4,754 Matching Annotations
  1. Last 7 days
    1. Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      A. General Statements

      We thank the reviewers for their constructive feedback. We have made significant revisions to the mathematical modelling section of the manuscript to address your concerns. Therefore, some of the specific issues and concerns raised in previous reviews no longer apply. Where that is the case, please see the relevant context in the revision as indicated in the point-by-point description section below. We summarize the key points in the revised manuscript as follows.

      1. The key finding of our study, involving experimental measurements and mathematical modelling, is plasticity in the MinD concentration gradient, which results from spatial differences in molecular interactions and is an intrinsic property of the Min system during cell growth. This study reveals not only the role of the MinD concentration gradient in modulating bacterial cell division site placement but also showcasing an example of cellular components in the form of a concentration gradient in fundamental cellular processes, a concept crucial in cell biology. This work provides conceptual advancement in a quantitative understanding of MinD oscillations in the cellular environment and provides implications for bacterial cell division regulation for further studies in the field.

      2. The reviewer requested clarification on the differences between our study and previous studies involving experimental measurements and mathematical modelling of Min oscillations in cells. We would like to emphasize that although the goal of the previous works was to measure the spatiotemporal distribution of oscillating MinD concentration gradients as a function of cell length, these works conceived the problem differently and therefore used different experimental designs and execution methods, which differentiates our key conclusions from theirs. This is also true for mathematical modelling. Although similar observations can be found in some respects, they are not directly comparable due to the different mathematics and assumptions used in the simulations. For example, our model was built to adequately investigate the biological question of the MinD concentration gradient during cell elongation but not to evaluate the impact of cell shape and confinement or the nucleation effect of MinD. Thus, our model cannot be generalized to other shapes, such as those observed in the study by Wu et al., 2015 (Wu et al, 2015). Therefore, we would like to draw attention to the experimental rigor and to the specific points and views that contribute to our understanding of Min systems. We now provide a comprehensive comparison between them in the Supplemental Information.

      3. We have re-run the simulation to refine and improve the modelling procedures and results, and the corresponding text and illustration are provided in the Results section of the main text (Lines 265-279, 614-653) and Fig. S6. In brief, we fixed the diffusion coefficients D_D and D_E from Meacci et al. (2006) (Meacci et al, 2006); the dissociation rate constant k_de from a previous simulation (Wu et al., 2015); and the experimentally measured MinD and MinE concentrations in this study. Meanwhile, the diffusion coefficients D_d and D_de were assumed values based on bacterial membrane protein diffusion (Schavemaker et al, 2018). This operation allowed us to probe for the general behaviours of the system. As a result, we were able to obtain a few parameter sets, including #2728, that generate features of the oscillation period, λ_N and I_Ratio, that highly mimic MinD oscillation in the cellular context (Figs. 4C, S7-9). We further tested the impact of different kinetic constants, k_de, k_dD, k_dE, k_D, and k_(ADP→ATP), which represent different molecular interactions influencing the oscillation period, λ_N and I_Ratio (Fig 4D-H). Our findings have provided us with a solid theoretical view of how oscillation features may be controlled by different molecular interactions. Furthermore, the modelling results help us understand the possible mechanisms associated with oscillation cycle maintenance and length-dependent variable concentration gradients.

      4. Regarding the inclusion or removal of results from more culture conditions, we decided to keep only one condition as in the previous version for the following reasons. In order to draw convincing conclusions, we consider it more important to characterize all aspects under the same growth condition and avoid manipulation. Therefore, the main conclusions are drawn from our experiments characterizing several aspects of MinD oscillations in cells growing with 0.4% glucose. In support of these observations, we decided to maintain only one other condition, 0.1% glucose. Further analysis of cells growing under other conditions will not change the main conclusions but will increase the difficulty of determining how the MinD concentration changes with cell growth.

      5. Studying the variable concentration gradient underlying the dynamic oscillations of the Min system may be of broad interest to cell biologists since the concentration gradient plays a fundamental role in various cellular processes, and the concept of concentration gradients is crucial in cell biology. Examples of related processes include passive and active transport, osmosis, cell signalling, and maintenance of cellular homeostasis. These processes allow cells to respond to their environment, regulate their internal conditions, and perform important functions required for survival and normal function. In addition, variable concentration gradients, characterized by the numerical descriptor λ_N and was reproduced in a simple mathematical model, demonstrate a nonlinear dynamics behaviour in physical biology. Therefore, the audience of this work can include the broader general audience of cell biology and physical biology rather than just the immediate specialized audience interested in the Min system. We will also reiterate the importance of specialized research, which often provides the basis for broader application and understanding.

      B. Point-by-point description of the revisions

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

      Summary: Parada et al. studied both experimentally and theoretically the MinD concentration distribution of Min waves during cell growth. The main finding was that (i) the gradient of MinD is steeper for longer cells and accordingly the MinD concentration at the middle of cell is lower, (ii) period of the oscillation is independent to the cell length, and (iii) those features are shared even under glucose starvation except the MinD gradient is steeper. (iv) Those results are supplemented by the analyses of the reaction-diffusion equations in which parameters that can reproduce the MinD concentration distribution are identified. I think the results are interesting; basically, as the cell grows, the contrast of the wave becomes clearer, such the MinD concentration at the cell centre decreases. The results may clarify the mechanism of FtsZ accumulation at the cell centre more quantitatively. The experiments were performed by measuring the fluorescent intensity of MinD during cell growth and analysing the intensity distribution along the long axis of the cell. The theoretical results were based on the analyses of the reaction-diffusion model. Both approaches are already well established and the results sound. Nevertheless, I do not think the novelty of this work is not well highlighted in the current manuscript; I think most of the results, except (iii) and (iv), have already been shown explicitly or implicitly in the previous studies. Min oscillations in a growing cell have been analysed both theoretically and experimentally in (Meacci 2005) and [1] (Fischer-Friedrich et al, 2010). The concentration distribution and period of the oscillation were measured. The complete results were presented in [2] (Meacci et al., 2006), and I am not aware of those results in scientific journals (the thesis is available online). Nevertheless, I think it is fair to cite those studies and compare the current results with them. In fact, in [2], it was shown that the concentration of MinD near the cell centre decreases as the cell grows, the total MinD concentration is approximately constant during the growth (therefore, the number of the molecules increases), and that the variance of the period becomes smaller as the cell grows. I do not think those previous studies spoil this work, and this work deserves publication somewhere. Still, the authors should highlight the novelty of this study more clearly.

