem excessively large, we still only witnessed EADs in 171 of the 950 samples. Especially border cases, which are close to exhibiting EADs are hardest to estimate for the NN emulator. As suggested, we now include the study with the full 950 samples (non-EAD & EAD) and classify the emulator AP into one of the labels for each sample. The mentioned 72.5% now represent the sensitivity, whereas our accuracy in such a scenario becomes 90.8% (total ratio of correct classifications):

"The data set #3 was used second and Appendix C shows all emulated APs, both containing the EAD and non-EAD cases. The emulation of all 950 APs took 0.76s on the GPU specified in Section 2.2.3 We show the emulation of all maximum conductances and the classification of the emulation. The comparison with the actual EAD classification (based on the criterion outlined in Appendix A) results in true-positive (EAD both in the simulation and emulation), false-negative (EAD in the simulation, but not in the emulation), false-positive (EAD in the emulation, but not in the simulation) and true-negative (no EAD both in the emulation and simulation). The emulations achieved 72.5% sensitivity (EAD cases correctly classified) and 94.9% specificity (non-EAD cases correctly classified), with an overall accuracy of 90.8% (total samples correctly classified). A substantial amount of wrongly classified APs showcase a notable proximity to the threshold of manifesting EADs. Figure 7 illustrates the distribution of RMSEs in the EAD APs between emulated and ground truth drugged APs. The average RMSE over all EAD APs was 14.5mV with 37.1mV being the maximum. Largest mismatches were located in phase 3 of the AP, in particular in emulated APs that did not fully repolarize." (Section 3.1.1)

R I-(f) Figure 1 - I think a large number of readers will understand the mathematical notation describing inputs/outputs; that said, there may be a substantial number of readers who may find that hard to read (e.g. lab-based researchers, or simulation-based researchers not familiar with machine learning). At the same time, this is a very important part of the paper to explain what is done where, so I wonder whether using words to describe the inputs/outputs would not be more practical and easier to understand (e.g. ”drug-based conductance scaling factor” instead of ”s” ?). It’s just an idea - it needs to be tried to see if it wouldn’t make the figure too cluttered.

We agree that the mathematical notation may be confusing to some readers. As a compromise between using verbose wording and mathematical notation, we introduced a legend in the lower right corner of the figure that shortly describes the notation in order to help with interpreting the figure.

R I-(g) ”APs with a transmembrane potential difference of more than 10% of the amplitude between t = 0 and 1000 ms were excluded” - I’m not sure I understand what exactly you mean here - could you clarify?

With this criterion, we try to discard data that is far away from fully repolarizing within the given time frame, which applies to 116 APs in data set #1 and 50 APs in data set #3. We added a small side note into the text:

"APs with a transmembrane potential difference of more than 10% of the amplitude between t = 0 and 1000ms (indicative of an AP that is far away from full repolarization) were excluded." (Section 2.2.1)

R I-(h) Speculation (for the future) - it looks like a tool like this could be equally well used to predict current traces, as well as action potentials. I wonder, would there be a likely benefit in feeding back the currents-traces predictions on the input of the AP predictor to provide additional information? Then again, this might be already encoded within the network - not sure.

Although not possible with the chosen architecture (see also R I-(b)), it is worth thinking about an implementation in future works and to study differences to the current emulator.

Entirely minor points:

R I-(i) ”principle component analysis” → principal component analysis

Fixed

R I-(j) The paper will be probably typeset by elife anyway, but the figures are often quite far from their sections, with Results figures even overflowing into Discussion. This can be often fixed by using the !htb parameters (\begin{figure}[!htb]), or potentially by using ”\usepackage[section]{placeins}” and then ”\FloatBarrier” at the start and end of each section (or subsection) - this prevents floating objects from passing such barriers.

Thank you for these helpful suggestions. We tried reducing the spacing between the figures and their references in the text, hopefully improving the reader’s experience.

R I-(k) Alternans seems to be defined in Appendix A (as well as repo-/depolarization abnormalities), but is not really investigated. Or are you defining these just for the purpose of explaining what sorts of data were also included in the data?

We defined alternans since this was an exclusion criterion for generating simulation data.

Reviewer 2 - Recommendations

R II-(a) Justification for methods selection: Explain the rationale behind important choices, such as the selection of specific parameters and algorithms.

Thank you for this recommendation, we tried to increase transparency of our choices by introducing a separate data section that summarizes all data sets and their use cases in Section 2.2.1 and also collect many of the explanations there. Additionally we added an overview table (Table 1) of the utilized data.

Author response table 1.

Table 1: Summary of the data used in this study, along with their usage and the number of valid samples. Note that each AP is counted individually, also in cases of control/drug pairs.

R II-(b) Interpretation of the evaluation results: After presenting the evaluation results, consider interpretations or insights into what the results mean for the performance of the emulator. Explain whether the emulator achieved the desired accuracy or compare it with other existing methods. In the revised version, we tried to further expand the discussion on possible applications of our emulator (Section 4.2). See also our response to C I-(a). To the best of our knowledge, there are currently no out-of-the-box methods available for directly comparing all experiments we considered in our work.

Reviewer 3 - Recommendations

R III-(a) In the introduction (Page 3) and then also in the 2.1 paragraph authors speak about the ”limit cycle”: Do you mean steady state conditions? In that case, it is more common to use steady state.

When speaking about the limit cycle, we refer to what is also sometimes called the steady state, depending on the field of research and/or personal preference. We now mention both terms at the first occurence, but stick with the limit cycle terminology which can also be found in other works, see e.g. [Endresen and Skarland, 2000].

R III-(b) On page 3, while comparing NN with GP emulators, I still don’t understand the key reason why NN can solve the discontinuous functions with more precision than GP.

The potential problems in modeling sharp continuities using GPs is further explained in the referenced work [Ghosh et al., 2018] and further references therein:

"Statistical emulators such as Gaussian processes are frequently used to reduce the computational cost of uncertainty quantification, but discontinuities render a standard Gaussian process emulation approach unsuitable as these emulators assume a smooth and continuous response to changes in parameter values [...] Applying GPs to model discontinuous functions is largely an open problem. Although many advances (see the discussion about non-stationarity in [Shahriari et al., 2016] and the references in there) have been made towards solving this problem, a common solution has not yet emerged. In the recent GP literature there are two specific streams of work that have been proposed for modelling non-stationary response surfaces including those with discontinuities. The first approach is based on designing nonstationary processes [Snoek et al., 2014] whereas the other approach attempts to divide the input space into separate regions and build separate GP models for each of the segmented regions. [...]"([Ghosh et al., 2018])

We integrated a short segment of this explanation into Section 1.

R III-(c) Why do authors prefer to use CARPentry and not directly openCARP? The use of CARPentry is purely a practical choice since the simulation pipeline was already set up. As we now point out however in Sec. 2.1 (Simulator), simulations can also be performed using any openly available ionic simulation tool, such as Myokit [Clerx et al., 2016], OpenCOR [Garny and Hunter, 2015] and openCARP [Plank et al., 2021]. We emphasized this in the text.

"Note, that the simulations can also be performed using open-source software such as Myokit [Clerx et al., 2016], OpenCOR [Garny and Hunter, 2015] and openCARP [Plank et al., 2021]." (Section 2.1)

R III-(d) In paragraph 2.1:

(a) In this sentence: ”Various solver and sampling time steps were applied to generate APs and the biomarkers used in this study (see Appendix A)” this reviewer suggests putting the Appendix reference near “biomarkers”. In addition, a figure that shows the test of various solver vs. sampling time steps could be interesting and can be added to the Appendix as well.

(b) Why did the authors set the relative difference below 5% for all biomarkers? Please give a reference to that choice. Instead, why choose 2% for the time step?

1) We adjusted the reference to be closer to “biomarkers”. While we agree that further details on the influence of the sampling step would be of interest to some of the readers, we feel that it is far beyond the scope of this paper.

2) There is no specific reference we can provide for the choice. Our goal was to reach 5% relative difference, which we surpassed by the chosen time steps of 0.01 ms (solver) and 0.05 ms (sampling), leading to only 2% difference. We rephrased the sentence in question to make this clear.

"We considered the time steps with only 2% relative difference for all AP biomarkers (solver: 0.01ms; sampling: 0.05ms) to offer a sufficiently good approximation." (Section 2.1)

R III-(e) In the caption of Figure 1 authors should include the reference for AP experimental data (are they from Orvos et al. 2019 as reported in the Experimental Data section?)

We added the missing reference as requested. As correctly assumed, they are from [Orvos et al., 2019].

R III-(f) Why do authors not use experimental data in the emulator development/training?

For the supervised training of our NN emulator, we need to provide the maximum conductances of our chosen channels for each AP. While it would be beneficial to also include experimental data in the training to diversify the training data, the exact maximum conductances in our the considered retrospective experiments are not known. In the case such data would be available with low measurement uncertainty, it would be possible to include.

R III-(g) What is TP used in the Appendix B? I could not find the acronymous explanation.

We are sorry for the oversight, TP refers to the time-to-peak and is now described in Appendix A.

R III-(h) Are there any reasons for only using ST and no S1? Maybe are the same?

The global sensitivity analysis is further outlined in Appendix B, also showing S1 (first-order effects) and ST (variance of all interactions) together (Figure 11) [Herman and Usher, 2017] and their differences (e.g. in TP) Since S1 only captures first-order effects, it may fail to capture higher-order interactions between the maximum conductances, thus we favored ST.

R III-(i) In Training Section Page 8. It is not clear why it is necessary to resample data. Can you motivate?

The resampling part is motivated by exactly capturing the swift depolarization dynamics, whereas the output from CARPentry is uniformly sampled. This is now further highlighted in the text.

"Then, the data were non-uniformly resampled from the original uniformly simulated APs, to emphasize the depolarization slope with a high accuracy while lowering the number of repolarization samples. For this purpose, we resamled the APs [...]" (Section 2.2.1)

R III-(j) For the training of the neuronal network, the authors used the ADAM algorithm: have you tested any other algorithm?

For training neural networks, ADAM has become the current de-facto standard and is certainly a robust choice for training our emulator. While there may exist slightly faster, or better-suited training algorithms, we witnessed (qualitative) convergence in the training (Equation (2)). We thus strongly believe that the training algorithm is not a limiting factor in our study.

R III-(k) What is the amount of the drugs tested? Is the same dose reported in the description of the second data set or the values are only referring to experimental data? Moreover, it is not clear if in the description of experimental data, the authors are referring to newly acquired data (since they described in detail the protocol) or if they are obtained from Orvos et al. 2019 work.

In all scenarios, we tested 5 different drugs (cisapride, dofetilide, sotalol, terfenadine, verapamil). We revised our previous presentation of the data available, and now try to give a concise overview over the utilized data (Section 2.2.1 and table 1) and drug comparison with the CiPA distributions (Table 5, former 4). Note that in the latter case, the available expected channel scaling factors by the CiPA distributions vary, but are now clearly shown in Table 5.

R III-(l) In Figure 4, I will avoid the use of “control” in the legend since it is commonly associated with basal conditions and not with the drug administration.

The terminology “control” in this context is in line with works from the CiPA initiative, e.g. [Li et al., 2017] and refers to the state of cell conditions before the drug wash-in. We added a minor note the first time we use the term control in the introduction to emphasize that we refer to the state of the cell before administering any drugs

"To compute the drugged AP for given pharmacological parameters is a forward problem, while the corresponding inverse problem is to find pharmacological parameters for given control (before drug administration) and drugged AP." (Section 1)

R III-(m) In Table 1 when you referred to Britton et al. 2017 work, I suggest adding also 10.1371/journal.pcbi.1002061.

We added the suggested article as a reference.

R III-(n) For the minimization problem, only data set #1 has been used. Have you tested data set #2?

In the current scenario, we only tested the inverse problem for data set #2 (former #1). The main purpose for data set #3 (former #2), was to test the possibility to emulate EAD APs. Given the overall lower performance in comparison to data set #2 (former #1), we also expect deteriorated results in comparison to the existing inverse synthetic problem.

R III-(o) In Figure 6 you should have the same x-axis (we could not see any points in the large time scale for many biomarkers). Why dVmMax is not uniformed distributed compared to the others? Can you comment on that?

As suggested, we re-adjusted the x-range to show the center of distributions. Additionally, we denoted in each subplot the number of outliers which lie outside of the shown range. The error distribution on dVmMax exhibits a slightly off-center, left-tailed normal distribution, which we now describe a bit more in the revised text:

"While the mismatches in phase 3 were simply a result of imperfect emulation, the mismatches in phase 0 were a result of the difficulty in matching the depolarization time exactly. [...] Likewise, the difficulty in exactly matching the depolarization time leads to elevated errors and more outliers in the biomarkers influenced by the depolarization phase (TP and dVmMax)," (Section 3.1.1)

R III-(p) Page 14. Can the authors better clarify ”the average RMSE over all APs 13.6mV”: is it the mean for all histograms in Figure 7? (In Figure 5 is more evident the average RMSE).

The average RMSE uses the same definition for Figures 5 and 7: It is the average over all the RMSEs for each pair of traces (simulated/emulated), though the amount of samples is much lower for the EAD data set and not normal distributed.

R III-(q) In Table 4, the information on which drugs are considered should be added. For each channel, we added the names of the drugs for which respective data from the CiPA initiative were available.

R III-(r) Pag. 18, second paragraph, there is a repetition of ”and”.

Fixed

R III-(s) The pair’s combination of scaling factors for simulating synthetic drugs reported in Table 2, can be associated with some effects of real drugs? In this case, I suggest including the information or justifying the choice.

The scaling factors in Table 2 are used to create data set #3 (former #2), and is meant to provide several APs which expose EADs. This is described in more detail in the new data section, Section 2.2.1:

"Data set #3: The motivation for creating data set #3 was to test the emulator on data of abnormal APs showing the repolarization abnormality EAD. This is considered a particularly relevant AP abnormality in pharmacological studies because of their role in the genesis of drug-induced ventricular arrhythmia’s [Weiss et al., 2010]. Drug data were created using ten synthetic drugs with the hERG channel and the Cav1.2 channel as targets. To this end, ten samples with pharmacological parameters for GKr and PCa (Table 2) were generated and the synthetic drugs were applied to the entire synthetic cardiomyocyte population by scaling GKr and PCa with the corresponding pharmacological parameter. Of the 1000 APs simulated, we discarded APs with a transmembrane potential difference of more than 10% of the amplitude between t = 0 and 1000ms (checked for the last AP), indicative of an AP that does not repolarize within 1000ms. This left us with 950 APs, 171 of which exhibit EAD (see Appendix C)." (Section 2.2.1)

R III-(t) A general comment on the work is that the authors claim that their study highlights the potential of NN emulators as a powerful tool for increased efficiency in future quantitative systems pharmacology studies, but they wrote ”Larger inaccuracies were found in the inverse problem solutions on experimental data highlight inaccuracies in estimating the pharmacological parameters”: so, I was wondering how they can claim the robustness of NN use as a tool for more efficient computation in pharmacological studies.

The discussed robustness directly refers to efficiently emulating steady-state/limit cycle APs from a set of maximum conductances (forward problem, Section 3.1.1). We extensively evaluated the algorithm and feel that given the low emulation RMSE of APs (< 1 mV), the statement is warranted. The inverse estimation, enabled through this rapid evaluation, performs well on synthetic data, but shows difficulties for experimental data. Note however that at this point there are multiple potential sources for these problems as highlighted in the Evaluation section (Section 4.1) and Table 5 (former 4) highlights the difference in accuracy of estimating per-channel maximum conductances, revealing a potentially large discrepancy. The emulator also offers future possibilities to incorporate additional informations in the forms of either priors, or more detailed measurements (e.g. calcium transients) and can be potentially improved to a point where also the inverse problem can be satisfactorily solved in experimental preparations, though further analysis will be required.

References [Beck, 2017] Beck, A. (2017). First-order methods in optimization. SIAM.

[Britton et al., 2013] Britton, O. J., Bueno-Orovio, A., Ammel, K. V., Lu, H. R., Towart, R., Gallacher, D. J., and Rodriguez, B. (2013). Experimentally calibrated population of models predicts and explains intersubject variability in cardiac cellular electrophysiology. Proceedings of the National Academy of Sciences, 110(23).

[Chang et al., 2017] Chang, K. C., Dutta, S., Mirams, G. R., Beattie, K. A., Sheng, J., Tran, P. N., Wu, M., Wu, W. W., Colatsky, T., Strauss, D. G., and Li, Z. (2017). Uncertainty quantification reveals the importance of data variability and experimental design considerations for in silico proarrhythmia risk assessment. Frontiers in Physiology, 8.

[Clerx et al., 2016] Clerx, M., Collins, P., de Lange, E., and Volders, P. G. A. (2016). Myokit: A simple interface to cardiac cellular electrophysiology. Progress in Biophysics and Molecular Biology, 120(1):100–114.

[Endresen and Skarland, 2000] Endresen, L. and Skarland, N. (2000). Limit cycle oscillations in pacemaker cells. IEEE Transactions on Biomedical Engineering, 47(8):1134–1137.

[Garny and Hunter, 2015] Garny, A. and Hunter, P. J. (2015). OpenCOR: a modular and interoperable approach to computational biology. Frontiers in Physiology, 6.

[Gemmell et al., 2016] Gemmell, P., Burrage, K., Rodr´ıguez, B., and Quinn, T. A. (2016). Rabbit-specific computational modelling of ventricular cell electrophysiology: Using populations of models to explore variability in the response to ischemia. Progress in Biophysics and Molecular Biology, 121(2):169–184.

[Ghosh et al., 2018] Ghosh, S., Gavaghan, D. J., and Mirams, G. R. (2018). Gaussian process emulation for discontinuous response surfaces with applications for cardiac electrophysiology models.

[Herman and Usher, 2017] Herman, J. and Usher, W. (2017). SALib: An open-source python library for sensitivity analysis. J. Open Source Softw., 2(9):97.

[Johnstone et al., 2016] Johnstone, R. H., Chang, E. T., Bardenet, R., de Boer, T. P., Gavaghan, D. J., Pathmanathan, P., Clayton, R. H., and Mirams, G. R. (2016). Uncertainty and variability in models of the cardiac action potential: Can we build trustworthy models? Journal of Molecular and Cellular Cardiology, 96:49–62.

[Li et al., 2017] Li, Z., Dutta, S., Sheng, J., Tran, P. N., Wu, W., Chang, K., Mdluli, T., Strauss, D. G., and Colatsky, T. (2017). Improving the in silico assessment of proarrhythmia risk by combining hERG (human ether`a-go-go-related gene) channel–drug binding kinetics and multichannel pharmacology. Circulation: Arrhythmia and Electrophysiology, 10(2).

[Muszkiewicz et al., 2016] Muszkiewicz, A., Britton, O. J., Gemmell, P., Passini, E., S´anchez, C., Zhou, X., Carusi, A., Quinn, T. A., Burrage, K., Bueno-Orovio, A., and Rodriguez, B. (2016). Variability in cardiac electrophysiology: Using experimentally-calibrated populations of models to move beyond the single virtual physiological human paradigm. Progress in Biophysics and Molecular Biology, 120(1):115–127.

[Orvos et al., 2019] Orvos, P., Kohajda, Z., Szlov´ak, J., Gazdag, P., Arp´adffy-Lovas, T., T´oth, D., Geramipour, A.,´ T´alosi, L., Jost, N., Varr´o, A., and Vir´ag, L. (2019). Evaluation of possible proarrhythmic potency: Comparison of the effect of dofetilide, cisapride, sotalol, terfenadine, and verapamil on hERG and native iKr currents and on cardiac action potential. Toxicological Sciences, 168(2):365–380.

[Paszke et al., 2019] Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Kopf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., and Chintala, S. (2019). PyTorch: An Imperative Style, High-Performance Deep Learning Library. In Advances in Neural Information Processing Systems, volume 32. Curran Associates, Inc.

[Plank et al., 2021] Plank, G., Loewe, A., Neic, A., Augustin, C., Huang, Y.-L., Gsell, M. A., Karabelas, E., Nothstein, M., Prassl, A. J., S´anchez, J., Seemann, G., and Vigmond, E. J. (2021). The openCARP simulation environment for cardiac electrophysiology. Computer Methods and Programs in Biomedicine, 208:106223.

[Shahriari et al., 2016] Shahriari, B., Swersky, K., Wang, Z., Adams, R. P., and de Freitas, N. (2016). Taking the Human Out of the Loop: A Review of Bayesian Optimization. Proceedings of the IEEE, 104(1):148–175. Conference Name: Proceedings of the IEEE.

[Snoek et al., 2014] Snoek, J., Swersky, K., Zemel, R., and Adams, R. (2014). Input Warping for Bayesian Optimization of Non-Stationary Functions. In Proceedings of the 31st International Conference on Machine Learning, pages 1674–1682. PMLR. ISSN: 1938-7228.

[Weiss et al., 2010] Weiss, J. N., Garfinkel, A., Karagueuzian, H. S., Chen, P.-S., and Qu, Z. (2010). Early afterdepolarizations and cardiac arrhythmias. Heart Rhythm, 7(12):1891–1899.

Author Response

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

eLife assessment

The study by Ghafari et al. addresses a question that is highly relevant for the field of attention as it connects structural differences in subcortical regions with oscillatory modulations during attention allocation. Using a combination of magnetoencephalography (MEG) and magnetic resonance imaging (MRI) data in human subjects, inter-individual differences in the lateralization of alpha oscillations are explained by asymmetry of subcortical brain regions. The results are important, and the strength of the evidence is convincing. Yet, clarifying the rationale, reporting the data in full, a more comprehensive analysis, and a more detailed discussion of the implications will strengthen the manuscript further.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors re-analysed the data of a previous study in order to investigate the relation between asymmetries of subcortical brain structures and the hemispheric lateralization of alpha oscillations during visual spatial attention. The visual spatial attention task crossed the factors of target load and distractor salience, which made it possible to also test the specificity of the relation of subcortical asymmetries to lateralized alpha oscillations for specific attentional load conditions. Asymmetry of globus pallidus, caudate nucleus, and thalamus explained inter-individual differences in attentional alpha modulation in the left versus right hemisphere. Multivariate regression analysis revealed that the explanatory potential of these regions' asymmetries varies as a function of target load and distractor salience.

Strengths:

The analysis pipeline is straightforward and follows in large parts what the authors have previously used in Mazzetti et al (2019). The authors use an interesting study design, which allows for testing of effects specific to different dimensions of attentional load (target load/distractor salience). The results are largely convincing and in part replicate what has previously been shown. The article is well-written and easy to follow.

We thank the reviewer for their interest in our study.

Weaknesses:

While the article is interesting to read for researchers studying alpha oscillations in spatial attention, I am somewhat sceptical about whether this article is of high interest to a broader readership. Although I read the article with interest, the conceptual advance made here can be considered mostly incremental. As the authors describe, the present study's main advance is that it does not include reward associations (as in previous work) and includes different levels of attentional load. While these design features and the obtained results indeed improve our general understanding of how asymmetries of subcortical structures relate to lateralized alpha oscillations, the conceptual advance is somewhat limited.

We thank the reviewer for their constructive comment. We’d like to highlight that this is the first study to show relationship between subcortical structures asymmetry with attention-modulated alpha oscillation that did not involve any reward-associations- which is the most studied role of basal ganglia. We also believe there is value is having a second study linking the asymmetry in volume of subcortical structures to the modulation of alpha oscillations as this surprising finding also have important clinical implications (see below). We edited the manuscript as below to explain the advances made in this study:

Introduction (Line 112): “Our current findings broaden our understanding of how subcortical structures are involved in modulating alpha oscillations during top-down spatial attention, in the absence of any reward or value associations. “

Discussion (Line 301): “It has also been shown that the spatial extent of pathological change in subcortical structures can predict cognitive changes in Parkinson’s Disease (43). […] Changes in neocortical oscillatory activity have also been observed in neurological disorders which mainly are known to affect subcortical structures. For example, individuals with Alzheimer's Disease demonstrate an increase in slow oscillatory activities and a decrease in higher frequency oscillations (45). Moreover, in patients with Parkinson’s Disease, the power of beta oscillations increases relatively to when they are dopamine-depleted compared with when they are on dopaminergic medication (46).”

While the analysis of the relation of individual subcortical structures to alpha lateralization in different attentional load conditions is interesting, I am not convinced that the present analysis is suited to draw strong conclusions about the subcortical regions' specificity. For example, the Thalamus (Fig. 5) shows a significant negative beta estimate only in one condition (low-load target, non-salient distractor) but not in the other conditions. However, the actual specificity of the relation of thalamus asymmetry to lateralized alpha oscillations would require that the beta estimate for this one condition is significantly higher than the beta estimates for the other three conditions, which has not been tested as far as I understand.

We thank the reviewer for this constructive comment. We agree with the reviewer that we should compare the beta value amongst the conditions. We therefore determined to better harness the multivariate nature of our analysis. Multivariate regression analysis allows one to test the null hypothesis that a given predictor does not contribute to all the dependent variables. A rejection of this hypothesis would suggest that lateralization of a given region of interest significantly predicts variability across all 4 of the task conditions, whereas failure to reject the null would imply that the predictive relationship holds only for that single condition. We tested this global null hypothesis using a MANOVA test and found the following which we have added to the manuscript:

Results (Line 250): “To ascertain whether each predictor contributes to all conditions, we conducted statistical tests on the results of our MMR using the null hypothesis that a given regressor does not impact all dependent variables. We found that while, with marginal significancy, caudate nucleus can predict variability across all four of the task conditions (F(26,4) = 2.82, p-value = 0.046), the predictive relationships of thalamus (F(26,4) = 2.43, p-value = 0.073) with condition 1, and globus pallidus (F(26,4) = 2.29, p-value = 0.087) with conditions 2 and 3 hold only for these conditions. In sum, this demonstrates that when the task is easiest (condition 1), the thalamus is related to alpha modulation. When the task is most difficult (condition 4), the caudate nucleus relates to the alpha modulation, however, its contributions are substantial enough to predict outcomes across all conditions. For the conditions with medium difficulty (conditions 2 and 3) the globus pallidus is related to the alpha band modulation. “

Method (Line 599): “To examine the specificity of each regressor for lateralized alpha in each condition, we statistically assessed the results of the MMR against the null hypothesis that a particular predictor does not contribute to all dependent variables, employing a MANOVA test in RStudio (version 2022.02.2) (80).”

Discussion (Line 337): “Thalamus, Globus Pallidus, and Caudate nucleus play varying roles across different load conditions.”

Discussion (Line 361): “Although these findings highlight the varying contributions of different regions, they do not imply a lack of evidence for correlations between these subcortical structures and other load conditions.”

Discussion (Line 379): “Additionally, we refrained from directly comparing the contributions of subcortical structures to different conditions due to low statistical power. […] In future studies it would be interesting to design an experiment directly addressing which subcortical regions contribute to distractor and target load in terms of modulating the alpha band activity. In order to ensure sufficient statistical power for doing so possibly each factor needs to be addressed in different experiments.”

Reviewer #3 (Public Review):

Summary:

In this study, Ghafari et al. explored the correlation between hemispheric asymmetry in the volume of various subcortical regions and lateralization of posterior alpha-band oscillations in a spatial attention task with varying cognitive demands. To this end, they combined structural MRI and task MEG to investigate the relationship between hemispheric differences in the volume of basal ganglia, thalamus, hippocampus, and amygdala and hemisphere-specific modulation of alpha-band power. The authors report that differences in the thalamus, caudate nucleus, and globus pallidus volume are linked to the attention-related changes in alpha band oscillations with differential correlations for different regions in different conditions of the design (depending on the salience of the distractor and/or the target).

Strengths:

The manuscript contributes to filling an important gap in current research on attention allocation which commonly focuses exclusively on cortical structures. Because it is not possible to reliably measure subcortical activity with non-invasive electrophysiological methods, they correlate volumetric measurements of the relevant subcortical regions with cortical measurements of alpha band power. Specifically, they build on their own previous finding showing a correlation between hemispheric asymmetry of basal ganglia volumes and alpha lateralization by assessing a task without an explicit reward component. Furthermore, the authors use differences in saliency and perceptual load to disentangle the individual contributions of the subcortical regions.

We appreciate the reviewer’s interest in our study.

Weaknesses:

The theoretical bases of several aspects of the design and analyses remain unclear. Specifically, we missed statements in the introduction about why it is reasonable, from a theoretical perspective, to expect:

(i) a link between volumetric measurements and task activity;

We thank the reviewer for this constructive feedback. We have now addressed this concern in the revised manuscript.

Discussion (Line 293): “It has been demonstrated that extensive navigation experience enlarges the size of right hippocampus (40). Furthermore, in terms of neurological disorders, it is well established that shrinkage (atrophy) in specific regions is a predictor of a number of neurological and psychiatric conditions including Parkinson’s disease, dementia, and Huntington’s disease. […] It has also been shown that the spatial extent of pathological change in subcortical structures can predict cognitive changes in Parkinson’s Disease (43). […] Changes in neocortical oscillatory activity have also been observed in neurological disorders which mainly are known to affect subcortical structures. For example, individuals with Alzheimer's Disease demonstrate an increase in slow oscillatory activities and a decrease in higher frequency oscillations (45). Moreover, in patients with Parkinson’s Disease, the power of beta oscillations increase relatively to when they are dopamine-depleted compared with when they are on dopaminergic medication (46). “

(ii) a specific link with hemispheric asymmetry in subcortical structures (While focusing on hemispheric lateralization might circumvent the problem of differences in head size, it would be better to justify this focus theoretically, which requires for example a short review of evidence showing ipsilateral vs contralateral connections between the relevant subcortical and cortical structures);

We thank the reviewer for this helpful comment that resulted in clarification of the manuscript. We addressed this issue in the revised manuscript; we also now have complemented the revised manuscript with papers directly investigating asymmetry of subcortical regions in relation to neurological disorders:

Introduction (Line 102): “We utilized the hemispheric laterality of subcortical structures and alpha modulation to overcome issues related to individual variations in oscillatory power and head size.”

Discussion (Line 314): “Employing hemispheric lateralization was motivated by the organizational characteristic of structural asymmetry in healthy brain (47). Additionally, considering the effects of aging (48) and neurodegenerative disorders, such as Alzheimer's Disease (49), on brain symmetry influenced this approach. Furthermore, computing lateralization indices for individuals addresses the challenge of accommodating variations in both head size and the power of oscillatory activity.”

Discussion (Line 374): “Furthermore, in this study, our emphasis has been on assessing the size of subcortical structures. Future investigations could explore subcortical white matter connectivities and hemispheric asymmetries. This approach has previously been conducted on superior longitudinal fasciculus (SLF) (61,62) and holds potential for examining cortico-subcortical connectivity in the context of oscillatory asymmetries.”

(iii) effects not only in basal ganglia and thalamus, but also hippocampus and amygdala (a justification of selection of all ROIs);

We thank the reviewer for this comment. We assessed the hippocampus and amygdala because they are automatically segmented in the FIRST algorithm. As our analysis showed they did not show a relation to the modulation of alpha oscillations, these regions also provide a useful control for our approach. Therefore, we included all subcortical structures in the model and evaluated their predictive impact. This is now addressed in the revised manuscript.

Method (Line 477): “FIRST is an automated model-based tool that runs a two-stage affine transformation to MNI152 space, to achieve a robust pre-alignment of thalamus, caudate nucleus, putamen, globus pallidus, hippocampus, amygdala, and nucleus accumbens based on individual’s T1-weighted MR images.”

Method (Line 576): “The absence of a relationship between modulations of alpha oscillations and the hippocampus and amygdala was expected as these regions typically are not associated with the allocation of spatial attention and thus add validity to our approach. “

(iv) effects that depend on distractor versus target salience (a rationale for the specific two-factor design is missing);

We thank the reviewer for this comment that helped us clarify the manuscript. The two-factor design is to investigate how allocation of attentional resources specifically relates to mechanisms of excitability and suppression mechanism. For this reason, both the salience of the distractor (associated with suppression) and the perceptual load of the target (associated with excitability) had to be manipulated. We clarified the rationale in the revised version as below:

Introduction (Line 96): “We analyzed MEG and structural data from a previous study (27), in which spatial cues guided participants to covertly attend to one stimulus (target) and ignore the other (distractor). To investigate the relationship between the allocation of attentional resources and mechanisms of neural excitability and suppression, the target load and the visual saliency of the distractor were manipulated using a noise mask. This load/salience manipulation resulted in four conditions that affect the attentional demands of target and distractor.”

(v) effects in the absence of reward (why it is important to show that the effect seen previously in a task with reward is seen also in a task without reward);

We thank the reviewer for this clarification comment. We addressed this question in introduction and discussion as below:

Introduction (Line 107): “By examining their role in a task without explicit reward, we aim to elucidate the generalizability of the contributions of subcortical structures to spatial attention modulation. Such a finding would implicate a role for the basal ganglia in cognition beyond the well-studied realm of the estimation of choice values (33). Specifically, in a prior study (28), we observed that the contributions of the basal ganglia were most pronounced when the items in question were associated with a reward. Our current findings broaden our understanding of how subcortical structures are involved in modulating alpha oscillations during top-down spatial attention, in the absence of any reward or value associations. “

Discussion (Line 333): “This convergence of results not only corroborates the validity and consistency of our findings but also extends the empirical foundation supporting the predictive role of the asymmetry of globus pallidus in modulating alpha oscillations beyond reward valence and to the context of attention.”

(vi) effects on rapid frequency tagging.

We thank the reviewer for this constructive comment. We have now included this analysis and added the results to the revised manuscript.

Results (Line 224): “It is worth noting that neither the behavioural nor the rapid invisible frequency tagging (RIFT) measures showed significant relationships with LVs and HLM() (Supplementary material, Figure 1 and Table 3).”

Discussion (Line 396): “We did not find any association between the power of RIFT signal and the size asymmetry of subcortical structures. Since to Bayes factors were less than 0.1, we conclude that our RIFT null findings are robust, suggesting a dissociation between how alpha oscillations and neuronal excitability indexed by RIFT relate to subcortical structures.”

Method (Line 548): “We computed the modulation index (MI) for rapid invisible frequency tagging (RIFT) by averaging the power of the signal in sensors on the right when attention was directed to the right compared to when it was directed to the left. This calculation was also performed for sensors on the left. Consequently, we identified the top 5 sensors on each side with the highest MI as the Region of Interest (ROI). Utilizing the sensors within the ROI, we computed hemispheric lateralization modulation (HLM) of RIFT by summing the average MI(RIFT) of the right sensors and the average MI(RIFT) of the left sensors, obtaining one HLM(RIFT) value for each participant. For a more comprehensive analysis, refer to reference (24).”

Supplementary Materials (Line 839): “Figure 1. Lateralization volume of thalamus, caudate nucleus and globus pallidus in relation to hemispheric lateralization modulation of rapid invisible frequency tagging (HLM(RIFT)) on the right and behavioural asymmetry on the left. A and E, The beta coefficients for the best model (having three regressors) associated with a generalized linear model (GLM) where lateralization volume (LV) values were defined as explanatory variables for HLM(RIFT) (A) and behavioural asymmetry (E). Error bars indicate standard errors of mean (SEM). B and F, Partial regression plot showing the association between LVTh and HLM(RIFT) (B, p-value = 0.59) and behavioural asymmetry (F, p-value = 0.38) while controlling for LVGP and LVCN. C and G, Partial regression plot showing the association between LVGP and HLM(RIFT) (C, p-value = 0.16) and behavioural asymmetry (G, p-value = 0.80) while controlling for LVTh and LVCN . D and H, Partial regression plot showing the association between LVCN and HLM(RIFT) (D, p-value = 0.53) and behavioural asymmetry (H, p-value = 0.74) while controlling for LVTh and LVGP. Negative (or positive) LVs indices denote greater left (or right) volume for a given substructure; similarly negative HLM(RIFT) values indicate stronger modulation of RIFT power in the left compared with the right hemisphere, and vice versa; positive behavioural asymmetry value shows higher accuracy when the target was on the right as compared with left, and vice versa for negative behavioural asymmetry values. The dotted curves in B, C, D, F, G, and H indicate 95% confidence bounds for the regression line fitted on the plot in red.

Author response image 1.

Second, the results are not fully reported. The model space and the results from the model comparison are omitted. Behavioral data and rapid frequency tagging results are not shown. Without having access to the data or the results of the analyses, the reader cannot evaluate whether the null effect corresponds to the absence of evidence or (as claimed in the discussion) evidence of absence.

We thank the reviewer for this constructive suggestion. In the revised manuscript, we incorporated the model space, model comparisons, BIC values from the models, behavioral and rapid frequency tagging analysis methods, and their respective results. Additionally, we computed Bayes factors for our null findings to enhance the interpretability of our results.

Results (Line 199): “This model predicted the HLM(α) values significantly in the GLM (F3,29 = 7.4824, p = 0.0007, adjusted R2 = 0.376) as compared with an intercept-only null model (Figure 4A).”

Although, the beta estimate of LVGP only showed a positive trend, removing it from the regression resulted in worse models (AIC and BIC tables in supplementary material).

Supplementary materials (Line 827): “Table 1. Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values for all possible combinations of regressors (Lateralized Volume of subcortical structures). The selected model, with lowest AIC, is marked in green.

Author response table 1.

Author response table 2.

Author response table 3.

Bayes factors for correlation between hemispheric laterality of subcortical structures with hemispheric lateralization modulation of rapid invisible frequency tagging (HLM(RIFT)) and with behavioural asymmetry (BA). The Pearson correlation between each subcortical structure with HLM(RIFT) and behavioural asymmetry was calculated. The likelihood of the data under the alternative hypothesis (the evidence of correlation) were subsequently compared to the likelihood under null hypothesis (absence of correlation), given the data. As it is demonstrated in the table, all Bayes factors were below or very close to 1 indicating evidence for the null hypothesis.

For the results of frequency tagging signal, we have now included this analysis and added the results to the revised manuscript. We refer the reviewer to our response to the weakness (vi) from reviewer #3.

Third, it remains unclear whether the MMS is the best approach to analyzing effects as a function of target and distractor salience. To address the question of whether the effects of subcortical volumes on alpha lateralization vary with task demands (which we assume is the primary research question of interest, given the factorial design), we would like to evaluate some sort of omnibus interaction effect, e.g., by having target and distractor saliency interact with the subcortical volume factors to predict alpha lateralization. Without such analyses, the results are very hard to interpret. What are the implications of finding the differential effects of the different volumes for the different task conditions without directly assessing the effect of the task manipulation? Moreover, the report would benefit from a further breakdown of the effects into simple effects on unattended and attended alpha, to evaluate whether effects as a function of distractor (vs target) salience are indeed accompanied by effects on unattended (vs attended) alpha.

The reviewer is correct that we did not directly compare between task conditions when we assessed the predictive relationship between basal ganglia lateralization and alpha lateralization. We opted for the multivariate regression approach as this allowed us to simultaneously model the predictive relationship between our continuous predictors and HLM alpha in each condition, allowing us to be most efficient with our level of statistical power (N=33). Indeed, directly comparing between task conditions within one model would result in an extra 16 regressors (1 (intercept) + 4-1 to model the difference between conditions + 3 to model the regressors + 3 x 3 to model each region x task condition interaction). This approach would be underpowered given our sample size, and the ensuing results are likely to be unreliable.

However, we statistically analysed our regression results. Multivariate regression analysis allows one to test the null hypothesis that a given predictor does not contribute to all the dependent variables. A rejection of this hypothesis would suggest that lateralization of a given region of interest significantly predicts variability across all 4 of the task conditions, whereas failure to reject the null would imply that the predictive relationship holds only for that single condition. We tested this global null hypothesis using a MANOVA test and reported the findings in response to weakness two from reviewer #1.

Discussion (Line 384): “In future studies it would be interesting to design an experiment directly addressing which subcortical regions contribute to distractor and target load in terms of modulating the alpha band activity. In order to ensure sufficient statistical power for doing so possibly each factor needs to be addressed in different experiments. “

The fourth concern is that the discussion section is not quite ready to help the reader appreciate the implications of key aspects of the findings. What are the implications for our understanding of the roles of different subcortical structures in the various psychological component processes of spatial attention? Why does the volumetric asymmetry of different subcortical structures have diametrically opposite effects on alpha lateralization? Instead, the discussion section highlights that the different subcortical structures are connected in circuits: "Globus pallidus also has wide projections to the thalamus and can thereby impact the dorsal attentional networks by modulating prefrontal activities." If this is true, then why does the effect of the GP dissociate from that of the thalamus? Also, what is it about the current behavioural paradigm that makes the behavioral readout insensitive to variation in subcortical volume (or alpha lateralization?)?

We thank the reviewer for this feedback. These are indeed all good points, and we hope that our findings will inspire further research to address these issues. In the revised manuscript we now write:

Discussion (Line 349): “The opposite effect of the globus pallidus compared to the thalamus is striking, and possibly explained but the globus pallidus containing GABAergic interneurons. Thus the inhibitory nature of the globus pallidus projections to thalamus could explain why they are related to the alpha modulation in different manners (57).”

Discussion (Line 379): “Moreover, the current study faced methodological constraints, limiting the analysis to the entire thalamus. […] . It would be of great interest to conduct further investigations to quantify the distinct impacts of individual thalamic nuclei on the association between subcortical structures and the modulation of oscillatory activity.“

Discussion (Line 388): “Moreover, our failure to identify a relationship between the lateralized volume of subcortical structures and behavioural measures should be addressed in studies that are better designed to capture performance asymmetries (63). Individual preferences toward one hemifield, which were not addressed in the current study design, could potentially strengthen the power to detect correlations between structural variations in the subcortical structures and behavioural measures.”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Minor comment:

Between-subject correlation/regression analyses always rely on the assumption that the underlying dependent measures are reliable. While the reliability of asymmetries of subcortical structures can be assumed, the reliability of lateralized alpha oscillations during spatial attention can be questioned. It would be helpful if the authors could test the reliability of alpha lateralization, for instance by calculating HLM(a) in the first and second half of the experiment and correlating the resulting HLM(a) values (split-half reliability).

We appreciate the reviewer for their insightful comment. Acknowledging that the between-subject regression relies on the reliability of alpha lateralization. Nonetheless, a previous study has demonstrated consistent results regarding HLM(α). We have further elaborated on these aspects in the discussion section:

Discussion (Line 328): “Furthermore, our regression analysis outcomes align with the findings of Mazzetti et al. (28) underscoring the significant predictive influence exerted by the lateralized volume of globus pallidus on the modulation of hemispheric lateralization in alpha oscillations during spatial attention tasks. This convergence of results not only corroborates the validity and consistency of our findings but also extends the empirical foundation supporting the predictive role of the asymmetry of globus pallidus in modulating alpha oscillations within the context of attention.”

Reviewer #3 (Recommendations For The Authors):

We recommend that a revised version of the manuscript

• Clarifies the theoretical basis for the 6 key design & analysis choices that we have outlined above;

We thank the reviewer for their precision. We addressed the concerns outlined above in the previous section.

• Also clarifies the task description (perhaps referring to target and distractor salience instead of target load versus distractor salience might help);

Thank you for this constructive comment. We used the terms ‘load’ for target and ‘salience’ for distractor because the noise manipulation of the faces reduces the salience of the image which results in distractors being less distractive (easier) but targets being more perceptually loaded (harder). The explanation of these terms is made clear in the revised manuscript.

Method (Line 447): “Over trials, the perceptual load of targets was manipulated using a noise mask; noisy targets are harder to detect than clear targets and therefore incur greater perceptual load in their detection. The saliency of distractor stimuli was also manipulated using a noise mask; noisy distractor stimuli are less salient than clear distractors and therefore less disruptive to performance on the detection task. The noise mask was created by randomly swapping 50% of the stimulus pixels (Figure 1B). This manipulation resulted in four target-load/distractor-saliency conditions: (1) target: low load, distractor: low saliency (i.e., clear target, noisy distractor), (2) target: high load, distractor: low saliency (i.e., noisy target, noisy distractor), (3) target: low load, distractor: high saliency (i.e., clear target, clear distractor), (4) target: high load, distractor: high saliency (i.e., noisy target, clear distractor) (Figure 1B and C).”

• Fully reports all the data, including those of the model comparisons, the behavioural results, and the rapid frequency tagging results;

We thank the reviewer for this constructive comment. We refer the reviewer to our response to second comment and comment (vi) from reviewer #3.

• Reports interaction effects to directly test the modulating role of task demands in the link between volume and alpha, and break down the alpha lateralization indices into their simple effects on the ipsilateral and contralateral hemispheres;

task demands have been addressed in response to in response to weakness two from reviewer #1.

Regarding the second part of the comment, in our study, to compare the lateralized modulation of alpha oscillations between the right and left hemispheres, we computed hemispheric lateralization modulation. This involved dividing trials into attention right and attention left. Subsequently, we calculated the lateralization index separately for sensors on the right and left. Specifically, this entailed computing ipsilateral – contralateral for sensors on the right and contralateral – ipsilateral for sensors on the left side of the brain. We addressed this concern in methods section as below:

Method (Line 537): “As MI(α) consistently represents power of alpha in attention right versus attention left conditions, it entails the comparison between ipsilateral and contralateral alpha modulation power for sensors located on the right side of the head. The same comparison applies inversely for sensors situated on the left side of the brain.”

• Clarifies in the discussion section the specific implications of the results for our understanding of the link between distinct subcortical structures and distinct component processes of spatial attention.

We thank the reviewer for their constructive comment. This point is addressed in response to the fourth concern of reviewer #3.

More detailed specific recommendations are provided below:

• Line 40ff: In this paragraph, the theoretical framework concerning the function of the subcortical regions of interest is described. Here, the authors jump back and forth between the role of the basal ganglia and the role of the thalamus. For clarity, we would advise to describe the functions of these two structures one after the other. And include a justification for assessing the hippocampus and the amygdala.

We appreciate the reviewer’s preciseness in this comment. We put the description of these structures one after the other in the revised manuscript as below:

Introduction (Line 44): “For instance, it has been shown that the pulvinar plays an important role in the modulation of neocortical alpha oscillations associated with the allocation of attention (9). Studies in rats and non-human primates have shown that both the thalamus and superior colliculus, are involved in the control of spatial attention by contributing to the regulation of neocortical activity (9-11). Notably, when the largest nucleus of the thalamus, the pulvinar, was inactivated after muscimol infusion, the monkey’s ability to detect colour changes in attended stimuli was lowered. This behavioral deficit occurred when the target was in the receptive field of V4 neurons that were connected to lesioned pulvinar (12). The basal ganglia play a role in different aspects of cognitive control, encompassing attention (13,14), behavioural output (15), and conscious perception (16). Moreover, the basal ganglia contribute to visuospatial attention by linking with cortical regions like the prefrontal cortex via the thalamus.”

Justification for assessing the hippocampus and the amygdala has been addressed in response to weakness (iii) from reviewer #3.

• The authors mention they defined symmetric clusters of 5 sensors in each hemisphere that showed the highest modulation, but it is not clear how this number of sensors was determined a priori.

We thank the reviewer for their comment. We edited the revised manuscript as below:

Method (Line 536): “Ten sensors were selected to ensure sufficient coverage of the region exhibiting alpha modulation as judged from prior work (62).”

• In line 141, the abbreviation HLM is first mentioned but the concept of "hemispheric lateralization modulation of alpha power" is only mentioned in the following section. For the ease of the reader, the abbreviation could be mentioned together with this concept at the beginning of this paragraph.

We thank the reviewer for the attention. In the revised manuscript HLM() is now mentioned with its concept.

Results (Line 153): “Next, we computed the hemispheric lateralization modulation of alpha power (HLM()) in each individual.”

• In line 188 of the results section, it is mentioned that the table including the AIC values for model comparisons is in the supplementary material, however, we could not locate this table.

We thank the reviewer for their constructive feedback. The supplementary materials were uploaded in a separate file, and it must not have been available to the reviewers. We have now added the supplementary materials to the end of the manuscript for convenience.

• Figure 4 is missing the panel headers A, B, C, and D.

We thank the reviewer for their precision. This figure is now fixed.

Author response image 2.

• In lines 205 and 206, behavioral and rapid frequency tagging analysis are mentioned. For the behavioral analysis, the method is described, but no results are provided. For the rapid frequency tagging, neither the methods nor the results are described. To evaluate the strength of this (non)-evidence, we would advise to elaborate on these analysis steps and report the results in the supplementary material.

We thank the reviewer for this constructive comment. A brief explanation of the analysis method of rapid frequency tagging signal is added to the revised manuscript.

Method (Line 548): “We computed the modulation index (MI) for rapid invisible frequency tagging (RIFT) by averaging the power of the signal in sensors on the right when attention was directed to the right compared to when it was directed to the left. This calculation was also performed for sensors on the left. Consequently, we identified the top 5 sensors on each side with the highest MI as the Region of Interest (ROI). Utilizing the sensors within the ROI, we computed hemispheric lateralization modulation (HLM) of RIFT by summing the average MI(RIFT) of the right sensors and the average MI(RIFT) of the left sensors, obtaining one HLM(RIFT) value for each participant. For a more comprehensive analysis, refer to reference (24).” For a more detailed answer, we refer the reviewer to the second comment from reviewer #3.

• For the paragraph starting at line 209, we would recommend referring to Figure 1.

We thank the reviewer for their suggestion. This paragraph is now referring to Figure 1.

Results (Line 229): “To relate load and salience conditions of the task to the relationship between subcortical structures and the alpha activity, we combined low-load or high-load targets with high-saliency or low-saliency distractors to manipulate the perceptual load appointed to each trial (Method section, Figure 1). “

• Figure 5 as well as the report of the beta weights in this section shows a difference in the direction of the effect for the thalamus compared to the globus pallidus and caudate nucleus which is not discussed in this section.

We thank the reviewer for bringing this important point to our attention. We addressed this comment in the discussion section as below:

Discussion (Line 349): “The opposite effect of the globus pallidus compared to the thalamus is striking, and possibly explained by the globus pallidus containing GABAergic interneurons. Thus the inhibitory nature of the globus pallidus projections to thalamus could explain why they are related to the alpha modulation in different manners (54).”

Discussion (Line 379): “Moreover, the current study faced methodological constraints, limiting the analysis to the entire thalamus. […] It would be of great interest to conduct further investigations to quantify the distinct impacts of individual thalamic nuclei on the association between subcortical structures and the modulation of oscillatory activity.“

• Comment 2 on line 80 is addressed in the paragraph following 264 by describing volumetric changes in basal ganglia in neurodegenerative disorders such as PD or Huntington's. Still, the link of how a decrease in volume in this region could be causally linked to changes in alpha-band power could be better supported.

We thank the reviewer for their constructive feedback. We are here highlighting the significant correlation between subcortical structures and changes in attention modulated alpha oscillation. We added a few more references to the discussion supporting the relationship between size and function in relation to neurological disorders. We also edited the manuscript to make this point clearer as below:

Introduction (Line 113): “Our current findings broaden our understanding of how subcortical structures are involved in modulating alpha oscillations during top-down spatial attention, independent of any reward or value associations. “

Discussion (Line 305): “Changes in neocortical oscillatory activity have also been observed in neurological disorders which mainly are known to affect subcortical structures. For example, individuals with Alzheimer's Disease demonstrate an increase in slow oscillatory activities and a decrease in higher frequency oscillations (42). Moreover, in patients with Parkinson’s Disease, the power of beta oscillations increases relatively to when they are dopamine-depleted compared with when they are on dopaminergic medication (43). “

• Related to the previous comment on behavioral and rapid frequency tagging results, these are difficult to evaluate without mention of the methods and/or results.

We thank the reviewer for this comment. We refer the reviewer to our response to the second comment from reviewer #3.

• The authors show differential effects of target load and distractor saliency; however, we missed the description of how these two variables differ conceptually as they are both described as contributing to task difficulty and it is not described why we would expect differential effects for these concepts (or in other words, how the authors explain the differential effects).

We thank the reviewer for their comment. Directly comparing between task conditions within one model would result in an extra 16 regressors (1 (intercept) + 4-1 to model the difference between conditions + 3 to model the regressors + 3 x 3 to model each region x task condition interaction). Give our sample size, this study is underpowered to directly compare alpha lateralisation in contralateral versus ipsilateral conditions. For a more detailed answer please refer to our response to weakness two from reviewer #1.

• Line 364ff: Based on the description of the experimental design, it is not clear to us whether participants only had to report on the change in gaze for the stimulus in the cued hemifield.

We thank the reviewer for this comment, which prompted us to clarify the experimental design as below:

Method (Line 440): “Then followed a 1000 ms response interval where participants were asked to respond with their right or left index finger whether the gaze direction of the cued face shifted left or right.”

• Line 47ff: As mentioned above, the AIC table is not included. Further, as it is mentioned that BIC values led to similar results (indicating that they are not identical), it would be valuable to report both AIC and BIC values.

We thank the reviewer for their constructive feedback. The supplementary materials were uploaded in a separate file, and it must not have been available to the reviewers. We have now added the BIC values and attached the supplementary materials to the end of the manuscript for convenience.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary

This article by Zhai et al, investigates sterol transport in bacteria. Synthesis of sterols is rare in bacteria but occurs in some, such as M capsulatus where the sterols are found primarily in the outer membrane. In a previous paper the authors discovered an operon consisting of five genes, with two of these genes encoding demethylases involved in sterol demethylation. In this manuscript, the authors set out to investigate the functions of the other three genes in the operon. Interestingly, through a bioinformatic analysis, they show that they are an inner membrane transporter of the RND family, a periplasmic binding protein, and an outer membrane-associated protein, all potentially involved with lipid transport, so providing a means of transporting the lipids to the outer membrane. These proteins are then extensively investigated through lipid pulldowns, binding analysis on all three, and X-ray crystallography and docking of the latter two.

Strengths

The lipid pulldowns and associated MST binding analysis are convincing, clearly showing that sterols are able to bind to these proteins. The structures of BstB and BstC are high resolution with excellent maps that allow docking studies to be carried out. These structures are distinct from sterol-binding proteins in eukaryotes.

We thank the reviewer for their favorable impression of this work.

Weaknesses

While the docking and molecular dynamics studies are consistent with the binding of sterols to BstB and BstC, this is not backed up particularly well. The MST results of mutants in the binding pocket of BstB have relatively little effect, and while I agree with the authors this may be because of the extensive hydrophobic interactions that the ligand makes with the protein, it is difficult to make any firm conclusions about binding.

We agree with the reviewer that at this point, there is no experimental evidence to define the sterol binding site in BstB. While in the manuscript we allude to the extensive hydrophobic interactions as being especially stabilizing and difficult to eliminate with one or two mutations, we are now also aware that hydrogen-bonding interactions with the polar head of the sterols are quite important (see data on BstC, where disruption of that interaction significantly reduces the equilibrium affinity for sterols). Our MD simulations show that at least 3 protein amino acids can participate in H-bonding with the sterols. Moreover, recent work from our lab show that even ligand site waters can extend an H-bonding network around the polar head of the lipid (Zhai et al., ChemBioChem 2023, 24, e202300156), thereby enabling H-bonding with amino acids that are further away from the ligand site. It is therefore difficult to predict which mutations will sufficiently destabilize the binding. While this question is one we will tackle in future studies focused on obtaining high-resolution substrate-bound structures of BstB or homologs, the findings reported here are still relevant and timely, and we posit will spur the discovery of functional homologs, including some in organisms that are more tractable.

The authors also discuss the possibility of a secondary binding site in BstB based on a slight cavity in domain B next to a flexible loop. This is not backed up in any way and seems unlikely.

The reviewer is correct in that the evidence for this second binding site weak. While the crystallographic structure shows a highly hydrophobic region and the binding studies suggests cooperativity exists in the binding of the 4methylsterol substrate, the docking studies do not strongly support binding at that site. As such, we have clarified in the manuscript that a second hydrophobic cavity is observed, but that its role in ligand interaction remains unexplored.

Reviewer #2 (Public Review):

Summary:

In eukaryotes, sterols are crucial for signaling and regulating membrane fluidity, however, the mechanism governing cholesterol production and transport across the cell membrane in bacteria remains enigmatic. The manuscript by Zhai et al. sheds light on this topic by uncovering three potential cholesterol transport proteins. Through comprehensive bioinformatics analysis, the authors identified three genes bstA, bstB, and bstC encoding proteins which share homology with transporters, periplasmic binding proteins, and periplasmic components superfamily, respectively. Furthermore, the authors confirmed the specific interaction between these three proteins and C-4 methylated sterols and determined the structures of BstB and BstC. Combining these structural insights with molecular dynamics simulation, they postulated several plausible substrate binding sites within each protein.

Strengths:

The authors have identified 3 proteins that seem likely to be involved in sterol transport between the inner and outer membrane. The structures are of high quality, and the sterol binding experiments support a role for these proteins in sterol transport.

We thank the reviewer for this positive view of our work.

Weaknesses:

While the author's model is very plausible, direct evidence for a role of BstABC in transport, or that the 3 proteins function together in a single pathway, is limited.

The reviewer is correct that we were unable to demonstrate that the three proteins work together to transport 4methylsterols. This is not for lack of trying. We first attempted gene deletion studies, and as mentioned in the manuscript (with more details now provided in the experimental section), this appeared to be lethal. We then attempted in vitro exchange experiments, in which the proteins would be used to transfer sterols from sterol-loaded “heavy” liposomes to a sterol-free “light” liposomes – such exchange assays are frequently performed with eukaryotic sterol transporters (see Chung et al., Science 2015, https://doi.org/10.1126/science.aab1370). These assays were not successful because 1) sterols incorporated poorly into liposomes made with E. coli polar lipids and yielded leaky liposomes; 2) use of liposomes prepared with the TLE of M. capsulatus proved more stable, but no appreciable exchange was observed; we reasoned that this might be due to the absence of an energy source for BstA, the RND component for which we have expressed and purified only the soluble periplasmic domain. Given the technical difficulty of these in vitro transport experiments, we will continue to pursue in vivo demonstration of function as new homologs are identified.

Reviewer #3 (Public Review):

Summary:

The work in this manuscript builds on prior efforts by this team to understand how sterols are biosynthesized and utilized in bacteria. The study reports a new function for three genes encoded near sterol biosynthesis enzymes, suggesting the resulting proteins function as a sterol transport system. Biochemical and structural characterization of the two soluble components of the pathway establishes that both proteins can bind sterols, with a preference for 4methylated derivatives. High-resolution x-ray structures of the apoproteins reveal hydrophobic cavities of the appropriate size to accommodate these substrates. Docking and molecular dynamics simulations confirm this observation and provide specific insights into residues involved in substrate binding.

Strengths:

The manuscript is comprehensive and well-written. The annotation of a new function in a set of proteins related to bacterial sterol usage is exciting and likely to enable further study of this phenomenon - which is currently not well understood. The work also has implications for improving our understanding of lipid usage in general among bacterial organisms.

We thank the reviewer for this synopsis of our work.

Weaknesses:

The authors might consider moving some of the bioinformatics figures to the main text, given how much space is devoted to this topic in the results section.

We have taken this advice and moved Figure S1 to the main manuscript.

Reviewer #1 (Recommendations For The Authors):

1. In the analysis of the MST data, the authors quote Hill coefficients. How reliable are these numbers? For BstB, for instance, it seems unlikely that more than one molecule would bind. Can the analysis be done without needing to include Hill coefficients?

We used fits that did and did not invoke cooperativity – see below. We are certain that both BstA and BstB are better fit with cooperativity invoked.

Author response image 1.

1. In looking at the maps associated with the structures, which were included in the review package, I see that two citric acid molecules fit beautifully into the density where currently PEG has been modelled. This needs to be fixed and some comments may be appropriate in the manuscript.

We thank the reviewer for calling our attention to this. Citric acid has now been added to the model, and we reason that these are present in the structure because citric acid was used in the crystallization condition. The revised model is now present in the PDB.

1. It is not necessary to show the two molecules in the asymmetric unit in Figure 4 given that it is not a dimer. This doesn't add anything to the manuscript.

We now show a single molecule of BstC in Figure 4 (now Figure 5).

1. I wouldn't consider the loops shown in Figure S4 as disordered. They have slightly higher B-values but are not completely mobile.

We did not refer to these loops as disordered. In the text, we say they “exhibit poor electron densities, suggesting conformational sampling of more than one state (Fig. S4A).”

Reviewer #2 (Recommendations For The Authors):

pg 7, "hinting at an astounding distinction": I might suggest a word other than astounding that conveys how statistically unlikely, unusual, etc. this result is.

Thank you – we have removed “astounding”.

pg 7, paragraph 2: Here the authors show that in the SSN analysis, BstB proteins cluster separately and suggest this implies a distinction in function. However, they also show that PhnD homologs do not cluster separately (distributed across multiple clusters), yet presumably have similar functions. I am not familiar with SSN, but it seems to me that the second statement about PhnD implies that the first statement about BstB might not be valid, i.e., if PhnD doesn't cluster based on function, on what basis can we conclude that BstB does? On what basis does clustering occur in the SSN analysis? Might it be driven by things other than function? This comment also concerns the final paragraph of this section.

The reviewer is correct in that PhnD homologs occupy separate clusters of the SSN. Many of these homologs were crystallized with phosphate-like compounds, but it is possible that they have non-overlapping substrate scopes and are therefore functionally distinct. As for the basis of clustering, the SSN is fully sequence-based. What has been observed is that proteins with highly similar sequences can have similar functions – but this is not always true.

pg 8, paragraph 1: The authors suggest that BstABC may be essential. This is probably not a critical claim and it might be simplest to just remove it, but if it is mentioned, the authors should probably explain what was attempted that failed, so a reader can assess the strength of the evidence supporting essentiality. For example, I don't see anything in the methods about genetic manipulations of M. capsulatus, so currently, this falls within the realm of "Data not shown".

We have provided additional information about the experimental techniques used to do this. This statement was included so that it is understood that the reason for the experimental failure is unlikely to be technical in nature, as we have successfully deleted some sterol related genes while others remain intractable.

Fig. 2A: It is unclear to me what is being plotted here, perhaps more experimental detail is required in the form of labels and/or legend. Is this a quantification of each sterol in each fraction separated by GC? There are essentially no methods provided for the GC-MS experiments. A reference is provided, but I think providing detailed methods for these specific experiments will provide a higher degree of scientific rigor. I am not sure what is standard for GCMS, but perhaps showing spectra in the supplement that establish the identity of the bound molecules as species I and II would be appropriate?

Additional experimental details have been provided and the figure legend changed to be more clear. Moreover, we now clearly state that the chromatograms shown were used to identify lipids due to retention times for spectra that were previously published in Wei et al., 2016.

pg 10-11, comparison with PhnD structure: Perhaps it is worth mentioning a 3rd possible explanation for the relative opening/closing of the cleft is simply crystal packing? I don't think it necessarily has to imply anything about a difference in function. Also, the focus seems to be on this pairwise comparison, but perhaps more insights could be gleaned from an analysis that included a wider range of homologs, especially if any are thought to bind hydrophobic substrates.

This could be true, and we have included a statement to that effect. We are unaware of homologs shown to bind to large, hydrophobic molecules.

I think that BstB is shown upside-down in sup movies relative to other figures. If it isn't changed, perhaps adding some labels would help orient the reader.

We have rotated the movies to be more consistent with the figures.

Fig. S7: No units are indicated for Kds (uM?).

Thank you – this has been fixed.

pg 11, paragraph 2. "adjacent to three residues: Glu118, Tyr120 and Asn192": The residue number used in the text doesn't seem to match the numbering in the PDB file. I think these residues correspond to Glu98, Tyr100, and Asn172 in the PDB file.

We regret this error. The correct numbering for both structures is now present in the deposited PDB files (7T1M for BstB and 7T1S for BstC).

pg 12, final paragraph: The authors present binding data for BstB variants with mutations in the putative sterol binding pocket identified in the structural and MD analyses. However, these mutants had no effect on binding. The authors rationalize this in terms of the size of the interface and hydrophobic nature (which indeed, may be correct and is very plausible), and it is worth noting that many of their mutations are to Ala and would largely preserve the hydrophobic nature of the cleft. However, these mutants raise questions about where sterols actually bind. No experimental evidence is presented that substrates bind in the cleft, it is only hypothesized based on structural homology, MD simulations, etc. These mutations formally provide evidence against the hypothesis being tested; I think that has to be discussed a bit more directly, alongside the caveats the authors already discuss about hydrophobicity, etc.

This is a valid point by the reviewer, and it is one we have attempted to address with our statement in the manuscript and in our response to reviewer 1. We have modified the relevant text to more clearly state that there is as of yet no experimental evidence for the binding of sterols to the cavity identified via molecular docking.

pg 13: Presumably this is not the full-length lipoprotein, but has been truncated/mutated in some way? Some statement of roughly what was purified/crystallized should be stated.

The SI methods on protein purification states that the genes of BstB and BstC without their respective signal peptides were obtained.

pg 13, last paragraph "TN1 exhibits hybrid hydrophobicity, with the sides horizontal to cavities being hydrophobic while the vertical sides are more hydrophilic". I don't really follow the horizontal vs vertical sides. Perhaps this could be described in a different way.

Noted and changed to “TN1 is closer to the N-terminal face of the structure, while CA1 and CA2 are proximal to the C-terminal face and form two open hydrophobic pockets; TN1 exhibits a mixture of hydrophobic and hydrophilic amino acids (Fig. 4B and Fig. S9B, Table S4).”

pg 15-16, "Comparison to eukaryotic sterol transporters": Perhaps this would be better suited for the discussion section? Could also be streamlined; it is mostly discussing and comparing eukaryotic sterol binding domains to each other, not to BstABC.

Given that BstB and BstC are the first identified proteins (and putative transporters) for bacterial sterol engagement, we thought a careful description of the existing sterol transporters (which are all eukaryotic) was warranted.

Reviewer #3 (Recommendations For The Authors):

I have just two minor suggestions for the authors if they wish to comment on or address them.

1. Do the three proteins (BstA/B/C) form any sort of complex? Perhaps this property was not assessed - but it seemed possible that the B and C components might constitute a shuttle for the membrane-bound transporter?

This is an important observation – the unliganded version of these proteins show no appreciable affinity for each other. However, BstB (which would be expected to engage both with BstA and BstC) belongs to a family of proteins known to undergo significant conformational change upon substrate binding. It is possible that with substrate present, complexes are formed – we have yet to investigate this.

1. In Figure S1, panel C - it appears that the label for the BstC cluster may have migrated away from the intended location. In this figure, it might also be useful to indicate in the caption the meaning of the red coloring of the nodes?

The label is now fixed – thank you for drawing our attention to this.

Author Response

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

REVIEWER #1

Leanza et al. investigated the regulation of Wnt signaling factors in the bone tissue obtained from individuals with or without type 2 diabetes. They showed that typical canonical Wnt ligands and downstream factors (Wnt10b, LEF1) are down-regulated, while Wnt5a and sclerostin mRNA are unregulated in diabetic bone tissue. Further, Wnt5a and sclerostin associated with the content of AGEs and SOST mRNA levels also correlated with glycemic control and disease duration.

Strengths:

• A strength of the study is the investigation of Wnt signaling in bone tissue from humans with type 2 diabetes. Most studies measure only serum levels of Wnt inhibitors, but this study takes it further and looks into bone specifically.

• The measurement of AGEs and its correlation to the Wnt signaling molecules is interesting and important. The correlation of sclerostin and Wnt5a with AGEs and disease duration suggests that inhibited Wnt signaling is paralleled by higher AGE levels and potentially weaker bone.

• The methodology in terms of obtaining the bone samples and the rigorous evaluation of RNA integrity is great and provides a solid basis for further analyses.

Weaknesses:

• A weakness may include the rather limited number of samples. Especially for some sub-analyses (e.g. RNA analyses), only a subset of samples was used.

• How was the sample size determined? It seems like more samples might have been necessary to obtain significant results for methods with a higher standard deviation (e.g. histomorphometry).

We apology for the oversight in the description of the statistical analysis and we thank the reviewer for the careful reading. For sample size calculation of bone histomorphometry we used the cohort of the only paper analyzing trabecular bone in T2D postmenopausal women by dynamic histomorphometry (Manavalan JS et al, JCEM 2012). We performed a priori sample size calculation using G*Power 3.1.9.7., based on the t-test, difference between two independent groups setting. Analysis demonstrated that given an effect size of 2.2776769, we needed a total of 12 patients (6/group) to reach a power of 0.978. Regarding gene expression analyses, it was performed not in a subset of patients, but in all recruited subjects for this study. Based on the results of gene expression analysis on our main outcome (Wnt signaling), we demonstrated that for SOST gene the effect size was 1.2733824, with a power of 0.9490065, confirming that sample size was sufficient to achieve adequate statistical power.

• Why is the number of samples different for the mRNA measurements? In most cases, there were 9, but in some 8 and in some 10?

We sincerely thank the reviewer for the opportunity to clarify such important aspects. The number of samples used for mRNA quantification may differ between the different analyzed genes due to multiple reasons: First, we used for the real-time PCR only samples with high quality ratio (260/280) between 1.8-2.0 as stated in the method section of the manuscript (Page 8, lines 163-164). Moreover, we decided not to use the undetermined values, undetectable after the amplification cycles (40 cycles in total), as specified in the method section (Page 8, line 167).

Overall, this study validates findings from the group that reported similar findings in 2020. This validates their methodology and shows that alterations in Wnt signaling are reproducible in human bone tissue.

We thank the reviewer for the positive comment, we really value her/his opinion.

COMMENTS:

(1) The authors could provide more details on how much of the bone was analyzed for bone histomorphometry (what area?).

We truly thank the reviewer for allowing us to explain more in depth our methodology. First, a biopsy containing trabecular bone from the femoral head was fixed in 10% neutral buffered formalin for 24 h prior to storage in 70% ethanol. Tissues were embedded in methylmethacrylate and sectioned sagittally by the Washington University Musculoskeletal Histology and Morphometry Core. Sections were stained with Goldner’s trichrome. Then, a rectangular region of interest containing trabecular bone was chosen below the cartilage-lined joint surface and primary spongiosa. This region had an average dimension of 45 mm2. Tissue processing artifacts, such as folding and edges, were excluded from the ROI. A threshold was chosen using the BIOQUANT software to automatically select trabeculae and measure bone volume. Finally, Osteoid was highlighted in the software and quantified semi-automatically using a threshold and correcting with the brush tool (as shown in the image below).

We specify that in the methods section (Page 7, lines 146-152).

Author response image 1.

(2) Could the number of samples used for histomorphometry be increased? That may also lead to more significant results.

We sincerely appreciated this suggestion from the reviewer but unfortunately, all available samples for histomorphometry have been analyzed and we are not able to increase the number of recruited participants at this time. Recruitment of people with T2D undergoing hip replacement is extremely difficult giving the limited number of those approved for elective surgery and compliant with our inclusion criteria. Considering also the long time needed to process bone sample for gene expression and histology analysis would require several months to have a consistent increase in recruited subjects. However, we have previously calculated sample size for bone histomorphometry analysis using the only available data of trabecular bone in T2D postmenopausal women measured by dynamic histomorphometry (Manavalan JS et al, JCEM 2012). We performed a priori sample size calculation using G*Power 3.1.9.7., based on the t-test of two independent groups. Analysis demonstrated that given an effect size of 2.2776769, we needed a total of 12 patients (6/group) to reach a power of 0.978.

(3) It would have been interesting to assess the biomechanical behavior of the bone specimens. While it is known that BMD is often higher in patients with T2D, the resistance to fractures is lower. Ideally, bone strength measures could be correlated with Wnt molecule expression and AGEs.

We agree with the reviewer that the assessment of biomechanical parameters in our cohort would increase the importance of this study, giving more insights on the effect of downregulation of Wnt signaling on bone strength. Thus, we followed reviewer suggestion, and we performed bone compression tests on trabecular bone core. We found a significant decrease in bone plasticity of T2D compared to controls [Young’s Modulus 21.6 (13.46-30.10 MPa) vs. 76.24 (26.81-132.9 MPa); p=0.0025). We added results of bone compression test in a new paragraph (Page 8, lines 191-194). In order to assess the validity of our results, we performed a post-hoc power calculation using G*Power 3.1.9.7. We demonstrated that effect size was 1.4716626, with a power of 0.9730784, confirming that sample size was sufficient to achieve adequate statistical power. We added methods in the related section and biomechanical data in table 3; we modified the manuscript accordingly (modifications are shown in track changes). Moreover, we also performed correlation analysis between Wnt target genes, AGEs and biomechanical parameters showing significant correlations as reported in the added paragraph in the results section (Page 11, Lines 225-233).

REVIEWER #2

This study reports the levels of expression of selected genes implicated in Wnt signaling in trabecular bone from femur heads obtained after surgery from post-menopausal women with (15 women) or without (21 women) type 2 diabetes. They found higher expression levels of SOST and WNT5A, and lower expression levels of LEF-1 and WNT10B in tissues from subjects with T2D, correlating with glycemia and advanced glycation products. No significant differences in bone density were observed. Overall, this is a cross-sectional, observational study measuring a limited set of genes found to vary with glycemia in postmenopausal women undergoing hip surgery.

Strengths:

The study demonstrates the feasibility of measuring gene expression in post-surgical trabecular bone samples, and finds differences associated with glycemia despite a relatively small number of subjects. It can form the basis for further research on the causes and consequences of changes in elements of the WNT signaling pathway in bone biology and disease.

Weaknesses:

The small number of targeted genes does not provide a comprehensive view of the transcriptional landscape within which the effects are observed. The gene expression changes are not associated with cellular or physiological properties of the tissue, raising questions about the biological significance of the observations.

We thank the reviewer for the comment. Replying to his/her concerns we have increased the number of Wnt target genes including more interactors of Wnt/β-catenin pathway. We measured GSK3B, AXIN2, BETA-CATENIN and SFRP5 gene expression levels, showing a significant increase in GSK3B, in line with a downregulation of Wnt signaling in T2D. We modified the manuscript accordingly with this new analysis and updated the figure 1 panel (Page 10, lines 210-213). Unfortunately, in this paper we were not able to perform experiments on cellular or physiological properties. However, in order to analyze the biological effect of the analyzed genes on the phenotype, we measured bone strength by performing compression tests on trabecular bone cores (Page 10, lines 201-203 and table 3) and used biomechanical parameters for correlation analysis with targeted genes showing significant correlations of bone strength and Wnt genes. We modified adding a new paragraph in the result section and a new figure panel to the main manuscript (Page 11, lines 225-233 and figure 4).

COMMENTS:

(1) The small number of targeted genes does not provide a comprehensive view of the transcriptional landscape within which the effects are observed. Given the author's success in obtaining good-quality RNA from trabecular bone, a more comprehensive exploration would greatly improve the quality of the study.

We agree with the reviewer that increase the transcriptional landscape related to Wnt signaling would be of interest for this work and we really thank for this opportunity. We were able to increase the number of Wnt target genes including more interactors of Wnt/β-catenin pathway, using the same cohort of patients in which we performed the other analysis. We also measured GSK3B, AXIN2, BETA-CATENIN and SFRP5 gene expression levels, showing a significant increase in GSK3B, in line with a downregulation of Wnt signaling in T2D. We modified the manuscript accordingly with this new analysis and updated the figures panel (Page 10, lines 210-213 and Figure 1).

(2) The gene expression changes are not associated with cellular or physiological properties of the tissue, raising questions about the biological significance of the observations. Can the authors perform immunohistochemistry to associate the changes in gene expression with protein expression?

We sincerely acknowledge this comment for focusing the attention on a such important aspect. We have partially replied to this comment in the previous paragraph. Regarding immunohistochemistry analysis, it is not possible to further use the available samples. This is mainly due to the fact that non-decalcified bones were embedded in plastic to allow for separate analysis of newly formed osteoid and mineralized bone. This process leads to poor antigen preservation and unsuitable detection of most targets. Moreover, antibodies for Wnt are also unreliable due to the secreted nature of the protein. Overall, this approach is unlikely to work efficiently. Similarly, RNAscope is not possible due to the resin. Optimization and validation of these analyses will need to be saved for a future study with fresh specimens.

REVIEWER #3

The manuscript by Leanza and colleagues explores the regulation of Wnt signaling and its association with advanced glycation end products (AGEs) accumulation in postmenopausal women with type 2 diabetes (T2D). The paper provides valuable insights into the potential mechanisms underlying bone fragility in individuals with T2D. Overall, the manuscript is well-structured, and the methodology is sound. I would suggest some minor revisions to improve clarity.

Strengths:

The study addresses an important and clinically relevant question concerning the mechanisms underlying bone fragility in postmenopausal women with T2D.

The study's methodology appears sound, and the inclusion of postmenopausal women with and without T2D undergoing hip arthroplasty adds to the clinical relevance of the findings. Additionally, measuring gene expression and AGEs in bone samples provides direct insights into the study's objectives.

The manuscript presents data clearly, and the results are well-organized.

Weaknesses:

Title. The title could be more specific to better reflect the content of the study. Also, the abstract should concisely summarize the study's main findings, providing some figures.

We thank the reviewer for this suggestion, and we modified the title giving specific information on the main findings of this study. The new title is “Bone canonical Wnt signaling is downregulated in type 2 diabetes and associates with higher Advanced Glycation End-products (AGEs) content and reduced bone strength”. Moreover, we added as suggested a graphical abstract summarizing our study results.

Introduction: the introduction would benefit from the addition of a clearer, more focused statement of the research questions or hypotheses guiding this study.

We thank the reviewer for this opportunity and we reformulated the hypothesis of this study based on our data and new findings as follow:” we hypothesized that T2D and AGEs accumulation downregulate Wnt canonical signaling and negatively affect bone strength”. (page 6, lines 116-117).

Methods: more information is needed on the hystomorphometry analysis. Surgical samples from 8 T2D and 9 non-diabetic subjects were used for histomorphometry analysis. How did these subjects compare with the other subjects in the T2D and control groups? Were they representative? How were they selected?

We thank the reviewer for the opportunity to clarify this important point. The number of subjects included in the different analysis of the paper differ for multiple reasons. In particular, we used only bone specimen with enough trabecular bone material adequate to perform histomorphometry analysis. Therefore, the samples used in the histomorphometry analysis belong to the same subjects enrolled in the study and analyzed for the other experiments of this paper. However, we have previously calculated sample size for bone histomorphometry analysis using the only available data of trabecular bone in T2D postmenopausal women measured by dynamic histomorphometry (Manavalan JS et al, JCEM 2012). We performed a priori sample size calculation using G*Power 3.1.9.7., based on the t-test of two independent groups. Analysis demonstrated that given an effect size of 2.2776769, we needed a total of 12 patients (6/group) to reach a power of 0.978.

COMMENTS:

(1) In the Abstract, values and p-values for comparisons, and Spearman's rho and p-values for correlations should be provided. Most adverbs (thus, accordingly, importantly) could be omitted to improve conciseness and clarity.

We kindly thank the reviewers for this precise and careful comment. We changed the Abstract accordingly. According to the abstract style of the journal we initially reported only the main findings. We have now modified providing values and p values as requested. We defer to the wishes of the editor as to the format in which the abstract should be reported.

(2) Result presentation: 25th and 75th percentile should be provided rather than the interquartile range, to better reflect data distribution.

We thank the reviewer for the opportunity to better clarify this part of the results section. We changed the manuscript accordingly.

(3) Estimated glomerular filtration rate should be calculated and provided as a marker of renal function, rather than serum creatinine values.

We thank the reviewer for the comment, and we modify the manuscript accordingly, adding the eGFR values in table 1 and in the result section.

(4) The manuscript should include a statement confirming compliance with the Declaration of Helsinki, considering that human subjects were involved in the study.

We thank the reviewer for the comment. The study was conducted in accordance with the Declaration of Helsinki. Ethics Committee of Campus Bio-Medico University approved the present study. Informed consent was obtained from all subjects involved in the study. (Page 6, lines 134-137).

Author Response

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

Reviewer #1 (Public Review):

In this manuscript, the authors investigated the role of Elg1 in the regulation of telomere length. The main role of the Elg1/RLC complex is to unload the processivity factor PCNA, mainly after completion of synthesis of the Okazaki fragment in the lagging strand. They found that Elg1 physically interacts with the CST (Cdc13-Stn1- Ten1) and propose that Elg1 negatively regulates telomere length by mediating the interaction between Cdc13 and Stn1 in a pathway involving SUMOylation of both PCNA and Cdc13. Accumulation of SUMOylated PCNA upon deletion of ELG1 or overexpression of RAD30 leads to elongated telomeres. On the other hand, the interaction of Elg1 with Sten1 is SIM-dependent and occurs concurrently with telomere replication in late S phase. In contrast Elg1-Cdc13 interaction is mediated by PCNA-SUMO, is independent on the SIM of Elg1 but still dependent on Cdc13 SUMOylation. The authors present a model containing two main messages 1) PCNA- SUMO acts as a positive signal for telomerase activation 2) Elg1 promotes Cdc13/Stn1 interaction at the expense of Cdc13/Est1 interaction thus terminating telomerase action.

The manuscript contains a large amount of data that make a major inroad on a new type of link between telomere replication and regulation of the telomerase. Nevertheless, the detailed choreography of the events as well as the role of PCNA- SUMO remain elusive and the data do not fully explain the role of the Stn1/Elg1 interaction. The data presented do not sufficiently support the claim that SUMO- PCNA is a positive signal for telomerase activation.

We thank the reviewer for her/his review efforts and opinion. We have re-submitted a new version of the manuscript in which we clarify some of the criticisms presented. In a point-by-point letter we respond to all the specific queries.

Reviewer #2 (Public Review):

This paper purports to unveil a mechanism controlling telomere length through SUMO modifications controlling interactions between PCNA unloader Elg1 and the CST complex that functions at telomeres. This is an extremely interesting mechanism to understand, and this paper indeed reveals some interesting genetic results, leading to a compelling model, with potential impact on the field. The conclusions are largely supported by experiments examining protein-protein interactions at low resolution and ambiguous regarding directness of interactions like co-IP and yeast two-hybrid (Y2H) combined with genetics. However, some results appear contradictory and there's a lack of rigor in the experimental data needed to support claims. There is significant room for improvement and this work could certainly attain the quality needed to support the claims. The current version needs substantial revision and lacks the necessary experimental detail. Stronger support for the claims would add detail to help distinguish competing models.

We thank the reviewer for her/his positive opinion. We have re-submitted a new version of the manuscript in which we clarify some of the criticisms presented by thereferees, and added all the missing experimental details. In a point-by-point letter we respond to all the specific queries.

Reviewer #3 (Public Review):

This paper reveals interesting physical connections between Elg1 and CST proteins that suggest a model where Elg1-mediated PCNA unloading is linked to regulation of telomere length extension via Stn1, Cdc13, and presumably Ten1 proteins. Some of these interactions appear to be modulated by sumolyation and connected with Elg1's PCNA unloading activity. The strength of the paper is in the observations of new interactions between CST, Elg1, and PCNA. These interactions should be of interest to a broad audience interested in telomeres and DNA replication.

We thank the reviewer for her/his positive opinion. We have re-submitted a new version of the manuscript in which we clarify some of the criticisms presented. In a point-by-point letter we respond to all the specific queries.

What is not well demonstrated from the paper is the functional significance of the interactions described. The model presented by the authors is one interpretation of the data shown, and proposes that the role of sumolyation is temporally regulate the Elg1, PCNA and CST interactions at telomeres. This model makes some assumptions that are not demonstrated by this work (such as Stn1 sumolyation, as noted) and are left for future testing. Alternative models that envision sumolyation as a key in promoting spatial localization could also be proposed based on the data here (as mentioned in the discussion), in addition to or instead of a role for sumolyation in enforcing a series of switches governing a tightly sequenced series of interactions and events at telomeres. Critically, the telomere length data from the paper indicates that the proposed model depicts interactions that are not necessary for telomerase activation or inhibition, as telomeres in pol30-RR strains are normal length and telomeres in elg1∆ strains are not nearly as elongated as in stn1 strains. One possibility mentioned in the paper is the PCNAS and Elg1 interactions are contributing to the negative regulation of telomerase under certain conditions that are not defined in this work. Could it also be possible that the role of these interactions is not primarily directed toward modulating telomerase activity? It will be of interest to learn more about how these interactions and regulation by Sumo function intersect with regulation of telomere extension.

We present compelling evidence for a role of SUMOylated PCNA in telomere length regulation. Figure 1 shows that this modification is both necessary and sufficient to elongate the telomeres, indicating that PCNA SUMOylation plays a positive role in telomere elongation. The model we present is consistent with all our results. There are, of course, possible alternative models, but they usually fail to explain some of the results. We agree that the fact that pol30-RR presents normal-sized telomeres implies that SUMO-PCNA is not required for telomerase to solve the "end replication problem", but rather is needed for "sustained" activity of telomerase. Since elongated telomeres (by absence of Elg1 or by over-expression of SUMO-PCNA) was the phenotype monitored, this may require sustained telomerase activity. Similar results were seen in the past for Rnr1 (Maicher et al., 2017), and this mode depends on Mec1, rather than Tel1 (Harari and Kupiec, 2018). Telomere length regulation is complex, and we may not yet understand the whole picture. It appears that for normal “end replication problem” solution, very little telomerase activity may be needed, and spontaneous interactions at a low level may suffice. Future work may find the conditions at which telomerase switches from "end replication problem" to "sustained" activity. We have added further explanations on this subject to the Discussion section.

We suspect, but could not prove, a role for Stn1 SUMOylation in the interactions. SUMOylation is usually transient, and notoriously hard to detect, and despite the fact that many telomeric proteins are SUMOylated, Stn1 SUMOylation could not be shown directly by us and others (Hang et al, 2011).

Reviewer #1 (Recommendations For The Authors):

Suggestions for improved or additional experiments, data or analyses.

• My main concern is the claim that SUMOylated PCNA acts as a positive signal for telomerase activation. Yet the pol30-RR mutant has no impact on telomere length. The explanation of the authors is not entirely convincing.

We are aware that the regulation of telomere length is complex, and we may not fully understand it yet. Just consider the fact that ~500 genes participate in determining the final telomere length of a yeast (Askree et al., 2004). Since mutation in EACH of these genes has a phenotype, the implication is that the joint action of 500 players determines the outcome (a dialogue of 500 participants). Having said this, we clearly show in figure 1 that mutations that prevent PCNA SUMOylation prevent telomere length elongation in cells lacking Elg1, and overexpressing SUMOylated PCNA is enough to elongate the telomeres. Thus, SUMOylation of PCNA does act as a positive signal for elongation.

However, it appears that to fulfill the minimal requirement of dealing with the "end- replication problem", PCNA SUMOylation is not required, and only a "sustained activity" mode requires the S-PCNA signal (as we have also shown, surprisingly, for RNR1, Maicher et al. 2017). This sustained activity mode depends on Mec1, rather than Tel1 (Harari and Kupiec, 2018). Since elongated telomeres (by absence of Elg1 or by over-expression of SUMO-PCNA) was the phenotype monitored, this may require sustained telomerase activity. Telomere length regulation is complex, and we may not yet understand the whole picture. It appears that for normal “end replication problem” solution, very little telomerase activity may be needed, and spontaneous interactions at a low level may suffice (for example, unmodified PCNA may promote telomerase activity at a lower level than that of SUMO-PCNA. Future work may find the conditions at which telomerase switches from "end replication problem" to "sustained" activity.

We have added further explanations on this subject to the Discussion section.

• The model is entitled « Elg1 negatively regulates the telomere length by forming an interaction with the CST complex ». Nevertheless, expression of PCNA-RR completely reversed the long telomere phenotype of elg1∆ cells. Thus it appears that although the interaction between Stn1 and Cdc13 is reduced in the absence of Elg1, Elg1/Stn1 interaction is not instrumental in the formation of the CST complex and thus in the termination of telomerase activity. Does the elg1∆SIM mutant that does not interact with Stn1 impact telomere length?

• In the model part (lane 318), it is argued that the complex Elg1-Stn1 unloads SUMOylated PCNA. Elg1-Stn1 interaction depends on the SIM of Elg1. This SIM is however not required for Elg1's function in genome-wide SUMO-PCNA unloading, is it required specifically at telomeres?

The interactions between Elg1 and SUMOylated PCNA are carried out through both the SIM and the Threonines 386 and 387 (Shemesh et al, 2017). Consistently, the single elg1-SIM mutant has telomeres of normal length, and its effects on telomere length can only be seen when combined with mutations in the Threonines (elg1- TT386/7AA or elg1-TT386/7DD). Although the unloading of SUMOylated PCNA by Elg1 is important, the gene is not essential, and PCNA is either eventually unloaded by RFC, or spontaneously dis-assembles. This explains why the telomere length does not reach the same length in the absence of Elg1 as in the absence of, say, Stn1.

• The model suggests that Elg1 promotes the interaction between Cdc13 and Stn1. This is based on the data presented in Figure 5 E and F. This is an important result. Because the experiment has been done on cells synchronized in S phase and the Elg1/Stn1 interaction occurs specifically at the end of S-phase, the FACS profile should be shown or a control provided to show that the two conditions are comparable.

The FACS profile for this experiment is shown in Figure 5C.

• Does the interaction between Cdc13 and Pol30 depend on the SUMOyaltion of POL30 ?

Yes. We have added this as new Figure S2, and presented the results together with Figure 3 (Figure 3 is already too crowded).

Others points :

• Fig 1 : it should be mentioned in the Materials and Methods or in the figure legend how the average telomere lengths (horizontal bar) were calculated from the teloblot, as the position of the bar is not always intuitive

We estimate telomere length by using TelQuant (Rubinstein et al., 2014). We have added this to the Methods section.

-Fig 2 : Owing to the large span of telomere length in the stn1 mutants, the epistatic relationship between elg1∆ and stn1 mutants is poorly illustrated by the teloblot.

We repeated this experiment several times, and stn1 mutants consistently gave a very spread telomere length. In ALL the blots, however, the double mutants elg1 stn1 showed a telomere length similar to that of the single stn1 mutant, and never longer.

