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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This study performed whole genome sequencing (WGS) on a large cohort of hypoplastic left heart syndrome (HLHS) patients and their families to identify candidate. Nine candidate genes with rare, predicted damaging homozygous variants were identified. Of the candidate HLHS gene homologs tested, cardiac-specific knockdown (KD) of the mitochondrial contact site and cristae organization system (MICOS) complex subunit dCHCHD3/6 resulted in drastically compromised heart contractility, diminished levels of sarcomeric actin and myosin, reduced cardiac ATP levels, and mitochondrial fission-fusion defects. These heart defects were similar to those inflicted by cardiac KD of ATP synthase subunits of the electron transport chain (ETC), consistent with the MICOS complex's role in maintaining cristae morphology and ETC complex assembly. Analysis of 183 genomes of HLHS patient-parent trios revealed five additional HLHS probands with rare, predicted damaging variants in CHCHD3 or CHCHD6. Hypothesizing an oligogenic basis for HLHS, the authors tested 60 additional prioritized candidate genes in these cases for genetic interactions with CHCHD3/6 in sensitized fly hearts. Moderate KD of CHCHD3/6 in combination with Cdk12 (activator of RNA polymerase II), RNF149 (E3 ubiquitin ligase), or SPTBN1 (scaffolding protein) caused synergistic heart defects, suggesting the potential involvement of a diverse set of pathways in HLHS.

General Comments:

The authors performed an elegant series of experiments that implicate variants of dCHCHD3/6 in HLHS patients as contributing to mitochondrial and sarcomeric defects and contractile function defects. Demonstrating in Drosphilia the functional and biochemical implications of knocking out dCHCHD3/6 provides some potentially important insights into the functional and biochemical implications of dCHCHD3/6 variants in HLHS patients. The data is also complemented by hiPSC-CM studies in which knockdown of CHCHD6 and CHCHD3 showed similar alterations in ATP synthase and mitochondrial morphology.

The authors nicely show that knock down of the subunit dCHCHD3/6 resulted in drastically compromised heart contractility, diminished levels of sarcomeric actin and myosin, reduced cardiac ATP levels, and mitochondrial fission-fusion defects in the Drosphilia. What is not clear is how these changes mirror the phenotype of HLHS in humans. It would helpful to speculate to a greater extent as to how these changes would manifest as a decreased left ventricular development in HLHS.

Specific Comments:

Line 139: Figure 1A does not show echos from the siblings.

Line155: This table is listed as "Table 1" not Supplemental Table 1.

Significance

This is a highly significant study. The main audience would be pediatric cardiologists and geneticists.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript titled "mitochondrial MICOS complex genes, implicated in hypoplastic left heart syndrome, maintain cardiac contractility and actomyosin integrity", Katja Birker et al. reveal that CHCHD3/6 cardiac-specific KD caused reduced contractility and decreased sarcomeric F-Actin and Myosin staining in fly due to impaired ATP synthase. The findings shown in this manuscript are interesting. However, the additional experiments are needed to confirm the conclusion before publication.

Major comments:

1. The authors mentioned that the heart dysfunction observed upon CHCHD3/6 KD may be mediated via defects in ATP synthase. Then, how does CHCHD3/6 KD affect ATP synthase? Additionally, OPA1 also affects ATP synthase, why does OPA1 KD just reduce fractional shortening (S.T.2) without reducing F-actin staining?
2. It has been reported that CHCHD3 KD in HeLa cells causes fragmented mitochondria, so how does CHCHD3/6 KD caused mitochondrial aggregation? What is the mechanism?
3. The ultrastructure of mitochondria (especially aggregated mitochondria) in control and CHCHD3/6 KD heart of drosophila should be analyzed by TEM.

Significance

The manuscript partially illustrate the relationship between MICOS complex with Hypoplastic left heart syndrome (HLHS), which is intertesing to the reader.

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

Learn more at Review Commons

---

Reply to the reviewers

Reply to the Reviewers

We thank the reviewers for dedicating time to review our manuscript and providing highly valuable feedback. Please find below a point-by-point answer.

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

Summary:

In the present manuscript, van der Plas et al. compellingly illustrated a novel technique for engendering a whole-brain functional connectivity map from single-unit activities sampled through a large-scale neuroimaging technique. With some clever tweaks to the restricted Boltzmann Machine, the cRBM network is able to learn a low-dimensional representation of population activities, without relying on constrained priors found in some traditional methods. Notably, using some 200 hidden layer neurons, the employed model was able to capture the dynamics of over 40,000 simultaneously imaged neurons with a high degree of accuracy. The extracted features both illustrate the anatomical topography/connectivities and capture the temporal dynamics in the evolution of brain states. The illustrated technique has the potential for wide-spread applications spanning diverse recording techniques and animal species. Furthermore, the prospectives of modeling whole-brain network dynamics in 'neural trajectory' space and of generating artificial data in silico make for very enticing reasons to adopt cRBM.

Major comments:

1. Line 164. The authors claim that conventional methods "such as k-means, PCA and non-negative matrix factorization" cannot be quantitatively assessed for quality on the basis that they are unable to generate new artificial data. Though partly true, in most neuroscience applications, this is hardly cause for concern. Most dimensionality reduction methods (with few exceptions such as t-sne) allow new data points to be embedded into the reduced space. As such, quality of encoding can be assessed by cross-validation much in the same way as the authors described, and quantified using traditional metrics such as percentage explained variance. The authors should directly compare the performance of their proposed model against that of NNMF and variational auto-encoders. Doing so would offer a more compelling argument for the advantage of their proposed method over more widely-used methods in neuroscience applications. Furthermore, a direct comparison with rastermap, developed by Stringer lab at Janelia (https://github.com/MouseLand/rastermap), would be a nice addition. This method presents itself as a direct competitor to cRBM. Additionally, the use of GLM doesn't do complete justice to the comparison point used, since a smaller fraction of data were used for calculating performance using GLM, understandably due to its computationally intensive nature.

PLANNED REVISION #2

We thank the reviewer for the comment, and certainly agree that there are multiple methods for unsupervised feature extraction from data and that they can be validated for encoding quality by cross-validation. However, we stress that reconstructing through a low-dimensional, continuous bottleneck is a different (and arguably, easier) task, than generating whole distributions. Reconstruction delineates the manifold of possible configurations, whereas generative modeling must weigh such configurations adequately. Moreover, none of the methodologies mentioned can perform the same tasks as cRBMs. For instance, NNMF learns localized assemblies, but cannot faithfully model inhibitory connections since, by definition, only non-negative weights are learnt. Also, the connection between the learnt assemblies and the underlying connectivity is unclear. Similarly, rastermap is an algorithm for robustly i) sorting neurons along a set number of dimensions (typically 1 or 2) such that neighboring neurons are highly correlated, and ii) performing dimensionality reduction by clustering along these dimensions. Because Rastermap uses k-means as the basis for grouping together neurons, it does not quantify connections between neurons and assemblies, nor assign neurons to multiple assemblies. Moreover, it is not a generative model, and thus cannot predict perturbation experiments, infer connectivities or assign probabilities to configurations. Therefore, we do not believe that NNMF or Rastermap would be a suitable alternative for cRBM in our study. We nonetheless appreciate the reviewer’s suggestions and agree that we should motivate more clearly why these methods are not applicable for our purposes. Therefore, to emphasize the relative merit of cRBM with respect to other unsupervised algorithms, we now provide a table (Supplementary Table 2) that lists their specific characteristics. We stress that we do not claim that cRBM are consistently better than these classical tools for dimensionality reduction, but focus only on the properties relevant to our study.

Further, we agree that VAEs, which also jointly learn a representation and distribution of data, are close competitors of cRBMs. In Tubiana et al. Neural Computation 2019, we previously compared sparse VAEs with cRBMs for protein sequence modeling, and found that RBMs consistently outperformed VAEs in terms of the interpretability-performance trade-off. In the revised manuscript, we propose to repeat the comparison with VAE for zebrafish neural recordings, and expect similar conclusions.

As for GLM, it is true that the comparison involved subsampling of the neurons (due to the very high computational cost of GLM, where we could estimate the connectivity of 1000 neurons per day). This was already denoted in the relevant figure caption, as the reviewer has seen, but we have now also clarified this point in Methods 7.10.3. Still, we performed our GLM analysis on 5000 neurons (using all neurons as regressors), which is 10% of all neurons, and we believe this is a sufficient number for comparison. This emphasizes the ability of our optimized cRBMs to handle very large datasets, such as the presently used zebrafish whole brain recordings.

Line 26. The authors describe their model architecture as a formalization of cell assemblies. Cell assemblies, as originally formulated by Hebb, pertains to a set of neurons whose connectivity matrix is neither necessarily complete nor symmetric. Critically, in the physiological brain, the interactions between the individual neurons that are part of an assembly would occur over multiple orders of dependencies. In a restricted Boltzmann machine, neurons are not connected within the same layer. Instead, visible layer neurons are grouped into "assemblies" indirectly via a shared connection with a hidden layer neuron. Furthermore, a symmetrical weight matrix connects the bipartite graph, where no recurrent connectivities are made. As such, the proposed model still only elaborates symmetric connections pertaining to first-order interactions (as illustrated in Figure 4C). Such a network may not be likened with the concept of cell assemblies. The authors should refrain from detailing this analogy (of which there are multiple instances of throughout the text). It is true that many authors today refer to cell assemblies as any set of temporally-correlated neurons. However, saying "something could be a cell assembly" is not the same as saying "something is a cell assembly". How about sticking with cRBM-based cell assemblies (as used in section 2.3) and defining it beforehand?

We thank the reviewer for this excellent question. We agree that there is, in general, a discrepancy between computationally-defined assemblies and conceptual/neurophysiological definition of cell assemblies. We have added a clarification in Results 2.1 to clarify the use of this term when it first occurs in Results. However, we still believe that our work contributes to narrowing the gap. Indeed, our RBM-defined assemblies are i) localized, ii) overlapping, iii) rooted in connectivity patterns (both excitatory and inhibitory), and iv) cannot be reduced to a simple partitioning of the brain with full & uniform connectivity within and between partitions. This is unlike previous work based on clustering (no overlaps or heterogeneous weights), NNMF (no inhibition) or correlation network analysis (no low-dimensional representation).

Regarding the specific comments pointed here, we stress that:

• Effective interactions between neurons are not purely pairwise (“First order”), due to the usage of the non-quadratic potential. (see Eqn 12-13). If the reviewer means by “First-order” interactions the lack of hierarchical organization, we agree, to some extent: in the current formulation, correlations between assemblies are mediated by overlaps between their weights. Fully-hierarchical organization, e.g. by using Deep Boltzmann Machines or pairwise connections within the hidden layer is an interesting future direction, but on the other hand may make it hard to clearly identify assemblies as they might be spread out over multiple layers
• Neurons that participate in a given assembly (as defined by a specific hidden unit) are not all connected with one another with equal strength. Indeed, these neurons may participate in other assemblies, resulting in heterogeneity of connections (see Eqn. 15-17) and interactions between assemblies.
• We acknowledge that the constraint of symmetrical connections is a core limitation of our method. Arguably, asymmetric connections are critical for predicting temporal evolution but less important for inferring a steady-state distribution from data, as we do here. In the revised submission, we added a new paragraph in the discussion section (lines 351-357) in which these limitations are discussed, including the imposed symmetry of the connections and the lack of hierarchical structures, copied below. We trust that this addresses the reviewer’s criticism:

---

In sum, cRBM-inferred cell assemblies display many properties that one expects from physiological cell assemblies: they are anatomically localized, can overlap, encompass functionally identified neuronal circuits and underpin the collective neural dynamics (Harris, 2005, 2012; Eichenbaum, 2018). Yet, the cRBM bipartite architecture lacks many of the traits of neurophysiological circuits. In particular, cRBMs lack direct neuron-to-neuron connections, asymmetry in the connectivity weights and a hierarchical organization of functional dependencies beyond one hidden layer. Therefore, to what extent cRBM-inferred assemblies identify to neurophysiological cell assemblies, as postulated by Hebb (1949) and others, remains an open question.

---

I would strongly recommend adding a paragraph discussing the limitation of using the cRBM, things future researchers need to keep in mind before using this method. One such recommendation is moving the runtime-related discussion for cRBM, i.e. 8-12 hrs using 16 CPU from Methods to Discussion, since it's relevant for an algorithm like this. Additionally, a statement mentioning how this runtime will increase with the length of recordings and/or with the number of neurons might be helpful. What if the recordings were an hour-long rather than 25mins. This would help readers decide if they can easily use a method like this.

We thank the reviewer for the suggestion, and agree that it is important to cover the computational cost in the main text. Regarding the runtime for longer recordings, the general rule of thumb is that the model requires a fixed number of gradient updates to converge (20-80k depending on the data dimensionality) rather than a fixed number of epochs. Thus, runtime should not depend on recording length, as the number of epochs can be reduced for longer recordings. While we did not verify this rule for neural recordings, this is what we previously observed when modeling protein/DNA sequence data sets, whose size range from few hundreds to hundreds of thousands of samples (Tubiana et al., 2019, eLife; Tubiana et al. 2019, Neural Computation; Bravi et al. Cell Systems 2021; Bravi et al. PLOS CB 2021; Fernandez de Cossio Diaz et al. Arxiv 2022 Di Gioacchino et al. BiorXiv 2022). We have now added a summary of these points in Methods 7.7.2, also refer to this with explicit mention of the runtime in the Discussion, end of 2nd paragraph:

---

By implementing various algorithmic optimizations (Methods 7.7), cRBM models converged in approximately 8-12 hours on high-end desktop computers (also see Methods 7.7.2).

---

Line 515. A core feature of the proposed compositional RBM is the addition of a soft sparsity penalty over the weight matrix in the likelihood function. The authors claim that "directed graphical models" are limited by the a priori constraints that they impose on the data structure. Meanwhile, a more accurate statistical solution can be obtained using a RBM-based model, as outlined by the maximum entropy principle. The problem with this argument is that the maximum entropy principle no longer applies to the proposed model with the addition of the penalty term. In fact, the lambda regularization term, which was estimated from a set of data statistics motivated by the experimenter's research goals (Figure S1), serves to constrict the prior probability. Moreover, in Figure S1F, we clearly see that reconstruction quality suffers with a higher penalty, suggesting that the principle had indeed been violated. That being said, RBMs are notoriously hard to train, possibly due to the unconstrained nature of the optimization. I believe that cRBM can help bring RBM into wider practical applications. The authors could test their model on a few values of the free parameter and report this as a supplementary. I believe that different parameters of lambda could elaborate on different anatomical clusters and temporal dynamics. Readers who would like to implement this method for their own analysis would also benefit tremendously from an understanding of the effects of lambda on the interpretation of their data. Item (1) on line 35 (and other instances throughout the text) should be corrected to reflect that cRBM replaces the hard constraints found in many popular methods with a soft penalty term, which allows for more accurate statistical models to be obtained.

We thank the reviewer for their analysis and suggestion. Indeed, adding the regularization term - not present in the classical formulation of the RBM (Hinton & Salakhutdinov, 2006, Science) - was critical for significantly enhancing its performance, which allowed us to implement this model on our large scale datasets (~50K visible units). We agree that providing more information on the effect of the regularization term will benefit readers who would like to use this method, and we propose to add this in the revised manuscript, which would implement the reviewer’s suggestion. See “PLANNED REVISION #1”.

The reviewer’s comment on the Maximum Entropy issue calls for some clarification. The maximum entropy principle is a recipe for finding the least constrained model that reproduces specified data-dependent moments. However, it cannot determine which moments are statistically meaningful in a finite-sized data set. A general practice is to only include low-order moments (1st and 2nd), but this is sometimes already too much for biological data. Regularization provides a practical means to select stable moments to be fitted and others to be ignored. This can be seen from the optimality condition, which writes, e.g., for the weights wi,mu:

| i h,mu>data - i h,mu>model | i,mu = 0.

i h,mu>data - i h,mu>model | = lambda sign(wi,mu) if |wi,mu| > 0.

Essentially, this lets the training decide which subset of the constraints should actually be used. Thus, regularized models are closer to the uniform distribution (g=w=0), and actually have higher entropy than unregularized one (see, e.g., Fanthomme et al. Journal of Statistical Mechanics, 2022). Therefore, we believe that a regularized maximum entropy model can still be considered a bona fide MaxEnt model. This formulation should not be confused with another formulation (that perhaps the reviewer has in mind) where a weighted sum of the entropy and the regularization term is maximized under the same moment-matching constraints. In this case, we agree that maximum entropy principle (MaxEnt) would be violated.

The choice of regularization value should be dictated by bias-variance trade-off considerations. Ideally, we would use the same criterion as for training, i.e., maximization of log-likelihood for the held-out test set, but it is intractable. Thus, we used a consensus between several tractable performance metrics as a surrogate; we believe this consensus to be principally independent of the research goal. While the reconstruction error indeed increases for large regularization values, this is simply because too few constraints are retained at high regularizations.

Essentially, the parameters selected by likelihood maximization find the finest assembly scale that can be accommodated by the data presented. Thus, the number and size of the assemblies are not specified by the complexity of the data set alone. Rather, the temporal resolution and length of the recordings play a key role; higher resolution recordings will allow the inference of a larger number of smaller assemblies, and enable the study of their hierarchical organization.

That being said, we fully agree that the regularization strength and number of hidden units have a strong impact on the nature of the representation learnt. In the revised manuscript, we will follow the reviewer’s suggestion and provide additional insights on the effect of these parameters on the representation learnt (please see revision plan).

Minor comments:

From a neuroscience point of view, it might be interesting to show what results are achieved using different values of M (say 100 or 300), rather than M=200, while still maintaining the compositional phase. Is there any similarity between the cRBM-based cell assemblies generated at different values of M? Is there a higher chance of capturing certain dynamics either functional or structural using cRBM? For example, did certain cRBM-based cell assemblies pop up more frequently than others at all values of M (100,200,300)?

This point will be addressed in the future, as detailed in our response to reviewer 2 (see PLANNED REVISION #1).

The authors have mentioned that this approach can be readily applied to data obtained in other animal models and using different recording techniques. It might be nice to see a demonstration of that.

We agree that showing additional data analysis would be interesting, but we feel that it would overburden the supplementary section of the manuscript, which is already lengthy. In previous works, we and collaborators have used cRBMs for analyzing MNIST data (Tubiana & Monasson, 2017, PRL; Roussel et al. 2022 PRE), protein sequence data (Tubiana et al., 2019, eLife; Tubiana et al. 2019, Neural Computation; Bravi et al. Cell Systems 2021; Bravi et al. PLOS CB 2021; Fernandez de Cossio Diaz et al. Arxiv 2022), DNA sequences (Di Gioacchino et al. BiorXiv 2022), spin systems (Harsh et al. J. Phys. A 2020), etc. Many are included as example notebooks - next to the zebrafish data - in the linked code repository. For neural data, we have recently shared our code with another research group working on mice auditory cortex (2-photon, few thousands of neurons, Léger & Bourdieu). Preliminary results are encouraging, but not ready for publication yet.

Line 237. The justification for employing a dReLU transfer function as opposed to ReLU is unclear, at least within the context of neurobiology. Given that this gives rise to a bimodal distribution for the activity of HUs, the rationale should be clearly outlined to facilitate interpretability.

We thank the reviewer for the question. As we detail in the manuscript (Methods), the dReLU potential is one of the sufficient requirements for the RBM to achieve the compositional phase. The compositional phase is characterized by localized assemblies that co-activate to generate the whole-brain neural dynamics. This property reflects neurobiological systems (Harris, 2005, Neuron), which is one of the reasons why we employed compositional RBMs for this study.

As the reviewer points out, the HUs that we infer exhibit bimodal activity (Figure 4). Importantly, the HU activity is not constrained by the model to take this shape, as dReLU potentials allow for several activity distributions (see Methods 7.5.4; “Choice of HU potential”). In fact, ReLU potentials are a special case of dReLU (by $(\gamma_{\mu, -} \to \infty)$), so our model allows HU potentials to behave like ReLUs, but in practice they converge to a double-well potential for almost all HUs, leading to bimodal activity distributions.

Following the suggestion of the reviewer, we have now added this detail for clarity in Methods 7.5.4 and referenced this Methods section at line 237.

Reviewer #1 (Significance (Required)):

van der Plas et al. highlighted a novel dimensionality reduction technique that can be used with success for discerning functional connectivities in large-scale single-unit recordings. The proposed model belongs to a large collection of dimensionality reduction techniques (for review, Cunningham, J., Yu, B. Dimensionality reduction for large-scale neural recordings. Nat Neurosci 17, 1500-1509 (2014). https://doi.org/10.1038/nn.3776; Paninski, L., & Cunningham, J. P. (2018). Neural data science: accelerating the experiment-analysis-theory cycle in large-scale neuroscience. Current opinion in neurobiology, 50, 232-241.). The authors themselves highlighted some of the key methods, such as PCA, ICA, NNMF, variational auto-encoders, etc. The proposed cRBM model has also been published a few times by the same authors in previous works, although specifically pertaining to protein sequences. The use of RBM-like methods in uncovering functional connectivities is not novel either (see Hjelm RD, Calhoun VD, Salakhutdinov R, Allen EA, Adali T, Plis SM. Restricted Boltzmann machines for neuroimaging: an application in identifying intrinsic networks. Neuroimage. 2014 Aug 1;96:245-60. doi: 10.1016/j.neuroimage.2014.03.048.). However, given that the authors make a substantial improvement on the RBM network and have demonstrated the value of their model using physiological data, I believe that this paper would present itself as an attractive alternative to all readers who are seeking better solutions to interpret their data. However, as I mentioned in my comments, I would like to see more definitive evidence that the proposed solution has a serious advantage over other equivalent methods.

Reviewer's expertise:

This review was conducted jointly by three researchers whose combined expertise includes single-unit electrophysiology and two-photon calcium imaging, using which our lab studies the neurobiology of learning and memory and spatial navigation. We also have extensive experience in computational neuroscience, artificial neural network models, and machine learning methods for the analysis of neurobiological data. We are however limited in our knowledge of mathematics and engineering principles. Therefore, our combined expertise is insufficient to evaluate the correctness of the mathematical developments.

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

In their manuscript, van der Plas et al. present a generative model of neuron-assembly interaction. The model is a restricted Boltzmann machine with its visible units corresponding to neurons and hidden units to neural assemblies. After fitting their model to whole-brain neural activity data from larval zebrafish, the authors demonstrate that their model is able to replicate several data statistics. In particular, it was able to replicate the pairwise correlations between neurons as well as assemblies that it was not trained on. Moreover, the model allows the authors to extract neural assemblies that govern the population activity and compose functional circuits and can be assigned to anatomical structures. Finally, the authors construct functional connectivity maps from their model that are then shown to correlate with established structural connectivity maps.

Overall, the authors present convincing evidence for their claims. Furthermore, the authors state that their code to train their restricted Boltzmann machine models is already available on GitHub and that the data underlying the results presented in this manuscript will be made publicly available upon publication, which will allow people to reproduce the results and apply the methods to their data.

One thing the authors could maybe discuss a bit more is the "right" parameter value M, especially since they used the optimal value of 200 found for one sample also for all the others. More specifically, how sensitive are the results to this value?

PLANNED REVISION #1

In the following we jointly address three of the reviewers’ questions (2 from reviewer 1, and 1 from reviewer 2).

Shortly summarized, the cRBM model has 2 free parameters; the number of hidden units M and the regularization parameter lambda. In figures 2 and S1 we optimize their values through cross-validation, and then perform our further analyses on models with these optimal values. The reviewers ask us to examine the outcome of the model for slightly different values of both parameters, in particular in relation to the sensitivity of the cRBM results to selecting the optimal parameters and the change in inferred assemblies and their dynamics.

We thank the reviewers for these questions and appreciate their curiosity to understand the effects of changing either of these two free parameters.

We inspected these models when we performed the model selection (of Figures 2 and S1), but did not formalize our findings into figures for the manuscript. We found that small changes in the parameter setting did not abruptly change the inferred assemblies (e.g., M=100) apart from slightly changing in size, so we expect that the statistics that we intend to include in the proposed supplementary figure would reflect that, and it would definitely benefit the manuscript to include this analysis. Very-low-M settings are interesting to include, because assemblies are much larger - essentially merging smaller assemblies of higher-M models - at the cost of model performance.

We propose to create additional supplementary figures that address these questions. As suggested, we will pick a few example cRBMs with different parameter settings (below-optimal M, above-optimal M, and same for lambda), as well as very low M settings (M~20 or 50). We will then show example assemblies and assembly dynamics, as well as the relevant statistics (assembly size, dynamics time scale etc) that describe them.

And, what happens if one would successively increase that number, would the number of assemblies (in the sense of hidden units that strongly couple to some of the visible units) eventually saturate?

This point will be addressed by inspecting models at different M values (see Revision Plan #1). We would like to further answer this question by referring to past work. In Tubiana et al., 2019, elife (Appendix 1) we have done this analysis, and the result is consistent with the reviewer’s intuition. Because of the sparsity regularization, if M becomes larger than its optimum, the assemblies further sparsify without benefiting model performance, and eventually new assemblies duplicate previous assemblies or become totally sparse (i.e., all weights = 0) to not further induce a sparsity penalty in the loss function. So the ‘effective’ number of assemblies indeed saturates for high M.

Moreover, regarding the presentation, I have a few minor suggestions and comments that the authors also might want to consider:

* In Figure 6C, instead of logarithmic axes, it might be better to put the logarithmic connectivity on a linear axis. This way the axes can be directly related to the colour bars in Figures 6A and B.

We agree and thank you for the suggestion. We have changed this accordingly (and also in the equivalent plots in figure S6).

* In Equation (8), instead of $\Gamma_{\mu}(I)$ it should be $\Gamma_{\mu}(I_{\mu}(v))$.

Done, thank you.

* In Section 7.0.5, it might make sense to have the subsection about the marginal distributions before the ones about the conditional distributions. The reason would be that if one wants to confirm Equation (8) one necessarily has to compute the marginal distribution in Equation (12) first.

We thank the reviewer for the suggestion, but respectfully propose to leave the section ordering as is. We understand what the reviewer means, but Equation (8) can also be obtained by factorizing P(v,h) Equation (7) and removing the v_i dependency. In Equation (8), \Gamma can then be obtained by normalization. We believe this flow aligns better with the main text (where conditionals come first, when used for sampling, followed by the marginal of P(v) used for the functional connectivity inference).

* In Line 647f, the operation the authors are referring to is strictly speaking not an L1-norm of the matrix block. It might be better to refer to that e.g. as a normalised L1-norm of the matrix block elements.

Done, thank you.

* In Line 22, when mentioning dimensionality reduction methods to identify assemblies, it might make sense to also reference the work by Lopes-dos-Santos et al. (2013, J. Neurosci. Methods 220).

Done, thank you for the suggestion.

Reviewer #2 (Significance (Required)):

The work presented in this manuscript is very interesting for two reasons. First, it has long been suggested that assemblies are a fundamental part of neural activity and this work seems to support that by showing that one can generate realistic whole-brain population activity imposing underlying assembly dynamics. Second, in recent years much work has been devoted to developing methods to find and extract neural assemblies from data and this work and the modelling approach can also be seen as a new method to achieve that. As such, I believe this work is relevant for anyone interested in neural population activity and specifically neural assemblies and certainly merits publication.

Regarding my field of expertise, I used to work on data analysis of neural population activity and in particular on the question of how one can extract neural assemblies from data. I have to say that I have not much experience with fitting statistical models to data, so I can't provide any in-depth comments on that part of the work, although what has been done seems plausible.

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

Summary: Understanding the organization of neural population activities in the brain is one of the most important questions in neuroscience. Recent technique advance has enabled researchers to record a large number of neurons and some times the whole brain. Interpreting and extracting meaningful insights from such data sets are challenging. van der Pals \textit{et al} applied a generative model called compositional Restricted Boltzmann Machine (cRBM) to discover neuron assemblies from spontaneous activities of zebra fish brain. They found that neurons can be grouped into around 200 assemblies. Many of them have clear neurophysiological meaning, for example, they are anatomically localized and overlapped with known neural circuits. The authors also inferred a coarse-grained functional connectivity which is similar to known structural connectivity.

The structure of the paper is well organized, the conclusion seems well supported by their numerical results. While this study provides a compelling demonstration that cRBM can be used to uncover meaningful structures from large neural recordings, the following issues limit my enthusiasm.

Major:

1) The overall implication is not clear to me. Although the authors mentioned this briefly in the discussion. It is not clear what else do we learn from discovered assemblies beyond stating that they are consist with previous study. For example, the author could have more analysis of the assembly dynamics, such as whether there are low dimensional structure etc.

First, we will comment on our analysis of the assemblies, before we continue to discuss the main implications of our work, which we believe are the inferred generative model of the zebrafish brain and the perturbation-based connectivity matrix that we discovered. Further, we have implemented the reviewer’s suggestion of analyzing the low-dimensional structure of the hidden unit activity, as further detailed in Question 7.

Indeed, the example assemblies that we show in Figure 3 have been thoroughly characterized in previous studies, which is why we chose to showcase these examples. Previous studies (including our own) typically focused on particular behaviors or sensory modalities, and aimed at identifying the involved neural circuit. Here, we demonstrate that by using cRBM on spontaneous activity recordings, one can simultaneously identify many of those circuits. In other words, these functional circuits/assemblies activate spontaneously, but in many different combinations and perhaps infrequently, so that it is very difficult to infer them from the full neural dynamics that they generate. cRBM has been able to do so, and Figure 3 (and supplementary video 1) serve to illustrate the variety of (known) circuits and assemblies that it inferred, some of which may represent true but not yet characterized circuits, which thus provide hypotheses for subsequent studies.

Further, we believe that the implication of our study goes beyond the properties of the assemblies we’ve identified, in several ways.

We demonstrate the power of cRBM’s generative capacity for inferring low-dimensional representations in neuroscience. Unlike standard dimensional reductionality methods, generative models can be assessed by comparing the statistics of experimental vs in-silico generated data. This is a powerful approach to validate a model, rarely used in neuroscience because of the scarcity of generative models compatible with large-scale data, and we hope that our study will inspire the use of this method in the field. We have made our cRBM code available, including notebook tutorials, to facilitate this.

The generative aspect of our model allowed us to predict the effect of single-neuron perturbations between all ~ 10^9 pairs of neurons per fish, resulting in a functional connectivity matrix.

We believe that the functional connectivity matrix is a major result for the field, similar to the structural connectivity matrix from Kunst et al., 2019, Neuron. The relation between functional and structural connectivity is unknown and of strong interest to the community (e.g., Das & Fiete, 2020, Nature Neuroscience). Our results allowed for a direct comparison of whole-brain region-by-region structural and functional connectivity. We were thus able to quantify the similarity between these two maps, and to identify specific region-pair matches and non-matches of functional and structural connectivity - which will be of particular interest to the zebrafish neuroscience community for developing future research questions.

Further, using these trained models - that will be made public upon publication - anyone can perform any type of in silico perturbation experiments, or generate endless artificial data with matching data statistics to the in vivo neural recordings.

We hope that this may convince the reviewer of the multiple directions of impact of our study. We will further address their comment on analysis of assembly dynamics below (question 7).

2) The learning algorithm of cRBM can be interpreted as matching certain statistics between the model and the experiment. For a general audience, it is not easy to understand $\langle h_{\mu} \rangle_{data}, \langle v_ih_{\mu}\rangle_{data}$. Since these are not directly calculated from experimental observed activities $v_i$, but rather the average is conditioned on the empirical distribution of $p(v_i)$. For example, the meaning of $\langle v_ih_{\mu}\rangle_{data}$ means

\begin{equation}

\langle v_ih_{\mu}\rangle_{data} = \frac{1}{l}\sum_{\mathbf{v}\in S} \mathbb{E}{p(\mathbf{h|v})}(v_ih{\mu}),

\end{equation}

where $S$ is the set of all observed neural activities: $S = {\mathbf{v}^1, \cdots, \mathbf{v}^l}$. The authors should explain this in the main text or method, since they are heavily loaded in figure 2.

We thank the reviewer for their suggestion, and have now implemented this. Their mathematics are correct; and we agree that it is not easy to understand without going through the full derivation of the (c)RBM. At the same time, we have tried not to alienate readers who might be more interested in the neuroscience findings than in understanding the computational method used. Therefore, we have kept mathematical details in the main text to a minimum (and have used schematics to indicate the statistics in Figures 2C-G), while explaining it in detail in Methods.

Accordingly, we have now extended section 7.10.2 (“Assessment of data statistics”) that explains how the data statistics were computed in Methods (and have referenced this in Results and in Methods 7.5.5), using the fact that we already explain the process of conditioning on __v __in Methods 7.5.1. The following sentences were added:

---

[..] However, because (c)RBM learn to match data statistics to model statistics (see Methods

7.5.5), we can directly compare these to assess model performance. [..]

[..]

For each statistic 〈 fk〉 we computed its value based on empirical data 〈 fk〉_data and on the model 〈 fk〉_model, which we then quantitatively compared to assess model performance. Data statistics 〈 fk〉_data were calculated on withheld test data (30% of recording). Naturally, the neural recordings consisted only of neural data v and not of HU data h. We therefore computed the expected value of __ht at each time point t conditioned on the empirical data _v_t, as further detailed in Methods 7.5.1.

[..]

---

3) As a modeling paper, it would be great to have some testable predictions.

We thank the reviewer for the enthusiasm and suggestion. We agree, and that is why we have included this in the form of functional connectivity matrices in Figures 5 and 6. To achieve this, we leveraged the generative aspect of the cRBM to perform in silico single-neuron perturbation experiments, which we aggregated to connectivity matrices. In other words, we have used our model to predict the functional connectivity between brain regions using the influence of single-neuron perturbations.

Obtaining a measure of functional connectivity/influence using single-neuron perturbations is also possible using state-of-the-art neuro-imaging experiments (e.g., Chettih & Harvey, 2019, Nature), though not at the scale of our in silico experiments. We therefore verify our predictions using structural data from Kunst et al., 2019, which we have extended substantially. We provide our functional connectivity result in full, and hope that this can inspire future zebrafish research by predicting which regions are functionally connected, which includes many pairs of regions that have not yet directly been studied in vivo.

Minor:

1) The assembly is defined by the neurons that are strongly connected with a given hidden unit. Thus, some neurons may enter different assemblies. A statistics of such overlap would be helpful. For example, a ven diagram in figure 1 that shows how many of them assigned to 1, 2, etc assemblies.

We thank the reviewer for this excellent suggestion. Indeed, neurons can be embedded in multiple assemblies. This is an important property of cRBMs, which deserves to be quantified in the manuscript. We have now added this analysis as a new supplementary figure 4. Neurons are embedded in an assembly if their connecting weight w_{i, \mu} is ‘significantly’ non-zero, depending on what threshold one uses. We have therefore shown this statistic for 3 values of the threshold (0.001, 0.01 and 0.1) - demonstrating that most neurons are strongly embedded in at least 1 assembly and that many neurons connect to more than 1 assembly.

Updated text in Results:

---

Further, we quantified the number of assemblies that each neuron was embedded in, which showed that increasing the embedding threshold did not notably affect the fraction of neurons embedded in at least 1 assembly (93% to 94%, see Figure S4).

---

2) What does the link between hidden units in Figure 1B right panel mean?

Thank you for the question, and we apologize for the confusion: if we understand the question right, the reviewer asks why the colored circles under the title ‘Neuronal assemblies of Hidden Units’ are linked. This schematic shows the same network of neurons as shown in gray at the left side of Fig 1B, but now colored by the assembly ‘membership’ of each neuron. Hence, the circles shown are still neurons (and not HUs), and their links still represent synaptic connections between neurons. We apologize for the confusion, and have updated the caption of Fig 1B to explain this better:

---

“[..] The neurons that connect to a given HU (and thus belong to the associated assembly), are depicted by the corresponding color labeling (right panel).[..]”.

---

3) A side-by-side comparison of neural activity predicted by model and the experimentally recorded activities would help the readers to appreciate the performance of the model. Such comparison can be done at both single neuron level or assembly level.

We thank the reviewer for this suggestion. The cRBM model is a statistical model, meaning that it fits the statistics of the data, and not the dynamics. The data that it generates therefore (should) adhere to the statistics of the training data, but does not reflect their dynamics. We believe that showing generated activity side-by-side of empirical activity is therefore not a meaningful example of generated data, as this would exemplify the dynamics, which this model is not designed to capture. Instead, in Figure 2, we show the statistics of the generated data versus the statistics of the empirical data (e.g., Fig 2C for the mean activity of all neurons). We believe that this is a better example representation of the generative performance of the model.

4) Definition of reconstruction quality in line 130.

We thank the reviewer for the suggestion, and have added the following sentence after line 130:

---

The reconstruction quality is defined as the log-likelihood of reconstructed neural data v___{recon} (i.e., __v that is first transformed to the low-dimensional h, and then back again to the high-dimensional __v___{recon}, see Methods 7.10.2).

---

Further, please note that Methods describes the definition in detail (Eq 18 of the submitted manuscript), although we agree with the reviewer that more detail was required in the Results section at line 130.

5) Line 165. If PCA is compared with cRBM, why other dimensionality reduction methods, such as k-means and non-negative matrix factorization, can not be compared in terms of the sparsity?

Please see answer to question 1 from R1 and Revision Plan #2.

6) Line 260, please provide minimum information about how the functional connectivity is defined based on assemblies discovered by cRBM.

We apologize if this was not clear. The first paragraph of this section (lines 248-259) of the submitted manuscript, provides the detail that the reviewer asks for, and we realize that the sentence of line 260 is better placed in the first paragraph, as it has come across as a very minimal explanation of how functional connectivity is defined.

We have now moved this sentence to the preceding paragraph, as well as specified the Method references (as suggested by this reviewer below), for additional clarity. We thank the reviewer for pointing out this sentence.

7) Some analysis of the hidden units population activities. Such as whether or not there are interesting low dimensional structure from figure 4A.

We thank the reviewer for their suggestion. In our manuscript we have used the cRBM model to create a low-dimensional (M=200) representation of zebrafish neural recordings (N=50,000). The richness of this model owes to possible overlaps between HUs/assemblies that can result in significant correlation in their activities. The latter is illustrated in Figure 4A-C: the activity of some HUs can be strongly correlated.

The reviewer’s suggestion is similar; to perform some form of dimensionality reduction on the low-dimensional HU activity shown in Fig 4. We have now added a PCA analysis to Figure 4 to quantify the degree of low-dimensional structure in the HU dynamics, and show the results in a new panel Figure 4D.

The following text has been added to the Results section:

---

These clusters of HUs with strongly correlated activity suggest that much of the HU variance could be captured using only a small number of variables. We quantified this by performing PCA on the HU dynamics, finding that indeed 52% of the variance was captured by the first 3 PCs, and 85% by the first 20 PCs (Figure 4D).

---

We believe that further visualization of these results, such as plotting the PC trajectories, would not further benefit the manuscript. The manuscript focuses on cRBM, and the assemblies/HUs it infers. Unlike PCA, these are not ranked/quantified by how much variance they explain individually, but rather they together ‘compose’ the entire system and explain its (co)variance (Figure 2). Breaking up a dominant activity mode (as found by PCA), such as the ARTR dynamics, into multiple HUs/assemblies, allows for some variation in activity of individual parts of the ARTR circuit (such as tail movement and eye movement generation), even though at most times the activity of these HUs is coordinated. We hope the reviewer agrees with our motivation to keep the manuscript focused on the nature of cRBM-inferred HUs.

8) Figure 4B right panel, how did the authors annotate the cluster manually? As certain assembly may overlap with several different brain regions, for example, figure 4D.

We thank the reviewer for this question, and we presume they meant to reference figure 3D as an example? For figure 4, as well as Figure 3, we used the ZBrain Atlas (Randlett et al., 2015) for definition of brain regions. This atlas presents a hierarchy of brain regions: for example, many brain regions are part of the rhombencephalon/hindbrain. This is what we used for midbrain/hindbrain/diencephalon. Further, many assemblies are solely confined to Optic Tectum (see Fig 3L), which we therefore used (split by hemisphere). Then, many brain regions are (partly) connected to the ARTR circuit, such as the example assembly of Figure 3D that the reviewer mentions. These we have all labeled as ARTR (left or right), though technically only part of their assembly is the ARTR. These two clusters therefore rather mean ‘ARTR-related’, in particular because their activity is locked to the rhythm of the ARTR (see Fig 4A). The final category is ‘miscellaneous’ (like Figure 3G).

However we agree that this wasn’t clear from the manuscript text, so we have changed the figure 4C caption to mention that ‘ARTR’ stands for ARTR-related assemblies, which we hope clarifies that ARTR-clustered assemblies can exist of multiple, disjoint groups of neurons, which relate to the ARTR circuit.

9) Better reference of the methods cited in the main text. The method part is quite long, it would be helpful to cite the section number when referring it in the main text.

We thank the reviewer for this helpful suggestion, we agree that it would benefit the manuscript to reference specific sections of the Methods. We have now changed all references to Methods to incorporate this.

10) Some discussion about the limitation of cRBM would be great.

We thank the reviewer for this suggestion, and have now included this. As Reviewer 1 had the same suggestion, we refer our answer to questions 2 and 3 from R1 for more detail.

Reviewer #3 (Significance (Required)):

This work provides a timely new technique to extract meaningful neural assemblies from large scale recordings. This study should be interested to both researchers doing either experiments and computation/theory. I am a computational neuroscientist.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Understanding the organization of neural population activities in the brain is one of the most important questions in neuroscience. Recent technique advance has enabled researchers to record a large number of neurons and some times the whole brain. Interpreting and extracting meaningful insights from such data sets are challenging. van der Pals \textit{et al} applied a generative model called compositional Restricted Boltzmann Machine (cRBM) to discover neuron assemblies from spontaneous activities of zebra fish brain. They found that neurons can be grouped into around 200 assemblies. Many of them have clear neurophysiological meaning, for example, they are anatomically localized and overlapped with known neural circuits. The authors also inferred a coarse-grained functional connectivity which is similar to known structural connectivity.

The structure of the paper is well organized, the conclusion seems well supported by their numerical results. While this study provides a compelling demonstration that cRBM can be used to uncover meaningful structures from large neural recordings, the following issues limit my enthusiasm.

Major:

1. The overall implication is not clear to me. Although the authors mentioned this briefly in the discussion. It is not clear what else do we learn from discovered assemblies beyond stating that they are consist with previous study. For example, the author could have more analysis of the assembly dynamics, such as whether there are low dimensional structure etc.
2. The learning algorithm of cRBM can be interpreted as matching certain statistics between the model and the experiment. For a general audience, it is not easy to understand $\langle h_{\mu} \rangle_{data}, \langle v_ih_{\mu}\rangle_{data}$. Since these are not directly calculated from experimental observed activities $v_i$, but rather the average is conditioned on the empirical distribution of $p(v_i)$. For example, the meaning of $\langle v_ih_{\mu}\rangle_{data}$ means \begin{equation} \langle v_ih_{\mu}\rangle_{data} = \frac{1}{l}\sum_{\mathbf{v}\in S} \mathbb{E}{p(\mathbf{h|v})}(v_ih{\mu}), \end{equation} where $S$ is the set of all observed neural activities: $S = {\mathbf{v}^1, \cdots, \mathbf{v}^l}$. The authors should explain this in the main text or method, since they are heavily loaded in figure 2.
3. As a modeling paper, it would be great to have some testable predictions.

Minor:

1. The assembly is defined by the neurons that are strongly connected with a given hidden unit. Thus, some neurons may enter different assemblies. A statistics of such overlap would be helpful. For example, a ven diagram in figure 1 that shows how many of them assigned to 1, 2, etc assemblies.
2. What does the link between hidden units in Figure 1B right panel mean?
3. A side-by-side comparison of neural activity predicted by model and the experimentally recorded activities would help the readers to appreciate the performance of the model. Such comparison can be done at both single neuron level or assembly level.
4. Definition of reconstruction quality in line 130.
5. Line 165. If PCA is compared with cRBM, why other dimentionality reduction methods, such as k-means and non-negative matrix factorization, can not be compared in terms of the sparsity?
6. Line 260, please provide minimum information about how the functional connectivity is defined based on assemblies discovered by cRBM.
7. Some analysis of the hidden units population activities. Such as whether or not there are interesting low dimensional structure from figure 4A.
8. Figure 4B right panel, how did the authors annotate the cluster manually? As certain assembly may overlap with several different brain regions, for example, figure 4D.
9. Better reference of the methods cited in the main text. The method part is quite long, it would be helpful to cite the section number when referring it in the main text.
10. Some discussion about the limitation of cRBM would be great.

Significance

This work provides a timely new technique to extract meaningful neural assemblies from large scale recordings. This study should be interested to both researchers doing either experiments and computation/theory. I am a computational neuroscientist.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In their manuscript, van der Plas et al. present a generative model of neuron-assembly interaction. The model is a restricted Boltzmann machine with its visible units corresponding to neurons and hidden units to neural assemblies. After fitting their model to whole-brain neural activity data from larval zebrafish, the authors demonstrate that their model is able to replicate several data statistics. In particular, it was able to replicate the pairwise correlations between neurons as well as assemblies that it was not trained on. Moreover, the model allows the authors to extract neural assemblies that govern the population activity and compose functional circuits and can be assigned to anatomical structures. Finally, the authors construct functional connectivity maps from their model that are then shown to correlate with established structural connectivity maps.

Overall, the authors present convincing evidence for their claims. Furthermore, the authors state that their code to train their restricted Boltzmann machine models is already available on GitHub and that the data underlying the results presented in this manuscript will be made publicly available upon publication, which will allow people to reproduce the results and apply the methods to their data.

One thing the authors could maybe discuss a bit more is the "right" parameter value M, especially since they used the optimal value of 200 found for one sample also for all the others. More specifically, how sensitive are the results to this value? And, what happens if one would successively increase that number, would the number of assemblies (in the sense of hidden units that strongly couple to some of the visible units) eventually saturate?

Moreover, regarding the presentation, I have a few minor suggestions and comments that the authors also might want to consider: - In Figure 6C, instead of logarithmic axes, it might be better to put the logarithmic connectivity on a linear axis. This way the axes can be directly related to the colour bars in Figures 6A and B. - In Equation (8), instead of $\Gamma_{\mu}(I)$ it should be $\Gamma_{\mu}(I_{\mu}(v))$. - In Section 7.0.5, it might make sense to have the subsection about the marginal distributions before the ones about the conditional distributions. The reason would be that if one wants to confirm Equation (8) one necessarily has to compute the marginal distribution in Equation (12) first. - In Line 647f, the operation the authors are referring to is strictly speaking not an L1-norm of the matrix block. It might be better to refer to that e.g. as a normalised L1-norm of the matrix block elements. - In Line 22, when mentioning dimensionality reduction methods to identify assemblies, it might make sense to also reference the work by Lopes-dos-Santos et al. (2013, J. Neurosci. Methods 220).

Significance

The work presented in this manuscript is very interesting for two reasons. First, it has long been suggested that assemblies are a fundamental part of neural activity and this work seems to support that by showing that one can generate realistic whole-brain population activity imposing underlying assembly dynamics. Second, in recent years much work has been devoted to developing methods to find and extract neural assemblies from data and this work and the modelling approach can also be seen as a new method to achieve that. As such, I believe this work is relevant for anyone interested in neural population activity and specifically neural assemblies and certainly merits publication.

Regarding my field of expertise, I used to work on data analysis of neural population activity and in particular on the question of how one can extract neural assemblies from data. I have to say that I have not much experience with fitting statistical models to data, so I can't provide any in-depth comments on that part of the work, although what has been done seems plausible.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

In the present manuscript, van der Plas et al. compellingly illustrated a novel technique for engendering a whole-brain functional connectivity map from single-unit activities sampled through a large-scale neuroimaging technique. With some clever tweaks to the restricted Boltzmann Machine, the cRBM network is able to learn a low-dimensional representation of population activities, without relying on constrained priors found in some traditional methods. Notably, using some 200 hidden layer neurons, the employed model was able to capture the dynamics of over 40,000 simultaneously imaged neurons with a high degree of accuracy. The extracted features both illustrate the anatomical topography/connectivities and capture the temporal dynamics in the evolution of brain states. The illustrated technique has the potential for wide-spread applications spanning diverse recording techniques and animal species. Furthermore, the prospectives of modeling whole-brain network dynamics in 'neural trajectory' space and of generating artificial data in silico make for very enticing reasons to adopt cRBM.

Major comments:

Line 164. The authors claim that conventional methods "such as k-means, PCA and non-negative matrix factorization" cannot be quantitatively assessed for quality on the basis that they are unable to generate new artificial data. Though partly true, in most neuroscience applications, this is hardly cause for concern. Most dimensionality reduction methods (with few exceptions such as t-sne) allow new data points to be embedded into the reduced space. As such, quality of encoding can be assessed by cross-validation much in the same way as the authors described, and quantified using traditional metrics such as percentage explained variance. The authors should directly compare the performance of their proposed model against that of NNMF and variational auto-encoders. Doing so would offer a more compelling argument for the advantage of their proposed method over more widely-used methods in neuroscience applications. Furthermore, a direct comparison with rastermap, developed by Stringer lab at Janelia (https://github.com/MouseLand/rastermap), would be a nice addition. This method presents itself as a direct competitor to cRBM. Additionally, the use of GLM doesn't do complete justice to the comparison point used, since a smaller fraction of data were used for calculating performance using GLM, understandably due to its computationally intensive nature.

Line 26. The authors describe their model architecture as a formalization of cell assemblies. Cell assemblies, as originally formulated by Hebb, pertains to a set of neurons whose connectivity matrix is neither necessarily complete nor symmetric. Critically, in the physiological brain, the interactions between the individual neurons that are part of an assembly would occur over multiple orders of dependencies. In a restricted Boltzmann machine, neurons are not connected within the same layer. Instead, visible layer neurons are grouped into "assemblies" indirectly via a shared connection with a hidden layer neuron. Furthermore, a symmetrical weight matrix connects the bipartite graph, where no recurrent connectivities are made. As such, the proposed model still only elaborates symmetric connections pertaining to first-order interactions (as illustrated in Figure 4C). Such a network may not be likened with the concept of cell assemblies. The authors should refrain from detailing this analogy (of which there are multiple instances of throughout the text). It is true that many authors today refer to cell assemblies as any set of temporally-correlated neurons. However, saying "something could be a cell assembly" is not the same as saying "something is a cell assembly". How about sticking with cRBM-based cell assemblies (as used in section 2.3) and defining it beforehand?

I would strongly recommend adding a paragraph discussing the limitation of using the cRBM, things future researchers need to keep in mind before using this method. One such recommendation is moving the runtime-related discussion for cRBM, i.e. 8-12 hrs using 16 CPU from Methods to Discussion, since it's relevant for an algorithm like this. Additionally, a statement mentioning how this runtime will increase with the length of recordings and/or with the number of neurons might be helpful. What if the recordings were an hour-long rather than 25mins. This would help readers decide if they can easily use a method like this.

Line 515. A core feature of the proposed compositional RBM is the addition of a soft sparsity penalty over the weight matrix in the likelihood function. The authors claim that "directed graphical models" are limited by the a priori constraints that they impose on the data structure. Meanwhile, a more accurate statistical solution can be obtained using a RBM-based model, as outlined by the maximum entropy principle. The problem with this argument is that the maximum entropy principle no longer applies to the proposed model with the addition of the penalty term. In fact, the lambda regularization term, which was estimated from a set of data statistics motivated by the experimenter's research goals (Figure S1), serves to constrict the prior probability. Moreover, in Figure S1F, we clearly see that reconstruction quality suffers with a higher penalty, suggesting that the principle had indeed been violated. That being said, RBMs are notoriously hard to train, possibly due to the unconstrained nature of the optimization. I believe that cRBM can help bring RBM into wider practical applications. The authors could test their model on a few values of the free parameter and report this as a supplementary. I believe that different parameters of lambda could elaborate on different anatomical clusters and temporal dynamics. Readers who would like to implement this method for their own analysis would also benefit tremendously from an understanding of the effects of lambda on the interpretation of their data. Item (1) on line 35 (and other instances throughout the text) should be corrected to reflect that cRBM replaces the hard constraints found in many popular methods with a soft penalty term, which allows for more accurate statistical models to be obtained.

Minor comments:

From a neuroscience point of view, it might be interesting to show what results are achieved using different values of M (say 100 or 300), rather than M=200, while still maintaining the compositional phase. Is there any similarity between the cRBM-based cell assemblies generated at different values of M? Is there a higher chance of capturing certain dynamics either functional or structural using cRBM? For example, did certain cRBM-based cell assemblies pop up more frequently than others at all values of M (100,200,300)?

The authors have mentioned that this approach can be readily applied to data obtained in other animal models and using different recording techniques. It might be nice to see a demonstration of that.

Line 237. The justification for employing a dReLU transfer function as opposed to ReLU is unclear, at least within the context of neurobiology. Given that this gives rise to a bimodal distribution for the activity of HUs, the rationale should be clearly outlined to facilitate interpretability.

Significance

van der Plas et al. highlighted a novel dimensionality reduction technique that can be used with success for discerning functional connectivities in large-scale single-unit recordings. The proposed model belongs to a large collection of dimensionality reduction techniques (for review, Cunningham, J., Yu, B. Dimensionality reduction for large-scale neural recordings. Nat Neurosci 17, 1500-1509 (2014). https://doi.org/10.1038/nn.3776; Paninski, L., & Cunningham, J. P. (2018). Neural data science: accelerating the experiment-analysis-theory cycle in large-scale neuroscience. Current opinion in neurobiology, 50, 232-241.). The authors themselves highlighted some of the key methods, such as PCA, ICA, NNMF, variational auto-encoders, etc. The proposed cRBM model has also been published a few times by the same authors in previous works, although specifically pertaining to protein sequences. The use of RBM-like methods in uncovering functional connectivities is not novel either (see Hjelm RD, Calhoun VD, Salakhutdinov R, Allen EA, Adali T, Plis SM. Restricted Boltzmann machines for neuroimaging: an application in identifying intrinsic networks. Neuroimage. 2014 Aug 1;96:245-60. doi: 10.1016/j.neuroimage.2014.03.048.). However, given that the authors make a substantial improvement on the RBM network and have demonstrated the value of their model using physiological data, I believe that this paper would present itself as an attractive alternative to all readers who are seeking better solutions to interpret their data. However, as I mentioned in my comments, I would like to see more definitive evidence that the proposed solution has a serious advantage over other equivalent methods.

Reviewer's expertise:

This review was conducted jointly by three researchers whose combined expertise includes single-unit electrophysiology and two-photon calcium imaging, using which our lab studies the neurobiology of learning and memory and spatial navigation. We also have extensive experience in computational neuroscience, artificial neural network models, and machine learning methods for the analysis of neurobiological data. We are however limited in our knowledge of mathematics and engineering principles. Therefore, our combined expertise is insufficient to evaluate the correctness of the mathematical developments.

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

Learn more at Review Commons

---

Reply to the reviewers

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

In this paper, Staneva et al describe a novel complex found at RNA PolII promoters that they term the SPARC. The manuscript focuses on defining the core components of the complex and the pivotal role of SET27 in defining its function, and role in PolII transcription. This manuscript is a logical follow on from an initial paper (Staneva et al, 2021) by the same authors where they systematically analyzed chromatin factors, and their role in both transcription start and termination. What is also very clear, is that this complex is one made of histone readers and writers which suggests its function is to change the chromatin structure around a PolII promoters. The authors show that this complex is necessary for the correct positioning of PolII and directionality of transcription.

This was a well-designed study and well written and clear manuscript that provides fascinating insight transcription control in bloodstream form parasites.

I have no major comments only a few minor ones.

1) Localisation of the different SPARC components appears to be either nuclear or nuclear and cytoplasmic. - Both SET27 and CRD1 show a nuclear and cytoplasmic localisation in the bloodstream form IFA (Supplementary Fig 1B), but only a nuclear localisation procyclic form.

Did the authors attempt C terminally tagging SET27, CRD1 to see if this resulted in a change in the pattern?

We have not tagged either protein at the C terminus, however SET27 (Tb927.9.13470) has been tagged both N- and C-terminally in procyclic form (PF) cells as part of the TrypTag project (http://tryptag.org). In both cases, SET27 localized to the nucleus, suggesting that the differences in localization we observe for SET27 depend on the life cycle stage, and not on the position of the tag. One caveat is that in the TrypTag project proteins are tagged with mNeonGreen whereas in our study proteins were tagged with YFP. Based on our images, CRD1 appears to be predominantly nuclear in both bloodstream form (BF) and PF parasites. CRD1 (Tb927.7.4540) has been tagged only N-terminally in PF cells as part of the TrypTag project where it has also been classified as mostly nuclear with only 10% of cells showing cytoplasmic localization for CRD1.

We are well aware that tags can alter the behaviour of a protein. Absolute confirmation of location will require the generation of antibodies that detect untagged proteins. However, this is a longer-term undertaking. We have added the following statement to the Results section to address the point raised:

“We tagged the proteins on their N termini to preserve 3′ UTR sequences involved in regulating mRNA stability (Clayton, 2019). We note, however, that the presence of the YFP tag and/or its position (N- or C-terminal) might affect protein expression and localization patterns”.

• The point is made that JBP2 shows a 'distinct cytoplasmic localisation' in PF cells. by this logic, the SET27 localisation in BF is also distinctly cytoplasmic and a nuclear enrichment is not clear.

Indeed the reviewer is correct - we have inadvertently over accentuated the significance of this difference in the text. We had emphasized the predominantly cytoplasmic localization of JBP2 in PF trypanosomes as potentially related to its weaker association with other (predominantly nuclear) SPARC components in the mass spectrometry experiments. The presence of SET27 in the nuclei of both BF and PF cells is confirmed by a positive ChIP signal. We have revised the manuscript text by changing “distinct cytoplasmic” to “predominantly cytoplasmic” to describe JBP2 localization in PF cells. We hope that this resolves the issue.

• Why would the localisation pattern change between life cycle stages? Surely PolII transcription should remain the same?

Although our analysis suggests that there may be some shift in SET27 and JBP2 localization between BF and PF stages, sufficient amounts of these proteins may be present in the nucleus for proper SPARC assembly and RNAPII transcription regulation in both life cycle forms. The proportion of SET27 and JBP2 proteins that localizes to the cytoplasm may have functions unrelated to transcription.

2) Several of the images in Supplementary Fig 1B seem to show foci in the nucleus (CSD1, PWWP1, CRD1). Do you see foci throughout the cell cycle or just in G1/S phase cells as shown here?

We have not systematically investigated protein localization at different cell cycle stages, so we do not have microscopy images for all proteins at all stages of the cell cycle. However, the images we did collect suggest the punctate pattern is preserved for CRD1 in the G2 phase in both BF and PF cells (see below) as we showed in Supplemental Figure S1B for cells with 1 kinetoplast and 1 nucleus (G1/S phase cells). The significance of these puncta remains to be determined.

3) In Figure 6, what does 'TE' stand for?

TE denotes transposable elements. We have added this to the figure legend.

4) The authors show this interesting link between SPARC complex and subtelomeric VSG gene silencing. -In the CRD1 ChIP or RBP1 ChIP, are there any other peaks in telomere adjacent regions in the WT cells similar to that seen on chromosome 9A? And does the sequence at this point resemble a PolII promoter?

Apart from peaks located on Chromosome 9_3A, there are other CRD1 and RPB1 ChIP peaks in chromosomal regions adjacent to telomeres in WT cells. We observed broadening of RPB1 distribution in these regions upon SET27 deletion, similar to what we show for Chromosome 9_3A. In particular, wider RPB1 distribution on Chromosome 8_5A coincides with upregulation of 10 VSG transcripts. These two loci explain most of the differentially expessed genes (DEGs) detected, but other subtelomeric regions show a similar pattern. We have added the following statement to the Results section to highlight that the phenotype shown for Chromosome 9_3A is not unique:

“We also observed a similar phenotype at other subtelomeric regions, such as Chromosome 8_5A where 10 VSGs and a gene encoding a hypothetical protein were upregulated upon SET27 deletion (Supplemental Table S3)”.

Cordon-Obras et al. (2022) have recently defined key sequence elements present at one RNAPII promoter. We searched for similar sequence motifs but failed to identify them as underlying CRD1 and RPB1 ChIP peaks, highlighting the likely sequence heterogeneity amongst trypanosome RNAPII promoters. To address this point, we have added the following sentence to the Discussion:

“Sequence-specific elements have recently been found to drive RNAPII transcription from a T. brucei promoter (Cordon-Obras et al., 2022), however, we were unable to identify similar motifs underlying CRD1 or RPB1 ChIP-seq peaks, suggesting that T. brucei promoters are perhaps heterogeneous in composition”.

-In the FLAG-CRD1 IP (Figure 3B), the VSG's seen here are not represented (as far as I can tell) in Figure 6B and C. If my reading is correct could, is this a difference in the FC cut off for what is significant in these experiments?

The VSGs detected in the FLAG-CRD1 IP from set27D/D cells are indeed different from the ones shown in Figure 6 (even after setting the same fold change cutoffs). We have highlighted this by adding the following statement to the Results section: “Gene ontology analysis of the upregulated mRNA set revealed strong enrichment for normally silent VSG genes (Figure 6B-D) which were distinct from the VSG proteins detected in the FLAG-CRD1 immunoprecipitations from set27D/D cells (Figure 3B)”.

The VSGs in the mass spectrometry experiments likely represent unspecific interactors of FLAG-CRD1. To clarify this, we have added the following statement to the Results section: ”Instead, several VSG proteins were detected as being associated with FLAG-CRD1 in set27D/D cells, though it is likely that these represent unspecific interactions”.

Reviewer #1 (Significance (Required)):

Trypanosomes are unusual in the way that they transcribe protein coding genes. Recent advances have defined the chromatin composition at the TSS and TTS, and the recent publication of a PolII promoter sequence(s) further adds to our understanding of how transcription here is regulated. Defining the SPARC complex now add to this understanding and highlights the role of potential histone readers and writers. I think that this will be of interest to the kinetoplastid community especially those working on control of gene expression.

Our lab studies gene expression and antigenic variation in T. brucei.

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

In this manuscript, the authors identify a six-membered chromatin-associated protein complex termed SPARC that localizes to Transcription Start Regions (TSRs) and co-localizes with and (directly or indirectly) interacts with RNA polymerase II subunits. Careful deletion studies of one of its components, SET27, convincingly show the functional importance of this complex for the genomic localization, accuracy, and directionality of transcription initiation. Overall, the experiments are well and logically designed and executed, the results are well presented, and the manuscript is easy to read.

There are a few minor points that would benefit from clarification and/or from a more detailed discussion:

1) The concomitant expression of many VSGs (37) in a SET27 deletion strain is remarkable and has important implications for their normally monoallelic expression. It is well established that VSG expression in wild-type T. brucei can only occur from one of ~15 subtelomeric bloodstream expression sites, which include the ESAGs. This result implies that VSG genes are also transcribed from "archival VSG sites" in the genome, not only from expression sites. Are there VSGs from the silent BESs among the upregulated VSGs? Is there precedence in the literature for the expression of VSGs from chromosomal regions besides the subtelomeric expression sites?

Our analysis of differentially expressed genes (DEGs) revealed that 43 VSG genes (37 of which are subtelomeric) and 2 ESAG genes are upregulated in the absence of SET27. Both ESAGs but none of the upregulated VSGs in set27D/D cells are annotated as located in BES regions. While it is possible that recombination events have resulted in gene rearrangements between the reference strain and our laboratory’s strain, at least some of the upregulated VSGs are likely to be transcribed from non-BES archival sites. VSG transcript upregulation from non-BES regions was also recently described by López-Escobar et al (2022).

We note that the upregulated mRNAs in set27D/D are still relatively lowly expressed (Figure 6C). This is presumably insufficient to coat the surface of T. brucei, and expression from BES sites instead may be required to achieve this. We have revised the manuscript Discussion section to make these points more clear:

“Bloodstream form trypanosomes normally express only a single VSG gene from 1 of ~15 telomere-adjacent bloodstream expression sites (BESs). In contrast, in set27D/D cells we detected upregulation of 43 VSG transcripts, none of which were annotated as located in BES regions. Recently, López-Escobar et al (2022) have also observed VSG mRNA upregulation from non-BES locations, suggesting that VSGs might sometimes be transcribed from other regions of the genome. However, the VSG transcripts we detect as upregulated in set27D/D were relatively lowly expressed (Figure 6C) and may not be translated to protein or be translated at low levels compared to a VSG transcribed from a BES site”.

2) The role of SPARC in defining transcription initiation is compelling. It's less clear to the reviewer if the observed transcriptional silencing within subtelomeric regions can also ascribed to SPARC. Have the authors considered the possibility that some components of the SPARC may be shared by other chromatin complexes, which could be responsible for the transcriptional activation of silent genes in SET27 deletion mutants?

We cannot rule out indirect effects through the participation of some SPARC components in other complexes operating independently of SPARC. Indeed, the transcriptional defect within the main body of chromosomes appears to be somewhat different from that observed at subtelomeric regions, particularly with respect to distance from SPARC. We have added a statement in the Discussion section to highlight the possibility raised by the reviewer:

“However, an alternative possibility is that transcriptional repression in subtelomeric regions is mediated by different protein complexes which share some of their subunits with SPARC, or whose activity is influenced by it”.

3) The authors mention that the observed interaction of FLAG-CRD1 with VSGs in the immunoprecipitations (Fig. 3B) is evidence for the actual expression of normally silent VSGs on the protein level. This is true, but it should be spelled out that this interaction is nevertheless likely an artifact, at least the physiological relevance of these interactions is questionable.

We agree that these are likely background associations and have added the following statement to the Results section to clarify this point:

“Instead, several VSG proteins were detected as associated with FLAG-CRD1 in set27D/D cells, though it is likely that these represent unspecific interactions”.

To avoid unnecessary confusion we have also removed the following sentence from the revised Discussion:

“The interactions of FLAG-CRD1 with VSGs in the affinity selections from set27Δ/Δ cells indicate that some of the normally silent VSG genes are also translated into proteins in the absence of SET27”.

4) "ophistokont" is misspelled in the introduction

Thanks for noticing. We have corrected it to “Opisthokonta”.

Reviewer #2 (Significance (Required)):

The manuscript by Staneva et al. addresses the fundamental regulatory mechanism of gene transcription in the protozoan parasite Trypanosoma brucei, a highly divergent eukaryotic organism that is renowned for unusual features and mechanisms in gene regulation, metabolism, and other cellular processes. While post-transcriptional regulation is prevalent and relatively well established in T. brucei, much less is known about the mechanism of transcription initiation and transcriptional control, in part due to the general paucity of well-defined conventional promoter regions in this organism (only very few have been identified thus far). In this context, the work by Staneva et al. is highly significant and represents an important contribution to the field of gene regulation and chromatin biology in T. brucei and other related kinetoplastid parasites.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, the authors identify a six-membered chromatin-associated protein complex termed SPARC that localizes to Transcription Start Regions (TSRs) and co-localizes with and (directly or indirectly) interacts with RNA polymerase II subunits. Careful deletion studies of one of its components, SET27, convincingly show the functional importance of this complex for the genomic localization, accuracy, and directionality of transcription initiation. Overall, the experiments are well and logically designed and executed, the results are well presented, and the manuscript is easy to read.

There are a few minor points that would benefit from clarification and/or from a more detailed discussion:

1. The concomitant expression of many VSGs (37) in a SET27 deletion strain is remarkable and has important implications for their normally monoallelic expression. It is well established that VSG expression in wild-type T. brucei can only occur from one of ~15 subtelomeric bloodstream expression sites, which include the ESAGs. This result implies that VSG genes are also transcribed from "archival VSG sites" in the genome, not only from expression sites. Are there VSGs from the silent BESs among the upregulated VSGs? Is there precedence in the literature for the expression of VSGs from chromosomal regions besides the subtelomeric expression sites?
2. The role of SPARC in defining transcription initiation is compelling. It's less clear to the reviewer if the observed transcriptional silencing within subtelomeric regions can also ascribed to SPARC. Have the authors considered the possibility that some components of the SPARC may be shared by other chromatin complexes, which could be responsible for the transcriptional activation of silent genes in SET27 deletion mutants?
3. The authors mention that the observed interaction of FLAG-CRD1 with VSGs in the immunoprecipitations (Fig. 3B) is evidence for the actual expression of normally silent VSGs on the protein level. This is true, but it should be spelled out that this interaction is nevertheless likely an artifact, at least the physiological relevance of these interactions is questionable.
4. "ophistokont" is misspelled in the introduction

Significance

The manuscript by Staneva et al. addresses the fundamental regulatory mechanism of gene transcription in the protozoan parasite Trypanosoma brucei, a highly divergent eukaryotic organism that is renowned for unusual features and mechanisms in gene regulation, metabolism, and other cellular processes. While post-transcriptional regulation is prevalent and relatively well established in T. brucei, much less is known about the mechanism of transcription initiation and transcriptional control, in part due to the general paucity of well-defined conventional promoter regions in this organism (only very few have been identified thus far). In this context, the work by Staneva et al. is highly significant and represents an important contribution to the field of gene regulation and chromatin biology in T. brucei and other related kinetoplastid parasites.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this paper, Staneva et al describe a novel complex found at RNA PolII promoters that they term the SPARC. The manuscript focuses on defining the core components of the complex and the pivotal role of SET27 in defining its function, and role in PolII transcription. This manuscript is a logical follow on from an initial paper (Staneva et al, 2021) by the same authors where they systematically analyzed chromatin factors, and their role in both transcription start and termination. What is also very clear, is that this complex is one made of histone readers and writers which suggests its function is to change the chromatin structure around a PolII promoters. The authors show that this complex is necessary for the correct positioning of PolII and directionality of transcription.

This was a well-designed study and well written and clear manuscript that provides fascinating insight transcription control in bloodstream form parasites.

I have no major comments only a few minor ones.

1. Localisation of the different SPARC components appears to be either nuclear or nuclear and cytoplasmic.
   • Both SET27 and CRD1 show a nuclear and cytoplasmic localisation in the bloodstream form IFA (Supplementary Fig 1B), but only a nuclear localisation procyclic form. Did the authors attempt C terminally tagging SET27, CRD1 to see if this resulted in a change in the pattern?
   • The point is made that JBP2 shows a 'distinct cytoplasmic localisation' in PF cells. by this logic, the SET27 localisation in BF is also distinctly cytoplasmic and a nuclear enrichment is not clear.
   • Why would the localisation pattern change between life cycle stages? Surely PolII transcription should remain the same?
2. Several of the images in Supplementary Fig 1B seem to show foci in the nucleus (CSD1, PWWP1, CRD1). Do you see foci throughout the cell cycle or just in G1/S phase cells as shown here?
3. In Figure 6, what does 'TE' stand for?
4. The authors show this interesting link between SPARC complex and subtelomeric VSG gene silencing.
   • In the CRD1 ChIP or RBP1 ChIP, are there any other peaks in telomere adjacent regions in the WT cells similar to that seen on chromosome 9A? And does the sequence at this point resemble a PolII promoter?
   • In the FLAG-CRD1 IP (Figure 3B), the VSG's seen here are not represented (as far as I can tell) in Figure 6B and C. If my reading is correct could, is this a difference in the FC cut off for what is significant in these experiments?

Significance

Trypanosomes are unusual in the way that they transcribe protein coding genes. Recent advances have defined the chromatin composition at the TSS and TTS, and the recent publication of a PolII promoter sequence(s) further adds to our understanding of how transcription here is regulated. Defining the SPARC complex now add to this understanding and highlights the role of potential histone readers and writers. I think that this will be of interest to the kinetoplastid community especially those working on control of gene expression.

Our lab studies gene expression and antigenic variation in T. brucei.

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

Learn more at Review Commons

---

Reply to the reviewers

Reviewers 1 and 2 are very positive about our manuscript, while reviewer 3 is surprisingly critical.

However, except for the first observation, most of reviewer 3´s comments are based on incorrect interpretations of our results.

We have integrated the useful comment into our revised version, and we will discuss in the following sections why reviewer 3’s remaining criticisms should be disregarded.

Reviewer 1:

Reviewer 1 has only minor suggestions and is satisfied that we prove convincingly our claims. The reviewer also finds our results reinforce our previously proposed hypothesis that the glands and the trachea evolved from common metamerically repeated ancient primordia.

We have introduced the following changes to the text to accommodate Reviewer’s 1 minor suggestions.

Main suggestion: Insert a paragraph in discussion explaining the relevance of new insights to more basal insects that do not form a ring gland.

RESPONSE:We have introduced at the end of Discussion the following paragraph:

“Our analysis of snail activation in the CA and PG shows that these glands and the trachea share similar upstream regulators, reinforcing the hypothesis that both diverged from an ancient segmentally repeated organ. In Drosophila melanogaster the CA and the PG primordia experiment a very active migration after which they fuse to the corpora cardiaca forming the ring gland (Sanchez-Higueras and Hombria, 2016). This differs from more basal insects where the CA fuses to the corpora cardiaca but not to the PG, and from the Crustacea where the three equivalent glands are independent of each other (Chang and O'Connor, 1977; Laufer et al., 1987; Nijhout, 1994; Wigglesworth, 1954). As the mechanisms we here describe relate to the early specification of the glandular primordia in Drosophila, it will be interesting to investigate if the equivalent genes are also involved in the endocrine gland specification of more distant arthropods”.

Additional comment 1: Introduction, pg 3, a paragraph starting with "In comparison to the extensive knowledge we have of ..." - consider omitting or greatly shortening, this text breaks a flow as it is focused on tracheal development. I understand the authors' logic, but this information distracts from the main focus on CA and PG. RESPONSE:We agree that the trachea description paragraph breaks the flow of the introduction to gland development. As suggested by the reviewer, we have deleted most of the descriptive text on trachea development but left all the references so that interested readers can find the information.

Additional comment 2: Beginning of discussion, pg 11: - change 2nd sentence to: " Our results indicate that the HH and the Wnt pathways act indirectly to negatively regulate the spatial activation ..." - the following sentence, starting with "Engrailed activation off hh transcription ...." is way too long and hard to follow, consider breaking into two sentences. RESPONSE:We have changed both sentences as suggested

Additional comment 3: In Fig 4B, mx and lb segments should be labeled so this panel is consistent with labeling in 4A. RESPONSE:We have changed Fig.4B labels to be consistent with 4A

Additional comment 4: In Fig 6, reduce a font size for labels on right-hand side (A1, A2, A1+A2 proximal, etc), so that they are visualy distinct from panel labels on left-hand side (A, B, C,..).

RESPONSE:We have changed Fig.6 Font size as suggested

Reviewer 2

The reviewer is positive and agrees that the results we present in “this paper add to our understanding of how the CA and PG primordia are specified and highlights important similarities with the specification of the tracheal primordia”. The reviewer’s comments focus specially on the activation vs. maintenance of sna.

Specific Comment a): Referring to Fig 1G-J, the reviewer says: It is not clear to me from either this figure or from the text whether the initial pattern of expression of the sna-rg reporter in stage 11 embryos is WT and then disappears at stage 12, or whether it is always defective. In trying to understand the activation process, I think it would be important to know for sure whether rg enhancer activity during the initiation phase in stage 11 is WT or not.

RESPONSE: As suggested by the reviewer, we have included st11 embryos in Fig. 1 as panels G,J which illustrate that early sna-rg activation occurs normally in snaΔrgR2embryos prior to apoptosis kicking in. To make space for these images, we have taken out the st12 embryos that we had in our previous submitted version. This does not affect the manuscript’s message, as st12 phenotypes are similar to those at st13 which are presented in Fig. 1H,J.

Moreover, in this revised version, the embryos in Fig. 1G-J have also been double stained with the apoptosis marker DCP1 to highlight the cell death observed in the gland primordia of snaΔrgR2 embryos (Fig. 1G’-J’).

Specific Comment b) The authors argue that the rg deletion removes the only region driving sna expression in CA/PG. I'm not convinced that necessity necessarily implies sufficiency with respect to the requirements for rescue. While the sna-rg reporter is expressed in a pattern that seems to mimic the endogenous gene, do we know that a rg-sna transgene would fully rescue the rg deletion mutant?

RESPONSE: In our previous paper (Sanchez-Higueras 2014) we presented evidence that in sna null embryos, a Snail BAC gene lacking the sna-rg CRM can fully rescue the mesoderm phenotypes but not the ring gland ones. This proved that in the BAC transgene there was no shadow CRM capable of rescuing the gland formation in the absence of sna-rg. In the current paper we show that deleting the endogenous sna-rg CRM in the sna locus results in the absence of sna transcription from the gland primordia.

Making a sna-rg- construct expressing sna to test if this rescues the snaΔrgR2 homozygous mutants could be done, but it will delay this publication without adding much to the paper: we already know that sna-rg is sufficient to drive activation in all the CA and the PG cells (Sanchez-Higueras 2014 Fig 2J-M) and it would be expected to rescue the gland formation in snaΔrgR2 homozygous mutants.

Having said that, we have changed the wording in the manuscript to one that may be acceptable to the reviewer.

Instead of:

“These results prove that snaΔrgR2 deletes the only regulatory region driving sna expression in the CA and PG gland primordia…”

We now say:

“These results prove that the snaΔrgR2 deletes mutation inactivates the only regulatory region driving sna expression in the CA and PG gland primordia…”

Specific Comment c) is Sna required for maintaining sna expression?

RESPONSE:This experiment is relevant to the maintenance mechanism of sna expression in the ring gland, and not to its activation which is the main focus of this paper.

The search for the maintenance mechanisms is currently been followed in the laboratory and we prefer not deal with it in this paper. Providing a negative answer to this question would not be satisfactory, as we would need to search for the factors controlling sna’s maintenance.

Specific comment d) The authors show that there is an expansion in the number of sna-rg reporter expressing cells along the AP axis when upd is ectopically expressed using a sal-Gal4 driver. Though not mentioned in the text at this juncture, sal is expressed in the PG primordia, while seven-up (svp) is expressed in the CA primordia. I assume that the upd induced expansion is only observed for the PG primorida (LB) and not the CA primordia (Mx)-at least this is what the figure looks like. (…) How about svp driven upd-assuming there is a svp-Gal4 driver-does it cause an expansion of Ca but not PG.

RESPONSE: As the reviewer has noticed, there is a stronger expansion of sna-rg-GFP expression in the labial segment than in the maxillary segment. This is not due to the use of the sal-Gal4 line. We see the same effect with arm-Gal4 which drives similar expression on the maxilla and the labium. To illustrate this point, we have included two new panels (Fig.5D-E) where the ectopic expression of Upd has been induced with arm-Gal4. These embryos have been stained with anti-Sal to label the PG. This experiment shows clearly that the PG has expanded much more than the CA.

There are several reasons why expansion of the glands could be more efficient in the labium than in the maxilla. One possible reason is the temporal response to Upd activation. Upd induction by the arm-Gal4 and sal-Gal4 lines may occur after the cells in the maxilla are no longer capable of activating sna-rg but still capable of activating it in the labium. This temporal hypothesis is based on our results showing that the CA expresses more transiently the upd gene and that STAT activation lasts for longer in the labium than in the maxilla (Fig. 4A-D)].

A second possibility, that we favour, is the existence of dorso-ventral repressor genes modulating sna-rg expression intrasegmentally. Some of our results point towards the sna-rg CRM receiving repressor inputs that modulate intrasegmental spatial expression in the dorso-vental axis. When we delete the A2 distal region of the sna-rg enhancer, its expression in the labium expands ventrally (Fig. 6E,G and Sup.Fig. 4D). If a similar repressor was also modulating sna-rg in the maxilla it could be blocking its expansion. However, at this stage we have no solid data to support any of these hypotheses. As explained before for the maintenance mechanisms of sna-rg expression, our ongoing work aims to isolate and characterize further elements controlling the ring gland gene network, including these negative regulators.

In the revised manuscript we now describe the different effects of Upd ectopic activation on the expression of sna-rg in the maxilla and the labium (underlined text is new to this revised version):

“To test if generalised Upd expression in the maxilla and labium can activate sna-rg expression independently of other upstream positive or negative inputs, we induced UAS-upd with either the sal-Gal4 or the arm-Gal4 lines. We observe that, these embryos have expanded sna-rg expression along the antero-posterior axis in the maxillary and labial segments (Fig. 5C). Analysis of Sal expression, which labels the PG primordium (Sanchez-Higueras et al., 2014), shows that Upd ectopic expression induces a moderate expansion of the CA primordia while resulting a much larger increase of the PG primordium (Fig. 5D-E). This expansion occurs mostly in the anterior and posterior axis from cells where the Hh and the Wnt pathways are normally blocking sna-rg expression, while expansion is less noticeable in the dorso-ventral axis. This indicates that most of the antero-posterior intrasegmental inputs provided by the segment polarity genes converge on Upd transcription but that the dorso-ventral information is registered downstream of Upd.”

The differential response of sna-rg to Upd activation in the maxillary and labial segments is also mentioned in Fig. 5 legend. (see Continuation comment d).

* Continuation comment d) “It looks to me also like the vvl domain is expanding as well. This information should be clarified.*”

RESPONSE: Yes, ectopic upd expression also expands vvl1+2 expression. We have previously published that vvl1+2 is a direct target of JAK/STAT signalling in the trachea (Sanchez-Higueras 2019 and Sotillos et al. 2010 Dev.Biol). Although vvl1+2 expands dorsally in the Mx, those cells do not activate sna-rg dorsally. The ventral restriction of sna-rg in the maxilla is controlled by Dfd while in the labium its dorsal expression depends on Scr. We explain this in Fig.5’s figure legend where we now say (underlined text is new to this revised version):

(C) Ectopic Upd expression driven with sal-Gal4 induces ectopic sna-rg and vvl1+2 expression in the gnathal segments, which for sna-rg is more pronounced in the labium than in the maxilla. Note that in the maxillary segment Upd can induce ectopic dorsal vvl1+2 but not sna-rg expression, this is expected as Dfd only induces sna-rg ventrally in the maxilla. (D-E) sna-rg-GFP embryos stained with anti-GFP (green) and anti-Sal (red). In control embryos (D) Sal labels the PG primordium but not the CA. In arm-Gal4 embryos ectopically expressing Upd, the PG is more expanded than the CA as shown by number of cells co-expressing Sal and GFP.

Specific Comment e) The authors note a difference between CA and PG in the requirement for STAT binding sites in the enhancers. Is that related to the fact that svp is expressed in CA and sal is expressed in PG? Would driving svp expression using the sal-Gal4 driver maintain sna-rg expression.

RESPONSE: During our preliminary ongoing experiments on sna maintenance mechanisms we looked in svp mutants and did not notice a change in sna-rg expression, thus it is unlikely that Svp is responsible for the difference. As said above, we continue looking for genes involved in gland formation. Sal could be involved in the maintenance of sna in the PG, but as Sal is expressed in the maxilla and labial segments before gland formation, it is difficult to disentangle if Sal is required for sna activation or maintenance (or both).

Specific Comment f) Do svp or sal have a role in initiating sna expression when upd is present or maintaining sna expression after upd disappears? Presumably there is already published data that would answer these questions.

RESPONSE: As explained above we did not find any effect of svp on activation of sna-rg, however we find that in sal mutants the labium does not express sna-rg. This shows that sal is likely to be another positive input. As in sal mutants both trh and Ubx become ectopically expressed in the Lb (Casanova1989 Roux's archives of developmental biology 198: 137-140; Castelli-Gair 1998 IJDB42:437-444) we have done the experiment in sal trh double mutants and in sal Ubx,abdA,Abd-B mutants. In both cases we still see a failure of sna activation in the Lb reinforcing the idea that Sal is an additional positive input. However, we prefer not to add the sal experiments as they would complicate the paper which currently focuses on the similar requirement of the Wnt, Hh and JAK/STAT signalling pathways.

Reviewer 3

Reviewer is very critical. We accept some of the points raised and have modified the manuscript accordingly. However, as we detail below, the most serious criticisms are incorrect and do not affect the conclusions reached by our work.

We agree with the following comment:

“In the Dfd Scr double mutant, both the CA and PG expression of the snail-rg-GFP reporter is still there - admittedly, the gland cells look abnormal at late stages, but this reporter that is supposed to function as a proxy for gland induction is still expressed. That either means that expression of sna-rg-GFP is not a proxy or that the glands are still being specified in the absence of the Hox genes that are proposed to specify these organs. The reporter should not be expressed if these Hox genes are what specify these endocrine organs.”

RESPONSE: The reviewer has made a good observation. The expression of sna-rg-GFP is not completely absent in Dfd Scr mutant embryos (Fig. 5F in this revised version), which indicates that although the Hox genes are required to activate upd in the maxilla and labium and in their absence the gland primordia become apoptotic, there must be other positive inputs to the enhancer. However, this does not mean the Hox gene input is irrelevant for gland specification. Not only the Hox genes are required to keep normal levels of upd expression in the Mx and Lb primordia and gland viability, but previously we also showed that cephalic Hox genes influence the dorso-ventral position inside the vvl1+2 expressing cells where the sna-rg enhancer is activated: in the maxilla Dfd induces the ventral vvl1+2 expressing cells to activate sna-rg, while in the labium Scr induces the dorsal vvl1+2 cells to activate sna-rg (Sanchez Higueras 2014). The data presented in this paper indicate that the input of both Dfd and Scr over sna-rg CRM activation are indirect.

As a result of the reviewer’s criticism, we have tested if the additional positive input could be provided by Ci. In our previous submitted version, we showed that the repressor form of Ci blocks sna-rg activation. In this revised version, we have tested what is the effect of expressing the activator form of Ci. In embryos overexpressing the activator CiPKA isoform, we have observed that the expression of sna-rg and upd are expanded, indicating that Ci can provide the additional Hox-independent positive input. In the revised version we present these new results as Fig.3G and Fig. 4I. We have modified accordingly the scheme that appears in panel 3I to include this. In the main text we describe the result in the Hh regulation section where we have added:

“Although the above results indicate Ci is not absolutely required for sna-rg expression, we observed that overexpression of CiPKA, the active form of Ci, causes a non-fully penetrant expansion of sna-rg expression (Fig. 3G) suggesting the possibility that sna-rg may be responsive to Ci and to a second activator.”

… and in the “Regulation of Upd ligand expression by the Wg and Hh pathways” section

where we say:

“We also found that ectopic expression of the activator Ci protein results in a non-fully penetrant expansion of upd expression in stage 10 embryos (Fig. 4H-I).”

We have also modified the final scheme in Fig. 7 to mention that Dfd and Scr prevent the apoptosis of the gland primordia, and that there must be an additional positive input controlling upd activation besides the Hox input. However, in the figure we do not define Ci as the activating input as we would like to have additional evidence before making such claim.

To clarify that the Hox input is not absolutely required we have modified the text in several places. Where we said:

“Expression of the sna-rg reporter in the maxilla and the labium requires Dfd and Scr function …”

We now say:

“Development of the CA and PG and normal expression of the sna-rg reporter in the maxilla and the labium require Dfd and Scr function …”

We also mention this in Fig. 5 legend where we have added:

“In Dfd Scr mutant embryos (F), although the gland primordia become apoptotic, residual GFP expression indicates that there must exist Hox independent inputs activating the sna-rg enhancer.”

As a result of reviewer 3’s comment, we have noticed a further example of similarity between the gland and the trachea specification, which we have commented in the revised discussion where we added the following paragraph:

“Another interesting similarity between glands and trachea is that, although ectopic Hox gene expression can ectopically induce sna-rg and trh outside their normal domain, the lack of Hox expression does not completely abolish their endogenous expression, indicating that in both cases a second positive input can compensate for the absence of Hox mediated activation. Our results suggest that, in the glands, this redundant input could be provided by the activating Ci form (Figs. 3G and 4I), but further analysis to confirm this possibility and discard alternative sna-rg activators should be performed.”

We disagree with the following comments:

The finding that the CA and PGs form in slightly different DV positions from each other and slightly different DV positions from the trachea (based on the vvl1+2 mCherry reporter staining combined with that of the sna-rg-GFP reporter staining in Figure 5A, where staining does not overlap except where the CA cells have started to migrate over the vvl1+2 mCherry expressing cells) argues pretty strongly against the CA and PG being homologous to each other or absolutely homologous to the trachea primordia

RESPONSE: This erroneous claim was based on Fig. 5A, that showed a double stained embryo where co-expression is difficult to appreciate without separating the channels. Co-expression of these two reporter lines in the ring gland has been previously documented beyond doubt in our 2014 publication, cited throughout the manuscript, where we presented eight different panels of glands clearly co-expressing both markers at various developmental stages (Current Biology 2014 Fig.2B-I). To prevent any readers reaching the same conclusion as the reviewer, we have modified Fig. 5A to show a double stained sna-rg-GFP vvl1+2-mCherry embryo alongside with the two separate channels (panels 5A’ and A’’) to make the co-expression evident.

Although we are not including it in this manuscript, the reviewer will also be able to find images in the same 2014 Current Biology publication (Fig.3), where the ectopic activation of Dfd in the trunk leads to the activation of the sna-rg-GFP reporter in the vvl1+2 tracheal cells, proving that the glands and the trachea are formed at homologous positions.

Having made clear that sna-rg activation in both the CA and the PG occurs in vvl1+2 expressing cells, we now refute a second criticism: The reviewer is puzzled that despite the glands being formed at different dorso-ventral positions in the vvl1+2 expressing patch of cells, we claim both groups of cells are homologous to the trachea.

We are not saying that the CA are formed at homologous positions to those giving rise to the PG. What we say is that both the CA and the PG are formed at positions homologous to those giving rise to the trachea in the trunk segments.

To make this clear in the revised version, we have changed the wording of a sentence in the Introduction section that might have originated the confusion.

Instead of saying:

“First, the CA, the PG and the traqueal primordia are specified in the lateral ectoderm at homologous positions”.

Now, it reads:

“First, the CA and the PG are specified in the cephalic lateral ectoderm at homologous positions to those forming the tracheal primordia in more posterior trunk segments.”