      ANS: We thank the reviewer for recognizing the soundness of our experimental and theoretical approaches and results. The key finding of our study, involving experimental measurements and mathematical modelling, is plasticity in the MinD concentration gradient, which results from spatial differences in molecular interactions and is an intrinsic property of the Min system during cell growth. This study reveals not only the role of the MinD concentration gradient in modulating bacterial cell division site placement but also showcasing an example of cellular components in the form of a concentration gradient in fundamental cellular processes, a concept crucial in cell biology. We believe that the established techniques and methods are integral to a broad range of works and provide confidence in improving them and using them to test hypotheses and obtain results. We also appreciate the reviewer for pointing out that Meacci's PhD thesis entitled "Physical aspects of Min oscillations in Escherichia coli" (Meacci & Kruse, 2005) is available online for public access. This thesis, along with two publications (Meacci & Kruse, 2005) (Meacci et al., 2006), explored Min oscillations in growing cells and used mathematical models. These two published works are cited in the previous version of the manuscript because we agree that these earlier works provide valuable context. As recommended, we went through these works again and the work by Fischer-Friedrich et al. (2010) (Fischer-Friedrich et al., 2010) to compare their wet experiments and mathematical models with ours, which are detailed in the Supplemental Information (Lines 26-147). Here, we emphasize that although the published works and our work set the goal of measuring the spatiotemporal distribution of oscillating MinD concentration gradients as a function of cell length, we conceived the problem differently and therefore used different experimental designs and analysis approaches, which have led to the key conclusions that differentiate our work from theirs.

      Major comments: (i) In (Meacci 2005) and [1,2], it was claimed that the standard deviation of the period is comparable with the mean period, particularly for the shorter cell. Therefore, they did not claim the period is independent to the cell length. As far as I understood, the variance arises from the variance of the total protein concentration in the assemble of cells. I am wondering how the authors are able to conclude the constant period in different cell length. I also point out that in the theoretical part of (Meacci 2005), the period is, in fact, increasing as the cell grows and suddenly decreases at the length in which cell division occurs.

      ANS: In our experiments, we found that the oscillation periods ranged from 36.8 to 65.6 sec, as measured from a population of cells (length of 1.9-4.5 µm; main text, Fig. 1E). Moreover, the standard deviations of the period ranged from 5.4% to 34.8% of the period, with larger standard deviations more common in shorter cells (Fig. 1D), indicating that regular interpolar oscillations are more likely to occur in longer cells. This observation echoes the study by Fischer-Friedrich et al. (2010) (Fischer-Friedrich et al., 2010), who reported stochastic switching MinD oscillation between two cell poles in cells below 2.5 μm. MinD starts to oscillate regularly from pole-to-pole between 2.5-3 μm with an oscillation period of 80 sec. Above 3.5 μm, MinD invariably undergoes regular oscillation with an initial period of 87 sec and then decreases to 70 sec at the end. In their study, they focused on the length-dependent switching from stochastic to regular oscillation states and speculated that the amount of MinE bound to the membrane critically influenced the shift from stochastic to regular interpolar oscillations. In addition, their observation of a longer period at the initial phase and a shorter period after the cells grew beyond 3.5 μm somewhat coincided with our simulation results, as shown in Fig. 4C-H, left. In Meacci's work (Thesis: Figure 2.14; Meacci and Kruse (2005) (Meacci & Kruse, 2005): Figure 5(b)), the temporal oscillation periods were measured from 40 to 120 sec when focusing on cells with lengths similar to those in our measurements (black dots in Meacci's chart). Our measurements of oscillation periods clearly show much smaller fluctuations than those in Meacci's study and are more comparable to Fischer-Friedrich's measurements. Differences can arise across different bacterial strains and culture conditions that may significantly affect the amount and quality of protein expressed in individual studies. In short, all three works differ in terms of experimental design and execution. Although similar observations can be found in some aspects, they are not directly comparable. Therefore, we would like to draw attention to the experimental rigor and specific points and views that contribute to our understanding of the Min system. We have changed the wording from 'constant period' to 'fairly stable period' throughout the manuscript. This description is based on our experimental measurements (Fig. 1D, E) and is also supported by our mathematical modelling (Fig. 4C-H, left). In response to the statement from the theoretical model of (Meacci & Kruse, 2005): "the period is increasing as the cell grows and suddenly decreases at the length in which cell division occurs." First, our simulation results revealed a mild increase in the oscillation period during cell elongation (Fig. 4C). The increase is adjustable by varying the reaction rate constants in the simulation (Fig. 4D-H). Second, although we did not simulate dividing cells, our experimental measurements clearly showed that this period increased in newborn cells (Fig. S4). As mentioned above, although similar observations can be found in different studies, they are not directly comparable because the experiments were performed differently for different purposes. We have added comparison of different models in the Supplemental Information (Lines 26-147).

      (ii) I do not think the explanations of the reaction-diffusion model were well described. The authors mentioned that they studied a one-dimensional model and used the delta function to describe the membrane reaction. Did the authors study 1D cytosol and 0D membrane? Then, why the surface diffusion term exists in (4) and (5)? I believe the authors simply assumed that both the membrane and the cytosol are 1D (with larger diffusion constants for cytosolic Min concentrations). Then, the delta functions in (1)-(5) are not necessary. In (Wu 2015), the delta function was used in order to treat a 2D membrane embedded in 3D space.

      Besides that, there is no description of the initial conditions for the concentration fields to solve the reaction-diffusion equations. I think the description of the no-flux boundary condition is better put in the Methods rather than supplementary materials.

      ANS: Thank you for your suggestions to improve the description of the numerical model. As summarized below, we have rewritten this section of 'Simulating the dynamic MinD concentration gradient in growing cells' in the manuscript (Lines 237-279). We have specified the dimensionality of the rate and diffusion constants of each molecule, where applicable, in our 1D model from Lines 237-264. Their dimensionality can also be conceived from their units, as listed in Tables 2 and S4. We have specified the initial 'no-flux' boundary conditions in Lines 267, 630, and 647. We agree that the delta function is not necessary and have removed it from the equations.

      (iii) As in the previous comment, the current model did not take into account the geometry of the system; namely, cytosol is in 3D and membrane is on 2D. Recent theoretical studies can handle the effect, and also the effect of confinement. I would appreciate it if the authors would make a comment on whether those issues are relevant or not for the conclusion of this work.

      ANS: Thank you for pointing out this interesting aspect of cell geometry as investigated in Wu et al., 2015 (Wu et al., 2015). Our model is built to adequately describe changes in the MinD concentration gradient during cell elongation under the assumption that a 1D description is sufficient. Thus, our model cannot be generalized to other shapes, such as those observed in Wu et al., 2015 (Wu et al., 2015). This point is now commented upon in Supplemental Information, lines 120-123.

      (iv) I would appreciate it if the authors would describe the screening process more clearly. I did understand the first screening is a finite imaginary part and a positive real part at the first mode of spatial inhomogeneity in the eigenvalues. However, I did not understand the other processes clearly. The second screening is based on \lambda_N and I_Ratio, but its criteria is not clear. I think both quantities fluctuated in experimental results and I am not sure what to define numerical results match them. The third process is based on a fitting error using the fitting function of linear increase plus a constant. I am not sure why we need to exclude, for example, the bottom right example in Fig.S6 because it shows no oscillation until the cell length of 3um but then the gradient linearly increases. Please clarify how to justify the criteria. The same argument applies to the fourth screening process. It is not clear why the slope should be smaller than 2.

      ANS: Thank you for your suggestions to improve the description of the screening process. We have re-run the simulation to refine and improve the screening process, and the corresponding text and illustration are provided in the Results section of the main text (Lines 237-279, 614-653) and Fig. S6.