• It is mentioned that other mutants in the collection showed epistasis. Are any of these mutants related to telomere replication or the proposed model?

Since we used the collection of non-essential mutants (so far), it was quite devoid of genes involved in DNA replication, which are mostly essential. An exception was siz1, which showed epistasis with elg1Δ.

• The section entitled « Elg1's functional activity is essential for its interaction with Cdc13 » (lane 205) is difficult to follow. The hierarchy between the different mutants of Elg1 on their capacity to unload PCNA is not totally in agreement with the data published in Itzkovich et al 2023 and Shemesh et al. 2017. In particular it appears to me from these papers that elg1-WalkerA 238 (KK343/4AA) mutant did not show a defect in contrast to elg1-WalkerA 238(KK343/4DD).

We are sorry for the typo in the results. We used the elg1-WalkerA (KK343/4DD) allele, which has a normal SIM but no activity. In a nutshell, we used mutants that either did or did not show unloading activity and/or SIM. The results clearly show that you need to unload PCNA in order for the N-ter of Elg1 to interact with Cdc13.

• Are the synchronization done at 30{degree sign}C ?

Yes. We have added the information to the Methods section.

• ChIP experiments are not described in the Materials and Methods

We apologize for this. They are now described.

• In the figure 6, the PCNA rings are curiously placed at the beginning of the Okasaki fragments.

We thank the referee for noticing, we have corrected the figure.

Reviewer #2 (Recommendations For The Authors):

This paper purports to unveil a mechanism controlling telomere length through SUMO modifications controlling interactions between PCNA unloader Elg1 and the CST complex that functions at telomeres. This is an extremely interesting mechanism to understand, and this paper indeed reveals some interesting genetic results, leading to a compelling model, with potential impact on the field. The conclusions are largely supported by experiments examining protein-protein interactions at low resolution and ambiguous regarding directness of interactions like co-IP and yeast two-hybrid (Y2H) combined with genetics. However, some results appear contradictory and there's a lack of rigor in the experimental data needed to support claims. There is significant room for improvement and this work could certainly attain the quality needed to support the claims. The current version needs substantial revision and lacks necessary experimental detail. Stronger support for the claims would add detail to help distinguish competing models.

Specific comments:

Insufficient technical detail: I could find no explanation of how overexpression was achieved. No description of how teloChIP is performed, either for the PCNA IP or how the sequence analysis is performed. Too limited details on growth like exact temperatures for the cell cycle time course.

We have significantly expanded the Methods section to include all the technical information.

Please do not bold and underline text for emphasis-EVER

We have removed those from the text.

Lines 130-132: they have not shown "accumulation of SUMOylated PCNA" anywhere; this is an inference.

We have modified the text, it says: ”show that SUMOylated PCNA, and not unmodified or ubiquitinated PCNA, is both necessary and sufficient for telomere elongation in the presence or in the absence of Elg1.”

Fig 2A Can authors show any other very long-telomere mutant like stn1 that does show enhancement in combination with elg1∆ to show feasibility of such phenotype?

We don't think it is appropriate for the paper, but we have systematically created double mutants with elg1Δ and found many additive and even synergistic interactions. Here is an example. in Author response image 1, taken from the PhD thesis of Taly Ben-Shitrit, a PhD student in the lab.

Author response image 1.

What about cdc13 or ten1? Epistatic?

We did not test telomere length in combination with Ten1. Combining elg1 with cdc13-50 resulted in synergistic elongation. Given the complex genetic relationship between Stn1/Ten1 and Cdc13, it is hard to interpret this result.

Seems tenuous to use Y2H to decipher protein-protein interactions occurring out of context (i.e., not at telomere but at reporter gene promoter)

Y2H is a great method to detect interactions, even if they are transient. Whenever possible, we confirm our findings using co-IP or telo-ChIP.

Lines 268-270: It would be more accurate to state "can be" instead of "becomes" or "is" as they have not shown that SUMOylation or PCNA unloading have occurred.

We agree, and have changed the text.

Cdc13snm protein level?

Unfortunately our Western blot is not presentable, but the level of Cdc13snm was similar to that of the wt Cdc13, and this result has been already published by Hang et al., 2011.

Fig S3A: If SUMOylated Cdc13 mediates the Stn1-Elg1 interaction, why is Stn1-Elg1 interaction maintained in cdc13snm strain? This result seems to directly contradict the premise and overall conclusion of this section that Cdc13-SUMO mediates the (Y2H) interaction of Elg1 and Stn1.

According to our model, the interaction between Stn1 and Elg1 takes place upstream, and only then this complex interacts with SUMOylated Cdc13. Hence, if Cdc13 cannot be SUMOylated, the interaction Elg1-Stn1 is not lost, although Stn1 fails to interact with Cdc13, leading to a telomeric phenotype.

Line 279: which data establishes Stn1-Elg1 interaction as direct? Fig 2B co-Ip indicates physical but not necessarily direct interaction, but later the authors suggest that the interaction requires a SUMOylated intermediary, and Y2H in Fig. S3B doesn't demonstrate direct interaction.

We have changed the text, taking out the word "direct".

Co-Ip shows that interaction of Elg1 with Stn1 occurs mainly during later Sphase and with an overall delay compared to initial Elg1-Pol3 interaction.Co-IP Interaction between Cdc13 and Stn1 is reduced in the absence of Elg1

The subsection title: "The interaction of Elg1 with Stn1 takes place at telomeres only at late S-phase" is not well supported by the data. I agree the data are consistent with the idea of the interactions occurring at telomeres but there's no direct evidence of this.

We have changed the subsection title. It now reads: " The interaction of Elg1 with Stn1 takes place only at late S-phase"

Model: Is unloading happening at the fork? Doesn't PCNA unloading have to follow its loading which occurred behind the fork particularly on the lagging strand? Model now suggest that Stn1 itself is SUMOylated.

Yes, according to the model Elg1 moves with the fork, unloading PCNA from the lagging strand. Once Elg1 reaches the telomeres, it interacts with Stn1 (Figure 5). This interaction requires SUMOylation of Stn1 or of some other protein, which is not PCNA (Figure 3D) nor Cdc13 (Figure S3A) and could be Stn1 itself or another telomeric protein (Hang et al., 2011)

Title is rather vague.

We think it summarizes what we present in the paper.

Abstract:

"We report that SUMOylated PCNA acts as a signal that positively regulates telomerase activity."

I don't think this is supported or a good description of what they find

Figure 1B clearly shows that SUMO-PCNA is both necessary and sufficient for telomere elongation.

"and dissected the mechanism by which Elg1 and Stn1 negatively regulates telomere elongation, coordinated by SUMO."

Again, I don't think this is sufficiently supported and the model invokes SUMOylation events not demonstrated like Stn1, which might be a significant step forward.

On the positive side, their model makes several predictions that they could test much more directly and rigorously: for example, examining the impact of the relevant mutations in the recruitment of proteins to the telomere.

We have dissected the mechanism, and future work will be devoted to examining the impact of the relevant mutations in the recruitment of proteins to the telomere.

Reviewer #3 (Recommendations For The Authors):

Comments:

1) The telomere length analysis data presented here is consistent with an interpretation that Stn1 and Elg1 play roles in a similar telomere maintenance pathway because the telomere restriction fragment pattern in the double mutants are not longer than the stn1 single mutants. No comment is made with respect to the yellow bars in Figure 2 that presumably measure telomere length appearing to be slightly shorter than in the stn1 single mutants. It may be interesting and informative if the double mutants do in fact have some phenotype distinct from the single stn1 mutants. Is there an impact on viability in the double mutant?

Given the variable telomeric phenotype of the single stn1 mutants, slight variations in the measurement of the median telomere size are expected. The difference observed is not likely to be significant. What is important is that the double mutants with elg1 do not show longer telomeres. In terms of fitness, the stn1 mutants grow slightly slowly, but the elg1 mutation does not slow them down further.

2) It is somewhat surprising that no additional telomere length analysis is included that actually tests the proposed model, including whether this path could be operational only under certain conditions. Maybe this is a topic of the next paper?

Indeed, future work will explore the conditions under which PCNA SUMOylation is essential, and those under which is only needed.

3) Were the error bars in Figure 5F determined only from the experiment in E? Does this represent error in measuring the data from one biological replicate? The type of error should be made clear to avoid readers assuming the data represents measurements from more than one sample in more than one experiment. The data would be stronger if it represented measurements from multiple experiments.

The graph was made with data from three biological replicates. We show the best blot in Figure 5E. We have now stressed this in the Figure Legend.

4) Why was only one two hybrid reporter shown? Having the multiple reporters can give confidence in interactions. (Not a big deal here given the nice co-IP data.)

We thought that it is enough to show one reporter, as the results with a different reporter (B-gal assay) led to the same conclusions. since this did not add information and made the paper too lengthy (and boring), we took them out. In any case all data was verified by co-IP.

5) Line 414 - what are the 32P-radio labeled PCR fragments? Are these solely comprised of TG1-3 repeats of some length? A bit more detail in this aspect of the method could be helpful.

We have added an explanation on the probe in the Methods section.

6) Line 432-433 - which anti-HA or anti-My antibodies are these? (very minor detail)

We have added the details.

Author response:

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

Public Reviews: 

Reviewer #1 (Public review): 

Summary: 

Liu and colleagues applied the hidden Markov model on fMRI to show three brain states underlying speech comprehension. Many interesting findings were presented: brain state dynamics were related to various speech and semantic properties, timely expression of brain states (rather than their occurrence probabilities) was correlated with better comprehension, and the estimated brain states were specific to speech comprehension but not at rest or when listening to non-comprehensible speech. 

Strengths: 

Recently, the HMM has been applied to many fMRI studies, including movie watching and rest. The authors cleverly used the HMM to test the external/linguistic/internal processing theory that was suggested in comprehension literature. I appreciated the way the authors theoretically grounded their hypotheses and reviewed relevant papers that used the HMM on other naturalistic datasets. The manuscript was well written, the analyses were sound, and the results had clear implications. 

Weaknesses: 

Further details are needed for the experimental procedure, adjustments needed for statistics/analyses, and the interpretation/rationale is needed for the results. 

For the Experimental Procedure, we will provide a more detailed description about stimuli, and the comprehension test, and upload the audio files and corresponding transcriptions as the supplementary dataset. 

For statistics/analyses, we have reproduced the states' spatial maps using unnormalized activity pattern. For the resting state, we observed a state resembling the baseline state described in Song, Shim, & Rosenberg (2023). However, for the speech comprehension task, all three states were characterized by network activities varying largely from zero. In addition, we have re-generated the null distribution for behaviorbrain state correlations using circular shift. The results are largely consistent with the previous findings. We have also made some other adjustment to the analyses or add some new analyses as recommended by the reviewer. We will revise the manuscript to incorporate these changes.

For the interpretation/rationale: We will add a more detailed interpretation for the association between state occurrence and semantic coherence. Briefly speaking, higher semantic coherence may allow for the brain to better accumulate information over time.

State #2 seems to be involved in the integration of information at shorter timescales (hundreds of milliseconds) while State #3 seems to be involved in the longer timescales (seconds). 

We greatly appreciate the reviewer for the insightful comments and constructive suggestions.  

Reviewer #2 (Public review): 

Liu et al. applied hidden Markov models (HMM) to fMRI data from 64 participants listening to audio stories. The authors identified three brain states, characterized by specific patterns of activity and connectivity, that the brain transitions between during story listening. Drawing on a theoretical framework proposed by Berwick et al. (TICS 2023), the authors interpret these states as corresponding to external sensory-motor processing (State 1), lexical processing (State 2), and internal mental representations (State 3). States 1 and 3 were more likely to transition to State 2 than between one another, suggesting that State 2 acts as a transition hub between states. Participants whose brain state trajectories closely matched those of an individual with high comprehension scores tended to have higher comprehension scores themselves, suggesting that optimal transitions between brain states facilitated narrative comprehension. 

Overall, the conclusions of the paper are well-supported by the data. Several recent studies (e.g., Song, Shim, and Rosenberg, eLife, 2023) have found that the brain transitions between a small number of states; however, the functional role of these states remains under-explored. An important contribution of this paper is that it relates the expression of brain states to specific features of the stimulus in a manner that is consistent with theoretical predictions. 

(1) It is worth noting, however, that the correlation between narrative features and brain state expression (as shown in Figure 3) is relatively low (~0.03). Additionally, it was unclear if the temporal correlation of the brain state expression was considered when generating the null distribution. It would be helpful to clarify whether the brain state expression time courses were circularly shifted when generating the null. 

In the revision, we generated the null distribution by circularly shifting the state time courses. The results remain consistent with our previous findings: p = 0.002 for the speech envelope, p = 0.007 for word-level coherence, and p = 0.001 for clause-level coherence.

We note that in other studies which examined the relationship between brain activity and word embedding features, the group-mean correlation values are similarly low but statistically significant and theoretically meaningful (e.g., Fernandino et al., 2022; Oota et al., 2022). We think these relatively low correlations are primarily due to the high level of noise inherent in neural data. Brain activity fluctuations are shaped by a variety of factors, including task-related cognitive processing, internal thoughts, physiological states, as well as arousal and vigilance. Additionally, the narrative features we measured may account for only a small portion of the cognitive processes occurring during the task. As a result, the variance in narrative features can only explain a limited portion of the overall variance in brain activity fluctuations.

We will replace Figure 3 and the related supplementary figures with new ones, in which the null distribution is generated via circular shift. Furthermore, we will expand our discussion to address why the observed brain-stimuli correlations are relatively small, despite their statistical significance.

(2) A strength of the paper is that the authors repeated the HMM analyses across different tasks (Figure 5) and an independent dataset (Figure S3) and found that the data was consistently best fit by 3 brain states. However, it was not entirely clear to me how well the 3 states identified in these other analyses matched the brain states reported in the main analyses. In particular, the confusion matrices shown in Figure 5 and Figure S3 suggests that that states were confusable across studies (State 2 vs. State 3 in Fig. 5A and S3A, State 1 vs. State 2 in Figure 5B). I don't think this takes away from the main results, but it does call into question the generalizability of the brain states across tasks and populations. 

We identified matching states across analyses based on similarity in the activity patterns of the nine networks. For each candidate state identified in other analyses, we calculate the correlation between its network activity pattern and the three predefined states from the main analysis, and set the one it most closely resembled to be its matching state. For instance, if a candidate state showed the highest correlation with State #1, it was labelled State #1 accordingly. 

Each column in the confusion matrix depicts the similarity of each candidate state with the three predefined states. In Figure S3 (analysis for the replication dataset), the highest similarity occurred along the diagonal of the confusion matrix. This means that each of the three candidate states was best matched to State #1, State #2, and State #3, respectively, maintaining a one-to-one correspondence between the states from two analyses.

For the comparison of speech comprehension task with the resting and the incomprehensible speech condition, there was some degree of overlap or "confusion."

In Figure 5A, there were two candidate states showing the highest similarity to State #2. In this case, we labelled the candidate state with the strongest similarity as State #2, while the other candidate state is assigned as State #3 based on the ranking of similarity. This strategy was also applied to naming of states for the incomprehensible condition. The observed confusion supports the idea that the tripartite-state space is not an intrinsic, task-free property. To make the labeling clearer in the presentation of results, we will use a prime symbol (e.g., State #3') to indicate cases where such confusion occurred, helping to distinguish these ambiguous matches.

(3) The three states identified in the manuscript correspond rather well to areas with short, medium, and long temporal timescales (see Hasson, Chen & Honey, TiCs, 2015).

Given the relationship with behavior, where State 1 responds to acoustic properties, State 2 responds to word-level properties, and State 3 responds to clause-level properties, the authors may want to consider a "single-process" account where the states differ in terms of the temporal window for which one needs to integrate information over, rather than a multi-process account where the states correspond to distinct processes. 

The temporal window hypothesis provides a more fitting explanation for our results. Based on the spatial maps and their modulation by speech features, States #1, #2, and #3 seem to correspond to short, medium, and long processing timescales, respectively. We will update the discussion to reflect this interpretation.

We sincerely appreciate the constructive suggestions from the two anonymous reviewers, which have been highly valuable in improving the quality of the manuscript.  

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors): 

(1) The "Participants and experimental procedure" section deserves more details. I've checked Liu et al. (2020), and the dataset contained 43 participants aged 20-75 years, whereas this study contained data from 64 young adults and 30 old adult samples. The previous dataset seems to have two stories, whereas this study seems to have three. Please be specific, given that the dataset does not seem the same. Could the authors also include more descriptions of what the auditory stories were? For example, what were the contents, and how were they recorded? 

The citation is partially incorrect. The dataset of young adults is shared with our work published in (2022). The 64 participants listened to one of three stories told by a female college student in Mandarin, recounting her real-life experience of hiking, a graduate admission interview, and her first time taking a flight, respectively. The sample of older adults is from our work published in (2020), which includes 30 older adults and additionally 13 young adults. The stimuli in this case were two stories told by an older woman in a Chinese dialect, describing her experience in Thailand and riding a warship, respectively. Since we aim to explore whether the main results can be replicated on a different age group, we excluded the 13 young adults from the analysis. 

All the stories were recorded during fMRI scanning using a noise-canceling microphone (FOMRI-III; Optoacoustics Ltd, Or-Yehuda, Israel) positioned above the speaker’s mouth. The audio recordings were subsequently processed offline with Adobe Audition 3.0 (Adobe Systems Inc., USA) to further eliminate MRI scanner noise.

In the revised manuscript, we have updated the citation, and provided a more detailed description of the stimuli in the supplementary material. We have also uploaded the audio files along with their corresponding transcriptions to GitHub.

(2) I am curious about individual differences in comprehension scores. Did participants have less comprehension of the audio-narrated story because the story was a hard-tocomprehend narrative or because the audio quality was low? Could the authors share examples of comprehension tests? 

We believe two factors contribute to the individual differences in comprehension scores. First, the audio quality is indeed moderately lower than in dailylife story-listening conditions. This is because those stories were recorded and played during fMRI scanning. Although a noise-canceling equipment was used, there were still some noises accompanying the speech, which may have made speech perception and comprehension more difficult than usual.

Second, the comprehension test measured how much information about the story (including both main themes and details) participants could recall. Specifically, participants were asked to retell the stories in detail immediately after the scanning session. Following this free recall, the experimenters posed a few additional questions drawn from a pre-prepared list, targeting information not mentioned in their recall. If participants experienced lapses of attention or did not store the incoming information into memory promptly, they might fail to recall the relevant content. In several studies, such a task has been called a narrative recall test. However, memory plays a crucial role in real-time speech comprehension, while comprehension affects the depth of processing during memory encoding, thereby influencing subsequent recall performance. To align with prior work (e.g., Stephens et al., 2010) and our previous publications, we chose to referred to this task as narrative comprehension. 

In the revised manuscript, we have provided a detailed description about the comprehension test (Line 907-933) and share the examples on GitHub. 

(3) Regarding Figure 3, what does it mean for a state occurrence to follow semantic coherence? Is there a theoretical reason why semantic coherence was measured and related to brain state dynamics? A related empirical question is: is it more likely for the brain states to transition from one state to another when nearby time points share low semantic similarity compared to chance? 

We analyzed semantic coherence and sound envelope as they capture different layers of linguistic and acoustic structure that unfold over varying temporal scales. Changes in the sound envelope typically occur on the order of milliseconds to a few hundred milliseconds, changes in word-level semantic coherence span approximately 0.24 ± 0.15 seconds, and changes in clause-level semantic coherence extend to 3.2 ± 1.7 seconds. Previous theory and empirical studies suggest that the timescales of information accumulation vary hierarchically, progressing from early sensory areas to higher-order areas (Hasson et al., 2015; Lerner et al., 2011). Based on this work, we anticipate that the three brain states, which are respectively associated with the auditory and sensory motor network, the language network and the DMN, would be selectively modulated by these speech properties corresponding to distinct timescales. 

Accordingly, when a state occurrence aligns with (clause-level) semantic coherence, it suggests that this state is engaged in processing information accumulated at the clause level (i.e., its semantic relationship). Higher coherence facilitates better accumulation, making it more likely for the associated brain state to be activated. 

We analyzed the relationship between state transition probability and semantic coherence, but did not find significant results. Here, the transition probability was calculated as Gamma(t) – Gamma(t-1), where Gamma refers to the state occurrence probability. The lack of significant findings may be because brain state transitions are driven primarily by more slowly changing factors. Indeed, we found the average dwell time of the three states ranges from 9.66 to 15.29s, which is a much slower temporal dynamics compared to the relatively rapid shifts in acoustic/semantic properties. 

In the revised version, we have updated the Introduction to clarify the rational for selecting the three speech properties and to explore their relationship with brain dynamics (Line 111-118)

(4) When running the HMM, the authors iterated K of 2 to 10 and K = 4, 10, and 12. However, the input features of the model consist of only 9 functional networks. Given that the HMM is designed to find low-dimensional latent state sequences, the choice of the number of latent states being higher than the number of input features sounds odd to me - to my speculation, it is bound to generate almost the exact same states as 9 networks and/or duplicates of the same state. I suggest limiting the K iterations from 2 to 8. For replication with Yeo et al.'s 7 networks, K iteration should also be limited to K of less than 7, or optionally, Yeo's 7 network scheme could be replaced with a 17network scheme. 

We understand your concern. However, the determination of the number (K) of hidden states is not directly related to the number of features (in this case, the number of networks), but rather depends on the complexity of the time series and the number of underlying patterns. Given that each state corresponds to a distinct combination of the features, even a small number of features can be used to model a system with complex temporal behaviors and multiple states. For instance, for a system with n features, assuming each is a binary variable (0 or 1), there are maximally 2ⁿ possible underlying states. 

In our study, we recorded brain activity over 300 time points and used the 9 networks as features. At different time points, the brain can exhibit distinct spatial configurations, reflected in the relative activity levels of the nine networks and their interactions. To accurately capture the temporal dynamics of brain activity, it is essential to explore models that allow for more states than the number of features. We note that in other HMM studies, researchers have also explored states more than the number of networks to find the best number of hidden states (e.g., Ahrends et al., 2022; Stevner et al., 2019). 

Furthermore, Ahrends et al. (2022) suggested that “Based on the HCP-dataset, we estimate as a rule of thumb that the ratio of observations to free parameters per state should not be inferior to 200”, where free parameters per state is [𝐾 ∗(𝐾 −1)+ (𝐾 −1)+𝐾 ∗𝑁 ∗(𝑁 +1)/2]/𝐾. According to this, there should be above 10, 980 observations when the number of states (K) is 10 (the maximal number in our study) and the number of networks (N) is 9. In our group-level HMM model, there were 64 (valid runs) * 300 (TR) = 19200 observations for young adults, and 50 (valid runs) * 210 (TR) = 10500 observations for older adults. Aside from the older adults' data being slightly insufficient (4.37% less than the suggestion), all other hyperparameter combinations in this study meet the recommended number of observations. 

(5) In Figure 2, the authors write that the states' spatial maps were normalized for visualization purposes. Could the authors also show visualization of brain states that are not normalized? The reason why I ask is, for example, in Song, Shim, & Rosenberg (2023), the base state was observed which had activity levels all close to the mean (which is 0 because the BOLD activity was normalized). If the activity patterns of this brain state were to be normalized after state estimation, the base state would have looked drastically different than what is reported. 

We derived the spatial maps of the states using unnormalized activity patterns, with the BOLD signals Z-score normalized to a mean of zero. Under the speech comprehension task, the three states exhibited relatively large fluctuations in network activity levels. The activity ranges were as follows: [-0.71 to 0.51] for State #1, [-0.26 to 0.30] for State #2, and [-0.82 to 0.40] for State #3. For the resting state, we observed a state resembling the baseline state as described in Song, Shim, & Rosenberg (2023), with activity values ranging from -0.133 to 0.09. 

In the revision, we have replaced the states' spatial maps with versions showing unnormalized activity patterns. 

(6) In line 297, the authors speculate that "This may be because there is too much heterogeneity among the older adults". To support this speculation, the authors can calculate the overall ISC of brain state dynamics among older adults and compare it to the ISC estimated from younger adults.  

We analyzed the overall ISC of brain state dynamics, and found the ISC was indeed significantly lower among the older adults than that among the younger adults. We have revised this statement as follows:

These factors can diminish the inter-subject correlation of brain state dynamics— indeed, ISCs among older adults were significantly lower than those among younger adults (Figure S5)—and reduce ISC's sensitivity to individual differences in task performance (Line 321-326).

Other comments: 

(7) In Figure 4, the authors showed a significant positive correlation between head movement ISC with the best performer and comprehension scores. Does the average head movement of all individuals negatively correlate with comprehension scores, given that the authors argue that "greater task engagement is accompanied by decreased movement"? 

We examined the relationship between participants' average head movement across the comprehension task and their comprehension scores. There was no significant correlation (r = 0.041, p = 0.74). In the literature (e.g. ,Ballenghein et al., 2019) , the relationship between task engagement and head movement was also assessed at the moment-by-moment level, rather than by using time-averaged data.

Real-time head movements reflect fluctuations in task engagement and cognitive state. In contrast, mean head movement, as a static measure, fails to capture these changes, and thus is not effective in predicting task performance.

(8) The authors write the older adults sample, the "independent dataset". Technically, however, this dataset cannot be independent because they were collected at the same time by the same research group. I would advise replacing the word independent to something like second dataset or replication dataset. 

We have replaced the phrase “independent dataset” with “replication dataset”. 

(9) Pertaining to a paragraph starting in line 586: For non-parametric permutation tests, the authors note that the time courses of brain state expression were "randomly shuffled". How was this random shuffling done: was this circular-shifted randomly, or were the values within the time course literally shuffled? The latter approach, literal shuffling of the values, does not make a fair null distribution because it does not retain temporal regularities (autocorrelation) that are intrinsic to the fMRI signals. Thus, I suggest replacing all non-parametric permutation tests with random circular shifting of the time series (np. roll in python).  

In the original manuscript, the time course was literally shuffled. In the revised version, we circular-shifted the time course randomly (circshift.m in Matlab) to generate the null distribution. The results remain consistent with our previous findings: p = 0.002 for the speech envelope, p = 0.007 for word-level coherence, and p = 0.001 for clause-level coherence (Line 230-235). 

(10) The p value calculation should be p = (1+#(chance>=observed))/(1+#iterations) for one-tailed test and p = (1+#(abs(chance)>=abs(observed)))/(1+#iterations) for twotailed test. Thus, if 5,000 iterations were run and none of the chances were higher than the actual observation, the p-value is p = 1/5001, which is the minimal value it can achieve. 

Have corrected. 

(11) State 3 in Figure S2 does not resemble State 3 of the main result. Could the authors explain why they corresponded State 3 of the Yeo-7 scheme to State 3 of the nineparcellation scheme, perhaps using evidence of spatial overlap? 

The correspondence of states between the two schemes was established using evidence of state expression time course. 

To assess temporal overlap, we calculated Pearson’s correlation between each candidate state obtained by the Yeo-7 scheme and the three predefined states obtained by the nine-network parcellation scheme in terms of state expression probabilities. The time courses of the 64 participants were concatenated, resulting in 19200 (300*64) time points for each state. The one that the candidate state most closely resembled was set to be its corresponding state. For instance, if a candidate state showed the highest correlation with State #1, it was labelled State #1 accordingly. As demonstrated in the confusion matrix, each of the three candidate states was best matched to State #1, State #2, and State #3, respectively, maintaining a one-to-one correspondence between the states from the two schemes.

We also assessed the spatial overlap between the two schemes. First, a state activity value was assigned to each voxel across the whole brain (including a total of 34,892 voxels covered by both parcellation schemes). This is done for each brain state. Next, we calculated Spearman’s correlation between each candidate state obtained by the Yeo-7 scheme and the three predefined states obtained by the nine-network scheme in terms of whole-brain activities. The pattern of spatial overlap is consistent with the pattern of temporal overlap, such that each of the three candidate states was best matched to State #1, State #2, and State #3, respectively.

Author response image 1.

We noted that the networks between the two schemes are not well aligned in their spatial location, especially for the DMN (as shown below). This may lead to the low spatial overlap of State #3, which is dominated by DMN activity. Consequently, establishing state correspondence based on temporal information is more appropriate in this context. We therefore only reported the results of temporal overlap in the manuscript. 

We have added a paragraph in the main text for “Establishing state correspondence between analyses” (Line 672-699). We have also updated the associated figures (Fig.S2, Fig.S3 and Fig.5)

Author response image 2.

(12) Line 839: gamma parameter, on a step size of? 

(16) Figure 3. Please add a legend in the "Sound envelope" graph what green and blue lines indicate. The authors write Coh(t) and Coh(t, t+1) at the top and Coh(t) and Coh(t+1) at the bottom. Please be consistent with the labeling. Shouldn't they be Coh(t-1, t) and Coh(t, t+1) to be exact for both? 

Have corrected. 

(17) In line 226, is this one-sample t-test compared to zero? If so, please write it inside the parentheses. In line 227, the authors write "slightly weaker"; however, since this is not statistically warranted, I suggest removing the word "slightly weaker" and just noting significance in both States 1 and 2.  

Have corrected.

(18) In line 288, please fix "we also whether". 

Have corrected. 

(19) In Figure 2C, what do pink lines in the transition matrix indicate? Are they colored just to show authors' interests, or do they indicate statistical significance? Please write it in the figure legend.   

Yes, the pink lines indicate a meaningful trend, showing that the between-state transition probabilities are significantly higher than those in permutation.

We have added this information to the figure legend. 

Reviewer #2 (Recommendations for the authors):

(1) It is unclear how the correspondence between states across different conditions and datasets was computed. Given the spatial autocorrelation of brain maps, I recommend reporting the Dice coefficient along with a spin-test permutation to test for statistical significance.  

The state correspondence between different conditions and between the two datasets are established using evidence of spatial overlap. The spatial overlap between states was quantified by Pearson’s correlation using the activity values (derived from HMM) of the nine networks. For each candidate state identified in other analyses (for the Rest, MG and older-adult datasets), we calculate the correlation between its network activity pattern and the three predefined states from the main analysis (for the young-adults dataset), and set the one it most closely resembled to be its matching state. For instance, if a candidate state showed the highest correlation with State #1, it was labelled State #1 accordingly. 

For the comparison between the young and older adults’ datasets (as shown below), the largest spatial overlap occurred along the diagonal of the confusion matrix, with high correlation values. This means that each of the three candidate states was best matched to State #1, State #2, and State #3, respectively, maintaining a one-to-one correspondence between the states from the two datasets. As the HMM is modelled at the level of networks which lack accurate coordinates, we did not apply the spin-test to assess the statistical significance of overlap. Instead, we extracted the state activity patterns from the 1000 permutations (wherein the original BOLD time courses were circularly shifted and an HMM was conducted) for the older-adults dataset. Applying the similar state-correspondence strategy, we generated a null distribution of spatial overlap. The real overlap of the three states was greater than and 97.97%, 95.34% and 92.39% instances from the permutation (as shown below). 

Author response image 3.

For the comparison of main task with the resting and the incomprehensible speech condition, there was some degree of confusion: there were two candidate states showing the highest similarity to State #2. In this case, we labeled the most similar candidate as State #2. The other candidate was then assigned to the predefined state with which it had the second-highest correlation. We used a prime symbol (e.g., State #3') to denote cases where such confusion occurred. These findings support our conclusion that the tripartite-organization of brain states is not a task-free, intrinsic property.

When establishing the correspondence between the Yeo-7 network and the ninenetwork parcellation schemes, we primarily relied on evidence from temporal overlap measures, as a clear network-level alignment between the two parcellation schemes is lacking. Temporal overlap was quantified by calculating the correlation of state occurrence probabilities between the two schemes. To achieve this, we concatenated the time courses of 64 participants, resulting in a time series consisting of 19,200 time points (300 time points per participant) for each state. Each of the three candidate states from the Yeo-7 network scheme was best matched to State #1, State #2, and State #3 from the main analyses, respectively. To determine the statistical significance of the temporal overlap, we circular shifted each participant’s time course of state expression obtained from the Yeo-7network scheme for 1000 times. Applying the same strategy to find the matching states, we generated a null distribution of overlap. The real overlap was much higher than the instances from permutation. 

Author response image 4.

In the revision, we have provided detailed description for how the state correspondence is established and reported the statistical significance of those correspondence (Line 671-699). The associated figures have also been updated (Fig.5, Fig. S2 and Fig.S3).  

(2) Please clarify if circle-shifting was applied to the state expression time course when generating the null distribution for behavior-brain state correlations reported in Figure (3). This seems important to control for the temporal autocorrelation in the time courses.  

We have updated the results by using circle-shifting to generated the null distribution. The results are largely consistent with the previous on without circular shifting (Line 230-242). 

(3) Figure 3: What does the green shaded area around the sound envelope represent? In the caption, specify whether the red line in the null distributions indicates the mean or median R between brain state expression and narrative features. It would also be beneficial to report this value in the main text. 

The green shaded area indicated the original amplitude of speech signal, while blue line indicates the smoothed, low-frequency contour of amplitude changes over time (i.e., speech envelope). We have updated the figure and explained this in the figure caption. 

The red line in the null distributions indicates the R between brain state expression and narrative features for the real data. and reported the mean R of the permutation in the main text. 

(4) The manuscript is missing a data availability statement (https://elifesciences.org/inside-elife/51839f0a/for-authors-updates-to-elife-s-datasharing-policies). 

We have added a statement of data availability in the revision, as follows: 

“The raw and processed fMRI data are available on OpenNeuro: https://openneuro.org/datasets/ds005623. The experimental stimuli, behavioral data and main scripts used in the analyses are provided on Github. ”

(5) There is a typo in line 102 ("perceptual alalyses"). 

Have corrected. 

We sincerely thank the two reviewers for their constructive feedback, thorough review, and the time they dedicated to improving our work.

Reference: 

Ahrends, C., Stevner, A., Pervaiz, U., Kringelbach, M. L., Vuust, P., Woolrich, M. W., & Vidaurre, D. (2022). Data and model considerations for estimating timevarying functional connectivity in fMRI. Neuroimage, 252, 119026. 

Ballenghein, U., Megalakaki, O., & Baccino, T. (2019). Cognitive engagement in emotional text reading: concurrent recordings of eye movements and head motion. Cognition and Emotion. 

Fernandino, L., Tong, J.-Q., Conant, L. L., Humphries, C. J., & Binder, J. R. (2022). Decoding the information structure underlying the neural representation of concepts. Proceedings of the national academy of sciences, 119(6), e2108091119. https://doi.org/10.1073/pnas.2108091119  

Hasson, U., Chen, J., & Honey, C. J. (2015). Hierarchical process memory: memory as an integral component of information processing. Trends in Cognitive Sciences, 19(6), 304-313. 

Lerner, Y., Honey, C. J., Silbert, L. J., & Hasson, U. (2011). Topographic mapping of a hierarchy of temporal receptive windows using a narrated story [Article]. Journal of Neuroscience, 31(8), 2906-2915. https://doi.org/10.1523/JNEUROSCI.3684-10.2011  

Liu, L., Li, H., Ren, Z., Zhou, Q., Zhang, Y., Lu, C., Qiu, J., Chen, H., & Ding, G. (2022). The “two-brain” approach reveals the active role of task-deactivated default mode network in speech comprehension. Cerebral Cortex, 32(21), 4869-4884. 

Liu, L., Zhang, Y., Zhou, Q., Garrett, D. D., Lu, C., Chen, A., Qiu, J., & Ding, G. (2020). Auditory–Articulatory Neural Alignment between Listener and Speaker during Verbal Communication. Cerebral Cortex, 30(3), 942-951. https://doi.org/10.1093/cercor/bhz138

Author response:

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

Reviewer #1 (Public Review):

The authors show that SVZ-derived astrocytes respond to a middle carotid artery occlusion (MCAO) hypoxia lesion by secreting and modulating hyaluronan at the edge of the lesion (penumbra) and that hyaluronan is a chemoattractant to SVZ astrocytes. They use lineage tracing of SVZ cells to determine their origin. They also find that SVZ-derived astrocytes express Thbs-4 but astrocytes at the MCAO-induced scar do not. Also, they demonstrate that decreased HA in the SVZ is correlated with gliogenesis. While much of the paper is descriptive/correlative they do overexpress Hyaluronan synthase 2 via viral vectors and show this is sufficient to recruit astrocytes to the injury. Interestingly, astrocytes preferred to migrate to the MCAO than to the region of overexpressed HAS2.

Strengths:

The field has largely ignored the gliogenic response of the SVZ, especially with regard to astrocytic function. These cells and especially newborn cells may provide support for regeneration. Emigrated cells from the SVZ have been shown to be neuroprotective via creating pro-survival environments, but their expression and deposition of beneficial extracellular matrix molecules are poorly understood. Therefore, this study is timely and important. The paper is very well written and the flow of results is logical.

Weaknesses:

The main problem is that they do not show that Hyaluronan is necessary for SVZ astrogenesis and or migration to MCAO lesions. Such loss of function studies have been carried out by studies they cite (e.g. Girard et al., 2014 and Benner et al., 2013). Similar approaches seem to be necessary in this work. 

We appreciate the comments by the reviewer. The article is, indeed, largely descriptive since we attempt to describe in detail what happens to newborn astrocytes after MCAO. Still, we have not attempted any modification to the model, such as amelioration of ischemic damage. This is a limitation of the study that we do not hide. However, we use several experimental approaches, such as lineage tracing and hyaluronan modification, to strengthen our conclusions.

Regarding the weaknesses found by the reviewer, we do not claim that hyaluronan is necessary for SVZ astrogenesis. Indeed, we observe that when the MCAO stimulus (i.e. inflammation) is present, the HMW-HA (AAV-Has2) stimulus is less powerful (we discuss this in line 330-332). We do claim, and we believe we successfully demonstrate, the reverse situation: that SVZ astrocytes modulate hyaluronan, not at the SVZ but at the site of MCAO, i.e. the scar. However, regarding whether hyaluronan is necessary for SVZ astrogenesis, we only show a correlation between its degradation and the time-course of astrogenesis. We suggest this result as a starting point for a follow-up study. We have included a phrase in the discussion (line 310), stating that further experiments are needed to fully establish a link between hyaluronan and astrogenesis in the SVZ.

Major points:

(1) How good of a marker for newborn astrocytes is Thbs4? Did you co-label with B cell markers like EGFr? Is the Thbs4 gene expressed in B cells? Do scRNAseq papers show it is expressed in B cells? Are they B1 or B2 cells?

We chose Thbs4 as a marker of newborn astrocytes based on published research (Beckervordersanforth et al., 2010; Benner et al., 2013; Llorens-Bobadilla et al. 2015, Codega et al, 2014; Basak et al., 2018; Mizrak et al., 2019; Kjell et al., 2020; Cebrian-Silla et al., 2021). From those studies, at least 3 associate Thbs4 to B-type cells based on scRNAseq data (LlorensBobadilla et al. 2015; Cebrian-Silla et al., 2021; Basak et al., 2018). We have included a sentence about this and the associated references, in line 92. 

We co-label Thbs4 with EGFR, but in the context of MCAO. We observed an increase of EGFR expression with MCAO, similar to the increase in Thbs4 alongside ischemia (see author ). We did not include this figure in the manuscript since we did not have available tissue from all the time points we used (7d, 60d post-ischemia). 

Author response image 1.

Thbs4 cells, in basal and ischemic conditions, only represent a small amount of IdU-positive cells (Fig 3F), suggesting that they are mostly quiescent cells, i.e., B1 cells. However, the scRNAseq literature is not consistent about this.

(2) It is curious that there was no increase in Type C cells after MCAO - do the authors propose a direct NSC-astrocyte differentiation?

Type C cells are fast-proliferating cells, and our BrdU/IdU experiment (Fig. 3) suggests that Thbs4 cells are slow-proliferating cells. Some authors suggest (Encinas lab, Spain) that when the hippocampus is challenged by a harsh stimulus, such as kainate-induced epilepsy, the NSCs differentiate directly into reactive astrocytes and deplete the DG neurogenic niche (Encinas et al., 2011, Cell Stem Cell; Sierra et al., 2015, Cell Stem Cell). We believe this might be the case in our MCAO model and the SVZ niche, since we observe a decrease in DCX labeling in the olfactory bulb (Fig S5) and an increase in astrocytes in the SVZ, which migrate to the ischemic lesion. We did not want to overcomplicate an already complicated paper, dwelling with direct NSC-astrocyte differentiation or with the reactive status of these newborn astrocytes. 

(3) The paper would be strengthened with orthogonal views of z projections to show colocalization.

We thank the reviewer for this observation. We have now included orthogonal projections in the critical colocalization IF of CD44 and hyaluronan (hyaluronan internalization) in Fig S6D, and a zoomed-in inset. Hyaluronan membrane synthesis is already depicted with orthogonal projection in Fig 6F.

(4) It is not clear why the dorsal SVZ is analysed and focused on in Figure 4. This region emanates from the developmental pallium (cerebral cortex anlagen). It generates some excitatory neurons early postnatally and is thought to have differential signalling such as Wnt (Raineteau group).

We decided to analyze in depth the dorsal SVZ after the BrdU experiment (Fig S3), where we observed an increase in BrdU+/Thbs4+ cells mostly in the dorsal area. Hence, the electrodes for electroporation were oriented in such a way as to label the dorsal area. We appreciate the paper by Raineteau lab, but we assume that this region may potentially exploit other roles (apart from excitatory neurons generated early postnatally) depending on the developmental stage (our model is in adults) and/or pathological conditions (MCAO). 

(5) Several of the images show the lesion and penumbra as being quite close to the SVZ. Did any of the lesions contact the SVZ? If so, I would strongly recommend excluding them from the analysis as such contact is known to hyperactivate the SVZ.

We thank the referee for the suggestion to exclude the harsher MCAO-lesioned animals from the analysis. Indeed, the MCAO ischemia, methodologically, can generate different tissue damages that cannot be easily controlled. Thus, based on TTC staining, we had already excluded the more severe tissue damage that contacted the SVZ, based on TTC staining.

(6) The authors switch to a rat in vitro analysis towards the end of the study. This needs to be better justified. How similar are the molecules involved between mouse and rat?

We chose the rat culture since it is a culture that we have already established in our lab, and that in our own hands, is much more reproducible than the mouse brain cell culture that we occasionally use (for transgenic animals only). Benito-Muñoz et al., Glia. 2016; Cavaliere et al., Front Cell Neurosci. 2013. It is true that there could be differences between the rat and mouse Thbs4-cell physiology, despite a 96% identity between rat and mouse Thbs4 protein sequence (BLASTp). In vitro, we only confirm the capacity of astrocytes to internalize hyaluronan, which was a finding that we did not expect in our in vivo experiments. Indeed, these observations, notwithstanding the obvious differences between in vivo and in vitro scenarios, suggest that the HA internalization by astrocytes is a cross-species event, at least in rodents. Regarding HA, hyaluronan is similar in all species, since it’s a glycan (this is why there are no antibodies against HA, and ones has to rely on binding proteins such as HABP to label it).

(7) Similar comment for overexpression of naked mole rat HA.

We chose the naked mole rat Hyaluronan synthase (HAS), because it is a HAS that produces HA of very high molecular weight, similar to the one found accumulated in the glial scar, at the lesion border. The naked-mole rat HAS used in mice (Gorbunova Lab) is a known tool in the ECM field. (Zhang et al, 2023, Nature; Tian et al., 2013, Nature).

Reviewer 1 (Recommendation to authors):

(1) Line 22: most of the cells that migrate out of the SVZ are not stem cells but cells further along in the lineage - neuroblasts and glioblasts.

We thank the reviewer for this clarification. We have modified the abstract accordingly. 

(2) In Figure 3d the MCAO group staining with GFAP looks suspiciously like ependymal cells which have been shown to be dramatically activated by stroke models.

The picture does show ependymal cells, which are located next to the ventricle and are indeed very proliferative in stroke. However, these cells do not express Thbs4 (Shah et al., 2018, Cell). In the quantifications from the SVZ of BrdU and IdU injected animals (Fig 3e and f), we only take into account Thbs4+ GFAP+ cells, no GFAP+ only. 

(3) The TTC injury shown in Figure 5c is too low mag.

We apologize for the low mag. We have increased the magnification two-fold without compromising resolution. The problem might also have arisen from the compression of TIF into JPEG in the PDF export process. We will address this in the revised version by carefully selecting export settings. The images we used are all publication quality (300 ppi).

(4) How specific to HA is HABP?

Hyaluronic Acid Binding Protein is a canonical marker for hyaluronan that is used also in ELISA to quantify it specifically, since it does not bind other glycosaminoglycans. The label has been used for years in the field for immunochemistry, and some controls and validations have been published: Deepa et al., 2006, JBC performed appropriate controls of HABP-biotin labeling using hyaluronidase (destroys labeling) and chondroitinase (preserves labeling). Soria et al., 2020, Nat Commun checked that (i) streptavidin does not label unspecifically, and (ii) that HABP staining is reduced after hyaluronan depletion in vivo with HAS inhibitor 4MU.

(5) A number of images are out of focus and thus difficult to interpret (e.g. SFig. 4e).

This is true. We realized that the PDF conversion process for the preprint version has severely compressed the larger images, such as the one found in Fig. S4e. We have submitted a revised version in a better-quality PDF (the final paper will have the original TIFF files). We apologize for the technical problem.

(6) "restructuration" is not a word.

We apologize for the mistake and thank the reviewer for the correction. We corrected “restructuration” with “reorganization” in line 67.

(7) While much of the manuscript is well-written and logical it could use an in-depth edit to remove awkward words and phrasings.

A native English speaker has revised the manuscript to correct these awkward phrases. All changes are labeled in red in the revised version.

(8) Please describe why and how you used skeleton analysis for HABP in the methods, this will be unfamiliar to most readers. The one-sentence description in the methods is insufficient.

We have modified the text accordingly, explaining in depth the logic behind the skeleton analysis. (Line 204). We also added several lines of text describing in detail the image analysis (CD44/HABP spots, fractal dimension, masks for membranal HABP, among others, in lines 484494) 

Reviewer #2 (Public Review)

Summary:

In their manuscript, Ardaya et al have addressed the impact of ischemia-induced gliogenesis from the adult SVZ and their effect on the remodeling of the extracellular matrix (ECM) in the glial scar. They use Thbs4, a marker previously identified to be expressed in astrocytes of the SVZ, to understand its role in ischemia-induced gliogenesis. First, the authors show that Thbs4 is expressed in the SVZ and that its expression levels increase upon ischemia. Next, they claim that ischemia induces the generation of newborn astrocyte from SVZ neural stem cells (NSCs), which migrate toward the ischemic regions to accumulate at the glial scar. Thbs4-expressing astrocytes are recruited to the lesion by Hyaluronan where they modulate ECM homeostasis.

Strengths:

The findings of these studies are in principle interesting and the experiments are in principle good.

Weaknesses:

The manuscript suffers from an evident lack of clarity and precision in regard to their findings and their interpretation.

We thank the reviewer for the valuable feedback. We hope the changes proposed improve clarity and precision throughout the manuscript.

(1) The authors talk about Thbs4 expression in NSCs and astrocytes, but neither of both is shown in Figure 1, nor have they used cell type-specific markers.

As we reported also to Referee #1 (major point 1), Thbs4 is widely considered in literature as a valid marker for newly formed astrocytes (Beckervordersanforth et al., 2010; Benner et al., 2013; Llorens-Bobadilla et al. 2015, Codega et al, 2014; Basak et al., 2018; Mizrak et al., 2019; Kjell et al., 2020; Cebrian-Silla et al., 2021). Some of the studies mentioned here and discussed in the manuscript text, also associate Thbs4 to B-type cells based on scRNAseq data (LlorensBobadilla et al. 2015; Cebrian-Silla et al., 2021; Basak et al., 2018). Moreover, we also showed colocalization of Thbs4 with activated stem cells marker nestin (Fig.2), glial marker GFAP (Fig. 3) and with dorsal NSCs marker tdTOM (from electroporation, Fig. 4). 

(2) Very important for all following experiments is to show that Thbs4 is not expressed outside of the SVZ, specifically in the areas where the lesion will take place. If Thbs4 was expressed there, the conclusion that Thbs4+ cells come from the SVZ to migrate to the lesion would be entirely wrong.

In Figure 1a, we show that Thbs4 is expressed in the telencephalon, exclusively in the neurogenic regions like SVZ, RMS and OB, together with cerebellum and VTA, which are likely not directly topographically connected to the damaged area (cortex and striatum). Regarding the origin of Thbs4+ cells, we demonstrated their SVZ origin by lineage tracking experiments after in vivo cell labeling (Fig. 4).

(3) Next, the authors want to confirm the expression level of Thbs4 by electroporation of pThbs4-eGFP at P1 and write that this results in 20% of total cells expressing GFP, especially in the rostral SVZ. I do not understand the benefit of this sentence. This may be a confirmation of expression, but it also shows that the GFP+ cells derive from early postnatal NSCs.

Furthermore, these cells look all like astrocytes, so the authors could have made a point here that indeed early postnatal NSCs expressing Thbs4 generate astrocytes alongside development. Here, it would have been interesting to see how many of the GFP+ cells are still NSCs.

We thank the reviewer for this useful remark. We have rephrased this paragraph in the results section (Line 99).

(4) In the next chapter, the authors show that Thbs4 increases in expression after brain injury. I do not understand the meaning of the graphs showing expression levels of distinct cell types of the neuronal lineage. Please specify why this is interesting and what to conclude from that.

Also here, the expression of Thbs4 should be shown outside of the SVZ as well.

In Fig 2, we show the temporal expression of two markers (besides Thbs4) in the SVZ. Nestin and DCX are the gold standard markers for NSCs, with DCX present in neuroblasts. This is already explained in line 119. What we didn’t explain, and now we say in line 124, is that Nestin and DCX decrease immediately after ischemia (7d time-point). This probably means that the NSCs stop differentiating into neuroblast to favor glioblast formation. This is also supported by the experiments in the olfactory bulb depicted in Fig. S5C-H.

(5) Next, the origin of newborn astrocytes from the SVZ upon ischemia is revealed. The graphs indicate that the authors perfused at different time points after tMCAO. Did they also show the data of the early time points? If only of the 30dpi, they should remove the additional time points indicated in the graph. In line 127 they talk about the origin of newborn astrocytes. Until now they have not even mentioned that new astrocytes are generated. Furthermore, the following sentences are imprecise: first they write that the number of slow proliferation NSCs is increased, then they talk about astrocytes. How exactly did they identify astrocytes and separate them from NSCs? Morphologically? Because both cell types express GFAP and Thbs4.

The same problem also occurs throughout the next chapter.

We thank the reviewer for this interesting comment. The experiment in Fig 3 combines BrdU and IdU. This is a tricky experiment, since chronic BrdU is normally analyzed after 30d, since the experimenter must wait for the wash out of BrdU (it labels slow-proliferating cells). Since we also wanted to label fast proliferative cells with IdU, we used IP injections of this nucleotide at the different time points, and perfused the day after. It wouldn’t make sense to show BrdU at earlier time points. We do so in Fig 3e, just to colocalize with Thbs4 to read the tendency of the experiment. However, the quantification of BrdU (not of IdU) is done only at 30 DPI, which is explained in the methods (line 407).

“In line 127, they talk about the origin of newborn astrocytes…” 

Indeed, we wanted to introduce in the paragraph title that ischemia induced the generation of new astrocytes, which is more clearly described in the text. We changed the paragraph title with “Characterization of Ischemia-induced cell populations”

“How exactly did they identify astrocytes and separate them from NSC?” 

With this experiment and using two different protocols to label proliferating cells (BrdU vs IdU) we wanted to track the precursor cells that derivate to astrocytes and that already expressed the marker Thbs4. Indeed, the different increase and rate of proliferation is only related to the progenitor cells that lately will differentiate in astrocytes. In this experiment we only referred to the astrocytes in the last sentence “These results suggest that, after ischemia, Thbs4positive astrocytes derive from the slow proliferative type B cells”

(6) "These results suggest that ischemia-induced astrogliogenesis in the SVZ occurs in type B cells from the dorsal region, and that these newborn Thbs4-positive astrocytes migrate to the ischemic areas." This sentence is a bit dangerous and bares at least one conceptual difficulty: if NSCs generate astrocytes under normal conditions and along the cause of postnatal development (which they do), then local astrocytes  (expressing the tdTom because they stem from a postnatal NSC ), may also react to MCAO and proliferate locally. So the astrocytes along the scar do not necessarily come from adult NSCs upon injury but from local astrocytes.  If the authors state that NSCs generate astrocytes that migrate to the lesion, I would like to see that no astrocytes inside the striatum carry the tdTom reporter before MCAO is committed.

We understand the referee’s concern about the postnatal origin of astrocytes that can also be labeled with tdTom. Our hypothesis, tested at the beginning of the paper, is that SVZ-derived astrocytes derive from slow proliferative NSC. Thus, it is reasonable that Tom+ cells can reach the cortical region in such a short time frame. This is why we assumed that local astrocytes can’t be positive for tdTom. We characterized the expression of tfTom in sham animals and we observed few tdTom+ cells in the cortex and striatum (Author response image 2 and Figure S4). The expression of tdTom mainly remains in the SVZ and the corpus callosum under physiological conditions. However, proliferation of local astrocytes labeled with tdTom expression (early postnatally astrocytes) could explain the small percentage of tdTom+ cells in the ischemic regions that do not express Thbs4, even though this percentage could represent other cell types such as OPCs or oligodendrocytes. 

Author response image 2.

(7) If astrocytes outside the SVZ do not express Thbs4, I would like to see it.  Otherwise, the discrimination of SVZ-derive GFAP+/Thbs4+ astrocytes and local astrocytes expressing only GFAP is shaky.

Regarding Thbs4 outside the SVZ, we already answered this in point 2 (please refer to Fig 1A). We also quantified the expression of Thbs4+/GFAP+ astrocytes in the corpus callosum, cortex and striatum of sham and MCAO mice (Figure S5a-b) and we did not observe that local astrocytes express Thbs4 under physiological conditions.

(8) Please briefly explain what a Skeleton analysis and a Fractal dimension analysis is, and what it is good for.

We apologized for the brief information on Skeleton and Fractal dimension analysis. We included a detailed explanation of these analyses in methods (line 484-494).

(9) The chapter on HA is again a bit difficult to follow. Please rewrite to clarify who produces HA and who removes it by again showing all astrocyte subtypes (GFAP+/Thbs4+ and GFAP+/Thbs4-).

We apologize for the lack of clarity. We rewrote some passages of those chapters (changes in red), trying to convey the ideas more clearly. We also changed a panel in Figure S6b-c to clarify all astrocytes subtypes that internalize hyaluronan (Thbs4+/GFAP+ and Thbs4-/GFAP+). See Author response image 3.

Author response image 3.

(10) Why did the authors separate dorsal, medial, and ventral SVZ so carefully? Do they comment on it? As far as I remember, astrogenesis in physiological conditions has some local preferences (dorsal?)

We performed the electroporation protocol in the dorsal SVZ based on previous results (Figure 3 and Figure S3). NSC produce specific neurons in the olfactory bulb according to their location in the SVZ. However, postnatal production of astrocytes mainly occurs through local astrocytes proliferation and the SVZ contribution is very limited at this time point. 

Reviewer #3 (Public Review)

Summary:

The authors aimed to study the activation of gliogenesis and the role of newborn astrocytes in a post-ischemic scenario. Combining immunofluorescence, BrdU-tracing, and genetic cellular labelling, they tracked the migration of newborn astrocytes (expressing Thbs4) and found that Thbs4-positive astrocytes modulate the extracellular matrix at the lesion border by synthesis but also degradation of hyaluronan. Their results point to a relevant function of SVZ newborn astrocytes in the modulation of the glial scar after brain ischemia. This work's major strength is the fact that it is tackling the function of SVZ newborn astrocytes, whose role is undisclosed so far.

Strengths:

The article is innovative, of good quality, and clearly written, with properly described Materials and Methods, data analysis, and presentation. In general, the methods are designed properly to answer the main question of the authors, being a major strength. Interpretation of the data is also in general well done, with results supporting the main conclusions of this article.

Weaknesses:

However, there are some points of this article that still need clarification to further improve this work.

(1) As a first general comment, is it possible that the increase in Thbs4-positive astrocytes can also happen locally close to the glia scar, through the proliferation of local astrocytes or even from local astrocytes at the SVZ? As it was shown in published articles most of the newborn astrocytes in the adult brain actually derive from proliferating astrocytes, and a smaller percentage is derived from NSCs. How can the authors rule out a contribution of local astrocytes to the increase of Thbs4-positive astrocytes? The authors also observed that only about one-third of the astrocytes in the glial scar derived from the SVZ.

We thank the reviewer for the interesting comment. We have extended the discussion about this topic in the manuscript, (lines 333-342), including the statement about a third of glial scar astrocytes being from the SVZ and not downplaying the role of local astrocytes.  Whether the glial scar is populated by newborn astrocytes derived from SVZ or from local astrocytes is under debate, since there are groups that found astrocytes contribution from local astrocytes (Frisèn group, Magnusson et al., 2014) but there are others that observed the opposite (Li et al., 2010; Benner et al., 2013; Faiz et al., 2015; Laug et al., 2019 & Pous et al., 2020). 

In our study we observed that Thbs4 expression is almost absent in the cortex and striatum of sham mice. To demonstrate that new-born astrocytes are derived from SVZ we used two techniques: the chronic BrdU treatment and the cell tracing which mainly labels SVZ neural stem cells. Fast proliferating cells lose BrdU quickly so local astrocytes under ischemic conditions do not express BrdU. In addition, we injected IdU the day before perfusion in order to see if local astrocytes express Thbs4 when they respond to the brain ischemia. However, we did not observe proliferating local astrocytes expressing Thbs4 after MCAO (see Author response image 4)

Author response image 4.

As mentioned in the response for reviewer 2, the cell tracing technique could label early postnatal astrocytes. We characterized the technique and only a small percentage of tdTom expression was found in the cortex and striatum of sham animals.  This tdTom population could explain the percentage of tdTom+ cells in the ischemic regions that do not express Thbs4 even though this percentage could represent other cell types such as OPCs or oligodendrocytes. Taking all together, evidences suggest that Thbs4+ astrocyte population derived from the SVZ. 

We indeed observed a small contribution of Thbs4+ astrocytes to the glial scar. However, Thbs4+ astrocytes arrive at the lesion at a critical temporal window - when local hyper-reactive astrocytes die or lose their function. We hypothesized that Thbs4+ astrocytes could help local astrocytes or replace them in reorganizing the extracellular space and the glial scar, an instrumental process for the recovery of the ischemic area. 

(2) It is known that the local, GFAP-reactive astrocytes at the scar can form the required ECM. The authors propose a role of Thbs4-positive astrocytes in the modulation, and perhaps maintenance, of the ECM at the scar, thus participating in scar formation likewise. So, this means that the function of newborn astrocytes is only to help the local astrocytes in the scar formation and thus contribute to tissue regeneration. Why do we need specifically the Thbs4positive astrocytes migrating from the SVZ to help the local astrocytes? Can you discuss this further?

Unfortunately, we could not demonstrate which molecular machinery is involved in these mechanisms, and we can only speculate the functional meaning of a second wave of glial activation. We added a lengthy discussion in lines 333-342.

(3) The authors observed that the number of BrdU- and DCX-positive cells decreased 15 dpi in all OB layers (Fig. S5). They further suggest that ischemia-induced a change in the neuroblasts ectopic migratory pathway, depriving the OB layers of the SVZ newborn neurons. Are the authors suggesting that these BrdU/DCX-positive cells now migrate also to the ischemic scar, or do they die? In fact, they see an increase in caspase-3 positive cells in the SVZ after ischemia, but they do not analyse which type of cells are dying. Alternatively, is there a change in the fate of the cells, and astrogliogenesis is increased at the expense of neurogenesis?  The authors should understand which cells are Cleaved-caspase-3 positive at the SVZ and clarify if there is a change in cell fate. Also please clarify what happens to the BrdU/DCX-positive cells that are born at the SVZ but do not migrate properly to the OB layers.

Actually, we cannot demonstrate the fate of missing BrdU/DCX cells in the OB. We can reasonably speculate that following the ischemic insult, the neurogenic machinery steers toward investing more energy in generating glial cells to support the lesion. We didn’t analyze the fate of the DCX that originally should migrate and differentiate to the OB, whether they die or if there is a shift in the differentiation program in the SVZ, since we consider that question is out of the study’s scope.   

(4) The authors showed decreased Nestin protein levels at 15 dpi by western blot and immunostaining shows a decrease already at 7div (Figure 2). These results mean that there is at least a transient depletion of NSCs due to the promotion of astrogliogenesis. However, the authors show that at 30dpi there is an increase of slow proliferating NSCs (Figure 3). Does this mean, that there is a reestablishment of the SVZ cytogenic process?  How does it happen, more specifically, how NSCs number is promoted at 30dpi?  Please explain how are the NSCs modulated throughout time after ischemia induction and its impact on the cytogenic process.

Based on the chronic BrdU treatment, results suggested a restoration of SVZ cytogenic process (also observed in the nestin and DCX proteins expression at 30dpi). However, we did not analyze how it happens (from asymmetric or symmetric divisions). As suggested by Encinas group, we hypothesized that the brain ischemia induces the exhaustion of the neurogenic niche of the SVZ by symmetric divisions of NSC into reactive astrocytes.

(5) The authors performed a classification of Thbs4-positive cells in the SVZ according to their morphology. This should be confirmed with markers expressed by each of the cell subtypes.

We thank the referee for the comment. Classifying NSC based on different markers could also be tricky because different NSC cell types share markers. This classification was made considering the specific morphology of each NSC cell type. In addition, Thbs4 expression in Btype cells is also observed in other studies (Llorens-Bobadilla et al. 2015; Cebrian-Silla et al.,

2021; Basak et al., 2018).

(6) In Figure S6, the authors quantified HABP spots inside Thbs4-positive astrocytes. Please show a higher magnification picture to show how this quantification was done.

We quantified HABP area and HABP spots inside Thbs4+ astrocytes with a custom FIJI script.

Thbs4 cell mask was done via automatic thresholding within the GFAP cell mask. Threshold for HABP marker was performed and binary image was processed with 1 pixel median filter (to eliminate 1 px noise-related spots). “Analyze particles” tool was used to sort HABP spots in the cell ROI. HABP spot number per compartment and population was exported to excel and data was normalized dividing HABP spots per ROI by total HABP spots. See Author response image 5.

Author response image 5.

Author Response

We thank all three Reviewers and the editors for the time and effort they put in reading and critiquing the manuscript. Our revised manuscript includes new data and analyses that address the original concerns. These include, 1) a new Supplemental Figure characterizing Cre expression and cellular phenotypes in the hippocampus, 2) new tables that give a more comprehensive picture of the EEG recordings and statistical analyses, 3) addition of whole cell electrophysiology data, and 4) rewriting to ensure that we do not state that either mTORC1 or mTORC2 hyperactivation is sufficient to cause epilepsy. We discuss the issue of statistical power to detect reduction in generalized seizure rate in the responses below. These suggestions and additions have improved the paper and we hope they will raise both significance and strength of support for the conclusions.

Reviewer #1 (Public Review):

Hyperactivation of mTOR signaling causes epilepsy. It has long been assumed that this occurs through overactivation of mTORC1, since treatment with the mTORC1 inhibitor rapamycin suppresses seizures in multiple animal models. However, the recent finding that genetic inhibition of mTORC1 via Raptor deletion did not stop seizures while inhibition of mTORC2 did, challenged this view (Chen et al, Nat Med, 2019). In the present study, the authors tested whether mTORC1 or mTORC2 inhibition alone was sufficient to block the disease phenotypes in a model of somatic Pten loss-of-function (a negative regulator of mTOR). They found that inactivation of either mTORC1 or mTORC2 alone normalized brain pathology but did not prevent seizures, whereas dual inactivation of mTORC1 and mTORC2 prevented seizures. As the functions of mTORC1 versus mTORC2 in epilepsy remain unclear, this study provides important insight into the roles of mTORC1 and mTORC2 in epilepsy caused by Pten loss and adds to the emerging body of evidence supporting a role for both complexes in the disease development.

Strengths:

The animal models and the experimental design employed in this study allow for a direct comparison between the effects of mTORC1, mTORC2, and mTORC1/mTORC2 inactivation (i.e., same animal background, same strategy and timing of gene inactivation, same brain region, etc.). Additionally, the conclusions on brain epileptic activity are supported by analysis of multiple EEG parameters, including seizure frequencies, sharp wave discharges, interictal spiking, and total power analyses.

Weaknesses:

(1) The sample size of the study is small and does not allow for the assessment of whether mTORC1 or mTORC2 inactivation reduces seizure frequency or incidence. This is a limitation of the study.

We agree that this is a minor limitation of the present study, however, for several reasons we decided not to pursue this question by increasing the number of animals. First, we performed a power analysis of the existing data. This analysis showed that we would need to use 89 animals per group to detect a significant difference (0.8 Power, p= 0.05, Mann-Whitney test) in the frequency of generalized seizures in the Pten-Raptor group and 31 animals per group in the Pten-Rictor group versus Pten alone. It is simply not feasible to perform video-EEG monitoring on this many animals for a single study. Second, even if we did do enough experiments to detect a reduction in seizure frequency, it is clear that neither Rptor nor Rictor deletion provides the kind normalization in brain activity that we seek in a targeted treatment. Both Pten-Rptor and Pten-Rictor animals still have very frequent spike-wave events (Fig. 3D) and highly abnormal interictal EEGs (Fig. 4), suggesting that even if generalized seizures were reduced, epileptic brain activity persists. This is in contrast to the triple KO animals, which have no increase in SWD above control level and very normal interictal EEG.

(2) The authors describe that they inactivated mTORC1 and mTORC2 in a new model of somatic Pten loss-of-function in the cortex. This is slightly misleading since Cre expression was found both in the cortex and the underlying hippocampus, as shown in Figure 1. Throughout the manuscript, they provide supporting histological data from the cortex. However, since Pten loss-of-function in the hippocampus can lead to hippocampal overgrowth and seizures, data showing the impact of the genetic rescue in the hippocampus would further strengthen the claim that neither mTORC1 nor mTORC2 inactivation prevents seizures.

Thank you for pointing out this issue. Cre expression was observed in both the cortex and the dorsal hippocampus in most animals, and we agree that differences in cortical versus hippocampal mTOR signaling could have differential contributions to epilepsy. We initially focused our studies on the cortex because spike-and-wave discharge, the most frequent and fully penetrant EEG phenotype in our model, is associated with cortical dysfunction. In our revised submission we have included a new Figure that quantifies Cre expression in the hippocampal subfields, as well as pS6, pAkt and soma size. These new data show that the amount of Cre expression in the hippocampus is not related to the occurrence of generalized seizures. The pattern of cell size changes in hippocampal neurons is the same as observed in cortical neurons. The levels of pS6 and pAkt are not much changed in the hippocampus, likely due to the sparse Cre expression there. We interpret these findings as supporting the conclusion that the reason we do not see seizure prevention by mTORC1 or mTORC2 inactivation is not due to hippocampal-specific dysfunction.

(3)Some of the methods for the EEG seizure analysis are unclear. The authors describe that for control and Pten-Raptor-Rictor LOF animals, all 10-second epochs in which signal amplitude exceeded 400 μV at two time-points at least 1 second apart were manually reviewed, whereas, for the Pten LOF, Pten-Raptor LOF, and Pten-Rictor LOF animals, at least 100 of the highest- amplitude traces were manually reviewed. Does this mean that not all flagged epochs were reviewed? This could potentially lead to missed seizures.

We reviewed at least 48 hours of data from each animal manually. All seizures that were identified during manual review were also identified by the automated detection program. It is possible but unlikely that there are missed seizures in the remaining data. We have added these details to the Methods of the revised submission.

(4) Additionally, the inclusion of how many consecutive hours were recorded among the ~150 hours of recording per animal would help readers with the interpretation of the data.

Thank you for this recommendation. Our revised submission includes a table with more information about the EEG recordings in the revised version of the manuscript. The number of consecutive hours recorded varied because the wireless system depends on battery life, which was inconsistent, but each animal was recorded for at least 48 consecutive hours on at least two occasions.

(5) Finally, it is surprising that mTORC2 inactivation completely rescued cortical thickness since such pathological phenotypes are thought to be conserved down the mTORC1 pathway. Additional comments on these findings in the Discussion would be interesting and useful to the readers.

We agree that the relationship between mTORC2, cortical thickness, and growth in general is an interesting topic with conflicting results in the literature. We didn’t add anything to the Discussion along these lines because we are up against word limits, but comment here that soma size was increased 120% by Pten inactivation and partially normalized to a 60% increase from Controls by mTORC2 inactivation (Fig. 2C). We and others have previously shown that mTORC2 inactivation (Rictor deletion) in neurons reduces brain size, neuron soma size, and dendritic outgrowth (PMIDs: 36526374, 32125271, 23569215). In our revised submission we also include new data showing that the membrane capacitance of Pten-Ric LOF neurons is normal. Thus, we do not find it completely surprising that mTORC2 inactivation reduces the cortical thickness increase caused by Pten loss. There may still be a slight increase in cortical thickness in Pten-Rictor animals, but it is statistically indistinguishable from Controls.

Reviewer #2 (Public Review):

Summary:

The study by Cullen et al presents intriguing data regarding the contribution of mTOR complex 1 (mTORC1) versus mTORC2 or both in Pten-null-induced macrocephaly and epileptiform activity. The role of mTORC2 in mTORopathies, and in particular Pten loss-off-function (LOF)-induced pathology and seizures, is understudied and controversial. In addition, recent data provided evidence against the role of mTORC1 in PtenLOF-induced seizures. To address these controversies and the contribution of these mTOR complexes in PtenLOF-induced pathology and seizures, the authors injected a AAV9-Cre into the cortex of conditional single, double, and triple transgenic mice at postnatal day 0 to remove Pten, Pten+Raptor or Rictor, and Pten+raptor+rictor. Raptor and Rictor are essentially binding partners of mTORC1 and mTORC2, respectively. One major finding is that despite preventing mild macrocephaly and increased cell size, Raptor knockout (KO, decreased mTORC1 activity) did not prevent the occurrence of seizures and the rate of SWD event, and aggravated seizure duration. Similarly, Rictor KO (decreased mTORC2 activity) partially prevented mild macrocephaly and increased cell size but did not prevent the occurrence of seizures and did not affect seizure duration. However, Rictor KO reduced the rate of SWD events. Finally, the pathology and seizure/SWD activity were fully prevented in the double KO. These data suggest the contribution of both increased mTORC1 and mTORC2 in the pathology and epileptic activity of Pten LOF mice, emphasizing the importance of blocking both complexes for seizure treatment. Whether these data apply to other mTORopathies due to Tsc1, Tsc2, mTOR, AKT or other gene variants remains to be examined.

Strengths:

The strengths are as follows: 1) they address an important and controversial question that has clinical application, 2) the study uses a reliable and relatively easy method to KO specific genes in cortical neurons, based on AAV9 injections in pups. 2) they perform careful video-EEG analyses correlated with some aspects of cellular pathology.

Weaknesses:

The study has nevertheless a few weaknesses: 1) the conclusions are perhaps a bit overstated. The data do not show that increased mTORC1 or mTORC2 are sufficient to cause epilepsy. However the data clearly show that both increased mTORC1 and mTORC2 activity contribute to the pathology and seizure activity and as such are necessary for seizures to occur.

We agree that our findings do not directly show that either mTORC1 or mTORC2 hyperactivity are sufficient to cause seizures, as we do not individually hyperactivate each complex in the absence of any other manipulation. We interpreted our findings in this model as suggesting that either is sufficient based on the result that there is no epileptic activity when both are inactivated, and thus assume that there is not a third, mTOR-independent, mechanism that is contributing to epilepsy in Pten, Pten-Raptor, and Pten-Rictor animals. In addition, the histological data show that Raptor and Rictor loss each normalize activity through mTORC1 and mTORC2 respectively, suggesting that one in the absence of the other is sufficient. However, we agree that there could be other potential mTOR-independent pathways downstream of Pten loss that contribute to epilepsy. We have revised the manuscript to reflect this.

(2) The data related to the EEG would benefit from having more mice. Adding more mice would have helped determine whether there was a decrease in seizure activity with the Rictor or Raptor KO.

Please see response to Reviewer 1’s first Weakness.

(3) It would have been interesting to examine the impact of mTORC2 and mTORC1 overexpression related to point #1 above.

We are not sure that overexpression of individual components of mTORC1 or mTORC2 would result in their hyperactivation or lead to increases in downstream signaling. We believe that cleanly and directly hyperactivating mTORC1 or especially mTORC2 in vivo without affecting the other complex or other potential interacting pathways is a difficult task. Previous studies have used mTOR gain-of-function mutations as a means to selectively activate mTORC1 or pharmacological agents to selectively activate mTORC2, but it not clear to us that the former does not affect mTORC2 activity as well, or that the latter achieves activation of mTORC2 targets other than p-Akt 473, or that it is truly selective. We agree that these would be key experiments to further test the sufficiency hypothesis, but that the amount of work that would be required to perform them is more that what we can do in this Short Report.

Reviewer #3 (Public Review):

Summary: This study investigated the role of mTORC1 and 2 in a mouse model of developmental epilepsy which simulates epilepsy in cortical malformations. Given activation of genes such as PTEN activates TORC1, and this is considered to be excessive in cortical malformations, the authors asked whether inactivating mTORC1 and 2 would ameliorate the seizures and malformation in the mouse model. The work is highly significant because a new mouse model is used where Raptor and Rictor, which regulate mTORC1 and 2 respectively, were inactivated in one hemisphere of the cortex. The work is also significant because the deletion of both Raptor and Rictor improved the epilepsy and malformation. In the mouse model, the seizures were generalized or there were spike-wave discharges (SWD). They also examined the interictal EEG. The malformation was manifested by increased cortical thickness and soma size.

Strengths: The presentation and writing are strong. The quality of data is strong. The data support the conclusions for the most part. The results are significant: Generalized seizures and SWDs were reduced when both Torc1 and 2 were inactivated but not when one was inactivated.

Weaknesses: One of the limitations is that it is not clear whether the area of cortex where Raptor or Rictor were affected was the same in each animal.

Our revised submission includes new data showing that the area of affected cortex and hippocampus are similar across groups. (Figure 1A and Supplementary Figure 1)

Also, it is not clear which cortical cells were measured for soma size.

Soma size was measured by dividing Nissl stain images into a 10 mm2 grid. The somas of all GFP-expressing cells fully within three randomly selected grid squares in Layer II/III were manually traced. Three sections per animal at approximately Bregma -1.6, -2,1, and -2.6 were used. As Cre expression was driven by the hSyn promoter these cells include both excitatory and inhibitory cortical neurons.

Another limitation is that the hippocampus was affected as well as the cortex. One does not know the role of cortex vs. hippocampus. Any discussion about that would be good to add.

See response to Reviewer 1’s second Weakness.

It would also be useful to know if Raptor and Rictor are in glia, blood vessels, etc.

Raptor and Rictor are thought to be ubiquitously active in mammalian cells including glia and endothelial cells. Previous studies have shown that mTOR manipulation can affect astrocyte function and blood vessel organization, however, our study induced gene knockout using an AAV that expressed Cre under control of the hSyn promoter, which has previously been shown to be selective for neurons. Manual assessment of Cre expression compared with DAPI, NeuN, and GFAP stains suggested that only neurons were affected.

Recommendations for the authors: please note that you control which revisions to undertake from the public reviews and recommendations for the authors

Reviewer #1 (Recommendations For The Authors):

In addition to the comments in the public review, it is recommended that the authors provide a more representative figure for p-Akt staining in the Pten LOF condition in Figure 1 D2. The current figure is not convincing.

Thanks for the suggestion. We have replaced the images with zoomed in panels that beter demonstrate the difference.

Additionally, in the last paragraph of the discussion, there is a reference error to an incorrect paper (reference 18) that should be corrected.

Thanks, corrected.

Reviewer #2 (Recommendations For The Authors):

Major comments:

Comment 1: Some statements need clarifications or changes.

(1) Abstract: "spontaneous seizures and epileptiform activity persisted despite mTORC1 or mTORC2 inactivation alone but inactivating both mTORC1 and mTORC2 normalized pathology." Did inactivation of one only also normalized the pathology? Did inactivating both normalized the seizures? Pathology is not equal to seizures.

We have altered this statement to avoid ambiguity.

(2) Abstract: "These results suggest that hyperactivity of both mTORC1 and mTORC2 are sufficient to cause epilepsy,". Based on the abstract, it is not clear that it is sufficient. It is necessary.

We have altered this statement by removing the term “sufficient.”

(3) "Thus, there is strong evidence that hyperactivation of mTORC1 downstream of PTEN disruption causes the macrocephaly, epilepsy, early mortality, and synaptic dysregulation observed in humans and model organisms [17]" I would suggest adding that the strongest evidence is that mTOR GOF mutations lead to the same pathology and epilepsy, suggesting mTORC1 is sufficient. The other findings suggest that it is necessary.

Unless we misunderstand the Reviewer’s point, we believe this viewpoint is already encompassed by the proceeding text that “These phenotypes resemble those observed in models of mTORC1- specific hyperactivation.”

(4) Introduction (end): "suggesting that hyperactivity of either complex can lead to neuronal hyperexcitability and epilepsy".

Comment 2: I do not agree with the title based on comment 1 above. You did not provide evidence that the mTORCs cause seizures. Your data suggest that they are necessary for seizures or contribute to seizures, but there is no evidence that mTORC2 can induce seizure.

We softened the title by replacing “cause” with “mediate.”

Comment 3: Fig. 1B. Could you beter describe the affected regions. I can see other regions than just the cortex and hippocampus.

Almost all affected cell bodies were in the cortex and hippocampus. The virus in the image is cell-filling and as such projections from affected neurons throughout the brain can also be seen. We have clarified this in the figure legend.

Comment 4: I feel unease about the number of animals recorded for EEG to assess seizure frequency. There is not enough power to draw clear conclusions. So, please make sure to not oversell your findings since it is all-or-nothing data (seizure or no seizure) in this case and the seizure frequency could very well be decreased with single mTOR LOF, but it is impossible to conclude. Maybe discuss this limitation of your study.

We have addressed this in the public comments response.

Minor:

(1) Pten LOF: define the abbreviation.

Done

(2) Make sure that gene name in mice are not capitalized and italicized.

OK

(3) Fig 1C: could you specify in the results where the analysis was done.

Detail added to Methods (to keep Results concise for word limit)

(4) In the subtitle: "Concurrent mTORC1/2 inactivation, but neither alone, rescues epilepsy and interictal EEG abnormalities in focal Pten LOF". Replace "rescues" but prevents. This is not a rescue experiment since the LOF is done at the same time.

OK

(5) "GS did not appear to be correlated with mTOR pathway activity (Supplementary Figure 2)." Please can you do proper correlation analysis, by plotting all the values as a function of seizure frequency independent of the condition? There is also no correlation between pAKt and seizures.

Here are those data in Author response image 1. They are now part of Supplementary Figure 2.

Author response image 1.

Reviewer #3 (Recommendations For The Authors):

Figures 1 D, and E show images that are too small to judge. Where are the layers? Please add marks.

We replaced these images with larger zoomed in images to show group differences more clearly. The images no longer show multiple differentiable cortical layers.

If Fig 1 characterizes the model, where is the seizure data? When did they start? Where did they start? Was the focus of the cortical area affected by PTEN loss of function?

Updated figure name to reflect content. Information about the seizure phenotypes is included in Figure 3.

Figure 2 The font size for the calibration is too small. The correlations are hard to see. Colors are not easy to discriminate.

We edited the figure to correct these problems.

Figure 3 shows a clear effect on generalized seizures but the text of the Results does not reflect that.

We wanted to be cautious about interpreting these data based on the issue raised by other reviewers that they are underpowered to detect seizure reduction in the Pten-Raptor and Pten-Rictor groups. We have updated the language to atempt to strike a beter balance between over- and under-interpretation. We also performed an additional analysis of the occurrence of generalized seizures to emphasize that only Control and PtRapRic animals have significantly lower seizure occurrence that Pten LOF mice (Fig 3C).

For interictal power, was the same behavioral state chosen? Was a particular band affected?

Epochs to be analyzed were selected automatically and were agnostic to behavioral state. Band-specific effects are outlined in Figure 4B and Table [2].

There is no information about whether the model exhibits altered sleep, food intake, weight, etc.

We didn’t collect information on food intake. It would be possible to look at sleep from the EEG, but that is not something that we are prepared to do at this point. Weight at endpoint was not different between genotypes but we did not collect longitudinal data on weight.

Were the sexes different?

Included in new Table [1]

Where were EEG electrodes and were they subdural or not?

Additional detail on this has been added to Methods. The screws are placed in the skull but above the dura.

How long were continuous EEG records- the method just says 150 hr. per mouse in total.

Included in new Table [1]

The statistics don't discuss power, normality, whether variance was checked to ensure it did not differ significantly between groups, or whether data are mean +- sem or sd. For ANOVAs, were there multifactorial comparisons and what were F, df, and p values? Exact p for post hoc tests?

We have added a new table (Table [3]) that gives information on the exact test used, F, p values, and exact p for post hoc tests. Information regarding power, normality, variance, post- tests and multiple comparisons corrections have been added to Methods section “Statistical Analysis.”

Author Response

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

eLife assessment

This valuable study addresses the long-term effect of warming and altered precipitation on microbial growth, as a proxy for understanding the impact of global warming. While the methods are compelling and the evidence supporting the claims is solid, additional analysis of the data would strengthen the study, which should be of broad interest to microbial ecologists and microbiologists.

We sincerely appreciate your assessment and thoughtful comments, which are valuable and very helpful for improving our manuscript. We have carefully considered all comments, and made extensive, thorough corrections and additional analysis of the data, which we hope to meet with approval.

Reviewer #1 (Public Review):

Warming and precipitation regime change significantly influences both above-ground and below-ground processes across Earth's ecosystems. Soil microbial communities, which underpin the biogeochemical processes that often shape ecosystem function, are no exception to this, and although research shows they can adapt to this warming, population dynamics and ecophysiological responses to these disturbances are not currently known. The Qinghai-Tibet Plateau, the Third Pole of the Earth, is considered among the most sensitive ecosystems to climate change. The manuscript described an integrated, trait-based understanding of these dynamics with the qSIP data. The experimental design and methods appear to be of sufficient quality. The data and analyses are of great value to the larger microbial ecological community and may help advance our understanding of how microbial systems will respond to global change. There are very few studies in which the growth rates of bacterial populations from multifactorial manipulation experiments on the Qinghai-Tibet Plateau have been investigated via qSIP, and the large quantity of data that comprises the study described in this manuscript, will substantially advance our knowledge of bacterial responses to warming and precipitation manipulations.

We appreciate the encouragement and positive comments.

Specific comments:

(1) Please add some names of microbial groups with most common for the growth rates.

We have added the sentence “The members in Solirubrobacter and Pseudonocardia genera had high growth rates under changed climate regimes” In the Abstract (Line 57-58).

(2) L47-48, consider changing "microbial growth and death" to "microbial eco-physiological processes (e.g., growth and death)", and changing "such eco-physiological traits" to "such processes".

Done (Line 47 and 48).

(3) L50-51, the author estimated bacterial growth in alpine meadow soils of the Tibetan Plateau after warming and altered precipitation manipulation in situ. Actually, the soil samples were collected and incubated in the laboratory rather than in the field like the previous experiment conducted by Purcell et al. (2021, Global Change Biology). "In situ" would lead me to believe that the qSIP incubation was conducted in the field, so I think the use of the word in situ is inappropriate here. [https://onlinelibrary.wiley.com/doi/full/10.1111/gcb.15911]

Agreed. We have deleted “in situ”.

(4) L52, what does "interactive global change factors" mean?

We have revised this sentence to “the growth of major taxa was suppressed by the single and combined effects of temperature and precipitation” (Line 52-53).

(5) L61, in my opinion, "Microbial diversity" belongs to the category of species composition, rather than ecosystem functional services. Please revise it.

Agree. We have deleted it.

(6) L69, consider changing "further" to "thus".

Done (Line 70).

(7) L82, delete "The evidence is overwhelming that".

Done.

(8) L85-90, these two sentences have similar meanings, please express them concisely.

We have deleted the sentence “Altered precipitation, particularly drought or heavy precipitation events, also tends to negatively influence soil processes and biodiversity”.

(9) L91, the effect of drought on soil microorganisms is lacking here.

We have added the sentence “Reduced precipitation affects soil processes notably by directly stressing soil organisms, and also altering the supply of substrates to microbes via dissolution, diffusion, and transport” in the Introduction (Line87-89).

(10) L102, "Growth" should be highlighted here, as changes in relative abundance can also be classified as population dynamics. The use of the term "population dynamics" will eliminate the highlight of this study in calculating the growth rate of microbial species in in-situ soil based on qSIP. Consider changing "population dynamics" to "population-growth responses" or something like that.

Done (Line 98).

(11) L105, please note that this citation focuses on plant physiological characteristics.

We have revised the reference (Line 102).

(12) L115, "soil temperature, water availability" should be considered as a direct impact of climate change, rather than an indirect impact on microorganisms.

We have deleted them.

(13) L134-135, please clarify the interaction types between which climate factors.

We have deleted this sentence.

(14) L135-138, suggest modifying or deleting this sentence. The results in this study are already eco-physiological data and do not need to be further "understood and predicted".

We have deleted this sentence.

(15) L150, "The experimental design has been described in previously". I think this refers to another study and not the actual incubations in this study. Also in L198, suggest a change to "Incubation conditions were similar to those previously described". So, it's clear it's not the same experiment.

We have revised these sentences to “has been described previously in (Ma et al., 2017)” (Line 136) and “according to a previous publication” (Line 194).

Reference:

Ma, Z., Liu, H., Mi, Z., Zhang, Z., Wang, Y., Xu, W. et al. (2017). Climate warming reduces the temporal stability of plant community biomass production. Nature Communications, 8, 15378.

(16) L188, change "pre-wet soil samples" to "pre-wet samples" and change "soil samples for 48h incubation" to "incubation samples". What does "pre-wet" mean? Does it represent soil pre-cultivation?

Done. The pre-wet samples, i.e., the soil samples before incubation (T = 0 d), were used to estimate the initial microbial composition. "pre-wet" does not mean soil pre-cultivation. We have added the description “A portion of the air-dried soil samples was taken as the pre-wet treatment (i.e., before incubation without H2O addition)” in MATERIALS AND METHODS (Line 174-175).

(17) Unify the time unit of incubation (hour or day). Consider changing "48 h" to "2 d" in Materials and Methods.

Done.

(18) L247, what version of RDP Classifier was used?

We used RDP v16 database for taxonomic annotation. We have added this information in the revision (Line 246).

(19) L270, "average molecular weights".

Done (Line 268).

(20) L272-275, based on the preceding description, it appears that the culture period was limited to 48 hours. Please confirm it.

Apologize for this mistake. We have revised it (Line 273).

(21) L297, switch the order of the first two sentences of this paragraph.

Done (Line 297).

(22) L331, change "smaller-than-additive" to "smaller than their expected additive effect".

Done (Line 331).

(23) L374 and 381, I struggle with why "larger combined effects" than single factor effects represent higher degree of antoninism, and I think it should be "smaller combined effects".

Agree. We have revised it according to this suggestion (Line 369 and 374).

(24) L375, remove "than that of drought and warming".

Done.

(25) L405, simplify the expression, change "between different warming and rainfall regimes" to "between climate regimes"

We have deleted this sentence.

(26) L406-408, species are already on the phylogenetic tree and they can not "clustered at the phylogenetic branches", but the functional traits of microbes can. Please revise it.

We have revised this sentence to “Overall, the most incorporators whose growth was influenced by the antagonistic interaction of T × P showed significant phylogenetic clustering (i.e., species clustered at the phylogenetic branches; NTI > 0, P < 0.05)” (Line 402-404).

(27) L409, the same as above, and consider removing "The incorporators subjected to". We have revised this sentence to “The incorporators whose growth subjected to the additive interaction of warming × drought also showed significant phylogenetic clustering (P < 0.05)” (Line 404-406).

(28) L412, consider changing "incorporators subjected to the synergistic interaction" to "the synergistic growth responses under multifactorial changes".

We have revised the sentence to “incorporators whose growth is influenced by the synergistic interaction showed phylogenetically random distribution under both climate scenarios (P > 0.05)” (Line 407-409).

(29) L505-506, please add a reference for this sentence.

Done (Line 488).

(30) L511-514, It should be noted that the production of MBC does not necessarily imply a net change in the C pool size. The accelerated growth rates may result in expedited turnover of MBC, rather than an increase in carbon sequestration.

Thanks. We have deleted this sentence.

(31) Language precision. In the discussion section there must be some additional caveats introduced to some of the claims the authors are making. For instance, L518, the author should clarify that "in this study, the bacterial growth in alpine grassland may be influenced by antagonistic interactions between multiple climatic factors after a decadal-long experiment". Because other studies may exhibit different results due to the focus on different ecosystem functions as well as environmental conditions. As such, softening of the language is recommended- lines are noted below- and these will not adjust the outcomes of this study, but support more precise interpretation.

We have revised the sentence to “In this study, a decade-long experiment revealed that bacterial growth in alpine meadows is primarily influenced by the antagonistic interaction between T × P” (Line 497-499).