It has been shown that each tracheal primordium (which are labelled by vvl1+2-mCherry) gives rise to different tracheal branches depending on the positions where they are specified: the dorsal cells give rise to the dorsal tracheal branches, the ventral cells to the ganglionic branches, the medial cells to the dorsal trunk etc. (for illustration see Fig.12 in Manning and Krasnow 1993). Each of these tracheal branches have a different shape and migrate to different positions. We believe that a similar positional specification occurs in the vvl1+2 cells in the maxilla and the labium. In the maxilla only the vvl1+2 ventral cells activate sna and svp (among other genes) to give rise to the CA. In the labium vvl1+2 dorsal cells activate sna, sal, phm (among other genes) to give rise to the PG. This regionalization is similar to what happens during tracheal branch specification, with the only difference that the interaction with Dfd and with Scr is what makes the positional outcome in the maxilla and the labium different (see in our Current Biology 2014 publication Fig. 3E-F and H-J). Thus, when the reviewer considers the equivalence between the CA/PG/trachea homology with that of the wing/haltere or that of the thoracic leg1/2/3 saying: “Indeed, the situation with these endocrine glands and the trachea is completely unlike the situation with the wing and haltere, wherein both structures arise from the same DV position in adjacent segments, or with legs 1, 2 and 3, which arise from the same DV position in adjacent segments”

…the reviewer should think about the coxa and the tarsi in the legs. The coxa in T1 is not homologous to the tarsi in T2 or T3, but when considering the leg structure as a whole, the coxa and the tarsi form part of the same homologous structure in T1, T2 and T3 despite being formed at different positions inside the leg primordia.

The reviewer also doubts that the activation of upd occurs in the sna-rg primordium when saying: “Likewise, the STAT10X-GFP staining does not overlap with the sna-rg-mCherry staining (I see red cells and I see green cells - there are no yellow cells). If activation of snail is through Upd activation of STAT signaling, we should see that the snail reporter expression is within the domain of STAT10X-GFP expression.”

RESPONSE: This is due to the fact that upd activation in the CA is extremely transient, leading to the loss of the x10STAT-GFP expression before the sna-rg-mCherry levels are robust enough in the maxilla. This criticism does not apply to the PG where due to upd expression lasting longer, co-expression of sna-rg-mCherry and x10STAT-GFP in panel 4B should be evident to the reviewer.

To try to sort the CA co-expression problem, we are currently repeating the experiment but instead of analysing sna-rg-mCherry activation with the RFP antibody, we will do an mcherry RNA in situ. We hope that the mcherry transcript will be detectable earlier than the protein and the co-expression will be evident.

We strongly disagree when the reviewer says: “This paper provides a strong basis for arguing that the CA and PG are induced independently of Jak/Stat signaling, whereas trachea require this signaling pathway.”

RESPONSE: When making this claim, the reviewer is ignoring a large number of experiments presented in the manuscript. If the CA and the PG are induced independently of JAK/STAT signalling:

(1) Why sna-rg expression disappears from the glands in mutants lacking the Upd ligands (Fig. 5B and 6K)?

(2) Why deleting the region containing the putative STAT binding sites in the sna-rg enhancer causes the loss of enhancer expression (Fig. 6C)?

(3) Why the smaller enhancer mentioned in point (2) recovers gland expression when adding a STAT binding site from an unrelated gene (Fig. 6G)?

(4) Why the regained expression of the construct mentioned in (3) is lost by the mutation of two bases affecting this single STAT site (Fig. 6H)?

The reviewer’s conclusion rests on giving an excessive importance to his reservations to CA co-expression in panel 4A while, surprisingly, disregarding the co-expression in the PG shown in panel 4B and all the experiments presented in Fig. 5 and Fig.6.

Reviewer 3 Minor comments: RESPONSE: Both comments have been taken into account in the revised version.

In summary, in this revised version we have answered most queries raised by reviewers 1 and 2. Moreover, reviewers 1 and 2 agree that the results presented in this manuscript reinforce the hypothesis that the CA and the PG glands and the trachea derive from the divergent evolution of a metamerically repeated homologous organ.

Reviewer 3 has made a good point that we have taken into account and has improved the revised submission.

However, reviewer 3 is wrong when concluding:

“This paper provides a strong basis for arguing that the CA, PG and trachea are not homologous structures”, and when saying: “the CA and PG are induced independently of Jak/Stat signaling, whereas trachea require this signaling pathway”.

As we argue above, these conclusions are erroneous because:

(1) Are based on the incorrect interpretation of Fig 5A and ignore previous published evidence cited throughout the manuscript.

(2) It does not take into account key experiments presented in this work, while giving too much weigh to a result that can be easily interpreted.

(3) It misinterprets the arguments justifying the positional homology between the CA/PG glands and trachea primordia.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

This paper focuses on the specification of two endocrine glands that form from head ectoderm, the corpora allata (CA), which forms in the maxillary segment and secretes Juvenile hormone, and the prothoracic glands (PG), which form in the labial segment and secrete Ecdysone. Secretion of both hormones results in a larval molt. Secretion of only Ecdysone induces metamorphosis, the transition of the larvae into the adult forms. Both the CA and PGs form in positions homologous to the tracheal primordia (approximately) and previous reports indicate that ectopic expression of the appropriate Hox genes can result in homeotic transformations of the glands into tracheal primordia and of tracheal primordia into glands. Using a GFP reporter construct for the snail gene as a proxy for gland specification, the authors show that CA and PG formation is regulated by two segment polarity genes: Hh and Wnt, with Hh signaling activating reporter gene expression and Wnt signaling inhibiting reporter gene expression. They also suggest that their endocrine gland GFP reporter is regulated by the two Hox proteins expressed in those segments: Dfd (maxillary) and Scr (labial) (although figure 5D,E argue against this conclusion). They presumably show that reporter gene regulation by Wnt signaling and Hh signaling is indirect and through localized transcriptional activation of the JAK/STAT signaling pathway ligand gene upd (however, the STAT reporter and the snail reporter are expressed in different cells (fig 4B) - so I'm not so convinced of this conclusion). The authors also find that the CA and PG primordia form at slightly different dorsal ventral positions and that DV positional information is controlled downstream of upd JAK/STAT signaling.

Major comments:

The paper is well written and makes for a nice story, but the corresponding data are not supportive of most of the conclusions drawn by the authors.

First, in the Dfd Scr double mutant, both the CA and PG expression of the snail-rg-GFP reporter is still there - admittedly, the gland cells look abnormal at late stages, but this reporter that is supposed to function as a proxy for gland induction is still expressed. That either means that expression of sna-rg-GFP is not a proxy or that the glands are still being specified in the absence of the Hox genes that are proposed to specify these organs. The reporter should not be expressed if these Hox genes are what specify these endocrine organs. This finding might explain why mutating the Hox consensus binding sites had no effect on expression of the smaller snail reporters.

The finding that the CA and PGs form in slightly different DV positions from each other and slightly different DV positions from the trachea (based on the vvl1+2 mCherry reporter staining combined with that of the sna-rg-GFP reporter staining in Figure 5A, where staining does not overlap except where the CA cells have started to migrate over the vvl1+2 mCherry expressing cells) argues pretty strongly against the CA and PG being homologous to each other or absolutely homologous to the trachea primordia. Likewise, the STAT10X-GFP staining does not overlap with the sna-rg-mCherry staining (I see red cells and I see green cells - there are no yellow cells). If activation of snail is through Upd activation of STAT signaling, we should see that the snail reporter expression is within the domain of STAT10X-GFP expression. This would be consistent with observing a loss of upd mRNA in the maxillary and labial segments with loss of Dfd and Scr, but not seeing a loss of the sna-rg-GFP reporter. This would also argue against the proposed homology between the glands and the trachea. Indeed, the situation with these endocrine glands and the trachea is completely unlike the situation with the wing and haltere, wherein both structures arise from the same DV position in adjacent segments, or with legs 1, 2 and 3, which arise from the same DV position in adjacent segments. This paper provides a strong basis for arguing that the CA, PG and trachea are not homologous structures and that the CA and PG are induced independently of Jak/Stat signaling, whereas trachea require this signaling pathway.

Minor comments:

Page 3: tracheal is misspelled in the first paragraph, line 3.

Page 5, end of first sentence in first full paragraph: "lethal" should be changed to "non-viable". I think the authors mean that homozygous embryos die, not that they cause the death of other life forms.

Significance

Nature of significance of advance:

I think the significant finding is that the CA, PG, and trachea are not homologous structures. But that is not what the authors are concluding. The only findings consistent with the data provided are that Wg signaling represses expression of the snail reporter and Hh signaling activates its expression (Figures 1 - 3). Most of the other conclusions do not seem to be sufficiently supported by the data.

Context of the work:

These authors have published that the CA and PG are structures specified in homologous positions to the trachea. It has already been published that CA, PG and trachea primordia express the Vvl transcription factor - although I did not go back to see how that was determined. It has already been published that ectopic expression of specific Hox genes can transform the gland primordia into trachea and vice versa (these experiments may also warrant a closer look). So, idea that CA, PG and TR arose from divergent evolution of a segmentally repeated ancient structure has been proposed.

Best target audience:

With the findings that are consistent with the story line (figures 1 - 3), Drosophila embryologists working on the formation of these glands would be interested.

My field of expertise:

Drosophila development.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This manuscript from Garcia-Ferres et al., describes studies aimed at understanding the specification of precursor cells to two ring gland organs, the corpora allata (CA) and the prothoracic gland (PG). These two glands are specified in the lateral ectoderm at positions that match the tracheae in more posterior segments, and like tracheal cells, the CA/PG primordia express ventral veinless (Vvl). The specification of CA/PG primordia requires the Hox genes Deformed (Dfd) and Sex combs reduced (Scr), while the trachea dependent on BX-C complex genes. CA/PG precursor cells also differ from tracheal precursors in that the former requires the expression of "mesodermal" gene snail (sna), while tracheal cells require trachealess. The sna gene has a complex regulatory region, with distinct enhancers for expression in the mesoderm and in the CA/PG primordia. Garcia-Ferres et al have used a reporter carrying the sna CA/PG enhancer, rg, to studying the mechanisms of CA/PG specification.

In the first set of experiments, using a Crispr deletion the authors showed that the rg enhancer in the endogenous Sna gene is essential for CA/PG specification. The next used the sna-rg to examine the effects of mutations in (potentially) upstream signaling pathways on CA/PG specification. The CA/PG primordia are located outside of the wingless (wg) (parasegmental) expression domain; however, the authors found that two sets of lateral ectodermal cells express sna-rg in wg mutants. Conversely ectopic expression of wg or armS10 in the maxillia and labium eliminates sna-rg expression indicating that wg is a negative regulator of sna. Unlike wg, mutations in hedgehog (hh) and engrailed (en) eliminate sna-rg expression, indicate that both of these genes are required to promote CA/PG specification. The failure to express sna-rg is due at least in part to repressive activities of Cubitus interruptus (Ci) as sna-rg expression is restored in an en, ci double mutant. Sna-rg expression also depends upon JAK/STAT signaling and is in a deficiency that deletes the three upd genes. A upd dependent reporter is activated in CA/PG primordia in stage10/11 embryo, while upd itself appears to be express in the same cells. Consistent with the idea that the JAK/STAT maybe controlling sna-rg expression in response to wg and hh signaling, the authors finding that expression of upd expands in wg mutants, while it disappears in hh mutants. upd expression appears to be controlled by inputs not only from wg and hh, but also by Dfd and Scr as Dfd Scr mutant embryos lack upd expression. In contrast, the pattern of wg and en expression in the Dfd Scr mutant is normal.

To confirm that JAK/STAT signaling is required for activating the sna-rg reporter, the authors undertook a functional dissection of the rg enhancers. When they split a truncated rg enhancer, rgR2, into two fragments, A1 and A2, there was no expression of the reporter. As fragment A2 has three STAT binding sites, the authors tested whether expression could be rescued by adding a generic STAT site to A1-it could be. On the other hand, when they mutated the three STAT sites in A2, they found that only PG expression is lost. This result suggests that there is a Upd responsive element in A1+A2; however, activation is likely indirect since the activity of this element is tissue specific. They also putative Hox-Exd-Hth bindings sites in the rgR2 enhancer; however. these sites do not appear to be required for reporter expression.

This paper adds to our understanding of how the CA and PG primordia are specified and highlights important similarities with the specification of the tracheal primordia. There are some questions that should be addressed.

Specific Comments:

There are two phases to sna expression in CA/PG primordia. In the first phase (stage 11) hh signaling to neighboring cells prevents of the Ci repressor protein which would in turn activate upd expression in these cells. However, wg expression anterior in anterior cells blocks Upd expression so that it is turned on only a single set of cells posterior to the hh expressing cells. Upd in turn activates the sna via the STAT binding sites in the rg enhancer. upd expression also depends on the Hox genes Dfd and Scr, and when they are mutant upd is not expressed and sna is not turned on. Upd expression is only transient, and so after it disappears a maintenance mechanism ensures that that sna is expressed until at least stage 16. The authors don't really address the maintenance mechanism so it isn't clear what elements or factors are needed to keep the rg enhancer active after upd expression disappears.

• a) In Fig. 1, G-J the authors show sna-rg reporter expression in WT and in their sna-rg deletion mutant. The images in G and I are of stage 12 embryos. At this point there is clearly little if any rg reporter expression in the deletion mutant, while high levels are observed in the WT. However, this is after sna expression is supposed to be activated by Upd, which is in stage 11. Moreover, upd expression is turned off in stage 12. (Assuming I didn't miss something) It is not clear to me from either this figure or from the text whether the initial pattern of expression of the sna-rg reporter in stage 11 embryos is WT and then disappears at stage 12, or whether it is always defective. In trying to understand the activation process, I think it would be important to know for sure whether rg enhancer activity during the initiation phase in stage 11 is WT or not. The expectation-at least from what is written in the manuscript-is that the initial expression of the sna-rg reporter will be the same as WT in the sna-rg deletion mutant.
• b) The authors argue that the rg deletion removes the only region driving sna expression in CA/PG. I'm not convinced that necessity necessarily implies sufficiency with respect to the requirements for rescue. While the sna-rg reporter is expressed in a pattern that seems to mimic the endogenous gene, do we know that a rg-sna transgene would fully rescue the rg deletion mutant?
• c) If sna protein is not required for initiating (see a) sna expression, is it required for maintaining sna expression?
• d) The authors show that there is an expansion in the number of sna-rg reporter expressing cells along the AP axis when upd is ectopically expressed using a sal-Gal4 driver. Though not mentioned in the text at this juncture, sal is expressed in the PG primordia, while seven-up (svp) is expressed in the CA primordia. I assume that the upd induced expansion is only observed for the PG primorida (LB) and not the CA primordia (Mx)-at least this is what the figure looks like.<br /> It looks to me also like the vvl domain is expanding as well. This information should be clarified. How about svp driven upd-assuming there is a svp-Gal4 driver-does it cause an expansion of Ca but not PG.
• e) The authors note a difference between CA and PG in the requirement for STAT binding sites in the enhancers. Is that related to the fact that svp is expressed in CA and sal is expressed in PG? Would driving svp expression using the sal-Gal4 driver maintain sna-rg expression
• f) Do svp or sal have a role in initiating sna expression when upd is present or maintaining sna expression after upd disappears? Presumably there is already published data that would answer these questions.

Significance

This manuscript will be of interest to scientists seeking to understand fate specification-and how the same pathways/interactions can generate completely different organs-- in this case ring glands as opposed to trachea. The paper does however leave many questions unanswered. This is to be expected given that not all of the key players have been identified, and for those that have, the functions are not fully understood.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this study, the authors focus on understanding the regulation of development of corpora allata (CA) and prothoracic gland (PG) in Drosophila. Through a series of well designed experiments they convincingly show that interactions between Scr, Dfd, and JAK/STAT pathways induce localized expression of unpaired (upd) gene, which in turn controls snail (a key regulator of early gland development). Overall, this is an excellent study that extends authors' previous work on evolution and divergence of glands (in head) and respiratory organs (in trunk) from common, metamerically repeated primordia. Specifically, this work provides new information regarding CA and PG specification and as such will be of interest to a broad range of developmental biologists. Text is very well written, and figures are well organized and convey results in a clear way. I have several small comments (detailed below), but other than that this manuscript is ready for publication.

Aleksandar Popadić

Main suggestion:

As the formation of the ring gland is an exclusively dipteran trait, it would be helpful to insert a paragraph in discussion explaining the relevance of new insights to other, more basal insects. In a sense, studies of a ring gland present a tail end of evolution, what do results obtain tell us about the regulation of the PG and CA development in other insects?

Additional comments:

1. Introduction, pg 3, a paragraph starting with "In comparison to the extensive knowledge we have of ..." - consider omitting or greatly shortening, this text breaks a flow as it is focused on tracheal development. I understand the authors' logic, but this information distracts from the main focus on CA and PG.
2. Beginning of discussion, pg 11:
   • change 2nd sentence to: " Our results indicate that the HH and the Wnt pathways act indirectly to negatively regulate the spatial activation ..."
   • the following sentence, starting with "Engrailed activation off hh transcription ...." is way too long and hard to follow, consider breaking into two sentences.
3. In Fig 4B, mx and lb segments should be labeled so this panel is consistent with labeling in 4A.
4. In Fig 6, reduce a font size for labels on right-hand side (A1, A2, A1+A2 proximal, etc), so that they are visualy distinct from panel labels on left-hand side (A, B, C,..).

Significance

While this is a strictly Drosophila study, it does provide a significant new insight into development of corpora allot and prothoracic gland (which are critical organs for insect growth and development). As such this work will be of interest to a wide audience of biologists (please see my comments below for details).

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

Learn more at Review Commons

---

Reply to the reviewers

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

*In this manuscript, Dr. Huiping Liu and colleagues investigate the role of CD81 in breast cancer metastasis, cancer stem cells, and extracellular vesicles (EVs). CD81 is a tetraspanin protein that has unclear roles in cancer. The authors discover that CD81 can form a complex with CD44 on cell surface and instigate cell clustering (important for CTC dissemination), self renewal and metastasis using multiple cell lines and xenografts. Multi-omic studies led them to CD81-regualted EVs, which they found to play critical roles in driving stemness and cell clustering. Furthermore, they found that CD81 is co-expressed with CD44 and affects patient outcomes. Overall, the novelty of the work is high, and the amount of data is impressive and of high quality. The experiments presented are well controlled and rigorous. However, there are some concerns as listed below: *

We appreciate the positive comments on the high novelty and high quality of the work. The concerns below have been addressed with point-to-point answers.

*Major: 1. The conclusion of EVs made by TICs to reprogram non-TICs into TICs is based on adding large amounts of EVs isolated from WT tumor cells. The amount may not be achievable in physiological conditions. Further, if the EVs released by TICs are sufficient to reprogram neighboring tumor cells, this event would be expected to self-propagate and all the tumor cells would become TICs. *

We are thankful to the reviewer for the valuable thoughts and input. While the EVs used to educate and reprogram non-TICs may seem to be large amounts, the doses are justified as below and in the Method section to address pathophysiologically relevant questions using testable models.

First, we add this section of “Exosome or sEV education” to the Method section.

The EV education/reprogramming doses were determined based on literature reports and our own measurements. From published reports (reference PMID: 27599779), EVs could be detected in human/mouse plasma with high levels in the patients with cancer (>1x109/mL higher in cancer patients than healthy controls), as well as mouse plasma (>1x109/ml higher EV counts in the tumor bearing mice than non-tumor mice). In our studies, triple negative breast cancer (TNBC) cells do secret large amounts of EVs, with a yield of 0.2~1x109 EVs /mL supernatant (MDA-MB-231 or 4T1 cell culture) as measured on NTA as well as Apogee micro flow vesiclometer (MFV). Please note the small EV collecting efficiency is only ~10% by ultracentrifugation at 100,000xg for 70 min during the EV purification and we end up obtaining 1-2% EVs after two spins (one or two washes) (200 mL culture to yield 60-75 µg EV protein). Therefore, we utilized an amount of the EVs purified from 50x larger volume of culture supernatants for EV educating culture. For example, 10-15 µg purified EVs purified from ~50mL supernatant were used for 1 mL educating culture to make 2~5x108 EVs/mL for reprograming CD81KO cells. The dose is pathophysiologically relevant to the EV concentrations in the plasma and culture media.

Second, we admit that the boundary and heterogeneity between TICs and non-TICs are more drastic in patient/mouse tumors in vivo than cultured tumor cells in vitro. Considering the complexity that the educational effects of cancer EVs on surrounding cells in vivo interplay with many other variables such as spatial restrictions, EV-dilution by interstitial fluids, EV release into blood, and mixed tumor-suppressing factors in the microenvironment, we have adopted the simplified EV education models in vitro to quantify the reprogramming effects of TIC EVs on non-TIC tumor cells (CD81KO cells). This model is mainly utilized to examine the effects of EV proteins CD44 and CD81 in measurable reprograming activities. Notably, we observed that CD44+CD81+ TICs tend to gradually dominate in culture whereas non-TIC populations gradually lose and thus a high percentage of cells may become tumorigenic TICs. Our data demonstrate that EV-education effects can contribute to the outcomes in a CD44- and CD81-dependent manner.

• The conclusion that CD44 switched from membrane to intracellular locations adherent vs. suspension cultures is not well supported by the data presented in Fig 1E. It appears that the cells expressing membrane CD44 did not change much, but the cells population expressing intracellular CD44 expanded in the suspension culture. *

We appreciate the comment and have clarified the observation and the conclusion accordingly in updated Figure 1C-D:

“CD44 was observed mostly on the cytoplasmic membrane in WT MDA-MB-231 cells, both adherent and in suspension, but the intracellular CD44 accumulated specifically in the cells in suspension (P=0.03) (Figure 1C-D)”

• Fig 6E-H experiments are a bit flawed. Metastasis degree is unlikely to be proportional to the primary tumor size, so it is not proper to normalize the metastatic burden to the primary tumor weight. These data can be removed without significantly weakening the major conclusions. *

Thanks for the suggestion. We have removed the normalized panel of original Fig 6H and updated as below with tumor weight and lung metastasis signals.

• Fig 3J appears to have different lanes pasted together. ¬¬Therefore, it is unclear whether the results could be compared, calling into the question of the statements involving the proteins studied including OCT4, pSTAT3 and FAK. *

Sorry for the confusion. We must clarify that all the lanes are from the same blot which has been stripped and reblotted for multiple proteins. The full blots of previous Fig 3J (new supplementary Fig S8A) are included below for reference.

• The authors state that EVs were comparable between WT and 44ko and 84ko cells partly based on the western blot presented in Fig 3E. While TSG101 appears similar, the other markers appear quite different. This needs explanation. *

We conducted three independent immunoblots for EV proteins in Fig 3E and quantified the protein intensities in 3F which demonstrates that both CD44 and CD81 are significantly lost in the EVs of CD81KO cells and CD44KO cells, CD63 slightly reduced in the EVs of CD44KO cells, whereas no significant differences in EV markers TSG101 and LAMP2b comparing the KO EVs with the WT EVs (the levels normalized by b-actin loading control). These data suggest that CD81 and CD44 are required for each other’s localization or packaging to the EV.

• The rationale for different cut-offs in Fig 4A-C is unclear. Auto cut-offs should not be used to maximize the detection of difference. The exact public dataset used should be stated. *

Thanks for the insightful suggestions. In Supplementary Figure S8, we have added new Kaplan-Meier plots based on Liu_2014 cohort (TNBC) with cutoffs at lower 25%, median and upper 25% of CD81 protein levels, respectively, most of which show significant differences in overall survival (OS), relapse-free survival (RFS), and distant metastasis-free survival (DMSF) between CD81 high and low groups (CD81 as an unfavorable marker). These are consistent with and strengthening the data of the KM plots with auto cut offs in Fig 4A-C.

• Minor The description regarding the screen leading to CD81 is confusing. Including a diagram may help. *

Thanks again and we have added a schematic to Supplementary Figure. S1A with the experimental procedure details of sorting CD44 +/- cells and performing CD44 knockdown in PDX tumors for mass spectrometry comparisons. CD81 is one of 38 overlapped proteins differently shown in two comparisons.

• Fig 1E, the phrase "ratio of cells expressing" is bit confusing. Did the authors mean % of cells expressing their indicated markers? *

Yes, the y axis label is equivalent to % (1=100%) and now specified as “Proportion of CD44+ or CD88+ cells” in Figure 1E.

• Cryo-EM shows detects impaired membrane integrity of EVs of 81ko but 44ko. Yet, Fig 2 shows that ko of either 44 or 81 disruptions the localization and therefore the function of the other. Explanation is needed for why 44ko did not affect 81 regulation of EVs. *

We appreciate the thoughtful comments. After repeating and reanalyzing the experimental data from multiple cryo-EM and immunoblotting experiments, we updated Figure 2 which shows both CD81KO and CD44KO impaired the membrane integrity of EVs. Furthermore, the immunoblot validated that CD44 is deficient in the EVs of CD81KO cells, suggesting a possible role of CD81 and in recruiting CD44 to EV and strengthening the EV membrane integrity. In the meantime, CD81 localization on the membrane is dependent on the presence of membrane CD44 in Fig 1E.

• Fig S8C is not a robust correlation analysis. A scatter plot should be presented. *

A scatter plot is included (new Fig S9C) with R = 0.29 and p = 0.015.

• There are a few typos. For example "cytoplastic membranes" should be cytoplasmic membranes. The following sentence is awkward: "Among the siCD81-upregulated genes, SEMA7a, a glycosylphosphatidylinositol membrane anchor promoting osteoclast and blood cell differentiation (58, 59), when further depleted in CD81KO cells, siSEMA7a partially rescued or restored mammosphere formation in these cells (Supplementary Figure S2D-F),.."* Typos were corrected. The sentence has been updated to:

Among the siCD81-upregulated genes, SEMA7a, a glycosylphosphatidylinositol membrane anchor promoting osteoclast and blood cell differentiation (50, 51), was depleted in CD81KO cells by siSEMA7a which transfection partially rescued or restored mammosphere formation in these cells (Supplementary Figure S2D-F), suggesting a novel role of SEMA7a in inhibiting self-renewal of breast cancer cells.

Reviewer #2

The experiments are generally well performed and convincing, except that all mouse experiments seem to have been performed only once, with small groups of mice (between 4 and 6 with several sites of tumor injection per mouse). Reproducing at least once these experiments should be shown.

We appreciate the positive comments as well as instructive suggestions. We would like to clarify that the mouse studies in Figures 5 and 6 were originally completed with two to three different models using both human and mouse TNBC cells. While one of the tumorigenic experiments in Figure 5 (where three different models were used) was done once due to the pandemic disruption and increased costs for NSG mice, Figure 6 data was originally generated by multiple in vivo experiments. Nevertheless, we have managed to repeat the in vivo tumorigenic experiments. We updated Figure 5 (see next page) with new results from 2-3 experiments for each panel/model with increased number of injections to 8-20 in 2-5 mice per group in each of three different tumor models. Furthermore, we also selectively repeated 4T1 experiments in Figure 6M with updated legend showing the experiment was consistently repeated.

“These experiments were repeated at least twice to show consistent conclusions”.

*Another unsatisfying aspect is the claim that CD44 and CD81 are specifically required for secretion of the subtype of EVs called exosomes, which forms in intracellular multivesicular endosomes. This claim is based on the (wrong) assertion that CD81 (together with CD9, CD63 and TSG101) is an exosome marker, based on ref 46,47 published in 2006 and 2011: the field has evolved a lot since then, and it is now becoming clear that CD63 (which normally accumulates in multivesicular endosomes) may be enriched in exosomes, but that neither CD9 nor CD81 are, since they mainly localize at the plasma membrane, and thus probably are released more prominently in small microvesicles (see Kowal, J., et al. (2016). Proc Natl Acad Sci U S A 113: E968. and Mathieu, M., et al. (2021). Nat Commun 12(1): 4389.). *

Thanks for the comment on the evolving literature of CD81. To avoid confusion and follow the EV nomenclature guidelines, we used the more acceptable and general term “extracellular vesicles (EVs)” instead of “exosomes” in our manuscript. We have also cited the two new publications in PNAS 2016 and Nat Commu 2021 (ref 38 and 39) with modified statement on CD81 as below.

“We performed mass spectrometry proteomic profiling of TNBC patient-derived xenografts (PDX) tumor cells that cluster and discovered that one of the altered proteins upon CD44 depletion was CD81, a tetraspanin protein enriched in extracellular vesicles (EVs) (38-39).”

*In figure 3, the authors show empty internal compartments in cells with deleted CD44 or CD81, which (if these compartments are altered MVBs) should then lead to fewer exosomes recovered extracellularly. However, the authors observe instead more particles recovered from these cells. When analyzing the composition of these EVs, the authors show maybe a decrease in CD63 in EVs from both CD44 and CD81 ko cells, but contradictory effects on the presence of Lamp2 (which is a more convincing marker of late endosome/lysosome-derived exosomes): increased in CD44 but decreased in CD81 ko. These results, however, are not really interpretable, since the authors show only a single Western Blot, thus no evidence of reliable changes in protein composition. In any case, I would suggest that the authors do not, in the current state of the article, try to claim any specific subcellular origin of the EVs affected by CD44 and CD81. *

We appreciate the comments and would like to clarify our observations. When the CD44/CD81 KO cells show increased empty internal compartments, it is a sign for loss of cellular mass. Therefore, it might be surprising but also reasonable to link such phenotype with an increased quantity of draining EVs in low quality (disrupted membrane) from CD44KO and CD81KO cells that show active endocytosis and/or exocytosis pathways.

To better quantify the proteins of the EVs derived from WT, CD44KO and CD81KO cells, we repeated the western blots three times and in new Fig 3F we reported the levels of CD44, CD81, and other EV proteins (CD63, TSP101, and LAMP2b). Compared to WT EVs, CD44 and CD81 are relatively compromised in the EVs of CD81KO and CD44KO, respectively. CD63 slightly decreases in CD44KO EVs whereas LAMP2b shows no significant changes in both KO cells. We modified the text below reporting the EV phenotypes without specifying subcellular origin of EVs.

After purified from the culture supernatants of CD44KO and CD81KO cells via 100,000 xg ultracentrifugation (Supplementary Fig S6A), the sizes of small EVs (ev44KO and ev81KO) were relatively comparable to the WT control, as characterized by nanoparticle tracking analysis (NTA) and immunoblotting with EV markers (Figure 3E-F, Supplementary Figure S6B). However, when examined by cryo-EM, ev81KO displayed impaired membrane integrity (Figure 3E-F), indicating an essential role for CD81 in modulating EV biogenesis and packaging of membrane proteins.

*In addition, the EVs are only quantified by the vesicle flow cytometry established by the authors, but then, in functional assay, and in Western blots, EVs are quantified in terms of proteins. It would be important to show if CD44 and CD81 ko also decrease the amount of proteins recovered in the EV preparations, and the number of particles quantified by NTA, as they apparently decrease the number of events detected by vesicle flow cytometry, or not, and if not, why did the authors chose to show only the vesicle flow cytometry results (this is not a very commonly used technic in the field). *

Thanks for the suggestion. We have quantified the EVs by vesicle flow cytometry (MFV) and NTA which show consistent EV counts of WT, CD44KO, and CD81 KO (see table below). We utilize EV protein to normalize the EV education as all the samples had about 6x108 EVs/ µg protein.

We have also calculated the EV production data (counts/cell) as measured by NTA in Supplementary Fig SD which is relatively consistent with the MFV analysis in Fig 3C, demonstrating that CD44KO and CD81KO cells release a higher number of EVs than WT cells.

We have also calculated the EV production data (counts/cell) as measured by NTA in Supplementary Fig SD which is relatively consistent with the MFV analysis in Fig 3C, demonstrating that CD44KO and CD81KO cells release a higher number of EVs than WT cells.

*Finally, a somehow frustrating aspect of the paper is that the link between the observed effect of CD44 or CD81 ko on EV release and their in vitro functions on mammosphere formation (fig3), and the effect on in vivo tumor growth and metastasis (fig 5-6) end up as two separate observations (the authors rightly do not claim that impaired exosome release in vivo is responsible for the impaired metastasis). The observation also of clustered CD81 and CD44+ circulating tumor cells in patients (fig4) is also somehow a separate observation. Thus there are several stories put together in this article. *

We appreciate the comment and apologize for disconnected data presentation. We have reorganized the paper to highlight machine learning-assisted discoveries of CD81 functions and molecular network in partnership with CD44 in promoting cancer stemness which is connected to endocytosis-related EV phenotypes. We admit that in addition to massive manpower and financial support, there are technical limitations in the EV-mediated functional studies in animal models. As proof-of-concept, we therefore utilized the simulated models in vitro to test the hypothesis of EV-CD44 and EV-CD81 in promoting cancer stemness of recipient cells. Our follow-up studies on EV-educated animals are ongoing but beyond the scope of the current manuscript. We are also open to the suggestion leaving the EV part out if that’s recommended by all editors and all reviewers.

Please see updated title “Machine learning-assisted elucidation of CD81-CD44 interactions in promoting cancer stemness and extracellular vesicle integrity” and the abstract.

Tumor-initiating cells with reprogramming plasticity or stem-progenitor cell properties (stemness) are thought to be essential for cancer development and metastatic regeneration in many cancers; however, elucidation of the underlying molecular network and pathways remains demanding. Combining machine learning and experimental investigation, here we report CD81, a tetraspanin transmembrane protein known to be enriched in extracellular vesicles (EVs), as a newly identified driver of breast cancer stemness and metastasis. Using protein structure modeling and interface prediction-guided mutagenesis, we demonstrate that membrane CD81 interacts with CD44 through their extracellular regions in promoting tumor cell cluster formation and lung metastasis of triple negative breast cancer (TNBC). In-depth global and phosphoproteomic analyses of tumor cells deficient with CD81 or CD44 unveils endocytosis-related pathway alterations, leading to further identification of a quality-keeping role of CD44 and CD81 in EV secretion as well as in EV-associated stemness-promoting function. CD81 is co-expressed along with CD44 in human circulating tumor cells (CTCs) and enriched in clustered CTCs that promote cancer stemness and metastasis, supporting the clinical significance of CD81 in association with patient outcomes. Our study highlights machine learning as a powerful tool in facilitating the molecular understanding of new molecular targets in regulating stemness and metastasis of TNBC.

Reviewer #2 (Significance (Required)):

These results are interesting as showing a novel molecule whose high expression may promote tumor progressions (CD81, as a cluster with CD44). The novelty lies in the functional interaction between CD44 and CD81 leading to the pro-metastatic effect described. Interaction between CD44 and CD81 had been previously observed in a generic proteomic study of EVs (Perez-Hernandez, D., et al. (2013). J Biol Chem 288: 11649), but not more explored in terms of consequent functions. A pro-metastatic effect of CD81 expression in tumors, especially TNBC, has also been recently demonstrated (Vences-Catalan, F., et al. (2021). Proc Natl Acad Sci U S A 118.). These two articles should be quoted in the current paper. My field of expertise is extracellular vesicles, and their roles in cancer progression. I can only judge superficially the modeling part of the article,

Thank you for highlighting the novelty of CD81 interaction with CD44 as a cluster in promoting tumor progression. We are grateful to the reviewer for providing the CD81 literature information which has been included and cited in the discussion (ref 65, 68), serving as a cross-validation for part of our work.

A study by Perez-Hernandez et al. also observed CD44 among the EV protein interactome network pulled down by CD81 peptides without exploring their relevance to EV functions (65)… A potential anti-CD81 therapeutic strategy was identified that may block the pro-metastatic effect of CD81 in animal studies (68).

Reviewer #3

*Major Comments:

The authors suggest that CD44 and CD81 interact and colocalize inside breast cancer cells. However, staining data presented shows very modest co-localization and immunoprecipitation experiments only employed beads as a negative control. To reinforce their conclusions, the authors should quantify co-localization using standard methods (Mander's or Pearson's coefficients). In addition, the authors should perform negative control immunoprecipitation experiments with antibodies against a target protein not expected to interact with CD81 to show that CD44 binding is specific. *

We are grateful for the instructive suggestions. To reinforce our conclusion about the membrane CD44-CD81 interactions, we repeated the Co-IP using anti-CD44 along with two negative controls (new Fig 1F), one of which is the IgG-bead control and the other is CD44KO cell lysate negative control). CD81 was only detected in the protein complex of the WT lysate (TN1 PDX or MDA-MB-231) pulled down by anti-CD44 (new Fig 1F). We also quantified the colocalization using Pearson’s coefficients with average r=0.57 from three different experiments which is now included in Supplementary Fig S2A.

From TEM and quantification in figures 3A, B the authors conclude that there is increased vacuolization with cells. They suggest that purple arrows point to multivesicular endosomes and yellow arrows vacuoles. In fact, the electron dense organelles indicated by purple arrows look more like lysosomes, whereas the yellow arrows appear more like early endosomes/endocytic vesicles. The authors should reassess their vesicle classification system and also provide a breakdown of the proportions of these structures within the graph.

Thanks for the comments. We agree with the reviewer that yellow arrows in Fig 3A could be vacuoles of early endosomes. And we added lysosome images and quantification in Supplementary Fig S6C, showing significant differences among three types of cells (WT, CD44KO, and CD81KO). That may help explain the phenotypes of altered EV release in CD44KO and CD81KO cells.

*Given that CD81+/CD44+ EVs are proposed to drive the aggregation and self-renewal of tumor initiating cells, it is somewhat counterintuitive that CD44 and CD81 KO cells show increased levels of EV secretion relative to WT controls. Additionally, it is perplexing that CD44 and CD81 secretion in EVs is unaffected (or my even increase) in knockout cells despite the fact these membrane proteins are supposed to interact and are mutually required for proper expression/localization. How do the authors reconcile these potentially contradictory observations? *

We appreciate the diligence and apologize for the lack of quantification and normalized loading in the original western blotting. We have repeated the EV western blots for protein density quantifications three times in new Fig 3E-F (see next page) that demonstrate a dramatic loss of CD44 in CD81 KO-EVs and a partial loss of CD81 in CD44KO EVs.

*The EV rescue experiments in figure 3H, I show recovery of mammosphere formation in CD81 KO cells treated with EVs from WT cells, suggesting that WT EVs are sufficient to rescue self-renewal. However, it's unclear from their studies whether exosome secretion of CD81/CD44 is actually necessary for aggregation and self-renewal phenotypes. Since CD81 has also been shown to be important for the trafficking of membrane proteins, the loss of self self renewal could relate to cell autonomous alterations in vesicular trafficking in CD81 KO cells. To reconcile between these possibilities, the authors should evaluate how depletion of factors necessary for CD81 secretion (e.g. Rab27a (PMID: 26305877; supplemental data), ESCRT components (PMID: 32049272), or others?) affects mammosphere formation. In either case the results would be extremely interesting and help to determine whether self-renewal is controlled by CD81 via cell autonomous or non-cell autonomous mechanisms. *

We are thankful for the extremely intriguing question about cell autonomous and non-cell autonomous roles of CD81 in controlling self-renewal. To address the question, we did transfect siRNAs to knock down Rab27a levels in TNBC cells and found a decreased efficiency in mammosphere formation of these cells in comparison to scrambled (Scr) control cells (Supplementary Fig S8D) (see next page), suggesting a possible non-cell autonomous mechanism of CD81 promoted self-renewal.

Furthermore, when the expression of the EV secretion-regulating gene Rab27a was downregulated in MDA-MB-231 cells by siRNA mediated transfection, the mammosphere formation was significantly reduced (The authors observe that CD81 depletion profoundly impairs primary tumor development and metastasis in pre-clinical models. In fact, the defect in metastasis appears to be secondary to the robust impairment in tumorigenesis. What is the fate of CD81 KO cells in mammosphere assays and transplant models? Do CD81 KO cells have reduced viability and/or proliferation in vitro and in vivo? Can the defects in mammosphere formation and tumorigenesis in CD81 KO cells be rescued via re-introduction of wtCD81? What about mutant CD81 that is deficient for CD44 binding? These studies will help to delineate the role of CD81 in primary tumor development and whether interaction with CD44 is require for this process.

Thanks for the comments and questions. We included the data on slightly slower cell proliferation of human CD81KO and mouse Cd81KO cells (no obvious cell death) in Supplementary Fig S1E-F (see next page). We have also reintroduced wtCD81 in two sets of distinct vectors into CD81KO cells and observed wtCD81 in both HA and GFP vectors rescued the defects of CD81KO cells while mutant CD81 deficient for CD44 binding failed to do so, demonstrating the role of CD81-CD44 interaction in promoting self-renewal.

*In the authors model, they show CD81 interactions with CD44 facilitating EV secretion which enhances self-renewal and CD44 alone facilitates tumor cell clustering. However, they show that CD81 can also binding CD44. Does the CD81-CD44 interaction also serve to facilitate clustering between tumor cells? *

Yes, we conducted the tumor cell clustering experiment and added the new Supplementary Fig 12D which demonstrates that CD81 WT rescues the clustering of CD81KO cells and the CD81 truncated mutant does not.

*Minor Points:

In a number of figures (e.g. Fig, 1C; Fig 3E, H) the authors show representative immunoblots but there are no indications in the legends of how many times the experiment was performed. Presumably at least 3 independent experiments were performed. The authors should also include quantification of these data to support that these effects are reproducible and significant. All data points should be plotted within graphs so that the reader can note the distribution of the data. *

Thanks for the suggestions. We have included all raw data points in all quantified bar graphs and quantified the western blots from at least three independent experiments. Representative is shown below (next page).

*While the manuscript was generally well written, there are a couple of grammatical mistakes that can be fixed. *

We appreciate the positive comment and have corrected the grammatical mistakes by native /professional writers.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

The manuscript of Ramos et al describes studies focused on understanding the relationship between membrane proteins CD44 and CD81 in the self-renewal of tumor initiating cells and breast cancer metastasis. They demonstrate CD81 and CD44 are protein binding partners, and that loss of either component affects the clustering and self-renewal of tumor initiating cells. Proteomics from CD44 and CD81 deficient cells revealed alterations in levels and phosphorylation status of proteins associated with endocytosis and lysosomes, suggesting that impaired tumor cell self-renewal was related to perturbations in the endolysosome pathway. Digging deeper into this connection, the authors found that loss of CD44 and CD81 resulted in increased vacuolization inside cells and greater release of EVs. Moreover, EVs from wildtype, but not CD44 or CD81 deficient cells, modestly rescued the self-renewal defects observed in cells lacking CD81. Importantly, Ramos et al find that CD81 is enriched within aggregated CTCs from TNBC patients with metastatic disease and its levels are negatively correlated with patient survival. Consistent with an important role for CD81 in TNBC, the authors find that depletion of CD81 severely impairs tumorigenesis and metastasis in pre-clinical models using PDX, 4T1 and MDA-MB-231 cells. Altogether, these observations lead the authors to propose that CD81 partners with CD44 to promote exosome biogenesis, tumor cluster formation, and lung metastasis in triple negative breast cancer.

Major Comments:

The authors suggest that CD44 and CD81 interact and colocalize inside breast cancer cells. However, staining data presented shows very modest co-localization and immunoprecipitation experiments only employed beads as a negative control. To reinforce their conclusions, the authors should quantify co-localization using standard methods (Mander's or Pearson's coefficients). In addition, the authors should perform negative control immunoprecipitation experiments with antibodies against a target protein not expected to interact with CD81 to show that CD44 binding is specific.

From TEM and quantification in figures 3A,B the authors conclude that there is increased vacuolization with cells. They suggest that purple arrows point to multivesicular endosomes and yellow arrows vacuoles. In fact, the electron dense organelles indicated by purple arrows look more like lysosomes, whereas the yellow arrows appear more like early endosomes/endocytic vesicles. The authors should reassess their vesicle classification system and also provide a breakdown of the proportions of these structures within the graph.

Given that CD81+/CD44+ EVs are proposed to drive the aggregation and self-renewal of tumor initiating cells, it is somewhat counterintuitive that CD44 and CD81 KO cells show increased levels of EV secretion relative to WT controls. Additionally, it is perplexing that CD44 and CD81 secretion in EVs is unaffected (or my even increase) in knockout cells despite the fact these membrane proteins are supposed to interact and are mutually required for proper expression/localization. How do the authors reconcile these potentially contradictory observations?

The EV rescue experiments in figure 3H, I show recovery of mammosphere formation in CD81 KO cells treated with EVs from WT cells, suggesting that WT EVs are sufficient to rescue self-renewal. However, it's unclear from their studies whether exosome secretion of CD81/CD44 is actually necessary for aggregation and self-renewal phenotypes. Since CD81 has also been shown to be important for the trafficking of membrane proteins, the loss of self self renewal could relate to cell autonomous alterations in vesicular trafficking in CD81 KO cells. To reconcile between these possibilities, the authors should evaluate how depletion of factors necessary for CD81 secretion (e.g. Rab27a (PMID: 26305877; supplemental data), ESCRT components (PMID: 32049272), or others?) affects mammosphere formation. In either case the results would be extremely interesting and help to determine whether self-renewal is controlled by CD81 via cell autonomous or non-cell autonomous mechanisms.

The authors observe that CD81 depletion profoundly impairs primary tumor development and metastasis in pre-clinical models. In fact, the defect in metastasis appears to be secondary to the robust impairment in tumorigenesis. What is the fate of CD81 KO cells in mammosphere assays and transplant models? Do CD81 KO cells have reduced viability and/or proliferation in vitro and in vivo? Can the defects in mammosphere formation and tumorigenesis in CD81 KO cells be rescued via re-introduction of wtCD81? What about mutant CD81 that is deficient for CD44 binding? These studies will help to delineate the role of CD81 in primary tumor development and whether interaction with CD44 is require for this process.

In the authors model, they show CD81 interactions with CD44 facilitating EV secretion which enhances self-renewal and CD44 alone facilitates tumor cell clustering. However, they show that CD81 can also binding CD44. Does the CD81-CD44 interaction also serve to facilitate clustering between tumor cells?

Minor Points:

In a number of figures (e.g. Fig, 1C; Fig 3E, H) the authors show representative immunoblots but there are no indications in the legends of how many times the experiment was performed. Presumably at least 3 independent experiments were performed. The authors should also include quantification of these data to support that these effects are reproducible and significant.

All data points should be plotted within graphs so that the reader can note the distribution of the data.

While the manuscript was generally well written, there are a couple of grammatical mistakes that can be fixed.

Significance

Given that the complex mechanisms contributing to self-renewal of tumor initiating cells and disease progression in TNBC are poorly understood, this work is both clinically important and addresses the biology from a unique perspective. In addition, this reviewer finds the study to be reasonably logical and well controlled. Nevertheless, there are a number of areas where additional experiments are necessary in order to strengthen the conclusions of the manuscript and provide support for the authors model.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The article by Liu et al describes a novel partner interaction between two transmembrane proteins, CD44 and CD81, which associate in tumor cells and in extracellular vesicles released by these cells. The authors have previously shown, by KD or KO strategies, that CD44 expression by tumors participates in their migration and metastatic properties. Here the authors show that CD81 deletion in tumors similarly decreases both tumor local growth and formation of metastasis. Since CD81 has been described on small EVs, the authors also evaluate the effect of CD44 or CD81 (or both) KO in the capacity of tumors to release small EVs, and in the ability of these EVs to reconstitute migration of tumor cells depleted of CD81 in vitro. Finally, the authors show a negative correlation between CD81 expression and clinical outcome in breast cancer patients, and observe clusters of circulating tumor cells expressing CD81 and CD44 in these patients.

The experiments are generally well performed and convincing, except that all mouse experiments seem to have been performed only once, with small groups of mice (between 4 and 6 with several sites of tumor injection per mouse). Reproducing at least once these experiments should be shown.

Another unsatisfying aspect is the claim that CD44 and CD81 are specifically required for secretion of the subtype of EVs called exosomes, which forms in intracellular multivesicular endosomes. This claim is based on the (wrong) assertion that CD81 (together with CD9, CD63 and TSG101) is an exosome marker, based on ref 46,47 published in 2006 and 2011: the field has evolved a lot since then, and it is now becoming clear that CD63 (which normally accumulates in multivesicular endosomes) may be enriched in exosomes, but that neither CD9 nor CD81 are, since they mainly localize at the plasma membrane, and thus probably are released more prominently in small microvesicles (see Kowal, J., et al. (2016). Proc Natl Acad Sci U S A 113: E968. and Mathieu, M., et al. (2021). Nat Commun 12(1): 4389.). In figure 3, the authors show empty internal compartments in cells with deleted CD44 or CD81, which (if these compartments are altered MVBs) should then lead to fewer exosomes recovered extracellularly. However, the authors observe instead more particles recovered from these cells. When analyzing the composition of these EVs, the authors show maybe a decrease in CD63 in EVs from both CD44 and CD81 ko cells, but contradictory effects on the presence of Lamp2 (which is a more convincing marker of late endosome/lysosome-derived exosomes): increased in CD44 but decreased in CD81 ko. These results, however, are not really interpretable, since the authors show only a single Western Blot, thus no evidence of reliable changes in protein composition. In any case, I would suggest that the authors do not, in the current state of the article, try to claim any specific subcellular origin of the EVs affected by CD44 and CD81. In addition, the EVs are only quantified by the vesicle flow cytometry established by the authors, but then, in functional assay, and in Western blots, EVs are quantified in terms of proteins. It would be important to show if CD44 and CD81 ko also decrease the amount of proteins recovered in the EV preparations, and the number of particles quantified by NTA, as they apparently decrease the number of events detected by vesicle flow cytometry, or not, and if not, why did the authors chose to show only the vesicle flow cytometry results (this is not a very commonly used technic in the field). Another important point to change is the presentation of results as bar graphs: all such graphs must be replaced by graphs showing the position of individual replicates, to illustrate the reproducibility of the presented results (eg fig1B, S1C, 3C, 3G-H, 5C, 5E, 5G, 6B, 6H, 6K, etc). (as explained in Weissgerber et al, Plos Biol 2015 13(4): e1002128).

Finally, a somehow frustrating aspect of the paper is that the link between the observed effect of CD44 or CD81 ko on EV release and their in vitro functions on mammosphere formation (fig3), and the effect on in vivo tumor growth and metastasis (fig 5-6) end up as two separate observations (the authors rightly do not claim that impaired exosome release in vivo is responsible for the impaired metastasis). The observation also of clustered CD81 and CD44+ circulating tumor cells in patients (fig4) is also somehow a separate observation. Thus there are several stories put together in this article.

Significance

These results are interesting as showing a novel molecule whose high expression may promote tumor progressions (CD81, as a cluster with CD44). The novelty lies in the functional interaction between CD44 and CD81 leading to the pro-metastatic effect described. Interaction between CD44 and CD81 had been previously observed in a generic proteomic study of EVs (Perez-Hernandez, D., et al. (2013). J Biol Chem 288: 11649), but not more explored in terms of consequent functions. A pro-metastatic effect of CD81 expression in tumors, especially TNBC, has also been recently demonstrated (Vences-Catalan, F., et al. (2021). Proc Natl Acad Sci U S A 118.). These two articles should be quoted in the current paper.

My field of expertise is extracellular vesicles, and their roles in cancer progression. I can only judge superficially the modeling part of the article,

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript, Dr. Huiping Liu and colleagues investigate the role of CD81 in breast cancer metastasis, cancer stem cells, and extracellular vesicles (EVs). CD81 is a tetraspanin protein that has unclear roles in cancer. The authors discover that CD81 can form a complex with CD44 on cell surface and instigate cell clustering (important for CTC dissemination), self renewal and metastasis using multiple cell lines and xenografts. Multi-omic studies led them to CD81-regualted EVs, which they found to play critical roles in driving stemness and cell clustering. Furthermore, they found that CD81 is co-expressed with CD44 and affects patient outcomes. Overall, the novelty of the work is high, and the amount of data is impressive and of high quality. The experiments presented are well controlled and rigorous. However, there are some concerns as listed below:

Major:

1. The conclusion of EVs made by TICs to reprogram non-TICs into TICs is based on adding large amounts of EVs isolated from WT tumor cells. The amount may not be achievable in physiological conditions. Further, if the EVs released by TICs are sufficient to reprogram neighboring tumor cells, this event would be expected to self-propagate and all the tumor cells would become TICs.
2. The conclusion that CD44 switched from membrane to intracellular locations adherent vs. suspension cultures is not well supported by the data presented in Fig 1E. It appears that the cells expressing membrane CD44 did not change much, but the cells population expressing intracellular CD44 expanded in the suspension culture.
3. Fig 6E-H experiments are a bit flawed. Metastasis degree is unlikely to be proportional to the primary tumor size, so it is not proper to normalize the metastatic burden to the primary tumor weight. These data can be removed without significantly weakening the major conclusions.
4. Fig 3J appears to have different lanes pasted together. ¬¬Therefore, it is unclear whether the results could be compared, calling into the question of the statements involving the proteins studied including OCT4, pSTAT3 and FAK.
5. The authors state that EVs were comparable between WT and 44ko and 84ko cells partly based on the western blot presented in Fig 3E. While TSG101 appears similar, the other markers appear quite different. This needs explanation.
6. The rationale for different cut-offs in Fig 4A-C is unclear. Auto cut-offs should not be used to maximize the detection of difference. The exact public dataset used should be stated.

Minor

1. The description regarding the screen leading to CD81 is confusing. Including a diagram may help.
2. Fig 1E, the phrase "ratio of cells expressing" is bit confusing. Did the authors mean % of cells expressing their indicated markers?
3. Cryo-EM shows detects impaired membrane integrity of EVs of 81ko but 44ko. Yet, Fig 2 shows that ko of either 44 or 81 disruptions the localization and therefore the function of the other. Explanation is needed for why 44ko did not affect 81 regulation of EVs.
4. Fig S8C is not a robust correlation analysis. A scatter plot should be presented.
5. There are a few typos. For example "cytoplastic membranes" should be cytoplasmic membranes. The following sentence is awkward: "Among the siCD81-upregulated genes, SEMA7a, a glycosylphosphatidylinositol membrane anchor promoting osteoclast and blood cell differentiation (58, 59), when further depleted in CD81KO cells, siSEMA7a partially rescued or restored mammosphere formation in these cells (Supplementary Figure S2D-F),.."

Significance

This work has high significance in CTC formation and metastasis. While CD44 has been reported by this group previously in 2019 Cancer Discovery to promote clustering formation and collective dissemination, this work takes a step further to identify a new interaction partner of CD81 on cell membrane and implicate the regulation of EVs as mechanism to promote stemness. This manuscript has broad interest in the Cancer Research community. This reviewer expertise is breast cancer and rodent models of breast cancer.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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

Learn more at Review Commons

---

Reply to the reviewers

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

Summary: Klein and colleagues generate an ES cell model system with inducible FACT depletion to understand how loss of FACT affects gene regulation in ES cells. They find that FACT is critical for ES cell maintenance through multiple mechanisms including direct regulation of key pluripotency transcription factors (Sox2, Oct4, and Nanog), maintaining open chromatin at enhancers, and regulated enhancer RNA transcription. The paper is well-written, the experiments are generally well-controlled and appropriately interpreted and placed within the context of the field.

We appreciate the Reviewer’s support of this manuscript.

Major comments: 1. In general, the ChIP-seq and CUT&RUN data are not that similar. Although correlation seems reasonable (S2A), looking at the heatmaps in S2B/C these seem pretty different. It's not very clear if this is a case where CUT&RUN has higher specificity (and signal-to-noise, which is very clear from example tracks) or if these two methods are picking up biologically different sites. Could the authors include some overlap analysis of peaks and comment on these discrepancies. Looking at the example tracks in Figure 2B, it seems likely that prior SPT16 and SSRP1 ChIP-seq were relatively high-noise.

We have identified overlapping peaks between the two techniques, and while CUT&RUN identified substantially more peaks overall, percentage of peaks shared between datasets were relatively consistent (1-6% of total) between the individual ChIP-seq datasets and the CUT&RUN dataset (Response Figure 1). We note that the biological classes identified through all datasets were remarkably consistent (Fig. 2D), and therefore attribute the discrepancies to the greater number of reproducible peaks called from CUT&RUN data. As discussed in the paper, peak calling algorithms designed for the specific data types were used, and therefore peak calling could also contribute to differences.

Response Figure 1. ChIP-seq and CUT&RUN peak overlap. Pie chart depicting the unique and overlapping peaks called from V5-SPT16 CUT&RUN data and FACT ChIP-seq data. These data are included in the revised manuscript (as a new Figure panel 2E). Peaks must have been identified in at least two technical or biological replicates.

Are motifs described in Figure 2E CUT&RUN only, and do prior ChIP-seq experiments also identify these motifs?

The motifs shown in Figure 2E (now 2F) are indeed CUT&RUN peaks only. We were unable to confidently assign enriched motifs to the ChIP-seq datasets (the most enriched motifs were approximately p = 10-18). By analyzing all SPT16 ChIP-seq peaks, rather than only intersected SPT16 ChIP-seq peaks, we were able to identify motifs recognized by two of the top three CUT&RUN motif hits (SOX2 and OCT4/SOX2/TCF/NANOG); however, enrichment was quite poor (p = 10-3). By limiting the analysis to intergenic regions, we were able to identify strong enrichment for motifs recognized by CTCF and BORIS (p = 10-58 and 10-51, respectively). As validation, we also called motifs from peak files published as supplementary material to the original Tessarz lab manuscript but were still unable to confidently call motifs (all p > 10-7 for SPT16 peaks, p > 10-15 for SSRP1 peaks). Related to major comment 1, we suspect that the weak motif enrichment is due to high background in ChIP-seq datasets compared to CUT&RUN datasets.

The authors state that FACT depletion affects eRNA transcription and measured this using TT-seq. The analysis in Figure 3B seems to be all the different types of sites looked at together (genes, PROMPTs, etc). Is there evidence that eRNAs specifically are regulated by FACT loss.

We apologize for the confusion and have clarified that Figure 3B (now 3A) is referring to mRNAs only in the text and figure. Our analysis of eRNA regulation by FACT is predominantly contained within Fig. 4B (TT-seq from DHSs, but no histone mark overlap assessment), Supp Fig. S4 (as in Fig 4B, but at DHSs overlapping H3K27ac or H3K4me1), Fig. 5E (FACT localization to putative enhancers, defined as in S4), and Fig. 6D (ATAC-seq demonstrating loss of accessibility at putative enhancers upon FACT depletion). Based on these results, we believe there are many eRNAs specifically misregulated by FACT loss and that potential direct targets (based on change in depletion and containing FACT binding) are in Fig 5E.

Could these be compared to DHS sites that lack FACT binding to support a direct role for FACT at these sites?

We appreciate the suggestion and have performed this analysis (see Response Figure 2). Relatedly, we analyzed putative silencers, defined as DHSs marked by H3K27me3, for FACT binding and expression changes (measured by TT-seq) following FACT depletion (Supp Fig. S7). As expected, FACT does not bind these putative silencer DHSs and transcription does not markedly increase or decrease from these regions after FACT depletion. Complicating the matter, FACT binds at many DHSs, even those that did not to meet our stringent peak-calling criteria (see Response Figure 2, middle cluster).

__Response Figure 2. Overlap between FACT binding sites and gene-distal DHSs. __Individual clusters are sorted by V5-SPT16 binding. Clusters were assigned based on direct overlap between called V5-SPT16 peaks and assigned gene-distal DHSs. Overall, 17.6% of DHSs overlapped a FACT peak identified in at least one CUT&RUN replicate (8.5% of DHSs overlapped a peak present in multiple replicates).

One mechanism proposed for how FACT regulates enhancers is that it is required for maintaining a nucleosome free area, and when FACT is depleted nucleosomes invade the site (Figure 7). It wasn't clear if they compared distal DHS sites were FACT normal bound to those without FACT binding in the MNase experiments, which could help support the direct role or specificity of FACT in regulating those enhancers (or a subset of them).

We have subset the V5-SPT16 CUT&RUN peaks and distal DHSs into groups and have identified increased nucleosome occupancy after depletion at both FACT-bound and FACT-unbound DHSs suggesting both direct and indirect regulation (Fig. 6A, D). There is disruption to nucleosome arrays at non-FACT-bound DHSs (although more modest relative to the FACT bound locations), and therefore we speculate that a nucleosome remodeler is involved downstream of FACT (possibly CHD1, per recent work out of Patrick Cramer and François Robert’s labs, among others).

1. Data quality for nucleosome occupancy was a little strange (Figure 7F), where the two clones had very different MNase patterns at TSS sites. Could the authors comment on why there is such a strong difference between clones here.

We agree that the trends identified by visualizing differential MNase-seq signal near TSSs do not fully replicate; however, in examining the nondifferential MNase-seq heatmaps, we see a more expected distribution (see new Figure 7A). Per our newly-added Supp Fig. S9B, all MNase-seq replicates had a pairwise Pearson correlation value of at least 0.73 (SPT16-depleted clone 1/rep 1 vs untagged rep 3), and the vast majority of samples had pairwise correlations of above 0.85, suggesting that these discrepancies are not due to strong differences in sequencing depth or MNase-protected regions. We therefore suspect that the clonal distinctions are a result of different background occupancy of nucleosomes near the TSS, resulting in an array with increased occupancy in one clone and more generalized increased occupancy in the other clone. We also added the MNase-seq data over TSSs in a non-differential form in Fig 7A, and believe the difference between the clones is due to the differential analysis, and have commented accordingly in the revised manuscript.

More details on some of the analysis steps would be really helpful in evaluating the experiments. Specifically, was any normalization done other than depth normalization? I ask this because the baseline levels for many samples in metaplots look quite different. For example, see Figure 7B where either clone 1 has a globally elevated (at least out 2kb) ratio of nucleosome in the IAA samples relative to the EtOH, or there is some technical difference in MNase. One suggestion is to look at methods in the CSAW R package to allow TMM based normalization strategies which may help.

We appreciate the suggestion – we have expanded our explanation of normalization methodology in the paper. We initially used quartile and RPGC normalizations to attempt to mitigate technical differences in MNase-seq data. Size distribution plots did not suggest differences in MNase digestion between samples, and neither quartile/RPGC nor TMM-based normalization fully resolved this issue. Because our ATAC-seq datasets agree with the general trends identified by MNase-seq (which are consistent, despite technical differences between clones), we do not believe that the differences constitute true biological difference, but rather experimental noise.

1. I appreciated the speculation section, and the possible relationship between FACT and paused RNAPII is interesting. While further experiments may be outside the scope of this work and I am not suggesting they do them, I am wondering if others have information on locations of paused RNAPII in ESC that would allow them to test if genes with paused RNAPII have a special requirement for FACT that they could use their current data to assess.

We agree that experiments to test the relationship between paused RNAPII and FACT are an intriguing next step, and plan to dissect those in the near future.

Minor comments: 1. When describing the peaks found in the text related to Figure 2 they refer to 'nonunique' peaks. Does this mean the intersection of the independent peak calls? Could they clarify this.

We apologize for the confusion and have clarified in the text that nonunique peaks does indeed refer to the intersection of independent peak calls (now specified on manuscript page 8, line 15).

In the text they refer to H3K56ac data in S2D and I don't see that panel. The color scheme for the 1D heatmaps (Figure 5A) is tough to appreciate the differences. I'd suggest something more linear rather than this spectral one might be easier to see.

We apologize for the confusion and removed the remaining H3K56ac-related data and references in the text. We appreciate the suggestion regarding the 1D heatmap color scheme and have adjusted the colors to a linear (white à red) scheme.

For the 2D heatmaps of binding, could they include the number of elements they are looking at for each group?

We appreciate the suggestion and have included numbers of elements visualized wherever applicable in the figure panels and legends.

1. Also for 2D heatmaps, I think the scale is Log2 (IAA/EtoH), but could they confirm that and include it in the figure?

We apologize for the confusion; the only heatmaps displaying log2(IAA:EtOH) are those in Fig. 6; for those panels, we have clarified the scale in the figure and legend.

Reviewer #1 (Significance (Required)):

• The use of degrader based approaches to depleting a protein allows refined kinetic and temporal assays which I think are important. Several papers showed a rapid invasion of nucleosomes after SWI/SNF loss using these kinds of approaches and revealed surprisingly fast replacement of SWI/SNF. This paper is consistent with those models, showing that another remodeler behaves the same, suggesting there may be general requirements for active chromatin remodeling to maintain the expression of these genes. It also highlights a key gap in how specificity works to target these enzymes remains somewhat unknown.

• This work will be of interest to those studying detailed mechanisms of gene regulation. Compared to some other chromatin regulators, FACT is understudied and so this work will allow comparison between different chromatin remodeling complexes.

• My experience: chromatin, gene regulation, cancer, genomics

We appreciate the thorough review and hope that we have sufficiently addressed your concerns.

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

The authors propose that the FACT complex can regulate pluripotency factors along with their regulatory targets through non-genic locations. They find that acute depletion of FACT leads to a "reduction" in pluripotency in mouse embryonic stem cell by disrupting transcription of master regulators of pluripotency. They also show FACT depletion affected the transcription of gene distal regulatory sites, but not silencers. They also stated that SPT16 depletion resulted in both, a reduction of chromatin accessibility and increase of nucleosome occupancy over FACT bound sites.

Overall the study appears technically well executed. The use of an Auxin induced depletion system is a good model to study the acute effects of FACT depletion. However, I have a number of concerns relating to specificity and interpretation of the results that need to be addressed. We appreciate the careful review and have addressed your comments below:

Major points o Authors claimed that depletion of the FACT complex "triggers a reduction in pluripotency". As evidence supporting this statement they present images of alkaline phosphatase assays of a time course performed upon depletion of FACT. These experiments indeed show that ESCs are destabilized in the absence of SPT16. However, some key questions regarding the phenotype remain unresolved: o What is are the kinetics of expression of selected naïve pluripotency and early differentiation markers? Are differentiation markers upregulated, consistent with normal differentiation upon FACT depletion?

We appreciate the suggestion and have emphasized the decrease in pluripotency factor expression, accompanied by an increase in differentiation marker expression across all three germ layers. We graphed 7 pluripotency factors and 7 differentiation markers for each germ layer; generally speaking, pluripotency factors are decreased while differentiation markers are increased (Response Figure 4; pluripotency factors are included in the new Fig. 3B, while differentiation markers are included in the new Supp Fig. S3 F-H).

We have also performed an immunocytochemistry (ICC) timecourse, per Reviewer 3’s suggestion. This ICC timecourse allows us to orthogonally assess decreased pluripotency factor expression, to pair with the OCT4 Western blot shown in Supp Fig. S1B. These new ICC data are shown in the new Fig. 1D and included here for convenience (Response Figure 5). In addition, we have added alkaline phosphatase staining at 12 hours of depletion to Fig. 1C.

__Response Figure 4. Plots of DESeq2 analysis across experimental timecourse. __Shown are lineage markers denoting: A. Pluripotency B. Endoderm C. Mesoderm and D. Ectoderm. Generally, expression of pluripotency factors decrease over time, while differentiation markers of each lineage increase over time. These data are shown in Figure 3B and Supplemental Figure S3F-H.

__Response Figure 5. Immunocytochemistry timecourse depicting DAPI staining (left panels, blue) and OCT4 immunofluorescence (right panels, green). __Images are representative of plate-wide immunofluorescence changes.

O Is only ESC identity affected or does loss of FACT impair viability also of cells that have exited pluripotency? To address this, growth curves and/or cell cycle analysis upon FACT depletion could be performed. Alternatively, the authors could utilize surface markers to distinguish naïve pluripotent form differentiated cells in the cell cycle analysis experiments to identify a potential differential response of pluripotent and differentiated cells to FACT depletion.

We have performed a growth curve with FACT depletion as suggested; as the two points are related, we will explain further below:

o Another key question is whether it is only the metastable pluripotent state of ESCs in heterogeneous FCS/LIF conditions which is affected by FACT loss, and whether cells cultured in the more homogeneous and more robust 2i-LIF conditions can tolerate FACT removal. If that is indeed the case it would enable the authors to address one main concern I have with this manuscript, which is that it is nearly impossible to distinguish the direct effect of FACT loss from differences induced by differentiation (and maybe cell death, see comment above). This is a critical concern that needs to be addressed and discussed appropriately.

We apologize for the confusion – all original experiments for this project were performed in the presence of LIF as well as GSK and MEK inhibitors CHIR99021 and PD0325091, respectively (2i+LIF conditions). To address the reviewers question, we have now performed a timecourse growth assay under both LIF-only and 2i+LIF conditions (Response Figure 6 and new Supp Fig S1F), and as suggested by the reviewer, observe a stronger effect of FACT depletion on cell viability in LIF-alone (FACT-depletion results in ~90% death within ~24 hours, with differences in growth observed by 12 hours) than in 2i+LIF (FACT-depletion results ~80% death within 48 hours, with differences in growth observed starting around 18 hours). Overall, ES cells in LIF alone are indeed more sensitive to FACT loss, supporting our decision to perform the experiments throughout the manuscript in 2i+LIF conditions.

LIF alone LIF + 2i

Response Figure 6. __Growth assays in LIF (left) and 2i+LIF (right) conditions. __Cells were treated with either EtOH or 3-IAA and counted at the indicated times. Viability was assessed using trypan blue exclusion. Error bars indicate standard deviation for biological triplicate experiments.

o A further major concern is about the specificity of the effect of FACT depletion. The authors claim that FACT is required to maintain pluripotency. From the data presented this is unclear. FACT appears to be part of the general transcription machinery in ESCs. It appears generally associated with active promoters and active genes, according to the data in this manuscript. Whether there is any specific link to pluripotency remains to be shown. It is unclear how enrichment analyses have been performed. If they haven't been performed using a background list of genes actively transcribed in ES cells, they will obviously show enrichment of ESC specific GO categories, because ESCs express ESC specific genes robustly expressed in ESCs?

We apologize for the confusion and have updated our methods section to include more comprehensive details on our pathway enrichment analyses. We have confirmed that pluripotency-related categories are still highly enriched in FACT-regulated DEGs, even when using a background dataset of all transcribed genes, per our TT-seq datasets (baseMean ≥ 1 in DESeq2 output).

In line with this: the authors show that FACT bound loci well overlap with Oct4 bound regions. But which proportion of FACT targets loci are actually Oct4 bound too?Is FACT binding exclusive to Oct4 regulated enhancers and promoters? In other words, will FACT be recruited to all actively transcribed genes in ES cells? In that case, a specific effect on pluripotency network regulation cannot be claimed.

We appreciate the suggestion, and have added the number of OCT4/SOX2/NANOG-bound FACT peaks and vice versa in the text and legend of Fig 3E-F. We have also summarized this information in Response Table 1, below (and included these data as Table 2 in the revised manuscript).

OCT4 peaks

Sox2 Peaks

Nanog Peaks

Any of OSN

V5 Peaks

8,544

5,948

5,307

9,682

OSN Peaks

45,476

19,211

16,817

52,899

% of OSN peaks bound by FACT

18.33%

30.72%

31.40%

17.91%

% of V5 peaks bound by pluripotency factor(s)

52.41%

36.85%

32.94%

59.63%

V5-bound promoters

4,261

2,719

2,327

4,452

OSN-bound promoters

6,550

1,542

666

6,948

V5- and OSN-bound promoters

2,040

801

343

2,202

OSN-bound gene-distal peaks

38,926

17,669

16,151

45,938

V5-bound gene-distal OSN peaks

6,504

5,147

4,964

7,480

__Response Table 1. Overlapping CUT&RUN and ChIP-seq peaks shared between OCT4, SOX2, NANOG, and V5-SPT16 under various stratifications. __Shown are numbers or percentages of peaks overlapping between V5 and OSN. The last column are peaks containing any of OCT4, SOX2, and/or NANOG. The first four rows include all peaks, regardless of location, and the last five rows are broken down by promoter (as defined by an annotated mRNA) or gene-distal location (defined by a minimum of +/- 1kb from a gene).

Of the 45,865 OCT4 peaks, 3,688 are located at promoters, and 1,209 of these peaks are bound by V5-SPT16 (32.8%). Inversely, 13,228 of 42,177 gene-distal OCT4 peaks are called as SPT16-V5 peaks in at least one CUT&RUN replicate (31.36%), suggesting a relationship between OCT4 binding and FACT binding, which has long been identified with genic transcription, but has roles extending beyond gene-proximal regulation. We observe similar trends with NANOG and SOX2.

o It is disappointing that neither raw data (GEO submission set to private) nor any Supplemental Tables containing differentially expressed transcripts and ChIP or Cut and Run peaks and associated genes were made available. This strongly reduces the depth of review that can be performed.

We apologize if the reviewer token in the cover letter was not accessible. The GEO datasets (including differentially expressed transcripts, raw fastq files, and analyzed datasets) will be made public upon publication; in the meantime, the GEO entry (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE181624) can still be accessed using the previously provided reviewer token: wvkvwmwynjeffux.

o To what extent do FACT bound loci overlap with genes differentially expressed 24h after FACT depletion? This analysis would help determine the direct targets of FACS regulation.

We appreciate the suggestion. This analysis can be found in the original Figure S6, broken down by FACT-repressed (expression increased upon FACT depletion), unchanged, and FACT-stimulated (expression decreased upon FACT depletion) DESeq2 results (ordered left-to-right, respectively). Figure S6A-C shows that V5-SPT16 binding is enriched, but not exclusive to, genes with FACT-regulated expression, while Fig. S6D-F shows TT-seq data for each group, sorted by log2-fold change assigned by DESeq2.

o The paper mainly relies on NGS analysis. Therefore, it is crucial that authors show as Supplemental Material some basic QC of these data. PCA analyses to show congruency of replicates are the minimum requirement.

We appreciate the suggestion and have included a new Supp. Fig S9, with pairwise comparative Pearson correlation scatterplots and heatmaps for replicates in each dataset, in addition to the scatterplots shown for CUT&RUN and ChIP-seq data in the original Supp Fig. S2A.

o Did the authors perform any filtering for gene expression levels before analysis? Are genes in the analysis robustly expressed in at least one of the conditions?

We apologize for the confusion. Due to the sensitive nature of TT-seq and the germ layer-inconsistent pattern of cell differentiation following FACT depletion, we did not perform filtering for gene expression prior to any analyses. For the vast majority of genes analyzed, however, we are able to identify transcription via TT-seq, even in those that do not significantly change expression upon FACT depletion (see Supp Fig S6E). As discussed above, we did include a cutoff for expressed genes in our revised pathway analysis.

o Wherever p values were reported for enrichment analyses, adjusted p values should be used

We apologize for the oversight; the p values were in fact adjusted p values and have updated the text and figures to make it explicit that the adjusted p values were used wherever applicable.

o I cannot follow the logic used by the authors to explain discrepant results from Chen et al about the role of FACT in ESCs. Chen et al showed that FACT disruption by SSRP1 depletion is compatible with ESC survival and leads to ERV deregulation. The authors of the present study attribute these differences to potential FACT independent roles of SSRP1. However, I would assume that if there are indeed FACT independent roles of SSRP1, then the phenotype of SSRP1 KOs in which FACT and other processes should be dysfunctional should be even stronger than a plain FACT KO. This needs a proper and careful explanation.

We apologize that our discussion of FACT-independent roles of SSRP1 was not clear and have clarified our wording in the text (page 4, line 49 – page 5, line 4)in the revised manuscript); we intended to reconcile the results of Chen et al. 2020 with Goswami et al. 2022 and Cao et al. 2003; despite SSRP1 knockout viability in embryonic stem cells, SSRP1 knockout is lethal in mice between 5-40 weeks and general SSRP1 knockout is lethal 3.5 days post-conception (per Goswami et al. 2022). We therefore posit that the general requirement for SSRP1 may be due to distinct roles from those carried out by the FACT complex in ES cells, as discussed by Spencer et al. 1999, Zeng et al. 2002, Li et al. 2007, and Marciano et al. 2018.

We note that our findings are in agreement with papers from the Gurova lab and others in that depletion of mRNA or protein of SPT16 leads to concomitant loss of SSRP1; we therefore do not expect total SSRP1 loss to have a stronger effect than SPT16 depletion. We therefore expect, and confirmed via Western blotting (Figure 1B, Supplemental Figure 1), that depletion of SPT16 leads to loss of both FACT subunits, and therefore all FACT subunit activity, complex-dependent or -independent.

Also, did the authors observe any evidence for ERV deregulation upon acute SPT16 depletion?

We did indeed observe ERV deregulation upon SPT16 depletion. When reviewing our TT-seq datasets, 7.1% of ERVs were derepressed, while 2.4% decreased in expression upon 24h FACT depletion (mm10 ERVs sourced from gEVE, Nakagawa and Takahashi, 2016). Further, we identified increased chromatin accessibility after FACT depletion at annotated LTR elements, as shown in the table below (Response Table 2). Here we are displaying the calculated enrichment score for accessibility detected at these locations. A negative value indicates lower accessibility than expected by region size, while a positive score indicates that reads are more enriched than expected at the indicated region.

ATAC-seq enrichment score for locations losing accessibility with FACT depletion

3h

6h

12h

24h

LTR Enrichment

-1.445

-1.299

-0.917

-0.559

Intergenic Enrichment

-6.046

-4.765

-3.926

-2.972

Promoter Enrichment

3.335

2.789

2.726

2.233

ATAC-seq enrichment score for locations gaining accessibility with FACT depletion

3h

6h

12h

24h

LTR Enrichment

-1

-0.436

1.103

1.13

Intergenic Enrichment

-1

0.134

0.435

0.236

Promoter Enrichment

-1

-3.585

1.171

1.39

__Response Table 2. Changes in ATAC-seq peak enrichment for selected regions, annotated via HOMER. __At regions differentially accessible between SPT16-depleted and SPT16-undepleted samples, regions were assigned to an annotated genomic feature using HOMER annotatePeaks.pl and assigned an enrichment score based on the ratio of ATAC-seq signal to region size. Over time, LTR elements become more enriched among the ATAC-seq peaks both gaining and losing accessibility, indicating a role for FACT in maintaining LTR accessibility.

We do wish to note, however, that Lopez et al. 2016 identified SPT16-independent regulation of LEDGF/HIV-1 replication by SSRP1, and therefore cannot rule out effects on ERV dysregulation due to SSRP1 loss that accompanies SPT16 depletion.

Minor points o Figure S2A is very small and resolution is low. Page 10: "...while all four Yamanaka factors (Pou5f1, Sox2, Klf4, and Myc) and Nanog were significantly 24 reduced after 24 hours (Fig. 3A, S3A-B)". No data for myc is being shown.

We apologize for the figure resolution and have included a larger image. Because pairwise comparative scatterplots are not space-efficient, we opted to display the Pearson correlations for the datasets including more samples (TT-seq and ATAC-seq timecourses) as heatmaps in the new Supp Fig S9. We have added Myc labeling to the volcano plot (now in Fig. 3A) and included a trace of Myc expression over time to the new pluripotency factor graph in Fig. 3B.

o Are the two bands in the middle in figure 1B is supposed to be a ladder? This should be clarified.

We thank the reviewer for noticing this and apologize for the oversight.

o Figure 3C- This Figure is complicated to read. Also, information appears redundant with the Table 1, I recommend to remove this panel.

We have moved the panel to the supplement (now Supp Fig. S3A). While the information is somewhat redundant with Table 1, we chose to include the former panel 3C as a visual representation of the consistent deregulation over depletion time across transcript categories.

o Figure 6 and figure 7 could be presented in one single figure since both aspects are complementary and target related aspects.

While we thank the reviewer for this suggestion, we do not believe that the information contained in Figs. 6 and 7 can effectively be conveyed in a single figure. While both figures focus on chromatin accessibility and nucleosome occupancy, Fig. 6 is designed to address the changes in chromatin accessibility over time, while Fig.­­­ 7 is more relevant to the biological mechanism through which FACT co-regulates targets of the core pluripotency network (OCT4/SOX2/NANOG) after 24 hours of depletion.

o Are the authors certain that the effects observed are directly linked to the FACT complex in contrast to FACT independent roles of SPT16, if any exist? The experiment to address this would be to deplete SSRP1 and investigate whether the effects are identical, which would be the hypothesis to be tested.

We thank the reviewer for this suggestion. We did attempt to create additional SSRP1-AID-tagged lines; however, generating these lines proved to be technically challenging, and comparison of the FACT-dependent and -independent roles of the individual subunits is beyond the scope of this work. Further complicating the matter, SSRP1 is effectively depleted within 6 hours of 3-IAA addition in SPT16-AID lines due to the interdependence of FACT subunits. We again thank the reviewer for their suggestion and will consider this work for a future study.

Reviewer #2 (Significance (Required)):

My expertise is pluripotency and GRNs.

I would judge the significance of the study as presented as low, mainly because at this moment it remains unclear what FACT indeed does concerning regulation of pluripotency.

We respect the reviewer’s opinion and hope that our revisions have made more clear how the FACT complex prevents nonspecific differentiation from occurring, thereby maintaining pluripotency and self-renewal in embryonic stem cells. Importantly, neither untagged cells treated with 3-IAA nor tagged cells treated with vehicle display the growth defects, loss of pluripotency factor expression, increased differentiation marker expression, phenotypic evidence of differentiation, and reduced alkaline phosphatase staining that the FACT-depleted cells do, highlighting a key requirement for FACT in pluripotent cells. Beyond this, we believe the novel gene distal regulatory role we have identified for FACT presents an exciting new role for this complex in gene regulation.

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

In this manuscript, Klein, et al. addressed function of FACT complex in mouse ESCs, using cut&run, TT-seq, ATAC-seq, MNase-seq, together with Auxin-mediated FACT degradation system. The authors first reported that efficient and acute depletion of SPT16 with the Auxin-mediated degradation system resulted in over 5,000 up- and 5,000 down-regulated genes within 24 hours, including down-regulation of pluripotent gens. Then, they demonstrated that many of FACT binding sites overlap with Oct4, Sox2, Nanog binding sites by Cut&Run, and those loci increase nucleosome occupancy 24 hour after removal of FACT.

The Auxin-mediated degradation system seems to be working very well (while I would like to see an over exposed version of Western blotting), and efficient degradation might explain the different phenotypes from the previous reported phenotypes by shRNA and the chemical inhibitor, which might not deplete FACT function completely and/or might have off-target effects. The Cut&Run data also have much sharper peaks than previously reported SSRP1, SPT16 ChIP-seq data. Doing ATAC-seq, MNase-seq upon removal of FACT is excellent. WIth the excellnet degradation system, depletion of FACT resulted in loss and gain of gene expression and differentiation. However, unfortunately it was not very clear to me what was the direct consequences of FACT removal and its mechanisms, waht was consequence of differentiation.

We appreciate the kind words regarding our choice and execution of techniques and the reviewer’s time spent on this manuscript. We have made a number of changes to the manuscript in order to clarify the direct role of FACT and the consequences of FACT loss on embryonic stem cells.

Although we did not develop the blots for a longer period when we performed the Westerns, we have artificially overexposed our V5-SPT16 Western blot from Figure S1 (in Adobe Illustrator) to highlight the more subtle bands at later depletion timepoints; we hope that this helps to clarify the effectiveness of the degron system.

Response Figure 7. V5-SPT16 Western blot with adjusted exposure. We manually adjusted the entire blots’ exposures using Adobe Illustrator. L indicates ladders, and the timecourse depletion is shown above the blot.

In my opinion, doing many of the analysis 24 hours after FACT depletion, where differential expressed (coding) genes (DEGs) are >10,000 (Table 1)), is too late to understand what the direct consequences are. Seeing 214 up- and 174 down-regulated DEGs 6 hours after FACT depletion, I do agree that FCAT seems to do both suppression and activation of target genes. It could have been really interesting to investigate what % of FACT bindign sites change chromatin accesibility and nucleosome occupancy at that time point, if those loci are close to any of the up- or down-regualted DEGs.

We appreciate the suggestion and have included more information regarding the percentage of FACT binding sites with altered chromatin accessibility, as well as included some analyses to address the directness of FACT’s contribution to DEGs at all timepoints (see Supp Figs S4, S6). We would like to note that, we performed the TT-seq and ATAC-seq experiments at 0, 3, 6, 12, and 24 hours post 3-IAA treatment in order for us to explore the progressive change in both the transcriptome and chromatin accessibility, with only the MNase-seq limited to 24 hours. As originally shown in our Sankey plot in Supp Fig 4, we see a progressive change in expression for a small subset of genes over our timecourse running from 0-24 hours, with the largest effect observed at 24 hours, once the FACT protein levels are almost entirely depleted. Similarly, we see a progressive change in ATAC-seq signal over the same regions, with the strongest effects over the same regions visible at 24 hours post-depletion. Due to our observation that SPT16 is not depleted at 3 or 6 hours, with significant depletion seen at 24 hours (see Response Figure 7) and because we intended to study the FACT complex’s role in preventing differentiation, we were most interested in the effects at 24 hours of depletion, which allow us to analyze both the disruption of pluripotency factor expression and the facilitation of differentiation marker expression across all three germ layers (see Response Figure 4).

Followings are reasons of above my judgement and suggestions to improve the manuscript.

Major points 1. Figure 1. ALP staining is not very sensitive way to evaluate ESC differentiation. I recommend Immunofluorescence for pluripotency genes (NANOG and/or SOX2) and quantification. Or present changes of pluripotency genes in graphs over time course from RNA-seq data.

We appreciate the suggestions and have taken both into account. We have included a new panel in Figure 3 (new 3B) to display the changes of pluripotency factor expression over our timecourse. We have also included some data showing differentiation factors as part of a response to Reviewer 1, which can be found above (Response Figure 4). In addition, we performed immunocytochemistry to examine OCT4 abundance over a depletion timecourse and have added a 12-hour to our alkaline phosphatase assay to address the sensitivity of differentiation over time (Figure 1C, D and Response Figure 5).

1. Fig 2A, 3E, 3F. How many transcription start sites are shown here? (Throughout the manuscript, it is hard to know how many loci are shown in the heatmaps. It should be described within the figures)

We apologize for the omission and have added numbers of loci shown to relevant Figure panels throughout the paper.

It is nice to see nascent transcription high sites have high FACT binding, but can you also show actual nascent transcription of these loci as a heatmap, before and after FACT depletion? These heatmaps should be shown in a descending order of FACT Cut&Run signalling, as FACT binding is important in this manuscript.

We appreciate the suggestion and have plotted those data below (see Response Figure 8).

Response Figure 8. Nascent transcription from sites with high FACT binding. Top: TPM-normalized TT-seq signal after 12-hour treatment, oriented to mRNA strand and plotted as entire mRNA length, ± 500 bp. Data are sorted by SPT16 CUT&RUN signal over 1kb upstream of annotated TSSs. n = 1 over 22,597 rows (RefSeq Select mRNAs). Bottom: TPM-normalized TT-seq signal after 24-hour treatment, oriented to mRNA strand and plotted as entire mRNA length, ± 500 bp. Data are sorted by SPT16 CUT&RUN signal over 1kb upstream of annotated TSSs. n = 3 (mean) over 22,597 rows (RefSeq Select mRNAs).

Strong FACT binding sites have strong transcription. Is FACT really supressing transcription?

We agree that it is very difficult to disentangle FACT function due to its binding correlation with transcription; however, we see a clear trend of FACT binding at promoters that are sensitive to FACT depletion (Supp Fig. S6A/D and C/F). Intriguingly, the genes that see the greatest derepression by DESeq2 analysis are those that are directly bound by FACT (per ChIP-seq and CUT&RUN; Supplemental Figure S6A/D), while the greatest decrease in expression occurs at genes that are less bound by FACT (Supp Fig S6C/F). In our opinion, this trend lends credence to both direct repression by FACT and distal gene regulation. We note that others (e.g., Kolundzic et al. 2018) have shown direct repression of gene expression by FACT, in line with that aspect of our data.

1. Fig 3ABD. It is more important to show 3h, 6h 12 h time points. The same apply to Fig 4. What %, how many of DEGs (coding and non-coding) at each time point had FACT binding nearby in ESCs?

We agree that the early timepoints are important and have added volcano plots to the supplemental material for earlier timepoints, with genes of interest specifically annotated. We have also examined pluripotency and differentiation markers at earlier timepoints, per other reviewers’ suggestions, and have included the percentage of DEGs with nearby FACT binding in the manuscript. Specifically, 2013 replicated V5 peaks (out of 16,054; 12.54%) occurred within 1000 bp of a RefSeq Select TSS.

Timepoint

Total DEGs (up)

V5-bound DEGs (up)

Total DEGs (down)

V5-bound DEGs (down)

3h

58

16 (27.59%)

5

1 (20%)

6h

214

38 (17.76%)

174

31 (17.82%)

12h

1366

123 (9.00%)

1932

281 (14.54%)

24h

5398

431 (7.98%)

5000

663 (13.26%)

__Response Table 3. Table of DESeq2-assigned DEGs that are bound by SPT16-V5. __To be defined as V5-SPT16-bound, a DEG must have SPT16-V5 binding within 1000 bp upstream of its RefSeq-select annotated TSS.

We believe that these earliest depletion timepoints are in line with FACT-mediated gene regulation occurring distal to the regulated genes’ promoters.

Fig 3EF. Interesting data and the overlap between SPT16 binding sites and pluripotency binding sites look very strong. But it is difficult to know what % is overlapping from these figures.

We appreciate the difficulty in quantifying the overlap between pluripotency factor binding sites and FACT binding sites; we have added those data to the manuscript below Figure 3E for OCT4; for other pluripotency factors, these data can be found in Response Figure 9 and Response Table 1. Briefly, 18.33% of OCT4 ChIP-seq peaks are bound by V5-SPT16 and 52.41% of V5-SPT16 peaks are bound by OCT4. Interestingly, 34.6% of gene-distal OCT4 ChIP-seq peaks are bound by V5-SPT16, implying greater convergence between FACT and pluripotency factors at gene-distal sites, in line with known trends for OCT4 binding. Overall, 59.63% of V5-SPT16 peaks are co-bound by at least one of OCT4, SOX2, or NANOG.