      (v) The authors claimed that the steeper gradient of MinD under glucose starvation results in cell division for shorter cells. I do not think the claim is convincing. It is necessary to measure the correlation between the length at the cell division and the gradient. It would also be nicer to show the correlation under other parameters. I think those studies truly support the authors' claim and the novelty of this work.

      ANS: Thank you for the comments. We would like to draw your attention to the right side of the graph shown in Fig. 3B, E, where measurements were obtained from cells prior to division. Our claim that "the steeper gradient of MinD under glucose starvation results in cell division for shorter cells" is also supported by the wave slope (λ_N range): 0.4% glucose of 1.49-2.66 (cell length range: 1.7-4.5 µm) and glucose starvation of 1.34-3.54 (cell length range: 2.1-3.8 µm). Therefore, under glucose starvation, λ_N increases more significantly with increasing length, allowing us to speculate on the contribution of steeper concentration gradient in stressed shorter cell to division. In the revised manuscript, the statement is kept in the Results section (Lines 217-218), but removed from the abstract. About the correlation between the concentration gradient and cell length at division under different conditions, we consider it more important to characterize all aspects under the same growth condition and avoid manipulation. In this study, the main conclusions are drawn from our experiments characterizing several aspects of MinD oscillations in cells growing with 0.4% glucose. In support of these observations, we decided to maintain only one other condition, 0.1% glucose. Further analysis of cells growing under other conditions will not change the main conclusions but will increase the difficulty of determining how the MinD concentration changes with cell growth.

      (vi) The conclusion at Line 346 "This plasticity arises from spatial differences in molecular interactions between MinD and MinE, as demonstrated..." looks unclear to me. My understanding is that (i) by screening the randomly sampled parameters in the reaction-diffusion model, the authors found the parameters that "match" experimental results, and (ii) the parameters after screening show the correlation between them (k_dD-k_dE and k_D-k_ATP->ADP). The logic heavily relies on the reaction-diffusion model is quantitatively correct. First, I think it is better to explain the logic more explicitly, that is, the claim of the molecular interaction is not based on the experimental facts. Second, I personally think the reaction-diffusion model used in this work does not reproduce quantitatively the experimental results, as discussed in (iii) and also (iv). Please make some discussions on how to justify the comparison between the model and experiments.

      ANS: Thank you for your constructive comments. To address these questions, we have re-run the simulation to refine and improve the results, and the corresponding text and illustration are provided in the Results section of the main text (Lines 237-279, 614-653) and Fig. S6. The kinetic parameters used in this study are described in the main text, lines 258-264: 'To randomly search for combinations of the parameter sets k_dD, k_dE, k_D, and k_(ADP→ATP), the following parameters were fixed in the simulation: the diffusion coefficients D_d and D_de were assumed values based on bacterial membrane proteins (Schavemaker et al., 2018), the diffusion coefficients D_D and D_E were from Meacci et al. (2006) (Meacci et al., 2006), and the dissociation rate constant k_de were from a previous simulation (Wu et al., 2015). This operation allowed us to probe for the general behaviours of the system.' Lines 277-279: 'This screening process reduced the parameter sets to 23, including set #2827, which, judging by the correlation plots for length vs. period, λ_N, and I_Ratio (Figs. S7-S9), showed features similar to those of the experimental data (Figs. 1E, 3B, C).' Based on the parameters of set #2827, we rigorously tested the impact of different kinetic constants that represent different molecular interactions on the oscillation period, λ_N and I_Ratio (Fig 4D-H). The results are described in the section of 'Effect of the kinetic rate constant on the MinD concentration gradient' of the main text, lines 323-349. This effort has provided us with a solid theoretical view of how oscillation features may be controlled by different molecular interactions. In addition, a comparison between our modelling and experimental results is described in the main text, section 'In silico oscillation resembles oscillation in a cellular context', lines 300-321.

      (vii) I did not capture the point why the authors can claim "... further distinguishing in vivo and in vitro observations. " at Line 350. I did not find the results comparing with vitro studies. I would appreciate a demonstration of vitro results and/or references.

      ANS: To avoid confusion, this sentence has been removed.

      Minor comments: (1) Line 214: It should be "Fange and Elf".

      ANS: Line 238 in the revised manuscript: This has been corrected.

      (2) I think it is better to show sampled points in Fig. 4C and 4D to show how dense the authors sampled in the parameter space.

      ANS: Since we have rewritten this part, the suggested revision is no longer applicable.

      REFERENCES: [1] Fischer-Friedrich, Elisabeth / Meacci, Giovanni / Lutkenhaus, Joe / Chaté, Hugues / Kruse, Karsten, "Intra- and intercellular fluctuations in Min-protein dynamics decrease with cell length", Proceedings of the National Academy of Sciences, 107, 6134-6139 (2010). [2] Meacci, Giovanni, "Physical Aspects of Min Oscillations in Escherichia Coli", PhD thesis (2006) available at

      Reviewer #1 (Significance (Required)):

      General assessment: I think the strength of this study is that it potentially shows the quantitative correlation between the MinD concentration gradient during the oscillation and the cell length when it divides. However, the current data of glucose starvation is not convincing enough. The model parts are interesting but their connection to the experiments is not clear in the current manuscript.

      ANS: Thank you for your comment. The key finding of our study, involving experimental measurements and mathematical modelling, is plasticity in the MinD concentration gradient, which results from spatial differences in molecular interactions and is an intrinsic property of the Min system during cell growth. We hypothesized that if the plasticity of the MinD concentration gradient is an intrinsic property of the system, then this property would be robust and show consistent behaviour under different growth conditions. Therefore, we tested this hypothesis by studying MinD oscillations under a low-glucose condition, and the results strengthened the main conclusion derived from experiments under the regular growth condition containing 0.4 % glucose. We believe that further analysis of cells growing under other conditions will not change the main conclusions but may increase the difficulty of determining how the MinD concentration changes with cell growth. Therefore, we decide to make this section concise, containing only one additional condition, even though we have more data than presented here. As mentioned earlier in this response letter, we have re-run the simulation to refine and improve the results, and the corresponding text and illustration are provided in the Results section of the main text (Lines 237-279, 614-653) and Fig. S6. This operation allowed us to probe for the general behaviours of the system. As a result, we were able to obtain a few parameter sets, including #2728, that generate features of the oscillation period, λ_N and I_Ratio, that strongly mimic MinD oscillation in the cellular context (Figs. 4C, S7-9). We further tested the impact of different kinetic constants, k_de, k_dD, k_dE, k_D, and k_(ADP→ATP), which represent different molecular interactions influencing the oscillation period, λ_N and I_Ratio (Figs. 4D-H). This effort has provided us with a solid theoretical view of how oscillation features may be controlled by different molecular interactions.

      Advance: The advance of this study is to measure the MinD concentration gradient under glucose starvation, and to compare the experimental results with the (simplified) model under a wide range of parameters. I do not think the advance in the current manuscript looks conceptual level because the conceptual conclusions are not really convincing from the results. In this respect, the advance of this work may be technical.