(32) Picrust analysis is a good way to connect species and their functions, especially Picrust2, which updated the reference database and optimized the algorithm to improve its prediction accuracy (Douglas et al., 2020, Nature Biotechnology). However, the link between microbial taxonomy and microbial metabolism is still not straightforward, especially in diverse microbial communities like soils. The authors should introduce caveats within discussion that they know the limitations of their methods. For context, as a reader who does metabolisms in soils, I found myself somewhat disappointed when piecrust data was introduced and not properly caveated. Particularly, it might be helpful to introduce briefly in the last paragraph of the results. These caveats are necessary to not potentially overstate the author's findings, and to make sure the reader knows the authors understand the very clear limitations of these methods. [https://www.nature.com/articles/s41587-020-0548-6]

Thanks. We have introduced caveats in DISCUSSION, that is “This is, however, still to be verified, as the functional output from PICRUSt2 is less likely to resolve rare environment-specific functions (Douglas et al., 2020)” (Line 540-542).

Reference:

Douglas, G., Maffei, V., Zaneveld, J., Yurgel, S., Brown, J., Taylor, C. et al. (2020). PICRUSt2 for prediction of metagenome functions. Nature Biotechnology, 38, 1-5.

(33) Although the author has explained the potential causes for the negative effects of different climate change factors (i.e., warming, drought, and wet) on microbial growth, there seems to be a lack of a summary assertion and an extension on how climate change affects microbial growth and related ecosystem functions. It is recommended to make a general summary of the results in the last part of Discussion.

We have added a general summary in the last paragraph of DISCUSSION, that is “Our results demonstrated that both warming and altered precipitation negatively affect the growth of grassland bacteria; However, the combined effects of warming and altered precipitation on the growth of ~70% soil bacterial taxa were smaller than the single-factor effects, suggesting antagonistic interaction. This suggests the development of multifactor manipulation experiments in precise prediction of future ecosystem services and feedbacks under climate change scenarios” (Line 552-558).

(34) L546, please add the taxonomic information for "OTU 14".

Done (Line 533).

(35) L800, change "The phylogenetic tree" to "A phylogenetic tree".

Done (Line 762).

Reviewer #2 (Public Review):

Summary:

The authors aimed to describe the effect of different temperature and precipitation regimes on microbial growth responses in an alpine grassland ecosystem using quantitative 18O stable isotope probing. It was found that all climate manipulations had negative effects on microbial growth, and that single-factor manipulations exerted larger negative effects as compared to combined-factor manipulations. The degree of antagonism between factors was analyzed in detail, as well as the differential effect of these divergent antagonistic responses on microbial taxa that incorporated the isotope. Finally, a hypothetical functional profiling was performed based on taxonomic affiliations. This work gives additional evidence that altered warming and precipitation regimes negatively impact microbial growth.

Strengths:

A long term experiment with a thorough experimental design in apparently field conditions is a plus for this work, making the results potentially generalisable to the alpine grassland ecosystem. Also, the implementation of a qSIP approach to determine microbial growth ensures that only active members of the community are assessed. Finally, particular attention was given to the interaction between factors and a robust approach was implemented to quantify the weight of the combined-factor manipulations on microbial growth.

We appreciate the reviewer’s positive comments.

Weaknesses:

The methodology does not mention whether the samples taken for the incubations were rhizosphere soil, bulk soil or a mix between both type of soils. If the samples were taken from rhizosphere soil, I wonder how the plants were affected by the infrared heaters and if the resulting shadow (also in the controls with dummy heaters) had an effect on the plants and the root exudates of the parcels as compared to plants outside the blocks? If the samples were bulk soil, are the results generalisable for a grassland ecosystem? In my opinion, it is needed to add more info on the origin of the soil samples and how these were taken.

The samples taken for the incubations can be considered as a mixture of rhizosphere and bulk soils. During soil sampling, we did not use conventional rhizosphere soil collection methods. However, there is a certain proportion of fragmented roots in the soil samples we collected, indicating that soil properties are influenced by plants. We have added this description in MATERIALS AND METHODS (Line 158).

To minimize the impact of physical shading on the plants, each sampling point was as far away from infrared heaters as possible. We have added this information of soil collection in MATERIALS AND METHODS, that is “In each plot, three soil cores of the topsoil (0-5 cm in depth) were randomly collected and combined as a composite sample, which can be considered as a mixture of rhizosphere and bulk soils. Each sampling point was as far away from infrared heaters as possible to minimize the impact of physical shading on the plants. The fresh soil samples were shipped to the laboratory and sieved (2-mm) to remove root fragments and stones.” (Line 157-162).

Previous studies based on our field experiment assessed the effects of warming and altered precipitation on soil microbial communities (Zhang et al., 2016), the temporal stability of plant community biomass (Ma et al., 2017), shifting plant species composition and grassland primary production (Liu et al., 2018). These studies provide guidance for the experiment design and execution.

Reference:

Zhang, KP., Shi, Y., Jing, X. et al. (2016). Effects of Short-Term Warming and Altered Precipitation on Soil Microbial Communities in Alpine Grassland of the Tibetan Plateau. Frontiers in Microbiology, 7, 1-11.

Ma ZY., Liu, HY., Mi, ZR. et al. (2017). Climate warming reduces the temporal stabilityof plant community biomass production. Nature Communications, 8, 15378.

Liu, HY., Mi, ZR., Lin, L. et al. (2018). Shifting plant species composition in response to climate change stabilizes grassland primary production. Proceedings of the National Academy of Sciences, 115, 4051-4056.

The qSIP calculations reported in the methodology for this work are rather superficial and the reader must be experienced in this technique to understand how the incorporators were identified and their growth quantified. For instance, the GC content of taxa was calculated for reads clustered in OTUs, and it is not discussed in the text the validity of such approach working at genus level.

We have added the description of qSIP calculations in Supplementary Materials.

The approach of GC content calculation can be used at genus level (Koch et al., 2018). The GC content of each bacterial taxon (Gi) was calculated using the mean density for the unlabeled (WLIGHTi) treatments (Hungate et al. 2015), rather than OTU sequence information. We have revised the sentence in MATERIALS AND METHODS, that is “the number of 16S rRNA gene copies per OTU taxon (e.g., genus or OTU) in each density fraction was calculated by multiplying the relative abundance (acquisition by sequencing) by the total number of 16S rRNA gene copies (acquisition by qPCR)” (Line 255-258).

Reference:

Hungate, B., Mau, R., Schwartz, E., Caporaso, J., Dijkstra, P., Van Gestel, N. et al. (2015). Quantitative microbial ecology through stable isotope probing. Applied and Environmental Microbiology, 81, 7570-7581.

Koch, B., McHugh, T., Hayer, M., Schwartz, E., Blazewicz, S., Dijkstra, P. et al. (2018). Estimating taxon-specific population dynamics in diverse microbial communities. Ecosphere, 9, e02090.

The selection of V4-V5 region over V3-V4 region to quantify the number of copies of the 16S rRNA gene should be substantiated in the text. Classic works determined one decade ago that primer pairs that amplify V3-V4 are most suitable to assess soil bacterial communities. Hungate et al. (2015), worked with the V3-V4 region when establishing the qSIP method. Maybe the number of unassigned OTUs is related with the selection of this region.

Both primer sets (V3-V4 and V4-V5 regions), are widely used across various sample sets, with highly similar in representing the total microbial community composition (Fadeev et al., 2021; Zhang et al., 2018).

A previous study based on our Field Research Station of Alpine Grassland Ecosystem used V4-V5 primer pairs to investigated the effect of warming and altered precipitation on the overall bacterial community composition (Zhang et al., 2016).

Another reason for choosing the V4-V5 primer set in this study was to integrate and compare the data with that of two previous qSIP studies (Ruan et al., 2023; Guo et al., submitted), both of them focused on the growth responses of active species to global change and used V4-V5 primer pairs.

We have added an explanation about primer selection as “The V4-V5 primer pairs were chosen to facilitate integration and comparison with data from previous studies (Ruan et al., 2023; Zhang et al., 2016)” (Line 213-215).

Reference:

Fadeev, E., Cardozo-Mino, M.G., Rapp, J.Z. et al. (2021). Comparison of Two 16S rRNA Primers (V3–V4 and V4–V5) for Studies of Arctic Microbial Communities. Frontiers in Microbiology, 12

Zhang, J.Y., Ding, X., Guan, R. et al. (2018). Evaluation of different 16S rRNA gene V regions for exploring bacterial diversity in a eutrophic freshwater lake. Science of The Total Environment, 618, 1254-1267.

Zhang, K.P., Shi, Y., Jing, X. et al. (2016). Effects of Short-Term Warming and Altered Precipitation on Soil Microbial Communities in Alpine Grassland of the Tibetan Plateau. Frontiers in Microbiology, 7, 1-11.

Ruan, Y., Kuzyakov, Y., Liu, X. et al. (2023). Elevated temperature and CO2 strongly affect the growth strategies of soil bacteria. Nature Communications, 14, 1-12.

Guo, J.J., Kuzyakov, Y., Li, L. et al. (2023). Bacterial growth acclimation to long-term nitrogen input in soil. The ISME Journal, Submitted.

Report of preprocessing and processing of the sequences does not comply state of the art standards. More info on how the sequences were handled is needed, taking into account that a significant part of the manuscript relies on taxonomic classification of such sequences. Also, an OTU approach for an almost species-dependent analysis (GC contents) should be replaced or complemented with an ASV or subOTUs approach, using denoisers such as DADA2 or deblur. Usage of functional prediction tools underestimates gene frequencies, including those related with biogeochemical significance for soil-carbon and nitrogen cycling.

(1) We have complemented the information about sequence processing as “The raw sequences were quality-filtered using the USEARCH v.11.0 (Edgar, 2010). In brief, the paired-end sequences were merged and quality filtered with “fastq_mergepairs” and “fastq_filter” commands, respectively. Sequences < 370 bp and total expected errors > 0.5 were removed. Next, “fastx_uniques” command was implemented to remove redundant sequences. Subsequently, high-quality sequences were clustered into operational taxonomic units (OTUs) with “cluster_otus” commandat a 97% identity threshold, and the most abundant sequence from each OTU was selected as a representative sequence.” (Line 238-245).

(2) We have complemented the zero-radius OTU (ZOTU) analysis by the unoise3 command in USEARCH (https://drive5.com/usearch/manual/pipe_otus.html), as shown in Fig. S1-S2. The results showed that overall growth responses of soil bacteria to warming and precipitation changes were similar based on OTU and ZOTU analyses, i.e., warming and altered precipitation tend to negatively affect the growth of grassland bacteria and the prevalence of antagonistic interactions of T × P. The similarity of results between the different methods is reflected at the overall community level, the phylum level, the genus level and the species (i.e., OTU or ZOTU) level (Fig. S1 and S2).

Author response image 1.

The growth responses of grassland bacteria to warming and altered precipitation based on ZOTU analysis. The results of growth rates at the community level (A), the phylum level (B), and the ZOTU level (C and D) were similar to those based on OTU analysis. C the single and combined factor effects of climate factors on species growth, by comparing with the growth rates in T0nP. D the proportions of species growth influenced by different interaction types of T × P. T0-P represents the ambient temperature and decreased precipitation; T+-P represents warming and decreased precipitation; T0cP represents ambient temperature and precipitation; T+cP represents warming and ambient precipitation; T0+P represents ambient temperature and enhanced precipitation; T++P represents warming and enhanced precipitation. Values represent mean and the error bars represent standard deviation. Different letters indicate significant differences between climate treatments.

Author response image 2.

The growth responses of grassland bacteria at the genus level to warming and altered precipitation based on OTU analysis (A and C) and ZOTU analysis (B and D). A and B the single and combined factor effects of climate factors on growth in genera, by comparing with those in T0nP. C and D the proportions of genera whose growth influenced by different interaction types of T × P.

(3) Agreed. We have introduced the caveat about the limitation of usage of functional prediction tools to the end of DISCUSSION, that is “This is, however, still to be verified, as the functional output from PICRUSt2 is less likely to resolve rare environment-specific functions (Douglas et al., 2020)” (Line 540-542). The caveat ensures that the reader knows the limitations of these methods, and are not potentially overstate our findings.

Reference:

Douglas, G.M., Maffei, V.J., Zaneveld, J.R. et al. (2020) PICRUSt2 for prediction of metagenome functions. Nat Biotechnol. 38, 685–688.

Reviewer #2 (Recommendations For The Authors):

General suggestions:

Regarding the qSIP method, would be of help to see the differences in density vs number of 16S rRNA gene abundance for the most responsive bacterial groups in the different treatments, taking into account that with only 7 fractions the entire change in bacterial growth was resolved.

We have selected three representative bacterial taxa (OTU1 belonging to Bradyrhizobium, OTU14 belonging to Solirubrobacter, OTU15 belonging to Pseudoxanthomonas), which have high growth rates in climate change treatments. The result showed that the peaks in the 18O treatment are shifted "backwards" (greater average weighted buoyancy density) compared to the 16O treatment, indicating that these species assimilates the 18O isotope into the DNA molecules during growth.

Author response image 3.

The distribution of 16S rRNA gene abundance of three representative bacterial taxa (OTU1- Bradyrhizobium, OTU14-Solirubrobacter, and OTU15-Pseudoxanthomonas) in different buoyant density fractions. Values represent mean and the error bars represent standard deviation.

Seven fractionated DNA samples were selected for sequencing because they contained more than 99% gene copy numbers of each samples (please see the Figure below). The DNA concentrations of other fractions were too low to construct sequencing libraries.

Author response image 4.

Relative abundance of 16S rRNA gene copies in each fraction. The fractions with density between 1.703 and 1.727 g ml-1 were selected because they contained more than 99% gene copy numbers of each sample. T0-P represents the ambient temperature and decreased precipitation; T+-P represents warming and decreased precipitation; T0cP represents ambient temperature and precipitation; T+cP represents warming and ambient precipitation; T0+P represents ambient temperature and enhanced precipitation; T++P represents warming and enhanced precipitation. Values represent mean and the error bars represent standard deviation.

With such dataset additional multivariate analysis would be of help to better interpret the ecological framework.

Thanks for the suggestion. Interpreting the ecological framework is meaningful for understanding microbial responses to environmental changes.

The main objective of this study is to investigate the growth response of soil microbes in alpine grasslands to the temperature and precipitation changes, and the interaction between climate factors. Our results, as well as the results of complementary analyses (based on subOTU analyses, SHOWN BELOW), indicate that warming and altered precipitation tend to negatively affect the growth of grassland bacteria, and the prevalence of antagonistic interactions of T × P.

We have emphasized our research objectives and main conclusions in the revised manuscript: “The goal of current study is to comprehensively estimate taxon-specific growth responses of soil bacteria following a decade of warming and altered precipitation manipulation on the alpine grassland of the Tibetan Plateau” (Line 112-114);

“Our results demonstrated that both warming and altered precipitation negatively affect the growth of grassland bacteria; However, the combined effects of warming and altered precipitation on the growth of ~70% soil bacterial taxa were smaller than the single-factor effects, suggesting antagonistic interaction” (Line 552-556).

Extension of interaction analysis and its conclusions should be shortened, summarizing the most relevant findings. In my opinion, it becomes a bit redundant.

We have shortened the discussion of Extension of interaction analysis by deleting the little relevant contents.

Below are some, but not all, examples that have been deleted or revised in the Discussion,

(1) Deleted “This result supports our second hypothesis that the interactive effects between warming and altered precipitation on soil microbial growth are not simply additive”;

(2) Deleted “A previous study suggested that multiple global change factors had negative effects on soil microbial diversity (Yang et al., 2021)”;

(3) Revised “A meta‐analysis of experimental manipulation revealed that the combined effects of different climate factors were usually less than expected additive effects, revealing antagonistic interactions on soil C storage and nutrient cycling processes (Dieleman et al., 2012; Wu et al., 2011). Moreover, two experimental studies on N cycling and net primary productivity demonstrated that the majority of interactions among multiple factors are antagonistic rather than additive or synergistic, thereby dampening the net effects (Larsen et al., 2011; Shaw et al., 2002)” to “A range of ecosystem processes have been revealed to be potentially subject to antagonistic interactions between climate factors, for instance, net primary productivity (Shaw et al., 2002), soil C storage and nutrient cycling processes (Dieleman et al., 2012; Wu et al., 2011; Larsen et al., 2011)” (Line 499-503);

(4) Revised “Previous evidences suggest that warming has a negative impact on soil carbon pools (Jansson & Hofmockel, 2020; Purcell et al., 2022). During the first phase of soil warming (~ 10 years), microbial activity increased, resulting in rapid soil carbon mineralization and respiration (Melillo et al., 2017)” to “Previous evidences suggest that warming has a negative impact on soil carbon pools (Jansson & Hofmockel, 2020; Purcell et al., 2022), mainly because of the rapid soil carbon mineralization and respiration (Melillo et al., 2017)” (Line 464-466).

I strongly suggest a functional analysis based on shotgun sequencing or RNAseq approaches. With this approach this work would be able to answer who is growing under altered T and Precipitation regimes and what are those that are growing doing.

Thanks for the suggestion. Metagenomic sequencing is a popular approach to evaluate potential functions of microbial communities in environment. However, there are two main reasons that limit the application of metagenomic or metatranscriptomic sequencing in this study: 1) Most of the fractionated samples in SIP experiment have low DNA concentration and do not meet the requirement of library construction for sequencing; 2) Metagenome and metatranscriptomics usually have relatively low sensitivity to rare species, reducing the diversity of detected active species.

This study focused on active microbial taxa and their growth in response to multifactorial climate change. We have added the prospect in DISCUSSION, that is “This suggests the development of methods combining qSIP with metagenomes and metatranscriptomes to assess the functional shifts of active microorganisms under global change scenarios” (Line 542-544).

Minor suggestions:

L121. _As

We have deleted this sentence and relocated the hypotheses in the last paragraph of INTRODUCTION (according to the suggestion of the reviewer #3).

Line150. Described previously in.

Done (Line 136).

Line500. Check whether it is better to use the word acclimatization (Coordinated response to several simultaneous stressors) in exchange of acclimation

We have revised it according to this suggestion (Line 481).

Fig.4C Drought

Done (Line 761).

Reviewer #3 (Public Review):

Summary:

In this paper, Ruan et al. studied the long-term impact of warming and altered precipitations on the composition and growth of the soil microbial community. The researchers adopted an experimental approach to assess the impact of climate change on microbial diversity and functionality. This study was carried out within a controlled environment, wherein two primary factors were assessed: temperature (in two distinct levels) and humidity (across three different levels). These factors were manipulated in a full factorial design, resulting in a total of six treatments. This experimental setup was maintained for ten years. To analyze the active microbial community, the researchers employed a technique involving the incorporation of radiolabeled water into biomolecules (particularly DNA) through quantitative stable isotope probing. This allowed for the tracking of the active fraction of microbes, accomplished via isopycnic centrifugation, followed by Illumina sequencing of the denser fraction. This study was followed by a series of statistical analysis to identify the impact of these two variables on the whole community and specific taxonomic groups. The full factorial design arrangement enabled the researchers to discern both individual contributions as well as potential interactions among the variables

Strengths:

This work presents a timely study that assesses in a controlled fashion the potential impact of global warming and altered precipitations on microbial populations. The experimental setup, experimental approach and data analysis seem to be overall solid. I consider the paper of high interest for the whole community as it provides a baseline to the assessment of global warming on microbial diversity.

Thanks for the encouragement and positive comments.

Weaknesses:

While taxonomic information is interesting, it would have been highly valuable to include transcriptomics data as well. This would allow us to understand what active pathways become enriched under warming and altered precipitations. Non-metabolic OTUs hold significance as well. The authors could have potentially described these non-incorporators and derived hypotheses from the gathered information. The work would have benefited from using more biological replicates of each treatment.

Thanks for the valuable suggestions.

(1) Metatranscriptomics can assess the functional profiles of the community, but it has relatively low sensitivity to rare species, which is difficult to correlate the function pathways with the assignment to the numerous active taxa identified by qSIP. Additionally, due to the low DNA concentration, most fractionated samples are difficult to construct sequencing libraries, while amplicon based sequencing analyses were allowed. This study therefore focused on active microbial taxa and their growth in response to multifactorial climate change. We have added the prospect in DISCUSSION, that is “This suggests the development of methods combining qSIP with metagenomes and metatranscriptomes to assess the functional shifts of active microorganisms under global change scenarios” (Line 542-544).

(2) 18O-qSIP can identify the growing microbial species (i.e., 18O incorporators) in the environment rather than metabolically active taxa. These non-incorporators in our study were likely to be metabolically active, i.e., maintaining life activities without reproduction, or recently deceased (Blazewicz et al., 2013). Therefore, it is hard to distinguish whether these non-incorporators possess metabolic activity.

(3) Agreed. The qSIP experiments involve the use of isotopes and the sequencing of a large number of DNA samples (90 samples per biological replicate in this study). Considering its high cost, we selected three replicates for analysis. We have explained this issue in MATERIALS AND METHODS, that is “Considering the cost of qSIP experiment (i.e., the use of isotopes and the sequencing of a large number of DNA samples), we randomly selected three out of the six plots, serving as three replicates for each treatment” (Line 154-157).

Reference:

Nuccio, E.E., Starr, E., Karaoz, U. et al. (2020) Niche differentiation is spatially and temporally regulated in the rhizosphere. ISME J 14, 999–1014.

Blazewicz, S.J., Barnard, R.L., Daly, R.A., Firestone, M.K (2013). Evaluating rRNA as an indicator of microbial activity in environmental communities: limitations and uses. The ISME Journal, 7, 2061–2068.

Reviewer #3 (Recommendations For The Authors):

Major comments:

The manuscript should be written in a clearer way. The language should be more direct, so the message is conveyed faster and clearer. Some sentences, for instance, could be shortened or re-organized. Below, you will find some examples.

We have rewritten the sentences to make the manuscript clearer. Below are some, but not all, examples that have been revised:

(1) Deleted “(reduced precipitation, hereafter ‘drought’, or enhanced precipitation, hereafter ‘wet’)” in INTRODUCTION;

(2) Deleted “Controlled experiments simulating climate change have investigated changes in microbial community composition as measured by shifts in the relative abundances (Evans & Wallenstein, 2014; Barnard et al., 2015). However, changes in relative abundances may be poor indicators of population responses to environmental change in some cases (Blazewicz et al., 2020). Another challenge is the presence of a large number of inactive microbial cells in the soil, which hinders the direct, quantitative measure of the ecological drivers in population dynamics (Fierer, 2017; Lennon & Jones, 2011).” in DISCUSSION;

(3) Deleted “This result supports our second hypothesis that the interactive effects between warming and altered precipitation on soil microbial growth are not simply additive” in DISCUSSION;

(4) Deleted “A previous study suggested that multiple global change factors had negative effects on soil microbial diversity (Yang et al., 2021)” in DISCUSSION;

(5) Revised “A meta‐analysis of experimental manipulation revealed that the combined effects of different climate factors were usually less than expected additive effects, revealing antagonistic interactions on soil C storage and nutrient cycling processes (Dieleman et al., 2012; Wu et al., 2011). Moreover, two experimental studies on N cycling and net primary productivity demonstrated that the majority of interactions among multiple factors are antagonistic rather than additive or synergistic, thereby dampening the net effects (Larsen et al., 2011; Shaw et al., 2002)” to “A range of ecosystem processes have been revealed to be potentially subject to antagonistic interactions between climate factors, for instance, net primary productivity (Shaw et al., 2002), soil C storage and nutrient cycling processes (Dieleman et al., 2012; Wu et al., 2011; Larsen et al., 2011)” in DISCUSSION (Line 499-503);

(6) Revised “Previous evidences suggest that warming has a negative impact on soil carbon pools (Jansson & Hofmockel, 2020; Purcell et al., 2022). During the first phase of soil warming (~ 10 years), microbial activity increased, resulting in rapid soil carbon mineralization and respiration (Melillo et al., 2017)” to “Previous evidences suggest that warming has a negative impact on soil carbon pools (Jansson & Hofmockel, 2020; Purcell et al., 2022), mainly because of the rapid soil carbon mineralization and respiration (Melillo et al., 2017)” in DISCUSSION (Line 464-466).

I'm curious about why, even though there were six replicates of the experiment, only three samples were collected for analysis. Metagenomic analyses tend to display high variability.

The qSIP experiments involve the use of isotopes and the sequencing of a large number of DNA samples (90 samples per biological replicate in this study). Considering its high cost, we selected three replicates for analysis..

In Fig. 3A, the absolute growth rates (16S copies/d*g) are shown. How do you know that the efficiency of DNA extraction was similar across all treatments and therefore the absolute numbers are comparable?

To avoid differences in extraction efficiency caused by experimental procedures, all DNA samples were extracted by the same person (the first author) within 2-3 hours, and a unifying procedure of cell lysis and DNA extraction was used, i.e., the mechanical cell destruction was attained by multi-size beads beating at 6 m s-1 for 40 s, and then FastDNA™ SPIN Kit for Soil (MP Biomedicals, Cleveland, OH, USA) was used for DNA extraction.

We have measured the concentration of extracted DNA and found no significant difference between treatments (Table for the response letter).

Author response table 1.

Soil DNA concentration in climate change treatments after qSIP incubation (measured by Qubit® DNA HS Assay Kits).

Values represent mean and standard deviation. T0-P represents the ambient temperature and decreased precipitation; T+-P represents warming and decreased precipitation; T0cP represents ambient temperature and precipitation; T+cP represents warming and ambient precipitation; T0+P represents ambient temperature and enhanced precipitation; T++P represents warming and enhanced precipitation. The results of ANOVA indicated no significant difference of extracted DNA concentration between treatments (p > 0.05).

We have introduced the caveat in the DISCUSSION, that is “Note that the experimental parameters such as DNA extraction and PCR amplification efficiencies also have significant effects on the accuracy of growth assessment. This alerts the need to standardize experimental practices to ensure more realistic and reliable results” (Line 544-547).

Line 96-99 and 121-124: "Hypotheses are typically placed at the end of the final paragraph in the Introduction section. It is advisable to relocate them there and provide a clearer description of the paper's main goal."

We have relocated the hypotheses at the end of INTRODUCTION, and the main goal of this study, that is “The goal of current study is to comprehensively estimate taxon-specific growth responses of soil bacteria following a decade of warming and altered precipitation manipulation on the alpine grassland of the Tibetan Plateau, by using the 18O-quantitative stable isotope probing (18O-qSIP)” (Line 112-115).

Line 399: Although you describe the classification among antagonistic interactions in the Methods section, I think you should describe this in further detail here. Can you clarify how you carried out this categorization and how these results were interpreted considering the phylogenetic classification.

We have added the description of antagonistic interactions, that is “The interaction type of T × P on the growth of ~70% incorporators was antagonistic (i.e., the combined effect size is smaller than the additive expectation) (Fig. 4C)” (Line 388-390).

The interaction types between factors can be classified into three categories: additive, synergistic and antagonistic. Additive interactions are those in which the combined effect size of factors is equal to the sum of the single effects of the factors (i.e., additive expectation, Fig. 1B). Synergistic interactions refer to the effect size was larger than the additive expectation by the combined manipulation of factors. On the contrary, antagonistic interactions refer to the combined effect size of factors is smaller than the additive expectation. In this study, the antagonistic interactions were further divided into three sub-categories: weak antagonistic interaction, strong antagonistic interaction, and neutralizing effect (Fig. 1B). The weak antagonistic interaction refers to the combined effect size smaller than the additive expectation and larger than any of the single factor effects. The strong antagonistic interaction refers to that the combined effect size is smaller than any of the single factor effects but larger than 0. The neutralizing effect refers to that the combined effect size is equal to 0, implying that the effects of different factors cancel each other out.

Methodologically, the single and combined effects of two climate factors and their interaction effects were calculated by the natural logarithm of response ratio (lnRR) and Hedges’ d, respectively (Yue et al., 2017).

We have added the result interpretation about the phylogenetic distribution patterns of incorporators, that is “The degree of phylogenetic relatedness can indicate the processes that influenced community assembly, like the extent a community is shaped by environmental filtering (clustered by phylogeny) or competitive interactions (life strategy is phylogenetically random distribution) (Evans & Wallenstein, 2014; Webb et al., 2002).The results showed that the incorporators whose growth was influenced by the antagonistic interaction of T × P showed significant phylogenetic relatedness, indicating the occurrence of taxa more likely shaped by environment filtering (i.e., selection pressure caused by changes in temperature and moisture conditions). In contrast, the growing taxa affected by synergistic interactions of T × P showed random phylogenetic distributions (Table S1), which may be explained by competition between taxa with similar eco-physiological traits or changes in genotypes (possibly through horizontal gene transfer) (Evans & Wallenstein, 2014). We also found that the extent of phylogenetic relatedness to which taxa groups of T × P interaction types varied by climate scenarios, suggesting that different climate history processes influenced the ways bacteria survive temperature and moisture stress” (Line 515-529).

Reference:

Evans, S.E. & Wallenstein, M.D. (2014). Climate change alters ecological strategies of soil bacteria. Ecology Letters, 17, 155-164.

Webb, C.O., Ackerly, D.D., McPeek, M.A. & Donoghue, M.J. (2002). Phylogenies and Community Ecology. Annual Review of Ecology and Systematics, 33, 475-505.

Yue, K., Fornara, D.A., Yang, W., Peng, Y., Peng, C., Liu, Z. et al. (2017). Influence of multiple global change drivers on terrestrial carbon storage: additive effects are common. Ecology Letters, 20, 663-672.

Line 407-8: What do you mean with "...clustered at the phylogenetic branches" and Line 410: "cluster near the tips of the phylogenetic tree". Can you please clarify?

Sorry for the unclear statement. We have added the explanation of NTI, that is “Nearest taxon index (NTI) was used to determine whether the species in a particular growth response are more phylogenetically related to one another than to other species (i.e., close or clustering on phylogenetic tree). NTI is an indicator of the extent of terminal clustering, or clustering near the tips of the tree (Evans & Wallenstein, 2014; Webb et al., 2002)” (Line 397-401).

Reference:

Evans, S.E. & Wallenstein, M.D. (2014). Climate change alters ecological strategies of soil bacteria. Ecology Letters, 17, 155-164.

Webb, C.O., Ackerly, D.D., McPeek, M.A. & Donoghue, M.J. (2002). Phylogenies and Community Ecology. Annual Review of Ecology and Systematics, 33, 475-505.

Could you provide some info about the biochemistry of the incorporation of heavy water into DNA molecules? What specific enzymes are typically involved?

Due to the low DNA concentration in most fractionated samples (less than 10 ng/μL, measured by Qubit DNA HS Assay Kits), only amplicon based sequencing analyses were allowed. This study therefore focused only on active microbial taxa and their growth in response to multifactorial climate change.

What might be the impact of soil desiccation on bacterial survival and subsequent water uptake?

Slow dehydration and air drying of soil is a very common phenomenon in nature (Koch et al., 2018). In this process, microorganisms will reduce metabolism, and shift towards a potentially active state (Blagodatskaya and Kuzyakov, 2013). A previous study suggested that the potentially active microbial population permanently existing in soil between the active and dormant physiological states. Even under long-term starvation the potentially active microorganisms maintain ‘physiological alertness’ to be ready to occasional substrate input (Blagodatskaya and Kuzyakov, 2013). These microorganisms are important participants in the biogeochemical cycle is the focus of this study.

Replacing the environmental water in the soil with 18O-labelled water is a typical practice for qSIP studies (Hungate et al. 2015; Koch et al., 2018). This process may cause disturbance to the microbial community. In this study, the soil samples were placed in a thermostatic incubator (14℃ and 16℃), rather than air-drying at 25℃ (as used in most studies). The incubation temperature is relatively low (compared to 25℃) and there is no violent air convection in the incubator, resulting slower evaporation and no significant discoloration caused by severe soil dehydration after 48 h. The process of soil drying in this study simulated the natural phenomenon, i.e., slow water loss in soil.

We have added the description in MATERIALS AND METHODS, that is “There is no violent air convection in the incubator and the incubation temperature is relatively low (compared to 25℃ used in previous studies), resulting slower evaporation and no significant discoloration caused by severe soil dehydration after 48 h” (Line 171-174).

Reference:

Blagodatskaya, E. & Kuzyakov, Y. (2013) Active microorganisms in soil: Critical review of estimation criteria and approaches. Soil Biology and Biochemistry, 67, 192-211.

Hungate, B., Mau, R., Schwartz, E., Caporaso, J., Dijkstra, P., Van Gestel, N. et al. (2015). Quantitative microbial ecology through stable isotope probing. Applied and Environmental Microbiology, 81, 7570-7581.

Koch, B., McHugh, T., Hayer, M., Schwartz, E., Blazewicz, S., Dijkstra, P. et al. (2018). Estimating taxon-specific population dynamics in diverse microbial communities. Ecosphere, 9, e02090.

The analysis of the 180 incorporators is interesting as it defines what microbes are metabolically active and hence growing under the different conditions tested. Should not be worth to analyze the non-incorporators? Is it possible to identify a pattern to generate a hypothesis of why they are metabolically inactive based on this information? In the Methods section, the authors state that they identified a total of 6,938 OTUs, of which only 1,373 were found to be incorporators.

Microbes exist in a range of metabolic states: growing, active (non-growth), dormant and recently deceased (Blazewicz et al., 2013), and there is still a lack of clear threshold for their identification. 18O-DNA qSIP can identified the growing microbial species (i.e., 18O incorporators) rather than all metabolic active taxa, because some cells are measurably metabolizing (catabolic and/or anabolic processes) without reproduction. Therefore, the non-incorporators in our study may be metabolically active, or not (recently deceased microorganisms). This study focuses on the growing microorganisms identified by 18O-qSIP.

In this study, ~20% microbial taxa (1,373/6,938) were identified as 18O incorporators. Microorganisms in soils suffer from resource and energy constraints frequently (Blagodatskaya and Kuzyakov, 2013). The energy requirements of species in the growing state are much higher (~30 fold) than those in the non-growing state, so the percentage of growing bacterial taxa in soil tends to be low.

Reference:

Blazewicz, S.J., Barnard, R.L., Daly, R.A., Firestone, M.K (2013). Evaluating rRNA as an indicator of microbial activity in environmental communities: limitations and uses. The ISME Journal, 7, 2061–2068.

Blagodatskaya, E. & Kuzyakov, Y. (2013) Active microorganisms in soil: Critical review of estimation criteria and approaches. Soil Biology and Biochemistry, 67, 192-211.

Minor comments:

Fig. 3A and 3B. Please show the results of the multiple comparisons.

Done.

Author response image 5.

Bacterial growth responses to climate change and the interaction types between warming and altered precipitation. The growth rates (A), and responses (LnRR) of soil bacteria to warming and altered precipitation (B) at the whole community level. The growth rates (C), and responses of the dominant bacterial phyla (D) had similar trends with that of the whole community. Values represent mean and the error bars represent standard deviation. Different letters indicate significant differences between climate treatments.

Fig. 4. This figure should be self-explanatory. This diagram is challenging to understand.

We have revised Fig. 4 to improve clarity.

Author response image 6.

The growth responses and phylogenetic relationship of incorporators subjected to different interaction types under two climate scenarios. A phylogenetic tree of all incorporators observed in the grassland soils (A). The inner heatmap represents the single and combined factor effects of climate factors on species growth, by comparing with the growth rates in T0nP. The outer heatmap represents the interaction types between warming and altered precipitation under two climate change scenarios. The proportions of positive or negative responses in species growth to single and combined manipulation of climate factors by summarizing the data from the inner heatmap (B). The proportions of species growth influenced by different interaction types of T × P by summarizing the data from the outer heatmap (C).

Fig. 4. It says "Dorought" instead of "drought"

Done (Line 760).

Line 109: "relieves" instead of "relieved"

Done (Line 102).

Line 129: Should be: "We classified the interaction types as additive, synergistic, antagonistic, null and neutralizing."

Done (Line 117).

Line 233: How were the 16S rRNA sequences from each density fraction analyzed?

(1) Raw sequencing data processing:

The raw 16S rRNA gene sequences of each density fraction were quality-filtered using the USEARCH v.11.0 (Edgar, 2010). The paired-end sequences were merged and quality filtered with “fastq_mergepairs” and “fastq_filter” commands, respectively. Sequences < 370 bp and total expected errors > 0.5 were removed. Next, “fastx_uniques” command was implemented to identify the unique sequences. Subsequently, high-quality sequences were clustered into operational taxonomic units (OTUs) with “cluster_otus” commandat a 97% identity threshold, and the most abundant sequence from each OTU was selected as a representative sequence. The taxonomic affiliation of the representative sequence was determined using the RDP classifier (Wang et al., 2007).

(2) qSIP calculation:

Sequencing data reflects the relative abundance of taxa in community. We multiply the OTU’s relative abundance (acquisition by sequencing) and the number of 16S rRNA gene copies (acquisition by qPCR) to obtain the number of gene copies per OTU in each fraction. Then, the proportion of gene copies of a specific OTU of each fraction relative to the total amount of gene copies in one sample was calculated and used as a weight value for further calculation of the average weighted buoyant density (the critical parameter for assessing microbial growth).

Line 366: "Three single-factor ... between warming and altered precipitation" -> "The individual impact of warming, drought, and wet conditions resulted in the most substantial negative effects on bacterial growth compared with the effects of warming x drought and warming x wet. A result that illustrates the negative interactions between warming and modified precipitations patterns."

Done (Line 365-368).

Line 376: "Similar with the result of whole growth of bacteria community, the growth responses of the major bacterial phyla were also negatively influenced by single climate factors". This sentence is hard to read. Maybe something like this: "Growth of the major bacterial phyla was also negatively influenced by the individual climate factors".

Done (Line 371-372).

Line 383: "In particular, the effects of wet and warming neutralized each other, resulting the net effects became zero on the growth rates of the phyla Actinobacteria and Bacteroidetes". "In Actinobacteria and Bacteroidetes, the effect of wet and warming neutralized each other, as the combined effect of these two factors had no effect on growth".

Done (Line 377-379).

Line 390: "The individual warming treatment (T+nP) reduced the growth rates of 75% incorporators..." "Warming (T+nP) reduced the growth of 75% of the taxonomic groups, which was followed by drought and wet.

Done (Line 384-385).

Line 392: "The combined manipulations of warming and altered precipitation lowered the percentages of incorporators with negative responses compared with single factor manipulation, especially warming and enhanced precipitation manipulation" -> "Warming x drought and warming x wet had a smaller impact on the growth of incorporators, compared with single effects."

Done (Line 385-387).

Line 468. This sentence "To the best ..." is not necessary.

We have deleted this sentence.

Line 476. Is it really "synthesis" the word you want to use?

We have deleted this sentence.

Line 477. Maybe should written like this: "Consistent with our findings, a recent experimental study demonstrated that 15 years of warming reduced the growth rate of soil bacteria in a montane meadow in northern Arizona."

Done (Line 459-461).

Line 490 and 502. Consider using "however" only once in a paragraph.

We have deleted the second “however” (Line 483).

Line 555-559. Based on genomic data you cannot predict the functional role of microbes in the environment. These sentences are speculative. Please, consider using less strong affirmations and focus more on the pathways that are enriched in the incorporators.

Agreed. We have deleted this part of content.

Author Response

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

We thank the reviewers for their careful, critical, and insightful evaluation of our manuscript.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The preprint by Laganowsky and co-workers describes the use of mutant cycles to dissect the thermodynamic profile of specific lipid recognition by the ABC transporter MsbA. The authors use native mass spectrometry with a variable temperature source to monitor lipid binding to the native protein dimer solubilized in detergent. Analysis of the peak intensities (that is, relative abundance) of 1-3 bound lipids as a function of solution temperature and lipid concentration yields temperature-dependent Kds. The authors use these to then generate van't Hoff plots, from which they calculate the enthalpy and entropy contributions to binding of one, two, and in some cases, three lipids to MsbA.

The authors then employ mutant cycles, in which basic residues involved in headgroup binding are mutated to alanine. By comparing the thermodynamic signatures of single and double (and in one instance triple) mutants, they aim to identify cooperativity between the different positions. They furthermore use inward and outward locking conditions which should control access to the different binding sites determined previously.

The main conclusion is that lipid binding to MsbA is driven mainly by energetically favorable entropy increase upon binding, which stems from the release of ordered water molecules that normally coordinate the basic residues, which helps to overcome the enthalpic barrier of lipid binding. The authors also report an increase in lipid binding at higher temperatures which they attribute to a non-uniform heat capacity of the protein. Although they find that most residue pairs display some degree of cooperativity, particularly between the inner and outer lipid binding sites, they do not provide a structural interpretation of these results.

Strengths:

The use of double mutant cycles and mass spectrometry to dissect lipid binding is novel and interesting. For example, the observation that mutating a basic residue in the inner and one in the outer binding site abolishes lipid binding to a greater extent than the individual mutations is highly informative even without having to break it down into thermodynamic terms (see "weaknesses" section). In this sense, the method and data reported here opens new avenues for the structure/activity relationship of MsbA. The "mutant cycle" approach is in principle widely applicable to other membrane proteins with complex lipid interactions.

Weaknesses:

The use of double mutant cycles to dissect binding energies is well-established, and has, as the authors point out, been employed in combination with mass spectrometry to study protein-protein interactions. Its application to extract thermodynamic parameters is robust in cases where a single binding event is monitored, e.g. the formation of a complex with well-defined stoichiometry, where dissociation constants can be determined with high confidence. It is, however, complicated significantly by the fact that for MsbA-lipid interactions, we are not looking at a single binding event, but a stochastic distribution of lipids across different sites. Even if the protein is locked in a specific conformation, the observation of a single lipid adduct does not guarantee that the one lipid is always bound to a specific site. In some of the complexes detected by MS, the lipid is likely bound somewhere else. Lipid binding Kds from mass spectrometry, although helpful in some instances as a proxy for global binding affinities, should therefore be taken with a grain of salt.

We agree with the reviewer in that while we will measure binding of lipid (mass shift) we do not know the binding location(s). Given this issue, we have added to the discussion section on this important point and elaborate more broadly on this problem in the context of membrane protein-lipid interactions. Tackling this issue represents a frontier challenge for the field.

The authors analyze the difference in binding upon mutating binding sites (ddG etc). Here, another complicating factor comes into play, the fact that mutation of a binding site (which the authors show reduces lipid binding) may instead allow the lipid to bind to a lower-affinity site elsewhere. Unfortunately, the authors do not specify the protein concentration, but assuming it is in the single-digit micromolar range, as common for native MS experiments, lipid and protein concentrations are almost equal for most of the data points, resulting in competition between binding sites for free lipids. As a rule of thumb, for Kd measurements, the concentration of the constant component, the protein, should be far below the Kd, to avoid working in the "titration" regime rather than the "binding" regime (see Jarmoskaite et al, eLife 2020). I cannot determine whether this is the case here. The way I understand the double mutant cycle approach, reliable Kd measurements are required to accurately determine dH and TdS, so I would encourage the authors to confirm their Kd values using complementary methods before in-depth interpretations of the thermodynamic components.

The reviewer references an article in eLife by Jarmoskaite and co-workers describing “titration” vs “binding” regimes. Below we paste a snippet from this article:

Author response image 1.

Equation 4a is an expression for the fraction of protein bound to ligand, which universally holds, i.e., if we know the concentration of molecules at equilibrium (including those unbound or free) then one can obtain the special ratio or equilibrium constant at a given temperature. Jarmoskaite et al. note that in practice (using traditional biophysical approaches) one cannot readily distinguish protein that is free or bound to ligand (see highlighted part above). While this assumption is basis of their eLife assessment, it does NOT apply to native mass spectrometry data. It is important to realize that the mole fraction (or concentration) of apo and each lipid bound states, i.e., [P], [PL], [PL2], …, [PLn+1], can readily be obtained directly from the deconvoluted mass spectrum. This is unlike other biophysical methods that are ensemble measurements, which measures the amount of heat or fraction of total ligand bound to protein. Since we can discern each lipid bound state, including the free protein and free ligand concentrations, the equilibrium binding constants can be directly calculated, and the protein and ligand concentration becomes irrelevant. In principle, equilibrium constants for protein-lipid interactions can be calculated from one mass spectrum. To increase transparency, we have updated the results section to highlight the important difference of the native MS approach compared to less robust traditional approaches that are riddled with underlying issues/assumptions.

We appreciated the reviewer’s suggestion of using complementary methods to confirm Kd values. In our previous report [1], we determined binding thermodynamics for soluble protein-ligand interactions using native MS, surface plasmon resonance (SPR), and isothermal calorimetry (ITC) and found the techniques yield similar binding constants and thermodynamic parameters. The use of soluble proteins with defined ligand binding studies was rather straightforward to carry out a complementary study. We have also shown consistent findings for native MS and SPR of membrane protein interaction with a soluble, regulatory protein [2]. However, in the case of membrane proteins they can bind the first few lipids very specifically and, with the addition of more lipid, bind even more lipids that represent rather weak binding. Thus, traditional approaches would report on the ensemble of lipids bound to membranes and specific lipid binding sites (such as inner and outer LPS binding sites in MsbA) are saturable but also additional binding will be observed, i.e., doesn’t follow traditional soluble protein-ligand binding studies. In the past we have used a fluorescent-lipid competition binding assay [3] to corroborate native MS results for Kir3.2, which showed a direct correlation. The disadvantage of this complementary approach is using a non-natural, fluorescent-modified lipid. Unfortunately, there is no commercial source for a fluorophore modified KDL.

It is somewhat counterintuitive that for many double mutants, and the triple mutant, the entropic component becomes more favorable compared to the WT protein. If the increase in entropy upon lipid binding comes from the release of ordered water molecules around the basic residues (a reasonable assumption) why does this apply even more in proteins where several basic residues have been changed to alanine, which coordinate far fewer water molecules?

There are many factors that contribute to the change in entropy of the system, beyond solvation entropy, and deciphering the entropic contributions of the various components remains a challenging task. We have revised the manuscript to emphasize that solvation is one component of the entropic term and other components are likely at play.

The authors could devote more attention to the fact that they use detergent micelles as a vehicle for lipid binding studies. To a limited extent, detergents compete with lipids for binding, and are present in extreme excess over the lipid. The micelle likely changes its behavior in response to temperature changes. For example, the packing around the protein loosens up upon heating, which may increase the chance for lipids to bind. In this case, the increase in binding at higher temperatures may not be related to a change in heat capacity. This question could be addressed by MD simulations, if it's not already in the literature.

The detergent and its concentration are consistent for all the different MsbA proteins in this study. In fact, we observe linear van’t Hoff plots with positive and negative slopes as well as non-linear curves that are convex or concave. The MsbA protein (wt or mutant), trapped or not, all display unique temperature-dependent responses. The reviewers comment of increasing temperature to loosen packing of detergent to promote lipid binding is clearly NOT that simple. If detergent was significantly influencing lipid binding (as suggested by reviewer) then increasing its concentration should impact lipid binding. In a previous study, we found no difference in membrane protein-lipid thermodynamics even when the concentration of detergent was increased five-fold [1]. We repeated similar experiments for MsbA and find the increased detergent concentration does not impact the abundances of lipid bound states. The figure to the right shows MsbA in the presence of lipid in 2x CMC (panel a and b) and 10x CMC (panel c and d). As you will see, no appreciably difference in the lipid bound signal is observed.

Author response image 2.

We applaud the suggestion of MD simulation. However, it is far beyond the scope of this paper and its not clear what will really be learned.

Reviewer #2 (Public Review):

Summary:

This is a solid study that dissects the thermodynamics of lipopolysaccharide (LPS) transporter MsbA and LPS. Native ESI-MS and the novel strategies developed by the authors were employed to quantify the affinities of LPS-MsbA interactions and its temperature dependence. Here, the equilibrium of lipid-protein interactions occurs in the micellar phase. The double-/triple-mutant cycle analysis and van't Hoff analysis allowed a full thermodynamic description of the lipid-protein interactions and the analysis of thermodynamic coupling between LPS binding sites. The most notable result would be that LPS-MsbA interaction is largely driven by entropy involving the negative heat capacity, a signature of the solvent reorganization effect (here authors attribute the solvent effect to "water" reorganization). The entropy driven lipid binding has been previously reported by the same authors for Kir1,2-PIP2 interactions.

Strengths:

1. This is overall a very thorough and rigorous study providing the detailed thermodynamic principles of LPS-MsbA interaction.

2. The double and triple-mutant cycle approaches are newly applied to lipid-protein interactions, enabling detailed thermodynamics between LPS binding sites.

3. The entropy-driven protein-lipid interaction is surprising. The binding seems to be mainly mediated by the electrostatic interaction between the positively charged residues on the protein and the negatively charged or polar headgroup of LPS, which could be thought of as "enthalpic" (making of a strong bond relative to that with solvent).

Weaknesses:

1. This study is a good contribution to the field, but it was difficult to find novel biological insights or methodological novelty from this study.

1a. Thermodynamic analysis of lipid-protein interactions, an example of entropy-driven lipid-protein interactions, and the cooperativity between lipid binding sites have been reported by the author's group. Also, the cooperativity between binding sites in general have been reported from numerous studies of biomolecular interactions.

We appreciate the reviewer for highlighting our previous work. Of course, a single study does not establish a pattern, such as entropy-driven lipid-protein interactions.

While we agree with the reviewer that cooperativity in biomolecular interactions has been established for many soluble protein systems, by no means do we have a detailed understanding of membrane protein-lipid interactions. This work is an important contribution to expanding on classical work on soluble protein systems to more challenging membrane protein systems and their interactions with lipids.

1b. It is not clear how this study provides new insights into the understanding of LPS transport mechanisms. Probably, authors could strengthen the Discussion by providing biological insights-how the residue coupling.

The thermodynamics provides us with a deeper insight into the chemical principles that drive specific membrane protein-lipid interactions. We have revised the discussion to highlight the importance of thermodynamics and the implication of individual residues to KDL binding, and the inner and outer LPS binding sites appear to be coupled, something that is new.

1. One to three LPS molecules bind to MsbA, but it is unclear whether bound KDL occupies inner or outer cavities, or both and how a specific mutation affects the affinity of specific LPS (i.e., to inner or to outer cavities). Based on the known structures, the maximal number of LPS is three. It is possible that the inner and outer cavities have different LPS affinities. Also, there can be multiple one-LPS-bound states, two-LPS-bound states if LPS strictly binds to the binding sites indicated by the structures. This aspect is beyond the scope of this study and difficult to address, but without this information, it seems hard to tell what is going on in the system.

In our response above, we note that lipids will bind to membrane proteins at specific site(s) and weaker sites, often described as non-annular lipids. The revision includes this discussion point.

1. If a single mutation is introduced to the inner cavity, its effect will be "doubled" because the inner cavity is shared by two identical subunits. This effect needs to be clarified in the result section.

Great point. In addition, an outer mutant will also impact not one but both outer binding site(s)s. The revised manuscript makes note of this point.

1. In the result section, "Mutant cycle analysis of KDL binding to vanadate-trapped MsbA.":

4a. It seems necessary to show the mass spectra for Msb-ADP-vanadate complex as well as its lipid bound forms.

In the original submission, the mass spectra of vanadate trapped MsbA with KDL binding was provided in Supplementary Figures 10 and 11.

4b. The rationale of this section (i.e., what mechanistic insights can be obtained from this study) is unclear. For example, it is not sure what meaningful information can be obtained from a single type (ADP/vanadate) of the bound state regarding the ATP-driven function of MsbA.

MsbA is a dynamic, populates different conformations. Trapping with vanadate locks the transporter in an outwardfacing state with NDB interacting. This provides the opportunity to characterize binding to the exterior site. We revised the manuscript to note this point.

Reviewer #3 (Public Review):

Summary:

In this paper presented by Liu et al, native MS on the lipid A transporter MsbA was used to obtain thermodynamic insight into protein-lipid interactions. By performing the analyses at different lipid A concentrations and temperatures, dissociation constants for 2-3 lipid A binding sites were determined, as well as enthalpies were calculated using nonlinear van't Hoff fitting. Changes in free Gibb's energies were then calculated based on the determined dissociation constants, and together with the enthalpy values obtained via van' t Hoff analysis, the entropic contribution to lipid binding (DeltaS*T) was indirectly determined.

Strengths:

This is an extensive high quality native MS dataset that provides unique opportunities to gain insights into the thermodynamic parameters underlying lipid A binding. In addition, it provides coupling energies between mutations introduced into MsbA, that are implicated in lipid A binding.

Weaknesses:

The data all rely on the accuracy of determining KD values for lipid binding to MsbA. For the weaker binding sites, the range of lipid concentrations probed were in fact too low to generate highly accurate data. Another weakness is a lack of clear evidence, which KD values belong to which of the possible lipid A binding sites.

See our detailed response to reviewer 1 regarding Kd determination using native MS compared to other techniques. We chose to focus on the first three lipid binding events and adjusted the concentrations accordingly to titrate these three. As noted above, the Kd values can be determined from one mass spectrum. For rigor, we include different titration points and fit sequential binding model to the data – the fits are shown in supplemental and quite reasonable.

Regarding multiple lipids binding to different site(s), we have been able to distinguish high-affinity vs low-affinity PIP binding to Kir3.2 in a previous study [4]. This was apparent by the mole fraction curves for some lipid bound states not returning back to zero. We agree binding to multiple sites can be an issue. However, other techniques report on the ensemble of binding and, hence, no real useful information is obtained. Native MS enables one step in the right direction by dissecting the different lipid bound states. Future directions will need to further address this forefront question in the field, which we make point of now in discussion.

Reviewer #1 (Recommendations For The Authors):

Experiments/analysis: In short, there should be a proof of principle experiment that the thermodynamic constants determined by MS are accurate. Once that is done, the authors can add a more engaging structural interpretation of the results from the mutant cycles (which they seem to consciously avoid in the present manuscript?). How are cooperative residues coupled? Why?

See our detailed response to reviewer 1 above.

The manuscript is well-written, but Figures 3-5 are somewhat repetitive and require a lot of time to understand. Schematics of the main findings in each figure would help the uninitiated reader.

We agree the illustrations are complex but there is rich data being shown.

Figure 2 C contains an x-axis label error.

Corrected.

Reviewer #2 (Recommendations For The Authors):

1. Lines 128-129: "Like other mutant cycle studies, we assume the single- and double-mutations do not disrupt binding at specific sites on MsbA."

This statement is obscure and needs to be clarified. Does this mean that the mutations still allow binding of KDL, or the mutations do not disrupt the conformational integrity of the binding sites?

This statement has been removed.

1. Lines 137-139: "More specifically, R78 coordinates one of the characteristic phosphoglucosamine (P-GlcN) substituents of KDL whereas K299 interacts with a carboxylic acid group in the headgroup of KDL."

Two identical subunits form a dimer interface that forms an LPS binding site. Thus, a single mutation on the inner cavity will disrupt two binding sites on LPS. One R78 to P-ClcN and the other to a sugar backbone. Also, one K299 interacts with a carboxylic acid group in the headgroup and the other to an unknown (not clear in the figure).

Also noted above, mutation of the outer site will also impact the two outer sites. We have made note of this caveat.

1. Lines 171-172: "leading to an increase in ΔG by ~4 kJ/mol (Fig. 2d)"

Relative to what?

Corrected.

1. Lines 172-173: "Mutant cycle analysis indicates a coupling energy (ΔΔGint) of 1.7 (plus minus) 0.4 kJ/mol that contributes to the stability of KDL-MsbA complex."

The sign of DDG (DDH,DDS)_int is a bit confusing. I recommend that authors define the meaning of negative or positive sign of DDG_int (DDH,DDS) at this point. Here, a positive sign means favorable cooperation between the two mutated residues. Sometimes, researchers designate a positive cooperativity as a negative sign.

The literature on mutant cycles does not appear to follow a consensus on the sign. Here, we have revised the manuscript to note positive sign means favorable cooperation and follow the formalism recently described by Horovitz, Sharon, and co-workers [5].

1. Lines 182-185: "Enthalpy and entropy for KDL binding MsbA R188A was largely similar to the wild-type protein (Fig 3a). However, the R243A mutation resulted in an increase in entropy, compensated for by an increase in positive enthalpy (Fig 3a)."

The thermodynamic parameters for R243A mutation change in a similar manner to WT and R188A. It is R238A, not R243A, whose DH-DS interplay shows a distinct pattern from WT. Please, reword this sentence.

The sentence has been revised.

1. Lines 252-253: Solvation of polar groups in aqueous solvent has been ascribed to positive heat capacities whereas negative for apolar solvation.

This statement is not precise. More precisely, the collapse of apolar molecules from their solvated state leads to the negative "change" in heat capacity.

The sentence has been corrected.

1. Line 262-267: "These hydrophilic patches will be highly solvated, which will be desolvated upon binding lipids contributing favorably to entropy. In the case of MsbA, the selected lysine and arginine residues (based alpha carbon position) are separated by about 9 to 18 Å (PDB 8DMM). This distance could result in overlap of solvation shells that collectively contribute to the positive coupling enthalpy observed for MsbA-KDL interactions."

This statement is too speculative without presenting the degree of solvation of the residues targeted for mutation. More quantitative arguments seem to be needed.

We have removed the speculative statement.

Reviewer #3 (Recommendations For The Authors):

In this paper presented by Liu et al, native MS on the lipid A transporter MsbA was used to obtain thermodynamic insight into protein-lipid interactions. By performing the analyses at different lipid A concentrations and temperatures, dissociation constants for 2-3 lipid A binding sites were determined, as well as enthalpies were calculated using nonlinear van't Hoff fitting.

Changes in free Gibb's energies were then calculated based on the determined dissociation constants, and together with the enthalpy values obtained via van' t Hoff analysis the entropic contribution to lipid binding (DeltaS*T) was indirectly determined.

Correction – In the case on linear van’t Hoff plots, dH and dS were determined directly from the plot. For the nonlinear form of the van’t Hoff equation, which does not include an entropy fitting parameter, we back calculated dS using dH and dG at a given temperature.

The authors then included single, double and triple mutants of residues known based on cryo-EM and X-ray structures to interact with Lipid A either in the large inward-facing cavity or at a secondary binding site accessible at the surface of outward-facing MsbA, and determined the thermodynamic parameters of these mutants alone and combined to gain access to coupling energies of pairwise interactions. This method has its roots in studying pair-wise interactions of protein-protein interfaces, generally known as thermodynamic mutant cycle analysis.

Having the main expertise in ABC transporter structure-function, I will judge the paper mostly from the standpoint of what I can learn as a transporter expert from this study and whether the insights are of value for researchers with average biophysical knowledge.

My overall impression of the manuscript is that, while it contains a wealth of experimental data using the innovative and unique method of native mass spectrometry, it is hard to understand what one can learn from this analysis beyond their interesting key finding that entropy plays an important role in lipid binding (but only at certain temperatures). In particular, the lessons learned from the coupling energy analysis of the introduced mutations is hard to grasp/digest for me with regards to what I can learn from these numbers (other than learning that there are such coupling effects).

We agree the thermodynamic data is rich. Often a ddGint of zero is reported as having no coupling/significance but here the value is due to compensating ddH and d-dTS terms. In our view, this work forms the foundation of additional studies to better understand the coupling energetic terms, beyond ddGint.

In some instances, the text/figure legends are a bit unclear or contain some typos; but this part can easily be handled in a revision. The discussion is well written and embeds the main findings in the (still rather limited) literature on thermodynamic analyses of lipid binding of membrane proteins.

Major points

1. The authors may have clarified the following point in a previous paper; but at least in this paper, it is unclear to me how they purified MsbA without lipid A. The reason I am asking is that in our experience, if one purifies MsbA expressed from E. coli with standard detergents (e.g. beta-DDM) one will find a perfect density for Lipid A when determining an inward-facing structure by cryo-EM. According to the Methods, MsbA is purified initially in DDM, and rebuffered to C10E5 during size exclusion chromatography. When looking at Fig. 2b, the authors state (or assume?) that if no lipid A is added, MsbA has 0 % lipid A bound.

We have previously reported details of MsbA sample prep and optimization [6]. The revised manuscript makes note of this previous work and refers the reader to the publication. Yes, we see no appreciable signal for lipid A bound to MsbA (see Fig 2b).

We also note that samples of MsbA prepared using DDM is highly heterogenous, contaminated by a battery of small molecules (that we suspect are co-purified lipids). These contaminants will inadvertently impact biochemical studies.

1. A second topic where further clarification is in my view needed is the question of the conformations that were probed and the lipid binding sites. If I get the experimental rationale correctly, most of the data were determined in the absence of nucleotides, and only a small subset (Fig. 5) of data were determined in the presence of ATP-vanadate. However, structural evidence for the cytosolic lipid A binding site has been only determined for outward-facing MsbA (PDB: 8DMM), but has thus far not been seen in any of the inward-facing cryo-EM structures of MsbA, including recent well-resolved cryo-EM structures showing excellent density for the lipid A bound to the inward-facing cavity (PDB: 7PH2). Further, there is only one lipid A molecule that can be accommodated by the inward-facing cavity, whereas (owing to the symmetry of the homodimer) two lipid A can be bound sideways to outward-facing MsbA. Now, my understanding problem is why one does see up to three lipid A molecules bound to inward-facing apo MsbA, e.g. Fig. 2b and elsewhere. Where are they expected to bind? And what is the evidence supporting these additional binding sites?

See our detailed response to reviewer 1. If we add more lipid, we see more lipid binding to MsbA, like every other membrane protein we have studied. This data clearly indicates that there are more KDL binding site(s) – deciphering the affinity of these site(s) represents a problem on the horizon.

A further question is which lipid A binding sites are present in vanadate-trapped MsbA. Here, there are two identical binding sites (at the surface of each MsbA molecule), and it is therefore surprising to see that the affinities for the first and the second binding site are so different (see e.g. Supplementary Fig. 13).

Great point. A logical explanation (described for other biochemical systems) is the two exterior LPS binding sites display negative cooperativity i.e., binding at one site weakens the affinity at the other site.

Finally, what is the evidence that in vanadate-trapped MsbA, all molecules have closed NBDs and thus assume the outward-facing conformation? It is not uncommon that vanadate trapping leads to NBD closure only in a subfraction of all transporters (hence not in 100 % of them).

Yes, the native mass spectrum shows no appreciable signal for MsbA not trapped with vanadate/ADP. In our previous cryoEM study [6], using the vanadate-trapped transporter, we did not observe particles with NDBs dissociated in space. Regarding samples from other labs, a native mass spectrum could shed light into the population of untrapped protein – however, most studies use SDS-PAGE for quality control of their purified samples. This technology is not sufficient to address underlying biochemical issues.

We do have a new report in preparation describing a new discovery regarding trapping efficiency of MsbA.

1. The key parameter that is underlying the entire thermodynamic analysis of wt and mutant MsbA is the dissociation/association constant, which are used to calculate free Gibb's energy and, via van't Hoff analysis, enthalpy. Entropy is not determined directly, but in fact indirectly from these two numbers both depending on the measurement quality of dissociation/association constant. Now, when looking at the fitted curves as shown in Figure 2b (and in the supplement), determination of the dissociation constant for KDL1 (blue curves) look reasonable and the determined KDs are within the range of measured points. However, for KDL2 (red) and even more so KDL3 (yellow), the determined KD values (Supplementary Table 5), the measured KD values are typically higher than highest KDL conc used in the assay (1.5 uM). For this reason, and despite the fact that error bars of the fits look reasonably small, I still have doubts about the reliability of these KD values for KDL2 and KDL3.

Hence, the surprisingly strong changes of enthalpy/entropy values for different mutants/temperatures may have their origin in incorrectly determined KD values.

The increase in binding affinity of subsequent lipid binding events is consistent with many reports from our group [1, 2, 4, 6-9] and that of Prof. Robinson [10, 11] on this topic. As noted above, we indeed observe linear van’t Hoff plots with positive and negative slopes as well as non-linear curves that are convex or concave. The MsbA protein (wt or mutant), trapped or not, all display unique temperature-dependent responses. If the reviewer suggestion that the Kd values are incorrectly or randomly determined, then none of the binding data should follow thermodynamic van’t Hoff equations. This is simply not the case - the error bars and fits are reasonable. Backing up even further, looking the raw native mass spectra (see supplemental figure 1-3 and 10-11) one can see different temperature-dependence of lipid binding.

Minor points

1. Lines 116-131: this section reads as an extended introduction/aims, and does not contain any results.

This section has been moved to introduction.

1. Lines 137-139: suggested to check whether these interactions are also present in recently determined cryo-EM structures determined at fairly high resolution (PDB: 7PH2)

The interactions of MsbA and LPS (bound at the interior site) are comparable for PDB 7PH2 and 6BPL.

1. Lines 144-146: suggested to elude in more detail on the fitting procedure here, as the KD values determined in this way are the foundation of all quantitative assessments.

Details of data analysis and the fitting procedure are provided in methods.

1. Figure legend, Fig. 2: Technically, MsbA was solubilized and purified in DDM and detergent exchange was done on SEC to C10E5.

Corrected.

1. Figure legend, Fig. 4: description in a) on deconvoluted mass spec data is incorrect. Letter below needs to be adjusted accordingly.

Corrected.

1. Figure legend, Fig. 5: suggested to mention in Figure legend title that here we look at ADP-vanadate trapped MsbA.

Corrected.

References 1. Cong, X., et al., Determining Membrane Protein–Lipid Binding Thermodynamics Using Native Mass Spectrometry. Journal of the American Chemical Society, 2016. 138(13): p. 4346-4349.

1. Cong, X., et al., Allosteric modulation of protein-protein interactions by individual lipid binding events. Nat Commun, 2017. 8(1): p. 2203.

2. Qiao, P., et al., Insight into the Selectivity of Kir3.2 toward Phosphatidylinositides. Biochemistry, 2020. 59(22): p. 2089-2099.

3. Qiao, P., et al., Entropy in the Molecular Recognition of Membrane Protein-Lipid Interactions. J Phys Chem Lett, 2021. 12(51): p. 12218-12224.

4. Sokolovski, M., et al., Measuring inter-protein pairwise interaction energies from a single native mass spectrum by double-mutant cycle analysis. Nat Commun, 2017. 8(1): p. 212.

5. Lyu, J., et al., Structural basis for lipid and copper regulation of the ABC transporter MsbA. Nat Commun, 2022. 13(1): p. 7291.

6. Patrick, J.W., et al., Allostery revealed within lipid binding events to membrane proteins. Proc Natl Acad Sci U S A, 2018. 115(12): p. 2976-2981.

7. Schrecke, S., et al., Selective regulation of human TRAAK channels by biologically active phospholipids. Nature Chemical Biology, 2021. 17(1): p. 89-95.

8. Zhu, Y., et al., Cupric Ions Selectively Modulate TRAAK-Phosphatidylserine Interactions. J Am Chem Soc, 2022. 144(16): p. 7048-7053.

9. Tang, H., et al., The solute carrier SPNS2 recruits PI(4,5)P(2) to synergistically regulate transport of sphingosine1-phosphate. Mol Cell, 2023. 83(15): p. 2739-2752 e5.

10. Yen, H.Y., et al., PtdIns(4,5)P(2) stabilizes active states of GPCRs and enhances selectivity of G-protein coupling. Nature, 2018. 559(7714): p. 423-427.

Author response:

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

Reviewer #1 (Recommendations For The Authors):

I have one major concern regarding this draft of the manuscript:

(1) In the manuscript (lines 130-31) it is stated that "About 55% (8/15) of mice with unilateral AAV-hM3Dq centered in the PMv showed an increase in LH release above 0.5ng/ml within 10-20 min following the CNO injection" However, data at time zero are not shown for 4 of the 8 "LH peak" animals. The missing data at time zero seems problematic for the analysis of the CNO-stimulated cohort. As mentioned in the manuscript, the area under the curve was calculated between the range of -10 to 20min post-injection. Because diestrus animals have spontaneous LH pulses, it is highly possible that an LH pulse is initiated in the10 minutes prior to drug delivery, as seen in the AAV-mCherry group in 1D, and similarly in 2C. Given the current form of analysis, it seems possible that a spontaneous LH pulse initiated anywhere up to 10 minutes prior to drug delivery could conceivably count as an experimentally induced "LH peak". Can you address this concern?

We understand the reviewer’s concern about the spontaneous LH pulses. This is the reason we have been very strict on our analysis and have taken multiple approaches to analyze these data. In our hM3Dq group 55% of the animals responded to CNO with an increase in LH, while 0 responded in the negative control group. But also, in the clozapine group, where no time 0 points were missing, 100% of the animals with hM3Dq showed an LH increase after the injection while only 28% (2/7) showed the increase in the negative control group. Rigorously, the DREADDs approach doubled the chances of LH increase. Note that the spontaneous LH peaks observed in negative controls or during baseline show a very sharp increase and decrease at the next time point, while the 4 “PMv hits” without time 0 and increase in LH in the CNO-hM3Dq group showed a sustained rise after the 10 min or prolonged high LH levels (above 1ng/ml) even 30 min after the injection. But, ultimately, the cFOS levels in the PMv of CNO-hM3Dq group with increase in LH are significantly higher than in any other group and the number of cFOS neurons are highly correlated to LH levels. Another important aspect that should not be dismissed is that in this experimental design, we used unilateral injection in animals that are in a fed state, therefore the leptin role in rising LH levels is probably dampened.

We have added a statement to clarify this issue.

The following are minor concerns:

a) Figure 4 a-d, it is clear that Vglut2 is absent in the VMH, but it seems more relevant to show this expression pattern in the PMv.

We chose the VMH because it has a very dense collection of either LeprCre;VGlut2 or Vglut2 only cells and it illustrates very well the conditional Vglut2 deletion at small and high magnifications. In the PMv, however, the distribution of these cells is sparse. The reviewer is correct that for the current study, the PMv is more relevant and therefore, we have included images of the PMv showing a control and a LeprCre-Vglut2floxed animal in higher magnification.

b) Methods section, targeting PMv: please check the injection coordinate: "dura-mater [dorsoventral -0.54]"

Thank you for noticing this mistake, all coordinates for the injection have now been corrected (-5.4 mm, ±0.5 and -5.4mm)

Reviewer #2 (Recommendations For The Authors):

This is a very well-written manuscript by Saenz de Meira and colleagues on a careful study reporting on the key role of glutamate transporter vGlut2 expression in the neurons of the ventral perimammillary nucleus (PMv) of the hypothalamus expressing the leptin receptor LepRb in energy homeostasis, puberty, and estrous cyclicity. The authors first show using cre-dependent chemogenetic viral tools that the selective activation of the PMv LepRb induces luteinizing hormone (LH) release. Then the authors demonstrate that the selective invalidation of vGlut2 in LepRb-expressing cells in the all body induces obesity and mild alteration of sexual maturation in both sexes and blunted estrous cyclicity in females. Finally, the authors knock out vGlut2 in PMv neurons in which they reintroduce LepRb expression in an otherwise LepRb-null background using an AAV Cre approach. This latter very elegant experiment shows that while the sole re-expression of LepRb in PMv neurons in LepRb-null mice was shown before to restore puberty onset, deleting vGlut2 in LepRb-expressing PMv neurons blunts this effect.

My specific comments are as follows. Please note that none of them require additional experiments and that they can be answered by amending the text.

(1) Please provide information on the serotypes and promoters of the AAVs used in the study to enhance reproducibility.

Thank you, serotypes and promoters have been added for all AAVs.

(2) Please reformulate lines 220-221. Indeed, this reviewer does not agree with the fact that balanopreputial separation (BPS) is a sign of puberty completion. BPS is merely a sign of the advancement of sexual maturation, akin to vaginal opening in females. In certain mouse strains, BPS coincides with mini puberty rather than puberty. The definitive sign of puberty completion involves the presence of spermatozoa in the vas deferens (equivalent to the first ovulation/first estrus in females).

Thank you for this remark, this statement has now been modified.

(3) The authors convincingly show that the potential contamination of the arcuate nucleus of the hypothalamus (ARH) with the AAV injections targeted to the PMv should not account for the DREADD-mediated activation of LH release. However, do the authors believe that DREADD activation of LepRb-expressing PMv neurons, inducing cFOS expression in these neurons, could also activate ARH kisspeptin neurons (which do not express LepRb) via transsynaptic action? Alternatively, do they posit direct activation of GnRH cell bodies in the preoptic region or GnRH axon/dendrites in the ARH/median eminence region?

Thank you for this comment. We don’t have enough evidence from this DREADDs experiment to make a strong prediction on the downstream pathways. However, as discussed, from the DREADDs khrGFP females, we observed very few kisspeptin cells expressing cFOS, reducing the evidence for a PMv to ARH kisspeptin action in this case. With the evidence from our LepR-Cre;Vglut2flox animals that showed no alterations in kiss1 gene expression but a strong decrease in GnRH release, we hypothesize that this acute activation of LH is mediated by direct inputs from PMv to GnRH neurons, while acknowledging the possible existence of alternative pathways. These arguments have been added to the discussion. 

(4) This reviewer finds it intriguing that glutamatergic signaling is required for LepRb re-expression in the PMv to restore fertility. Given that the authors and others have shown that PMv neurons heavily express NOS1, the activity of which is known to heavily rely on glutamatergic NMDAR activation, the authors may want to contextualize their results in light of the recent study showing that NOS1 is found to be a new causative gene in people with congenital hypogonadotropic hypogonadism.

Thank you for the advice, we have added a paragraph discussing the possible involvement of nNos from PMv neurons in the discussion.

(5) Does the absence of vGlut2 have any impact on the obesity phenotype in mice where LepRb is selectively re-expressed in the PMv?

We have followed the weight of these animals after the AAV injections. However, due to the difficulty of generating dual homozygous (LepRnull homozygous are infertile) and producing adequate stereotaxic injections with minimum contamination of adjacent nuclei, the groups could not be run all together and thus, we refrained from performing comparative analysis of energy balance. Analysis of body weight in LepRnull mice with reactivation of LepR in PMv neurons have been published before (Donato et al., 2011 using the Flp/Frt model and Mahany et al., 2018 using the Cre/loxP system). No difference in body weight was observed in both studies. Below is the progression of body weight in mice with reactivation of LepR and deletion of Vglut2 in PMv neurons. We added a comment on this regard.

Author response image 1.

Reviewer #3 (Recommendations For The Authors):

The authors examined the effects of glutamate release from PMv LepR neurons in the regulation of puberty and reproduction in female mice. Multiple genetic mouse models were utilized to either manipulate PMv LepR neuron activities, or to delete glutamate vesicle transporters from LepR neurons. The authors have been quite rigorous in validating these models and exploring potential contaminations. Most of the data presented are solid and convincing, and support the conclusion. This reviewer has the following suggestions for the authors to further improve this work and the manuscript.

(1) The DREADD study had some issues. For example, "2 out of 7 control mice with no AAV showed an increase in LH...", indicating that LH increase may just happen randomly. More importantly, 45% of PMv-hit mice did not show LH response to CNO, making it hard to interpret the positive LH responses from the other 55% PMv-hit mice undergoing the same treatment. Overall, there are just too many variabilities in these DREADD data for anyone to come up with a clean and convincing conclusion. This reviewer suggests repeating these experiments or removing the DREADD data altogether. After all, the rest of the results are much more convincing and stand alone to support the role of glutamate release from these PMv LepR neurons.

We appreciate the reviewer’s concern. Indeed, LH shows spontaneous pulsatility which is one of the biggest challenges in our field. We have answered this concern for Reviewer 1 above and modified the text accordingly. We decided to keep the data in the publication because we believe that this is very important evidence supporting our observations since this is the only experiment that approaches the role of the PMv in a free-moving, ad libitum fed mouse model that is not deficient for leptin signaling or glutamatergic neurotransmission. Altogether this paper strongly supports a role for glutamate signaling on leptin’s action in reproductive function. Evidence for this role were dismissive or contentious until now.

(2) The mCherry signals in Figure 3 are of low quality and do not look like cell bodies.

We have now equally increased the contrast and brightness in all higher magnification images of mCherry neurons (Fig 3F, G, I and J) to improve their visibility. The lower magnification images are high quality images of areas with high density of mCherry positive neurons. Thick section (30µm) at low magnification compromises the focus at different Z-axis levels. We feel that images 3E and 3H are important to define the location of cells in the arcuate nucleus. Colocalization and mCherry expression are clear in high magnification images.

(3) The validation of Vglut2 deletion in LepR neurons (Fig. 4A-D) is very nice and convincing, but the images are from the VMH region. Why not show the PMv region?

As mentioned to Reviewer 1, we chose the VMH because it has a very dense collection of either LeprCre;VGlut2 or Vglut2 only cells and it illustrates very well the Vglut2 deletion at small and high magnifications. In the PMv, however, the distribution of these cells is sparce. The reviewer is correct that for the current study, the PMv is more relevant and therefore, we have included images of the PMv showing a control and a LeprCre-Vglut2floxed animal in higher magnification.

(4) Figures 4-5 used LepR-Cre as controls, while Figure 6 used Vglut2flox as controls. Why? Also, how did the authors set up the breedings to generate "littermates" in each of these studies?

We used the LepR-Cre as controls for our experiments since we need Cre homozygous for proper Cre expression and we had the LepR-Cre homozygous colony from the DREADDs experiment. Also, these mice had previously been thoroughly evaluated and no metabolic and/or reproductive disruption were noticed (please, see lines 213-214 of the original submission). However, our LepR-Cre colony had to be drastically reduced during COVID and suffered from unexpected Δ recombination leading to loss of Vglut2 homozygotes. To overcome these issues, we used VGlut2-floxed controls for the gene expression and GnRH immunoreactivity experiments. These mice had previously been used as controls for metabolic experiments with the LepCre-Vglut2fl genotype (Xu et al., 2013 Mol Metab), showing no deficiencies in the metabolic phenotype.

As described in the methods section (lines 464-466 of the original preprint), to inactivate glutamate in leptin responsive cells, LepRb-Cre mice were crossed with mice carrying loxP-modified Vglut2 alleles. Our experimental mice were homozygous for the LepRb-Cre allele (LepRb_cre/cre_) and homozygous for the Vglut2-loxP allele (Vglut2_fl/fl_). Our controls consisted of mice homozygous for the Cre allele (LepRb_cre/cre_;Vglut2_+/+, named LepRb-Cre) or homozygous for the Vglut2-loxP allele (LepRb+/+;Vglut2_fl/fl, named Vglut2_flox_). Both experimental (LepRb_cre/cre_;Vglut2_fl/fl_, named LepRbΔVglut2) and control mice were derived from the same litters with parents homozygous for one of the genes and heterozygous for the other gene (LepRb_cre/cre_;Vglut2_fl/+or LepRb_cre/+;Vglut2_fl/fl_). Mice were genotyped at weaning (21 days) and again at the end of the experiments.

(5) The labeling of Figures 5E-F is missing, making it hard to read.

We have confirmed that Figure 5E and F were mentioned in the figure legends and in the results text. To improve the analysis of the figure we have added the Y axis titles to Figure 5 C,D, E and F, previously only shown in Fig 5A and B.

(6) The last experiment was very nice confirming the role of glutamate release from PMv LepR neurons. However, the key phenotypes (puberty development, pregnancy) were not graphed and only stated in the text.

Thank you for your comment. Since the key result is that none the LeprLoxTb;Vglut2flox animals showed vaginal opening or pregnancy, we don’t feel the need to graph this. All the details of the reproductive and metabolic phenotyping of the Lepr-loxTB with re-expression of LepR in the PMV were described in Mahany et al., 2018.

Author Response

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

This important study shows that two methods of sleep induction in the fly, optogenetically activation of the dorsal fan-shaped body (which is rapidly reversible and maintains a neuronal activity signature similar to wakefulness), and Gaboxadol-induced sleep (which shuts down neuronal activity), produce distinct forms of sleep and have different effects on brain-wide neural activity. The majority of the conclusions of the paper are supported by compelling data, but the evidence supporting the claim that the two interventions trigger distinct transcriptional responses is incomplete.

Thank you for the helpful and detailed reviews. We feel that these have improved the manuscript considerably, and hopefully the additional figures in this Reply letter will help further convince our readers.

Public Review

In this study, Anthoney and coworkers continue an important, unique, and technologically innovative line of inquiry from the van Swinderen lab aimed at furthering our understanding of the different sleep stages that may exist in Drosophila. Here, they compare the physiological and transcriptional hallmarks of sleep that have been induced by two distinct means, a pharmacological block of GABA signaling and optogenetic activation of dorsal fan-shaped-body neurons. They first employ an incredibly impressive fly-on-the-ball 2-photon functional imaging setup to monitor neural activity during these interventions, and then perform bulk RNA sequencing of fly brains at different stages. These transcriptomic analyses leads them to (a) knocking out nicotinic acetyl-choline receptor subunits and (b) knocking down AkhR throughout the fly brain testing the impact of these genetic interventions on sleep behaviors in flies. Based on this work, the authors present evidence that optogenetically and pharmacologically induced sleep produces highly distinct brain-wide effects on physiology and transcription. The study is of significant interest, is easy to read, and the figures are mostly informative. However there are features of the experimental design and the interpretation of results that diminish enthusiasm.

a- Conditions under which sleep is induced for behavioral vs neural and transcriptional studies

1- There is a major conceptual concern regarding the relationships between the physiological and transcriptomic effects of optogenetic and pharmacological sleep promotion, and the effects that these manipulations have on sleep behavior. The authors show that these two means of sleep-induction produce remarkably distinct physiological and transcriptional responses, however, they also show that they produce highly similar effects on sleep behavior, causing an increase in sleep through increases in the duration of sleep bouts. If dFB neurons were promoting active sleep, the sleep it produces should be more fragmented than the sleep induced by the drug, because the latter is supposed to produce quiet sleep. Yet both manipulations seem to be biasing behavior toward quiet sleep.

This is a correct observation, which is already evident in our sleep architecture data (Figure 2E-H): chronic optogenetic sleep induction promotes longer sleep bouts that are similar in structure (bout number vs bout duration) to those produced by THIP feeding. Since our plots in Figure 2E-H follow the 5min sleep criterion cutoff, upon the Reviewer’s advice we re-analyzed our optogenetic experiments for short (1-5min) sleep. These are graphed below in Author response image 1. As can be seen, and as suspected by the Reviewer, the optogenetic manipulation does not increase the total amount of short sleep; indeed, it decreases it compared to baseline (these are for the exact same data as in Figure 2). Optogenetic sleep induction does not create a bunch of short sleep bouts.

Author response image 1.

Short sleep in optogenetic experiments. A. Average baseline (±SEM) 1-5min sleep across a day and night. B. Average (±SEM) 1-5min sleep in optogenenetically-activated flies, across a day and night.

We agree with the reviewer that this observation might seem inconsistent with the idea that optogenetic activation promotes active sleep, and that short sleep is active sleep. However, it does not necessarily follow that optogenetic activation has to produce short sleep. Indeed, we know from our brain imaging data (and the associated behavioral analysis) that active sleep will persist for as long as we induce it with red light. While we have not induced it for longer than 15 minutes (Tainton-Heap et al, Current Biology, 2021; Troup et al, J. of Neuroscience, 2023), this is already clearly longer than a <5min sleep bout. So our interpretation is that the longer sleep bouts induced by optogenetic activation are prolonged active sleep, rather than quiet sleep. In other words, this artificial sleep manipulation induces prolonged active sleep, rather than many short sleep bouts. This is of course different than what happens during spontaneous sleep. We have tried to be clearer about sleep bout durations in the revised manuscript (e.g., the new Figure 3), and we now admit early in the results (lines 376-380) that that we don’t know what optogenetic activation looks like in the fly brain beyond 15 minutes.

2- The authors show that the pharmacological block of GABA signaling and the optogenetic activation of dorsal fan-shaped-body neurons cause different responses on brain activity. Based on these recordings and the behavioral and brain transcriptomic data they then claim that these responses correspond to different sleep states and are associated with the expression and repression of a different constellation of genes. Nevertheless, neural activity in animals was recorded following short stimulations whereas behavioral and transcriptomic data were obtained following chronic stimulation. In this regard, it would be interesting to determine how the 12-hour pharmacological intervention they employed for their transcriptomic analysis changes neural activity throughout the brain - 12 hours will likely be too long for the open-cuticle preps, but an in-between time-point (e.g. 1h) would probably be equally informative.

The longest we’ve imaged brain activity for optogenetic sleep induction is 15 minutes, as discussed above. We see no changes in activity across this time, which would normally have led to a quiet sleep stage in spontaneous sleep recordings. Whole-brain imaging after 10 hours of optogenetic sleep induction (our RNA collection timepoint) is not realistic, and even 1 hour is difficult. We have however conducted overnight electrophysiological recordings (with multichannel silicon probes), where we activated the same R23E10 neurons for successive 20-minute bouts (alternating with 20min of no red light). We are preparing this work for publication (Van De Poll, et al). We see no evidence of optogenetic activation of this circuit ever producing anything resembling quiet sleep. Since we are not in a position to provide this new electrophysiological data in the current study, we are careful to clarify that we have not investigated what brain imaging looks like after chronic optogenetic activation (lines 376-380). We are showing through diverse lines of evidence that what is called sleep can look different in flies.

b- Efficiency of THIP treatment under different conditions

1- There are no data to quantify how THIP alters food consumption. It is evident that flies consume it otherwise they would not show increased sleep. However, they may consume different amounts of food overall than the minus THIP controls. This might have an influence on the animal's metabolism, which could at least explain the fact that metabolism-related genes are regulated (Figure 5). Therefore, in the current state, it is not possible to be certain that gene regulation events measured in this experiment are solely due to THIP effects on sleep.

We have two arguments against this reasonable criticism. First, as discussed above, the optogenetic flies are sleeping at least as much as the THIP-fed flies, so in principle they also might be feeding less. But we see no metabolic gene downregulation in the optogenetic dataset. We include this counterargument in the discussion (lines 752-756). Then, together with our co-author Paul Shaw we have shown that THIP-fed flies are not eating less compared to controls (Dissel et al, Current Biology, 2015), by tracking dye consumption. We show those results again below in Author response image 2 to support our reasoning that feeding is not an issue.

Author response image 2.

Flies were fed blue dye in their food while being sleep deprived (SD), or while being induced to sleep with 0.1mg/ml THIP in their food, or both. Dye consumption was measured in triplicate for pooled groups of 16 flies. Average absorbance at 625nm (±stan dev) is shown. Experiments were not significantly different (ANOVA of means).

2- A similar problem exists in the sleep deprivation experiments. If flies are snapped every 20 seconds, they may not have the freedom to consume appropriate amounts of food, and therefore their consumption of THIP or ATR may be smaller than in non-sleep deprived controls. Thus, it would be crucial to know whether the flies that are sleep-deprived (i.e. shaken every 20 seconds for 12 hours) actually consume comparable amounts of food (and therefore THIP) as those that are undisturbed. If not, then perhaps the transcriptional differences between the two groups are not sleep-specific, but instead reflect varying degrees of exposure to THIP.

Please see our response to the similar critique above, and how Figure R2 addresses this concern.

3- The authors should further discuss the slow action of THIP perfusion vs dFB activation, especially as flies only seem to fall asleep several minutes after THIP is being washed away. Is it a technical artifact? If not, it may not be unreasonable to hypothesize that THIP, at the concentration used, could prevent flies from falling asleep, and that its removal may lower the concentration to a point that allows its sleep-promoting action. The authors could easily test this by extending THIP treatment for another 4-5 minutes.

The reviewer is partially correct in suggesting a technical artifact: THIP does not get washed away immediately after 5min of perfusion. The drip system we employ means that THIP concentration will slowly increase to the maximum concentration of 0.2mg/ml, and then slowly get diluted away at a rate of 1.25ml/minute (this is all in the Methods). In a previous study (Yap et al, Nature Communications, 2017) we used this exact same perfusion procedure to test a range of THIP concentrations, and settled on 0.2mg/ml as the lowest that reliably induced quiet sleep within 5 minutes. Higher concentrations induced quiet sleep faster, so the alternate explanation proposed by the Reviewer is not supported. We feel that our previous electrophysiological study provided the necessary groundwork for using the same approach and dosage here for our whole-brain imaging readout.

c- Comments regarding the behavioral assays

1- L319-322: the authors conclude that dFB stimulation and THIP consumption have similar behavioral effects on sleep. However, this is inaccurate as in Figure S1 they explain that one increases bout number in both day and night and the other one only during the day.

We have now added a caveat about night bout architecture being different (lines 353-356). Figure S1 is now Figure 3.

2- The behavioral definitions used for active and quiet sleep do not fit well with strong evidence that deep sleep (defined by lowered metabolic rates) is probably most closely associated with bouts of inactivity that are much longer than the >5min duration used here, i.e., probably 30min and longer (Stahl et al. 2017 Sleep 40: zsx084). Given that the authors are providing evidence that quiet sleep is correlated with changes in the expression of metabolism related genes, they should at least discuss the fact that reductions in metabolism have been shown to occur after relatively long bouts of inactivity and might reconsider their behavioral sleep analysis (i.e., their criteria for sleep state) with this in mind.

Interestingly, induced sleep bout durations are on average longer for the optogenetic manipulation (40min vs 25min); this was evident in Figure S1C vs S1F (now Figure 3). So as discussed above, this provides a counterargument for sleep bout duration alone being indicative of metabolic processes associated with quiet sleep: the optogenetic dataset did not uncover metabolic-related pathways as relevant to that sleep manipulation. We refer to Stahl et al, Sleep, 2017, in our discussion (lines 748-750), making exactly this point about metabolic rates being decreased in longer sleep bouts, and flowing up with our observation that optogenetic flies sleep just as much, and their bouts are actually longer. So clearly different processes must be involved.

d- Comments regarding the recordings of neuronal activity

1- There is an additional concern regarding the proposed active and quiet sleep states that rest at the heart of this study. Here these two states in the fly are compared to the REM and NREM sleep states observed in mammals and the parallels between active fly sleep and REM and quiet fly sleep and NREM provide the framework for the study. The establishment of such parallel sleep states in the fly is highly significant and identifying the physiological and molecular correlates of distinct sleep stages in the fly is of critical importance to the field. However, the proposal that the dorsal fan shaped body (dFB) neurons promote active sleep runs counter to the prevailing model that these neurons act as a major site of sleep homeostasis. If quiet sleep were akin to NREM, wouldn't we expect the major site of sleep homeostasis in the brain to promote it? Furthermore, the authors state that the effects of dFB neuron excitation on transcription have "almost no overlap" (line 500) with the transcriptomic effects of sleep deprivation (Supplementary Table 3), which is not what would be expected if dFB neurons are tracking sleep pressure and promoting sleep, as suggested by a growing body of convergent work summarized on page four of the manuscript. Wouldn't the 10h excitation of the dFB neurons be predicted to mimic the effects of sleep deprivation if these neurons "...serve as the discharge circuit for the insect's sleep homeostat..." (line 60)? Shouldn't their prolonged excitation produce an artificial increase in sleep drive (even during sleep) that would favor deep, restorative sleep? How do the authors interpret their results with regard to the current prevailing model that dFB neurons act as a major site of sleep homeostasis? This study could be seen as evidence against it, but the authors do not discuss this in their Discussion.