Can you show 1 heatmap split into 3 groups, a. SPT16-V5 unique, common between SPT16-V5 and Oct4 ChIP-seq, Oct4 ChIP-seq unique, with indication of numbers each group has? Also make the same figures for Sox2 and Nanog. (E is less important. If the authors want, they can use the published FACT ChIP-seq data in the same loci.)

We appreciate the suggestion and have plotted V5-SPT16 CUT&RUN data and pluripotency factor ChIP-seq over unique and shared regions for OCT4 (top) SOX2 (middle) and NANOG (bottom). Interestingly, although some peaks in the non-overlapping cluster were not called as peaks by the algorithms’ threshold, one can observe that a subset do seem to have overlapping binding. We again appreciate the suggestion and think that this was an excellent way to display the data and have included these data as a new panel (Fig. 3E) but also show below in Response Figure 9.

Fig. 5. Basic information what % (how many) of SPT16-V5 CUT&RUN peaks belong to this 'enhancer' category is missing.

We apologize for the oversight and have added numbers to the figure and legend.

I am not sure the meaning of separating enhancers and TSS of coding genes in the analyses, though. If majority of SPT16-V5 CUT&RUN peaks overlap with Oct4 binding sites, it is not surprising that SPT16-V5 CUT&RUN peaks overlaps with ATAC-seq signal and enhancer marks.

We agree that it is unsurprising that V5-SPT16 overlaps with accessible chromatin and enhancers, given the extensive overlap with OCT4 ChIP-seq peaks. We wanted to emphasize our novel finding of gene-distal FACT binding, given the more established trend of binding at promoters.

1. Fig 6A. I could not figure out what % of DHSs overlaps with FACT binding sites.

We have added this percentage to Fig 5C and included an analysis of altered chromatin accessibility in a new Table 3 (page 20). Briefly, 11,234 replicated V5-SPT16 peaks (out of 16,043; 70%) directly overlap a gene distal DHS. Orthogonally, 11,234 DHSs (out of 132,555; 8.5%) directly overlap a V5-SPT16 peak.

I do not see the point of showing DHSs which do not overlap with FACT binding sites.

In agreement with Reviewer 1, we believe that it is important to include FACT-unbound DHSs for a clearer understanding of the direct vs indirect effects of FACT depletion. We have condensed some of these data into a single heatmap, clustered between FACT-bound DHSs, non-FACT-bound DHSs, and FACT-bound non-DHS sites to streamline the information (now shown in Fig 3E).

Response Figure 9. Heatmaps of clustered SPT16 and OSN binding data. Shown are clustered heatmaps depicting V5-SPT16 CUT&RUN binding overlapping ChIP-seq peaks for OCT4 (top) SOX2 (middle) and NANOG (bottom). In each set of heatmaps the top cluster is pluripotency factor-unique, the middle cluster is shared, and the V5-unique cluster is on the bottom. Each cluster is sorted by descending strength of V5-SPT16 binding (CUT&RUN). Clusters were assigned by directly overlapping peaks.

How ATAC-seq signal changes upon depletion of FACT at FACT binding sites (Fig 6B) is important. Can you explain why ATAC-seq signals increase at the FACT binding site flanking regions (across +/- 2kb) where FACT binding is strong (without changing the chromatin accessibility at the FACT binding sites)? Perhaps authors need to show actual ATAC-seq track with EtOH or 3-IAA treatment over ~10kb regions flanking FACT binding sites. It is difficult to understand what is happening seeing only the changes (ratio) of ATAC-seq read counts, how big the differences are.

We agree that the local window and ratio of ATAC-seq signal somewhat muddles the true biological trends. We have plotted non-differential ATAC-seq signal for each SPT16-AID clone over V5 binding sites, ±10 kb, to more accurately depict the local chromatin status (shown below in Response Figure 10). There is an apparent trend at V5-SPT16 CUT&RUN peaks of accessible chromatin, and this high local accessibility very likely contributes to the high ATAC-seq signal immediately flanking V5 binding sites; over the binding sites themselves, however, FACT depletion consistently triggers decreased accessibility (see Fig. 6).

Can you identify differentially open loci based on 3-IAA- and Et-OH treated ATAC-seq data at each time point, and then how many of them overlap with FACT binding sites? There are a few tools to identify differential open regions with ATAC-seq data. That could help to understand the direct roles of FACT binding.

We appreciate the suggestion and have performed this analysis using a combination of PEPATAC and HOMER (see Response Tables 4-6 below). FACT depletion leads to the following accessibility changes:

3-hour

6-hour

12-hour

24-hour

Decreased accessibility

220 (0.35%)

3,713 (5.99%)

6,885 (11.11%)

8,441 (13.62%)

Increased accessibility

2 (0.00%)

12 (0.02%)

276 (0.45%)

6,031 (9.73%)

Response Table 4. Accessibility changes over consensus ATAC-seq peaks. Consensus ATAC-seq peaks were defined per PEPATAC standards (peaks called by MACS2 in (n/2)+1 samples, irrespective of condition.

3-hour

6-hour

12-hour

24-hour

Decreased accessibility

848 (1.64%)

1870 (3.51%)

2525 (4.83%)

4,092 (7.90%)

Increased accessibility

107 (0.21%)

283 (0.55%)

534 (1.03%)

2,449 (4.73%)

Response Table 5. Accessibility changes over regions bound by V5-SPT16.

Response Figure 10. ATAC-seq data shown over a 20kb window. Heatmaps depicting non-differential ATAC-seq data over FACT binding sites for SPT16-AID clones 1 (top) and 2 (bottom). Data are sorted by V5-SPT16 binding strength.

All

3-hour

6-hour

12-hour

24-hour

Decreased accessibility

3,294 (2.46%)

3,175 (2.37%)

3,636 (2.71%)

7,018 (5.23%)

Increased accessibility

102 (0.08%)

313 (0.23%)

1,797 (1.34%)

5,975 (4.45%)

V5-bound DHSs (11,234 total)

3-hour

6-hour

12-hour

24-hour

Decreased accessibility

1 (0.01%)

9 (0.08%)

96 (0.85%)

2006 (17.86%)

Increased accessibility

5 (0.04%)

28 (0.25%)

71 (0.63%)

87 (0.77%)

Response Table 6. Accessibility changes over gene-distal DHSs and over only FACT-bound gene-distal DHSs.

Together with Fig 1A and Fig 6C, do they mean the more FACT binding, the more transcription (Fig 1A). Also the higher transcription rate, the more increased chromatin accessibility upon depletion of FACT (Fig 6C)?

While we do see that FACT binding correlates with transcription and with FACT-dependent chromatin accessibility, we do not wish to make the argument that FACT binding alone is indicative of high transcription, nor that transcription is necessarily the deciding factor in FACT-depleted chromatin accessibility changes. We do want to note that transcriptional disruption is a likely contributor to increased chromatin accessibility in the absence of FACT as it pertains to paused RNAPII, as speculated in our discussion, but that experiments to truly test this hypothesis are beyond the scope of this work. That being said, in response to Reviewer 1, we did assess the potential correlation of FACT binding to locations with greater paused RNAPII (Response Figure 3) and see a connection. We are excited to explore this more in future work.

Perhaps plotting nascent transcripts at 12hr, 24 hr of FACT depletion next to these heatmaps might show if it colleates with transcription changes as well?

We appreciate the suggestion, and have included this plot in Response Figure 8, sorted by FACT binding to gene promoters; however, we find it difficult to visualize differences in transcription with non-differential heatmaps.

Sites losing chromatin accessibility (bottom half of Fig 6C) seem not to have FACT binding (bottom half of Fig 1A), thus it is likely to be indirect effects. It is better to make figures focussing on 'direct effects'.

We agree that there are sites with reduced chromatin accessibility upon FACT depletion that are not bound by FACT; however, given the extensive binding of FACT at gene-distal regulatory regions (F2D, F4A, F5, F6A/D), we would suggest that these “indirect” effects are possibly the result of FACT-dependent gene-distal regulation.

Fig 1A and Fig 6C indicated that FACT binding sites (i.e. TSS) decrease chromatin accessibility. I thought it does not fit with the idea of increasing nucleosome occupancy. But actually the data (Fig 7F) shows that TSS does not show increased nucleosome occupancy unlike Fig 7A-E. In fact, Fig 6B showed that about bottom 50% of weaker V5 binding sites decreased chromatin accessibility at 24 hr, which fits with increased nucleosome occupancy in Fig 7A. But then if you looked at only top 50% of stronger V5 binding sites, which did not decrease chromatin accessibility, nucleosome occupancy did not change as well? Why don't you make heatmap of MNase-seq next to Fig 6B?

We have added heatmaps of non-differential MNase-seq data to Fig. 7A to address both concerns. Regarding Figure 6B, we note that the V5-SPT16 peaks themselves invariantly show decreased chromatin accessibility, and that it is the surrounding chromatin, not the V5-SPT16 peak itself, that shifts from increased to decreased chromatin accessibility at 12-24 hours of depletion. We would also like to clarify that the original heatmaps in Fig 6B were sorted by change in chromatin accessibility at 24h, rather than V5 binding.

We disagree that the TSSs do not show increased nucleosome occupancy in Fig. 7F, as there is an increase in signal above background directly over the TSS in both replicates, per the differential metaplot shown in Fig. 7B, that is specific to the AID-tagged lines. However, the two clones did show variable results. To address this, we have plotted the non-differential MNase-seq plots (Fig. 7A), which show more consistent trends; it appears that the transformation of the data into differential at this location was the cause of the slightly variable plots over TSSs.

1. I could not follow based on which data the model in Fig 8 is made. Again it is better to focus in the direct effects.

Thank you for the suggestion; we have updated our model to focus more on the direct effects.

Minor points. 10. Line 1 page 5, Kolundzic paper did not have MEF reprograming data. They reported human fibroblast reprogramming was enhanced by FACT KD.

We appreciate the correction and have clarified the language to specify that the work of Kolundzic et al. included human fibroblast reprogramming and Shen et al. performed MEF reprogramming.

1. Line 3, I disagree with "these data establish FACT as essential in pluripotent cells". One paper said FACT KD increased proliferation of mESCs, the other paper said chemical inhibition of FACT was necessary for passaging ESCs, but not proliferation. Importance of FACT in pluripotent cells was very unclear to me.

We have clarified our language to specify that pluripotent cells have a FACT dependency that differentiated cells do not. We note that we were unable to recapitulate a relationship between FACT and trypsinization/passaging of ES cells, suggesting a more nuanced role for FACT in pluripotent cells, in line with work from the Tessarz and Gurova labs.

Line 7 Page 7, reference the paper with the ChIP-seq data.

We apologize for the oversight and have added the reference.

Line 16, Page 7. It doesn't seem the the Cut&run and previously published ChIP-seq data agree well.. >50% look different. It is nothing the authors can do, but can you show venn diagram of peak overlap?

In response to Reviewer 1, we have generated Response Figure 1 where we display a pie chart of the overlap. In addition to displaying this again to the right in Response Figure 11 this, we have included another analysis below in Response Figure 11, to address this comment. Specifically, we have plotted peak overlaps as a Venn diagram to compare peaks identified in at least two experimental replicates from either the CUT&RUN or ChIP-seq data (left). We have also overlapped replicated peaks between the individual targets and displayed them as a pie chart (right; same as Response Figure 1). While the CUT&RUN data do display a greater signal:noise ratio and call far more peaks, we note that more peak conservation between experiments is relatively consistent (1-6%) between all datasets, including the ChIP-seq experiments profiling opposite factors.

Overall, we see strongly reproducible trends (albeit with less sharp definition in the ChIP-seq), complemented by highly similar biological feature assignment in Fig. 2D and Pearson correlation values of between 0.76 and 0.78 between SPT16 ChIP-seq and V5-SPT16 CUT&RUN (Supp Fig. S2A).

__Response Figure 11. Overlaps between SPT16-V5 CUT&RUN, SPT16 ChIP-seq, and SSRP1 ChIP-seq. __Called peaks were compared between V5-SPT16 CUT&RUN, SPT16 ChIP-seq, and SSRP1 ChIP-seq, using both our own analysis pipeline (left) and the peaks published with the original manuscript by Tessarz et al. (2018; right). While our ChIP-seq peak-calling appears to have applied more stringent thresholds, trends are generally agreeable.

Line 12, 22 page 10. Fig.3AB is 24 hrs. Do not match with the text.

We apologize for the error and have changed the references in the text to the new panel 3C.

1. Line 23, 24, page 10, Highlight Klf4 and Myc in the volcano plot.

We have added KLF4 and MYC annotation to the volcano plot in Fig. 3A, as well as plotted their log2FC over time in the new panel 3B.

1. Line 18, 19, page 16. This is not accurate statement. Sample 2 increased the accessibility at 6 hours. Sample 1 decreased, but even the control did so.

We apologize for the unclear wording; we intended to suggest that all timepoints after 6 hours (i.e., 12 and 24 hours) display decreased accessibility directly over the DHS. We have corrected the text.

1. Line 48-50, page 16. Two replicates show very different patterns. Difficult to agree with the statement based on the figure.

We agree that the differential replicate patterns are not ideal; however, both replicates display an increase in nucleosome-sized reads over the promoter region, consistent with our ATAC-seq results presented in Fig 6C. Size distribution plots did not suggest differences in MNase digestion between samples, and neither quartile/RPGC nor TMM-based normalization fully solved this issue. Because our ATAC-seq datasets agree with the general trends identified by MNase-seq (which are consistent, despite technical differences between clones), we do not believe that the differences constitute biological difference, but rather experimental noise. We have included a heatmap of non-differential MNase-seq signal around TSSs in Fig 7A to highlight the experimental reproducibility between replicates. Based on this analysis it appears that the transformation of the data into differential at this location was the cause of the slightly variable plots over TSSs.

1. Line 15, page 19. Where does "1.5 times" come from? which is 1.5 times more, and is that different from the proportion of those?

We apologize for the unclear reference to the altered transcripts in Table 1 and have changed our wording to be more precise.

1. Line 32, page 19. Is Fig S2B correct figure?

We appreciate the correction; the text should have referred to Fig. 4 and has been fixed.

Line 35-39, page 21. I understand FACT does not bind to silenced loci. If FACT does not bind, it is not surprising that expression from those loci does not change upon FACT deletion. I do not understand what the authors said.

We agree that a lack of binding and unchanged expression after FACT depletion at putative silencers are unsurprising; given FACT’s extensive genic and gene-distal binding, we wished to show a class of transcribed regions unbound by FACT as a control, to show that non-FACT-regulated transcription was not affected by FACT transcription. We have clarified our wording in the text to emphasize that a lack of change was expected at silencers.

Reviewer #3 (Significance (Required)):

Previously it has been shown that Oct4 physically interacts with the FAcilitates Chromatin Transactions (FACT) complex. Seemingly contradicting phenotypes have been reporting upon suppression of FACT function in the maintenance and induction of pluripotent cells. Mylonas has reported that knockdown of SSRP1, a component of FACT complex, increased ESC proliferation (2018). Shen has described that chemical inhibition of FACT complex affected passaging of ESCs, but proliferation was not affected without passaging. Kolundzic has found that both SSRP1 and SUPT16H, another component of FACT complex, enhance human fibroblast reprogramming into iPSCs (2018), while Shen has reported that chemical inhibition of FACT blocks mouse iPSC generation form MEFs.

My expertise lies on pluripotent stem cells and transcriptional regulations. I did like the Auxin-mediated FACT degradation system these authors used and acute depletion of FACT is an excellent way of evaluating FACT function in ESC, compared to previously published shRNA based knockdown or use of a chemical inhibitor. However, as I described above, it was not very clear what could the direct effects and I feel looking at 24 hours after depletion might be to late to address this question.

We appreciate the review and agree that acute depletion of FACT has great potential to understand the complex’s function in ES cells. We understand that the nature of gene-distal regulation does make it difficult to cleanly elucidate direct regulation, and hope that our revisions have clarified that our goal was to examine direct, gene-distal regulation, rather than indirect effects. We would like to note that we examined transcription and chromatin accessibility after 3, 6, 12, and 24 hours of 3-IAA treatment, with all these data included in the original manuscript, and saw minimal change (likely because FACT was not fully depleted until later timepoints); to capture the true biological effects of FACT depletion, we explored most thoroughly the 24 hour 3-IAA treatment to understand the downstream effects between FACT loss and cellular differentiation. However, we have expanded discussion and analyses of the earlier timepoints in this revised manuscript.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this manuscript, Klein, et al. addressed function of FACT complex in mouse ESCs, using cut&run, TT-seq, ATAC-seq, MNase-seq, together with Auxin-mediated FACT degradation system. The authors first reported that efficient and acute depletion of SPT16 with the Auxin-mediated degradation system resulted in over 5,000 up- and 5,000 down-regulated genes within 24 hours, including down-regulation of pluripotent gens. Then, they demonstrated that many of FACT binding sites overlap with Oct4, Sox2, Nanog binding sites by Cut&Run, and those loci increase nucleosome occupancy 24 hour after removal of FACT.

The Auxin-mediated degradation system seems to be working very well (while I would like to see an over exposed version of Western blotting), and efficient degradation might explain the different phenotypes from the previous reported phenotypes by shRNA and the chemical inhibitor, which might not deplete FACT function completely and/or might have off-target effects. The Cut&Run data also have much sharper peaks than previously reported SSRP1, SPT16 ChIP-seq data. Doing ATAC-seq, MNase-seq upon removal of FACT is excellent. WIth the excellnet degradation system, depletion of FACT resulted in loss and gain of gene expression and differentiation. However, unfortunately it was not very clear to me what was the direct consequences of FACT removal and its mechanisms, waht was consequence of differentiation.

In my opinion, doing many of the analysis 24 hours after FACT depletion, where differential expressed (coding) genes (DEGs) are >10,000 (Table 1)), is too late to understand what the direct consequences are. Seeing 214 up- and 174 down-regulated DEGs 6 hours after FACT depletion, I do agree that FCAT seems to do both suppression and activation of target genes. It could have been really interesting to investigate what % of FACT bindign sites change chromatin accesibility and nucleosome occupancy at that time point, if those loci are close to any of the up- or down-regualted DEGs.

Followings are reasons of above my judgement and suggestions to improve the manuscript.

Major points

1. Figure 1. ALP staining is not very sensitive way to evaluate ESC differentiation. I recommend Immunofluorescence for pluripotency genes (NANOG and/or SOX2) and quantification. Or present changes of pluripotency genes in graphs over time course from RNA-seq data.
2. Fig 2A, 3E, 3F. How many transcription start sites are shown here? (Throughout the manuscript, it is hard to know how many loci are shown in the heatmaps. It should be described within the figures) It is nice to see nascent transcription high sites have high FACT binding, but can you also show actual nascent transcription of these loci as a heatmap, before and after FACT depletion? These heatmaps should be shown in a descending order of FACT Cut&Run signalling, as FACT binding is important in this manuscript. Strong FACT binding sites have strong transcription. Is FACT really supressing transcription?
3. Fig 3ABD. It is more important to show 3h, 6h 12 h time points. The same apply to Fig 4. What %, how many of DEGs (coding and non-coding) at each time point had FACT binding nearby in ESCs?
4. Fig 3EF. Interesting data and the overlap between SPT16 binding sites and pluripotency binding sites look very strong. But it is difficult to know what % is overlapping from these figures. Can you show 1 heatmap split into 3 groups, a. SPT16-V5 unique, common between SPT16-V5 and Oct4 ChIP-seq, Oct4 ChIP-seq unique, with indication of numbers each group has? Also make the same figures for Sox2 and Nanog. (E is less important. If the authors want, they can use the published FACT ChIP-seq data in the same loci.)
5. Fig. 5. Basic information what % (how many) of SPT16-V5 CUT&RUN peaks belong to this 'enhancer' category is missing. I am not sure the meaning of separating enhancers and TSS of coding genes in the analyses, though. If majority of SPT16-V5 CUT&RUN peaks overlap with Oct4 binding sites, it is not surprising that SPT16-V5 CUT&RUN peaks overlaps with ATAC-seq signal and enhancer marks.
6. Fig 6A. I could not figure out what % of DHSs overlaps with FACT binding sites. I do not see the point of showing DHSs which do not overlap with FACT binding sites. How ATAC-seq signal changes upon depletion of FACT at FACT binding sites (Fig 6B) is important. Can you explain why ATAC-seq signals increase at the FACT binding site flanking regions (across +/- 2kb) where FACT binding is strong (without changing the chromatin accessibility at the FACT binding sites)? Perhaps authors need to show actual ATAC-seq track with EtOH or 3-IAA treatment over ~10kb regions flanking FACT binding sites. It is difficult to understand what is happening seeing only the changes (ratio) of ATAC-seq read counts, how big the differences are. Can you identify differentially open loci based on 3-IAA- and Et-OH treated ATAC-seq data at each time point, and then how many of them overlap with FACT binding sites? There are a few tools to identify differential open regions with ATAC-seq data. That could help to understand the direct roles of FACT binding.
7. Together with Fig 1A and Fig 6C, do they mean the more FACT binding, the more transcription (Fig 1A). Also the higher transcription rate, the more increased chromatin accessibility upon depletion of FACT (Fig 6C)? Perhaps plotting nascent transcripts at 12hr, 24 hr of FACT depletion next to these heatmaps might show if it colleates with transcription changes as well? Sites losing chromatin accessibility (bottom half of Fig 6C) seem not to have FACT binding (bottom half of Fig 1A), thus it is likely to be indirect effects. I beleive it is better to make figures focussing on 'direct effects'.
8. Fig 1A and Fig 6C indicated that FACT binding sites (i.e. TSS) decrease chromatin accessibility. I thought it does not fit with the idea of increasing nucleosome occupancy. But actually the data (Fig 7F) shows that TSS does not show increased nucleosome occupancy unlike Fig 7A-E. In fact, Fig 6B showed that about bottom 50% of weaker V5 binding sites decreased chromatin accessibility at 24 hr, which fits with increased nucleosome occupancy in Fig 7A. But then if you looked at only top 50% of stronger V5 binding sites, which did not decrease chromatin accessibility, nucleosome occupancy did not change as well? Why don't you make heatmap of MNase-seq next to Fig 6B?
9. I could not follow based on which data the model in Fig 8 is made. Again it is better to focus in the direct effects.

Minor points.

1. Line 1 page 5, Kolundzic paper did not have MEF reprograming data. They reported human fibroblast reprogramming was enhanced by FACT KD.
2. Line 3, I disagree with "these data establish FACT as essential in pluripotent cells". One paper said FACT KD increased proliferation of mESCs, the other paper said chemical inhibition of FACT was necessary for passaging ESCs, but not proliferation. Importance of FACT in pluripotent cells was very unclear to me.
3. Line 7 Page 7, reference the paper with the ChIP-seq data.
4. Line 16, Page 7. It doesn't seem the the Cut&run and previously published ChIP-seq data agree well.. >50% look different. It is nothing the authors can do, but can you show venn diagram of peak overlap?
5. Line 12, 22 page 10. Fig.3AB is 24 hrs. Do not match with the text.
6. Line 23, 24, page 10, Highlight Klf4 and Myc in the volcano plot.
7. Line 18, 19, page 16. This is not accurate statement. Sample 2 increased the accessibility at 6 hours. Sample 1 decreased, but even the control did so.
8. Line 48-50, page 16. Two replicates show very different patterns. Difficult to agree with the statement based on the figure.
9. Line 15, page 19. Where does "1.5 times" come from? which is 1.5 times more, and is that different from the proportion of those?
10. Line 32, page 19. Is Fig S2B correct figure?
11. Line 35-39, page 21. I understand FACT does not bind to silenced loci. If FACT does not bind, it is not surprising that expression from those loci does not change upon FACT deletion. I do not understand what the authors said.

Significance

Previously it has been shown that Oct4 physically interacts with the FAcilitates Chromatin Transactions (FACT) complex. Seemingly contradicting phenotypes have been reporting upon suppression of FACT function in the maintenance and induction of pluripotent cells. Mylonas has reported that knockdown of SSRP1, a component of FACT complex, increased ESC proliferation (2018). Shen has described that chemical inhibition of FACT complex affected passaging of ESCs, but proliferation was not affected without passaging. Kolundzic has found that both SSRP1 and SUPT16H, another component of FACT complex, enhance human fibroblast reprogramming into iPSCs (2018), while Shen has reported that chemical inhibition of FACT blocks mouse iPSC generation form MEFs.

My expertise lies on pluripotent stem cells and transcriptional regulations. I did like the Auxin-mediated FACT degradation system these authors used and acute depletion of FACT is an excellent way of evaluating FACT function in ESC, compared to previously published shRNA based knockdown or use of a chemical inhibitor. However, as I described above, it was not very clear what could the direct effects and I feel looking at 24 hours after depletion might be to late to address this question.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The authors propose that the FACT complex can regulate pluripotency factors along with their regulatory targets through non-genic locations. They find that acute depletion of FACT leads to a "reduction" in pluripotency in mouse embryonic stem cell by disrupting transcription of master regulators of pluripotency. They also show FACT depletion affected the transcription of gene distal regulatory sites, but not silencers. They also stated that SPT16 depletion resulted in both, a reduction of chromatin accessibility and increase of nucleosome occupancy over FACT bound sites.

Overall the study appears technically well executed. The use of an Auxin induced depletion system is a good model to study the acute effects of FACT depletion. However, I have a number of concerns relating to specificity and interpretation of the results that need to be addressed.

Major points

• Authors claimed that depletion of the FACT complex "triggers a reduction in pluripotency". As evidence supporting this statement they present images of alkaline phosphatase assays of a time course performed upon depletion of FACT. These experiments indeed show that ESCs are destabilized in the absence of SPT16. However, some key questions regarding the phenotype remain unresolved:
• What is are the kinetics of expression of selected naïve pluripotency and early differentiation markers? Are differentiation markers upregulated, consistent with normal differentiation upon FACT depletion?
• Is only ESC identity affected or does loss of FACT impair viability also of cells that have exited pluripotency? To address this, growth curves and/or cell cycle analysis upon FACT depletion could be performed. Alternatively, the authors could utilize surface markers to distinguish naïve pluripotent form differentiated cells in the cell cycle analysis experiments to identify a potential differential response of pluripotent and differentiated cells to FACT depletion.
• Another key question is whether it is only the metastable pluripotent state of ESCs in heterogeneous FCS/LIF conditions which is affected by FACT loss, and whether cells cultured in the more homogeneous and more robust 2i-LIF conditions can tolerate FACT removal. If that is indeed the case it would enable the authors to address one main concern I have with this manuscript, which is that it is nearly impossible to distinguish the direct effect of FACT loss from differences induced by differentiation (and maybe cell death, see comment above). This is a critical concern that needs to be addressed and discussed appropriately.
• A further major concern is about the specificity of the effect of FACT depletion. The authors claim that FACT is required to maintain pluripotency. From the data presented this is unclear. FACT appears to be part of the general transcription machinery in ESCs. It appears generally associated with active promoters and active genes, according to the data in this manuscript. Whether there is any specific link to pluripotency remains to be shown. It is unclear how enrichment analyses have been performed. If they haven't been performed using a background list of genes actively transcribed in ES cells, they will obviously show enrichment of ESC specific GO categories, because ESCs express ESC specific genes Will results hold true if these experiments are performed using background lists of genes robustly expressed in ESCs? In line with this: the authors show that FACT bound loci well overlap with Oct4 bound regions. But which proportion of FACT targets loci are actually Oct4 bound too? Is FACT binding exclusive to Oct4 regulated enhancers and promoters? In other words, will FACT be recruited to all actively transcribed genes in ES cells? In that case, a specific effect on pluripotency network regulation cannot be claimed.
• It is disappointing that neither raw data (GEO submission set to private) nor any Supplemental Tables containing differentially expressed transcripts and ChIP or Cut and Run peaks and associated genes were made available. This strongly reduces the depth of review that can be performed.
• To what extent do FACT bound loci overlap with genes differentially expressed 24h after FACT depletion? This analysis would help determine the direct targets of FACS regulation.
• The paper mainly relies on NGS analysis. Therefore, it is crucial that authors show as Supplemental Material some basic QC of these data. PCA analyses to show congruency of replicates are the minimum requirement.
• Did the authors perform any filtering for gene expression levels before analysis? Are genes in the analysis robustly expressed in at least one of the conditions?
• Wherever p values were reported for enrichment analyses, adjusted p values should be used
• I cannot follow the logic used by the authors to explain discrepant results from Chen et al about the role of FACT in ESCs. Chen et al showed that FACT disruption by SSRP1 depletion is compatible with ESC survival and leads to ERV deregulation. The authors of the present study attribute these differences to potential FACT independent roles of SSRP1. However, I would assume that if there are indeed FACT independent roles of SSRP1, then the phenotype of SSRP1 KOs in which FACT and other processes should be dysfunctional should be even stronger than a plain FACT KO. This needs a proper and careful explanation. Also, did the authors observe any
• evidence for ERV deregulation upon acute SPT16 depletion?

Minor points

• Figure S2A is very small and resolution is low. Page 10: "...while all four Yamanaka factors (Pou5f1, Sox2, Klf4, and Myc) and Nanog were significantly 24 reduced after 24 hours (Fig. 3A, S3A-B)". No data for myc is being shown.
• Are the two bands in the middle in figure 1B is supposed to be a ladder? This should be clarified.
• Figure 3C- This Figure is complicated to read. Also, information appears redundant with the Table 1, I recommend to remove this panel.
• Figure 6 and figure 7 could be presented in one single figure since both aspects are complementary and target related aspects.
• Are the authors certain that the effects observed are directly linked to the FACT complex in contrast to FACT independent roles of SPT16, if any exist? The experiment to address this would be to deplete SSRP1 and investigate whether the effects are identical, which would be the hypothesis to be tested.

Significance

My expertise is pluripotency and GRNs.

I would judge the significance of the study as presented as low, mainly because at this moment it remains unclear what FACT indeed does concerning regulation of pluripotency.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Klein and colleagues generate an ES cell model system with inducible FACT depletion to understand how loss of FACT affects gene regulation in ES cells. They find that FACT is critical for ES cell maintenance through multiple mechanisms including direct regulation of key pluripotency transcription factors (Sox2, Oct4, and Nanog), maintaining open chromatin at enhancers, and regulated enhancer RNA transcription. The paper is well-written, the experiments are generally well-controlled and appropriately interpreted and placed within the context of the field.

Major comments:

1. In general, the ChIP-seq and CUT&RUN data are not that similar. Although correlation seems reasonable (S2A), looking at the heatmaps in S2B/C these seem pretty different. It's not very clear if this is a case where CUT&RUN has higher specificity (and signal-to-noise, which is very clear from example tracks) or if these two methods are picking up biologically different sites. Could the authors include some overlap analysis of peaks and comment on these discrepancies. Looking at the example tracks in Figure 2B, it seems likely that prior SPT16 and SSRP1 ChIP-seq were relatively high-noise.
2. Are motifs described in Figure 2E CUT&RUN only, and do prior ChIP-seq experiments also identify these motifs?
3. The authors state that FACT depletion affects eRNA transcription and measured this using TT-seq. The analysis in Figure 3B seems to be all the different types of sites looked at together (genes, PROMPTs, etc). Is there evidence that eRNAs specifically are regulated by FACT loss. Could these be compared to DHS sites that lack FACT binding to support a direct role for FACT at these sites?
4. One mechanism proposed for how FACT regulates enhancers is that it is required for maintaining a nucleosome free area, and when FACT is depleted nucleosomes invade the site (Figure 7). It wasn't clear if they compared distal DHS sites were FACT normal bound to those without FACT binding in the MNase experiments, which could help support the direct role or specificity of FACT in regulating those enhancers (or a subset of them).
5. Data quality for nucleosome occupancy was a little strange (Figure 7F), where the two clones had very different MNase patterns at TSS sites. Could the authors comment on why there is such a strong difference between clones here.
6. More details on some of the analysis steps would be really helpful in evaluating the experiments. Specifically, was any normalization done other than depth normalization? I ask this because the baseline levels for many samples in metaplots look quite different. For example, see Figure 7B where either clone 1 has a globally elevated (at least out 2kb) ratio of nucleosome in the IAA samples relative to the EtOH, or there is some technical difference in MNase. One suggestion is to look at methods in the CSAW R package to allow TMM based normalization strategies which may help.
7. I appreciated the speculation section, and the possible relationship between FACT and paused RNAPII is interesting. While further experiments may be outside the scope of this work and I am not suggesting they do them, I am wondering if others have information on locations of paused RNAPII in ESC that would allow them to test if genes with paused RNAPII have a special requirement for FACT that they could use their current data to assess.

Minor comments:

1. When describing the peaks found in the text related to Figure 2 they refer to 'nonunique' peaks. Does this mean the intersection of the independent peak calls? Could they clarify this.
2. In the text they refer to H3K56ac data in S2D and I don't see that panel. The color scheme for the 1D heatmaps (Figure 5A) is tough to appreciate the differences. I'd suggest something more linear rather than this spectral one might be easier to see.
3. For the 2D heatmaps of binding, could they include the number of elements they are looking at for each group?
4. Also for 2D heatmaps, I think the scale is Log2 (IAA/EtoH), but could they confirm that and include it in the figure?

Significance

• The use of degrader based approaches to depleting a protein allows refined kinetic and temporal assays which I think are important. Several papers showed a rapid invasion of nucleosomes after SWI/SNF loss using these kinds of approaches and revealed surprisingly fast replacement of SWI/SNF. This paper is consistent with those models, showing that another remodeler behaves the same, suggesting there may be general requirements for active chromatin remodeling to maintain the expression of these genes. It also highlights a key gap in how specificity works to target these enzymes remains somewhat unknown.
• This work will be of interest to those studying detailed mechanisms of gene regulation. Compared to some other chromatin regulators, FACT is understudied and so this work will allow comparison between different chromatin remodeling complexes.
• My experience: chromatin, gene regulation, cancer, genomics

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

Learn more at Review Commons

---

Reply to the reviewers

We were pleased with the overall very positive comments by the reviewers considering our study as convincing, well written and of importance for researchers not only in the fields of adrenal and gonadal disease, but also for endocrinology, tumorigenesis, sexual dimorphism and beyond.

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

Lyraki et.al. investigated the molecular basis of sexual dimorphism in adrenocortical hyperplasia. They use genetically modified mice, ectopically overexpressing R-spondin in SF1-experessing tissues (Sf1-Rspo1-gain-of-function [GOF]), causing overactivation of WNT signalling. Effects of R-spondin overexpression is nicely visualized using RNA Scope in situ hybridization and immunofluorescence; Ectopic R-spondin overexpression resulted in sex-specific adrenocortical hyperplasia (female) and degeneration (male), without observed changes in overall steroidogenic activity. Comparisons of the transcriptome of both male and female GOF and control mice of 4 weeks of age using RNA-sequencing demonstrated differential expression of genes involved in the cell cycle and immune response, in a sex-specific manner. Consequently, a BrdU proliferation assay confirmed increased proliferation in the inner cortex of female GOF mice but not male GOF mice. In contrast, male GOF mice showed increased CD68 staining, suggestive for increased macrophage infiltration. Next, using a sex reversal mice model, they show that the sexual dimorphism in adrenal hyperplasia is dependent on testicular androgens rather than chromosomal sex. When female GOF mice were treated daily with dihydrotestosterone, their adrenal weight and adrenal proliferation (BrdU assay) reduced. Moreover, knocking out the androgen receptor in the adrenal cortex of both female and male mice overexpressing R-spondin resulted in increased adrenal weight in male mice to a level comparable to the weight of female adrenals.

Altogether, I believe this manuscript convincingly shows that androgens act on adrenocortical cells and contribute to the sex-dependent susceptibility to adrenocortical hyperplasia. I have a few comments:

1. Can the authors please briefly specify why for Figure 1D, three different statistical tests were performed? Based on the graphs it seems probable that this decided based on tests for normality and unequal variances? Were tests for normality and equal variances performed?

We apologize for not further explaining the reasons for choosing different tests. Different versions of one-way ANOVA were performed for this figure to assess whether adrenal weight differs between groups of different sex and genotype. The choice of statistical test was based on different normality distributions and unequal variances. More specifically, most groups for the 3 weeks, 6 weeks, and 6 months timepoints displayed Gaussian distributions according to Shapiro-Wilk test. On the contrary, all the groups in the 12-month timepoint failed the Shapiro-Wilk normality test, prompting us to use a non-parametric test instead of ANOVA for this specific timepoint (the Kruskal-Wallis test). In addition, we tested for equality of variance using the Brown-Forsythe test. While the groups in the other timepoints display more or less equal variances, we confirmed significantly unequal standard deviations among the different groups in the 6-week timepoint (pThe manuscript will be amended accordingly.

Would it be an idea to perform a two-way ANOVA (or equivalent test) instead to study the interactive effect of ectopic R-spondin overexpression and sex on adrenal weight as well as BrdU expression?

We will perform the two-way ANOVA test as suggested.

Based on the images of Axin2 expression in Figure 3B, Axin2 expression seems higher in the outer cortex of the female adrenal, compared to the outer cortex of the male adrenal (and the opposite in the inner cortex). Can you please explain why the opposite trend is seen in Figure 3D for the outer cortex?

For our analysis we separated the adrenal into two regions with the outer cortex representing the region

Please clearly state the number of mice used for each experiment, either in the results section or in the methods section. At line 146, it is stated that n=6 mice "are analysed". At the other sentences just the n was provided. Therefore, I was wondering if not all mice were analysed here? How many mice were used? Also, at line 169 it is stated that 9 female Sf1-Rspo1GOF mice were used, while at line 170, only 8 aging Sf1-Rspo1GOF mice weer checked for carcinoma. Please explain.

We apologize for not having been more specific. The statement “n=6 mice analysed” in line 146 refers to the total number of animals analysed in this age group. We will clarify this by rephrasing the sentence: “At 6 weeks, all the male Sf1-Rspo1GOF adrenals (n=6) exhibited these degenerative changes to a varying degree….’

We will also review the entire manuscript and provide the number of samples used for each experiment.

Because of the low number, I am (based on this data only) not convinced that the presence of adrenocortical carcinoma in 1 of 8 female adrenals versus 0 of 8 male adrenals suggests sexual dimorphism (line 172).

The massive hyperplasia observed in female, but not in male mice in combination with the observed tumour in the female cohort strongly suggests that tumour formation in Rspo1GOF mice is sex specific. However, we agree with the reviewer that the low number of mice analysed does not allow us to draw a firm conclusion. We will therefore soften our statement and instead discuss that the increased proliferation is very likely to lead to a higher risk of developing adrenal tumours in females.

At line 159, please specify the age of the mice when steroid hormone levels were quantified.

Steroid hormone levels were quantified in 6 weeks old animals and this information will be included in the revised version of our manuscript.

I would suggest rephrasing sentence 186-188. Principle component analysis indeed nicely separates the samples on sex and Rspondin overexpression. However, I am not sure if one could say that sex was the “second most important factor” underlying the variation in gene expression.

We will change the respective phrase to: “Strikingly, sex was responsible for 26.3% of the variation in gene expression patterns among our experimental groups…”

The manuscript is well written, although the manuscript (especially the introduction) is quit lengthy and could be written more concisely.

We will review and shorten the manuscript.

Reviewer #1 (Significance (Required)): Of note, I am working in a specific endocrinological research niche and a bit distanced from this field. However, sexual dimorphism in (adrenal) diseases is a well-recognized phenomenon and the role of androgens and the androgen receptor in adrenal proliferation have previously been studied. Novel studies, like this study, are required to understand the mechanisms of sexual dimorphism in adrenal disease. This study is interesting especially (but not exclusively) for researchers working on adrenal or gonadal diseases or sexual dimorphism.

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

Summary: In this paper, the authors focus on the molecular cause of the sexual dysmorphism observed in the adrenocortical phenotype induced by moderate Wnt/b-catenin activation due to the gain of function of Rspo1 (Sf1-Rspo1GOF mice). Indeed, while the phenotype is similar before puberty, the females later develop adrenocortical hyperplasia with increase in proliferation and adrenal weight whereas the adrenal cortex of the males becomes thinner with ages and presents recruitment of macrophages and monocytes. Interestingly, the proliferation in the adrenal cortex is suppressed by androgen treatment in female and increases by Ar KO in male supporting an important role of androgen in the regulation of adrenocortical proliferation.

Major comments: - The authors describe an increase of Axin 2 expression in the outer part of the adrenal cortex in males compared to females but could the authors add a higher magnification of the outer cells in Sf1-Rspo1GOF to better illustrate this difference? Also, is this difference driven by an increase of dot per cell or an increase of the numbers of Axin 2 positive cells in this area?

High power views of the Axin2 RNA-Scope analysis will be included. The metric for the data that we provide is dots/cell and thus reflects the expression of Axin2 per cell. This information will be added to the revised version of the manuscript.

• The authors show that DHT injection in females leads to decrease proliferation but also increase the expression of Axin 2, a b-catenin target gene. As b-catenin is known to increase cell proliferation, did the authors look for instance, at the consequence on the expression of CCND1 or others b-catenin target genes involved in cell cycle regulation?

MYC expression (a b-catenin target gene) was analysed, but was not found to be changed upon DHT treatment. qPCR analysis for CCND1, as suggested by this reviewer, will be performed.

• Does DHT treatment in females or Ar KO in males affect the nucleo-cytoplasmic accumulation of b-catenin compared to controls? These results could help determine if Ar acts through b-catenin signaling or independently.

Measuring b-catenin activity in DHT treated females and AR KO is also a very good suggestion. We will analyse samples by immunofluorescent staining, and (as nuclear b-catenin is notoriously difficult to detect) by Axin2 expression which can be used as a readout for canonical b-catenin signalling.

• The authors did not mention if AS-RspoGOF males have macrophages and monocytes accumulation in the adrenal cortex as observed in Sf1-RspoGOF, this is an important information to better understand their origin and to know if they are due to mature or early embryonic dysregulation as AS-cre is activated later in time than Sf1-cre.

This is an interesting point and we will provide immunostaining for CD68 and IBA1 on AS-RspoGOF animals.

Minor comments: - In figure S8, could the authors add a H&E staining of AS-RspoGOF male adrenals in order to have all the controls?

This information will be added to Figure S8

• Based on the results presented here, the macrophage and monocyte observed in Sf1-RSPO1GOF males does not sound to be due to androgens as they are not observed in females treated with DHT. The authors should discuss these results and potential hypothesis explaining the macrophage recruitment.

A sentence discussing this point will be included.

• The authors should add a sentence or two to integrate their results in the ones previously published by Dumontet and collaborators (PMID: 29367455) regarding the consequences of DHT treatment on the regulation of Wnt signaling.

A sentence putting our study into context with findings in Dumontet et al. will be added.

Reviewer #2 (Significance (Required)):

Although the potential role of DHT in the sexual dysmorphism has previously been suggested in mouse models developing adrenocortical tumors, the demonstration of the direct role of the androgen receptor in this regulation demonstrated here is an important key in the understanding of molecular causes of this phenomenon that is observed both in human and mice adrenocortical tumorigenesis. Moreover, this model opens new perspectives to study the importance of the immune system in the regulation of the adrenal cortex homeostasis. Based on my expertise on adrenocortical homeostasis, I know that this manuscript will be of particular interest for researchers in the field of adrenocortical function and tumorigenesis but also, more generally for people working on the consequences of DHT and sexual dysmorphism.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In this paper, the authors focus on the molecular cause of the sexual dysmorphism observed in the adrenocortical phenotype induced by moderate Wnt/b-catenin activation due to the gain of function of Rspo1 (Sf1-Rspo1GOF mice). Indeed, while the phenotype is similar before puberty, the females later develop adrenocortical hyperplasia with increase in proliferation and adrenal weight whereas the adrenal cortex of the males becomes thinner with ages and presents recruitment of macrophages and monocytes. Interestingly, the proliferation in the adrenal cortex is suppressed by androgen treatment in female and increases by Ar KO in male supporting an important role of androgen in the regulation of adrenocortical proliferation.

Major comments:

• The authors describe an increase of Axin 2 expression in the outer part of the adrenal cortex in males compared to females but could the authors add a higher magnification of the outer cells in Sf1-Rspo1GOF to better illustrate this difference? Also, is this difference driven by an increase of dot per cell or an increase of the numbers of Axin 2 positive cells in this area?
• The authors show that DHT injection in females leads to decrease proliferation but also increase the expression of Axin 2, a b-catenin target gene. As b-catenin is known to increase cell proliferation, did the authors look for instance, at the consequence on the expression of CCND1 or others b-catenin target genes involved in cell cycle regulation?
• Does DHT treatment in females or Ar KO in males affect the nucleo-cytoplasmic accumulation of b-catenin compared to controls? These results could help determine if Ar acts through b-catenin signaling or independently.
• The authors did not mention if AS-RspoGOF males have macrophages and monocytes accumulation in the adrenal cortex as observed in Sf1-RspoGOF, this is an important information to better understand their origin and to know if they are due to mature or early embryonic dysregulation as AS-cre is activated later in time than Sf1-cre.

Minor comments:

• In figure S8, could the authors add a H&E staining of AS-RspoGOF male adrenals in order to have all the controls?
• Based on the results presented here, the macrophage and monocyte observed in Sf1-RSPO1GOF males does not sound to be due to androgens as they are not observed in females treated with DHT. The authors should discuss these results and potential hypothesis explaining the macrophage recruitment.
• The authors should add a sentence or two to integrate their results in the ones previously published by Dumontet and collaborators (PMID: 29367455) regarding the consequences of DHT treatment on the regulation of Wnt signaling.

Significance

Although the potential role of DHT in the sexual dysmorphism has previously been suggested in mouse models developing adrenocortical tumors, the demonstration of the direct role of the androgen receptor in this regulation demonstrated here is an important key in the understanding of molecular causes of this phenomenon that is observed both in human and mice adrenocortical tumorigenesis. Moreover, this model opens new perspectives to study the importance of the immune system in the regulation of the adrenal cortex homeostasis. Based on my expertise on adrenocortical homeostasis, I know that this manuscript will be of particular interest for researchers in the field of adrenocortical function and tumorigenesis but also, more generally for people working on the consequences of DHT and sexual dysmorphism.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Lyraki et.al. investigated the molecular basis of sexual dimorphism in adrenocortical hyperplasia. They use genetically modified mice, ectopically overexpressing R-spondin in SF1-experessing tissues (Sf1-Rspo1-gain-of-function [GOF]), causing overactivation of WNT signalling. Effects of R-spondin overexpression is nicely visualized using RNA Scope in situ hybridization and immunofluorescence; Ectopic R-spondin overexpression resulted in sex-specific adrenocortical hyperplasia (female) and degeneration (male), without observed changes in overall steroidogenic activity. Comparisons of the transcriptome of both male and female GOF and control mice of 4 weeks of age using RNA-sequencing demonstrated differential expression of genes involved in the cell cycle and immune response, in a sex-specific manner. Consequently, a BrdU proliferation assay confirmed increased proliferation in the inner cortex of female GOF mice but not male GOF mice. In contrast, male GOF mice showed increased CD68 staining, suggestive for increased macrophage infiltration.

Next, using a sex reversal mice model, they show that the sexual dimorphism in adrenal hyperplasia is dependent on testicular androgens rather than chromosomal sex. When female GOF mice were treated daily with dihydrotestosterone, their adrenal weight and adrenal proliferation (BrdU assay) reduced. Moreover, knocking out the androgen receptor in the adrenal cortex of both female and male mice overexpressing R-spondin resulted in increased adrenal weight in male mice to a level comparable to the weight of female adrenals.

Altogether, I believe this manuscript convincingly shows that androgens act on adrenocortical cells and contribute to the sex-dependent susceptibility to adrenocortical hyperplasia. I have a few comments:

1. Can the authors please briefly specify why for Figure 1D, three different statistical tests were performed? Based on the graphs it seems probable that this decided based on tests for normality and unequal variances? Were tests for normality and equal variances performed?

Would it be an idea to perform a two-way ANOVA (or equivalent test) instead to study the interactive effect of ectopic R-spondin overexpression and sex on adrenal weight as well as BrdU expression? 2. Based on the images of Axin2 expression in Figure 3B, Axin2 expression seems higher in the outer cortex of the female adrenal, compared to the outer cortex of the male adrenal (and the opposite in the inner cortex). Can you please explain why the opposite trend is seen in Figure 3D for the outer cortex? 3. Please clearly state the number of mice used for each experiment, either in the results section or in the methods section. At line 146, it is stated that n=6 mice "are analysed". At the other sentences just the n was provided. Therefore, I was wondering if not all mice were analysed here? How many mice were used? Also, at line 169 it is stated that 9 female Sf1-Rspo1GOF mice were used, while at line 170, only 8 aging Sf1-Rspo1GOF mice weer checked for carcinoma. Please explain. 4. Because of the low number, I am (based on this data only) not convinced that the presence of adrenocortical carcinoma in 1 of 8 female adrenals versus 0 of 8 male adrenals suggests sexual dimorphism (line 172). 5. At line 159, please specify the age of the mice when steroid hormone levels were quantified. 6. I would suggest rephrasing sentence 186-188. Principle component analysis indeed nicely separates the samples on sex and Rspondin overexpression. However, I am not sure if one could say that sex was the "second most important factor" underlying the variation in gene expression. 7. The manuscript is well written, although the manuscript (especially the introduction) is quit lengthy and could be written more concisely.

Significance

Of note, I am working in a specific endocrinological research niche and a bit distanced from this field. However, sexual dimorphism in (adrenal) diseases is a well-recognized phenomenon and the role of androgens and the androgen receptor in adrenal proliferation have previously been studied. Novel studies, like this study, are required to understand the mechanisms of sexual dimorphism in adrenal disease. This study is interesting especially (but not exclusively) for researchers working on adrenal or gonadal diseases or sexual dimorphism.

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

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this work, Sil et al. use fluorescent microscopy and biochemical reconstitution to study the ribonucleoparticles formed by one of the two L1 transposon proteins: ORF1p. Authors show that fluorescently tagged ORF1p forms puncta in HeLa cells within several hours of induced expression. The authors use this system to test how various mutations in the ORF1p affect its ability to form puncta in cells. They then correlate this property to the ability to induce transposition, which is quantified using a reporter system. Mutants that fail to form puncta are also unable to induce transposition. This leads the authors to conclude that condensation of ORF1p is required for L1 retrotransposition. A two-color colocalization assay demonstrates that ORF1p is immobile within the observed puncta, showing no evidence of exchange and mixing with a co-expressed ORF1p labeled with a different fluorescent protein. In addition, the authors purify the ORF1p protein, and various mutant variants, and test their ability to undergo phase separation in vitro in various conditions where they vary the concentrations of ORF1p, the salt, and RNA. The simple phase separation assays are complemented with a biophysical characterization of the condensates, where the post-fusion relaxation into a circular shape of the droplet is quantified and used to determine the inverse capillary velocity, which reflects the condensate viscosity and surface tension. These properties and then correlated with the ability of the variants to form puncta and facilitate retrotransposition.

It is an interesting and well-written article. The figures are neat and well documented. The experimental methods are described in sufficient detail. However, I believe that the conclusions made are not sufficiently supported by the presented evidence. The authors show correlation, not causation, of the ORF1p condensation and transposition. The evidence that the ORF1p particles form co-translationally, that they are condensates, and that they directly mediate transposition is insufficient. The in vitro work is interesting, but too preliminary and needs a more careful quantification. I encourage the authors to address my comments experimentally as much as they can. Where not possible, they could tone down the language and address the comments in writing and point out the limitations in the article text.

Major comments:

1. What is the evidence that ORF1p forms condensates in an endogenous situation? A more thorough discussion of the evidence, based on the literature is needed. Alternatively, authors could use antibodies (if available) to demonstrate that such structures indeed exist in cell culture of tissues.
2. The model that the observed puncta form co-translationally through co-condensation of ORF1p and its encoding mRNA is intriguing and would indeed provide an elegant biophysical explanation for the discussed cis preference of transposition. In my opinion, this idea is the strongest part of the paper. I would advise the authors to provide more compelling evidence for this idea, as currently, it is not well-supported by the data. At the least, the authors need to show that the L1 mRNA is actually present in the studied condensates (for example, using smFISH on fixed cells). This will also allow the determination of the number of L1 mRNAs present in each condensate.
3. If authors have access to a microscope that can perform FRAP measurements, I would strongly suggest such an assay, where the individual cytoplasmic and nuclear ORF1p puncta can be examined for their material properties as a function of time (compare 6 hours post-induction and 72 hours post-induction).
4. Please provide a more detailed analysis of the formation of nuclear ORF1p condensates. How much later do they appear? The nucleus is the place where transposition occurs. Do the authors suggest that the co-translationally formed condensates enter the nucleus? Or do they form there de-novo? Is there also no colocalization in the nuclear foci? This could be addressed by a quantitative time-course.
5. The in vitro assays only use the L1 mRNA fragment. Do other RNAs (for example total RNA, rRNA, mRNA) similarly affect ORF1p condensates? Other studies showed that the presence of specific RNA could nucleate the formation of condensates in vitro, particularly where non-specific RNA is also present, mimicking the cellular environment (Maharana et al. Science 2018, PMID: 29650702; Elguindy and Mendell, Nature 2021, PMID: 34108682). The authors should test if the observed effect of L1 mRNA fragment is sequence-specific. Length dependence should also be addressed, as it may be the key parameter for the "co-translational assembly and gelation" model.

Minor comments:

1. The K3/K4 and R261 variants don't form puncta and do not promote transposition, yet phase separate at a similar concentration in vitro. The stammer mutants phase separate less efficiently in vitro, yet form puncta and promote transposition. This suggests that the in vitro phase separation assay is not very informative of the protein's behavior in cells. To me, it suggests that the puncta observed in cells might not be formed through phase separation. Other mechanisms of puncta formation should be explored.
2. Based on the fluorescent images, can the authors estimate what percent of the ORF1p protein is actually present in distinct condensates and how much is diffuse in the cytoplasm or nucleoplasm? How does the outside (diffuse) concentration change upon increased expression or ORF1p? Is there any evidence of a saturation concentration?
3. Does the ORF1-Halo and ORF1-mNG2 colocalization change at longer time-points where larger condensates are observed?
4. The authors often refer to the "the total area of condensed phase". This parameter is not very useful, as it highly depends on the experimental condition. Instead, authors should determine the apparent saturation concentration for each studied mutant in the presence and absence of RNA at a relevant RNA concentration. This requires increasing the resolution at the protein concentration axis and an unbiased analysis pipeline.
5. It is shown that decreasing ORF1p protein concentration at a fixed salt concentration decreased the total condensed phase area but increased the protein partition coefficient. The DNA/RNA binding mutant R261A does not show this trend. Moreover, it is the only mutant that shows a change in the phase diagram upon the addition of RNA. One explanation is that there are nucleic acid contaminants present in the protein prep. In fact, the R261A mutant seems to also have a lower 260 nm peak relative to 280nm peak at the chromatogram. That the enrichment of the ORF-1p protein changes with increasing concentration strongly suggests that we are already looking at a multi-component system here, where the contaminant would be a second component. The authors do include an extra step in the purification protocol to reduce nucleic acid contamination. However, they could also run an ion-exchange chromatography to improve the purity. Alternatively, they could test if adding benzonase, RNAse or DNAse changes the phase diagram of the ORF1p alone.
6. It would be great to see how the StammerDel behaves in vitro. The authors could at least try the purification with their current protocol. Full-length proteins often behave very differently than the fragments alone.

Significance

The model that the proteins encoded by the L1 transposon form condensates co-translationally and that these assemblies are functional units of the transposon that explain the cis-preference is a significant, important and interesting concept. In my opinion, this idea is the strongest part of the paper. However, unless supported by more evidence (such as experiments and analysis suggested above), it remains just an idea. This work would be of interest of the phase separation community as well as general cell biology and genetics field.

My field of expertise is: biomolecular phase separation, quantitative microscopy, cell biology, protein biochemistry, developmental biology and genetics.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This paper revisits aggregate formation by ORF1p, a nucleic acid (NA) binding protein encoded by the L1 retrotransposon. This topic dates to 1996 (Hohjoh and Singer - 1996 Embo J 15: 630) and was extensively examined again in 2012 using highly purified ORF1p by Callahan et al (Callahan et al - 2012 Nucleic Acids Res., 40, 813), to determine the effect of salt and nucleic acid on this process. The earlier studies employed chemical cross linking and gel electrophoresis to examine ORF1p aggregates in the presence and absence of NA and neither were cited in the present study. As ORF1p contains several intrinsically disordered regions (IDRs) ORF1p aggregates can form phase separated condensates (droplets) which were characterized microscopically in the present study, and the authors assume that condensate formation is intrinsic to the function of ORF1p in retrotransposition, or as they state on page 2: "...we hypothesized that ORF1p undergoes condensation to carry out its roles in L1 RNP formation...". The authors attempt to correlate the ability of L1 encoded ORF1p complexed or not with RNA to form phase separated condensates in parallel with retrotransposition assays. They couple these observations with in vitro studies on condensate formation by the purified protein.

I have the following major comments:

• (A) The functional relevance of condensate formation by IDR-containing proteins has been questioned (Martin, E. W. and A. S. Holehouse - 2020; Emerging Topics in Life Sciences 4: 307). These authors conclude their review as follows: "In summary, IDRs are ubiquitous and play a wide range of functional roles across the full spectrum of biology, and in a large number (likely the majority) of cases their biological function has nothing to do with the ability to form large macroscopic liquid droplets. The notion that the presence of an IDR means a protein has evolved to phase separate is an inaccurate inference that has unfortunately been used to justify questionable lines of inquiry and questionable experimental design." And in terms of ORF1p this admonition is exemplified by the findings of Newton et al (2021, Biophys J 120;2181) cited by the present authors. This study showed that phase separated condensates readily form by just the N-terminal 152 amino acids (NTD + coiled coil). As this region of ORF1p cannot bind NA, condensate formation is indifferent to RNA binding, an obviously critical function of ORF1p.
• (B) Earlier studies (Ostertag et al - 2000; NAR 28:1418) showed that sufficient retrotransposition events have occurred by 48 hours after introduction of an L1 retrotransposition reporter to be readily detectable by whole cell staining for the retrotransposition-generated reporter gene product. The 48-hour lag presumably reflects the time to accumulate sufficient L1RNPs or their retrotransposed products to be detectable. Does this mean that the puncta (Fig 1F) accumulating during the first 24 hours after introduction of their full-length L1 retrotransposition reporter (Fig 1C) are the L1RNPs generated by the reporter? If not, what are they? If they are L1RNPs, are they thought to be or expected to exhibit the properties of phase separated condensates or are such properties just a feature of disembodied ORF1p that the authors posit could form an active L1RNP? The Ostertag paper should be cited here given its relevance to this issue.
• (C) Four of the IDRs in ORF1p harbor or are juxtaposed to phosphorylation sites essential for retrotransposition (their citation - Cook et al, 2015). As the authors expressed their purified proteins in E. coli, it is not phosphorylated and would not only be inactive for retrotransposition and given the structural effects of phosphorylation (e.g., Bah, A., et al.;2015; Nature, 510, 106) it would differ significantly from the structure of the active protein. As variables they introduce into ORF1p several not too subtle mutations particularly regarding the ORF1 coiled coil. They thereby aim to assess the role or particulars of ORF1p condensate formation for L1 retrotransposition. In their Abstract they state (p.1, l. 11) "...we propose that ORF1p oligomerization on L1 RNA drives the formation of a dynamic L1 condensate that is essential for retrotransposition."
• (D) Although the authors provide no direct experimental evidence for the above statement and whatever the authors mean by "dynamic L1 condensate" how does this conclusion materially differ from the conclusions published by Naufer et al, in 2016 (NAR; 44,281), which also was not cited by the authors. Naufer et al used single molecule studies and highly purified ORF1p that had been expressed in insect cells (and thus was fully phosphorylated, Cook et al, 2015). They showed that oligomerization of nucleic acid (NA)-ORF1p complexes to a compacted stably bound polymer was positively correlated with retrotransposition. Both properties could be eliminated by coiled coil mutations that had no effect on biochemical assays of ORF1p activity - high affinity NA binding and NA chaperone activity. As both properties map to the carboxy terminal-half of ORF1p, the inactivating coiled coil mutations are an example of the numerous instances of strong epistasis exerted by amino acid substitutions in the coiled coil on the retrotransposition activity of ORF1p. In some cases epistasis is exerted at the single residue level (e.g., Martin,et al - 2008, Nucleic Acids Res., 36, 5845; Furano, et al. - 2020, PLOS Genetics 16 e1008991.)

While the authors are apparently also not mindful of the PLOS Genetics paper examining the effect of a single inactivating coiled coil substitution at the level of microscopically observed condensates could have provided compelling evidence linking their formation and retrotransposition. On the other hand, lack of a condensate-based readout for single amino acid inactivating coiled coil mutations would question the validity of equating ORF1p condensates with retrotransposition competence. - (E) The afore mentioned Callahan et al study (2012, NAR, 40, 813) in addition to producing results partly recapitulated in Fig. 2 of the present paper, showed that ORF1p polymerization was mediated by interactions between the highly conserved RRM-containing region of ORF1p. This observation is consistent with previous studies showing RRM-mediated protein interactions of other proteins (Clery, et al 2008, Curr. Opin. Struct. Biol., 18, 290; Kielkopf, et al Genes Dev., 18, 1513)

As well as including the missing citations of the L1 literature, implications of the above considerations need to be addressed before publication.

I have the following additional comments and issues:

1. p.2. l. 8, the citation to TPRT should include Luan,et al.- 1993, Cell 72: 595
2. p. 5, middle of 2nd para - what does "different diffusivity" mean? - what are "stereotyped puncta"?

Any invocation of cis preference should cite the foundational study by Kaplan, N., et al. (1985). "Evolution and extinction of transposable elements in Mendelian populations." Genetics 109 459. 3. p.10 middle paragraph, the authors state: "The decreased phase separation of the R261A mutant was unexpected, as we predicted that mutating a core RNA-binding residue would only affect condensation in the presence of RNA. We also noted that the protein partition coefficients of the R261A condensed phases were higher than their counterparts for WT and K3A/K4A. Taken together, these experiments showed that K3/K4 and R261 are not essential for protein condensation in vitro."

these findings would have been predicted by the afore mentioned findings of Newton et al, which should be cited here. 4. p. 14, first paragraph "we predicted that stammer-deleted ORF1p would maintain an elongated coiled coil conformation that might disfavor trimer- trimer interactions that are mediated by the N terminal half of the protein (Figure 4A, left two cartoons)."

It seems that the authors are stating that different fully formed trimers can form larger complexes mediated by interactions between their coiled coils, an idea apparently based on results published by Khazina and Weichenreider (2018). This paper states that "Additional biophysical characterizations suggest that L1ORF1p trimers form a semi-stable structure that can partially open up, indicating how trimers could form larger assemblies of L1ORF1p on LINE-1 RNA." However, the cited Khazina structural data ((PDB) entry 6FIA)) were derived from coiled coils that had been solubilized to monomers in guanidinium HCl from inclusion bodies (insoluble aggregates) that had accumulated during their synthesis in E. coli...a common condition for highly expressed proteins. Fully denatured ORF1p coiled coils such as these, which also lack the entire NTD are an in vitro artifact and never exist in "nature". It is almost certain that ORF1p monomers trimerize while being synthesized on adjacent ribosomes (e.g., Bertolini et al.- 2021; Science 371: 57). I am not aware of any biochemical evidence from the Martin laboratory on mouse ORF1p or the Weichenrieder or Furano laboratories on human ORF1p indicating that the coiled coils of fully formed trimers synthesized in vivo can unravel to mediate interactions between different trimers. In fact, the authors' results in Fig 1F supports this contention. 5. p.10, Legend to Figure 1G The cells were stained simultaneously with two Halo ligand dyes (Halo-JF549 and Halo- JF646), giving a positive control for colocalization.

Why is staining the same ligand (Halo) with two different dyes a colocalization control? 6. The authors conclude their paper with the statement "The L1 system characterized in this work employs a uniquely powerful combination of biochemical reconstitution, live-cell imaging, and functional phenotyping in cells. In vitro reconstitution allows us to study the biophysical properties of condensates in a minimal and controllable system."

However, there are several instances where the in vitro biochemical properties of ORF1p variants are somewhat discordant with their in vivo results. In the case of their coiled coil mutants. replacement of the coiled coil stammer, MEL (uniquely invariant for more than 50 Myr of primate coiled coil evolution) with AAA or AEA exhibited reduced retrotransposition that was not accompanied by a corresponding reduction in condensate formation (Fig 4). In another instance, while mutation of the highly conserved residue (R261) necessary for RNA binding eliminated retrotransposition it did not have a corresponding effect on condensate formation even in the presence of RNA (Fig 3).

General comments on the Figures - Although I rather liked the cartoon version of ORF1p (Fig 1B) and when used to show the location of mutated site, versions that purport to show the effect of mutations on structure (Fig 4A) are misleading and should be eliminated.

Closing Comment:

Overall, I enjoyed reading this paper, and feel that when the issues I raised are appropriately addressed and the relevant missing citations are included it would make a useful contribution. However, it seems that the authors could make a more compelling case that dissociates condensate formation of ORF1p and its activity in retrotransposition, consistent with the Martin and Holehouse review cited above. So, I urge them to reconsider their conclusions. I did not find the highly speculative discussion about the relevance of phase separation / condensate formation to cis preference at all convincing as it is just as it is just as likely (maybe more so) to be enforced at the level of selection - evolutionary failures, by definition, are not propagated.

Significance

Although this paper addresses a long-studied topic in L1 biology, namely how the L1 encoded proteins assemble into an L1 RNP (the retrotransposition intermediate), the authors posit that the formation of phase/separated protein condensates (visible as microscopic droplets) are involved. Such droplets are a currently popular biochemical feature exhibited by some proteins, but their functional relevance is a currently a contentious topic in protein biochemistry. I do not think that the authors make a convincing case that condensate formation is involved, rather I think that their evidence provides reasonable evidence that condensation has no role. I urge the authors to consider this possibility, but whatever which conclusion proves to be correct, their study would make a useful contribution to the field.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This study shows that the ORF1 protein of the LINE-1 retroelement forms puncta in vivo that they define as cytoplasmic biomolecular condensates based on the characterization of the biophysical properties of ORF1p condensates in vitro.

Defective retrotransposition of some ORF1p mutants correlates with defects in puncta formation in vivo and alteration of biophysical properties of in vitro condensates leading the authors to conclude that condensation of ORF1p is required for retrotransposition.

The study combines biochemical reconstitution, biophysic analysis and live-cell imaging. In particular, the authors take advantage of a new powerful tool they have developed based on the tagging of ORF1 within a functional L1 reporter element. The fluorescent tag allows following the dynamics of ORF1p by live-cell imaging.

The key conclusion is that ORF1p condensation is important for L1 retrotransposition. The correlation is clearly shown but raises several questions: Is the defect in ORF1p condensation the only explanation for the retrotransposition defects of the ORF1p mutants analyzed here? Can we exclude that the mutations in ORF1p affect other functions of the protein such as its binding to RNA (as in the case of the R261 mutant) and cis-preference, or its binding to other factors involved in L1 replication? Could the loss of these functions affect L1 retrotransposition independently of ORF1p condensation?

Major comments:

On several occasions, the authors propose that ORF1p-HALO dynamics in vivo is linked to its co-translational association with L1 RNA. However, they never show the presence of L1 RNAs in ORF1p-HALO puncta in vivo. To strengthen the conclusion that the puncta observed in vivo are L1 RNPs, the authors should add experiments showing the presence of L1 RNA in the cytoplasmic puncta (by RNA FISH) or that the puncta are dependent on the presence of L1 RNA (expressing ORF1p-HALO alone should not be sufficient for puncta formation). These experiments seem to be realistic in few weeks with the tools already available in the laboratory.

Apart from this comment, the authors are cautious in their conclusions. It is clear, as they indicate in the Discussion, that showing that ORF1p condensation is also required for the mobility of other retroelements will strengthen the implication of ORF1p condensation in L1 replication.

The data are well presented and the methods described in detail so that others should be able to use them. The experiments seem to be adequately replicated and the statistical analysis adequate.

Minor comments:

Figure 1F: Having the pictures of cell nuclei (like in Figure 1D) would be nice to know how many cells we are looking at in this panel.

Figure 2E: it is surprising that there is no correlation between the ORF1p:RNA ratio and the number of individual fusion events (i.e. curves of ORF1p+RNA 10000:1 and 1000:1 overlap while 3000:1 is different). Could the authors discuss this point?

Previous studies are appropriately referenced. Text and figures are clear and precise.

Referees cross-commenting

The main critical points shared by all reviewers are: 1) the need to show the presence of LINE1 RNAs in ORF1p condensates in vivo and 2) the lack of evidence for causality between ORF1 condensate formation and L1 transposition efficiency (At this stage, the authors should moderate their conclusions, especially in the abstract). Regarding the other reviews, we noticed the need to cite additional relevant studies in the field (reviewer #2) and the interesting points raised by reviewer #3 to investigate the formation of ORF1 condensates in an endogenous situation, and whether other RNAs do affect ORF1p condensates.

Significance

The study is technically interesting in that it describes a new system for tracking ORF1p puncta formation in vivo. The findings are not unexpected because it comes after the publication of Newton et al. in 2021 (PMID : 33798566), describing that ORF1p does phase separation in vitro. Furthermore, the fact that RNPs form "membrane-less" structures is already established in other situations as the authors point out. Compared to Newton et al., condensates are better-defined biochemically, especially for RNA association features and in vivo dynamics.

The ORF1 protein is widely studied for its role in L1 retrotransposition. The protein forms a homotrimer in vitro, binds to L1 mRNA in a cis-preferential manner, and is required for retrotransposition. On the other hand, RNA-binding proteins are often involved in the formation of membrane-less organelles (stress granules, RNA processing bodies...). These observations suggest that ORF1p may also form RNP condensates required for L1 retrotransposition. A study published in Biophysical Journal in 2021 (Newton et al. PMID: 33798566) has already reported the phase separation of the LINE-1 ORF1p that is mediated by the N-terminus and coiled-coil domain. This former study was based on in vitro microscopy and NMR approaches and is cited in the submitted manuscript. The study submitted to Rev commons goes further by analyzing the biochemical properties of ORF1p condensates in the presence of L1 RNA and by following in vivo condensates of ORF1p (WT or mutants) expressed from a functional L1 reporter element by live-cell imaging. The findings will interest a wide audience investigating the biology of retroelements and more particularly scientists who study the L1 retrotransposon.

I am an expert in retrotransposon biology but I do not work on L1. I am not expert enough to assess the quality and relevance of the biophysical experiments in the paper. In particular, panels 2D, 3B and 3D were difficult to analyze.

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

Learn more at Review Commons

---

Reply to the reviewers

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

• *

The authors proposed that the stable and opened membrane neck that connects the bud to the cytoplasm may persist for a long time in the infected cell during active RNA production. The viral ring-shaped nsPs is supposed to have an important role of maintaining this stable high-curvature membrane neck. It is suggested that the nsP1 dodecamer may pull together the membrane inner surface in the neck region via electrostatic interactions. Namely the authors observed that in the absence of negatively charged membrane lipids nsP1 did not bind appreciably to the membrane. The presented experimental data and theoretical consideration suggest that the CHIKV spherule consists of a membrane bud filled with viral RNA, and has a macromolecular complex gating the opening of this bud to the cytoplasm.

The presented results are interesting, the manuscript is well written and can be published after revision. The following comments are offered to the authors' consideration.

We thank the reviewer for this positive overall assessment.

1.Since there is no protein coating over the curved surface of the membrane bud, the authors concluded that the membrane neck must be stabilised by specific mechanism involving nsP1. It was further assumed that the viral protein nsP1 serves as a base for the assembly of of a larger protein complex at the neck of the membrane bud. In addition to suggested mechanism of the neck stabilization, thin highly curved membrane neck can be stabilised also by accumulation of the membrane components having the appropriate membrane curvature (i. E. negative intrinsic curvature or anisotropic intrinsic curvature), see Kralj-Iglic et al., Eur. Phys. J. B., 10: 5-8 (1999), https://doi.org/10.1007/s100510050822.

Please discuss this issue in the manuscript.

This is a good point, thank you for making it. In the revised manuscript we discuss both the possibility of lipid sorting into the neck region by nsP1 (lines 217-222), and the mentioned paper regarding anisotropic inclusions (lines 268-271).

• *

2.In Eq. (1) the Gaussian curvature term (appearing in Helfrich bending energy term) is not included. Usually this term is omitted in the case of closed membrane shapes (i.e. so-called spherical topology) due to validity of the Gauss-Bonnet theorem. In the present manuscript/work the shape equation was solved for the membrane patch. Can you therefore please explain shortly to the reader why you can omit the Gaussian curvature term from Eq.(1). For example due fixed inclination angle and foxed curvature at the boundary, .....

Thanks for finding this omission. We have now revised the manuscript to describe why we can omit the Gaussian curvature term (lines 241-245).

• *

• *

3.«Sigma« and »P« can be considered also as global Lagrange multipliers for the constraint of the fixed total membrane area of the bud (including the neck membrane) and the constraint of the fixed volume of the bud. If you then take into account separately also the equation for the fixed membrane area you could predict different shapes of the bud (by solving the shape equation) at fixed area of the bud, calculated for different values of the model parameters (and different boundary conditions) - in this case Sigma is the result of variational procedure (as well P if you consider also the constraint for the fixed volume of the bud). See for example Medical & Biological Engineering & Computing, vol. 37, pp. 125-129, 1999 and J. Phys. Condens. Matter, vol. 4, pp. 1647-1657, 1992. Can you please shortly discuss in the manuscript also this issue.

This is an interesting point. We now discuss this and cite the mentioned papers at the end of the theory section in the supplementary information (lines 203-205) as well as briefly mentioning it when discussing Eq. 1 (lines 240-242).

• *

**Referees cross-commenting**

I agree as well.

• *

Reviewer #1 (Significance (Required)):

The presented experimental and theoretical results are interesting, the manuscript is well written and can be published after revision.

We thank the reviewer for this appreciative comment.

• *

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

Summary:

In their manuscript "Architecture of the chikungunya virus replication organelle" Laurent and colleagues show:

• *

- the 3D structure of the "neck complex" that forms the gateway between the Chikungunya virus replication/transcription organelle (termed "spherule") and the cytoplasm of infected cells. The structure was obtained by native electron cryo-tomography and sub-tomogram averaging of BHK cells infected with a single-cycle replicon system encoding all components of the viral replication machinery. The nominal resolution of the structure is 28 Å. The viral nsP1 protein, for which two high-resolution structures have previously been published, could unambiguously be located within the density of the neck complex.

• *

- nsP1 interaction with membranes relies on lipids with a single negative net charge, such as POPS, POPG and PI, whereas two different PIPs with a negative net charge greater than one support nsP1 binding less efficiently. These membrane determinants for nsP1 binding were elucidated using two complementary methods: multilamellar vesicle pulldown assays and confocal imaging of fluorescently labeled giant unilamellar vesicles in the presence of fluorescently labeled nsP1. Purified nsP1 was produced in E. coli.

• *

- nsP1 recruits nsP2 (another component of the neck complex) to membranes with suitable lipid composition. This observation was made using the same multilamellar vesicle pulldown assay.

• *

- the 3D organization of the viral genome within the spherule, demonstrating that each spherule contains one copy of the genome as a double-stranded RNA molecule. This analysis was carried out by segmentation of the same tomograms that were used to visualize the neck complex.

• *

- the force exerted by RNA polymerization within the spherules is sufficient to drive membrane remodeling. This is a theoretical argument based on mathematical modelling.

• *

Major comments:

The article is written clearly and all major claims seem justified. The biochemical assays are presented in duplicates or triplicates, which is sufficient to derive the provided conclusions. The workflow for electron cryo-tomography analysis seems sound, even though the low number of individual particles (=64) for sub-tomogram averaging of the neck complex limits the resolution of its final structure. Given the strong competition in the field, and considering the high experimental workload that would be required for further improvement of the resolution, I do not recommend any additional benchwork for this paper.

We thank the reviewer for this assessment, especially for recognising the challenge in obtaining a larger number of spherule subtomograms under the complex replicon particle protocol we had to use in order to study the BSL3 CHIKV under BSL2 conditions.

• *

My only concern is the accuracy of the experimental genome length measurements, which has important implications for their mechanistic interpretation. The type of tomograms that have been recorded here inherently suffers from anisotropy with respect to both resolution and contrast. This makes accurate tracing of tangled filaments very challenging, and in this light, I congratulate the authors for the impressively good agreement of their average experimentally determined genome length with the theoretical genome length (Figure 4C). As to be expected, however, the second supplementary video clearly shows multiple gaps in the traced genome, implying that there must necessarily be errors in the length measurements. Unless there is a possibility to confidently estimate the magnitude of these errors, my preferred interpretation would be that the vast majority of imaged spherules - regardless of their temporary volume in the moment of sample freezing - likely contains precisely one copy of the double-stranded RNA genome, and not fractions thereof as is suggested in the text (for example, line 305: "Analysis of the cryo-electron tomograms gave a clear answer to the question of the membrane bud contents: the lumen of full-size spherules consistently contains 0.8-0.9 copies."). I feel that this subject deserves more discussion in the manuscript. If the authors prefer to keep their original interpretation that the majority of spherules contains only fractions of full genomes, I invite them to provide an explanation for why their experimental genome length measurements are sufficiently accurate to favor this rather surprising conclusion over my more trivial interpretation. If I understand correctly, my preferred interpretation has implications for the mathematical model for membrane remodeling (Equation 2).

This is a good point. In fact, we agree that our original manuscript and wording was unclear and we agree with the reviewer’s interpretation (“my preferred interpretation would be that the vast majority of imaged spherules - regardless of their temporary volume in the moment of sample freezing - likely contains precisely one copy of the double-stranded RNA genome”). We have now changed the text to reflect that we believe we have a 10-20% false negative rate in the filament tracing and that the most likely interpretation is indeed that each spherule has exactly one genome copy (lines 207-210). In addition, we looked at the possible consequences of the slight underestimation of the filament length for the mathematical model, and describe on lines 257-264 why this in fact would have no impact on the conclusions of the modeling.

• *

Minor comments:

Virus taxa should be capitalized and written in italics wherever applicable. I recommend adhering to the following rules:

https://talk.ictvonline.org/information/w/faq/386/how-to-write-virus-species-and-other-taxa-names

Thank you for helping us clarify this. In response to this we have now italicized and capitalized all virus taxa.

Figure 2I looks as if the pink cross-section of nsP1 has not been scaled correctly. Comparison to Figure 2H gives me the impression that the diameter of the pink nsP1 ring in Figure 2I should be scaled down relative to the greyscale neck complex.

We would like to than the reviewer for their keen eye. There was indeed a scaling problem, which we have now solved in the updated Fig. 2.

• *

The caption of Figure 2 calls more panels than are provided in the figure. The caption "panel E" seems to be obsolete.

Thanks for finding this mistake. We have now revised Fig. 2 and its legend.

• *

In the methods, centrifugation speed should be given in units of relative centrifugal force (rcf) instead of revolutions per minute (rpm), especially for the MLV pulldown assay where no rotor is indicated.

We agree and have changed this on lines 482,490,524,531,543 and 597 of the manuscript

• *

In the methods for the MLV assay, the lipid:protein ratio is given with 500:1. It should be specified whether this is a mass ratio or a molar ratio.

It was molar ratio which we have now specified on line 595.

In the methods, the buffer composition for the mass photometry measurement should be indicated.

Good point. We added this on lines 632-633.

• *

**Referees cross-commenting**

• *

I agree to the other reviewers' remarks.

• *

Reviewer #2 (Significance (Required)):

• *

Chikungunya virus is a very important human pathogen, and research on the architecture of its replication/transcription organelle holds great promise for the development of future therapies. Laurent and colleagues advanced this field by providing pioneering low-resolution 3D structures of the membrane-bound viral protein complex and the viral RNA content of this organelle in situ. In addition, they also assessed the lipid requirements for membrane interaction of the primary viral membrane anchor of this complex, nsP1, in vitro. Underlining the importance of these results, a competing group submitted a partially overlapping study to BioRXiv three months ahead (https://doi.org/10.1101/2022.04.08.487651). Whereas the competing group describes the structure of the neck complex at a much higher resolution, it neither analyzes the RNA content of the spherules nor does it address the lipid preferences of nsP1. The present study by Laurent and colleagues should therefore be of great interest to many virologists and cellular biologists.

• *

I am a structural virologist with a focus on envelope glycoproteins. Of relevance to this review, I have experience with cellular electron cryo-tomography and sub-tomogram averaging, as well as in-vitro protein/liposome interaction assays. I do not feel qualified to evaluate the details of the mathematical model for membrane remodeling that is used in the last results section of this manuscript.

We thank reviewer 2 for this positive overall assessment of our work.

• *

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

• *

This is an interesting and well written paper describing the replication spherules generated by Chikungunya virus. Cryo-electron tomography was used to determine a low-resolution structure of the spherule, suggesting that nsP1 is located at the neck of the spherule. Segmentation of the tomograms combined with mathematical modeling was used to produce a structural model for RNA organization in the spherule, suggesting that each spherule contained approximately one copy of a full double-stranded RNA genome. I have a few minor comments:

• *

We are thankful for this positive overall assessment of our work.

• *

The structural studies were complemented with lipid binding assays, showing that nsP1 has an affinity for anionic lipids. While interesting, the connection of these experiments to the rest of the study seems tenuous. There is no further mention of them in the discussion or how they relate to the tomography and their replication model.

We agree that those data were not as well integrated into the paper as they could have been, and are thankful that the reviewer pointed this out. To improve the integration of these data into the manuscript, we have expanded on two ways in which the reconstitution data relate to the rest of the paper: (i) the tomography led us to hypothesise that nsP1 recruits other nsPs to the membrane, which we could confirm with the reconstitution (lines 151-152, and throughout that parapgraph), and (ii) the lipid preferences of nsP1 that we could measure using the titrating pulldown experiments inform the possible models for how the spherule memebrane is remodeled since nsP1 binds lipids that cannot on their own stabilize a neck shape (lines 217-222). We have also slightly expanded the discussion of the biochemistry and its relation to other data in the paper (lines 307-311).

• *

It is a nice match between the calculated length of the RNA (assumed to be ds) and the length of the vector, but the segmentation of the RNA is not completely convincing based on the provided images. It is difficult to distinguish the RNA strands from the noise and other components in the spherule and, at least by eye, the segments do not seem very connected. Please provide some more details on the tracing algorithm. Has it been validated on a known system?

We appreciate this comment and recognise that we did not sufficiently explain the tracing algorithm. This software was in fact custom written (by others, ca 10 years ago) for cryo-electron tomography and has since been used by others in several studies of cellular cryo-electron tomograms, e.g. to study actin cytoskeleton. We now mention this in the results (lines 195-196) and methods (lines 462-463).

The tomogram video is nice, but it would be good to see a raw image as well, preferably covering a wider view that includes the whole cell, as well as a tomogram that represents the entire field of the reconstruction.

This is a good suggestion. We unfortunately cannot provide images covering the entire cells since this is beyond the field of view of the electron microscope (and an image montage was not acquired at the time of data collection). However, we are now providing an additional supplementary movie that shows the entire field of view of the tomogram. In addition, we have uploaded two of the tomograms (including the uncropped tomogram from Figure 1) to EMDB where they will be downloadable by everyone after publication. We hope the reviewer appreciates that this is all that is technically possible at the moment.

• *

In figure 2, the panels are mislabeled relative to the legend, which refers to the color guide as its own panel.

Thanks for pointing this out, we have rectified this in the revised Fig. 2 and its legend.

Line 405: C36 symmetry? Why? Shouldn't it be C12 symmetry?

36-fold symmetry was applied to the lipid membrane part to smoothen it further. The membrane part of the structure is simply outlining the neck shape and this is better visualised in this smoothened representation as also done e.g. in the study of the coronavirus neck complex (Wolff et al, Science 2020). We changed the methods text to make this more clear (line 449).

• *

Line 409: "fit" should be "fitted"

Thanks, Corrected in the revised manuscript line 454.

**Referees cross-commenting**

• *

I think we are all in good agreement, and I believe that the concerns raised can be addressed though a better explanation of the methods and improved discussion of their results.

We also agree and believe we have addressed all of the remaining concerns in the revised manuscript.

• *

• *

Reviewer #3 (Significance (Required)):

• *

This is a rather focused study, showing tomography data on the alphavirus replication complex. The main significance of the study is the description of the spherule's dimension and its relationship to the nature of the RNA, which provided a model for the replication process. While somewhat narrow in scope, the study should be of interest to people working in the virus replication and virus structure field. The lipid data are interesting, but does not seem well integrated with the rest of the study.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This is an interesting and well written paper describing the replication spherules generated by Chikungunya virus. Cryo-electron tomography was used to determine a low-resolution structure of the spherule, suggesting that nsP1 is located at the neck of the spherule. Segmentation of the tomograms combined with mathematical modeling was used to produce a structural model for RNA organization in the spherule, suggesting that each spherule contained approximately one copy of a full double-stranded RNA genome. I have a few minor comments:

The structural studies were complemented with lipid binding assays, showing that nsP1 has an affinity for anionic lipids. While interesting, the connection of these experiments to the rest of the study seems tenuous. There is no further mention of them in the discussion or how they relate to the tomography and their replication model.

It is a nice match between the calculated length of the RNA (assumed to be ds) and the length of the vector, but the segmentation of the RNA is not completely convincing based on the provided images. It is difficult to distinguish the RNA strands from the noise and other components in the spherule and, at least by eye, the segments do not seem very connected. Please provide some more details on the tracing algorithm. Has it been validated on a known system?

The tomogram video is nice, but it would be good to see a raw image as well, preferably covering a wider view that includes the whole cell, as well as a tomogram that represents the entire field of the reconstruction.

In figure 2, the panels are mislabeled relative to the legend, which refers to the color guide as its own panel.

Line 405: C36 symmetry? Why? Shouldn't it be C12 symmetry? Line 409: "fit" should be "fitted"

Referees cross-commenting

I think we are all in good agreement, and I believe that the concerns raised can be addressed though a better explanation of the methods and improved discussion of their results.

Significance

This is a rather focused study, showing tomography data on the alphavirus replication complex. The main significance of the study is the description of the spherule's dimension and its relationship to the nature of the RNA, which provided a model for the replication process. While somewhat narrow in scope, the study should be of interest to people working in the virus replication and virus structure field. The lipid data are interesting, but does not seem well integrated with the rest of the study.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In their manuscript "Architecture of the chikungunya virus replication organelle" Laurent and colleagues show:

• the 3D structure of the "neck complex" that forms the gateway between the Chikungunya virus replication/transcription organelle (termed "spherule") and the cytoplasm of infected cells. The structure was obtained by native electron cryo-tomography and sub-tomogram averaging of BHK cells infected with a single-cycle replicon system encoding all components of the viral replication machinery. The nominal resolution of the structure is 28 Å. The viral nsP1 protein, for which two high-resolution structures have previously been published, could unambiguously be located within the density of the neck complex.
• nsP1 interaction with membranes relies on lipids with a single negative net charge, such as POPS, POPG and PI, whereas two different PIPs with a negative net charge greater than one support nsP1 binding less efficiently. These membrane determinants for nsP1 binding were elucidated using two complementary methods: multilamellar vesicle pulldown assays and confocal imaging of fluorescently labeled giant unilamellar vesicles in the presence of fluorescently labeled nsP1. Purified nsP1 was produced in E. coli.
• nsP1 recruits nsP2 (another component of the neck complex) to membranes with suitable lipid composition. This observation was made using the same multilamellar vesicle pulldown assay.
• the 3D organization of the viral genome within the spherule, demonstrating that each spherule contains one copy of the genome as a double-stranded RNA molecule. This analysis was carried out by segmentation of the same tomograms that were used to visualize the neck complex.
• the force exerted by RNA polymerization within the spherules is sufficient to drive membrane remodeling. This is a theoretical argument based on mathematical modelling.

Major comments:

The article is written clearly and all major claims seem justified. The biochemical assays are presented in duplicates or triplicates, which is sufficient to derive the provided conclusions. The workflow for electron cryo-tomography analysis seems sound, even though the low number of individual particles (=64) for sub-tomogram averaging of the neck complex limits the resolution of its final structure. Given the strong competition in the field, and considering the high experimental workload that would be required for further improvement of the resolution, I do not recommend any additional benchwork for this paper.

My only concern is the accuracy of the experimental genome length measurements, which has important implications for their mechanistic interpretation. The type of tomograms that have been recorded here inherently suffers from anisotropy with respect to both resolution and contrast. This makes accurate tracing of tangled filaments very challenging, and in this light, I congratulate the authors for the impressively good agreement of their average experimentally determined genome length with the theoretical genome length (Figure 4C). As to be expected, however, the second supplementary video clearly shows multiple gaps in the traced genome, implying that there must necessarily be errors in the length measurements. Unless there is a possibility to confidently estimate the magnitude of these errors, my preferred interpretation would be that the vast majority of imaged spherules - regardless of their temporary volume in the moment of sample freezing - likely contains precisely one copy of the double-stranded RNA genome, and not fractions thereof as is suggested in the text (for example, line 305: "Analysis of the cryo-electron tomograms gave a clear answer to the question of the membrane bud contents: the lumen of full-size spherules consistently contains 0.8-0.9 copies."). I feel that this subject deserves more discussion in the manuscript. If the authors prefer to keep their original interpretation that the majority of spherules contains only fractions of full genomes, I invite them to provide an explanation for why their experimental genome length measurements are sufficiently accurate to favor this rather surprising conclusion over my more trivial interpretation. If I understand correctly, my preferred interpretation has implications for the mathematical model for membrane remodeling (Equation 2).

Minor comments:

Virus taxa should be capitalized and written in italics wherever applicable. I recommend adhering to the following rules: https://talk.ictvonline.org/information/w/faq/386/how-to-write-virus-species-and-other-taxa-names

Figure 2I looks as if the pink cross-section of nsP1 has not been scaled correctly. Comparison to Figure 2H gives me the impression that the diameter of the pink nsP1 ring in Figure 2I should be scaled down relative to the greyscale neck complex.