      ANS: Thank you for this constructive comment and have responded as follows. In combination with both experimental and theoretical efforts in the revised manuscript, this work provides conceptual advancement in a quantitative understanding of MinD oscillations in the cellular environment and provides implications for bacterial cell division regulation for further studies in the field. Specifically, we would like to emphasize that this work revealed the inherent plasticity and adaptability of the MinD concentration gradient that contributes to division site selection. The mathematical modelling provided us with a solid theoretical view of how oscillation features may be controlled by different molecular interactions.

      Audience: As a theoretician working on biophysics, including the model of the Min system, I think a specialised audience would be interested in this study. People who are studying the mechanism of the Min oscillation and resulting cell division, particularly those who are interested in both experiments and models, would be interested in this work. For the broad audience, I do not think the novelty of this study is well described.

      ANS: Thank you for your comment. We would like to point out that studying the variable concentration gradient underlying the dynamic oscillations of the Min system may be of broad interest to cell biologists since the concentration gradient plays a fundamental role in various cellular processes, and the concept of concentration gradients is crucial in cell biology. Examples include passive and active transport, osmosis, cell signalling, and maintenance of cellular homeostasis. These processes allow cells to respond to their environment, regulate their internal conditions, and perform important functions required for survival and normal function. In addition, the variable concentration gradient, characterized by the numerical descriptor λ_N and reproduced in a simple mathematical model, demonstrates a nonlinear dynamics behaviour in physical biology. Therefore, the audience of this work may include the broader general audience of cell biology and physical biology rather than just the immediate specialized audience interested in the Min system. We will also reiterate the importance of specialized research, which often provides the basis for broader application and understanding.

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

      Summary: This work by Parada et al showed that in the oscillatory Min System, MinD gradient was steeper in longer e.coli cells, while period was stable. This behavior was recapitulated in a mathematical model and it also revealed coordinated reaction rates in a wide range of parameter space.

      ANS: We thank the reviewer for the concise summary of our work.

      Major comments: 1. There were some inconsistencies between experimental and modeling data. Wave slope (𝜆𝑁) plateaued at ~3um in the model but not shown in the experiment (Fig.3B). The period was much less in the model (Fig. S8) than in the experiment (Fig. 1B).

      ANS: Thank you for pointing out this problem. We have re-run the simulation to refine and improve the results, and the corresponding text and illustration are provided in the Results section of the main text (Lines 237-279, 614-653) and Fig. S6. This operation allowed us to probe for the general behaviours of the system. As a result, we were able to obtain a few parameter sets, including #2728, that generate features of the oscillation period, λ_N and I_Ratio, that highly mimic MinD oscillation in the cellular context (Figs. 4C, S7-9). Regarding oscillation period, the simulation result was shorter than the experimental measurements. Even though, based on the parameters of set #2827, we rigorously tested the impact of different kinetic constants that represent different molecular interactions on the oscillation period, λ_N and I_Ratio (Main text, lines 323-349; Fig 4D-H). This effort has provided us with a theoretical view of how oscillation features may be controlled by different molecular interactions. We found that the rate constants k_de, representing detachment of the MinDE complex from the membrane, and k_(ADP→ATP), representing recharging of MinD-ADP with ATP, more significantly affected the oscillation period. The results suggested that the oscillation cycle time is tunable. In response to the question of the wave slope (λ_N) plateaued at ~3um in the modelling (Fig. 3B) but not shown in the experiment (Fig. 1D), we think this is due to experimental examination of a heterogenous population of cells versus simulating a growing bacterial cell. We came up with conclusions and hypotheses through wet experiments, which were further strengthened using mathematical modelling, providing insights into kinetic properties of the Min system.

      1. Generally, I found that the data of starved condition added little to the major message. Unless the model can recapitulate the even steeper gradient in such condition by tuning starvation-related parameters, it may be removed.

      ANS: We thank the reviewer for this suggestion. The key finding of our study, involving experimental measurements and mathematical modelling, is plasticity in the MinD concentration gradient, which results from spatial differences in molecular interactions and is an intrinsic property of the Min system during cell growth. We hypothesized that if the plasticity of the MinD concentration gradient is an intrinsic property of the system, then this property would be robust and show consistent behaviour under different growth conditions. Therefore, we tested this hypothesis by studying MinD oscillations under a low-glucose condition, and the results strengthened the main conclusion derived from experiments under the regular growth condition containing 0.4 % glucose. We agree that further analysis of cells growing under other conditions will not change the main conclusions but may increase the difficulty of determining how the MinD concentration changes with cell growth. Therefore, we decide to make this section concise, containing only one additional condition, even though we have more data than presented here.

      1. The authors need to compare what was different/novel between the model in this study and previous models such as Wu, et al 2015 and highlight the uniqueness of this work.

      ANS: Thank you for this suggestion. We now provide a comprehensive comparison between them in the Supplemental Information (Lines 26-147). We would like to emphasize that although the goal of the previous works was to measure the spatiotemporal distribution of oscillating MinD concentration gradients as a function of cell length, these works conceived the problem differently and therefore used different experimental designs and execution methods, which differentiates our key conclusions from theirs. This is also true for mathematical modelling. Although similar observations can be found in some respects, they are not directly comparable due to the different mathematics and assumptions used in the simulations. Therefore, we would like to draw attention to the experimental rigor and to the specific points and views that contribute to our understanding of Min systems.

      1. The model explored parameter space of reaction rates and found 60 sets. The KdE, KD, KdD, KADP-ATP ranged 6 orders of magnitude. It is interesting data in itself, but cells were not likely to vary that much for reaction rates. The relevance should be discussed.

      ANS: Thank you for pointing out this problem. For this revision, we re-ran the simulation to refine and improve the results, allowing us to identify parameter sets that generate features resembling the experimental measurements. Using set #2728 as an example, the variations in the five rate constants k_de, k_dD, k_dE, k_D, and k_(ADP→ATP) fall within a small range (Table 2, S4), eliminating the concern that arose from the previous version of the manuscript. We found that this parameter set allows for maximum utilization of MinD and MinE molecules, which are fixed in number according to experimental measurements, to drive membrane-associated oscillations in the simulation.

      Minor comments: 1. Fig.1B colors were conflicting. The legend was different than diagram. Fig.1C no scale for x axis.

      ANS: We have resolved the colour conflict in Fig. 1B, and a time range has been added to Fig. 1C.

      1. Fig.S6A How the 638 oscillatory parameter sets were matched with experimental data and screened to 174 sets was not clear. Data of fitting errorANS: Thank you for your suggestions to improve the description of the screening process. In this revision, we have re-run the simulation to refine and improve the results, and the corresponding text and illustration are provided in the Results section of the main text (Lines 237-279, 614-653) and Fig. S6. This operation allowed us to probe for the general behaviours of the system. The mentioned filter no longer applies.

      2. Significant digits were not used properly. For example, the period (table 1) was showed as 46.00 sec, but the imaging interval was 12 sec, the 2 decimal digits were thus meaningless. The same argument goes for length measurement at 2.84 um, while the optical resolution of the microscope used should be no good than 200nm.