These are all excellent and thoughtful points, which have made us re-think parts of our discussion. First off, the potential comparison with REM and NREM is entirely speculative, and we have tried to make that more obvious in introduction) and the discussion (e.g, see lines 43, 708, 818). The evidence that the FB neurons (and maybe others) are involved in the homeostatic regulation of sleep is well-supported in the literature, so that part of the discussion holds. However, we concede that the timing of our sleep manipulations could benefit from more explanation. We conducted these during the flies’ subjective day, after the animals had presumably had a good night’s sleep. This means that we induced either kind of sleep for 10 daytime hours, which presumably replaced whatever behavioural states would ‘naturally’ be happening during the day. Female flies sleep less during the day than at night, and we have shown in previous work that daytime sleep quality is different than night-time sleep (van Alphen et al, Journal of Neuroscience, 2013), leading us to suggest that most ‘deep’ or quiet sleep happens at night, for flies. Following this reasoning, daytime optogenetic activation might not be depriving flies of much quiet sleep, or accumulating a deep sleep drive as the Reviewer proposes. Rather, both induced sleep manipulations could be providing 10 hours of either kind of sleep that the flies don’t really ‘need’. Why did we design it this way? Firstly, we were interested in simply asking what these chronic sleep manipulations do to gene expression in rested flies, and how they might be similar or different. We focussed on daytime manipulations to avoid precisely the confound of sleep pressure, and also because we observed red-light artifacts at night for our optogenetic experiments (which we reported). Our sleep deprivation strategy was designed specifically as a control for the THIP (Gaboxadol) experiments, to control for non-sleep related effects of the drug (see below our rationale for why this was less crucial for the optogenetic experiments). In conclusion, we had a logical rationale for how the experiments were done, centred on the straightforward question of whether these two different approaches to sleep induction were having similar effects in well-rested flies. In retrospect, we were not anticipating the Reviewer’s thoughtful logic regarding the dFB’s potential role in also regulating deep sleep homeostasis. We now provide some discussion along these lines to make readers aware of this line of reasoning, as well as our rationale for why prolonged optogenetic sleep induction was not sleep-depriving (lines 768-777).

2- Regarding the physiological effects of Gaboxadol, to what extent is the quieting induced by this drug reminiscent of physiology of the brains of flies spontaneously meeting the behavioral criterion for quiet sleep? Given the relatively high dose of the drug being delivered to the de-sheathed brain in the imaging experiments (at least when compared to the dose used in the fly food), one worries that the authors may be inducing a highly abnormal brain state that might bear very little resemblance to the deeply sleeping brain under normal conditions. As the authors acknowledge, it is difficult to compare these two situations. Comparing the physiological state of brains put to sleep by Gaboxadol and brains that have spontaneously entered a deep sleep state therefore seems critical.

As discussed above, our Gaboxadol (THIP) perfusion concentration (0.2mg/ml) was the minimal dosage that effectively induced sleep within 5 minutes, based upon previously published work (Yap et al, Nature Communications, 2017). Lower concentrations were unreliable, with some never inducing sleep at all. Comparisons with feeding THIP are tenuous, and we make that clear in our discussion (lines 731-735). Nevertheless, the Reviewer makes an excellent point about comparisons with spontaneous ‘quiet’ sleep. Here, we feel well supported (please see Author response image 3 below, comparing THIP-induced sleep (this work, B) and spontaneous sleep (A) from previous study). In our previous study (Tainton-Heap et al, 2021) we showed that neural activity and connectivity decreases during spontaneous quiet sleep. This is what we also see with THIP perfusion. In contrast, in Troup et al, J. of Neuroscience (2023) we confirm that neither neural activity nor connectivity changes during optogenetic R23E10 activation, and general anesthesia – unlike THIP – does NOT produce a quiet brain state. Our finding that THIP effects are nothing like general anesthesia (at the level of brain activity levels) suggests a physiological sleep state closer to spontaneous quiet sleep. We elaborate on this important observation in our results, also pointing to crucial differences with general anesthesia (lines 411-415).

Author response image 3.

THIP-induced sleep resembles quiet spontaneous sleep. A. Calcium imaging data from spontaneously sleeping flies, taken from Tainton-Heap et al, 2021. Left, percent neurons active; right, mean degree, a measure connectivity among active neurons. Both measures decrease during later stages of sleep. B. Calcium imaging data from flies induced to sleep with 5min of 0.2mg/ml THIP perfusion (this study). Left, percent neurons active; right, mean degree. Both measures are significantly decreased, resembling the later stages of spontaneous sleep, which we have termed ‘quiet sleep. Hence THIP-induced sleep resembles quiet sleep. Note that the genetic background is different in A and B, hence the different baseline activity levels.

3- There are some issues with Figure 3, in particular 3C-D. It is not clear whether these panels show representative traces or an average, however both the baseline activity and fluorescence are different between C and D, in particular in their amplitude. Therefore, it is difficult to attribute the differences between C and D to the stimulation itself or to the previously different baseline. In addition, the fact that flies with dFB activation seem to keep a basal level of locomotor activity whereas THIP-treated ones don't is quite striking, however it is not being discussed. Finally, the authors claim that the flies eventually wake up from THIP-induced sleep (L360-361), however there are no data to support this statement.

These are representative traces, which is a way of showing the raw calcium data (Cell ID) so readers can see for themselves that one manipulation silences whereas the other does not – even though flies become inactive for both. The Y-axis scale is standard deviation of the experiment mean. Since THIP decreases neural activity, then the baseline is comparatively higher. Since optogenetic activation does not change average neural activity levels, the baseline is centered on zero. This is an outcome of our analysis method and does not reflect any ‘true’ baseline. We have now clarified this in our figure legend. We now also confess that flies rendered asleep optogenetically can be ‘twitchy’ (line 374). Finally, we show data for 3 flies that were recorded until they woke up. The rest were verified behaviorally, after the experiment. This is now explained in the Methods.

4- In Figure 4C, it is strange that the SEM is always exactly the same across the whole experiment. Readers should be aware that there might have been an issue when plotting the figure.

This is not a mistake, the standard errors are just all quite close (between 0.17 and 0.22). This is because of the way we did the analysis, asking how many flies responded to each stimulus event, with incremental levels of responsiveness. This is explained in the Methods. The figure makes the important point of sleep and recovery.

e- Comments regarding the transcript analyses

1- General comment: the title of this manuscript is inaccurate - the "transcriptome" commonly refers to the entirety of all transcripts in a cell/tissue/organ/animal (including genes that are not differentially expressed following their interventions), and it is therefore impossible to "engage two non-overlapping transcriptomes" in the same tissue. Perhaps the word "transcriptional programs" or transcriptional profiles" would be more accurate here?

We thank the Reviewer for this advice and have changed the title as proposed.

2- Given the sensitivity of transcriptomic methods, there is a significant concern that the optogenetic experiments are not as well controlled as they could be. Given the need for supplemental all-trans retinal (ATR) for functional light gating of channelrhodopsins in the fly, it is convenient to use flies with Gal4-driven opsin that have not been given supplemental ATR as a negative control, particularly as a control for the effects of light. However, there is another critical control to do here. Flies bearing the UAS-opsin responder element but lacking the GAL4 driver and that have been fed ATR are critical for confirming that the observed effects of optogenetic stimulation are indeed caused by the specific excitation of the targeted neurons and not due to leaky opsin expression, or the effect of ATR feeding under light stimulation or some combination of these factors. Given the sensitivity of transcriptomic methods, it would be good to see that the candidate transcripts identified by comparing ATR+ and ATR- R23E10GAL4/UAS-Chrimson flies are also apparent when comparing R23E10GAL4/UAS-Chrimson (ATR+) with UAS-Chrimson (ATR+) alone.

We have not done these experiments on UAS-Chrimson/+ controls. Like many others in our field, we viewed non-ATR flies as the best controls, because this involves identical genotypes. Since we were however aware that ATR feeding itself could be affect gene expression, we specifically checked for this with our early (1hour) collection timepoint. We only found 26 gene expression differences between ATR and -ATR flies at this early timepoint, compared with 277 for the 10-hour timepoint. We detail this rationale in our results, explaining why this is a convincing control for ATR feeding. If there was leaky opsin expression / activity, this would have been evident in our design. Regarding the cumulative effect of light, this would also have been accounted in our design, as only 1 hour would have elapsed in our first timepoint compared to 10 hours in our second. While the Reviewer is correct in saying that parental controls are called for in many Drosophila experiments, this becomes quickly unmanageable in transcriptomic studies, which is exactly why well-designed +ATR vs -ATR comparisons in the exact same strain are most appropriate. We feel that our 1-hr timepoint mostly addresses this concern.

3- Figures about qPCR experiments (5G and 6G) are problematic. First, whereas the authors seem satisfied with the 'good correspondence' between their RNA-seq and qPCR results, this is true for only ~9/19 genes in 5G and 2/6 genes in 6G. Whereas discrepancies are not rare between RNA-seq and qPCR, the text in L460-461 and 540-541 is misleading. In addition, it is unclear whether the n=19 in L458 refers to the number of genes tested or the number of replicates. If the qPCR includes replicates, this should be more clearly mentioned, and error bars should be added to the corresponding figures.

We consider that our qPCR validations were convincing, as they were all mostly changed in the ‘right’ direction. We agree that are some discrepancies, so have modified our language to reflect this. We have also clarified that 19 refers to the number of genes validated by qPCR in that THIP dataset. All qPCRs involved three technical replicates. We prefer to keep these histograms the way they are to convey these simple trends. For complete transparency, we now provide a supplemental Excel worksheet with all of the qPCR data, alongside corresponding RNAseq data and stats for the selected genes (Supplementary Table 9).

4- There is a lack of error bars for all their RNAseq and qPCR comparisons, which is particularly surprising because the authors went to great lengths and analyzed an applaudably large amount of independent biological replicates, yet the variability observed in the corresponding molecular data is not reported.

The genes reported in each of our datasets and associated supplemental figures and tables were all significant, as determined by criteria outlined in the Methods. However, we appreciate that readers might want to get a sense of the values and variances involved, as well as access to the entire gene datasets. We now provide all of these as additional ‘sheets’ in our existing supplemental tables (S2-S7), so this should be very easy to navigate and evaluate. In addition to the previously provided lists for significant genes, in the second Excel sheet (‘All genes’) readers will be able to see the data for all 5 replicates, for the significant genes as well as all other ~15,000 genes (listed in alphabetical order). We feel that this will be a helpful resource, because admittedly significance thresholds can still be a little arbitrary and some readers might want to look up ‘their’ genes of interest.

Comments to authors

Other comments

1- Text in L441 & 606 is misleading. According to ref 52, AkhR is involved specifically in starvation-induced sleep loss, and not in general sleep regulation.

Corrected.

2- The language used in L568-570 and 573-574 is confusing. The authors should specify that the knock down of cholinergic subunits, rather than the subunits themselves is what causes sleep to increase or decrease.

Corrected.

3- The authors' investigation of cholinergic receptor subunits function is very preliminary, and it is difficult to draw any conclusion from what is presented here. In particular, their behavioral data is difficult to reconcile with the RNA-seq data showing overexpression of both short sleep increasing and short sleep decreasing subunits. Without knowing where in the brain these subunits are required for controlling sleep, the data in Figure 7 is difficult to appreciate.

We have now conducted additional experiments where we specifically knocked down these alpha receptor subunits (all 7 of them) in the R23E10 neurons. This seemed an obvious knockdown location, to determine if any of these subunits regulated activity in the same sleep promoting neurons that were the focus of this study. We found that alpha1 knockdown in these neurons had similar sleep phenotypes, which we believe is an important result. Since this functional localisation is a logical ending for the paper, we have now made it the final figure.

Suggestions & comments

1- It would be interesting if the authors could discuss their findings that metabolism genes are downregulated in THIP flies in the context of recent work that showed upregulation of mitochondrial ROS after sleep deprivation (Kempf et al, 2019).

We now add the Kempf 2019 reference and allude to how those findings could be consistent with ours.

2- The fact that THIP-induced sleep persists long after THIP removal (Fig 3D) is very intriguing and interesting. This suggests that the drug might trigger a sleep-inducing pathway that can continue on its own without the drug, once activated.

This is correct, and in stark contrast to the optogenetic manipulation we employ, which does not appear to show such sleep inertia. We have now added a sentence highlighting this interesting difference (lines 394-396).

3- The authors identify many new genes regulated in response to specific methods for sleep induction. These are all potentially interesting candidates for further studies investigating the molecular basis of sleep. It would be interesting to know which of these genes are already known to display circadian expression patterns.

By providing all of the gene lists, these are now available to ask questions such as these. We hesitate however to delve into this domain for this work, as our main goal was to compare these two kinds of sleep in flies.

4- The brain-wide monitoring of neural activity invites a number of very exciting follow-up experiments - most importantly, it would be fascinating to establish, which neurons are active in the different phases the authors describe! Are these neurons that are involved in transmitting external visual stimuli to the central brain? Do they also project into the central complex? They could make use of the large collection of existing driver lines in the fly and they could also exploit the extraordinary knowledge of the connectome and transcriptome of the fly brain.

Thank you for sharing our enthusiasm for these likely future directions.

5- The Dalpha2,3,4,6 and 7 Knock-out strains they generate will be a useful reagent for the Drosophila neuroscience community once the efficiency/success of the knock-out has been confirmed by qPCR.

These knockout strains have all been confirmed by our co-authors Hang Luong, Trent Perry, and Philip Batterham. These knockout confirmations are outlined in publications that we reference (Perry et al, 2021).

Materials and methods:

1- This study has employed custom-built apparatus and custom-written code/scripts, but these do not appear to be available to the reader. For the sake of replicability, the authors should make these available.

The code/scripts are available via the University of Queensland research data management system as described in the Methods, and can be sent by the Lead Contact. The imaging hardware and analysis code are identical to what was described in a previous publication, and available as directed therein (Tainton-Heap et al, 2021).

2- Also, the authors should give details on the food used to rear their flies. Fly media comes in several common forms and sleep is sensitive to diet.

This has now been elaborated in the beginning of the Methods.

3- The light regime used for optogenetic excitation of dFB neurons consists of 12h of uninterrupted bright red LED light. Most optogenetic stimulations consist of pulsed high frequency flashes interlaced with pauses in illumination. Can dFB neurons be driven constitutively with 12 hours of bright light?

We showed in Tainton-Heap (2021) that 7Hz pulsed red light had exactly the same effect on R23E10/Chrimson readouts as continuous red light, which is why we opted here to provide continuous red light. That optogenetic sleep induction can be driven continuously for 12 hours is evident by our 24-hour sleep profiles. However, we agree that one could question whether sleep quality is similar after 12 hours. To address this, we did an additional experiment where we stimulated the flies hourly, to determine if their behavioural responsiveness to mechanical stimuli changed over the course of continued sleep induction, for both optogenetic and THIP-induced sleep. We present the data below in Author response image 4. As can be seen in these new analyses, while optogenetic sleep induction persists across 12 daytime hours (speed is close to zero throughout), flies do indeed become more responsive later in the day. This could have two different interpretations: either some sleep functions are being satisfied over time, or the activation regime is becoming less effective over time. Either way, these data show that at our 10-hour daytime timepoint, unstimulated flies are still largely inactive, even though their arousal thresholds might have gradually changed; so the uninterrupted red-light regime is still effective. The comparison with THIP is interesting: here there does not seem to be a change in responsiveness over time; the drug just decreases behavioral responsiveness throughout. Together, these experiments support our view that both approaches are sleep-promoting throughout the 12-hour day, although we appreciate that sleep quality is not identical.

Author response image 4.

A) The average speed of baseline (grey) and optogenetically-activated flies (green) across 24 hours. Red dots indicate vibration stimulus times. B) The average speed of control (grey) and THIP-fed flies (blue) across 24 hours. Flies are all R23E10/Chrimson. N= 87 for optogenetic, n=88 for -THIP, n=85 for +THIP.

4- The authors use the SNAP apparatus to prevent THIP-treated flies from sleeping to tease out possible sleep-independent effects. This is an excellent control. Why have the authors not done the same with the optogenetic treatment? It's surprising not to see this control given the concern the authors express (lines 501 - 502) that the dFB manipulation might be paralyzing awake flies, which certainly seems possible given the light regimes used. Why not test this directly with SNAP?

We appreciate that this may have been a valuable additional control. However, we designed this control for the THIP experiments specifically because of concerns about THIP’s (yet unknown) mechanism of action in flies. THIP is a gabaergic drug with most likely many off-target effects that have little to do with sleep, hence the need for a control where we compare to flies that ingested THIP but have been prevented from sleeping. In contrast, R23E10-driven sleep induction is exactly that, a circuit when activated that induces sleep. Whatever specific neurons might really be involved, the Gal4 circuit is sleep-inducing. This is well supported by multiple publications. The most appropriate control for assessing transcriptomic effects during optogenetic sleep here is not preventing sleep, but rather no increased sleep in flies that have not ingested ATR, and comparing that to effects of ATR alone, which is what we have done. Adding a sleep-deprivation layer onto both of these analyses may have been interesting, but a lot more analyses and not strictly required to identify relevant sleep-related genes. We have rephrased the misleading sentence about paralyzing flies, to instead clarify that lack of overlap with the SD dataset suggests that optogenetic activation is not preventing sleep functions from being engaged.

5- A pairwise comparison of ZT01 and ZT10 does not address circadian expression cycles in a meaningful way. There will be strong effects of the LD cycle here. I suggest toning this down. (Though it is gratifying to see the expected changes in the core clock genes.)

We have changed the language from ‘circadian’ to ‘light-dark’ to address this, although have kept the word ‘circadian’ when referring specifically to genes such as per, clock, timeless, etc.

6- Line 109: There is a reference missing.

We now provide the relevant reference.

Results

1- General comment regarding the figures: a general effort could be made to improve the design and quality of the figures and make them more readable. There are a lot of issues such as stretched or misaligned text, badly drawn frames, etc.

We think we know which figures this might relate to (e.g., Figures 3,4B), so we have adjusted where appropriate.

2- Instead of 'dFB-induced' (e.g., L77) it would be more accurate to use 'optogenetically-induced'

Thank you for this helpful advice. We have changed our language throughout to say ‘optognetically-induced’

3- Figure S1 should be integrated in the main figure to make the quantification more easily 4accessible.

We have integrated Figure S1 into the main figures. It is now Figure 3.

5- It would be good to include red light controls in Figure 2C, E, G.

Making Figure S1 a main figure has better highlighted the fact that we have done red light controls (‘baseline’).

6- line 313: Fig2E-H - these graphs would benefit if the authors made it more obvious where the maximum sleep amount would fall - i.e. the combination of bouts and minutes that add up to 12 hours (and therefore the entire day/night)

If a fly were to sleep uninterrupted for all 12 hours of a day or night, that would amount to a sleep bout 720 minutes long. We do not feel that identifying this maximum on these graphs would be helpful. It should be clear from the data that a floor is reached with very few sleep bouts exceeding 60 minutes in our paradigm. To help orient the reader though, we now clarify in the figure legend that the maximum is 720 minutes or 12 hours.

7- Fig. 2B, D: It was not clear why the authors took the 3-day average here. Doesn't that lead to a whole range of very different behaviors? I could, perhaps naively, imagine that a fly's behavior changes after 2 days of almost-permanent sleep?

We took the 3-day average because the effect of THIP on each successive day was not significantly different (see Author response image 5, below). Flies wake up enough to have a good feed (see Author response image 2) and then go back to sleep. Since this is however an important point raised by the reviewer, we now mention in the Methods that sleep duration was not different among the 3 averaged days and nights (lines 193-195).

Author response image 5.

Data from THIP feeding experiment (Figure 2B) in manuscript, separated into 3 successive days and nights, with THIP-fed flies (blue) compared to controls (white). Averages  SD are shown, samples sizes are the same as in Figure 2D. No THIP data was significantly different across days and nights (ANOVA of means).

8- In Figure 2C the authors compare optogenetically induced to "spontaneous sleep," which I think refers to baseline sleep before stimulation, according to the figure. I think the proper comparison would be to the red light control (ATR-); though see the comment above regarding optogenetic controls).

This information was provided in Figure S1. We now provide it as a main Figure 3, as requested above.

We also made a point about red light having an effect at night, which is why we focussed on daytime effects for our transcriptomic comparisons. We feel that the ATR-fed flies (minus red light) are an appropriate control here for optogenetically-induced sleep: same exact genotype and ATR feeding, just no optogenetic activation. We therefor would prefer to keep these graphs as they are, especially since we show -ATR data subsequently.

9- Figures 3A and 4A are redundant; Figure 3B has some active ROIs that are outside of the brain. I am not sure how this is possible?

We have removed the redundant 4A and replaced it with the THIP molecule to clearly signal what this figure is focussed on. In Figure 3B (now 4B), the brain mask is a visual estimate made from the middle of the image stack. Some neurons in other layers are outside this single-layer estimate. All neurons were all accounted for.

10- Figure 4B is confusing. It took me a while to understand and so it can do with re-drawing in a more accessible way.

We agree that this was confusing, e.g. there were too many arrows. We have redrawn and simplified (Now 5A).

11- The authors state that flies wake up from THIP-induced sleep on the ball, but in Figure 4D there appears to be fewer samples for flies who have woken up from THIP (3) compared to those observed before THIP administration. Are flies dying?

None of the flies died. Most flies were removed from imaging to confirm recovery, while 3 were left in our imaging setup to measure brain activity upon recovery. These results are in Figure 5C and now clarified in the Methods.

12- Fig5C,D: I'm surprised that by far the most significant changes (in terms of log2-FC and p-val) occur in the sleep-deprived flies? It is not clear to me what the authors mean by effects that "relate waking process"? Perhaps they could elaborate on this?

We have removed the phrase ‘relates to waking processes’. We now also remark on the high level of fold-change in many of these genes but refrain from discussing this further in the results. It is interesting though.

13- The sentence in L425-428 is unclear - it would be good to rephrase this.

We have rephrased this sentence, hopefully it’s clearer now.

14- Text in L544-545 is confusing. What do you mean by 'less clear'?

We have replaced ‘less clear’ with ‘not dominated by a single category’.

15- It is unclear what is the control in Fig 7A. It would be good to mention what strain was used.

Different knockout strains had different controls. These are identified in the figure legend and Methods.

16- L579-581: it would be helpful to include this data in a supplementary figure.

We now provide this as a supplementary figure as requested (Supplementary Figure 6).

17- There is no information about R57C10 in the methods - it would be good to explain which neurons this line labels, and why you chose it.

We now clarify in the methods that R57C10-Gal4 is a pan-neural driver, and provide a reference.

18- Table S5 - If I'm not mistaken then the first line should say 1h, not 10h.

Corrected

Author response:

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

We would like to thank the reviewers for helping us improve our article and software. The feedback that we received was very helpful and constructive, and we hope that the changes that we have made are indeed effective at making the software more accessible, the manuscript clearer, and the online documentation more insightful as well. A number of comments related to shared concerns, such as:

• the need to describe various processing steps more clearly (e.g. particle picking, or the nature of ‘dust’ in segmentations)

• describing the features of Ais more clearly, and explaining how it can interface with existing tools that are commonly used in cryoET

• a degree of subjectivity in the discussion of results (e.g. about Pix2pix performing better than other networks in some cases.)

We have now addressed these important points, with a focus on streamlining not only the workflow within Ais but also making interfacing between Ais and other tools easier. For instance, we explain more clearly which file types Ais uses and we have added the option to export .star files for use in, e.g., Relion, or meshes instead of coordinate lists. We also include information in the manuscript about how the particle picking process is implemented, and how false positives (‘dust’) can be avoided. Finally, all reviewers commented on our notion that Pix2pix can work ‘better’ despite reaching a higher loss after training. As suggested, we included a brief discussion about this idea in the supplementary information (Fig. S6) and used it to illustrate how Ais enables iteratively improving segmentation results. 

Since receiving the reviews we have also made a number of other changes to the software that are not discussed below but that we nonetheless hope have made the software more reliable and easier to use. These include expanding the available settings, slight changes to the image processing that can help speed it up or avoid artefacts in some cases, improving the GUI-free usability of Ais, and incorporating various tools that should help make it easier to use Ais with remote data (e.g. doing annotation on an office PC, but model training on a more powerful remote PC). We have also been in contact with a number of users of the software, who reported issues or suggested various other miscellaneous improvements, and many of whom had found the software via the reviewed preprint.

Reviewer 1 (Public Review):

This paper describes "Ais", a new software tool for machine-learning-based segmentation and particle picking of electron tomograms. The software can visualise tomograms as slices and allows manual annotation for the training of a provided set of various types of neural networks. New networks can be added, provided they adhere to a Python file with an (undescribed) format. Once networks have been trained on manually annotated tomograms, they can be used to segment new tomograms within the same software. The authors also set up an online repository to which users can upload their models, so they might be re-used by others with similar needs. By logically combining the results from different types of segmentations, they further improve the detection of distinct features. The authors demonstrate the usefulness of their software on various data sets. Thus, the software appears to be a valuable tool for the cryo-ET community that will lower the boundaries of using a variety of machine-learning methods to help interpret tomograms. 

We thank the reviewer for their kind feedback and for taking the time to review our article. On the basis of their  comments, we have made a number of changes to the software, article, and documentation, that we think have helped improve the project and render it more accessible (especially for interfacing with different tools, e.g. the suggestions to describe the file formats in more detail). We respond to all individual comments one-by-one below.

Recommendations:

I would consider raising the level of evidence that this program is useful to *convincing* if the authors would adequately address the suggestions for improvement below.

(1) It would be helpful to describe the format of the Python files that are used to import networks, possibly in a supplement to the paper. 

We have now included this information in both the online documentation and as a supplementary note (Supplementary Note 1). 

(2) Likewise, it would be helpful to describe the format in which particle coordinates are produced. How can they be used in subsequent sub-tomogram averaging pipelines? Are segmentations saved as MRC volumes? Or could they be saved as triangulations as well? More implementation details like this would be good to have in the paper, so readers don't have to go into the code to investigate. 

Coordinates: previously, we only exported arrays of coordinates as tab-separated .txt files, compatible with e.g. EMAN2. We now added a selection menu where users can specify whether to export either .star files or tsv .txt files, which together we think should cover most software suites for subtomogram averaging. 

Triangulations: We have now improved the functionality for exporting triangulations. In the particle picking menu, there is now the option to output either coordinates or meshes (as .obj files). This was previously possible in the Rendering tab, but with the inclusion in the picking menu exporting triangulations can now be done for all tomograms at once rather than manually one by one.

Edits in the text: the output formats were previously not clear in the text. We have now included this information in the introduction:

“[…] To ensure compatibility with other popular cryoET data processing suites, Ais employs file formats that are common in the field, using .mrc files for volumes, tab-separated .txt or .star files for particle datasets, and the .obj file format for exporting 3D meshes.”

(3) In Table 2, pix2pix has much higher losses than alternatives, yet the text states it achieves fewer false negatives and fewer false positives. An explanation is needed as to why that is. Also, it is mentioned that a higher number of epochs may have improved the results. Then why wasn't this attempted? 

The architecture of Pix2pix is quite different from that of the other networks included in the test. Whereas all others are trained to minimize a binary cross entropy (BCE) loss, Pix2pix uses a composite loss function that is a weighted combination of the generator loss and a discriminator penalty, neither of which employ BCE. However, to be able to compare loss values, we do compute a BCE loss value for the Pix2pix generator after every training epoch. This is the value reported in the manuscript and in the software. Although Pix2pix’ BCE loss does indeed diminish during training, the model is not actually optimized to minimize this particular value and a comparison by BCE loss is therefore not entirely fair to Pix2pix. This is pointed out (in brief) in the legend to the able: 

“Unlike the other architectures, Pix2pix is not trained to minimize the bce loss but uses a different loss function instead. The bce loss values shown here were computed after training and may not be entirely comparable.”

Regarding the extra number of epochs for Pix2pix: here, we initially ran in to the problem that the number of samples in the training data was low for the number of parameters in Pix2pix, leading to divergence later during training. This problem did not occur for most other models, so we decided to keep the data for the discussion around Table 1 and Figure 2 limited to that initial training dataset. After that, we increased the sample size (from 58 to 170 positive samples) and trained the model for longer. The resulting model was used in the subsequent analyses. This was previously implicit in the text but is now mentioned explicitly and in a new supplementary figure. 

“For the antibody platform, the model that would be expected to be one of the worst based on the loss values, Pix2pix, actually generates segmentations that are seem well-suited for the downstream processing tasks. It also output fewer false positive segmentations for sections of membranes than many other models, including the lowest-loss model UNet. Moreover, since Pix2pix is a relatively large network, it might also be improved further by increasing the number of training epochs. We thus decided to use Pix2pix for the segmentation of antibody platforms, and increased the size of the antibody platform training dataset (from 58 to 170 positive samples) to train a much improved second iteration of the network for use in the following analyses (Fig. S6).”

(4) It is not so clear what absorb and emit mean in the text about model interactions. A few explanatory sentences would be useful here. 

We have expanded this paragraph to include some more detail.

“Besides these specific interactions between two models, the software also enables pitching multiple models against one another in what we call ‘model competition’. Models can be set to ‘emit’ and/or ‘absorb’ competition from other models. Here, to emit competition means that a model’s prediction value is included in a list of competing models. To absorb competition means that a model’s prediction value will be compared to all values in that list, and that this model’s prediction value for any pixel will be set to zero if any of the competing models’ prediction value is higher. On a pixel-by-pixel basis, all models that absorb competition are thus suppressed whenever their prediction value for a pixel is lower than that of any of the emitting models.”

(5) Under Figure 4, the main text states "the model interactions described above", but because multiple interactions were described it is not clear which ones they were. Better to just specify again. 

Changed as follows:

“The antibody platform and antibody-C1 complex models were then applied to the respective datasets, in combination with the membrane and carbon models and the model interactions described above (Fig. 4b): the membrane avoiding carbon, and the antibody platforms colocalizing with the resulting membranes”.

(6) The next paragraph mentions a "batch particle picking process to determine lists of particle coordinates", but the algorithm for how coordinates are obtained from segmented volumes is not described. 

We have added a paragraph to the main text to describe the picking process:

“This picking step comprises a number of processing steps (Fig. S7). First, the segmented (.mrc) volumes are thresholded at a user-specified level. Second, a distance transform of the resulting binary volume is computed, in which every nonzero pixel in the binary volume is assigned a new value, equal to the distance of that pixel to the nearest zero-valued pixel in the mask. Third, a watershed transform is applied to the resulting volume, so that the sets of pixels closest to any local maximum in the distance transformed volume are assigned to one group. Fourth, groups that are smaller than a user-specified minimum volume are discarded. Fifth, groups are assigned a weight value, equal to the sum of the prediction value (i.e. the corresponding pixel value in the input .mrc volume) of the pixels in the group. For every group found within close proximity to another group (using a user-specified value for the minimum particle spacing), the group with the lower weight value is discarded. Finally, the centroid coordinate of the grouped pixels is considered the final particle coordinate, and the list of all

coordinates is saved in a tab-separated text file.

“As an alternative output format, segmentations can also be converted to and saved as triangulated meshes, which can then be used for, e.g., membrane-guided particle picking. After picking particles, the resulting coordinates are immediately available for inspection in the Ais 3D renderer (Fig. S8).“

The two supplementary figures are pasted below for convenience. Fig. S7 is new, while Fig. S8 was previously Fig. S10 -the reference to this figure was originally missing in the main text, but is now included.

(7) In the Methods section, it is stated that no validation splits are used "in order to make full use of an input set". This sounds like an odd decision, given the importance of validation sets in the training of many neural networks. Then how is overfitting monitored or prevented? This sounds like a major limitation of the method. 

In our experience, the best way of preparing a suitable model is to (iteratively) annotate a set of training images and visually inspect the result. Since the manual annotation step is the bottleneck in this process, we decided not to use validation split in order to make full use of an annotated training dataset (i.e. a validation split of 20% would mean that 20% of the manually annotated training data is not used for training)

We do recognize the importance of using separate data for validation, or at least offering the possibility of doing so. We have now added a parameter to the settings (and made a Settings menu item available in the top menu bar) where users can specify what fraction (0, 10, 20, or 50%) of training datasets should be set aside for validation. If the chosen value is not 0%, the software reports the validation loss as well as the size of the split during training, rather than (as was done previously) the training loss. We have, however, set the default value for the validation split to 0%, for the same reason as before. We also added a section to the online documentation about using validation splits, and edited the corresponding paragraph in the methods section:

“The reported loss is that calculated on the training dataset itself, i.e., no validation split was applied. During regular use of the software, users can specify whether to use a validation split or not. By default, a validation split is not applied, in order to make full use of an input set of ground truth annotations. Depending on the chosen split size, the software reports either the overall training loss or the validation loss during training.”

(8) Related to this point: how is the training of the models in the software modelled? It might be helpful to add a paragraph to the paper in which this process is described, together with indicators of what to look out for when training a model, e.g. when should one stop training? 

We have expanded the paragraph where we write about the utility of comparing different networks architectures to also include a note on how Ais facilitates monitoring the output of a model during training:

“When taking the training and processing speeds in to account as well as the segmentation results, there is no overall best architecture. We therefore included multiple well-performing model architectures in the final library, in order to allow users to select from these models to find one that works well for their specific datasets. Although it is not necessary to screen different network architectures and users may simply opt to use the default (VGGNet), these results thus show that it can be useful to test different networks in order to identify one that is best. Moreover, these results also highlight the utility of preparing well-performing models by iteratively improving training datasets and re-training models in a streamlined interface. To aid in this process, the software displays the loss value of a network during training and allows for the application of models to datasets during training. Thus, users can inspect how a model’s output changes during training and decide whether to interrupt training and improve the training data or choose a different architecture.”

(9) Figure 1 legend: define the colours of the different segmentations. 

Done

(10) It may be better to colour Figure 2B with the same colours as Figure 2A. 

We tried this, but the effect is that the underlying density is much harder to see. We think the current grayscale image paired with the various segmentations underneath is better for visually identifying which density corresponds to membranes, carbon film, or antibody platforms.

Reviewer 2 (Public Review):

Summary: 

Last et al. present Ais, a new deep learning-based software package for the segmentation of cryo-electron tomography data sets. The distinguishing factor of this package is its orientation to the joint use of different models, rather than the implementation of a given approach. Notably, the software is supported by an online repository of segmentation models, open to contributions from the community. 

The usefulness of handling different models in one single environment is showcased with a comparative study on how different models perform on a given data set; then with an explanation of how the results of several models can be manually merged by the interactive tools inside Ais. 

The manuscripts present two applications of Ais on real data sets; one is oriented to showcase its particlepicking capacities on a study previously completed by the authors; the second one refers to a complex segmentation problem on two different data sets (representing different geometries as bacterial cilia and mitochondria in a mouse neuron), both from public databases. 

The software described in the paper is compactly documented on its website, additionally providing links to some YouTube videos (less than an hour in total) where the authors videocapture and comment on major workflows. 

In short, the manuscript describes a valuable resource for the community of tomography practitioners. 

Strengths: 

A public repository of segmentation models; easiness of working with several models and comparing/merging the results. 

Weaknesses: 

A certain lack of concretion when describing the overall features of the software that differentiate it from others. 

We thank the reviewer for their kind and constructive feedback. Following the suggestion to use the Pix2pix results to illustrate the utility of Ais for analyzing results, we have added a new supplementary figure (Fig. S6) and brief discussion, showing the use of Ais in iteratively improving segmentation results. We have also expanded the online documentation and included a note in the supplementary information about how models are saved/loaded (Supplemetary note 1) 

Recommendations:

I would like to ask the authors about some concerns about the Ais project as a whole: 

(1) The website that accompanies the paper (aiscryoet.org), albeit functional, seems to be in its first steps. Is it planned to extend it? In particular, one of the major contributions of the paper (the maintenance of an open repository of models) could use better documentation describing the expected formats to submit models. This could even be discussed in the supplementary material of the manuscript, as this feature is possibly the most distinctive one of the paper. Engaging third-party users would require giving them an easier entry point, and the superficial mention of this aspect in the online documentation could be much more generous.

We have added a new page to the online documentation, titled ‘Sharing models’ where we include an explanation of the structure of model files and demonstrate the upload page. We also added a note to the Supplementary Information that explains the file format for models, and how they are loaded/saved (i.e., that these standard keras model obects). 

To make it easier to interface Ais with other tools, we have now also made some of the core functionality available (e.g. training models, batch segmentation) via the command line interface. Information on how to use this is included in the online documentation. All file formats are common formats used in cryoET, so that using Ais in a workflow with, e.g. AreTomo -> Ais -> Relion should now be more straightforward.

(2) A different major line advanced by the authors to underpin the novelty of the software, is its claimed flexibility and modularity. In particular, the restrictions of other packages in terms of visualization and user interaction are mentioned. Although in the manuscript it is also mentioned that most of the functionalities in Ais are already available in major established packages, as a reader I am left confused about what exactly makes the offer of Ais different from others in terms of operation and interaction: is it just the two aspects developed in the manuscript (possibility of using different models and tools to operate model interaction)? If so, it should probably be stated; but if the authors want to pinpoint other aspects of the capacity of Ais to drive smoothly the interactions, they should be listed and described, instead of leaving it as an unspecific comment. As a potential user of Ais, I would suggest the authors add (maybe in the supplementary material) a listing of such features. Figure 1 does indeed carry the name "overview of (...) functionalities", but it is not clear to me which functionalities I can expect to be absent or differently solved on the other tools they mention.

We have rewritten the part of the introduction where we previously listed the features as below. We think it should now be clearer for the reader to know what features to expect, as well as how Ais can interface with other software (i.e. what the inputs and outputs are). We have also edited the caption for Figure 1 to make it explicit that panels A to C represent the annotation, model preparation, and rendering steps of the Ais workflow and that the images are screenshots from the software.

“In this report we present Ais, an open-source tool that is designed to enable any cryoET user – whether experienced with software and segmentation or a novice – to quickly and accurately segment their cryoET data in a streamlined and largely automated fashion. Ais comprises a comprehensive and accessible user interface within which all steps of segmentation can be performed, including: the annotation of tomograms and compiling datasets for the training of convolutional neural networks (CNNs), training and monitoring performance of CNNs for automated segmentation, 3D visualization of segmentations, and exporting particle coordinates or meshes for use in downstream processes. To help generate accurate segmentations, the software contains a library of various neural network architectures and implements a system of configurable interactions between different models. Overall, the software thus aims to enable a streamlined workflow where users can interactively test, improve, and employ CNNs for automated segmentation. To ensure compatibility with other popular cryoET data processing suites, Ais employs file formats that are common in the field, using .mrc files for volumes, tab-separated .txt or .star files for particle datasets, and the .obj file format for exporting 3D meshes.”

“Figure 1 – an overview of the user interface and functionalities. The various panels represent sequential stages in the Ais processing workflow, including annotation (a), testing CNNs (b), visualizing segmentation (c). These images (a-c) are unedited screenshots of the software. a) […]”

(3) Table 1 could have the names of the three last columns. The table has enough empty space in the other columns to accommodate this. 

Done.

(4) The comment about Pix2pix needing a larger number of training epochs (being a larger model than the other ones considered) is interesting. It also lends itself for the authors to illustrate the ability of their software to precisely do this: allow the users to flexibly analyze results and test hypothesis

Please see the response to Reviewer 1 comment #3. We agree that this is a useful example of the ability to iterate between annotation and training, and have added an explicit mention of this in the text:

“Moreover, since Pix2pix is a relatively large network, it might also be improved further by increasing the number of training epochs. In a second iteration of annotation and training, we thus increased the size of the antibody platform training dataset (from 58 to 170 positive samples) and generated an improved Pix2pix model for use in the following analyses.”

Reviewer 3 (Public Review):

We appreciate the reviewer’s extensive and very helpful feedback and are glad to read that they consider Ais potentially quite useful for the users. To address the reviewer’s comments, we have made various edits to the text, figures, and documentation, that we think have helped improve the clarity of our work. We list all edits below. 

Summary

In this manuscript, Last and colleagues describe Ais, an open-source software package for the semi-automated segmentation of cryo-electron tomography (cryo-ET) maps. Specifically, Ais provides a graphical user interface (GUI) for the manual segmentation and annotation of specific features of interest. These manual annotations are then used as input ground-truth data for training a convolutional neural network (CNN) model, which can then be used for automatic segmentation. Ais provides the option of several CNNs so that users can compare their performance on their structures of interest in order to determine the CNN that best suits their needs. Additionally, pre-trained models can be uploaded and shared to an online database. 

Algorithms are also provided to characterize "model interactions" which allows users to define heuristic rules on how the different segmentations interact. For instance, a membrane-adjacent protein can have rules where it must colocalize a certain distance away from a membrane segmentation. Such rules can help reduce false positives; as in the case above, false negatives predicted away from membranes are eliminated. 

The authors then show how Ais can be used for particle picking and subsequent subtomogram averaging and for the segmentation of cellular tomograms for visual analysis. For subtomogram averaging, they used a previously published dataset and compared the averages of their automated picking with the published manual picking. Analysis of cellular tomogram segmentation was primarily visual. 

Strengths:

CNN-based segmentation of cryo-ET data is a rapidly developing area of research, as it promises substantially faster results than manual segmentation as well as the possibility for higher accuracy. However, this field is still very much in the development and the overall performance of these approaches, even across different algorithms, still leaves much to be desired. In this context, I think Ais is an interesting package, as it aims to provide both new and experienced users with streamlined approaches for manual annotation, access to a number of CNNs, and methods to refine the outputs of CNN models against each other. I think this can be quite useful for users, particularly as these methods develop. 

Weaknesses: 

Whilst overall I am enthusiastic about this manuscript, I still have a number of comments: 

(1) On page 5, paragraph 1, there is a discussion on human judgement of these results. I think a more detailed discussion is required here, as from looking at the figures, I don't know that I agree with the authors' statement that Pix2pix is better. I acknowledge that this is extremely subjective, which is the problem. I think that a manual segmentation should also be shown in a figure so that the reader has a better way to gauge the performance of the automated segmentation.

Please see the answer to Reviewer 1’s comment #3.

(2) On page 7, the authors mention terms such as "emit" and "absorb" but never properly define them, such that I feel like I'm guessing at their meaning. Precise definitions of these terms should be provided. 

We have expanded this paragraph to include some more detail:

“Besides these specific interactions between two models, the software also enables pitching multiple models against one another in what we call ‘model competition’. Models can be set to ‘emit’ and/or ‘absorb’ competition from other models. Here, to emit competition means that a model’s prediction value is included in a list of competing models. To absorb competition means that a model’s prediction value will be compared to all values in that list, and that this model’s prediction value for any pixel will be set to zero if any of the competing models’ prediction value is higher. On a pixel-by-pixel basis, all models that absorb competition are thus suppressed whenever their prediction value for a pixel is lower than that of any of the emitting models.” 

(3) For Figure 3, it's unclear if the parent models shown (particularly the carbon model) are binary or not.

The figure looks to be grey values, which would imply that it's the visualization of some prediction score. If so, how is this thresholded? This can also be made clearer in the text. 

The figures show the grayscale output of the parent model, but this grayscale output is thresholded to produce a binary mask that is used in an interaction. We have edited the text to include a mention of thresholding at a user-specified threshold value:

“These interactions are implemented as follows: first, a binary mask is generated by thresholding the parent model’s predictions using a user-specified threshold value. Next, the mask is then dilated using a circular kernel with a radius 𝑅, a parameter that we call the interaction radius. Finally, the child model’s prediction values are multiplied with this mask.”

To avoid confusion, we have also edited the figure to show the binary masks rather than the grayscale segmentations. 

(4) Figure 3D was produced in ChimeraX using the hide dust function. I think some discussion on the nature of this "dust" is in order, e.g. how much is there and how large does it need to be to be considered dust? Given that these segmentations can be used for particle picking, this seems like it may be a major contributor to false positives. 

‘Dust’ in segmentations is essentially unavoidable; it would require a perfect model that does not produce any false positives. However, when models are sufficiently accurate, the volume of false positives is typically smaller than that of the structures that were intended to be segmented. In these cases, discarding particles based on size is a practical way of filtering the segmentation results. Since it is difficult to generalize when to consider something ‘dust’ we decided to include this additional text in the Method’s section rather than in the main text:

“… with the use of the ‘hide dust’ function (the same settings were used for each panel, different settings used for each feature).

This ‘dust’ corresponds to small (in comparison to the segmented structures of interest) volumes of false positive segmentations, which are present in the data due to imperfections in the used models. The rate and volume of false positives can be reduced either by improving the models (typically by including more examples of the images of what would be false negatives or positives in the training data) or, if the dust particles are indeed smaller than the structures of interest, they can simply be discarded by filtering particles based on their volume, as applied here. In particle picking a ‘minimum particle volume’ is specified – particles with a smaller volume are considered ‘dust’.

In combination with the newly included text about the method of converting volumes into lists of coordinates (see Reviewer 1’s comment #6).

“Third, a watershed transform is applied to the resulting volume, so that the sets of pixels closest to any local maximum in the distance transformed volume are assigned to one group. Fourth, groups that are smaller than a user-specified minimum volume are discarded…”

We think it should now be clearer that (some form of) discarding ‘dust’ is a step that is typically included in the particle picking process.

(5) Page 9 contains the following sentence: "After selecting these values, we then launched a batch particle picking process to determine lists of particle coordinates based on the segmented volumes." Given how important this is, I feel like this requires significant description, e.g. how are densities thresholded, how are centers determined, and what if there are overlapping segmentations? 

Please see the response to Reviewer 1’s comment #6.

(6) The FSC shown in Figure S6 for the auto-picked maps is concerning. First, a horizontal line at FSC = 0 should be added. It seems that starting at a frequency of ~0.045, the FSC of the autopicked map increases above zero and stays there. Since this is not present in the FSC of the manually picked averages, this suggests the automatic approach is also finding some sort of consistent features. This needs to be discussed. 

Thank you for pointing this out. Awkwardly, this was due to a mistake made while formatting the figure. In the two separate original plots, the Y axes had slightly different ranges, but this was missed when they were combined to prepare the joint supplementary figure. As a result, the FSC values for the autopicked half maps are displayed incorrectly. The original separate plots are shown below to illustrate the discrepancy:

Author response image 1.

The corrected figure is Figure S9 in the manuscript. The values of 44 Å and 46 Å were not determined from the graph and remain unchanged.

(7) Page 11 contains the statement "the segmented volumes found no immediately apparent false positive predictions of these pores". This is quite subjective and I don't know that I agree with this assessment. Unless the authors decide to quantify this through subtomogram classification, I don't think this statement is appropriate. 

We originally included this statement and the supplementary figure because we wanted to show another example of automated picking, this time in the more crowded environment of the cell. We do agree that it requires better substantiation, but also think that the demonstration of automated picking of the antibody platforms and IgG3-C1 complexes for subtomogram averaging suffices to demonstrate Ais’ picking capabilities. Since the supplementary information includes an example of picked coordinates rendered in the Ais 3D viewer (Figure S7) that also used the pore dataset, we still include the supplementary figure (S10) but have edited the statement to read:

“Moreover, we could identify the molecular pores within the DMV, and pick sets of particles that might be suitable for use in subtomogram averaging (see Fig. S11).”

We have also expanded the text that accompanies the supplementary figure to emphasize that results from automated picking are likely to require further curation, e.g. by classification in subtomogram averaging, and that the selection of particles is highly dependent on the thresholds used in the conversion from volumes to lists of coordinates.

(8) In the methods, the authors note that particle picking is explained in detail in the online documentation. Given that this is a key feature of this software, such an explanation should be in the manuscript. 

Please see the response to Reviewer 1’s comment #6. 

Recommendations:

(9) The word "model" seems to be used quite ambiguously. Sometimes it seems to refer to the manual segmentations, the CNN architectures, the trained models, or the output predictions. More precision in this language would greatly improve the readability of the manuscript.

This was indeed quite ambiguous, especially in the introduction. We have edited the text to be clearer on these differences. The word ‘model’ is now only used to refer to trained CNNs that segment a particular feature (as in ‘membrane model’ or ‘model interactions’). Where we used terms such as ‘3D models’ to describe scenes rendered in 3D, we now use ‘3D visualizations’ or similar terms. Where we previously used the term ‘models’ to refer to CNN architectures, we now use terms such as ‘neural network architectures’ or ‘architecture’. Some examples:

… with which one can automatically segment the same or any other dataset …

Moreover, since Pix2pix is a relatively large network, …       

… to generate a 3D visualization of ten distinct cellular …

… with the use of the same training datasets for all network architectures …

In Figure 1, the text in panels D and E is illegible. 

We have edited the figure to show the text more clearly (the previous images were unedited screenshots of the website).

(10) Prior to the section on model interactions, I was under the impression that all annotations were performed simultaneously. I think it could be clarified that models are generated per annotation type. 

Multiple different features can be annotated (i.e. drawn by hand by the user) at the same time, but each trained CNN only segments one feature. CNNs that output segmentations for multiple features can be implemented straightforwardly, but this introduces the need to provide training data where for every grayscale image, every feature is annotated. This can make preparing the training data much more cumbersome. Reusability of the models is also hampered. We now mention the separateness of the networks explicitly in the introduction:

“Multiple features, such as membranes, microtubules, ribosomes, and phosphate crystals, can be segmented and edited at the same time across multiple datasets (even hundreds). These annotations are then extracted and used as ground truth labels upon which to condition multiple separate neural networks, …”

(11) On page 6, there is the text "some features are assigned a high segmentation value by multiple of the networks, leading to ambiguity in the results". Do they mean some false features? 

To avoid ambiguity of the word ‘features’, we have edited the sentence to read:

“… some parts of the image are assigned a high segmentation value by multiple of the networks, leading to false classifications and ambiguity in the results.”

(12) Figures 2 and 3 would be easier to follow if they had consistent coloring. 

We have changed the colouring in Figure 2 to match that of Figure 3 better:

(13) For Figure 3D, I'm confused as to why the authors showed results from the tomogram in Figure 2B. It seems like the tomogram in Figure 3C would be a more obvious choice, as we would be able to see how the 2D slices look in 3D. This would also make it easier to see the effect of interactions on false negatives. Also, since the orientation of the tomogram in 2B is quite different than that shown in 3D, it's a bit difficult to relate the two.

We chose to show this dataset because it exemplifies the effects of both model competition and model interactions better than the tomogram in Figure 3C. See Figure 3D and Author response image 2 for a comparison:

Author response image 2.

(14) I'm confused as to why the tomographic data shown in Figures 4D, E, and F are black on white while all other cryo-ET data is shown as white on black. 

The images in Figure 4DEF are now inverted.

(15) For Figure 5, there needs to be better visual cueing to emphasize which tomographic slices are related to the segmentations in Panels A and B. 

We have edited the figure to show more clearly which grayscale image corresponds to which segmentation:

(16) I don't understand what I should be taking away from Figures S1 and S2. There are a lot of boxes around membrane areas and I don't know what these boxes mean. 

We have added a more descriptive text to these figures. The boxes are placed by the user to select areas of the image that will be sampled when saving training datasets.

Author response:

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

Reviewer #1 (Public review): 

The study examines how pyruvate, a key product of glycolysis that influences TCA metabolism and gluconeogenesis, impacts cellular metabolism and cell size. It primarily utilizes the Drosophila liver-like fat body, which is composed of large post-mitotic cells that are metabolically very active. The study focuses on the key observations that overexpression of the pyruvate importer MPC complex (which imports pyruvate from the cytoplasm into mitochondria) can reduce cell size in a cell-autonomous manner. They find this is by metabolic rewiring that shunts pyruvate away from TCA metabolism and into gluconeogenesis. Surprisingly, mTORC and Myc pathways are also hyper-active in this background, despite the decreased cell size, suggesting a non-canonical cell size regulation signaling pathway. They also show a similar cell size reduction in HepG2 organoids. Metabolic analysis reveals that enhanced gluconeogenesis suppresses protein synthesis. Their working model is that elevated pyruvate mitochondrial import drives oxaloacetate production and fuels gluconeogenesis during late larval development, thus reducing amino acid production and thus reducing protein synthesis. 

Strengths: 

The study is significant because stem cells and many cancers exhibit metabolic rewiring of pyruvate metabolism. It provides new insights into how the fate of pyruvate can be tuned to influence Drosophila biomass accrual, and how pyruvate pools can influence the balance between carbohydrate and protein biosynthesis. Strengths include its rigorous dissection of metabolic rewiring and use of Drosophila and mammalian cell systems to dissect carbohydrate:protein crosstalk. 

Weaknesses: 

However, questions on how these two pathways crosstalk, and how this interfaces with canonical Myc and mTORC machinery remain. There are also questions related to how this protein:carbohydrate crosstalk interfaces with lipid biosynthesis. Addressing these will increase the overall impact of the study. 

We thank the reviewer for recognizing the significance of our work and for providing constructive feedback. Our findings indicate that elevated pyruvate transport into mitochondria acts independently of canonical pathways, such as mTORC1 or Myc signaling, to regulate cell size. To investigate these pathways, we utilized immunofluorescence with well-validated surrogate measures (p-S6 and p-4EBP1) in clonal analyses of MPC expression, as well as RNAseq analyses in whole fat body tissues expressing MPC. These methods revealed surprising hyperactivation of mTORC1 and Myc signaling in Drosophila fat body cells expressing MPC, which are dramatically smaller than control cells. One explanation of these seemingly contradictory observations could be an excess of nutrients that activate mTORC1 or Myc pathways. However, our data is inconsistent with a nutrient surplus that could explain this hyperactivation. Instead, we observed reduced amino acid abundance upon MPC expression, which is very surprising given the observed hyperactivation of mTORC1. This led us to hypothesize the existence of a feedback mechanism that senses an inappropriate reduction in cell size and activates signaling pathways to promote cell growth. The best-characterized “sizer” pathway for mammalian cells is the Cyclin D/CDK4 complex, which has been well studied in the context of cell size regulation of the cell cycle (PMID 10970848, 34022133). However, the mechanisms that sense cell size in post-mitotic cells, such as fat body cells and hepatocytes, remain poorly understood. Investigating the hypothesized size-sensing mechanisms at play here is a fascinating direction for future research.

For the current study, we conducted epistatic analyses with mTORC1 pathway members by overexpressing PI3K and knocking down the TORC1 inhibitor Tuberous Sclerosis Complex 1 (Tsc1). These manipulations increased the size of control fat body cells but not those overexpressing the MPC (Supplementary Fig. 3c, 3d). Regarding Myc, its overexpression increased the size of both control and MPC+ clones (Supplementary Fig. 3e), but Myc knockdown had no additional effect on cell size in MPC+ clones (Supplementary Fig. 3f). These results suggest that neither mTORC1, PI3K, nor Myc is epistatic to the cell size effects of MPC expression. Consequently, we shifted our focus to metabolic mechanisms regulating biomass production and cell size.

When analyzing cellular biomolecules contributing to biomass, we observed a significant impact on protein levels in Drosophila fat body cells and mammalian MPC-expressing HepG2 spheroids. Triglyceride abundance in MPC-expressing HepG2 spheroids and whole fat body cells showed a statistically insignificant decrease compared to controls. Furthermore, lipid droplets in fat body cells were comparable in MPC-expressing clones when normalized to cell size.

Interestingly, RNA-seq analysis revealed modestly increased expression of fatty acid and cholesterol biosynthesis pathways in MPC-expressing fat body cells. Upregulated genes included major SREBP targets, such as ATPCL (2.08-fold), FASN1 (1.15-fold), FASN2 (1.07-fold), and ACC (1.26-fold). Since mTORC1 promotes SREBP activation and MPC-expressing cells showed elevated mTOR activity and upregulation of SREBP targets, we hypothesize that SREBP is modestly activated in these cells. Nonetheless, our data on amino acid abundance and its impact on protein synthesis activity suggest that protein abundance is likely to play a prominent causal role in regulating cell size in response to increased pyruvate transport into mitochondria.

Reviewer #2 (Public review): 

In this manuscript, the authors leverage multiple cellular models including the drosophila fat body and cultured hepatocytes to investigate the metabolic programs governing cell size. By profiling gene programs in the larval fat body during the third instar stage - in which cells cease proliferation and initiate a period of cell growth - the authors uncover a coordinated downregulation of genes involved in mitochondrial pyruvate import and metabolism. Enforced expression of the mitochondrial pyruvate carrier restrains cell size, despite active signaling of mTORC1 and other pathways viewed as traditional determinants of cell size. Mechanistically, the authors find that mitochondrial pyruvate import restrains cell size by fueling gluconeogenesis through the combined action of pyruvate carboxylase and phosphoenolpyruvate carboxykinase. Pyruvate conversion to oxaloacetate and use as a gluconeogenic substrate restrains cell growth by siphoning oxaloacetate away from aspartate and other amino acid biosynthesis, revealing a tradeoff between gluconeogenesis and provision of amino acids required to sustain protein biosynthesis. Overall, this manuscript is extremely rigorous, with each point interrogated through a variety of genetic and pharmacologic assays. The major conceptual advance is uncovering the regulation of cell size as a consequence of compartmentalized metabolism, which is dominant even over traditional signaling inputs. The work has implications for understanding cell size control in cell types that engage in gluconeogenesis but more broadly raise the possibility that metabolic tradeoffs determine cell size control in a variety of contexts. 

We thank the reviewer for their thoughtful recognition of our efforts, and we are honored by the enthusiasm the reviewer expressed for the findings and the significance of our research. We share the reviewer’s opinion that our work might help to unravel metabolic mechanisms that regulate biomass gain independent of the well-known signaling pathways.

Reviewer #3 (Public review): 

Summary: 

In this article, Toshniwal et al. investigate the role of pyruvate metabolism in controlling cell growth. They find that elevated expression of the mitochondrial pyruvate carrier (MPC) leads to decreased cell size in the Drosophila fat body, a transformed human hepatocyte cell line (HepG2), and primary rat hepatocytes. Using genetic approaches and metabolic assays, the authors find that elevated pyruvate import into cells with forced expression of MPC increases the cellular NADH/NAD+ ratio, which drives the production of oxaloacetate via pyruvate carboxylase. Genetic, pharmacological, and metabolic approaches suggest that oxaloacetate is used to support gluconeogenesis rather than amino acid synthesis in cells over-expressing MPC. The reduction in cellular amino acids impairs protein synthesis, leading to impaired cell growth. 

Strengths: 

This study shows that the metabolic program of a cell, and especially its NADH/NAD+ ratio, can play a dominant role in regulating cell growth.

The combination of complementary approaches, ranging from Drosophila genetics to metabolic flux measurements in mammalian cells, strengthens the findings of the paper and shows a conservation of MPC effects across evolution.

Weaknesses: 

In general, the strengths of this paper outweigh its weaknesses. However, some areas of inconsistency and rigor deserve further attention. 

Thank you for reviewing our manuscript and offering constructive feedback. We appreciate your recognition of the significance of our work and your acknowledgment of the compelling evidence we have presented. We have carefully revised the manuscript in line with the reviewers' recommendations.

The authors comment that MPC overrides hormonal controls on gluconeogenesis and cell size (Discussion, paragraph 3). Such a claim cannot be made for mammalian experiments that are conducted with immortalized cell lines or primary hepatocytes. 

We appreciate the reviewer’s insightful comment. Pyruvate is a primary substrate for gluconeogenesis, and our findings suggest that increased pyruvate transport into mitochondria increases the NADH-to-NAD+ ratio, and thereby elevates gluconeogenesis. Notably, we did not observe any changes in the expression of key glucagon targets, such as PC, PEPCK2, and G6PC, suggesting that the glucagon response is not activated upon MPC expression. By the statement referenced by the reviewer, we intended to highlight that excess pyruvate import into mitochondria drives gluconeogenesis independently of hormonal and physiological regulation. 

It seems the reviewer might also have been expressing the sentiment that our in vitro models may not fully reflect the in vivo situation, and we completely agree.  Moving forward, we plan to perform similar analyses in mammalian models to test the in vivo relevance of this mechanism. For now, we will refine the language in the manuscript to clarify this point.

Nuclear size looks to be decreased in fat body cells with elevated MPC levels, consistent with reduced endoreplication, a process that drives growth in these cells. However, acute, ex vivo EdU labeling and measures of tissue DNA content are equivalent in wild-type and MPC+ fat body cells. This is surprising - how do the authors interpret these apparently contradictory phenotypes? 

We thank the reviewer for raising this important issue. The size of the nucleus is regulated by DNA content and various factors, including the physical properties of DNA, chromatin condensation, the nuclear lamina, and other structural components (PMID 32997613). Additionally, cytoplasmic and cellular volume also impact nuclear size, as extensively documented during development (PMID 17998401, PMID 32473090).

In MPC-expressing cells, it is plausible that the reduced cellular volume impacts chromatin condensation or the nuclear lamina in a way that slightly decreases nuclear size without altering DNA content. Specifically, in our whole-fat body experiments using CG-Gal4 (as shown in Supplementary Figure 2a-c), we noted that after 12 hours of MPC expression, cell size was significantly reduced (Supplementary Figure 2c and Author Response Figure 1A). However, the reduction in nuclear size is modestly different at 24 hours and significantly different at 36 hours (Author Response Figure 1B), suggesting that the reduction in cell size is a more acute response to MPC expression, followed only later by effects on nuclear size.

In clonal analyses, this relationship was further clarified. MPC-expressing cells with a size greater than 1000 µm² displayed nuclear sizes comparable to control cells, whereas those with a drastic reduction in cell size (less than 1000 µm²) exhibited smaller nuclei (Author Response Figure 1C and 1D). These observations collectively suggest that changes in nuclear size are more likely to be downstream rather than upstream of cell size reduction. Given that DNA content remains unaffected, we focused on investigating the rate of protein synthesis. Our findings suggest that protein synthesis might play a causal role in regulating cell size, thereby reinforcing the connection between cellular and nuclear size in this context.

Author response image 1.<br />

Cell Size vs. Nuclear Size in MPC-Expressing Fat Body Cells A. Cell size comparison between control (blue, ay-GFP) and MPC+ (red, ay-MPC) fat body cells over time, measured in hours after MPC expression induction. B. Nuclear area measurements from the same fat body cells in ay-GFP and ay-MPC groups. C. Scatter plot of nuclear area vs. cell area for control (ay-GFP) cells, including the corresponding R² value. D. Scatter plot of nuclear area vs. cell area for MPC-expressing (ay-MPC) cells, with the respective R² value.

This figure highlights the relationship between nuclear and cell size in MPC-expressing fat body cells, emphasizing the distinct cellular responses observed following MPC induction.

In Figure 4d, oxygen consumption rates are measured in control cells and those overexpressing MPC. Values are normalized to protein levels, but protein is reduced in MPC+ cells. Is oxygen consumption changed by MPC expression on a per-cell basis? 

As described in the manuscript, MPC-expressing cells are smaller in size. In this context, we felt that it was most appropriate to normalize oxygen consumption rates (OCR) to cellular mass to enable an accurate interpretation of metabolic activity. Therefore, we normalized OCR with protein content to account for variations in cellular size and (probably) mitochondrial mass. 

Trehalose is the main circulating sugar in Drosophila and should be measured in addition to hemolymph glucose. Additionally, the units in Figure 4h should be related to hemolymph volume - it is not clear that they are. 

We appreciate this valuable suggestion. In the revised manuscript, we have quantified trehalose abundance in circulation and within fat bodies. As described in the Methods section and following the approach outlined in Ugrankar-Banerjee et al. (2023, we bled 10 larvae (either control or MPC-expressing) using forceps onto parafilm. From this, 2 microliters of hemolymph were collected for glucose measurement. The hemolymph was treated with trehalase overnight, and the resulting glucose derived from trehalose was measured. We have observed that trehalose levels were also elevated in hemolymph of fat body-specific MPC-expressing larvae, further supporting our conclusion that MPC expression in fat body induces a hyperglycemic state. These data are now included in Figure 4h of the revised manuscript, and the details are further mentioned in the revised materials and methods.  

Measurements of NADH/NAD ratios in conditions where these are manipulated genetically and pharmacologically (Figure 5) would strengthen the findings of the paper. Along the same lines, expression of manipulated genes - whether by RT-qPCR or Western blotting - would be helpful to assess the degree of knockdown/knockout in a cell population (for example, Got2 manipulations in Figures 6 and S8). 

We appreciate this suggestion, which will provide additional rigor to our study. We have already quantified NADH/NAD+ ratios in HepG2 cells under UK5099, NMN, and Asp supplementation, as presented in Figure 6k. As suggested, we have quantified the expression of Got2 manipulations mentioned in Figure 6j using RT-qPCR, this data is presented in revised Supplementary Figure 8f-h. In addition, Supplementary Figure 8i has been updated with western blot analysis of Got2 expression in knock-out cells used to perform the size analysis in HepG2 cells.

Additionally, we have also analysed the efficiency of pcb (Supplementary Figure 6a-c), pdha (Supplementary Figure 6f-h), dlat (Supplementary Figure 6f, g and i), pepck2 (Supplementary Figure 6n-p), fbp  (Supplementary Figure 6n, m, q)  manipulations used to modulate the expression of these genes. These validations will ensure the robustness of our findings and strengthen the conclusions of our study.

Reviewer #1 (Recommendations for the authors): 

General questions: 

(1) MPC over-expression in HepG2 cells altered the redox balance and the NADH/NAD+ ratio. This is suggested to help drive the metabolic rewiring from protein to carbohydrate biosynthesis. In line with this overexpression of Nmnat (which makes NAD+) or NDX rescues cell size and elevates protein biosynthesis. However, mechanistically it is unclear exactly how these redox NAD+ changes directly impact protein biosynthesis. Some additional explanations will strengthen this portion of the study. 

Our data indicate that the altered redox state of the cell, particularly elevated NADH levels, affects the rate of protein synthesis. A similar relationship between redox balance and protein synthesis has been observed during embryonic development (PMID: 39879975), although the underlying mechanism remains uncharacterized. Our study suggests that increased NADH levels reprogram cellular carbohydrate metabolism, shifting it from glycolysis toward gluconeogenesis. This metabolic shift necessitates the use of oxaloacetate by PEPCK2, instead of its diversion toward GTP-mediated aspartate synthesis. Aspartate, which can be anaplerotically converted into glutamate and proline, plays a critical role in protein biosynthesis. Thus, the conversion of oxaloacetate to phosphoenolpyruvate represents a key metabolic node influencing protein synthesis under altered redox conditions. Additionally, since aspartate serves as a precursor for NAD biosynthesis, this may suggest a feedforward loop reinforcing the metabolic rewiring. Nonetheless, the precise relationship between NADH concentration and redox status and the regulation of protein synthesis warrants further investigation in future studies.

(2) In the MPC1/2 (MPC+) over-expression background, can blocking of gluconeogenesis downstream in the carbohydrate synthesis pathway rescue the phenotype? 

We knocked down FBPase (Drosophila fbp) using an RNAi construct, achieving approximately 60% reduction in FBPase expression in Drosophila. Notably, FBPase knockdown in fat body cells overexpressing MPC rescued the reduced cell size phenotype. These findings are presented in Figure 4o and Supplementary Figures 6n–q.

(3) Biomass accrual and cell size are also influenced by lipogenesis. The study suggests mTORC and Myc are uncoupled to cell size determination per se, but how lipogenesis regulatory pathways like SREBP are impacted by MPC overexpression is not really explored. How lipid membrane synthesis inter-relates to this protein/carbohydrate crosstalk would add to the understanding of the system. 

As mentioned above - When analyzing cellular biomolecules contributing to biomass, we observed a significant impact on protein levels in Drosophila fat body cells and mammalian MPC-expressing HepG2 spheroids. Triglyceride abundance in MPC-expressing HepG2 spheroids and whole fat body cells showed a statistically insignificant decrease compared to controls. Furthermore, lipid droplets in fat body cells were comparable in MPC-expressing clones when normalized to cell size.

Interestingly, RNA-seq analysis revealed increased expression of fatty acid and cholesterol biosynthesis pathways in MPC-expressing fat body cells. Upregulated genes included major SREBP targets, such as ATPCL (2.08-fold), FASN1 (1.15-fold), FASN2 (1.07-fold), and ACC (1.26-fold). Since mTOR promotes SREBP activation and MPC-expressing cells showed elevated mTOR activity and upregulation of SREBP targets, we hypothesize that SREBP is modestly activated in these cells. Nonetheless, our data on amino acid abundance and its impact on protein synthesis activity suggest that protein abundance, rather than lipids, is likely to play a larger causal role in regulating cell size in response to increased pyruvate transport into mitochondria.

Reviewer #2 (Recommendations for the authors): 

I have only minor suggestions for the authors to consider. 

Minor points 

(1) Wherever possible, scale bars should be labeled with units or indicated comparisons (e.x. Supplementary Fig. 1). To make the data as accessible as possible, it would be helpful for the authors to include the data presented in Supplementary Figure 1 as an associated table as well. 

We have corrected this in the revised manuscript and included the table. 

(2) To support the conclusions about TCA cycle flux (lines 280-284), it will be helpful for the authors to consider relative metabolite pool sizes (which they should have on hand) in addition to labeling rate and fraction. 

We thank the reviewer for this suggestion. We have included the metabolite counts with fractional abundance changes side by side in Supplementary Figure 5. 

(3) believe (?) there is a typo in lines 326-328; PEPCK KO increases (not decreases) the size of spheroids/cells. 

We thank the reviewer for pointing out this error. We have corrected this in the revised manuscript.

(4) Supplementary Figure 7b: PHD has 3 phospho sites that have independent regulation; the specific phosphosite queried should be listed on the figure and unless all 3 sites are probed the claims about lack of change in phosphorylation (line 337) should be removed. 

We thank the reviewer for bringing this to our attention. We have included this in the revised manuscript.

(5) (Optional) I appreciate the effort the authors undertook to acquire cytoplasmic and mitochondrial ratios of NADH/NAD. While I recognize that many labs perform this assay, it is difficult for this reviewer to envision how accurately these values reflects the ratios present in the intact cell given how quickly these redox couples interconvert and significant post-harvest metabolic flux (see for ex PMID: 31767181), even with the extremely rapid fractionation protocol described in the methods. The present data certainly support the notion that MPC+ cells are more reduced, but these ratios may reflect a capacity for reductive metabolism rather than a bona fide NADH/NAD ratio; for example, Figure 7f shows almost identical NADH/NAD ratios in the cytoplasm and mitochondria, even though these compartments are frequently considered to have (sometimes vastly) different redox states. If the authors are willing, I would support them by including a brief discussion of the caveat of this method for new readers in the field. 

We agree with this important note from the reviewers. This is an important caveat of the technique that we used for these analyses. We have included a description of this caveat in the manuscript (Revised Manuscripts lines 393 to 395).

Reviewer #3 (Recommendations for the authors): 

Minor points: 

(1) Line 327 - "smaller" should be "bigger". 

We thank the reviewer for pointing out this error. We have corrected this in the revised manuscript.

(2) For Figure 7 - references to panels e and f in the text, and descriptions of e and f in the Figure Legend are switched with regard to the Figure itself. 

We thank the reviewer for pointing out this error. We have corrected this in the revised manuscript.

(3) Line 449 - "reduced" is missing its R 

We thank the reviewer for pointing out this error. We have corrected this in the revised manuscript.

(4) Some additional, careful proofreading is needed - several other punctuation errors were found. 

We thank the reviewer for pointing out these errors. 

We thank the reviewer for bringing this to our attention. We have conducted very careful proofreading and corrected errors.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Jin, Briggs, and colleagues use light sheet imaging to reconstruct the islet threedimensional Ca2+ network. The authors find that early/late responding (leader) cells are dynamic over time, and located at the islet periphery. By contrast, highly connected or hub cells are stable and located toward the islet center. Suggesting that the two subpopulations are differentially regulated by fuel input, glucokinase activation only influences leader cell phenotype, whereas hubs remain stable.

Strengths:

The studies are novel in providing the first three-dimensional snapshot of the beta cell functional network, as well as determining the localization of some of the different subpopulations identified to date. The studies also provide some consensus as to the origin, stability, and role of such subpopulations in islet function.

We thank the reviewers for their positive assessment.

Weaknesses:

Experiments with metabolic enzyme activators do not take into account the influence of cell viability on the observed Ca2+ network data. Limitations of the imaging approach used need to be recognized and evaluated/discussed.

We worked very hard to make sure the islets remained stable and healthy over the duration of imaging time course. We imaged the islet in 3D and observed that all betacells displayed glucose-dependent oscillations, which can only arise from functioning cells. From the raw calcium traces (displayed in the figures) we observed no detectable loss of signal over 60 min of continuous imaging regardless of drug treatment; this is because the laser excitation is below the bleach threshold for GCaMP6s, and it is bleaching that generates phototoxicity. To demonstrate this clearly, we performed a bleach test using 6x laser power; in this case calcium amplitude dropped 30% over a 60 min of imaging, however islet calcium oscillatory behavior was preserved. Light-sheet is well documented to be 1000x more gentle than other optical sectioning techniques, which is why it was chosen for this application.

Regarding the limitations of imaging approach, we recognized studying islets ex vivo is necessarily performed in the absence of native surrounding tissue, as highlighted in the discussion.

Reviewer #2 (Public Review):

The manuscript by Erli Jin, Jennifer Briggs et al. utilizes light sheet microscopy to image islet beta cell calcium oscillations in 3D and determine where beta cell populations are located that begin and coordinate glucose-stimulated calcium oscillations. The light sheet technique allowed clear 3D mapping of beta cell calcium responses to glucose, glucokinase activation, and pyruvate kinase activation. The manuscript finds that synchronized beta-cells are found at the islet center, that leader beta cells showing the first calcium responses are located on the islet periphery, that glucokinase activation helped maintain beta cells that lead calcium responses, and that pyruvate kinase activation primarily increases islet calcium oscillation frequency. The study is well-designed, contains a significant amount of high-quality data, and the conclusions are largely supported by the results.

It has recently been shown that beta cells within islets containing intact vasculature (such as those in a pancreatic slice) show different calcium responses compared to isolated islets (such as that shown in PMID: 35559734). It would be important to include some discussion about the potential in vitro artifacts in calcium that arise following islet isolation (this could be included in the discussion about the limitations of the study).

Although isolated islets reproduce the slow oscillatory calcium behavior observed in vivo, we agree that missing elements such as blood flow, cholinergic innervation, and surrounding tissues may each impact islet calcium responses. Pancreatic regional blood flow also links the endocrine and exocrine signaling which can directly influence the behavior of beta cells. We have highlighted some of these issues in the discussion “In addition to α-cells, vasculature may also impact islet Ca2+ responses, and may induce additional heterogeneity in vivo.” (see line 375, Ref. 46).

Reviewer #3 (Public Review):

Summary:

Jin, Briggs et al. made use of light-sheet 3D imaging and data analysis to assess the collective network activity in isolated mouse islets. The major advantage of using whole islet imaging, despite compromising on the speed of acquisition, is that it provides a complete description of the network, while 2D networks are only an approximation of the islet network. In static-incubation conditions, excluding the effects of perfusion, they assessed two subpopulations of beta cells and their spatial consistency and metabolic dependence.

Strengths:

The authors confirmed that coordinated Ca2+ oscillations are important for glycemic control. In addition, they definitively disproved the role of individual privileged cells, which were suggested to lead or coordinate Ca²⁺ oscillations. They provided evidence for differential regional stability, confirming the previously described stochastic nature of the beta cells that act as strongly connected hubs as well as beta cells in initiating regions (doi.org/10.1103/PhysRevLett.127.168101).

The fact that islet cores contain beta cells that are more active and more coordinated has also been readily observed in high-frequency 2D recordings (e.g. DOI: 10.2337/db22-0952), suggesting that the high-speed capture of fast activity can partially compensate for incomplete topological information.

They also found an increased metabolic sensitivity of mantle regions of an islet with a subpopulation of beta cells with a high probability of leading the islet activity which can be entrained by fuel input. They discuss a potential role of alpha/delta cell interaction, however relative lack of beta cells in the islet border region could also be a factor contributing to less connectivity and higher excitability.

The Methods section contains a useful series of direct instructions on how to approach fast 3D imaging with currently available hardware and software.

The Discussion is clear and includes most of the issues regarding the interpretation of the presented results.

Some issues concerning inconsistencies between data presented and statements made as well as statistical analysis need to be addressed.

Taken together it is a strong technical paper to demonstrate the stochasticity regarding the functions subpopulations of beta cells in the islets may have and how less well-resolved approaches (both missing spatial resolution as well as missing temporal resolution) led us to jump to unjustified conclusions regarding the fixed roles of individual beta cells within an islet.

We thank the reviewers for the comments on the many strengths of the manuscript and address the specific critiques below.

Recommendations for the authors:

Reviewing Editor Comments:

Essential revisions:

(1) How useful is GK activation as a subpopulation-level perturbation, given that all beta cells would be affected? Previous studies by the authors have shown that GK gradients likely dictate subpopulation behaviour, so the concern here is that GK activation across all cells might mask the influence of such gradients i.e. a U-shaped effect. Also, does the GK activator differentially penetrate the islet such that first responders/leaders are more vulnerable than hubs?

As we previously published, non-saturating concentrations of GK activator (as used here) have the same effect on calcium oscillations as raising glucose (PMID:33147484). In other words, the activator boosts the activity of the endogenous GK. To the second point, recent ex vivo islet studies (PMID: 28380380) document the islet penetration of a fluorescent glucose analogue within seconds even under static conditions, and in our study the islets calcium oscillations reached steady state, so we are not concerned about drug penetration. The real limitation with any drug study in the islet is that non-beta cells are also activated; this limitation is included in the discussion along with the recommendation that genetic tools are needed to assess the effect of GK activation in the various endocrine subpopulations. 

An additional concern with the GK activation experiment is that GK activation might push beta cells into a more stressed state such that they are more susceptible to phototoxicity. Although the authors state that photobleaching is low, they provide no data to support such a statement. Given the long duration of imaging and acquisition rate, phototoxicity might be more of an issue, especially with GK activation. Some further analysis (e.g. apoptosis) would be useful here to exclude an effect of beta cell viability versus GK activation on the observed phenotype of the different subpopulations.

Acute GK activation (for 30min) does not stress the islet; the drug has the same effect as raising glucose (PMID: 33147484). To determine whether photobleaching was impacted by GK activation, we examined the peak of consecutive oscillations in response to vehicle and GK activator. The average photobleaching was less than 2% of the calcium fluorescence over 30min of continuous imaging. Furthermore, GKa activation did not significantly increase photobleaching (see Author response image 1). 

Author response image 1.

To the reviewer’s second point, apoptosis cannot occur on the timescale of the drug treatment (30min), and raw calcium traces are included showing that all beta cells display oscillatory behavior throughout the course of the experiment.

(2) The authors show that glucokinase activation increases the duration of islet calcium oscillations and in some islets (3 of 15 islets) causes "a Ca2+ plateau." The authors indicate that "Glucokinase, as the 'glucose sensor' for the β-cell, controls the input of glucose carbons into glycolysis, and opens KATP channels." It would be nice to have some experimental evidence that the change in oscillation rate caused by the glucokinase activator is due to KATP activation. This could be accomplished by treating islets with subthreshold KATP activators (e.g., diazoxide) or subthreshold KATP inhibitors (e.g., tolbutamide).

The statement that glucokinase activation opens KATP channels was a typo; glucose metabolism closes KATP channels by raising the ATP/ADP ratio. We now include additional citations that document the relationship between GK and KATP and the oscillatory behavior. See Ref 22 (PMID: 33147484) and Ref 34 (PMID: 33147484).

The manuscript finds that "Early phase cells were maintained to a greater degree upon GKa application." Yet GKa is proposed to activate KATP. Some discussion about how the early phase is maintained in cell populations by GKa activation in the context of KATP activity would be useful.

As discussed above, we meant to say that GKa will close KATP and apologize for the confusion. As we mentioned in the discussion, early phase cells are most likely maintained to a great degree following GK activation as result of enhanced GK gradient and reduced effect of stochastic alpha cell input. 

(3) Membrane potential depolarization precedes calcium channel activation and subsequent calcium entry. In many cases, electrical coupling across beta cells happens on millisecond timescale. It would be good to confirm that the calcium is showing the same time scale in terms of elevation following beta cell membrane potential depolarization. One concern is that the islet beta cells could be depolarizing at the same speed and lagging in terms of calcium channel activation and calcium entry.

We thank the reviewer for making this point, which is almost certainly true, particularly since plasma membrane calcium influx is not the sole source of intracellular calcium. Previously published “simultaneous” recordings of Vm and calcium show their same phase relationship but do not have sufficient time resolution to capture depolarization of each cell. A quantification of phase lag would require the field to generate mice with voltage sensors expressed in beta cells; these tools are not yet available.  

A related issue: in the text, the authors discuss changes in membrane potential (not been measured in this study), while in the figures they exclusively describe Ca2+ oscillations (which were measured). Examples are on lines 149, 150, 153, 154, 263. It is recommended that the silent and active phases in the Results section describe processes actually measured in this study as shown in 6A.

To clarify, we did not use the term ‘membrane potential’ anywhere in the manuscript. We do sometimes refer to calcium influx as a proxy for membrane depolarization; we think this is valid given the abundant evidence that these processes are interdependent in beta cells.

(4) It would be good to include the timing of the phases of calcium entry. When was the beta cell calcium entry monitored for the response time? Were the response times between the late and early phases consistent for each oscillation? It looks as if the start of the calcium upstroke was similar for many beta cells (such as for the Figure 2I traces). It would be nice to include a shorter time duration graph of calcium oscillation traces right when the upstroke starts. This would allow the community to observe the differences in the start time of calcium entry. 

We agree this is an important point. We now include an inset showing the expanded time scale of the calcium upstroke in Fig.2I. The response time spread between early and late phase cells is now shown in Fig.7F (and in Author response image 2). We also quantified the coefficient of variation in the response time spread (0 = no variation and 1 = maximal variation) and found no significant differences between metabolic activators (Author response image 2). 

Author response image 2.

Also, for most of the GCaMP6s traces shown, the authors indicate that they are plotted as F/F0. However, this normalization (F/F0) is not done for the actual traces shown. For example, Figure 2D shows the traces starting from what looks to be 0 to 0.3 F/F0, but the traces for an F/F0 group should all start at 1. Please change this for all representative oscillations so the start of calcium entry for example traces all line up.

This has been corrected in Fig. 2D, I and Fig. 3B. Also Fig.6 should be F not F/F0

Reviewer #1 (Recommendations for the authors):

(1) Line 53: "Silencing the electrical activity of these hub cells with optogenetics was found to abolish the coordination within that plane of the islet". The authors should acknowledge that studies also showed that beta cell transcription factor (Pdx1/Mafa) dosage was important for hub cell phenotype and islet function.

Thank you, this reference to Nasteska et al. (PMID: 33514698, Ref. 16) has been added to the discussion.

(2) Light sheet imaging is used to image the 3D islet volume. Whilst speed is undoubtedly an advantage of this technique, axial resolution is ~1.1 µm over 4 µm z-step size. How confident are the authors that single nuclei can be reliably identified given their ~6 µm size in a beta cell (e.g. do some elongated nuclear appear, which could be "doublets")?

The axial resolution of 1.1 µm exceeds the resolution needed for the Nyquist criterion (i.e. sampling every 2-3 µm). As a practical matter, it is not possible to doublecount nuclei because the software will exclude nuclei that occupy the same volume. Only a very elongated nucleus (>10 µm) would be double counted and this does not occur.

(3) The authors discuss the advantages of the light sheet imaging approach used, including speed and phototoxicity. Some more balance is needed here since other approaches such as two-photon excitation achieve similar speeds with much better axial resolution (see dozens of neural circuit studies).

We are careful to point out that two-photon excitation has better axial resolution, better tissue penetration, and often higher speeds (kHz using linescans) – however these neuronal studies are limited to the cells in a few planes and the laser power is orders of magnitude higher than lightsheet. For this reason, two photon imaging has not been used to image islet calcium in three dimensions. The bottom line is lightsheet trades axial resolution for gentle volumetric imaging. 

(4) Line 340: "Laser ablation or optogenetic inactivation of these early phase cells would be predicted to have little impact on islet function, as suggested previously by electrophysiological studies in which surface β-cells have been voltage-clamped with no impact on β-cell oscillations". This statement is slightly ambiguous since the authors showed in their previous studies that laser ablation of first responder cells/leaders was able to influence the Ca2+ network. Do the authors mean that laser ablation would only temporarily influence islet function before another cell picked up the role of a first responder/leader? As written, the sentence seems to imply that first responders/leaders are unimportant for the islet function.

We intended to imply that the oscillatory system is sufficiently robust that a new cell take over when leader cells are ablated. We also cite Korosak et al. (PMID:34723613, Ref. 40) and Dwulet et al. (PMID: 33939712, Ref. 15) to make this point, although to clarify we are not examining first responders in this study.

(5) Line 369: "In contrast with leader cells, we found that the highly synchronized cells are both spatially and temporally stable." The sentence needs qualifying- what would spatiotemporal stability be expected to confer on such a subpopulation?

We believe that the spatiotemporal stability of highly synchronized cells is a consequence of beta cells in the center of the islet lacking the stochastic input of nearby alpha cells; we raise this point in the discussion: “The preponderance of α-cells on the periphery of mouse islets, which influence β-cell oscillation frequency, would be expected to disrupt β-cell synchronization on the periphery and stabilize it in the islet center – which is precisely the pattern of network activity we observed.” (see line 372). 

(6) Line 370: "However, in conflict with the description of hub cells as intermingled with other cells throughout the islet, the location of such cells in 3D space is close to the center." The study by Johnston et al did not have the axial resolution to exclude that some cells might have been grouped together.

We agree and have included the reviewer’s comment in the text (See line 384); that’s an important reason for conducting this 3D study.  

(7) Line 380: "One explanation may be that paracrine communication within the islet determines which region of cells will show high or low degree. For example, more peripheral cells that are in contact with nearby δ-cells may show some suppression in their Ca2+ dynamics, and thus reduced synchronization." A potentially exciting future study. Should however probably cite DOI s41467-022-31373-6 here.

We thank the reviewer for their input. This reference to Ren et al. (PMID:35764654) was previously included as Ref. 42 (now Ref. 45)

Reviewer #3 (Recommendations for the authors):

(1) There are in fact no radially oriented networks in the core of an islet (l. 130, Figure 4) apart from the fact that every hub has somewhat radially oriented edges. For radiality to have some general meaning, the normalized distance from the geometric center would need to be lower than 0.4. The networks are centrally located, which does not change the major conclusions of the study.

Thank you for pointing out this imprecise language. We did not intend to imply that the functional network is orientated radially. We corrected the text (see line 131, 145) to indicate that the cells with high and low synchronization are distributed in a radial pattern. 

(2) The study would benefit from acknowledging that Ca2+ influx is not a sole mechanism to drive insulin secretion and that KATP channels are not the sole target sensitive to changes in the cytosolic (global or local) ADP and ATP concentration or that there is an absolute concentration-dependence of these ligands on KATP channels. The relatively small conductance changes that have been found to be associated with active and silent phases (closing and opening of the KATP channels as interpreted by the authors, respectively, doi: 10.1152/ajpendo.00046.2013) and should be due to metabolic factors, could be also associated to desensitization of KATP channels to ATP due to the increase in cytosolic Ca2+ changes after intracellular Ca2+ flux (DOI: 10.1210/endo.143.2.8625) as they have been found to operate also at time scales, significantly faster (DOI: 10.2337/db22-0952) than reported before (refs. 21,22). Metabolic changes influence intracellular Ca2+ flux as well.

The reviewer is absolutely correct that there are amplifying factors and other sources of calcium beyond plasma membrane influx and there are other mechanisms that regulate insulin secretion beyond calcium levels. These alternative mechanisms are introduced in Refs. 1-2, however they are not the focus of this study. 

(3) There is no explanation for why KL divergence is so different between the pre-test regional consistency of the islets used to test the vehicle compared to those where GKa and PKa have been tested.

We thank the reviewer for their careful observation. This arises because there are larger differences between preparations than within a preparation. This has been described previously (PMID: 16306370 and 20037650) and could be expected to account for the differences in KL divergence between animals. 

(4) Statistical analysis would profit from testing the normality of the data distribution before choosing the statistical test and then learning the difference between parametric and nonparametric tests. For example, in Figures 3CD and 5EF, the data density is lower at the calculated mean than below and above this value and there are other examples in other figures too.

We thank the reviewer for this very important comment, and we apologize for the oversight on our part. To address this comment, we conducted two normality tests: Anderson-Darling and Kolmogorov-Smirnov on all statistical analyses in the manuscript. If the data were not normally distributed, we changed the analysis to Wilcoxon matchedpairs signed rank test (non-parametric version of t-tests) or the Friedman test (nonparametric version of ANOVA). Three results were changed based on this statistical correction: Figure 4D, also 5F 3D (from P=0.01 to P=0.0526), Figure 5F  ¼ z-depth (P = 0.005 to P = 0.012). We have updated the manuscript methods, results, and figures accordingly. Importantly, these results did not change the main points of the paper.