The caption of Figure 2 calls more panels than are provided in the figure. The caption "panel E" seems to be obsolete.

In the methods, centrifugation speed should be given in units of relative centrifugal force (rcf) instead of revolutions per minute (rpm), especially for the MLV pulldown assay where no rotor is indicated.

In the methods for the MLV assay, the lipid:protein ratio is given with 500:1. It should be specified whether this is a mass ratio or a molar ratio.

In the methods, the buffer composition for the mass photometry measurement should be indicated.

Referees cross-commenting

I agree to the other reviewers' remarks.

Significance

Chikungunya virus is a very important human pathogen, and research on the architecture of its replication/transcription organelle holds great promise for the development of future therapies. Laurent and colleagues advanced this field by providing pioneering low-resolution 3D structures of the membrane-bound viral protein complex and the viral RNA content of this organelle in situ. In addition, they also assessed the lipid requirements for membrane interaction of the primary viral membrane anchor of this complex, nsP1, in vitro. Underlining the importance of these results, a competing group submitted a partially overlapping study to BioRXiv three months ahead (https://doi.org/10.1101/2022.04.08.487651). Whereas the competing group describes the structure of the neck complex at a much higher resolution, it neither analyzes the RNA content of the spherules nor does it address the lipid preferences of nsP1. The present study by Laurent and colleagues should therefore be of great interest to many virologists and cellular biologists.

I am a structural virologist with a focus on envelope glycoproteins. Of relevance to this review, I have experience with cellular electron cryo-tomography and sub-tomogram averaging, as well as in-vitro protein/liposome interaction assays. I do not feel qualified to evaluate the details of the mathematical model for membrane remodeling that is used in the last results section of this manuscript.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The authors proposed that the stable and opened membrane neck that connects the bud to the cytoplasm may persist for a long time in the infected cell during active RNA production. The viral ring-shaped nsPs is supposed to have an important role of maintaining this stable high-curvature membrane neck. It is suggested that the nsP1 dodecamer may pull together the membrane inner surface in the neck region via electrostatic interactions. Namely the authors observed that in the absence of negatively charged membrane lipids nsP1 did not bind appreciably to the membrane. The presented experimental data and theoretical consideration suggest that the CHIKV spherule consists of a membrane bud filled with viral RNA, and has a macromolecular complex gating the opening of this bud to the cytoplasm.

The presented results are interesting, the manuscript is well written and can be published after revision. The following comments are offered to the authors' consideration.

1. Since there is no protein coating over the curved surface of the membrane bud, the authors concluded that the membrane neck must be stabilised by specific mechanism involving nsP1. It was further assumed that the viral protein nsP1 serves as a base for the assembly of of a larger protein complex at the neck of the membrane bud. In addition to suggested mechanism of the neck stabilization, thin highly curved membrane neck can be stabilised also by accumulation of the membrane components having the appropriate membrane curvature (i. E. negative intrinsic curvature or anisotropic intrinsic curvature), see Kralj-Iglic et al., Eur. Phys. J. B., 10: 5-8 (1999), https://doi.org/10.1007/s100510050822.<br /> Please discuss this issue in the manuscript.
2. In Eq. (1) the Gaussian curvature term (appearing in Helfrich bending energy term) is not included. Usually this term is omitted in the case of closed membrane shapes (i.e. so-called spherical topology) due to validity of the Gauss-Bonnet theorem. In the present manuscript/work the shape equation was solved for the membrane patch. Can you therefore please explain shortly to the reader why you can omit the Gaussian curvature term from Eq.(1). For example due fixed inclination angle and foxed curvature at the boundary, .....
3. «Sigma« and »P« can be considered also as global Lagrange multipliers for the constraint of the fixed total membrane area of the bud (including the neck membrane) and the constraint of the fixed volume of the bud. If you then take into account separately also the equation for the fixed membrane area you could predict different shapes of the bud (by solving the shape equation) at fixed area of the bud, calculated for different values of the model parameters (and different boundary conditions) - in this case Sigma is the result of variational procedure (as well P if you consider also the constraint for the fixed volume of the bud). See for example Medical & Biological Engineering & Computing, vol. 37, pp. 125-129, 1999 and J. Phys. Condens. Matter, vol. 4, pp. 1647-1657, 1992. Can you please shortly discuss in the manuscript also this issue.

Referees cross-commenting

I agree as well.

Significance

The presented experimental and theoretical results are interesting, the manuscript is well written and can be published after revision.

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

Learn more at Review Commons

---

Reply to the reviewers

The current resubmission is our revision plan only. Therefore the authors do not wish to provide a response at this time. We will include our response to reviewers with our full resubmission.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Nguyen and Goetz here explore the roles of Tau tubulin kinase 2 (TTBK2) in ciliary stability. Using a murine conditional null MEF model, they ablate TTBK2 from cells with established cilia and monitor the impact on ciliary structures. They observe loss of cilia over time, acoompanied by changes in centriolar satellites, axonemal microtubule composition and intraflagellar transport protein levels at the basal body. They note changes in autophagy proteins in the absence of TTBK2 and find that pharmacological manipulation of actin dynamics impacts on the ciliary phenotypes seen in TTBK2 deficient cells.

Major comments:

1. The phrase 'centrosomal compartment' is potentially ambiguous. This is not a generally-used term and its use to mean the base of the cilium, PCM and satellites is spatially uninformative; there is not a 'compartment' meant here in the sense of a discrete structure. Especially in the context of the title, I suggest this phrase should be revised for greater clarity, but it may also be useful to rephrase it in the Discussion.
2. As a question for discussion/ consideration: Is TTBK2 a satellite component, or is it envisaged that these functions are all derived from a CEP164-associated fraction? This point is related to the point raised above about a 'compartment', in that different populations of TTBK2 might be involved differently in cilium regulation. An experimental approach to this might be to use a CEP164-TTBK2 fusion (such as was described by Cajanek and Nigg in PNAS 2014) and test if this can rescue the TTBK2 deficiency.
3. The statement on p.8 that 'Loss of TTBK2 results in increased actin activity' is not directly supported by the data, so should be revised. The actin analyses are indirect and show attenuated effects, so some caution is warranted in the interpretation of these findings (as is the case in the Discussion). With that point in mind, I am unsure if the title's inclusion of an actin-based mechanism for TTBK2 functions in cilium stability is optimal.
4. A key technical issue is that the description of how intensity measurements were made should be improved. It is not clear what area or volume was used for this in the various experiments that measured axonemal/ centrosomal/ peri-centrosomal regions, particularly when the centrioles are further apart from one another. As the intensity measurements form a key part of the paper, this should be clarified throughout.
5. From the images presented in Fig. 2A, the classification of the number of buds/ axonemal breaks is not clear. Improved images of the different outcomes of TTBK2 removal should be shown to make the basis for the proposed differentiation between these phenotypes more convincing. This is visible in the movies, but the still images are not ideal.
6. A control should be presented for the loss of TTBK2 in the drug treatment experiments in Figure 4, to confirm that there is no impact on the recombination.
7. A control for the relative expression level of the rescuing TTBK2-GFP protein should be provided in support of the data in Fig. S1D. This should also be included for the data in Fig. S4.

Minor comments:

1. Fig. 1B- 'lambda tubulin' should be corrected. Fig 7A should also correct the tubulin designation.
2. It would be helpful to indicate in individual figures throughout that the bar graphs show means +/- SEM (it is stated in the Methods, but it would be desirable to have the Figures be entirely self-contained).
3. Fig S1E does not show individual experiments as data points; this should be corrected for consistency.
4. The phospho-Aurora A staining in Fig. S2C should be quantitated; there appears to be an increased level of this signal in the absence of TTBK2.
5. The Ac-Tub staining shown in Fig. 3A is confusing, given the intensity measurements presented. More representative images should be shown.
6. It is unclear whether there is a decline in PCM1 intensity levels over the timecourse of the (vehicle only) experiment in Fig. 5B. This should be tested for. It should also be specified in the Figure Legend that these cells remain serum starved for the duration of the timecourse (assuming the experiment follows the design outlined in Fig. 1A).
7. The images in Figure 7A and 7C are not at sufficient resolution to distinguish IFT88 or IFT140 signals; blow-up panels should be included.
8. Scale bars should be included in Figs. 2, S2, 3, 5, S3, S4, 6, 7.
9. Details of the serum starvation regime should be provided in the Methods (% serum).

Referees cross-commenting

The comments from Reviewers 1 and 2 are detailed and constructive. There is good convergence between all three reviewers on a requirement for additional data on the involvement of actin and on the analysis of the centriolar satellites. As these are the main themes of the study, such information seems essential to support the principal conclusions drawn.

I question Reviewer 1's stipulation that a quantitative proteomic analysis of PCM1 interactors is needed to sustain the conclusion that the satellites are substantively altered upon TTBK2 depletion. This would be a very strong experiment, but I feel that the analysis of a selection of individual satellite components, as done here, is sufficient to support the conclusion that the satellites are impacted by TTBK loss. Obviously, the more detailed the analysis (i.e., the more proteins examined), the better, but I am not convinced that this will bring us so much closer to the mechanism of TTBK2. I concur with the point raised for revision by Reviewer 2, that the actual role of the satellites in cilium stability should be addressed more robustly, however.

Significance

The importance of the primary cilium as a signalling organelle makes TTBK2 function a theme of general interest. A potential role for TTBK2 in the maintenance of cilium stability through a link to the centriolar satellites is new. Readership would include people working on centrosomes/ cilia and related themes.

My expertise: cell biology of centrosomes/ cilia.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

This study by Nguyen and Goetz, explores the role of Tau tubulin kinase 2 (TTBK2) in cilium integrity in cultured cells. TTBK2 is a serine-threonine kinase that plays a key role in cilium initiation and has been implicated in maintaining assembled cilia in adult mice. The authors developed an inducible system to deplete TTBK2 once cilia are assembled to investigate the role of this kinase in cilium stability in a cell culture system. Using this system and live- and fluorescent- imaging approaches, the authors find that TTBK2 promotes cilium stability.

The authors suggest three parallel pathways by which TTBK2 maintains cilia: 1) through tubulin polyglutamylation and actin dynamics, 2) through regulation of IFT pools at the centrosome, and 3) through regulation of centriolar satellite composition. Some of these conclusions (especially 1 and 3) need to be revisited and need some additional experiments (detailed below).

Major comments:

1. In the abstract and discussion, the authors suggest that TTBK2 functions via centriolar satellites to promote cilium stability. While the data support a role for TTBK2 in centriolar satellite homeostasis, it is unclear whether satellites contribute to cilium stability. Given that restoration of centriolar satellite composition using rapamycin does not rescue cilia loss upon TTBK2 depletion, these sections need to be revised.
2. The authors show that TTBK2 regulates centriolar satellite composition via its kinase activity. Is TTBK2's kinase activity important for cilium stability, actin dynamics and IFT localization?
3. Do changes in tubulin poly-glutamylation precede the loss of cilia phenotype observed upon depletion of TTBK2? Currently changes in tubulin post-translational modifications are assessed at 72h post-tamoxifen treatment. In order to support the hypothesis that changes in tubulin poly-glutamylation drives cilia loss upon TTBK2 depletion (as suggested in the abstract), these experiments must be repeated at an earlier time points, such as 24-48 h post tamoxifin treatment).
4. By using small molecule inhibitors that alter actin dynamics, the authors suggest that loss of TTBK2 regulates actin polymerization leading to cilium instability. Are there any observable changes in cellular F-actin upon loss of TTBK2? Phalloidin staining and/or a biochemical assay to directly assess soluble vs. filamentous actin would be helpful to bolster the claim that TTBK2 regulates actin dynamics.

Minor comments:

The manuscript is well written, and easy to follow. Prior studies are referenced appropriately. I have some minor points that would improve the presentation and clarity of the manuscript:

1. I would recommend the authors improve the presentation of figures by making the size of graphs more consistent across figures. For example, graphs in Figure 4 should be increased in size as they are currently very hard to read. Importantly, scale bars are missing from most immunofluorescence images.
2. Statistical tests are missing in Figure S1 A and S1 E.
3. Are the additional puncta observed around the centrosome in the TTBK2 immunofluorescence images (Figure 1B) non-specific signals? Also, gamma-tubulin is mislabeled as lambda-tubulin in the figure.
4. Insets would be helpful for images Figure 1C, S1C, 2A, 4B, 7A, 7C.
5. Quantification of fluorescence intensity is missing for Figure S2 C.
6. The title for Figure S2 is misleading as it suggests there is a change in cilia disassembly factors upon loss of TTBK2, while the data show no changes in any of the factors assessed.

Significance

Although we now have some understanding of how cilium assembly is initiated, how the cilium is maintained at steady state remains poorly understood. Therefore, this study exploring the role of TTBK2 in ciliary structure maintenance is timely and will be of interest to cilia biologists. TTBK2 has previously been implicated in cilium maintenance, and the link between TTBK2 and IFT recruitment, and tubulin post-translational modifications has been previously described using hypomorphic mutants of TTBK2. Though this study specifically looks TTBK2's role in cilium maintenance at steady state, these previous studies (referenced in the manuscript) do diminish the novelty of the manuscript. Mechanistic details of how TTBK2 regulates actin, IFT dynamics and tubulin post-translational modifications to control cilium stability remain unknown but are important avenues for future research.

Reviewers' expertise: cilia and centrosome biology, microscopy.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

TTBK2 is kinase mutated in spinocerebellar ataxia type 11 (SCA11). It is well characterized for its functions and molecular mechanism of action during cilium assembly and ciliary signaling. In this manuscript, Nguyen et al. investigated the role of TTBK2 during cilium stability using mouse embryonic fibroblasts derived from Ttbk2cmut embryos. This system allowed them to inducibly deplete TTBK2 in ciliated cells and thus, address the specific functions of TTBK2 during cilium maintenance and stability in ciliated cells. Upon TTBK2 loss, the ciliary axoneme was destabilized in part via an increased frequency of cilia breaks and primary cilia was lost, which was shown by live imaging experiments. To dissect the mechanism of axoneme destabilization, they performed rescue experiments with drug treatments as well as quantified of basal body, satellite and ciliary abundance of key ciliogenesis factors and tubulin modifications. Based on results from these experiments, the authors concluded that altered actin dynamics, reduction in axonemal polyglutamylation as well as changes in satellite-associated PCM1 and OFD1 and IFT levels at the basal body together underlies the axonemal destabilization phenotypes associated with TTBK2. Overall, the manuscript is well-written and the presented data is robust.

I list below three major concerns I have on the manuscript along with detailed explanation. Although questions addressed in the manuscript and the tools the authors generated will be of general interest, the presented data falls short in supporting the major conclusions of the manuscript that pertain to the mechanisms by which TTBK2 regulate cilium stability and maintenance.

1. Previous papers from the Goetz lab showed that TTBK2 is important for the structure and stability of the ciliary axoneme (i.e. reduced polyglutamylation of cilia in Ttbk2 hypomorphic mutants and null cells, disorganized axonemal microtubules). Although authors study the roles of TTBK2 in cilium stability with temporal control and identify altered centriolar satellite composition and actin dynamics as potential mechanisms, TTBK2's function in this process is not unexpected.
2. The authors suggest that TTBK2 regulate cilium stability through parallel pathways that operate via actin, centriolar satellites, autophagy and IFT machinery. The presented data is not sufficient to assess whether these are direct or indirect consequences of TTBK2 depletion, which results in lack of a coherent model for how TTBK2 operates in ciliated cells. For example, what is the regulatory/functional link between TTBK2, actin and myosin VI during cilium maintenance? The authors discuss that TTBK2 BioID data includes actin-binding proteins and suggest actin polymerization as one potential mechanism. To gain insight into how TTBK2 alters actin dynamics, they can follow-up on the BioID hits to explain how TTBK2 depletion alters actin dynamics. Alternatively, they can treat cells with specific inhibitors of actin polymerization to determine whether axonemal destabilization phenotypes are rescued.
3. The authors define changes in centriolar satellite composition as a consequence of TTBK2 depletion based on reduced PCM1 and elevated OFD1 and CEP290 intensity at the pericentrosomal region in TTBK2-depleted cells. Centriolar satellites are composed of about 200 proteins and a significant number of these proteins are implicated in ciliogenesis including the previously characterized interactors of TTBK2. Changes in PCM1, OFD1 and CEP290 levels in the 1 uM ROI authors defined around the basal body is not sufficient to conclude that satellite composition is altered and that this change underlies the axoneme destabilization and disassembly. Proteomic pulldown of PCM1 before and after tamoxifen addition will reveal how the satellite proteome is affected by TTBK2 depletion and will strengthen authors' conclusions.

Below are other comments I have on the data and its analysis and presentation:

1. Fig. 1B: TTBK2 at the basal body was assessed as "positive" or "negative". Instead of classification into two groups, quantification of the basal body levels of TTBK2 in a time course manner will be more informative in correlating phenotype with TTBK2 depletion.
2. Fig. 1C: In addition to percentage of ciliated cells, the cilium length should also be quantified in a time course manner to determine how TTBK2 depletion leads to cilium disassembly by 48-72 h. For representative images presented, insets are required.
3. Fig. S1A: Western blot quantification of TTBK2 levels in addition to mRNA levels will be informative in assessing changes in protein levels upon tamoxifen addition.
4. Statistical analysis for Fig. 2B is required. How many cilia / experimental replicates were quantified in Fig. 2B?
5. Fig. S2: Bowie et al. 2018 paper reported increased Kif2a levels at the basal body as one possible mechanism for cilium instability in TTBK2 mutant cells. However, TTBK2 depletion in ciliated cells does not have a similar effect. How do the authors explain this difference on effects of TTBK2 in basal body Kif2a levels?
6. How did the authors quantify between budding and axonemal events in Fig. 2? In general, the methods section for analysis of the microscopy data should be written in a more detailed way to be able to assess the presented data.
7. Acetylated tubulin, but not glut tubulin, ciliary levels are decreased upon TTBK2 depletion. Can such change directly affect cilium stability? What triggers changes in glutamylation upon TTBK2 depletion?
8. "Loss of TTBK2 results in an increase in actin activity" (pg.8) is an overconclusion based on the presented data. What is the effect of TTBK2 depletion on actin levels and organization? Specifically, the authors can also stain for actin in the cilia to determine whether it underlies the increased excision events reported by live imaging of cilia. Given that Nager et al. 2017 paper identified actin-myosin activity as a mechanism for ciliary ectocytosis, this will be interesting to test using similar assays to quantify actin dynamics in this paper. Moreover, Yeyati et al. 2018 paper on dissecting KDM3A's function during actin dynamics and cilium stability used quantitative approaches that can be adapted by the authors.
9. Quantification of centriolar satellite levels of PCM1, OFD1 and CEP290 can be done in a different way to more specifically identify the satellite pools of these proteins. The quantification of their pericentrosomal levels was done by drawing 1 uM ROI around the centrosome. Therefore, the levels might represent changes in the centrosomal pool of these proteins, but not the satellite pool, which would then change the author's conclusions. A method that the authors can adapt is the Gupta et al. Cell 2015 paper. Such quantification will ensure that the authors are not drawing their conclusions based on changes in basal body levels of the proteins. Moreover, the representative images presented for centriolar satellite pools of these proteins in Fig. 5, Fig. 6 and Fig. S4 should be modified to include not only the basal body pool but the whole cell including an inset. Since satellites are distributed throughout the cell, presenting whole cell images is important to assess the remaining pool beyond the basal body.
10. Does TTBK2 depletion alter the cellular abundance of CEP290, OFD1 and PCM1? This might underlie the changes in the basal body abundance of these proteins.
11. Fig. 4C-E: Graphs are too small compared to figures.
12. In multiple parts of the manuscript, the authors stated that the mechanisms of TTBK2 during cilium initiation is not known. Including major papers from the Goetz lab, there are many studies in literature that defines how TTBK2 regulates cilium initiation. The major unknowns relates to its functions during cilium maintenance and disassembly.

Significance

Although cilium assembly has been studied with extensive detail, relatively less is known about them mechanisms of cilium maintenance. This is in part due to lack of tools to manipulate protein expression in ciliated cells. The MEF cells used in this study is an elegant tool to address questions related to cilium maintenance and stability. Moreover, the live imaging experiments performed in TTBK2 depleted ciliated cells are excellent experiments in showing the spatiotemporal events that lead to axoneme destabilization. By identifying TTBK2 as a critical regulator of these processes, the results of the manuscript advances our understanding of how cilium is maintained as well as to the molecular defects that underlie SCA11. The topic is also of general interest to cell and developmental biologists.

As a reviewer, my expertise is on questions that pertain to the biogenesis of centrosome and cilium and we extensively use cell biology, proteomics and biochemistry approaches.

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

Learn more at Review Commons

---

Reply to the reviewers

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

The manuscript by Neville et al addresses the link between the localization and the activity of the so-called "Pins complex" or "LGN complex", that has been shown to regulate mitotic spindle orientation in most animal cell types and tissues. In most cell types, the polarized localization of the complex in the mitotic cell (which can vary between apical and basolateral, depending on the context) localizes pulling forces to dictate the orientation. The authors reexplore the notion that this polarized localization of the complex is sufficient to dictate spindle orientation, and propose that an additional step of "activation" of the complex is necessary to refine positioning of the spindle.

The experiments are performed in the follicular epithelium (FE), an epithelial sheet of cell that surrounds the drosophila developing oocyte and nurse cells in the ovarium. Like in many other epithelia, cell divisions in the FE are planar (the cell divides in the plane of the epithelium). The authors first confirm that planar divisions in this epithelium depends on the function of Pins and its partner mud, and that the interaction between the two partners is necessary, like in many other epithelial structures. Planar divisions are often associated with a lateral/basolateral "ring" of the Pins complex during mitosis. The authors show that in the FE, Pins is essentially apical in interphase and becomes enriched at the lateral cortex during mitosis, however a significant apical component remains, whereas mud is almost entirely absent from the apical cortex. Pins being "upstream" of mud in the complex, this is a first hint that the localization of Pins is not sufficient to dictate the localization of mud and of the pulling forces.

The authors then replace wt Pins, whose cortical anchoring strongly relies on its interaction with Gai subunits, with a constitutively membrane anchored version (via a N-terminal myristylation). They show that the localization of myr-Pins mimics that of wt-Pins, with a lateral enrichment in mitosis, and a significant apical component. Since a Myr-RFP alone shows a similar distribution, they conclude that the restricted localization of Pins in mitosis is a consequence of general membrane characteristics in mitosis, rather than the result of a dedicated mechanism of Pins subcellular restriction. Remarkably, Myr-Pins also rescues Pins loss-of-function spindle orientation defects.

They further show that the cortical localization of Pins does not require its interaction with Dlg (unlike what has been suggested in other epithelia). However, spindle orientation requires Dlg, and in particular it requires the direct Dlg/Pins interaction. The activity of Dlg in the FE appears to be independent from khc73 and Gukholder, two of its partners involved in its activity in microtubule capture and spindle orientation in other cell types. Based on all these observations, the authors propose that Dlg serves as an activator that controls Pins activity in a subregion of its localization domain (in this case, the lateral cortex of the mitotic FE cell). They propose to test this idea by relocalizing Pins at the apical cortex, using Inscuteable ectopic expression. With the tools that they use to drive Inscuteable expression, they obtain two populations of cells. One population has a stronger apical that basolateral Insc distribution, and the spindle is reoriented along the apical-basal axis; the other population has higher basolateral than apical levels of Insc distribution, and the spindle remains planar. The authors write that Pins localization is unchanged between the two subsets of cells (although I do not entirely agree with them on that point, see below), and that although Mud is modestly recruited to the apical cortex in the first population, it remains essentially basolateral in both. In this situation, the localization of Insc in the cell is therefore a better predictor of spindle orientation than that of Pins or Mud. Remarkably, removing Dlg in an Insc overexpression context leads to a dramatic shift towards apical-basal reorientation of the spindle, suggesting that loss of Dlg-dependent activation of the lateral Pins complex reveals an Insc-dependent apical activation of the complex. Overall, I find the demonstration convincing and the conclusion appropriate. One of the limitations of the study is the use of different drivers and reporters for the localization of Pins, which makes it hard to compare different situations, but not to the point that it would jeopardize the main conclusions. I do not have major remarks on the paper, only a few minor observations and suggestion of simple experiments that would complete the study.

Minor:

What happens to Pins and Mud in Dlg mutant cells that overexpress Insc and behave as InscA? Are they still essentially lateral, or are they more efficiently recruited to the apical cortex?

This is a terrific question. Of course we would love to know and intend to find out.

One way to do this (consistent with the manuscript) would be to generate flies that are Dlg[1P20], FRT19A/RFP-nls, hsflp, FRT19A; TJ-GAL4/+; Pins-Tom, GFP-Mud/UAS-Insc. (Note that these flies would only allow us to image Mud; we would have to repeat the experiment using GFP FRT19A; hsflp 38 to see Pins. This isn’t ideal given that we’d like to image both together). Generating these flies is a major technical challenge because of the number of transgenes and chromosomes involved.

Our preferred way to do this would be to generate flies that are Dlg[1P20]/Dlg[2]; TJ-GAL4/+; Pins-Tom, GFP-Mud/UAS-Insc. So far, we’ve been unsuccessful. We are now undertaking a modified crossing scheme that we hope will solve the problem, though we aren’t overly optimistic about the outcome. We find that the temperature-sensitive mutation Dlg[2] presents an activation barrier; while we are able to generate flies that are Dlg[2] / FM7 in combination with transgenes and/or mutations on other chromosomes, we do not always recover the Dlg[2] / Y males (which must develop at 18degrees) from these complex genotypes.

In the longer term (outside the scope of revision), we are working to develop more tools for imaging Mud and Pins that we hope will help answer this question.

Regarding the competition between Pins and Insc for dictating the apical versus basolateral localization of Insc, the Insc-expression threshold model could be easily tested in Pins62/62 mutants, where it is expected that only InscA localization should be observed, even at 25{degree sign}C (unless Pins is required for the cortical recruitment of Insc, as it is the case in NBs - see Yu et al 2000 for example).

This is another great experiment and one we’d love to carry out. Again, the genetics are currently challenging, only because both UAS-Inscuteable and FRT82B pinsp62 are on the third chromosome. (Right now we’re trying to hop UAS-Inscuteable to the second).

However, we do have another idea for testing the threshold model, which is to repeat the experiment in which we express UAS-Insc in cells that are DlgIP20/IP20 at 25oC. Because the relevant cells (UAS-Insc OX in Dlg mitotic clones) are relatively rare, we have not yet been able to collect enough examples to make a firm conclusion. However, our preliminary results (only six cells so far!) suggest that more InscB cells are observed at the lower temperature, consistent with the threshold model.

I do not agree with the authors on P.10 and Figure 6A-D, when they claim that the apical enrichment of Pins is equivalent in both InscA and InscB cells. The number of measured cells is very low, and the ratio of apical/lateral Pins differs between the two sets of cells. The number of cells should be increased and the ratios compared with a relevant statistic method.

Totally fair. We are working to add more data to these panels (6B and 6D). The trend observed in 6D may be softening in agreement with the reviewer’s prediction, although we currently don’t yet have enough new data points to be confident in that conclusion. Therefore, we have not yet updated the manuscript, though we expect to do so during the revision period. We will also add a statistical comparison. Importantly, as the reviewer suggested, this does not alter our conclusions.

A lot of the claims on Pins localization rely on overexpression (generally in a Pins null background) of tagged Pins expressed from different promoters or drivers, and fused to different fluorescent tags. Therefore, it is difficult to evaluate to which extent the localization reflects an endogenous expression level, and to compare the different situations. As the cortical localization of Pins relies on interaction with cortical partners (mostly GDP-bound Gai) which are themselves in limiting quantity in the cell, and in the case of Gai-GDP, regulated by Pins GDI activity, this poses a problem when comparing their distribution, because the expression level of Pins may contribute to its cortical/cytoplasmic ratio, but also to its lateral/apical distribution. Although I understand that the authors have been using tools that were already available for this study, I think it would be more convincing if all the Pins localization studies were performed with endogenously tagged Pins, even those with Myr localization sequences. In an age of CRISPR-Cas-dependent homologous recombination, I think the generation of such alleles should have been possible. Although this would probably not change the main claims of the paper, it would have made a more convincing case for the localization studies.

We don’t disagree at all with this point. We did indeed try to stick with the published UAS-Pins-myr-GFP, not only for convenience but because it allows us to make comparisons to other studies using the same tool (Chanet et al Current Biology 2017 and Camuglia et al eLife 2022). Another consideration is that we used only one driver across our experiments (Traffic jam-GAL4). It is quite weak at the developmental stages that we examine, meaning that overexpression is not a major concern. (Indeed we have struggled with the opposite problem).

We certainly take the reviewer’s comment seriously and we therefore described it in the manuscript. We are currently working to develop endogenous tools using CRISPR.

Paragraph added to Discussion – Limitations of our Study:

“Another technical consideration is that our work makes use of transgenes under the control of Traffic jam-GAL4. While this strategy allows us to compare our results with previous work employing the same or similar tools, a drawback is that we cannot guarantee that Traffic jam-GAL4 drives equivalent expression to the endogenous Pins promoter (Chanet et al., 2017, Camuglia et al., 2022). However, given that Traffic jam-GAL4 is fairly weak at the developmental stages examined, we are not especially concerned about overexpression effects.”

The authors should indicate in the figure legends or in the methods that the spindle orientation measurements for controls for Pins62/62 are reused between figures 1, 3, 4, 5, 6 , and between figure 3, 4 and 5, respectively.

Absolutely. Added to the Methods section.

Reviewer #1 (Significance (Required)):

Altogether, this study makes a convincing case that the localization of the core members of the pulling force complex, Pins and Mud, is not entirely sufficient to localize active force generation, and that the complex must be activated locally, at least in the FE. The notion of activation of the Pins/LGN complex has probably been on many people's mind for years: Pins/LGN works as a closed/open switch depending on the number of Gai subunits it interacts with, it must be phosphorylated, etc... suggesting that not all cortical Pins/LGN was active and involved in force generation. However the study presented here shows an interesting case where localization and activation are clearly disconnected. The authors show how Dlg plays this role in physiological conditions in the FE, and use ectopic expression of Insc to show that, at least in an artificial context, Insc can have the same "activating activity" (or at least an activating activity that is stronger than its apical recruitment capability and stronger than Dlg's activating activity). It is to my knowledge the first case of such a clear dissociation. In their discussion, the authors are careful not to generalize the observation to other tissues. Although I did not reexplore all that has been published on the Pins/LGN-NuMA/Mud complex over the last 20 years, my understanding is that despite interesting cases of distribution of the complex like that of Mud in the tricellular junction in the notum, the localization model can still explain most of the phenotypes that have been described without invoking an activation step. If it is the case, then the activation model is another variation (an interesting one!) on the regulation of the core machinery, which are plentiful as the authors indicate in their introduction, and is maybe specific to the FE; if not, then it would be interesting to push the discussion further by reexamining previous results in other systems, and pinpointing those phenotypes that could be better explained with an activation step.

Overall, I find this is an elegant piece of work, which should be of interest to many cell and developmental biologists beyond the community of spindle orientation aficionados.

Reviewer #2 (Evidence, reproducibility and clarity (Required)): Summary: The manuscript by Neville et al. addressed the mechanism how conserved spindle regulators (Pins/Mud/Gai/Dynein) control spindle orientation in the proliferating epithelia by revising "the canonical model", using the Drosophila follicular epithelium (FE). The authors examined the epistatic relationship among Pins, Mud and Dlg in FE and found that Pins controls the cortical localization of Mud by utilizing mutant analyses, and suggested their localization does not fully overlap using the newly generated knock-in allele. They also showed that Pins relocalization during mitosis depends on cortical remodeling, or passive model, where Pins localization changes with other membrane-anchored proteins. Their data further suggest that Pins cortical localization is not influenced by Dlg, but Pins-interacting domain of Dlg does affect spindle orientation. Based on these results, the authors propose that Dlg controls spindle orientation not by redistributing Pins, but by promoting (or "activating" from their definition) Pins-dependent spindle orientation. Interestingly, ectopic expression of Inscuteable (Insc) suggested that Insc localization, either apical or lateral, correlates with spindle orientation, and their localization is a dominant indicator of spindle orientation, compared to the localization of Pins and Mud, implicating potentially distinct roles of activation and localization of the spindle complex. Overall their genetic experiments are well-designed and provide stimulation for future research. However, their evidence is suggestive, but not conclusive for their proposal. I have several concerns about their conclusion and would like to request more detailed information as well as to propose additional experiments.

Major concerns: 1. This report lacks technical and experimental details. As the typical fly paper, the authors need to show the exact genotypes of flies they used for experiments. This needs to be addressed for Figures 1-6, and Supplemental Figures. Especially, which Gal4 drivers were used for UAS-Pins wt or mutant constructs in Figure 4 with pins mutant background, Khc73, GUKH mutant backgrounds. Which exact flies were used for mutant clone experiments for Supplemental Figure 3? (A for typical mosaic, and B for MARCM). Without these details, it is impossible to evaluate results and reproduce by others.

We take this concern very seriously!

• We listed the GAL4 driver (Traffic jam-GAL4) in the first section of the Materials and Methods: Expression was driven by Traffic Jam-GAL4 (Olivieri et al., 2010). The transgene and relevant citation have been added to Table 1.
• We explained the background stock for the MARCM experiment in the Materials and Methods: Mosaic Analysis with a Repressible Cell Marker (after the method of Lee and Luo) was carried out using GFP-mCD8 (under control of an actin promoter) as the marker. The transgene and relevant citation have been added to Table 1.
• In line with other fly studies (eg. Nakajima et al., Nature 2013) and our own Drosophila work (Bergstralh et al Current Biology 2013, Bergstralh*, Lovegrove*, St Johnston NCB 2015, Bergstralh et al Development 2016, Finegan et al EMBO J 2019, Cammarota*, Finegan* et al Current Biology 2020) we were careful to show the relevant genotype components in each figure.
• We included a fully referenced Supplementary Table (Table 1 – Drosophila genetics) listing every mutant allele or transgene with a citation and a note about availability. We have expanded this table in response to the author’s concern (see above).

  Related to the comment 1, how did the author perform "clonal expression of Ubi-Pin-YFP" in page 5? As far as I understand, Ubi-Pin-YFP is expressed ubiquitously by the ubiquitin promoter.

The reviewer makes a good point. We regret that we did not make this experiment more clear. Ubi-Pins-YFP was recombined onto an FRT chromosome (FRT82B). We made mitotic clones.

We have clarified this in the Methods section as follows:

“Mitotic clones of Ubi-Pins-YFP were made by recombining the Ubi-Pins-YFP transgene onto the FRT82B chromosome”

1. In page 6, if Pins relocalization is passive and is associated with membrane-anchored protein remodeling during mitosis, its relocalization can be suppressed by disrupting the process of mitotic remodeling (mitotic rounding). The authors should test this by either genetic disruption or pharmacological treatment for the actomyosin should cause defects in Pins relocalization, which bolster their conclusion.

We agree that this is a cool experiment and are happy to give it another shot. However, we do note that interpretation could be difficult. We don’t know that mitotic rounding and membrane-anchored protein remodeling during mitosis are inextricably linked. Notably, the remodeling we describe reflects cell polarity; apical components are evidently moved to the lateral cortex. This is contrary to understanding of rounding, which reflects isotropic actomyosin activity (Chanet et al., (2017) Curr.Biol. & Rosa et al., (2015) Dev. Cell.). Therefore we don’t understand what a “negative” result would mean, or for that matter that a “positive” result would be safe to interpret.

We have attempted many strategies to prevent cell rounding in the follicular epithelium, none of which have successfully prevented rounding. 1) We attempted to genetically knockdown Moesin in the FE and did not see an effect on cell rounding. However we couldn’t confirm knockdown and therefore are not confident in this manipulation. 2) It is difficult to interpret the result of genetically disrupting Myosin, because it causes pleiotropic effects, such as inhibition of the cell cycle, and disruption of monolayer architecture. 3) We treated egg chambers with Y-27632 (a Rok-inhibitor) and examined its effect on mitotic cell rounding and on cytokinesis, which are Rok-dependent processes. Our experiments were performed using manually-dissociated ovarioles treated for 45 minutes in Schneider Cell Medium supplemented with insulin. Even at our maximum concentration of 1mM Y-27632, several orders of magnitude above the Ki, we are unable to see any effect on mitotic cell shape or actin accumulation at the mitotic cortex and did not observe any evidence of defective cytokinesis. We also did not observe defects in spindle organization or orientation, as would be expected from failed rounding. We therefore do not believe that the inhibitor works in this tissue. One possible explanation is that the follicle cells are secretory, and likely to pass molecules taken up from the media quickly into the germline. Therefore, we do not anticipate that we can perform this experiment to our satisfaction.

1. The critical message in this manuscript is that the core spindle complex mediated by Pins-Mud controls spindle orientation by "activation", but not localization. The findings that Pins and Mud localization is not influenced by Insc and that ecotpic Insc expression and genetic Mud depletion (Figure 6) might support their proposal, but these results just suggest their localization does not matter. I wonder how the authors could conclude and define "activation". What does this activation mean in the context of spindle orientation? Can the authors test activation by enzymatic activity or assess dynamics of spindle alignment?

We intend for the critical message of the manuscript to be that “The spindle orienting machinery requires activation, not just localization.” We absolutely do not make the claim that localization is not important, only that it is not sufficient. The reviewer recognizes this point here and in a subsequent comment: “The authors showed that Pins and Mud localization themselves are not sufficient for the control of spindle orientation with genetic analyses.”

We also do not claim that Pins and/or Mud localization are not impacted by Inscuteable. On the contrary, we plainly see and report that they are; the intensity profiles in Figure 6 are distinct from those in Figure 2, as discussed in the text.

We appreciate the reviewer’s point about activation. Since we do not understand these proteins to be enzymes, we aren’t sure what enzymatic activity would be tested. The dynamics of spindle alignment in this slowly developing system are prohibitively difficult to measure: the mitotic index is very low (~2%) and only a very small fraction of those cells will be in a focal plane that permits accurate live imaging in the apical-basal axis. Alternative modes of activation include conformational change and/or a connection with other important molecules. The simplest possibility would be that Dlg allows Pins to bind Mud, but so far our data do not support it. We have added the following paragraph to our discussion:

“The mechanism of activation remains unclear. While the most straightforward possibility is that Dlg promotes interaction between Pins and Mud, our results show that Mud is recruited to the cortex even when Dlg is disrupted (Figure 4D). Alternatively, Discs large may promote a conformational change in the spindle-orientation complex and/or a change in complex composition. Furthermore, the Inscuteable mechanism is not likely to work in the same way. Dlg binds to a conserved phosphosite in the central linker domain of Pins and should therefore allow for Pins to simultaneously interact with Mud (Johnston et al., 2009). Contrastingly, binding between Pins and Inscuteable is mediated by the TPR domains of Pins, meaning that Mud is excluded (Culurgioni et al., 2011; 2018). While a stable Pins-Inscuteable complex has been suggested to promote localization of a separate Pins-Mud-dynein complex, our work raises the possibility that it might also or instead promote activation.”

1. In page 7-8, although Pins-S436D rescue spindle orientation, but Pins-S436A does not in pins null clones background, Pins localization is not influenced by Dlg. This questions how exactly Pins and Dlg can interact, and how Dlg affect Pins function. Related to this observation, in the embryonic Pins:Tom localization in dlg mutant does not provide strong evidence to support their conclusion given the experimental context is different from previous study (Chanet et al., 2017).

We agree with the reviewer. Our data (this paper and previous papers) and the work of others indicate that this interaction is important for spindle orientation (Bergstralh et al., 2013a; Saadaoui et al., 2014; Chanet et al., 2017). However, we show here that Dlg doesn’t obviously impact Pins localization (as proposed in our earlier paper), but does impact the ability of the spindle orientation machinery to work (hence activity).

The reviewer makes a very good point. Our experimental context is different from the previous study concerning Pins and Dlg in embryos: Chanet et al (2017) performed their work in the embryonic head, whereas we look at divisions in the ventral embryonic ectoderm. These are distinct mitotic zones (Foe et al. (1989) Development) and exhibit distinct epithelial morphologies. We show that Pins:Tom localizes at the mitotic cell cortex in Dlg[2]/Dlg[1P20] in cells in the ventral embryonic ectoderm. Our only conclusion from this experiment is that Pins:Tom can localize without the Dlg GUK domain in another cell type (outside the follicular epithelium). In the current preliminary revision we have softened our claim as follows:

“We also examined the relationship between Pins and Dlg in the Drosophila embryo. A previous study showed that cortical localization of Pins in embryonic head epithelial cells is lost when Dlg mRNA is knocked down (Chanet et al., 2017). We find that Pins:Tom localizes to the cortex in the ventral ectoderm of early embryos from Dlg1P20/Dlg2 mothers, indicating that Pins localization in the ventral embryonic ectoderm epithelium does not require direct interaction with Dlg. We therefore speculate that Dlg plays an additional role in that tissue, upstream of Pins (Figure 4G).”

Our intention is to elaborate on our findings with additional data from embryos. To this end we have already acquired preliminary control data investigating the spindle angle with respect to the plane of the epithelium, and are in the process of examining spindle angles in dlg mutant embryonic tissue.

In page 11, the authors state "... that activation of pulling in the FE requires Dlg". I was not convinced by anything related to "pulling". There is no evidence to support "pulling" or such dynamic in this paper, just showing Mud localization, correct?

We appreciate the reviewer’s concern. The original sentence read that “We interpret [our data] to mean that interaction between Pins and Dlg, which is required for pulling, stabilizes the lateral pulling machinery even if Dlg is not a direct anchor.” This statement is based on work across multiple systems, including the C. elegans embryo (Grill et al Nature 2001), the Drosophila pupal notum (Bosveld et al, Nature 2016), and HeLa cells (Okumura et al eLife 2018), which shows that Mud/dynein-mediated pulling (on astral microtubules) orients/positions spindles. This is described in the introduction.

To address the reviewer’s particular concern, we have replaced “pulling” with “spindle-orentation machinery,” so that this sentence now reads …“activation of the spindle-orientation machinery in the FE requires Dlg.”

1. Ectopic expression of Insc (Figure 6) provided a new idea and hypothesis, but the conclusion is more complicated given that Insc is not expressed in normal FE. For example, the statement that "Inscuteable and Dlg mediate distinct and competitive mechanism for activation of the spindle-orienting machinery in follicle cells" is probably right, but it does not show anything meaningful since Insc does not exist in normal FE. Is Dlg in a competitive situation during mitosis of FE? If so, which molecules are competitive against Dlg? The important issue is to provide a new interpretation of how spindle orientation is controlled epithelial cells. I strongly recommend to add models in this manuscript for clarity.

We considered the addition of model cartoons very carefully in preparing the original manuscript, and again after review. While we are certainly not going to “dig in” on this point, our concern is that model figures would obscure rather than clarify the message. As the reviewer points out, we do not understand how activation works, and as discussed in the manuscript we don’t think it’s likely to work the same way in follicle cells (Dlg) as it does in neuroblasts (Insc). Therefore model figure(s) are premature.

We do not agree with the statement that "Inscuteable and Dlg mediate distinct and competitive mechanism for activation of the spindle-orienting machinery in follicle cells… does not show anything meaningful.” This is a remarkable finding because it suggests that there is more than one way to activate Pins. Given the critical importance of spindle orientation in the developing nervous system, and the evolutionary history of the Dlg-Pins interaction, we think that this finding supports a model in which the Dlg-Pins interaction evolved in basal organisms, and a second Inscuteable-Pins interaction evolved subsequently to support neural complexity. These ideas are raised in the Discussion.

The reviewer also writes that “The important issue is to provide a new interpretation of how spindle orientation is controlled epithelial cells.” We find this concern perplexing, since the reviewer clearly recognizes that we have provided a new interpretation: Dlg is not a localization factor but rather a licensing factor for Pins-dependent spindle orientation.

Minor comments: 8. Some sections were not written well in the manuscript. "It does not" in page 6. "These predictions are not met". I just couldn't understand what they stand for. Their writing has to be improved.

Again, we are not going to dig in here, but we would prefer to retain the original language, which in our opinion is fairly clear. Our study is hypothesis-driven and based on assumptions made by the current model. We used direct language to help the reviewer understand what happened when we tested those assumptions.

1. In page 9, Supplementary Figure 4 should be cited in the paragraph (A potential strategy for..), not Supplemental Figure 1A, and 1B.

Good catch, thank you! We have corrected this.

1. In page 10, the authors examine aPKC localization in Insc expressing context of FE. Does aPKC localization correlate with Insc localization (Insc dictates aPKC?)? aPKC is not involved in spindle orientation from the author's report (Bergstralh et al., 2013), so it does not likely provide any supportive evidence.

I’m afraid we don’t entirely understand this comment. The interdependent relationship between aPKC and Inscuteable localization is long-established in the literature and was previously addressed in the follicle epithelium (Bergstralh et al. 2016). We do not make the claim here that aPKC governs spindle orientation. We are emphasizing that the difference between InscA and InscB cells extends to the relocalization of polarity components involved in Insc localization. As described in the manuscript, these data are provided to support our threshold model:

“In agreement with interdependence between Inscuteable and the Par complex, we find that aPKC is stabilized at the apical cortex in InscA cells but enriched at the lateral cortex in InscB cells (Figure 6E). This finding is consistent with an Inscuteable-expression threshold model; below the threshold, Pins dictates lateral localization of Inscuteable and aPKC. Above the threshold, Inscuteable dictates apical localization of Pins and aPKC.”

1. In Dicussion page 12, "In addition, we find that while the LGN S408D (Drosophila S436D) variant is reported to act as a phosphomimetic, expression of this variant has no obvious effect on division orientation (Johnston et al., 2012)". Where is the evidence for this? I interpret that this phosphomimetic form can rescue like wt-Pins not like unphospholatable mutant S436A, so it means that have an effect on spindle orientation, correct?

The reviewer makes a good point. We regret the confusion. We mean to point out that the S436D variant is no different from the wild type. We have amended the text to clarify:

“In addition, we find that while the LGN S408D (Drosophila 436D) variant is reported to act as a phosphomimetic, this variant does not cause an obvious mutant phenotype in the follicular epithelium (Johnston et al., 2012). What then is the purpose of this modification? Since the phosphosite is highly conserved through metazoans, one possibility to consider is that the phosphorylation regulates the spindle orientation role of Pins, whereas unphosphorylated Pins plays a different role (Schiller and Bergstralh, 2021).”

Reviewer #2 (Significance (Required)):

The authors showed that Pins and Mud localization themselves are not sufficient for the control of spindle orientation with genetic analyses. While the authors tried to challenge the concept of "canonical model", there is no clear demonstration of "activation" of the spindle complex. I appreciate their genetic evidence and new results, and understand the message that Pins and Mud effects are beyond localization, but there is no alternative mechanism to support their model. At the current stage, their evidence provides more hypothesis, not conclusion. Based on my expertise in Developmental and Cell biology, I suggest that the work has an interest in audience who studies spindle machinery, but for general audience.

We think that the reviewer fundamentally shares our perspective on the study. Our work tests assumptions made by the canonical model and shows that they aren’t always met (meaning that the question of how spindle orientation works in epithelia at least is still unsolved), and that in the FE at least one component (Dlg) has been misunderstood. We reach a major conclusion, which is that localization of Pins is not enough to predict spindle orientation in the FE.

It’s absolutely true that the precise molecular role of Dlg has not been solved by our study. This is a major question for the lab, and we are currently undertaking biochemical work to address it. It’s probably more work than we can (or should) do on our own, which is just one reason to share our current results with colleagues.

One fundamental reason for undertaking this study is that 25 years of spindle orientation studies released into an environment in which “positive” conclusions are the bar for publication success may have burdened the field with claims that are overly-speculative. We appear to have contributed to this problem ourselves in 2013. With that in mind we contend that providing an alternative molecular mechanism at this point is premature and would impair rather than improve the literature.

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

Neville et al re-examine the role and regulation of Pins/LGN in Drosophila follicular epithelial cells. They argue that polar or bipolar enrichment of Pins localisation at the plasma membrane is not crucial for spindle orientation, and therefore propose that Pins must be somehow activated to function. These interpretations are not supported by the data. However, the data strongly suggest an alternative interpretation which is of major biological significance.

As an initial point, we disagree with the summary above. We do not argue that enrichment of Pins is not crucial for spindle orientation. We argue that enrichment of Pins is not sufficient. This is why we titled the paper “The spindle orienting machinery requires activation, not just localization” instead of “The spindle orienting machinery requires activation, not localization.”

Although we disagree with reviewer, we appreciate their criticism of our manuscript and are glad for the opportunity to clarify our findings. In our responses to the specific comments (below) we explain why our data contradict the reviewer’s model and what we will do to improve the manuscript.

Comments:

1. In the experiments on Dlg mutants (Fig 4D, S3) visualising Pins:Tom, the wild-type needs to be shown next to the Dlg mutant image, otherwise a comparison cannot be made. For example, Pins:Tom looks strongly enriched at the lateral membranes in the wild-type shown in Fig 2B&C, but much more weakly localised at the lateral membranes in Dlg1P20/2 mutants in Fig 4D. Thus, it looks like the Dlg GUK domain is required for full enrichment of Pins:Tom at lateral membranes, even if some low level of Pins can still bind to the plasma membrane in the absence of the Dlg GUK domain. Quantification would likely show a reduction in Pins:Tom lateral enrichment in the Dlg1P20/2 mutants - consistent with the spindle misorientation phenotype in these mutants.

The reviewer raises a reasonable concern about Figure 4D. We noted the difficulty of imaging Pins:Tom, which is exceedingly faint, in our original manuscript. For technical reasons, only one copy of the transgene was imaged in the experiment presented in 4G (two copies were used in Figure 2B), and the lack of signal presented an even greater challenge. In the manuscript we went with the clearest image. To address the reviewer’s concern, we have added signal intensity plots to this figure showing that Pins:Tom and Pins-myr are both laterally enriched at mitosis in Dlg[1P20]/Dlg[2] mutants. These data have been added as a new panel (E) in Figure 4. We were also able to replace the pictures in 4D with new ones generated after review.

1. In Fig 4E, the phosphomutant PinsS436A-GFP looks more strongly apical and less strongly lateral than the wildtype Pins-GFP, consistent with the spindle misorientation phenotype in S436A rescued pins mutants.

The reviewer has an eagle eye! We did not detect a difference in localization across the three transgenes, though we were certainly looking for it (that’s why we generated these flies in the first place). Again, the strength of signal was a major challenge in these experiments, and we therefore went with the cleanest image. In response to the reviewer’s concern, we note that the S436A and S436D examples shown have equivalent apical signal, but only the S436A fails to rescue spindle orientation.

Together, Reviewer Comments 1 and 2 suggest a model in which Dlg is required for lateral enrichment of Pins at mitosis. As described in the manuscript, this is the very model proposed in our own previous study (Bergstralh, Lovegrove, and St Johnston; 2013), and reiterated in a subsequent review article (Bergstralh, Dawney, and St Johnston; 2017). We point these publications out because the senior author of the current manuscript is not especially enthusiastic about showing himself to be wrong (twice!) in the literature. He therefore insisted on seeing multiple lines of evidence before making the counterargument:

• The reviewer’s model (the 2013 model) is firstly challenged by work shown in Figure 3. We find that membrane-anchored proteins (even just myristoylated RFP!) demonstrate lateral enrichment at mitosis, regardless of whether or not they interact with the Dlg-GUK domain.
• Even stronger evidence is shown in Figure 4F. Pins-myr-GFP is very plainly enriched at the lateral cortex in Dlg[IP20]/Dlg[2] mutant cells (now demonstrated with signal intensity plots in FIGURE 4E). However, the spindle doesn’t orient correctly (quantified in 4C). Since Dlg is impacting spindle orientation independently of Pins localization, these data support our “claim in the final sentence of the abstract ‘Local enrichment of Pins is not sufficient to determine spindle orientation; an activation step is also necessary’.”

  In the InscA examples, Pins:Tom looks apical. In the InscB examples, Pins:Tom looks more laterally localised, consistent with the spindle orientations in these experiments.

These figures (6A-D) don’t only show/quantify Pins:Tom localization. They also show localization of GFP-Mud. Whereas Pins:Tom is certainly apically enriched in the InscA examples, the interesting finding is that GFP-Mud is not. In strong contrast, it instead shows a weak apical localization and a strong lateral enrichment. As described in the manuscript, this pattern of Mud localization predicts normal spindle orientation, which is not observed in these cells.

Thus, these data appear to support the existing model that Pins enrichment at the plasma membrane is a key factor directing mitotic spindle orientation in these cells. The author's claim in the final sentence of the abstract "Local enrichment of Pins is not sufficient to determine spindle orientation; an activation step is also necessary" is not supported by the data.

We are pleased that the reviewer shared this quote; our claim is that Pins localization is not sufficient, not that it is unnecessary (see above). We absolutely do not dispute that “Pins enrichment at the plasma membrane is a key factor directing mitotic spindle orientation.”

The open question posed by the data is why GFP-Mud is excluded apically & basally during mitosis, while Pins:Tom is not. The simple alternative model is that Mud only localises to the plasma membrane where Pins is most strongly concentrated, such that Mud strongly amplifies any Pins asymmetry. Thus, even myr-Pins can still rescue a pins n mutant, because myr-Pins is still enriched laterally compared to apically (or basally).

Thus, I would strongly suggest re-titling the manuscript to: "Mud/NuMA amplifies minor asymmetries in Pins localisation to orient the mitotic spindle".

Well, that is a good-looking title, and we’re therefore sorry to decline the suggestion. However, as described above, Figure 4D shows that Pins enrichment does not always predict spindle orientation. More importantly, Figure 6A (cited by the reviewer in Comment 3) very plainly shows that Mud does not “only locali[ze] to the plasma membrane where Pins is most strongly concentrated.” In this picture – and across multiple InscA cells (Figure 6B) - Pins is strongly concentrated at the apical surface, whereas Mud is not.

Mud/NuMA presumably achieves this amplification by binding to the plasma membrane only where Pins is concentrated above a critical threshold level. This would mean a non-linear model based on cooperativity among Pins monomers that increases the binding avidity to Mud above the threshold concentration of Pins monomers.

This is essentially a minor revision of the standard model, which we expected would hold true in the FE. As described above, it is not supported by our data.

Reviewer #3 (Significance (Required)):

The manuscript is focused on the question of mitotic spindle orientation in epithelial cells, which is a fundamental unsolved problem in biology. The data reported are impressive and important, providing new insights into how the key spindle orientation factors Mud/NuMA and Pins/LGN localise during mitosis in epithelia. I recommend publication after major revisions.

We are delighted that the reviewer finds our data impressive and important, and our experiments insightful. We understand that the “major revisions” requested are meant to bring the paper in line with their model (our own earlier model). Since the data in our original manuscript contradict that model, the revisions are instead focused on clarifying and strengthening our message.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Neville et al re-examine the role and regulation of Pins/LGN in Drosophila follicular epithelial cells. They argue that polar or bipolar enrichment of Pins localisation at the plasma membrane is not crucial for spindle orientation, and therefore propose that Pins must be somehow activated to function. These interpretations are not supported by the data. However, the data strongly suggest an alternative interpretation which is of major biological significance.

Comments:

1. In the experiments on Dlg mutants (Fig 4D, S3) visualising Pins:Tom, the wild-type needs to be shown next to the Dlg mutant image, otherwise a comparison cannot be made. For example, Pins:Tom looks strongly enriched at the lateral membranes in the wild-type shown in Fig 2B&C, but much more weakly localised at the lateral membranes in Dlg1P20/2 mutants in Fig 4D. Thus, it looks like the Dlg GUK domain is required for full enrichment of Pins:Tom at lateral membranes, even if some low level of Pins can still bind to the plasma membrane in the absence of the Dlg GUK domain. Quantification would likely show a reduction in Pins:Tom lateral enrichment in the Dlg1P20/2 mutants - consistent with the spindle misorientation phenotype in these mutants.
2. In Fig 4E, the phosphomutant PinsS436A-GFP looks more strongly apical and less strongly lateral than the wildtype Pins-GFP, consistent with the spindle misorientation phenotype in S436A rescued pins mutants.
3. In the InscA examples, Pins:Tom looks apical. In the InscB examples, Pins:Tom looks more laterally localised, consistent with the spindle orientations in these experiments.

Thus, these data appear to support the existing model that Pins enrichment at the plasma membrane is a key factor directing mitotic spindle orientation in these cells. The author's claim in the final sentence of the abstract "Local enrichment of Pins is not sufficient to determine spindle orientation; an activation step is also necessary" is not supported by the data.

The open question posed by the data is why GFP-Mud is excluded apically & basally during mitosis, while Pins:Tom is not. The simple alternative model is that Mud only localises to the plasma membrane where Pins is most strongly concentrated, such that Mud strongly amplifies any Pins asymmetry. Thus, even myr-Pins can still rescue a pins mutant, because myr-Pins is still enriched laterally compared to apically (or basally).

Thus, I would strongly suggest re-titling the manuscript to: "Mud/NuMA amplifies minor asymmetries in Pins localisation to orient the mitotic spindle".

Mud/NuMA presumably achieves this amplification by binding to the plasma membrane only where Pins is concentrated above a critical threshold level. This would mean a non-linear model based on cooperativity among Pins monomers that increases the binding avidity to Mud above the threshold concentration of Pins monomers.

Significance

The manuscript is focused on the question of mitotic spindle orientation in epithelial cells, which is a fundamental unsolved problem in biology. The data reported are impressive and important, providing new insights into how the key spindle orientation factors Mud/NuMA and Pins/LGN localise during mitosis in epithelia. I recommend publication after major revisions.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

The manuscript by Neville et al. addressed the mechanism how conserved spindle regulators (Pins/Mud/Gai/Dynein) control spindle orientation in the proliferating epithelia by revising "the canonical model", using the Drosophila follicular epithelium (FE). The authors examined the epistatic relationship among Pins, Mud and Dlg in FE and found that Pins controls the cortical localization of Mud by utilizing mutant analyses, and suggested their localization does not fully overlap using the newly generated knock-in allele. They also showed that Pins relocalization during mitosis depends on cortical remodeling, or passive model, where Pins localization changes with other membrane-anchored proteins. Their data further suggest that Pins cortical localization is not influenced by Dlg, but Pins-interacting domain of Dlg does affect spindle orientation. Based on these results, the authors propose that Dlg controls spindle orientation not by redistributing Pins, but by promoting (or "activating" from their definition) Pins-dependent spindle orientation. Interestingly, ectopic expression of Inscuteable (Insc) suggested that Insc localization, either apical or lateral, correlates with spindle orientation, and their localization is a dominant indicator of spindle orientation, compared to the localization of Pins and Mud, implicating potentially distinct roles of activation and localization of the spindle complex. Overall their genetic experiments are well-designed and provide stimulation for future research. However, their evidence is suggestive, but not conclusive for their proposal. I have several concerns about their conclusion and would like to request more detailed information as well as to propose additional experiments.

Major concerns:

1. This report lacks technical and experimental details. As the typical fly paper, the authors need to show the exact genotypes of flies they used for experiments. This needs to be addressed for Figures 1-6, and Supplemental Figures. Especially, which Gal4 drivers were used for UAS-Pins wt or mutant constructs in Figure 4 with pins mutant background, Khc73, GUKH mutant backgrounds. Which exact flies were used for mutant clone experiments for Supplemental Figure 3? (A for typical mosaic, and B for MARCM). Without these details, it is impossible to evaluate results and reproduce by others.
2. Related to the comment 1, how did the author perform "clonal expression of Ubi-Pin-YFP" in page 5? As far as I understand, Ubi-Pin-YFP is expressed ubiquitously by the ubiquitin promoter.
3. In page 6, if Pins relocalization is passive and is associated with membrane-anchored protein remodeling during mitosis, its relocalization can be suppressed by disrupting the process of mitotic remodeling (mitotic rounding). The authors should test this by either genetic disruption or pharmacological treatment for the actomyosin should cause defects in Pins relocalization, which bolster their conclusion.
4. The critical message in this manuscript is that the core spindle complex mediated by Pins-Mud controls spindle orientation by "activation", but not localization. The findings that Pins and Mud localization is not influenced by Insc and that ecotpic Insc expression and genetic Mud depletion (Figure 6) might support their proposal, but these results just suggest their localization does not matter. I wonder how the authors could conclude and define "activation". What does this activation mean in the context of spindle orientation? Can the authors test activation by enzymatic activity or assess dynamics of spindle alignment?
5. In page 7-8, although Pins-S436D rescue spindle orientation, but Pins-S436A does not in pins null clones background, Pins localization is not influenced by Dlg. This questions how exactly Pins and Dlg can interact, and how Dlg affect Pins function. Related to this observation, in the embryonic Pins:Tom localization in dlg mutant does not provide strong evidence to support their conclusion given the experimental context is different from previous study (Chanet et al., 2017).
6. In page 11, the authors state "... that activation of pulling in the FE requires Dlg". I was not convinced by anything related to "pulling". There is no evidence to support "pulling" or such dynamic in this paper, just showing Mud localization, correct?
7. Ectopic expression of Insc (Figure 6) provided a new idea and hypothesis, but the conclusion is more complicated given that Insc is not expressed in normal FE. For example, the statement that "Inscuteable and Dlg mediate distinct and competitive mechanism for activation of the spindle-orienting machinery in follicle cells" is probably right, but it does not show anything meaningful since Insc does not exist in normal FE. Is Dlg in a competitive situation during mitosis of FE? If so, which molecules are competitive against Dlg? The important issue is to provide a new interpretation of how spindle orientation is controlled epithelial cells. I strongly recommend to add models in this manuscript for clarity.

Minor comments:

1. Some sections were not written well in the manuscript. "It does not" in page 6. "These predictions are not met". I just couldn't understand what they stand for. Their writing has to be improved.
2. In page 9, Supplementary Figure 4 should be cited in the paragraph (A potential strategy for..), not Supplemental Figure 1A, and 1B.
3. In page 10, the authors examine aPKC localization in Insc expressing context of FE. Does aPKC localization correlate with Insc localization (Insc dictates aPKC?)? aPKC is not involved in spindle orientation from the author's report (Bergstralh et al., 2013), so it does not likely provide any supportive evidence.
4. In Dicussion page 12, "In addition, we find that while the LGN S408D (Drosophila S436D) variant is reported to act as a phosphomimetic, expression of this variant has no obvious effect on division orientation (Johnston et al., 2012)". Where is the evidence for this? I interpret that this phosphomimetic form can rescue like wt-Pins not like unphospholatable mutant S436A, so it means that have an effect on spindle orientation, correct?

Significance

The authors showed that Pins and Mud localization themselves are not sufficient for the control of spindle orientation with genetic analyses. While the authors tried to challenge the concept of "canonical model", there is no clear demonstration of "activation" of the spindle complex. I appreciate their genetic evidence and new results, and understand the message that Pins and Mud effects are beyond localization, but there is no alternative mechanism to support their model. At the current stage, their evidence provides more hypothesis, not conclusion. Based on my expertise in Developmental and Cell biology, I suggest that the work has an interest in audience who studies spindle machinery, but for general audience.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Neville et al addresses the link between the localization and the activity of the so-called "Pins complex" or "LGN complex", that has been shown to regulate mitotic spindle orientation in most animal cell types and tissues. In most cell types, the polarized localization of the complex in the mitotic cell (which can vary between apical and basolateral, depending on the context) localizes pulling forces to dictate the orientation The authors reexplore the notion that this polarized localization of the complex is sufficient to dictate spindle orientation, and propose that an additional step of "activation" of the complex is necessary to refine positioning of the spindle.

The experiments are performed in the follicular epithelium (FE), an epithelial sheet of cell that surrounds the drosophila developing oocyte and nurse cells in the ovarium. Like in many other epithelia, cell divisions in the FE are planar (the cell divides in the plane of the epithelium). The authors first confirm that planar divisions in this epithelium depends on the function of Pins and its partner mud, and that the interaction between the two partners is necessary, like in many other epithelial structures. Planar divisions are often associated with a lateral/basolateral "ring" of the Pins complex during mitosis. The authors show that in the FE, Pins is essentially apical in interphase and becomes enriched at the lateral cortex during mitosis, however a significant apical component remains, whereas mud is almost entirely absent from the apical cortex. Pins being "upstream" of mud in the complex, this is a first hint that the localization of Pins is not sufficient to dictate the localization of mud and of the pulling forces. The authors then replace wt Pins, whose cortical anchoring strongly relies on its interaction with Gai subunits, with a constitutively membrane anchored version (via a N-terminal myristylation). They show that the localization of myr-Pins mimics that of wt-Pins, with a lateral enrichment in mitosis, and a significant apical component. Since a Myr-RFP alone shows a similar distribution, they conclude that the restricted localization of Pins in mitosis is a consequence of general membrane characteristics in mitosis, rather than the result of a dedicated mechanism of Pins subcellular restriction. Remarkably, Myr-Pins also rescues Pins loss-of-function spindle orientation defects. They further show that the cortical localization of Pins does not require its interaction with Dlg (unlike what has been suggested in other epithelia). However, spindle orientation requires Dlg, and in particular it requires the direct Dlg/Pins interaction. The activity of Dlg in the FE appears to be independent from khc73 and Gukholder, two of its partners involved in its activity in microtubule capture and spindle orientation in other cell types. Based on all these observations, the authors propose that Dlg serves as an activator that controls Pins activity in a subregion of its localization domain (in this case, the lateral cortex of the mitotic FE cell). They propose to test this idea by relocalizing Pins at the apical cortex, using Inscuteable ectopic expression. With the tools that they use to drive Inscuteable expression, they obtain two populations of cells. One population has a stronger apical that basolateral Insc distribution, and the spindle is reoriented along the apical-basal axis; the other population has higher basolateral than apical levels of Insc distribution, and the spindle remains planar. The authors write that Pins localization is unchanged between the two subsets of cells (although I do not entirely agree with them on that point, see below), and that although Mud is modestly recruited to the apical cortex in the first population, it remains essentially basolateral in both. In this situation, the localization of Insc in the cell is therefore a better predictor of spindle orientation than that of Pins or Mud. Remarkably, removing Dlg in an Insc overexpression context leads to a dramatic shift towards apical-basal reorientation of the spindle, suggesting that loss of Dlg-dependent activation of the lateral Pins complex reveals an Insc-dependent apical activation of the complex.

Overall, I find the demonstration convincing and the conclusion appropriate. One of the limitations of the study is the use of different drivers and reporters for the localization of Pins, which makes it hard to compare different situations, but not to the point that it would jeopardize the main conclusions. I do not have major remarks on the paper, only a few minor observations and suggestion of simple experiments that would complete the study

Minor:

What happens to Pins and Mud in Dlg mutant cells that overexpress Insc and behave as InscA? Are they still essentially lateral, or are they more efficiently recruited to the apical cortex?

Regarding the competition between Pins and Insc for dictating the apical versus basolateral localization of Insc, the Insc-expression threshold model could be easily tested in Pins62/62 mutants, where it is expected that only InscA localization should be observed, even at 25{degree sign}C (unless Pins is required for the cortical recruitment of Insc, as it is the case in NBs - see Yu et al 2000 for example)

I do not agree with the authors on P.10 and Figure 6A-D, when they claim that the apical enrichment of Pins is equivalent in both InscA and InscB cells. The number of measured cells is very low, and the ratio of apical/lateral Pins differs between the two sets of cells. The number of cells should be increased and the ratios compared with a relevant statistic method.

A lot of the claims on Pins localization rely on overexpression (generally in a Pins null background) of tagged Pins expressed from different promoters or drivers, and fused to different fluorescent tags. Therefore, it is difficult to evaluate to which extent the localization reflects an endogenous expression level, and to compare the different situations. As the cortical localization of Pins relies on interaction with cortical partners (mostly GDP-bound Gai) which are themselves in limiting quantity in the cell, and in the case of Gai-GDP, regulated by Pins GDI activity, this poses a problem when comparing their distribution, because the expression level of Pins may contribute to its cortical/cytoplasmic ratio, but also to its lateral/apical distribution. Although I understand that the authors have been using tools that were already available for this study, I think it would be more convincing if all the Pins localization studies were performed with endogenously tagged Pins, even those with Myr localization sequences. In an age of CRISPR-Cas-dependent homologous recombination, I think the generation of such alleles should have been possible. Although this would probably not change the main claims of the paper, it would have made a more convincing case for the localization studies.

The authors should indicate in the figure legends or in the methods that the spindle orientation measurements for controls or Pins62/62 are reused between figures 1, 3, 4, 5, 6 , and between figure 3, 4 and 5, respectively

Significance

Altogether, this study makes a convincing case that the localization of the core members of the pulling force complex, Pins and Mud, is not entirely sufficient to localize active force generation, and that the complex must be activated locally, at least in the FE.

The notion of activation of the Pins/LGN complex has probably been in many people's mind for year: Pins/LGN works as a closed/open switch depending on the number of Gai subunits it interacts with, it must be phosphorylated, etc... suggesting that not all cortical Pins/LGN was active and involved in force generation. However the study presented here shows an interesting case where localization and activation are clearly disconnected. The authors show how Dlg plays this role in physiological conditions in the FE, and use ectopic expression of Insc to show that, at least in an artificial context, Insc can have the same "activating activity" (or at least an activating activity that is stronger than its apical recruitment capability and stronger than Dlg's activating activity). It is to my knowledge the first case of such a clear dissociation. In their discussion, the authors are careful not to generalize the observation to other tissues. Although I did not reexplore all that has been published on the Pins/LGN-NuMA/Mud complex over the last 20 years, my understanding is that despite interesting cases of distribution of the complex like that of Mud in the tricellular junction in the notum, the localization model can still explain most of the phenotypes that have been described without invoking an activation step. If it is the case, then the activation model is another variation (an interesting one!) on the regulation of the core machinery, which are plentiful as the authors indicate in their introduction, and is maybe specific to the FE; if not, then it would be interesting to push the discussion further by reexamining previous results in other systems, and pinpointing those phenotypes that could be better explained with an activation step.

Overall, I find this is an elegant piece of work, which should be of interest to many cell and developmental biologists beyond the community of spindle orientation aficionados.

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

Learn more at Review Commons

---

Reply to the reviewers

We thank for reviewers for their feedback and were pleased they think that the manuscript is “of great interest to the scientific community”. The reviewers agree that the manuscript addresses an important question and that the identification of ASNS as a potential vulnerability of late-stage colorectal cancer is significant. The reviewers agree that our findings would be substantially strengthened by validation in state-of-the-art organoid model systems. We agree with this and are currently liaising with collaborators (Owen Sansom, Beatson Institute and Laura Thomas, Swansea University) to replicate our findings in both mouse and human colorectal organoid models. We will determine the sensitivity of colorectal organoid models to ASNS inhibition across a range of tumorigenicities and mutational profiles representing different stages of the adenoma-carcinoma progression. We believe these experiments will substantially strengthen the manuscript and lend weight to our finding that late-stage adenocarcinoma cells are vulnerable to ASNS inhibition.

This is the predominant concern across reviewers, we are confident we can address this and all other, relatively minor, concerns as detailed below.

Please find below a point-by-point reply to the reviewer’s comments. Reviewer comments are in italicized text and our responses follow.

Reviewer #1

• All of the findings in this manuscript are limited to in vitro observations, we know that most of the in vitro findings can not be translated in vivo. The manuscript would significantly benefit from in vivo experiments using the cells described in Fig.1 A and confirming the in vitro findings.*

We agree that validation of our results in a more physiological context would significantly elevate our manuscript. In order to address this, we intend to use both human and mouse colorectal organoid models (please see detailed description of this in response to reviewer 2). We have decided to take this approach rather than conduct in vivoexperiments using the AA series (C1, SB, 10C and M) for two main reasons. Firstly, the C1 and SB cell lines do not form tumours in mice, consistent with them representing early colorectal adenoma cells. As such, we are not able to use the entire series in in vivo experiments. Secondly, we are keen to demonstrate replication of our findings in an alternative model. An organoid model would offer increased functional relevance, whilst allowing us to retain the ability to validate our observed metabolic dependencies across the adenoma to carcinoma sequence. We hope the reviewer agrees that these experiments would address their concerns.

• The authors should provide proliferation data for the cell lines they used in this manuscript (C1, SB, 10C and M). In Fig. 1 B they show clear differences in EACR, can the authors provide data on glucose uptake differences in these analyzed cell lines.*

We agree that proliferation and glucose uptake data would be a useful addition to the manuscript. We will provide doubling times for the cell lines used in this study and will measure glucose uptake by analysing extracellular glucose levels in the cell culture media from each of the cell lines.

• In Figure 2 C the authors should provide isotope tracing data for the upper glycolysis (e.g. glucose and glucose-6-P) and alanine. In Figure 2 D the authors should provide the isotope tracing data for glutamine and glutamate.*

We have data for glycolytic intermediates; glycerol-3-phosphate and dihydroxyacetone phosphate (DHAP) and alanine and will add them to the figures as requested.

• Do the authors see any sign of reductive carboxylation in their U-13C glutamine experiments?*

We observe only a low level of reductive carboxylation across the AA series cell lines (

• Can the authors speculate how the C1, SB, 10C and M cell lines would react when glucose would be replaced with galactose in the culture environment and forcing the cells to increase oxidative phosphorylation (OXPHOS).*

We would speculate that the cells would react similarly to our experiments in low glucose conditions displayed in Fig 3A-K. Given that M cells were shown to be the most flexible with regards to fuel source, we would expect them to be able to survive and proliferate more efficiently than the other cell lines in challenging culture conditions. Additionally, we would expect the C1s to survive well in galactose conditions, given that they rely less on glycolysis for ATP production and have significantly higher spare respiratory capacity compared to the more progressed cell lines.

• Can the authors comment whether C1, SB, 10C and M cell lines show differences in coping with oxidative stress?*

Again, we would speculate that the M cells would cope with exposure to oxidative stress best, given their metabolic flexibility. However, we would aim to test this by measuring the cellular response to hydrogen peroxide (which would induce oxidative stress) across all cell lines.

• In the ASNS knockdown experiments do the authors detect an increase in glucose uptake in ASNS deficient cells.*

This is an interesting question; we will address it by comparing extracellular glucose levels in C1 and M cells transfected with control and siRNA targeting ASNS.

• Can the authors provide gene expression data that would explain the metabolic switch between early and late-stage adenocarcinoma? Do the authors detect any differences in mTORC1 activation among the C1, SB, 10C and M cell lines? ASNS is an ATF4 target, can the authors provide any expression data on ATF4 in their cell lines.*

To address the first question, using our proteomics data, we have generated heatmaps showing protein abundance data from key metabolic pathways including glycolysis, the TCA cycle and the electron transport chain in the C1, SB and M cell lines. These data show an array of variation in protein expression of these pathways between the C1, SB and M cells, with no clear up or downregulation of these pathways as a whole, but rather more intricate regulation of clusters of proteins within the pathways. These data align well with the metabolomic data presented in Figure 2 and will allow us to investigate the mechanisms underlying the metabolic switch. These heat maps will be incorporated into the manuscript. Using the heatmaps we will identify and discuss key nodes we predict to explain the metabolic switch between early and late-stage adenocarcinoma. We will then determine whether manipulation of these nodes impact the metabolic phenotype of the cells experimentally. For example, the heat maps have highlighted that ENO3 or enolase 3 is strongly upregulated in the SB and M cells in comparison to the C1 cells and may be involved in driving the metabolic switch. Indeed, ENO3 has previously been found to promote colorectal cancer progression by enhancing glycolysis (Chen et al, Med Oncol, 2022), consistent with what we see here. To test this, we will knock down ENO3 across the cell line series and determine the impact on cellular phenotype and metabolism (using Seahorse extracellular flux analysis).

With regards to mTORC1 activation, we have further analysed our proteomics data from C1, SB and M cells and have found that the M cells show significantly higher serine 235/236 phosphorylation of ribosomal S6 protein, a common readout for mTORC1 activation, compared to C1 and SB cells. Further, we aim to carry out immunoblotting across the four cell lines to analyse phospho-S6 (relative to total S6), 4E-BP1 and phospho-ULK-1 (relative to total ULK-1) levels.

With regards to ATF4, using our proteomics data we have generated a heatmap of gene expression changes of ATF4 target genes in C1, SB and M cells that we will provide in supplementary material . These data suggest that there does not appear to be any clear pattern of enhanced or reduced ATF4 transcriptional activity across the cell lines, with different clusters of genes within this signature up or downregulated across the series. Moreover, Ingenuity Pathway Analysis (IPA) revealed that the ATF4 pathway showed an activation z-score of -0.41 (p=0.0134) in SB versus C1 cells, and 0.35 (p=0.00051) in M versus C1 cells (where a threshold of +/- 2 indicates activation/suppression of the pathway, respectively), confirming there is no clear regulation of this pathway between the cell lines. In addition, we will carry out immunoblotting for ATF4 expression levels across the cell line series.

Reviewer #2

*Major comments: *

*Early CRC *

*Molecular understanding of CRC is obviously of great interest and importance for the clinics. However, tumors of early stages are almost exclusively resected and not treated with systemic agents. Hence, the argument by the authors that the metabolic understanding of early CRC is of clinical relevance is somewhat misleading. Overall, it would have been much more clinically relevant to investigate the multiple steps of later stages during CRC progression. How about metabolic changes during metastasis. Deep mechanistic understanding of process during metastasis has striking clinical relevance. *

We agree with the reviewer that understanding metastatic progression is of clinical relevance and should indeed be investigated in more detail. Using our model, we do shed light on a vulnerability of late-stage adenocarcinoma cells (sensitivity to asparagine synthetase (ASNS) inhibition). Indeed, we show that ASNS expression is elevated in both colorectal tumour and metastatic tissue in comparison to normal suggesting that our study may have revealed a vulnerability with utility for treating late stage (and potentially metastatic) tumours. The reviewer raises an important issue with the way we frame the utility of the model in the manuscript text. We will rewrite this to emphasise its utility in identifying late-stage vulnerabilities and the clinical value of this approach. We maintain that the molecular understanding of colorectal cancer across all stages of its progression will provide a valuable contribution to the field but agree that we should be more specific with regards to the clinical utility of our findings.

*Model system *

The cell lines used in this study are not state-of-the-art to investigate the complex process during CRC progression. The original paper is from 1993 in which the cell lines were generated does not allow understanding of the characteristics of these cell lines. Recently, multiple models have been established, for example in organoids, to investigate the progression of CRC much more reliably. There are systems that use CRISPR/CAS9 edited human organoids that follow the genetic alterations of CRC progression with accompanied phenotypes. Further, extensive biobanks of organoids from patients are available (also commercially) which better represent the stages of CRC. Similarly, the question raised above of how representative this progression cell line set is needs to addressed. The mutagen-induced progression could generate various alterations that are not detected in patients, hence create an artificial system. Overall, biological replicates are missing.

We thank the reviewer for their critique and agree that our manuscript would be significantly strengthened if we were able to replicate our key findings in another model. We agree that the cell line series we have used here has limitations and we will make sure these are discussed by adding a ‘Limitations’ section to the ‘Discussion’. We maintain that the cell line series is a valuable tool in which to effectively identify metabolic vulnerabilities for further research. A key advantage of this system is that it is a human cell line series of the same lineage. In addition, we can easily conduct metabolomics and stable isotope tracer analysis allowing us to investigate cellular metabolic activity and manipulate any identified pathways easily. As such, the cell line series is an effective tool in which to identify potential vulnerabilities, but we agree that these vulnerabilities need to be validated in state-of-the-art organoid systems for the impact of the work to be clearer.

To address this, in collaboration with Owen Sansom (Beatson Institute) and Laura Thomas (Swansea University), we aim to validate our identified metabolic dependency in mouse and human colorectal organoids respectively. We will determine the sensitivity of colorectal organoid models across a range of tumorigenicities and mutational profiles representing different stages of the adenoma-carcinoma progression to asparagine synthetase (ASNS) inhibition. We believe these experiments will substantially strengthen the manuscript and lend weight to our finding that late-stage adenocarcinoma cells are vulnerable to ASNS inhibition.

*Gene Expression analysis *

In Figure 5 C and D is the expression of ASNS to stages and overall survival from online available datasets correlated. Its unclear what the difference between tumor and metastatic in C means. The labelling in D is too small. Is the difference between the two groups significant? Are these patients only at a specific stage? It seems not that ASNS is a strong prognosticator; further stratification is needed to clarify the role of ASNS in CRC.

The data displayed in Fig 5C and 5D are from separate datasets so are not correlated. In Fig 5C ‘Tumour’ refers to gene expression from the primary tumour site (in this case the colorectum), whereas ‘Metastatic’ refers to gene expression from a metastatic tumour (from which the primary tumour was of colorectal origin). We will make this clearer in the text and figure legend. We will also make the labelling on the survival plot in D clearer, indicating that the difference between the two groups is significant and displaying the p value clearly.

The data included in the survival plots in 5D encompass all tumour stages. We have further analysed these data, adjusting for tumour stage. We found that high ASNS expression in later stage tumours (stage 3 and 4) is associated with poorer overall survival, whereas there is no significant difference in overall survival in earlier stage tumours (stage 1 and 2) in relation to ASNS expression. We plan to add this to the supplementary materials and discuss in the main text as it is consistent with our findings from the AA cell line series.

*Western Blot controls *

For the Western Blots in Figure 6 A and C the total S6 and ULK1 controls are missing what is needed to assess the effect on pS6 and pULK1 correctly.

We will add total S6 and ULK1 controls to these figures.

In the same panels, the KO efficacy is not very high in A (-ASN). However, this is crucial to make the conclusion that this cell line (C1) is not dependent on ASNS.

The average knockdown efficiency in the C1 cells is 72% across n=3 experiments. Therefore, levels of ASNS are significantly reduced. However, to further validate this finding, we will use L-Albizziine, a competitive inhibitor of ASNS, at the same concentration in both C1 and M cells to eliminate any issues surrounding variation in knockdown efficiency and to replicate the results obtained using ASNS siRNA. These data will be included in supplementary material.

*Minor comments: *

*Statistical analysis of proliferation assays *

The statistical significance for proliferation assays are missing.

The statistical significance at the final timepoints of the proliferation assays are displayed on bar graphs in Supplementary Figure 5 (Figure S5B and C). We will add these to the proliferation curves in the main figure.

Reviewer #3

*A major concern is the model used in this study: *

Sodium butyrate and the carcinogen N-methyl-N-nitro-nitrosoguanidine (MNNG) were used for the transformation. I believe this model was developed by one of the co-authors of the study, A.C. Williams in the 1990s. The relevance of the model for in vivo colon carcinogenesis is not entirely clear to me and I miss information why in particular sodium butyrate and MNNG were used. I am not an expert on colon carcinogenesis but I did not have the impression that this model has been widely adopted and I miss detailed information on the model itself as well as a critical discussion of its limitations.

We thank the reviewer for raising these concerns and will include a ‘Limitations’ section in the manuscript ‘Discussion’ to elaborate on both the utility and the limitations of this model system. As described in response to concerns raised by reviewer #1 and reviewer #2, we plan to validate our findings in organoid models of colorectal tumourigenesis to strengthen the discoveries made using the AA cell line series.

With regards to the use of sodium butyrate and MNNG for transformation of the C1 cells, justification was provided in the original paper describing generation of the cell line model series (Williams et al, Cancer Research. 1990). Sodium butyrate is naturally occurring in the gut and was used for the transformation of the C1 cells as it had been proposed to play a role in promoting colorectal tumorigenesis through upregulating carcinoembryonic antigen (CEA) expression and enhancing proliferation in adenoma cells able to resist growth arrest following treatment (Berry et al, Carcinogenesis. 1988). At the time of generating the cell line series, few reagents were known to induce transformation in human epithelial cells. However, MNNG was one of those and had been previously used to transform keratinocytes (Rhim et al, Science. 1986). Crucially, tumours formed in mice from xenografted 10C cells were found to be heterogeneous, displaying areas of differentiation with glandular organisation, the presence of functional goblet cells enabling mucin production, as well as areas of poorly or undifferentiated cells. Furthermore, cytogenetic analyses revealed that genetic changes in the cell line progression model such as chromosome 18q loss and KRAS activation replicate those seen in CRC patients (Williams et al, Oncogene. 1993). Together, these characteristics recapitulate human tumours in vivo, validating the use of sodium butyrate and MNNG in generating an in vitro CRC cell line model that represents human colorectal tumorigenesis.

Figure 6: total levels of ribosomal S6 protein and ULK1 should be detected, quantified and used for normalization.

We agree with the reviewer, we will add total S6 and ULK1 controls to these figures.

Can you measure ASN upon inhibition of autophagy? Does it go down further?

This is an interesting question, and we will address this experimentally by measuring ASN levels following treatment with chloroquine in the C1 and M cell lines. We will do this using stable isotope labelling and mass spectrometry and include the results in supplementary material.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Legge and colleagues metabolically characterize an in vitro colon carcinogenesis model based on the development of the adenoma cell line PC/AA to a tumorigenic phenotype. They find differences in the use of glucose and glutamine in the different cell lines and a dependency of what they call late-stage adenocarcinoma cells on asparagine (ASN) synthesis, which is not present in their early stage cell line. The study is well-written and interesting.

A major concern is the model used in this study:

Sodium butyrate and the carcinogen N-methyl-N-nitro-nitrosoguanidine (MNNG) were used for the transformation. I believe this model was developed by one of the co-authors of the study, A.C. Williams in the 1990s. The relevance of the model for in vivo colon carcinogenesis is not entirely clear to me and I miss information why in particular sodium butyrate and MNNG were used. I am not an expert on colon carcinogenesis but I did not have the impression that this model has been widely adopted and I miss detailed information on the model itself as well as a critical discussion of its limitations.

The seahorse and the stable isotope labelling experiments appear fine to me.

Figure 6: total levels of ribosomal S6 protein and ULK1 should be detected, quantified and used for normalization.

Can you measure ASN upon inhibition of autophagy? Does it go down further?

Significance

The identification of ASN synthesis as a potential vulnerability of advanced colon cancer is potentially significant but needs to be confirmed in other models.

Differences in ASN sensing between early and late stage colon cancer cells as well as the role of autophagy are potentially interesting and merit further investigation.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The manuscript by Legge and colleagues describes the metabolic rewiring during colorectal cancer (CRC) progression. While earlier stages depend on glycolysis but maintain oxidative metabolism later stages are more plastic and can be maintained in nutrient-poor environments. The study addresses a very important open question towards the progression of CRC, however the stated clinical relevance is relatively little. The strongest limitation is the used model system to describe the step-wise progression of CRC what makes it difficult to understand how important the findings of this study are for the very heterogonous disease CRC.

Major comments:

Early CRC

Molecular understanding of CRC is obviously of great interest and importance for the clinics. However, tumors of early stages are almost exclusively resected and not treated with systemic agents. Hence, the argument by the authors that the metabolic understanding of early CRC is of clinical relevance is somewhat misleading. Overall, it would have been much more clinically relevant to investigate the multiple steps of later stages during CRC progression. How about metabolic changes during metastasis. Deep mechanistic understanding of process during metastasis has striking clinical relevance.

Model system

The cell lines used in this study are not state-of-the-art to investigate the complex process during CRC progression. The original paper is from 1993 in which the cell lines were generated does not allow understanding of the characteristics of these cell lines. Recently, multiple models have been established, for example in organoids, to investigate the progression of CRC much more reliably. There are systems that use CRISPR/CAS9 edited human organoids that follow the genetic alterations of CRC progression with accompanied phenotypes. Further, extensive biobanks of organoids from patients are available (also commercially) which better represent the stages of CRC. Similarly, the question raised above of how representative this progression cell line set is needs to addressed. The mutagen-induced progression could generate various alterations that are not detected in patients, hence create an artificial system. Overall, biological replicates are missing.

Gene Expression analysis

In Figure 5 C and D is the expression of ASNS to stages and overall survival from online available datasets correlated. Its unclear what the difference between tumor and metastatic in C means. The labelling in D is too small. Is the difference between the two groups significant? Are these patients only at a specific stage? It seems not that ASNS is a strong prognosticator; further stratification is needed to clarify the role of ASNS in CRC.

Western Blot controls

For the Western Blots in Figure 6 A and C the total S6 and ULK1 controls are missing what is needed to assess the effect on pS6 and pULK1 correctly. In the same panels, the KO efficacy is not very high in A (-ASN). However, this is crucial to make the conclusion that this cell line (C1) is not dependent on ASNS

Minor comments:

Statistical analysis of proliferation assays

The statistical significance for proliferation assays are missing.

Significance

The described topic is very relevant and of great interest to the field. However, I see the major limitation in the applied system to decode metabolic dependencies. If the key points were validated for example in state-of-the-art organoid systems, the impact of the work would be much clearer. I have to note that my expertise in the field of metabolism is relative little. My expertise lays in modelling CRC and its plasticity/heterogeneity with a strong asset to the translational space.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Legge et al. provide novel metabolic liabilities in colorectal cancer progression. They identified that during colorectal tumor progression cancer cells shift from glycolytic to oxidative metabolism that enables cancer cells to advance. Using stable isotope tracing experiments with U-13C glucose and U-13C glutamine they showed that the contribution of these carbon sources changes from colorectal adenoma to carcinoma progression. In their in vitro model they identified that late-stage colorectal adenocarcinoma cells (M) display metabolic plasticity which leads to resistance to changes in nutrient supply compared to early stage adenoma cells. Proteomic analysis revealed that amino acid metabolism is highly dysregulated during carcinoma progression. The authors identified ASNS and regulation of asparagine availability as novel metabolic target for late-stage adenocarcinoma.

Overall, this is a very interesting and important manuscript describing a previously unknown metabolic liability in late-stage adenocarcinoma. The authors also nicely demonstrate how cancer metabolism is adapting and changing during cancer progression. These findings are clearly of high relevance in the context of better understanding of how metastatic cancer cells gain abilities to metastasize. Therefore, the findings presented here would advance the field. However, there are some open questions in the study design and the current set of data as listed below.