      ANS: We have corrected this significant digit throughout the manuscript.

      1. For scatter plot like Fig.1D-G, generally smaller dots would show trend more obvious.

      ANS: We have modified the plots and used smaller dots in Figs. 1D-G, 3B, C, E, F, S3D, and S5B, C.

      1. The molecular mechanism of why MinD gradient increases with length was not the scope of the current study, but better to be discussed.

      ANS: Let me address this comment in another way. The key finding of our study, involving experimental measurements and mathematical modelling, is plasticity in the MinD concentration gradient, which results from spatial differences in molecular interactions and is an intrinsic property of the Min system during cell growth. In the revised manuscript, we have re-run the simulation to refine and improve the modelling procedures and results, and the corresponding text and illustration are provided in the Results section of the main text (Lines 265-279, 614-653) and Fig. S6. In brief, we fixed the diffusion coefficients D_D and D_Efrom Meacci et al. (2006) (Meacci et al., 2006); the dissociation rate constant k_de from a previous simulation (Wu et al., 2015); and the experimentally measured MinD and MinE concentrations in this study. Meanwhile, the diffusion coefficients D_d and D_de were assumed values based on bacterial membrane protein diffusion (Schavemaker et al., 2018). This operation allowed us to probe for the general behaviours of the system. As a result, we were able to obtain a few parameter sets, including #2728, that generate features of the oscillation period, λ_N and I_Ratio, that highly mimic MinD oscillation in the cellular context (Figs. 4C, S7-9). We further tested the impact of different kinetic constants, k_de, k_dD, k_dE, k_D, and k_(ADP→ATP), which represent different molecular interactions influencing the oscillation period, λ_N and I_Ratio (Fig 4D-H). Our findings have provided us with a solid theoretical view of how oscillation features may be controlled by different molecular interactions. Furthermore, the modelling results help us understand the possible mechanisms associated with oscillation cycle maintenance and length-dependent variable concentration gradients.

      1. Fig. S8, why sudden jump in period in many of the sets of both groups?

      ANS: This supplemental figure is now Fig. S7. A slower oscillation at the initiation of oscillation appears to be a common property in our simulation.

      Reviewer #2 (Significance (Required)):

      Min system was well-studied oscillation mechanism to restrict FtsZ at cell center. Previous work has shown how the system work molecularly, simulated the behavior and reconstituted many different patterns in vitro. The major new information from this work was: 1. the rigorously measured endogenous level of MinD and MinE; 2. gradient increased with length; 3. a model recapitulated this relationship and explored parameter space of reaction rates. The paper was well presented, experiments and analysis were rigorous, and the conclusions were not overstated. It should interest specialized cell biologists studying cell size, oscillation pattern.

      ANS: Many thanks to Reviewer 2 for recognizing the contributions of our work to the understanding of the Min system and its role in cell division. We also thank you for identifying professional cell biologists studying cell size and oscillation patterns as readers of our paper. We would like to emphasize that cellular concentration gradients play a fundamental role in various cellular processes and that the concept of concentration gradients is crucial in cell biology. These concentration gradient-mediated processes allow cells to respond to their environment, regulate their internal conditions and perform important functions required for survival. In addition, the variable concentration gradient, characterized by the numerical descriptor λ_N and reproduced in a simple mathematical model, demonstrates a nonlinear dynamics behaviour in physical biology. Therefore, the audience of this work may include a broader audience in the field of cell biology and physical biology rather than just an immediate specialist audience. We will also reiterate the importance of specialized research, which often provides the basis for broader application and understanding.

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

      The manuscript shows that the concentration of MinD does not change during the division cycle of E. coli. Due to the oscillation pattern the concentration of MinD decreases at the mid-cell which makes it favorable for the division. The mid-cell decrease in concentration of MinD is majorly length dependent. The oscillation pattern is not due to the change in concentration of MinD, but due to the plasticity arises from the spatial differences in molecular interactions between MinD and MinE. The manuscript is well written, the experiments are performed carefully and the results will be of interest to readers from variety of field. However, there are several concerns need explanation.

      ANS: We greatly appreciate the positive feedback from the reviewer, and we address the specific concerns below.

      Major concerns: One of my major concern is these interactions are not shown experimentally but explained using either the previously published literature or mathematical models. Further, the previous literatures are shown on in vitro models which does not mimic the in vivo system fully.

      ANS: We thank the reviewer for the important point that reaction rates in previous studies and in our model of Min oscillations have not been experimentally tested. We are aware of the lack of experimental measurements, but these reaction rates cannot be measured in batch reactions using classical biochemical methods. To accurately measure these reaction rates, the experiments require advanced techniques and methods to handle spatial and temporal resolution, which is beyond the scope of our current study. However, in the revised manuscript, we have re-run the simulation to refine and improve the results, and the corresponding text and illustration are provided in the Results section of the main text (Lines 237-279, 614-653) and Fig. S6. In our simulation, we fixed the diffusion coefficients D_D and D_E from Meacci et al. (2006) (Meacci et al., 2006); the dissociation rate constant k_de from a previous simulation (Wu et al., 2015); and the experimentally measured MinD and MinE concentrations in this study. Meanwhile, the diffusion coefficients D_d and D_de were assumed values based on bacterial membrane protein diffusion (Schavemaker et al., 2018). This operation allowed us to probe for the general behaviours of the system. As a result, we were able to obtain a few parameter sets, including #2728, that generate features of the oscillation period, λ_N and I_Ratio, that highly mimic MinD oscillation in the cellular context (Figs. 4C, S7-9). Interestingly, we found that this parameter set allows for maximum utilization of MinD and MinE molecules, which are fixed numbers from experimental measurements, to drive membrane-associated oscillations in the simulation. We further tested the impact of different kinetic constants, k_de, k_dD, k_dE, k_D, and k_(ADP→ATP), which represent different molecular interactions influencing the oscillation period, λ_N and I_Ratio (Figs. 4D-H). Our findings have provided us with a solid theoretical view of how oscillation features may be controlled by different molecular interactions, and help us understand the possible mechanisms associated with oscillation cycle maintenance and length-dependent variable concentration gradients.

      The concentration of MinD does not change with the increasing length of the cell. Is the MinD concentration (or copy numbers) is different in the case of cells growing in low glucose and when compared to the cells growing at high glucose?

      ANS: Thank you for the comments. As shown in Figs. 2B, C, the concentration of MinD changed with cell length, but the number of MinD molecules per unit area did not change significantly with cell length. Although how the number of MinD molecules changes when cells are grown under low-glucose conditions is unclear, this number does not appear to be essential for the following reasons. We focused on studying Min oscillations during the normal growth cycle, minimizing experimental manipulations to analyse oscillation dynamics. Measurements of oscillations in cells grown under low-glucose conditions support the primary measurements. We think that further analysis of MinD concentration changes in growing cells under low-glucose conditions will not change the main conclusion of this manuscript: 'plasticity in the MinD concentration gradient is an intrinsic property of the Min system during cell growth',

      As per the current study a particular I-ratio at the mid-cell is required to initiate the cell division. In the case of cells growing at low glucose, how this required I-ratio is achieved at the mid-cell?