Author response:

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

Reviewer #1 (Public Review):

Summary:

In this manuscript the authors re-examine the developmental origin of cortical oligodendrocyte (OL) lineage cells using a combination of strategies, focussing on the question of whether the LGE generates cortical OL cells. The paper is interesting to myelin biologists, the methods used are appropriate and, in general, the study is well-executed, thorough, and persuasive, but not 100% convincing.

Thank you very much for approving our paper.

Strengths, weaknesses, and recommendations:

The first evidence presented that the LGE does not generate OLs for the cortex is that there are no OL precursors 'streaming' from the LGE during embryogenesis, unlike the MGE (Figure 1A). This in itself is not strong evidence, as they might be more dispersed. In fact, in the images shown, there is no obvious 'streaming' from the MGE either. Note that in Figure 1 there is no reference to the star that is shown in the figure.

We totally agree with you. While OPC migration stream is not strong evidence to support that the LGE does not generate OPCs for the cortex, when considering our additional evidence, the absence of obvious 'streaming' from LGE to cortex provided supplementary support for this conclusion. Finally, we have removed the star in the figure.

The authors then electroporate a reporter into the LGE at E13.5 and examine the fate of the electroporated cells (Figures 1C-E). They find that electroporated cells became neurons in the striatum and in the cortex but no OLs for the cortex. There are two issues with this: first, there is no quantification, which means there might indeed be a small contribution from the LGE that is not immediately obvious from snapshot images. Second, it is unexpected to find labelled neurons in the cortex at all since the LGE does not normally generate neurons for the cortex. Electroporations are quite crude experiments as targeting is imprecise and variable and not always discernible at later stages. For example, in Figure 1D, one can see tdTOM+ cells near the AEP, as well as the striatum. Hence, IUE cannot on its own be taken as proof that there is no contribution of the LGE to the cortical OL population.

Thank you for your constructive suggestions.

(1) Following the reviewer's suggestion, we have added these statistics, please see Figure 1F.

(2) The reviewer raised a good point. We occasionally found a very small number of electroporated cells in the MGE/AEP VZ in our IUE system. Therefore, we can identify these electroporated cells in the cortex, most of them expressed the neuronal marker NeuN. We suspect these are MGE-derived cortical interneurons. It's worth noting that these electroporated cells (MGE-derived) are not glia cells. The probable reason may be that MGE/AEP generate cortical OPCs mainly before E13.5 (in this study we performed IUE at E13.5).

The authors then use an alternative fate-mapping approach, again with E13.5 electroporations (Figure 2). They find only a few GFP+ cells in the cortex at E18 (Figures 2C-D) and P10 (Figure 2E) and these are mainly neurons, not OL lineage cells. Again, there is no quantification.

Thank you very much for your suggestions. Actually, in this fate-mapping approach, the electroporated cells in the cortex is very few. We analyzed four mice, and found that all GFP positive cells (139 GFP+) did not express OLIG2, SOX10 and PDGFRA.

Figure 3 is more convincing, but the experiments are incomplete. Here the authors generate triple-transgenic mice expressing Cre in the cortex (Emx1-Cre) and the MGE (Nkx2.1-Cre) as well as a strong nuclear reporter (H2B-GFP). They find that at P0 and P10, 97-98% of OL-lineage cells (SOX10+ or PDGFRA+) in the cortex are labelled with GFP (Figure 3). This is a more convincing argument that the LGE/CGE might not contribute significant numbers of OL lineage cells to the cortex, in contrast to the Kessaris et at. (2006) paper, which showed that Gsh2-Cre mice label ~50% of SOX10+ve cells in the motor cortex at P10. The authors of the present paper suggest that the discrepancy between their study and that of Kessaris et al. (2006) is based on the authors' previous observation (Zhang et al 2020) (https://doi.org/10.1016/j.celrep.2020.03.027) that GSH2 is expressed in intermediate precursors of the cortex from E18 onwards. If correct, then Kessaris et al. might have mistakenly attributed Gsh2-Cre+ lineages to the LGE/CGE when they were in fact intrinsic to the cortex. However, the evidence from Zhang et al 2020 that GSH2 is expressed by cortical intermediate precursors seems to rest solely on their location within the developing cortex; a more convincing demonstration would be to show that the GSH2+ putative cortical precursors co-label for EMX1 (by immunohistochemistry or in situ hybridization), or that they co-label with a reporter in Emx1-driven reporter mice. This demonstration should be simple for the authors as they have all the necessary reagents to hand. Without these additional data, the assertion that GSX2+ve cells in the cortex are derived from the cortical VZ relies partly on an act of faith on the part of the reader. Note that Tripathi et al. (2011, "Dorsally- and ventrally-derived oligodendrocytes have similar electrical properties but myelinate preferred tracts." J. Neurosci. 31, 6809-6819) found that the Gsh-Cre+ OL lineage contributed only ~20% of OLs to the mature cortex, not ~50% as reported by Kessaris et al. (2006). If it is correct that these Gsh2-derived OLs are from the cortical anlagen as the current paper claims, then it would raise the possibility that the ventricular precursors of GSH2+ intermediate progenitors are not uniformly distributed through the cortical VZ but are perhaps localized to some part of it. Then the contribution of Gsh2-derived OLs to the cortical population could depend on precisely where one looks relative to that localized source. It would be a nice addition to the current manuscript if the authors could explore the distribution of their GSH2+ intermediate precursors throughout the developing cortex. In any case, Tripathi et al. (2011) should be cited.

Thank you for your constructive suggestions.

(1) We used the Emx1Cre; RosaH2B-GFP mouse and found that nearly all GSX2+ cells in the cortical SVZ are derived from the Emx1+ lineage at P0 (Please see our new Figure 3-supplement 1A-C). 

(2) According to your suggestion, we have cited this paper (Tripathi et al.) in our revised manuscript.

(3) The study conducted by Kessaris et al. (2006) revealed that roughly 50% of cortical oligodendrocytes (OLs) originate from the Gsx2 lineage (LGE/CGE-derived). In contrast, Tripathi et al. (2011) observed that Gsx2-derived OLs contribute only around 20% to the corpus callosum (CC). To investigate the reasons behind these disparate findings, we conducted three experiments. Firstly, using Emx1Cre; RosaH2B-GFP mice, we found that approximately 89% of lateral CC (LCC) OLs originate from the Emx1 lineage, with only around 11% derived from the ventral source (refer to Author response image 1A and B below). Secondly, employing Nkx2-1Cre; RosaH2B-GFP mice, we determined that approximately 11% of LCC OLs originate from the Nkx2.1 lineage (refer to pictures C and D below). Finally, we found that approximately 98.3% of lateral LCC OLs originate from both Emx1 and Nkx2.1 lineages, with only around 1.7% possibly derived from the LGE (see Author response image 1E and F below). Taken together, our results indicate that approximately 89% of LCC OLs originate from the Emx1 lineage, while 11% of LCC OLs are derived from the medial ganglionic eminence (MGE).

It is worth noting that OLs from Emx1 and Nkx2.1 lineages were equally distributed in the medial CC (mCC) (see Author response image 1G below). This finding suggests that MGE-derived OLs exhibit spatial heterogeneity in their distribution within the CC. These results provide evidence that the contribution of the lateral ganglionic eminence (LGE) and caudal ganglionic eminence (CGE) to CC OLs is minimal.

Author response image 1.

Finally, the authors deleted Olig2 in the MGE and found a dramatic reduction of PDGFRA+ and SOX10+ cells in the cortex at E14 and E16 (Figure 4A-F). This further supports their conclusion that, at least at E16, there is no significant contribution of OLs from ventral sources other than the MGE/AEP. This does not exclude the possibility that the LGE/CGE generates OLs for the cortex at later stages. Hence, on its own, this is not completely convincing evidence that the LGE generates no OL lineage cells for the cortex.

There are three reasons why we didn't analyze Olig2-NCKO mice after E16.5. 1. The expression of Nkx2.1Cre is lower within the dorsal-most region of the MGE than other Nkx2.1-expressing regions. Even at E15.5, we can still find a small number of OPCs in the lateral cortex. We speculate that these OPCs are derived from dorsal MGE. 2. Considering the possibility of incomplete recombination in Olig2 gene locus, we guess OPCs (Olig2+) in the lateral cortex are derived from MGE. Indeed, we found a few OPCs in the MGE/AEP in the Olig2-NCKO mice (Figure 4F). 3. The recent study (bioRxiv preprint doi: https://doi.org/10.1101/2024.01.23.576886) showed that the contribution of LGE/CGE to cortical OPCs is minimal, which further supporting our findings. Taken together, our results provide additional evidence supporting the limited contribution of the LGE/CGE to cortical OPCs (OLs).

Reviewer #2 (Public Review):

Traditional thinking has been that cortical oligodendrocyte progenitor cells (OPCs) arise in the development of the brain from the medial ganglionic eminence (MGE), lateral/caudal ganglionic eminence (LGE/CGE), and cortical radial glial cells (RGCs). Indeed a landmark study demonstrated some time ago that cortical OPCs are generated in three waves, starting with a ventral wave derived from the medial ganglionic eminence (MGE) or the anterior entopeduncular area (AEP) at embryonic day E12.5 (Nkx2.1+ lineage), followed by a second wave of cortical OLs derived from the lateral/caudal ganglionic eminences (LGE/CGE) at E15.5 (Gsx2+/Nkx2.1- lineage), and then a final wave occurring at P0, when OPCs originate from cortical glial progenitor cells (Emx1+ lineage). However, the authors challenge the idea in this paper that cortical progenitors are produced from the LGE. They have found previously that cortical glial progenitor cells were also found to express Gsx2, suggesting this may not have been the best marker for LGE-derived OPCs. They have used fate mapping experiments and lineage analyses to suggest that cortical OPCs do not derive from the LGE.

Strengths:

(1) The data is high quality and very well presented, and experiments are thoughtful and elegant to address the questions being raised.

(2) The authors use two elegant approaches to lineage trace LGE derived cells, namely fate mapping of LGE-derived OPCs by combining IUE (intrauterine electroporation) with a Cre recombinase-dependent IS reporter, and Lineage tracing of LGE-derived OPCs by combining IUE with the PiggyBac transposon system. Both approaches show convincingly that labelled LGE-derived cells that enter the cortex do not express OPC markers, but that those co-labelling with oligodendrocyte markers remain in the striatum.

(3) The authors then use further approaches to confirm their findings. Firstly they lineage trace Emx1-Cre; Nkx2.1-Cre; H2B-GFP mice. Emx1-Cre is expressed in cortical RGCs and Nkx2.1-Cre is specifically expressed in MGE/AEP RGCs. They find that close to 98% of OPCs in the cortex co-label with GFP at later times, suggesting the contribution of OPCs from LGE is minimal.

(4) They use one further approach to strengthen the findings yet further. They cross Nkx2.1-Cre mice with Olig2 F/+ mice to eliminate Olig2 expression in the SVZ/VZ of the MGE/AEP (Figures 4A-B). The generation of MGE/AEP-derived OPCs is inhibited in these Olig2-NCKO conditional mice. They find that the number of cortical progenitors at E16.5 is reduced 10-fold in these mice, suggesting that LGE contribution to cortical OPCs is minimal.

We thank the reviewer for summarizing the strengths of our manuscript.

Weaknesses:

(1) The authors use IUE in experiments mentioned in point 2 of 'Strengths' above (Figures 1 and 2) and claim that the reporter was delivered specifically into LGE VZ at E13.5 using this IUE. It would be nice to see some sort of time course of delivery after IUE to show the reporter is limited to LGE VZ at early times post-IUE.

Thank you very much for your suggestions. Indeed, when using IUE in our system, we occasionally found a small number of electroporated cells in the MGE/AEP VZ. Thus, we can find very few electroporated cells (MGE/AEP-derived) in the cortex and these electroporated cells are neuron (perhaps interneuron).

(2) In the experiments mentioned in point 3 of 'Strengths' (Figure 3), statistical analysis showed that only approximately 2% of OPCs were GFP-negative cells. This 2% could possibly be derived from the LGE/CGE so does not totally rule out that LGE contributes some cortical OPCs.

Thank you for your constructive suggestions. We apologize for any imprecise descriptions. Despite we suspect that this 2% may originate from MGE {Considering the possibility of incomplete recombination in Olig2 gene locus, we guess the OPCs (Olig2+) may be derived from MGE. Indeed, we found a few OPCs in the MGE/AEP in the Olig2-NCKO mice (Figure 4F)} or from the dMGE (The expression of Nkx2.1Cre is lower within the dorsal-most region of the MGE than in other Nkx2.1-expressing regions). Anyway, we have softened the assertion everywhere in our revised manuscript.

(3) In the experiments mentioned in point 4 of 'Strengths' (Figure 4), they do still find cortical OPCs at E16.5 in the Olig2-NCKO conditional mice. It is unclear whether this is due to the recombination efficiency of the CRE enzyme not being 100%, or whether there is some LGE contribution to the cortical OPCs.

This experiment alone may not provide strong evidence to support that LGE do not contribute to the cortical OPCs during development. However, when combing our other results with this result, we can confirm that the contribution of LGE to cortical OPCs is minimal. Furthermore, a recent study reported that LGE/CGE-derived OLs make minimum contributions to the neocortex and corpus callosum，which further supporting the reliability of our conclusion.

We would like to thank the reviewers and editors for their valuable comments and suggestions again.

Impact of Study:

The authors show elegantly and convincingly that the contribution of the LGE to the pool of cortical OPCs is minimal. The title should perhaps be that the LGE contribution is minimal rather than no contribution at all, as they are not able to rule out some small contribution from the LGE. These findings challenge the traditional belief that the LGE contributes to the pool of cortical OPCs. The authors do show that the LGE does produce OPCs, but that they tend to remain in the striatum rather than migrate into the cortex. It is interesting to wonder why their migration patterns may be different from the MGE-derived OPCs which migrate to the cortex. The functional significance of these different sources of OPCs for adult cortex in homeostatic or disease states remains unclear though.

Recommendations for the authors:

Reviewer #1 (Recommendations for The Authors):

(1) Change the title to e.g. 'limited contribution of the LGE to cortical oligodendrocytes'. Alternatively, It might be more useful to highlight where they come from, e.g. "Cortical oligodendrocytes originate predominantly or exclusively from the MGE and cortical VZ"

As suggested, we have changed the old title to the following: The lateral/caudal ganglionic eminence makes a limited contribution to cortical oligodendrocytes

(2) Demonstrate using lineage tracing that GSH2+ cells in the cortex are derived from the Emx1-lineage, e.g. using immunohistochemistry for GSX2 and a reporter in Emx1-Cre mice crossed to a reporter.

In our revised manuscript, we have added a new figure (Figure 3-supplement 1A-C) to demonstrate that the GSX2+ cells in the cortex are derived from the Emx1-lineage.

(3) Make it clear in their discussion that they have not explored the CGE so it is possible that this region generates some OLs.

The Emx1Cre; Nkx2.1Cre; H2B-GFP mice showed that only ~2% cortical OLs are derived from LGE/CGE. Actually, considering the efficiency of Cre enzyme recombination and the relatively low Cre activity in the dMGE of Nkx2.1Cre, the actual contribution of LGE/CGE-derived cortical OLs could be even lower than our current observation. Therefore, our results demonstrate that the LGE/CGE generate very few，possibly even no，OLs for the cortex.

(4) Soften the assertion that the LGE does not generate any OL lineage cells that reach the cortex by e.g. changing the word 'sole' to 'predominant' (line 88) and, elsewhere in the paper, leaving open the possibility that small numbers of LGE-derived OLs might enter the cortex, depending on where exactly one looks.

As suggested, we have softened the assertion everywhere in our manuscript.

(5) Lines 255-260: 'First, the time window during which the MGE generates OLs is very brief, perhaps occurring before MGE neurogenesis. The high level of SHH in the MGE allows for the production of a small population of cortical OPCs around E12.5. Subsequently, multipotent intermediate progenitors begin to express DLX transcription factors resulting in ending the generation of OPCs in the MGE'. What is the evidence that OL genesis precedes neurogenesis? If there is none (as I suspect) then this statement should be removed.

The editors raised a good point. We have no strong evidence to support that OL genesis precedes neurogenesis in MGE, thus, we removed these sentences in our manuscript.

(6) Figure 1E should show quantification of cells as a % of electroporated cells and as a % of PDGFRA+ or OLIG2+ or SOX10+ cells, so that the reader might have a clear view of the extent of labelling.

Done.

(7) Figure 4: This is interesting but incomplete. At E14.5 the authors show the presence of PDGFRA+cells in the telencephalon. However, at E16.5 they show images only of the dorsal-most region of the cortex. If the LGE/CGE begins to generate OLPs for the early cortex, they would be expected to appear near the cortico-striatal boundary, as shown in Kessaris 2006 Fig1g-h. In the current manuscript, the authors do not show these regions, or the LGE and CGE, in their images. It is essential to show PDGFRA immunolabelling at the cortico-striatal boundary and also in the LGE and CGE at E16.5 in control and Olig2 mutant mice. It is also necessary to extend this analysis to E18.5, perhaps showing PDGFRA+ cells streaming from the cortical VZ/SVZ.

There are three reasons why we didn't analyze Olig2-NCKO mice after E16.5. 1.Frankly, the expression of Nkx2.1Cre is lower within the dorsal-most region of the MGE than other Nkx2.1-expressing regions. Even at E15.5, we can still find a small number of OPCs in the lateral cortex. We guess these OPCs are derived from dMGE. 2. Considering the possibility of incomplete recombination in Olig2 gene locus, we guess OPCs (Olig2+) are derived from MGE. In fact, we found a few OPCs in the MGE/AEP in the Olig2-NCKO mice (Figure 4F). 3. The recent study (bioRxiv preprint doi: https://doi.org/10.1101/2024.01.23.576886) showed that the contribution of LGE/CGE to cortical OPCs is minimal. Taken together, our results provide additional evidence supporting the limited contribution of the LGE/CGE to cortical OPCs (OLs).

(8) Cite Tripathi et al. (2011) and mention the disparity between the findings of that paper and Kessaris et al. (2006) and possible reasons - see main review above.

Done.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The manuscript "Self-inhibiting percolation and viral spreading in epithelial tissue" describes a model based on 5-state cellular automata of development of an infection. The model is motivated and qualitatively justified by time-resolved measurements of expression levels of viral, interferon-producing, and antiviral genes. The model is set up in such a way that the crucial difference in outcomes (infection spreading vs. confinement) depends on the initial fraction of special virus-sensing cells. Those cells (denoted as 'type a') cannot be infected and do not support the propagation of infection, but rather inhibit it in a somewhat autocatalytic way. Presumably, such feedback makes the transition between two outcomes very sharp: a minor variation in concentration of ``a' cells results in qualitative change from one outcome to another. As in any percolation-like system, the transition between propagation and inhibition of infection goes through a critical state with all its attributes. A power-law distribution of the cluster size (corresponding to the fraction of infected cells) with a fairly universal exponent and a cutoff at the upper limit of this distribution.

Strengths:

The proposed model suggests an explanation for the apparent diversity of outcomes of viral infections such as COVID.

Author response: We thank the referee for the concise and accurate summary of our work.

Weaknesses:

Those are not real points of weakness, though I think addressing them would substantially improve the manuscript.

Author response: Below we will address these point by point.

The key point in the manuscript is the reduction of actual biochemical processes to the NOVAa rules. I think more could be said about it, be it referring to a set of well-known connections between expression states of cells and their reaction to infection or justifying it as an educated guess.

Author response: We have now improved this part in the model section. We have added a few sentences explaining how the cell state transitions are motivated by the UMAP results:

“The cell state transitions triggered by IFN signaling or viral replication are known in viral infection, but how exactly the transitions are orchestrated for specific infections is poorly understood. The UMAP cell state distribution hints at possible preferred transitions between states. The closer two cell states are on the UMAP, the more likely transitions between them are, all else being equal. For instance, the antiviral state (𝐴) is easily established from a susceptible cell (𝑂), but not from the fully virus-hijacked cell (𝑉 ). The IFN-secreting cell state (𝑁) requires the co-presence of the viral and antiviral genes and thus the cell cluster is located between the antiviral state (𝐴) and virus-infected state (𝑉 ) but distant from the susceptible cells (𝑂).

Inspired by the UMAP data visualization (Fig. 1a), we propose the following transitions between five main discrete cell states”

Another aspect where the manuscript could be improved would be to look a little beyond the strange and 'not-so-relevant for a biomedical audience' focus on the percolation critical state. While the presented calculation of the precise percolation threshold and the critical exponent confirm the numerical skills of the authors, the probability that an actual infected tissue is right at the threshold is negligible. So in addition to the critical properties, it would be interesting to learn about the system not exactly at the threshold: For example, how the speed of propagation of infection depends on subcritical p_a and what is the cluster size distribution for supercritical p_a.

Author response: We agree that further exploring the model away from the critical threshold is worthwhile. While our main focus has been on explaining the large degree of heterogeneity in outcomes – readily explained as a consequence of the sharp threshold-like behavior – we now include plots of the time-evolution of the infection (as well as the remaining states) over time for subcritical values of pa. The plots can be found in Figure S4 of the supplement.

Reviewer #2 (Public Review):

Xu et al. introduce a cellular automaton model to investigate the spatiotemporal spreading of viral infection. In this study, the author first analyzes the single-cell RNA sequencing data from experiments and identifies four clusters of cells at 48 hours post-viral infection, including susceptible cells (O), infected cells (V), IFN-secreting cells (N), and antiviral cells (A). Next, a cellular automaton model (NOVAa model) is introduced by assuming the existence of a transient pre-antiviral state (a). The model consists of an LxL lattice; each site represents one cell. The cells change their state following the rules depending on the interaction of neighboring cells. The model introduces a key parameter, p_a, representing the fraction of pre-antiviral state cells. Cell apoptosis is omitted in the model. Model simulations show a threshold-like behavior of the final attack rate of the virus when p_a changes continuously. There is a critical value p_c, so that when p_a < p_c, infections typically spread to the entire system, while at a higher p_a > p_c, the propagation of the infected state is inhibited. Moreover, the radius R that quantifies the diffusion range of N cells may affect the critical value p_c; a larger R yields a smaller value of the critical value p_c. The structure of clusters is different for different values of R; greater R leads to a different microscopic structure with fewer A and N cells in the final state. Compared with the single-cell RNA seq data, which implies a low fraction of IFN-positive cells - around 1.7% - the model simulation suggests R=5. The authors also explored a simplified version of the model, the OVA model, with only three states. The OVA model also has an outbreak size. The OVA model shows dynamics similar to the NOVAa model. However, the change in microstructure as a function of the IFN range R observed in the NOVAa model is not observed in the OVA model.

Author response: We thank the referee for the comprehensive summary of our work.

Data and model simulation mainly support the conclusions of this paper, but some weaknesses should be considered or clarified.

Author response: Thank you - we will address these point by point below.

(1) In the automaton model, the authors introduce a parameter p_a, representing the fraction of pre-antiviral state cells. The authors wrote: ``The parameter p_a can also be understood as the probability that an O cell will switch to the N or A state when exposed to the virus of IFNs, respectively.' Nevertheless, biologically, the fraction of pre-antiviral state cells does not mean the same value as the probability that an O cell switches to the N or A state. Moreover, in the numerical scheme, the cell state changes according to the deterministic role N(O)=a and N(a)=A. Hence, the probability p_a did not apply to the model simulation. It may need to clarify the exact meaning of the parameter p_a.

Author response: We acknowledge that this was an imprecise formulation, and have now changed it.

What we tried to convey with that comment was that, alternatively to having a certain fraction of cells be in the a state initially, one could instead have devised a model in which We should note that even the current model has a level of stochasticity, since we choose the cells to be updated with a constant probability rate - we choose N cells to update in each timestep, with replacement.

However, based on your suggestion, we simulated a version of the dynamics which included stochastic conversion, i.e. each action of a cell on a nearby cell happens only with a probability p_conv (and the original model is recovered as the p_conv=1 scenario). Of course, this slows down the dynamics (or effectively rescales time by a factor p_conv), but crucially we find that it does not appreciably affect the location of the threshold p_c. Below we include a parameter scan across p_a values for R=1 and p_conv=0.5, which shows that the threshold continues to appear at around p_a=27%. each O-state cell simply had a probability to act as an a-state cell upon exposure to the virus or to interferons, i.e. to switch to an N state (if exposed to virus) or to the A state (if exposed to interferons). In this simplified model, there would be no functional difference, since it would simply amount to whether each cell had a probability to be designated an a-cell initially (as in our model), or upon exposure. So our remark mainly served to explain that the role of the p_a parameter is simply to encode that a certain fraction of virus-naive cells behave this way (whether predetermined or not).

(2) The current model is deterministic. However, biologically, considering the probabilistic model may be more realistic. Are the results valid when the probability update strategy is considered? By the probability model, the cells change their state randomly to the state of the neighbor cells. The probability of cell state changes may be relevant for the threshold of p_a. It is interesting to know how the random response of cells may affect the main results and the critical value of p_a.

Author response: This is a good point - we are firm believers in the importance of stochasticity. We should note that even the current model has a level of stochasticity, since we choose the cells to be updated with a constant probability rate - we choose N cells to update in each timestep, with replacement.

However, based on your suggestion, we simulated a version of the dynamics which included stochastic conversion, i.e. each action of a cell on a nearby cell happens only with a probability p_conv (and the original model is recovered as the p_conv=1 scenario). Of course, this slows down the dynamics (or effectively rescales time by a factor p_conv), but crucially we find that it does not appreciably affect the location of the threshold p_c. Below we include a parameter scan across p_a values for R=1 and p_conv=0.5, which shows that the threshold continues to appear at around p_a=27%.

We now discuss these findings in the supplement and include the figure below as Fig. S5.

Author response image 1.

(3) Figure 2 shows a critical value p_c = 27.8% following a simulation on a lattice with dimension L = 1000. However, it is unclear if dimension changes may affect the critical value.

Author response: Re-running the simulations on a lattice 4x as large (i.e. L=2000) yields a similar critical value of 27-28% for R=1, so we are confident that finite size effects do not play a major role at L=1000 and beyond. For R=5, however, we find that a minimum lattice size greater than L=1000 is necessary to determine the critical threshold. Concretely, we find that the threshold value pc for R=5 changes somewhat when the lattice size is increased from 1000 to 2000, but is invariant under a change from 2000 to 3000, so we conclude that L=2000 is sufficient for R=5. The pc value for R=5 cited in the manuscript (~0.4%) was determined from simulations at L=2000.

Reviewer #3 (Public Review):

Summary:

This study considers how to model distinct host cell states that correspond to different stages of a viral infection: from naïve and susceptible cells to infected cells and a minority of important interferon-secreting cells that are the first line of defense against viral spread. The study first considers the distinct host cell states by analyzing previously published single-cell RNAseq data. Then an agent-based model on a square lattice is used to probe the dependence of the system on various parameters. Finally, a simplified version of the model is explored, and shown to have some similarity with the more complex model, yet lacks the dependence on the interferon range. By exploring these models one gains an intuitive understanding of the system, and the model may be used to generate hypotheses that could be tested experimentally, telling us "when to be surprised" if the biological system deviates from the model predictions.

Author response: Thank you for the summary! We agree with the role that you describe for a model such as this one.

Strengths:

-  Clear presentation of the experimental findings and a clear logical progression from these experimental findings to the modeling.

-  The modeling results are easy to understand, revealing interesting behavior and percolation-like features.

-  The scaling results presented span several decades and are therefore compelling. - The results presented suggest several interesting directions for theoretical follow-up work, as well as possible experiments to probe the system (e.g. by stimulating or blocking IFN secretion).

Weaknesses:

-  Since the "range" of IFN is an important parameter, it makes sense to consider lattice geometries other than the square lattice, which is somewhat pathological. Perhaps a hexagonal lattice would generalize better.

-  Tissues are typically three-dimensional, not two-dimensional. (Epithelium is an exception). It would be interesting to see how the modeling translates to the three-dimensional case. Percolation transitions are known to be very sensitive to the dimensionality of the system.

Author response: We agree that probing different lattice geometries (2- and 3-dimensional alike) would be interesting and worthwhile. However, for this manuscript, we prefer to confine the analysis to the current, simple case. We do agree, however, that an extensive exploration of the role of geometry is an interesting future possibility.

-  The fixed time-step of the agent-based modeling may introduce biases. I would consider simulating the system with Gillespie dynamics where the reaction rates depend on the ambient system parameters.

-  Single-cell RNAseq data typically involves data imputation due to the high sparsity of the measured gene expression. More information could be provided on this crucial data processing step since it may significantly alter the experimental findings.

Justification of claims and conclusions:

The claims and conclusions are well justified.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

It is necessary to explain what UMAP does. Is clustering done in the space of twenty-something original dimensions or 2D? How UMAP1 and UMAP2 are selected and are those the same in all plots?

Author response: We have now added a few sentences to clarify the point raised above - the second snippet explains how clustering is performed:

“As a dimension reduction algorithm, UMAP is a manifold learning technique that favors the preservation of local distances over global distances (McInnes et al., 2018; Becht et al., 2019). It constructs a weighted graph from the data points and optimizes the graph layout in the low-dimensional space.”

“We cluster the cells with the principal components analysis (PCA) results from their gene expression. With the first 16 principal components, we calculate k-nearest neighbors and construct the shared nearest neighbor graph of the cells then optimize the modularity function to determine clusters. We present the cluster information on the UMAP plane and use the same UMAP coordinates for all the plots in this paper hereafter.”

Figure 1, what do bars in the upper right corners of panels d,e,f, and g indicate? ``Averaged' refers to time average? Something is missing in ``Cell proportions are labeled with corresponding colors in a)' .

Author response: Thank you - we have now modified the figure caption. The bars in the upper right corners of panels d, e, f are color keys for gene expression, the brighter the color is, the higher the gene expression is.

“Averaged” gene expression refers to the mean expression of that particular gene across the cells within each indicated cluster.

The lines in c) correspond to cell proportions in different states at different time points. The same state in 1) and c) is shown in the same color.

Line 46, ``However' does not sound right in this context. Would ``Also' be better?

Author response: We agree and have corrected it in the revised manuscript.

Line 96``The viral genes are also partially expressed in these cells, but different from the 𝑁 cluster, the antiviral genes are fully expressed (Fig. S1 and S2).' The sentence needs to be rephrased.

Author response: We have rephrased the sentence: “As in the N cluster, the viral gene E is barely detected in these cells, indicating incomplete viral replication. However, in contrast to the N cluster, the antiviral genes are expressed to their full extent (Fig. S1 and S2).”

Line 126, missing "be", ``large' -> ``larger'.

Author response: Thank you, we have now corrected these typos.

Line 139-140 The logical link between ignoring apoptosis and the diffusion of IFN is unclear.

Author response: We modified the sentence as “Here, we assume that the secretion of IFNs by the 𝑁 cells is a faster process than possible apoptosis (Wen et al., 1997; Tesfaigzi, 2006) of these cells and that the diffusion of IFNs to the neighborhood is not significantly affected by apoptosis.”

Fig. 2a Do the yellow arrows show the effect of IFN and the purple arrows the propagation of viral infection?

Author response: That is correct. We have added this information to the figure caption: “The straight black arrows indicate transitions between cell states. The curved yellow arrows indicate the effects of IFNs on activating antiviral states. The curved purple arrows indicate viral spread to cells with 𝑂 and 𝑎 states.”

Fig. 3, n(s) as the axis label vs P(s) in the text? How do the curves in panel a) look when the p_a is well above or below p_c?

Author response: Thank you. We have edited the labels in the figure to reflect the symbols used in the text.

Boundary conditions? From Fig. 4, apparently periodic?

Author response: Yes, we use periodic boundary conditions in the model. We clarify it in the model section now (last sentence).

It will be good to see a plot with time dependences of all cell types for a couple of values of p_a, illustrating propagation and cessation of the infection.

Author response: We agree, and have added a Figure S4 in the supplement which explores exactly that. Thank you for the suggestion.

A verbal qualitative description of why p_a has such importance and how the infection is terminated for large p_a would help.

Reviewer #2 (Recommendations For The Authors):

Below are two minor comments:

(1) In the single-cell RNA sequencing data analysis, the authors describe the cell clusters O, V, A, and N. However, showing how the clusters are identified from the data might be more straightforward.

Author response: Technically, we cluster the cells using principal components analysis (PCA) results of their gene expression. With the first 16 principal components, we calculate k-nearest neighbors and construct the shared nearest neighbor graph of the cells and then optimize the modularity function to determine clusters. We manually annotate the clusters with O, V, A, and N based on the detected abundance of viral genes, antiviral genes, and IFNs.

(2) In Figure 3, what does n(s) mean in Figure 3a? And what is the meaning of the distribution P(s) of infection clusters? It may be stated clearly.

Author response: The use of n(s) was inconsistent, and we have now edited the figure to instead say P(s), to harmonize it with the text. P(s) is the distribution of cluster sizes, s, expressed as a fraction of the whole system. In other words, once a cluster has reached its final size, we record s=(N+V)/L^2 where N and V are the number of N and V state cells in the cluster (note that, by design, each simulation leads to a single cluster, since we seed the infection in one lattice point). We now indicate more clearly in the caption and the main text what exactly P(s) and s refer to.

Reviewer #3 (Recommendations For The Authors):

- Would the authors kindly share the simulation code with the community? Also, the data analysis code should be shared to follow current best practices. This needs to be standard practice in all publications. I would go as far as to say that in 2024 publishing a data analysis / simulation study without sharing the relevant code should be ostracized by the community.

Author response: We absolutely agree and have created a GitHub repository in which we share the C++ source code for the simulations and a Python notebook for plotting. The public repository can be found at https://github.com/BjarkeFN/ViralPercolation. We add this information in supplement under section “Code availability”.

­

- I would avoid the use of the wording "critical" threshold since this is almost guaranteed to infuriate a certain type of reader.

­

- Line 265 has a curious use of " ... " which should be replaced with something more appropriate.

Author response: Thank you for pointing it out! We have checked the typos.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1:

The manuscript suggests the zebrafish homolog of ctla-4 and generates a new mutant in it. However, the locus that is mutated is confusingly annotated as both CD28 (current main annotation in ZFIN) and CTLA-4/CD152 (one publication from 2020), see: https://zfin.org/ZDB-GENE-070912-128. Both human CTLA-4 and CD28 align with relatively similar scores to this gene. There seem to be other orthologs of these receptors in the zebrafish genome, including CD28-like (https://zfin.org/ZDB-GENE-070912-309) which neighbors the gene annotated as CD28 (exhibiting similar synteny as human CD28 and CTLA-4). It would be helpful to provide more information to distinguish between this family of genes and to further strengthen the evidence that this mutant is in ctla-4, not cd28. Also, is one of these genes in the zebrafish genome (e.g. cd28l) potentially a second homolog of CTLA-4? Is this why this mutant is viable in zebrafish and not mammals? Some suggestions:

(a) A more extensive sequence alignment that considers both CTLA-4 and CD28, potentially identifying the best homolog of each human gene, especially taking into account any regions that are known to produce the functional differences between these receptors in mammals and effectively assigns identities to the two genes annotated as "cd28" and "cd28l" as well as the gene "si:dkey-1H24.6" that your CD28 ORF primers seem to bind to in zebrafish.

In response to the reviewer's insightful suggestions, we have conducted more extensive sequence alignment and phylogenetic analyses that consider both CTLA-4, CD28, and CD28-like molecules, taking into account key regions crucial for the functionalities and functional differences between these molecules across various species, including mammals and zebrafish.

Identification of zebrafish Ctla-4: We identified zebrafish Ctla-4 as a homolog of mammalian CTLA-4 based on key conserved structural and functional characteristics. Structurally, the Ctla-4 gene shares similar exon organization compared to mammalian CTLA-4. Ctla-4 is a type I transmembrane protein with typical immunoglobulin superfamily features. Multiple amino acid sequence alignments revealed that Ctla-4 contains a ¹¹³LFPPPY¹¹⁸ motif and a ¹²³GNGT¹²⁶ motif in the ectodomain, and a tyrosine-based ²⁰⁶YVKF²⁰⁹ motif in the distal C-terminal region. These motifs closely resemble MYPPPY, GNGT, and YVKM motifs in mammalian CTLA-4s, which are essential for binding to CD80/CD86 ligands and molecular internalization and signaling inhibition. Despite only 23.7% sequence identity to human CTLA-4, zebrafish Ctla-4 exhibits a similar tertiary structure with a two-layer β-sandwich architecture in its extracellular IgV-like domain. Four cysteine residues responsible for the formation of two pairs of disulfide bonds (Cys²⁰-Cys⁹¹/Cys⁴⁶-Cys⁶⁵ in zebrafish and Cys²¹-Cys⁹²/Cys⁴⁸-Cys⁶⁶ in humans) that connect the two-layer β-sandwich are conserved. Additionally, a separate cysteine residue (Cys¹²⁰ in zebrafish and Cys¹²⁰ in humans) involved in dimerization is also present, and Western blot analysis under reducing and non-reducing conditions confirmed Ctla-4’s dimerization. Phylogenetically, Ctla-4 clusters with other known CTLA-4 homologs from different species with high bootstrap probability, while zebrafish Cd28 groups separately with other CD28s. Functionally, Ctla-4 is predominantly expressed on CD4⁺ T and CD8⁺ T cells in zebrafish. It plays a pivotal inhibitory role in T cell activation by competing with CD28 for binding to CD80/86, as validated through a series of both in vitro and in vivo assays, including microscale thermophoresis assays which demonstrated that Ctla-4 exhibits a significantly higher affinity for Cd80/86 than Cd28 (KD = 0.50 ± 0.25 μM vs. KD = 2.64 ± 0.45 μM). These findings confirm Ctla-4 as an immune checkpoint molecule, reinforcing its identification within the CTLA-4 family.

Comparison between zebrafish Cd28 and "Cd28l": Zebrafish Cd28 contains an extracellular SYPPPF motif and an intracellular FYIQ motif. The extracellular SYPPPF motif is essential for binding to Cd80/CD86, while the intracellular FYIQ motif likely mediates kinase recruitment and co-stimulatory signaling. In contrast, the "Cd28l" molecule lacks the SYPPPF motif, which is critical for Cd80/CD86 binding, and exhibits strong similarity in its C-terminal 79 amino acids to Ctla-4 rather than Cd28. Consequently, "Cd28l" resembles an atypical Ctla-4-like molecule but fails to exhibit Cd80/CD86 binding activity.

We have incorporated the relevant analysis results into the main text of the revised manuscript and updated Supplementary Figure 1. Additionally, we provide key supplementary analyses here for the reviewer's convenience.  

Author response image 1.

Illustrates the alignment of Ctla-4 (XP_005167576.1) and Ctla-4-like (XP_005167567.1, previously referred to as "Cd28l") in zebrafish, generated using ClustalX and Jalview. Conserved and partially conserved amino acid residues are highlighted in color gradients ranging from carnation to red, respectively. The B7-binding motif is encircled with a red square.

(b) Clearer description in the main text of such an analysis to better establish that the mutated gene is a homolog of ctla-4, NOT cd28.

We appreciate the reviewer's advice. Additional confirmation of zebrafish Ctla-4 is detailed in lines 119-126 of the revised manuscript.

(c) Are there mammalian anti-ctla-4 and/or anti-cd28 antibodies that are expected to bind to these zebrafish proteins? If so, looking to see whether staining is lost (or western blotting is lost) in your mutants could be additionally informative. (Our understanding is that your mouse anti-Ctla-4 antibody is raised against recombinant protein generated from this same locus, and so is an elegant demonstration that your mutant eliminates the production of the protein, but unfortunately does not contribute additional information to help establish its homology to mammalian proteins).

This suggestion holds significant value. However, a major challenge in fish immunology research is the limited availability of antibodies suitable for use in fish species; antibodies developed for mammals are generally not applicable. We attempted to use human and mouse anti-CTLA-4 and anti-CD28 antibodies to identify Ctla-4 and Cd28 in zebrafish, but the results were inconclusive, with no expected signals. This outcome likely arises from the low sequence identity between human/mouse CTLA-4 and CD28 and their zebrafish homologs (ranging from 21.3% to 23.7% for CTLA-4 and 21.2% to 24.0% for CD28). Therefore, developing specific antibodies against zebrafish Ctla-4 is essential for advancing this research.

The methods section is generally insufficient and doesn't describe many of the experiments performed in this manuscript. Some examples:

(a) No description of antibodies used for staining or Western blots (Figure1C, 1D, 1F).

(b) No description of immunofluorescence protocol (Figure 1D, 1F).

(c) No description of Western blot protocol (Figure 1C, 2C).

(d) No description of electron microscopy approach (Figure 2K).

(e) No description of the approach for determining microbial diversity (Entirety of Figure 6).

(f) No description of PHA/CFSE/Flow experiments (Figure 7A-E).

(g) No description of AlphaFold approach (Figures 7F-G).

(h) No description of co-IP approach (Figure 7H).

(i) No description of MST assay or experiment (Figure 7I).

(j) No description of purification of recombinant proteins, generation of anti-Ctla-4 antibody, or molecular interaction assays (Figures S2 and S6).

We apologize for this oversight. The methods section was inadvertently incomplete due to an error during the file upload process at submission. This issue has been addressed in the revised manuscript. We appreciate your understanding.

Figure 5 suggests that there are more Th2 cells 1, Th2 cells 2, and NKT cells in ctla-4 mutants through scRNA-seq. However, as the cell numbers for these are low in both genotypes, there is only a single replicate for each genotype scRNA-seq experiment, and dissociation stress can skew cell-type proportions, this finding would be much more convincing if another method that does not depend on dissociation was used to verify these results. Furthermore, while Th2 cells 2 are almost absent in WT scRNA-seq, KEGG analysis suggests that a major contributor to their clustering may be ribosomal genes (Fig. 5I). Since no batch correction was described in the methods, it would be beneficial to verify the presence of this cluster in ctla-4 mutants and WT animals through other means, such as in situ hybridization or transgenic lines.   

We are grateful for the insightful comments provided by the reviewer. Given that research on T cell subpopulations in fish is still in its nascent stages, the availability of specific marker antibodies and relevant transgenic strains remains limited. Our single-cell RNA sequencing (scRNA-seq) analysis revealed that a distinct Th2 subset 2 was predominantly observed in Ctla-4 mutants but was rare in wild-type zebrafish, it suggests that this subset may primarily arise under pathological conditions associated with Ctla-4 mutation. Due to the near absence of Th2 subset 2 in wild-type samples, KEGG enrichment analysis was performed exclusively on this subset from Ctla-4-deficient intestines. The ribosome pathway was significantly enriched, suggesting that these cells may be activated to fulfill their effector functions. However, confirming the presence of Th2 subset 2 using in situ hybridization or transgenic zebrafish lines is currently challenging due to the lack of lineage-specific markers for detailed classification of Th2 cell subsets and the preliminary nature of scRNA-seq predictions.

To address the reviewers' suggestion to confirm compositional changes in Th2 and NKT cells using dissociation-independent methods, we quantified mRNA levels of Th2 (il4, il13, and gata3) and NKT (nkl.2, nkl.4, and prf1.1) cell marker genes via RT-qPCR in intestines from wild-type and mutant zebrafish. As shown in Figure S7B and S7C, these markers were significantly upregulated in Ctla-4-deficient intestines compared to wild-type controls. This indicates an overall increase in Th2 and NKT cell activity in mutant zebrafish, aligning with our scRNA-seq analysis and supports the validity of our initial findings.

Before analyzing the scRNA-seq data, we performed batch correction using the Harmony algorithm via cloud-based Cumulus v1.0 on the aggregated gene-count matrices. This methodological detail has been included in the “Materials and Methods” section of the revised manuscript. Moreover, the RT-qPCR results are presented in Supplementary Figures S7B and S7C.

Quality control (e.g., no. of UMIs, no. of genes, etc.) metrics of the scRNAseq experiments should be presented in the supplementary information for each sample to help support that observed differential expression is not merely an outcome of different sequencing depths of the two samples.

As illustrated in Fig. S5, the quality control data have been supplemented to include the effective cell number of the sample, along with pre- and post-filtering metrics such as nFeature_RNA, nCount_RNA and mitochondrial percentage (percent.mito). Furthermore, scatter plots comparing the basic information of the sample cells before and after filtering are provided.

Some references to prior research lack citations. Examples:

(a)"Given that Ctla-4 is primarily expressed on T cells (Figure 1E-F), and its absence has been shown to result in intestinal immune dysregulation, indicating a crucial role of this molecule as a conserved immune checkpoint in T cell inhibition."

The references were incorporated into line 71 of the revised manuscript.

(b) Line 83: Cite evidence/review for the high degree of conservation in adaptive immunity.

The references were incorporated into line 93 of the revised manuscript.

(c) Lines 100-102: Cite the evidence that MYPPPY is a CD80/86 binding motif.

The references were incorporated into line 117 of the revised manuscript.

The text associated with Figure 8 (Lines 280-289) does not clearly state that rescue experiments are being done in mutant zebrafish.

We have provided a clear explanation of the rescue experiments conducted in Ctla-4-deficient zebrafish. This revision has been incorporated into line 319.

Line 102: Is there evidence from other animals that LFPPPY can function as a binding site for CD80/CD86? Does CD28 also have this same motif?

The extracellular domains of CTLA-4 and CD28, which bind to CD80/CD86, are largely conserved across various species. This conservation is exemplified by a central PPP core motif, although the flanking amino acids exhibit slight variations. In mammals, both CTLA-4 and CD28 feature the conserved MYPPPY motif. By contrast, in teleost fish, such as rainbow trout, CTLA-4 contains an LYPPPY motif, while CD28 has an MYPPPI motif (Ref. 1). Grass carp CTLA-4 displays an LFPPPY motif, whereas its CD28 variant bears an IYPPPF motif. Yeast two-hybrid assays confirm that these motifs facilitate interactions between grass carp CTLA-4 and CD28 with CD80/CD86 (Ref. 2). Similarly, zebrafish Ctla-4 contains the LFPPPY motif observed in grass carp, while Cd28 exhibits a closely related SYPPPF motif.

References:

(1) Bernard, D et al. (2006) Costimulatory Receptors in a Teleost Fish: Typical CD28, Elusive CTLA-4. J Immunol. 176: 4191-4200.

(2) Lu T Z et al. (2022) Molecular and Functional Analyses of the Primordial Costimulatory Molecule CD80/86 and Its Receptors CD28 and CD152 (CTLA-4) in a Teleost Fish. Frontiers in Immunology. 13:885005.

Line 110-111: Suggest adding citation of these previously published scRNAseq data to the main text in addition to the current description in the Figure legend.

The reference has been added in line 129 in the main text.

Figure 3B: It would be helpful to label a few of the top differentially expressed genes in Panel B?

The top differentially expressed genes have been labeled in Figure 3B.

Figure 3G: It's unclear how this analysis was conducted, what this figure is supposed to demonstrate, and in its current form it is illegible.

Figure 3G displays a protein-protein interaction network constructed from differentially expressed genes. The densely connected nodes, representing physical interactions among proteins, provide valuable insights for basic scientific inquiry and biological or biomedical applications. As proteins are crucial to diverse biological functions, their interactions illuminate the molecular and cellular mechanisms that govern both healthy and diseased states in organisms. Consequently, these networks facilitate the understanding of pathogenic and physiological processes involved in disease onset and progression.

To construct this network, we first utilized the STRING database (https://string-db.org) to generate an initial network diagram using the differentially expressed genes. This diagram was subsequently imported into Cytoscape (version 3.9.1) for visualization and further analysis. Node size and color intensity reflect the density of interactions, indicating the relative importance of each protein. Figure 3G illustrates that IL1β was a central cytokine hub in the disease process of intestinal inflammation in Ctla-4-deficient zebrafish.

Expression scale labeling:

(a) Most gene expression scales are not clearly labeled: do they represent mean expression or scaled expression? Has the expression been log-transformed, and if so, which log (natural log? Log10? Log2?). See: Figure 3E, 3I, 4D, 4E, 5B, 5G, 5H, 6I.

The gene expression scales are detailed in the figure legends. Specifically, Figures 3E, 3I, and 6I present heatmaps depicting row-scaled expression levels for the corresponding genes. In contrast, Figures 4D and 4E display heatmaps illustrating the mean expression of these genes. Additionally, the dot plots in Figures 5B, 5G, and 5H visualize the mean expression levels of the respective genes.

(b) For some plots, diverging color schemes (i.e. with white/yellow in the middle) are used for non-diverging scales and would be better represented with a sequential color scale. See: 4D, 4E, and potentially others (not fully clear because of the previous point).

The color schemes in Figures 4D and 4E have been updated to a sequential color scale. The gene expression data depicted in these figures represent mean expression values and have not undergone log transformation. This information has been incorporated into the figure legend for clarity.

Lines 186-187: Though it is merely suggested, apoptotic gene expression can be upregulated as part of the dissociation process for single-cell RNAseq. This would be much stronger if supported by a staining, such as anti-Caspase 3.

Following the reviewer's insightful recommendations, we conducted a TUNEL assay to evaluate apoptosis in the posterior intestinal epithelial cells of both wild-type and Ctla-4-deficient zebrafish. As expected, our results demonstrate a significant increase in epithelial cell apoptosis in Ctla-4-deficient zebrafish compared with wild-type fish. The corresponding data are presented in Figure S6D and have been incorporated into the manuscript. Detailed protocols for the TUNEL assay have also been included in the Materials and Methods section.

Author response image 2.

Illustrates the quantification of TUNEL-positive cells per 1 × 10⁴ μm^2/⁻ in the posterior intestines of both wild-type (WT) and ctla-4^⁻/⁻ zebrafish (n = 5). The data demonstrate a comparative analysis of apoptotic cell density between the two genotypes.

Lines 248-251: This manuscript demonstrates gut inflammation and also changes in microbial diversity, but I don't think it demonstrates an association between them, which would require an experiment that for instance rescues one of these changes and shows that it ameliorates the other change, despite still being a ctla-4 mutant.

We appreciate the valuable comments from the reviewer. Recently, the relationship between inflammatory bowel disease (IBD) and gut microbial diversity has garnered considerable attention, with several key findings emerging from human IBD studies. For instance, patients with IBD (including ulcerative colitis and Crohn's disease) exhibit reduced microbial diversity, which is correlated with disease severity. This decrease in microbial richness is thought to stem from the loss of normal anaerobic bacteria, such as Bacteroides, Eubacterium, and Lactobacillus (Refs. 1-6). Research using mouse models has shown that inflammation increases oxygen and nitrate levels within the intestinal lumen, along with elevated host-derived electron acceptors, thereby promoting anaerobic respiration and overgrowth of Enterobacteriaceae (Ref 7). Consistent with these findings, our study observed a significant enrichment of Enterobacteriaceae in the inflamed intestines of Ctla-4-deficient zebrafish, which supporting the observations in mice. Despite this progress, the zebrafish model for intestinal inflammation remains under development, with limitations in available techniques for manipulating intestinal inflammation and reconstructing gut microbiota. These challenges hinder investigations into the association between intestinal inflammation and changes in microbial diversity. We plan to address these issues through ongoing technological advancements and further research. We thank the reviewer for their understanding.

References:

(1) Ott S J, Musfeldt M, Wenderoth D F, Hampe J, Brant O, Fölsch U R et al. (2004) Reduction in diversity of the colonic mucosa associated bacterial microflora in patients with active inflammatory bowel disease. Gut 53:685-693.

(2) Manichanh C, Rigottier-Gois L, Bonnaud E, Gloux K, Pelletier E, Frangeul L et al. (2006) Reduced diversity of faecal microbiota in Crohn's disease revealed by a metagenomic approach. Gut 55:205-211.

(3) Qin J J, Li R Q, Raes J, Arumugam M, Burgdorf K S, Manichanh C et al. (2010) A human gut microbial gene catalogue established by metagenomic sequencing. Nature 464:59-U70.

(4) Sha S M, Xu B, Wang X, Zhang Y G, Wang H H, Kong X Y et al. (2013) The biodiversity and composition of the dominant fecal microbiota in patients with inflammatory bowel disease. Diagn Micr Infec Dis 75:245-251.

(5) Ray K. (2015) IBD. Gut microbiota in IBD goes viral. Nat Rev Gastroenterol Hepatol 12:122.

(6) Papa E, Docktor M, Smillie C, Weber S, Preheim S P, Gevers D et al. (2012) Non-Invasive Mapping of the Gastrointestinal Microbiota Identifies Children with Inflammatory Bowel Disease. Plos One 7: e39242-39254.

(7) Hughes E R, Winter M G, Duerkop B A, Spiga L, de Carvalho T F, Zhu W H et al. (2017) Microbial Respiration and Formate Oxidation as Metabolic Signatures of Inflammation-Associated Dysbiosis. Cell Host Microbe 21:208-219.

Lines 270-272 say that interaction between Cd28/ctla-4 and Cd80/86 was demonstrated through bioinformatics, flow-cytometry, and Co-IP. Does this need to reference Fig S6D for the flow data? Figures 7F-G are very hard to read or comprehend as they are very small. Figure 7H is the most compelling evidence of this interaction and might stand out better if emphasized with a sentence referencing it on its own in the manuscript. 

In this study, we utilized an integrated approach combining bioinformatics prediction, flow cytometry, and co-immunoprecipitation (Co-IP) to comprehensively investigate and validate the interactions between Cd28/Ctla-4 and Cd80/86. Flow cytometry analysis, as depicted in Supplementary Figure 6D (revised as Supplementary Figure 8F), demonstrated the surface expression of Cd80/86 on HEK293T cells and quantified their interactions with Cd28 and Ctla-4. These experiments not only validated the interactions between Cd80/86 and Cd28/Ctla-4 but also revealed a dose-dependent relationship, providing robust supplementary evidence for the molecular interactions under investigation. Furthermore, in Figure 7F-G, the axis font sizes were enlarged to improve readability. Additionally, in response to reviewers' feedback, we have emphasized Figure 7H, which presents the most compelling evidence for molecular interactions, by including a standalone sentence in the text to enhance its prominence.

For Figure 7A-E, for non-immunologists, it is unclear what experiment was performed here - it would be helpful to add a 1-sentence summary of the assay to the main text or figure legend.

We apologize for this oversight. Figures 7A–E illustrate the functional assessment of the inhibitory role of Ctla-4 in Cd80/86 and Cd28-mediated T cell activation. A detailed description of the methodologies associated with Figures 7A–E is provided in the ‘Materials and Methods’ section of the revised manuscript.

For Figure 7F-G, it is extremely hard to read the heat map legends and the X and Y-axis. Also, what the heatmaps show and how that fits the overall narrative can be elaborated significantly.

We regret this oversight. To enhance clarity, we have increased the font size of the heatmap legends and the X and Y-axes, as shown in the following figure. Additionally, a detailed analysis of these figures is provided in lines 299–306 of the main text.

In general, the main text that accompanies Figure 7 should be expanded to more clearly describe these experiments/analyses and their results.

We have conducted a detailed analysis of the experiments and results presented in Figure 7. This analysis is described in lines 278-314.

Reviewer #2:

The scRNASeq assay is missing some basic characterization: how many WT and mutant fish were assayed in the experiment? how many WT and mutant cells were subject to sequencing? Before going to the immune cell types, are intestinal cell types comparable between the two conditions? Are there specific regions in the tSNE plot in Figure 4A abundant of WT or ctla-4 mutant cells?

In the experiment, we analyzed 30 wild-type and 30 mutant zebrafish for scRNA-seq, with an initial dataset comprising 8,047 cells in the wild-type group and 8,321 cells in the mutant group. Sample preparation details are provided on lines 620-652. Due to the relatively high expression of mitochondrial genes in intestinal tissue, quality control filtering yielded 3,263 cells in the wild-type group and 4,276 cells in the mutant group. Given that the intestinal tissues were dissociated using identical protocols, the resulting cell types are comparable between the two conditions. Both the wild-type and Ctla-4-deficient groups contained enterocytes, enteroendocrine cells, smooth muscle cells, neutrophils, macrophages, B cells, and a cluster of T/NK/ILC-like cells. Notably, no distinct regions were enriched for either condition in the tSNE plot (Figure 4A).

The cell proliferation experiment using PHA stimulation assay demonstrated the role of Ctla-4 in cell proliferation, while the transcriptomic evidence points towards activation rather than an overall expansion of T-cell numbers. This should be discussed towards a more comprehensive model of how subtypes of cells can be differentially proliferating in the disease model.

In the PHA-stimulated T cell proliferation assay, we aimed to investigate the regulatory roles of Ctla-4, Cd28, and Cd80/86 in T cell activation, focusing on validating Ctla-4's inhibitory function as an immune checkpoint. While our study examined general regulatory mechanisms, it did not specifically address the distinct roles of Ctla-4 in different T cell subsets. We appreciate the reviewer's suggestion to develop a more comprehensive model that elucidates differential T cell activation across various subsets in disease models. However, due to the nascent stage of research on fish T cell subsets and limitations in lineage-specific antibodies and transgenic strains, such investigations are currently challenging. We plan to pursue these studies in the future. Despite these constraints, our single-cell RNA sequencing data revealed an increased proportion of Th2 subset cells in Ctla-4-deficient zebrafish, as evidenced by elevated expression levels of Th2 markers (Il4, Il13, and Gata3) via RT-qPCR (see Figures S7B). Notably, recent studies in mouse models have shown that naïve T cells from CTLA-4-deficient mice tend to differentiate into Th2 cells post-proliferation, with activated Th2 cells secreting higher levels of cytokines like IL-4, IL-5, and IL-13, thereby exerting their effector functions (Refs. 1-2). Consequently, our findings align with observations in mice, suggesting conserved CTLA-4 functions across species. We have expanded the "Discussion" section to clarify these points.

References:

(1) Bour-Jordan H, Grogan J L, Tang Q Z, Auger J A, Locksley R M, Bluestone J A et al. (2003) CTLA-4 regulates the requirement for cytokine-induced signals in T_H2 lineage commitment. Nature Immunology 4: 182-188.

(2) Khattri Roli, Auger, Julie A, Griffin Matthew D, Sharpe Arlene H, Bluestone Jeffrey A et al. (1999) Lymphoproliferative Disorder in CTLA-4 Knockout Mice Is Characterized by CD28-Regulated Activation of Th2 Responses. The Journal of Immunology 162:5784-5791.

It would be nice if the authors could also demonstrate whether other tissues in the zebrafish have an inflammation response, to show whether the model is specific to IBD.

In addition to intestinal tissues, we also performed histological analysis on the liver of Ctla-4-deficient zebrafish. The results showed that Ctla-4 deficiency led to mild edema in a few hepatocytes, and lymphocyte infiltration was not significant. Compared to the liver, we consider intestinal inflammation to be more pronounced.

Some minor comments on terminology

(a) "multiomics" usually refers to omics experiments with different modalities (e.g. transcriptomics, proteomics, metabolomics etc), while the current paper only has transcriptomics assays. I wouldn't call it "multiomics" analysis.

We appreciate the reviewer's attention to this issue. The "multi-omics" has been revised to "transcriptomics".

(b) In several parts of the figure legend the author mentioned "tSNE nonlinear clustering" (Figures 4A and 5A). tSNE is an embedding method rather than a clustering method.

The "tSNE nonlinear clustering" has been revised to "tSNE embedding”.

(c) Figure 1E is a UMAP rather than tSNE.

The "tSNE" has been revised to "UMAP" in the figure legend in line 1043.

Reviewer #3: 

Line 28: The link is not directly reflected in this sentence describing CTLA-4 knockout mice.

We appreciate the reviewer for bringing this issue to our attention. We have expanded our description of CTLA-4 knockout mice on lines 77-84.

Line 80-83: There is a lack of details about the CTLA-4-deficient mice. The factor that Th2 response could be induced has been revealed in mouse model. See the reference entitled "CTLA-4 regulates the requirement for cytokine-induced signals in TH2 lineage commitment" published in Nature Immunology.

We thank the reviewer for providing valuable references. We have added descriptions detailing the differentiation of T cells into Th2 cells in CTLA-4-deficient mice on lines 78–81, and the relevant references have been cited in the revised manuscript.

To better introduce the CTLA-4 immunobiology, the paper entitled "Current Understanding of Cytotoxic T Lymphocyte Antigen-4 (CTLA-4) Signaling in T-Cell Biology and Disease Therapy" published in Molecules and Cells should be referred.

We have provided additional details on CTLA-4 immunology (lines 75-84) and have included the relevant reference in the revised manuscript.

In current results, there are many sentences that should be moved to the discussion, such as lines 123-124, lines 152-153, lines 199-200, and lines 206-207. So, the result sections just describe the results, and the discussions should be put together in the discussion.

We have relocated these sentences to the 'Discussion' section and refined the writing.

In the discussion, the zebrafish enteritis model, such as DSS/TNBS and SBMIE models, should also be compared with the current CTLA-4 knockout model. Also, the comparison between the current fish IBD model and the previous mouse model should also be included, to enlighten the usage of CTLA-4 knockout zebrafish IBD model.

We compared the phenotypes of our current Ctla-4-knockout zebrafish IBD model with other models, including DSS-induced IBD models in zebrafish and mice, as well as TNBS- and SBM-induced IBD models in zebrafish. The details are included in the "Discussion" section (lines 353-365).

As to the writing, the structure of the discussion is poor. The paragraphs are very long and hard to follow. Many findings from current results were not yet discussed. I just can't find any discussion about the alteration of intestinal microbiota.

In response to the reviewers' constructive feedback, we have revised and enhanced the discussion section. Furthermore, we have integrated the most recent research findings relevant to this study into the discussion to improve its relevance and comprehensiveness.

In the discussion, the aerobic-related bacteria in 16s rRNA sequencing results should be focused on echoing the histopathological findings, such as the emptier gut of CTLA-4 knockout zebrafish.

As mentioned above, the discussion section has been revised and expanded to provide a better understanding of the potential interplay among intestinal inflammatory pathology, gut microbiota alterations, and immune cell dysregulation in Ctla-4-deficient zebrafish. Furthermore, promising avenues for future research that warrant further investigation were also discussed.

In the current method, there are no descriptions for many used methods, which already generated results, such as WB, MLR, MST, Co-IP, AlphaFold2 prediction, and how to make currently used anti-zfCTLA4 antibody. Also, there is a lack of description of the method of the husbandry of knockout zebrafish line.

We regret these flaws. The methods section was inadvertently incomplete due to an error during the file upload process at submission. This issue has been rectified in the revised manuscript. Additionally, Ctla-4-deficient zebrafish were reared under the same conditions as wild-type zebrafish, and the rearing methods are now described in the "Generation of Ctla-4-deficient zebrafish" section of the Materials and Methods.

Line 360: the experimental zebrafish with different ages could be a risk for unstable intestinal health. See the reference entitled "The immunoregulatory role of fish-specific type II SOCS via inhibiting metaflammation in the gut-liver axis" published in Water Biology and Security. The age-related differences in zebrafish could be observed in the gut.

We appreciate the reviewers' reminders. The Ctla-4 mutant zebrafish used in our experiments were 4 months old, while the wild-type zebrafish ranged from 4 to 6 months old. These experimental fish were relatively young and uniformly distributed in age. During our study, we examined the morphological structures of the intestines in zebrafish aged 4 to 6 months and observed no significant abnormalities. These findings align with previous research indicating no significant difference in intestinal health between 3-month-old and 6-month-old wild-type zebrafish (Ref. 1). Consequently, we conclude that there is no notable aging-related change in the intestines of zebrafish aged 4 to 6 months. This reduces the risk associated with age-related variables in our study. We have added an explanation stating that the Ctla-4 mutant zebrafish used in the experiments were 4 months old (Line 449) in the revised manuscript.

Reference

(1) Shan Junwei, Wang Guangxin, Li Heng, Zhao Xuyang et al. (2023) The immunoregulatory role of fish-specific type II SOCS via inhibiting metaflammation in the gut-liver axis. Water Biology and Security 2: 100131-100144.

Section "Generation of Ctla-4-deficient zebrafish": There is a lack of description of PCR condition for the genotyping.

The target DNA sequence was amplified at 94 °C for 4 min, followed by 35 cycles at 94°C for 30 s, 58°C for 30 s and 72°C for 30 s, culminating in a final extension at 72 °C for 10 min. The polymerase chain reaction (PCR) conditions are described in lines 458-460.

How old of the used mutant fish? There should be a section "sampling" to provide the sampling details.

The "Sampling" information has been incorporated into the "Materials and Methods" section of the revised manuscript. Wild-type and Ctla-4-deficient zebrafish of varying months were housed in separate tanks, each labeled with its corresponding birth date. Experiments utilized Ctla-4-deficient zebrafish aged 4 months and wild-type zebrafish aged between 4 to 6 months.

Line 378-380: The index for the histopathological analysis should be detailed, rather than just provide a reference. I don't think these indexes are good enough to specifically describe the pathological changes of intestinal villi and mucosa. It is suggested to improve with detailed parameters. As described in the paper entitled "Pathology of Gastric Intestinal Metaplasia: Clinical Implications" published in Am J Gastroenterol., histochemical, normal gastric mucins are pH neutral, and they stain magenta with periodic acid-Schiff (PAS). In an inflamed gut, acid mucins replace the original gastric mucins and are stained blue with Alcian blue (AB). So, to reveal the pathological changes of goblet cells and involved mucin components, AB staining should be added. Also, for the number of goblet cells in the inflammatory intestine, combining PAS and AB staining is the best way to reveal all the goblet cells. In Figure 2, there were very few goblet cells. The infiltration of lymphocytes and the empty intestinal lumen could be observed. Thus, the ratio between the length of intestinal villi and the intestinal ring radius should calculated.

In response to the reviewers’ valuable suggestions, we have augmented the manuscript by providing additional parameters related to the pathological changes observed in the Ctlta-4-deficient zebrafish intestines, including the mucin component changes identified through PAS and AB-PAS staining, the variations in the number of goblet cells evaluated by AB-PAS staining, and the ratio of intestinal villi length to the intestinal ring radius, as illustrated in the following figures. These new findings are detailed in the "Materials and Methods" (lines 563-566) and "Results" (lines 143-146) sections, along with Supplementary Figure S3 of the revised manuscript.

Section "Quantitative real-time PCR": What's the machine used for qPCR? How about the qPCR validation of RNA seq data? I did not see any related description of data and methods for qPCR validation. In addition, beta-actin is not a stable internal reference gene, to analyze inflammation and immune-related gene expression. See the reference entitled "Actin, a reliable marker of internal control?" published in Clin Chim Acta. Other stable housekeeping genes, such as EF1alpha and 18s, could be better internal references.

RT-qPCR experiments were conducted using a PCR thermocycler device (CFX Connect Real-Time PCR Detection System with Precision Melt Analysis^TM Software, Bio-Rad, Cat. No. 1855200EM1). This information has been incorporated into lines 608-610 of the "Materials and Methods" section. In these experiments, key gene sequences of interest, including il13, mpx, and il1β, were extracted from RNA-seq data for RT-qPCR validation. To ensure accurate normalization, potential internal controls were evaluated, and β-actin was identified as a suitable candidate due to its consistent expression levels in the intestines of both wild-type and Ctla-4-deficient zebrafish. The use of β-actin as an internal control is further supported by its application in recent studies on intestinal inflammation (Refs 1–2).

References:

(1) Tang Duozhuang, Zeng Ting, Wang Yiting, Cui Hui et al. (2020) Dietary restriction increases protective gut bacteria to rescue lethal methotrexate-induced intestinal toxicity. Gut Microbes 12: 1714401-1714422.

(2) Malik Ankit, Sharma Deepika et al. (2023) Epithelial IFNγ signaling and compartmentalized antigen presentation orchestrate gut immunity. Nature 623: 1044-1052.

How to generate sCtla-4-Ig, Cd28-Ig and Cd80/86? No method could be found.

We apologize for the omission of these methods. The detailed protocols have now been added to the "Materials and Methods" section of the revised manuscript (lines 464-481).

Figure 5: As reviewed in the paper entitled "Teleost T and NK cell immunity" published in Fish and Shellfsh Immunology, two types of NK cell homologues have been described in fish: non-specific cytotoxic cells and NK-like cells. There is no NKT cell identified in the teleost yet. Therefore, "NKT-like" could be better to describe this cell type.

We refer to "NKT" cells as "NKT-like" cells, as suggested.

For the supplementary data of scRNA-seq, there lacks the details of expression level.

The expression levels of the corresponding genes are provided in Supplemental Table 4.

Supplemental Table 1: There are no accession numbers of amplified genes.

The accession numbers of the amplified genes are included in Supplemental Table 1.

The English needs further editing.

We have made efforts to enhance the English to meet the reviewers' expectations.

Line 32: The tense should be the past.

This tense error has been corrected.

Line 363-365: The letter of this approval should be provided as an attachment.

The approval document is provided as an attachment.

Line 376: How to distinguish the different intestinal parts? Were they judged as the first third, second third, and last third parts of the whole intestine?

The differences among the three segments of zebrafish intestine are apparent. The intestinal tube narrows progressively from the anterior to the mid-intestine and then to the posterior intestine. Moreover, the boundaries between the intestinal segments are well-defined, facilitating the isolation of each segment.

Line 404: Which version of Cytoscape was used?

The version of Cytoscape used in this study is 3.9.1. Information about the Cytoscape version is provided on line 603.

The product information of both percoll and cell strainer should be provided.

The information regarding Percoll and cell strainers has been added on lines 626 and 628, respectively.

Line 814: Here should be a full name to tell what is MST.

The acronym MST stands for "Microscale Thermophoresis", a technique that has been referenced on lines 1157-1158.

Author Response

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

In this manuscript, Xie et al report the development of SCA-seq, a multiOME mapping method that can obtain chromatin accessibility, methylation, and 3D genome information at the same time. This method is highly relevant to a few previously reported long read sequencing technologies. Specifically, NanoNome, SMAC-seq, and Fiber-seq have been reported to use m6A or GpC methyltransferase accessibility to map open chromatin, or open chromatin together with CpG methylation; Pore-C and MC-3C have been reported to use long read sequencing to map multiplex chromatin interactions, or together with CpG methylation. Therefore, as a combination of NanoNome/SMAC-seq/Fiber-seq and Pore-C/MC-3C, SCA-seq is one step forward. The authors tested SCA-seq in 293T cells and performed benchmark analyses testing the performance of SCA-seq in generating each data module (open chromatin and 3D genome). The QC metrics appear to be good and the methods, data and analyses broadly support the claims. However, there are some concerns regarding data analysis and conclusions, and some important information seems to be missing.

1. The chromatin accessibility tracks from SCA-seq seem to be noisy, with higher background than DNase-seq and ATAC-seq (Fig. 2f, Fig. 4a and Fig. S5). Also, SCA-seq is much less sensitive than both DNase-seq and ATAC-seq (Figs. 2a and 2b). This and other limitations of SCA-seq (high background, high sequencing cost, requirement of specific equipment, etc) need to be carefully discussed.

We thank the reviewer for the important comment about noisy GpC methylation signal in SCA-seq. We acknowledge that the SCA-seq signal presented in Fig. 2f, Fig. 4a, and Fig. S5 in our first draft was indeed noisy, as we present the raw 1D genomic signal. In this revision, we have taken steps to reduce the noise in GpC methylation signal by identifying the accessible regions on each segment of every single molecule. For each segment, we performed the sliding window analysis (50bp window sliding by a 10 bp step) with binomial test to identify accessible windows that significantly deviate from background GpC methylation ratio. The overlapping accessible windows (p < 0.05 for binomial test and contain at least two GpC sites) on the single fragments are merged as accessible region. Then we retain the GpC methylation signal inside the accessible region to reduce the background noise (Sfig 5ab). The details of the noise filtering steps are described in the Methods section (page 22 lines 13-23).

Visually, we can observe from the updated exemplary view of 1D signal track that the noise is dramatically reduced in filtered SCA-seq GpC methylation signal compared to the raw signal (Sfig5c). The clean SCA-seq GpC methylation 1D signals were also updated (Fig2f and Fig4a). We have observed an increase in the TSS enrichment score, which is a commonly used metric for assessing the signal-to-noise ratios in ATAC-seq data quality control. Specifically, the TSS enrichment score increased to 2.74 when using the filtered signal, compared to 1.93 when using the raw signal (Sfig5d). After noise filtering, 80% of SCA-seq 1D peaks overlaps with peaks called by ATAC-seq and/or DNase-seq (Fig2ab), compared to 74% from the raw signal in the first draft.

We thank the reviewer for raising up the concern about the sequencing cost and requirement of specific equipment. The sequencing cost is approximately 1300 USD per sample to sequence 30X depth human sample and obtain saturated GpC methylation signal (Sfig4d) as well as loop signal similar to the NGS-based Hi-C (Fig3gh). Considering that SCA-seq simultaneously provides higher-order chromatin structure and chromatin accessibility at single molecule resolution, we believe the cost is acceptable. However, it is worth noting that SCA-seq requires a regular Oxford nanopore sequencer with R9.4.1 chip, which is currently available but might be discontinued by Oxford Nanopore in the future. We have addressed all these concerns in the discussion section.

1. In Fig. 2f, many smaller peaks are present besides the major peaks. Are they caused by baseline DNA methylation? How many of the small methylation signals are called peaks? In Fig. 4a, it seems that the authors define many more enhancers from SCA-seq data than what will be defined from ATAC-seq or DHS. Are those additional enhancers false positives? Also, it is difficult to distinguish the gray "inaccessible segments" from the light purple "accessible segments.

We thank the reviewer for bringing up these concerns.

Regarding the smaller peaks in the 1D genomic GpC methylation signal, we have addressed this issue by implementing the noise filtering in this revision, the small peaks on 1D tracks are greatly reduced (Fig2f, Sfig5c). It is important to note that SCA-seq generates accessibility signals specifically on ligation junctions, which differs from the one-dimensional (1D) signals obtained through ATAC-seq or DNase-seq. The presence of remaining small peaks in the SCA-seq data can be attributed to the varied sequencing depth, which is influenced by the enriched spatial interactions occurring in regions of the genome that are enriched with ligation junctions. In general, the SCA-seq 1D peaks are well correlated with the high confidence peaks from 1D track of ATAC-seq and DNase-seq (Fig2b).

We apologize for the lack of clarity in our enhancer annotation. The enhancer regions were obtained from The Ensembl Regulatory Build (PMID: 25887522). We have now included this information in the method section (page 24 line 16).

We thank the reviewer for pointing out this visualization problem. The color scheme has been revised, with purple now representing the inaccessible segments and yellow representing the accessible segments.

1. For 3D genome analysis, it is important to provide information about data yield from SCA-seq. With 30X sequencing depth, how many contacts are obtained (with long-read sequencing, this should be the number of ligation junctions)? How is the number compared to Hi-C.

We thank the reviewer for raising up this crucial point about the sequencing yield that we missed. We have now included this information in the revised result section (page 11, lines 11-14).

We have checked the public data of a successful HEK293T Hi-C run (PMID: 34400762). The Hi-C experiment produced 699,464,541 reads (105G base), and we obtained 388,031,859 contacts.

From 100G bases of HEK293T SCA-seq data, we obtained 81,229,369 ligation junctions and 378,848,187 virtual pairwise contacts (3.8M pairwise contacts per Gb). The SCA-seq performance of virtual pairwise contact number per Gb is similar to that of PORE-C (PMID: 35637420).

1. Fig 3j. Because SCA-seq only do GpC methylation, the capability to detect the footprint at individual CTCF peaks depends on the density of GpC nearby. Have the authors taken GpC density into account when defining CTCF sites with or without footprint?

We appreciate the reviewer for bringing up the concern about the GpC site density at CTCF site. We would like to highlight that Battaglia et al. have demonstrated the feasibility of identifying transcription factor binding events using GpC labeling (PMID: 36195755). In our study, we have implemented a high-resolution sliding window approach to enhance the sensitivity of CTCF binding detection. We have taken GpC density into account by performing a sliding window (50 bp window, 10 bp step) binomial test on every single molecule overlapping with CTCF site to call accessible region. The detailed steps to call accessible region has been described in the answer of the first question. Based on the pattern in Fig3j, we identify CTCF footprints if the accessible regions are called nearby the CTCF sites (at least 20 bp away from the center of CTCF sites) but not on the CTCF sites.

To ensure that the GpC site density is sufficient for binomial test of each sliding window of the regions around CTCF site genome-wide, we examined the number of GpC sites in each window. Our analysis revealed that GpC sites are evenly distributed, and over 87% of the windows contain at least 2 GpC sites, which qualifies them for a binomial test (Author response image 1). This indicates that we are able to detect the CTCF footprint at most of the CTCF sites, taking into consideration the GpC density.

Author response image 1.

Genome wide GpC site density at CTCF site centered region. Distribution of the number of GpC sites (y-axis) at each 50 bp sliding window region (x-axis) was presented in violin plots.

1. This study only performs higher resolution chromatin interaction analysis based on individual read concatenates. It is unclear to me if the data have enough depth to perform loop analysis with Hi-C pipelines.

We thank the reviewer for highlighting this important concern about the depth of data for performing loop analysis. We have performed Aggregate peak analysis for SCA-seq and Hi-C side-by-side using hiccups function in Juicer (v1.9.9) (PMID: 27467249). We acknowledge that the level of loop signal enrichment is relatively weaker (one-fold less) in SCA-seq compared to Hi-C (Fig3h). This difference can be attributed to the lower sequencing yield per Gb in SCA-seq, which resulted in 4.93M pairwise contacts per Gb, compared to the 7M contacts per Gb in Hi-C. Despite this discrepancy, we were still able to observe the clear genome-wide loop enrichment pattern in SCA-seq (Fig3gh).

1. It appears that SCA-seq is of low efficiency in detecting chromatin interactions. As shown in Fig. S7a, 65.4% of sequenced reads contained only one restriction enzyme (RE) fragment/segment (with no genomic contact), which is much higher than that reported in published PORE-C methods. In addition, Fig. S7g is very confusing and in conflict with Fig. S7a. For example, in Fig. S7g, 21.4% and 22.2% of CSA-seq concatemers contain one and two segments, whereas the numbers are 65.4% and 14.7% in Fig. S7a, respectively. Please explain.

We apologize for the confusion in sfig7a and sfig7g.

Sfig7a was intended to illustrate the cardinality count of concatemers with only chr7 segments included, representing the intra-chromosome cardinality instead of the genome-wide cardinality. We have revised sfig7a and its corresponding figure legend to clarify that the figure describes segments of intra-chromosome interactions.

On the other hand, sfig7g shows the concatemers including both intra-chromosome and inter-chromosome segments, which explains the differences in the percentages of different cardinality ranges compared to Figure S7a. Moreover, the percentages reported in Figure S7g are similar to what is typically reported in PORE-C methods when considering both intra- and inter-chromosome interactions.

To provide a comprehensive view of the genome-wide concatemer cardinality distribution, we have also included a histogram in Fig3k, which demonstrates the detailed distribution of cardinality for genome-wide concatemers.

1. I disagree with the rationale of the entire Fig. S9. Biologically there is no evidence that chromatin accessibility will change due to genome interactions (the opposite is more likely), therefore the definition of "expected chromatin accessibility" is hard to believe. If the authors truly believe this is possible, they will need to test their hypothesis by deleting cohesin and check if the chromatin accessibility driven by "power center" are truly abolished. The math in Fig. S9 is also confusing. Firstly, the dimension of the contact matrix in Fig. S9 appears to be wrong, it should have 8 rows. Secondly, I don't understand why the interaction matrix is not symmetric. Third, if I understand correctly the diagonal of the matrix should be all 1, it is also hard to understand why the matrix only has 1, 0 or -1. It appears that the authors assume that the observed accessibility is a simple sum of the expected accessibility of all its interacting regions; this is wrong. In my opinion, the whole Fig. S9 should be deleted unless the authors can make sense of it and ideally also provide more evidence.

I apologize for any confusion caused by the rationale and figures in Fig. S9. The purpose of the hypothesis presented in the figure is to explore the potential relationship between chromatin accessibility and genome interactions. While there is currently no direct biological evidence supporting this hypothesis, it is a possibility that warrants further investigation.

Regarding the suggestion to delete Fig. S9 unless more evidence is provided, it is important to note that this paper primarily focuses on the methodology and theoretical framework. Experimental validation of the hypothesis falls outside the scope of this particular study.

We have made corrections to the schematic matrix in Fig. S9 to accurately represent the dimensions and symmetry. The numbers in the matrix represent mean accessible values of the contacts. Specifically, accessible-accessible contacts are represented by 2, accessible-inaccessible contacts are represented by 0, and inaccessible-inaccessible contacts are represented by -2.

Minor concerns:

1. The authors may want to clearly demonstrate the specificity and sensitivity of the ATAC part and the efficiency of the Hi-C part of SCA-seq.

We appreciate the reviewer’s suggestion to demonstrate the specificity and sensitivity of the ATAC-seq part and the efficiency of the Hi-C part in SCA-seq.

We considered the non-peak region genomic bins shared by ATAC-seq and DNase-seq as true negatives and the overlapping peaks of ATAC-seq and DNase-seq as true positives. Based on these criteria, the specificity of SCA-seq 1D peaks is calculated as TN / N, where TN represents the number of true negatives (89107) and N represents the sum of true negatives and false positives (89107 + 9345). The resulting specificity is 0.91. The sensitivity of SCA-seq 1D peaks is calculated as TP / P, where TP represents the number of true positives (33190) and P represents the sum of true positives and false negatives (33190 + 11758). The resulting sensitivity is 0.73.

We evaluate the efficiency of spatial interaction by the restriction enzyme digested fragments recovered in the pairwise contacts that contain ligation junctions. In SCA-seq, the efficiency is calculated as the number of dpnII digested fragments recovered by pairwise contacts (5625908) divided by the total number of in silico dpnII digested fragments (7127633). The resulting efficiency is 0.79.

We have now included this information in the revised result section (page 8 lines 15-18)

1. Fig 4g, colors with apparent differences might be used to clearly discriminate the three types of interactions (I-I, I-A and A-A).

We appreciate the reviewer for bringing up the issue regarding the visualization in Fig 4g. The color scheme has been revised, with purple now representing I-I interactions, orange representing I-A interactions, and red representing A-A interactions. We believe that these modifications have significantly improved the clarity.

1. Fig. 4c, when fitting an unknown curve, R-square becomes meaningless.

We appreciate the reviewer for pointing out the issue regarding the interpretation of R-square. We have removed the R-square value from Fig. 4c.

1. Fig 5a, "oCGIs comprised 65% CGIs that did not directly contact enhancers or promoters". Should it be "oCGIs comprised 65% of all CGIs"?

We appreciate the reviewer for pointing out the clarification needed in Fig 5a. We have revised the phrase in the figure legend to accurately state that “oCGIs comprised 65% of all CGIs”. Thank you for bringing this to our attention.

1. Page 15 lines 5-8, "By examining the methylation status on reads, as expected, these read segments demonstrated lower CpG methylation and higher chromatin accessibility (GpC methylation), which further supports their roles in gene activation (Fig 5b)". This statement seems to be inconsistent with the figure legend.

We appreciate the reviewer for pointing out the inconsistency in the legend of Fig 5b. We have revised the legend of Fig 5b to accurately highlight the low CpG methylation on oCGI regions. Thank you for bringing this to our attention.

1. Language editing and proof reading are needed.

I apologize for any errors or mistakes in the language. We have carefully reviewed the manuscript and made the necessary language editing and proofreading revisions to ensure its quality for publication.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1:

(1) The mechanism by which fenofibrate rescues memory loss in Kallistatin-transgenic mice is unclear. As a PPARalpha agonist, does fenofibrate target the Kallistatin pathway directly or indirectly? Please provide a discussion based on literature supporting either possibility.

Thank you for your important suggestion. Fenofibrate is indeed acting as a PPARα agonist. Fenofibrate has been shown to protect memory and cognitive function by downregulating α- and β-secretases[1]. Activation of PPARα can reduce Aβ plaques by upregulating ADAM10, thereby protecting memory and cognition[2]. Whereas, Fenofibrate can also act through a PPARα-independent pathway[3]. In our previous study, we proved that Fenofibrate can directly down-regulate the expression of Kallistatin in hepatocytes[4]. Here, our findings showed that Kallistatin induces cognitive memory deterioration by increasing amyloid-β plaques accumulation and tau protein hyperphosphorylation (Fig. 1-3), and Fenofibrate can directly down-regulate the serum level of Kallistatin (Fig. 8G). In addition, the expression of PPARα in the hippocampal tissue of Kallistatin (KAL-TG) mice showed no significant difference compared to the WT group (Author response image 1A-B). Therefore, we think Fenofibrate may improve memory and cognitive function at least in part through a PPARα-independent effect, which provides a new mechanism of Fenofibrate in AD with elevated Kallistatin levels.

Author response image 1.

(A-B) Protein levels of PPARα were tested by western blot analysis in hippocampal tissue, then statistically analyzed the above results.

(2) The current study exclusively investigated the hippocampus. What about other cognitive memory-related regions, such as the prefrontal cortex? Including data from these regions or discussing the possibility of their involvement could provide a more comprehensive understanding of the role of Kallistatin in memory impairment.

Thank you for your suggestion. In addition to hippocampal tissue analysis, we performed immunohistochemical detection of Aβ and phosphorylated Tau levels in the prefrontal cortex. Our findings revealed that KAL-TG mice exhibited significantly elevated Aβ and phosphorylated Tau levels in the prefrontal cortex compared to WT mice. These observations align with the pathological patterns observed in hippocampal tissues, demonstrating consistent neurodegenerative pathology across both the hippocampus and prefrontal cortex. The data for this part are seen as follows.

Author response image 2.

(A-B) Immunofluorescence staining of Aβ and phosphorylated tau (p-tau T231) was carried out in the prefrontal cortex tissue of KAL-TG and WT mice. Error bars represented the Standard Error of Mean (SEM); **p < 0.01. Scale bar, 100 μm.

(3) Fenofibrate rescued phenotypes in Kallistatin-transgenic mice while rosiglitazone, a PPARgamma agonist, did not. This result contradicts the manuscript's emphasis on a PPARgamma-associated mechanism. Please address this inconsistency.

Thank you for the reminder. In fact, our results showed a trend towards improved memory and cognitive function in KAL-TG mice treated with Rosiglitazone, although its effect is not as significant as that of Fenofibrate. Several studies have reported that Rosiglitazone has a beneficial effect on memory and cognitive function in mouse models of dementia, while these studies involve treatment periods of 3 to 4 months[5, 6], whereas our treatment period was only one month. Extending the treatment period with Rosiglitazone may result in a more pronounced improvement. In addition, Fenofibrate may have a PPAR-independent pathway by downregulating Kallistatin directly as discussed above and then show stronger effects.

(4) Most of the immunohistochemistry images are unclear. Inserts have similar magnification to the original representative images, making judgments difficult. Please provide larger inserts with higher resolution.

According to your suggestion, we provided larger inserts with higher resolution in Fig 3A and Fig 4B, as follows:

(5) The immunohistochemistry images in different figures were taken from different hippocampal subregions with different magnifications. Please maintain consistency, or explain why CA1, CA3, or DG was analyzed in each experiment.

Thank you for your advice. The trends of changes in different brain regions(including CA1, CA3, or DG) are consistent. Following your suggestion, we have now selected the DG region replaced the different hippocampal subregions with the DG area, and re-conducted the statistical analysis in Fig 5I & 6C, as follows. Due to the significant deposition of Aβ only in the CA1 region, Fig 2A was not replaced.

(6) Figure 5B is missing a title. Please add a title to maintain consistency with other graphs.

Thanks for your suggestion. We have added a title to Figure 5B, as follows:

(7) Please list statistical methods used in the figure legends, such as t-test or One-way ANOVA with post-hoc tests.

Thanks for your suggestion. We have listed the statistical methods used in the figure legends.

Reviewer #2:

(1) It was suggested that Kallistatin is primarily produced by the liver. The study demonstrates increased Kallistatin levels in the hippocampus tissue of AD mice. It would be valuable to clarify if Kallistatin is also increased in the liver of AD mice, providing a comprehensive understanding of its distribution in disease states.

Thank you for your suggestion. We extracted liver tissue from APP/PS1 mice, and the Western blot results indicated that the expression of Kallistatin in the liver of APP/PS1 mice was elevated, as follows:

Author response image 3.

(A-B) Protein levels of Kallistatin were tested by western blot analysis in the liver tissue, then statistically analyzed the above results. Error bars represented the Standard Error of Mean (SEM); **p < 0.01.

(2) Does Kallistatin interact directly with Notch1 ligands? Clarifying this interaction mechanism would enhance understanding of how Kallistatin influences Notch1 signaling in AD pathology.

Thank you for your suggestion. This study reveals that Kallistatin directly binds to Notch1 and contributes to the activation of the Noch1-HES1 signaling pathway. As for whether Kallistatin can bind to the ligands of Notch1, it needs to conduct further investigations in future studies. Our preliminary data showed that Jagged1 was upregulated in the hippocampal tissues of KAL-TG mice by qPCR and Western blot analyses.

Author response image 4.

Kallistatin promoted Notch ligand Jagged1 expression to activate Notch1 signaling. (A) QPCR analysis of Notch ligands (Dll1, Dll3, Jagged1, Jagged2) expression in the 9 months hippocampus tissue. (B) Western blotting analysis of Notch ligand Jagged1 expression in the hippocampus tissue. (C) Western blotting analysis of Notch ligand Jagged1 expression in the hippocampus primary neuron. β-actin served as the loading control. Error bars represented the Standard Error of Mean (SEM); *p < 0.05.

(3) Is there any observed difference in AD phenotype between male and female Kallistatin-transgenic (KAL-TG) mice? Including this information would address potential gender-specific effects on cognitive decline and pathology.

Thank you for your suggestion. Actually, we have previously used female mice for Morris Water Maze experiments, and the results showed that both male and female KAL-TG mice exhibited a phenotype of decreased memory and cognitive function compared to the gender-matched WT group, while there was no significant difference between male and female KAL-TG mice as follows:

Author response image 5.