1. All of the findings in this manuscript are limited to in vitro observations, we know that most of the in vitro findings can not be translated in vivo. The manuscript would significantly benefit from in vivo experiments using the cells described in Fig.1 A and confirming the in vitro findings.
2. The authors should provide proliferation data for the cell lines they used in this manuscript (C1, SB, 10C and M). In Fig. 1 B they show clear differences in EACR, can the authors provide data on glucose uptake differences in these analyzed cell lines.
3. In Figure 2 C the authors should provide isotope tracing data for the upper glycolysis (e.g. glucose and glucose-6-P) and alanine. In Figure 2 D the authors should provide the isotope tracing data for glutamine and glutamate.
4. Do the authors see any sign of reductive carboxylation in their U-13C glutamine experiments?
5. Can the authors speculate how the C1, SB, 10C and M cell lines would react when glucose would be replaced with galactose in the culture environment and forcing the cells to increase oxidative phosphorylation (OXPHOS).
6. Can the authors comment whether C1, SB, 10C and M cell lines show differences in coping with oxidative stress?
7. In the ASNS knockdown experiments do the authors detect an increase in glucose uptake in ASNS deficient cells.
8. Can the authors provide gene expression data that would explain the metabolic switch between early and late-stage adenocarcinoma? Do the authors detect any differences in mTORC1 activation among the C1, SB, 10C and M cell lines? ASNS is an ATF4 target, can the authors provide any expression data on ATF4 in their cell lines.

Referees cross-commenting

The comments from the other reviewers are fair and would significantly improve the quality of the manuscript. Overall all three reviewers think that this manuscript is of great interest for the scientific community.

Significance

Overall, this is a very interesting and important manuscript describing a previously unknown metabolic liability in late-stage adenocarcinoma. The authors also nicely demonstrate how cancer metabolism is adapting and changing during cancer progression. These findings are clearly of high relevance in the context of better understanding of how metastatic cancer cells gain abilities to metastasize. Therefore, the findings presented here would advance the field.

The findings are novel and have not been described in this detail before.

The findings would be of importance for a broad audience and ideally will help to identify novel pharmacological inhibitors that could be used in patients.

My expertise is in cancer metastasis, cancer metabolism and translational/clinical medicine.

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

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary

This manuscript attempts to link different aspects of HDAC1 function to Plasmodium falciparum biology. HDAC1 is essential so is likely to have important functions in parasite development.

The emphasis is upon the potential gene regulation aspects of HDAC1 function, but it is well known that acetylation of other proteins is regulated by HDAC1 orthologues. While they examine the genome occupancy of HDAC1, it's not clear whether the phenotypic effects described can be ascribed to effects upon histone modifications. For RNA-seq analysis and ChIPseq, they generally use one time point so they have not controlled for potential differences in cell cycle to explain differences in gene expression or genome occupancy. These weaknesses in the experimental design make it difficult to evaluate the significance of their data with artemesinin and drug resistant lines.

The authors suggest that CKII is important for regulating the function of HDAC1. This is biologically plausible, but the link could be more convincing. In addition, the evidence that the potential gene regulation effects are critical for the phenotype observed could be stronger.

Major comments:

Figure 1: They perform phosphorylation studies with recombinant CKII and HDAC1, but they do not demonstrate whether the phosphorylated residues correspond to the predicted residues S391, S397 and S440 or if mutation of the predicted residues affects activity.

The inhibitor data are consistent with the predicted effects, but kinase inhibitors do not always have the same target in vivo or in cells as they do in protein assays. Concentrations of inhibitors used should be provided in the materials and methods.

They also claim that CK2 and HDAC1 interact in parasites (p5). They do not provide data to support this statement, nor do they provide any data about other proteins that might be interacting with HDAC1. If they were able to purify enough HDAC1 for mass spec identification, they should provide further documentation about interacting proteins and potential post-translational modifications.

In addition, they should provide more detailed characterization with Western/IFA of when HDAC1 is expressed and whether CKII is always expressed at the same time.

Overall the importance and significance of CKII in regulation of HDAC1 activity is not clear and would be much strengthened if experiments performed with recombinant protein could be replicated in IP parasite lysates with appropriate controls and a time series.

Figure 2: Using an HDAC1 GFP line they perform ChIP-seq. The ChIP-seq experiments seem to be well performed with high correlation between replicates but were performed at a single time point in the life cycle of erythrocytic stages. It's not clear if the distribution or abundance of HDAC1 changes during the cell cycle, though they suggest it does, and given changes noted in genome occupancy, one cannot determine if the differences seen could be completely explained by parasites being in different stages of the cell cycle with different levels of HDAC1. They show enrichment of different pathways, but do not comment on whether these are just pathways that are enriched in trophozoites.

Figure 3 They characterize the growth rate of parasites treated with sublethal concentrations of HDAC1 inhibitor and see effects. The images presented in panel A are not good quality and parasite morphology is difficult to evaluate. They perform RNA-seq at a single time point and the choice of time point and drug concentration used is not justified. Changes are reported but again with a single time point, it's difficult to interpret the significance of the changes-are these dying parasites or parasites slowly progressing through the life cycle? To really understand the effects of these drugs a better characterization of dose response and time point series is needed.

Figure 4 Upon overexpression of PfHDAC1-GFPglmS there appear to be more parasites. It is unclear if this due to more merozoites per schizont, better invasion with more rings. Again, better characterization of time points would be helpful to understand how overexpression of HDAC1 affects proliferation.

Figure 5. They state that there is less HDAC1 in art resistant lines, but given that they have not provided any information about cell cycle expression of HDAC1 and growth of these lines in comparison to wild-type, it is unclear if there are differences in biology or if the cells differ where they are in the cell cycle.

This is particularly important because of the known differences of artemisinin effects depending upon cell cycle stage.

Figure 6 Genome occupancy data are difficult to interpret given possible differences in cell cycle.

Minor comments:

The general quality of images and gels should be improved.

More information should be provided about the validation and specificity of the in house HDAC1 antibodies.

Concentrations of inhibitors used should be provided.

Referees cross-commenting

There is consensus amongst all reviewers that the experiments as presented cannot be readily interpreted and are lacking adequate controls. The amount of experimental work and further analysis is considerable.

Significance

Understanding gene expression and the role of HDAC1 is potentially significant, particularly if these can be linked to important biological processes such as artemisinin resistance. Potentially the audience would be broad. The link between these processes is not well supported by the data as currently presented.

Expertise: epigenetics, parasite gene expression.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this study, Kanyal et al present a functional analysis of the Plasmodium falciparum Histone deacetylase 1 (PfHDAC1). PfHDAC1 is the only predicted class I HDAC in P. falciparum and has been shown to be the target of several established and novel compounds with anti-malarial activity across parasite stages. In this work, the authors showed that recombinant HDAC1 is phosphorylated in the presence of CKII in vitro and can de-acetylate P. falciparum histones (although no loading control was presented for this latter assay). ChIPseq of GFP tagged HDAC1 identifies target sites relating to diverse cellular processes, and sublethal treatment with the proposed HDAC1 inhibitor rhomidepsin has an impact on cell cycle progression, suggesting that HDAC1 may act in cell cycle control. However, overexpression of Histone deacetylase seems to enhance parasite multiplication by increasing invasion gene expression, which seems counterintuitive as overexpression is expected to cause decreased histone acetylation and thereby gene repression, and hence this pattern may be due to indirect effects, which the authors acknowledge but the relevance of which they do not discuss further. Interestingly, HDAC1 expression is reduced in Art resistant parasites and inhibition of HDAC1 (at what concentrations?) increases ART resitance both in wt and in K13 mutant parasites, suggesting regulation of HDAC1 may be involved in adapting to artemisinin treatment. ChIPseq of artemisinin resistant versus sensitive parasites suggests that HDAC1 is relocated to various different loci, although replicates for this experiment seem to be missing and hence the validity of these results would need further support, particularly because ChIP results conducted with anti-HDAC1 antibodies and anti-GFP antibodies seem to diverge considerably. Lastly, the authors propose that artemisinin treatment results in mistrageting of HDAC1 but find no correlation with gene expression. Generally, the study raises some interesting aspects related to a function of HDAC1 in artemisinin resistance but would benefit from more rigorous analyses and comparisons of the NGS data presented in correlation to each other (e.g. ChIP anti-GFP vs ChIP anti-HDAC1) as well as to published data sets (e.g. Huang et al). Sometimes it is difficult to assess which data set the authors refer to and whether transcriptional data were derived from RNA harvested in parallel to ChIPseq (matched) or whether they were performed independently or by others. Also, many assays seem to lack replicates and controls as outlined below.

Major comments:

• Transcriptomic data of parasites treated with Romidepsin are presented as a proxy for HDAC1 function and indicate deregulation of invasion pathways, however what is the evidence that romidepsin targets (exlusively) HDAC1? This could for example be addressed by comparing the Romidepsin IC50 in HDAC1 overexpressing parasites versus parasites with WT levels of HDAC1.
• How do the rhomidepsin treatment data correlate with JX21108 RNAseq results, a validated HDAC1 targeting compound? The authors need to thoroughly cross evaluate their data with the RNAseq data set from HDAC1 knockdown parasites and JX21108 treated parasites presented in Huang et al, 2020.
• What is the overlap between genes deregulated after rhomidepsin treatment and ChIPseq targets?
• What are the target genes that show strong enrichment in the gene body in Fig. 2E? How are the data sorted? It is expected that HDAC1 may affect gene expression differently when it is present in the gene body to when it is present in the promoter region, therefore it would be useful to stratify the target genes by peak position relative to genetic elements.
• How do the anti-GFP ChIPseq data in K13WT strains (496 target genes, Fig 5) correlate with the anti-HDAC1 ChIPseq data (1409 target genes Fig 2 and 6)? There seems to be limited overlap in the target sites in number and quality, but it is difficult to assess just from looking at gene numbers and GO analyses. The data sets need to be more thoroughly cross-validated. What proportion of peaks overlap and where?
• How many replicates were performed for each experiment? Many of the Figures showing recombinant assays and ChIPseq assays seem to represent only a single biological replicate (e.g. Fig 1E histone deacetylation assay, Fig. 5A, Fig 6C: ChIP under Artemisinin treatment).
• Several critical controls are missing, for example Figure 1E/ Suppl. Figure 2D loading control (anti-H3). How was densitometry normalized?
• What was the parasite age in RNAseq of HDAC1-GFP-GlmS parasites? Were the two data sets from different parasite lines adjusted for parasite age? How many replicates for RNAseq?
• The data in Figure 6A Lanes 1-5 are evidently the same as shown in Fig 1D. The presentation of Art treated data as a single lane 6 without direct reference is not convincing as this does not allow a direct comparison of loading and between conditions.
• How does histone acetylation change in response to Art treatment?

Minor comments:

• Page 4 top paragraph: check whether Ref 16 and 17 are correctly cited here.
• In all Figures: specify drug concentrations and number of replicates.
• What concentrations of etinostat, DHA, Romidepsin were used to treat parasites? Please provide exact concentrations of treatments ( not just +, ++, for example for TBB Fig 1, Artemisinin 6A), what was the "continuous sublethal dosage of romidepsin" exactly, what is the IC50 of romidepsin?
• What is referred to as control in Fig. 1E?
• Fig 4F please specify in the figure legend what the control was
• Fig 4G the labelling of the circus plot is unreadably small.
• Figure 5G and H: what RNAseq data set is shown here? Are these matched RNAseq data from these ChIP assays or other?
• The calculation of how the growth curves were corrected as "GFP-glmS corrected growth curves" is unclear, please provide exact formula. Generally, the multiplication rates even in untreated conditions appear rather low in all experimetns (for example Fig 4D, only less than 2-fold growth after 1 cycle, 4 fold growth after 2 cycles.... Do the parasites under the normal growth conditions really only duplicate in each cycle? This seems a very low multiplication rate even for static in vitro culture of P. falciparum.
• What is the relevance of the 2xFKBP in the tagging construct?

Referees cross-commenting

All reviewers agree that the manuscript in its current form would benefit from the addition of controls and replicates as well as additional time points for RNAseq and ChIP experiements.

Significance

Generally, the study raises some interesting aspects related to a function of HDAC1 in artemisinin resistance but would benefit from more rigorous analyses and comparisons of the NGS data presented in correlation to each other (e.g. ChIP anti-GFP vs ChIP anti-HDAC1) as well as to published data sets (e.g. Huang et al). Sometimes it is difficult to assess which data set the authors refer to and whether transcriptional data were derived from RNA harvested in parallel to ChIPseq (matched) or whether they were performed independently or by others. Also, many assays seem to lack replicates and controls as outlined below.

My personal field of research is chromatin biology and antigenic variation in malaria parasites.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This study characterises a Plasmodium class I Histone deacetylase (PfHDAC1). The manuscript reports a wide range of experiments - some of them complex and involved, but not all of these experiments appear to be well controlled, and some are insufficiently described to know if they have been appropriately designed and interpreted. A link to HDAC1 regulation and artemisinin resistance is advanced, but the evidence here is very indirect and inconclusive.

The study shows that HDAC1 interacts with PfCKII- a homologue of the mammalian casein kinase known to interact with mammalian HDAC1. They also demonstrate that, at least in vitro, HDAC1 can serve as a substrate for phosphorylation by PfCKII, and that this phosphorylation impacts HDAC1's deacetylation of histones. Such assays where a kinase is provided with a single, abundant substrate in vitro, are not always rigourous tests for kinase specificity, but do in this case at least indicate that HDAC1 associated with its activity.

Major issues:

1. The authors conduct CHiP seq experiments on a GFP tagged HDAC. It is unclear from the methods and results section what control is used in these experiments. The ENCODE consortium has established minimum standards (Landt et al 2012) for conducting and reporting CHiP seq experiments, and states that the "recommended control for epitope-tagged measurements is an immunoprecipitation using the same antibody against the epitope tag in otherwise identical cells that do not express the tagged factor.". These experiments appear to lack that control and the enrichments reported should be approached with caution in the absence of such a control.
2. The genes with apparently altered ChiP seq were subjected to gene ontology enrichment analysis, and the authors report potential enrichments - which appear to impact a range of unconnected biological pathways throughout the parasite and throughout the lifecycle, despite the CHIP seq being conducted only at a single time stage. No mention is made of correction for multiple hypothesis testing, known to present a considerable problem for such analyses, and no correction is described for background GO distributions in the P. falciparum genome, so again it's unknown if or how that was performed. The reported enriched categories must be also treated with considerable caution given the absence of description of these crucial steps. The authors report from this section that HDAC1 is associated with stress responses, but really, by their criteria, HDAC1 is associated with 1/3 of the whole genome, so it's a bit selective to regard it as a stress regulator
3. The authors preform a well-designed series of transfection experiments with modulation of HDAC1 to show that an overexpression of HDAC1 leads to increased growth rate, and that this increase reduces when the overexpression of HDAC1 is inducibly repressed. However, I found the presentation of results from these experiments difficult to understand and there is considerable transformation of the data prior to plotting - they would be easier to understand if no background subtraction to normalise for GFP were conducted, and if all strains were plotted on the same axes. A potential confounding factor in this experiment is that many lines overexpressing GFP grow more slowly, and that this growth defect can be localisation dependent, so that over-expression of GFP alone may cause a different growth penalty than GFP on a nuclear protein. I am uncertain that the conclusion of 50% faster growth is a safe one based on these graphs - at some time intervals the over-expressor appears to grow just as slow or even slower (as a percentage of the previous timepoint) than the control, and these appear to have been based on technical replicates of a single biological experiment. The authors contend that the growth rate is due to changed expression of invasion genes (among many other substrate gene categories) giving rise to enhanced invasion - such a phenomenon is readily testable, and the authors should dissect this if they wish to substantiate the frankly surprising claim that overexpression of HDAC leads to increased growth rate.
4. The authors also report an apparent down regulation of HDAC abundance in artemisinin resistant parasites. This conflicts with previous global proteomic analyses of artemisinin resistant parasites which found no such change in HDAC1 regulation or abundance (eg Siddiqui et al 2017, Yang et al 2019). Stage matching is a particular challenge in such experiments given the differences in cycle progression between ARTR and ARTS parasites, and it isn't clear that this has been adequately controlled for to have confidence in these results, particularly given their contradiction of previous analyses. The abundance of PfHDAC1 changes considerably throughout the asexual intraerythrocytic cycle, (out of synch with the control used here actin), so potential stage-mismatch might contribute to apparent differences here. Again, explicit mention of replicates is lacking. The authors also mention genes regulated by HDAC1 as including genes related to processes related to artemisin resistance, but this is hard to sustain - indeed with so many genes apparently substrates of HDAC1 it would be highly surprising if there were no overlap with some genes in pathways related to artemisin resistance. An accompanying experiment demonstrating an increase in survival (of both ART resistant and ART sensitive lines) in an artemisinin ring stage survival assay is intriguing, after using a possible inhibitor of HDAC but these results are hard to reconcile with a dynamic transcriptional response. (Why was this done with an uncharacterised inhibitor, rather than the more specific HDAC1 overexpressor/knockdown system? An accompanying RNAseq analysis is described, but the analysis is piecemeal and selective, with the authors pointing out candidate genes representing categories plausibly linked to artemisinin resistance. I found this section unconvincing and indirect - lots of genes are changed in these experiments, and so they inevitably include some that are feasibly linked to artemisinin resistance, but the one gene convincingly known to modulate resistance, K13, isn't mentioned, and presumably wasn't specifically changed in this analysis.
5. A previous study by the laboratory of Christian Doerig (Eukaryot Cell. 2010 Jun; 9(6): 952-959.) reported that HDAC1 activity (unclear which of the HDACs) is associated with Pfcrk-3). This activity may not correspond to the HDAC1 characterised here, but deserves some discussion.
6. The Western blots are letterboxed and in some cases appear to crop out bands on the limit of the image (eg Fig 5, 6). Please provide fuller pictures of the blots and indicate the relevant bands if there are several background bands.

Minor issues

The text uses breaking spaces for the gap between genus abbreviation and species throughout. Replace with non-breaking spaces. Abstract: "is correlated with parasitemia progression" - Unclear meaning. Reword. Introduction "closes in on 400,000 deaths annually" Unclear meaning/vernacular usage. Reword. Very long paragraph on pages 3-4. Reorder logical flow and break into smaller paragraphs to make this more easily read. "Given the evidence of the role of HDAC inhibition in the emergence of chemotherapeutic resistance in mammalian system" - needs a reference - no mention of this phenomenon up until this point of the manuscript

Referees cross-commenting

I agree with the other reviewers comments. Although the manuscript contains a very large number of complex experiments, necessary controls, sufficient replicates, and appropriate analysis are missing from many of the experiments.

I appreciate that the experiments referred to would require a very substantial time and resource commitment to complete, but in their current form, many of these experiments are not safely interpretable.

Significance

This manuscript makes major claims for HDAC1, in particular for its role in artemisinin resistance. Such a link would be significant, but I regard few of these claims as having been robustly substantiated in this manuscript. The CHIP-seq evidence is of interest as a useful dataset, particularly if accompanied by relevant controls

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

Learn more at Review Commons

---

Reply to the reviewers

General Comments

We thank all reviewers for providing very detailed, knowledgeable, and informative reviews.

All reviewers were complementary about the data and the rigor of the study. Reviewers 2 and 3 commented on the significance of the work, and their assessments were complementary, specifically about the fact that it bridges several previous studies and links these to kinase-phosphatase regulation on the BUB complex. We agree that this a major strength of the work. That is why we also believe the comment by reviewer 1 that “most of the phenotypes/observations are consistent with the literature and not surprising” is actually a strength and not a weakness. Sometimes manuscripts that bring together various different findings into one conceptual model can be very powerful, even if each finding in isolation is not so surprising. In this case, the concept that a dual kinase-phosphatase module integrates two major mitotic processes will, we believe, prove to be a significant breakthrough that helps to explain how these processes are properly integrated at kinetochores.

The main criticism of all reviewers related to the interpretations and writing style, which in general, we felt were valid. We will take on board all these comments, reword the manuscript during revision, and provide a detailed response to each of these points at resubmission.

In terms of points requiring new experiments, there were 3 in total:

1) Reviewers 1 and 3 raised an important issue about the feedback loop which will be addressed with new experiments to uncouple the feedback.

2) Reviewer 2 made an important point about KNL1 levels, including a good suggestion to perform FRAP analysis to examine BUB complex dynamics when MELT numbers are increased. We will carry out this experiment prior to revision.

3) Reviewer 1 had a second major comment regarding the modulation of MELT number and how this cannot be directly linked to PLK1/PP2A levels. We have 3 new experiments to add regarding this, performed already, which are discussed in the section below.

All other comments were textual points that in most case we felt were valid. They showed that all reviewers had a very good grasp of the paper, the concepts, and the field in general. So, we finish by thanking all reviewers again for their thorough and detailed assessments of our manuscript. The comments they raised will help us to improve the manuscript after revision.

Description of the planned revisions

Three main points:

1) The role of the feedback loop [reviewers 1 and 3]:

The general issue is explained succinctly by reviewer 3’s comment:

“The argument linking the negative feedback loop to biological functions is not straightforward. The authors provide evidence in Figure 1 for regulatory pathways between docked PLK1 and bound PP2A. However, their assays in Figure 2 bypass the feedback loop by directly modulating PP2A activity. These experiment supports an argument that the kinase/phosphatase activity balance is important, but do not address the feedback loop specifically (which could potentially be done using mutations that disrupt the feedback regulation). The claim that "a homeostatic feedback loop maintains an optimal balance of PLK1 and PP2A on the BUB complex" is too strong because there is no direct evidence connecting the feedback loop to optimal function.”

This is a good point that we will address at revision. We demonstrate that the enzymes regulate each other on the BUB complex in figure 1 (PLK1 recruits PP2A, and PP2A removes PLK1), which balances their levels on the BUB complex. To determine consequences of upsetting this balance we locked either the kinase-bound or phosphatase-bound states (Figure 2). Importantly, this is required to assess direct phenotypes associated with each, but it does not directly demonstrate the role of the feedback loop. To do this we will generate mutants, as suggested, and analyse their phenotypes.

We will mutate the PLK1 binding site (T620A) and recover the PLK1-regulated sites in the KARD motif to phospho-mimicking aspartates (S676D/T680D), analyzing effects on PLK1/PP2A recruitment, chromosome alignment and SAC strength. We predict that this will remove PLK1 and recover some PP2A, but to lower levels overall than the BUBR1-B56 fusion. In that case the phenotypes will probably be milder, but that would not change the overall conclusions.

We maintain that locking PLK1 on its phospho-binding site (in BUBR1-DPP2A) is the ideal scenario to test direct PLK1 roles, but we will also now create alanine mutants of the PLK1 site (S676A/T680A) and the CDK1 site (S670A) to address the feedback loops controlled by CDK1 and PLK1. Our prediction is that these will skew the balance towards PLK1, without fully removing PP2A, again likely to produce milder intermediate phenotypes.

It is definitely worth testing these predictions, because it would directly address the role of the feedback loop and it would avoid relying solely on “artificially high levels” as mentioned by reviewer 1. One final point on this however, the PLK1 recruitment in DPP2A cells is not artificial – it is PLK1 bound to its native phospho-motif when PP2A is unbound (without any feedback from PP2A this phospho-site and PLK1 binding increase to the observed maximal levels). The fusion of B56 is admittedly less optimal, but this does still lock the phosphatase-bound state in a set stoichiometry, crucially in the absence of kinase. This is required to assess direct phosphatase effects. These PLK1/PP2A levels may well be higher than observed physiologically on the BUB complex when considering the behavior of all BUBR1 molecules, since we doubt they ever reach 1:1 stoichiometry with either PLK1 or PP2A. However, the feedback loop is operating within individual molecules (figure 1), which may well individually flip between PLK1 or PP2A bound states. This may occur on certain molecules at specific times. Therefore, locking the PLK1.PP2A-bound state on all molecules is, in our opinion, still a valid and useful perturbation to assess function of these two states.

2) The increase recruitment of BUB1-PLK1/PP2A when MELT numbers are increased [reviewer 2]

"While in the 12x and 19x mutant conditions there are more molecules of BUBs per Knl1, the overall BUB levels are the same as in wild-type controls. Since the MELT repeat used throughout the paper is a consensus sequence that is likely optimal for BUB binding, it is possible that the phenotypes of the 12x and 19x mutants are explained because of an increase in the affinity of BUBs for Knl1 rather than overall levels. This would also help explain why Knl1 and BUBs are observed at the spindle midzone in the 19x mutant (Fig. S4)"

The reviewer raises an important issue here, when stating that increasing MELT numbers decreases KNL1 kinetochore recruitment. This has the net effect of normalizing overall BUB1-PLK1/PP2A kinetochore levels, even though BUB1-PLK1/PP2A recruitment per KNL1 molecule is increased. That is why we were careful to state BUB1-PLK1/PP2A were increased “on each KNL1 molecule” and not “on kinetochores” when referring to the effect in the 12x/19x MELT mutant. However, this could easily be misinterpreted so this point will be clarified at revision.

The issue of why the phenotype is so dependent on kinase/phosphatase level per KNL1 molecules is an important one however, which has puzzled us until now. We think the suggestion to look at turnover by FRAP is a good one, because enhanced binding strength could underlie the phenotype here, and potentially explain the lack of disassembly at anaphase. We will perform these experiments at revision to see if they can clarify the issue.

3) The link between MELT number and PLK1/PP2A levels [Reviewer 1]

“My second comment relates to the fact that the two parts of the paper are not directly linked although the authors try to do this. They nicely manipulate the MELT repeats on KNL1 to change the number of Bub complexes. However, they cannot directly link the data to changes in Plk1 and PP2A-B56 levels only as many other things are changing. By increasing MELT numbers Bub complex and Mad1/Mad2 levels increase as well as an example and this makes interpretations complicated. To me these experiments are not addressing the main conclusions of the paper.”

We do not agree with this overall assessment, but there are two elements to this comment: the effect of modulating MELT number on SAC strength (and its link to PLK1) or on KT-MT stability (and link to P2A). We will therefore discuss each separately:

For SAC regulation, we feel that the data is clear and the interpretations are justified, although we will add new data to support this point after revision. Increasing MELT number causes defects in MELT-BUB dissociation and SAC silencing (4a-c). Importantly, these phenotypes can be completely rescued by inhibiting PLK1 (4d-e). So, we do link the effects of high MELT number to PLK1 activity. Our interpretation is that when MELT numbers are increased the ability of PLK1 to phosphorylate these motifs and maintain the SAC platform is enhanced (when MPS1 is inhibited pharmacologically or upon KT-MT attachment). So, whilst it is true that many factors, such as the kinetochore levels of BUB/MAD1/MAD2, are crucial for the SAC, the ability of PLK1 to maintain these levels (via pMELT-BUB1) is crucial and that changes as MELT number increases. This contributes directly to the observe SAC silencing phenotype, as confirmed by the complete rescue of this phenotype after PLK1 inhibition.

We did also explore the possibility that increased BUB1 activity could also contribute to SAC strengthening, for example, by enhancing Aurora B recruitment to centromeres. However, BUB1 inhibition did not alter SAC strength or MELT dephosphorylation kinetics. We will add this data after revision.

We also evaluated the levels of phosphorylated MAD1-pT716, which is important for MCC assembly (Ji et al. 2017, Ji et al. 2018, Faesen, 2017). Our data show that WT and 19xMELT exhibit similar MAD1-pT716 levels during a nocodazole arrest and following MPS1 inhibition. In summary, the main changes we observe are elevated BUB1 levels due to MELT phosphorylation, and increased BUB1 phosphorylation on pT461 (as shown in Figure 4h). All this points towards a localized effect of PLK1 on/around the BUB complex. We will add this data and make this point clearly at revision.

For KT-MT attachment regulation, we agree that we do not have a similar way to inhibit PP2A-B56 activity to rescue hyperstable microtubule attachment when MELT numbers are high. For this, we require a way to rapidly inhibit PP2A-B56 activity after attachments have formed, something that is not technical feasible at the present. We can also not say for certain that reduced MELT numbers destabilize microtubule due to lack of PP2A, however we feel this is the most like interpretation for the following reasons. The phenotype of removing PP2A from BUBR1 or removing the MELT from KNL1 (along with all associated factors), is identical: mutant cells have comparable chromosome misalignment due to unattached kinetochores (compare 2F-I with 5A-D). Therefore, the additional factors lost by removing the MELTs cannot be having such a strong impact in KT-MT attachment. The obvious factor that could affect attachment strength is again BUB1, via Aurora B recruitment to centromeres. However, loss of BUB1 (after MELT removal) is predicted to enhance attachment stability (reduced Aurora B) and not decrease it, as we observe. So, whilst we cannot definitely conclude that modulating MELT number affect attachment stability via PP2A, we feel that this is certainly the most likely explanation. We will state this clearly in the revised text.

Description of analyses that authors prefer not to carry out

“SAC strength of BubR1 WT, ΔC and B56γ was analysed in the presence of nocodazole + MPS1i. It would be interesting to see what the phenotypes are without MPS1i [Reviewer 1]”

In the absence of MPS1i basal MELT phosphorylation increases (DC) or decreases (B56g) as predicted (Figure 2d; compare timepoint 0 all conditions). This does not cause any change to SAC strength when all kinetochores are unattached in nocodazole (not shown). The sensitize SAC assay (nocodazole + MPSi) has been used by many groups (originally Santaguida et al, 2011; Saurin et al, 2011), because it reduces SAC signals from all unattached kinetochores which would otherwise produce a saturated response. In this case, we specifically chose a dose of MPS1 inhibitor that gave a partial SAC response from which we could observe either strengthening or weakening – a key point of the assay. Indeed, this showed that the SAC was strengthened (DC) or weakened (B56g), as predicted (Figure 2E). The only other way to do this, which has been used by some in the literature, is to use a low dose of nocodazole which prevents all kinetochore from signaling to the SAC. We specifically wanted to avoid this situation because then you cannot untangle the effects on SAC and KT-MT attachment stability – this was crucial in our case.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

A dividing cell relies on both error correction and spindle assembly checkpoints (SACs) to ensure accurate chromosome segregation. The PLK1-PP2A pair resides at the interface of the two pathways and makes these pathways robust and responsive. By generating fusion proteins and mutagenesis to perturb PLK1 and PP2A activities in HeLa cells, the authors found that both enzymes constitute a negative feedback loop on the scaffold BubR1. The docked PLK1 phosphorylates the PP2A binding site to recruit PP2A, while the PLK1 docking site is dephosphorylated after PP2A recruitment. They also manipulate kinetochore recruitment of PP2A to show effects on the SAC and on kinetochore-microtubule attachments. Independently, by varying active MELT repeats on KNL1, the authors found that decreasing MELT repeats weakens the SAC with unstable attachments, while increasing MELT repeats strengthens the SACs with hyperstable attachments. Thus, the recruited level of BubR1 determines SAC strength and attachment stability. The data included in this paper are solid, and the concepts are interesting, but these potential strengths of the manuscript are obscured because the logical progression and presentation are difficult to follow.

Major comments:

1. The first three paragraphs of the introduction lead up to the question: "Why the same phosphatase complex is used for both process is still not clearly understood". One conceptually simple answer is that the SAC should be silenced as attachments are stabilized, so it makes sense to use the same enzyme to accomplish both tasks. If the authors have something else in mind, they should clarify.
2. The argument linking the negative feedback loop to biological functions is not straightforward. The authors provide evidence in Figure 1 for regulatory pathways between docked PLK1 and bound PP2A. However, their assays in Figure 2 bypass the feedback loop by directly modulating PP2A activity. These experiment supports an argument that the kinase/phosphatase activity balance is important, but do not address the feedback loop specifically (which could potentially be done using mutations that disrupt the feedback regulation). The claim that "a homeostatic feedback loop maintains an optimal balance of PLK1 and PP2A on the BUB complex" is too strong because there is no direct evidence connecting the feedback loop to optimal function.
3. The paragraph starting with "Given the roles of PLK1 and PP2A ...": How does the kinase-dominant situation destabilize KT-MT attachments? Is it by inhibiting PP2A or by phosphorylating some kinetochore components? The former pathway is unclear because the authors show that PLK1 promotes PP2A recruitment in Figure 1.
4. The paragraph starting with "In summary, a balanced recruitment of PLK1 and PP2A ...": What is meant by "phosphorylation sites that block KT-MT attachment" (used twice in the paragraph)? Block means prevent binding, as opposed to activities that destabilize attachment by promoting unbinding. Which do the authors mean? The following is also unclear: "kinetochores are no longer responsive to MT attachment". If attachment is blocked, as stated in the previous sentence, then what does it mean to say that kinetochores are not responsive to attachment (which never occurred if it was blocked)?
5. Figures 1-2 and Figures 3-5 are separate concepts, but this is never explained clearly in the manuscript. Specifically, Figures 1-2 focus on the antagonism between PLK1 and PP2A activities, whereas Figures 3-5 focus on changing both activities in the same direction (either increase or decrease). There is no transition from one part to the other. Both concepts should be explained and the hypotheses stated clearly.
6. Several words are used ambiguously. "Homeostasis" is vague, and it is unclear what exactly the authors mean. As discussed above, the meaning seems different for figures 1-2 vs figures 3-5. "Reciprocal changes" is also unclear (and seems misleading) because the perturbations of PLK1 and PP2A levels are in the same direction (more MELT motifs means more binding sites for both). For "preserved" (description of figure 4F), it's unclear if the authors refer to the Bub1 and BubR1 levels at metaphase or the change between prometaphase and metaphase. The authors should clarify what they mean and how it is measured.
7. "too much kinase-phosphatase module would cause a strong SAC and hyperstable KT-MT attachment, and too little would cause a weak SAC and hypostable KT-MT attachments": The reasoning behind these predictions is not clear. For example, why does high phosphatase not silence the SAC as explained earlier in the manuscript?
8. The mitotic exit assay in Figure 4G is hard to interpret. Mitotic duration depends on establishing correct attachments and then silencing the SAC. 19xMELT could affect both. A better measurement of SAC silencing would be time from metaphase alignment to anaphase.
9. The paragraph starting "In summary, PLK1 is able to phosphorylate MELTs to recruit Bub1": The authors should clarify what was already known and what advance they are making. Similarly, the sentence starting with "Therefore, the number of MELT motifs ..." should also clarify the advance relative to previous findings.

Minor comments:

1. Please include pages numbers (and line numbers are also helpful).
2. Previous literature (PMID 17785528) suggests that phosphorylation of pT620 is important for KT-MT regulation but not SAC signaling. Can the authors comment on this?
3. SAC silencing seems more appropriate than "mitotic exit" in the last sentence of the second paragraph.
4. Figure 1 would be clearer with images and the relevant quantifications together in the same panel.
5. In the first paragraph of Results, the authors primarily explain the impacts of PP2A on SAC silencing but not on KT-MT attachment, even though the topic sentence seems equally weighted to both.
6. Some terms are not defined when first introduced, such as the KARD domain and B56gamma.
7. Figure 1JK: how were mitotic cells enriched and harvested?
8. Figure 2C: a log scale may help show changes in both directions.
9. Chromosome alignment assays in Figure 2 are not so informative because perturbation either way can generate misaligned chromosomes. The primary figures are dense with data, so these results can be made supplemental.
10. Figure 3F: can the authors comment on the B56gamma decrease between nocodazole and MG132 conditions?
11. Figure 4A: the data are difficult to interpret as presented. It is not clear whether SAC signaling changes monotonically with number of MELT repeats. It would be better to plot MELT number vs a summary statistic (such as time to 50% mitotic exit or something else that the authors find informative).
12. Figure 6. "uncouplr" should be "uncouple".
13. Missing references in bibliography: Ghongane et al 2014, Roy et al. 2020.

Significance

The proposed model is conceptually significant, and this mechanistic work bridges several previous findings. Biochemical studies suggest that KNL1 harbors PLK1-PP2A and Bub complex, and functional studies suggest that truncation on MELT motifs generates mitotic errors (PMID 24363448, 24344183). Furthermore, biochemical and functional assays suggest that PLK1 is a key regulator of the SAC (PMID 33125045). In this work, the authors link functional consequences to biochemical interactions among KNL1, BubR1, PLK1, and PP2A, which is an advance.

This work should appeal to the kinetochore and cell cycle communities, but the logical flow needs to be improved.

My most relevant expertise is in mitotic kinases and regulation of kinetochore microtubules.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript by Corno et al the authors investigate how the kinase/phosphatase balance is regulated at kinetochores during mitosis. When kinetochores are unattached, mitotic kinases signal through the spindle assembly checkpoint (SAC) to prevent progression through mitosis. Once attachments have occurred, the activities of kinetochore phosphatases silence the SAC and stabilize kinetochore-microtubule attachments to promote exit from mitosis. Key regulators of this signaling are the BUB proteins Bub1 and BubR1, which recruit the kinase Plk1. BubR1 also recruits the phosphatase PP2A-B56 in a Plk1-dependent manner. By analyzing the recruitment of Plk1 and PP2A-B56 at unattached kinetochores, the authors uncovered a mechanism by which Plk1 and PP2A-B56 negatively regulate each other's recruitment, thereby maintaining an optimal kinase/phosphatase balance. Disrupting this balance can lead to either a hyperactive SAC or premature stabilization of erroneous attachments.

Overall, this is a very careful study and the data is clear and consistent. My only comments are in regards to the interpretation of the Knl1 data on the manuscript.

Comments

• In figures 3 and 4, the authors engineer a system to manipulate the levels of BUBs at the kinetochore by modulating the number of MELT repeats on Knl1. For this, they mutate all 19 MELT repeats on Knl1 and then add back discreet number of MELT motifs that follow a consensus sequence. They find that a 6x mutant, which contains 6 repeats of this consensus MELT motif, is sufficient to rescue the functions of Knl1 on the SAC and on chromosome alignment. However, 12x and 19x MELT repeat versions show an increased SAC response and increased stability of kinetochore-microtubule attachments. The authors interpret this as a result of increased kinetochore levels of BUBs, and therefore, increased levels of both Plk1 and PP2A-B56 (Fig. 3). However, from their representative images in Fig. S3A, it does not appear as if BUB levels are significantly increased in cells expressing the 12x and 19x Knl1 mutants, compared to wild-type controls. Considering that in the 12x and 19x mutants Knl1 recruitment is reduced by more than half of controls (Fig. S3B) and because the authors normalized their BUB kinetochore intensity levels by the Knl1 values, it makes it seem as if BUB kinetochore levels are increased in the cell under these conditions. While in the 12x and 19x mutant conditions there are more molecules of BUBs per Knl1, the overall BUB levels are the same as in wild-type controls. Since the MELT repeat used throughout the paper is a consensus sequence that is likely optimal for BUB binding, it is possible that the phenotypes of the 12x and 19x mutants are explained because of an increase in the affinity of BUBs for Knl1 rather than overall levels. This would also help explain why Knl1 and BUBs are observed at the spindle midzone in the 19x mutant (Fig. S4). To distinguish between these possibilities, the authors might consider doing FRAP experiments using fluorescently labelled Bub1 or BubR1 and measure BUB protein dynamics at kinetochores.
• Also in regards to the 12x and 19x mutants, because these reduce Knl1 kinetochore levels, this fact alone might explain some of the observed phenotypes, such as the mild defects in chromosome alignment (Fig. S5).

Significance

This is a significant advancement in our understanding of how the spindle assembly checkpoint is regulated by kinases and phosphatases at the kinetochore. The authors used very precise manipulations to dissect how these components are balanced and they uncovered a very interesting negative feedback mechanism. They also provided significant evidence for the importance of this balance in normal mitotic progression. This work will be of broad interest to cell biologists, as well as cancer biologists.

My expertise is on the fields of cell biology, mitosis and cell division mechanisms.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The proper segregation of chromosomes during mitosis depends on correct kinetochore-microtubule (KT-MT) attachments which is monitored by the spindle assembly checkpoint (SAC). The Bub complex composed of the Bub1-Bub3 and BubR1-Bub3 complexes plays an important role in regulating KT-MT by recruiting Plk1 and PP2A-B56 as well as the SAC components Mad1/Mad2 and Cdc20. The Bub complex is recruited to kinetochores by Bub3 binding directly to phosphorylated MELT repeats on KNL1. The MELT repeats are predominantly phosphorylated by Mps1 but can also be phosphorylated by Plk1.

Here the authors investigate the Plk1 and PP2A-B56 module on BubR1 further. It is already known that Plk1 is recruited to BubR1 T620 (a Cdk1 site) and that this is important for proper chromosome segregation. Furthermore, BubR1 binds to PP2A-B56 through and LxxIxE motif that is phosphorylated by Plk1 to stimulate binding. The main message of the paper is that the Plk1-PP2A-B56 module on BubR1 is crucial for integrating SAC and KT-MT attachments and that a homeostatic negative feedback loop between Plk1 and PP2A-B56 exists to limit the levels of these enzymes. As outlined below I do not think that their experimental evidence/setup can be used to draw these conclusions. Overall, I have limited comments on the technical execution of experiments as this is overall done well but more on interpretations of results and what conclusions can be drawn. Also, most of the phenotypes/observations are consistent with the literature and not suprising.

Major comments:

The authors generate a scenario where high levels of Plk1 are recruited to BubR1 by removing the entire C-terminus to remove the PP2A-B56 binding site and a situation of high PP2A-B56 by replacing the BubR1 C-term with B56gamma. Firstly, these are crude mutants generating extreme situations - specifically the fusion of B56gamma which likely creates artificially high levels of BubR1-B56gamma making it difficult to make conclusions on the physiological levels. These mutants are then analysed in a number of assays and phenotypic consequences analysed in the presence of nocodazole + Mps1 to evaluate SAC strength. It would be interesting to see what the phenotypes are without Mps1 inhibition. I cannot see in any way how the authors from two end points (high kinase or high phosphatase) can reach a conclusion that a homeostatic negative feedback loop exists between these enzymes that is critical for integrating SAC and KT-MT interactions. They can only conclude on what happens if you have too much kinase or phosphatase but no data are addressing what happens if you manipulate the cross-talk on BubR1. The experiments to do must be to carefully tune the proposed cross talk and then monitor what happens. This can be done by bypassing Plk1 regulation of the PP2A-B56 binding site or increasing the strength of the Plk1 binding site. Furthermore, there is no data to show that cross-talk is changing in response to changes in KT-MT attachment status - is the levels and activities of Plk1/PP2A-B56 on BubR1 changing. Yes Plk1 and PP2A-B56 can regulate each other on BubR1 but is this regulated as proposed.

My second comment relates to the fact that the two parts of the paper are not directly linked although the authors try to do this. They nicely manipulate the MELT repeats on KNL1 to change the number of Bub complexes. However, they cannot directly link the data to changes in Plk1 and PP2A-B56 levels only as many other things are changing. By increasing MELT numbers Bub complex and Mad1/Mad2 levels increase as well as an example and this makes interpretations complicated. To me these experiments are not addressing the main conclusions of the paper.

Specific comments:

1. I would caution the interpretation of phenotypes being suppressed by Plk1 inhibitors. This does not address whether it is BubR1 bound Plk1 that is specifically affected - several Plk1 substrates could be contributing. Similar for Mad1 phosphorylation they cannot conclude that it is Plk1 bound to BubR1 phosphorylating Mad1.
2. On page 4 the authors write: "We sought to modulate MELT numbers in a way that would allow BUB complex levels to be increased or decreased in a graded manner, thereby causing reciprocal changes to PLK1 and PP2A levels." I do not see how this will result in reciprocal levels as total Bub complex levels are increased.
3. Page 7 first paragraph authors write: "remained high despite inhibition of Mps1". This is not completely correct as levels are dropping dramatically after 5/10 min of Mps1 inhibition. Total drop seems more than WT situation.
4. Page 7 second paragraph authors write: "implying the elevated PLK1 in this situation (Figure 3E) is also able to better amplify MELT signalling..." This cannot be concluded as the starting levels are so much higher in 19XMELT than WT and there is no data to show this is due to more Plk1 bound to BuBR1.
5. Page 7 second section authors write: "Figure 5B shows that KNL1deltaMELT causes severe chromosome misalignments, as expected, given the PP2A-B56 binding to KNL1 is inhibited in this situation". Multiple things are changing so this cannot be interpreted only as a readout of PP2A-B56. With no MELTs there is no recruitment of SAC proteins.
6. Page 7 second section authors write: "Therefore, the number of MELT motifs is crucial for determining the stability of KT-MT attachments, most probably by setting the levels of PP2A-B56..." Again many things are changing so impossible to interpret data in light of PP2A-B56.
7. One possibility not mentioned by authors is that PP1 bound to KNL1 cannot act as efficiently on some MELT repeats and that the dephosphorylation by PP1 of the 19xMELT is different from KNL1 WT which would impact on their results.

Significance

See above.

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

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This study finds that probucol enhances mitophagy and mitochondrial quality control, suggesting its target ABCA1 may be a good target for the treatment of Parkinson's disease. Using an AI approach based on reported enhancers of mitophagy, they identify a list of 79 compounds to test in an assay based on Parkin- and CCCP-mediated mitophagy. One of the hits is probucol, an inhibitor of ABCA1, an ATP-dependent lipid transporter. Probucol causes a slight increase in mitochondrial turnover in HeLa cells, and increases Paraquat-driven mitophagy in flies. To explore the effects of lipids on the probucol-induced effects on mitophagy, it is observed that overexpression of human ABCA1 in flies suppresses the effect of Paraquat, and that inhibitors of DGAT reduces mitophagy as well, which may result from altered lipid droplets metabolism. Probucol alleviates motor phenotypes in fly and zebrafish PD models induced with mitochondrial dysfunction. The rough-eye fly phenotype observed with overexpression of Rho-7 (PARL ortholog) is modulated by ABCA expression, in addition to modulation of Parkin/PINK1. Finally, probucol appears to work at late stage mitophagy by modify LC3 lipidation and increasing lysosomal areas.

Overall, this is an interesting paper, which makes a good case for the consideration of probucol as a potential treatment for PD. This is an improved version of a manuscript from the same group I had reviewed for another journal, and many of the issues raised have been resolved since. Yet, I remain to be convinced about some points:

Major points

1. The data provided does not support the claim that the AI selection of compounds had anything to do with the "high-rate" of success. To be able to claim this, there should be a comparison with a randomly selected set of ~80 compounds from the 3231 candidates, and show that for those, there is a lower chance of finding hits, if any at al. This section needs to be improved and a more compelling case presented. Just to be clear, it doesn't take anything away from the rest of the paper, but the authors make claim about this AI method that I don't believe are justified. Either provide more data or change the text and tone down the claims about the AI method.
2. The effects of probucol on LD area and mitophagy in cells are interesting, but the data presented are not sufficient to claim that the effects are ABCA1-dependent. To do this, they must measure LD area and mitophagy in cells with shRNA against ABCA1, followed by CCCP and probucol (or DMSO). Probucol should have no effect when ABCA1 is knocked down. The experiments shown in Appendix Fig. S4 don't exactly show that; S4C lacks probucol treatment, and S4D/E measure mitochondrial volume, which is not mitophagy. Furthermore, it would important to show at least one dose-response curve for probucol in mitophagy, in cells. What is the IC50? Does it match that of the known Kd for ABCA1? This would be further evidence that the effects of probucol are "on-target".
3. In Figure 5A, what is the baseline climbing % WITHOUT paraquat? This is important to show to provide a comparison point for the level of rescue induced by probucol.
4. The epistatic relationships for the rough eye phenotype (REP) are confusing and need to be better explained/presented. A diagram showing the effect of each gene on the phenotype would be useful. Also, given that the effect of probucol in cells are Parkin/PINK1-independent, it is somewhat confusing to find that hABC1 effects are Parkin-dependent for the REP. Please clarify.
5. In Figure 7E, again, data for the baseline climbing % is lacking. Furthermore, it is not clear why the % climbing of ABCA RNAi+DMSO+PQ (bottom graph) is higher than the control condition mCherry RNAi+DMSO (top graph). Are these % for the bottom and top graphs comparable? If not, then the bottom graph should also include a baseline condition without paraquat. Finally, why is paraquat not reducing climbing in ABCA RNAi or Atg7 RNAi? Controls are lacking in these experiments (as in Figure 5A above).

Minor point

1. In Figure 5C, experiment with SR3677 is shown, but mentioned nowhere in the text. I understand this is a ROCK inhibitor, but this should be mentioned.

Significance

Nature and significance: This is an interesting report, but there is no major conceptual advance in our understanding of mitophagy and neurodegeneration. Nonetheless, the discovery of a new potential target (ABCA1) for treating mitochondrial dysfunction of worth reporting.

Comparison with existing knowledge: the report is original, but there have been a few reports of the effect of perbucol in Parkinson's model (e.g. Ray et al 2014, Cell Death Discov, PMID 24407237), or ABCA1 variants (Ya & Lu, 2017, Med Sci Monitor). Those should be acknowledged.

Audience: researchers in the field of Parkinson's research, but also mitophagy and lipid metabolism.

My expertise: Parkin/PINK1 pathway, mitophagy, pharmacology, structural biology.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In their manuscript, Moskal et al. performed an AI-based screen to find potential mitophagy inducers and subsequently test them in model organisms of human neurodegenerative disorders, whereby mitophagy induction might be beneficial. The authors found that probucol, a retired lipid-lowering drug, is a candidate mitophagy inducer at least in cell culture and flies. Furthermore, they reported physiological benefits in association with probucol treatment to flies fed paraquat, a toxin that causes loss of dopaminergic neurons. This manuscript has some potential especially as it implicates an unknown target for mitophagy, ABCA. However, the paper can't be published in its current form for the several reasons:

Major comments:

The sequence in which the experiments were presented in this paper is illogical and sometimes confusing: (i) in page 5, when the mitophagy-inducing effect of probucol was shown in paraquat-treated flies, the corresponding physiological relevance of this observation, ie effects on climbing ability and survival, should have been reported immediately and not later on in another section of the manuscript. Since paraquat induces mitophagy, It is not very intuitive to imagine that further increase in response to probucol is beneficial. This leaves the reader wondering whether further increase in mitophagy is beneficial or detrimental. In fact, excessive mitochondrial clearance has been shown to render caloric restriction - the most robust antiaging and autophagy-inducing intervention, detrimental (PMID: 30929899). (ii) Similarly, causality testing of the role of ABCA-autophagy axis in mediating the salutary effects of probucol in flies, which were reported in the very end of the manuscript should have been mentioned early on upon introducing such health-promoting effects. (iii) Another example is also LC3 lipidation and lysosome abundance, which support the pro-autophagic effect of probucol. Instead of being reported with other evidence supporting mitophagy induction at the beginning of the manuscript (page 5) were kept for no obvious reason to be mentioned in the very end of the results (page 9) after all other mechanistic and in vivo physiological testing results were reported!

As such, the manuscript needs extensive re-writing, thereby also removing unnecessary data that only distracts from the main message. By that, I mainly refer to the rough eye model induced by expression of rhomboid-7 in the fly eye. The sudden shift towards using the rho-7 fly eye model along with extensive characterization of the role of pink1 and parkin is distracting from the main objective of the study: finding a novel mitophagy inducer in probucol and describe its mechanism of action. Instead of reporting causality testing of autophagy in paraquat-treated flies, where probucol induced mitophagy, the authors decided to discuss the fly eye model and go on to show the role of mitophagy, pink1 and parkin in this model. I suggest removing this model from the current manuscript and dedicate a future manuscript to the experiments performed there because they do not serve the purpose of the study. In fact, mitophagy induction in this model is harmful. As such, this model only dilutes the message and lowers confidence in the robustness of the reported cascade of events involving probucol>ABCA>mitophagy>physiological benefits.

Regardless, the experiments testing causality of autophagy and ABCA in Fig. 7e are misleading. There are no corresponding negative controls showing the benefits of probucol in absence of ABCA and Atg7 RNAi. The authors should not rely on a different set of experiments (Fig. 5a-b) done at a different time and another cohort of flies.

The authors applied an interesting screening approach using semantic textual similarity to a pre-defined list of positive controls. These positive controls comprised of 7 mitophagy inducers that primarily act on the NAD+-sirtuin pathway. NAD+, as also sirtuins, induce a myriad of effects, not only mitophagy. In fact, some might even argue that autophagy/mitophagy are only partly involved in the effects of these compounds (PMID: 34843394). Why did not the authors expand their positive control list to include other classes of mitophagy inducers? the reference (ref. 10) used has many other potential positive controls that do not necessarily increase apoptosis and mitochondrial damage. Furthermore, the sematic screen has been limited to papers published until 2014. Why was that the case? In fact, a crude Pubmed (not Medline) search using the term "mitophagy" returns 935 hits (till 2014), while from 2015 onwards it returns more than 5800 papers! This clearly shows that much more has been done after 2014. The search should be expanded, otherwise the authors are missing out on all major development that happened in the field.

The authors also reported that they filtered out the candidate molecules with any association to either the term "apoptosis" or "mitochondrial damage". How could you differentiate between causation or protection from mitochondrial damage when a compound is mention in the context of apoptosis/mito damage using semantic fingerprint?

How does ABCA1 KO and overexpression affect the pro-mitophagic action of probucol? This has been shown for DGAT, but not for ABCA which is a central finding in this manuscript.

• In page 1, the following statement is problematic: "we focused on the ultimate endpoint of mitophagy-clearance of damaged mitochondria from cells. Ultimately, if this step is improved, then the negative consequences of mitochondrial damage in the dopaminergic neurons may be mitigated". Increasing the clearance of damaged mitochondria is not necessarily beneficial. In fact, excessive clearance of damaged mitochondria renders autophagy inducing interventions harmful as mentioned above. It is also not clear whether and how the authors could tell that only damaged mitochondria were sequestrated. Even in absence of mitophagy induction at baseline (in absence of CCCP), this does not imply that only damaged mitochondria are cleared when damage is induced. Lack of evidence is not an evidence for lack of an effect.
• In page 4, the authors report that "the degradation of VDAC1 was increased at several time points by probucol treatment, but not in the absence of mitochondrial damage (Figure 2D, E)" the graph shows DSMO and probucol not CCCP vs DSMO, is this a typo?
• Fig 7A-B is very confusing. Why did not the authors use protease inhibitors to properly evaluate autophagy flux if that was the purpose? also how come starvation does not induce autophagy in these cells? what was the time point tested? Were the cells starved as a positive (or worse: negative) control?
• The authors reported that their screen could efficiently predict olaparbi upon leaving it out (top 3.9% of the 3231 compounds screened). How did the other 6 positive controls fair in this cross validation using the leave-one-out approach? please report this for the other 6 positive compounds as well.

Minor

• rather small sample size in most of the experiments with n=3-4 (eg, Fig 5 D, E F and G).
• Some figures do not show the data points and thus it is not possible to tell how big was the exact sample size (eg, Fig. 5A)
• how do the authors explain the discrepancy between Fig. 4G and Fig. 3D where PQ increased % of red-only/total mito area in one but not in the other?
• How can the results in Fig 6 quantified? If qualitative then many more representative images should have been presented.
• There appears to be a deleted lane from the western blot in Fig S2A. This needs to be declared in the legend, along with a justification.
• Western blots in general do not show the molecular weight. Please add this throughout the manuscript
• A lot of abbreviations are mentioned without spelling out what they stand for, eg, CCCP, ABCA, DGAT.
• Typo in page 1: the loss of which is (not are) responsible for the classical motor..
• Line numbers were not included in the submitted manuscript, thus I could not provide the authors with the exact position of any of the issues mentioned above.

Significance

Despite reporting novel findings, the manuscript has major flaws in terms of experimental approach and presentation of the results. The authors need to revise their screen, as also restructure the manuscript to be better reflect the findings, thereby improving the significance of both the physiological and mechanistic value of the study.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

In the manuscript by Moskal et al, the authors utilize an AI-based approach to identify mitophagy modulating compounds based on parameters from already FDA-approved drugs. From this screen, the authors find that the compound Probucol enhances stress-induced mitophagy. It appears to do this via its canonical target, ABCA1, via a PINK1/Parkin-independent mechanism. Importantly, the authors show that probucol improves phenotypes in fly and animal models with mitochondrial perturbations, suggesting that this pathway may hold some therapeutic value.

Major comments:

• I will mention that I have already reviewed a version of this manuscript at another location and here the authors have addressed my previous concerns.
• The key conclusions are robust and shown in multiple cell types and models.
• I do not have any requests for additional experiments, given that those present already sufficiently support the claims of the paper.

Minor comments:

• I am a little confused over the descriptions of the rough eye phenotype in Fig. 6. In the text on p7, the authors state that rho-7 overexpression causes the eyes to appear rough. Yet in the figure the eyes appear smoother - indeed the legend says they have "glossy appearing eyes". This seems the opposite to rough - unless I am missing something? Perhaps rewording the text would help with this.

Significance

• Here the authors identify a small molecule, probucol, that enhances mitophagy and in doing so implicate the ABCA1 pathway of lipid transport in this process. A great strength of the work is that in vivo data (in flies and fish) are obtained, which has clear disease-relevance. The manuscript is timely, in that lipid signaling and lipid droplets are being shown to have a strong regulatory role in autophagy.
• The authors use a novel AI-based approach to identify mitophagy-inducing compounds, and while this seems a good approach, I do not feel I have the expertise here to critically review this aspect. It is also noteworthy that the initial assay was designed to look at modulators of Parkin-dependent mitophagy, but the compound discovered appears to act independently of PINK1/Parkin.
• My main expertise is in autophagy and mitophagy cell biology and I do think the findings in this manuscript will be of interest to this field and those focussed on therapeutic approaches for diseases where impaired mitochondrial function has been implicated.
• The manuscript does lack a mechanistic understanding of the mode of action of probucol - how it is enhancing mitophagy is not clear. However, it is up to the authors and editors of the destination journal as to whether more work is done here, or in a follow-up study.

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

Learn more at Review Commons

---

Reply to the reviewers

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

Summary:

Ciliates extensively rearrange their somatic genome every time a new somatic nucleus develops from the zygotic germline nucleus. In this manuscript, Feng et al report the sequencing, assembly and annotation of the germline and somatic genomes of Euplotes woodruffi and the germline genome of Tetmemena sp. (whose somatic genome was sequenced and assembled by the same lab in 2015). They present a comparative analysis of developmentally programmed genome rearrangements in these two species and in the model ciliate Oxytricha trifallax. Their major findings are that:

(i) E. woodruffi and Tetmemena sp. eliminate a smaller fraction of their germline genome (~54%) from their somatic macronucleus (MAC) than O. trifallax (>80%)

(ii) Transposable elements (TE) represent a smaller fraction of the germline genome (~2%) in the first two ciliates than in O. trifallax (~15%). TEs are mainly located at the boundaries of germline chromosomes and in intergenic regions, but can also be found inside IESs

(iii) Several thousands of genes are scrambled in the germline genome of all three species

The authors have also addressed the possible origin of gene scrambling. They report an interesting association with local paralogy and propose a model for the emergence of the odd-even pattern of gene unscrambling between two paralogous copies.

Major comments:

1. Based on the statistics presented in Table 1, genome assemblies are of good quality, with a reasonable N50 size of germline (MIC) contigs. It seems, however, that no entire MIC chromosome could be assembled, since no two-telomere contig is mentioned in the list. As proposed by the authors (p.7) the presence of numerous TEs at the boundaries of MIC contigs (Fig S1) may have hindered the assembly of MIC chromosome ends. I would have appreciated to have more information on the "other repeats" (which seem to differ from tandem repeats according to Fig 2) and their location along MIC contigs.

   Subcategories of “other repeats” were included in Table S2 based on Repeatmasker annotations. We now analyzed the locations of other repeats in MIC contigs and include those as well in new Figure S1B. About 30% of “other” transposable elements are present at the boundaries of MIC contigs, which may also hinder the assembly. Notably, 35-45% of “other TEs” are in assembled, intergenic regions.

The definition of "Internal Eliminated Sequences" (IES) is not clear. The authors make a distinction between IESs and TEs. I understand that IESs are DNA segments that separate two macronuclear-destined sequences (MDS) in the germline genome. Thus they appear to be restricted to those regions that eventually yield gene-sized MAC chromosomes. IESs are eliminated between two pointers that may not be identical on both sides in case of scrambled genes. Some clarification is needed here.

To illustrate my point: I found the statement "with many TE insertions within IESs, suggesting that TE insertions may have generated IESs" particularly confusing (p. 9 lines 5-6). Does this mean that IESs extend beyond the ends of inserted TEs? The legend of Fig S1 should also be clarified.

We clarified the text and legend. IESs can extend beyond the ends of inserted TEs, even if the original IES is a decayed TE, due to subsequent sequence evolution at the boundaries or if the original insertion was into an existing IES. David Prescott referred to sequence evolution at the edges of IESs as “pointer sliding” (ref.36).

p. 10 lines 2-4 and Fig S2: Could the authors explain the difference they make between MDS (in the text) and CDS (in Fig S2)? My understanding is that a CDS is the entire gene coding sequence and may be made of multiple MDSs. If this is correct, the sentence should read "We compared the number of MDSs between single-copy orthologs for single-gene MAC chromosomes across the three species and found that the orthologs have similar CDS lengths".

Yes, we made the correction.

p. 12 lines 10-15: the discovery that paralogous MDSs can be found in scrambled genomic loci is interesting. If the two paralogs can be distinguished based on the number of substitutions, it would be informative to go back to individual reads and check whether each of the two copies can be incorporated in the unscrambled CDS (and at which frequency). Would the pointers be compatible with this?

The paralogous MDSs in the MIC are often not identical. The copy with the highest similarity is assigned as “preliminary match” by SDRAP (ref. 52), and others are assigned as “additional matches”. To validate SDRAP assignments, we did pairwise BLASTN alignments (“-task megablast”) of paralogous MIC MDSs and their corresponding MAC MDSs. We confirmed that in the three species, the preliminary match has the best or equally best pid (percentage of identity) in most cases. Therefore, the MDS assigned as preliminary match is more likely the paralog incorporated into the MAC chromosome.

We used genome assemblies of Euplotes woodruffi, which had the highest Nanopore coverage, to further investigate the frequency of MDS incorporation. We followed the reviewer’s suggestion and called SNP variants on both MAC and MIC genomes. For MAC SNP calling, we used Illumina reads as input for freebayes (ref a). For MIC SNP calling, we used Nanopore reads, instead of Illumina reads, to avoid non-specific short-read mapping on paralogous MDSs and to avoid the presence of any contaminating MAC reads. Variants were called and phased by PEPPER-Margin-DeepVariant (ref b), a new tool published in 2021 in Nature Methods, which has been reported to have similar accuracy to Illumina read variant calling, especially at high read coverage. We used the parameter “--pepper_min_coverage_threshold 20” to call confident variants when at least 20 reads cover the position. Only 92 MIC SNPs in the paralogous MDSs passed all filters of the program. Using this small set of MIC SNPs, we were unfortunately unable to distinguish which paralogous MIC MDS was incorporated into the MAC. Therefore, we cannot infer with what frequency one paralogous MDS is incorporated over another, until they become sufficiently diverged, which is compatible with the model.

a. Garrison E, Marth G. Haplotype-based variant detection from short-read sequencing. arXiv preprint arXiv:1207.3907. 2012 Jul 17.

b. Shafin K, Pesout T, Chang PC, Nattestad M, Kolesnikov A, Goel S, Baid G, Kolmogorov M, Eizenga JM, Miga KH, Carnevali P. Haplotype-aware variant calling with PEPPER-Margin-DeepVariant enables high accuracy in nanopore long-reads. Nature methods. 2021 Nov;18(11):1322-32.

The hypothesis that odd-even scrambled loci have evolved from paralogous genes in E. woodruffi is supported by the existence of paralogous MDSs, length conservation of MDS/IES pairs and sequence similarity between corresponding MDS and IES in a pair. The correlations presented for Oxytricha and Tetmemena are much less convincing (Fig S5D and E). I recommend that the authors are even more cautious in their statement on p.13 ("For Oxytricha and Tememena, the MDS and IES lengths for such MDS/IES pairs also correlate positively, but more moderately").

Thank you, we rephrased the text.

p. 15 last paragraph: Why did the authors focus only on TBEs inserted in non-scrambled IESs to look for orthologous TBE insertions? Is there a reason to believe that no recent TBE insertion occurred at other genomic loci? Or was it only for practical reasons? It is also not clear to me whether the authors have considered full-length TBEs or the presence of at least one TBE ORF.

This analysis was limited for practical reasons, because we identify position conservation of TBEs by aligning protein sequences of MAC genes. We only consider TBEs inserted in non-scrambled IESs in exons. It would be difficult and less meaningful to align completely non-coding MIC-limited regions.

Partial TBEs are also included if they contain at least one TBE ORF (detected by BLAST).

Furthermore, TE insertion cannot explain the origin of scrambled IESs, and TEs rarely map to scrambled IESs (Figure S1A), but there is a clear evolutionary model for the origin of nonscrambled IESs from decay of TBEs (ref. 49). Initial purifying selection would act on the TE to maintain its ability to self-excise, whereas we advocate for a different model for the origin of scrambled IESs by decay of paralogous MDSs.

p. 16: the authors report that some introns of E. woodruffi map "near" Oxytricha/Tetmemena pointers. How near? Based on the information provided by the authors, I don't think this observation necessarily implies that IESs were converted to introns (or reciprocally) during evolution. If this were true, shouldn't at least one intron boundary coincide exactly with a pointer? The authors should clarify this (also in the discussion, on p. 20, top paragraph).

We used a 20bp window (~7 amino acids), as described in the Methods, and added that to the Results. Full detail is provided in the Methods section, “Ortholog comparison pipeline and Monte Carlo simulations”. 103 E. woodruffi introns are within 20bp from the midpoint of Oxytricha/Tetmemena pointers. Among these, 43 intron boundaries overlap an Oxytricha or Tetmemena pointer. We observed 306 cases of precisely matching boundaries between any two species, where the exon junction of one species maps inside the MDS/IES pointer of another species, although we would only expect the boundaries of introns and IESs to coincide so precisely if they were recent conversions. Hence we feel that a window analysis is informative.

p. 19 2nd paragraph: the suggested mechanism explaining the 5' bias of IESs in E. woodruffi genes is unclear. How could germline recombination take place between a MIC chromosome and a MAC reverse transcript or nanochromosome? This would imply that DNA could be imported in the MIC. Is there evidence that this might occur?

The ability of TEs to invade the MIC demonstrates that even foreign DNA can be incorporated into the MIC. Since MAC DNA is present at high copy number, it offers a potential source for a recombination template that could erase IESs, as could an errant reverse transcript of one of the long noncoding template RNAs. Any of these would be infrequent events that would matter on an evolutionary time scale even if developmentally rare.

According to Figure 1, no scrambled genes have been reported in Paramecium tetraurelia. Within the frame of the proposed model, this is somewhat unexpected because this ciliate went through several whole genome duplications during evolution and harbors many paralogous gene pairs. Is there a reason why no gene scrambling took place in Paramecium?

Paramecium uses only TA dinucleotide pointers for IES elimination, unlike the rich diversity of pointers in spirotrichous ciliates. This limitation in its machinery may explain why no scrambled loci have been observed in Paramecium, despite the abundance of paralogs. Our model suggests that local MIC paralogy is associated with the origin of scrambling. But most of the paralogy reported in Paramecium is at the level of whole chromosomes in the MAC (ref. 104) rather than local MIC paralogy.

Minor comments:

p. 4 (4th bottom line): To my knowledge, ref #28 presents a draft (incomplete) MIC assembly of the Paramecium genome.

Thank you, we added reference 29 and adjusted the wording describing the quality of MIC genome draft assemblies.

p. 7 (last paragraph): "encoding" should be replaced by "carrying"

Thank you, we made the change.

p. 10 (2nd paragraph): insert a missing "o" into "nanochromosomes"

Thank you, corrected.

p. 10 (same paragraph): the weak 5' bias of IES distribution in Tetmemena should be shown (either as an additional panel in Fig 3 or in a Sup Figure.

Thank you, we added it as Figure S2C.

p. 24 2nd paragraph: "a" is missing in "Trinity, which is a software..."

Thank you, we made the correction.

CROSS-CONSULTATION COMMENTS

I agree with most comments of reviewer 3.

The authors have actually defined "TE" in the introduction (p. 6). Depending on the journal's rules for abbreviation use, it may not be necessary to define it again in the results section

Reviewer #1 (Significance (Required)):

Ciliates are unicellular models to study developmentally programmed genome rearrangements at the mechanistic, genome-wide and evolutionary levels. These aspects have so far mostly been addressed in three species: P. tetraurelia and Tetrahymena thermophila on the one hand, the spirotrichous ciliate O. trifallax on the other.

One new piece of information that can be found in the present manuscript is the assembly and annotation of the germline genome of two novel species: Tetmemena sp, closely related to Oxytricha, and the more distant E. woodruffi. Feng et al establish that, similar to other ciliates, Tetmemena and Euplotes eliminate TEs and other germline-specific sequences during programmed genome rearrangements. They also undergo extensive gene unscrambling, which results in IES removal and MDS reordering to assemble coding sequences.

A TE origin was discussed previously for Paramecium (Arnaiz et al PLoS Genet; Sellis et al 2021 PLoS Biol) and Tetrahymena IESs (Hamilton et al 2016 eLife). While this may also hold true in spirotrichous ciliatesThe present manuscript proposes a completely new evolutionary scenario for IESs from scrambled genes. Here, Feng et al establish that scrambled genes of spirotrichous ciliates tend to be associated with local paralogy. They provide evidence supporting that IESs from scrambled genes may have evolved from paralogous MDSs.

Although I am more an expert in the molecular mechanisms involved in genome rearrangements, I feel that the work reported here should draw the attention of a broader audience interested in genome dynamics and evolution, beyond the specific field of spirotrichous ciliate biology.

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

Feng et al. provide a solid analysis of the evolution of genome rearrangement in spirotrich ciliates. The authors applied a variety of state-of-the-art sequencing and bioinformatic methods to investigate the intriguing and extremely complex patterns of genome architecture in this protist lineage. Methods (including statistical analyses) are adequate and explained in detail. Results and discussions reflect careful, clever analysis of the data and excellent linkage with the literature. Figures and tables complement the text in a compelling way. I have only minor suggestions:

Summary: more gradually introduce Spirotrichea and the phylogenetic relationship among the three species analyzed. This would better position the reader to understand the evolutionary context you are working in. Also, it would be helpful to more clearly differentiate novel vs. existing data. A suggestion: "This study focuses on three spirotrich species: two in the family Oxytrichidae (Oxytricha trifallax and Tetmemena sp) and Euplotes woodruffi as an outgroup. To complement existing data, we sequenced, assembled and annotated the germiline and somatic genomes of E. woodruffi and the germline genome of Tetmemena sp."

Thank you, we clarified the summary (abstract).

Introduction, first paragraph: Replace "The species in this study..." for a more precise statement, such as "The three spirotrich species studied here..."

Thank you, we have made this statement more precise.

p. 4: This sentence is unclear: "These useful tools provide partial insight to guide selection of species for full genome sequencing, which allows construction of complete rearrangement maps of a MIC genome onto a MAC genome for a reference species."

Thank you, we have clarified this sentence.

p. 8: define TE on first mention.

Defined on page 6.

Table 1. Indicate which MIC and MAC data are from this study.

References are included for published data and a note has been added to indicate data from this study.

Reviewer #3 (Significance (Required)):

The present work represents a significant advance in the field of evolutionary genomics. The focus of the paper is on ciliates, an ancient (2 billion-year old) and highly diverse eukaryotic phylum that presents many peculiarities, including sex, nuclear dimorphism, genome rearrangement, high numbers of paralogs and transposons, etc. While some data exist on a few model ciliates of disparate phylogenetic position, this work focuses on two species taxonomically placed in the same family, plus a more distant outgroup within the same class. This gives a novel dimension to this study, that goes beyond exploring genome architecture in a single clade. Instead, it allows to explore evolutionary trends in genome rearrangement among relatively closely related species. This paper should be of high interest not only for ciliate biologists (like me), but also in relation to comparative genomics of protists/eukaryotes and germ-soma biology. I highly recommend publication.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Feng et al. provide a solid analysis of the evolution of genome rearrangement in spirotrich ciliates. The authors applied a variety of state-of-the-art sequencing and bioinformatic methods to investigate the intriguing and extremely complex patterns of genome architecture in this protist lineage. Methods (including statistical analyses) are adequate and explained in detail. Results and discussions reflect careful, clever analysis of the data and excellent linkage with the literature. Figures and tables complement the text in a compelling way. I have only minor suggestions:

• Summary: more gradually introduce Spirotrichea and the phylogenetic relationship among the three species analyzed. This would better position the reader to understand the evolutionary context you are working in. Also, it would be helpful to more clearly differentiate novel vs. existing data. A suggestion: "This study focuses on three spirotrich species: two in the family Oxytrichidae (Oxytricha trifallax and Tetmemena sp) and Euplotes woodruffi as an outgroup. To complement existing data, we sequenced, assembled and annotated the germiline and somatic genomes of E. woodruffi and the germline genome of Tetmemena sp."

• Introduction, first paragraph: Replace "The species in this study..." for a more precise statement, such as "The three spirotrich species studied here..."

• p. 4: This sentence is unclear: "These useful tools provide partial insight to guide selection of species for full genome sequencing, which allows construction of complete rearrangement maps of a MIC genome onto a MAC genome for a reference species."

• p. 8: define TE on first mention.

• Table 1. Indicate which MIC and MAC data are from this study.

Significance

The present work represents a significant advance in the field of evolutionary genomics. The focus of the paper is on ciliates, an ancient (2 billion-year old) and highly diverse eukaryotic phylum that presents many peculiarities, including sex, nuclear dimorphism, genome rearrangement, high numbers of paralogs and transposons, etc. While some data exist on a few model ciliates of disparate phylogenetic position, this work focuses on two species taxonomically placed in the same family, plus a more distant outgroup within the same class. This gives a novel dimension to this study, that goes beyond exploring genome architecture in a single clade. Instead, it allows to explore evolutionary trends in genome rearrangement among relatively closely related species. This paper should be of high interest not only for ciliate biologists (like me), but also in relation to comparative genomics of protists/eukaryotes and germ-soma biology. I highly recommend publication.

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

Learn more at Review Commons

---

Referee #2

This reviewer did not leave any comments

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Ciliates extensively rearrange their somatic genome every time a new somatic nucleus develops from the zygotic germline nucleus. In this manuscript, Feng et al report the sequencing, assembly and annotation of the germline and somatic genomes of Euplotes woodruffi and the germline genome of Tetmemena sp. (whose somatic genome was sequenced and assembled by the same lab in 2015). They present a comparative analysis of developmentally programmed genome rearrangements in these two species and in the model ciliate Oxytricha trifallax. Their major findings are that:

(1) E. woodruffi and Tetmemena sp. eliminate a smaller fraction of their germline genome (~54%) from their somatic macronucleus (MAC) than O. trifallax (>80%)

(2) Transposable elements (TE) represent a smaller fraction of the germline genome (~2%) in the first two ciliates than in O. trifallax (~15%). TEs are mainly located at the boundaries of germline chromosomes and in intergenic regions, but can also be found inside IESs

(3) Several thousands of genes are scrambled in the germline genome of all three species

The authors have also addressed the possible origin of gene scrambling. They report an interesting association with local paralogy and propose a model for the emergence of the odd-even pattern of gene unscrambling between two paralogous copies.

Major comments:

(1) Based on the statistics presented in Table 1, genome assemblies are of good quality, with a reasonable N50 size of germline (MIC) contigs. It seems, however, that no entire MIC chromosome could be assembled, since no two-telomere contig is mentioned in the list. As proposed by the authors (p.7) the presence of numerous TEs at the boundaries of MIC contigs (Fig S1) may have hindered the assembly of MIC chromosome ends. I would have appreciated to have more information on the "other repeats" (which seem to differ from tandem repeats according to Fig 2) and their location along MIC contigs.

(2) The definition of "Internal Eliminated Sequences" (IES) is not clear. The authors make a distinction between IESs and TEs. I understand that IESs are DNA segments that separate two macronuclear-destined sequences (MDS) in the germline genome. Thus they appear to be restricted to those regions that eventually yield gene-sized MAC chromosomes. IESs are eliminated between two pointers that may not be identical on both sides in case of scrambled genes. Some clarification is needed here.

To illustrate my point: I found the statement "with many TE insertions within IESs, suggesting that TE insertions may have generated IESs" particularly confusing (p. 9 lines 5-6). Does this mean that IESs extend beyond the ends of inserted TEs? The legend of Fig S1 should also be clarified.

(3) p. 10 lines 2-4 and Fig S2: Could the authors explain the difference they make between MDS (in the text) and CDS (in Fig S2)? My understanding is that a CDS is the entire gene coding sequence and may be made of multiple MDSs. If this is correct, the sentence should read "We compared the number of MDSs between single-copy orthologs for single-gene MAC chromosomes across the three species and found that the orthologs have similar CDS lengths".

(4) p. 12 lines 10-15: the discovery that paralogous MDSs can be found in scrambled genomic loci is interesting. If the two paralogs can be distinguished based on the number of substitutions, it would be informative to go back to individual reads and check whether each of the two copies can be incorporated in the unscrambled CDS (and at which frequency). Would the pointers be compatible with this?

(5) The hypothesis that odd-even scrambled loci have evolved from paralogous genes in E. woodruffi is supported by the existence of paralogous MDSs, length conservation of MDS/IES pairs and sequence similarity between corresponding MDS and IES in a pair. The correlations presented for Oxytricha and Tetmemena are much less convincing (Fig S5D and E). I recommend that the authors are even more cautious in their statement on p.13 ("For Oxytricha and Tememena, the MDS and IES lengths for such MDS/IES pairs also correlate positively, but more moderately").

(6) p. 15 last paragraph: Why did the authors focus only on TBEs inserted in non-scrambled IESs to look for orthologous TBE insertions? Is there a reason to believe that no recent TBE insertion occurred at other genomic loci? Or was it only for practical reasons? It is also not clear to me whether the authors have considered full-length TBEs or the presence of at least one TBE ORF.

(7) p. 16: the authors report that some introns of E. woodruffi map "near" Oxytricha/Tetmemena pointers. How near? Based on the information provided by the authors, I don't think this observation necessarily implies that IESs were converted to introns (or reciprocally) during evolution. If this were true, shouldn't at least one intron boundary coincide exactly with a pointer? The authors should clarify this (also in the discussion, on p. 20, top paragraph).

(8) p. 19 2nd paragraph: the suggested mechanism explaining the 5' bias of IESs in E. woodruffi genes is unclear. How could germline recombination take place between a MIC chromosome and a MAC reverse transcript or nanochromosome? This would imply that DNA could be imported in the MIC. Is there evidence that this might occur?

(9) According to Figure 1, no scrambled genes have been reported in Paramecium tetraurelia. Within the frame of the proposed model, this is somewhat unexpected because this ciliate went through several whole genome duplications during evolution and harbors many paralogous gene pairs. Is there a reason why no gene scrambling took place in Paramecium?

Minor comments:

• p. 4 (4th bottom line): To my knowledge, ref #28 presents a draft (incomplete) MIC assembly of the Paramecium genome.

• p. 7 (last paragraph): "encoding" should be replaced by "carrying"

• p. 10 (2nd paragraph): insert a missing "o" into "nanochromosomes"

• p. 10 (same paragraph): the weak 5' bias of IES distribution in Tetmemena should be shown (either as an additional panel in Fig 3 or in a Sup Figure.

• p. 24 2nd paragraph: "a" is missing in "Trinity, which is a software..."

CROSS-CONSULTATION COMMENTS

I agree with most comments of reviewer 3.

The authors have actually defined "TE" in the introduction (p. 6). Depending on the journal's rules for abbreviation use, it may not be necessary to define it again in the results section

Significance

• Ciliates are unicellular models to study developmentally programmed genome rearrangements at the mechanistic, genome-wide and evolutionary levels. These aspects have so far mostly been addressed in three species: P. tetraurelia and Tetrahymena thermophila on the one hand, the spirotrichous ciliate O. trifallax on the other.

• One new piece of information that can be found in the present manuscript is the assembly and annotation of the germline genome of two novel species: Tetmemena sp, closely related to Oxytricha, and the more distant E. woodruffi. Feng et al establish that, similar to other ciliates, Tetmemena and Euplotes eliminate TEs and other germline-specific sequences during programmed genome rearrangements. They also undergo extensive gene unscrambling, which results in IES removal and MDS reordering to assemble coding sequences.

• A TE origin was discussed previously for Paramecium (Arnaiz et al PLoS Genet; Sellis et al 2021 PLoS Biol) and Tetrahymena IESs (Hamilton et al 2016 eLife). While this may also hold true in spirotrichous ciliatesThe present manuscript proposes a completely new evolutionary scenario for IESs from scrambled genes. Here, Feng et al establish that scrambled genes of spirotrichous ciliates tend to be associated with local paralogy. They provide evidence supporting that IESs from scrambled genes may have evolved from paralogous MDSs.

• Although I am more an expert in the molecular mechanisms involved in genome rearrangements, I feel that the work reported here should draw the attention of a broader audience interested in genome dynamics and evolution, beyond the specific field of spirotrichous ciliate biology.

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

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2022-01541

Corresponding author(s): Hubert Hilbi

1. General Statements

Upon infection of eukaryotic host cells, Legionella pneumophila forms a unique compartment, the Legionella-containing vacuole (LCV). While the role of vesicle trafficking pathways for LCV formation has been quite extensively studied, the role of putative membrane contact sites (MCS) between the LCV and the ER has been barely addressed. In our study, we provide a comprehensive analysis of the localization and function of protein and lipid components of LCV-ER MCS in the genetically tractable amoeba Dictyostelium discoideum.

We would like to thank the 3 reviewers for their thorough and constructive reviews. Overall, the reviewers state that the study is of interest to researchers in the field of Legionella and other intracellular pathogens (Reviewer 2), as well as to cell biologists (Reviewer 3). Reviewer 1 does not ask for additional experiments but is critical about the overall structure of the manuscript and the proteomics approach. As requested by the reviewer, we have substantially restructured the revised manuscript, now clearly outline the hypotheses put forward in the study and streamlined the proteomics data. Reviewer 2 asks for additional experiments to support our model of LCV-ER MCS. In the revised manuscript, we have included additional experiments addressing lipid exchange at the MCS, and we plan to perform further co-localization experiments. Reviewer 3 appreciates the comprehensive LCV proteomics and asks for only minor revisions, which we have incorporated in the revised version of the manuscript. We include below a point-by-point response to all the comments made by the reviewers.

2. Description of the planned revisions

Reviewer #2

Major comment

1) MCS contain protein complexes or a group of proteins, but the proteins here are studied in isolation and do not support the model shown in Figure 7. Co-localization studies of the putative LCV-ER MCS proteins are critical, especially given that the authors hypothesize the proteins are working together to modulate PI(4)P levels.

Response: As suggested by the reviewer, we will perform additional co-localization experiments with MCS components. To this end, we will construct mCherry-Vap, and we will co-transfect the parental D. discoideum strain Ax3 with plasmids producing mCherry-Vap and OSBP8-GFP or GFP-OSBP11. Using these dually fluorescence labelled D. discoideum strains, the co-localization of Vap with the OSBPs will be assessed at 1, 2, and 8 h post infection. The data will be presented as fluorescence micrographs, and co-localization of Vap with the OSBPs will be quantified using Pearson’s correlation coefficient and fluorescence intensity profiles. The data will be outlined in the text (l. 258 ff.) and shown in the new Fig. 2 and__ Fig. S4__.

3. Description of the revisions that have already been incorporated in the transferred manuscript

Reviewer #1 (Evidence, reproducibility and clarity):

In the manuscript by Vormittag, et al., the authors perform proteomics identification of proteins associated with the Legionella-containing vacuole (LCV) in the model amoeba Dictyostelium discoideum comparing WT to atlastin knockout mutants. The authors find approximately half the D. discoideum proteome associated with the LCV, but there was enrichment of some proteins on the WT relative to the mutant. They focus on proteins involved in forming membrane contact sites (MCS) that previously were shown to be important for expansion of the Chlamydia-containing vacuole. Most significant are the oxysterol binding proteins (OSBP) and VapA (similar to that seen in Chlamydia). The authors show differential association of these proteins with either the LCV or presumably the ER associated with the LCV. Using a linear scale over 8 days, they show that mutations in some of the MCS reduce yields in two of the OSPB knockout mutants and the growth rate of the vap mutant is slowed but ultimate yield is increased. Using some nice microscopy techniques, they measure LCV size, and the osbK mutant appears particular small relative to other strains, whereas the osbH mutant generates large vacuoles. This doesn't necessarily correlate with the PI4P quantities on the vacuoles (which is higher in all of them), but I am not totally sure how this is measured, and whether is it PI4P/pixel or PI4P/LCV. In all cases, this was reduced by Sac1 mutation. Surprisingly, even though there was uniform increase in PI4P in each of the mutants, loss of PI4P only affects localization of some of the proteins. Finally, in what seems to be a peripherally related experiment, the authors show that a pair of Legionella translocated effectors are required to maintain PI4P levels, although it is not clear how this is related to the other data in the manuscript.

It is not clear from the manuscript if the authors are just cataloging things or trying to test a hypothesis. This is an extremely difficult manuscript to read and reconstruct what the authors showed. I really think that the only people who will understand what is written are people who are familiar with the work in Chlamydia starting in 2011 in Engel's and Derre's laboratories, which clearly showed that MCS and most specifically Vap/OSBPs are involved in vacuole expansion. If the authors could rewrite the manuscript along these lines, perhaps comparing their data to the Chlamydia data it would help a lot. Otherwise, I don't think anyone else will understand why they are focusing on these things. I don't recommend new experiments (although re-analyzing data is necessary), but the manuscript has to be taken apart and claims removed, and data be interpreted properly. Otherwise, the manuscript seems like just a clearing house for data.

Response: Thank you for the concise summary of our data and pointing out the need to restructure the manuscript and to clearly outline the hypotheses underlying the study. According to the reviewer’s suggestions, we have now re-structured the manuscript. In the revised manuscript the story unfolds from the observation that the ER tightly associates with (isolated) LCVs, and the proteomics approach is used as a validation of the presence of MCS proteins at the LCV-ER MCS.

As suggested by the reviewer, we now highlight the seminal work on Chlamydia by the Engel and Derré laboratories not in the Discussion section (as in the original version of the manuscript) but already in the Introduction section (l. 142-148). We believe that it makes a stronger case to start out an analysis of LCV-ER MCS with a Legionella-specific cell biological finding (LCV-ER association) and an unbiased proteomics approach, as compared to a more derivative and defensive approach starting out with what is known about Chlamydia.

The reviewer’s comment “This is an extremely difficult manuscript to read” appears overly harsh and conflicts with the positive evaluation of Reviewer #2 and Reviewer #3. Finally, we respectfully disagree with the reviewer’s statement that experiments characterizing L. pneumophila effectors implicated in the formation and function of LCV-ER MCS are peripheral. These experiments significantly contribute to a mechanistic understanding of how L. pneumophila forms and exploits LCV-ER MCS, and they are central for studies on pathogen-host interactions. The studies are analogous to the work on Chlamydia effectors by the Engel and Derré laboratories, but the mode of action of Legionella and Chlamydia effectors is obviously different. Another important distinction of our work to the studies on Chlamydia is the use of the genetically tractable amoeba, D. discoideum, which allows an analysis of LCV-ER MCS by fluorescence microscopy at high spatial resolution.

Specific comments

1. The problems start with the first figure, in which the authors state that almost half the D. discoideum proteome is LCV-associated. I doubt that this is correct, and they should base this on some selective criterion. Furthermore in Fig. 1A, they show Venn diagrams for how they whittled this down, but the Supplemental Dataset gives us no clue on how this was done. I can only sit down myself with the dataset and try to figure that out, but that is an unreasonable expectation for the reader. The dataset provided should have a series of sheets, describing how the large protein set was whittled down and how they were sorted, so the reader can evaluate how robust the final results were. To me (at least), if they said: "look we got this surprising result that suggests MCS are involved in promoting LCV formation, and although this is well recognized in Chlamydia but poorly recognized in Legionella", that would be satisfactory to me.

Response: According to the reviewer’s suggestions, we have now thoroughly re-structured the manuscript. In the revised manuscript the story unfolds from the observation that the ER tightly associates with LCVs in infected cells and with isolated LCVs. The proteomics approach is now used as a validation of the presence of MCS proteins at the LCV-ER MCS and relegated to the Supplementary Information section (former Fig. 1, now Fig. S3).

For the proteomics analysis, all protein identifications have been filtered for robustness applying a constant FDR (false discovery rates) of protein and PSM (peptide spectrum match) of 0.01, which is a commonly accepted threshold in the field. Moreover, two identified unique peptides were required for protein identification. The parallel application of both filter criteria results in very robust and reliable data sets. This is outlined in the Material and Methods section (l. 683-693).

In the data set of LCV-associated proteins, 2,434 D. discoideum proteins have been identified (Table S1). This is 18.5% of the total of 13,126 predicted D. discoideum proteins (UniprotKB) and considerably less than “almost half the D. discoideum proteome”, as stated by the reviewer. Moreover, 1,224 L. pneumophila proteins have been identified (among 3,024 predicted L. pneumophila proteins in the database). This is a reasonable number of proteins identified from an intracellular vacuolar pathogen, given the LCV isolation and proteomics methods applied. We now outline these findings more extensively in the Results section (l. 207-213). Moreover, to render Table S1 more reader-friendly, we added to the datasheet “All data” the datasheets “Dictyostelium”, “Legionella” and “Info”.

The Venn diagram in Fig. S3A (previously Fig. 1A) does not show a subset of proteins “whittled down” from the entire proteomes, but simply summarizes LCV-associated proteins, which were either identified exclusively in the parental strain Ax3 but not in the Δsey1 mutant strain, or only in Δsey1 but not in Ax3, thus identifying possible candidates relevant for the LCV-ER MCS. This information is now outlined more clearly in the text (l. 238-241). Moreover, we now explicitly define in the Material and Methods section (l. 697-704) the “on” and “off” proteins shown in Fig. S3A.