      ANS: Thank you for the excellent question. As described in the main text, lines 199-201, I_Ratio is defined as the ratio of the minimum intensity to the maximum intensity measured from the experimental data, which gradually decreases as the cell length increases (Fig. 3C). Since the minimum and maximum intensities were measured from the concentration gradient, which is characterized by the slope of the concentration gradient (λ_N), there exists a correlation between I_Ratio and λ_N. That is, a larger λ_N will result in a smaller I_Ratio, and vice versa. When comparing measurements made from cells grown with 0.4% and 0.1% glucose (Fig. 3B, C, E, F), the changes in λ_N are more drastic within a shorter length under low-glucose condition, which is accompanied by more drastic changes in I_Ratio. Furthermore, when the I_Ratio value was approximately 0.5, the corresponding cell length was significantly shorter under low-glucose condition. Therefore, we speculate that there may be an effective I_Ratio that is low enough for stable FtsZ ring formation. This effective I_Ratio can occur at any cell length, allowing us to see that bacteria divide at shorter cell lengths under low-glucose conditions. This property necessitates a faster reduction in the concentration gradient to reach the effective I_Ratio for cells dividing at shorter lengths. As a result, by adjusting λ_N as a function of length, the steepness of the I_Ratio reduction can be altered. Please see the main text, lines 389-406.

      There is decrease in the MinD oscillation time observed in low glucose condition. As explained by the authors the MinD oscillation is mainly guided by the FtsE induced removal of MinD from the membrane, how the authors can explain this decrease?

      ANS: Thank you for raising the question of how the MinE-induced detachment of membrane-bound MinD contributes to the oscillation time of MinD under low-glucose conditions. Although this is an interesting question, determining what regulates MinE-induced detachment of membrane-bound MinD under low-glucose conditions is beyond the scope of the current study. This unknown regulatory mechanism that regulates MinD-MinE interactions in growing cells under low glucose conditions is worthy of further investigation. However, our modelling results have provided a theoretical view of how oscillation features may be controlled by different molecular interactions between MinD and MinE and may guide future experiments investigating the underlying mechanism involved. Please refer to the Results section: 'Spatiotemporal distribution of the concentration gradient' in the main text, lines 351-373.

      Further, it is explained that the concentration of cellular ATP is in much higher concentration compared to the required amount for this oscillation. As the Iratio is majorly dependent on the cell length, what could be the reason for the differential N in the case of low and high glucose condition?

      ANS: Please refer to the previous answer to the question: 'As per the current study a particular I-ratio at the mid-cell is required to initiate the cell division. In the case of cells growing at low glucose, how this required I-ratio is achieved at the mid-cell?'. (this letter, Lines 764-779) In addition, our modelling in search of parameter sets that generate characteristics of MinD oscillation resembling oscillation in vivo allowed us to evaluate the impact of different molecular interactions, as represented by different rate constants (Fig. 4), which has provided important information for future mechanistic investigations, although not in the present study. Please see the Results section: 'Effect of the kinetic rate constant on the MinD concentration gradient' in the main text, lines 323-349.

      MinD is a highly insoluble protein. It also has an amphipathic helix and thus most of the time it binds to the membrane. The method used by the author to determine the cellular MinD concentration (mentioned in Fig S1) will only give the concentration of the soluble MinD and not of the total MinD. How the authors justify this as the total concentration. This is also the same in the case of MinE copy number calculation. Authors may need to perform the transcriptome analysis and compare both the data.

      ANS: We thank the reviewer for the comments. Since the attachment of MinD and MinE to the membrane is transient and MinD-membrane interactions require ATP, we expected that most of the protein would be released from the membrane into the cytoplasm after cell disruption, sufficiently representing the total MinD concentration. Furthermore, our measurements of molecule numbers are within the range of previous measurements (Di Ventura & Sourjik, 2011; Juarez & Margolin, 2010; Meacci & Kruse, 2005; Tostevin & Howard, 2006; Touhami et al, 2006). Thus, we believe that our current measurements are reliable and sufficient for subsequent interpretation.

      One of the main question asked by the authors in the abstract is. "How the intracellular Min protein concentration gradients are coordinated with cell growth to achieve spatiotemporal accuracy of cell division is unknown". Although the authors have shown that there is a change in concentration gradient during cell growth, the mechanism for the same is not very well explained. Authors have not provided any specific explanation for the increase in the velocity of the MinD oscillation and the gradient formation. How the velocity of MinD is increasing although there is no increase in the MinD concentration.

      ANS: We have changed 'the mechanism' to 'the exact way' in the abstract (Abstract, line 28). Moreover, in the revised manuscript, we have improved the mathematical model and performed a thorough investigation of the variations in the kinetic constants. This effort has provided us with a solid theoretical view of how oscillation features may be controlled by different molecular interactions. The results may guide future experiments investigating the underlying mechanism involved. Please refer the answers to previous questions above.

      Figure 2B: shows the overall concentration of MinD in a single cell varies between 1180 - 1160 molecules/um2. In Fig 2C it is mentioned that mid-cell has a MinD concentration of 120-20 molecuels/ um2. Further, Fig3C and 3F shows I-ratio values varies between 0.6-0.4. Considering the values given the I-ratio (I min/ I max) should be between 0.1- 0.01. Authors need to explain the same. Figure 2C: The data in both the Y-axes are not matching and needs more clarification in the legend. Whether the number of molecules were counted only in the marked 200 nm area? If so, why the Y-axis 1 (molecules/um2) is decreasing 7 times, whereas, Y-axis 2 (molecules) is only by 2 times.

      ANS: In this work, we measured sfGFP-MinD intensity through fluorescence microscopy. The fluorescence intensity was converted into molecular numbers based on estimates from Western blot analyses (Fig. S1). This number of molecules for MinD and MinE was assumed to be the mean number, which was fit into the midpoint of the doubling time (Fig. 2B, black dashed line; main text, lines 166-167). Fig. 2C was obtained by further processing the same dataset to restrict the region of analysis to the midcell zone. Please refer to the main text, lines 158-178. However, the λ_N and I_Ratio values were calculated from the processed intensity data (Fig. S2; main text, lines 190-209, 533-559). Because of the conversion from intensity to molecule number in Figs. S2B, C and the image processing procedure applied to the calculation of λ_N and I_Ratio, it is not possible to directly compare the fold change and the upper and lower limits between molecule numbers and the λ_N and I_Ratio values.

      Other comments: Line 84: Requires reference for this statement.

      ANS: A recent review article has been added in the main text, line 84: '(Cameron & Margolin, 2024)'.

      Line 96: Can authors provide other evidence or validation for the determination of the copy numbers such as transcriptome analysis.

      ANS: We thank the reviewer for this suggestion. However, we believe that direct measurement of cellular protein abundance is reliable and sufficient for our purposes. Furthermore, transcriptome-measured RNA abundance does not translate directly to protein abundance in living cells because posttranscriptional processing, translation, posttranslational processing, and protein stability issues complicate the interpretation. Therefore, protein abundance measurement from cell extracts is straightforward for our purpose.