(A-D) Behavioral performance was assessed through the Morris water maze test. (A) The escape latency time was presented during 1-5 days. (B-D) Cognitive functions were evaluated by spatial probe test on day 6, then analyzing each group of mice crossing platform times(B), time percent in the targeted area (C), and the path traces heatmap (D). Error bars represented the Standard Error of Mean (SEM); F represents Female, M represents Male, and TG refers to KAL-TG; *p < 0.05.

(4) It is recommended to include molecular size markers in Western blots for clarity and accuracy in protein size determination.

Thank you for your reminder. We have shown the molecular weight of each bolt.

(5) The language should be revised for enhanced readability and clarity, ensuring that complex scientific concepts are communicated effectively to a broader audience.

According to your suggestion, we have polished the article for enhancing readability and clarity.

Reviewer #3:

(1) The authors did not illustrate whether the protective effect of fenofibrate against AD depends on Kallistatin.

Thank you for your important suggestion. Fenofibrate is indeed acting as a PPARα agonist. Fenofibrate has been shown to protect memory and cognitive function by downregulating α- and β-secretases[1]. Activation of PPARα can reduce Aβ plaques by upregulating ADAM10, thereby protecting memory and cognition[2]. Whereas, Fenofibrate can also act through a PPARα-independent pathway[3]. In our previous study，we proved Fenofibrate can directly down-regulate the expression of KAL in hepatocytes[4]. Here, our findings showed that Kallistatin induces cognitive memory deterioration by increasing amyloid-β plaques accumulation and tau protein hyperphosphorylation (Fig. 1-3), and Fenofibrate can directly down-regulate the serum level of Kallistatin (Fig. 8G). In addition, the expression of PPARα in the hippocampal tissue of Kallistatin (KAL-TG) mice showed no significant difference compared to the WT group (Author response image 1-B). Therefore, we think Fenofibrate may improve memory and cognitive function at least in part through downregulatin Kallistatin. To conclusively determine whether fenofibrate’s therapeutic effects depend on Kallistatin, future studies should employ Kallistatin-knockout AD animal models to evaluate fenofibrate’s impact on cognitive and memory functions. These investigations will further clarify the mechanistic underpinnings of fenofibrate in AD therapy.

(2) The conclusions are supported by the results, but the quality of some results should be improved.

Thank you for your kind suggestion. We have updated the magnified images in the immunohistochemistry section of the article, ensuring that the fields of view for the immunohistochemistry are within the same brain region, and have shown the molecular weights in each bolt. Additionally, we have conducted a quantitative analysis of the protein levels in the Western blot results presented in Fig6&8.

(3) Figures 2c, 3c, and 4a present the Western blot results of p-tau from mice of different ages on one membrane, showing age-dependent expression. The authors analyzed the results of mice of different ages in one statistical chart, which will create ambiguity with the results of the representative images. For example, the expression of p-tau 396 in the blot was lower in the WT-12 M group than in the WT-9 M group (Figure 3c), which is contradictory to the statistical analysis.

Thank you for your reminder. The statistical presentation here does not match the figure. At that time, the WB experiments for the hippocampal tissue at each age group were conducted separately, and it was not appropriate to compare different age groups together. This graph cannot illustrate age dependency. We have replaced the statistical graph in Figure 3B&D, as follows:

(4) Figure 4b shows that KAL-TG-9 M had greater BACE1 expression than KAL-TG-12 M. Furthermore, the nuclei are not uniformly colored. Please provide more representative figures.

Thank you for your reminder. Due to the fact that these sets of data were not processed in a single batch, the ages in the graph are not comparable. Regarding the issue of inconsistent nuclear staining, we have provided another representative image from this group, as follows:

(5) Unclear why the BACE1 and Aβ levels seems less with KAL+shHES1 treatment than GFP+shNC treatment (Fig 6H)? This finding contradicts the conclusion.

Thank you for your reminder. This experiment was repeated three times, and here, we have represented the representative results along with the corresponding statistical data. There are no difference between KAL+shHES1 treatment and GFP+shNC treatment. We have updated the Fig. 6H.

(6) The Western blot results in figure 6e-h, 8h-i, and S3-S5 were not quantified.

Thank you for your reminder. We have added statistical graphs and original images of the pictures in figure 6e-h, 8h-i, and S3-S5.

(7) The authors did not provide the detection range of the Aβ42 ELISA kit.

Thank you for your suggestion. The Aβ42 ELISA kit is from the IBL, with the product number 27721. Its standard range is 1.56 - 100 pg/mL, and the sensitivity is 0.05 pg/mL.

(8)The authors did not specify the sex of the mice. This is important since sex could have had a dramatic impact on the results.

Thank you for your suggestion. The results we present in the text are all statistically obtained from male mice. Actually, we have previously used female mice for Morris Water Maze experiments, and the results showed that both male and female KAL-TG mice exhibited a phenotype of decreased memory and cognitive function compared to the gender-matched WT group, while there was no significant difference between male and female KAL-TG mice (Author response image 5).

Minor:

(1) In Figure 2b, there are no units for the vertical coordinates of the statistical graph.

Thank you for your reminder. We have added units for the vertical coordinates in Figure 2b.

(2) In Figure 2c, the left Y-axis title is lacking in the statistic chart.

Thank you for your reminder. We have added the left Y-axis title in the statistic chart.

Reference：

(1) Assaf N, El-Shamarka ME, Salem NA, Khadrawy YA, El Sayed NS. Neuroprotective effect of PPAR alpha and gamma agonists in a mouse model of amyloidogenesis through modulation of the Wnt/beta catenin pathway via targeting alpha- and beta-secretases. Progress in Neuro-Psychopharmacology and Biological Psychiatry 2020, 97: 109793.

(2) Rangasamy SB, Jana M, Dasarathi S, Kundu M, Pahan K. Treadmill workout activates PPARα in the hippocampus to upregulate ADAM10, decrease plaques and improve cognitive functions in 5XFAD mouse model of Alzheimer’s disease. Brain, Behavior, and Immunity 2023, 109: 204-218.

(3) Yuan J, Tan JTM, Rajamani K, Solly EL, King EJ, Lecce L, et al. Fenofibrate Rescues Diabetes-Related Impairment of Ischemia-Mediated Angiogenesis by PPARα-Independent Modulation of Thioredoxin-Interacting Protein. Diabetes 2019, 68(5): 1040-1053.

(4) Fang Z, Shen G, Wang Y, Hong F, Tang X, Zeng Y, et al. Elevated Kallistatin promotes the occurrence and progression of non-alcoholic fatty liver disease. Signal Transduct Target Ther 2024, 9(1): 66.

(5) Nelson ML, Pfeifer JA, Hickey JP, Collins AE, Kalisch BE. Exploring Rosiglitazone's Potential to Treat Alzheimer's Disease through the Modulation of Brain-Derived Neurotrophic Factor. Biology (Basel) 2023, 12(7).

(6) Pedersen WA, McMillan PJ, Kulstad JJ, Leverenz JB, Craft S, Haynatzki GR. Rosiglitazone attenuates learning and memory deficits in Tg2576 Alzheimer mice. Exp Neurol 2006, 199(2): 265-273.

Author response:

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

Reviewer #1 (Public review):

Weiler, Teichert, and Margrie systematically analyzed long-range cortical connectivity, using a retrograde viral tracing strategy to identify layer and region-specific cortical projections onto the primary visual, primary somatosensory, and primary motor cortices. Their analysis revealed several hundred thousand inputs into each region, with inputs originating from almost all cortical regions but dominated in number by connections within cortical sub-networks (e.g. anatomical modules). Generally, the relative areal distribution of contralateral inputs followed the distribution of corresponding ipsilateral inputs. The largest proportion of inputs originated from layer 6a cells, and this layer 6 dominance was more pronounced for contralateral than ipsilateral inputs, which suggests that these connections provide predominantly feedback inputs. The hierarchical organization of input regions was similar between ipsi- and contralateral regions, except for within-module connections, where ipsilateral connections were much more feed-forward than contralateral. These results contrast earlier studies which suggested that contralateral inputs only come from the same region (e.g. V1 to V1) and from L2/3 neurons. Thus, these results provide valuable data supporting a view of interhemispheric connectivity in which layer 6 neurons play an important role in providing modulatory feedback.

The conclusions of this paper are mostly well-supported by the data and analysis, but additional consideration of possible experimental biases is needed.

We thank the reviewer for their positive feedback on our manuscript.

Further discussion or analysis is needed about possible biases in uptake efficiency for different cell types. Is it possible that the nuclear retro-AAV has a tropism for layer 6 axons? Quantitative comparisons with results obtained with alternative methods such as rabies virus (Yao et al., 2023) or anterograde tracing (Harris et al., 2019) may be helpful for this.

We appreciate this technical comment. For the reasons indicated below we are confident that our AAV approach successfully and rather comprehensively labels inputs to the three target areas. Firstly, in the brains in which we injected our retrograde nuclear-AAV tracer into VISp, SSp-bfd or MOp we found several instances where layer 5 and/or layer 2/3 as was the dominant cortical projection layer (please see e.g. Figure 3 heatmaps). This was true for both ipsilateral and contralateral projection. 

Secondly, by way of comparison Yao et al., 2023 injected rabies virus into VISp (but not in SSp-bfd or MOp) and their results show notable similarities to ours: 1) They show that contralateral inputs to VISp (and higher visual areas) were mainly located in Layers 5 and 6. 2) Retrogradely labelled neurons in higher visual areas revealed anatomical hierarchy that reflects the known functional hierarchy of the mouse cortical visual system and that shown by our retro-AAV approach. Thus, as AAV and rabies based tracing lead to similar results, this is further evidence against bias via tropism of our AAV tracer. That said, direct comparisons of the results between our study and the Yao et al., 2023 study should be viewed with some caution since Yao et. al.  injected rabies virus into specific Cre-driver lines in which the rabies virus targets individual genetically defined cell types in specific layers. Importantly, because of the lack of a specific cre-driver line, L6 cortico-cortical (L6 CC) cells could not be targeted by their approach. Thus, the dataset in Yao et al., overlook the contribution of L6 CCs due to the lack of available Cre-lines. 

Thirdly, in a recent study (Weiler et al., 2024) we found that in a specific pathway (SSp-bfd→ VISp) both retro-AAV and the more traditional non-viral tracer cholera toxin subunit B (CTB) identified neurons in Layer 6 as the main source of projection neurons. The same results for the same pathway was shown by Bieler et al., 2019 (Bieler et al., 2017) using Fluorogold for retrograde tracing. Thus, the described dominance of Layer 6 projection neurons in specific pathways is likely not the result of a tropism of retro-AAV tracers. 

Please also see that we have now further extended the summary of these points in our revised manuscript in the discussion section (e.g. lines 457-463): 

Quantitative analysis of the injection sites should be included to account for possible biases. For example, L6 neurons are known to be the main target of contralateral inputs into the visual cortex (Yao et al., 2023). Thus, if the injections are biased towards or against layer 6 neurons, this may change the layer distribution of retrogradely labeled input cells. Comparison across biological replicates may help reveal sensitivity to particular characteristics of the injections.

In response to the reviewers' feedback, please see we have now quantified the injection volume per cortical layer, as shown in the revised Fig. S3D. Our results indicate that the injections were not biased toward Layer 6. Instead, the injected tracer volumes in Layers 1, 4, 5, and 6 were similar across all animals and injected areas. However, we observed that the injected tracer volume in Layer 2/3 tended to be higher than in other layers. Although the tracer volumes in Layers 2/3 appeared to be higher, the proportion of input neurons located in Layers 2/3 for most of the cortical projection areas was consistently lower than that from Layer 6. These findings provide strong evidence against injection bias towards L6 inputs.

The possibility of labelling axons of passage within the white matter should be addressed. This could potentially lead to false positive connections, contributing to the broad connectivity from most cortical regions that were observed.

For clarification, please see Fig.S2B in our revised manuscript. In this panel we plot the average percentage volume of the viral boli in the target areas and in all other nearby structures including the white matter. The percentage of virus injected into the white matter (WM) was 0.0824 ± 0.0759% for VISp and 0.0650 ± 0.0481 for SSp-bfd injections. Notably, injections into MOp showed no leakage into white matter (0%). These minimal volumes of virus in the white matter are unlikely to significantly influence the observed profile of widespread connectivity. Please see we have added a sentence to the Results section (lines 84-86) where we state that we only used brains that had a transduction of the white matter below 0.1%.

Reviewer #2 (Public review):

Summary:

Weiler et al use retrograde tracers, two-photon tomography, and automatic cell detection to provide a detailed quantitative description of the laminar and area sources of ipsi- and contralateral cortico-cortical inputs to two primary sensory areas and a primary motor area. They found considerable bilateral symmetry in the areas providing cortico-cortical inputs. However, although the same regions in both hemispheres tended to supply inputs, a larger proportion of inputs from contralateral areas originated from deeper layers (L5 and L6).

Strengths:

The study applies state-of-the-art anatomical methods, and the data is very effectively presented and carefully analyzed. The results provide many novel insights into the similarities and differences of inputs from the two hemispheres. While over the past decade there have been many studies quantitatively and comprehensively describing cortico-cortical connections, by directly comparing inputs from the ipsi and contralateral hemispheres, this study fills in an important gap in the field. It should be of great utility and an important reference for future studies on inter-hemispheric interactions.

We thank the reviewer for this encouraging feedback on our manuscript.

Weaknesses:

Overall, I do not find any major weakness in the analyses or their interpretation. However, one must keep in mind that the study only analyses inputs projecting to three areas. This is not an inherent flaw of the study; however, it warrants caution when extrapolating the results to callosal projections terminating in other areas. As inputs to two primary sensory areas and one is the primary motor cortex are studied, some of the conclusions could potentially be different for inputs terminating in high-order sensory and motor areas. Given that primary areas were injected, there are few instances of feedforward connections sampled in the ipsilateral hemisphere. The study finds that while ipsi-projections from the visual cortex to the barrel cortex are feedforward given its fILN values, those from the contralateral visual cortex are feedback instead. One is left to wonder whether this is due to the cross-modal nature of these particular inputs and whether the same rule (that contralateral inputs consistently exhibit feedback characteristics regardless of the hierarchical relationship of their ipsilateral counterparts with the target area,) would also apply to feedforward inputs within the same sensory cortices.

We acknowledge that what we find for primary sensory and motor target areas may not hold for other functionally different areas such as anterior cingulate cortex, retrosplenial cortex or frontal lobe that might be expected to receive strong feedforward cortical input. To begin to understand the organization of the global cortical input we have however first explored with primary sensory and motor areas. Please see that we have now added a sentence to the Discussion section of our manuscript that highlights the importance of investigating the hierarchical organization of intra and interhemispheric input onto higher cortical areas or within subregions of a given sensory area.

Another issue that is left unexplored is that, in the current analyses the barrel and primary visual cortex are analyzed as a uniform structure. It is well established that both the laminar sources of callosal inputs and their terminations differ in the monocular and binocular areas of the visual cortex (border with V2L). Similarly, callosal projections differ when terminating the border of S1 (a row of whiskers), and then in other parts of S1. Thus, some of the conclusions regarding the laminar sources of callosal inputs might depend on whether one is analyzing inputs terminating or originating in these border regions.

The aim of the present study was to analyse the global projectome to the VISp, SSp-bfd and MOp, irrespective of which subregions were included. Importantly, we purposely injected rather large bolus volumes to achieve large sample sizes of target neurons in each cortical layer.  For SSp-bfd, we utilised our previously reconstructed barrel map (Weiler et al., 2024) to precisely map our viral injection sites onto the barrels (Author response image 1). Analysis revealed that the six injection sites consistently encompassed 7–13 barrels (Author response image 1, three exemplary injection sites). Additionally, we determined the centres of mass for each injection site and mapped them onto the barrel map. Four of the injection sites were located in the lateral part of SSp-bfd, two in the central region, and none in the medial part. Notably, the injection sites within SSp-bfd exhibited significant overlap. As a result, a selective analysis of callosal projections targeting these injection sites would likely not yield distinct projection patterns, as the projectomes would inevitably include projections to surrounding barrels, leading to contamination.

Author response image 1.

Left: exemplary Injection sites mapped onto the 3D barrel map of SSp-bfd within the Mouse Allen Brain Atlas. Barrels were reconstructed using a specialized software as described previously (Weiler et al., 2024) Right: Centres of mass of all SSp-bfd injection sites mapped onto the 3D barrel map.

Due to the fact we covered a significant proportion of the respective target primary sensory area any further subdivision of these data is not possible and requires more tailored injections into clearly defined subareas. Investigating the separate projectomes onto these subregions (e.g. onto V1M and V1B) remains an important interesting research question that we, at least in part, will address in a future study.

Finally, while the paper emphasizes that projections from L6 "dominate" intra and contralateral cortico-cortical inputs, the data shows a more nuanced scenario. While it is true that the areas for which L6 neurons are the most common source of cortico-cortical projections are the most abundant, the picture becomes less clear when considering the number of neurons sending these connections. In fact, inputs from L2/3 and L5 combined are more abundant than those from L6 (Figure 3B), challenging the view that projections from L6 dominate ipsi- and contralateral projecting cortico-cortical inputs.

We agree in the case of the barrel cortex, layer 5 significantly contributes in terms of the number of brain areas projecting from within the ipsilateral and contralateral hemispheres. Please see we have replaced the term “dominates” in the title, abstract and in the manuscript where relevant.

Recommendations for the authors:  

Reviewer #1 (Recommendations for the authors):

The sections analyzing the role of L6 towards feedback (pg. 11-13, Figure 6) were a bit verbose and confusing to me. Three possible models are proposed:

(1) a decrease in L23 projections, (2) an increase in L56 projections, or (3) both.

However, what is being quantified appears to be the fractions inputs, with L23. L5, and L6 summing to 1. Thus, a decrease in L23 would necessarily result in an increase in L56 projections. It seems like it would make more sense to quantify the percent change in the total number of inputs (rather than fractional) from each layer so that the 3 models are actually independent possibilities.

The issue with the suggested analysis is that, with one exception (one area projecting to MOp), the number of projection neurons in contralateral areas is always ~60-80% lower compared to their ipsilateral counterparts. Consequently, this is also true for the number of projection neurons in the different cortical layers. Thus, quantifying the percentage change from the ipsilateral to the contralateral hemisphere in the total number of inputs from each layer will always result in negative values. 

Nevertheless, we addressed the reviewer’s issue by calculating the preservation index (1(ipsi-contra)/(ipsi+contra)) for the sensory-motor areas independently for the absolute number of neurons within L2/3, 5 and 6 for the cortical areas projecting to VISp, SSp-bfd and MOp (see Author response image 2). When analysing the shift from the ipsilateral to the contralateral hemisphere, we observed that significantly more projection neurons were preserved in L6 compared to L2/3 for VISp and SSp-bfd. This shows that the number of L6 projection neurons declines less from the ipsilateral to the contralateral hemisphere compared to L2/3. However, our focus was on the fraction of projection neurons within each layer relative to the other layers per hemisphere (see Fig.6 of our manuscript). This measure is critical for distinguishing between feedforward and feedback connectivity. Calculating the change for each layer independently unfortunately does not provide insights into this comparison, as it does not capture the relative distribution of projection neurons across layers, which is central to our analysis. Therefore, we chose to present the data as layer fractions normalised within each hemisphere separately, enabling a comparison of relative changes between hemispheres, as shown in Fig.6 in the manuscript. We agree that with our approach a decrease in the fraction of L2/3 neurons would necessarily lead to an increase in the fraction of L5+6 neurons. However, as we analysed the fractional change for L5 and L6 separately, we found that the fraction of projection neurons in L5 generally showed only minor changes, while the fraction of L6 projection neurons increased substantially (Fig.6C). In addition, excluding L5 from the ipsi- or contralateral default network had significant effects on the fILN in only a relatively small number of projection areas. Excluding L6 resulted in significant changes in many more projection areas than layer 5.

Author response image 2.

Preservation index for L2/3, L5 and L6 of the 24 sensory-motor areas projecting onto the three target areas VISp, SSp-bfd and MOp.

Reviewer #2 (Recommendations for the authors):

I feel that there are a few conclusions that could be strengthened in the paper:

(1) The laminar sources of callosal inputs and their terminations differ in the monocular and binocular areas of the visual cortex (border with V2L. Similarly, callosal inputs are different close to the border of S1 with S2 than in the rest of the barrel cortex. From the methods sections and Figure S2, it seems that some injections targeted the V1 binocular zone while others were aimed at the monocular zone. Thus, it would be of interest to compare the laminar distribution and fILM of the contra inputs in inputs to the binocular and monocular zones (and S1 border vs the rest, if possible within this dataset).

Please see the answer for the reviewer’s second point in the public review (above).

(2) The results are currently a bit unclear on whether the contra inputs reflect the cortical hierarchy. Figure 4E-F makes it clear that the ipsi and contra fILMs do not always match. However, it seems from the plots in Figure 4D and Figure S6 that, while the contra fILM values are always higher, there might be a correlation between the ipsi and contra fILM. This could be addressed by directly plotting contra vs ipsi fILM.

Similarly, it would be useful to directly address if there is any hint of the visual hierarchy, as calculated in Figure S5 for the contra inputs.

Regarding the first point of the reviewer: We appreciate this comment. We do indeed find a positive correlation between the fILN ipsilateral and fILN contralateral across the individual cortical areas for all three targets. (please see Author response image 3 below). This is indeed an interesting observation that indicates a high degree of preservation concerning the rank order of the anatomical hierarchy within the input arising from both hemispheres. Please see we have included this new figure 4F into the manuscript and added a sentence in the results (lines 282-288): 

Regarding the second point of the reviewer: For visual hierarchy, although weaker, we find that the hierarchical ranking of the higher cortical visual areas is preserved for the contralateral hemisphere (see Author response image 3 below). 

Author response image 3.

Rank ordered average fILN values (± sem) of higher visual cortical areas of the ventral and dorsal visual stream for the ipsilateral and contralateral hemisphere.

(3) I find the emphasis in the title and other parts of the paper on Layer 6 corticocortical cells dominating the anatomical organization of intra and interhemispheric feedback a bit of an overstatement. While it is true that the areas for which L6 is the most abundant source of cortico-cortical projections are the most abundant (Figure 3C), when just focusing on the number of neurons sending corticocortical connections (Figure 3B), this is less clear. Ipsi connections are roughly divided 1/3, 1/3 , 1/3 between L2/3 , L5 and L6. In the contra, while projections from L6 neurons are the most abundant, there are not a majority and are less than those of L2/3 and L5 together. I suggest revising the statement about L6 cells dominating cortico-cortical connections to more accurately reflect these nuances.

(4) The observations from Figure 3 discussed above suggest that L6 inputs dominate in areas with less abundant projections to the injected areas. Is this the case? Is the fraction of L6 inputs inversely correlated with the number of inputs from that area?

Please see the following correlation plots for the total number of inputs versus the fraction of L6 inputs per area for both the ipsilateral and contralateral hemisphere. We do find on the ipsilateral hemisphere a negative correlation between the total number of inputs and the L6 input fraction for VISp and to a lesser degree for SSp-bfd. Interestingly, we find the opposite correlation for the ipsilateral MOp, contralateral VISp, SSp-bfd and MOp (Author response image 4, Author response table 1). While this is an interesting finding, the correlations often appeared to be weak and often absent within the individual animals and across the three target areas (Author response table 1). Thus, these correlations are seemingly not a general feature of cortical connectivity.

Author response image 4.

Total number of cells versus fraction of cells within L6 per cortical brain area (average across animals) for the ipsilateral (top) and contralateral (bottom) hemisphere for the three target areas VISp, SSp-bfd and MOp.

Author response table 1: Respective correlations between total numbers of cells and fraction of cells within L6 per cortical brain area for the ipsilateral and contralateral hemisphere for the three target areas (significant correlations highlighted with green).

Minor issues:

(5) Where was the mouse in Figure 3A injected?

In this exemplary mouse the retrograde tracer was injected into VISp. We added this information in the Figure legend of Figure 3A1. 

(6) Clarify in panel 4F that the position of the circle corresponds to the area location.

Done as suggested. 

References

Bieler M, Sieben K, Cichon N, Schildt S, Röder B, Hanganu-Opatz IL. 2017. Rate and Temporal Coding Convey Multisensory Information in Primary Sensory Cortices. eNeuro 4. doi:10.1523/ENEURO.0037-17.2017

Weiler S, Rahmati V, Isstas M, Wutke J, Stark AW, Franke C, Graf J, Geis C, Witte OW, Hübener M, Bolz J, Margrie TW, Holthoff K, Teichert M. 2024. A primary sensory cortical interareal feedforward inhibitory circuit for tacto-visual integration. Nat Commun 15:3081. doi:10.1038/s41467-024-47459-2

Yao S, Wang Q, Hirokawa KE, Ouellette B, Ahmed R, Bomben J, Brouner K, Casal L, Caldejon S, Cho A, Dotson NI, Daigle TL, Egdorf T, Enstrom R, Gary A, Gelfand E, Gorham M, Griffin F, Gu H, Hancock N, Howard R, Kuan L, Lambert S, Lee EK, Luviano J, Mace K, Maxwell M, Mortrud MT, Naeemi M, Nayan C, Ngo N-K, Nguyen T, North K, Ransford S, Ruiz A, Seid S, Swapp J, Taormina MJ, Wakeman W, Zhou T, Nicovich PR, Williford A, Potekhina L, McGraw M, Ng L, Groblewski PA, Tasic B, Mihalas S, Harris JA, Cetin A, Zeng H. 2023. A whole-brain monosynaptic input connectome to neuron classes in mouse visual cortex. Nat Neurosci 26:350–364. doi:10.1038/s41593-022-01219-x

Author Response

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

eLife assessment

The authors analyzed the causative association between circulating immune cells and periodontitis, and reported three risk immune cells related to periodontitis. The significance of the findings is fundamental, which substantially advances our understanding of periodontitis. The strength of evidence is convincing.

Reviewer #1 (Public Review):

Ye et al. used Mendelian randomization method to evaluate the causative association between circulating immune cells and periodontitis and finally screened out three risk immune cells related to periodontitis. Overall, this is an important and novel piece of work that has the potential to contribute to our understanding of the causal relationship between circulating immune cells related to periodontitis. However, there are still some concerns that need to be addressed.

We sincerely appreciate the constructive feedback from the editor and reviewers, which has been instrumental in enhancing the quality of our manuscript.

(1) The authors used 1e-9 as the threshold to select effective instrumental variables (IVs), which should give the corresponding references. Meanwhile, the authors should test and discuss the potential impact of inconsistent thresholds for exposure (1e-9, 5e-6 were selected by the author respectively) and outcome IVs (5e-8) on the robustness of the results.

Thank you for your insightful comments. We have selected two GWAS databases as the data sources for the exposure group: the BCC Consortium with a sample size of 563,946, and the Sardinian cohort of 3,757. The considerable disparity in sample size between them may result in variations in outcomes, primarily showcased in the differences in positive SNP numbers. We, therefore, adopted an unconventional (non 5e-8) yet rigorously controlled screening strategy, an approach that is widely accepted in MR studies (Li et al., 2022; Liu et al., 2023). We believe that the present thresholds are sufficiently rigorous to guarantee the validity of the subsequent Mendelian randomization analysis.

However, employing two distinct methods in exposure screening is not typical, and we posit that this method can be viewed as an innovative strategy, providing a reference for future research dealing with two databases with significant discrepancies (Huang et al., 2023; Kong et al., 2023). As you perceptively noted, we acknowledge that this strategy may exert a certain influence on the research outcomes, and we have factored this potential limitation into our manuscript. “Third, the considerable variation in sample size between the two exposure databases contributes to the discrepancies in the number of positive SNPs. Despite our exploration of multiple selection thresholds for IVs, the inconsistency in screening methods and the discrepancy in the included SNPs could potentially introduce bias.” (Page 14)

As for the "outcome IVs with 5e-8" you mentioned, we didn't implement this screening threshold in the outcome IVs. Indeed, we applied the same screening criteria as specified at 5e-06 (refer to Stable 2). Is the statement that you're referring to the following: "Additionally, SNPs that displayed a direct association with the outcome would also be excluded to uphold the third MR assumption (P < 5e-8)" (Page 6)? In this context, we adopted a standard criterion in the IVs screening process to remove SNPs directly associated with the outcome.

Reference

Huang W, Wang Z, Zou C, Liu Y, Pan Y, Lu J, Zhou K, Jiao F, Zhong S, Jiang G. 2023. Effects of metabolic factors in mediating the relationship between Type 2 diabetes and depression in East Asian populations: A two-step, two-sample Mendelian randomization study. J Affect Disorders 335:120–128. doi:10.1016/j.jad.2023.04.114

Kong L, Ye C, Wang Y, Zheng J, Zhao Z, Li M, Xu Y, Lu J, Chen Y, Xu M, Wang W, Ning G, Bi Y, Wang T. 2023. Causal effect of lower birthweight on non-alcoholic fatty liver disease and mediating roles of insulin resistance and metabolites. Liver Int 43:829–839. doi:10.1111/liv.15532

Li P, Wang H, Guo L, Gou X, Chen G, Lin D, Fan D, Guo X, Liu Z. 2022. Association between gut microbiota and preeclampsia-eclampsia: a two-sample Mendelian randomization study. Bmc Med 20:443. doi:10.1186/s12916-022-02657-x Liu B, Lyu L, Zhou W, Song J, Ye D, Mao Y, Chen G-B, Sun X. 2023. Associations of the circulating levels of cytokines with risk of amyotrophic lateral sclerosis: a Mendelian randomization study. Bmc Med 21:39. doi:10.1186/s12916-023-02736-7

(2) What is the reference for selecting Smoking, Fasting plasma glucose, and BMI as covariates? They do not seem to be directly related to immune cells as confounding factors.

The variables of Smoking, Fasting Plasma Glucose (FPG), and Body Mass Index (BMI) are commonly used as covariates in multivariable Mendelian randomization studies (Kong et al., 2023; Liu et al., 2023). The association between Smoking, FPG, and BMI with immune cells may not be immediately apparent. However, these factors have been identified as potential confounders that could impact overall health, which in turn may indirectly modulate systemic immune responses, susceptibility, and inflammation.

(1) . Smoking: It has been well-documented that smoking can cause inflammation and impair immune function, thereby increasing individual's susceptibility to infections and diseases (Shiels et al., 2014). As such, smoking is recognized as a covariate that could potentially influence the outcomes of an investigation into immune cells.

(2) FPG: Elevated FPG levels indicate poor glycemic control, potentially leading to conditions like diabetes (Choi et al., 2018). Consequently, studies have demonstrated that elevated FPG levels can compromise the immune system's ability to combat infections.

(3) BMI: It is a measure of body fat that takes into account a person's weight and height. Both obesities, characterized by a high BMI, and underweights, characterized by a low BMI, have been associated with a range of health issues, inclusive of a compromised immune system (Piñeiro-Salvador et al., 2022). Consequently, BMI is factored in as a covariate in this study.

We have thus incorporated these factors as covariates in our study to mitigate their potential confounding effects. The selection of these covariates is primarily guided by previous research and established knowledge concerning the potential influences on immune function. We appreciate your query and will ensure to clarify this point in our revised manuscript. “We have incorporated covariates, including the number of cigarettes smoked, fasting plasma glucose (FPG) levels, and body mass index (BMI) into the MVMR analysis, given that these factors could indirectly affect systemic immune responses and inflammation (Liu et al., 2023).” (Page 6-7)

Reference

Choi S-C, Titov AA, Abboud G, Seay HR, Brusko TM, Roopenian DC, Salek-Ardakani S, Morel L. 2018. Inhibition of glucose metabolism selectively targets autoreactive follicular helper T cells. Nat Commun 9:4369. doi:10.1038/s41467-018-06686-0

Kong L, Ye C, Wang Y, Zheng J, Zhao Z, Li M, Xu Y, Lu J, Chen Y, Xu M, Wang W, Ning G, Bi Y, Wang T. 2023. Causal effect of lower birthweight on non-alcoholic fatty liver disease and mediating roles of insulin resistance and metabolites. Liver Int 43:829–839. doi:10.1111/liv.15532

Liu Y, Lai H, Zhang R, Xia L, Liu L. 2023. Causal relationship between gastro-esophageal reflux disease and risk of lung cancer: insights from multivariable Mendelian randomization and mediation analysis. Int J Epidemiol 52:1435–1447. doi:10.1093/ije/dyad090

Piñeiro-Salvador R, Vazquez-Garza E, Cruz-Cardenas JA, Licona-Cassani C, García-Rivas G, Moreno-Vásquez J, Alcorta-García MR, Lara-Diaz VJ, Brunck MEG. 2022. A cross-sectional study evidences regulations of leukocytes in the colostrum of mothers with obesity. BMC Med 20:388. doi:10.1186/s12916-022-02575-y

Shiels MS, Katki HA, Freedman ND, Purdue MP, Wentzensen N, Trabert B, Kitahara CM, Furr M, Li Y, Kemp TJ, Goedert JJ, Chang CM, Engels EA, Caporaso NE, Pinto LA, Hildesheim A, Chaturvedi AK. 2014. Cigarette smoking and variations in systemic immune and inflammation markers. J Natl Cancer Inst 106:dju294. doi:10.1093/jnci/dju294

(3) It is not entirely clear about the correction of P-value for the total number of independent statistical tests.

In our study, we used the Bonferroni correction to adjust the P-values for multiple comparisons. The adjusted P-value is calculated as the original P-value times the total number of independent statistical tests. Specifically, we applied multiple corrections in the following two aspects: First, we corrected the results of the FUSION algorithm in TWAS, with a correction value of P < 6.27 ×10-6 (0.05/7,890 genes) (Page 8). Second, we performed multiple corrections on the initial results of MR (P < 0.05/17 traits = 0.003). However, none of the results met the criteria after the correction, which is one of the limitations detailed in the discussion section of our study (Page 14).

(4) The author used whole blood data to apply FUSION algorithm. Although whole blood is a representative site, the authors should add FUSION testing of periodontally relevant tissues, such as oral mucosa.

We appreciate your insightful comments and suggestions. We concur that employing periodontally relevant tissues, like oral mucosa, for FUSION testing might yield more precise and pertinent results. However, in the Genotype-Tissue Expression project (GTEx) database, we could not find transcriptome data related to oral tissues, such as gums, oral mucosa, and alveolar bone (Review Table 1). Owing to the limitations of the database, in the context of our study, we primarily relied on whole blood data, given its availability and the extensive precedent documented in the literature for its utilization (Xu et al., 2023; Yuan et al., 2022).

We acknowledge that this is a limitation of our study and will certainly consider incorporating periodontally relevant tissues in our future research. In the revised manuscript, we have explicitly stated this limitation and underscored the necessity for additional studies to corroborate our findings with periodontally relevant tissues. Fifth, we relied on the whole blood data For FUSION algorithm due to the lack of transcriptome data associated with oral tissues (such as gums, oral mucosa, and alveolar bone) in the GTEx database. “Fifth, we relied on the whole blood data For FUSION algorithm due to the lack of transcriptome data associated with oral tissues (such as gums, oral mucosa, and alveolar bone) in the GTEx database. This has led to an excessive focus on systemic immunological changes, thereby overlooking the significance of alterations in local periodontal tissue immunity. Such an oversight could potentially compromise the precision and pertinence of our research findings.” (Page 15)

Author response table 1.

Organizations and Samplesize in the GTEx database

Reference

Xu J, Si H, Zeng Y, Wu Y, Zhang S, Shen B. 2023. Transcriptome-wide association study reveals candidate causal genes for lumbar spinal stenosis. Bone Joint Res 12:387–396. doi:10.1302/2046-3758.126.BJR-2022-0160.R1

Yuan J, Wang T, Wang L, Li P, Shen H, Mo Y, Zhang Q, Ni C. 2022. Transcriptome‐wide association study identifies PSMB9 as a susceptibility gene for coal workers’ pneumoconiosis. Environmental Toxicology 37:2103–2114. doi:10.1002/tox.23554

(5) The authors chose gingival hyperplasia as a secondary validation phenotype of periodontitis in this study. However, gingival recession, as another important phenotype associated with periodontitis, should also be tested and discussed.

We appreciate your insightful feedback highlighting the significance of incorporating gingival recession as a phenotype in periodontitis studies. Our emphasis on gingival hyperplasia in the study was primarily dictated by the initial study design and the data available from FinnGen R9K11. Notwithstanding the lack of gingival recession data in the available databases, we identified chronic gingivitis data in an earlier version of the Finnish database (FinnGen R5K11) as an alternative. We performed a Mendelian Randomization analysis on this dataset, with the results integrated into Supplementary Table 10. Concurrently, Table 1, Supplementary Table 1, Figure 4, and the corresponding descriptions in the manuscript were updated. We trust this adjustment can address the limitations identified in our research. We are confident that this not only augments the comprehensiveness of our study but also fosters a more holistic comprehension of periodontal disease.

(6) This study used GLIDE data as a replicated validation, but the results were inconsistent with FinnGen's dataset.

Thank you for your insightful comments and for bringing this issue to our attention. Indeed, it is of utmost importance to ensure the validity and reliability of our findings across various datasets. The observed inconsistency between the GLIDE data and FinnGen's dataset could be attributed to several reasons.

Firstly, this discrepancy might originate from the differences in population composition. The former is grounded on a comprehensive meta-analysis of cohorts focusing on periodontitis, whereas the latter utilizes a dataset from a full-phenotype cohort. In the former, the ratio of periodontitis to the control groups is approximately 1:2. In contrast, the ratio in the latter seems to be minuscule. The sample size in the FinnGen data may not suffice to detect the effects observed in the GLIDE dataset, given that larger exposure sizes enhance the ability to detect genuine associations.

Moreover, the heterogeneity of periodontitis can potentially result in variable outcomes. Phenotypic definition methods differ between the two databases. The GLIDE database diagnoses based on the criteria of Centers for Disease Control and Prevention/American Academy of Periodontology (CDC/AAP) and Community Periodontal Index (CPI) for physical signs. While the FinnGen database adopts the International Classification of Diseases (ICD) 10 standard for a comprehensive diagnosis. The former database employs a more practical yet broader standard for periodontitis, which might encompass pseudo-periodontitis.

Finally, the observed differences could be attributed to the variations in immune responses at distinct stages of periodontitis. During the initial stages of periodontitis, neutrophils and macrophages primarily mediate the immune response. With the progression of the disease, the involvement of T cells and B cells increases, thereby leading to a more intricate immune response (Darveau, 2010). Besides, the immune system's response to these oral health conditions is not uniform and can be influenced by multiple factors, including the individual's overall health, genetics, and lifestyle, potentially impacting the results (Hung et al., 2023).

Reference

Darveau RP. 2010. Periodontitis: a polymicrobial disruption of host homeostasis. Nat Rev Microbiol 8:481–490. doi:10.1038/nrmicro2337

Hung M, Kelly R, Mohajeri A, Reese L, Badawi S, Frost C, Sevathas T, Lipsky MS. 2023. Factors Associated with Periodontitis in Younger Individuals: A Scoping Review. J Clin Med 12:6442. doi:10.3390/jcm12206442

Reviewer #2 (Public Review):

This manuscript presents a well-designed study that combines multiple Mendelian randomization analyses to investigate the causal relationship between circulating immune cells and periodontitis. The main conclusions of the manuscript are appropriately supported by the statistics, and the methodologies used are comprehensive and rigorous.

These findings have significant implications for periodontal care and highlight the potential for systemic immunomodulation management on periodontitis, which is of interest to readers in the fields of periodontology, immunology, and epidemiology.

We greatly appreciate the positive feedback and valuable insights provided by the reviewer, which have significantly contributed to the improvement of our manuscript.

Reviewer #2 (Recommendations for The Authors):

*Abstract

Line 30-32: "Two-sample bidirectional univariable MR followed by sensitivity testing, multivariable MR, subgroup analysis, and the Bayesian model averaging (MR-BMA) were performed to explore the causal association between them. " What does the term "them" refer to here, please clarify it. The research method here is unclear, please reorganize it.

Line 39: "S100A9 and S100A12" here should be italic.

We appreciate your meticulous suggestions and have revised the methods section accordingly. Additionally, the two genes have been highlighted in italics for emphasis.

"Univariable MR, multivariable MR, subgroup analysis, reverse MR, and Bayesian model averaging (MR-BMA) were utilized to investigate the causal relationships. Furthermore, transcriptome-wide association study (TWAS) and colocalization analysis were deployed to pinpoint the underlying genes." (Page 1)

Introduction

Line 78-80: "As reported, the number of immune cells in periodontal tissue changes as periodontitis progresses, featuring an increase in monocytes, and B cells and a decrease in T cells." Does the author mean that both monocytes and B cells increase as periodontitis progresses?

We are grateful for your meticulous reading and perceptive inquiries. We would like to confirm the accuracy of your understanding. In lines 78-80, our intended message was to communicate that with the progression of periodontitis, there is an increase in both monocytes and B cells in the periodontal tissue. This represents a typical immune response to the infection, where these cells play a pivotal role in counteracting periodontal pathogens. To enhance clarity, we have revised these lines in the manuscript as follows:

"With the progression of periodontitis, there is a significant alteration in the quantity of immune cells present within the periodontal tissue. Specifically, an increase in the count of both monocytes and B cells is observed, whereas a decrease is noted in the count of T cells." (Page 3)

Method

Line 164-165: "As the main test, the MVMR-IVW method, offered by the MVMR-least absolute shrinkage and selection operator (MVMR-LASSO), and the MVMR-Egger method were chosen." The author's expression here is ambiguous.

In response to your comment on the ambiguity in lines 164-165, we have revised the sentence for clarity. We hope this addresses your concern and clarifies our point more effectively.

"The MVMR-IVW method was utilized as the primary test, supplemented by the MVMR-least absolute shrinkage and selection operator (MVMR-LASSO) and the MVMR-Egger method." (Page 7)

Table 1: FinnGen has a greater sample size and more SNPs than GLIDE; why do authors choose the latter as the primary analysis?

Our choice to utilize GLIDE as the primary analysis tool, instead of FinnGen, was mainly guided by the specific research question we aimed to address. Despite FinnGen offering a larger sample size and more SNPs, GLIDE offers a more specialized and targeted dataset that suits the unique requirements of our study. In most MR studies, a similar strategy is adopted, wherein a large database of disease GWAS meta is utilized for exploration, followed by validation in full phenotype cohort (such as UKBiobank and FinnGen) (Liu et al., 2023; Yuan et al., 2023). To summarize, the reasons may primarily include the following:

Firstly, GLIDE offers a concentrated and targeted methodology for examining genetic data pertinent to periodontitis. This dataset is grounded in a comprehensive meta-analysis of cohorts centered on periodontitis, wherein the ratio of periodontitis cases to control groups is approximately 1:2. Conversely, the proportion in FinnGen seems to be negligible, given that it employs a dataset derived from a comprehensive phenotype cohort. Consequently, employing the GLIDE database as a primary investigative tool can generate more abundant genetic information associated with periodontitis.

Furthermore, the methodological facets of GLIDE align more accurately with the analytical framework of our study. For instance, the diagnostic criteria methods vary between the two databases. The GLIDE database derives its basis from the Centers for Disease Control and Prevention/American Academy of Periodontology (CDC/AAP) and Community Periodontal Index (CPI) for physical indicators. In contrast, the FinnGen database employs the International Classification of Diseases (ICD) 10 standard for an exhaustive diagnosis. The former adopts a more pragmatic, yet broader, standard for diagnosing periodontitis. The latter continues to use concepts of diseases such as "chronic periodontitis", which have been replaced by "periodontitis" in the latest disease classification from the "2017 World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions" in the periodontal field (Caton et al., 2018).

Reference

Caton JG, Armitage G, Berglundh T, Chapple ILC, Jepsen S, Kornman KS, Mealey BL, Papapanou PN, Sanz M, Tonetti MS. 2018. A new classification scheme for periodontal and peri-implant diseases and conditions - Introduction and key changes from the 1999 classification. J Clin Periodontol 45 Suppl 20:S1–S8. doi:10.1111/jcpe.12935

Liu Y, Lai H, Zhang R, Xia L, Liu L. 2023. Causal relationship between gastro-esophageal reflux disease and risk of lung cancer: insights from multivariable Mendelian randomization and mediation analysis. Int J Epidemiol 52:1435–1447. doi:10.1093/ije/dyad090

Yuan S, Xu F, Li X, Chen J, Zheng J, Mantzoros CS, Larsson SC. 2023. Plasma proteins and onset of type 2 diabetes and diabetic complications: Proteome-wide Mendelian randomization and colocalization analyses. Cell Rep Med 4:101174. doi:10.1016/j.xcrm.2023.101174

Result

Line 224: "The observed significant results remained robust after removing pleiotropic SNPs." It is not clear what the authors mean by "remain robust".

Line 229-231: "The causal relationship between neutrophils and periodontitis remained stable with no evidence of heterogeneity or pleiotropy." It is also not clear what the authors mean by "remain stable". How does the author get to the conclusion that there is no evidence of heterogeneity or pleiotropy?

Figure S5: Please offer a brief explanation on how to investigate outlier or influential changes using scatter plots and Cochran's Q test and Cook's distance.

Line 224: We apologize for the confusion caused by the term "remain robust". In the revised manuscript, we clarified this by stating, "The observed significant results are considered 'robust' if the effect of sensitivity analyses was identical to that of Inverse Variance Weighted (IVW) method, yielding a P-value less than 0.05." (Page 6)

Line 229-231: We used the terms "remain stable" and "remain robust" interchangeably to express the same idea. To clarify, we have now unified the expression in the revised manuscript. As for the conclusion of "no evidence of heterogeneity or pleiotropy", it is derived from the results of Cochran's Q and Egger's intercept tests (P<0.05). We have added this explanation to the revised manuscript for better clarity.

Figure S5: In the revised manuscript and Table, we have provided a succinct explanation regarding the investigation of outliers or influential changes as follows: " A genetic variant was defined as either an outlier or an influential variant if it possessed a q-value exceeding 10 or if its Cook's distance surpassed the median of the corresponding F-distribution. " (Page 7)

We have made all the necessary changes in the revised manuscript based on your comments. We hope our responses and revisions adequately address your concerns.

Discussion

I have consulted several pieces of literature to ensure a thorough explanation, which may be helpful for your writing.

(1) Hajishengallis G, Li X, Divaris K, Chavakis T. Maladaptive trained immunity and clonal hematopoiesis as potential mechanistic links between periodontitis and inflammatory comorbidities. Periodontol 2000. 2022;89(1):215-230. doi:10.1111/prd.12421

(2) Hajishengallis G, Chavakis T. Mechanisms and Therapeutic Modulation of Neutrophil-Mediated Inflammation. J Dent Res. 2022;101(13):1563-1571. doi:10.1177/00220345221107602

We appreciate your valuable feedback and the additional references you provided to enrich our manuscript. Upon receiving your comments, we have meticulously reviewed and incorporated the suggested literature into our revised manuscript. These references have furnished insightful information, which has been assimilated into the revised manuscript (Page 12) to enhance the explanation of the mechanisms of neutrophil-mediated inflammation and the potential association between periodontitis and inflammatory comorbidities.

"The quantity and functionality of neutrophils both act as critical indicators of inflammation severity. The reduction in neutrophil count and inflammatory mediators, observed after successful periodontitis treatment, suggests a reduction in systemic inflammation (Hajishengallis , 2022)." (Page 12)

"Trained myeloid cells have the potential to amplify the functionality of neutrophils, thereby fortifying the body's defense against subsequent infections. Nevertheless, within the framework of chronic inflammation, these cells could potentially intensify tissue damage (Hajishengallis and Chavakis, 2022)." (Page 12)

Author response:

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

eLife Assessment

This manuscript reveals important insights into the role of ipsilateral descending pathways in locomotion, especially following unilateral spinal cord injury. The study provides solid evidence that this method improves the injured side's ability to support weight, and as such the findings may lead to new treatments for stroke, spinal cord injuries, or unilateral cerebral injuries. However, the methods and results need to be better detailed, and some of the statistical analysis enhanced.

Thank you for your assessment. We incorporated various text improvements in the final version of the manuscript to address the weaknesses you have pointed out. The specific improvements are outlined below.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This manuscript provides potentially important new information about ipsilateral cortical impact on locomotion. A number of issues need to be addressed.

Strengths:

The primary appeal and contribution of this manuscript are that it provides a range of different measures of ipsilateral cortical impact on locomotion in the setting of impaired contralateral control. While the pathways and mechanisms underlying these various measures are not fully defined and their functional impacts remain uncertain, they comprise a rich body of results that can inform and guide future efforts to understand cortical control of locomotion and to develop more effective rehabilitation protocols.

Weaknesses:

(1) The authors state that they used a cortical stimulation location that produced the largest ankle flexion response (lines 102-104). Did other stimulation locations always produce similar, but smaller responses (aside from the two rats that showed ipsilateral neuromodulation)? Was there any site-specific difference in response to stimulation location?

We derived motor maps in each rat, akin to the representation depicted in Fig 6. In each rat, alternative cortical sites did, indeed, produce distal or proximal contralateral leg flexion responses. Distal responses were more likely to be evoked in the rostral portion of the array, similarly to proximal responses early after injury. This distribution in responses across different cortical sites is reported in this study (Fig. 6) and is consistent with our prior work. The Results section has been revised to provide additional clarification of the passage you indicated and context for the data presented in Figure 6:

On page 4, we have clarified: “Stimulation through these channels produced a strong whole-leg flexion movement, with an evident distal component. From visual inspection, all responding electrodes in the array produced contralateral leg flexion, although with different strength of contraction for a fixed stimulation intensity (100μA). Moreover, some sites did not present a distal movement component, failing in eliciting ankle flexion and resulting in a generally weaker proximal flexion.”

On page 12, we have further noted: “By visually inspecting the responses elicited by stimulation delivered through each of the array electrodes, we categorized movements as proximal or distal. This classification was based on whether the ankle participated in the evoked response or if the movement was restricted to the proximal hindlimb. Each leg was scored independently.”

(2) Figure 2: There does not appear to be a strong relationship between the percentage of spared tissue and the ladder score. For example, the animal with the mild injury (based on its ladder score) in the lower left corner of Figure 2A has less than 50% spared tissue, which is less spared tissue than in any animal other than the two severe injuries with the most tissue loss. Is it possible that the ladder test does not capture the deficits produced by this spinal cord injury? Have the authors looked for a region of the spinal cord that correlates better with the deficits that the ladder test produces? The extent of damage to the region at the base of the dorsal column containing the corticospinal tract would be an appropriate target area to quantify and compare with functional measures.

In Fig. S6 of our 2021 publication "Bonizzato and Martinez, Science Translational Medicine", we investigated the predictive value of tissue sparing in specific sub-regions of the spinal cord for ladder performance. Among others, we examined the correlation between the accuracy of left leg ladder performance in the acute state and the preservation of the corticospinal tract (CST). Our results indicated that dorsal CST sparing serves as a mild predictor for ladder deficits, confirming the results obtained in this study.

(3) Lines 219-221: The authors state that "phase-coherent stimulation reinstated the function of this muscle, leading to increased burst duration (90{plus minus}18% of the deficit, p=0.004, t-test, Fig. 4B) and total activation (56{plus minus}13% of the deficit, p=0.014, t-test, Fig. 3B). This way of expressing the data is unclear. For example, the previous sentence states that after SCI, burst duration decreased by 72%. Does this mean that the burst duration after stimulation was 90% higher than the -72% level seen with SCI alone, i.e., 90% + -72% = +18%? Or does it mean that the stimulation recovered 90% of the portion of the burst duration that had been lost after SCI, i.e., -72% * (100%-90%)= -7%? The data in Figure 4 suggests the latter. It would be clearer to express both these SCI alone and SCI plus stimulation results in the text as a percent of the pre-SCI results, as done in Figure 4.

Your assessment is correct; we intended to report that the stimulation recovered 90% of the portion of the burst duration that had been lost after SCI. This point has been clarified (see page 9):

“…leading to increased burst duration (recovered 90±18% of the lost burst duration, p=0.004, t-test, Fig. 4B) and total activation (recovered 56±13% of the total activation, p=0.014, t-test, Fig. 3B)”

(4) Lines 227-229: The authors claim that the phase-dependent stimulation effects in SCI rats are immediate, but they don't say how long it takes for these effects to be expressed. Are these effects evident in the response to the first stimulus train, or does it take seconds or minutes for the effects to be expressed? After the initial expression of these effects, are there any gradual changes in the responses over time, e.g., habituation or potentiation?

The effects are immediately expressed at the very first occurrence of stimulation. We never tested a rat completely naïve to stimuli, as each treadmill session involves prior cortical mapping to identify a suitable active site for involvement in locomotor experiments. Yet, as demonstrated in Supplementary Video 1 accompanying our 2021 publication on contralateral effects of cortical stimulation, "Bonizzato and Martinez, Science Translational Medicine," the impact of phase-dependent cortical stimulation on movement modulation is instantaneous and ceases promptly upon discontinuation of the stimulation. We did not quantify potential gradual changes in responsiveness over time, but we cannot exclude that for long stimulation sessions (e.g., 30 min or more), stimulus amplitude may need to be slightly increased over time to compensate habituation.

(5) Awake motor maps (lines 250-277): The analysis of the motor maps appears to be based on measurements of the percentage of channels in which a response can be detected. This analytic approach seems incomplete in that it only assesses the spatial aspect of the cortical drive to the musculature. One channel could have a just-above-threshold response, while another could have a large response; in either case, the two channels would be treated as the same positive result. An additional analysis that takes response intensity into account would add further insight into the data, and might even correlate with the measures of functional recovery. Also, a single stimulation intensity was used; the results may have been different at different stimulus intensities.

We confirm that maps of cortical stimulation responsiveness may vary at different stimulus amplitudes. To establish an objective metric of excitability, we identified 100µA as a reliable stimulation amplitude across rats and used this value to build the ipsilateral motor representation results in Figure 6. This choice allows direct comparison with Figure 6 of our 2021 article, related to contralateral motor representation. The comparison reveals a lack of correlation with functional recovery metrics in the ipsilateral case, in contrast to the successful correlation achieved in the contralateral case.

Regarding the incorporation of stimulation amplitudes into the analysis, as detailed in the Method section (lines 770-771), we systematically tested various stimulation amplitudes to determine the minimal threshold required for eliciting a muscle twitch, identified as the threshold value. This process was conducted for each electrode site.

Upon reviewing these data, we considered the possibility of presenting an additional assessment of ipsilateral cortical motor representation based on stimulation thresholds. However, the representation depicted in the figure did not differ significantly from the data presented in Figure 6A. Furthermore, this representation introduced an additional weakness, as it was unclear how to represent the absence of a response in the threshold scale. We chose to arbitrarily designate it as zero on the inverse logarithmic scale, where, for reference, 100 µA is positioned at 0.2 and 50 µA at 0.5.

In conclusion, we believe that the conclusions drawn from this analysis align substantially with those in the text. The addition of the threshold analysis, in our assessment, would not contribute significantly to improving the manuscript.

Author response image 1.

Threshold analysis

Author response image 2.

Occurrence probability analysis, for comparison.

(6) Lines 858-860: The authors state that "All tests were one-sided because all hypotheses were strictly defined in the direction of motor improvement." By using the one-sided test, the authors are using a lower standard for assessing statistical significance that the overwhelming majority of studies in this field use. More importantly, ipsilateral stimulation of particular kinds or particular sites might conceivably impair function, and that is ignored if the analysis is confined to detecting improvement. Thus, a two-sided analysis or comparable method should be used. This appropriate change would not greatly modify the authors' current conclusions about improvements.

Our original hypothesis, drawn from previous studies involving cortical stimulation in rats and cats, as well as other neurostimulation research for movement restoration, posited a favorable impact of neurostimulation on movement. Consistent with this hypothesis, we designed our experiments with a focus on enhancing movement, emphasizing a strict direction of improvement.

It's important to note that a one-sided test is the appropriate match for a one-sided hypothesis, and it is not a lower standard in statistics. Each experiment we conducted was constructed around a strictly one-sided hypothesis: the inclusion of an extensor-inducing stimulus would enhance extension, and the inclusion of a flexion-inducing stimulus would enhance flexion. This rationale guided our choice of the appropriate statistical test.

We acknowledge your concern regarding the potential for ipsilateral stimulation to have negative effects on locomotion, which might not be captured when designing experiments based on one-sided hypotheses. That is, when hypothesizing that an extensor stimulus would enhance extension (a one-sided hypothesis) in a functional task, and finding an opposite result (inhibition), statistical rigor would impose that we cannot present that result as significant. This concern is valid, and we explicitly mentioned our design choice it in the method section, Quantification and statistical analyses:

“All tests were one-sided, as our hypotheses were strictly defined to predict motor improvement. Specifically, we hypothesized that delivering an extension-inducing stimulus would enhance leg extension, and delivering a flexion-inducing stimulus would enhance leg flexion. Consequently, any potentially statistically significant result in the opposite direction (e.g., inhibition) would not be considered. However, no such occurrences were observed.”

As a final note, even if such opposite observations were made, they could serve as the basis for triggering an ad-hoc follow-up study.

Reviewer #1 also provided several detailed suggestions in the section “Recommendations for the authors”. We estimated that each of them was beneficial for the correctness or for the readability of the text, and thus all were incorporated into the final version.

Reviewer #2 (Public Review):

Summary:

The authors' long-term goals are to understand the utility of precisely phased cortex stimulation regimes on recovery of function after spinal cord injury (SCI). In prior work, the authors explored the effects of contralesion cortex stimulation. Here, they explore ipsilesion cortex stimulation in which the corticospinal fibers that cross at the pyramidal decussation are spared. The authors explore the effects of such stimulation in intact rats and rats with a hemisection lesion at the thoracic level ipsilateral to the stimulated cortex. The appropriately phased microstimulation enhances contralateral flexion and ipsilateral extension, presumably through lumbar spinal cord crossed-extension interneuron systems. This microstimulation improves weight bearing in the ipsilesion hindlimb soon after injury, before any normal recovery of function would be seen. The contralateral homologous cortex can be lesioned in intact rats without impacting the microstimulation effect on flexion and extension during gait. In two rats ipsilateral flexion responses are noted, but these are not clearly demonstrated to be independent of the contralateral homologous cortex remaining intact.

Strengths:

This paper adds to prior data on cortical microstimulation by the laboratory in interesting ways. First, the strong effects of the spared crossed fibers from the ipsi-lesional cortex in parts of the ipsi-lesion leg's step cycle and weight support function are solidly demonstrated. This raises the interesting possibility that stimulating the contra-lesion cortex as reported previously may execute some of its effects through callosal coordination with the ipsi-lesion cortex tested here. This is not fully discussed by the authors but may represent a significant aspect of these data. The authors demonstrate solidly that ablation of the contra-lesional cortex does not impede the effects reported here. I believe this has not been shown for the contra-lesional cortex microstimulation effects reported earlier, but I may be wrong. Effects and neuroprosthetic control of these effects are explored well in the ipsi-lesion cortex tests here.

In the revised version of the manuscript, we incorporated various text improvements to address the points you have highlighted in your review. Additionally, we have integrated the suggested discussion topic on callosal coordination related to contralateral cortical stimulation. The discussion section now incorporates:

“Since bi-cortical interactions in sculpting descending commands are known (Brus-Ramer et al., 2009), and in light of the changes we report in ipsilesional motor cortex excitability, the role of the ipsilateral cortex in mediating or supporting functional descending commands from the contralateral cortex, particularly the immediate increase in flexion of the affected hindlimb and long-term recovery of functional control (Bonizzato & Martinez, 2021), could be further explored.”

The localization of the specific channels closest to the interhemispheric fissure (Fig. 7D) may suggest the involvement of transcallosal interactions in mediating the transmission of the cortical command generated in the ipsilateral motor cortex (Brus-Ramer, Carmel, & Martin, 2009). “While ablation experiments (Fig. 8) refute this hypothesis for ipsilateral extension control, they do not conclusively determine whether a different efferent pathway is involved in ipsilateral flexion control in this specific case."

Weaknesses:

Some data is based on very few rats. For example (N=2) for ipsilateral flexion effects of microstimulation. N=3 for homologous cortex ablation, and only ipsi extension is tested it seems. There is no explicit demonstration that the ipsilateral flexion effects in only 2 rats reported can survive the contra-lateral cortex ablation.

We agree with this assessment. The ipsilateral flexion representation is here reported as a rare but consistent phenomenon, which we believe to have robustly described with Figure 7 experiments. We underlined in the text that the ablation experiment did not conclude on the unilateral-cortical nature of ipsilateral flexion effects, by replacing the sentence with the following:

“While ablation experiments (Fig. 8) refute this hypothesis for ipsilateral extension control, they do not conclusively determine whether a different efferent pathway is involved in ipsilateral flexion control in this specific case."

Some improvements in clarity and precision of descriptions are needed, as well as fuller definitions of terms and algorithms.

Likely Impacts: This data adds in significant ways to prior work by the authors, and an understanding of how phased stimulation in cortical neuroprosthetics may aid in recovery of function after SCI, especially if a few ambiguities in writing and interpretation are fully resolved.

The manuscript text has been revised in its final version, and we sought to eliminate all ambiguity in writing and data interpretation.

In the section “Recommendations for the authors” Reviewer #2 also suggested to better define multiple terms throughout the manuscript. A clarification was added for each.

The Reviewer pointed out that we might have overlooked a correlation between locomotor recovery and motor maps increase in Figure 6. We re-approached this evaluation and found that the reviewer is correct. We were led to think that there was no correlation by “horizontally” looking at whether motor map size across rats would predict locomotor scores (as it did in the case of contralateral cortex mapping, Bonizzato and Martinez, 2021). However we now found a strong correlation between changes that happen over time for each rat and locomotor recovery, a result that was only hinted with no appropriate quantification in the previous version of the manuscript. We have now reformulated the results of Figure 6 on page 12, to include this result, and we would like to thank the reviewer for having noticed this opportunity.

Finally, we have expanded the discussion to include the following points:

The possibility that hemi-cortex coordination of contralesional microstimulation inputs may explain the Sci Transl Med results for contralesional cortex ICMS, which warrants further investigation.

The recognition that the ablation experiments do not provide conclusive evidence regarding ipsilateral flexion control and whether an alternative efferent pathway might be involved in this specific case.

Reviewer #3 (Public Review):

Summary:

This article aims to investigate the impact of neuroprosthesis (intracortical microstimulation) implanted unilaterally on the lesion side in the context of locomotor recovery following unilateral thoracic spinal cord injury.

Strength:

The study reveals that stimulating the left motor cortex, on the same side as the lesion, not only activates the expected right (contralateral) muscle activity but also influences unexpected muscle activity on the left (ipsilateral) side. These muscle activities resulted in a substantial enhancement in lift during the swing phase of the contralateral limb and improved trunk-limb support for the ipsilateral limb. They used different experimental and stimulation conditions to show the ipsilateral limb control evoked by the stimulation. This outcome holds significance, shedding light on the engagement of the "contralateral projecting" corticospinal tract in activating not only the contralateral but also the ipsilateral spinal network.

The experimental design and findings align with the investigation of the stimulation effect of contralateral projecting corticospinal tracts. They carefully examined the recovery of ipsilateral limb control with motor maps. They also tested the effective sites of cortical stimulation. The study successfully demonstrates the impact of electrical stimulation on the contralateral projecting neurons on ipsilateral limb control during locomotion, as well as identifying important stimulation spots for such an effect. These results contribute to our understanding of how these neurons influence bilateral spinal circuitry. The study's findings contribute valuable insights to the broader neuroscience and rehabilitation communities.

Thank you for your assessment of this manuscript. The final version of the manuscript incoporates your suggestions for improving term clarity and we enhanced the discussion on the mechanisms of spinal network engagement, as outlined below.

Weakness:

The term "ipsilateral" lacks a clear definition in the title, abstract, introduction, and discussion, potentially causing confusion for the reader.

[and later] However, in my opinion, readers can easily link the ipsilateral cortical network to the ipsilateral-projecting corticospinal tract, which is less likely to play a role in ipsilateral limb control in this study since this tract is disrupted by the thoracic spinal injury.

In order to mitigate the risk of having readers linking the effects of ipsilateral cortical stimulation with ipsilateral-projecting corticospinal tract, we specified:

In the abstract, we precise that our goal was: “to investigate the functional role of the ipsilateral motor cortex in rat movement through spared contralesional pathways.”

In the introduction: “In most cases, this lesion also disrupts all spinal tracts descending on the same side as the cortex under investigation at the thoracic level, meaning that the transmission of cortical commands to the ipsilesional hindlimb must depend on crossed descending tracts (Fig. S1).”

The unexpected ipsilateral (left) muscle activity is most likely due to the left corticospinal neurons recruiting not only the right spinal network but also the left spinal network. This is probably due to the joint efforts of the neuroprosthesis and activation of spinal motor networks which work bilaterally at the spinal level.

We agree with your assessment and the discussion section now emphasizes the effects of supraspinal drive onto spinal circuits.

In the section “Recommendations for the authors” Reviewer #3 suggested to provide an early reminder to the reader that the focus is on exploring the control of the ipsilateral limb through the corticospinal tract of the same side, projecting contralaterally. We did so in the abstract and introduction, as presented above.

The reviewer also suggested that the discussion could be shorter. While we recognize it covers diverse subjects that may appeal to different readers, we believe omitting some sections could limit its overall scope. The manuscript underwent three revisions and a thorough dialogue with reviewers from diverse backgrounds, and we are hesitant to undo some of these improvements.

Moreover, the section falls short of fully exploring the involvement of contralateral projecting corticospinal neurons in spinal networks for diverse motor behaviors. It could potentially delve into aspects like the potential impact of corticospinal inputs on gating the cross-extensor reflex loop and elucidating the mechanisms underlying the recruitment of the ipsilateral spinal network for generating ipsilateral limb movements. Is it a direct control on motor neurons or via existing spinal circuits?

The discussion section now includes the potential spinal circuits through which corticospinal neurons may affect motor control and reflexes.

Reviewer #3 also provided several detailed suggestions in the sub-section “Minor points”. We estimated that all of them were beneficial for the correctness or for the readability of the text, and thus were incorporated into the final version. Some of the questions raised were answered directly in the text (defining “% of chronic map” and rephrasing the original Line 479). We would like to answer here below two remaining questions:

Fig. 3C I wonder what is the average latency between stimulation onset and onset of right ankle flexor activity. Is the latency fixed, or variable (which probably indicates that the Cortical activation signal is integrated with spinal CPG activity.)

ICMS trains, unfortunately, do not allow for precise dissection of transmission timing. Single pulses at 100 µA are insufficient to generate motoneuron responses and require multiple pulses to build up cortical transmission. Alstermark et al. (Journal of Neurophysiology, 2004) used two to four stimuli with higher amplitudes to investigate forelimb transmission timing. In our 2021 Science Translational Medicine paper, we employed single pulses at 1 mA to establish transmission delays from the contralateral cortex to the ankle flexor. However, the circuits recruited at 1 mA are not directly comparable to those activated by shorter trains.

In this study, we used cortical trains of approximately 14 pulses, typical of ICMS protocols. Each pulse could potentially be the first to generate a response volley in the ankle flexor, with delays measured at 30 to 60 ms from ICMS train onset. While we believe that cortical commands are necessarily integrated with spinal CPG activity—as indicated in Figures 1B and 3D, where timing is crucial and descending commands can be gated out if delivered off-phase—the variability in latency that we recorded could be attributed to any of the following factors: cortical activation build-up, integration within reticular relay networks, or CPG integration.

Fig. 4A. Why is the activity of under contralateral ankle flexor intact condition is later than the stimulation condition?

We timed the stimulation to coincide with the contralateral leg lift and did not adjust its onset relative to spontaneous walking in SCI rats. Although stimulation could induce leg lift, as shown in Fig. 4A, SCI rats exhibited a slightly earlier and stronger activation of the right (contralateral) ankle flexor muscle even during spontaneous walking. This phenomenon is attributed to the deficits observed on the left side. The stronger right leg bears the body weight, as illustrated in Fig. 3, and thus, during body advancement, the right leg is engaged sooner and more rapidly (with a shorter swing phase) to provide support (right foot forward).

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

(1) Although there are many citations acknowledging relevant previous work, there often isn't a very granular attribution of individual previous findings to their sources. In the results section, it's sometimes ambiguous when the paper is recapping established background and when it is breaking new ground. For example, around equation 8 in the results (sv = r - rho*t), it would be good to refer to previous places where versions of this equation have been presented. Offhand, McNamara 1982 (Theoretical Population Biology) is one early instance and Fawcett et al. 2012 (Behavioural Processes) is a later one. Line 922 of the discussion seems to imply this formulation is novel here.

We would like to clarify that original manuscript equation 8, , as we derive, is not new, as it is similarly expressed in prior foundational work by McNamara (1982), and we thank the reviewer for drawing our attention to the extension of this form by Fawcett, McNamara, Houston (2012).

We now so properly acknowledge this foundational work and extension in the results section…

“This global reward-rate equivalent immediate reward (see Figure 4) is the subjective value of a pursuit, svPursuit (or simply, sv, when the referenced pursuit can be inferred), as similarly expressed in prior foundational work (McNamara 1982), and subsequent extensions (see (Fawcett, McNamara, Houston (2012)).”

…and in the Discussion section at the location referenced by the reviewer:

“From it, we re-expressed the pursuit’s worth in terms of its global reward rate-equivalent immediate reward, i.e., its ‘subjective value’, reprising McNamara’s foundational formulation (McNamara 1982).”

(2) The choice environments that are considered in detail in the paper are very simple. The simplicity facilitates concrete examples and visualizations, but it would be worth further consideration of whether and how the conclusions generalize to more complex environments. The paper considers "forgo" scenario in which the agent can choose between sequences of pursuits like A-B-A-B (engaging with option B at all opportunities, which are interleaved with a default pursuit A) and A-A-A-A (forgoing option B). It considers "choice" scenarios where the agent can choose between sequences like A-B-A-B and A-C-A-C (where B and C are larger-later and smaller-sooner rewards, either of which can be interleaved with the default pursuit). Several forms of additional complexity would be valuable to consider. [A] One would be a greater number of unique pursuits, not repeated identically in a predictable sequence, akin to a prey-selection paradigm. It seems to me this would cause t_out and r_out (the time and reward outside of the focal prospect) to be policy-dependent, making the 'apportionment cost' more challenging to ascertain. Another relevant form of complexity would be if there were [B] variance or uncertainty in reward magnitudes or temporal durations or if [C] the agent had the ability to discontinue a pursuit such as in patch-departure scenarios.

A) We would like to note that the section “Deriving Optimal Policy from Forgo Decision-making worlds”, addresses the reviewer’s scenario of n-number of pursuits”, each occurring at their own frequency, as in prey selection, not repeating identically in a predictable sequence. Within our subsection “Parceling the world…”, we introduce the concept of dividing a world (such as that) into the considered pursuit type, and everything outside of it. ‘Outside’ would include any number of other pursuits currently part of any policy, as the reviewer intuits, thus making t^out and r^out policy dependent. Nonetheless, a process of excluding (forgoing) pursuits by comparing the ‘in’ to the ‘out’ reward rate (section “Reward-rate optimizing forgo policy…”) or its equivalent sv (section “The forgo decision can also be made from subjective value), would iteratively lead to the global reward rate maximizing policy. This manner of parceling into ‘in’ and ‘out’ thus simplifies visualization of what can be complex worlds. Simpler cases that resemble common experimental designs are given in the manuscript to enhance intuition.

We thank the reviewer for this keen suggestion. We now include example figures (Supplemental 1 & 2) for multi-pursuit worlds which have the same (Supplemental 1) and different pursuit frequencies (Supplemental 2), which illustrate how this evaluation leads to reward-rate optimization. This addition demonstrates how an iterative policy would lead to reward rate maximization and emphasizes how parcellating a world into ‘in’ and ‘out’ of the pursuit type applies and is a useful device for understanding the worth of any given pursuit in more complex worlds. The policy achieving the greatest global reward rate can be realized through an iterative process where pursuits with lower reward rates than the reward rate obtained from everything other than the considered pursuit type are sequentially removed from the policy.

B) We would also emphasize that the formulation here contends with variance or uncertainty in the reward magnitudes or temporal durations. The ‘in’ pursuit is the average reward and the average time of the considered pursuit type, as is the ‘out’ the average reward and average time outside of the considered pursuit type.

C) In this work, we consider the worth of initiating one-or-another pursuit (from having completed a prior one), and not the issue of continuing within a pursuit (having already engaged it), as in patch/give-up. Handling worlds in which the agent may depart from within a pursuit, which is to say ‘give-up’ (as in patch foraging), is outside the scope of this work.

(3) I had a hard time arriving at a solid conceptual understanding of the 'apportionment cost' around Figure 5. I understand the arithmetic, but it would help if it were possible to formulate a more succinct verbal description of what makes the apportionment cost a useful and meaningful quality to focus on.

We thank the reviewer for pressing for a succinct and intuitive verbal description.

We added the following succinct verbal description of apportionment cost… “Apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration.” This definition appears in new paragraphs (as below) describing apportionment cost in the results section “Time’s cost: opportunity & apportionment costs determine a pursuit’s subjective value”, and is accompanied by equations for apportionment cost, and a figure giving its geometric depiction (Figure 5). We also expanded original figure 5 and its legend (so as to illustrate the apportionment scaling factor and the apportionment cost), and its accompanying main text, to further illustrate and clarify apportionment cost, and its relationship to opportunity cost, and time’s cost.

“What, then, is the amount of reward by which the opportunity cost-subtracted reward is scaled down to equal the sv of the pursuit? This amount is the apportionment cost of time. The apportionment cost of time (height of the brown vertical bar, Figure 5F) is the global reward rate after taking into account the opportunity cost (slope of the magenta-gold dashed line in Figure 5F) times the time of the considered pursuit. Equally, the difference between the inside and outside reward rates, times the time of the pursuit, is the apportionment cost when scaled by the pursuit’s weight, i.e., the fraction that the considered pursuit is to the total time to traverse the world (Equation 9, right hand side). From the perspective of decision-making policies, apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration (Equation 9 center, Figure 5F).

Equation 9. Apportionment Cost.

While this difference is the apportionment cost of time, the opportunity cost of time is the amount that would be expected from a policy of not taking the considered pursuit over a time equal to the considered pursuit’s duration. Together, they sum to Time’s Cost (Figure 5G). Expressing a pursuit’s worth in terms of the global reward rate obtained under a policy of accepting the pursuit type (Figure 5 left column), or from the perspective of the outside reward and time (Figure 5 right column), are equivalent. However, the latter expresses sv in terms that are independent of one another, conveys the constituents giving rise to global reward rate, and provides the added insight that time’s cost comprises an apportionment as well as an opportunity cost.”

The above definition of apportionment cost adds to other stated relationships of apportionment cost found throughout the paper (original lines 434,435,447,450).

I think Figure 6C relates to this, but I had difficulty relating the axis labels to the points, lines, and patterned regions in the plot.

We thank the reviewer for pointing out that this figure can be made to be more easily understood.

We have done so by breaking its key features over a greater number of plots so that no single panel is overloaded. We have also changed text in the legend to clarify how apportionment and opportunity costs add to constitute time’s cost, and also correspondingly in the main text.

I also was a bit confused by how the mathematical formulation was presented. As I understood it, the apportionment cost essentially involves scaling the rest of the SV expression by t^out/(tⁱⁿ + t^out).

The reviewer’s understanding is correct: the amount of reward of the pursuit that remains after subtracting the opportunity cost, when so scaled, is equivalent to the subjective value of that pursuit. The amount by which that scaling decreases the rest of the SV expression is equal to the apportionment cost of time.

The way this scaling factor is written in Figure 5C, as 1/(1 + (1/t^out) tⁱⁿ), seems less clear than it could be.

To be sure, we present the formula in original Figure 5C in this manner to emphasize the opportunity cost subtraction as separable from the apportionment rescaling, expressing the opportunity cost subtraction and the apportionment scaling component of the equation as their own terms in parentheses.

But we understand the reviewer to be referring to the manner by which we chose to express the scaling term. We presented it in this way in the original manuscript, (rather than its more elegant form recognized by the reviewer) to make direct connection to temporal discounting literature. In this literature, discounting commonly takes the same mathematical form as our apportionment cost scaling, but whereas the steepness of discounting in this literature is controlled by a free fit parameter, k, we show how for a reward rate maximizing agent, the equivalent k term isn’t a free fit parameter, but rather is the reciprocal of the time spent outside the considered pursuit type.

We take the reviewer’s advice to heart, and now first express subjective value in the format that emphasizes opportunity cost subtraction followed by an apportionment downscaling, identifying the apportionment scaling term, t^out/(t^out + tⁱⁿ), ie the outside weight. Figure 5 now shows the geometric representation of apportionment scaling and apportionment cost. Only subsequently in the discounting function section then do we now in the revised manuscript rearrange this subjective value expression to resemble the standard discounting function form.

Also, the apportionment cost is described in the text as being subtracted from sv rather than as a multiplicative scaling factor.

What we describe in the original text is how apportionment cost is a component of time’s cost, and how sv is the reward less time’s cost. It would be correct to say that apportionment cost and opportunity cost are subtracted from the pursuit’s reward to yield the subjective value of the pursuit. This is what we show in the original Figure 5D graphically. Original Figure 5 and accompanying formulas at its bottom show the equivalence of expressing sv in terms of subtracting time’s cost as calculated from the global reward rate under a policy of accepting the considered pursuit, or, of subtracting opportunity cost and then scaling the opportunity cost subtracted reward by the apportionment scaling term, thereby accounting for the apportionment cost of time.

The revision of original figure 5, its figure legend, and accompanying text now make clear the meaning of apportionment cost, how it can be considered a subtraction from the reward of a pursuit, or, equivalently, how it can be thought of as the result of scaling down of opportunity cost subtracted reward.

It could be written as a subtraction, by subtracting a second copy of the rest of the SV expression scaled by t_in/(t_in + t_out). But that shows the apportionment cost to depend on the opportunity cost, which is odd because the original motivation on line 404 was to resolve the lack of independence between terms in the SV expression.

On line 404 of the original manuscript, we point out that the simple equation―which is a reprisal of McNamara’s insight―is problematic in that its terms on the RHS are not independent: the global reward rate is dependent on the considered pursuit’s reward (see Fig5B). The alternative expression for subjective value that we derive expresses sv in terms that are all independent of one another. We may have unintentionally obscured that fact by having already defined rhoⁱⁿ as rⁱⁿ/ tⁱⁿ and rho^out as r^out/t^out on lines 306 and 307.

Therefore, in the revision, Ap 8 is expressed so to keep clear that it uses terms that are all independent of one another, and only subsequently express this formula with the simplifying substitution, rho^out.

That all said, we understand the reviewer’s point to be that the parenthetical terms relating the opportunity cost and the apportionment rescaling both contain within them the parameter t^out, and in this way these concepts we put forward to understand the alternative equation are non-independent. That is correct, but it isn’t at odds with our objective to express SV in terms that are independent with one another (which we do). Our motivation in introducing these concepts is to provide insight and intuition into the cost of time (especially now with a clear and simple definition of apportionment cost stated). We go to lengths to demonstrate their relationship to each other.

(4) In the analysis of discounting functions (line 664 and beyond), the paper doesn't say much about the fact that many discounting studies take specific measures to distinguish true time preferences from opportunity costs and reward-rate maximization.

We understand the reviewer’s comment to connote that temporal decision-making worlds in which delay time does not preclude reward from outside the current pursuit is a means to distinguish time preference from the impact of opportunity cost. One contribution of this work is to demonstrate that, from a reward-rate maximization framework, an accounting of opportunity cost is not sufficient to understand apparent time preferences as distinguishable from reward-rate maximization. The apportionment cost of time must also be considered to have a full appreciation of the cost of time. For instance, let us consider a temporal decision-making world in which there is no reward received outside the considered pursuit. In such a world, there is no opportunity cost of time, so apparent temporal discounting functions would appear as if purely hyperbolic as a consequence of the apportionment cost of time alone. Time preference, as revealed experimentally by the choices made between a SS and a LL reward, then, seem confounding, as preference can reverse from a SS to a LL option as the displacement of those options (maintaining their difference in time) increases (Green, Fristoe, and Myerson 1994; Kirby and Herrnstein 1995). While this shift, the so-called “Delay effect”, could potentially arise as a consequence of some inherent time preference bias of an agent, we demonstrate that a reward-rate maximal agent exhibits hyperbolic discounting, and therefore it would also exhibit the Delay effect, even though it has no time preference.

In the revision we now make reference to the Delay Effect (in abstract, results new section “The Delay Effect” with new figure 14, and in the discussion), which is taken as evidence of time preference in human and animal literature, and note explicitly how a reward-rate maximizing agent would also exhibit this behavior as a consequence of apparent hyperbolic discounting.

In many of the human studies, delay time doesn't preclude other activities.

Our framework is generalizable to worlds in which being in pursuit does not preclude an agent from receiving reward during that time at the outside reward rate. Original Ap 13 solves for such a condition, and shows that in this context, the opportunity cost of time drops out of the SV equation, leaving only the consequences of the apportionment cost of time. We made reference to this case on lines 1032-1034 of the original manuscript: “In this way, such hyperbolic discounting models [models that do not make an accounting of opportunity cost] are only appropriate in worlds with no “outside” reward, or, where being in a pursuit does not exclude the agent from receiving rewards at the rate that occurs outside of it (Ap. 13).”

The note and reference is fleeting in the original work. We take the reviewer’s suggestion and now add paragraphs in the discussion on the difference between humans and animals in apparent discounting, making specific note of human studies in which delay time doesn’t preclude receiving outside reward while engaged in a pursuit. Relatedly, hyperbolic discounting is oft considered to be less steep in humans than in animals. As the reviewer points out, these assessments are frequently made under conditions in which being in a pursuit does not preclude receiving reward from outside the pursuit. When humans are tested under conditions in which outside rewards are precluded, they exhibit far steeper discounting. We now include citation to that observation (Jimura et al. 2009). We handle such conditions in original AP 13, and show how, in such worlds, the opportunity cost of time drops out of the equation. The consequence of this is that the apparent discounting function would become less steep (the agent would appear as if more patient), consistent with reports.

“Relating to the treatment of opportunity cost, we also note that many investigations into temporal discounting do not make an explicit distinction between situations in which 1) subjects continue to receive the usual rewards from the environment during the delay to a chosen pursuit, and 2) situations in which during a chosen pursuit’s delay no other rewards or opportunities will occur (Kable & Glimcher, 2007; Kirby & Maraković, 1996; McClure, Laibson, Loewenstein, & Cohen, 2004). Commonly, human subjects are asked to answer questions about their preferences between options for amounts they will not actually earn after delays they will not actually have to wait, during which it is unclear whether they are really investing time away from other options or not (Rosati et al., 2007). In contrast, in most animal experiments, subjects actually receive reward after different delays during which they do not receive new options or rewards. By our formulation, when a pursuit does not exclude the agent from receiving rewards at the rate that occurs outside, the opportunity cost of time drops out of the subjective value equation (Ap 12).

Equation 10. The value of initiating a pursuit when pursuit does not exclude receiving rewards at the outside rate (Ap 12)

Therefore, the reward-rate maximizing discounting function in these worlds is functionally equivalent to the situation in which the outside reward rate is zero, and will―lacking an opportunity cost―be less steep. This rationalizes why human discounting functions are often reported to be longer (gentler) than animal discounting functions: they are typically tested in conditions that negate opportunity cost, whereas animals are typically tested in conditions that enforce opportunity costs. Indeed, when humans are made to wait for actually received reward, their observed discounting functions are much steeper (Jimura et al. 2009). “

In animal studies, rate maximization can serve as a baseline against which to measure additional effects of temporal discounting. This is an important caveat to claims about discounting anomalies being rational under rate maximization (e.g., line 1024).

We agree that the purpose of this reward-rate maximizing framework is to serve as a point of comparison in which effects of temporal intervals and rewards that define the environment can be analyzed to better understand the manner in which animals and humans deviate from this ideal behavior. Our interest in this work is in part motivated by a desire to have a deeper understanding of what “true” time preference means. Using the reward-rate maximizing framework here provides a means to speak about time preferences (ie biases) in terms of deviation from optimality. From this perspective, a reward-rate maximal agent doesn’t exhibit time preference: its actions are guided solely by reward-rate optimizing valuation. Therefore, one contribution of this work is to show that purported signs of time preference (hyperbolic discounting, magnitude, sign, and (now) delay effect) can be explained without invoking time preference. What errors from optimality that remain following an proper accounting of reward-rate maximizing behavior should then, and only then, be considered from the lens of time preference (bias).

(5) The paper doesn't feature any very concrete engagement with empirical data sets. This is ok for a theoretical paper, but some of the characterizations of empirical results that the model aims to match seem oversimplified. An example is the contention that real decision-makers are optimal in accept/reject decisions (line 816 and elsewhere). This isn't always true; sometimes there is evidence of overharvesting, for example.

We would like to note that the scope of this paper is limited to examining the value of initiating a pursuit, rather than the value of continuing within a pursuit. The issue of continuing within a pursuit constitutes a third fundamental topology, which could be called give-up or patch-foraging, and is complex and warrants its own paper. In Give-up topologies, which are distinct from Forgo, and Choice topologies, the reviewer is correct in pointing out that the preponderance of evidence demonstrates that animals and humans are as if overpatient, adopting a policy of investing too much time within a pursuit, than is warranted_._ In Forgo instances, however, the evidence supports near optimality.

(6) Related to the point above, it would be helpful to discuss more concretely how some of this paper's theoretical proposals could be empirically evaluated in the future. Regarding the magnitude and sign effects of discounting, there is not a very thorough overview of the several other explanations that have been proposed in the literature. It would be helpful to engage more deeply with previous proposals and consider how the present hypothesis might make unique predictions and could be evaluated against them.

We appreciate the reviewer’s point that there are many existing explanations for these various ‘anomalous’ effects. We hold that the point of this work is to demonstrate that these effects are consistent with a reward-rate maximizing framework so do not require additional assumptions, like separate processes for small and large rewards, or the inclusion of a utility function.

Nonetheless, there is a diversity of explanations for the sign and magnitude effect, and, (now with its explicit inclusion in the revision) the delay effect. Therefore, we now also include reference to additional work which proffers alternative explanations for the sign and magnitude effects, (as reviewed by (Kalenscher and Pennartz 2008; Frederick et al. 2002)), as well as a scalar timing account of non-stationary time preference (Gibbon, 1977).

With respect to making predictions, this framework makes the following in regards to the magnitude, sign, and (now in the revision) delay effect: in Discussion, Magnitude effect subsection: “The Magnitude Effect should be observed, experimentally, to diminish when 1) increasing the outside time while holding the outside reward constant, (thus decreasing the outside reward rate), or when 2) decreasing the outside reward while holding the outside time constant (thus decreasing the outside reward rate). However, 3) the Magnitude Effect would exaggerate as the outside time increased while holding the outside reward rate constant.”, in Sign effect subsection: “…we then also predict that the size of the Sign effect would diminish as the outside reward rate decreases (and as the outside time increases), and in fact would invert should the outside reward rate turn negative (become net punishing), such that punishments would appear to discount more steeply than rewards.” Delay effect subsection: “...a sign of irrationality is that a preference reversal occurs at delays greater than what a reward-rate-maximizing agent would exhibit.”

A similar point applies to the 'malapportionment hypothesis' although in this case there is a very helpful section on comparisons to prior models (line 1163). The idea being proposed here seems to have a lot in common conceptually with Blanchard et al. 2013, so it would be worth saying more about how data could be used to test or reconcile these proposals.

We thank the reviewer for holding that the section of model comparisons to be very helpful. We believe the text previously dedicated to this issue to be sufficient in this regard. We have, however, adding substantively to the Malapportionment Hypothesis section (Discussion) and its accompanying figure, to make explicit a number of predictions from the Malapportionment hypothesis as it relates to Hyperbolic discounting, the Delay Effect, and the Sign and Magnitude Effects.

Reviewer #1 Recommendations

(1) As a general note about the figures, it would be helpful to specify, either graphically or in the caption, what fixed values of reward sizes and time intervals are being assumed for each illustration.

Thank you for the suggestion. We attempted to keep graphs as uncluttered as possible, but agree that for original figures 4,5,16, and 17, which didn’t have numbered axes, that we should provide the amounts in the captions in the revised figures (4,5, and now 17,18). These figures did not have numerics as their shapes and display are to illustrate the form of the relationship between vectors, being general to the values they may take.

We now include in the captions for these figures the parameter amounts used.

(2) Should Equation 2 have t in the denominator instead of r?

Indeed. We thank the reviewer for catching this typographical error.

We have corrected it in the revision.

(3) General recommendation:

My view is that in order for the paper's eLife assessment to improve, it would be necessary to resolve points 1 through 4 listed under "weaknesses" in my public review, which pertain to clarity and acknowledgement of prior work. I think a lot hinges on whether the authors can respond to point #3 by making a more compelling case for the usefulness and generality of the 'apportionment cost' concept, since that idea is central to the paper's contribution.

We believe these critical points (1-4) to improve the paper will now have been addressed to the reviewer’s satisfaction.

Reviewer #2 (Public review):

While the details of the paper are compelling, the authors' presentation of their results is often unclear or incomplete:

(1) The mathematical details of the paper are correct but contain numerous notation errors and are presented as a solid block of subtle equation manipulations. This makes the details of the authors' approach (the main contribution of the paper to the field) highly difficult to understand.