The overall rational for the comparative proteomics approach was our previous finding that compared to the D. discoideum parental strain Ax3, the Δsey1 mutant strain accumulates less ER around LCVs (PMID: 28835546, 33583106). This finding suggests that formation of the LCV-ER MCS might be compromised in the Δsey1 mutant strain. This hypothesis is now outlined at the beginning of the Results paragraph (l. 204-207).

I am clueless regarding how Fig. 6 fits with the rest of the manuscript. If this is about MCS, there is no demonstration these effectors are directly involved in MCS other than the somewhat diffuse argument that there is some correlative connection to PI4P levels, that I am not particularly convinced by.

Response: The PtdIns(4)P gradient between two different cellular membranes is an intrinsic feature of MCS. To date, a quantification of PtdIns(4)P levels on LCVs in response to the presence or absence of specific L. pneumophila effectors is lacking. Accordingly, we opted for quantifying the PtdIns(4)P levels on LCVs in presence and absence of an L. pneumophila effector putatively generating PtdIns(4)P on LCVs, the phosphoinositide 4-kinase LepB, or titrating PtdIns(4)P on LCVs, the PtdIns(4)P-binding ubiquitin ligase SidC. To address the concerns of Reviewer 1 and Reviewer 3 (see below), we now outline in detail the rational to assess the role of LepB and SidC for MCS function (l. 385-387). Importantly, we now also provide data that at LCV-ER MCS PtdIns(4)P/cholesterol lipid exchange is functionally important (new Fig. 6 and Fig. S10). In the revised version of the manuscript, this new data is preceding the experiments with the L. pneumophila effectors, which should render our choice of effectors more comprehensible to the reader and increase the flow of the manuscript.

Line 146 and associated paragraph. We don't need a catalog of proteins in narrative. There is more detail in the narrative than there is in the tables and figures, which would be a more appropriate way to present the data.

Response: As suggested by the reviewer, we summarized the LCV-associated D. discoideum proteins and considerably reduced the list in the text (l. 214-230).

Line 186. There is nothing wrong with pursuing MCS based on the idea that this was seen before with Chlamydia and you wanted to test if this was a previously unappreciated aspect of Legionella biology. I don't see the rationale based on the proteomics, partly because I don't understand how the proteomics dataset was parsed.

Response: As suggested by the reviewer, we thoroughly re-structured the manuscript and now highlight the seminal work on Chlamydia by the Engel and Derré laboratories already in the Introduction section (not in the Discussion section as in the original version of the manuscript). We believe that it makes a stronger case to start out an analysis of LCV-ER MCS with a Legionella-specific cell biological finding (LCV-ER association) and an unbiased proteomics approach, as compared to a more derivative and defensive approach starting out with what is known about Chlamydia.

Figure 3: These growth curves are super-weird. I am not used to looking at 8 days of logarithmic growth in a linear scale and seeing no (apparent) growth for 4 days. Considering all the microscopy data are performed in the first 18 hrs of infection, it’s hard to see how this is related to data at 8 days post infection. If this were plotted in logarithmic scale, as microbiologists are used to doing, then perhaps we could see a connection. Also, in some cases, it might be helpful to calculate a growth rate, because it’s possible the author may now see some effects by comparing logarithmic growth rates.

Response: We have been characterizing growth of L. pneumophila in D. discoideum in several studies using growth curves with RFU vs. time plotted in linear scale (e.g., Finsel et al., 2013, Cell Host Microbe 14:38; Rothmeier et al., 2013, PloS Pathog 9: e1003598; Swart et al., 2020, mBio 11: e00405-20). The D. discoideum-L. pneumophila infection model is peculiar, since the amoebae do not survive temperatures beyond 26 degC. This is substantially below the optimal growth temperature of L. pneumophila (35-40 degC). This means that - due to the many genetic tools available - D. discoideum is an excellent model to investigate cell biological aspects of the infection at early time points (ca. 1-18 h p.i.), but the amoebae are not an optimal system to quantify (several rounds) of intracellular growth.

Figure 2: The images don't necessarily show what the bar graphs show. In particular, look at Osp8. That image doesn't make sense to me.

Response: The individual channels of the merged images in Fig. 1 (formerly Fig. 2) are shown in Fig. S2. By looking at the individual channels, it becomes clear that OSBP8-GFP co-localizes with calnexin-mCherry (overlapping signals), but not with P4C-mCherry or AmtA-mCherry (adjacent signals). Co-localization was quantified in a non-biased manner by Pearson’s correlation coefficient. To further visualize co-localization, we now also provide fluorescence intensity profiles for all confocal micrographs (amended Fig. 1).

In summary, I think the authors hit on something that is probably important for Legionella biology, but it’s not clear what they want to show. They are very invested in connecting everything to PI4P levels, which may or may not be correct, but it seems to me that perhaps taking more care in showing the importance of the Vap/OSPB nexus in supporting Legionella growth should be the first priority.

Response: Given the importance of the PtdIns(4)P gradient for lipid exchange at MCS, we believe it is justified to put considerable emphasis on this lipid. To further substantiate a functional role of PtdIns(4)P at LCV-ER MCS, we now also show that an increase in PtdIns(4)P at the LCV correlates with a decrease of cholesterol (new Fig. 6 and Fig. S10). The inverse correlation of these two lipids is in agreement with the notion that cholesterol is a counter lipid of PtdIns(4)P at LCV-ER MCS.

It is not clear from the manuscript if the authors are just cataloging things or trying to test a hypothesis.

Response: In the revised version of the manuscript, we put forward several specific hypotheses, which we then tested in our study (l. 152-155).

If I understand Fig. 1, only one of the candidates (VapA) was verified as being more enriched in WT relative to atlastin mutants. This argues even more strongly that the authors have to describe their criteria for choosing these candidates.

Response: As outlined above (specific point 1), we have now re-structured the manuscript according to the reviewer’s suggestions. In the revised manuscript the story unfolds from the observation that the ER tightly associates with LCVs in infected cells and with isolated LCVs. The proteomics approach is now used as a validation of the presence of MCS proteins at the LCV-ER MCS and relegated to the Supplementary Information section (formerly Fig. 1, now Fig. S3). We consider the proteomics approach a powerful hypothesis generator, and the experimental identification of several MCS proteins by proteomics validated the cell biological and bioinformatics insights.

Reviewer #1 (Significance (Required)):

As stated above, the manuscript can't decide if it’s about MCS or PI4P, and I would argue strongly that the emphasis on PI4P detracts from the manuscript, as well as its inability to draw connection to previous work that is likely to be important.

Response: We respectfully disagree with the reviewer on this important point and hold that proteins as well as lipids are crucial functional determinants of MCS. The PtdIns(4)P gradient is a pivotal process for lipid exchange at MCS. Therefore, we believe it is justified to put considerable emphasis on this lipid. In the Introduction section, we now specify several hypotheses on the localization and function of lipids and proteins at LCV-ER MCS (l. 152-155). Moreover, we now also refer to the previous work on Chlamydia MCS in the Introduction section (l. 142-148).

Reviewer #2 (Evidence, reproducibility and clarity):

Summary of paper and major findings

Membrane contact sites (MCS) are locations where two membranes are in close proximity (10-80nm). MCS have a defined protein composition which tether the membranes together and function in small molecule and lipid exchange. Typically, MCS proteins contain structural (e.g., tethers) and functional (e.g., exchange lipids) proteins, in addition to proteins which regulate the structure and function of the MCS. In this manuscript, Vormittag et al describe protein components of MCS between the Legionella-containing vacuole (LCV) and the host endoplasmic reticulum (ER) in the amoeba Dictyostelium. Proteomics of isolated LCVs followed by microscopy analysis identified several proteins which localize to either the LCV-associated ER (OSBP8), the LCV (OSBP11), or both (VAP and Sac1). The mammalian homologs of these proteins have been shown to play important roles in ER MCS, with VAP serving a structural role, Sac1 a PI(4P) phosphatase regulating PI(4)P levels, and OSBP8 and OSBP11 lipid transferring proteins. Given the importance of PI(4)P in formation and maintenance of the Legionella-containing vacuole, the authors used dicty mutants to determine the importance of these proteins in bacterial growth, LCV size, and PI(4)P levels on the LCV. While VAP and OSBP11 appear to promote Legionella infection, OSBP8 appears to restriction infection, although all identified MCS components appear to play a role in decreasing PI(4P) shortly after infection. Finally, VAP and OSBP8 localization to the LCV is PI(4)P-dependent. Overall, the authors conclude that these MCS components play a role in modulating PI(4)P levels on the LCV.

Overall, this is an interesting study further exploring the role of PI(4)P in LCV-ER interactions, and how PI(4)P levels are regulated. The figures are clearly presented, there is an impressive amount of data, and rigor appears to be strong with appropriate replicates and statistical analysis. The phenotypes are often mild, but the authors are careful to not overinterpret the data. While this is an interesting study, additional experiments are necessary to support the overall model and the text needs to put the findings into the larger context.

Response: We would like to thank the reviewer for this positive and constructive assessment. We performed and planned additional experiments to further strengthen the study and support our model.

Major comments

1) MCS contain protein complexes or a group of proteins, but the proteins here are studied in isolation and do not support the model shown in Figure 7. Co-localization studies of the putative LCV-ER MCS proteins are critical, especially given that the authors hypothesize the proteins are working together to modulate PI(4)P levels.

Response: To further explore the possible interactions between Vap and OSBP proteins, we plan co-localization experiments using D. discoideum strains producing mCherry-Vap and either OSBP8-GFP or GFP-OSBP11, as outlined above (Section 2, new__ Fig. 2__ and Fig. S4).

Moreover, we included additional data on PtdIns(4)P/cholesterol lipid exchange (Fig. 6 __and Fig. S10__), which have been incorporated into the model (amended Fig. 8). Based on the available data, we do not postulate direct interactions between Vap and OSBP proteins. The previous model, which now has been amended, might have been misleading in that respect.

2) The phenotypes are relatively mild, suggesting functional redundancy. Double knockouts, particularly in VAP and OSBP11, may generate a stronger phenotype that better supports the hypothesis and demonstrate the importance during infection.

Response: Thank you for this interesting suggestion. Please see Section 4 below for our arguments, why we believe that this intriguing approach is beyond the scope of the current study.

3) The timing of PI(4)P and MCS protein localization during infection is critical to understanding how MCS might be functioning. Based on Figure 6C, PI(4)P levels decrease on the LCV during infection, but this is not fully explained in the context of what's known in the literature and what is observed the previous figures. How does localization of different MCS components change during infection, and does this correlate with the changes in growth or LCV size? A better description in the Introduction on LCV-associated PI(4)P levels would be beneficial in orienting the reader to why PI(4)P levels are modulated.

Response: As suggested by the reviewer, we added to the Introduction section more detail about the kinetics of PtdIns(4)P accumulation on LCVs (l. 65-71), and we discuss the limited spatial resolution of the IFC approach (formerly Fig. 6C, now Fig. 7C; l. 407-408). Importantly, we also provide new data showing that within 2 h p.i. an increase in PtdIns(4)P at the LCV coincides with a decrease of cholesterol (new Fig. 6 and Fig. S10). The new data is put into this context in the Discussion section (l. 449-454).

4) OSW-1 has other targets besides OSBPs, and depleting Sac1 and Arf1 in A549 cells is not specifically targeting the MCS, as these proteins have other functions. The data in mammalian cells is not convincing and should be removed.

Response: As suggested by the reviewer, we removed the data on depleting Sac1 in A549 cells (Fig. 3D, and Fig. S6BC). We propose to leave the pharmacological data on inhibition of L. pneumophila replication by OSW-1 in the manuscript, but to clearly point out that OSW-1 has other targets besides OSBPs (l. 297-299).

Minor comments

1) Figure 2 is missing details on number of experiments/replicates and statistical analysis.

Response: Thank you for having noted this oversight. The number of independent experiments and statistical analysis have now been added to Fig. 1 (formerly Fig. 2) (l. 1009-1010).

2) Can the authors hypothesize why VAP promotes growth early during infection, but appears to restrict growth at later timepoints (Figure 3A)?

Response: Thank you for raising this intriguing point. The opposite effects on growth of Vap at early and later timepoints during infection might be explained by interactions with antagonistic OSBPs. Vap likely co-localizes with OSBP8 as well as with OSBP11 on the limiting LCV membrane or the ER, respectively (experiment to be performed; Fig. 2 and__ Fig. S4__). The absence of OSBP8 (ΔosbH) or OSBP11 (ΔosbK) causes larger or smaller LCVs, and increased or reduced intracellular replication of L. pneumophila, respectively. Thus, OSBP8 seems to restrict and OSBP 11 seems to promote intracellular replication. Accordingly, if Vap affects or interacts with OSBP11 early and with OSBP8 later during infection, opposite effects on growth of Vap might be explained. These reflections are now outlined in the Discussion section (l. 431-441).

3) There is a large amount of data, which makes it difficult at times to follow. I suggest adding additional information to table 1, including LCV size and whether or not the protein's localization is PI(4)P-dependent.

Response: Thank you for this suggestion. As proposed by the reviewer, we added the additional information to Table 1 (PtdIns(4)P-dependency of protein localization, LCV size).

Reviewer #2 (Significance (Required)):

Membrane contact sites during bacterial infection are a growing area of research. In Legionella, several papers point to the presence of MCS. Further, PI(4)P is known to be an important component on the LCV. This paper shows that MCS protein members are important in modulating LCV PI(4)P levels. The model as presented is not completely supported by the data as co-localization experiments are needed, along with more detailed analysis of how PI(4)P levels change over infection and the role of these MCS proteins in that process. This study will be of interest to those studying Legionella and other vacuolar pathogens. Area of expertise is on membrane contact sites and lipid biology.

Response: Thank you very much for the overall positive and constructive evaluation.

Reviewer #3 (Evidence, reproducibility and clarity):

The authors perform proteomic analysis of Legionella-containing vacuoles. The observe association of membrane contact site (MCS) proteins including VAP, OSBPs, and Sac1. Functional data indicates that these proteins contribute to PI4P levels on LCVs and their ability to acquire lipid from the ER to enable LCV expansion/stability. Overall, the paper is an important contribution to the field and builds upon a growing appreciation for MCS in establishment of intracellular niches by microbial pathogens. I have only minor comments for the authors consideration.

Response: We would like to thank the reviewer for this enthusiastic assessment.

Minor comments:

-line 145, "This approach revealed 3658 host or bacterial proteins identified on LCVs...". This number seems high... how does it compare to prior proteomic studies of pathogen-containing vacuoles?

Response: As outlined above (reviewer 1, point 1), we have now changed the text (l. 207-213): “This approach revealed 2,434 LCV-associated D. discoideum proteins (Table S1), of a total of 13,126 predicted D. discoideum proteins (UniprotKB). Moreover, 1,224 L. pneumophila proteins were identified (among 3,024 predicted L. pneumophila proteins), which is a reasonable number of proteins identified from an intracellular bacterial pathogen within its vacuole with the proteomics methods applied (Herweg et al, 2015; Schmölders et al., 2017).”

• line 160. Can the authors comment on why mitochondrial proteins are observed in their proteomic analysis? Are these non-specific background signals or reflecting relevant organelle contact?

Response: The dynamics of mitochondrial interactions with LCVs and the effects of L. pneumophila infection on mitochondrial functions have been thoroughly analyzed (PMID: 28867389). This seminal work is now cited in the text (l. 227-230).

• line 268. It is reported that LCVs are smaller with MCS disruption at 2 and 8 h p,i.. Does this also lead to instability or rupture of LCVs? And related to this why would LCVs be bigger at 16h with MCS disruption?

Response: MCS components affect LCV size positively or negatively. E.g., the absence of OSBP8 (ΔosbH) or OSBP11 (ΔosbK) causes larger or smaller LCVs, and increased or reduced intracellular replication of L. pneumophila, respectively. However, as outlined in the Discussion section (l. 442-454), we believe that the relatively small size likely reflects a structural remodeling of the pathogen vacuole rather than a substantial LCV expansion. LCV rupture takes place only very late in the infection cycle (beyond 48 h) and is followed by lysis of the host amoeba (PMID: 34314090).

• lines 288 and 299 "data not shown" this data should be included in a supplemental figure.

Response: The data on the localization of GFP-Sac1 and GFP-Sac1_ΔTMD are included in the Figs. 1A, 4A, 5AD, S2A, S7A, and__ S9__ (l. 328, l. 339).

• line 327. The authors choose to focus on the role of LepB and SidC in MCS modulation. The rationale for choosing these two amongst the ca 330 effectors was not given. Were other effectors also examined?

Response: LepB and SidC were chosen due to their activities producing or titrating PtdIns(4)P, respectively, and their LCV localization. This rational is now given in the text (l. 385-387). No other effectors were examined up to this point.

Reviewer #3 (Significance (Required)):

Comprehensive LCV proteomics of interest to field of cellular microbiology. Studies of MCS broadly relevant to cell biologists.

Response: Thank you very much for the overall very positive evaluation.

4. Description of analyses that authors prefer not to carry out

Reviewer #2

Major comment

2) The phenotypes are relatively mild, suggesting functional redundancy. Double knockouts, particularly in VAP and OSBP11, may generate a stronger phenotype that better supports the hypothesis and demonstrate the importance during infection.

Response: Thank you for raising the important question of functional redundancy. We now outline this concept in the Discussion section (l. 427-429). A further analysis of the genetic and biochemical relationship between Vap and OSBP11 or OSBP8 are without doubt some of the most interesting aspects of further studies on the topic of LCV-ER MCS.

The construction of a D. discoideum double mutant strain is time consuming and usually takes 1-2 months. Provided that a Vap/OSBP11 double deletion mutant strain is viable and can be generated, it takes another 1-2 months to thoroughly characterize the strain regarding intracellular replication of L. pneumophila (Fig. 3), LCV size (Fig. 4), and PtdIns(4)P score (Fig. 5). Moreover, there is already a large amount of data in the paper (to quote Reviewer #2), and therefore, adding new data might makes it even harder to follow the story and focus on the key points. Finally, we believe that the planned colocalization experiments (Reviewer #2, point 1) and the new data on lipid exchange kinetics (new Fig. 6 and Fig. S10) fit the current story more coherently, and thus, are more straightforward and informative than the generation and characterization of double mutant strains. For these reasons, we believe that the generation and characterization of D. discoideum double mutant strains is beyond the scope of the current study.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The authors perform proteomic analysis of Legionella containing vacuoles. The observe association of membrane contact site (MCS) proteins including VAP, OSBPs, and Sac1. Functional data indicates that these proteins contribute to PI4P levels on LCVs and their ability to acquire lipid from the ER to enable LCV expansion/stability. Overall the paper is an important contribution to the field and builds upon a growing appreciation for MCS in establishment of intracellular niches by microbial pathogens. I have only minor comments for the authors consideration.

Minor comments:

• line 145, "This approach revealed 3658 host or bacterial proteins identified on LCVs...". This number seems high... how does it compare to prior proteomic studies of pathogen-containing vacuoles?
• line 160. Can the authors comment on why mitochondrial proteins are observed in their proteomic analysis? Are these non-specific background signals or reflecting relevant organelle contact?
• line 268. It is reported that LCVs are smaller with MCS disruption at 2 and 8 h p,i.. Does this also lead to instability or rupture of LCVs? And related to this why would LCVs be bigger at 16h with MCS disruption?
• lines 288 and 299 "data not shown" this data should be included in a supplemental figure
• line 327. The authors choose to focus on the role of LepB and SidC in MCS modulation. The rationale for choosing these two amongst the ca 330 effectors was not given. Were other effectors also examined?

Significance

Comprehensive LCV proteomics of interest to field of cellular microbiology. Studies of MCS broadly relevant to cell biologists.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary of paper and major findings

Membrane contact sites (MCS) are locations where two membranes are in close proximity (10-80nm). MCS have a defined protein composition which tether the membranes together and function in small molecule and lipid exchange. Typically, MCS proteins contain structural (e.g., tethers) and functional (e.g., exchange lipids) proteins, in addition to proteins which regulate the structure and function of the MCS. In this manuscript, Vormittag et al describe protein components of MCS between the Legionella containing vacuole (LCV) and the host endoplasmic reticulum (ER) in the amoeba Dictyostelium. Proteomics of isolated LCVs followed by microscopy analysis identified several proteins which localize to either the LCV-associated ER (OSBP8), the LCV (OSBP11), or both (VAP and Sac1). The mammalian homologs of these proteins have been shown to play important roles in ER MCS, with VAP serving a structural role, Sac1 a PI(4P) phosphatase regulating PI(4)P levels, and OSBP8 and OSBP11 lipid transferring proteins. Given the importance of PI(4)P in formation and maintenance of the Legionella Containing Vacuole, the authors used dicty mutants to determine the importance of these proteins in bacterial growth, LCV size, and PI(4)P levels on the LCV. While VAP and OSBP11 appear to promote Legionella infection, OSBP8 appears to restriction infection, although all identified MCS components appear to play a role in decreasing PI(4P) shortly after infection. Finally, VAP and OSBP8 localization to the LCV is PI(4)P-dependent. Overall, the authors conclude that these MCS components play a role in modulating PI(4)P levels on the LCV.

Overall, this is an interesting study further exploring the role of PI(4)P in LCV-ER interactions, and how PI(4)P levels are regulated. The figures are clearly presented, there is an impressive amount of data, and rigor appears to be strong with appropriate replicates and statistical analysis. The phenotypes are often mild, but the authors are careful to not overinterpret the data. While this is an interesting study, additional experiments are necessary to support the overall model and the text needs to put the findings into the larger context.

Major comments

1. MCS contain protein complexes or a group of proteins, but the proteins here are studied in isolation and do not support the model shown in Figure 7. Co-localization studies of the putative LCV-ER MCS proteins are critical, especially given that the authors hypothesize the proteins are working together to modulate PI(4)P levels.
2. The phenotypes are relatively mild, suggesting functional redundancy. Double knockouts, particularly in VAP and OSBP11, may generate a stronger phenotype that better supports the hypothesis and demonstrate the importance during infection.
3. The timing of PI(4)P and MCS protein localization during infection is critical to understanding how MCS might be functioning. Based on Figure 6C, PI(4)P levels decrease on the LCV during infection, but this is not fully explained in the context of what's known in the literature and what is observed the previous figures. How does localization of different MCS components change during infection, and does this correlate with the changes in growth or LCV size? A better description in the Introduction on LCV-associated PI(4)P levels would be beneficial in orienting the reader to why PI(4)P levels are modulated.
4. OSW-7 has other targets besides OSBPs, and depleting Sac1 and Arf1 in A549 cells is not specifically targeting the MCS, as these proteins have other functions. The data in mammalian cells is not convincing and should be removed.

Minor comments

1. Figure 2 is missing details on number of experiments/replicates and statistical analysis.
2. Can the authors hypothesize why VAP promotes growth early during infection, but appears to restrict growth at later timepoints (Figure 3A)?
3. There is a large amount of data, which makes it difficult at times to follow. I suggest adding additional information to table 1, including LCV size and whether or not the protein's localization is PI(4)P-dependent.

Significance

Membrane contact sites during bacterial infection are a growing area of research. In Legionella, several papers point to the presence of MCS. Further, PI(4)P is known to be an important component on the LCV. This paper shows that MCS protein members are important in modulating LCV PI(4)P levels. The model as presented is not completely supported by the data as co-localization experiments are needed, along with more detailed analysis of how PI(4)P levels change over infection and the role of these MCS proteins in that process. This study will be of interest to those studying Legionella and other vacuolar pathogens.

Area of expertise is on membrane contact sites and lipid biology.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In the manuscript by Vormittag, et al., the authors perform proteomics identification of proteins associated with the Legionella-containing vacuole (LCV) in the model amoeba Dictyostelium discoideum comparing WT to atlastin knockout mutants. The authors find approximately half the D. discoideum proteome associated with the LCV, but there was enrichment of some proteins on the WT relative to the mutant. They focus on proteins involved in forming membrane contact sites (MCS) that previously were shown to be important for expansion of the Chlamydia-containing vacuole. Most significant are the oxysterol binding proteins (OSBP) and VapA (similar to that seen in Chlamydia). The authors show differential association of these proteins with either the LCV or presumably the ER associated with the LCV. Using a linear scale over 8days, they show that mutations in some of the MCS reduce yields in two of the OSPB knockout mutants and the growth rate of the vap mutant is slowed but ultimate yield is increased. Using some nice microscopy techniques, they measure LCV size, and the osbK mutant appears particular small relative to other strains, whereas the osbH mutant generates large vacuoles. This doesn't necessarily correlate with the PI4P quantities on the vacuoles (which is higher in all of them), but I am not totally sure how this is measured, and whether is it PI4P/pixel or PI4P/LCV. In all cases, this was reduces by Sac1 mutation. Surprisingly, even though there was uniform increase in PI4P in each of the mutants, loss of PI4P only affects localization of some of the proteins. Finally, in what seems to be a peripherally related experiment, the authors show that a pair of Legionella translocated effectors are required to maintain PIF4P levels, although it is not clear how this is related to the other data in the manuscript.

It is not clear from the manuscript if the authors are just cataloging things or trying to test a hypothesis. This is an extremely difficult manuscript to read and reconstruct what the authors showed. I really think that the only people who will understand what is written are people who are familiar with the work in Chlamydia starting in 2011 in Engel's and Derre's laboratories, which clearly showed that MCS and most specifically Vap/OSBPs are involved in vacuole expansion. If the authors could rewrite the manuscript along these lines, perhaps comparing their data to the Chlamydia data it would help a log. Otherwise, I don't think anyone else will understand why they are focusing on these things. I don't recommend new experiments (although re-analyzing data is necessary), but the manuscript has to be taken apart and claims removed, and data be interpreted properly. Otherwise, the manuscript seems like just a clearing house for data.

1. The problems start with the first figure, in which the authors state that almost half the D. discoideum proteome is LCV-associated. I doubt that this is correct, and they should base this on some selective criterion. Furthermore in Fig. 1A, they show Venn diagrams for how they whittled this down, but the Supplemental Dataset gives us no clue on how this was done. I can only sit down myself with the dataset and try to figure that out, but that is an unreasonable expectation for the reader. The dataset provided should have a series of sheets, describing how the large protein set was whittled down and how they were sorted, so the reader can evaluate how robust the final results were. To me (at least), if they said: "look we got this surprising result that suggests MCS are involved in promoting LCV formation, and although this is well recognized in Chlamydia but poorly recognized in Legionella", that would be satisfactory to me.
2. I am clueless regarding how Fig. 6 fits with the rest of the manuscript. If this is about MCS, there is no demonstration these effectors are directly involved in MCS other than the somewhat diffuse argument that there is some correlative connection to PI4P levels, that I am not particularly convinced by.
3. Lin 146 and associated paragraph. We don't need a catalog of proteins in narrative. There is more detail in the narrative than there is in the tables and figures, which would be a more appropriate way to present the data.
4. Line 186. There is nothing wrong with pursuing MCS based on the idea that this was seen before with Chlamydia and you wanted to test if this was a previously unappreciated aspect of Legionella biology. I don't see the rationale based on the proteomics, partly because I don't understand how the proteomics dataset was parsed.
5. Figure 3: These growth curves are super-weird. I am not used to looking at 8 days of logarithmic growth in a linear scale, and seeing no (apparent) growth for 4 days. Considering all the microscopy data are performed in the first 18 hrs of infection, its hard to see how this is related to data at 8 days post infection. If this were plotted in logarithmic scale, as microbiologists are used to doing, then perhaps we could see a connection. Also, in some cases, it might be helpful to calculate a growth rate, because its possible the author may now see some effects by comparing logarithmic growth rates.
6. Figure 2: The images don't necessarily show what the bar graphs show. In particular, look at Osp8. That image doesn't make sense to me.

In summary, I think the authors hit on something that is probably important for Legionella biology, but its not clear what they want to show. They are very invested in connecting everything to PI4P levels, which may or may not be correct, but it seems to me that perhaps taking more care in showing the importance of the Vap/OSPB nexus in supporting Legionella growth should be the first priority.

It is not clear from the manuscript if the authors are just cataloging things or trying to test a hypothesis.

If I understand Fig. 1, only one of the candidates (VapA) was verified as being more enriched in WT relative to atlastin mutants. This argues even more strongly that the authors have to describe their criteria for choosing these candidates

Significance

As stated above, the mansucript can't decide if its about MCS or PI4P, and I would argue strongly that the emphasis on PI4P detracts from the manuscript, as well as its inability to draw connection to previous work that is likely to be important.

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

Learn more at Review Commons

---

Reply to the reviewers

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

This paper examines the formation and repair of micronuclei in non-cancerous cells, specifically in mouse embryonic fibroblasts. This work was performed completely in culture and used a combination of western blot, confocal and superresolution microscopy to assess the contents of micronuclei over a repair period of 5 hours after 2 hours of induction of double strand breaks by treatment with etoposide. The authors found that the bodies colocalised with LC3, Beclin 1 and lysosomes suggestive of autophagy. However no evidence of autophagic flux has been demonstrated.

Major issues are as follows:

Figure 2

A - Any sense of the autophagic flux? LC3B - I and LC3B - II seem to be in equal quantities most of the time. Maybe using the tandem LC3 in this system could provide further insight. Also remove the violin plots from this graph and from G and H, as there are too few data points.

Thank you for your comment. We have evidence of a functional autophagic flux, since we observed an increasing number of acidic vesicles stained with Lysotracker in response to DNA damage, which were reduced after DNA repair. Some of the micronuclei were also co-stained with Lysotracker, suggesting their lysosomal degradation. We reorganized the data in the revised figure 2A to communicate better these observations. We reproduce here the dynamic of Lysotracker stain, please notice an increase in the abundance of acidic vesicles after 2h of DNA damage. A further evidence of activation of functional autophagy is the dynamic intracellular distribution of both LC3 and BECN1, indicative of autophagy induction. Please notice in revised Figure 2A that LC3 surrounding vesicles increases after 2h of DNA damage and diminish when DNA is repaired. BECN1 in control MEFs is highly concentrated inside the nucleus, predominantly at the nucleolus, and after DNA damage it redistributes towards the cytoplasm. Finally after DNA repair, BECN1 appears highly concentrated at the nucleus again. These dynamic changes correlate with autophagosomes formation and successful fusion with lysosomes. In the revised manuscript we removed the violin plot as suggested. Since the elimination of nuclear components occurs in a subset of cells, the role of the autophagic machinery needs to be analyzed cell by cell. We considered better to eliminate also the Western blot, as an analysis of the whole population does not provide information relevant for this study.

• Can you reduce the brightness in the merge image, as I cannot see DAPI nor a convincing Beclin-1/LC3 co-localisation.

Thank you for the observation. We improved the quality of the images and reorganized Figure 2 to convincingly show BECN1 and LC3 co-localization, together with Lysotracker, in nuclear alterations (buds and micronuclei). We modified the results text accordingly.

• Although the data is convincing, It would be clearer if the brightness of the merge image was reduced.

Thank you for your comment. We improved the images shown, these data is now integrated in new Figure 2A.

• Is the significant result the difference between 5h R Control si and 5h R Atg7? if so, there is no significant change in micronuclei as the same time point, can you explain this disconnect? are the buds being degraded prior to becoming micronuclei?

That is correct, we found no statistical significant difference in the number of micronuclei formed silencing Atg7, although there was a trend to reduce them. To consolidate the role of autophagy in nuclear buds and micronuclei formation, we studied Atg4-/- MEFs. We confirmed a statistical significant reduction of buds formation when autophagy is impaired (new Figure 2G). However, we observed that the number of micronuclei increased after 2h of DNA damage in Atg4-/- MEFs, suggesting that autophagy does not contribute to micronuclei formation but elimination. Together, our results suggest that the origin of buds and micronuclei are mechanistically different. A difference in the biogenesis of buds and micronuclei has been previously suggested studying cells cultured under strong stress conditions that induce DNA amplification, as well as in cells under folic acid deficiency. While interstitial DNA without telomere was more prevalent in buds than in micronuclei, telomeric DNA was more frequently observed in micronuclei (Fennech et al. 2011, Mutagenesis 26:125-132). We agree with the reviewer, it seems that not all the buds become micronuclei.

Figure 3 A - nice microscopy showing the co-localisation of TOP2A and LC3-GFP. I'm interested in DAPI being on some bodies and not others. Do you have any sense of the dynamics of this?

Thank you for the interesting question. Since removal of nuclear alterations as nuclear buds and micronuclei is a very dynamic process, we detect nuclear damaged material in the cytoplasm are at different degradation stages. Nucleases could be degrading DNA in micronuclei. Another possibility to the lack of DAPI signal in some micronuclei containing TOP2A and GFP-LC·is that TOP2A could be expelled from the nucleus with undetectable fragments of DNA or even without DNA, as a renewal process. We believe that nuclear buds can form without extruding DNA in some cases, perhaps to modulate proteostasis in addition to protect genome stability. In the revised manuscript we discuss this possibility further.

G - c shows a strand of mostly TOP2B coming from the nucleus. Is there any evidence that this occurs using either confocal microscopy or super resolution approaches. Could you try Z-stack to find these?

Thank you for the suggestion, we analyzed Z-stack images and tried to observe it also by immunofluorescence. We could detect some tubular signal connecting the nucleolus with a micronucleus containing TOP2B and BECN1 (arrow head in Fig 3B reproduced below), although we cannot be certain we are detecting the same nuclear extrusion mechanism by Electron Microscopy than by immunofluorescence.

Figure 4 C - is there a significant increase in FBL negative bodies, this would make sense if FBN is being degraded in the micronuclei during the repair process

We found that the number of micronuclei without FBL increased with statistical significant difference by Two-way-ANOVA followed by Dunnett´s multiple comparison test (P=0.463 comparing cells with 2h of DNA damage with control cells; P=0.0017 comparing cells after 5h of DNA repair with untreated cells; n=5). We agree with the reviewer, a possible explanation is that FBL is being degraded in micronuclei during the repair process. Although it could also be possible that nucleolar is less sensitive to Etoposide poisoning, or that nucleolar DDR is mechanistically different.

• Would it be possible to increase the n of these experiments to confirm either no change in FBL/LC3 co-loc, or evidence of increase?

Thank you for the suggestion. We repeated the experiment two more times to increase the n to 5. We found no statistical difference in the number of nuclear buds or micronuclei containing both FBL and LC3 during DNA damage and repair. Therefore it seems that the release of nucleolar components is not enhanced by Etoposide-induced DSB, suggesting that nucleolar DDR is a unique response, independent of DDR elsewhere in the genome (reviewed in Nucleic Acids Research, 2020, Vol. 48, No. 17 9449–9461 doi: 10.1093/nar/gkaa713).

Minor issues:

Figure 4 and 5 legends are in a different font.

Thank you. We correct the font in the current manuscript.

Reviewer #1 (Significance (Required)):

There is little specific data on the role of autophagy in clearing micronuclei in cancer cells, so this may be suggestive of a new mechanism that occur during normal cellular homeostasis. There are known links between lamin A defects and the formation of micronuclei, but not explicitly that the micronuclei are also Lamin A positive. it is likely that analogous processes occur in both cancer and non-cancer, so the impact of these data is not clear to me. This paper may be of interest to researchers interested in nuclear structure and DNA damage, but based on the data presented the significance is limited.

The significance of the present work is to discover that autophagy is relevant both during physiological DNA damage and in response to an exogenous DNA damaging agent, to extrude damaged DNA, TOP2cc and Fibrillarin from the nucleus. This knowledge is relevant since insufficiencies on autophagy imply a risk of genomic instability, which in turn could drive the cell into a senescent or malignant state. We present data showing that autophagy regulates the dynamic formation and elimination of nuclear buds and micronuclei in a mechanistically differentiated way. While autophagy contributes to nuclear buds formation, it is necessary for micronuclei elimination. Our data suggest that nucleophagy could be also a mechanism to alleviate basal nucleolar stress. As the reviewer noticed, some micronuclei did not have DNA. It is conceivable then that nuclear buds and micronuclei form also for a proteostatic function, not necessarily involving DNA damage elimination. We believe the significance of our work contributes to our understanding of the cell, as well as to cancer research. Whether common mechanisms between cancerous and normal cells occur is relevant to know, to consider the specificity of potential therapeutic approaches.

I don't have sufficient expertise to evaluate the super resolution microscopy beyond assessing the images.

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

Peer review of the manuscript with the number RC-2021-01181 by Muciño-Hernandez G et. al. at Review Commons and with the tittle "Nucleophagy contributes to genome stability 1 though TOP2cc and nucleolar components degradation"

1. Summary Muciño-Hernandez G et. al. show in this manuscript that mouse embryonic fibroblasts (MEFs) have basal levels of nuclear buds and micronuclei, which are indicators of genomic DNA damage. These basal levels of nuclear buds and micronuclei in MEFs increased after Etoposide treatment, which is known to induce DNA Double stranded Breaks (DSD). Interestingly, the nuclear buds and micronuclei co-localize with makers for nucleophagy (BECN1 and LC3) and acidic vesicles, suggesting that they are cleared by nucleophagy. The authors propose that basal levels of nucleophagy clear basal levels of genomic DNA damage that occurs as result from DNA-dependent biological processes in the cell nucleus, thereby contributing to nuclear stability of MEFs under physiological conditions. These basal levels of nucleophagy increase after the action of factors that induce DNA damage and nuclear stress. The concepts proposed by Muciño-Hernandez G et. al. are novel, since most of the current published data on nucleophagy related to DNA damage have been obtained under pathological conditions, e.g. implementing cancer cells.

The authors use in their manuscript various molecular biology techniques to obtain data that support their claims, including Western Blot analysis of protein extracts from MEFs, immunostaining on MEFs and neutral comet assays, complemented with state of the art imaging techniques, such as confocal microscopy, immunoelectron microscopy and super resolution microscopy. The quality of the data is sound. The structure of the manuscript support the understanding of the reader. However, I would like to suggest several improvements that will help to increase the quality of the manuscript, in order that fits to the standards of articles recently published in journals affiliated to Review Commons, such as the Journal of Cell Biology, the EMBO Journal or eLife.

1. Major comments

2.1 The authors have to improve the description of the results. Especially the description of those Figure panels containing plots that were generated using data from several experiments has to be improved.

One example is the description of the Figure 1D, which is in the lanes 137-151 of the current version of the manuscript. Whereas the authors describe in lanes 137-147 observations related to representative pictures of confocal microscopy after immunostaining presented in Figure 1D (left), the description of the quantification from 9 independent experiments presented in the plots in Figure 1D (right) comes relatively short in lanes 147-150 without mentioning any of the values implemented for creating the plots.

"Interestingly, while the frequency of nuclear buds gradually increased after DNA damage and during DNA repair, the frequency of micronuclei also increased after DNA damage, but diminished upon DNA repair."

The other plots presented in the different figure panels across the manuscript are described in a similar manner. I would like to suggest to the authors to improve their manuscript by including during the description of their results the values that were implemented for the degeneration of the plots presented in the manuscript. For example, in the specific case of Figure 1D above:

"Interestingly, the percentage of MEFs with nuclear buds gradually increased from XY% ({plus minus} XY SD) in control non-treated (Ctrl) MEFs to XY% ({plus minus} XY SD; P=XY) after 2 h Etoposide-induced DSB in MEFs and XY% ({plus minus} XY SD; P=XY) after DNA repair take place in MEFSs 5 h upon stop of Etoposide treatment (Figure 1D, right). In contrast, the percentage of MEFs with micronuclei significantly increased from XY% ({plus minus} XY SD) in Ctrl MEFs to XY% ({plus minus} XY SD; P=XY) after 2 h Etoposide-induced DSB, whereas it was reduced to XY% ({plus minus} XY SD; P=XY) 5 h after stop of Etoposide treatment (Figure 1D, right)."

Descriptions of the plots as mentioned above will make the text more intuitive for the reader, and they will make possible to read the Results Section without switching to the Figure Legends or the Material and Methods Section or to Supplementary Files. Even though the representative pictures from different microscopy techniques presented in the manuscript are of good quality and support the claims of the authors, it is important to mention that the quantifications presented in the plots demonstrate the statistical significance of these representative pictures. Thus, the authors should consistently include in the manuscript during the description of theirs results all the information (mean values, standard error of the means, P values, n values, etc.) that support their interpretation of the results and demonstrate the statistical significance of their claims.

Thank you for your clear and valuable advice. We followed it and in the revised manuscript we included the data in the results section.

2.2 Following a similar line of argumentation as in the previous point, the authors should provide as Supplementary Material an Excel file containing a statistical summary, including all statistical relevant information from each one of the plots presented in each Figure panel, such as n values, P values, Test implemented, values used for the plots, numbers of experiments, etc. The information could be organized in the Excel file in different data sheets according to the Figure panels, in order that the reader can easily navigate through the data. In the current version of the manuscript, one cannot find the values used for the generation of the plots presented in the manuscript in any of the submitted files.

Thank your for this suggestion. We have included in Table S1 an Excel file with a data sheet for each Figure panel, containing all the data collected and the statistical analysis performed.

Minor comments

3.1 In general, prior studies were appropriately referenced. Only few references has to be added.

Line 48: Add to the already included reference "Dobersch et al., 2021" also the reference Singh et al., 2015 PMID 26045162.

Thank you, we added this reference.

Line 53: Add the corresponding reference after the word "respectively".

We added the corresponding reference.

Line 82: Add the corresponding reference after the word "them".

We added the corresponding reference.

Line 125: Add the corresponding reference after the word "cells".

We added the corresponding reference.

Line 130: The expression "...by analyzing the recruitment of the phosphorylated histone γH2AX..." is the first time that the authors mention in the manuscript the DNA damage maker γH2AX. I suggest that is better introduced as " ... by analyzing the recruitment of the DNA damage marker γH2AX (histone variant H2A.X phosphorylated a serine 139, Rogakou EP, et al., 1998, PMID 9488723) to DSB sites."

Thank you very much for your suggestion. In the revised manuscript we corrected the text as suggested.

Line 199: Add the corresponding reference after the word "formation".

We added the corresponding reference.

Line 205: Add the corresponding reference after the word "cells".

We added the corresponding reference.

3.2 The use of the English language is appropriate throughout the manuscript. However, there are minor errors in the use of punctuation marks, in the use of prepositions and typos. I will list some of them below. However, I would like to recommend that manuscript is corrected by an English native speaker.

Thank you for your careful review of our manuscript. We corrected all the errors listed. A college proficient in English has reviewed the revised manuscript.

Line 41: "...and reproductive systems; genome instability also..." the semicolon can be replaced by a period.

√

Line 43: "Since early in development DNA is under constant endogenous..." between "development" and "DNA" there should a comma.

√

The sentence in lanes 53-55 has to be rephrased.

√

Lines 62-63: the expression "...throughout life." should be substituted.

√

Line 70: The abbreviation "rDNA" has to be explained the first time that is used.

√

Lines 81-82: It has to be explained for the scientist that is not specialized in the field of nucleophagy, how the integrity of the genome is threatened by micronuclei and nuclei-derived material.

√ Lines 106-110: The sentence is long. It would be easier to understand for the reader if this sentence is divided into two sentences.

√

Lines 121-122: The subtitle should be rephrased.

√

Lines 132-138: The sentence is long. It would be easier to understand for the reader if this sentence is divided into two sentences, e.g. with a period before the word "hence".

√

Lines 143-144: "... in a subpopulation of healthy, untreated cells...". The interpretation of "healthy" might be subjective. I would like to suggest substituting in the complete manuscript the word "healthy" by "control".

√

Line 163: The abbreviation for γH2AX was already introduced in line 130.

√

Line 182: A comma after "cell lines" is missed.

√

Line 183: delete "either". √ Lines 190-194: The sentence is long. It would be easier to understand for the reader if this sentence is divided into two sentences, e.g. with a period after the word "decreased" in line 191.

√

Line 218: I assume that instead of "bus", it should be "buds".

√

Line 220: I assume that instead of "iRNA", it should be "siRNA". In addition, it is the first time that the abbreviation is used. Thus, I suggest introducing it as "...was silenced by specific small interfering RNA (siRNA) previous to ..."

√

Line 327: delete the word "chronic".

√

Line 344: I assume that instead of "(figures 4C)", it should be "(Figure 4D)".

√

3.3 The structure of the Figures is ok for the peer review process and it might be optimized during editing of the manuscript. Nevertheless, I would like to suggest to the authors to increase the lettering size throughout all the figures. It will make the figures more intuitive.

Thank you for the suggestion. We increase the font size of the figures.

Reviewer #2 (Significance (Required)):

Significance

The work presented by Muciño-Hernandez G et. al. will be clearly a significant contribution to the scientific community working on autophagy, DNA damage repair and cancer, among others. It will be of interest to a broad spectrum of scientists, as I will elaborate in the following lines. The authors propose that MEFs have basal levels of genomic DNA damage under physiological conditions, which are cleared by basal levels of nucleophagy. On one hand, these findings are in line with various publications demonstrating that DNA-dependent biological processes in the cell nucleus, such as transcription, replication, recombination, and repair, involve intermediates with DNA breaks that may compromise the integrity of DNA. Thus, there must be mechanisms that ensure the integrity of the genome during these processes under physiological conditions, one of them seems to be nucleophagy. This perspective might explain the fact that proteins and histone modifications that were initially characterized during DNA repair also play a role during transcription, recombination, and replication. For example, phosphorylated H2AX at S139 (γH2AX) is often used as a marker for DNA-DSB [PMID 9488723]. However, accumulating evidences suggest additional functions of this histone modification [PMIDs 19377486; 22628289; 23382544]. In addition, McManus et al. [PMID 16030261] analyzed the dynamics of γH2AX in normal growing mammalian cells and found γH2AX in all phases of cell cycle with a maximum during M phase, suggesting that γH2AX may contribute to the fidelity of the mitotic process, even in the absence of ectopic- induced DNA damage. Further, Singh et al [PMID 26045162] and Dobersch et al [PMID 33594057] report that γH2AX plays a role in transcriptional activation in response to TGFB-signaling. Moreover, classical DNA-repair complexes have been linked to DNA demethylation and transcriptional activation [PMIDs 17268471; 28512237; 25901318], and DNA-DSB is known to induce ectopic transcription that is essential for repair, supporting a tight mechanistic correlation between transcription, DNA damage, and repair [PMID 24207023]. Perhaps, the authors might consider introducing several of the aspects and the citations written above into the Discussion section of the revised version of their manuscript. On the other hand, most of the published data related to nucleophagy have been obtained from cancer cells. Muciño-Hernandez G et. al. obtained their data implementing MEFs to demonstrate that the proposed mechanisms take also place under non-pathological conditions, what is one of the novel aspects of the present work.

I hope that my suggestions help the authors to improve their manuscript, thereby reaching the standards of manuscripts recently published in journals affiliated to Review Commons AND increasing the impact of their contribution to the scientific community.

Thank you very much for your suggestions. They helped us to present now a much-improved manuscript. We hope the revised work is now suitable for publication in the Journal of Cell Science.

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

In this manuscript, Muciño-Hernández and colleagues suggest that basal formation of nuclear buds and micronuclei increases in primary mouse embryonic fibroblasts following etoposide-induced double strand breaks (DSBs). The study combines the use of biochemical methodologies with confocal and super resolution microscopy in an effort to explore the contribution of nucleophagy to genome stability. The authors provide evidence that autophagy is induced upon etoposide treatment. They detected GFP-LC3 and BECN1 signals in nuclear buds and micronuclei even in untreated control and to a higher extent in etoposide-treated cells. Then, the authors examined whether nucleophagy is required for the removal of nuclear buds and micronuclei, by treating fibroblasts with control and Atg7 siRNA. The authors claim that the percentage of cells with micronuclei or nuclear buds decrease upon Atg7 knockdown, suggesting that components of the autophagy machinery induce the formation of these nuclear abnormalities. Moreover, Type II DNA Topoisomerases (TOP2A and TOP2B) and the ribosomal protein fibrillarin were detected in nuclear buds and micronuclei in fibroblasts treated or not with etoposide. Again in this case, GFP-LC3 was detected in fibrillarin-containing nuclear alterations. Based on these observations, the authors suggest that nucleophagy contributes to the elimination of chromosomal fragments or nucleolar bodies exiting the nucleus under DNA damage -inducing conditions. Specifically, they propose a key role for nucleophagy in maintaining genome stability by eliminating Type II DNA Topoisomerase cleavage complex (TOP2cc) and nucleolar components such as fibrillarin.

While it seems that there is a relationship between nuclear-extruded TOP2 with endogenous BECN1 and GFP-LC3 suggesting autophagic engagement, inconsistencies of fluorescent images between different figures indicate possible technical problems/limitations (please see specific comments, below), compromising authors' claims. LC3 immunoblotting and GFP-LC3 localization results appear over-interpreted (comments below). Neither TOP2 nor Fibrillarin have been shown to be actual autophagic substrates. Also, the link between genomic stability, micronuclei formation and autophagy has been previously reported (Zhao et al., PMID: 33752561).

An additional major concern is relates to nucleophagy being a selective type of autophagy. As such it requires efficient recognition and sequestration of the nuclear material destined to be degraded. Cargo specificity is mediated by receptor proteins, but no evidence for such receptors is provided in this study. Moreover, there is no real mechanistic insight on how nucleophagy mediates genome stability and how this can be interpreted in terms of cell survival under physiological and stress conditions. In other words, the biological significance of the findings presented has not been addressed.

Specific comments are summarized below:

The authors suggest that autophagy is induced after etoposide treatment and during the DNA repair process. However, the Western blot presented in Fig. 2A is not convincing and quantification does not support a significant autophagy induction in any of these cases. Autophagy appears to be induced 1h after etoposide removal, as evidenced LC3II/LC3 I increase (Fig. 2A and S2A). Nevertheless, all these changes should be more rigorously assessed.

Thank you for the observation. We removed the analysis of LC3II/LC3I by Western blot in the revised manuscript because a basal and induced elimination of nuclear components by the autophagic machinery occurs only in a subset of cells. It needs to be analyzed cell by cell. Pooling together all the cells dilutes the observation. Nevertheless, the dynamic intracellular distribution of both LC3 and BECN1 indicate autophagy induction. Please notice in revised Figure 2A that LC3 surrounding vesicles increases after 2h of DNA damage and diminish when DNA is repaired. BECN1 in control MEFs is highly concentrated inside the nucleus, at the nucleolus as it co-localized with Fibrillarin (new Figure 4E), and after DNA damage it redistributes towards the cytoplasm. Finally after DNA repair, BECN1 appears highly concentrated at the nucleus again. A further evidence of a functional autophagic flux, is the observation of an increasing number of acidic vesicles stained with Lysotracker in response to DNA damage, which were reduced after DNA repair. Some of the micronuclei were also co-stained with Lysotracker, suggesting their lysosomal degradation.

Line 190 and Fig. 2A: It is totally unclear whether "autophagy activation" takes place during the two waves described. There is no LC3B-I to LC3B-II conversion to initially suggest "autophagy activation". It rather suggests that autophagy is stalled. Fig. 2F shows that GFP-LC3 is strongly fluorescent into the lysotracker-stained lysosomes, further pointing to possible functional or technical problems.

As pointed out by reviewer 1, the images presented in original Figure 2F were over-exposed. In the current version we replaced those images with new images of better quality. We also reorganized the presentation of the data, and in revised Figure 2A we present photos where more convincingly can be observed a co-localization of BECN1 with LC3, with o without Lysotracker signal in nuclear buds and micronuclei. We also performed immunolocalization of endogenous LC3 (new Figure 2D) to rule out a possible misinterpretation of GFP-LC3 aggregates. As explained before, we removed original Figure2A.

Fig. 2B and Sup. Fig. 2B: BECN1 staining looks problematic. There is extreme BECN1 accumulation in the nucleus. Are those nuclear patterns of endogenous BECN1 and GFP-LC3 normal (see also minor comment 6 and 7)? Is there literature supporting such a distribution?

Yes, it has been documented BECN1 localization in the nucleus during development and in response to DNA damage stimuli such as ionizing radiation, and with a function related to DNA repair alternative to autophagosome formation (Fei Xu, et al. 2017, Scientific Reports | 7:45385 | DOI: 10.1038/srep45385). In the current manuscript we also detected endogenous LC3, to avoid a possible artifact with GFP-LC3 expression. We observed endogenous LC3 also localized in the nucleus (new Figure 2D).

It is hard to imagine how BECL1 is implicated in a (here hypothetical) nuclear lamina degradation event driven by LC3-lamin B1 direct interaction (Dou et al., 2015). BECL1 is an upstream to LC3 component and is a subunit of the PI3K complex catalyzing the local PI3P generation. The above should cause recruitment of the downstream autophagic machinery. Other subunits of the same complex or downstream effectors should be identified at the same spots to support authors' claims.

Our proposal that BECN1 is contributing to nucleophagy is supported by its co-localization with LC3 and Lysotracker stained vesicles (new figure 2A), as well as with TOP2 (Figure 3A-C). We appreciate the interesting idea of the reviewer; we certainly did not analyze the presence of BECN1 interacting partners. We agree, further studies analyzing their localization could complement our current findings. Supporting our work, others have observed UVRAG in the nucleus, specifically in centromeric regions, and it also has a role in DNA repair through its interaction with DNA-PK (Dev Cell. 2012 May 15; 22(5): 1001–1016. doi: 10.1016/j.devcel.2011.12.027). Given the anti-tumorigenic role of several autophagic molecules, it is tempting to speculate that several of them could have triple roles in the nucleus: directly interacting with DNA repair machinery, eliminating unrepairable DNA damaged and preventing excessive protein accumulation in the nucleus. Further experiments are necessary to probe this hypothesis, but are beyond the scope of the present manuscript.

U, 2h D and 5h R images of whole cells are necessary. The authors should also provide representative images of cells under different conditions i.e. control, etoposide-treatment and during DNA repair. Along similar lines, untreated control cells are not included in Fig. 2E and F. These images are needed for a better comparison between normal and DNA damage-inducing conditions.

The reviewer is right. In the revised Figure 2 we included representative images of control cell, Etoposide-treatment and during DNA repair cells. Images of whole cells are now shown in supplementary Figure 2S.

The authors state that autophagy is required for nuclear buds and micronuclei formation. However, the data shown in Fig. 2G and H are hardly convincing given that the statistical difference between cells treated with control and Atg7 siRNA is not strong (for example, *p˂0.5, 5h after etoposide removal). To provide further support to this notion, they should use cells from autophagy defective mutants and examine the appearance of nuclear abnormalities across different conditions compared to control cells.

We agree with the reviewer and followed his/her suggestion. We established collaboration with Dr. Sandra Cabrera, who kindly shared with us Atg4b-/- mice from which we isolated MEFs to compare side by side with WT MEFs the appearance of nuclear abnormalities. We confirmed a statistical significant reduction in the formation of nuclear buds in both conditions: silencing the expression of Atg7 by siRNA and in Atg4b-/- MEFs, suggesting that the autophagic machinery contributes to buds formation (new Figure 2F-G). Interestingly, we observed a different result analyzing micronuclei. While we found no statistical significant difference in the percentage of cells with micronuclei silencing the expression of Atg7 by siRNA, we found a statistical significant increment of cells with micronuclei in Atg4b-/- MEFs (new Figure 2F-G). This apparently discrepant result suggests that nuclear buds and micronuclei have a different mechanistic origin. A difference in the biogenesis of buds and micronuclei has been previously suggested studying cells cultured under strong stress conditions that induce DNA amplification, as well as in cells under folic acid deficiency. While interstitial DNA without telomere was more prevalent in buds than in micronuclei, telomeric DNA was more frequently observed in micronuclei (Fennech et al. 2011, Mutagenesis 26:125-132).

Lines 223-228: The role of autophagic machinery in the formation of nuclear buds is not supported and furthermore hard to conceptualize. How the components of autophagy are implicated during the nuclear buds and micronuclei formation? Colocalization of autophagic proteins might mean that autophagy is engaged at some point after or during the above formation. The causal, mechanistic and temporal aspects of the above budding and nucleophagic events need experimental support and/or more accurate interpretation.

We agree with the reviewer, and now we expressed our interpretation with more caution. The role of autophagic machinery in the formation of nuclear buds is supported by the following findings: a) the localization of LC3 and BECN1 in nuclear buds; b) the inhibition of Atg7 expression by specific siRNAs reduced the number of cells with buds and c) Atg4b-/- MEFs had reduced number of cells with buds (new Figure 2G). How the components of autophagic machinery are implicated in nuclear buds formation is an interesting question and deserves further investigation, beyond the scope of the present manuscript.

The authors claim that nucleophagy eliminates topoisomerase cleavage complex because TOP2A and TOP2B appear to more extensively co-localize with GFP-LC3 and BECN1 after etoposide-induced DSBs. However, the quantification presented in Fig. 3D-F to support this statement does not, in general, show a statistically significant difference in fibroblasts across different conditions (normal, etoposide treatment, etoposide removal).

Autophagic elimination of TOP2 protein is supported by the following findings: 1) both BECN1 and LC3 were detected in micronuclei in acidic vesicles (labeled with Lysotracker), which is indicative of the autolysosomal nature of the cytoplasmic compartment containing TOP2 (Figure 2A); 2) TOP2B was found by electron microscopy in some cells exiting the nucleus surrounded by LC3 (Figure 3G); 3) TOP2B accumulated in cells lacking ATG4, as expected if it is degraded by autophagy (Figure 3H).

Why would BECLIN colocalise with TOP2B in Figure 3g, given that beclin is involved in the initiation process?

We think that BECN1 is involved in additional functions to the initiation process of bud formation. For example, it has been shown by others that BECN interacts with TOP2 (Dev Cell. 2012 May 15; 22(5): 1001–1016. doi: 10.1016/j.devcel.2011.12.027). It could be working as an autophagic receptor targeting TOP2cc to buds and micronuclei. We are aware that further studies are necessary to test this hypothesis, but they are beyond the scope of this manuscript.

Fig. 4A and B: There is no enrichment of GFP-LC3 in "the nuclear alterations containing Fibrillarin" as stated in lines 341-343 comparing to the rest of the cellular GFP fluorescence.

It is true that there is not a local enrichment of GFP-LC3 as those normally reported as LC3 puncta in response to autophagy induction by starvation, for example. Nevertheless we are confident of the specificity of the observation, as not every nuclear alteration was found having GFP-LC3. We detected GFP-LC3 in 72% (mean ± 3.61 SD) of the nuclear alterations containing Fibrillarin in untreated cells, in 65.7% (mean ± 1.97 SD) of cells with 2h of DNA damage and in 90.33% (mean ±6.36 SD) after 5 h of DNA repair (in 5 independent experiments).

Moreover, there is no statistical significance in Fig. 4C and D measurements limiting the safety of authors' conclusions in lines 341-346.

We agree with reviewer´s observation. We repeated these experiments two more times and did not find a statistical significant difference in the percentage of cells with nuclear lesions containing Fibrillarin and GFP-LC3 after DNA damage nor after DNA repair. These results suggest that nucleolar DDR is a particular response, independent of DDR elsewhere in the genome, as has been suggested (reviewed in Nucleic Acids Research, 2020, Vol. 48, No. 17 9449–9461; doi: 10.1093/nar/gkaa713). An alternative is that the release of nucleolar components is not enhanced by Etoposide at the dose and time used in this work.

Lines 368-370: As discussed by the authors and reported in previous publication (Xu et al., 2017), "BECN1 interacts directly with TOP2B, which leads to the activation of DNA repair proteins, and the formation of NR and DNA-PK repair complexes", independent of its role in autophagy. Currently, there are no rigorous findings supporting the contribution of BECN1 (as a functional constituent of the core autophagic machinery) to nuclear damaged material extrusion (lines 382-384).

We agree with the reviewer in that we did not perform an assay to demonstrate that BECN1 is contributing to TOP2 nuclear extrusion as a functional constituent of the core autophagic machinery. Nevertheless, the following data support the proposal of an autophagic elimination of TOP2cc: 1) TOP2B was detected in micronuclei containing BECN1 (Figure 3B); 2) BECN1 was found in micronuclei containing LC3 and in an acidic vesicle (labeled with Lysotracker), indicative of the autolysosomal nature of the compartment (Figure 2A); 3) TOP2 was found in some cells exiting the nucleus surrounded by LC3 (Figure 3G); d) TOP2 accumulated in cells lacking ATG4, suggesting its autophagic degradation (Figure 3H).

Lines 435-441 and Fig. 5: The current findings do not support the proposed model. It is hard to support and conceptualize the statement "proteasome and nucelophagy function in a dynamic way inside the nucleus".

The reviewer is right. We made a mistake integrating an interpretation within the summary of the actual findings of this work. We correct the text in the current version.

In Fig. 5, LC3 appears to decorate inner nuclear membrane and probably to interact with some of the other proteins depicted, which is misleading.

We agree with the reviewer. We removed the scheme in the current manuscript.

Beclin-1 appears to interact with Fibrillarin (Nucleolus).

This is correct. We observed by immunofluorescence a co-localization of BECN1 with Fibrillarin (new Figure E), and demonstrated by co-immunoprecipitation that they are constituents of a complex (new Figure F).

Most of the differences in Sup. Fig. 3 lack statistical significance compromising the authors' claims.

We agree with the reviewer. To perform a separated statistical analysis of the percentage of cells with nuclear buds or micrnonuclei did not provide further information. We eliminated this analysis in the current version.

Many conclusions are drawn by colocalisation-immunofluorescence analysis. Co-immunoprecipitation experiments should also be performed to show that TOP2B and fibrillarin interact with LC3/autophagic machinery.

Thank you for your suggestion. We performed immunoprecipitation analysis and confirmed an interaction of Fibrillarin with BECN1, this result is now presented in Figure 4F. We found no co-immunoprecipitation of LC3 with either Fibrillarin or TOP2A, nor of TOP2B with BECN1.

Additionally, colocalisation analysis should be performed using tools such as Pearson's correlation and is an initial indication of nucleophagy. In the case of fibrillarin, immunofluorescence images do not indicate colocalisation, they need to be repeated.

The transport of Fibrillarin out of the nucleus by micronuclei formation and its autophagic degradation implies that both proteins are contained in the same vesicular compartment, it does not necessarily requires a direct interaction of Fibrillarin with LC3. Therefore, a co-localization detected by Pearson´s analysis is not a necessary confirmation of the nucelophagic degradation of Fibrillarin. Actually, Fibrillarin does not seem to interact with LC3, since we could not detect both proteins by co-immunoprecipitation. Nevertheless, we observed a nucleolar localization of BECN1 overlapping with Fibrillarin (new Figure 4E), and we confirm by co-immunoprecipitation the presence of both BECN1 and Fibrillarin in a complex (new Figure 4F). Following reviewer´s advice, we repeated two more times the analysis of Fibrillarin immunolocalization. We corroborated its localization in micronuclei and nuclear buds in 5.86% (mean ± 5.03 SD) of untreated cells, indicating a basal level of nucleolar material exclusion from the nucleus. Interestingly, the percentage of cells with Fibrillarin in nuclear alterations did not increased with statistical significance with Etoposide treatment. At 2 h of DNA damage we observed only a slight increase to 6.8% (mean ± 4.03 SD) of cells having nuclear buds and micronuclei with Fibrillarin, while the number of cells with nuclear lesions increased to 30.6% (mean ± 4.2 SD). Similarly, the proportion of cells having Fibrillarin in nuclear lesions after 5 h of DNA repair increased only to 7.66 % (mean ±6.08 SD), while the total number of cells having nuclear buds and micronuclei increased to 38.42% (mean ± 9.3SD). These results suggest that nucleolar components are constantly sent out of the nucleus as a homeostatic process, and not significantly in response to Etoposide-induced DSB.

Measurement of LC3/fibrillarin positive puncta should be performed, under basal conditions, genotoxic, and nucleolar stress under control and Atg7 knockdown conditions.

Since we observed no statistical significant change in the number of micronuclei with Fibrillarin under Etoposide-induced DSB nor DNA repair, we did not perform the suggested experiment.

Moreover, if nuclear proteins described are substrates of autophagy, then their levels would decrease upon autophagic induction i.e. starvation or in this case DNA damage and nucleolar stress. Thus, western blot analysis of relative protein levels can be performed.

Thank you for the suggestion. Since only 5% of the cells have micronuclei with Fibrillarin, and this proportion did not increased significantly in response to DNA damage, it is unlikely to detect a difference in the amount of Fibrillarin in response to autophagy manipulation performing a population analysis (as it is in a Western blot). Nevertheless, we compared Fibrillarin abundance by Western blot in WT MEFs vs. Atg4-/- MEFs untreated (U), treated for 2 h with Etoposide (D) and after 5 h of DNA repair (5) shown in the top panel of the follow figure. As expected, we found no statistical significant difference determined by 2way-ANOVA followed by Sidak´s multiple comparisons test (n=3). Ajusted P values are shown for each comparison (left graph).

On the other hand, since the percentage of cells with TOP2B in micronuclei and nuclear buds increased in response to DNA damage and during DNA repair, it was possible to detect a statistical significant accumulation of TOP2B in cells lacking ATG4 after 5h of DNA repair (bottom panel and right graph in the figure above). This observation is now included in new Figure 3H. Supporting our finding, TOP2A is reduced in cancerous cells grown under glucose deprivation (Alchanati, I., et al. 2009. PLoS One. 4:e8104).

Endogenous LC3 nuclear buds should also be detected to verify nucleophagy as GFP-LC3 has been shown to aggregate, causing artifacts under certain conditions.

We agree with the reviewer. We detected endogenous LC3 by immunofluorescence. This result is now included in Figure 2D.

Minor comments

In the Discussion section, the paragraph focused on the role of the ubiquitin-proteasome system is not substantiated by the data presented in the manuscript. Along similar lines, formation of aggresomes following etoposide treatment and their subsequent removal has not been monitored.

We apologized for the confusion, we corrected the text to now clearly distinguish which are our findings and which are published data that we just attempt to relate.

Western blots of better quality should be provided with assigned markers of protein size.

The Western blots shown have markers of protein size.

There are several language errors in the text that need to be corrected. Several sentences are too long and confusing or must be re-phrased. For example, see the lines: 123-125, 209-210,212, 218,221-222.

We apologize for our language errors. We corrected all errors indicated and asked colleges proficient in English to review our text.

Fig. 1B. Place "μm" into parenthesis.

√

Sup. Fig. 1B: Replace "gH2AX" with "γH2AX".

√

Fig. 1D: Separate DAPI and γH2AX channel images would be informative.

We now show also separated channels.

Fig. 2E: Enlarged separate DAPI, GFP-LC3 and lamin A/C channel images would be informative.

We now show also separated channels.

Line 218: Replace "bus" with "buds".

√

Fig. 2B, 2E, 2F, 3A and probably Sup. Fig. 2B represent MEFs treated for 2h with etoposide. The pattern of GFP-LC3 in 2B looks extensively nuclear and almost absent from cytoplasm.

We confirmed our finding detecting endogenous LC3.

In addition, Fig. 2B and 3B represent MEFs treated for 2h with Etoposide. The pattern of endogenous BECN1 in Fig. 2B looks extensively nuclear and almost absent from cytoplasm. In Fig. 3B the pattern is notably different.

BECN1 pattern of distribution is rather similar, predominantly in the nucleolus. We demonstrate it further by detecting BECN1 overlapping localization with Fibrillarin (new Figure 4E) and co-immunoprecipitation (new Figure 4F).

Sup. Fig. 2C: Index box is not properly aligned.

Thank you. We reviewed the alignment of each index box and reorganized the figure in the revised manuscript to add the whole blots of the new experiments we performed to analyze MEFs Atg4-/-.

Lines 154, 343 and 837: Replace "DBS" with "DSB".

Thank you, we corrected these typos.

Fig. 4 panels are not clearly cited at the text.

We apologize, we reviewed that they are clearly cited now.

Line 220: siRNA

Thank you, we corrected the text.

Lines 373-374: References "Lenain et al., 2015" and "Li et al., 2019" are missing.

Thank you for noticing it, we added the missing references. We use EndNote X9, we did not expect it to fail.

Lines 400-401 and 407: Probably the second "Latonen, 2011" reference needs "et al".

It is correct. We now cite this paper properly.

Line 427: Do authors refer to Fig. 1E rather than Fig. 2B?

Yes, we are sorry for this mistake. Thank you for pointing it out.

Line 434: Correct "clearance" spelling.

Thank you, we corrected it.

Reviewer #3 (Significance (Required)):

The authors suggest that nucleophagy contributes to the elimination of chromosomal fragments or nucleolar bodies exiting the nucleus under DNA damage -inducing conditions. Specifically, they propose a key role for nucleophagy in maintaining genome stability by eliminating Type II DNA Topoisomerase cleavage complex (TOP2cc) and nucleolar components such as fibrillarin.

However, neither TOP2 nor Fibrillarin have been shown to be actual autophagic substrates. Also, the link between genomic stability, micronuclei formation and autophagy has been previously reported (Zhao et al., PMID: 33752561).

We found nuclear buds and micronuclei with markers of different stages of the autophagic pathway, suggesting an active role of autophagy proteins in buds formation, and micronuclei removal. We detected TOP2 and Fibrillarin in micronuclei and propose their elimination by nucleophagy by the following findings: 1) both BECN1 and LC3 were detected in micronuclei in acidic vesicles (labeled with Lysotracker), which is indicative of autolysosomes (Figure 2A); 2) TOP2B was found by electron microscopy in some cells exiting the nucleus surrounded by LC3 (Figure 3G); 3) TOP2B accumulated in cells lacking ATG4, as expected if it is degraded by autophagy (Figure 3H); 4) BECN1 has a dynamic cytoplasmic-nucelar traffic in response to DNA damage; 5) BECN1co-localized with Fibrillaron in nucleolus and both proteins were co-immunoprecupitated.

The link between genomic stability, micronuclei formation and autophagy has been previously reported only in cancerous cells. Considering that physiological DNA damage occurs constantly in the cell, basal nucleophagy is potentially fundamental to maintain cells healthy.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this manuscript, Muciño-Hernández and colleagues suggest that basal formation of nuclear buds and micronuclei increases in primary mouse embryonic fibroblasts following etoposide-induced double strand breaks (DSBs). The study combines the use of biochemical methodologies with confocal and super resolution microscopy in an effort to explore the contribution of nucleophagy to genome stability. The authors provide evidence that autophagy is induced upon etoposide treatment. They detected GFP-LC3 and BECN1 signals in nuclear buds and micronuclei even in untreated control and to a higher extent in etoposide-treated cells. Then, the authors examined whether nucleophagy is required for the removal of nuclear buds and micronuclei, by treating fibroblasts with control and Atg7 siRNA. The authors claim that the percentage of cells with micronuclei or nuclear buds decrease upon Atg7 knockdown, suggesting that components of the autophagy machinery induce the formation of these nuclear abnormalities. Moreover, Type II DNA Topoisomerases (TOP2A and TOP2B) and the ribosomal protein fibrillarin were detected in nuclear buds and micronuclei in fibroblasts treated or not with etoposide. Again in this case, GFP-LC3 was detected in fibrillarin-containing nuclear alterations. Based on these observations, the authors suggest that nucleophagy contributes to the elimination of chromosomal fragments or nucleolar bodies exiting the nucleus under DNA damage -inducing conditions. Specifically, they propose a key role for nucleophagy in maintaining genome stability by eliminating Type II DNA Topoisomerase cleavage complex (TOP2cc) and nucleolar components such as fibrillarin.

While it seems that there is a relationship between nuclear-extruded TOP2 with endogenous BECN1 and GFP-LC3 suggesting autophagic engagement, inconsistencies of fluorescent images between different figures indicate possible technical problems/limitations (please see specific comments, below), compromising authors' claims. LC3 immunoblotting and GFP-LC3 localization results appear over-interpreted (comments below). Neither TOP2 nor Fibrillarin have been shown to be actual autophagic substrates. Also, the link between genomic stability, micronuclei formation and autophagy has been previously reported (Zhao et al., PMID: 33752561).

An additional major concern is relates to nucleophagy being a selective type of autophagy. As such it requires efficient recognition and sequestration of the nuclear material destined to be degraded. Cargo specificity is mediated by receptor proteins, but no evidence for such receptors is provided in this study. Moreover, there is no real mechanistic insight on how nucleophagy mediates genome stability and how this can be interpreted in terms of cell survival under physiological and stress conditions. In other words, the biological significance of the findings presented has not been addressed.

Specific comments are summarized below:

The authors suggest that autophagy is induced after etoposide treatment and during the DNA repair process. However, the Western blot presented in Fig. 2A is not convincing and quantification does not support a significant autophagy induction in any of these cases. Autophagy appears to be induced 1h after etoposide removal, as evidenced LC3II/LC3 I increase (Fig. 2A and S2A). Nevertheless, all these changes should be more rigorously assessed.

Line 190 and Fig. 2A: It is totally unclear whether "autophagy activation" takes place during the two waves described. There is no LC3B-I to LC3B-II conversion to initially suggest "autophagy activation". It rather suggests that autophagy is stalled. Fig. 2F shows that GFP-LC3 is strongly fluorescent into the lysotracker-stained lysosomes, further pointing to possible functional or technical problems.

Fig. 2B and Sup. Fig. 2B: BECN1 staining looks problematic. There is extreme BECN1 accumulation in the nucleus. Are those nuclear patterns of endogenous BECN1 and GFP-LC3 normal (see also minor comment 6 and 7)? Is there literature supporting such a distribution? It is hard to imagine how BECL1 is implicated in a (here hypothetical) nuclear lamina degradation event driven by LC3-lamin B1 direct interaction (Dou et al., 2015). BECL1 is an upstream to LC3 component and is a subunit of the PI3K complex catalyzing the local PI3P generation. The above should cause recruitment of the downstream autophagic machinery. Other subunits of the same complex or downstream effectors should be identified at the same spots to support authors' claims. U, 2h D and 5h R images of whole cells are necessary. The authors should also provide representative images of cells under different conditions i.e. control, etoposide-treatment and during DNA repair. Along similar lines, untreated control cells are not included in Fig. 2E and F. These images are needed for a better comparison between normal and DNA damage-inducing conditions.

The authors state that autophagy is required for nuclear buds and micronuclei formation. However, the data shown in Fig. 2G and H are hardly convincing given that the statistical difference between cells treated with control and Atg7 siRNA is not strong (for example, *p˂0.5, 5h after etoposide removal). To provide further support to this notion, they should use cells from autophagy defective mutants and examine the appearance of nuclear abnormalities across different conditions compared to control cells.

Lines 223-228: The role of autophagic machinery in the formation of nuclear buds is not supported and furthermore hard to conceptualize. How the components of autophagy are implicated during the nuclear buds and micronuclei formation? Colocalization of autophagic proteins might mean that autophagy is engaged at some point after or during the above formation. The causal, mechanistic and temporal aspects of the above budding and nucleophagic events need experimental support and/or more accurate interpretation.

The authors claim that nucleophagy eliminates topoisomerase cleavage complex because TOP2A and TOP2B appear to more extensively co-localize with GFP-LC3 and BECN1 after etoposide-induced DSBs. However, the quantification presented in Fig. 3D-F to support this statement does not, in general, show a statistically significant difference in fibroblasts across different conditions (normal, etoposide treatment, etoposide removal). Why would BECLIN colocalise with TOP2B in Figure 3g, given that beclin is involved in the initiation process?

Fig. 4A and B: There is no enrichment of GFP-LC3 in "the nuclear alterations containing Fibrillarin" as stated in lines 341-343 comparing to the rest of the cellular GFP fluorescence. Moreover, there is no statistical significance in Fig. 4C and D measurements limiting the safety of authors' conclusions in lines 341-346.

Lines 368-370: As discussed by the authors and reported in previous publication (Xu et al., 2017), "BECN1 interacts directly with TOP2B, which leads to the activation of DNA repair proteins, and the formation of NR and DNA-PK repair complexes", independent of its role in autophagy. Currently, there are no rigorous findings supporting the contribution of BECN1 (as a functional constituent of the core autophagic machinery) to nuclear damaged material extrusion (lines 382-384).

Lines 435-441 and Fig. 5: The current findings do not support the proposed model. It is hard to support and conceptualize the statement "proteasome and nucelophagy function in a dynamic way inside the nucleus". In Fig. 5, LC3 appears to decorate inner nuclear membrane and probably to interact with some of the other proteins depicted, which is misleading. Beclin-1 appears to interact with Fibrillarin (Nucleolus).

Most of the differences in Sup. Fig. 3 lack statistical significance compromising the authors' claims.

Many conclusions are drawn by colocalisation-immunofluorescence analysis. Co-immunoprecipitation experiments should also be performed to show that TOP2B and fibrillarin interact with LC3/autophagic machinery. Additionally, colocalisation analysis should be performed using tools such as Pearson's correlation and is an initial indication of nucleophagy. In the case of fibrillarin, immunofluorescence images do not indicate colocalisation, they need to be repeated. Measurement of LC3/fibrillarin positive puncta should be performed, under basal conditions, genotoxic, and nucleolar stress under control and Atg7 knockdown conditions. Moreover, if nuclear proteins described are substrates of autophagy, then their levels would decrease upon autophagic induction i.e. starvation or in this case DNA damage and nucleolar stress. Thus, western blot analysis of relative protein levels can be performed.

Endogenous LC3 nuclear buds should also be detected to verify nucleophagy as GFP-LC3 has been shown to aggregate, causing artifacts under certain conditions.

Minor comments

In the Discussion section, the paragraph focused on the role of the ubiquitin-proteasome system is not substantiated by the data presented in the manuscript. Along similar lines, formation of aggresomes following etoposide treatment and their subsequent removal has not been monitored.

Western blots of better quality should be provided with assigned markers of protein size.

There are several language errors in the text that need to be corrected. Several sentences are too long and confusing or must be re-phrased. For example, see the lines: 123-125, 209-210,212, 218,221-222.

Fig. 1B. Place "μm" into parenthesis.

Sup. Fig. 1B: Replace "gH2AX" with "γH2AX".

Fig. 1D: Separate DAPI and γH2AX channel images would be informative.

Fig. 2E: Enlarged separate DAPI, GFP-LC3 and lamin A/C channel images would be informative.

Line 218: Replace "bus" with "buds".

Fig. 2B, 2E, 2F, 3A and probably Sup. Fig. 2B represent MEFs treated for 2h with etoposide. The pattern of GFP-LC3 in 2B looks extensively nuclear and almost absent from cytoplasm.

In addition, Fig. 2B and 3B represent MEFs treated for 2h with Etoposide. The pattern of endogenous BECN1 in Fig. 2B looks extensively nuclear and almost absent from cytoplasm. In Fig. 3B the pattern is notably different.

Sup. Fig. 2C: Index box is not properly aligned.

Lines 154, 343 and 837: Replace "DBS" with "DSB".

Fig. 4 panels are not clearly cited at the text.

Line 220: siRNA

Lines 373-374: References "Lenain et al., 2015" and "Li et al., 2019" are missing.

Lines 400-401 and 407: Probably the second "Latonen, 2011" reference needs "et al".

Line 427: Do authors refer to Fig. 1E rather than Fig. 2B?

Line 434: Correct "clearance" spelling.

Significance

The authors suggest that nucleophagy contributes to the elimination of chromosomal fragments or nucleolar bodies exiting the nucleus under DNA damage -inducing conditions. Specifically, they propose a key role for nucleophagy in maintaining genome stability by eliminating Type II DNA Topoisomerase cleavage complex (TOP2cc) and nucleolar components such as fibrillarin.

However, neither TOP2 nor Fibrillarin have been shown to be actual autophagic substrates. Also, the link between genomic stability, micronuclei formation and autophagy has been previously reported (Zhao et al., PMID: 33752561).

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Peer review of the manuscript with the number RC-2021-01181 by Muciño-Hernandez G et. al. at Review Commons and with the tittle "Nucleophagy contributes to genome stability 1 though TOP2cc and nucleolar components degradation"

1. Summary

Muciño-Hernandez G et. al. show in this manuscript that mouse embryonic fibroblasts (MEFs) have basal levels of nuclear buds and micronuclei, which are indicators of genomic DNA damage. These basal levels of nuclear buds and micronuclei in MEFs increased after Etoposide treatment, which is known to induce DNA Double stranded Breaks (DSD). Interestingly, the nuclear buds and micronuclei co-localize with makers for nucleophagy (BECN1 and LC3) and acidic vesicles, suggesting that they are cleared by nucleophagy. The authors propose that basal levels of nucleophagy clear basal levels of genomic DNA damage that occurs as result from DNA-dependent biological processes in the cell nucleus, thereby contributing to nuclear stability of MEFs under physiological conditions. These basal levels of nucleophagy increase after the action of factors that induce DNA damage and nuclear stress. The concepts proposed by Muciño-Hernandez G et. al. are novel, since most of the current published data on nucleophagy related to DNA damage have been obtained under pathological conditions, e.g. implementing cancer cells.

The authors use in their manuscript various molecular biology techniques to obtain data that support their claims, including Western Blot analysis of protein extracts from MEFs, immunostaining on MEFs and neutral comet assays, complemented with state of the art imaging techniques, such as confocal microscopy, immunoelectron microscopy and super resolution microscopy. The quality of the data is sound. The structure of the manuscript support the understanding of the reader. However, I would like to suggest several improvements that will help to increase the quality of the manuscript, in order that fits to the standards of articles recently published in journals affiliated to Review Commons, such as the Journal of Cell Biology, the EMBO Journal or eLife.

2. Major comments

2.1 The authors have to improve the description of the results. Especially the description of those Figure panels containing plots that were generated using data from several experiments has to be improved.

One example is the description of the Figure 1D, which is in the lanes 137-151 of the current version of the manuscript. Whereas the authors describe in lanes 137-147 observations related to representative pictures of confocal microscopy after immunostaining presented in Figure 1D (left), the description of the quantification from 9 independent experiments presented in the plots in Figure 1D (right) comes relatively short in lanes 147-150 without mentioning any of the values implemented for creating the plots.

"Interestingly, while the frequency of nuclear buds gradually increased after DNA damage and during DNA repair, the frequency of micronuclei also increased after DNA damage, but diminished upon DNA repair."

The other plots presented in the different figure panels across the manuscript are described in a similar manner. I would like to suggest to the authors to improve their manuscript by including during the description of their results the values that were implemented for the degeneration of the plots presented in the manuscript. For example, in the specific case of Figure 1D above:

"Interestingly, the percentage of MEFs with nuclear buds gradually increased from XY% ({plus minus} XY SD) in control non-treated (Ctrl) MEFs to XY% ({plus minus} XY SD; P=XY) after 2 h Etoposide-induced DSB in MEFs and XY% ({plus minus} XY SD; P=XY) after DNA repair take place in MEFSs 5 h upon stop of Etoposide treatment (Figure 1D, right). In contrast, the percentage of MEFs with micronuclei significantly increased from XY% ({plus minus} XY SD) in Ctrl MEFs to XY% ({plus minus} XY SD; P=XY) after 2 h Etoposide-induced DSB, whereas it was reduced to XY% ({plus minus} XY SD; P=XY) 5 h after stop of Etoposide treatment (Figure 1D, right)."

Descriptions of the plots as mentioned above will make the text more intuitive for the reader, and they will make possible to read the Results Section without switching to the Figure Legends or the Material and Methods Section or to Supplementary Files. Even though the representative pictures from different microscopy techniques presented in the manuscript are of good quality and support the claims of the authors, it is important to mention that the quantifications presented in the plots demonstrate the statistical significance of these representative pictures. Thus, the authors should consistently include in the manuscript during the description of theirs results all the information (mean values, standard error of the means, P values, n values, etc.) that support their interpretation of the results and demonstrate the statistical significance of their claims.

2.2 Following a similar line of argumentation as in the previous point, the authors should provide as Supplementary Material an Excel file containing a statistical summary, including all statistical relevant information from each one of the plots presented in each Figure panel, such as n values, P values, Test implemented, values used for the plots, numbers of experiments, etc. The information could be organized in the Excel file in different data sheets according to the Figure panels, in order that the reader can easily navigate through the data. In the current version of the manuscript, one cannot find the values used for the generation of the plots presented in the manuscript in any of the submitted files.

3. Minor comments

3.1 In general, prior studies were appropriately referenced. Only few references has to be added.

Line 48: Add to the already included reference "Dobersch et al., 2021" also the reference Singh et al., 2015 PMID 26045162.

Line 53: Add the corresponding reference after the word "respectively".

Line 82: Add the corresponding reference after the word "them".

Line 125: Add the corresponding reference after the word "cells".

Line 130: The expression "...by analyzing the recruitment of the phosphorylated histone γH2AX..." is the first time that the authors mention in the manuscript the DNA damage maker γH2AX. I suggest that is better introduced as " ... by analyzing the recruitment of the DNA damage marker γH2AX (histone variant H2A.X phosphorylated a serine 139, Rogakou EP, et al., 1998, PMID 9488723) to DSB sites."

Line 199: Add the corresponding reference after the word "formation".

Line 205: Add the corresponding reference after the word "cells".

3.2 The use of the English language is appropriate throughout the manuscript. However, there are minor errors in the use of punctuation marks, in the use of prepositions and typos. I will list some of them below. However, I would like to recommend that manuscript is corrected by an English native speaker.

Line 41: "...and reproductive systems; genome instability also..." the semicolon can be replaced by a period.

Line 43: "Since early in development DNA is under constant endogenous..." between "development" and "DNA" there should a comma.

The sentence in lanes 53-55 has to be rephrased.

Lines 62-63: the expression "...throughout life." should be substituted.

Line 70: The abbreviation "rDNA" has to be explained the first time that is used.

Lines 81-82: It has to be explained for the scientist that is not specialized in the field of nucleophagy, how the integrity of the genome is threatened by micronuclei and nuclei-derived material.

Lines 106-110: The sentence is long. It would be easier to understand for the reader if this sentence is divided into two sentences.

Lines 121-122: The subtitle should be rephrased.

Lines 132-138: The sentence is long. It would be easier to understand for the reader if this sentence is divided into two sentences, e.g. with a period before the word "hence".

Lines 143-144: "... in a subpopulation of healthy, untreated cells...". The interpretation of "healthy" might be subjective. I would like to suggest substituting in the complete manuscript the word "healthy" by "control".

Line 163: The abbreviation for γH2AX was already introduced in line 130.

Line 182: A comma after "cell lines" is missed.

Line 183: delete "either".

Lines 190-194: The sentence is long. It would be easier to understand for the reader if this sentence is divided into two sentences, e.g. with a period after the word "decreased" in line 191.

Line 218: I assume that instead of "bus", it should be "buds".

Line 220: I assume that instead of "iRNA", it should be "siRNA". In addition, it is the first time that the abbreviation is used. Thus, I suggest introducing it as "...was silenced by specific small interfering RNA (siRNA) previous to ..."

Line 327: delete the word "chronic".

Line 344: I assume that instead of "(figures 4C)", it should be "(Figure 4D)".

3.3 The structure of the Figures is ok for the peer review process and it might be optimized during editing of the manuscript. Nevertheless, I would like to suggest to the authors to increase the lettering size throughout all the figures. It will make the figures more intuitive.

Significance

4. Significance

The work presented by Muciño-Hernandez G et. al. will be clearly a significant contribution to the scientific community working on autophagy, DNA damage repair and cancer, among others. It will be of interest to a broad spectrum of scientists, as I will elaborate in the following lines. The authors propose that MEFs have basal levels of genomic DNA damage under physiological conditions, which are cleared by basal levels of nucleophagy. On one hand, these findings are in line with various publications demonstrating that DNA-dependent biological processes in the cell nucleus, such as transcription, replication, recombination, and repair, involve intermediates with DNA breaks that may compromise the integrity of DNA. Thus, there must be mechanisms that ensure the integrity of the genome during these processes under physiological conditions, one of them seems to be nucleophagy. This perspective might explain the fact that proteins and histone modifications that were initially characterized during DNA repair also play a role during transcription, recombination, and replication. For example, phosphorylated H2AX at S139 (γH2AX) is often used as a marker for DNA-DSB [PMID 9488723]. However, accumulating evidences suggest additional functions of this histone modification [PMIDs 19377486; 22628289; 23382544]. In addition, McManus et al. [PMID 16030261] analyzed the dynamics of γH2AX in normal growing mammalian cells and found γH2AX in all phases of cell cycle with a maximum during M phase, suggesting that γH2AX may contribute to the fidelity of the mitotic process, even in the absence of ectopic- induced DNA damage. Further, Singh et al [PMID 26045162] and Dobersch et al [PMID 33594057] report that γH2AX plays a role in transcriptional activation in response to TGFB-signaling. Moreover, classical DNA-repair complexes have been linked to DNA demethylation and transcriptional activation [PMIDs 17268471; 28512237; 25901318], and DNA-DSB is known to induce ectopic transcription that is essential for repair, supporting a tight mechanistic correlation between transcription, DNA damage, and repair [PMID 24207023]. Perhaps, the authors might consider introducing several of the aspects and the citations written above into the Discussion section of the revised version of their manuscript. On the other hand, most of the published data related to nucleophagy have been obtained from cancer cells. Muciño-Hernandez G et. al. obtained their data implementing MEFs to demonstrate that the proposed mechanisms take also place under non-pathological conditions, what is one of the novel aspects of the present work.

I hope that my suggestions help the authors to improve their manuscript, thereby reaching the standards of manuscripts recently published in journals affiliated to Review Commons AND increasing the impact of their contribution to the scientific community.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This paper examines the formation and repair of micronuclei in non-cancerous cells, specifically in mouse embryonic fibroblasts. This work was performed completely in culture and used a combination of western blot, confocal and superresolution microscopy to assess the contents of micronuclei over a repair period of 5 hours after 2 hours of induction of double strand breaks by treatment with etoposide. The authors found that the bodies colocalised with LC3, Beclin 1 and lysosomes suggestive of autophagy. However no evidence of autophagic flux has been demonstrated.