      Fig 1C: what is the units of time in Fig 1C? Is it equal for all the cell lengths?

      ANS: As described in the main text, lines 511-512, 'Time-lapse images of sfGFP-MinD were acquired at 12-sec intervals for 10 min or before the fluorescence diminished'. This condition is applied to all the acquired images in this work.

      Page 6, line 136-138: what could be the possible mechanism for change in velocity at different cell cycle time?

      ANS: To avoid confusion, we have modified the text and tone down the velocity when mentioned. This is because the mentioned velocity is inferred from the measured oscillation period and cell length but not from direct measurements; our emphasis is on understanding how the oscillation period remains fairly stable during cell growth rather than how the velocity changes. In the revised manuscript, we used modelling results to elucidate the possible mechanism related to period maintenance. The corresponding text and illustration are provided in the Results section (Lines 300-373) and the Discussion section of the main text (Lines 407-446) and Figs. 4, 5. In brief, this simulation allowed us to probe for general behaviours of the system, allowing us to obtain a few parameter sets that generate features of the oscillation period, λ_N and I_Ratio highly mimicking MinD oscillation in the cellular context (Fig 4C, S7-9). We further tested the impact of different kinetic constants, k_de, k_dD, k_dE, k_D, and k_(ADP→ATP), which represent different molecular interactions influencing the oscillation period, λ_N and I_Ratio (Fig 4D-H). This effort has provided us with a solid theoretical view of how oscillation features may be controlled by different molecular interactions. Please see the Results section: 'Effect of the kinetic rate constant on the MinD concentration gradient' in the main text, lines 323-349.

      Page 7, line 155: Any evidence for claiming the same?

      ANS: The sentence has been modified as follows: 'Thus, the fairly stable oscillation period and variable velocity did not change the precision of the septum placement.' (Main text, lines 155-156)

      Page 7, line 156: Is there any proof authors can show that burst MinD synthesis occurs during the division? If not in the case of MinD, is it shown in any other protein?

      ANS: The text is now in line 168-171: 'Interestingly, the value after division was not doubled, which could indicate a balanced outcome between de novo synthesis and degradation or a burst of MinD synthesis at cell division followed by constant synthesis.' In previous studies by Männik et al. (2018) (Mannik et al, 2018) and Vischer et al. (2015) (Vischer et al, 2015), the division protein FtsZ increased the cellular concentration throughout the cell cycle under slow growth conditions and degraded rapidly at the end of the cell cycle, a process controlled by the ClpXP protease. Because we do not know the relevance of these observations to our study, which focused on the plasticity of the MinD concentration gradient, we decided not to discuss them in the manuscript.

      Page 9, line 217: The Fig 4A is not explained clearly and all the terms mentioned needs to be explained. This figure is used to explain the differential concentration of MinD at the poles and the mid-cell, thus needs to be explain more clearly.

      ANS: Thank you for your comments. Please refer to the above answer to the question: 'One of my major concern is these interactions are not shown experimentally but explained using either the previously published literature or mathematical models. Further, the previous literatures are shown on in vitro models which does not mimic the in vivo system fully.', in this letter, lines 691-715.

      Page 12, line 285: What is meaning of default speed of MinD oscillation in new-born cells? Do the authors observed any specific velocity in the new-born cells? What is the explanation for length dependent oscillation velocity for MinD?

      ANS: Thank you for the questions. As mentioned earlier, the emphasis of this study is on understanding how the oscillation period remains relatively stable while showing plasticity of the concentration gradient during cell growth. The velocity is inferred from the oscillation period and cell length but is not a direct measurement. To avoid confusion, we have modified the text and placed less emphasis on the velocity when mentioned.

      Reviewer #3 (Significance (Required)):

      General assessment: Major work of the manuscript is relying on the mathematical models, whereas the audience are majorly from the biology fields and thus simplified explanations are required in many places. Many of the legends in the figures require more explanation for better understanding. If possible more experimental data can be added, specifically to explain the model mentioned in figure 4A.

      ANS: We have modified the figure legends to include more explanations. As mentioned above, we have also revised Fig. 4 to include improvements in modelling results to better fit the experimental data and to examine the impacts of the kinetics constants of the reaction steps in the Min system. Please refer to lines 691-715 in this letter.

      Advance: The study is adding to the existing knowledge and will be helpful to fill the conceptual gaps in understanding the mid-cell MinD concentration and what may favor the initiation of bacterial division. Audience: Majorly the microbiology community will be interested in the study. This will also be interest to Physicists and mathematical persons working to understand bacterial division.

      ANS: We thank the reviewer for this positive comment.

      Reviewer #4 (Evidence, reproducibility and clarity (Required)):

      The study by Parada et al. illuminates the intricate interplay between Min proteins, exemplified by MinD, and cell growth in E. coli. Their findings demonstrate that the MinD concentration gradient steepens progressively as cells elongate, potentially influencing FtsZ ring formation via MinC. Moreover, their comprehensive reaction-diffusion model not only corroborates experimental observations of length-dependent concentration gradients but also underscores the critical role of kinetic interactions involving Min proteins, the membrane, and ATP. This elucidation significantly advances our understanding of the oscillatory mechanisms within the Min system. Both the experimental and simulation data are robust, and the manuscript is exceptionally well-written. I express my full support for publication pending the satisfactory resolution of the outlined concerns.

      ANS: We appreciate the reviewer's positive feedback and have addressed most issues to the best of our ability.

      1. Remove the dot in front of "Min" in line 57.

      ANS: This has now been removed.

      1. In lines 82-84, the statement "...The distribution of the division inhibitor MinC may be synchronized with spatiotemporal differences in MinD concentrations, leading to a stable placement of the FtsZ ring at the midcell..." suggests a potential synchronization between MinC and MinD oscillations. It is crucial to investigate if sfGFP-MinC exhibits similar concentration gradient oscillatory behavior in vivo as observed with MinD.

      ANS: Thank you for bringing up this question. The key finding of our study, involving experimental measurements and mathematical modelling, is plasticity in the MinD concentration gradient, which results from spatial differences in molecular interactions and is an intrinsic property of the Min system during cell growth. With many investigations already covered in this manuscript, we prefer to investigate sfGFP-MinC in future studies, which will have different focuses on how MinC dynamics are coupled with the variable MinD concentration gradient to directly impact FtsZ ring formation.

      1. Ensure consistent significant digits throughout the text. For instance, 1.95{plus minus}0.16 μM in line 97, 1.4{plus minus}0.13 μM in line 98, and 1.9 {plus minus} 0.2 μM in line 100 have varying precision. Consider using integers for molecules.

      ANS: We have corrected the significant digits in the main text and supplemental information.

      1. Address the discrepancy in expression levels of MinD and MinE between strain FW1541 and its parental strain W3110. Given the labeling effect, it is possible that MinD expression levels differ. However, MinC's expression level should be approximately the same. Conduct whole-genome sequencing of both strains to identify any additional mutations.