We thank the reviewers for having detected typographical errors regarding three equations. They have been corrected. The first typographical error in the original main text (Line 277) regards equation 2 and will be corrected so that equation 2 appears correctly as

The second typo regards the definition of the considered pursuit’s reward rate which appear in the original main text (line 306), and has been corrected to appear as

The third typographical error occurred in conversion from Google Sheets to Microsoft Word appearing in the original main text (line 703) and regards the subjective value expression when no reward is received in an intertrial interval (ITI). It has been corrected to appear as

(2) One of the main contributions of the paper is the notion that time’s cost in decision-making contains an apportionment cost that reflects the allocation of decision time relative to the world. The authors use this cost to pose a hypothesis as to why subjects exhibit sub-optimal behavior in choice decisions. However, the equation for the apportionment cost is never clearly defined in the paper, which is a significant oversight that hampers the effectiveness of the authors' claims.

We thank the reviewer for pressing on this critical point. Reviewers commonly identified a need to provide a concise and intuitive definition of apportionment cost, and to explicitly solve and provide for its mathematical expression.

We added the following succinct verbal description of apportionment cost… “Apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration.” This definition appears in new paragraphs (as below) describing apportionment cost in the results section “Time’s cost: opportunity & apportionment costs determine a pursuit’s subjective value”, and is accompanied by equations for apportionment cost, and a figure giving its geometric depiction (Figure 5). We also expanded original figure 5 and its legend (so as to illustrate the apportionment scaling factor and the apportionment cost), and its accompanying main text, to further illustrate and clarify apportionment cost, and its relationship to opportunity cost, and time’s cost.

“What, then, is the amount of reward by which the opportunity cost-subtracted reward is scaled down to equal the sv of the pursuit? This amount is the apportionment cost of time. The apportionment cost of time (height of the brown vertical bar, Figure 5F) is the global reward rate after taking into account the opportunity cost (slope of the magenta-gold dashed line in Figure 5F) times the time of the considered pursuit. Equally, the difference between the inside and outside reward rates, times the time of the pursuit, is the apportionment cost when scaled by the pursuit’s weight, i.e., the fraction that the considered pursuit is to the total time to traverse the world (Equation 9, right hand side). From the perspective of decision-making policies, apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration (Equation 9 center, Figure 5F).

Equation 9. Apportionment Cost.

While this difference is the apportionment cost of time, the opportunity cost of time is the amount that would be expected from a policy of not taking the considered pursuit over a time equal to the considered pursuit’s duration. Together, they sum to Time’s Cost (Figure 5G). Expressing a pursuit’s worth in terms of the global reward rate obtained under a policy of accepting the pursuit type (Figure 5 left column), or from the perspective of the outside reward and time (Figure 5 right column), are equivalent. However, the latter expresses sv in terms that are independent of one another, conveys the constituents giving rise to global reward rate, and provides the added insight that time’s cost comprises an apportionment as well as an opportunity cost.”

(3) Many of the paper's figures are visually busy and not clearly detailed in the captions (for example, Figures 6-8). Because of the geometric nature of the authors' approach, the figures should be as clean and intuitive as possible, as in their current state, they undercut the utility of a geometric argument.

We endeavored to make our figures as simple as possible. We have made in the revision changes to figures that we believe improve their clarity. These include: 1) breaking some figures into more panels when more than one concept was being introduced (such as in revised Figure 5 , 6, 7, and 8), 2) using the left hand y axis for the outside reward, and the right hand axis for the inside reward when plotting the “in” and “outside” reward, and indicating their respective numerics (which run in opposite directions), 3) adding a legend to the figures themselves where needed (revised figures 10, 11, 12, 14) 4) adding the values used to the figure captions, where needed, and 5) ensuring all symbols are indicated in legends.

(4) The authors motivate their work by focusing on previously-observed behavior in decision experiments and tell the reader that their model is able to qualitatively replicate this data. This claim would be significantly strengthened by the inclusion of experimental data to directly compare to their model's behavior. Given the computational focus of the paper, I do not believe the authors need to conduct their own experiments to obtain this data; reproducing previously accepted data from the papers the authors' reference would be sufficient.

Our objective was not to fit experimentally observed data, as is commonly the goal of implementation/computational models. Rather, as a theory, our objective is to rationalize the broad, curious, and well-established pattern of temporal decision-making behaviors under a deeper understanding of reward-rate maximization, and from that understanding, identify the nature of the error being committed by whatever learning algorithm and representational architecture is actually being used by humans and animals. In doing so, we make a number of important contributions. By identifying and analyzing reward-rate-maximizing equations, we 1) provide insight into what composes time’s cost and how the temporal structure of the world in which it is embedded (its ‘context’) impacts the value of a pursuit, 2) rationalize a diverse assortment of temporal decision-making behaviors (e.g., Hyperbolic discounting, the Magnitude Effect, the Sign Effect, and the Delay effect), explaining them with no assumed free-fit parameter, and then, by analyzing error in parameters enabling reward-rate maximization, 3) identify the likely source of error and propose the Malapportionment Hypothesis. The Malapportionment Hypothesis identifies the underweighting of a considered pursuit’s “outside”, and not error in pursuit’s reward rates, as the source of error committed by humans and animals. It explains why animals and humans can present as suboptimally ‘impatient’ in Choice, but as optimal in Forgo. At the same time, it concords with numerous and diverse observations in decision making regarding whether to initiate a pursuit. The nature of this error also, then, makes numerous predictions. These insights inform future computational and experimental work by providing strong constraints on the nature of the algorithm and representational architecture used to learn and represent the values of pursuits. Rigorous test of the Malapportionment Hypothesis will require wholly new experiments.

In the revision, we also now emphasize and add predictions of the Malapportionment Hypothesis, updated its figure (Figure 21), its legend, and its paragraphs in the discussion.

“We term this reckoning of the source of error committed by animals and humans the Malapportionment Hypothesis, which identifies the underweighting of the time spent outside versus inside a considered pursuit but not the misestimation of pursuit rates, as the source of error committed by animals and humans (Figure 21). This hypothesis therefore captures previously published behavioral observations (Figure 21A) showing that animals can make decisions to take or forgo reward options that optimize reward accumulation (Krebs et al., 1977; Stephens and Krebs, 1986; Blanchard and Hayden, 2014), but make suboptimal decisions when presented with simultaneous and mutually exclusive choices between rewards of different delays (Logue et al., 1985; Blanchard and Hayden, 2015; Carter and Redish, 2016; Kane et al., 2019). The Malapportionment Hypothesis further predicts that apparent discounting functions will present with greater curvature than what a reward-rate-maximizing agent would exhibit (Figure 21B). While experimentally observed temporal discounting would have greater curvature, the Malapportionment Hypothesis also predicts that the Magnitude (Figure 21C) and Sign effect (Figure 21D) would be less pronounced than what a reward-rate-maximizing agent would exhibit, with these effects becoming less pronounced the greater the underweighting. Finally, with regards to the Delay Effect (Figure 21E), the Malapportionment Hypothesis predicts that preference reversal would occur at delays greater than that exhibited by a reward-rate-maximizing agent, with the delay becoming more pronounced the greater the underweighting outside versus inside the considered pursuit by the agent.”

(5) While the authors reference a good portion of the decision-making literature in their paper, they largely ignore the evidence-accumulation portion of the literature, which has been discussing time-based discounting functions for some years. Several papers that are both experimentally-(Cisek et al. 2009, Thurs et al. 2012, Holmes et al. 2016) and theoretically-(Drugowitsch et al. 2012, Tajima et al. 2019, Barendregt et al. 22) driven exist, and I would encourage the authors to discuss how their results relate to those in different areas of the field.

In this manuscript, we consider the worth of initiating one or another pursuit having completed a prior one, and not the issue of continuing within a pursuit having already engaged in it. The worth of continuing a pursuit, as in patch-foraging/give-up tasks, constitutes a third fundamental time decision-making topology which is outside the scope of the current work. It engages a large and important literature, encompassing evidence accumulation, and requires a paper on the value of continuing a pursuit in temporal decision making, in its own right, that can use the concepts and framework developed here. The excellent works suggested by the reviewer will be most relevant to that future work concerning patch-foraging/give-up topologies.

Reviewer #2 Recommendations:

(1) In Equation 1, the term rho_d is referred to as the reward rate of the default pursuit, when it should be the reward of the default pursuit.

Regarding Equation 1, it is formulated to calculate the average reward received and average time spent per unit time spent in the default pursuit. So, f_i is the encounter rate of pursuit i for one unit of time spent in the default pursuit (lines 259-262). Added to the summation in the numerator, we have the average reward obtained in the default pursuit per unit time () and in the denominator we have the time spent in the default pursuit per unit time (1).

We have added clarifying text to assist in meaning of the equation in Ap 1, and thank the reviewer for pointing out this need.

(2) The notation for "in" and "out" of a considered pursuit type begins as being used to describe the contribution from a single pursuit (without inter-trial interval) towards global reward rate and the contribution of all other factors (other possible pursuits and inter-trial interval) towards global reward rate, respectively, but is then used to describe the pursuit's contribution and the inter-trial interval's contribution, respectively, to the global reward rate. This should be cleaned up to be consistent throughout, or at the very least, it should be addressed when this special case is considered the default.

As understood by the reviewer, “in” and “out” of the considered pursuit type describes the general form by which a world can be cleaved into these two parts: the average time and reward received outside of the considered pursuit type for the average time and reward received within that pursuit type. A specific, simple, and common experimental instance would be a world composed of one or another pursuit and an intertrial interval.

We now make clear how such a world composed of a considered pursuit and an inter trial interval would be but one special case. In example cases where t^out represents the special case of an inter-trial interval, this is now stated clearly. For instance, we do so when discussing how a purely hyperbolic discounting function would apply in worlds in which no reward is received in t^out, stating that this is often the case common to experimental designs where t^out represents an intertrial interval with no reward. Importantly, by the new inclusion of illustrated worlds in the revision that have n-number pursuits that could occur from a default pursuit and 1) equal frequency (Supplemental 1), and 2) at differing frequencies (Supplemental 2), we make more clear the generalizability and utility of this t^out/tin concept.

(3) Figure 5 should make clear the decomposition of time's cost both graphically and functionally. As it stands, the figure does not define the apportionment cost.

In the revision of original fig 5, we now further decompose the figure to effectively convey 1) what opportunity cost, and (especially) 2) the apportionment cost is, both graphically and mathematically, 3) how time’s cost is comprised by them, 4) how the apportionment scaling term scales the opportunity-cost-subtracted reward by time’s allocation to equal the subjective value, and 4) the equivalence between the expression of time’s cost using terms that are not independent of one another with the expression of time’s cost using terms that are independent of one another.

(4) Figures 6-8 do not clearly define the dots and annuli used in panels B and C.

We have further decomposed figures 6-8 so that the functional form of opportunity, apportionment, and time’s cost can be more clearly appreciated, and what their interrelationship is with respect to changing outside reward and outside time, and clearly identify symbols used in the corresponding legends.

(5) The meaning of a negative subjective value should be specifically stated. Is it the amount a subject would pay to avoid taking the considered pursuit?

As the reviewer intuits, negative subjective value can be considered the amount an agent ought be willing to pay to avoid taking the considered pursuit.

We now include the following lines in “The forgo decision can also be made from subjective value” section in reference to negative subjective value…

“A negative subjective value thus indicates that a policy of taking the considered pursuit would result in a global reward rate that is less than a policy of forgoing the considered pursuit. Equivalently, a negative subjective value can be considered the amount an agent ought be willing to pay to avoid having to take the considered pursuit.”

(6) Why do you define the discounting function as the normalized subjective value? This choice should be justified, via literature citations or a well-described logical argument.

The reward magnitude normalized subjective value-time function is commonly referred to as the temporal discounting function as it permits comparison of the discount rate isolated from a difference in reward magnitude and/or sign and is deeply rooted in historical precedent. As the reviewer points out, the term is overloaded, however, as investigations in which comparisons between the form of subjective value-time functions is not needed tend to refer to these functions as temporal discounting functions as well.

We make clear in the revised text in the introduction our meaning and use of the term, the justification in doing so, and its historical roots.

“Historically, temporal decision-making has been examined using a temporal discounting function to describe how delays in rewards influence their valuation. Temporal discounting functions describe the subjective value of an offered reward as a function of when the offered reward is realized. To isolate the form of discount rate from any difference in reward magnitude and sign, subjective value is commonly normalized by the reward magnitude when comparing subjective value-time functions (Strotz, 1956, Jimura, 2009). Therefore, we use the convention that temporal discounting functions are the magnitude-normalized subjective value-time function (Strotz, 1956).”

Special addition. In investigating the historical roots of the discounting function prompted by the reviewer, we learned (Grüne-Yanoff 2015) that it was Mazur that simply added the “1+k” in the denominator of the hyperbolic discounting function. Our derivation for the reward-rate optimal agent makes clear why apparent temporal discounting functions ought have this general form.

Therefore, we add the following to the “Hyperbolic Temporal Discounting Function section in the discussion…

“It was Ainslie (Ainslie, 1975) who first understood that the empirically observed “preference reversals” between SS and LL pursuits could be explained if temporal discounting took on a hyperbolic form, which he initially conjectured to arise simply from the ratio of reward to delay (Grüne-Yanoff 2015). This was problematic, however, on two fronts: 1) as the time nears zero, the value curve goes to infinity, and 2) there is no accommodation of differences observed within and between subjects regarding the steepness of discounting. Mazur (Mazur, 1987) addressed these issues by introducing 1 + k into the denominator, providing for the now standard hyperbolic discounting function, . Introduction of “1” solved the first issue, though “it never became fully clear how to interpret this 1” (Grüne-Yanoff 2015; interviewing Ainslie). Introduction of the free-fit parameter, k, accommodated the variability observed across and within subjects by controlling the curvature of temporal discounting, and has become widely interpreted as a psychological trait, such as patience, or willingness to delay gratification (Frederick et al., 2002).”

…continuing later in that section to explain why the reward-rate optimal agent would exhibit this general form…

“Regarding form, our analysis reveals that the apparent discounting function of a reward-rate-maximizing agent is a hyperbolic function…

…which resembles the standard hyperbolic discounting function, , in the denominator, where . Whereas Mazur introduced 1 + k to t in the denominator to 1) force the function to behave as t approaches zero, and 2) provide a means to accommodate differences observed within and between subjects, our derivation gives cause to the terms 1 and k, their relationship to one another, and to t in the denominator. First, from our derivation, “1” actually signifies taking t_out amount of time expressed in units of t_out (t_out/t_out=1) and adding it to t_in  amount of time expressed in units of t_out (ie, the total time to make a full pass through the world expressed in terms of how the agent apportions its time under a policy of accepting the considered pursuit).”

Additional Correction. In revising the section, “Hyperbolic Temporal Discounting Functions” in the discussion, we also detected an error in our description of the meaning of suboptimal bias for SS. In the revision, the sentence now reads…

“More precisely, what is meant by this suboptimal bias for SS is that the switch in preference from LL to SS occurs at an outside reward rate that is lower—and/or an outside time that is greater —than what an optimal agent would exhibit.”

(7) Figure 15B should have negative axes defined for the pursuit's now negative reward.

Yes- excellent point.

To remove ambiguity regarding the valence of inside and outside reward magnitudes, we have changed all such figures so that the left hand y-axis is used to signify the outside reward magnitude and sign, and so that the right hand y-axis is used to signify the inside reward magnitude and sign.

With respect to the revision of original 15B, this change now makes clear that the inside reward label and numerics on the right hand side of the graph run from positive (top) to negative (bottom) values so that it can now be understood that the magnitude of the inside reward is negative in this figure (ie, a punishment). The left hand y-axis labeling the outside reward magnitude has numerics that run in the opposite direction, from negative (top) to positive (bottom). In this figure, the outside reward rate is positive whereas the inside reward rate is negative.

(8) When comparing your discounting function to the TIMERR and Heuristic models, it would be useful to include a schematic plot illustrating the different obtainable behaviors from all models rather than just telling the reader the differences.

We hold that the descriptions and references are sufficient to address these comparisons.

(9) I would strongly suggest cleaning up all appendices for notation…

The typographical errors that have been noted in these reviews have all been corrected. We believe the reviewer to be referring here to the manner that we had cross-referenced Equations in the appendices and main text which can lead to confusion between whether an equation number being referenced is in regard to its occurrence in the main text or its occurrence in the appendices.

In the revision, we eliminate numbering of equations in the appendices except where an equation occurs in an appendix that is referenced within the main text. In the main text, important equations are numbered sequentially and note the appendix from which they derive. If an equation in an appendix is referenced in the main text, it is noted within the appendix it derives.

…and replacing some of the small equation manipulations with written text describing the goal of each derivation.

To increase clarity, we have taken the reviewer’s helpful suggestion, adding helper text in the appendices were needed, and have bolded the equations of importance within the Appendices (rather than removing equation manipulations making clear steps of derivation).

(10) I would suggest moving the table in Appendix 11 to the main text where misestimation is referenced.

So moved. This appendix now appears in the main text as table 1 “Definitions of misestimating global reward rate-enabling parameters”.

Reviewer #3 (Public review):

One broad issue with the paper is readability. Admittedly, this is a complicated analysis involving many equations that are important to grasp to follow the analyses that subsequently build on top of previous analyses.

But, what's missing is intuitive interpretations behind some of the terms introduced, especially the apportionment cost without referencing the equations in the definition so the reader gets a sense of how the decision-maker thinks of this time cost in contrast with the opportunity cost of time.

We thank the reviewer for encouraging us to formulate a succinct and intuitive statement as to the nature of apportionment cost. We thank the reviewer for pressing for a succinct and intuitive verbal description.

We added the following succinct verbal description of apportionment cost… “Apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration.” This definition appears in a new paragraph (as below) describing apportionment cost in the results section “Time’s cost: opportunity & apportionment costs determine a pursuit’s subjective value”, and is accompanied by equations for apportionment cost, and a figure giving its geometric depiction (Figure 5). We also expanded original figure 5 and its legend (so as to illustrate the apportionment scaling factor and the apportionment cost), and its accompanying main text, to further illustrate and clarify apportionment cost, and its relationship to opportunity cost, and time’s cost.

“What, then, is the amount of reward by which the opportunity cost-subtracted reward is scaled down to equal the sv of the pursuit? This amount is the apportionment cost of time. The apportionment cost of time (height of the brown vertical bar, Figure 5F) is the global reward rate after taking into account the opportunity cost (slope of the magenta-gold dashed line in Figure 5F) times the time of the considered pursuit. Equally, the difference between the inside and outside reward rates, times the time of the pursuit, is the apportionment cost when scaled by the pursuit’s weight, i.e., the fraction that the considered pursuit is to the total time to traverse the world (Equation 9, right hand side). From the perspective of decision-making policies, apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration (Equation 9 center, Figure 5F).

Equation 9. Apportionment Cost.

While this difference is the apportionment cost of time, the opportunity cost of time is the amount that would be expected from a policy of not taking the considered pursuit over a time equal to the considered pursuit’s duration. Together, they sum to Time’s Cost (Figure 5G). Expressing a pursuit’s worth in terms of the global reward rate obtained under a policy of accepting the pursuit type (Figure 5 left column), or from the perspective of the outside reward and time (Figure 5 right column), are equivalent. However, the latter expresses sv in terms that are independent of one another, conveys the constituents giving rise to global reward rate, and provides the added insight that time’s cost comprises an apportionment as well as an opportunity cost.”

The above definition of apportionment cost adds to other stated relationships of apportionment cost found throughout the paper (original lines 434,435,447,450).

Re-analysis of some existing empirical data through the lens of their presented objective functions, especially later when they describe sources of error in behavior.

Our objective was not to fit experimentally observed data, as is commonly the goal of implementation/computational models. Rather, as a theory, our objective is to rationalize the broad, curious, and well-established pattern of temporal decision-making behaviors under a deeper understanding of reward-rate maximization, and from that understanding, identify the nature of the error being committed by whatever learning algorithm and representational architecture is actually being used by humans and animals. In doing so, we make a number of important contributions. By identifying and analyzing reward-rate-maximizing equations, we 1) provide insight into what composes time’s cost and how the temporal structure of the world in which it is embedded (its ‘context’) impacts the value of a pursuit, 2) rationalize a diverse assortment of temporal decision-making behaviors (e.g., Hyperbolic discounting, the Magnitude Effect, the Sign Effect, and the Delay effect), explaining them with no assumed free-fit parameter, and then, by analyzing error in parameters enabling reward-rate maximization, 3) identify the likely source of error and propose the Malapportionment Hypothesis. The Malapportionment Hypothesis identifies the underweighting of a considered pursuit’s “outside”, and not error in pursuit’s reward rates, as the source of error committed by humans and animals. It explains why animals and humans can present as suboptimally ‘impatient’ in Choice, but as optimal in Forgo. At the same time, it concords with numerous and diverse observations in decision making regarding whether to initiate a pursuit. The nature of this error also, then, makes numerous predictions. These insights inform future computational and experimental work by providing strong constraints on the nature of the algorithm and representational architecture used to learn and represent the values of pursuits. Rigorous test of the Malapportionment Hypothesis will require wholly new experiments.

In the revision, we also now emphasize and add predictions of the Malapportionment Hypothesis, augmenting its figure (Figure 21), its legend, and its paragraphs in the discussion.

“We term this reckoning of the source of error committed by animals and humans the Malapportionment Hypothesis, which identifies the underweighting of the time spent outside versus inside a considered pursuit but not the misestimation of pursuit rates, as the source of error committed by animals and humans (Figure 21). This hypothesis therefore captures previously published behavioral observations (Figure 21A) showing that animals can make decisions to take or forgo reward options that optimize reward accumulation (Krebs et al., 1977; Stephens and Krebs, 1986; Blanchard and Hayden, 2014), but make suboptimal decisions when presented with simultaneous and mutually exclusive choices between rewards of different delays (Logue et al., 1985; Blanchard and Hayden, 2015; Carter and Redish, 2016; Kane et al., 2019). The Malapportionment Hypothesis further predicts that apparent discounting functions will present with greater curvature than what a reward-rate-maximizing agent would exhibit (Figure 21B). While experimentally observed temporal discounting would have greater curvature, the Malapportionment Hypothesis also predicts that the Magnitude (Figure 21C) and Sign effect (Figure 21D) would be less pronounced than what a reward-rate-maximizing agent would exhibit, with these effects becoming less pronounced the greater the underweighting. Finally, with regards to the Delay Effect (Figure 21E), the Malapportionment Hypothesis predicts that preference reversal would occur at delays greater than that exhibited by a reward-rate-maximizing agent, with the delay becoming more pronounced the greater the underweighting outside versus inside the considered pursuit by the agent.”

Reviewer #3 Recommendations:

As mentioned above, the readability of this paper should be improved so that the readers can follow the derivations and your analyses better. To this end, careful numbering of equations, following consistent equation numbering formats, and differentiating between appendix referencing and equation numbering would have gone a long way in improving the readability of this paper. Some specific questions are noted below.

To increase clarity, in the revision we eliminated numbering of equations in the appendices except where an equation occurs in an appendix that is referenced within the main text. In the main text, important equations are thus numbered sequentially as they appear and note the appendix from which they derive. If an equation in an appendix is referenced in the main text, it is noted within the appendix it derives.

(1) In general, it is unclear what the default pursuit is. From the schematic on the left (forgo decision), it appears to be the time spent in between reward-giving pursuits. However, this schematic also allows for smaller rewards to be attained during the default pursuit as do subsequent equations that reference a default reward rate. Here is where an example would have really benefited the authors in getting their point across as to what the default pursuit is in practice in the forgo decisions and how the default reward rate could be modulated.

(1) The description of the default pursuit has been modified in section “Forgo and Choice decision topologies” to now read… “After either the conclusion of the pursuit, if accepted, or immediately after rejection, the agent returns to a pursuit by default (the “default” pursuit). This default pursuit effectively can be a waiting period over which reward could be received, and reoccurs until the next pursuit opportunity becomes available.” (2) Additionally, helper text has been added to Ap1 regarding the meaning of time and reward spent in the default pursuit. Finally, (3) new figures concerning n-pursuits occurring at the same (Supplement 1) or different (Supplement 2) frequencies from a default pursuit is now added, providing examples as suggested by the reviewer.

(2) I want to clarify my understanding of the topologies in Figure 1. In the forgo, do they roam in the "gold" pursuit indefinitely before they are faced with the purple pursuit? In general, comparing the 2 topologies, it seems like in the forgo decision, they can roam indefinitely in the gold topology or choose the purple but must return to the gold.

The reviewer’s understanding of the topology is correct. The agent loops across one unit time in the default gold pursuit indefinitely, though the purple pursuit (or any pursuit that might exist in that world) occurs on exit from gold at its frequency per unit time. The default gold pursuit will then itself have an average duration in units of time spent in gold. As the reviewer states, the agent can re-enter into gold from having exited gold, and can enter gold from having exited purple, but cannot re-enter purple from having exited purple; rather, it must enter into the default pursuit.

…Another point here is that this topology is highly simplified (only one considered pursuit). So it may be helpful to either add a schematic for the full topology with multiple pursuits or alternatively, provide the corresponding equations (at least in appendix 1 and 2) for the simplified topology so you can drive home the intuition behind derived expressions in these equations.

We understand the reviewer to be noting that, while, the illustrated example is of the simple topology, the mathematical formulation handles the case of n-number pursuits, and that illustrating a world in which there are a greater number of pursuits, corresponding to original appendices 1&2, would assist readers in understanding the generality of these equations.

An excellent suggestion. We have now n-pursuit world illustrations where each pursuit occurs at the same (Supplemental Figure 1) and at different frequencies (Supplemental Figure 2) to the manuscript, and have added text to assist in understanding the form of the equation and its relationship to unit time in the default pursuit in the main and in the appendices.

(3) In Equation and Appendix 1, there are a few things that are unclear. Particularly, why is the expected time of the default option E(t_default )= 1/(∑_(i=1)^n f_i )? Similarly, why is the E(r_default )= ρ_d/(∑_(i=1)^n f_i )? Looking at the expression for E(r_default ), it implies that across all pursuits 1 through n, the default option is encountered only once. Ultimately, in Equation 1.4, (and Equation 1), the units of the two terms in the numerator don't seem to match. One is a reward rate (ρ_d) and the other is a reward value. This is the most important equation of the paper since the next several equations build upon this. Therefore, the lack of clarity here makes the reader less likely to follow along with the analysis in rigorous detail. Better explanations of the terms and better formatting will help alleviate some of these issues.

The equation is formulated to calculate the average reward received and average time spent per unit time spent in the default pursuit. So, f_i is the encounter rate of pursuit i for one unit of time spent in the default pursuit. Added to the summation in the numerator we have the average reward obtained in the default pursuit per unit time () and in the denominator we have the time spent in the default pursuit per unit time (1).

Text explaining the above equation has been added to Ap 1.

(4) In equation and appendix 2, I'm trying to relate the expressions for t_out and r_out to the definitions "average time spent outside the considered pursuit". If I understand the expression in Equation 2.4 on the right-hand side, the numerator is the total time spent in all of the pursuits in the environment and the denominator refers to the number of times the considered pursuit is encountered. It is unclear as to why this is the average time spent outside the considered pursuit. In my mind, the expression for average time spent outside the considered pursuit would look something like t_out=1+ ∑_(i≠in)〖p_i t_i 〗= 1+ ∑_(i≠in)〖f_i/(∑_(j=1)^n f_j ) * t_i 〗. It is unclear how these expressions are then equivalent.

Regarding the following equation,

f_i is the probability that pursuit i will be encountered during a single unit of time spent in the default pursuit. The numerator of the expression is the average amount of time spent across all pursuits, excepting the considered pursuit, per unit time spent in the default pursuit. Note that the + 1 in the numerator is accounting for the unit of time spent in the default pursuit and is added outside of the sum. Since f_in is the probability that the considered pursuit will be encountered per unit of time spent in the default pursuit, is the average amount of time spent in the default pursuit between encounters of the considered pursuit. By multiplying the average time spent across all outside pursuits per unit of time in the default pursuit by the average amount of time spent in the default pursuit between encounters of the considered pursuit, we get the average amount of time spent outside the considered pursuit per encounter of the considered pursuit. This is calculated as if the pursuit encounters are mutually exclusive within a single unit of time spent within the default pursuit, as this is the case as the length of our unit time (delta t) approaches zero.

The above text explaining the equation has been added to Ap 2.

(5) In Figure 3, one huge advantage of this separation into in-pursuit and out-of-pursuit patches is that the optimal reward rate maximizing rule becomes one that compares ρ_in and ρ_out. This contrasts with an optimal foraging rule which requires comparing to the global reward rate and therefore a circularity in solution. In practice, however, it is unclear how ρ_out will be estimated by the agent.

How, in practice, a human or animal estimates the reward rates―be they the outside and/or global reward rate under a policy of accepting a pursuit―is the crux of the matter. This work identifies equations that would enable a reward-rate maximizing agent to calculate and execute optimal policies and emphasizes that the effective reward rates and weights of pursuits must be accurately appreciated for global reward rate optimization. In so doing, it makes a reckoning of behaviors commonly but erroneously treated as suboptimal. Then, by examining the consequences of misestimation of these enabling parameters, it identifies mis-weighting pursuits as the nature of the error committed by whatever algorithm and representational architecture is being used by humans and animals (the Malapportionment Hypothesis). This curious pattern identified and analyzed in this work thus provides a clue into the nature of the learning algorithm and means of representing the temporal structure of the environment that is used by humans and animals―the subject of future work.

We note, however, that we do discuss existing models that grapple with how, in practice, how a human or animal may estimate the outside reward rate. Of particular importance is the TIMERR model, which estimates the outside reward rate from its past experience, and can make an accounting of many qualitative features widely observed. However, while appealing, it would mix prior ‘in’ and ‘outside’ experiences within that estimate, and so would fail to perform forgo tasks optimally. Something is still amiss, as this work demonstrates.

(6) The apportionment time cost needs to be explained a little bit more intuitively. For instance, it is clear that the opportunity cost of time is the cost of not spending time in the rest of the environment relative to the current pursuit. But given the definition of apportionment cost here in lines 447- 448 "The apportionment cost relates to time's allocation in the world: the time spent within a pursuit type relative to the time spent outside that pursuit type, appearing in the denominator." The reference to the equation (setting aside the confusion regarding which equation) within the definition makes it a bit harder to form an intuitive interpretation of this cost. Please reference the equation being referred to in lines 447-448, and again, an example may help the authors communicate their point much better

We thank the reviewer for pressing on this critical point.

Action: We added the following succinct verbal description of apportionment cost… “Apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration.” This definition appears in a new paragraph (as below) describing apportionment cost in the results section “Time’s cost: opportunity & apportionment costs determine a pursuit’s subjective value”, and is accompanied by equations for apportionment cost, and a figure giving its geometric depiction (Figure 5).

“What, then, is the amount of reward by which the opportunity cost-subtracted reward is scaled down to equal the sv of the pursuit? This amount is the apportionment cost of time. The apportionment cost of time (height of the brown vertical bar, Figure 5F) is the global reward rate after taking into account the opportunity cost (slope of the magenta-gold dashed line in Figure 5F) times the time of the considered pursuit. Equally, the difference between the inside and outside reward rates, times the time of the pursuit, is the apportionment cost when scaled by the pursuit’s weight, i.e., the fraction that the considered pursuit is to the total time to traverse the world (Equation 9, right hand side). From the perspective of decision-making policies, apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration (Equation 9 center, Figure 5F).

Equation 9. Apportionment Cost.

While this difference is the apportionment cost of time, the opportunity cost of time is the amount that would be expected from a policy of not taking the considered pursuit over a time equal to the considered pursuit’s duration. Together, they sum to Time’s Cost (Figure 5G). Expressing a pursuit’s worth in terms of the global reward rate obtained under a policy of accepting the pursuit type (Figure 5 left column), or from the perspective of the outside reward and time (Figure 5 right column), are equivalent. However, the latter expresses sv in terms that are independent of one another, conveys the constituents giving rise to global reward rate, and provides the added insight that time’s cost comprises an apportionment as well as an opportunity cost.”

(7) The analyses in Figures 6 and 7 give a nice visual representation of how the time costs are distributed as a function of outside reward and time spent. However, without an expression for apportionment cost it is hard to intuitively understand these visualizations. This also relates to the previous point of requiring a more intuitive explanation of apportionment costs in relation to the opportunity cost of time. Based on my quick math, it seems that an expression for apportionment cost would be as follows: (r_in- ρ_out*t_in)*(t_in⁄t_out )/(t_in⁄t_out +1 ). The condition described in Figure 7 seems like the perfect place to compute the value of just apportionment cost when the opportunity cost is zero. It would be helpful to introduce the equation here.

We designed original figure 7, as the reviewer appreciates, to emphasize that time has a cost even when there is no opportunity cost, being due entirely to the apportionment cost of time.

We now provide the mathematical expression of apportionment cost and apportionment scaling in Figure 5, the point in the main text of its first occurrence.

…and have expanded original figure 5, its legend (so as to illustrate the apportionment scaling factor and the apportionment cost), and its accompanying main text, to further illustrate and clarify apportionment cost, and its relationship to opportunity cost, and time’s cost.

(8) The analysis regarding choice decisions is relatively straightforward, pending the concerns for the main equations listed above for the forgo decisions. Legends certainly would have helped me grasp Figures 10-12 better.

We believe the reviewer is referring to missing labels for the Sooner Smaller pursuit, and the Larger Later Pursuit in these figures? We used the same conventions as in Figure 9, but we see now that adding these labels to these figures would be helpful, and add them in the revision.

We have now added to the figures themselves figure legends indicating the Sooner Small Pursuit and the Larger Later Pursuit. We have also added to the main text to emphasize the points made in these figures regarding the impact of opportunity cost and apportionment cost.

(9) The derivation of the temporal discounting function from subjective reward rate is much appreciated as it provides further evidence for potential equivalence between reward rate optimization and hyperbolic discounting, which is known to explain a slew of decision-making behaviors in the economics literature.

We thank and greatly appreciate the reviewer for this recognition.

In response to the reviewer’s comment, we have added text that further relates reward rate optimization to hyperbolic discounting…

(1) We add discussion of how our normative derivation gives explanation to Mazur’s ad hoc addition of 1 + k to Ainslie’s reward/time hyperbolic discounting conception. See new first paragraph under “Hyperbolic Temporal Discounting Functions” for the historical origins of the standard hyperbolic equation (which are decidedly not normatively derived). And then see our discussion (new second paragraph in sections “The apparent discounting function of global….”) of how our normative derivation gives explanation to “1”, “k”, and their relationship to each other.

(2) We add explicit treatment of the Delay Effect in a new “The Delay Effect” section of the results along with a figure, and in its corresponding Discussion section.

Minor comments:

(1) Typo in equation 2, should be t_i in the denominator within the summation, not r_i .

We thank the reviewer for catching this typo, and have corrected it in the revision.

(2) Before equation 6, typo when defining ρ_in= r_in/(t_in.). Should be t_in in the denominator, not r_out.

We thank the reviewer for catching this typo, and have corrected it in the revision.

(3) Please be consistent with equation numbers, placement of equation references, and the reason for placing appendix numbers. This will improve readability immensely.

To increase clarity, in the revision we eliminated numbering of equations in the appendices except where an equation occurs in an appendix that is referenced within the main text. In the main text, important equations are thus numbered sequentially and note the appendix from which they derive. If an equation in an appendix is referenced in the main text, it is noted within the appendix it derives.

(4) Line 505 - "dominants" should be dominates.

Typo fixed as indicated

(5) Figures 10-12: add legends to the figures.

Now so included.

(6) Lines 701-703: please rewrite the equation separately. It is highly unclear what rt is here.

We thank the reviewer for bringing attention to this error. The error arose in converting from Google Sheets to Microsoft Word.

The equation has now been corrected.

Additional citations noted in reply and appearing in Main text

Ainslie, George. 1975. “Specious Reward: A Behavioral Theory of Impulsiveness and Impulse Control.” Psychological Bulletin 59: 257–72.

Frederick, Shane, George Loewenstein, Ted O. Donoghue, and T. E. D. O. Donoghue. 2002. “Time Discounting and Time Preference : A Critical Review.” Journal of Economic Literature 40: 351–401.

Gibbon, John. 1977. “Scalar Expectancy Theory and Weber’s Law in Animal Timing.” Psychological Review 84: 279–325.

Green, Leonard, Nathanael Fristoe, and Joel Myerson. 1994. “Temporal Discounting and Preference Reversals in Choice between Delayed Outcomes.” Psychonomic Bulletin & Review 1: 383–89.

Grüne-Yanoff, Till. 2015. “Models of Temporal Discounting 1937-2000: An Interdisciplinary Exchange between Economics and Psychology.” Science in Context 28 (4): 675–713.

Jimura, Koji, Joel Myerson, Joseph Hilgard, Todd S. Braver, and Leonard Green. 2009. “Are People Really More Patient than Other Animals? Evidence from Human Discounting of Real Liquid Rewards.” Psychonomic Bulletin & Review 16: 1071–75.

Kalenscher, Tobias, and Cyriel M. A. Pennartz. 2008. “Is a Bird in the Hand Worth Two in the Future? The Neuroeconomics of Intertemporal Decision-Making.” Progress in Neurobiology 84 (3): 284–315.

Kirby, Kris N., and R. J. Herrnstein. 1995. “Preference Reversals Due to Myopic Discounting of Delayed Reward.” Psychological Science 6 (2): 83–89.

Mazur, James E. 1987. “An Adjusting Procedure for Studying Delayed Reinforcement.” In The Effect of Delay and of Intervening Events on Reinforcement Value., 55–73. Quantitative Analyses of Behavior, Vol. 5. Hillsdale, NJ, US: Lawrence Erlbaum Associates, Inc.

McNamara, John. 1982. “Optimal Patch Use in a Stochastic Environment.” Theoretical Population Biology 21 (2): 269–88.

Rosati, Alexandra G., Jeffrey R. Stevens, Brian Hare, and Marc D. Hauser. 2007. “The Evolutionary Origins of Human Patience: Temporal Preferences in Chimpanzees, Bonobos, and Human Adults.” Current Biology: CB 17: 1663–68.

Strotz, R. H. 1956. “Myopia and Inconsistency in Dynamic Utility Maximization.” The Review of Economic Studies 23: 165–80.

Author response:

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

This valuable study combines multidisciplinary approaches to examine the role of insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2) as a potential novel host dependency factor for Zika virus. The main claims are partially supported by the data, but remain incomplete. The evidence would be strengthened by improving the immunofluorescence analyses, addressing the role of IGF2BP2 in "milder" infections, and elucidating the role of IGF2BP2 in the biogenesis of the viral replication organelle. With the experimental evidence strengthened, this work will be of interest to virologists working on flaviviruses.

We thank the reviewers for their feedback and constructive suggestions. In this revised version of the manuscript, we have addressed the reviewer’s comments to the best of our ability as detailed below. We believe that the newly incorporated data strengthens our study and conclusions. We hope that this revised manuscript will satisfy the reviewers and will be of high interest to flavivirologists.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This study investigated the co-option of IGF2BP2, an RNA-binding protein by ZIKV proteins. Designed experiments evaluated if IFG2BP2 co-localized to sites of viral RNA replication, interacted with ZIKV proteins, and how ZIKV infection changed the IGF2BP2 interactome.

Strengths:

The authors have used multiple interdisciplinary techniques to address several questions regarding the interaction of ZIKV proteins and IGF2BP2.

The findings could be exciting, specifically regarding how ZIKV infection alters the interactome of IGF2BP2.

We thank the reviewer for acknowledging the multidisciplinary approach of our study and its exciting potential.

Weaknesses:

Significant concerns regarding the current state of the figures, descriptions in the figure legends, and the quality of the immunofluorescence and electron microscopy exist.

In this new version of the manuscript, we have improved the quality of the microscopy data and included the requested information in the figure legends as described below in the Recommendations section.

Reviewer #2 (Public Review):

Clément Mazeaud et al. identified the insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2) as a proviral cellular protein that regulates Zika virus RNA replication by modulating the biogenesis of virus-induced replication organelles.

The absence of IGF2BP2 specifically dampens ZIKV replication without having a major impact on DENV replication. The authors show that ZIKV infection changes IGF2BP2 cellular distribution, which relocates to the perinuclear viral replication compartment. These assays were conducted by infecting cells with an MOI of 10 for 48 hours. Considering the ZIKV life cycle, it is noteworthy that at this time there may be a cytopathic effect. One point of concern arises regarding how the authors can ascertain that the observed change in localization is a consequence of the infection rather than of the cytopathic effect. To address this concern, shorter infection periods (e.g., 24 hours post-infection) or additional controls, such as assessing cellular proteins that do not change their localization or infecting with another flavivirus lacking the IGF2BP2 effect, could be incorporated into their experiments.

We thank the reviewer for these relevant comments regarding the specificity of IGF2BP2 relocalization to the ZIKV replication compartment.

It is noteworthy that we chose the 2-day post-infection time point for our analyses because it corresponds to the peak of replication with much more titers produced compared to those at 24 hours post-infection (generally ~106 PFU/mL vs. ~104 PFU/mL). Consistently, the abundance of viral replication factories is more obvious at this time-point. A MOI of 5-10 was chosen to maximize the % of infected cells. That said, as suggested by the reviewer, we have analyzed the distribution of IGF2BP2 in ZIKV-infected cells at one-day post-infection, and we provide evidence in Figure S1 that IGF2BP2 relocalizes to the dsRNA-containing compartment at this time point.

Importantly, we now show in Figure S5 that in contrast to IGF2BP2, other host RNA-binding proteins such as LARP1 and DDX5 do not accumulate to ZIKV replication compartment at 2 days post-infection. LARP1 actually seems to be excluded from it while DDX5 remains nuclear. Of note, consistent with the ZIKV-induced decrease in expression observed in western blots (Fig 4A), the intensity of DDX5 signal decreases in infected cells. Altogether, this demonstrates that the IGF2BP2 relocalization phenotype is specific and is not due to ZIKV-induced cell death.

By performing co-immunoprecipitation assays on mock and infected cells that express HAtagged IGF2BP2, the authors propose that the observed change in IGF2BP2 localization results from its recruitment to the replication compartment by the viral NS5 polymerase and associated with the viral RNA. Given that both IGF2BP2 and NS5 are RNA-binding proteins, it is plausible that their interaction is mediated indirectly through the RNA molecule. Notably, the authors do not address the treatment of lysates with RNase before the IP assay, leaving open the possibility of this indirect interaction between IGF2BP2 and NS5.

We agree with the hypothesis of the reviewer. As suggested, we have performed coimmunoprecipitation assays following RNase A treatment of the cell lysates. As shown in new Fig S6, the abundance of ZIKV NS5 co-immunoprecipitating with IGF2BP2-HA is drastically decreased upon RNase A treatment compared to the untreated condition. This demonstrates that the IGF2BP2/NS5 interaction is mostly RNA-dependent, which is not surprising as RNA is often a structural component of ribonucleoprotein complexes. Of note, the same is observed with ATL2. This new set of data allows us to refine our model of Figure 11 and the discussion as they strongly suggest that the direct binding of IGF2BP2 to viral RNA (evidenced in vitro; Fig 5D) is required for subsequent association with NS5 and ER-shaping protein ATL2. This is in line with the fact that viral RNA is a co-factor in the biogenesis of ER-derived ZIKV vesicle packets (PMID: 32640225). However, we cannot exclude a contribution of cellular RNA in these processes as discussed.   

In in vitro binding assays, the authors demonstrate that the RNA-recognition motifs of the IGF2BP2 protein specifically bind to the 3' nontranslated region (NTR) of the ZIKV genome, excluding binding to the 5' NTR. However, they cannot rule out the possibility of this host protein associating with other regions of the viral genome. Using a reporter ZIKV subgenomic replicon system in IGF2BP2 knock-down cells, they additionally demonstrate that IGF2BP2 enhances viral genome replication. Despite its proviral function, the authors note that the "overexpression of IGF2BP2 had no impact on total vRNA levels." However, the authors do not delve into a discussion of this latter statement.

We agree with the reviewer’s comments. We now mention in the discussion that we cannot exclude the possibility that IGF2BP2 associates with RNA motifs within the coding region of the viral genomic RNA, especially considering that it contains N6A-methylated sequences (PMID: 27773535; 27773536; 29373715). Moreover, we discuss the observation that IGF2BP2 overexpression has no impact on vRNA levels (as well as titers). We believe that this is because endogenous IGF2BP2 is highly expressed in cancer cells such as the Huh7.5 and JEG-3 cells used here and is presumably not limiting for viral replication in our system (PMID: 38320625; 35111811; 34309973; 35023719; 37088822; 33224879; 35915142).

In this study, the authors extend their findings by illustrating that ZIKV infection triggers a remodeling of IGF2BP2 ribonucleoprotein complex. They initially evaluate the impact of ZIKV infection on IGF2BP2's interaction with its endogenous mRNA ligands. Their results reveal that viral infection alters the binding of specific mRNA ligands, yet the physiological consequences of this loss of binding in the cell remain unexplored. 

We acknowledge that it would be of interest to further study the physiological relevance of the modulation of IGF2BP2 ribo-interactome. Since we have focused here on the role of IGF2BP2 in viral replication, we feel that this will be the focus of future studies notably involving a larger omic-centered approach to identify the most impacted IGF2BP2 mRNA ligands. Of note, Gokhale and colleagues have already reported that CIRBP, TNRC6A and PUM2 proteins regulates the replication of Flaviviridae (PMID: 31810760).

Additionally, the authors demonstrate that ZIKV infection modifies the IGF2BP2 interactome. Through proteomic assays, they identified 62 altered partners of IGF2BP2 following ZIKV infection, with proteins associated with mRNA splicing and ribosome biogenesis being the most represented. In particular, the authors focused their research on the heightened interaction between IGF2BP2 and Atlastin 2, an ER-shaping protein reported to be involved in flavivirus vesicle packet formation. The validation of this interaction by Western blot assays prompted an analysis of the effect of ZIKV on organelle biogenesis using a newly described replication-independent vesicle packet induction system. Consequently, the authors demonstrate that IGF2BP2 plays a regulatory role in the biogenesis of ZIKV replication organelles.

Based on these findings and previously published data, the authors propose a model outlining the role of IGF2BP2 in ZIKV infectious cycle, detailing the changes in IGF2BP2 interactions with both cellular and viral proteins and RNAs that occur during viral infection.

The conclusions drawn in this paper are generally well substantiated by the data.

We thank the reviewers for this encouraging general comments on our study.

However, it is worth noting that the majority of infections were conducted at a high MOI for 48 hours, spanning more than one infectious cycle. To enhance the robustness of their findings and mitigate potential cell stress, it would be valuable to observe these effects at shorter time intervals, such as 24 hours post-infection.

As explained above, IGF2BP2 relocalization to the (dsRNA-enriched) replication compartment was also observed in ZIKV infected cells at one day post-infection.

Furthermore, the assertion regarding the association of IGF2BP2 with NS5 could be strengthened through additional immunoprecipitation (IP) assays. These assays, performed in the presence of RNAse treatment, would help exclude the possibility of an indirect interaction between IGF2BP2 and NS5 (both RNA-binding proteins) through viral RNA, thus providing more confidence in the observed association.

See above for our answer and the description of the new data of Fig. S7.

Reviewer #3 (Public Review):

Summary:

The manuscript by Mazeaud and colleagues pursued a small-scale screen of a targeted RNAi library to identify novel players involved in Zika (ZIKV) and dengue (DENV) virus replication. Loss-of-function of IGF2BP2 resulted in reduced titers for ZIKV of the Asian and African lineages in hepatic Huh7.5 cells, but not for either of the four DENV serotypes nor West Nile virus (WNV). The phenotype was further confirmed in two additional cell lines and using a ZIKV reporter virus. In addition, using immunoprecipitation assays the interaction between IGF2BP2 and ZIKV NS5 protein and RNA genome was detected. The work addressed the role of IGF2BP2 in the infected cell combining confocal microscopy imaging, and proteomic analysis. The approach indicated an altered distribution of IGF2BP2 in infected cells and changes in the protein interactome including disrupted association with partner mRNAs and modulation of the abundance of a specific set of protein partners in IGF2BP2 immunoprecipitated ribonucleoprotein (RNP) complexes. Finally, based on the changes in IGF2BP2 interactome and specifically the increment in the abundance of Atlastin 2, the biogenesis of ZIKV replication organelles (vRO) is investigated using a genetic system that allows virus replication-independent assembly of vRO. Electron microscopy showed that knockdown of IGF2BP2 expression reduced the number of cells with vRO.

Strengths:

The role of IGF2BP2 as a proviral factor for ZIKV replication is novel. The study follows a logical flow of experiments that altogether support the assembly of a specialized RNP complex containing IGF2BP2 and ZIKV NS5 and RNA genome.

We thank the reviewer for their positive feedback on our study and its novelty.

Weaknesses:

The statistical analysis should clearly indicate the number of biological replicates of experiments to support statistical significance.

This information has been included in all figure legends.

The claim that IGF2BP2 knockdown impairs de novo viral organelle biogenesis and viral RNA synthesis is built upon data that show a reduction in RNA synthesis <0.5-fold as assessed using a reporter replicon, thus suggesting a limited impact of the knockdown on RNA replication.

We agree that a 50% decrease in the replication of our reporter replicon might be considered mild. However, we want to pinpoint that in an infectious set-up, the phenotypes were higher as demonstrated by an 80% decrease in viral particle production even when IGF2BP2 levels were never depleted more that 80% compared to endogenous levels. Moreover, our findings were validated through the analysis of de novo vRO biogenesis by electron microscopy in a replication-independent set-up. Together, these experiments provide compelling evidence for a role for IGF2BP2 in the early stages of viral genome replication.

Validation of IGF2BP2 partners that are modulated upon ZIKV infection (i.e. virus yield in knocked down cells) can be relevant especially for partners such as Atlastin 2, as the hypothesis of a role for IGF2BP2 RNP in vRO biogenesis is based on the observed increase in the abundance of Atlastin 2 in the RNP complex preciìtated from infected cells.

First, we would like to emphasize that the proviral role of ATL2 in flavivirus replication, including links to vRO biogenesis, was already reported in two independent studies notably by one of the co-authors (PMID: 31636417; 31534046). Therefore, we have chosen to discuss these previous studies in the manuscript rather than repeating published experiments.  Second, we agree that it would be interesting to further interrogate the role of modulated IGF2BP2 protein partners in ZIKV replication. However, these experiments would constitute a new project per se involving fastidious RNAi-based phenotypic screening and subsequent functional characterization of the identified hits. Therefore, this will be the focus of follow-up studies.  

Recommendations for the Authors:

Reviewer #1 (Recommendations For The Authors):

All IFAs claimed that showing co-localization is minimal, this needs to be addressed.

We have performed colocalization analyses for relevant images in the revised manuscript (see below and Figs. 4B, 5A, S4A-C and S5A-D. Although this quantification increases confidence in our analysis, we were still cautious in our conclusions, stating that colocalization was partial and that IGF2BP2 accumulates in the replication compartment.

Western blots and IPs need to be quantified.

As requested, we have included WB quantification in Figs. 2A, 4A, 4D, 8B-D, S6C and S7D.

Figure 1: What is the strain background for the ZIKV reporter virus?

As indicated in the legend of Figure 1E of the primary submission, the Rluc-expressing ZIKV reporter virus (ZIKV-R2A) was based on the FSS13025 isolate (Asian lineage)(PMID: 27198478). To clarify this, we have also indicated the strain background in the main text of the Results and Material & Methods sections.

Figure 2A: If shGF2BP2 reduces viral titer, the NS3 should show a reduction in 2A, but it doesn't.

We agree with the reviewer. Although NS3 seems not to be decreased upon IGF2BP2 knockdown in the experiment initially shown in Figure 2A, it should be noted that our homemade rat anti-NS3 antibody is highly sensitive, leading to signal saturation that makes it challenging to distinguish changes in NS3 expression without diluting substantially the lysate sample before the PAGE-SDS. The initial reason for including Fig 2A was not to make a statement about viral protein expression but to validate IGF2BP2 knock-down efficiency. Conclusions about NS3 levels in the initial figure are further complicated by the high MOI of ZIKV was used in Huh7.5 cells which are not quantitative for viral replication measurements. To address this issue, we assessed the impact of IGF2BP2 knockdown on viral protein abundance (as a read-out of overall viral replication) with a lower MOI of ZIKV. The results of the repeat experiment (seen in the new Fig. 2A) show that IGF2BP2 knockdown leads to a decrease in the abundance of NS4A, NS5 and NS3, which is consistent with the titer decrease phenotypes.

Figure S3: The re-localization claimed is minimal and does not show overlap with NS3. The dsRNA is difficult to see here. Suggest improving the immunofluorescence images and reducing the claim for "strong" co-option of RNP complexes.

In addition to replication complexes, NS3 labels convoluted membranes which are devoid of dsRNA and IGF2BP2 and surround the cage-like replication compartment as large puncta (PMID: 27545046; 33432690; 28249158). The signal overlap is more obvious between IGF2BP2 and NS3/dsRNA-containing areas, which is reflected by the Mander’s coefficients that have been included in the revised version (Fig. S5C-D). We have also adjusted the text to conclude that the colocalization was partial and that IGF2BP2 accumulated in the replication compartment. We acknowledge that the dsRNA signal is weak, and we have updated the images (and others, when relevant) to better visualize this viral component. Moreover, we have rephrased the sentence to remove the word “strongly”.

Figure 4A: Western blot needs quantification.

This is now included in the figure.

Figure 4B: As in many of the IFAs, the co-localization is only partial. Additionally, the dsRNA is not visible. So the images need to be improved. The colocalization should be quantified across the cell diameter.

We changed the color and intensity of the dsRNA staining to make it more visible. Mander’s colocalization coefficients have been determined and included in Figures 4B and S5C-D.

Figure 4C: It is difficult to understand what the +/- is on the blots for the cell extracts and the anti-HA IP samples. It is not described in the figure legend or the text.

As already indicated on the right of the panel, the +/- indicates whether or not IGF2BP2-HA was overexpressed in the cells. In the revised version, this is clarified in the figure legend.

Figure 5A: Once again similar to other IFAs, the co-localization is only minimal and thus difficult to claim as "co-localization" is actually happening. It would be good to either improve the images or discuss this observation in the text and reduce the claim of colocalization. Specifically, since the two proteins might be co-localizing in specific regions which would make it a very interesting observation. Also, quantification of co-localizing regions would be beneficial.

We have included the requested colocalization analysis. We have been cautious to indicate that colocalization was only partial. It is noteworthy that, despite many efforts in the optimization of the cell permeabilization procedure, we noticed that the FISH probes were not very efficient in accessing the perinuclear area of the infected cells, where replication complexes accumulate. In that respect, it is likely that this imaging approach “miss” some of the IGF2BP2/vRNA complexes and that the determined colocalization factor is underestimated. This explains why the confirmation of the vRNA/IGF2BP2 complex with a biochemical approach (Fig. 5B) was very relevant.

Figure 5D: It is unclear what the blue squares represent. Clearer figure legends and text would be beneficial.

As stated in the initial figure, the blue squares indicate values obtained with the ZIKV 5’ UTR probe while the green circles involve a 3’ UTR probe. We have further emphasized this information in the figure legend to make it clearer.

Figure 6B. The graph is missing the data and X-axis label for shIGF2BP2.

We had initially omitted the values of the conditions with shIGF2BP2 and the replicationdead GAA replicon, since this viral system does not allow accumulation of viral genomes or proteins and was not relevant at the 48h time point. We thought that the inclusion of the shNT/GAA condition was enough an internal negative control of viral replication since values for shIGF2BP2/GAA did not exceed background. Nevertheless, we have now included this condition in the revised figure.

Figure 7D: It is unclear what the -/+ signs are in the cell extracts and the IP blots. Specifically, since there is an NS5 signal in the (-) lanes.

As explained above, the +/- indicates whether IGF2BP2-HA was overexpressed. The meaning of these symbols is now further clarified in the figure legend.

Figure 8C: The circles with the different colors are not clearly described. What does it mean?

As indicated in the figure (left part), the red and green circles identify the partners of the STRING network whose association with IGF2BP2 is decreased and increased during infection, respectively. We have included this information in the figure legend.

Figure 9: The electron microscopy to quantify vesicles should be carried out using whole-cell tomography in order to get the most accurate quantification of the vesicles following different treatments. This is because if you only look at one cell profile (slice), the number of vesicles might be less in that profile and more in another below or above it. It is unclear how many cell profiles were used for the quantification and how the calculations were carried out.

We agree with the reviewer that ideally, one should perform 3D electron tomography to precisely assess the morphology of VPs. Regardless the fact that we do not possess the imaging infrastructure to perform that type of analysis, such an approach would represent a tremendous amount of work if one would like to process at least 200-400 vesicles from > 50 cells and their whole cytoplasm (as we did). Despite not having 3D images, this number of data points is sufficient to see general changes in viral replication vesicle morphology, especially considering that Huh7-Lunet cells are relatively flat cells. (PMID: 32640225; 36700643; 34696522; 31636417). Furthermore, since IGF2BP2 knockdown decreases the abundance of VPs and does not impact their diameter, we believe that the addition of sophisticated 3D analysis would not bring any new and relevant information and that the TEM data stand by themselves for the conclusion we made. A more refined morphological analysis to determine how IGF2BP2 is structurally involved in virus-mediated membrane reorganization could be the focus of a future study.

We feel that we have already provided sufficient information about the quantification in the Material & Methods section of the first version of the manuscript: “Quantification was performed by systematically surveying cells and evaluating the presence of VPs. Only cells with >2 VPs were considered as positive. For each condition, >50 cells were surveyed over 4 biological replicas. All observed VPs were imaged, and VP diameters were determined using ImageJ by measuring the distance across two axes and averaging”.

Reviewer #2 (Recommendations For The Authors):

The inclusion of a control in the knock-down and infection assays with the reporter virus could enhance the validity of the findings. Introducing STAT2 knockdown, a recognized antiviral protein for ZIKV, as a control would provide a valuable benchmark to evaluate the extent of viral enhancement in the experiments. This additional control not only supports the proposed function of LARP1 in virus assembly/release but also strengthens the overall interpretation of the results.

We agree that adding a positive control could have been relevant for assessing the extent of replication modulation, especially for increases such as that observed with shLARP1. However, finding such control proteins in our system was a challenge. Indeed, STAT2 would not have been a good control for these experiments since we used Huh7.5 cells for the RNAi mini-screening, which do not express a functional RIG-I protein, and generally do not produce type I and III interferons. Thus, STAT2 knockdown is not expected to result in an increase in replication. That said, we feel that it was unnecessary to include a control for replication inhibition here given that only a few statistically reliable candidates we obtained. Instead, we have opted for an extensive secondary validation approach by assessing the proviral role of IGF2BP2 for multiple viruses - DENV1-2-3-4, WNV and SARS-CoV-2, and 3 ZIKV strains in three relevant cell types.

Additionally, in Figure S4, the authors employ an antibody against NS5 that specifically recognizes ZIKV NS5 but not DENV NS5. Given the objective of highlighting distinctions between these two viruses, it is advisable to use an antibody that detects DENV NS5 as well. This approach would contribute to a more comprehensive comparison, ensuring a balanced representation of both viruses in the experimental analysis.

We thank the reviewer for this relevant suggestion. We have repeated the coimmunoprecipitation assays using antibodies specific to DENV NS5 (Aithor response image 1). While we specifically pulled down ZIKV NS5 with IGF2BP2-HA as expected, this was not the case for DENV NS5 when using extracts from DENV-infected cells despite our multiple attempts. Indeed, the amount of pulled-down DENV NS5 with IGF2BP2-HA was always comparable to that in the negative control (“empty” pWPI lentivirus-transduced cells, “-“ condition), which corresponds to non-specific binding to the HA-resin. Thus, while the antibody was very efficient at detecting DENV NS5 in the cell extracts, no specific binding between DENV NS5 and IGF2BP2-HA could be evidenced. Consistent with our different replication phenotypes between DENV and ZIKV, this strongly supports that the NS5/IGF2BP2 interaction is specific to ZIKV. The specificity of the IGF2BP2 interaction with ZIKV NS5 compared to DENV NS5 is discussed in the updated manuscript.

Author response image 1.

DENV NS5 is not specifically co-immunoprecipitated with IGF2BP2-HA in contrast to ZIKV NS5. Huh7.5 cells stably expressing IGF2BP2-HA (+) and control cells (-) were infected with ZIKV H/PF/2013 at a MOI of 10 or left uninfected. Two days later, cell extracts were prepared and subjected to RNase A treatment (+) or not (-) before anti-HA immunoprecipitations. The resulting complexes were analyzed by western blotting for their abundance in the indicated proteins.

Reviewer #3 (Recommendations For The Authors):

(1) Statistical analysis. Please clearly indicate what columns and error bars represent for bar graphs such as those presented in Figures 1A-D and F, Figures 2B-C, and bottom panels in DE, Figure 3, Figure 5B, Figure 6B-C, and Figures 9B-D and F. For instance, the mean of n independent experiments and standard deviation.

Information about the number of replicates, error bars, and statistical tests has been added for all figures in the legends. 

(2) What is the scale in the Y-axis of Figure 2C? As shown, it is difficult to know what is the virus titer in knocked-down cells. Please use a linear scale or a log scale.

This is a linear scale of viral titers, which we have modified to make it clearer for the reader.

(3) Throughout the manuscript (e.g. Figures 1, 2, and 3) the fold reduction in titer is presented instead of the actual virus titers. I suggest showing the titer as it may be much more informative for the reader.

We prefer showing the data as fold reduction as they better reflect the IGF2BP2 knockdowninduced phenotypes across the independent biological replicates. Indeed, from one experiment to another, the reference titers in the control condition sometimes varies because of the cell passage or the lentiviral transduction efficiency for instance, especially when low multiplicities of infection are used. However, the reduction phenotype in foldchange observed upon IGF2BP2 knockdown was always consistent regardless of the titer value.  Of note, all considered experiments had reference titers above 105 PFU/mL.

(4) Is it possible to perform a colocalization analysis of confocal images showing overlapping signals?

This has been done and the results of these analyses are included in the updated figures 4B, 5A, S4 and S5.

(5)  Assessing the effect of Atlastin2 knockdown in virus yield and showing coimmunoprecipitation of Atlastin 2 with NS5 can add relevant information.

As mentioned in the discussion and above, ATL2 was already reported to be required for DENV and ZIKV replication in two independent studies (including one by one of the coauthors)(PMID: 31636417; 31534046). We have not tested whether ATL2 associates with NS5. However, new Fig. S7 of the revised manuscript shows that IGF2BP2/ATL2 is RNAdependent. This suggests that, as initially depicted in our model, IGF2BP2 associates with the ER (and thus, ATL2) after its binding to the viral RNA. Further interrogation into the role of atlastins in the flavivirus replication cycle is the focus of another ongoing IGF2BP2-unrelated study from one of the co-authors which will be reported elsewhere.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors investigate ligand and protein-binding processes in GPCRs (including dimerization) by the multiple walker supervised molecular dynamics method. The paper is interesting and it is very well written.

Strengths:

The authors' method is a powerful tool to gain insight into the structural basis for the pharmacology of G protein-coupled receptors.

Weaknesses:

Cholesterol may play a fundamental role in GPCR dimerization (as cited by the authors, Prasanna et al, "Cholesterol-Dependent Conformational Plasticity in GPCR Dimers"). Yet they do not use cholesterol in their simulations of the dimerization.

We thank Reviewer #1 for the positive comment on mwSuMD.

In the revised version of the manuscript, the section about the A_2A/D2 receptors dimerization has been removed because largely speculative. We agree that the lack of cholesterol in those simulations added uncertainty to the presented results.

Reviewer #2 (Public Review):

The study by Deganutti and co-workers is a methodological report on an adaptive sampling approach, multiple walker supervised molecular dynamics (mwSuMD), which represents an improved version of the previous SuMD.

Case-studies concern complex conformational transitions in a number of G protein Coupled Receptors (GPCRs) involving long time-scale motions such as binding-unbinding and collective motions of domains or portions. GPCRs are specialized GEFs (guanine nucleotide exchange factors) of heterotrimeric Gα proteins of the Ras GTPase superfamily. They constitute the largest superfamily of membrane proteins and are of central biomedical relevance as privileged targets of currently marketed drugs.

MwSuMD was exploited to address:

(1) Binding and unbinding of the arginine-vasopressin (AVP) cyclic peptide agonist to the V2 vasopressin receptor (V2R);

(2) Molecular recognition of the β2-adrenergic receptor (β2-AR) and heterotrimeric GDPbound Gs protein;

(3) Molecular recognition of the A1-adenosine receptor (A1R) and palmitoylated and geranylgeranylated membrane-anchored heterotrimeric GDP-bound Gi protein;

(4) The whole process of GDP release from membrane-anchored heterotrimeric Gs following interaction with the glucagon-like peptide 1 receptor (GLP1R), converted to the active state following interaction with the orthosteric non-peptide agonist danuglipron;

(5) The heterodimerization of D2 dopamine and A2A adenosine receptors (D2R and A2AR, respectively) and binding to a bi-valent ligand.

The mwSuMD method is solid and valuable, has wide applicability, and is compatible with the most world-widely used MD engines. It may be of interest to the computational structural biology community.

The huge amount of high-resolution data on GPCRs makes those systems suitable, although challenging, for method validation and development.

While the approach is less energy-biased than other enhanced sampling methods, knowledge, at the atomic detail, of binding sites/interfaces and conformational states is needed to define the supervised metrics, the higher the resolution of such metrics is the more accurate the outcome is expected to be. The definition of the metrics is a user- and system-dependent process.

The too many and ambitious case-studies undermine the accuracy of the output and reduce the important details needed for a methodological report. In some cases, the available CryoEM structures could have been exploited better.

The most consistent example concerns AVP binding/unbinding to V2R. The consistency with CryoEM data decreases with an increase in the complexity of the simulated process and involved molecular systems (e.g. receptor recognition by membrane-anchored G protein and the process of nucleotide exchange starting from agonist recognition by an inactive-state receptor). The last example, GPCR hetero-dimerization, and binding to a bi-valent ligand, is the most speculative one as it does not rely on high-resolution structural data for metrics supervision.

We praise Reviewer #2 for the detailed comment on the manuscript. In this revised version, the hetero-dimerization between A_2AR and D₂R has been removed. Also, results about GPCR case studies other than GLP-1R have been reduced and downgraded in importance to focus on the fundamental key points of the adaptive sampling method.  We agree that the consistency with cryoEM data tends to decrease with an increase in the complexity of the simulated process and involved molecular systems. While it is possible to approximate cryoEM results  our unbiased adaptive sampling technique finds its most interesting application in mechanistically unknown out-of-equilibrium processes rather than reproducing known experimental data perfectly. The simulated case studies we present showcase the versatility, speed and consistency of our adaptive method to explore energetically unbiased transitions.

Reviewer #3 (Public Review):

Summary:

In the present work, Deganutti et al. report a structural study on GPCR functional dynamics using a computational approach called supervised molecular dynamics.

Strengths:

The study has the potential to provide novel insight into GPCR functionality. An example is the interaction between loops of GPCR and G proteins, which are not resolved experimentally, or the interaction between D344 and R385 identified during the Gs coupling by GLP-1R. However, validation of the findings, even computationally through for instance in silico mutagenesis study, is advisable.

Weaknesses:

In its current form, the manuscript seems immature and in particular, the described results grasp only the surface of the complex molecular mechanisms underlying GPCR activation. No significant advance of the existing structural data on GPCR and GPCR/G protein coupling is provided. Most of the results are a reproduction of the previously reported structures.

We thank Reviewer #3 for the positive comment on the work. The revised manuscript focuses more on the GLP-1R and Gs case studies. We believe it addresses the weaknesses raised by showing the behaviour of key structural motifs and providing new hypotheses about GDP release.  

Reviewer #2 (Recommendations For The Authors):

In this methodological report, Deganutti and co-workers propose an improved version of supervised molecular dynamics (SuMD), named multiple walker SuMD (mwSuMD). Such an adaptive sampling method was challenged in simulations of complex transitions involving GPCRs, which are out of reach by classical MD.

Although less energy-biased than other enhanced sampling methods, mwSuMD requires knowledge of the atomic detail of the ligand-protein or protein-protein binding site/interfaces and the structural hallmarks of the states whose conversion the method is going to address. Such knowledge is, indeed, necessary to define the supervised metrics (e.g. distances, RMSD, etc), which is a user- and system-dependent process.

We classify mwSuMD as an adaptive, rather than enhanced, sampling method as it does not use any energy bias. We agree with the Reviewer that some knowledge of the system is required to productively set up the simulations, but this is the case for almost any MD advanced methods.  

The text requires improvement in the essential methodological details and cleaning of those parts is not properly instrumental in method validation.

While attempting to prove the widest possible applicability of the method, the authors exaggerated the number of examples, which, in spite of the increasing complexity were only summarily described. Please, limit the case studies to AVP binding/unbinding to V2R and the whole process of GDP release from membrane-anchored Gs following activation of GLP1R by danuglipron. The latter case, indeed, involves small ligand binding (danuglipron), small ligand dissociation (GDP), receptor activation, and activated receptor binding to membraneanchored G protein and G protein conformational transition instrumental to nucleotide depletion, which is already too much. In this framework, the cases of Gs-β2AR and Gi-A2R recognition are redundant. Most importantly, the case of D2R-A2AR heterodimerization and binding to a bi-valent ligand must be eliminated. The reason is that the case is not entirely based on the mwSuMD and the biased protein-protein interface does not rely on highresolution data (i.e. no structural model of D2R-A2AR dimer has been determined so far). Last but not least, the high intrinsic flexibility of the bi-valent ligand adds further indetermination to the computational experiment. Being too speculative, the case-study does not serve to model validation.

We thank the Reviewer for the suggestion. In the current revised form, the manuscript focuses on AVP binding/unbinding to V2R and the GLP-1R activation, Gs recognition and GDP release.

While eliminating the three case studies mentioned above, the remaining ones should be described more extensively and clearly, highlighting the most productive setup for each system. Incidentally, listing the performance parameters (e.g. distribution mode and minimum RMSD) of each simulation setting in Table S1 is worth doing.

More accuracy in the methodological description is needed.

As for the supervised metrics, the rationale behind the choice of a particular index and whether it is the outcome of a number of trials must be declared and the selected indices must be better defined. Here there are a few examples.

AVP-V2R case. It is not clear why the AVP centroids were computed on residues C1-Q4 (I suppose the Cα-atoms) and not on the Cα-atoms of the whole cyclic part (C1-C6). Along the same line, the choice of the Cα-atoms of four amino acid residues to compute the receptor binding-site centroids requires justification.

We have amended the text to clarify that all the heavy atoms of AVP residues C1-Q4, which are anticipated to bind deep into V₂R, were considered alongside V₂R residues part of the peptide binding site (Cα atoms only). From our experience, the choice of including side chains or not for the definition of centroids usually does not affect the supervision output. It should only affect the output of mwSuMD simulations based on the RMSD which considers the specific relative distance from the reference. However, a benchmark of the differences produced by divergent selections is beyond the scope of the present work.

GLP1R case. The statement: "Since the opening of TM1-ECL1 was observed in two replicas out of four, we placed the ligand in a favorable position for crossing that region of GLP-1R" is rather weak as a strategy to manually (?) define the input position of the ligand.

As stated in the manuscript, placing the agonist in that position was driven by preliminary 8 μs of classic MD simulations that pointed out the possible path for binding.  We agree with the Reviewer that there is still some degree of arbitrarity in it and for this reason, we have not presented structural details of the F06882961 binding path.

As for the supervised metrics, what does it mean "the distance between the ligand and GLP-1R TM7 residues L3797.34-F3817.36"? Was the distance computed between ligand and L379-F381 centroids? Also: "In the supervised stages, the distance between residues M386-L394 Gas of helix 5 (α5) and the GLP-1R intracellular residues R1762.46, R3486.37, S3526.41, and N4057.60 was monitored" was it an inter-centroid distance? Furthermore, "supervising the distance between AHD residues G70-R199 Gas and K300-L394Gas" was it the distance between the centroid of the AHD and the centroid of the C-terminal half of the Ras-like domain? In general, when more than two atoms are involved in distance calculation, please, specify if the distance is inter-centroid.

Also: "During the third phase, the RMSD of PF06882961, as well as the RMSD of ECL3 (residues A3686.57-T3787.33, Ca atoms), were supervised" was the RMSD computed without superimposing the ligand to estimate its roto-translations?

We have added details about the selections used for computing centroids throughout the methods section. For example, all the heavy atoms of F06882961 and the Ca atoms of L379-F381 were considered. RMSD values during GLP-1R activation were computed after superimposition on TM2, ECL1, and TM3 residues 170-240 (Ca atoms). This now has been specified in the text.

The authors considered the 7LCJ GLP1R-danuglipron complex as a fully active reference state instead of considering the receptor from a ternary complex with Gs. The ternary complex (7LCI) was indeed considered as a reference only in simulations of receptor-G protein recognition. 

7LCJ and 7LCI are both fully active states. The main difference is that in 7LCJ, Gs coordinates were not deposited. Indeed, their RMSD computed on the TMD Ca atoms and F06882961 is 0.63 Å and 0.54 Å, respectively.

Most importantly, the ternary complex chosen by the authors is not adequate as a reference for simulating the "opening" of the AHD because it bears a miniGs, hence, missing the AHD. In that framework, such an opening is rather vague and was not properly supervised by mwSuMD. The authors must repeat metrics supervisions by using, as a reference, the 6X1A ternary complex, which bears a displaced AHD. This would likely lead to a different path of GDP release.

To the best of our knowledge, there is no evidence that a specific open conformation of the AHD is linked to GDP release. In support, we note that in GPCR ternary complexes, the AHD is usually not modelled because of its high flexibility. The only body of evidence we are aware of is that AHD must open up to allow GDP release. For this reason,  we decided to supervise the distance between AHD and the Ras domain without using a reference.

In the statement: "The AHD opening was simulated starting from the best GLP-1R:Gs binding mwSuMD replica" the definition "best binding" requires clarification.

This has been amended, specifying that Replica 2 was considered the “best replica” due to the closed deviation to the cryoEM structure.

As for the case study on β2-AR-Gs recognition, I strongly suggest to eliminate it. However, I'd like to make some comments. The sentence: "the adrenergic β2 receptor (b2 AR) in an intermediate active state was downloaded from GPCRdb (https://gpcrdb.org/)" is vague as it does not indicate what intermediate active state structure was used. Since the goal of the case study was to probe the method in simulating receptor-G protein binding, it would have been better to start with a fully active state of the receptor like the 4LDO structure, employed by the authors only to extract epinephrine.

mwSuMD is designed to provide insights into structural transitions. We started from an intermediate active state of β2-AR in complex with adrenaline because resembling the most populated state stabilised by a full agonist according to NMR studies (DOI:10.1016/j.cell.2015.08.045); the fully-active β2-AR conformation is stabilized only after Gs binding. However, following the Reviewer’s suggestion, we have reduced the presented results for the β2-AR-Gs recognition.

Also in this case, it is not clear if the supervised receptor-G protein distance is between the centroid of the whole 7-helix bundle and the centroid of Gs α5. It is not clear why the TM6 RMSD concerned only the cytosolic end of the helix and did not include the kink region. With that selection, to estimate the outward displacement, RMSD should have been computed without superimposing the considered portion (once all remaining Cα-atoms of the receptors are superimposed).

As the Reviewer pointed out above, some knowledge of the system is required to set up mwSuMD. Using more generic metrics as we did in this case, like the distance between the whole TMD and Gs α5 represents a general approach applicable to other GPCRs, that should allow orthogonal metrics to evolve independently from the supervision.

As now specified in the text, the superimposition for RMSD calculation was performed on residues 40 to 140 Ca atoms, hence not considering TM6.

As for the A1R-Gi recognition, as already stated, I strongly suggest eliminating it. However, I'd like to add some comments. I would discourage the employment of an AlphaFold model for simulations deputed to model validation in general and, in particular, when highresolution structures are available. In this case, the authors would have used the 1GP2 structure of heterotrimeric Gi no matter if from the rat species.

Following the Reviewer’s suggestion, we have dramatically reduced the results presented for the A1R-Gi recognition. We considered 1GP2 for the simulations but H5 lacks the Cterminal six residues and therefore some extent of modelling was still necessary. However, we take the Reviewer’s comment on board and consider it for future work.

Also, the palmitoylation and geranylgeranylation process is quite tortuous and it is not clear why the NVT ensemble was employed in the second stage of equilibration. This is reflected also on the GLP1R case study.

We have amended the text to clarify this passage. The second NVT stage is required for stabilizing the G protein and its orientation in the simulation box. The figure below shows that a plateau of the Ca RMSD during the NVT step was reached after 700 ns for both Gi (black) and Gs (orange).

Author response image 1.

Here, it is not clear if the RMSD of α5 of Gi was computed with or without superposition.

The RMSD of α5  was computed after superimposing on A₁R residues 40-140 Ca atoms (the less flexible region of the receptor). We have now amended the text to report this information. 

Reviewer #3 (Recommendations For The Authors):  

Points to address:

(1) Root Mean Square Deviation (RMSD) data are often reported as minimum values. It would be useful to provide the average value along the stable part of the trajectories. From the plots in Figure 2ab, it seems that the minimum values reported in the paper are very far from the average ones and thus represent special cases that are seldom reached during simulation. The authors should clarify this point;

For the revised manuscript, we moved Figure 2 to the supplementary material and added average RMSD values for the most notable replicas in Figures 4e and S8a,b. As a reference, in the text, we now report RMSDs from our previous classic MD simulations (https://doi.org/10.1038/s41467-021-27760-0) of Gs:GLP-1R cryoEM structure (G_α = 6.18 ± 2.40 Å; G_β \= 7.22 ± 3.12 Å; G_γ = 9.30 ± 3.65 Å) which show how flexible G proteins bound to GPCRs are and give better context to the RMSD values we measured during mwSuMD simulations.

(2) The RMSD values reported in the paper always refer to single molecules or proteins. It would be useful to also report the RMSD computed over the whole complexes (ligand/GPCR or GPCR/G protein). It would provide a better metric for understanding the general distance between the results and the reference experimental structures;

We have now removed the results sections for A₁R and β₂ AR to focus on GLP-1R, whose RMSD is analyzed in detail in Figures 2, 3 and 4.

(3) A number of computational works investigated the GPCR/G protein interaction and these studies should be cited and discussed. Examples are the works from Mafi et al. 2023 (doi: 10.1038/s41557-023-01238-6), Fleetwood et al. 2020 (doi: 10.1021/acs.biochem.9b00842), Calderon et al. 2023 and 2024 (doi: 10.1021/acs.jcim.3c00805 and doi: 10.1021/acs.jcim.3c01574), Maria-Solano and Choi 2023 (doi: 10.7554/eLife.90773.1), Mitrovic et al. 2023 (doi: 10.1021/acs.jpcb.3c04897), and D'Amore et al. 2023 (doi: 10.1101/2023.09.14.557711). Many of these works focused on the activation of B2AR and the interaction with its G protein. In addition, Maria-Solano and Choi 2023 and D'Amore et al. 2023 also characterized the rotation of TM6 during the A1R and A2AR activation. Therefore, the claim "To the best of our knowledge, this is the first time an MD simulation captures the TM6 rotation upon receptor activation as results reported so far are largely limited to the TM6 opening and kinking55." is untimely;

We thank the Reviewer for the suggested references. We have added them to the introduction as examples of energy-biased (Calderon et al. 2023 and 2024, Maria-Solano and Choi, Mitrovic et al., D'Amore et al) or adaptive sampling (Fleetwood et al) approaches to GPCR. Since the above articles focus on β₂  AR and A₁R, we do not discuss them in detail because the results sections for A₁R and b₂ AR have been drastically reduced in the manuscript.

We note that among the suggested references, only Mafi et al report about a simulated G protein (in a pre-formed complex) and none of the work sampled TM6 rotation without input of energy. However, we have removed the claim from the text.   

(4) In the discussion section, the authors claim that a distance-based approach can be employed when the structural data of the endpoints is limited. However, the results obtained from the distance-based protocol during the validation of the approach, which was done using V2R as a reference, are unsatisfying, as acknowledged by the authors themselves. For instance, the RMSD mode value reported for the AVP C alpha atoms with respect to 7DW9 is high, 0.7 nm, whereas the minimum value is 0.38 nm. In addition, some side chains are not oriented in the experimental conformation and might have a different interaction pattern with the receptor if compared with the experimental structure. Considering that in this case the endpoint is known, it is plausible that the performance of the method would degrade even further when data about the target structure is limited. In a real case scenario, the ligand binding mode is unknown and in such a case no RMSD matrix can be used. This represents the major concern of this study that is no prediction is provided, but only - rather inaccurate - reproduction of the known structural data;

The goal of the first part of the work was to compare mwSuMD to SuMD to justify its application on ligand binding using a challenging case study like vasopressin. The general validation of the parent method SuMD as a predictive tool for ligand binding mode has been extensively reported over the years (a few examples: https://doi.org/10.1021/ci400766b ; https://doi.org/10.1021/acs.jcim.5b00702 ; https://doi.org/10.1038/s41598-020-77700-z) and fell beyond the scope of this work. 

(5) In the discussion, the authors write "A complete characterization of the possible interfaces between GPCR monomers, which falls beyond the goal of the present work, should be achieved by preparing different initial unbound states characterized by divergent relative orientations between monomers to dynamically dock." It would be useful for the reader to refer to and cite here advanced computational approaches that allow a comprehensive sampling of GPCR dimerization independently from the starting conformation of the receptors. One example is coarse-grained metadynamics as shown in doi: 10.1038/s41467-023-42082-z;

The A<sub>2A</sub/D<sub>2</sub receptors dimerization has been removed from the manuscript. 

(6) In many cases, it is not reported how residues missing from the experimental structures used to model the proteins were reconstructed. This information is important, considering that the authors comment on the results of their calculations on addressing these regions, such as in the case of B2AR. Furthermore, the authors did not report how their initial models were validated. The authors should also explain why they did not model the IC loops of A2AR and D2R;

In the current version of the manuscript, for V2R ECL2 and GLP-1R, we specify that we produced 10 solutions with Modeller and considered the best one in terms of the DOPE score. 

The only receptor model used,  β₂ AR, is now presented as preliminary data focusing on Gs and avoiding any structural detail of the Gs recognition. 

As reported above the A2A-D2 dimerization has been removed from the manuscript.

(7) In several cases, the authors state that residues never investigated before play an important role in the interaction between different proteins. An example is provided on page 6 for the B2AR/G protein association. Since this claim is quite significant, it would benefit from validation, at least for further calculations such as in silico mutagenesis studies. Another example is at the end of page 10 where the authors report a hidden interaction between D344 and R385 that is pivotal for Gs coupling by GLP-1R. Is there other evidence supporting this result (previously reported literature data, conservation rate of these residues, etc.)?;

We have removed the supplementary table reporting B2AR/G protein interactions to reduce speculations and added a reference that reports GLP-1 EC50 reduction upon mutation of position 344 to Ala (https://doi.org/10.1021/acscentsci.3c00063).

(8) The authors should provide a deeper discussion about the conformational rearrangement of GPCR and G protein during the coupling. In detail, the conformational changes of microswitich amino acids of GPCR (e.g., PIF, NPxxY, inactivating ionic lock) and alpha helix 5 of G proteins should be discussed in relation to the literature data and experimental structures;

We have removed the A1R and b2 AR results to focus on GLP-1R. Key structural motifs in the polar central network and TM6 kink are analyzed more in detail in Figure 3.

(9) The chronology of the conformational changes of GLP-1R is arbitrarily chosen. During the simulation, the RMSD values reported in Fig. 3 are high and do not demonstrate the full accomplishment of the simulation of the activation process of the receptor;

We agree with the Reviewer that the GLP-1R inactive to active transition was not fully accomplished, compared to other work on class A GPCRs.  Unlike class A, class B GPCRs represent a challenging system to work with in silico because inactive starting conformations (e.g 6LN2) are extremely distant from the active one (e.g 7LCJ, 7LCI or 6X18), as demonstrated in Figure S6 for GLP-1R. Here we report the first attempt to model a class B GPCR activation mechanism starting from the inactive state, and even if not fully achieved, we believe it represents state-of-the-art simulations for this class of receptors.

(10) It would be helpful for the reader not familiar with the employed technique that the authors explain in one sentence in the main text the pros and cons of using multiple walkers instead of single walker SuMD;

We thank the Reviewer for the excellent suggestion. In the Discussion, we have now commented that: “more extensive sampling obtainable by seeding multiple parallel short simulations instead of a single simulation for batch”, while in the Methods we explain that “mwSuMD is designed to increase the sampling from a specific configuration by seeding user-decided parallel replicas (walkers) rather than one short simulation as per SuMD. Since one replica for each batch of walkers is always considered productive, mwSuMD gives more control than SuMD on the total wall-clock time used for a simulation. On the flip side, mwSuMD requires multiple GPUs to be the most effective, although any multi-threaded GPU can run more walkers on the same hardware keeping the sampling variety.”.

Minor points to address:

(11) Page 3: the following sentence is duplicated (also found on page 2) "GPCRs preferentially couple to very few G proteins out of 23 possible counterparts";

(12) Page 20: Figure S13 refers to the QM validation of PF06882961 torsional angle, not to the image of the receptor conformational changes, which is instead Figure S14 (please correct figure caption).

We thank the Reviewer for the accurate reading of the manuscript. These typos have been corrected.

Author Response

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

eLife assessment

This important study convincingly shows that the less common D-serine stereoisomer is transported in the kidney by the neutral amino acid transporter ASCT2 and that it is a noncanonical substrate for sodium-coupled monocarboxylate transporter SMCTs. With a multihierarchical approach, this important study further shows that Ischemia-Reperfusion Injury in the kidney causes a specific increment in renal reabsorption carried out, in part, by ASCT2.

Public Reviews:

Reviewer #1 (Public Review):

Most amino acids are stereoisomers in the L-enantiomer, but natural D-serine has also been detected in mammals and its levels shown to be connected to a number of different pathologies. Here, the authors convincingly show that D-serine is transported in the kidney by the neutral amino acid transporter ASCT2 and as a non-canonical substrate for the sodium-coupled monocarboxylate transporter SMCTs. Although both transport D-serine, this important study further shows in a mouse model for acute kidney injury that ASCT2 has the dominant role.

Strengths:

The paper combines proteomics, animal models, ex vivo transport analyses, and in vitro transport assays using purified components. The exhaustive methods employed provide compelling evidence that both transporters can translocate D-serine in the kidney.

Weakness:

In the model for acute kidney injury, the SMCTs proteins were not showing a significant change in expression levels and were rather analysed based on other, circumstantial evidence. Although its clear SMCTs can transport D-serine its physiological role is less obvious compared to ASCT2.

We greatly value the reviewer's efforts and feedback in reviewing our manuscript. We acknowledge the reviewer's observation that the changes indicated by our proteomic results are not markedly pronounced. To reinforce our findings, we have incorporated an analysis of gene alterations at the single-cell level (snRNA-seq) from the publicly accessible IRI mouse model data (Figure supplement 7). The snRNA-seq data align with our proteomic data in terms of the general trend of gene/protein alterations, but reveal more substantial changes in both ASCT2 and SMCTs. These discrepancies might stem from the different quantification methods used, suggesting a possible underestimation in our label-free proteomic quantification. The differences we see between the functional changes in transporters and their quantification in proteomics can be explained by the unique challenges posed by membrane proteins. Post-translational modifications and the complex nature of multiple transmembrane domains often impact the accurate measurement of these proteins in proteomic studies. This complexity can lead to a mismatch between the actual functional changes occurring in the transporters and their perceived abundance or alterations as detected by proteomic methods (Figure 4A) (Schey KL et al. Biochemistry 2015, doi: 10.1021/bi301604j). However, this label-free quantitative proteomics approach is well-suited for our study, given its screening efficiency, compatibility with animal models, and the absence of a labeling requirement. We may consider incorporating alternative quantitative proteomic methods in future for a more thorough comparison. We have included these considerations in lines 351-356 of the revised manuscript.

Manuscript lines 351-356

“When evaluating the extent of gene/protein alterations between the control and IRI conditions, we observed that the gene alterations of both Asct2 and Smcts, as revealed by snRNAsequencing, are more pronounced than the protein alteration ratios obtained from proteomics. This discrepancy may stem from difficulty in the quantification method, especially for membrane transport proteins in label-free quantitative proteomics.”

Regarding the roles of ASCT2 and SMCTs in renal D-serine transport, snRNA-seq showed that ASCT2 expression in the controls is less than 10% of the cell population. We suggest that ASCT2 contributes to D-serine reabsorption because of its high affinity and SMCTs (SMCT1 and SMCT2) would play a role in D-serine reabsorption in the cells without ASCT2 expression. In addition, we included other factors (the turnover rate and the presence of local canonical substrates) that may determine the capability of D-serine reabsorption. We have included this suggestion in the Discussion lines 386-404.

Manuscript lines 386-404

“Kinetics analysis of D-serine transport revealed the high affinity by ASCT2 (Km 167 µM) (Foster et al., 2016) and low affinity by SMCT1 (Km 3.39 mM; Figure 5E). In addition to transport affinity, the expression levels and co-localization of multiple transporters within the same cells are critical for elucidating the physiological roles of transporters or transport systems (Sakaguchi et al., 2024). In our proteome data, the chromatogram intensities of Smct1 (2.9 x 109 AU) and Smct2 (1.6 x 108 AU) were significantly higher than that of Asct2 (1.5 x 107 AU) in control mice (Table 1: abundance in Sham). While direct intensity comparisons between different proteins in mass spectrometry analyses are not precise, they can provide a general indication of relative protein amounts. This finding aligns with the snRNA-seq data, where Asct2 expression was found to be minimal, present in less than 10% of cell populations under both control and IRI conditions, suggesting that many cells do not express Asct2. Conversely, Smct1 and Smct2 show high and ubiquitous expression in control conditions, but their levels are markedly reduced in IRI conditions (Figure supplement 7). Our ex vivo assays demonstrate that both ASCT2 and SMCTs mediate D-serine transport (Figure 7B). Consequently, Asct2 may contribute to D-serine reabsorption due to its high affinity, whereas Smcts, owing to their abundance, particularly in cells lacking Asct2, likely play a significant role in D-serine reabsorption. Moreover, factors such as transport turnover rate (Kcat) and the presence of local canonical substrates are also vital in defining the overall contribution of Dserine transport systems.”