Major issues are as follows:

Figure 2 A - Any sense of the autophagic flux? LC3B - I and LC3B - II seem to be in equal quantities most of the time. Maybe using the tandem LC3 in this system could provide further insight. Also remove the violin plots from this graph and from G and H, as there are too few data points. B. Can you reduce the brightness in the merge image, as I cannot see DAPI nor a convincing Beclin-1/LC3 co-localisation. F. Although the data is convincing, It would be clearer if the brightness of the merge image was reduced. G. Is the significant result the difference between 5h R Control si and 5h R Atg7? if so, there is no significant change in micronuclei as the same time point, can you explain this disconnect? are the buds being degraded prior to becoming micronuclei?

Figure 3 A - nice microscopy showing the co-localisation of TOP2A and LC3-GFP. I'm interested in DAPI being on some bodies and not others. Do you have any sense of the dynamics of this? G - c shows a strand of mostly TOP2B coming from the nucleus. Is there any evidence that this occurs using either confocal microscopy or super resolution approaches. Could you try Z-stack to find these?

Figure 4 C - is there a significant increase in FBL negative bodies, this would make sense if FBN is being degraded in the micronuclei during the repair process D. Would it be possible to increase the n of these experiments to confirm either no change in FBL/LC3 co-loc, or evidence of increase?

Minor issues:

Figure 4 and 5 legends are in a different font.

Significance

There is little specific data on the role of autophagy in clearing micronuclei in cancer cells, so this may be suggestive of a new mechanism that occur during normal cellular homeostasis. There are known links between lamin A defects and the formation of micronuclei, but not explicitly that the micronuclei are also Lamin A positive. it is likely that analogous processes occur in both cancer and non-cancer, so the impact of these data is not clear to me. This paper may be of interest to researchers interested in nuclear structure and DNA damage, but based on the data presented the significance is limited.

I don't have sufficient expertise to evaluate the super resolution microscopy beyond assessing the images.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this manuscript, Joshi et al investigate the intracellular behavior of PGRP-LE for the activation of the NF-kB pathway in Drosophila enterocytes undergoing pathogen infection. The authors identify that, upon enteric infection, PGRP-LE aggregates to form a microscopic structure of puncta, which colocalize with Rab5. The authors further analyze the role of Rab5 for the NF-kB pathway and suggest that Rab5-dependent pathway represents one of two distinct routes for the activation of the NF-kB pathway based on the observation that RNAi-mediated knockdown of Rab5 selectively downregulates PGRP-SC1, which results in systemic immune response. Generally, the manuscript provides convincing experimental results to support the authors' arguments raising an interesting cell biological aspect of PGRP-LE for the well-known immune pathway. However, in my opinion, there are some ambiguous points as well. I would like to have several suggestions to strengthen the manuscript.

Major comments

1. To identify the role of Rab5, the authors performed an RNAi-mediated knockdown experiment and found that the expression of PGRP-SC1 is downregulated but the expression of other target genes such as AttacinD are not affected. The authors concluded that the expression of PGRP-SC1 is under the control of a Rab5-dependent route while other targets are regulated by Rab5-independent route. However, an alternative interpretation would be that Rab5 is required for all target genes and the observed differential expression of the targets is due to residual activity of Rab5 after RNAi-mediated knockdown. If the authors show that RNAi-mediated Rab5 knockdown almost deplete Rab5 expression, it would be helpful for the authors' argument. Also, this alternative explanation is worth to be provided in the discussion section.
2. The authors show that enterocytes with Rab5 knockdown still produce enlarged puncta without any further characterization. However, the identity of this subcellular structure would be an important piece of information to support the authors' argument concerned with a Rab5-independent route, which is largely a speculation at the moment. So, I would recommend to investigate whether the enlarged puncta colocalize with any known endosome and/or autophagosome markers. This information will enable to understand the Rab5-independent NF-kB activation pathway (e.g. by manipulating this pathway) in enterocytes.

Minor comments

It would be helpful for general readers to have an additional figure with a simple drawing of the authors' working model.

Significance

This paper showed for the first time a Rab5-dependent PGRP-LE aggregation that act as a signaling hub to finely modulate NF-kB pathway. As NF-kappaB is an evolutionarily conserved transcription factor that is essential for the immune activation from Drosophila to mammals, the present information would be of interest to a broad audience.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

The manuscript presented by Joshi et al presents a body of results describing the aggregation of the peptidoglycan receptor PGRP-LE, which is an intracellular protein, in response to intestinal infection by oral ingestion in Drosophila. This study is based on the generation of two CRISPR/Cas9 mutant lines in which the PGRP-LE sequence has been fused to the V5 epitope (inserted into the PGRP domain) or the fluorescent protein eGFP (added at the C-terminal position). In each case, the "sensor protein" is expressed under the control of the endogenous promoter that ensures a physiological expression of the sensor.

As expected from the literature, the authors show that the expression of each of the two PGRP-LE sensors is strongly induced in the digestive tract by the ingestion of the bacterium Erwinia carotorova carotorova (E.cc), which is known to produce a strong activation of the NF-kB signaling cascade under these infection conditions. In this study, the authors show that PGRP-LE-V5 sensors form clusters in the immunocompetent domains of the gut, particularly in the R4 domain where NF-kB activation is known to be primarily dependent on PGRP-LE. This clustering is not observed in clones with little or no expression of PGRP-LE due to RNAi-mediated knockdown of gene expression. The transcription of endogenous PGRP-LE or that of the PGRP-LE-V5 and eGFP sensors is not increased by the infection, allowing the authors to propose that the PGRP-LE protein pre-existing in the intestinal cells relocalizes into clusters or aggregates. These aggregates are also marked by the Rab5 protein, a marker of early endosomes, but not by the Rab7 marker, a marker of late endosomes. The expression of the antimicrobial peptide AttD is similar in the presence of the sensors as in control flies, which indicates that the immune response is not drastically affected by these sensors. Moreover, the kinetics of receptor aggregation parallels that of NF-kB pathway activation followed by AttD expression.

Ingestion of E. coli or commensal bacteria or PGN, which do not induce a significant immune response according to the literature and data reproduced here by the authors, do not induce receptor aggregation either. Surprinsingly, heat-killed E.cc bacteria, which induce no or a very slight expression of AttD cause more but smaller aggregates of PGRP-LE. Moreover, these aggregates are not labeled by the Rab5 protein. The authors show that this aggregation of PGRP-LE is not affected by the down-regulation of the HH pathway, and is correctly induced by a uracil auxotrophic Ecc mutant. The expression of RNAi directed against the dFADD protein, an adaptor of the PGRP-LC membrane receptor contributing to the activation of the Imd/NF-kB pathway, does not alter this aggregation either. Finally, the authors observed that a set of genes whose expression in response to E.cc is dependent on PGRP-LE shows a differential dependence on Rab5 expression: while PGRP-SC1 expression is affected by Rab5 silencing, this is not the case for PGRP-LB or PGRP-SC2 expression. Furthermore, directed Rab5 knock-down in the adult gut induces an exacerbated immune response in the fat body. The combined action of PGRP-LE and Rab5 would therefore be necessary for the activation of PGRP-SC1 but not of PGRP-LB or PGRP-SC2. From these results the authors propose the existence of two pathways of activation of NF-kB target genes downstream of PGRP-LE, depending or not on an endosomal Rab5 signaling platform. The authors also propose that the amount of PGN may control the choice of Rab5-dependent or Rab5-independent pathway activation.

Major comments:

The authors have constructed beautiful genetic tools (PGRP-LE sensors). They present a set of convincing results concerning the formation of PGRP-LE protein aggregates in response to E.cc infection under different infection conditions or genetic backgrounds. Nevertheless, the study remains essentially descriptive and based on immunofluorescence and expression studies of a small set of genes responsive to the NF-kB pathway. To better support the hypotheses and conclusions, deep sequencing studies would be very powerful to reveal whether the differential expression observed for the target genes PGRP-SC1 versus PGRP-SC2 and PGRP-LB is also true for a large set of genes of the immune response, which would make the results more accurate. It would also be interesting to study more genetic conditions, e.g. affecting the endocytic pathway, proteasomal degradation or autophagy in order to determine the fate of aggregates and the mechanisms of their removal/resolution. Furthermore, biochemical studies, such as immunoblots, would allow following the fate of PGRP-LE at the protein level. The authors indeed show that the expression of PGRP-LE gene is not induced by E.cc but one can wonder if the protein is stabilized. They propose that PGRP-LE is not recycled because it does not colocalize with Rab7, but it might be also degraded by the lysosomal pathway rather than recycled. It would be interesting to test if aggregates are removed by the lysosomal pathway or by autophagy. Moreover, a recycling via Rab7 is maybe not expected for a protein that is not localized on the plasma membrane. A kinetic study including co-staining with Rab7 would better support the conclusion that there is no colocalization with Rab7. Otherwise, they may miss the right timing to observe this colocalization. Similarly, the absence of colocalization with Lamp1 at a given time does not allow concluding with certainty that PGRP-LE is not degraded by the lysosomal pathway. The 24h staining (Fig2A) sounds similar to a Lamp1 profile. One should therefore be more cautious in drawing conclusions about these co-staining experiments. Moreover, Rab7 and Lamp-1 staining are faint and miss RNAi controls to show the specificity of the staining.

In conclusion, a corpus of additional experiments would be necessary to significantly advance the field and demonstrates the existence of a Rab5 signalization platform causing differential expression of target genes of the immune response. The expression of a large set of genes could be tested, some of the RNAi lines used needs to be better characterized, complementary genetic and biochemistry experiments would help to understand the fate of PGRP-LE, the effect of the Imd pathway could be more documented with other RNAi than FADD... The role of other components of the endocytic pathway tan Rab5 could be assayed with other RNAi (Rab7, ESCRT, ... ) to block the endocytic pathway and observe if it interferes with the aggregates. The authors could also possibly test the proposed hypothesis on the amount of PGN/bacteria that would be at the origin of a differential response.

In the figure and figures legends and methods, the authors describe the aggregates as oligomers, but no experiment support this assumption. In the text, the authors stick with the nomenclature as clusters or aggregates which is more appropriate.

Minor comments:

• The abstract would benefit from being rewritten: the first half provides general information that is not strictly necessary, which prevents a more thorough description of the results. I disagree or misunderstand the statement "little is known about the subcellular events required to translate these early steps into downstream target gene transcription" because extensive studies of the fly immune response have been done.
• Two spellings in the intro: PeptidoGlycaN or PeptidoGlycan. I suggest peptidoglycan
• "the innate immune response that might otherwise be obscured by the action of the adaptive immune response": this is a rather archaic way of thinking because it is clear that the two responses are complex and intimately intertwined.
• "to visualize PGN detection by PGRP": correct "by PGRP-LE". -avoid "to our surprise". -"locus-directed": I suggest "tissue directed" or "in a localized manner in the digestive tract".
• Describe the purpose and procedure of smurf methodology.
• As noted above, do not describe clusters as oligomers in the methods and figures and figure legends. -"PGRP-LE recruits Rab5 protein": do the authors suggest a direct interaction between the two products? If so, it would be interesting to test this with co-IP experiments. However, it is possible that the aggregates are internalized in the endosomal compartment, independently of any Rab5/PGRP-LC interaction. Therefore, the term "recruits" is confusing here. -To make the results accessible to a broader audience, the authors may clarify the drosophila-specific genetic tools used in this study (Flpout clones, Gal80ts conditional expression...)
• In some cases, statistical analysis of RT-qPCR data are performed using a one-way ANOVA (fig 1H, 5A) whereas in others (fig 2H and L, 5B) a non-parametric Kruskal-Wallis test is used. The rationale for these discrepancies should be explained. Moreover, in all these experiments the data are compared to a control that is set to 100% and has no standard deviation. This violates some of the ANOVA assumptions (normality of the data points). To be correct, an outside control should be used to normalize the data (including the control to which the other genotypes are compared)
• Could the authors better explain the rationale for using PGRP-LE::V5 in some experiments and PGRP-LE::GFP in others? -Fig 1H: in this experiment, according to the legend, all the genotypes are infected. So it's not clear how the authors conclude that infection does not activate PGRP-LE expression in the absence of a non-infected control. We may have missed some points. Furthermore, as stated above, the authors could also perform a Western blot to ensure that PGRP-LE translation is not activated, or the protein stabilized, following infection.
• Fig 2A: The PGRP-LE aggregates a 24 hpi look different from the previous time points. It would be interesting to make a double staining with a Lamp1 antibody to check for colocalization at this late time points.
• Fig 2H: attD induction by hk E.cc is indicated as not significantly different from uninfected control and presumably not from E. coli and PGN. So the statement "hk E.cc which induced a weak AttD transcription" in the text is not correct.
• Fig 3: The RNAi lines used in this figure have no effect on PGRP-LE aggregation. To safely conclude that the corresponding proteins do not play a role in this process, the efficiency of the RNAi lines against their respective targets should be shown.
• Fig 3A,B : why no quantification of the aggregates are presented in this particular figure?
• Fig 4D: the pictures are too small, use the same magnification as in A and C

Significance

The studies presented in this manuscript are interesting and well done but remain mainly descriptive without sufficient data to support what could be a conceptual advance. Further work is needed to demonstrate that PGRP-LE would signal via two different pathways, dependent or not on Rab5 and the endocytic machinery. Further genetic and biochemical studies would allow to better describe these two putative signaling pathways leading to differential immune response genes expression, and/or the nature (oligomeric or not) and fate of PGRP-LE aggregates (endocytic-, lysosomal-, autophagic-patways,...). Such endosomal signaling platform has been described for the activation of the Toll pathway. Exacerbated immune response in the fat body following inactivation of Rab5, Fab1, and ESCRT components has been described earlier suggesting that accurate termination of IMD signaling also requires the endocytic machinery.

This study concerns fly scientists interested in the fine understanding of the signaling mechanisms of the innate immune response and may have a wider audience in the community of scientists interested in the molecular mechanisms of cell signaling in eukaryotic cells in response to external stimuli, and the role of endocytic trafficking in this response. Our expertise (reviewer and co-reviewer) covers the NF-kB-dependent immune response and some aspects of intracellular trafficking.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

The authors used CRISPR engineering to drop a V5 or GFP tag into teh PGRP-LE locus (protein fusions) to monitor the behavior of this intracellular peptidoglycan sensing receptor in the drosophila midgut. They show that upon immune stimulation with Ecc that PGRP-LE forms some sort of aggregate or punctae that is dynamic during the 24 hour of infection monitored. A similar response is not seen with live E. coli but a week and smaller response is observed with heat killed E. coli, for unclear reasons. These punctae appear to form independent of the classic IMD signaling components, suggesting it is upstream event in the pathway which is consistent with early studies showing the PGRP-LE multimerizes (infinitely) upon binding PGN and also that it forms amyloid fibrils doing signaling. The Ecc punctae tightly colocalize with Rab5 but not Rab7 or other early endosome markers, but in the absence of Rab5 the PGRP-LE punctae are greatly enlarged. Rab5 was found to critical for induction of PGRP-SC1 but not the classic IMD pathway AMP, Attain.

While the conclusions of the report are intriguing and the development of these tools is very exciting, the conclusions are not fully convincing. To start, the author wish to conclude that PGRP-LE localization is altered with Ecc infection but they have not excluded that the expression of the protein is sharply upregulated. I.e. in the uninfected animals there is not really any PGRP-LE observed (1D). The try to tackle this by looking at mRNA expression, but this data lacks the unaffected control. [In fact, the uninfected control is missing on most of the gene expression data, which is a troubling omission and makes it hard to really understand what the data shows.]. Moreover, the mRNA levels do not necessarily corresponding to the protein levels, i.e. there could be post translation control. So, overall, the authors need to provide more compelling evidence that PGRP-LE is relocalized upon Ecc challenge rather than upregulated.

Moreover, the paper contains some seeming contradictory findings that the authors make little effort explain. For example, they conclude "These results suggest that although smaller PGRP-LE aggregates can form normally in the absence of Rab5, the latter is required for proper bigger E.cc mediated PGRP-LE aggregates" because E. coli induced PGRP-LE clusters don't colocalize with Rab5, yet in the absence of Rab5, the Ecc cluster are super-enlarged (4F). This makes no sense with the conclusions.

Finally, the interaction and function of the Rab5 interaction is underdeveloped and lacks insight. For example, why is Rab5 required for the induction of one target gene but not another? And, why not characterize this more completely? Why is there not Rab5 vesicle with E. coli feeding or even uninfected? The cell biology requires more in-depth consideration. From 4E, the authors wish to conclude that the Rab5 vesicle are induced by Ecc (even in the absence of PGRP-LE) yet the uninfected control is not shown. IN a simple world, would not one would expect Rab5 endosomes in all cells, at least to some level?

And, focusing on the big picture, the authors claim that it is "not easily testible" if the PGRP-LE aggregates are amyloidal, as suggested by earlier publications. This could actually be tested by staining with amyloid specific dies and/or suitable mutants engineered int he RHIM domain. This would be very informative if the authors could extend this work to examine this question.

Minor comments:

All the colocalization data should be quantified as in 4B. It is not true that DAP = Gram negative. Gram-positive bacilli also have DAP PGN. The wording in the Introduction should be adjusted. The text needs a careful proofreading.

Referees cross-commenting

I think the comments from #2 and myself are aligned. Working is interesting, tools are especially exciting, but the studies are descriptive and under-developed. I will further add, I found the absence of uninfected controls for many assays a major problem.

Significance

The significance of this work lies in the development of powerful tools to track an intracellular innate immune receptor in an intact animals. The connection to Rab5 is curious and likely an important advance in our understanding of the cell biology of this pathway, but is under-developed. The significance is this difficult to know for certain. The Drosophila immunity field, and the insect immunity field more broadly, will be keenly interested in this study. The wider NF-κB/innate immune field will also be interested in these findings, given teh similarity between this pathway and NOD1/NOD2 immune sensing in mammals.

My area of expertise is the Drosophila immune response and this manuscript is very much in my wheelhouse.

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

Learn more at Review Commons

---

Reply to the reviewers

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

Avar et al report on the development of a high-throughput method to screen modifiers of prion replication in cell lines using a genome-wide siRNA library. They identified a number of hits and further studied one candidate, the ribonucleoprotein Hnrnpk. The authors convincingly show the interest of their method. However, the claims that the ribonucleoprotein Hnrnpk impact prion propagation need to be more quantitatively and statistically substantiated.

1. * A large part of the manuscript is dedicated to the validation of the high-throughput assay (called QUIPPER). QUIPPER is made in 384-plates and provides great technological improvement. It works with different prion-permissive cell lines and different prion strains. QUIPPER is an antibody-FRET-based assay that detects a specific population of PrPSc that resists phospholipase C (PIPLC) treatment. Historically, PIPLC has been shown to cleave cell surface PrPC while preserving PrPSc (which is endocytic or inaccessible). I would recommend that the authors quantify the proportion of PIPLC-resistant PrPSc (PrPPIPLC) versus total PrPSc in their different models. First, PrPPIPLC proportion may be cell and strain dependent. Second and most importantly, as siRNA effects are studied using PrPPIPLC as readout, it is crucial to know if this form is a bona fide surrogate of PrPSc and infectivity or only a specific, subcellular, potentially minor form of PrPSc. This is particularly important as the effects of Hnrnpk knock-down in QUIPPER and western blot sounds discordant; in QUIPPER, the effects are strong (> 5-fold) while by western blot, the effects are much more modest (We addressed this issue in several ways; firstly, we quantified the proportion of PIPLC-resistant PrP (PrPPLC) versus PrPSc in two different models (Fig. 1B and D). Secondly, we directly compared residual infectivity of cells treated with PK or PIPLC (Figure 1C), using the standard scrapie cell assay. The results show that infectivity is retained upon PIPLC treatment. In addition, we assessed the 161 hits obtained via QUIPPER using PrPSc as a readout (Fig. 3B).

To provide further data on the robustness of our PIPLC-based readout, we have performed western blotting of infected and uninfected cells upon PIPLC treatment and assessed the band patterns following PIPLC administration. This Figure is now incorporated in the manuscript as Supp. Fig. 1C and demonstrates that upon PIPLC digestion of NBH and RML infected CAD5 and GT-1/7 cells, PrP is barely detectable in the non-infected cells, while it is in the prion infected ones. The blots also show that the PIPLC-resistant PrP (PrPPLC) is resistant to PK digestion. These new data, together with those provided in Fig. 1B and Figure 1C, show that PrPPLC is equivalent to PrPSc in terms of PK resistance and infectivity.

The reviewer pointed out a discordance between Western Blotting and QUIPPER. Although it is not clearly stated, we think the reviewer may be suggesting a discordance based on Fig. 3D. We would like to point out that Fig. 3D does not report fold changes as the reviewer is suggesting, but Z-scores, measured by standard deviations from the mean, not allowing to infer fold-changes. We quantified the effect of NT and HNRNPK targeting siRNAs on prion levels (Fig. 4A) and saw a three-fold change. We believe that the quantifications provided in the new version of the manuscript alleviate the concerns regarding any discordance.

Technically, this is quite easy as it necessitates, after PIPLC treatment, the quantification of PrPSc in the supernatant versus PrPSc in the cell pellet. In Fig. 1C, the authors show that PrPPIPLC is infectious in a cell-scrapie assay. Using this approach, they could also quantify the infectivity of these species relative to the total infectivity content.

We addressed this in Supplementary Fig. 1C as depicted above. Supplementary Fig. 1C shows the alikeness of the PrP species measured via the QUIPPER vs. the canonical PK digestion: upon digestion with PIPLC following a PK treatment, we detect PrPSc. Therefore, the experiment demonstrates that PrPPLC is alike in nature to PrPSc. The difference between the PK digested (lanes 3&4) vs PIPLC treated then PK digested lanes (lanes 7&8) is the PrPSc that is released into the media following PIPLC digestion.

• *

• The authors identified a list of prion modifiers candidate. Surprisingly, the authors did not perform a pathways analysis to identify potential pathways that could impact prion propagation.*

Despite extensive efforts, there were no pathways that were enriched in our 40 hits, which is mentioned in the discussion part of the manuscript. Two analyses (for the 161 candidates and 40 hits) are now added to Supplementary Fig. 3C and pasted below.

• *

• The authors then studied in more details one hit, the ribonucleoprotein Hnrnpk. They studied the impact of Hnrnpk knock-down on PrPC and PrPres levels in different cell lines. These data (Fig 4 and Fig S4) lack quantitative (on a higher number of wells) and statistical analyses. The western blot that are shown suggest that PrPC levels are slightly increased by the siRNA and that the increase in PrPres levels is modest, barely significant given the western blot method. Same comment after PSA treatment, at least in PG127-infected hovS cells.*

We performed a quantification on the western blots for all figures mentioned by the reviewers throughout the manuscript. These are incorporated to the manuscript for the figures: Fig. 4A, Fig. 4B, Supplementary Fig. 4A, Supplementary Fig. 4C, Supplementary Fig. 4D, Supplementary Fig. 4F, Supplementary Fig. 4G.

Additionally, statistical analyses have been incorporated into the manuscript in these figures: Fig. 4C, Fig. 4D, Fig. 4E, Fig, 4F, Fig, 4G, Fig, 4H, Supplementary Fig. 4F. The analyses and the quantitative data demonstrate the effect of Hnrnpk downregulation and PSA treatment on prion levels to be significant. Moreover, we also addressed the regulation of prions via HNRNPK using vacuoles as a read-out as well as with a different mode of regulating HNRNPK expression using shRNAs. All these results, point to HNRNPK as a true modulator of PrPSc.

In Figure 4A and B, the use of POM1 and/or POM2 to detect PrPC / PrPres is confusing. POM2 is supposed to detect mostly full-length PrPC (Fig 4A top panel), but more than 3 glycoforms are detected. In Fig 4B, POM1 is used for PrPC but because it has a central epitope, it detects both PrPC and PrPSc.

Both antibodies are able to recognize both PrPC and PrPSc as it has been shown in many publications from the Aguzzi lab as well as other labs in the field. https://pubmed.ncbi.nlm.nih.gov/19060956/

Note also in Fig 4B, that DMSO alone seems to impact PrPC levels in PG127-infected hovS cells. This advocates again for a more quantitative analysis.

We have quantified the western blots using the DMSO control as standard value. As DMSO was used to dilute PSA, this should take into account potential effects coming from DMSO (Fig. 4D, Fig. 4F, Fig. 4H and Supplementary Fig. 4F).

• Psammaplysene A (PSA) is a pharmacological Hnrnpk binder. The authors used this molecule to further demonstrate that Hnrnpk is involved in prion propagation. I disagree with the author's conclusion that "PSA effect does seem to be limited when HNRNPK shRNAs are applied". In Fig S4D, 1µM PSA seems do decrease PrPres levels at similar levels whether the shRNA is applied or not. Again quantification and statistical analyses from several independent experiments would help supporting the authors conclusions.*

We assessed this point carefully by quantification of the western blots (Fig. 4H) and providing statistical data (Student’s t-test) from three experiments. As we see a threefold lower decrease of prions with and without Hnrnpk regulation when PSA is present, we concluded that the effect we see from PSA should be arising through Hnrnpk. However, we cannot conclusively delineate the effect of PSA, because Hnrnpk ablation is not possible due to essentiality of Hnnrpk. This has now been added to the discussion portion of our manuscript.

• The authors finally tested PSA on organotypic brain slices (in that case, they provide statistical results) and on flies infected with ovine PG137 prions. PSA administration significantly reduced the locomotor deficits prion-infected flies. The authors quantified the effects of PSA on prion accumulation in flies. Because the overall levels were not detectable by immunoblot, they used a cell-free assay termed RT-QuIC to address prion seeding activity in fly heads. I have specific comments about these experiments:
• Maybe I missed it, but I could not find which recombinant PrP is used in RT-QuIC assay.*

This information is provided in the M&M section of the manuscript at hand. The relevant section on P25 reads, where HaPrP23-231 refers to hamster PrP:

The reaction buffer of the RT-QuIC consisted of 1 mM EDTA (Life Technologies), 10 μM thioflavin T, 170 mM NaCl, and 1× PBS (incl. 130 mM NaCl) and HaPrP23-231 filtered using 100-kD centrifugal filters (Pall Nanosep OD100C34) at a concentration of 0.1 mg/ml.

In addition, we added this information to the main text as well.

- This is important as recombinant PrP self-polymerize after a period of time and here the authors have left the RT-QuIC assay running for unusually long period of times (RT-QuIC are stopped after 24h-48h).

For prions, long RT-QuIC experiments are often performed (also see: https://pubmed.ncbi.nlm.nih.gov/32598380/, https://journals.asm.org/doi/10.1128/mBio.02451-14, https://www.nature.com/articles/s41598-021-84527-9, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3458796/ and others).

In addition, this is controlled for in all experiments performed in the lab, as the prion-negative sample containing the same RT-QuIC substrate does not become positive after the entire duration of the assay (Fig. 5D).

- Instead of titrating prion seeding activity by endpoint titration, the authors quantified PSA activity by measuring the effect on another parameter of the RT-QuIC, the length of the lag phase before the conversion reaction is visible. While this is an interesting criterion, reduction of seeding activity must be shown to unequivocally demonstrate that PSA has delayed prion pathogenesis in flies.

Based on the data presented in the manuscript, we assessed prion pathogenesis in flies using a well-established climbing assay, demonstrating that treatment with PSA significantly improves locomotor behavior, which has been shown to be directly linked to prion levels and is known to have even greater sensitivity then the traditional mouse bioassay (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5998032/, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6113635/, https://link.springer.com/article/10.1007/s00441-022-03586-0).. The RT-QuIC represented here represents itself as a secondary read-out to the climbing assay, for which Lag-time quantification is used routinely (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3893511/, https://www.nature.com/articles/s41598-017-10922-w, https://journals.asm.org/doi/10.1128/mBio.02451-14, https://www.nature.com/articles/s41598-021-87295-8). Our results effectively highlight the overlap between the complementary read-outs.

- Can the authors exclude any interfering effect of PSA on the RT-QuIC reaction, given the amount of material used to seed the reaction (1:20 diluted head homogenates)?

We do not know how much PSA has reached the Drosophila brain, therefore, the experiment suggested by the reviewer cannot be tied to a 1:20 dilution. However, the concern of the reviewer is valid, and we therefore performed a spiking experiment of a prion positive sample using 1uM PSA (the highest amount used to treat cells, for which we saw a strong prion-reducing effect). We did not see an interference in the RT-QuIC signal due to PSA in the reaction. This has been incorporated into Figure 5D.

• could the authors comment on the fact that HNRNPK knock-out is not possible and that their siRNA and shRNA are not affecting the cell viability?*

To select hits during the screen process, we apply a viability filter, excluding siRNAs that reduce viability by more than 50% when compared to the non-targeting control siRNA (Supplementary Fig. 1F). For GT-1/7 cells we do not see any effect on viability of siRNA treatment after 96h. However, as downregulation of HNRNPK worsens the cytopathological vacuolation in the hovS model, as shown in Supp. Fig 4A, we do see an effect on cell fitness using both siRNA as well as shRNA. In addition, as knocking down HNRNPK will not lead to its complete loss, the remaining levels might be enough to sustain viability. Moreover, the longest knockdown experiment we performed is 7 days, we cannot exclude that longer exposure would have an impact on viability, but this question is not in the scope of the paper.

• In the discussion the authors do not discuss how Hnrnpk could impact prion propagation. This may deserve a comment as this protein is present in the nucleus. As PrPC has been also identified in this compartment, can this specific form be involved in prion pathogenesis?*

We additionally elaborated on potential ways of how Hnrnpk might impact prion propagation in the discussion, which includes potential nuclear PrPSc as well as with regards to our data obtained from the sequencing efforts shown in Fig. 4I. In addition, we investigated some functional targets of Hnrnpk how they are affected by PSA, which is now added to Supp. Fig 4G.

Reviewer #1 (Significance (Required)):

The QUIPPER method is a great conceptual and technological approach that could be applied to genome-wide analyses and screening for therapeutic molecules.

* The study will interest a general audience interested in neurodegenerative diseases linked to protein misfolding. There are commonalities in pathways and modifiers of the conversion. Further PrP has emerged as a receptor for alpha-synuclein (Parkinson disease) and A-beta peptides (Alzheimer's disease).

Expertise key words: prion diseases - prion pathogenesis in cell models*

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

Prions are protein-based infectious agents that underlie neurodegenerative disease. For prion diseases (e.g., mad cow disease), the infectious agent is the cellular prion protein (PrPc). It exists in a normal conformation and carries out its normal cellular function. However, when it becomes misfolded and aggregates it can adopt an altered conformation, referred to as the prion conformation, or PrPSc. PrPSc aggregates can template the conversion of other PrPc molecules into the PrPSc form. In this way the prions can propagate from one cell to the next and throughout an organism. Prion diseases are truly devastating and identifying ways of stopping prion propagation is of great interest. In this manuscript by Aguzzi and colleagues, the authors designed a way to screen for prion propagation modifiers in mammalian cells. They built a highly sensitive readout of PrPSc propagation and adapted it to a 384-well plate format in adherent cells. They then used this to perform a genomewide siRNA screen, looking for genes that increased or decreased PrPSc propagation when knocked down.

* They identified nearly 1,200 modulators of prion propagation and then subjected them to various validations and filtering to focus on only those hits that affected PrPSc but not PrPc (though hits that affect levels of PrPc could certainly be interesting). All this led to 40 genes (20 that increased and 20 that decreased prion propagation.*

* Among these 40, the authors focused on one hit, hnRNPK, an essential RNA-binding protein with diverse cellular functions. They provide evidence that reducing levels of hnRNPK leads to increase prion levels.*

* They next move to a marine compound called Psammaplysene A (PSA), which had previously been shown to have some neuroprotective properties and to be able to bind to hnRNPK. Because of the latter observation, the authors test if PSA can affect prion levels. They show that indeed treatment of their cell line prion infection model, or an organotypic slice model, or a fly model with PSA is sufficient to decrease prion levels.*

* The authors propose that PSA works to reduce prion levels by increasing the activity of hnRNPK and that this also implies a role of RNA (because hnRNPK is an RNA-binding protein) in prion propagation. * In a nutshell, in my opinion the design and execution of this genomewide screen is ingenious and has yielded a treasure trove of potential prion modifiers. The ability to distinguish between modifiers of Prpc and PrpSc is super powerful. However, the follow-up and focus on hnRNPK and its connections (which seem tenuous) to the marine compound PSA are incomplete and raise more questions than answers. In its present form, it is hard to assess the potential significance of hnRNPK in prion propagation. I have some comments and suggestions for the authors to consider.

* 1.To my eye, Fig. 4A looks like Hnrnpk siRNA leads to slightly increased levels of PrPc (detected with POM2 antibody) and this could explain the increase in PrPSc levels. Can the authors assess Prnp RNA levels and the effects of their siRNAs on Prnp expression? It would also be useful to provide quantification of immunoblots if possible.*

We quantified the western blots as mentioned in our response to reviewer 1. The quantifications are now provided for figures: Fig. 4A and Supplementary Fig. 4A, showing that the increase in prion levels is much stronger than that of PrPC. These confirm the results from the screen as seen in Fig. 3D. In addition, we would again like to point out that the use of shRNAs to knockdown HNRNPK did not yield the increase in PrPC levels aforementioned, as evident by Supplementary Fig. 4D which demonstrates a decrease of PrPC, despite increasing PrPSc levels. Moreover, we show quantification of RNA levels upon downregulation of Hnrnpk and with PSA, which show that downregulation of Hnrnpk via siRNAs indeed increases Prnp mRNA levels and that PSA does not change RNA levels of neither Hnrnpk nor Prnp (Fig. 4C).

• In Supplemental Fig. 4B it also looks like knocking down Hnrnpk results in decreased PrPc levels in this experiment and its not clear how robust the increase in PrPSc levels are. Quantification of these experiments, if possible, would be helpful.*

Please see response above. We now provide quantification to all western blots.

• The authors treat with PSA, which is supposed to bind to Hnrnpk. They state that this treatment does not affect PrPc levels but to my eye Supplemental Fig. 4C looks like highest doses of PSA cause a decrease in PrPc levels. Quantification of the immunoblots would also be useful here.*

Please see response above. We now provide quantification to all western blots and added a sentence to the manuscript.

• The authors use Hnrnpk knockdown along with PSA to test if the effects of PSA depend on Hnrnpk. They see PSA decreases PrPSc levels and that this is, to my eye, only slightly attenuated by Hnrnpk reduction. I interpret these results slightly different than the authors. To me, it seems that this result indicates that PSA's effects are (mostly) independent of Hnrnpk.*

Addressed in point 4 from reviewer one.

• In the original paper identifying PSA and hnRNPK physical interaction, RNA-binding was important. In the authors' assays, does Hnrnpk's effect on prions depend on RNA-binding? Specific mutations to the RNA-binding domains can be made to assess this.*

This is a very interesting point. We did try to obtain data to support this claim, however, due to the essentiality as well as tight control of Hnrnpk expression, we were not able to express different forms of Hnrnpk and acquire conclusive data. Therefore, it is currently being pursued how Hnrnpk might affect prion propagation in the scope of another publication.

• The genetic interaction in the vacuolation phenotype between Prnp and Hnrnpk that the authors report is very interesting (Supplemental Fig. 4A). It seems like this system and phenotype could be useful for the authors in exploring mechanisms by which HnrnpK is functioning.*

• *

We absolutely agree to the reviewer’s comment. As mentioned above a second publication is under way to investigate the mechanisms of Hnrnpk’s antiprion function, which is not in the scope of this study.

• The authors propose that PSA increases activity of Hnrnpk but does it change any Hnrnpk RNA targets from their RNA sequencing? Some functional readout of Hnrnpk function would be useful here to test this hypothesis.*

Although we do suspect RNA binding has an important role in the anti-prion function of Hnrnpk, we cannot exclude other modalities which Hnrnpk might be function through, such as DNA binding and protein-protein interactions. Therefore, to answer this question, a considerable effort that explores each of the potential of these modalities with regards to the anti-prion function of Hnrnpk would be needed. This extensive effort, however, is out of the scope of the manuscript at hand. However, we investigated the effect of PSA on some known functional targets of Hnrnpk (as suggested by the reviewer) from our sequencing efforts and added this analysis as Supplementary Fig. 4H to the manuscript. These results suggest that PSA leads to an increase of the expression of DNA targets of Hnrnpk, potentially suggesting a modality of action. Moreover, we amended the discussion with regards to potential pathways that might be yielding the effect seen as evidenced by the RNAseq data.

• In the Introduction, the authors mention two yeast papers in introducing the concept of using unicellular model organisms to perform modifier screens. The first paper (Outeiro and Lindquist, 2003) is a classic but does not contain a yeast screen. The other one does include a loss of function screen in yeast (for polyQ toxicity modifiers) but those results seems to be due to loss of the [RNQ+] prion from certain deletion strains instead of from specific roles of modifier genes, so that paper might not be the best exemplar of yeast modifier screens.*

We sincerely thank the reviewer for their careful readthrough of the manuscript, the portion that refers to the manuscripts as screens was amended and two new citations for appropriate yeast screens were added to the manuscript.

• The authors asked if any of their hits from their screen had human genetics connections to neurodegeneration. They mention one of their hits Dock3 right after saying that no hit reached statistical significance after multiple testing corrections. This seems a bit misleading since any time one makes a list of anything there will always be, by definition, one at the top of the list.*

We amended the wording to improve clarity of the manuscript.

• The authors perform RNA sequencing on prion infected cells that either had Hnrnpk siRNA or PSA and since these two treatments had opposite effects they looked for genes that went in the corresponding directions. They didn't find anything significant when looking for genes downregulated by Hnrnpk siRNA and upregulated by PSA. They did find glucose metabolism genes when looking in the opposite direction. The significance of this finding is unclear and the authors do not expand on it.*

Addressed in point 7 of reviewers 1 and 2, we expanded the discussion portion of the manuscript with regards to these results.

• To me, the data with PSA seem more robust than the Hnrnpk data and it seems that the authors are trying to perhaps over-fit them together. It is possible that PSA affects prion levels independent of Hnrnpk function. This would not dampen my enthusiasm at all for this finding and could be of interest to those in the prion field, in which the search for anti-prion compounds is of great interest.*

Upon statistical analysis of the result in Fig 4H, we see a three-fold decrease of PSA activity upon HNRNPK downregulation, suggesting PSA activity might be linked to HNRNPK. However, the reviewers point is well taken and we emphasized the value of understanding the function of PSA or mimicry of its effect as potential therapy in the future.

***Cross-commenting:**

All three reviewers seem to appreciate the novelty and impact of the new QUIPPER method the authors have developed to discover modifiers of prion propagation. All three reviewers also seem to be somewhat less convinced by the connection to hnRNPK, including how the compound PSA's anti-prion effects involve hnRNPK (or not).*

* In my opinion, this manuscript presents important and novel work and a really ingenious new method to study prion propagation, which will be broadly useful to the prion field. I feel that the hnRNPK data could be strengthened, especially with more quantitative analyses. The PSA treatment data are compelling but it seems that the effects might be independent of hnRNPK and that the authors are trying to force a connection which might not be there.*

* Reviewer #2 (Significance (Required)):*

* *** Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. ****

I have expertise in neurodegenerative disease, protein misfolding, yeast modifier screens, CRISPR modifier screens in human cells, and RNA-binding proteins. I have general knowledge about prions, including PrP, but I am not a prion expert.*

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

The authors conducted an arrayed RNAi-based genome-wide high-throughput screening of all protein-coding modifier genes that affect prion propagation in cultured cells (murine and human cell lines) using a novel quantitative high throughput QUIPPER assay that they developed. They identified 1191 genes, of which 40 selectively affect PrPSc. Half of the 40 genes seem to inhibit PrPSc (limiter) whereas the other half do the opposite (stabilizers). One of the strong limiters is Hnrnpk, is an essential small heterogeneous nuclear ribonucleoprotein that has been implicated in a few protein misfolding diseases. The biological relevance of the findings is demonstrated by the detection of previously reported modifier genes as well as thorough verification of Hnrnpk as an effective prion limiter that seems to be independent of the two prion strains or host species (mouse and human cell lines as well as Drosophila).

* The manuscript is very well written, the approach is novel, very well verified, and effective, the data are solid, and the main conclusions convincing.*

* Two issues need to be discussed.*

* Major comments:*

* First, some genes encoding proteins involved in PrP processing, such as ADAM10 and ADAM8, are known to affect PrPC levels, but they are not among the modifier genes identified. Based on Table 2, ADAM8 expression is very low in the GT-1/7 cells. This points to one of the caveats of the RNAi screening approach in that potential roles of low expressing genes in the cell lines used could be missed. Although it is beyond the scope of this manuscript, it would be helpful to add discussions on complimentary screening enhancing gene expression and the use of more cell lines that will allow identification of more modifiers.*

We thank the reviewer for their concern. The point regarding the screen being less sensitive for genes that are low-expressed in the cell line in question is valid. Upon advancing of the CRISPR-based technologies and the improvement of these technologies to be used in combination with prions, we see their value. We added a sentence to the discussion, talking about gene activation as a future alternative to perform a complimentary screen.

Second, the statement that PSA's anti-prion effect potentially arises through enhancing the activity of HNRNPK makes sense, but it is also possible that PSA can directly inhibit prion replication as well. It would be helpful to calculate the percentage of reduction in PrPSc by PSA treatment and the percentages compared between shNT and shHNRNK cells.

We thank the reviewer for the careful read through of the manuscript. The point was addressed for reviewer 1 point 4. In addition, if PSA is added to the RT-QuIC, it does not prevent aggregate formation, indicating that PSA is unlikely to directly inhibit prion replication, but rather depends on a cellular host-intrinsic molecule for its activity. However, we also elaborate more on the possibility of potential other mechanisms for Hnrnpk and PSA’s function on regulating prion levels in the discussion section of our manuscript.

Minor comments:

* First, Figure 1C shows that the relative intensity for RML CAD5 cell lysate infected cells is less than with PIPLC treated or PK treated, which seems to be the opposite of what is expected, because PIPLC or PK treatment should not increase infectivity. Please explain.*

We agree with the reviewer that the results were surprising. For the practicality of the screen, we wanted to show that the treatment does not eliminate the infectious species, which we were able to demonstrate. However, the increase of infectivity could stem from many different factors, e.g. the amount of duration of PK treatment might not harm but instead rather expose the infectious species, or PIPLC might remove cell surface molecules that could prevent infection of cells. However, as there are a plethora of possible scenarios and it was not relevant for the study at hand, we did not go into further detail.

Second, in Fig S1 e, the labels are too small to read. In Fig 3D, it would be easier to match the stabilizer or limiter genes with the corresponding Z score dots if the genes with a negative Z scores are labelled on the left side while genes with positive Z scores be labelled on the right side.

We amended the figures as per the reviewer’s suggestion.

Third, The following sentence on page 11 is confusing: "20 out of these 40 candidates reduce prion propagation upon silencing, and 20 candidates enhanced prion propagation, and henceforward are called stabilizers or limiters, respectively (Fig. 3D-E, Supplementary Table 1)." Did the author mean to say "....and 20 candidates enhanced prion propagation upon silencing, and hence..."?

We reworded the sentence according to the reviewer’s comment.

* Fourth, In the subheading "Hnrnpk expression limits of prion propagation in mouse and human cells", "of" should be deleted.*

We addressed this in the main manuscript file.

***Cross-commenting:**

I agree with Reviewer #2's assessment that more quantification will be helpful and the link between the effect of PSA treatment and hnRNPK can be strengthened. I want to stress that the knockdown data clearly shows the involvement of hnRNPK as a prion limiter in cultured cells. The question on PSA does affect the interpretation of the ex vivo and in vivo data.*

* The blot in Fig. S4c seems to show some decrease in PrPC levels in NBH-treated GT-1/7 cells. This blot needs to be quantified to confirm whether the PrPC level is changed by PSA treatments. Whether PSA directly inhibits prion replication can be relatively easily assessed in RT-QuIC reactions. Alternative to the use of PSA, RNAi-mediated hnRNPK knockdown can also be done on cultured tissue slices or in brain, but this will require a lot more time and efforts and may be too much to ask for in this manuscript.*

Quantifications for blots were added throughout the manuscript and the text was amended accordingly, and all the points mentioned have been addressed throughout this response letter.

Reviewer #3 (Significance (Required)):

* The findings are novel and very significant. They identified a large number of modifier genes, and established a solid foundation for future studies on prion modifier genes to study prion replication and pathogenesis and for novel therapies against prions and potentially some other protein misfolding diseases. HNRNPK seems to be good target for therapeutic intervention and PSA may be a good candidate for prion treatment. The novel QUIPPER assay can be used to screen for anti-prion compounds and potentially adapted to study other misfolding proteins associated with cells.*

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The authors conducted an arrayed RNAi-based genome-wide high-throughput screening of all protein-coding modifier genes that affect prion propagation in cultured cells (murine and human cell lines) using a novel quantitative high throughput QUIPPER assay that they developed. They identified 1191 genes, of which 40 selectively affect PrPSc. Half of the 40 genes seem to inhibit PrPSc (limiter) whereas the other half do the opposite (stabilizers). One of the strong limiters is Hnrnpk, is an essential small heterogeneous nuclear ribonucleoprotein that has been implicated in a few protein misfolding diseases. The biological relevance of the findings is demonstrated by the detection of previously reported modifier genes as well as thorough verification of Hnrnpk as an effective prion limiter that seems to be independent of the two prion strains or host species (mouse and human cell lines as well as Drosophila).

The manuscript is very well written, the approach is novel, very well verified, and effective, the data are solid, and the main conclusions convincing. Two issues need to be discussed.

Major comments:

First, some genes encoding proteins involved in PrP processing, such as ADAM10 and ADAM8, are known to affect PrPC levels, but they are not among the modifier genes identified. Based on Table 2, ADAM8 expression is very low in the GT-1/7 cells. This points to one of the caveats of the RNAi screening approach in that potential roles of low expressing genes in the cell lines used could be missed. Although it is beyond the scope of this manuscript, it would be helpful to add discussions on complimentary screening enhancing gene expression and the use of more cell lines that will allow identification of more modifiers.

Second, the statement that PSA's anti-prion effect potentially arises through enhancing the activity of HNRNPK makes sense, but it is also possible that PSA can directly inhibit prion replication as well. It would be helpful to calculate the percentage of reduction in PrPSc by PSA treatment and the percentages compared between shNT and shHNRNK cells.

Minor comments:

First, Figure 1C shows that the relative intensity for RML CAD5 cell lysate infected cells is less than with PIPLC treated or PK treated, which seems to be the opposite of what is expected, because PIPLC or PK treatment should not increase infectivity. Please explain.

Second, in Fig S1 e, the labels are too small to read. In Fig 3D, it would be easier to match the stabilizer or limiter genes with the corresponding Z score dots if the genes with a negative Z scores are labelled on the left side while genes with positive Z scores be labelled on the right side.

Third, The following sentence on page 11 is confusing: "20 out of these 40 candidates reduce prion propagation upon silencing, and 20 candidates enhanced prion propagation, and henceforward are called stabilizers or limiters, respectively (Fig. 3D-E, Supplementary Table 1)." Did the author mean to say "....and 20 candidates enhanced prion propagation upon silencing, and hence..."?

Fourth, In the subheading "Hnrnpk expression limits of prion propagation in mouse and human cells", "of" should be deleted.

Cross-commenting:

I agree with Reviewer #2's assessment that more quantification will be helpful and the link between the effect of PSA treatment and hnRNPK can be strengthened. I want to stress that the knockdown data clearly shows the involvement of hnRNPK as a prion limiter in cultured cells. The question on PSA does affect the interpretation of the ex vivo and in vivo data.

The blot in Fig. S4c seems to show some decrease in PrPC levels in NBH-treated GT-1/7 cells. This blot needs to be quantified to confirm whether the PrPC level is changed by PSA treatments. Whether PSA directly inhibits prion replication can be relatively easily assessed in RT-QuIC reactions. Alternative to the use of PSA, RNAi-mediated hnRNPK knockdown can also be done on cultured tissue slices or in brain, but this will require a lot more time and efforts and may be too much to ask for in this manuscript.

Significance

The findings are novel and very significant. They identified a large number of modifier genes, and established a solid foundation for future studies on prion modifier genes to study prion replication and pathogenesis and for novel therapies against prions and potentially some other protein misfolding diseases. HNRNPK seems to be good target for therapeutic intervention and PSA may be a good candidate for prion treatment. The novel QUIPPER assay can be used to screen for anti-prion compounds and potentially adapted to study other misfolding proteins associated with cells.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Prions are protein-based infectious agents that underlie neurodegenerative disease. For prion diseases (e.g., mad cow disease), the infectious agent is the cellular prion protein (PrPc). It exists in a normal conformation and carries out its normal cellular function. However, when it becomes misfolded and aggregates it can adopt an altered conformation, referred to as the prion conformation, or PrPSc. PrPSc aggregates can template the conversion of other PrPc molecules into the PrPSc form. In this way the prions can propagate from one cell to the next and throughout an organism. Prion diseases are truly devastating and identifying ways of stopping prion propagation is of great interest. In this manuscript by Aguzzi and colleagues, the authors designed a way to screen for prion propagation modifiers in mammalian cells. They built a highly sensitive readout of PrPSc propagation and adapted it to a 384-well plate format in adherent cells. They then used this to perform a genomewide siRNA screen, looking for genes that increased or decreased PrPSc propagation when knocked down.

They identified nearly 1,200 modulators of prion propagation and then subjected them to various validations and filtering to focus on only those hits that affected PrPSc but not PrPc (though hits that affect levels of PrPc could certainly be interesting). All this led to 40 genes (20 that increased and 20 that decreased prion propagation.

Among these 40, the authors focused on one hit, hnRNPK, an essential RNA-binding protein with diverse cellular functions. They provide evidence that reducing levels of hnRNPK leads to increase prion levels.

They next move to a marine compound called Psammaplysene A (PSA), which had previously been shown to have some neuroprotective properties and to be able to bind to hnRNPK. Because of the latter observation, the authors test if PSA can affect prion levels. They show that indeed treatment of their cell line prion infection model, or an organotypic slice model, or a fly model with PSA is sufficient to decrease prion levels.

The authors propose that PSA works to reduce prion levels by increasing the activity of hnRNPK and that this also implies a role of RNA (because hnRNPK is an RNA-binding protein) in prion propagation.

In a nutshell, in my opinion the design and execution of this genomewide screen is ingenious and has yielded a treasure trove of potential prion modifiers. The ability to distinguish between modifiers of Prpc and PrpSc is super powerful. However, the follow-up and focus on hnRNPK and its connections (which seem tenuous) to the marine compound PSA are incomplete and raise more questions than answers. In its present form, it is hard to assess the potential significance of hnRNPK in prion propagation. I have some comments and suggestions for the authors to consider.

1. To my eye, Fig. 4A looks like Hnrnpk siRNA leads to slightly increased levels of PrPc (detected with POM2 antibody) and this could explain the increase in PrPSc levels. Can the authors assess Prnp RNA levels and the effects of their siRNAs on Prnp expression? It would also be useful to provide quantification of immunoblots if possible.
2. In Supplemental Fig. 4B it also looks like knocking down Hnrnpk results in decreased PrPc levels in this experiment and its not clear how robust the increase in PrPSc levels are. Quantification of these experiments, if possible, would be helpful.
3. The authors treat with PSA, which is supposed to bind to Hnrnpk. They state that this treatment does not affect PrPc levels but to my eye Supplemental Fig. 4C looks like highest doses of PSA cause a decrease in PrPc levels. Quantification of the immunoblots would also be useful here.
4. The authors use Hnrnpk knockdown along with PSA to test if the effects of PSA depend on Hnrnpk. They see PSA decreases PrPSc levels and that this is, to my eye, only slightly attenuated by Hnrnpk reduction. I interpret these results slightly different than the authors. To me, it seems that this result indicates that PSA's effects are (mostly) independent of Hnrnpk.
5. In the original paper identifying PSA and hnRNPK physical interaction, RNA-binding was important. In the authors' assays, does Hnrnpk's effect on prions depend on RNA-binding? Specific mutations to the RNA-binding domains can be made to assess this.
6. The genetic interaction in the vacuolation phenotype between Prnp and Hnrnpk that the authors report is very interesting (Supplemental Fig. 4A). It seems like this system and phenotype could be useful for the authors in exploring mechanisms by which HnrnpK is functioning.
7. The authors propose that PSA increases activity of Hnrnpk but does it change any Hnrnpk RNA targets from their RNA sequencing? Some functional readout of Hnrnpk function would be useful here to test this hypothesis.
8. In the Introduction, the authors mention two yeast papers in introducing the concept of using unicellular model organisms to perform modifier screens. The first paper (Outeiro and Lindquist, 2003) is a classic but does not contain a yeast screen. The other one does include a loss of function screen in yeast (for polyQ toxicity modifiers) but those results seems to be due to loss of the [RNQ+] prion from certain deletion strains instead of from specific roles of modifier genes, so that paper might not be the best exemplar of yeast modifier screens.
9. The authors asked if any of their hits from their screen had human genetics connections to neurodegeneration. They mention one of their hits Dock3 right after saying that no hit reached statistical significance after multiple testing corrections. This seems a bit misleading since any time one makes a list of anything there will always be, by definition, one at the top of the list.
10. The authors perform RNA sequencing on prion infected cells that either had Hnrnpk siRNA or PSA and since these two treatments had opposite effects they looked for genes that went in the corresponding directions. They didn't find anything significant when looking for genes downregulated by Hnrnpk siRNA and upregulated by PSA. They did find glucose metabolism genes when looking in the opposite direction. The significance of this finding is unclear and the authors do not expand on it.
11. To me, the data with PSA seem more robust than the Hnrnpk data and it seems that the authors are trying to perhaps over-fit them together. It is possible that PSA affects prion levels independent of Hnrnpk function. This would not dampen my enthusiasm at all for this finding and could be of interest to those in the prion field, in which the search for anti-prion compounds is of great interest.

Cross-commenting:

All three reviewers seem to appreciate the novelty and impact of the new QUIPPER method the authors have developed to discover modifiers of prion propagation. All three reviewers also seem to be somewhat less convinced by the connection to hnRNPK, including how the compound PSA's anti-prion effects involve hnRNPK (or not).

In my opinion, this manuscript presents important and novel work and a really ingenious new method to study prion propagation, which will be broadly useful to the prion field. I feel that the hnRNPK data could be strengthened, especially with more quantitative analyses. The PSA treatment data are compelling but it seems that the effects might be independent of hnRNPK and that the authors are trying to force a connection which might not be there.

Significance

Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

I have expertise in neurodegenerative disease, protein misfolding, yeast modifier screens, CRISPR modifier screens in human cells, and RNA-binding proteins. I have general knowledge about prions, including PrP, but I am not a prion expert.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Avar et al report on the development of a high-throughput method to screen modifiers of prion replication in cell lines using a genome-wide siRNA library. They identified a number of hits and further studied one candidate, the ribonucleoprotein Hnrnpk. The authors convincingly show the interest of their method. However, the claims that the ribonucleoprotein Hnrnpk impact prion propagation need to be more quantitatively and statistically substantiated.

1. A large part of the manuscript is dedicated to the validation of the high-throughput assay (called QUIPPER). QUIPPER is made in 384-plates and provides great technological improvement. It works with different prion-permissive cell lines and different prion strains. QUIPPER is an antibody-FRET-based assay that detects a specific population of PrPSc that resists phospholipase C (PIPLC) treatment. Historically, PIPLC has been shown to cleave cell surface PrPC while preserving PrPSc (which is endocytic or inaccessible). I would recommend that the authors quantify the proportion of PIPLC-resistant PrPSc (PrPPIPLC) versus total PrPSc in their different models. First, PrPPIPLC proportion may be cell and strain dependent. Second and most importantly, as siRNA effects are studied using PrPPIPLC as readout, it is crucial to know if this form is a bona fide surrogate of PrPSc and infectivity or only a specific, subcellular, potentially minor form of PrPSc. This is particularly important as the effects of Hnrnpk knock-down in QUIPPER and western blot sounds discordant; in QUIPPER, the effects are strong (> 5-fold) while by western blot, the effects are much more modest (< 2-fold). Technically, this is quite easy as it necessitates, after PIPLC treatment, the quantification of PrPSc in the supernatant versus PrPSc in the cell pellet. In Fig. 1C, the authors show that PrPPIPLC is infectious in a cell-scrapie assay. Using this approach, they could also quantify the infectivity of these species relative to the total infectivity content.
2. The authors identified a list of prion modifiers candidate. Surprisingly, the authors did not perform a pathways analysis to identify potential pathways that could impact prion propagation.
3. The authors then studied in more details one hit, the ribonucleoprotein Hnrnpk. They studied the impact of Hnrnpk knock-down on PrPC and PrPres levels in different cell lines. These data (Fig 4 and Fig S4) lack quantitative (on a higher number of wells) and statistical analyses. The western blot that are shown suggest that PrPC levels are slightly increased by the siRNA and that the increase in PrPres levels is modest, barely significant given the western blot method. Same comment after PSA treatment, at least in PG127-infected hovS cells. In Figure 4A and B, the use of POM1 and/or POM2 to detect PrPC / PrPres is confusing. POM2 is supposed to detect mostly full-length PrPC (Fig 4A top panel), but more than 3 glycoforms are detected. In Fig 4B, POM1 is used for PrPC but because it has a central epitope, it detects both PrPC and PrPSc.

Note also in Fig 4B, that DMSO alone seems to impact PrPC levels in PG127-infected hovS cells. This advocates again for a more quantitative analysis. 4. Psammaplysene A (PSA) is a pharmacological Hnrnpk binder. The authors used this molecule to further demonstrate that Hnrnpk is involved in prion propagation. I disagree with the author's conclusion that "PSA effect does seem to be limited when HNRNPK shRNAs are applied". In Fig S4D, 1µM PSA seems do decrease PrPres levels at similar levels whether the shRNA is applied or not. Again quantification and statistical analyses from several independent experiments would help supporting the authors conclusions. 5. The authors finally tested PSA on organotypic brain slices (in that case, they provide statistical results) and on flies infected with ovine PG137 prions. PSA administration significantly reduced the locomotor deficits prion-infected flies. The authors quantified the effects of PSA on prion accumulation in flies. Because the overall levels were not detectable by immunoblot, they used a cell-free assay termed RT-QuIC to address prion seeding activity in fly heads. I have specific comments about these experiments: - Maybe I missed it, but I could not find which recombinant PrP is used in RT-QuIC assay. - This is important as recombinant PrP self-polymerize after a period of time and here the authors have left the RT-QuIC assay running for unusually long period of times (RT-QuIC are stopped after 24h-48h). - Instead of titrating prion seeding activity by endpoint titration, the authors quantified PSA activity by measuring the effect on another parameter of the RT-QuIC, the length of the lag phase before the conversion reaction is visible. While this is an interesting criterion, reduction of seeding activity must be shown to unequivocally demonstrate that PSA has delayed prion pathogenesis in flies. - Can the authors exclude any interfering effect of PSA on the RT-QuIC reaction, given the amount of material used to seed the reaction (1:20 diluted head homogenates)? 6. could the authors comment on the fact that HNRNPK knock-out is not possible and that their siRNA and shRNA are not affecting the cell viability? 7. In the discussion the authors do not discuss how Hnrnpk could impact prion propagation. This may deserve a comment as this protein is present in the nucleus. As PrPC has been also identified in this compartment, can this specific form be involved in prion pathogenesis?

Significance

The QUIPPER method is a great conceptual and technological approach that could be applied to genome-wide analyses and screening for therapeutic molecules.

The study will interest a general audience interested in neurodegenerative diseases linked to protein misfolding. There are commonalities in pathways and modifiers of the conversion. Further PrP has emerged as a receptor for alpha-synuclein (Parkinson disease) and A-beta peptides (Alzheimer's disease).

Expertise key words: prion diseases - prion pathogenesis in cell models

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

Learn more at Review Commons

---

Reply to the reviewers

All the Reviewer’s comments are reproduced below, with our responses interspersed in [[brackets]]. Citations from the revised manuscript are included in “quotation marks”. The website accepts input only as plain text. Consequently, we had to transform the mathematical expressions into plain text. We apologize for the reduced readability.

Reviewer #1

1) The authors state that: "the conductance density mediated by the expression of the mutant was 2.5 times smaller than the wild type, although we transfected the same amount of plasmid DNA (Fig. 2E). Assuming that protein expression is independent of the mutation, the observation suggested that the unitary proton flux ratio RC of wild type to mutant channel was equal to 2.5" (lines 82‐85).

Macroscopic conductance (G) depends on channel number (N), microscopic or unitary conductance (γ), and open probability (PO) by G=N γ PO. The authors assume that the level of WT and D174A mutant protein expression on plasma membrane, which determines N, are equal; however, this critical assumption does not appear to have been tested.

The fact that conductance density (nS/pF) is plotted in Fig. 2E does not alter this caveat because this procedure normalizes the data only for cell surface area (i.e., size). The authors' conclude that "The conductance density relationship (Fig. 2E) compares the maximal conduction of both constructs; this is the fully open channel (open probability ≈ 1)"(lines 87‐88). However, neither raw currents nor G‐V data are shown. Typically, currents measured at large, near‐saturating PO are used to compare the relative conductances of WT and mutant ion channels. The currents shown in Fig. 2A and 2B exhibit prominent 'droop' at even modest depolarizing potentials (+10 mV for D174A and +30 mV for WT), indicating that the proton gradient has been substantially perturbed by the flow of ge depolarizing voltages needed to drive channels to near‐maximal PO. Furthermore, there is no evidence that maximal PO itself is also not different in WT and D174A channels. Indeed, maximal PO for native Hv1 channels measured using variance analysis is reported by significantly smaller than 1.0, and assuming that PO = 1.0 for either WT or D174A is therefore not well supported. Maximal could be altered by the D174A mutation, which has a clear and strong effect on channel gating evidenced by the large (‐70 mV) negative shift in threshold potential reported both here and previously in the literature. Effects of mutations on maximal PO due to altered gating behavior could be separate and distinct from any change in plasma membrane channel number (N). 3 Lastly, because D174A channels have a much higher PO than WT at 0 mV, the mutant will necessarily conduct inward proton currents at the physiological resting membrane potential (RMP) in tsa‐201 cells (perhaps ‐30 mV?). Inwardly directed proton currents will therefore cause intracellular acidification under resting conditions.

The constitutive acid load in cells expressing D174A, but not WT, is likely to have a variety of physiological consequences, including decreased protein expression or plasma membrane targeting of D174A. There is evidence that another constitutively open Hv1 mutant (R205H) also generates smaller currents macroscopic conductance than WT, and this phenomenon is likely to result from decreased cell surface expression. To conclude that the microscopic conductances of WT and D174A are unequal, the authors must demonstrate that N is not different. The authors' conclusion that D174A "conducts protons at a lower rate" (line 89) is therefore not well supported by the experimental data.

[[

We toned down our conclusions from the experiments to accommodate the reviewer's criticism: (page 4): " Consequently, the mutant channel is nearly fully open (Fig. 2D), readily seen when the membrane potential is 0 mV and external voltage is absent. The high open probability of the D174 mutant under symmetrical pH conditions is readily seen in the tail current amplitude reaching a quasi-saturation (Fig. 2A). The resulting outward currents have a higher amplitude in the wild-type (Fig. 2A+B). Interestingly, the conductance density mediated by the expression of the mutant was 2.5 times smaller than the wild type, although we transfected the same amount of plasmid DNA (Fig. 2E). Our observation suggests a reduced flux through the mutant if we assume that protein abundance in the plasma membrane is independent of the mutation."]]

2) The authors indirectly measure apparent proton flux rates (λD) in LUVs containing WT and D174A mutant Hv1 channels using a fluorescence‐based approach, and conclude that λD is 2.4 times smaller for D174A than WT. However, the method for estimating λD is not performed under voltage clamp, and the driving force for proton current is neither known nor measured.

[[

The reviewer is mistaken. The method for estimating λD is performed under voltage clamp, and the driving force for proton current is known.

Page 6: “To obtain λD, we encapsulated c_k^i=150 mM KCl in the HV1 containing large unilamellar vesicles (LUVs) and exposed these vesicles to a buffer with a K+ concentration c_k^o= 3 mM. The addition of valinomycin facilitated K+ efflux, thereby inducing a membrane potential, ψ. ψ constituted the driving force for H+ uptake. It can be calculated according to the Goldman equation:

ψ = -RT/F ln ((c_k^i+(P_H/P_K ) c_H^i)/(c_k^o+(P_H/P_K ) c_H^o ))

(1)

The ratio of the HV1 mediated proton permeability P_H to the valinomycin-mediated potassium permeability P_K is always smaller than 0.04. We base our conclusion on the observation that the CCCP-mediated proton permeability represents an upper limit for P_H since CCCP always induces a faster vesicular proton uptake than HV1 (Fig. 3). Accordingly, the maximum value of P_H/P_K can be estimated as the ratio of valinomycin to CCCP conductivities. The respective values are equal to 1.6 10-3 Ω-1 cm2 [1] and 4 10-6 Ω-1 cm-2 [2]. At pH 7.5, we find c_H^o=10^(-7.5) M, i.e., c_k^o ≫ (P_H/P_K )c_H^o. Similarily, c_k^I ≫ (P_H/P_K ) c_H^i for a broad range of intravesicular pH. With these simplifications, Eq. 1 transforms into the Nernst equation yielding:

ψ = -RT/F ln (c_k^i)/(c_k^o )=-100 mV

(2)

ψ of such size may decrease intravesicular pH by nearly two units. Such acidification does not violate c_k^i ≫ (P_H/P_K ) c_H^i so that ψ remains constant throughout the experiment. That is, the vesicle experiments proceed under voltage clamp conditions. The simple explanation is that, due to the small proton concentration and the limited buffer capacity, the K+ conductance exceeds H+ conductance under all conditions. The conclusion is in line with simulations (32), confirming that the membrane potential is driven very near the Nernst potential for K+.”]]

The authors state that "Transmembrane voltage constituted the driving force for proton uptake into LUVs (Figure M). It resulted from facilitated K+ efflux out of the vesicles (30)", (lines 261‐262), but this voltage is unknown and not likely to equal the Nernst equilibrium potential for K+ once Hv1 channels begin to open.

[[

The reviewer is mistaken. The voltage is known (see the equations above). The opening of the HV1 channels does not alter the potential because c_k^o ≫ (P_H/P_K ) c_H^o and c_k^i ≫ (P_H/P_K ) c_H^i for a broad range of intravesicular pH (see above).]]

Once Hv1 channels begin to open, intra‐lumenal pH (pHi) will necessarily occur during the experiment. Such changes are likely exacerbated by a) the low proton buffering capacity of the system (5 mM HEPES) and b) the absence of any counter‐charge pathway to balance the effect of proton charge movement on the membrane potential.

[[

Vesicle acidification occurs. It signifies the presence of functional proton channels. Nevertheless, the membrane potential does not change (see Equation 1 above). The statement b) is not correct because the outward K+ movement counters the inward-directed proton charge movement.]]

Given the small volume of LUVs, even a relatively modest difference in either membrane potential or pHi could substantially alter the driving force for proton movement. Together, these factors are highly likely to result in a rapid and potentially large change in the driving force for proton flux.

[[

As outlined above, membrane potential stays invariant. Vesicle acidification changes the driving force for proton flux. The steady state is reached when the electrochemical potentials for protons on the two sides of the membrane are equal to each other.]]

Driving force changes may also be different for WT and D174A because their relative PO may be different under the experimental conditions used here. Because D174A activates at much more negative voltages, it is likely to open more quickly and to a higher PO than WT at early times after depolarization is initiated by addition of valinomycin (Fig. 3A). This fact will likely result in a larger initial inward current being carried by D174A than WT channels. The result would be a more rapid acidification of LUVs by D174A.

[[

The reviewer is mistaken. Assuming a transport rate of 20,000 potassium ions per second (G. Stark, B. Ketterer, R. Benz and P. Läuger; Biophys. J. 1971 Vol. 11 Pages 981-981) and a membrane capacity of 1 μF cm-2, it takes valinomycin about 10 ms to drive the vesicular potential to near Nernst values. Activation of the proton channel is at least 10 times slower. Thus, both mutant channel and wild type channel may open at roughly the same instant. The driving force is sufficient to open both channels to the same probability.]]

The experimental data in Fig. 3A are consistent with the expectation that the proton gradient and driving force more rapidly approach equilibrium for D174A than WT channels: the apparent rate of AMCA fluorescence change is slower in D174A. Although the authors correctly interpret the experimental data to mean that the apparent λD is slower for D174A, they do not rule out the artifactual explanation for the measured differences. Indeed, the observation in Fig. 3A that AMCA fluorescence change eventually reaches a plateau and is not affected by CCCP means that the proton gradient has become exhausted during the experiment, and directly demonstrates that the proton driving force is uncontrolled under the current experimental conditions.

[[

The reviewer's interpretation of our results is flawed. Instead of becoming exhausted, the proton gradient builds up during the experiment. Initially, extravesicular and intravesicular pH values are equal to each other. Valinomycin-mediated K+ efflux results in a membrane potential that drives Hv1-mediated H+ influx.

Page 8: “The number NC of reconstituted HV1 dimers per vesicle determines the acidification rate λ, i.e., the time that elapses before reaching the steady state. The final intraluminal pH is independent of NC. Similarly, CCCP addition in the steady state does not change the intraluminal pH of HV1-containing vesicles. But CCCP will affect the intraluminal pH of vesicles deprived of HV1 since H+ background permeability is too small to allow vesicle acidification within the time allotted for the experiment. Consequently, only HV1-free vesicles will acidify upon CCCP addition. That is, CCCP addition allows estimating the fraction of vesicles deprived of HV1.”]]

In contrast to the authors' statement that "Our experiments with the purified and reconstituted channels corroborated the conclusion (Fig. 3A)", (lines 92‐93) it is not clear that unitary proton flux rates/unitary conductances are actually different in WT and D174A.

[[

The reviewer is mistaken. Since we measured under voltage clamp conditions, ensured rapid installment of the membrane potential, and selected a potential large enough to allow for the same open probability of wild-type and mutant channels, the measured transport rates, λ, are valid. Moreover, we determined the number of HV1 channels per vesicle and thus calculated the transport rate of an individual channel, λD. Since λD is different for WT and D174A, the unitary proton flux rates/unitary conductances are actually different in the wild type and mutant.]]

3) The presumed differences in unitary conductances (i.e., 'transport rate') between WT and D174A are used to estimate Arrhenius activation energies (Ea): ("The difference in measures transport rates allows a rough estimation of the Arrhenius 128 activation energy Ea for HV1‐mediated proton flow. It amounts to 40 kJ/mol for the wild type and 23 kJ for the mutant. Thus, Ea exceeds the corresponding 15 kJ/mol barrier measured for gramicidin A (32, 33)", (lines 128‐130). The method for determining Ea in the current work is not well‐described. In Ref. 32, the authors estimate Arrhenius activation energy (Ea = 20 kJ/mol) for gramicidin D (not gramicidin A) from the slope of a line fit to measurements of currents at various temperatures. Here, the authors measure AMCA fluorescence decay rates at 4 °C and 23 °C and observe a similar temperature‐dependent difference in WT and D174A (Fig. S2). Given that the data indicate that WT and D174A are similarly temperature‐dependent, it is unclear how the authors arrive at different Ea values. The authors' conclusion that "The increment in Ea suggests that the transport mechanism may be different from a pure Grotthuss type, where the proton uses an uninterrupted water wire to cross the membrane", (lines 131‐133) therefore does not appear to be well‐supported.

[[

We removed both the calculation and discussion of activation energies. Knowledge and discussion of activation energies distract from the scope of the manuscript. We show the experiments at different temperatures solely to demonstrate that Hv1 and D174A facilitate proton transport at a decreased temperature where the background conductivity of the lipid bilayer to water is small.]]

4) The authors report no difference in water permeability in WT vs. D174A (Fig. 5 and S1) and interpret the results to mean that proton currents are not associated with measurable bulk water flow. A similar conclusion was reached for native Hv1 channels using deuterium substitution (DeCoursey & Cherny, 1997).

[[

The comment of the reviewer is misleading:

• Equal water permeabilities of WT and D174A would not exclude an association between proton currents and water flow. Accordingly, our manuscript does not contain the stipulated interpretation.
• DeCoursey & Cherny (1997) did not evaluate bulk water flow through proton channels. They compared D+ and H+ currents across the plasma membrane of rat alveolar epithelial cells. Page 2: “Comparing deuterium ion and proton currents through the plasma membrane of rat alveolar epithelial cells, DeCoursey & Cherny (22) found an isotope effect exceeding that for hydrogen bond cleavage in bulk water. It suggested the involvement of an amino acid side chain in proton conduction (22). Alternatively, altered properties of confined water could have been responsible for the higher isotope effect.”]]

However, the absence of bulk water flow does not itself rule out the possibility that 'trapped' waters within the Hv1 pore do not themselves carry the measured proton current. If intra‐pore water molecules are tethered by hydrogen bonds with protein atoms, they may not move when Hv1 channels open.

[[

The reviewer’s comment contains one misinterpretation and one unfounded statement:

1. We never stated that 'trapped' waters within the Hv1 pore do not themselves carry the measured proton current. On the contrary, we envisioned the trapped waters delivering the protons to one or more titratable amino acid side chains and accepting the protons from them.
2. The reviewer’s view that intra‐pore water molecules tethered by hydrogen bonds with protein atoms may not move when Hv1 channels open is a misconception. Page 12 bottom: “The contrasting opinion that instead of a channel obstruction hydrogen bonds may immobilize the pore water (19) is not convincing. First, the lifetime of a hydrogen bond is in the ps range while HV1’s mean open time exceeds 100 ms (41). Thus, hydrogen bonds may break more than 1011 times during the open state, rendering them unfit for tethering intraluminal water molecules. Second, the effect of hydrogen bonds between water molecules and pore residues is limited to decreased water mobility in narrow channels (23). Their number, NH, allows for predicting pf (26). Specifically, every H-bond donating or receiving pore-lining residue contributes an average increment ΔΔG╪ of 0.1 kcal/mol to the Gibbs free energy of activation ΔG╪ (24). Equation (1) allows the calculation of ΔG╪:

ΔG╪= N_H ΔΔG╪ + ΔΔG╪_i (13)

where ΔΔG╪_i = 2 kcal/mol (24). Since N_H = 6 (Fig. S1) in the open HV1 conformation, Eq. 1 predicts ΔG╪ = 2.6 kcal/mol. Eq. (2) allows calculating HV1’s pf from this value (42):

p_f = v_0 v_w exp(-ΔG╪/RT) (14)

where vw = 3 × 10−23 cm3 is the volume of one water molecule and ν0 is the universal attempt frequency, ν0 = kB∙T/h ≈ 6.2 × 1012 s−1 at room temperature (kB is Boltzmann’s and h is Planck’s constant).”]]

Proton transfer through a hydrogen‐bonded network of waters requires only that the electronic structure of the network be rearranged during proton transfer; water is not required. As in the previous study (DeCoursey & Cherny, 1997), the lack of water flux reported here demonstrates seems to reinforce the notion that H+ moves separately from its waters of hydration (i.e., hydronium, H3O+, is not the permeant species) and does not necessarily imply information about the mechanism of proton transfer (i.e., side chain ionization vs. Grotthuss‐type transfer in a water‐wire).

[[

The reviewer is mixing two unrelated issues. Of course, proton transport may be separated from mass transfer. Yet, charge transfer may or may not include one or several titratable amino acid side chains. If proton side chain ionization is not involved in proton transfer, a water wire must exist that connects the aqueous solutions on both sides of the membrane. In this case, an osmotic gradient will drive water molecules through the open channel. Since we did not observe such water flux, we conclude that the water wire is interrupted by at least one side chain. Thus, our experiments imply information about the mechanism of proton transfer.]]

The authors state that: 1) "every H‐bond donating or receiving pore‐lining residue would have contributed an increment ΔΔ𝐺‡ of 0.1 kcal/mol to the Gibbs free energy of activation Δ𝐺‡ (25)" (lines 145‐147), and 2) calculating NH from this Δ𝐺‡ allows estimation of the channel's unitary water permeability (Eqn. 2). Although hydrogen bonding patterns will undoubtedly alter the free energy for channel activation, this is not the same free energy change as that for proton transfer.

[[

The reviewer's remark is in line with the previous and the current versions of our manuscript.]]

Hv1 gating involves conformational changes that are both voltage and Δ pH-dependent, and the D174A mutation is known to alter the voltage dependence of gating (Fig. 2 and previous studies). The effect of D174A on Hv1 unitary conductance, however, is speculated but not unambiguous (see above).

[[

Our experiments unambiguously demonstrate the effect of D174A on Hv1 unitary conductance. The interpretation of the experiments is straightforward – there is no speculation involved. The contrasting opinion of the reviewer rests on his misinterpretations of (i) our measurements of proton transport rate λD for wild-type and mutant (see above) and the CCCP-effect (see above).]]

In the absence of definitive experimental data showing differences in the unitary conductance of WT vs. D174A, the authors' assumption that water permeability would be strongly temperature‐dependent (lines 154‐160) seems premature and their ensuing conclusion tenuous: "pore residues interrupt the HV1 spanning water wire, trapping the water molecules inside the HV1 channel. In contrast to water, protons cross the pore by hopping from one acidic residue to another through one or more bridging water molecules (Fig. 6)" (lines 161‐164).

[[

The reviewer chooses to misinterpret our lines. We did not assert that water permeability through the Hv1 channel would be strongly temperature‐dependent. We referred to the well-known fact that there is a strong temperature dependence of lipid bilayer water permeability - in contrast to the tiny effect of temperature on the water permeation across aqueous channels.

Page 11, bottom: “Considering the stark dependence of the activation energy for background water flow across lipid bilayers (24), we repeated the experiments at a decreased temperature of 4°C. Thanks to the low background water permeability at 4°C, even tiny contributions of HV1 to Pf should be detectable. Yet, the channels did not contribute to the water flow through the vesicular membrane even though channel water permeability but weakly depends on temperature (24).”]]

Furthermore, the authors calculate the number of hydrogen bonds (NH) that pore waters could form with pore lining residues based on an X‐ray structure of a chimeric proton channel protein (pdb: 3WKV) that is: a) manifests discontinuous transmembrane water density and is known to represent a non‐conductive conformation, b) contains residues from Ci‐VSP in the critical S2‐S3 linker that form part of the proton transfer pathway, and c) exhibits structural features (i.e., highly conserved ionizable residues such as D185 and R205, which like D174 are reported to dramatically alter Hv1 gating, are packed into a solvent‐free crevice) that are inconsistent with physiological function. Given that all Hv1 ionizable mutant combinations tested so far (the sole exception of D112V ‐ other nonionizable substitutions at D112 are tolerated) remain functional (Musset, Smith et al., 2011, Ramsey, Mokrab et al., 2010), the identities of water‐interacting residues speculative.

[[

We substituted the X‐ray structure of the chimeric proton channel protein for the AlphaFold structure. We now provide views of the open and closed conformations in the Supplement based on the homology structure (13). Microsecond-long molecular dynamics simulations have optimized the latter.

The experimental observation of mutants’ functionality (with the sole exception of D112V) supports our view that proton transfer occurs through a hydrogen‐bonded network of waters that is only once (at D112) interrupted by an amino acid side chain. The nature of the amino acids interacting with the proton transferring water molecules is of little importance.]]

Interpreting differences in the calculated NH based on pdb: 3WKV therefore seems unlikely to reveal fundamentally important insights into Hv1 function. The author's conclusion that "The observation rules out the formation of an uninterrupted water chain spanning the open channel from the aqueous solution at one side of the membrane to the other. NH would have governed water mobility if such a water wire had formed (24)", (lines 143‐145) therefore does not appear to be strongly supported.

[[

We did not base our conclusion of an obstructed water pathway on the analysis of structural models. In contrast, the conclusion is the result of our experiments. The structural models permitted the prediction of the expected water permeability. Depending on the model and the channel conformation, we find NH values between six and 16. All of these NH values translate into water permeabilities exceeding gramicidin’s water permeability. Thus, we would have been able to detect the water flux through an unobstructed proton channel.]]

Reviewer #2:

Summary: Voltage‐gated proton channels are peculiar members of the voltage‐gated ion channel family due to their absence of canonical pore. Instead, protons permeate through their voltage‐sensing domain. The mechanisms of proton permeation in Hv1 channels are still unclear, with currently two competing hypotheses: (i) hopping through titrable residues within the protein; or (ii) via Grotthuss mechanism involving proton jumping through a continuous water wire. So far, these hypotheses were only tackled by computation. The authors therefore aimed to experimentally test the two hypotheses. To do so, the authors measured the transport rates of protons and water through wild‐type and mutant D174A Hv1 reconstituted in lipid vesicles. Overall, the presented data are convincing and support their conclusion that proton conduction through the channel is not solely mediated by water transport. However, there are several aspects of the paper that I did not understand and would require clarification.

[[

We thank the reviewer for the positive evaluation.]]

Major comments: My major concern is about the relevance of using the D174A mutant. The authors explain at the beginning of the paper that Hv1‐D174A is open at 0 mV, which allows measuring proton flux in systems in which voltage cannot be controlled. However, it seems from the proton flux experiments that wild‐type Hv1 can conduct protons perfectly well in the used experimental paradigm. So why test a mutant? It is actually not clear why wild‐type Hv1 can conduct protons in the proton conduction assay.

[[

We introduced the D174A mutation to measure water flux in a setting where the membrane potential is zero. We only performed the proton flux measurements to show that our reconstituted HV1 channels are functional. HV1 can conduct protons because we establish a transmembrane potential in the proton conduction assay. That is, only initially, extravesicular and intravesicular pH values are equal. Valinomycin addition results in a K+ efflux that, in turn, generates a membrane potential. This potential drives the HV1-mediated H+ influx.]]

The authors should clearly state the trans‐membrane potential created by the K+ gradient across the vesicle, as well as the pH inside and outside the vesicle, and related these conditions to their electrophysiology data to give us an idea of the open probability of wild‐type Hv1 in the conditions used in the proton conduction assays. This is critical to be able to compare the relative rates of proton transport between the wild‐type and the mutant.

[[Page 6, bottom:

" ...we encapsulated c_k^i=150 mM KCl in the HV1 containing large unilamellar vesicles (LUVs) and exposed these vesicles to a buffer with a K+ concentration c_k^o= 3 mM. The addition of valinomycin facilitated K+ efflux, thereby inducing a membrane potential, ψ. ψ constituted the driving force for H+ uptake. It can be calculated according to the Goldman equation:

ψ = -RT/F ln ((c_k^i+(P_H/P_K ) c_H^i)/(c_k^o+(P_H/P_K ) c_H^o ))

(1)

The ratio of the HV1 mediated proton permeability P_H to the valinomycin-mediated potassium permeability P_K is always smaller than 0.04. We base our conclusion on the observation that the CCCP-mediated proton permeability represents an upper limit for P_H since CCCP always induces a faster vesicular proton uptake than HV1 (Fig. 3). Accordingly, the maximum value of P_H/P_K can be estimated as the ratio of valinomycin to CCCP conductivities. The respective values are equal to 1.6 10-3 Ω-1 cm2 [1] and 4 10-6 Ω-1 cm-2 [2]. At pH 7.5, we find c_H^o=10^(-7.5) M, i.e., c_k^o ≫ (P_H/P_K )c_H^o. Similarily, c_k^I ≫ (P_H/P_K ) c_H^i for a broad range of intravesicular pH. With these simplifications, Eq. 1 transforms into the Nernst equation yielding:

ψ = -RT/F ln (c_k^i)/(c_k^o )=-100 mV

(2)

ψ of such size may decrease intravesicular pH by nearly two units.

Such acidification does not violate so that remains constant throughout the experiment. That is, the vesicle experiments proceed under voltage clamp conditions. The simple explanation is that, due to the small proton concentration and the limited buffer capacity, the K+ conductance exceeds H+ conductance under all conditions. The conclusion is in line with simulations (32), confirming that the membrane potential is driven very near the Nernst potential for K+.”]]

Similarly, the buffers and pH used for the water transport assay are not explicitly mentioned. Are they the same as for the proton transport assay or are the buffers inside and outside the vesicle symmetrical?

[[

We added the information about buffers and pH used to the legend. Except for 150 mM sucrose, the internal and external solutions were identical: 150 mM KCl, 5 mM HEPES (pH 7.5), and 0.5 mM EGTA.]]

Finally, in the introduction the authors base their assumptions about water transport on an X‐ray structure of Hv1 in a closed conformation (3WKV). I do not think it is relevant to study permeation, which in theory should only happen in an open state. If the authors want to make assumptions about the number of hydrogen bonds in the pore and how many water molecules are in the pore (and I don't think they need to do it), they should rather base their assumptions on the computational models of Hv1 open state.

[[

We thank the reviewer for the advice. We added a figure to the Supplement. It shows Hv1 models from long-timescale molecular dynamics simulations (Geragotelis et al, Proc Natl Acad Sci U S A 2020 Vol. 117 Issue 24 Pages 13490-13498). The open structure reveals NH=6. We used this value for our calculations.]]

Minor comments:

1) Figure 6: the authors should precise that the model of proton conduction through Hv1 is just an assumption. The structural features of Hv1 open state are indeed unknown.

[[We modified the figure based on the simulation results of Geragotelis et al. We indicated in the legend that the scheme is based on HV1 homology models.]]

2) Page 9, lines 170‐171 "Drastically prolonged tail current kinetics might reflect a decreased voltage‐dependence of the deactivation in the D174 mutant". Or rather the prolonged kinetics reflect the stabilization of the open state by the mutation (as stated by the authors just after).

[[Page 14:

“Drastically prolonged tail current kinetics might reflect (i) a decreased voltage dependence of the deactivation in the D174A mutant or (ii) a stabilized open state (14).”]]

3) Supplementary figures are displayed in an odd fashion. Figure S3 should be placed before Figures S1 and S2.

[[We added two more Supplementary Figures and displayed them in the order of text mentionings.]]

4) In Figure 2, displaying the current trace corresponding to the 0 mV voltage step would improve readability of the figure, by showing that Hv1‐D174A mutants conduct protons at 0 mV and not wt Hv1.

[[

We show the current trace corresponding to the 0 mV voltage step for the D174A mutant in panel A and the trace for the wild-type in panel B of Fig. 2.]]

5) Figure 2 legend "Pronounced inward H+ currents activate negatively to the reversal potential (here ‐70 mV)". I think the authors mean "Here 0 mV", ‐70 mV is the threshold potential. Panel (c), I guess the EH vs Vrev plot is for D174A mutants but it is not mentioned in the legend

[[

We corrected the legend. “Pronounced inward H+ currents activate negatively (here – 70 mV) to reversal potential (here – 8 mV), indicating a high open probability of the D174A mutant at 0 mV.” And “Comparison of calculated Nernst potential for protons (EH) and measured reversal potential (Vrev) for the D174A mutant.”]]

6) Page 4, line 89: the fact that D174A conducts protons at a lower rate is, at this point, based on a lot on assumption. I would just correct the last sentence by saying "Thus, D174A, while opening with less depolarization, seems to conduct protons at a lower rate"

[[We toned down our statement and inserted a phrase very close to the one suggested.

Page 5: “Our observation suggests a reduced flux through the mutant if we assume that the protein expression level is independent of the mutation.”]]

7) Page 6, line 107. The word "therefore" is not necessary

[[ok]]

8) Page 7, line 128: "of" in "measures of transport" is missing

[[We deleted the paragraph.]]

9) Page 12, lines 261‐262: "Figure M" ??

[[“Inset of Figure 3A”]]

CROSS‐CONSULTATION COMMENTS I agree with the two other reviewer's comments. I think our reviews more or less raise the same weaknesses in the study.

Significance

This paper addresses a single question with a clearly defined experimental paradigm. Once the issues addressed, the paper should bring important significance to the field of voltage‐gated ion channels since the nature of proton conduction in Hv1 was not known. It could help explain ion conduction in some channelopathies involving ion conduction through the voltage‐sensing domain. The audience is mainly the voltage‐gated ion channel community, as well as the community of membrane permeation mechanisms My field of expertise is in ion channel structure‐function and pharmacology. I have little expertise in the described proton and water flow assays. Therefore I do not have sufficient expertise to evaluate the detailed experimental protocol that led to the measurements.

Reviewer #3:

Summary: This study addresses a fundamental question about the mechanism of proton conduction in the voltage gated proton channel Hv1 i.e., whether protons hop through an uninterrupted water wire, or move by other means involving titratable channel residues. The authors argue that an uninterrupted water wire entails a certain rate of water movement through the open channel, which they estimate to be around 10‐12 cm3s‐1 based on a structural model of Hv1 and previous work on other channels. They then measure water permeability of LUVs containing a purified Hv1 mutant expected to be open at 0 mV via light scattering, and proton flux using a pH sensitive fluorescent dye. They calculate a water permeability much lower than predicted and conclude that the water in the conduction pathway does not form an uninterrupted water wire. The manuscript is written clearly, and the experimental measurements are convincing.

[[We thank the reviewer for the positive evaluation.]]

There are nonetheless some ambiguities in the way the formation of water wires is discussed.

Major comments: A protein like Hv1 is larger and more complex than small peptides like gramicidin. In this context, transient water wires, frequently interrupted by titratable residues, or by steric hindrance from hydrophobic sidechains etc. are likely. Can the authors provide an estimate for the maximum frequency and lifetime of uninterrupted proton wires compatible with their measurements? This would be helpful to evaluate whether short‐lived uninterrupted water wires could contribute significantly to proton conduction or not. Trapping usually implies restricted movement. So, for how long do water molecules need to stay inside the channel in order to be considered trapped? Are the water molecules really trapped or simply forming broken wires?