Reviewer #2 (Public Review):

Summary:

The manuscript "A multi-hierarchical approach reveals D-1 serine as a hidden substrate of sodium-coupled monocarboxylate transporters" by Wiriyasermkul et al. is a resubmission of a manuscript, which focused first on the proteomic analysis of apical membrane isolated from mouse kidney with early Ischemia-Reperfusion Injury (IRI), a well-known acute kidney injury (AKI) model. In the second part, the transport of D-serine by Asct2, Smct1, and Smct2 has been characterized in detail in different model systems, such as transfected cells and proteoliposomes.

Strengths:

A major problem with the first submission was the explanation of the link between the two parts of the manuscript: it was not very clear why the focus on Asct2, Smct1, and Smct2 was a consequence of the proteomic analysis. In the present version of the manuscript, the authors have focused on the expression of membrane transporters in the proteome analysis, thus making the reason for studying Asct2, Smct1, and Smct2 transporters more clear. In addition, the authors used 2D-HPLC to measure plasma and urinary enantiomers of 20 amino acids in plasma and urine samples from sham and Ischemia-Reperfusion Injury (IRI) mice. The results of this analysis demonstrated the value of D-serine as a potential marker of renal injury. These changes have greatly improved the manuscript and made it more convincing.

We deeply appreciate the reviewer’s comments on the manuscript. We have responded to the recommendations one by one in the later section.

Reviewer #3 (Public Review):

Summary:

The main objective of this work has been to delve into the mechanisms underlying the increment of D-serine in serum, as a marker of renal injury.

Strengths:

With a multi-hierarchical approach, the work shows that Ischemia-Reperfusion Injury in the kidney causes a specific increment in renal reabsorption of D-serine that, at least in part, is due to the increased expression of the apical transporter ASCT2. In this way, the authors revealed that SMCT1 also transports D-serine.

The experimental approach and the identification of D-serine as a new substrate for SMCT1 merit publication in Elife.

The manuscript also supports that increased expression of ASCT2, even together with the parallel decreased expression of SMCT1, in renal proximal tubules underlies the increased reabsorption of D-serine responsible for the increment of this enantiomer in serum in a murine model of Ischemia-Reperfusion Injury.

Weaknesses:

Remains to be clarified whether ASCT2 has substantial stereospecificity in favor of D- versus L-serine to sustain a ~10-fold decrease in the ratio D-serine/L-serine in the urine of mice under Ischemia-Reperfusion Injury (IRI).

It is not clear how the increment in the expression of ASCT2, in parallel with the decreased expression of SMCT1, results in increased renal reabsorption of D-serine in IRI.

We thoughtfully appreciate the reviewer’s comment on the manuscript. Considering the alteration of D-/L-serine ratios, there are several factors including protein expression levels at both apical and basolateral sides, properties of the transporters (e.g. transport affinities, substrate stereoselectivities), and the expression of DAAO (D-amino acid oxidase) which selectively degrades D-amino acids. Moreover, the mechanism becomes more complicated when the transport systems of L- and D-enantiomers are different and have distinct stereoselectivities as in the case of serine. Future studies are required to complete the mechanism. However, we would like to explore the mechanism based on the current knowledge.

From this study, we identified ASCT2 and SMCTs (SMCT1 and SMCT2) as D-serine transport systems. We showed that SMCT1 prefers D-serine. Although we did not analyze ASCT2 stereoselectivity, based on the previous studies, ASCT2 recognizes both D- and Lserine with high affinities and slightly prefers L-enantiomer (Km of 18.4 µM for L-serine in oocyte expression system (Utsunomiya-Tate et al. J Biol Chem 1996) and 167 µM for Dserine in oocyte expression system (Foster et al. Plos ONE 2016), and the IC50 of 0.7 mM for L-serine and 4.9 mM for D-serine (in HEK293 expression systems, Foster et al. PLOS ONE 2016). The proteomics showed an increase of ASCT2 (1.6-fold increase) and a decrease of SMCTs (1.7-fold decrease in SMCT1, and 1.3-fold decrease in SMCT2) in IRI conditions. The table below summarizes D-serine transport by ASCT2 and SMCTs.

In the case of L-serine, ASCT2 and B0ATs (in particular B0AT3) have been revealed as L-serine transport systems in the kidneys (Bröer et al. Physiol Rev 2008; Singer et al. J Biol Chem 2009). Proteomics showed that B0ATs have higher expression levels than ASCT2 supporting the idea that B0ATs are the main L-serine transport system (Table S1: Abundance of B0AT1 = 1.34E+09, B0AT3 = 2.13E+08, ASCT2 = 1.46E+07). In IRI conditions, B0AT3 decreased 1.8 fold and B0AT1 decreased 1.1 fold. From these results, we included the contribution of B0ATs in L-serine transport in Author response table 1.

Author response table 1.

Taken together, we suggest that high ratios of D-/L-serine in IRI conditions are a combinational result of 1) increase of D-serine reabsorption by ASCT2 enhancement and SMCTs reduction and 2) decrease of L-serine reabsorption by B0ATs. We have included this suggestion in the Discussion lines 438-451.

Manuscript lines 438-451

“The enantiomeric profiles of serine revealed distinct plasma D/L-serine ratio, with low rations in the normal control but elevated ratios in IRI, despite the weak stereoselectivity of ASCT2 (Figure 1B). This observation suggested differential renal handling of D-serine compared to L-serine. While we identified SMCTs as a D-serine transport system, it has been reported that L-serine reabsorption is mediated by B0AT3 (Singer et al., 2009). We propose that the alterations in plasma and urinary D/L-serine ratios are the combined outcomes of: 1) transport systems for L-serine, and 2) transport systems for D-serine. In normal kidneys, the low plasma D/L-serine ratios could result from the efficient reabsorption of L-serine by B0AT3, coupled with the DAAO activity that degrades intracellular D-serine reabsorbed by SMCTs. In IRI conditions, our enantiomeric amino acid profiling revealed low plasma L-serine and high urinary L-serine (Figure supplements 1B, 2B). Additionally, the proteomic analysis indicated a reduction in B0AT3 levels (4h IRI/sham = 0.56 fold; 8h IRI/sham = 0.65 fold; Table S1). These observations suggest that the low L-serine reabsorption in IRI is a result of B0AT3 reduction.”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

This is a thorough study that was reviewed previously under the old system. I think the authors have strengthened their findings and have no further suggestions.

We appreciate reviewer 1 for his/her effort and comments, which greatly contributed to improving this manuscript.

Reviewer #2 (Recommendations For The Authors):

The experiments seem to me to have been well performed and the data are readily available.

Weaknesses:

More than weakness I would speak of discussion points: I have a few suggestions that may help to make the paper more accessible to a general audience.

(1) In the Introduction, when the authors introduce the term "micromolecules", it would be beneficial to provide a precise definition or clarification of what they mean by this term. Adding a brief explanation may help the reader to better understand the context.

Following the reviewer’s comment, we have included the explanation of the micromolecule and membrane transport proteins in lines 41-43.

Manuscript lines 41-43

“Membrane transport proteins function to transport micromolecules such as nutrients, ions, and metabolites across membranes, thereby playing a pivotal role in the regulation of micromolecular homeostasis.”

(2) In line 91, I suggest specifying that this is a renal IRI model.

Following the reviewer’s comment, we have added the information that it is a renal IRI model of AKI (lines 90-92).

Manuscript lines 90-92

“We applied 2D-HPLC to quantify the plasma and urinary enantiomers of 20 amino acids of renal ischemia-reperfusion injury (IRI) mice, a model of AKI and AKI-to-CKD transition (Sasabe et al., 2014; Fu et al., 2018).”

(3) Lines 167-168 state that Asct2 is localised to the apical side of the renal proximal tubules. Is there any expression of Asct2 in other nephron segments?

To our knowledge, there is no report of ASCT2 expression in other nephron segments. Our immunofluorescent data of the ASCT2 staining in the whole kidney at the low magnification and another region of Figure 3 (below) as well as immunohistochemistry from Human Protein Atlas (update: Jun 9th, 2023) did not show a strong signal of ASCT2 expression in other regions besides the proximal tubules. Thus, we conclude that ASCT2 is mainly expressed in proximal tubules, but not in other nephron regions.

Author response image 1.

(4) Lines 225-226: Have the authors expressed the candidate genes in HEK293 cells with ASCT2 knockdown?

This experiment was done by expressing the candidate genes in the presence of endogenous ASCT2. We have added the information in lines 225-227 to emphasize this process.

Manuscript lines 225-227

“Based on this finding, we utilized cell growth determination assay as the screening method even in the presence of endogenous ASCT2 expression. HEK293 cells were transfected with human candidate genes without ASCT2 knockdown.”

(5) Lines 254-255: why was D-serine transport enhanced by ASCT2 knockdown in FlpInTRSMCT1 or 2 cells?

We appreciate the reviewer to point out this data. We apologize for causing the confusion in the text. The total amount of D-serine uptake in the cells did not enhance but the net uptake (uptake subtracted from the background) was increased. This enhancement is a result of the lower background by ASCT2 knockdown. We have revised the texts and explained this result in more detail (lines 256-258).

Manuscript lines 256-258

“In the cells with ASCT2 knockdown, the background level was lower, thereby enhancing the D-[3H]serine transport contributed by both SMCT1 and SMCT2 (the net uptake after subtracted with background) (Figure 5C).”

(6) Line 265: The low affinity of SMCT1 for D-serine alone makes it an unlikely transporter for urinary D-serine.

We admitted the reviewer’s concern about the low affinity of SMCT1. However, Km at mM range is widely accepted for several low-affinity amino acid transporters such as proton-coupled amino acid transporter PAT1 (Km = 2 – 5 mM; Miyauchi et al. Biochem J 2010), cationic amino acid transporter CAT2A (Km = 3 – 4 mM; Closs et al. Biochem 1997), and large-neutral amino acid transporter LAT4 (Km = 17 mM; Bodoy et al. J Biol Chem 2005). In the kidneys, many compounds are well-known to be reabsorbed by the low-affinity but high-capacity (high-expression) transporters. Similarly, D-serine was reported to be reabsorbed by the low-affinity transporter (Kragh-Hansen and Sheikh, J Physiol 1984; Shimomura et al. BBA 1988; Silbernagl et al. Am J Physiol Renal Physiol 1999). Moreover, amino acid profile showed urinary D-serine in the range of 100 – 200 µM (Figure supplement 2). This concentration range could drive SMCT1 function (Figure 5). Combined with the high and ubiquitous expression of SMCT1, we propose that SMCT1 is a low-affinity but highcapacity D-serine transporter in the kidneys.

snRNA-seq is a method that can directly compare the expression levels between different genes within the same cells. From Figure supplement 7, expression of SMCT1 is much more abundant than ASCT2. ASCT2 was presented in less than 10% of cell population. It is possible that 90% of the cells that do not express ASCT2 use SMCT1 to reabsorb Dserine.

We have revised the Discussion regarding this comment (lines 386-404).

Manuscript lines 386-404

“Kinetics analysis of D-serine transport revealed the high affinity by ASCT2 (Km 167 µM) (Foster et al., 2016) and low affinity by SMCT1 (Km 3.39 mM; Figure 5E). In addition to transport affinity, the expression levels and co-localization of multiple transporters within the same cells are critical for elucidating the physiological roles of transporters or transport systems (Sakaguchi et al., 2024). In our proteome data, the chromatogram intensities of Smct1 (2.9 x 109 AU) and Smct2 (1.6 x 108 AU) were significantly higher than that of Asct2 (1.5 x 107 AU) in the control mice (Table 1: abundance in Sham). While direct intensity comparisons between different proteins in mass spectrometry analyses are not precise, they can provide a general indication of relative protein amounts. This finding aligns with the snRNA-seq data, where Asct2 expression was found to be minimal, present in less than 10% of cell populations under both control and IRI conditions, suggesting that many cells do not express Asct2. Conversely, Smct1 and Smct2 show high and ubiquitous expression in control conditions, but their levels are markedly reduced in IRI conditions (Figure supplement 7). Our ex vivo assays demonstrate that both ASCT2 and SMCTs mediate D-serine transport (Figure 7B). Consequently, Asct2 may contribute to D-serine reabsorption due to its high affinity, whereas Smcts, owing to their abundance, particularly in cells lacking Asct2, likely play a significant role in D-serine reabsorption. Moreover, factors such as transport turnover rate (Kcat) and the presence of local canonical substrates are also vital in defining the overall contribution of Dserine transport systems.”

(7) Line 316: The authors state that there is a high tubular D-serine reabsorption in IRI and in line 424 there is an inactivation of DAAO during the pathology. This suggests that there is a reabsorption of D-serine mediated by a transport system in the basolateral membrane domain of proximal tubular cells. Do the authors have any information about this transporter?

We agree with the reviewer that transporters at the basolateral membrane are important to complete the D-serine reabsorption in the kidney, and have included this issue in the original manuscript. We stated that transport systems at the basolateral side are necessary to be analyzed in order to complete the picture of D-serine transport systems in the kidney (lines 481-483 of the revised manuscript). However, we did not have any strong candidates for basolateral D-serine transport systems. Because we analyzed the proteome of BBMV, which concentrates on the apical membrane proteins, the analysis did not detect several transporters at the basolateral side.

(8) In lines 462-463, the authors state: "It is suggested that PAT1 is less active at the apical membrane where the luminal pH is neutral". However, the pH of urine in the proximal tubules is normally acidic due to the high activity of NH3. I suggest rewording this sentence.

Thank you for your comment. Proximal tubule (PT) is the first and the main region to maintain acid-base homeostasis in the kidney. In PT cells, NH3 secretes H+ to titrate luminal HCO3- and creates CO2, which is absorbed into PT cells and produces "new intracellular HCO3-", which is subsequently reabsorbed into the blood. Although ion fluxes in PT is to maintain the pH homeostasis, the pH regulation in both luminal and intracellular PT cells is highly dynamic. We totally agree with the reviewer and to follow that, we have revised the text by emphasizing the pH around PT segments, rather than the final urine pH, and leaving the discussion open for the possibility of PAT1 function in PT of normal kidneys (lines 474481).

Manuscript lines 474-481

“PAT1, a low-affinity proton-coupled amino acid transporter (Km in mM range), has been found at both sub-apical membranes of the S1 segment and inside of the epithelia (The Human Protein Atlas: https://www.proteinatlas.org; updated on Dec 7th, 2022) (Sagné et al., 2001; Vanslambrouck et al., 2010). PAT1 exhibits optimum function at pH 5 - 6 but very low activity at pH 7 (Miyauchi et al., 2005; Bröer, 2008b). Future research is required to address the significance of PAT1 on D-serine transport in the proximal tubule segments where pH regulation is known to be highly dynamic (Boron, 2006; Nakanishi et al., 2012; Bouchard and Mehta, 2022; Imenez Silva and Mohebbi, 2022).”

Reviewer #3 (Recommendations For The Authors):

The authors proposed that the increased expression of ASCT2, even together with the decreased expression of SMCT1/2, causes the increased renal reabsorption of D-serine that occurs in IRI. In the discussion, the main argument to sustain this hypothesis is the higher apparent affinity for D-serine of ASCT2 (<200 uM Km) versus SMCT1 (3.4 mM Km). In the Discussion section (page 18- 1st complete paragraph), the authors indicate that the Mass Spec intensities of SMCT1 and 2 are two and one order of magnitude higher respectively than that of ASCT2. This suggests that SMCT1 is clearly more expressed than ASCT2 in control conditions. IRI increments ASCT2 protein expression in brush-border membrane vesicle from kidney 1.6 folds and decreases that of SMCT1 0.6 folds. How this fold changes, even taking into account the lower Km of ASCT2 versus SMCT1 would explain the dramatic changes in the D-/L-serine ratios in plasma and urine in IRI? The authors might discuss whether other transport characteristics, even unknown (e.g., a higher turnover rate of ASCT2 vs SMCT1), would also contribute to the higher D-serine reabsorption in IRI.

SMCT1 shows some enantiomer selectivity for D- vs L-serine. At 50 uM concentration the transport is almost double for D. vs L-serine, but is ASCT2 stereoselective between the two enantiomers of serine? Some of the authors of this manuscript showed in a previous paper that the basolateral transporter Asc1 also participates in the accumulation of D-serine in serum caused by renal tubular damage. (Serum D-serine accumulation after proximal renal tubular damage involves neutral amino acid transporter Asc-1. Suzuki M et al. Sci Rep. 2019 Nov 13;9(1):16705 (PMID: 31723194)). Asc1 shows no stereoselectivity between L- and D-serine. Can the authors discuss possible mechanisms resulting in increased renal reabsorption of Dserine than L-serine in IRI with the participation of transporters with modest stereoselectivity for D- vs L-serine?

We appreciate the reviewer’s comments on the degree of protein alteration in proteomics, the functional contributions of ASCT2 and SMCTs, and the alteration of D/L ratios. We have included the possibilities of the technical concerns and the discussion on the roles of ASCT2 and SMCTs as follows.

• Regarding the expression levels, proteomics and snRNA-seq showed the same tendency that ASCT2 increase and SMCTs decrease in IRI conditions. However, the degrees of alterations are more contrast in snRNA-seq. This may be due to the difference in quantification methods and probably points out the underestimated quantification of membrane transport proteins in label-free proteomics. The accuracy of protein quantifications in the label-free proteomics are often impacted by the presence of post-translational modifications and multiple trans-membrane domains like in the case of the membrane transport proteins (Schey KL et al. Biochemistry 2015, doi: 10.1021/bi301604j). Alternative methods of quantitative proteomics may be added in the future for a more thorough comparison. We have added this issue in lines 351-356 of the revised version.

Manuscript lines 351-356

“When evaluating the extent of gene/protein alterations between the control and IRI conditions, we observed that the gene alterations of both Asct2 and Smcts, as revealed by snRNA-sequencing, are more pronounced than the protein alteration ratios obtained from proteomics. This discrepancy may stem from difficulty in the quantification method, especially for membrane transport proteins in label-free quantitative proteomics.”

• For the functional contributions of ASCT2 and SMCTs in the kidney, we admitted the reviewer’s concern about the low affinity of SMCT1. Following the reviewer’s comment, we have included other factors besides transport affinities, e.g. expression levels and turnover rates of the transporters. From the results of both proteomics and snRNA-seq, ASCT2 expression is significantly lower than SMCTs in the normal conditions. snRNA-seq showed that ASCT2 was presented in less than 10% of the cell population (Figure supplement 7). We propose that most of the cells that do not express ASCT2 may use SMCT1 to reabsorb D-serine. This topic was included in the revised manuscript lines 386-404.

Manuscript lines 386-404

“Kinetics analysis of D-serine transport revealed the high affinity by ASCT2 (Km 167 µM) (Foster et al., 2016) and low affinity by SMCT1 (Km 3.39 mM; Figure 5E). In addition to transport affinity, the expression levels and co-localization of multiple transporters within the same cells are critical for elucidating the physiological roles of transporters or transport systems (Sakaguchi et al., 2024). In our proteome data, the chromatogram intensities of Smct1 (2.9 x 109 AU) and Smct2 (1.6 x 108 AU) were significantly higher than that of Asct2 (1.5 x 107 AU) in the control mice (Table 1: abundance in Sham). While direct intensity comparisons between different proteins in mass spectrometry analyses are not precise, they can provide a general indication of relative protein amounts. This finding aligns with the snRNA-seq data, where Asct2 expression was found to be minimal, present in less than 10% of cell populations under both control and IRI conditions, suggesting that many cells do not express Asct2. Conversely, Smct1 and Smct2 show high and ubiquitous expression in control conditions, but their levels are markedly reduced in IRI conditions (Figure supplement 7). Our ex vivo assays demonstrate that both ASCT2 and SMCTs mediate D-serine transport (Figure 7B). Consequently, Asct2 may contribute to D-serine reabsorption due to its high affinity, whereas Smcts, owing to their abundance, particularly in cells lacking Asct2, likely play a significant role in D-serine reabsorption. Moreover, factors such as transport turnover rate (Kcat) and the presence of local canonical substrates are also vital in defining the overall contribution of D-serine transport systems.”

• As for the dramatic alterations of D/L-serine ratios juxtaposed with minimal changes in ASCT2 and SMCTs expression level, we cautiously refrain from drawing a definitive conclusion regarding the entire mechanism. This caution is grounded in the scientific understanding of a comprehensive elucidation of both L-serine transport systems and D-serine transport systems at both apical and basolateral membranes. Nevertheless, we would like to suggest a mechanism at the apical membrane based on the current knowledge.

For D-serine transport systems, we found ASCT2 and SMCTs contributions in this study. Meanwhile, L-serine was previously reported to be mediated mainly by the neutral amino acid transporters B0AT3 (in particular B0AT3; Bröer et al. Physiol Rev 2008; Singer et al. J Biol Chem 2009). Hence, the mechanism behind the alterations of D/L-serine ratios should include B0AT3 functions as well. In IRI conditions, B0AT3 decreased 1.8 fold. We suggest that high ratios of D-/L-serine in IRI conditions are a combined outcome of 1) increase of D-serine reabsorption by ASCT2 enhancement and SMCTs reduction, and 2) decrease of L-serine reabsorption by B0AT3. We have included this suggestion in the Discussion lines 438-451.

Manuscript lines 438-451

“The enantiomeric profiles of serine revealed distinct plasma D/L-serine ratios, with low ratios in the normal control but elevated ratios in IRI, despite the weak stereoselectivity of ASCT2 (Figure 1B). This observation suggested the differential renal handling of D-serine compared to L-serine. While we identified SMCTs as a Dserine transport system, it has been reported that L-serine reabsorption is mediated by B0AT3 (Singer et al., 2009). We propose that the alterations in plasma and urinary D/Lserine ratios are the combined outcomes of: 1) transport systems for L-serine, and 2) transport systems for D-serine. In normal kidneys, the low plasma D/L-serine ratios could result from the efficient reabsorption of L-serine by B0AT3, coupled with the DAAO activity that degrades intracellular D-serine reabsorbed by SMCTs. In IRI conditions, our enantiomeric amino acid profiling revealed low plasma L-serine and high urinary L-serine (Figure supplements 1B, 2B). Additionally, the proteomics analysis indicated a reduction in B0AT3 levels (4h IRI/sham = 0.56 fold; 8h IRI/sham = 0.65 fold; Table S1). These observations suggest that the low L-serine reabsorption in IRI is a result of B0AT3 reduction.”

• In the case of Asc-1, it was reported to be a D-serine transporter in the brain (Rosenberg et al. J Neurosci 2013). Suzuki et al. 2019 showed the increase of Asc-1 in cisplatin-induced tubular injury. Notably, the mRNA of Asc-1 is predominantly found in Henle’s loop, distal tubules, and collecting ducts but not in proximal tubules, and its protein expression level is dramatically low in the kidney (Human Protein Atlas: update on Jun 19, 2023). Furthermore, in this study, Asc-1 expression was not detected in the brush border membrane proteome. Consequently, we have decided not to include Asc-1 in the Discussion of this study, which primarily focuses on the proximal tubules.

Author response:

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

We thank the constructive criticism provided by the reviewers and editor. Based on these suggestions, we have thoroughly reworked the manuscript. More specifically but not limit:

(1) We have corrected the mistakes mentioned by the reviewers on a point-by-point basis.

(2) We have provided additional experimental evidences to explain the rationale behind selecting five miRNAs for q-PCR validation. Furthermore, we have elaborated on the reasons for focusing primarily on research related to cartilage.

(3) In response to concerns regarding overinterpretation in the manuscript, we have made more precise descriptions and revisions. Furthermore, we have added some details in our methods, including the addition of results showing the conservation of miR-199b-5p sequences between human and mouse species.

(4) We have provided additional details on the experiments, including the process for predicting target genes, timing of chondrocyte culture and other experimental operations.

(5) Finally, we have made additional revisions to the details of the figures to avoid any distortions and enhance the precision of the language.

Below please find our responses to the reviewers’ comments on a point-by-point basis. You also can track the changes in the modified manuscript. We believe that this revision has been substantially improved.

eLife assessment

The manuscript provides interesting evidence that miR-199b-5p regulates osteoarthritis and as such it may be considered as a potential therapeutic target. This finding may be useful to further advance the field.

Thank you for your positive comments.

Although the study is considered potentially clinically relevant, the evidence provided was deemed insufficient and incomplete to support the conclusions drawn by the authors.

Thank you for your critical comments and constructive advices. We have response point to point according to the reviewers’ questions and thoroughly re-working our manuscript. We hope the revised manuscript can be qualified to the criteria and be published on the journal of eLife.

Reviewer #1 (Public Review):

Summary:

In this manuscript, the authors observed that miR-199b-5p is elevated in osteoarthritis (OA) patients. They also found that overexpression of miR-199b-5p induced OA-like pathological changes in normal mice and inhibiting miR-199b-5p alleviated symptoms in knee OA mice. They concluded that miR-199b-5p is not only a potential micro-target for knee OA but also provides a potential strategy for the future identification of new molecular drugs.

Thanks for your comment.

Strengths:

The data are generated from both human patients and animal models.

Thanks for the positive comment.

Weaknesses:

The data presented in this manuscript is not solid enough to support their conclusions. There are several questions that need to be addressed to improve the quality of this study.

The following questions that need to be addressed to improve the quality of the study.

(1) Exosomes were characterized by electron microscopy and western blot analysis (for CD9, 264 CD63, and CD81). However, figure S1 only showed two sample WB results and there is no positive and negative control as well as the confused not clear WB figure.

Thank you for your suggestion. We acknowledge that a comprehensive identification of extracellular vesicles should include both positive and negative samples. However, in some of the initial studies we referenced, the positive and negative control were not mentioned1;2. In our study, we identified extracellular vesicles using a combination of electron microscopy, nanoparticle tracking analysis, and marker detection of exosomes. We agree that having negative samples would make our results more convincing, and we will include a negative control group in our future experiments. Additionally, we have provided clearer images in the revised version. (supplemental fig1 A)

Reference

(1) Ying W, Riopel M, Bandyopadhyay G, et al. Adipose Tissue Macrophage-Derived Exosomal miRNAs Can Modulate In Vivo and In Vitro Insulin Sensitivity. Cell. 2017;171(2).

(2) Fang T, Lv H, Lv G, et al. Tumor-derived exosomal miR-1247-3p induces cancer-associated fibroblast activation to foster lung metastasis of liver cancer. Nature Communications. 2018;9(1):191.

(2) The sequencing of miRNAs in serum exosomes showed that 88 miRNAs were upregulated and 89 miRNAs were downregulated in KOA patients compared with the control group based on fold change > 1.5 and p < 0.05. Figure 2 legend did not clearly elucidate what those represent and why the authors chose those five miRNAs to further validate although they did mention it with several words in line 108 'based on the p-value and exosomal'.

In fact, our study included two additional groups: the acupuncture treatment group (4 weeks of continuous acupuncture treatment) and the waiting treatment group (no intervention, followed by acupuncture treatment after 4 weeks), in addition to the healthy control and knee osteoarthritis (OA) patient groups. After comparing these four groups, we found that 11 genes (hsa-miR-504-3p, hsa-miR-1915-3p, hsa-miR-103a-2-5p, hsa-miR-887-3p, hsa-miR-1228-5p, hsa-miR-34c-3p, hsa-miR-3168, hsa-miR-518e-3p, hsa-miR-1296-5p, hsa-miR-338-3p, and hsa-miR-199b-5p) were upregulated in KOA patients but downregulated after acupuncture treatment, with no change in the waiting treatment group. Additionally, 7 genes (hsa-miR-448, hsa-miR-514a-3p, hsa-miR-4440, hsa-let-7f-5p, hsa-let-7a-5p, hsa-let-7d-5p, and hsa-miR-15b-3p) were downregulated in KOA patients but upregulated after acupuncture treatment, with no change in the waiting treatment group. Considering the improvement in clinical symptoms of KOA patients after acupuncture treatment, we believe that these 18 genes are of significant value. Based on overall expression abundance and species specificity, we finally selected 5 genes, namely the 5 genes mentioned in this article. Regarding this result, we have already included it in the supplementary fig5(fig. S5).

Author response image 1.

Venn diagram showing differentially expressed miRNAs in the OA group compared with healthy patients and patients who recovered after acupuncture treatment.

(3) In Figure 3 legend and methods, the authors did not mention how they performed the cell viability assay. What cell had been used? How long were they treated and all the details? Other figure legends have the same problem without detailed information.

Thank you for your suggestions. In Figure 3, cell viability was determined using the CCK-8 assay. We used second-generation chondrocytes for this analysis. The chondrocytes were obtained from young mice aged 3-5 days after birth. The cartilage tissues were extracted, and the cells were cultured in complete medium after digestion with collagenase. The detailed description of the cell viability assay, cell culture procedures, specific timing, and treatment methods of the cells used can be found in our revised manuscript. (page14-15，line304-313)

Besides, we have made thorough revisions to all figure legends to provide a clearer explanation of the relevant content.

(4) The authors claimed that Gcnt2 and Fzd6 are two target genes of miR-199b-5p. However, there is no convincing evidence such as western blot to support their bioinformatics prediction.

In the current study, we first identified six potential target genes by intersecting the predicted targets obtained from six bioinformatics websites. Subsequently, q-PCR was employed to test all six genes, revealing two genes with significant changes, namely Fzd6 and Gcnt2. We then predicted the binding sites of these genes and validated their existence through luciferase assays. Moreover, we examined the expression of these two potential targets in human KOA samples using a human database and found them to be expressed specifically in the samples. These results suggest that Fzd6 and Gcnt2 are potential target genes for KOA. However, we didn’t do western blot assay to verify the results. Based on your suggestions, we have further discussed the limitations of our study in this regard and proposed future research strategies.

(5) To verify the binding site on 3'UTR of two potential targets, the authors designed a mouse sequence for luciferase assay, but not sure if it is the same when using a human sequence.

Thank for your great advice. We carried out the comparative analysis of sequence conservatism between human and mouse, and find the binding site on 3'UTR matches to human sequence very well. The sequence conservation between hsa_miR-199b-5p and mmu_miR-199b-5p was as high as 95.65%. We added the methods and results in the revised manuscript. (page9, line181-184; page17, line361-365) (supplemental fig6).

In detail: Firstly, the sequence information of mmu_miRNA-199b-5p was used to locate the human homologous sequence in the UCSC database. The homologous sequence was found to be located in the human genome at chr9:128244721-128244830 (supplemental fig6 A). Based on this positional information and the source gene, a further comparison was conducted in miRbase to identify the nearest miRNA at the position of the human genome. It was discovered that hsa_miR-199b-5p is positionally conserved and located at chr9:128244721-128244830 (supplemental fig6 B). The sequence of hsa_miR-199b-5p was obtained from the miRbase database (supplemental fig6 C), and a comparative analysis was performed between the sequences of humans and mouse (supplemental fig6 D). Besides being positionally conserved, the sequence conservation between hsa_miR-199b-5p and mmu_miR-199b-5p was as high as 95.65%, indicating a good sequence conservation.

Author response image 2.

(A) By using the sequence information of mmu_miRNA-199b-5p, we located the position of its human homologous sequence in the UCSC database. (B) Based on the positional information and the source gene, we further aligned this position with the closest miRNA in miRbase. (C) We compared the sequences of hsa_miR-199b-5p and mmu_miR-199b-5p. (D) Conservation analysis was performed to compare the sequence conservation of miR-199b-5p.

Reviewer #2 (Public Review):

Summary:

The authors identified miR-199b-5p as a potential OA target gene using serum exosomal small RNA-seq from human healthy and OA patients. Their RNA-seq results were further compared with publicly available datasets to validate their finding of miR-199b-5p. In vitro chondrocyte culture with miR-199b-5p mimic/inhibitor and in vivo animal models were used to evaluate the function of miR-199b-5p in OA. The possible genes that were potentially regulated by miR-199b-5p were also predicted (i.e., Fzd6 and Gcnt2) and then validated by using Luciferase assays.

We greatly appreciate Reviewer #2 constructive comments.

Strengths:

(1) Strong in vivo animal models including pain tests.

(2) Validates the binding of miR-199b-5p with Fzd6 and binding of miR-199b-5p with Gcnt2.

Thanks for positive comment.

Weaknesses:

(1) The authors may overinterpret their results. The current work shows the possible bindings between miR-199b-5p and Fzd6 as well as bindings between miR-199b-5p and Gcnt2. However, whether miR-199b-5p truly functions through Fzd6 and/or Gcnt2 requires genetic knockdown of Fzd6 and Gcnt2 in the presence of miR-199b-5p.

In this study, we employed a comprehensive approach by integrating data from six bioinformatics databases to identify potential target genes for miR-199b-5p. Subsequent qPCR analysis revealed significant changes in two genes, Fzd6 and Gcnt2. We then utilized luciferase assays to validate the predicted binding sites and confirmed the interaction between miR-199b-5p and these genes. Additionally, we examined the expression profiles of these potential target genes in human KOA samples using a human database, which unveiled distinct expression patterns.

While our findings suggest that Fzd6 and Gcnt2 may serve as potential target genes for miR-199b-5p, we acknowledge the necessity for further experimental validation and in-depth functional characterization. Building upon your insightful recommendations, we have thoroughly addressed the research limitations and proposed potential research strategies for future investigations in our discussion. (page11，line227-231)

(2) In vitro chondrocyte experiments were conducted in a 2D manner, which led to chondrocyte de-differentiation and thus may not represent the chondrocyte response to the treatments.

We admit that 3D culture system will be more accurate and reliable. However, according to Liu Qianqian et al researches3, the 2D culture systems were also used and work well. Besides, the second-generation primary mice chondrocytes we used in the current study did not exhibit a significant dedifferentiated morphology. So, considering the experiment condition in our lab, we chose the second-generation cultured primary mouse chondrocytes in the whole process of cell experiment. To show the reliability of the cells, we provided more pictures in the supplement fig 7(fig. S7) In the future study, we will adopt 3D culture system for experiments. Thank you for your advices and we have added this limitation in the revised manuscript. (page11，line237-240)

Author response image 3.

Primary mice chondrocytes we cultured （P1）and the secondary generation cells(P2) we used in the following experiment.

References which used 2D ：

(3) Liu Q, Zhai L, Han M, et al. SH2 Domain-Containing Phosphatase 2 Inhibition Attenuates Osteoarthritis by Maintaining Homeostasis of Cartilage Metabolism via the Docking Protein 1/Uridine Phosphorylase 1/Uridine Cascade. Arthritis & Rheumatology (Hoboken, NJ). 2022;74(3):462-474.

(3) There is a lack of description for bioinformatic analysis.

Sorry for our neglection. We have added relevant descriptions and details. (Pages 14, line299-303)

(4) There are several errors in figure labeling.

We have revised. (Fig. 3, Fig. 4, Fig. 5 and Fig. 7)

Recommendations for the authors:

We appreciate the reviewers' feedback as we believe it has significantly contributed to the refinement of our manuscript. We are confident that our revisions have strengthened the quality and impact of our study, and we agree that the suggestions presented by the reviewers are valuable and appropriate for publication.

Reviewer #2 (Recommendations For The Authors):

I would like to thank the authors for investigating the functional role of miR-199b-5p in knee OA. While this study has the potential to provide valuable knowledge to the fields of miRNAs and joint diseases, significant improvements in several areas are required.

We appreciate your constructive comments, and we have made a substantial improvement to the manuscript. We thank all the reviewers for their advice as well as their criticisms.

Major concerns:

(1) According to the Authors, miR-199b-5p is identified by the results from their own miRNA-sequencing as well as comparison with other publicly available datasets (both synovium and cartilage datasets). It is unclear to me why the synovium dataset was used here as it appears that the entire manuscript was mainly focused on chondrocytes.

Thank you for your question. As we are aware, cartilage degradation is the initial pathological change in knee osteoarthritis (KOA), which subsequently leads to other pathological changes such as synovial inflammation4. These factors are interrelated, and current research on KOA encompasses cartilage, synovium, and system inflammation et al. Therefore, when we identified a large number of dysregulated miRNAs in extracellular vesicles isolated from serum, it was crucial to determine whether these dysregulated miRNAs were also altered in cartilage or synovium. To address this, we compared our findings with publicly available databases and found a higher overlap with the cartilage cell dataset, including miRNA-199b. Consequently, we decided to focus our subsequent investigations on cartilage-related research.

Reference

(4) Hunter D, Bierma-Zeinstra S. Osteoarthritis. Lancet (London, England). 2019;393(10182):1745-1759.

(2) Also, 169 of 177 differentially expressed exosome miRNAs were intersected with differentially expressed miRNAs from OA cartilage datasets. It is surprising that in the 5 selected miRNAs for further qRT-PCR validation, 3 out of 5 were not in the exosome miRNA dataset (i.e., hsa-mir-1296-5p, hsa-mir-15b-3p, and hsa-mir-338-3p; page 5, line 109 and Fig. 1B). Isn't that selecting the miRNAs that both differently expressed in exosome and cartilage datasets for validation more essential? Furthermore, from the Authors' exosome miRNA dataset, only 5 out of 15 KOA patients actually exhibited up-regulated miR-199b-5p vs. health controls. Please elaborate on how the target was determined.

In fact, our study included two additional groups: the acupuncture treatment group (4 weeks of continuous acupuncture treatment) and the waiting treatment group (no intervention, followed by acupuncture treatment after 4 weeks), in addition to the healthy control and knee osteoarthritis (OA) patient groups. After comparing these four groups, we found that 11 genes (hsa-miR-504-3p, hsa-miR-1915-3p, hsa-miR-103a-2-5p, hsa-miR-887-3p, hsa-miR-1228-5p, hsa-miR-34c-3p, hsa-miR-3168, hsa-miR-518e-3p, hsa-miR-1296-5p, hsa-miR-338-3p, and hsa-miR-199b-5p) were upregulated in KOA patients but downregulated after acupuncture treatment, with no change in the waiting treatment group. Additionally, 7 genes (hsa-miR-448, hsa-miR-514a-3p, hsa-miR-4440, hsa-let-7f-5p, hsa-let-7a-5p, hsa-let-7d-5p, and hsa-miR-15b-3p) were downregulated in KOA patients but upregulated after acupuncture treatment, with no change in the waiting treatment group. Considering the improvement in clinical symptoms of KOA patients after acupuncture treatment, we believe that these 18 genes are of significant value. Based on overall expression abundance and species specificity, we finally selected 5 genes, namely the 5 genes mentioned in this article. Regarding this result, we have already included it in the supplementary fig5(fig. S5).

Author response image 4.

Venn diagram showing differentially expressed miRNAs in the OA group compared with healthy patients and patients who recovered after acupuncture treatment.

(3) There is also a lack of description for bioinformatic analysis regarding how miRNA sequencing datasets were analyzed. What R/python packages or algorithms were used? What were the QC criteria?

We apologize for any confusion caused. We have now included a clear description of the method employed, and R was utilized for this data analysis (revised in Page14, Line301-305). To ensure consistency, we compared our findings with publicly available human serum data from the database (GSE105027) using a fold change threshold of > 1.5 and a significance level of p < 0.05. In the cartilage data (GSE175961), we observed a list of miRNAs with shared expression patterns, yet the precise differential values could not be determined.

(4) Another major concern is the chondrocyte culture method. Chondrocytes should be cultured in a 3D manner (i.e., a 3D pellet culture system or a micro mass culture method). 2D cultured chondrocytes tend to de-differentiate into MSC-like cells and thus lose their chondrocyte phenotype. This is evident from Fig. 3B and C. Cells started to spread out and only a few cells were positive for COL2A1 with a deep brown staining color. Thus, the results from the in vitro studies may not be representative of chondrocyte response to the treatments.

We admit that 3D culture system will be more accurate and reliable. However, according to Liu Qianqian et al researches3, the 2D culture systems were also used and work well. Besides, the second-generation primary mice chondrocytes we used in the current study did not exhibit a significant dedifferentiated morphology. So, considering the experiment condition in our lab, we chose the second-generation cultured primary mouse chondrocytes in the whole process of cell experiment. To show the reliability of the cells, we provided more pictures in the supplement fig 7(fig. S7) In the future study, we will adopt 3D culture system for experiments. Thank you for your advices and we have added this limitation in the revised manuscript. (page11, line237-240)

Author response image 5.

Primary mice chondrocytes we cultured （P1）and the secondary generation cells(P2) we used in the following experiment.

References which used 2D ：

(3) Liu Q, Zhai L, Han M, et al. SH2 Domain-Containing Phosphatase 2 Inhibition Attenuates Osteoarthritis by Maintaining Homeostasis of Cartilage Metabolism via the Docking Protein 1/Uridine Phosphorylase 1/Uridine Cascade. Arthritis & Rheumatology (Hoboken, NJ). 2022;74(3):462-474.

(5) Page 7, lines 148-149: "The cartilage of mice injected with the miR-199b-5p mimic was slightly degraded (p=0.02) (Fig. 4E, F)". However, there was no significance between the groups found in Fig. 4F. Also, from the histological images of Fig. 4E, it looks like mice with inhibitor injection had more cartilage damage than miR-199b-5p mimic.

We apologize for any confusion caused. Figures 4E and 4F represent the Safranin Fast Green Staining staining of the joint after the administration of miR-199b-5p inhibitor and mimic under physiological conditions. As you can see, there is minimal difference between these four images. There is no statistically significant difference. However, in Figures 5E and 5F, the MIA-induced KOA model was utilized, and noticeable differences can be observed after the administration of the inhibitor and mimic. In the revised version, we have emphasized that Figures 4E and 4F represent the results under physiological conditions, not under the MIA-induced model. (page 7, line 146-151)

(6) Page 7, lines 149-150: "Additionally, the articular surface showed insect erosion (Fig. 4G)." It is also unclear how micro-CT analysis will be able to demonstrate the erosion of cartilage. Or the authors actually indicate the trochlear groove. However, this could also be observed in the control group and the results were not quantified. It is also unclear if the cross-section images of micro-CT shown here are helpful at all without any further explanation in the manuscript.

Figure 4 G represents control, vehicle control, inhibitor, and mimic groups, while Figure 5 G represents model, model+vehicle control, model+inhibitor, and model+mimic groups. From Figure 4G, it can be observed that the simulator group showed the most obvious erosion appearance, while the inhibitor group did not exhibit this phenomenon5. From Figure 5G, it can be seen that the model group and model+mimic group exhibited the most pronounced erosion appearance, while the model+inhibitor group showed the best recovery. To highlight the pathological changes in the erosion appearance, we marked the typical locations with red arrows in the images for easy comparison and reading by the readers (Fig. 4G; Fig. 5G). We also made corresponding textual modifications in the original manuscript to address these findings (page 7, line 150-151; page 8, line 160-161). In addition, the 3D reconstruction of micro-CT is based on the synthesis of these cross-sectional images.

References

(5) Tao Y, Wang Z, Wang L, et al. Downregulation of miR-106b attenuates inflammatory responses and joint damage in collagen-induced arthritis. Rheumatology (Oxford, England). 2017;56(10):1804-1813.

(7) Page 17, line 309-310: "Before model establishment and at 3, 7, 10, 14, 21, and 28 days after model establishment." Please re-write this as this is not clear regarding the experimental procedure.

Thank you. We had to re-write the sentences as following：Baseline testing of behavioral pain thresholds was conducted prior to model establishment, followed by behavioral pain threshold testing on days 3, 7, 10, 14, 21, and 28 after model establishment. (pages15, line322-324)

(8) Fig. 5A. The M + inhibitor and Model images are not at the same plane as M + mimic and M + RNAnc images.

Thank you. We have modified.

(9) Fig. 5B. There are two lines both with circle markers (Control and M+inhibitor). Please correct.

We have corrected.

(10) Fig. 5F. Missing * sign.

We added *sign.

(11) Please elaborate how the potential binding sites between miR-199b-5p and Gcnt2 and between miR-199b-5p and Fzd6.

We apologize for any lack of clarity in the original text. In fact, we utilized targets to predict potential binding sites. Specifically, for the mouse species, we predicted that the 3'UTR of Fzd6 binds with miR-199b-5p at positions 2483-2490, 3244-3251, 3303-3309, and 3854-3860, while the 3'UTR of Gcnt2 binds with miR-199b-5p at positions 2755-2762 and 4144-4151. In the revised version, we provide a detailed description of the methodology used for predicting these sites and offer an elaborate explanation of the results. (pages16, line352)

Additionally, to demonstrate consistency with human binding sites, we not only predicted the binding sites of human miR with these two target genes but also found a high conservation of up to 95.65% between the human and mouse sequences of miR-199b-5p. We have included this information in the supplementary materials (Fig. S6). In Fig. 6E-F, we presented the potential binding sites between miR-199b-5p and Gcnt2, as well as between miR-199b-5p and Fzd6. In addition, we provide the predicted binding of human sequence to illustrate the binding sites. Furthermore, the predicted binding of human miR-199b-5p with fzd6 and gcnt2 showed a high degree of consistency. (The fluorescent labeling in the following text indicates the potential predicted binding sites.) (Supplement file 8)

hsa-miR-199b-5p MIMAT0000263

CCCAGUGUUUAGACUAUCUGUUC

NCBI Gene ID 8323 GenBank Accession NM_001164615

Gene Symbol FZD6 3' UTR Length 1368

Gene Description frizzled class receptor 6

3' UTR Sequence: agaacattttctctcgttactcagaagcaaatttgtgttacactggaagtgacctatgcactgttttgtaagaatcactgttacattcttcttttgcacttaaagttgcattgcctactgttatactggaaaaaatagagttcaagaataatatgactcatttcacacaaaggttaatgacaacaatatacctgaaaacagaaatgtgcaggttaataatatttttttaatagtgtgggaggacagagttagaggaatcttccttttctatttatgaagattctactcttggtaagagtattttaagatgtactatgctattttacttttttgatataaaatcaagatatttctttgctgaagtatttaaatcttatccttgtatctttttatacatatttgaaaataagcttatatgtatttgaacttttttgaaatcctattcaagtatttttatcatgctattgtgatattttagcactttggtagcttttacactgaatttctaagaaaattgtaaaatagtcttcttttatactgtaaaaaaagatataccaaaaagtcttataataggaatttaactttaaaaacccacttattgataccttaccatctaaaatgtgtgatttttatagtctcgttttaggaatttcacagatctaaattatgtaactgaaataaggtgcttactcaaagagtgtccactattgattgtattatgctgctcactgatccttctgcatatttaaaataaaatgtcctaaagggttagtagacaaaatgttagtcttttgtatattaggccaagtgcaattgacttcccttttttaatgtttcatgaccacccattgattgtattataaccacttacagttgcttatattttttgttttaacttttgttttttaacatttagaatattacattttgtattatacagtacctttctcagacattttgtagaattcatttcggcagctcactaggattttgctgaacattaaaaagtgtgatagcgatattagtgccaatcaaatggaaaaaaggtagttttaataaacaagacacaacgtttttatacaacatactttaaaatattaaggagttttcttaattttgtttcctattaagtattattctttgggcaagattttctgatgcttttgattttctctcaatttagcatttgcttttggtttttttctctatttagcattctgttaaggcacaaaaactatgtactgtatgggaaatgttgtaaatattaccttttccacattttaaacagacaactttgaatacaaaaactttgttttgtgtgatcttttcattaataaaattatctttgtataagaaaaaaaaaaaaaa

hsa-miR-199b-5p MIMAT0000263

CCCAGUGUUUAGACUAUCUGUUC

NCBI Gene ID 2651 GenBank Accession NM_001491

Gene Symbol GCNT2 3' UTR Length 2780

Gene Description glucosaminyl (N-acetyl) transferase 2 (I blood group)

3' UTR Sequence: gctattcatgagctactcatgactgaagggaaactgcagctgggaagaggagcctgtttttgtgagagacttttgccttcgtaatgttaaccgtttcaggaccacgtttatagcttcaggacctggctacgtaattatacttaaaatatccactggacactgtgaaatacactaacaggatggctgggtagagcaatctgggcactttggccaattttagtcttgctgtttcttgatgctcacctctatattagtttattgttaggatcaatgataaatttaaatgacctcagatctttgcaccagatactcatcatatacaaatgttttagtaaaaaagagaattgtagataatactgtctaggaaaataagaattaggtttctttgaagaaggaatcttttataacaccttaacagtcaccactgtgctcaaccagacagatagtgaaacagctttctgggtaattcaccaatttcctttaaaacataagctacctgaatggagaatacatcttgtttctgagtttcaacactagcatttttggcttactcatggacaaagttctgtatatagtataaagtcattaacaagaaacaggatatgctttaagacagaattcactgtctgttgcttcagtaaaaggacctcggggaataaaacatttctctcttatatgccagaatgtaggctggtccctatgtcatgtcttccattaagaacactaaaaagtccttgcaagaatggagatatgcattcaagagaggtgctatcacatagatctagtctgaagtctggaacactttcctcttctatgacccctctctccccagtattatcttacttgcaaaatggagaccaaattctatcctgtgaggcttttaattgcaccatagtatgctctgagtagctttacactgcctggtactgatagtagtggctcgatttttaagagccttcaattgtagatgaacatctctgttatttatccctcattcatccatccgttcattcattcagccttcaatcaacatctcttgagtgtctattatgtacaggacatgtactgagacaaaaaggaaacataagagctttttcactctaaaaatcttggcaataatgtcaacaccagaaagcctcctctggagaatcttacagagtgattgtagtttaatacaggaacacacagggctgtgtagcatgataccaggcccaggagatcagtaattacaaattaagggttaaatcagagattattcaacagagagggagaaaggaggagacagagggaggacctgttgtgttccagccattctggtattcctttatgtatctaatttcattcaaacctcacaacagtcttgtgaggcccttatataattactcccattttgcagatgaagtaactgaggcttagaaaggttaatagcaccggggaacaatttctctgggtgagaattgggactctgttgctggtcttctcagttcatttcctgaggtggatttactgagagaaggtgaaataaagccatatttagtataccagagaaggtagattttaagaatggtctcagtgttaatactgagaaaaagtcctgtcagttcagaaaaaatgtgaagtctactttagtattcctgtaatactaaaccgttgagtttctaaatatttatttattctaacaaaaagcaattactacaaatggatgacacatttaatgaacacaattttattttttttctgtaactgtgcttgttgaatgtcaatcatatttaaagggaatgactttgaagtaaaaccttttttcttgctactgaaaaaaatggagttgttttgggtggtaaagtgttaaggaatagggacagctggtcacacaaggaactcttgaaggccacatgtgaaaacctgtcacttgcacagaggccagtcccactaaggtgaccagagtgggctccaagcacaaactgccattggctatagatgggactgtgtccccccaaaattcatgtgttggagccttaaccctcaatgtgatggtatttgagatggggcctttggtaagggaagtttagatgaggtcacgagggtaggaccctcatgatgggatgagtccccttacaagacctctggcttgggccgggcgtggtggctcacacctgtaatcccaacactttgggaggccaaggcaggtagatcacttgatgccaggagttccagaccaggctggccgacatggtgaaaccccatctctactaaaaaatataaaaattagccgggctttgtggcatgtgcctgtaatcccagctatttggcaggctgaggcatgagaatcgcttgaacccaggaggtggaggttacagtgagctgagagtgccccactgcactccagcctgggtgacagagcgagactttgtcccaaaacaaaataggtgaggggatagcgaatgcactcagggtcagcagtggagtttaaaaattgtctcttttcaacttatttaaatgacagcacctgagaagaggaaccgttttacactggatgtttctcatgtagaacaagaaatctttctggaattgatgtttacatgtctgttgttggtcatctctcctgtgtcttaaatactttaatgttggaagagcatagtgtttgggctagtgggtttctgacagcccatgggaatgccctgaaactactgtatctgatgtttgttttcgatgaggttccatgttttgttttcttgggaataaattaatatattgttttccaaaaaaaaaaaaaaaaaaaa

(12) Page 10-11, Line 222-223: "Our findings indicate that miR-199b-5p plays a crucial role in KOA by targeting Fzd6 and Gcnt2". This is an overstatement. The current work shows the possible bindings of miR-199b-5p and Fzd6 as well as bindings of miR-199b-5p and Gcnnt2. Whether miR-199b-5p truly functions through Fzd6 and/or Gcnt2 requires genetic knockdown of Fzd6 and Gcnt2 in the presence of miR-199b-5p. Thus, please tune down this statement and the title of the manuscript.

We agree your opinion of our conclusion. Therefore, we delete the overstatement sentences and tune down the conclusion of the manuscript. (the title; page 8,179; page11, line227-228)

(13) The Schematic figure (the last figure). Please remove osteophyte as this was not quantified in the study.

We modified the schematic figure accordingly.

Minor concerns:

(1) Most figures were distorted.

We provide a new version of the figure to avoid distortions.

(2) Providing GO term numbers in Fig. 1C is not very helpful. Maybe show the GO term and corresponding numbers in the manuscript (Page 4, lines 79 - 82).

Thank you for your advice. We added the corresponding notes of the GO term numbers in the manuscript to explain each biological concept of it. (Page 4, line 77-89；Page 22，line 515-532)

(3) What were M-0.5 and M-1 in Fig. 2D? Different MIA concentrations?

Yes, these are different MIA concentrations, which we illustrate in the legend. (Page 23, line 535-536)

(4) Please follow the nomenclature of the gene symbol. For example, Fig. 3E-P should be mouse genes (?).

We modified the relevant gene symbol.

(5) Page 3, line 59. Not all chondrocytes are pathogenic cells in OA.

We are sorry for the mistake, now it has been modified. (Page 3, line 59)

(6) Typo. Page 3, line 55.

We changed the Typo.

(7) Page 4, line 78. These are differentially expressed miRNAs, not genes.

We have revised the unsuitable expression. (Page4, line75-76)

I wish the authors all the best with their continued work in this area.

Thank you for your wishes.

Author Response

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

Reviewer #1 (Public Review):

Summary:

The manuscript by Xia et al. investigated the mechanisms underlying Glucocorticoid-induced osteonecrosis of the femoral head (GONFH). The authors observed that abnormal osteogenesis and adipogenesis are associated with decreased β-catenin in the necrotic femoral head of GONFH patients, and that the inhibition of β-catenin signalling leads to abnormal osteogenesis and adipogenesis in GONFH rats. Of interest, the deletion of β-catenin in Col2-expressing cells rather than in Osx-expressing cells leads to a GONFH-like phenotype in the femoral head of mice.

Strengths:

A strength of the study is that it sets up a Col2-expressing cell-specific β-catenin knockout mouse model that mimics the full spectrum of osteonecrosis phenotype of GONFH. This is interesting and provides new insights into the understanding of GONFH. Overall, the data are solid and support their conclusions.

Reviewer #1 (Recommendations For The Authors):

1) Fig. 1I should be quantified and presented as bar graphs to make it consistent with other data, and the significance should be shown.

Reply: Thanks for your comments. We have provided the quantitative bar graph in the new version.

2) Fig. 2H, beta-catenin, ALP and FABP4 should be labled below the X axis. Moreover, the pattern of Fig. 2H is different from other bar graphs and the dots for individual samples are missing, so I could not judge the N values for the experiments. N values should also be provided for Fig. 3.

Reply: Thanks for your comments. We have added the labels of beta-catenin, ALP and FABP4 below the X axis in Fig. 2H. The modes of quantitative bar graphs were changed to show the N values in the each experiment.

3) Fig. 4 shows the fate mapping of Col2+ cells and Osx+ cells in the femoral head. In this regard, the authors presented images for Col2-expressing cells at all the indicated time points, i.e. 1, 3, 6, and 9 months, but only presented images for Osx-expressing cells for 1 month while those for 3, 6, and 9 months are missing.

Reply: Thanks for your comments. Here, we showed that the expression of Osx+ cells in the femoral head were total different with Col2+ cells at the age of 3, 6 month, further indicating they were two different progenitor lineage cells.

Author response image 1.

4) Some experiments may need to be described in more detail" e.g., ABH/Orange G staining, biomechanical testing, μCT analysis, et al.

Reply: Thanks for your comments. We have provided more information of experiment procedures.

5) This study proposed that Col2-expressing cells play a key role in the progression of GONFH, did the authors use Col2+ cells for the in vitro experiments?

Reply: As in vitro experiments could not reflect the location of Col2-expressing cells in the femoral head, therefore here we applied in vivo lineage tracing study. After as long as 9 month of linage trace, we thoroughly showed the self-renew ability and osteogenic commitment of Col2+ cells, as well as its space variation in the femoral head with age. Conditional knockout of β-catenin caused that Col2+ cells trans-differentiated into adipogenic cells instead of osteogenic cells, which directly clarified the mechanism of Col2+ cells leading to GONFH-like phenotype in mice.

6) A few typo errors, such as Line 13, "contribute" should be "contributes"; Line 118, "reveled" should be "revealed".

Reply: We have revised the grammar errors in the new manuscript.

Reviewer #2 (Public Review):

Summary:

In this manuscript, the authors reported a study to uncover that β-catenin inhibition disrupting the homeostasis of osteogenic/adipogenic differentiation contributes to the development of Glucocorticoid-induced osteonecrosis of the femoral head (GONFH). In this study, they first observed abnormal osteogenesis and adipogenesis associated with decreased β-catenin in the necrotic femoral head of GONFH patients, but the exact pathological mechanisms of GONFH remain unknown. They then performed in vivo and in vitro studies to further reveal that glucocorticoid exposure disrupted osteogenic/adipogenic differentiation of bone marrow stromal cells (BMSCs) by inhibiting β-catenin signaling in glucocorticoid-induced GONFH rats, and specific deletion of β-catenin in Col2+ cells shifted BMSCs commitment from osteoblasts to adipocytes, leading to a full spectrum of disease phenotype of GONFH in adult mice.

Strengths:

This innovative study provides strong evidence supporting that β-catenin inhibition disrupts the homeostasis of osteogenic/adipogenic differentiation that contributes to the development of GONFH. This study also identifies an ideal genetically modified mouse model of GONFH. Overall, the experiment is logically designed, the figures are clear, and the data generated from humans and animals is abundant supporting their conclusions.

Weaknesses:

There is a lack of discussion to explain how the Wnt agonist 1 works. There are several types of Wnt ligands. It is not clear if this agonist only targets Wnt1 or other Wnts as well. Also, why Wnt agonist 1 couldn't rescue the GONFH-like phenotype in β-cateninCol2ER mice needs to be discussed.

Reply: Thanks for your constructive comments. Wnt agonist 1 is a cell-permeating activator of the Wnt signaling pathway that induces transcriptional activity dependent on β-catenin (PMID: 25514428,18624906). In the present study, we aim to demonstrate that activation of β-catenin signaling could alleviate the phenotype of rat GONFH, thus only β-catenin and downstream targets (RUNX2, ALP, PPAR-γ, FABP4) expressions were detected after Wnt agonist 1 intervention. Conditional knockout β-catenin in Col2+ cells lead to an mouse GONFH-like phenotype. Wnt agonist 1 couldn't rescue this GONFH-like, as it did not activate β-catenin signaling. We have discussed them in the new version.

Reviewer #3 (Public Review):

Summary:

In this manuscript, the authors are trying to delineate the mechanism underlying the osteonecrosis of the femoral head.

Strengths:

The authors provided compelling in vivo and in vitro data to demonstrate Col2+ cells and Osx+ cells were differentially expressed in the femoral head. Moreover, inducible knockout of β-catenin in Col2+ cells but not Osx+ cells lead to a GONFH-like phenotype including fat accumulation, subchondral bone destruction, and femoral head collapse, indicating that imbalance of osteogenic/adipogenic differentiation of Col2+ cells plays an important role in GONFH pathogenesis. Therefore, this manuscript provided mechanistic insights into osteonecrosis as well as potential therapeutic targets for disease treatment.

Weaknesses:

However, additional in-depth discussion regarding the phenotype observed in mice is highly encouraged.

Reply: Thanks for your comments. Inducible knockout of β-catenin in Col2+ cells but not Osx+ cells lead to a GONFH-like phenotype. Lineage tracing data showed Col2+ cells and Osx+ cells were different cell populations, and we have discussed the potential mechanism caused the different phenotypes between β-cateninCol2ER mice and β-cateninOsxER mice.

1) Why did the authors use dexamethasone in the cellular experiments but methylprednisolone to induce the GONFH rat model?

Reply: Thanks for the comments. Here, we applied a dexamethasone (DEX)-treated BMSC model in vitro and a methylprednisolone (MPS)-induced rat model in vivo for GONFH study based on the published literatures (PMID: 37317020, 29662787, 29512684,35126710, 32835568).

2) Both bone damage and fat accumulation were observed in 3-month-old and 6-month-old β-cateninCol2ER mice, but the femoral head collapse (the feature of GONFH at the late stage) only occurred in the older β-catenin Col2ER mice. This interesting observation needs to be discussed. Reply: Thanks for the comments. Bone damage caused a poor mechanical support is the key to femoral head collapse. Despite of similar trabecular bone loss and fat accumulation in the 3-month-old and 6-month-old β-cateninCol2ER mice, the older mice also presented extensive subchondral bone destruction. Integrated subchondral bone provided a well mechanical support for femoral head morphology, therefore femoral head collapse were occurred in the older β-cateninCol2ER mice.

3) In the Materials and Methods, detailed information on the reagents should be provided.

Reply: We have provided detailed information of the important reagents.

4) As shown in Figure 4, β-cateninOsxER mice at 3 months of age did not show differences in lipid droplet area and empty lacunae rate, but there was a decrease in bone area. The authors should at least provide some necessary discussion of this phenomenon.

Reply: Thanks for your comments. In the present study, we found few lipid droplet and empty lacuna but a significant decrease of bone mass in the femoral heads of β-cateninOsxER mice. Previous studies showed that specific knockout of β-catenin in Osx-expressing cells promoted osteoclast formation and activity, leading to the bone mass loss (PMID: 29124436, 34973494). We discussed this phenomenon in the new version.

Author Response

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

eLife assessment

This study presents a potentially valuable discovery which indicates that activation of the P2RX7 pathway can reduce the lung fibrosis after its establishment by inflammatory damage. If confirmed, the study could clarify the role of specific immune networks in the establishment and progression of lung fibrosis. However, the presented data and analyses are incomplete as they primarily rely on limited pharmacological treatments with modest effect sizes. I hope you will be convinced by the validity of our approaches with the following explanation/information and I remain at your disposal to discuss

Public Reviews:

Reviewer #1 (Public Review):

In this revised preprint the authors investigate whether a presumably allosteric P2RX7 activating compound that they previously discovered reduces fibrosis in a bleomycin mouse model. They chose this particular model as publicly available mRNA data indicate that the P2RX7 pathway is downregulated in idiopathic pulmonary fibrosis patients compared to control individuals. In their revised manuscript, the authors use three proxies of lung damage, Ashcroft score, collagen fibers, and CD140a+ cells, to assess lung damage following the administration of bleomycin. These metrics are significantly reduced on HEI3090 treatment. Additional data implicate specific immune cell infiltrates and cytokines, namely inflammatory macrophages and damped release of IL-17A, as potential mechanistic links between their compound and reduced fibrosis. Finally, the researchers transplant splenocytes from WT, NLRP3-KO, and IL-18-KO mice into animals lacking the P2RX7 receptor to specifically ascertain how the transplanted splenocytes, which are WT for P2RX7 receptor, respond to HEI3090 (a P2RX7 agonist). Based on these results, the authors conclude that HEI3090 enhanced IL-18 production through the P2RX7-NLRP3 inflammasome axis to dampen fibrosis.

These findings could be interesting to the field, as there are conflicting results as to whether NLRP3 activation contributes to fibrosis and if so, at what stage(s) (e.g., acute damage phase versus progression). The revised manuscript is more convincing in that three orthogonal metrics for lung damage were quantified. However, major weaknesses of the study still include inconsistent and small effect sizes of HEI3090 treatment versus either batch effects from transplanted splenocytes or the effects of different genetic backgrounds. Moreover, the fundamental assumption that HEI3090 acts specifically and functionally through the P2RX7 pathway in this model cannot be directly tested, as the authors now provide results indicating that P2RX7 knockout mice do not establish lung fibrosis on bleomycin treatment.

I’m particularly concerned by the assumption made by reviewer 1 concerning the fact that P2RX7 knockout mice do not establish lung fibrosis on bleomycin treatment.

Indeed, what we showed in the point-to-point response is that BLM induces fibrosis in both WT and P2RX7 KO mice, but the intensity of the fibrosis is reduced in P2RX7KO mice, panel A. Therefore, as discussed in our first response, our results confirmed the previous publication of Riteau et al, that P2RX7 participates in BLM-induced lung fibrosis (see panel B).

Author response image 1.

Bleomycin induced lung fibrosis in WT versus p2rx7 KO mice. A: lung from BLM-treated mice were stained with HE and fibrosis was quantified using the Ashcroft protocol. Result showed that fibrosis induced by BLM in KO mice is reduced as compared to WT mice. B: Representative images of lung sections at day 14 after BLM treatment stained with H&E as published in Riteau et al. and illustrating that fibrosis induced by BLM in KO mice is reduced as compared to WT mice. WT mice vehicle (n=4) or p2rx7 KO (n=6) mice. Two-tailed Mann-Whitney test, p values: **p < 0.01.

Importantly, this lower intensity of lung fibrosis in P2RX7 KO mice, does not interfere with the capacity of our molecule to attenuate lung fibrosis, as demonstrated in the adoptive transfer of IL1B KO splenocytes in P2RX7 KO mice, in which HEI3090 decreases the Ashcroft score, the % of fibrosis and the collagen fibers (see below).

Author response image 2.

HEI3090 activity requires P2RX7’s expressing immune cells: Experimental design. p2rx7-/- mice were given 3.106 il1β-/- splenocytes i.v. one day prior to BLM delivery (i.n. 2.5 U/kg). Mice were treated daily i.p. with 1.5 mg/kg HEI3090 or vehicle for 14 days. (C) Representative images of lung sections at day 14 after treatment stained with H&E and Sirius Red with il1β-/- splenocytes, bar= 100 µm (left) and fibrosis score assessed by the Ashcroft method, the % of fibrosis and the content of collagen fibers (right). Each point represents one mouse (n=2 in WT and NLRP3 experiment, n =1 in IL18 and IL1B experiment), data represented as violin plot or mean±SEM, two-tailed Mann-Whitney test, *p < 0.05. WT: Wildtype, KO: P2RX7 knock-out

Importantly, in the same experimental setting, e.g adoptive transfer of splenocytes from different genetic backgrounds, HEI3090 decreases the fibrosis intensity only with WT and IL1B KO splenocytes and not with NLRP3 KO and IL18KO splenocytes.

Author response image 3.

HEI3090 activity requires P2RX7’s expressing immune cells: Experimental design. p2rx7-/- mice were given 3.106 WT, NLRP3-/-, IL18-/- or IL1β-/- splenocytes i.v. one day prior to BLM delivery (i.n. 2.5 U/kg). Mice were treated daily i.p. with 1.5 mg/kg HEI3090 or vehicle for 14 days. Fibrosis in whole lung was assessed by the % of fibrosis (upper panel) and the content of collagen fibers (lower panel). Each point represents one mouse (n=2 in WT and NLRP3 experiments, n =1 in IL18 and IL1B experiment). Data represented as violin plot or mean±SEM, two-tailed Mann-Whitney test, *p < 0.05. WT: Wildtype, KO: P2RX7 knock-out

In order to provide clear evidence that HEI3090 functions through P2RX7, a different lung fibrosis model that does not require P2RX7 would be necessary. For example, in such a system the authors could demonstrate a lack of HEI3090-mediated therapeutic effect on P2RX7 knockout.

Since BLM induces lung fibrosis in P2RX7 KO mice as we showed in this manuscript and as already published by Riteau in 2010, shown earlier in our response (first figure) and because HEI3090 is able to decrease the intensity of fibrosis in WT and IL1B-/- → P2RX7 KO mice but not in KO, NLRP3-/- → P2RX7 KO and IL18-/- → P2RX7 KO mice we believe that our data sustain the conclusion that

1. HEI3090 required the expression of P2RX7 in immune cells to mediate the antifibrotic activity,

2. IL1B is not a crucial effector mediating the antifibrotic effect of HEI3090.

Molecularly, additional evidence on specificity, such as thermal proteome profiling and direct biophysical binding experiments, would also enhance the authors' argument that the compound indeed binds P2RX7 directly and specifically. Since all small molecules have some degree of promiscuity, the absence of an additional P2RX7 modulator, or direct recombinant IL-18 administration (as suggested by another reviewer), is needed to orthogonally validate the functional importance of this pathway. Another way the authors could probe pathway specificity would involve co-administering α-IL-18 with HEI3090 in several key experiments (similar to Figure 4L).

At the moment we have no funds to do these experiments and given the high competition, we have decided to publish our story without these new data.

Reviewer #2 (Public Review):

In the study by Hreich et al, the potency of P2RX7-specific positive modulator HEI3090, developed by the authors, for the treatment of Idiopathic pulmonary fibrosis (IPF) was investigated. Recently, the authors have shown that HEI3090 can protect against lung cancer by stimulating dendritic cell P2RX7, resulting in IL-18 production that stimulates IFN-γ production by T and NK cells (DOI: 10.1038/s41467-021-20912-2). Interestingly, HEI3090 increases IL-18 levels only in the presence of high eATP. Since the treatment options for IPF are limited, new therapeutic strategies and targets are needed. The authors first show that P2RX7/IL-18/IFNG axis is downregulated in patients with IPF. Next, they used a bleomycin-induced lung fibrosis mouse model to show that the use of a positive modulator of P2RX7 leads to the activation of the P2RX7/IL-18 axis in immune cells that limits lung fibrosis onset or progression. Mechanistically, treatment with HEI3090 enhanced IL-18-dependent IFN-γ production by lung T cells leading to a decreased production of IL-17 and TGFβ, major drivers of IPF. The major novelty is the use of the small molecule HEI3090 to stimulate the immune system to limit lung fibrosis progression by targeting the P2RX7, which could be potentially combined with current therapies available. Overall, the study was well performed, and the manuscript is clear.

We thank the reviewer for this very positive comments.

However, there is need for more details on the description and interpretation of the adoptive transfer experiments, as well as the statistical analyses and number of replicate independent experiments.

I’m concerned by the reviewer’s comments, and I would like to bring additional information/explanation, which I hope will convince you on the validity of our approaches.

Author response image 4.

Adoptive transfer experiment. Adoptive transfer experiments are classically used to document which immune cells participate in immune cell responses (with more than 150 publications in pubmed with the key words adoptive transfer and onco immunology) and intravenous administration is a common route to trigger lungs (PMID: 23336716). To characterize the molecular effector (P2RX7, NLRP3, IL18 and IL1B) accounting for the antifibrotic effect of HEI3090 we purified splenocytes from donor mice and administrated them intra venously in P2RX7 KO mice. As shown in Author response image 4, HEI3090 has no antifibrotic activity when splenocyte isolated from mice invalidated for p2rx7 are iv into P2RX7 KO mice (KO in KO). By contrast, HEI3090 has antifibrotic activity when WT splenocytes expressing P2RX7 (isolated from WT mice) are transferred into P2RX7 KO mice (WT in KO).

This experiment brings strong evidence to demonstrate the efficacy of adoptive transfer approach to identify molecular effector required to mediate the antifibrotic effect of HEI3090.

Statistical analyses and number of replicate independent experiments

We thank the reviewer for his comment, and we apologize to not have been sufficiently clear in our previous response with this miss phrased statement “the experiment was stopped when significantly statistical results were observed” when we should have written “the experiment was stopped when each experimental group contained at least 5 mice”.

To define the size of experimental groups we did a pilot experiment, with 4 WT mice (e.g. 4 biological replicates) in each group (as shown aside), and a statistical forecasting based on the result of the pilot experiment (40% difference, standard error: 0.9, α risk: 0.05, power: 0.8). Since we focused on the effect of HEI3090 we based our statistical analysis on a one-way ANOVA analysis comparing in each experiment the vehicle and the treated group.

The pilot experiment and statistical forecasting indicated 4 mice per group to characterize the effect of HEI3090 on BLM-induced lung fibrosis. Each experiment was started with 6 to 8 mice per group. Being aware that 30% of mice can unexpectedly dye due to BLM treatment, we duplicated the experiment, when necessary, to include at least 5 mice in each group of each experiment meaning 5 biological replicates, knowing that 4 mice are sufficient to statistically analyze the results. In each experiment we have checked for the presence of outlier, using the ROULT method, and removed the outliers when necessary.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, Yang et al report a novel regulatory role of SIRT4 in the progression of kidney fibrosis. The authors showed that in the fibrotic kidney, SIRT4 exhibited an increased nuclear localization. Deletion of Sirt4 in renal tubule epithelium attenuated the extent of kidney fibrosis following injury, while overexpression of SIRT4 aggravates kidney fibrosis. Employing a battery of in vitro and in vivo experiments, the authors demonstrated that SIRT4 interacts with U2AF2 in the nucleus upon TGF-β1 stimulation or kidney injury and deacetylates U2AF2 at K413, resulting in elevated CCN2 expression through alternative splicing of Ccn2 gene to promote kidney fibrosis. The authors further showed that the translocation of SIRT4 is through the BAX/BAK pore complex and is dependent on the ERK1/2-mediated phosphorylation of SIRT4 at S36, and consequently the binding of SIRT4 to importin α1. This fundamental work substantially advances our understanding of the progression of kidney fibrosis and uncovers a novel SIRT4-U2AF2-CCN2 axis as a potential therapeutic target for kidney fibrosis.

Strengths:

Overall, this is an extensive, well-performed study. The results are convincing, and the conclusions are mostly well supported by the data. The message is interesting to a wider community working on kidney fibrosis, protein acetylation, and SIRT4 biology.

Weaknesses:

The manuscript could be further strengthened if the authors could address a few points listed below:

(1) In the results part 3.9, an in vitro deacetylation assay employing recombinant SIRT4 and U2AF2 should be included to support the conclusion that SIRT4 is a deacetylase of U2AF2. Similarly, an in vitro binding assay can be included to confirm whether SIRT4 and U2AF2 are directly interacted.

Thank you for your insightful comments and suggestions for improving our manuscript. We appreciate your recommendation to include an in vitro deacetylation assay employing recombinant SIRT4 and U2AF2 to support our conclusion regarding the deacetylase activity of SIRT4 on U2AF2.

We would like to clarify that the data demonstrating the effect of SIRT4 on U2AF2 acetylation were already included in our original submission. Specifically, Figure 5C illustrates that the TGF-β1-caused decreased acetylation of U2AF2 is attenuated by Sirt4 knockdown. Conversely, overexpression of SIRT4 (SIRT4 OE) enhances the deacetylation process of U2AF2 in the presence of TGF-β1. These results support that SIRT4 is a deacetylase for U2AF2.

Furthermore, we have already provided evidence of the direct interaction between SIRT4 and U2AF2 through a co-immunoprecipitation (CoIP) assay, which was shown in Figure 5B. This assay confirms the physical interaction between SIRT4 and U2AF2.

We believe that the existing data sufficiently address the points raised in your comments. We are grateful for the opportunity to clarify these aspects of our study and hope that our response has adequately addressed your concerns.

(2) In Figure 6D, the Western Blot data using U2AF2-K453Q is confusing and is quite disconnected from the rest of the data and not explained. This data can be removed or explained why U2AF2-K453Q is employed here.

Thank you for your inquiry regarding the rationale behind the K453Q mutation in our study.

In the study, we have predicted some acetylation sites. U2AF2-K453Q is another site mutation to mimic a hyperacetylated state of U2AF2, our results indicated that U2AF2 acetylation at K413 had little effects on CNN expression. Therefore, we found that only the U2AF2 acetylation at K413 can regulate CCN2 expression, not acetylation at other sites. In order not to cause ambiguity in the study, we have removed the results of U2AF2-K453Q in our revised manuscript.

(3) Although ERK inhibitor U0126 blocked the nuclear translocation of SIRT4 in vivo, have the authors checked whether treatment with U0126 could affect the expression of kidney fibrosis markers in UUO mice?

Thank you for your insightful question regarding the effects of the ERK inhibitor U0126 on the expression of kidney fibrosis markers in UUO mice.

In our study, we indeed conducted in vivo experiments using U0126 and observed that it effectively ameliorated kidney fibrosis markers, which is consistent with its established role in inhibiting the fibrotic process. Specifically, U0126 treatment significantly suppressed the SIRT4-mediated renal fibrosis, which was evidenced by the reduced expression of fibrosis markers (Author response image 1).

Author response image 1.

U0126 treatment alleviates renal fibrosis in UUO mice.

However, in the initial submission, we chose not to include these results in the main body of the manuscript based on the following reasons: 1) we intent to highlight the inhibitory effects of U0126 on ERK and its subsequent impact on kidney fibrosis might shift the focus of our study away from the central theme of SIRT4's role in renal fibrosis. 2) We aimed to maintain a clear narrative that emphasizes the novel findings related to SIRT4 and its regulation by the ERK pathway.

Nonetheless, we recognize the importance of these findings and are willing to include the relevant data in the revised manuscript if it aligns with the journal's editorial direction and contributes to the broader understanding of renal fibrosis treatment strategies.

We appreciate the opportunity to clarify this aspect of our research and are open to further suggestions from the editorial team.

(4) The format of gene and protein abbreviations in the manuscript should be standardized.

Thank you for your comment on the formatting of gene and protein abbreviations in our manuscript. We have carefully reviewed our formatting practices and confirmed that we have adhered to the standard conventions as follows:

(1) Mouse gene names are presented with an initial capital letter and in italics.

(2) Human gene names are written in uppercase and in italics.

(3) Protein names are in all capital letters and not italicized.

We understand the importance of consistency in scientific publications and have ensured that these standards are uniformly applied throughout the revised manuscript. If there were any discrepancies, we have corrected them to maintain the clarity and professionalism.

We appreciate the opportunity to refine our work and are committed to upholding the standards of scientific communication.

(5) There are a few grammar issues throughout the manuscript. The English/grammar could be stronger, thus improving the overall accessibility of the science to readers.

Thank you for bringing the grammar issues to our attention. We have made diligent efforts to revise and improve the manuscript's English and grammar throughout. We have also enlisted the support of a professional language editing service to ensure the clarity and accuracy of our scientific communication.

We are confident that these revisions have significantly enhanced the manuscript's accessibility to a broader readership and have addressed the language concerns raised.

We appreciate your guidance and are committed to delivering a manuscript of the highest quality.

Reviewer #2 (Public Review):

Summary:

This manuscript presents a novel and significant investigation into the role of SIRT4 For CCN2 expression in response to TGF-β by modulating U2AF2-mediated alternative splicing and its impact on the development of kidney fibrosis.

Strengths:

The authors' main conclusion is that SIRT4 plays a role in kidney fibrosis by regulating CCN2 expression via pre-mRNA splicing. Additionally, the study reveals that SIRT4 translocates from the mitochondria to the cytoplasm through the BAX/BAK pore under TGF-β stimulation. In the cytoplasm, TGF-β activated the ERK pathway and induced the phosphorylation of SIRT4 at Ser36, further promoting its interaction with importin α1 and subsequent nuclear translocation. In the nucleus, SIRT4 was found to deacetylate U2AF2 at K413, facilitating the splicing of CCN2 pre-mRNA to promote CCN2 protein expression. Overall, the findings are fully convincing. The current study, to some extent, shows potential importance in this field.

Weaknesses:

(1) Exosomes containing anti-SIRT4 antibodies were found to effectively mitigate UUO-induced kidney fibrosis in mice. While the protein loading capacity and loading methods were not mentioned.

We appreciate your inquiry about the protein loading capacity and methods for the exosomes. As you have correctly noted, these details are indeed essential for the comprehensive understanding of our experimental approach. We have provided these information in the electronic supplementary material, specifically in Section 2.17, where we describe the methodology used for loading the anti-SIRT4 antibodies into the exosomes and the capacity at which this was achieved.

We hope that this additional detail in the supplementary material addresses your concerns and enhances the clarity of our study's methodology.

(2) The method section is incomplete, and many methods like cell culture, cell transfection, gene expression profiling analysis, and splicing analysis, were not introduced in detail.

Thank you for your meticulous review and the feedback provided on our manuscript. We acknowledge your concern regarding the completeness of the methods section.

We would like to clarify that in our initial submission, all text and figures were compiled into a single document, with the supplementary methods detailed at the end, separate from the main text methods. This format was chosen to adhere to submission guidelines that prioritize the concise presentation of core methods in the main text while providing additional details in the supplementary material for comprehensiveness.

The detailed methodologies for cell culture, cell transfection, gene expression profiling analysis, and splicing analysis, which you inquired about, are now indeed included in the revised electronic supplementary material.

We apologize for any misunderstanding caused by the initial structure of our submission and appreciate the opportunity to clarify the comprehensive nature of our methodological reporting.

(3) The authors should compare their results with previous studies and mention clearly how their work is important in comparison to what has already been reported in the Discussion section.

We appreciate the opportunity to discuss the significance of our findings in the broader context of renal fibrosis research. In response to your suggestion, we have further refined our discussion to explicitly compare our results with those of previous studies and to clearly articulate the importance of our work.

(1) Novelty of SIRT4's Role in Renal Fibrosis: Our study introduces a novel concept in the field by demonstrating the nuclear translocation of SIRT4 as a key initiator of kidney fibrosis. This finding diverges from previous studies that have primarily focused on SIRT4's mitochondrial roles, highlighting a new dimension of SIRT4's function in renal pathophysiology.

(2) Mechanistic Insights: We provide a detailed mechanistic pathway, from the release of SIRT4 from mitochondria through the BAX/BAK pore to its subsequent nuclear translocation and impact on U2AF2 deacetylation. This pathway has not been previously described, offering a fresh perspective on the regulation of fibrogenic gene expression.

(3) Implications for Therapy: Our findings suggest potential therapeutic interventions targeting SIRT4 nuclear translocation, which could be a significant advancement over existing treatments that have shown limited efficacy in addressing the root causes of renal fibrosis.

(4) Epigenetic Regulation: By elucidating the role of SIRT4 in regulating alternative splicing of CCN2 pre-mRNA through U2AF2 deacetylation, our study contributes to the growing understanding of epigenetic mechanisms in renal fibrosis, a field that has been understudied compared to genetic factors.

Differential Cellular Roles of SIRT4: Our work indicates that SIRT4 may have distinct roles in different cell types, which is a complex and nuanced aspect of CKD pathophysiology that has not been fully explored in previous research.

Integration with Previous Research: We have compared our findings with existing literature, noting where our work aligns with and diverges from previous studies. This comparison underscores the value of our research in expanding the current paradigm of renal fibrosis.

In conclusion, we believe that our study provides critical insights into the pathogenesis of renal fibrosis and offers a potential therapeutic target. We have clarified these points in the discussion section of our manuscript to ensure that the significance of our work is clearly communicated to the readers.

Reviewer #3 (Public Review):

Summary:

Yang et al reported in this paper that TGF-beta induces SIRT4 activation, TGF-beta activated SIRT4 then modulates U2AF2 alternative splicing, U2AF2 in turn causes CCN2 for expression. The mechanism is described as this: mitochondrial SIRT4 transport into the cytoplasm in response to TGF-β stimulation, phosphorylated by ERK in the cytoplasm, and pathway and then undergo nuclear translocation by forming the complex with importin α1. In the nucleus, SIRT4 can then deacetylate U2AF2 at K413 to facilitate the splicing of CCN2 pre-mRNA to promote CCN2 protein expression. Moreover, they used exosomes to deliver Sirt4 antibodies to mitigate renal fibrosis in a mouse model. TGF-beta has been widely reported for its role in fibrosis induction.

Strengths:

TGF-beta induction of SIRT4 translocation from mitochondria to nuclei for epigenetics or gene regulation remains largely unknown. The findings presented here that SIRT4 is involved in U2AF2 deacetylation and CCN2 expression are interesting.

Weaknesses:

SIRT4 plays a critical role in mitochondria involved in respiratory chain reaction. This role of SIRT4 is critically involved in many cell functions. It is hard to rule out such a mitochondrial activity of SIRT4 in renal fibrosis. Moreover, the major concern is what kind of message mitochondrial SIRT4 proteins receive from TGF-beta. Although nuclear SIRT4 is increased in response to TNF treatment, it is likely de novo synthesized SIRT4 proteins can also undergo nuclear translocation upon cytokine stimulation. TGF-beta-induced mitochondrial calcium uptake and acetyl-CoA should be evaluated for calcium and acetyl-CoA may contribute to the gene expression regulation in nuclei.

Recommendations for the authors:

Reviewer #3 (Recommendations For The Authors):

(1) SIRT4 overall is a mitochondrial enzyme that indeed can undergo shuttling between mitochondria and cytoplasm. Renal fibrosis is a process of complex, SIRT4 deacetylates U2AF4 at K 413.

Thank you for your comment highlighting the known mitochondrial localization of SIRT4 and its role in renal fibrosis.

We concur with the literature that SIRT4 is predominantly a mitochondrial enzyme. However, our study expands upon this understanding by demonstrating a novel shuttling mechanism of SIRT4 between mitochondria and the nucleus in the context of renal fibrosis. Specifically, we observed that under conditions of obstructive nephropathy and renal ischemia reperfusion injury, SIRT4 significantly accumulates in the nucleus, which is a critical event in the fibrotic response.

Our findings reveal that upon TGF-β stimulation, a known inducer of fibrosis, SIRT4 is released from the mitochondria through the BAX/BAK pore and subsequently translocates to the nucleus. This translocation is mediated by the ERK1/2-dependent phosphorylation of SIRT4 at serine 36, which enhances its interaction with importin α1, a key component in nuclear import processes.

Once in the nucleus, SIRT4 exerts its effects on the alternative splicing of CCN2 pre-mRNA by deacetylating U2AF2 at lysine 413. This deacetylation event promotes the formation of the U2 small nuclear ribonucleoprotein (U2 snRNP) and facilitates the splicing of CCN2 pre-mRNA, leading to increased expression of the profibrotic protein CCN2.

Our study, therefore, not only confirms the mitochondrial association of SIRT4 but also uncovers its nuclear function in the regulation of gene expression during renal fibrosis. These findings underscore the complexity of SIRT4's role in cellular processes and its potential as a therapeutic target for fibrotic diseases.

(2) Figure 2 and Figure 3 should be combined.

Thank you for your suggestion to combine Figures 2 and 3 for potential improvement in presentation.

After careful consideration, we have found that merging these figures is not feasible due to space constraints on a standard A4 page, which is necessary to maintain the clarity and detail of the data presented in both figures. Each figure contains complex data that, when combined, would compromise the readability and the integrity of the individual elements.

We believe that the current presentation of Figures 2 and 3 provides a clear and detailed visualization of the data, which is essential for the reader's understanding of our study's findings.

(3) In Figure 4G, the mass spectrum of U2AF2 acetylation on K413 should be included rather than the alignment among species. Moreover, endogenous HAT1 on endogenous U2AF2 rather than exogenous FLAG-U2F2 should be examined.

Thank you for your thoughtful comments and for the suggestion to include the mass spectrum of U2AF2 acetylation on K413 in Figure 4G.

We appreciate the value that the mass spectrometry data would add to our study, providing a direct and definitive assessment of the acetylation status at this specific residue. However, we regret to inform you that our current facilities do not have access to the necessary mass spectrometry equipment to perform these analyses.

While we are unable to include this data in the present manuscript, we concur with the importance of such evidence and plan to undertake these studies in the future. We are in the process of establishing collaborations with laboratories that have the required facilities to perform mass spectrometry. Our intention is to incorporate these data into a follow-up study, which will further validate and expand upon the findings presented in this manuscript.

We believe that our current findings, although lacking the mass spectrometry confirmation, still provide valuable insights into the role of U2AF2 acetylation in [insert relevant biological process]. We have taken care to present our data rigorously and transparently, and we are committed to pursuing the highest standards of experimental validation in our future work.

We hope you will consider the merits of our study in the context of the current limitations and appreciate the opportunity to clarify our position.

Furthermore, regarding the examination of endogenous HAT1's effect on endogenous U2AF2 acetylation levels, we have conducted the necessary experiments. Our results demonstrate that overexpression of HAT1 leads to a significant increase in the acetylation of endogenous U2AF2 (Figure. R2). This new data set has been added to the revised manuscript and supports the role of HAT1 in the regulation of U2AF2 acetylation.

We believe that these revisions address your concerns and provide a more comprehensive understanding of the molecular mechanisms underlying the regulation of U2AF2 acetylation.

We appreciate the opportunity to improve our manuscript based on your constructive feedback and hope that our revisions meet with your satisfaction.

Author response image 2.

HAT1 OE reduces the acetylation of endogenous U2AF2

(4) Figure 6F. Does portien mean protein?

Thank you for your careful review and insightful comments on our manuscript. You are correct in pointing out the error regarding the term "portien" in Figure 6F. It was indeed a typographical oversight on our part, and we apologize for any confusion this may have caused.

We have made the necessary correction to ensure that "protein" is accurately used in place of "portien" in Figure 6F. We appreciate the opportunity to enhance the clarity and accuracy of our presentation.

(5) The authors should pay attention to their writing. There are many typos and other issues with the use of the English language and grammar.

Thank you for bringing the grammar issues to our attention. We have made diligent efforts to revise and improve the manuscript's English and grammar throughout. We have also enlisted the support of a professional language editing service to ensure the clarity and accuracy of our scientific communication.

We are confident that these revisions have significantly enhanced the manuscript's accessibility to a broader readership and have addressed the language concerns raised.

We appreciate your guidance and are committed to delivering a manuscript of the highest quality.