[[Page 13, bottom:

“The question arises whether the obstacle in the water pathway is permanent. HV1’s titratable residues or steric hindrance from fluctuating sidechains may frequently interrupt otherwise intact water wires. Yet, our calculations (Eqs. 7 – 11) show that proton diffusion from the bulk solution to the pore mouth is the transport limiting step. Undoubtedly, transient closure would have caused a detectable pore resistance because part of the protons arriving at the pore mouth could not enter the pore. If the pore was closed longer than one ps, an arriving H+ may diffuse out of the capture zone and vanish into the bulk:

t_c=(r_0^2)/6D = 10^(-16)/(6 × 8.65 × 10^(-5) ) s = 2 × 10^(-13) s

(16)

where tc denotes the time a proton requires to diffuse a distance equal to the capture radius r0. Since transient closures would give rise to experimentally undetected pore resistance, they must be ruled out. The observation agrees well with noise experiments, where Lorentzian time constants, albeit smaller than the time constants for H+ current activation but larger than 0.1 s were observed (41).

We provided the calculations showing the diffusion limitations on page 9:

“…we show that the transport limiting step is H+ diffusion to the pore (access resistance) and not transport through the pore. Therefore, we first calculate the maximum current Imax permitted by diffusion for a constantly open pore (35):

I_max=2π F r_o D_H c_H

(7)

where F, r0, DH, and cH are Faraday's constant, the capture radius, the H+ diffusion constant, and the H+ concentration, respectively. The only unknown parameter is r0. Taking the gA estimate r0 = 0.87 Å (36), disregarding buffer effects and assuming DH = 8.65×105 cm2s-1, we find:

I_max=2π (9.6 ×10^4 As)/mol × 0.87 × 10^(-8) cm × 8.65 x 10^(-5) (cm^2 s^(-1) × 4 × 10^(-7.5) mol)/(1000 cm^3 )

(8)

I_max=5.6 × 10^(-17) A

(9)

Eq. 8 considers that the approximately 25 % charged lipids in the bilayer induce an increase in surface proton concentration, i.e. it accounts for a surface potential of roughly -40 mV in 150 mM salt. The maximal unitary rate would then be equal to:

q_max = 5.6 × 10^(-17) C/s/1.6 × 10^(-19) C =348 s^(-1)

(10)

Here we used the r0 value determined for gA (36). Acidic moieties at the entrance of HV1 and proton surface migration along the lipid bilayer could serve to increase that value (37, 38). The observation suggests transport limitations by poor proton availability. Calculation of the channel resistance, Rch (35), confirms the hypothesis:

R_ch = R_pore+R_access =[l_ch+(π a_ch)/2] ρ/(π a_ch^2 )

(11)

where R_pore is the resistance of the pore proper and R_access is the access resistance. Assuming a channel radius, a_ch, of 0.15 nm, a length, l_ch of 4 nm and solution resistivity (H+ as the sole conducted ion at bulk pH of 7.5 and a surface potential of -40 mV), ρ, of 2×105 Ω cm, we find R_ch = 4×1013 Ω. Thus, the resulting current, Iρ, that we may expect for the vesicular membrane potential of 100 mV is equal to 3×10-15 A. Accordingly, Iρ exceeds Imax by more than one order of magnitude. Consequently, we may safely conclude that HV1 conductance is limited by proton availability under our conditions. ”]]

The main conclusion of the paper rests on the negative results from the water permeability assay of Fig. 5. It is recommended to include a positive control (e.g., with gramicidin A), run under the same conditions and similar number of channels per LUV, to show how the results should look like in case of significant water permeability.

[[We included the gramicidin measurements (Fig. 6) as requested.]]

Figure 6 show a simplified scheme of proton transport with trapped water molecules in Hv1. Panel A represents a resting state (nonconductive); panel B represents an open state (conductive), favored by the D174A mutation. So, what makes B conductive and A nonconductive? Is it the presence of two salt bridges in B vs. three salt bridges in A? This should be clarified.

[[

We modified the figure based on the simulation results of Geragotelis et al. We indicate with arrows the parts of the channel where the proton is free to move and crosses the sites with insurmountable energy barriers.

Legend to the figure (now Fig. 8): “In the region of the selectivity filter adjacent to D112, the channel is too narrow to let water molecules pass (see also Fig. S1). Yet, the proton may bypass the electrostatic barrier of the open channel at D112 (18), i.e., jump between the two neighboring water molecules. Removal of D174 shifts the voltage sensitivity so that most channels are already open at a transmembrane potential of 0 mV. B) The closed channel. It neither allows water nor proton transport. In its new location, D112 provides an insurmountable electrostatic barrier to proton passage.”]]

Minor comments: The interpretation of Fig. 2E strongly depends on the assumption that the D174A mutation does not alter membrane trafficking. It is recommended to check the validity of this assumption, e.g., by colocalization with a plasma membrane marker. Images of SDS‐PAGE results for the studied Hv1 proteins should be provided to show preparation purity.

[[

We toned down the interpretation of Fig. 2E. As it stands now, Fig. 2 shows that the mutant (i) is functional and (ii) has a high open probability at 0 mV. These conclusions are independent on membrane trafficking. We included images of SDS page results for the studied HV1 proteins in the Supplement.]]

CROSS‐CONSULTATION COMMENTS I agree with the comments from the other two reviewers. My major point is that refuting major water permeability in Hv1 is not the same thing as refuting that protons can be conducted by transient water wires, unless it is proved that the transient water wires cannot sustain enough proton movement to account for the single channel conductance. Reviewer #3 (Significance (Required)): The Hv1 channel plays important roles in the human body, including the immune, respiratory, and reproductive systems. Despite recent advances in understanding the mechanism of proton conduction by Hv1, whether or not protons hop within a continuous water wire in the open channel is a subject of debate (DeCoursey J. Physiol. 2017, Bennett & Ramsey J. Physiol. 2017). This work provides important insights on the debate by refuting the existence of a water wire that can sustain large water permeability. The findings reported here will be of interest to ion channel biophysicist like this reviewer, but also to biologists studying cellular pH homeostasis and the pathophysiology of Hv1.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

This study addresses a fundamental question about the mechanism of proton conduction in the voltage gated proton channel Hv1 i.e., whether protons hop through an uninterrupted water wire, or move by other means involving titratable channel residues. The authors argue that an uninterrupted water wire entails a certain rate of water movement through the open channel, which they estimate to be around 10-12 cm3s-1 based on a structural model of Hv1 and previous work on other channels. They then measure water permeability of LUVs containing a purified Hv1 mutant expected to be open at 0 mV via light scattering, and proton flux using a pH sensitive fluorescent dye. They calculate a water permeability much lower than predicted and conclude that the water in the conduction pathway does not form an uninterrupted water wire. The manuscript is written clearly, and the experimental measurements are convincing. There are nonetheless some ambiguities in the way the formation of water wires is discussed.

Major comments:

A protein like Hv1 is larger and more complex than small peptides like gramicidin. In this context, transient water wires, frequently interrupted by titratable residues, or by steric hindrance from hydrophobic sidechains etc. are likely. Can the authors provide an estimate for the maximum frequency and lifetime of uninterrupted proton wires compatible with their measurements? This would be helpful to evaluate whether short-lived uninterrupted water wires could contribute significantly to proton conduction or not.

Trapping usually implies restricted movement. So, for how long do water molecules need to stay inside the channel in order to be considered trapped? Are the water molecules really trapped or simply forming broken wires?

The main conclusion of the paper rests on the negative results from the water permeability assay of Fig. 5. It is recommended to include a positive control (e.g., with gramicidin A), run under the same conditions and similar number of channels per LUV, to show how the results should look like in case of significant water permeability.

Figure 6 show a simplified scheme of proton transport with trapped water molecules in Hv1. Panel A represents a resting state (nonconductive); panel B represents an open state (conductive), favored by the D174A mutation. So, what makes B conductive and A nonconductive? Is it the presence of two salt bridges in B vs. three salt bridges in A? This should be clarified.

Minor comments:

The interpretation of Fig. 2E strongly depends on the assumption that the D174A mutation does not alter membrane trafficking. It is recommended to check the validity of this assumption, e.g., by colocalization with a plasma membrane marker.

Images of SDS-PAGE results for the studied Hv1 proteins should be provided to show preparation purity.

Referees cross-commenting

I agree with the comments from the other two reviewers. My major point is that refuting major water permeability in Hv1 is not the same thing as refuting that protons can be conducted by transient water wires, unless it is proved that the transient water wires cannot sustain enough proton movement to account for the single channel conductance.

Significance

The Hv1 channel plays important roles in the human body, including the immune, respiratory, and reproductive systems. Despite recent advances in understanding the mechanism of proton conduction by Hv1, whether or not protons hop within a continuous water wire in the open channel is a subject of debate (DeCoursey J. Physiol. 2017, Bennett & Ramsey J. Physiol. 2017). This work provides important insights on the debate by refuting the existence of a water wire that can sustain large water permeability. The findings reported here will be of interest to ion channel biophysicist like this reviewer, but also to biologists studying cellular pH homeostasis and the pathophysiology of Hv1.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

Voltage-gated proton channels are peculiar members of the voltage-gated ion channel family due to their absence of canonical pore. Instead, protons permeate through their voltage-sensing domain. The mechanisms of proton permeation in Hv1 channels are still unclear, with currently two competing hypotheses: (i) hopping through titrable residues within the protein; or (ii) via Grotthuss mechanism involving proton jumping through a continuous water wire. So far, these hypotheses were only tackled by computation. The authors therefore aimed to experimentally test the two hypotheses. To do so, the authors measured the transport rates of protons and water through wild-type and mutant D174A Hv1 reconstituted in lipid vesicles. Overall, the presented data are convincing and support their conclusion that proton conduction through the channel is not solely mediated by water transport. However, there are several aspects of the paper that I did not understand and would require clarification.

Major comments:

My major concern is about the relevance of using the D174A mutant. The authors explain at the beginning of the paper that Hv1-D174A is open at 0 mV, which allows measuring proton flux in systems in which voltage cannot be controlled. However, it seems from the proton flux experiments that wild-typet Hv1 can conduct protons perfectly well in the used experimental paradigm. So why test a mutant? It is actually not clear why wild-type Hv1 can conduct protons in the proton conduction assay. The authors should clearly state the trans-membrane potential created by the K+ gradient across the vesicle, as well as the pH inside and outside the vesicle, and related these conditions to their electrophysiology data to give us an idea of the open probability of wild-type Hv1 in the conditions used in the proton conduction assays. This is critical to be able to compare the relative rates of proton transport between the wild-type and the mutant. Similarly, the buffers and pH used for the water transport assay are not explicitly mentioned. Are they the same as for the proton transport assay or are the buffers inside and outside the vesicle symmetrical? Finally, in the introduction the authors base their assumptions about water transport on an X-ray structure of Hv1 in a closed conformation (3WKV). I do not think it is relevant to study permeation, which in theory should only happen in an open state. If the authors want to make assumptions about the number of hydrogen bonds in the pore and how many water molecules are in the pore (and I don't think they need to do it), they should rather base their assumptions on the computational models of Hv1 open state.

Minor comments:

1. Figure 6: the authors should precise that the model of proton conduction through Hv1 is just an assumption. The structural features of Hv1 open state are indeed unknown.
2. Page 9, lines 170-171 "Drastically prolonged tail current kinetics might reflect a decreased voltage-dependence of the deactivation in the D174 mutant". Or rather the prolonged kinetics reflect the stabilization of the open state by the mutation (as stated by the authors just after).
3. Supplementary figures are displayed in an odd fashion. Figure S3 should be placed before Figures S1 and S2.
4. In Figure 2, displaying the current trace corresponding to the 0 mV voltage step would improve readability of the figure, by showing that Hv1-D174A mutants conduct protons at 0 mV and not wt Hv1.
5. Figure 2 legend "Pronounced inward H+ currents activate negatively to the reversal potential (here -70 mV)". I think the authors mean "Here 0 mV", -70 mV is the threshold potential. Panel (c), I guess the EH vs Vrev plot is for D174A mutants but it is not mentioned in the legend
6. Page 4, line 89: the fact that D174A conducts protons at a lower rate is, at this point, based on a lot on assumption. I would just correct the last sentence by saying "Thus, D174A, while opening with less depolarization, seems to conduct protons at a lower rate"
7. Page 6, line 107. The word "therefore" is not necessary
8. Page 7, line 128: "of" in "measures of transport" is missing
9. Page 12, lines 261-262: "Figure M" ??

Referees cross-commenting

I agree with the two other reviewer's comments. I think our reviews more or less raise the same weaknesses in the study.

Significance

This paper addresses a single question with a clearly defined experimental paradigm. Once the issues addressed, the paper should bring important significance to the field of voltage-gated ion channels since the nature of proton conduction in Hv1 was not known. It could help explain ion conduction in some channelopathies involving ion conduction through the voltage-sensing domain.

The audience is mainly the voltage-gated ion channel community, as well as the community of membrane permeation mechanisms.

My field of expertise is in ion channel structure-function and pharmacology. I have little expertise in the described proton and water flow assays. Therefore I do not have sufficient expertise to evaluate the detailed experimental protocol that led to the measurements.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

1. The authors state that: "the conductance density mediated by the expression of the mutant was 2.5 times smaller than the wild type, although we transfected the same amount of plasmid DNA (Fig. 2E). Assuming that protein expression is independent of the mutation, the observation suggested that the unitary proton flux ratio RC of wild type to mutant channel was equal to 2.5" (lines 82-85).

Macroscopic conductance (G) depends on channel number (N), microscopic or unitary conductance (), and open probability (PO) by G=NPO. The authors assume that the level of WT and D174A mutant protein expression on plasma membrane, which determines N, are equal; however, this critical assumption does not appear to have been tested. The fact that conductance density (nS/pF) is plotted in Fig. 2E does not alter this caveat because this procedure normalizes the data only for cell surface area (i.e., size).

The authors' conclude that "The conductance density relationship (Fig. 2E) compares the maximal conduction of both constructs; this is the fully open channel (open probability ≈ 1)"(lines 87-88). However, neither raw currents nor G-V data are shown. Typically, currents measured at large, near-saturating PO are used to compare the relative conductances of WT and mutant ion channels. The currents shown in Fig. 2A and 2B exhibit prominent 'droop' at even modest depolarizing potentials (+10 mV for D174A and +30 mV for WT), indicating that the proton gradient has been substantially perturbed by the flow of ge depolarizing voltages needed to drive channels to near-maximal PO. Furthermore, there is no evidence that maximal PO itself is also not different in WT and D174A channels. Indeed, maximal PO for native Hv1 channels measured using variance analysis is reported by significantly smaller than 1.0, and assuming that PO = 1.0 for either WT or D174A is therefore not well supported. Maximal could be altered by the D174A mutation, which has a clear and strong effect on channel gating evidenced by the large (-70 mV) negative shift in threshold potential reported both here and previously in the literature. Effects of mutations on maximal PO due to altered gating behavior could be separate and distinct from any change in plasma membrane channel number (N). Lastly, because D174A channels have a much higher PO than WT at 0 mV, the mutant will necessarily conduct inward proton currents at the physiological resting membrane potential (RMP) in tsa-201 cells (perhaps -30 mV?). Inwardly directed proton currents will therefore cause intracellular acidification under resting conditions. The constitutive acid load in cells expressing D174A, but not WT, is likely to have a variety of physiological consequences, including decreased protein expression or plasma membrane targeting of D174A. There is evidence that another-constitutively open Hv1 mutant (R205H) also generates smaller currents macroscopic conductance than WT, and this phenomenon is likely to result from decreased cell surface expression. To conclude that the microscopic conductances of WT and D174A are unequal, the authors must demonstrate that N is not different The authors' conclusion that D174A "conducts protons at a lower rate" (line 89) is therefore not well supported by the experimental data. 2. The authors indirectly measure apparent proton flux rates (D) in LUVs containing WT and D174A mutant Hv1 channels using a fluorescence-based approach, and conclude that D is 2.4 times smaller for D174A than WT. However, the method for estimating D is not performed under voltage clamp, and the driving force for proton current is neither known nor measured. The authors state that "Transmembrane voltage constituted the driving force for proton uptake into LUVs (Figure M). It resulted from facilitated K+ efflux out of the vesicles (30)", (lines 261-262), but this voltage is unknown and not likely to equal the Nernst equilibrium potential for K+ once Hv1 channels begin to open.

Once Hv1 channels begin to open, intra-lumenal pH (pHi) will necessarily occur during the experiment. Such changes are likely exacerbated by a) the low proton buffering capacity of the system (5 mM HEPES) and b) the absence of any counter-charge pathway to balance the effect of proton charge movement on the membrane potential. Given the small volume of LUVs, even a relatively modest difference in either membrane potential or pHi could substantially alter the driving force for proton movement. Together, these factors are highly likely to result in a rapid and potentially large change in the driving force for proton flux.

Driving force changes may also be different for WT and D174A because their relative PO may be different under the experimental conditions used here. Because D174A activates at much more negative voltages, it is likely to open more quickly and to a higher PO than WT at early times after depolarization is initiated by addition of valinomycin (Fig. 3A). This fact will likely result in a larger initial inward current being carried by D174A than WT channels. The result would be a more rapid acidification of LUVs by D174A.

The experimental data in Fig. 3A are consistent with the expectation that the proton gradient and driving force more rapidly approach equilibrium for D174A than WT channels: the apparent rate of AMCA fluorescence change is slower in D174A. Although the authors correctly interpret the experimental data to mean that the apparent D is slower for D174A, they do not rule out the artifactual explanation for the measured differences. Indeed, the observation in Fig. 3A that AMCA fluorescence change eventually reaches a plateau and is not affected by CCCP means that the proton gradient has become exhausted during the experiment, and directly demonstrates that the proton driving force is uncontrolled under the current experimental conditions.

In contrast to the authors' statement that "Our experiments with the purified and reconstituted channels corroborated the conclusion (Fig. 3A)", (lines 92-93) it is not clear that unitary proton flux rates/unitary conductances are actually different in WT and D174A. 3. The presumed differences in unitary conductances (i.e., 'transport rate') between WT and D174A are used to estimate Arrhenius activation energies (Ea): ("The difference in measures transport rates allows a rough estimation of the Arrhenius 128 activation energy Ea for HV1-mediated proton flow. It amounts to 40 kJ/mol for the wild type and 23 kJ for the mutant. Thus, Ea exceeds the corresponding 15 kJ/mol barrier measured for gramicidin A (32, 33)", (lines 128-130).

The method for determining Ea in the current work is not well-described. In Ref. 32, the authors estimate Arrhenius activation energy (Ea = 20 kJ/mol) for gramicidin D (not gramicidin A) from the slope of a line fit to measurements of currents at various temperatures. Here, the authors measure AMCA fluorescence decay rates at 4{degree sign}C and 23{degree sign}C and observe a similar temperature-dependent difference in WT and D174A (Fig. S2). Given that the data indicate that WT and D174A are similarly temperature-dependent, it is unclear how the authors arrive at different Ea values. The authors' conclusion that "The increment in Ea suggests that the transport mechanism may be different from a pure Grotthuss type, where the proton uses an uninterrupted water wire to cross the membrane", (lines 131-133) therefore does not appear to be well-supported. 4. The authors report no difference in water permeability in WT vs. D174A (Fig. 5 and S1) and interpret the results to mean that proton currents are not associated with measurable bulk water flow. A similar conclusion was reached for native Hv1 channels using deuterium substitution (DeCoursey & Cherny, 1997). However, the absence of bulk water flow does not itself rule out the possibility that 'trapped' waters within the Hv1 pore do not themselves carry the measured proton current. If intra-pore water molecules are tethered by hydrogen bonds with protein atoms, they may not move when Hv1 channels open. Proton transfer through a hydrogen-bonded network of waters requires only that the electronic structure of the network be rearranged during proton transfer; water is not required. As in the previous study (DeCoursey & Cherny, 1997), the lack of water flux reported here demonstrates seems to reinforce the notion that H+ moves separately from its waters of hydration (i.e., hydronium, H3O+, is not the permeant species) and does not necessarily imply information about the mechanism of proton transfer (i.e., side chain ionization vs. Grotthuss-type transfer in a water-wire).

The authors state that: 1) "every H-bond donating or receiving pore-lining residue would have contributed an increment ΔΔ𝐺‡ of 0.1 kcal/mol to the Gibbs free energy of activation Δ𝐺‡ (25)" (lines 145-147), and 2) calculating NH from this Δ𝐺‡ allows estimation of the channel's unitary water permeability (Eqn. 2). Although hydrogen bonding patterns will undoubtedly alter the free energy for channel activation, this is not the same free energy change as that for proton transfer. Hv1 gating involves conformational changes that are both voltage and pH-dependent, and the D174A mutation is known to alter the voltage dependence of gating (Fig. 2 and previous studies). The effect of D174A on Hv1 unitary conductance, however, is speculated but not unambiguous (see above). In the absence of definitive experimental data showing differences in the unitary conductance of WT vs. D174A, the authors' assumption that water permeability would be strongly temperature-dependent (lines 154-160) seems premature and their ensuing conclusion tenuous: "pore residues interrupt the HV1 spanning water wire, trapping the water molecules inside the HV1 channel. In contrast to water, protons cross the pore by hopping from one acidic residue to another through one or more bridging water molecules (Fig. 6)" (lines 161-164).

Furthermore, the authors calculate the number of hydrogen bonds (NH) that pore waters could form with pore-lining residues based on an X-ray structure of a chimeric proton channel protein (pdb: 3WKV) that is: a) manifests discontinuous transmembrane water density and is known to represent a non-conductive conformation, b) contains residues from Ci-VSP in the critical S2-S3 linker that form part of the proton transfer pathway, and c) exhibits structural features (i.e., highly conserved ionizable residues such as D185 and R205, which like D174 are reported to dramatically alter Hv1 gating, are packed into a solvent-free crevice) that are inconsistent with physiological function. Given that all Hv1 ionizable mutant combinations tested so far (the sole exception of D112V - other non-ionizable substitutions at D112 are tolerated) remain functional (Musset, Smith et al., 2011, Ramsey, Mokrab et al., 2010), the identities of water-interacting residues speculative. Interpreting differences in the calculated NH based on pdb: 3WKV therefore seems unlikely to reveal fundamentally important insights into Hv1 function. The author's conclusion that "The observation rules out the formation of an uninterrupted water chain spanning the open channel from the aqueous solution at one side of the membrane to the other. NH would have governed water mobility if such a water wire had formed (24)", (lines 143-145) therefore does not appear to be strongly supported.

References

Bennett AL, Ramsey IS (2017a) CrossTalk opposing view: proton transfer in Hv1 utilizes a water wire, and does not require transient protonation of a conserved aspartate in the S1 transmembrane helix. J Physiol

Bennett AL, Ramsey IS (2017b) Rebuttal from Ashley L. Bennett and Ian Scott Ramsey. J Physiol

De La Rosa V, Bennett AL, Ramsey IS (2018) Coupling between an electrostatic network and the Zn(2+) binding site modulates Hv1 activation. J Gen Physiol

De La Rosa V, Ramsey IS (2018) Gating Currents in the Hv1 Proton Channel. Biophys J 114: 2844-2854

DeCoursey TE (2017) CrossTalk proposal: Proton permeation through HV 1 requires transient protonation of a conserved aspartate in the S1 transmembrane helix. J Physiol 595: 6793-6795

DeCoursey TE, Cherny VV (1997) Deuterium isotope effects on permeation and gating of proton channels in rat alveolar epithelium. J Gen Physiol 109: 415-34

Musset B, Smith SM, Rajan S, Morgan D, Cherny VV, Decoursey TE (2011) Aspartate 112 is the selectivity filter of the human voltage-gated proton channel. Nature 480: 273-7

Ramsey IS, Mokrab Y, Carvacho I, Sands ZA, Sansom MS, Clapham DE (2010) An aqueous H+ permeation pathway in the voltage-gated proton channel Hv1. Nat Struct Mol Biol 17: 869-75

Ramsey IS, Moran MM, Chong JA, Clapham DE (2006) A voltage-gated proton-selective channel lacking the pore domain. Nature 440: 1213-6

Randolph AL, Mokrab Y, Bennett AL, Sansom MS, Ramsey IS (2016) Proton currents constrain structural models of voltage sensor activation. Elife 5: e18017

Significance

Here the authors attempt to ascertain whether water molecules may mediate proton transfer in the voltage-gated proton channel Hv1 using a combination of whole-cell voltage clamp electrophysiology, protein purification, reconstitution, and pH-dependent AMCA fluorescence measurement and estimates of water permeability, and hydrogen bond calculations based on an X-ray structure of a chimeric Hv1 proton channel model protein. The authors address an important question that is fundamental to the exquisitely proton-selective Hv1 channel and which may be applicable to other proton transporting proteins.

Although there is high potential for significance to a wide range of experimenters studying biologically fundamental mechanisms of proton transport, the experimental data fail to strongly support most of the authors main conclusions, and it is unclear whether the work represents a technial advance for the field. Previous work in the literature has described two main hypotheses for the proton transport mechanism in Hv1:

• A) an intra-protein transmembrane water wire that allows permeating H+ to move along a chain of hydrogen-bonded water molecules and does not require explicit ionization of any particular amino acid side chain (Bennett & Ramsey, 2017a, Bennett & Ramsey, 2017b, Ramsey et al., 2010), and
• B) Explicit ionization of a conserved side chain in the S1 helix (D112 in human Hv1) is required for proton transfer in Hv1 channels (DeCoursey, 2017, Musset et al., 2011). The Reviewer is an expert in the field, having originally identified and functionally characterized Hv1 channels in 2006 (Ramsey, Moran et al., 2006), contributed to the identification of key side chains and structural determinants of Hv1 function (De La Rosa, Bennett et al., 2018, Ramsey et al., 2010, Randolph, Mokrab et al., 2016), measured gating currents in Hv1 (De La Rosa & Ramsey, 2018), and authored the hypothesis that Hv1 utilizes a water-wire type mechanism for proton transfer (Ramsey et al., 2010).

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

Learn more at Review Commons

---

Reply to the reviewers

August 17, 2022

RE: Review Commons Refereed Preprint #RC-2022-01442

Dear Editor of the EMBO Journal,

Please find our updated manuscript and response to the reviewers’ comments. We appreciate the effort that the reviewers have put into the evaluation of our manuscript.

We are happy with the potential importance the reviewers realise in the study:

Reviewer 1: The finding that ubiquitination occurs inside mitochondria would be an important conceptual advance, which would open new perspectives both for ubiquitination and mitochondrial biology

Reviewer 2: This work would represent a significant/exceptional discovery if supported by compelling data.

Reviewer 3: the results are interesting and very important, as mentioned in the major comments section…

With regard to the major comments raised by the reviewers, you will find below our specific response point by point with explanations and suggested novel experiments (highlighted in yellow). In summary we suggest the following actions to fully support our model:

• We will perform a-complementation with ubiquitin (lacking the GG motif) fused at its C-terminus to the short fragment of b-galactosidase (a). Blue colonies with ωm will indicate import.
• As shown in Figure S2, now added to the manuscript, we show detection of ubiquitinated proteins and mono ubiquitin in extracts of mitochondria pre-treated with trypsin.
• A bio-archives address of our other manuscript will be provided.
• The use of a-complementation for protein localization was developed by us 15 years ago and since then has been used by us and other groups verifying its use as a screening tool. One point is clear, ωm or ωc do not leak into other subcellular compartments. Nevertheless, in the research of specific genes validation is important. Yes!!! ωm and ωc are exclusively located in mitochondria or the cytosol respectively.
• We will highly purify mitochondria on gradients and treat them with protease.
• We cannot be sure that we will be able to detect a protein with ubiquitin modifying activity which functions solely on certain proteins in mitochondria, so publication cannot rely on this.
• Repeat mass spectrometry with careful editing will be undertaken as suggested by the reviewer.
• We will attempt to perform protease protection assays in the presence of specific detergents.

Before tackling the very tough revision, we would like to know if EMBO Journal would positively consider acceptance of our manuscript based on the review and planned revision.

Prof. Ophry Pines Microbiology & Molecular Genetics Hebrew University of Jerusalem Jerusalem 91220 Israel

---

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

Summary:

In this manuscript, Zhang et al. investigate whether ubiquitination occurs inside mitochondria of the budding yeast S. cerevisiae. They first observe thanks to a sensitive complementation assay that several components of the yeast ubiquitination (and deubiquitination) machinery can localize inside mitochondria. To be able to specifically probe ubiquitin conjugates assembled inside mitochondria they fused HA-tagged ubiquitin to a mitochondrial targeting sequence. Using this construct, they demonstrate that ubiquitin conjugates can be assembled in mitochondria. A series of elegant experiments demonstrates that the pattern of ubiquitin conjugates depends on the mitochondrial localization and the activity of the ubiquitin conjugating enzyme Rad6. Altogether, these results convincingly demonstrate that ubiquitination can occur inside yeast mitochondria when ubiquitin is intentionally targeted inside this organelle. It however remains unclear whether mitochondrial ubiquitination occurs in endogenous conditions (without targeting ubiquitin into this compartment) and whether it affects mitochondrial functions.

Response: Regarding the question whether mitochondrial ubiquitination occurs in endogenous conditions, we feel that this is obvious based on our results. We detect numerous ubiquitination related enzymes (E1, E2, E3, DUB) eclipsed in mitochondria but none of the proteasome subunits. As pointed out by the reviewer “these results convincingly demonstrate that ubiquitination can occur inside yeast mitochondria”. With that said, additional data will be incorporated into the manuscript as suggested by the reviewer and can be seen below.

Major comments:

1) The materials and methods section is lacking important information (western blot protocol, details of antibodies, strains, plasmids...). It is thus difficult to evaluate how several experiments were performed and how their design (e.g. the promoters chosen to express tagged proteins) could impact the interpretation of the results. This is a major issue that needs to be corrected. The main text should also explicitly indicate whether tagged proteins used in the alpha-complementation assay are overexpressed or not.

Response: The materials and methods section will be updated accordingly.

2) Despite the previous comment, the data presented in the manuscript convincingly demonstrate that multiple components of the ubiquitination machinery can localize within mitochondria and that ubiquitin conjugates can be assembled in mitochondria when ubiquitin is modified to be intentionally targeted into this compartment. However, little data is shown to support the hypothesis that ubiquitin conjugates can be assembled in mitochondria when ubiquitin is not fused to a mitochondrial targeting sequence. Thus, in my opinion, the evidences presented in the current manuscript are not sufficient to conclude that ubiquitin conjugates are assembled in mitochondria in endogenous conditions (as this is done implicitly). Additional evidences are needed to draw this conclusion (see some experimental suggestions hereafter). Without further evidences, the speculative aspects of the claim that "ubiquitination occurs in the mitochondrial matrix" should be discussed explicitly.

Response: See the discussion above why we are confident that ubiquitination occurs in mitochondria. Our major problem with ubiquitin and the ubiquitination enzymes is that they are eclipsed in mitochondria. We propose as suggested by the reviewer (item 4 of his review) to perform a-complementation with ubiquitin fused at its C-terminus to the short fragment of b-galactosidase (a). Blue colonies with ωm will indicate import.

3) The authors used a mass spectrometry approach to identify mitochondrial ubiquitination substrates. However, they have not yet succeeded in identifying a substrate whose modification is specifically regulated by a given component of the mitochondrial ubiquitination machinery. They have also not identified a phenotype or process impacted by mitochondrial ubiquitination. Thus, at this stage, the biological consequences of mitochondrial ubiquitination remain elusive.

__Response: __We have not identified a substrate whose modification is dependent on a given component of the mitochondrial ubiquitination machinery, even though we have tried. Again, the problem is low levels of these proteins eclipsed in mitochondria. Even when we do find a protein that is ubiquitinated (e.g. Aco1) its ubiquitination is not exclusively dependent on Rad6. Thus, different ubiquitin enzymes may have the same substrates.

4) The authors have not directly investigated whether ubiquitin itself (without a mitochondrial targeting sequence) localizes in mitochondria. I encourage them to address this question since it would provide an important piece of evidence suggesting that mitochondrial ubiquitination can occur in endogenous conditions. This could be done using the alpha-complementation assay and the results could be presented within Figure 1. Ideally this experiment should be performed without overexpressing ubiquitin. Note that if the authors decide to use a C-terminally tagged form of ubiquitin for this experiment, the GG motif of ubiquitin should be mutated to avoid cleavage of the alpha tag by cellular DUBs. This form of ubiquitin will not be conjugatable, but this is not an issue for this experiment since its aim is to determine whether ubiquitin can be targeted to mitochondria, not to probe conjugates.

Response: We will perform experiments as suggested by the reviewer including ubiquitin fused at its C-terminus to the short fragment of b-galactosidase (a), see item 2. We have previously made a PreSu9-Ubi lacking a GG motif but now will look at a different combination of this and other constructs.

5) In the top panels of Figure 2 and S1, free ubiquitin is well detectable in the total and cytosolic fractions. It is however not clear to me whether it is also detectable in the concentrated mitochondrial fraction. If yes and if it would be resistant to trypsin digestion, it would provide additional evidence that endogenous ubiquitin can be targeted to the mitochondrial matrix (see previous comment).

Response: See Item 6.

6) The data shown in the top panel of Figure 2 and S1 also suggest that free ubiquitin is less concentrated in mitochondria than in the cytosol (since it is more difficult to detect in the concentrated mitochondrial fraction than in the cytosolic fraction, see previous comment). It is thus possible that the use of preSu9-HA-Ubi (or preFum1-HA-Ubi) lead to an artificially high intra-mitochondrial concentration of free ubiquitin. As the concentration of free ubiquitin is known to impact ubiquitination processes, I encourage the authors to compare the relative levels of free ubiquitin present in the mitochondrial fraction prepared from WT and preSu9-HA-Ubi (or preFum1-HA-Ubi) expressing cells. If free ubiquitin is detectable in mitochondrial fractions and resistant to trypsin (see previous comment), this could be done by repeating the experiment shown in Figure 3B and probing the blot with an antibody that recognizes free ubiquitin.

Response to 5 and 6: Detection of ubiquitin in mitochondria is extremely difficult even when mitochondria are 15-fold concentrated versus the cytosol and when HA-Ubi is overexpressed. Thus, ubiquitin is eclipsed in mitochondria. Nevertheless, as shown in the Figure below which was not part of the submitted manuscript yet was performed in parallel to experiments done early on, shows detection of very weak bands of free ubiquitin in extracts of mitochondria pre-treated with trypsin.

Endogenous ubiquitination pattern in mitochondria of _Δrad6 _cells is restored to normal by Rad6-α. __WT or Δrad6 cells containing a Rad6-α construct or an empty plasmid were subjected to subcellular fractionation. Mitochondrial fractions with or without trypsin treatment, were probed for ubiquitin by WB. Aco1 is a matrix mitochondrial protein, and Tom70 is a mitochondrial outer membrane protein (MOM) facing the cytosol.

7) I strongly encourage the authors to provide more data indicating that "ubiquitination occurs in mitochondria" by performing experiments that do not rely on the use of the preSu9-HA-Ubi or other forms of ubiquitin that are intentionally targeted to mitochondria. For instance, they could analyse the pattern of HA-Ubi conjugates of trypsin digested mitochondrial fractions prepared from wt, rad6-delta, and rad6-delta complemented with preSu9-Rad6-alpha-SL17. Note that if trypsin digested mitochondrial fractions are too contaminated by ubiquitinated proteins present outside mitochondria to perform this experiment, the authors may use the unspecific DUB Usp2 as an alternative protease to strip ubiquitinated proteins from the mitochondria periphery.

Response: Concentrated mitochondrial extracts from WT and Δrad6 cells untreated or treated with trypsin were probed with anti-ubiquitin antibodies (Figure above). A very weak band corresponding to free ubiquitin can be detected in extracts of mitochondria treated with trypsin but these are very weak and are on the limit of detection.

Minor comments:

1) Overall, the manuscript is well organized and easy to follow. The text is clearly written; the figures are well annotated.

2) The authors should provide full images of all the blots with anti-ubiquitin and anti-HA antibodies so that one can see the bands corresponding to free ubiquitin (or free HA-Ubi). For instance, in Figure 3B, it is not possible to see the presence (or absence) of the band corresponding to free HA-Ubi because the very bottom of the image is cut.

3) The authors should indicate whether the MTS of Su9 (and Fum1) are expected to be cleaved after import of preSu9-HA-Ubi (and preFum1-HA-Ubi) in mitochondria. They should also label on the corresponding immunoblots the presence (or absence) of the band corresponding to the free preSu9-HA-Ubi (and preFum1-HA-Ubi) (or HA-Ubi if the MTS is expected to be cleaved from these constructs).

4) In Figure 3B, the ubiquitin conjugates produced with preSu9-HA-Ubi and preFum1-HA-Ubi have different migration patterns. I think this should be explicitly mentioned and discussed. Could it be due to the presence of lysine residues in the Su9 or Fum1 MTS that could lead to the assembly of artificial ubiquitin chains?

5) The authors indicate that "endogenous Rad6 [...] is expressed at very low levels and can hardly be detected in the mitochondrial fraction by WB (Figure S5)". I did not manage to observe the band corresponding to endogenous Rad6 in the mitochondrial fraction in the pdf. The authors should provide a more contrasted or better quality image.

CROSS-CONSULTATION COMMENTS I agree with reviewer 2 that proper validation of the complementation assay is crucial for this manuscript. I was myself wondering whether it uses endogenously tagged proteins or whether it is based on an overexpression system. I imagine this information will be detailed in the manuscript in preparation mentioned by the authors. I am therefore wondering whether it would be possible to ask the authors to provide the draft of this manuscript (or at least the validation part).

Response: A bio-archives address of our other manuscript will be provided upon resubmission. See other issues referred to the response Reviewer 2.

I agree with most comments of reviewer 3. Regarding the hypothesis that preSu9-HA-Ubi could form aggregates on the cytosolic surface of the mitochondria, I think that the results presented on Figure 7B rather argue against it (since they indicate that Rad6 localized inside mitochondria can restore the pattern of ubiquitin conjugates). That's why (in my opinion) the major question the author now need to adress is whether intra-mitochondrial ubiquitination occurs in endogenous conditions (ie without forcing ubiquitin into this compartment and without E2 or E3 overexpression).

Response: See response to the other reviewers

Reviewer #1 (Significance (Required)):

The finding that ubiquitination occurs inside mitochondria would be an important conceptual advance, which would open new perspectives both for ubiquitination and mitochondrial biology research. However, the significance of the current manuscript is limited because the presented evidences heavily rely on the use of artificial conditions (ubiquitin tagged with a mitochondrial-targeting sequence) that may trigger irrelevant ubiquitination events. The significance would be much higher if the authors would provide further evidences indicating that intra-mitochondrial ubiquitination occurs in endogenous conditions and/or if they had identified a mitochondrial process specifically impacted by mitochondrial ubiquitination.

Expertise of the reviewer: Ubiquitination, Yeast biology, protein-protein interactions. No specific expertise in mitochondrial biology

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

In the manuscript by Yu et al., the authors test the concept that certain proteins are unevenly distributed within distinct cell compartments. Due to this localization discrepancy, protein detection in some subcellular compartments can be "eclipsed" by a predominant subset of specific protein localizing in another cell compartment their actual distribution. Therefore, tiny amounts of physiologically relevant proteins could be biologically relevant. Still, their function in some locations can be overlooked (or eclipsed) because of the high expression level of the same protein in another subcellular compartment(s). Although, this concept is not particularly novel. For example, it is already known that many different proteins can localize to distinct cellular locations (e.g., permanent mitochondrial and peroxisomal localization of many proteins or transient localization of particular proteins to separate cell compartments). The authors apply a yeast system and an α-complementation assay to test further the role of such eclipsed proteins in mitochondrial biology. Specifically, they focus on the ubiquitin (Ub, or as abbreviated incorrectly in this manuscript; Ubi) conjugation pathway, components of which have never been convincingly shown to localize inside the mitochondria. This work proposes that certain ubiquitination events can occur inside yeast mitochondria. This work would represent a significant/exceptional discovery if supported by compelling data. However, the major problem with this work is that the conclusions are based on the ectopic expression of distinct proteins. This approach is not failproof in precise protein expression/delivery to the specific subcellular locations and is likely to result in a non-specific localization. Thus, the problem of eclipsed proteins is addressed by the methodology that may lead to the artificial generation of eclipsed overexpressed proteins. A more effective approach would be if the authors found a way to study this issue with endogenous proteins. The need for overexpression of mitochondria-targeted ubiquitin makes it challenging to reconcile the physiological role of these fundings. In addition, some critical technical issues and omissions further reduce the potential impact of this work (see Specific comments below). For example, strong evidence of mitochondria fraction purity and additional evidence that all the essential constructs used in this work are not misdirected to a different compartment are needed.

Response: “Although, this concept is not particularly novel” is a very disappointing remark by the reviewer!! While dual targeting of proteins has been known for many, many years, how widespread the phenomenon was unknown and thought to be negligible. We are leaders for the last 30 years in the field of dual targeting and distribution and in particular distribution of single translation products. We coined the terms “echoforms” and “eclipsed distribution” and developed methods to detect and screen for dual targeting. The concept of eclipsed distribution and in particular eclipsed targeting to mitochondria is very new, and is leading to a novel perception of the mitochondrial proteome (see MS submission). While the reviewer appears to be an expert on ubiquitination, we are experts on dual targeting.

• Ub was abbreviated incorrectly in this manuscript, Ubi. __Response: __This will be corrected.

Other comments will be referred to in the response to Specific comments.

Specific comments 1. The authors should demonstrate beyond doubt that the ω components of their assay (ω-C, which supposedly stays in the cytosol-ONLY and the ω-M component, which seemingly remains in the mitochondria-ONLY) are in the compartment that the authors claim. These two proteins are transfected into yeast cells and overexpressed. Therefore it is possible that they leak to other, not intended, subcellular compartments. The authors assume that ω-M and ω-C are exclusively located either in the mitochondria or the cytosol. However, this should be shown as validation of the assay. The indicated reference from 2005 (Ref.13) and others are irrelevant since assays have variations and are often researcher/lab dependent. This validation is very important since a misallocation of the overexpressed ω-M or ω-C, leaking into other subcellular compartments, may cause misdetection of the α-constructs.

Response: The use of a-complementation for protein localization was developed by us 15 years ago and since then has been used by us and other groups verifying its use as a screening tool. One point is clear, ωm or ωc do not leak into other subcellular compartments. Nevertheless, in the research of specific genes validation is important. Yes!!! ωm and ωc are exclusively located in mitochondria or the cytosol respectively.

It is not surprising that Ub conjugates are detected in mitochondrial fractions. It could be due to ubiquitination of the OMM (coming from the cytosol) or perhaps since the subcellular fractions were not pure mitochondria free from contamination (the likely culprit could be the ER). The mitochondrial fractions in this work were obtained by 10,000 g separation between cytosolic and mitochondrial crude fractions. Indeed, these 10,000 g crude fractions are highly impure with membranes from other compartments (i.e., microsomes, lysosomes, and so on). Therefore, more sophisticated purification methods should be used. In addition, the authors should also test these fractions for non-mitochondrial proteins from other membrane organelles.

Response: We agree with the reviewer and therefore will take the following approaches:

1. i) We will treat isolated mitochondria with protease in order to remove adhering proteins and digest OMM proteins…… see attached figure.
2. ii) We will highly purify mitochondria on gradients and this will be straight forward since we are now employing such methods in other projects in the lab. iii) Matrix protein enrichment (by mass spec) is associated with IP for preSu9-HA-Ub conjugates which is three-fold higher than for HA-Ub. In any case the fact that we identify conjugates of proteins not known to be mitochondrial, strongly supports our thesis.

Figure 2. Coomassie blue staining does not show any signal in the "M" fraction. It can be interpreted that the authors do not get any mitochondria there, and therefore the lack of Ub signal is due to the absence of the protein in the samples. Using the same amount of protein from each fraction would probably reduce the necessity of 15x enrichment.

Response: The Coomassie blue staining does show a signal in the "M" fraction which is weak yet when a 15x enrichment is run, the protein level by Coomassie blue staining is similar to the cytosolic fraction.

Figure 3. It is puzzling why the HA-UBQ presence is so strong in the crude mitochondrial fraction, but the preSu9-HA-Ub signal (mito-matrix) is comparatively weak. These data suggest that the crude mito-fraction could be highly contaminated with OTHER membranes. On the other hand, the preSu9-HA-UBQ signal is no more than 1-5% of the total mitochondrial signal. The high enrichment of the HA-Ubi in both cytosols and the mitochondria could indicate the OMM ubiquitination or (again) contamination by other compartments. The constructs with MTS are detected in the mitochondria. However, the localization of tagged MTS-Ubi in a non-targeted compartment (e.g., cytosol) should be excluded by additional exposure times. Because the manuscript talks about eclipsed proteins, this is important.

Response: The HA-Ub is strong in the mitochondrial fraction, in the absence of trypsin, but is very weak in the presence of the protease indicating that most of the ubiquitinated proteins are externally attached to mitochondria. In contrast, PreSu9-HA-Ub is imported into the mitochondrial matrix and is protected from trypsin. This manuscript refers to “eclipsed in mitochondria” (not the cytosol) and this is true for ubiquitination enzymes as well as for ubiquitin.

Figure 3C-E. These data indeed suggest that the Ub-conjugates could be formed inside the mitochondria. However, the above-discussed possibility that other than mitochondria compartments co-sediment in the 10,000g fractions makes the data interpretation highly challenging.

__Response: __We will highly purify mitochondria on gradients and this will be straight forward since we are now employing such methods in other projects in the lab.

Figure 4. Unsurprisingly, mitochondrial targeting of Ub leads to detecting some co-immunoprecipitating mitochondrial proteins. However, these data do not support the notion that Ub conjugation machinery acts inside the mitochondria and that the target proteins are indeed conjugated with Ub (the interaction with Ub is not equal to being conjugated). At the minimum, the authors should provide a validation that some of the detected mitochondrial matrix proteins are indeed ubiquitinated. To this end, purified mitochondria could be used for the candidate protein IP under denaturing conditions and then blotted for the candidate protein and Ub.

__Response: __As shown in Table S2 and figure S7, forms of Ilv5, a mitochondrial protein, are ubiquitinated in WT and Drad6 cells. These modified forms of Ilv5 can be eluted from mitochondrial extracts of WT and Drad6 cells. However, the ubiquitination of ilv5 is not dependent or effected by the Drad6 mutation. We cannot be sure that we will be able to detect a protein with ubiquitin modifying activity which functions solely on certain proteins in mitochondria.

Figure 5. The knock-out of the E2 Rad6 causes a change in the mitochondria ubiquitination pattern. This is an interesting observation, but again it does not prove that the change in the mitochondrial ubiquitination is due to the activity of Rad6 inside of the mitochondria, as opposed to ubiquitination of the OMM proteins or contaminating fractions. One also wonders why overexpression of mitochondria-targeted Ub would be necessary to detect the ubiquitination if this process was physiologically relevant, especially given that detecting endogenous Ub is not challenging. Furthermore, the apparent increase in ubiquitination in E2 mutant cells (Fig. 5) should also be addressed in more detail. Finally, data from one WB is shown, and quantification of several independent experiments should also be provided.

__Response: __We show in the MS that RAD6 is exclusively targeted to mitochondria (Su9MTS) while unimported molecules are degraded (SL17; degron). This hybrid Rad6 can restore the WT ubiquitin pattern, while a rad6 active site mutant cannot.

Figure 6. Can the authors provide Western blot data showing the expression of Rad6? Furthermore, quantifying these rescue experiments is necessary to make this conclusion more solid.

Response: Even though we did not succeed in making good Rad6 antisera, we can clearly detect Rad6-a fusion proteins (Figure 7B).

Figure 7. The authors found that preSu9-Rad6-α have problems being imported into the mitochondria matrix; therefore, they rebuild it as a preSu9-Rad6-α-SL17 protein. SL17 is a degron that targets the cytosolic protein (not imported into the mitochondria) to the proteasome and degraded (Figs. 7A-B-C). These issues could be a red flag for the rest of the manuscript, suggesting that other constructs (that were not critically evaluated for their localization in this work) could leak to different cellular compartments.

Response: The wording used by the reviewer is particularly disturbing since current understanding in cell biology of eukaryotic cells does not accept “leaking” of proteins to different cellular compartments. One wouldn’t want DNAses, RNAses, Proteases etc leaking from one compartment to another. The localization of proteins to different cellular compartments involves very precise signals on the proteins, and specific cellular components, such as translocases, are required to target proteins to their exact destination. This is true for Rad6; it contains an MTS like sequence which when removed blocks import of the protein into mitochondria. Rad6 according to our analysis is an eclipsed dual targeted protein, so it no surprise that it is in two compartments and the trick with the SL17 degron solves the problem.

The manuscript needs to be carefully edited, some references are in the not correct format, and there are issues with figure labels.

Response: Careful editing will be undertaken as suggested by the reviewer.

CROSS-CONSULTATION COMMENTS I agree with a great summary by reviewer 1. This discovery should be validated by top-quality data.

Reviewer #2 (Significance (Required)):

In the manuscript by Yu et al., the authors test the concept that certain proteins are unevenly distributed within distinct cell compartments. Due to this localization discrepancy, protein detection in some subcellular compartments can be "eclipsed" by a predominant subset of specific protein localizing in another cell compartment their actual distribution. Therefore, tiny amounts of physiologically relevant proteins could be biologically relevant. Still, their function in some locations can be overlooked (or eclipsed) because of the high expression level of the same protein in another subcellular compartment(s). Although, this concept is not particularly novel. For example, it is already known that many different proteins can localize to distinct cellular locations (e.g., permanent mitochondrial and peroxisomal localization of many proteins or transient localization of particular proteins to separate cell compartments). The authors apply a yeast system and an α-complementation assay to test further the role of such eclipsed proteins in mitochondrial biology. Specifically, they focus on the ubiquitin (Ub, or as abbreviated incorrectly in this manuscript; Ubi) conjugation pathway, components of which have never been convincingly shown to localize inside the mitochondria. This work proposes that certain ubiquitination events can occur inside yeast mitochondria. This work would represent a significant/exceptional discovery if supported by compelling data. However, the major problem with this work is that the conclusions are based on the ectopic expression of distinct proteins. This approach is not failproof in precise protein expression/delivery to the specific subcellular locations and is likely to result in a non-specific localization. Thus, the problem of eclipsed proteins is addressed by the methodology that may lead to the artificial generation of eclipsed overexpressed proteins. A more effective approach would be if the authors found a way to study this issue with endogenous proteins. The need for overexpression of mitochondria-targeted ubiquitin makes it challenging to reconcile the physiological role of these fundings. In addition, some critical technical issues and omissions further reduce the potential impact of this work (see Specific comments above). For example, strong evidence of mitochondria fraction purity and additional evidence that all the essential constructs used in this work are not misdirected to a different compartment are needed.

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

Summary: In this study, the authors detected a set of components of a ubiquitination system in the mitochondrial matrix in budding yeast using the subcellular compartment-dependent α-complementation assay. The authors detected the conjugates of mitochondrial targeting signal sequence-directed HA-Ub (preSu9-HA-Ub) in the mitochondrial matrix. The immunoprecipitates of the preSu9-HA-Ubi conjugates were highly enriched for the mitochondrial matrix proteins. Subsequently, the authors focused on the Rad6 E2 ubiquitin conjugating enzyme in the mitochondrial matrix and evaluated its inactivation-altered ubiquitination pattern in the organelle. The authors conclude that ubiquitination occurs in the mitochondrial matrix because of the eclipsed targeted components of the ubiquitination machinery.

Major comments: The authors argued that the proteins that were modified with preSu9-HA-Ubi, which was forced to be imported into the mitochondria, are present in the mitochondrial matrix, because these species are resistant to trypsin digestion. However, it was possible that they formed severe aggregates on the cytosolic surface of the mitochondria, and hence, were resistant to the proteinase. In other words, a small amount of proteins that were not imported into the mitochondria could be deposited on the cytosolic surface of the mitochondria, where they were modified with preSu9-HA-Ubi by cytosolic Rad6. To confirm if the preSu9-HA-Ubi-modified proteins were really present in the mitochondrial matrix, they should perform the protease protection assay in the presence of an appropriate detergent (Figure 3D). In addition, subcellular fractionation of the organelle by density gradient centrifugation, indirect immunofluorescence microscopic analysis of the preSu9-HA-Ubi conjugates, and/or experiments on the in vitro import of preSu9-HA-Ubi and Rad6 into the mitochondria would strongly support the authors conclusion. Other experiments that might support the authors conclusion would be to test whether the band pattern for the preSu9-HA-Ubi conjugates changes when the mitochondrial import is impaired.

Response: We will attempt to perform 1) Protease protection assay in the presence of a detergent (Figure 3D). 2) Subcellular fractionation of the organelle by density gradient centrifugation. 3) In vitro import of Rad6 into the mitochondria.

Minor comments: In Figure 3B, the molecular weight distributions of the preSu9-HA-Ubi conjugates and those of the preFum-HA-Ubi conjugates are different. Is there any reason for this difference?

In Figure 3E, the position of "-" (MG132) for lane 1 is not correct.

In Figure 6A: The band pattern for preSu9-HA-Ubi (lane 13) in the rad6-delta cells expressing Ubc8-alpha is different from that of the wild-type cells expressing Ubc8-alpha (lane 12) as well as that obtained from the rad6-delta cells harboring empty plasmids (lane 9). Is there any explanation for this observation?

In Figure 7B and S6: The level of preSu9-Rad6-alpha-SL17 in the rad6-delta cells is always lower than that in the wild-type cells (compare lanes 13 and 10 in Figure 7B, and lanes 13 and 12 in Figure S6). Is there any explanation for this observation? The protease protection assay (with detergent control) is needed to fully confirm that preSu9-Rad6-alpha-SL17 is present in the mitochondria.

In Figure S7, the authors presented the matrix proteins, Ilv5 and Aco1, detected in the preSu9-HA-Ubi IPed samples and described this observation in the main text. However, the authors also showed the blots for Idh1 and Fum1, which were also pulled down with preSu9-HA-Ubi from the WT cells more than from the rad6-delta cells. Is this correct? If so, please elucidate this observation in the main text.

Figure 8D and 8E are not cited in the main text. Although there are no explanations for these figures in the main text, it looks like Rad6-deltaN11-alpha resides in the mitochondrial fraction. However, the alpha-complementation assay suggests that it resides in the cytosol. Please explain this discrepancy.

First page of the discussion section, item 6): E2 Rad6, but not E3 Rad6?

Figure S7: HA-Ub (cytosolic form) control is needed in addition to the empty vector control.

Figure S7, left panel: There is an unnecessary line break in "Hsp60" and "Ilv5."

Figure S7, right panel: There is an unnecessary line break in "Hsp60."

CROSS-CONSULTATION COMMENTS I agree with comments of reviewer 1 and 2. -Validation of the complementation assay. -I also think that it is important to address whether intra-mitochondrial ubiquitination can be observed with endogenous level of ubiquitin. If even a small amount of preSu9-HA-Ub is mistargeted to the cytosol, proteins at the cytosolic side of mitochondrial outer membrane could be ubiquitinated and detected in the mitochondrial fraction. -Preparation of mitochondria with more sophisticated purification methods (i.e. high resolution density gradient) would be needed to separate mitochondria from ER and other organelles. -More information is needed in the materials and methods section.

Reviewer #3 (Significance (Required)): Significance Although the results are interesting and very important, as mentioned in the major comments section, additional experiments are needed to support their model. However, researchers working on the mitochondrial biology and ubiquitin systems might be interested in and influenced by the reported findings.

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

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

In this study, the authors detected a set of components of a ubiquitination system in the mitochondrial matrix in budding yeast using the subcellular compartment-dependent α-complementation assay. The authors detected the conjugates of mitochondrial targeting signal sequence-directed HA-Ub (preSu9-HA-Ub) in the mitochondrial matrix. The immunoprecipitates of the preSu9-HA-Ubi conjugates were highly enriched for the mitochondrial matrix proteins. Subsequently, the authors focused on the Rad6 E2 ubiquitin conjugating enzyme in the mitochondrial matrix and evaluated its inactivation-altered ubiquitination pattern in the organelle. The authors conclude that ubiquitination occurs in the mitochondrial matrix because of the eclipsed targeted components of the ubiquitination machinery.

Major comments:

The authors argued that the proteins that were modified with preSu9-HA-Ubi, which was forced to be imported into the mitochondria, are present in the mitochondrial matrix, because these species are resistant to trypsin digestion. However, it was possible that they formed severe aggregates on the cytosolic surface of the mitochondria, and hence, were resistant to the proteinase. In other words, a small amount of proteins that were not imported into the mitochondria could be deposited on the cytosolic surface of the mitochondria, where they were modified with preSu9-HA-Ubi by cytosolic Rad6. To confirm if the preSu9-HA-Ubi-modified proteins were really present in the mitochondrial matrix, they should perform the protease protection assay in the presence of an appropriate detergent (Figure 3D). In addition, subcellular fractionation of the organelle by density gradient centrifugation, indirect immunofluorescence microscopic analysis of the preSu9-HA-Ubi conjugates, and/or experiments on the in vitro import of preSu9-HA-Ubi and Rad6 into the mitochondria would strongly support the authors conclusion. Other experiments that might support the authors conclusion would be to test whether the band pattern for the preSu9-HA-Ubi conjugates changes when the mitochondrial import is impaired.

Minor comments:

• In Figure 3B, the molecular weight distributions of the preSu9-HA-Ubi conjugates and those of the preFum-HA-Ubi conjugates are different. Is there any reason for this difference?

• In Figure 3E, the position of "-" (MG132) for lane 1 is not correct.

• In Figure 6A: The band pattern for preSu9-HA-Ubi (lane 13) in the rad6-delta cells expressing Ubc8-alpha is different from that of the wild-type cells expressing Ubc8-alpha (lane 12) as well as that obtained from the rad6-delta cells harboring empty plasmids (lane 9). Is there any explanation for this observation?

• In Figure 7B and S6: The level of preSu9-Rad6-alpha-SL17 in the rad6-delta cells is always lower than that in the wild-type cells (compare lanes 13 and 10 in Figure 7B, and lanes 13 and 12 in Figure S6). Is there any explanation for this observation? The protease protection assay (with detergent control) is needed to fully confirm that preSu9-Rad6-alpha-SL17 is present in the mitochondria.

• In Figure S7, the authors presented the matrix proteins, Ilv5 and Aco1, detected in the preSu9-HA-Ubi IPed samples and described this observation in the main text. However, the authors also showed the blots for Idh1 and Fum1, which were also pulled down with preSu9-HA-Ubi from the WT cells more than from the rad6-delta cells. Is this correct? If so, please elucidate this observation in the main text.

• Figure 8D and 8E are not cited in the main text. Although there are no explanations for these figures in the main text, it looks like Rad6-deltaN11-alpha resides in the mitochondrial fraction. However, the alpha-complementation assay suggests that it resides in the cytosol. Please explain this discrepancy.

• First page of the discussion section, item 6): E2 Rad6, but not E3 Rad6?

• Figure S7: HA-Ub (cytosolic form) control is needed in addition to the empty vector control.

• Figure S7, left panel: There is an unnecessary line break in "Hsp60" and "Ilv5."

• Figure S7, right panel: There is an unnecessary line break in "Hsp60."

CROSS-CONSULTATION COMMENTS

I agree with comments of reviewer 1 and 2.

• Validation of the complementation assay.
• I also think that it is important to address whether intra-mitochondrial ubiquitination can be observed with endogenous level of ubiquitin. If even a small amount of preSu9-HA-Ub is mistargeted to the cytosol, proteins at the cytosolic side of mitochondrial outer membrane could be ubiquitinated and detected in the mitochondrial fraction.
• Preparation of mitochondria with more sophisticated purification methods (i.e. high resolution density gradient) would be needed to separate mitochondria from ER and other organelles.
• More information is needed in the materials and methods section.

Significance

Significance

Although the results are interesting and very important, as mentioned in the major comments section, additional experiments are needed to support their model. However, researchers working on the mitochondrial biology and ubiquitin systems might be interested in and influenced by the reported findings.

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

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In the manuscript by Yu et al., the authors test the concept that certain proteins are unevenly distributed within distinct cell compartments. Due to this localization discrepancy, protein detection in some subcellular compartments can be "eclipsed" by a predominant subset of specific protein localizing in another cell compartment their actual distribution. Therefore, tiny amounts of physiologically relevant proteins could be biologically relevant. Still, their function in some locations can be overlooked (or eclipsed) because of the high expression level of the same protein in another subcellular compartment(s). Although, this concept is not particularly novel. For example, it is already known that many different proteins can localize to distinct cellular locations (e.g., permanent mitochondrial and peroxisomal localization of many proteins or transient localization of particular proteins to separate cell compartments). The authors apply a yeast system and an α-complementation assay to test further the role of such eclipsed proteins in mitochondrial biology. Specifically, they focus on the ubiquitin (Ub, or as abbreviated incorrectly in this manuscript; Ubi) conjugation pathway, components of which have never been convincingly shown to localize inside the mitochondria. This work proposes that certain ubiquitination events can occur inside yeast mitochondria. This work would represent a significant/exceptional discovery if supported by compelling data. However, the major problem with this work is that the conclusions are based on the ectopic expression of distinct proteins. This approach is not failproof in precise protein expression/delivery to the specific subcellular locations and is likely to result in a non-specific localization. Thus, the problem of eclipsed proteins is addressed by the methodology that may lead to the artificial generation of eclipsed overexpressed proteins. A more effective approach would be if the authors found a way to study this issue with endogenous proteins. The need for overexpression of mitochondria-targeted ubiquitin makes it challenging to reconcile the physiological role of these fundings. In addition, some critical technical issues and omissions further reduce the potential impact of this work (see Specific comments below). For example, strong evidence of mitochondria fraction purity and additional evidence that all the essential constructs used in this work are not misdirected to a different compartment are needed.

Specific comments

1. The authors should demonstrate beyond doubt that the ω components of their assay (ω-C, which supposedly stays in the cytosol-ONLY and the ω-M component, which seemingly remains in the mitochondria-ONLY) are in the compartment that the authors claim. These two proteins are transfected into yeast cells and overexpressed. Therefore it is possible that they leak to other, not intended, subcellular compartments. The authors assume that ω-M and ω-C are exclusively located either in the mitochondria or the cytosol. However, this should be shown as validation of the assay. The indicated reference from 2005 (Ref.13) and others are irrelevant since assays have variations and are often researcher/lab dependent. This validation is very important since a misallocation of the overexpressed ω-M or ω-C, leaking into other subcellular compartments, may cause misdetection of the α-constructs.
2. It is not surprising that Ub conjugates are detected in mitochondrial fractions. It could be due to ubiquitination of the OMM (coming from the cytosol) or perhaps since the subcellular fractions were not pure mitochondria free from contamination (the likely culprit could be the ER). The mitochondrial fractions in this work were obtained by 10,000 g separation between cytosolic and mitochondrial crude fractions. Indeed, these 10,000 g crude fractions are highly impure with membranes from other compartments (i.e., microsomes, lysosomes, and so on). Therefore, more sophisticated purification methods should be used. In addition, the authors should also test these fractions for non-mitochondrial proteins from other membrane organelles.
3. Figure 2. Coomassie blue staining does not show any signal in the "M" fraction. It can be interpreted that the authors do not get any mitochondria there, and therefore the lack of Ub signal is due to the absence of the protein in the samples. Using the same amount of protein from each fraction would probably reduce the necessity of 15x enrichment.
4. Figure 3. It is puzzling why the HA-UBQ presence is so strong in the crude mitochondrial fraction, but the preSu9-HA-Ub signal (mito-matrix) is comparatively weak. These data suggest that the crude mito-fraction could be highly contaminated with OTHER membranes. On the other hand, the preSu9-HA-UBQ signal is no more than 1-5% of the total mitochondrial signal. The high enrichment of the HA-Ubi in both cytosols and the mitochondria could indicate the OMM ubiquitination or (again) contamination by other compartments. The constructs with MTS are detected in the mitochondria. However, the localization of tagged MTS-Ubi in a non-targeted compartment (e.g., cytosol) should be excluded by additional exposure times. Because the manuscript talks about eclipsed proteins, this is important.
5. Figure 3C-E. These data indeed suggest that the Ub-conjugates could be formed inside the mitochondria. However, the above-discussed possibility that other than mitochondria compartments co-sediment in the 10,000g fractions makes the data interpretation highly challenging.
6. Figure 4. Unsurprisingly, mitochondrial targeting of Ub leads to detecting some co-immunoprecipitating mitochondrial proteins. However, these data do not support the notion that Ub conjugation machinery acts inside the mitochondria and that the target proteins are indeed conjugated with Ub (the interaction with Ub is not equal to being conjugated). At the minimum, the authors should provide a validation that some of the detected mitochondrial matrix proteins are indeed ubiquitinated. To this end, purified mitochondria could be used for the candidate protein IP under denaturing conditions and then blotted for the candidate protein and Ub.
7. Figure 5. The knock-out of the E2 Rad6 causes a change in the mitochondria ubiquitination pattern. This is an interesting observation, but again it does not prove that the change in the mitochondrial ubiquitination is due to the activity of Rad6 inside of the mitochondria, as opposed to ubiquitination of the OMM proteins or contaminating fractions. One also wonders why overexpression of mitochondria-targeted Ub would be necessary to detect the ubiquitination if this process was physiologically relevant, especially given that detecting endogenous Ub is not challenging. Furthermore, the apparent increase in ubiquitination in E2 mutant cells (Fig. 5) should also be addressed in more detail. Finally, data from one WB is shown, and quantification of several independent experiments should also be provided.
8. Figure 6. Can the authors provide Western blot data showing the expression of Rad6? Furthermore, quantifying these rescue experiments is necessary to make this conclusion more solid.
9. Figure 7. The authors found that preSu9-Rad6-α have problems being imported into the mitochondria matrix; therefore, they rebuild it as a preSu9-Rad6-α-SL17 protein. SL17 is a degron that targets the cytosolic protein (not imported into the mitochondria) to the proteasome and degraded (Figs. 7A-B-C). These issues could be a red flag for the rest of the manuscript, suggesting that other constructs (that were not critically evaluated for their localization in this work) could leak to different cellular compartments.
10. The manuscript needs to be carefully edited, some references are in the not correct format, and there are issues with figure labels.

CROSS-CONSULTATION COMMENTS

I agree with a great summary by reviewer 1. This discovery should be validated by top-quality data.

Significance

In the manuscript by Yu et al., the authors test the concept that certain proteins are unevenly distributed within distinct cell compartments. Due to this localization discrepancy, protein detection in some subcellular compartments can be "eclipsed" by a predominant subset of specific protein localizing in another cell compartment their actual distribution. Therefore, tiny amounts of physiologically relevant proteins could be biologically relevant. Still, their function in some locations can be overlooked (or eclipsed) because of the high expression level of the same protein in another subcellular compartment(s). Although, this concept is not particularly novel. For example, it is already known that many different proteins can localize to distinct cellular locations (e.g., permanent mitochondrial and peroxisomal localization of many proteins or transient localization of particular proteins to separate cell compartments). The authors apply a yeast system and an α-complementation assay to test further the role of such eclipsed proteins in mitochondrial biology. Specifically, they focus on the ubiquitin (Ub, or as abbreviated incorrectly in this manuscript; Ubi) conjugation pathway, components of which have never been convincingly shown to localize inside the mitochondria. This work proposes that certain ubiquitination events can occur inside yeast mitochondria. This work would represent a significant/exceptional discovery if supported by compelling data. However, the major problem with this work is that the conclusions are based on the ectopic expression of distinct proteins. This approach is not failproof in precise protein expression/delivery to the specific subcellular locations and is likely to result in a non-specific localization. Thus, the problem of eclipsed proteins is addressed by the methodology that may lead to the artificial generation of eclipsed overexpressed proteins. A more effective approach would be if the authors found a way to study this issue with endogenous proteins. The need for overexpression of mitochondria-targeted ubiquitin makes it challenging to reconcile the physiological role of these fundings. In addition, some critical technical issues and omissions further reduce the potential impact of this work (see Specific comments above). For example, strong evidence of mitochondria fraction purity and additional evidence that all the essential constructs used in this work are not misdirected to a different compartment are needed.

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

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

In this manuscript, Zhang et al. investigate whether ubiquitination occurs inside mitochondria of the budding yeast S. cerevisiae. They first observe thanks to a sensitive complementation assay that several components of the yeast ubiquitination (and deubiquitination) machinery can localize inside mitochondria. To be able to specifically probe ubiquitin conjugates assembled inside mitochondria they fused HA-tagged ubiquitin to a mitochondrial targeting sequence. Using this construct, they demonstrate that ubiquitin conjugates can be assembled in mitochondria. A series of elegant experiments demonstrates that the pattern of ubiquitin conjugates depends on the mitochondrial localization and the activity of the ubiquitin conjugating enzyme Rad6. Altogether, these results convincingly demonstrate that ubiquitination can occur inside yeast mitochondria when ubiquitin is intentionally targeted inside this organelle. It however remains unclear whether mitochondrial ubiquitination occurs in endogenous conditions (without targeting ubiquitin into this compartment) and whether it affects mitochondrial functions.

Major comments:

1) The materials and methods section is lacking important information (western blot protocol, details of antibodies, strains, plasmids...). It is thus difficult to evaluate how several experiments were performed and how their design (e.g. the promoters chosen to express tagged proteins) could impact the interpretation of the results. This is a major issue that needs to be corrected. The main text should also explicitly indicate whether tagged proteins used in the alpha-complementation assay are overexpressed or not.

2) Despite the previous comment, the data presented in the manuscript convincingly demonstrate that multiple components of the ubiquitination machinery can localize within mitochondria and that ubiquitin conjugates can be assembled in mitochondria when ubiquitin is modified to be intentionally targeted into this compartment. However, little data is shown to support the hypothesis that ubiquitin conjugates can be assembled in mitochondria when ubiquitin is not fused to a mitochondrial targeting sequence. Thus, in my opinion, the evidences presented in the current manuscript are not sufficient to conclude that ubiquitin conjugates are assembled in mitochondria in endogenous conditions (as this is done implicitly). Additional evidences are needed to draw this conclusion (see some experimental suggestions hereafter). Without further evidences, the speculative aspects of the claim that "ubiquitination occurs in the mitochondrial matrix" should be discussed explicitly.

3) The authors used a mass spectrometry approach to identify mitochondrial ubiquitination substrates. However, they have not yet succeeded in identifying a substrate whose modification is specifically regulated by a given component of the mitochondrial ubiquitination machinery. They have also not identified a phenotype or process impacted by mitochondrial ubiquitination. Thus, at this stage, the biological consequences of mitochondrial ubiquitination remain elusive.

4) The authors have not directly investigated whether ubiquitin itself (without a mitochondrial targeting sequence) localizes in mitochondria. I encourage them to address this question since it would provide an important piece of evidence suggesting that mitochondrial ubiquitination can occur in endogenous conditions. This could be done using the alpha-complementation assay and the results could be presented within Figure 1. Ideally this experiment should be performed without overexpressing ubiquitin. Note that if the authors decide to use a C-terminally tagged form of ubiquitin for this experiment, the GG motif of ubiquitin should be mutated to avoid cleavage of the alpha tag by cellular DUBs. This form of ubiquitin will not be conjugatable, but this is not an issue for this experiment since its aim is to determine whether ubiquitin can be targeted to mitochondria, not to probe conjugates.

5) In the top panels of Figure 2 and S1, free ubiquitin is well detectable in the total and cytosolic fractions. It is however not clear to me whether it is also detectable in the concentrated mitochondrial fraction. If yes and if it would be resistant to trypsin digestion, it would provide an additional evidence that endogenous ubiquitin can be targeted to the mitochondrial matrix (see previous comment).

6) The data shown in the top panel of Figure 2 and S1 also suggest that free ubiquitin is less concentrated in mitochondria than in the cytosol (since it is more difficult to detect in the concentrated mitochondrial fraction than in the cytosolic fraction, see previous comment). It is thus possible that the use of preSu9-HA-Ubi (or preFum1-HA-Ubi) lead to an artificially high intra-mitochondrial concentration of free ubiquitin. As the concentration of free ubiquitin is known to impact ubiquitination processes, I encourage the authors to compare the relative levels of free ubiquitin present in the mitochondrial fraction prepared from wt and preSu9-HA-Ubi (or preFum1-HA-Ubi) expressing cells. If free ubiquitin is detectable in mitochondrial fractions and resistant to trypsin (see previous comment), this could be done by repeating the experiment shown in Figure 3B and probing the blot with an antibody that recognizes free ubiquitin.

7) I strongly encourage the authors to provide more data indicating that "ubiquitination occurs in mitochondria" by performing experiments that do not rely on the use of the preSu9-HA-Ubi or other forms of ubiquitin that are intentionally targeted to mitochondria. For instance, they could analyse the pattern of HA-Ubi conjugates of trypsin digested mitochondrial fractions prepared from wt, rad6-delta, and rad6-delta complemented with preSu9-Rad6-alpha-SL17. Note that if trypsin digested mitochondrial fractions are too contaminated by ubiquitinated proteins present outside mitochondria to perform this experiment, the authors may use the unspecific DUB Usp2 as an alternative protease to strip ubiquitinated proteins from the mitochondria periphery.

Minor comments:

1) Overall, the manuscript is well organized and easy to follow. The text is clearly written; the figures are well annotated.

2) The authors should provide full images of all the blots with anti-ubiquitin and anti-HA antibodies so that one can see the bands corresponding to free ubiquitin (or free HA-Ubi). For instance, in Figure 3B, it is not possible to see the presence (or absence) of the band corresponding to free HA-Ubi because the very bottom of the image is cut.

3) The authors should indicate whether the MTS of Su9 (and Fum1) are expected to be cleaved after import of preSu9-HA-Ubi (and preFum1-HA-Ubi) in mitochondria. They should also label on the corresponding immunoblots the presence (or absence) of the band corresponding to the free preSu9-HA-Ubi (and preFum1-HA-Ubi) (or HA-Ubi if the MTS is expected to be cleaved from these constructs).

4) In Figure 3B, the ubiquitin conjugates produced with preSu9-HA-Ubi and preFum1-HA-Ubi have different migration patterns. I think this should be explicitly mentioned and discussed. Could it be due to the presence of lysine residues in the Su9 or Fum1 MTS that could lead to the assembly of artificial ubiquitin chains?

5) The authors indicate that "endogenous Rad6 [...] is expressed at very low levels and can hardly be detected in the mitochondrial fraction by WB (Figure S5)". I did not manage to observe the band corresponding to endogenous Rad6 in the mitochondrial fraction in the pdf. The authors should provide a more contrasted or better quality image.

CROSS-CONSULTATION COMMENTS

• I agree with reviewer 2 that proper validation of the complementation assay is crucial for this manuscript. I was myself wondering whether it uses endogenously tagged proteins or whether it is based on an overexpression system. I imagine this information will be detailed in the manuscript in preparation mentioned by the authors. I am therefore wondering whether it would be possible to ask the authors to provide the draft of this manuscript (or at least the validation part).

• I agree with most comments of reviewer 3. Regarding the hypothesis that preSu9-HA-Ubi could form aggregates on the cytosolic surface of the mitochondria, I think that the results presented on Figure 7B rather argue against it (since they indicate that Rad6 localized inside mitochondria can restore the pattern of ubiquitin conjugates). That's why (in my opinion) the major question the author now need to adress is whether intra-mitochondrial ubiquitination occurs in endogenous conditions (ie without forcing ubiquitin into this compartment and without E2 or E3 overexpression).

Significance

The finding that ubiquitination occurs inside mitochondria would be an important conceptual advance, which would open new perspectives both for ubiquitination and mitochondrial biology research. However, the significance of the current manuscript is limited because the presented evidences heavily rely on the use of artificial conditions (ubiquitin tagged with a mitochondrial-targeting sequence) that may trigger irrelevant ubiquitination events. The significance would be much higher if the authors would provide further evidences indicating that intra-mitochondrial ubiquitination occurs in endogenous conditions and/or if they had identified a mitochondrial process specifically impacted by mitochondrial ubiquitination.

Expertise of the reviewer: Ubiquitination, Yeast biology, protein-protein interactions. No specific expertise in mitochondrial biology

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

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

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1:

---

1) The authors could consider qualifying the observations as preliminary as no

mechanistic data or longer-term pathophysiology is investigated. Indeed, the latter is well

beyond the current scope and may require generation of cell-type specific STING ki mice.

• *

Thank you for the comment. We have qualified our observations as preliminary (line 662).

Indeed, generating cell-type specific STING ki mice is part of our future plans.

2) The authors consistently write "NF-kB/inflammasomes" - these two pathways (although

related) are quite distinct and should not be lumped together in such a way.

• *

Thank you for this important note, we now corrected the text (for example see section headings in the Results section, lines 338 and 415).

3) Line 79: "NRLP3" should be corrected to NLRP3.

• *

Line 210: age of "adult mice" in weeks should be state in the text and figure legend.

Thank you, corrected.

4) Line 262: In Figure 3B and D the images look very different and there is no indication of

what a positive inclusion is? This should be indicated on the image.

• *

Thank you for the suggestion. We replaced the corresponding panels with new images, where we show the nuclei with blue, and the Thioflavin S staining with magenta pseudo-color (current Figure 3E). We marked the outline of Thioflavin S positive cells with yellow. An inset showing the magnification of some neurons with inclusions is also presented.

5) Line 280: The data of Ifi44 should also be mentioned in the text.

• *

Thank you. We performed new experiments to show the gene expression changes in the

striatum and in the substantia nigra, therefore majority of the gene expression data from the cortex has been moved to the supplementary material (supplemental Figures 3 and 5), and is not discussed in detailed in the current manuscript.

6) Line 290: Figure 4, Examining IL-1B and Caspase-1 transcripts is not a readout of

inflammasome activation. pro-IL1B is upregulated in response to NFkB activity. Inflammasome

activation is commonly examined in other methods e.g. via ASC puncta formation (imaging

based), active IL-1B secretion (ELISA), Caspase-1 and IL-1B cleavage via western blot

Thank you for this suggestion, we performed new experiments and added the data as Figure 6.

• We performed Western blot analysis to detect IL-1β cleavage and NLRP3 proteins from the striatum (Figure 6C-E). 2) We quantified the number of ASC puncta within microglia and astroglia from striatal sections (Figure 6F-I). 3). 3) We also measured the protein levels of several additional immune mediators in the striatum of STING ki and KO animals (supplemental Figure 6, summary heatmap is on Figure 7A). 7) Line 310: The NF-kB subunit examined should be stated (p65?). Furthermore, IRF3

translocation might be a better readout for STING activation.

• *

We indeed detected the p65 subunit of NF-kB (antibody is listed in the supplemental Table), and now it is also indicated in the text (line 366). We also performed subcellular fractionation and quantified IRF3 in the nuclear and cytoplasmic fractions. The data is now added on Figure 6A, B.

8) Discussion: Given the findings here suggest a strong role for NF-kB, a short discussion

of IFN vs non-IFN responses from STING should be included. There have been a number of

seminal papers demonstrating the importance of non-IFN STING responses of late as well as

much evidence from SAVI mice to suggest some non-IFN driven pathologies.

• *

Thank you for the suggestion. The data on inflammasomes were given a separate section in the results (from line 395). In the discussion, from line 535 we discuss the IFN dependent response and from line 548 we discuss the non-IFN driven pathways.

9) Discussion: Is there any evidence from the human SAVI patients of neuroinflammation

etc. This should be mentioned either way in the discussion

• *

Thank you for this comment. The manifestation of neurological symptoms is not a core feature of the human SAVI disease. Some patients suffer from various neurological symptoms e.g. calcification of basal ganglia, spastic diplegia and episodes of seizure (Fremond et al., 2021). We inserted a short text in discussion (lines 532-534).

10) Discussion: There is a large body of work demonstrating STING-induced cell death in

numerous cell types. Despite this it is not mentioned nor discussed but should be. It could

represent how dopaminergic neurons are lost in the STING ki mice.

• *

Thank you for pointing out the gap in our discussion. We added additional text in lines 604-618.

11) The resolution/quality of some of the imaging is not great but this may be due to PDF

Compression

Thank you, we upload the figures with higher resolution.

Reviewer #2:

---

1) The authors base their conclusions (line 215-216) on the neuroinflammatory status of

their mice strongly on an assessment of the Iba1 and GFAP-positive area fraction. Increase of

Iba1 and GFAP areas does not necessarily correlate with an increased cytokine production and

release by the cells. Therefore, in addition the measurement of cytokine mRNAs it would be

necessary to measure cytokines also on protein level (see also #4 and #5).

• *

Thank you for this suggestion, we measured the protein levels of several immune mediators with LEGENDplex™ assay from the striatum, and the new data are included as Figure 7A and supplemental Figure 6.

2) In the same context: Is the increase of Iba1 and GFAP- covered area due to increased

proliferation of microglia and astrocytes or due to increased expression of these markers in

activated glia? How is the number of Iba1/GFAP-positive cells affected?

• *

We quantified the number of glia cells in the striatum and in the substantia nigra of adult STING WT and STING ki mice, and, parallel with higher immunoreactivity for the corresponding markers, we detected increased number of cells as well. The quantifications are now included in supplemental Figure 1.

3) Nowadays we know that microglia and astrocytes can exist in a variety of activated

states which can be either beneficial or detrimental. An analysis of disease-associated

microglial markers (Keren-Shaul et al. 2017) would give a good picture of the state microglia are

in.

• *

Thank you for the suggestion. In addition to the panel of immune modulators at the protein level (supplemental Figure 6), we performed qPCR analysis of additional “M1” marker (Nos2) and additional “M2” markers (Il4, Fizz2, Ym1) (Gong et al., 2019). The data is included in Figure 7A and shown in supplemental figure 6. The findings are described from line 431.

4) It also would be of interest to determine which cell type is responsible for the observed

neurodegeneration. Which cytokines are released by microglia or astrocytes upon STING

activation? Even in vitro experiments would help here to get a more profound understanding.

• *

We agree with the suggestion, however, the further in vitro experiments are beyond the scopes of this study and will be the basis of a future project.

5) In line 273 the authors describe that STING is known to activate NFkB and the

inflammasome. As proof that this is also occurring in their mouse, they perform qPCR analysis

of whole brain IL-1b, TNF-a and Casp1 expression. While this analysis indicates that there is

indeed an increased mRNA production of proinflammatory cytokines in the brains of STING ki

mice, it does not give any indication whether the inflammasome is active or not. The inflammasome is a protein complex largely regulated on protein level. Meaning an assessment

of the cleavage of Caspase 1 on protein level or the presence of cleaved IL-1b in comparison to

uncleaved Pro-IL-1b by Western Blot as well as a staining for the number of inflammasomes

would be required to draw these conclusions.

• *

Thank you for the suggestion. We performed additional experiments: 1) Western blot to detect pro-IL1b and IL1b and NLRP3 proteins from the striatum (Figure 6C-E), and 2) we quantified the number of ASC puncta within microglia and astroglia from striatal sections (Figure 6F-I).

6) To conclude that NFkb/inflammasome pathway is the most active/crucial in astrocytes

(line 354) a staining for ASC inflammasomes would be of importance, especially as astrocytes

normally do not express NLRP3.

• *

Thank you for this comment. We stained brain sections for ASC specks and for microglia (Iba1) and astroglia (GFAP) markers (Figure 6F-I). Although amount of ASC specks in astroglia was lower than in microglia, we found still a substantial amount of ASC specks in astroglia in the brains of STING ki animals.

7) As already shown for ALS (Yu et al., 2020) and Parkin KO (Sliter et al. 2018), the authors want to

further assess the relevance of the STING pathway to PD (line 27-28). Therefore, an in-depth analysis of

key PD hallmarks beyond phosphorylated a-synuclein, loss the other was parkin/PINK related (so TDP

deleted) of TH-stained neurons and dopamine reduction is needed. In the discussion the authors

hypothesize that autophagy (line 467) may be linked to the observed phenotype. Therefore,

assessment of autophagy/mitophagy as well as mitochondrial dysfunction and mtDNA should

be analysed. In the same line of thought it would be important to know if and how the observed

dopamine reduction effects mouse behaviour, thus mice should be subjected to the Rotarod or

pole or beam walk test.

• *

Thank you for these suggestions. In the work by Yu et al. and Sliter et al., the STING pathway was shown to mediate neurodegeneration resulting from TDP-43 pathology and mitochondrial damage. Our work is complementary by investigating the effects of constitutive activation of STING. We have therefore focused on the signaling pathways downstream of STING. As mentioned above, the most important next step will be to separate the contributions of neuronal and glial cells by generating cell type specific STING activation. Of course, it will be interesting to see at a later time point whether STING activation feeds back. We also speculate that STING activation may also cause TDP-43 pathology. Yet, this will be part of a future study. To acknowledge that the pathology is not specific to alpha-synuclein, we added a short statement from line 634.

With respect to the comprehensive analysis of the PD phenotype, our work includes the

classical parameters of TH neuron number, TH fiber density, dopamine concentration and

synuclein pathology. With respect to mouse behavior, we note that the STING ki mice have severe inflammation in the lung, kidney and other (peripheral) organs, reduced body weight and reduced lifespan (Luksch et al., 2019; Motwani et al., 2019; Siedel et al., 2020). Motor deficits cannot be attributed to dopamine neuron degeneration and for this reason were not included (stated in the Discussion, lines 624-625). In order to expand the description of the PD phenotype we now included measurements of cytosolic reactive oxygen species, mitochondrial oxygen species and nitric oxide, which result from inflammation and are known to affect dopaminergic neurons (new Figure 8).

Reviewer #3:

---

1) The method for quantification of TH-positive cells is not sufficient. They just described

how they stained every fifth sections but did not mention how they count. This is a critical point

and they should carefully provide information more than just referring their previous paper.

Counting of dopaminergic neurons and quantification of fibers was described in a dedicated section of the methods. This section has now been expanded (from line 154).

2) It is not persuasive that they did not investigate local inflammation in SN. They

presented increased microglia and astrocytes in the striatum but not analyzed these cells in SN

• *

Indeed, we measured neuroinflammation in the substantia nigra as well, however, although increased in STING ki mice, it was less pronounced than neuroinflammation in the striatum. We now include the quantification of area fraction as well as cell number counting of microglia and astroglia in the substantia nigra of STING WT and STING ki animals (supplemental Figure 1), and also the expression of inflammatory mediators in Figure 4.

3) In Figure 3, they analyzed alpha-synuclein phosphorylation and beta-sheet structure in

the striatum. This is funny from the aspect of Parkinson's disease, which dominantly affects SN.

They should perform similar experiments with SN samples. In a different aspect, the aggregates

detected by Thio S may not be alpha-synuclein and could be tau, TDP43 or other substances.

Phospho-synuclein of course does not mean aggregation, so they can consider electron

microscopy.

• *

We agree with the reviewer. To complement our data, we therefore performed solubility assay both from the striatum and from the substantia nigra to quantify the ratio of alpha-synuclein in the Triton X-100 soluble and insoluble fractions (Figure 3C, D) as previously (Szego et al., 2022; Szegő et al., 2019). Additionally, we quantified phosphorylated alpha-synuclein from the substantia nigra as well Figure 3A,B).

We also agree with the reviewer that the presence of Thioflavin S-positive inclusions may also contain other, beta-sheet forming proteins and noted this from line 634.

4) Figure 5, pSTAT3 increased in Iba1-negative cells, which seem neurons from the size of

nuclei. First, the authors should investigate the identity of pSTAT3-positive cells with GFAP and

MAP2. If pSTAT3 is actually increased in neurons, what does it mean in the pathology? For

instance, in viral infection, STAT3 activation triggers suicide of neurons to prevent further

proliferation of viral particles in neurons. Is it homologous or other function?

• *

Thank you for this suggestion. The brain sections were stained for Iba1 and GFAP. pSTAT3 nuclear staining indeed increased in non-glia cells, based on the morphology, we think in neurons. However, detailed characterization of the signal is out of the scopes of this (preliminary) study.

5) In Figure 6 and overall, cell types in which the activation of three signaling pathways,

were mixed up and hard to understand the actual situation in the brain.

• *

In our model, STING is activated in all cells. Consequently, we cannot determine the origin of immune mediators found elevated in the STING ki mice. This will require cell type specific STING activation. In order to react to the reviewer’s comment and be clearer, we have added more details about the brain region and age of mice used for each analysis also in the figures.

6) In the method section, the original paper for generation of heterozygous STING N153S

KI mice should be Warner et al, JEM 2017.

• *

We used a STING N153S ki mouse strain that was independently generated in the Technical University Dresden (Luksch et al., 2019).

7) NF-κB stains seem located in cytoplasm in Figure 5B.

• *

We agree: especially in the young STING ki mice, cytoplasmic NF-kB staining is increased

compared to STING WT mice. To quantify nuclear translocation, however, we counted the

number of those cells where NF-kB signal was overlapping with the nuclear Hoechst staining.

8) In Figure 4 and 6, why the authors evaluate gene expressions in frontal cortex instead of

SN or striatum.

• *

As noted in several comments, we show here that the STING-induced pathology involves

dopaminergic neurons, but believe that it is not specific for the dopaminergic system given that STING-ki is ubiquitously expressed. For practical reasons, we have used cortical samples for the expression analysis. For consistency, we now performed additional qPCR measurement from the striatum and from the substantia nigra and included them as new Figure 4 and supplemental Figure 6N-Q. The previous data from the cortex was moved to the supplemental Figures 3 and 5. Additionally, we measured the levels of several inflammatory modulators from the striatum of STING ki and KO animals (Figure 7A and supplemental figure 6A-M).

9) In some groups (Sting-ki;ifnar1-/- in Fig 6C, 6E), the values were separated to two

groups, which makes readers to doubt on soundness of their genotyping.

• *

Our genotyping protocol is highly standardized, and the genotype of the animals were correctly assigned. Here we provide an example of gel images showing the products after PCR reactions for the STING N153S allele (Figure 1a), STING WT allele (Figure 1b), Ifnara WT allele (Figure 1c) and lack of Ifnara allele (Figure 1d) of the same animals. We note that a bimodal distribution of phenotypes is often observed in Ifnar-/- mice.