      ANS: Thank you for the comments. As described in the main text (Lines 67-70), the most important aspect is the concentration ratio between MinD and MinE. Although the numbers are not the same, they are comparable to those in previous studies (Hale et al, 2001; Li et al, 2014; Schmidt et al, 2016; Shih et al, 2002) (Main text, lines 113-115). Furthermore, we performed whole-genome sequencing of the W3110 and FW1541 strains. We confirmed that sfGFP was correctly inserted. The sequence alignment of the minCDE locus is provided for your reference but not for publication. Although there are some sporatic point mutations, there is no obvious reason to believe that the mutations would impact Min protein expression. We will organize the deposition data as soon as I can.

      1. Clarify the apparent discrepancy between lines 112 and 127. Line 112 suggests that the periodic regularity of interpolar oscillations increases with cell length, as demonstrated in Fig 1B-C, 1E, Fig S5. However, in the subsequent section (starting from line 127), the authors state that oscillation periods remain relatively stable across cells of different lengths. Provide clarification on this apparent discrepancy.

      ANS: Thank you for pointing out this confusion caused by misuse of the term. In Lines 122-123, the statement has been modified as follows: '...the uniformity of the oscillation intervals appears to increase with length...' In line 139, 'The oscillation period' refers to the time required for the oscillation cycle. Since the correction in line 123 should suffice to clarify, we did not modify the statement in line 139.

      1. Specify if the analysis was limited to non-constricted cells. If so, state this explicitly in the text, as it could impact the interpretation of results, especially in relation to the linear dependence of cell length on time before constriction, as shown in Fig S3C.

      ANS: We did not specifically remove those constricted cells, but cells before splitting were considered one cell. We have added a statement to clarify in Lines 144-145.

      1. Improve clarity in Fig 2A by using distinct colors (e.g., green and red) for differentiation on the Y-axis.

      ANS: The Y axes of Fig. 2A have been modified.

      1. Correct "of" to "from" in line 223 for improved clarity and accuracy.

      ANS: Corrected.

      1. Include the missing "A" in Fig S6A for completeness and accuracy.

      ANS: This figure has been updated.

      1. Ensure consistency in referencing style (full names versus short names) throughout the manuscript.

      ANS: This has now been done.

      Reviewer #4 (Significance (Required)):

      While numerous commendable in vitro studies have explored the oscillatory behavior of the Min system, this work uniquely delves into the oscillation of MinD within live cells. It unveils the remarkable coordination between intracellular Min protein concentration gradients and cell growth, shedding light on the precise spatiotemporal regulation of cell division.

      ANS: We thank the reviewer for this positive comment.

      References Di Ventura B, Sourjik V (2011) Self-organized partitioning of dynamically localized proteins in bacterial cell division. Molecular systems biology 7: 457 Fischer-Friedrich E, Meacci G, Lutkenhaus J, Chate H, Kruse K (2010) Intra- and intercellular fluctuations in Min-protein dynamics decrease with cell length. Proceedings of the National Academy of Sciences of the United States of America 107: 6134-6139 Hale CA, Meinhardt H, de Boer PA (2001) Dynamic localization cycle of the cell division regulator MinE in Escherichia coli. The EMBO journal 20: 1563-1572 Juarez JR, Margolin W (2010) Changes in the Min oscillation pattern before and after cell birth. Journal of bacteriology 192: 4134-4142 Li GW, Burkhardt D, Gross C, Weissman JS (2014) Quantifying absolute protein synthesis rates reveals principles underlying allocation of cellular resources. Cell 157: 624-635 Mannik J, Walker BE, Mannik J (2018) Cell cycle-dependent regulation of FtsZ in Escherichia coli in slow growth conditions. Molecular microbiology 110: 1030-1044 Meacci G, Kruse K (2005) Min-oscillations in Escherichia coli induced by interactions of membrane-bound proteins. Phys Biol 2: 89-97 Meacci G, Ries J, Fischer-Friedrich E, Kahya N, Schwille P, Kruse K (2006) Mobility of Min-proteins in Escherichia coli measured by fluorescence correlation spectroscopy. Phys Biol 3: 255-263 Schavemaker PE, Boersma AJ, Poolman B (2018) How Important Is Protein Diffusion in Prokaryotes? Front Mol Biosci 5: 93 Schmidt A, Kochanowski K, Vedelaar S, Ahrne E, Volkmer B, Callipo L, Knoops K, Bauer M, Aebersold R, Heinemann M (2016) The quantitative and condition-dependent Escherichia coli proteome. Nature biotechnology 34: 104-110 Shih YL, Fu X, King GF, Le T, Rothfield L (2002) Division site placement in E. coli: mutations that prevent formation of the MinE ring lead to loss of the normal midcell arrest of growth of polar MinD membrane domains. The EMBO journal 21: 3347-3357 Tostevin F, Howard M (2006) A stochastic model of Min oscillations in Escherichia coli and Min protein segregation during cell division. Phys Biol 3: 1-12 Touhami A, Jericho M, Rutenberg AD (2006) Temperature dependence of MinD oscillation in Escherichia coli: running hot and fast. Journal of bacteriology 188: 7661-7667 Vischer NO, Verheul J, Postma M, van den Berg van Saparoea B, Galli E, Natale P, Gerdes K, Luirink J, Vollmer W, Vicente M, den Blaauwen T (2015) Cell age dependent concentration of Escherichia coli divisome proteins analyzed with ImageJ and ObjectJ. Front Microbiol 6: 586 Wu F, van Schie BG, Keymer JE, Dekker C (2015) Symmetry and scale orient Min protein patterns in shaped bacterial sculptures. Nature nanotechnology 10: 719-726

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

      Learn more at Review Commons


      Referee #4

      Evidence, reproducibility and clarity

      The study by Parada et al. illuminates the intricate interplay between Min proteins, exemplified by MinD, and cell growth in E. coli. Their findings demonstrate that the MinD concentration gradient steepens progressively as cells elongate, potentially influencing FtsZ ring formation via MinC. Moreover, their comprehensive reaction-diffusion model not only corroborates experimental observations of length-dependent concentration gradients but also underscores the critical role of kinetic interactions involving Min proteins, the membrane, and ATP. This elucidation significantly advances our understanding of the oscillatory mechanisms within the Min system. Both the experimental and simulation data are robust, and the manuscript is exceptionally well-written. I express my full support for publication pending the satisfactory resolution of the outlined concerns.

      1. Remove the dot in front of "Min" in line 57.
      2. In lines 82-84, the statement "...The distribution of the division inhibitor MinC may be synchronized with spatiotemporal differences in MinD concentrations, leading to a stable placement of the FtsZ ring at the midcell..." suggests a potential synchronization between MinC and MinD oscillations. It is crucial to investigate if sfGFP-MinC exhibits similar concentration gradient oscillatory behavior in vivo as observed with MinD.
      3. Ensure consistent significant digits throughout the text. For instance, 1.95{plus minus}0.16 μM in line 97, 1.4{plus minus}0.13 μM in line 98, and 1.9 {plus minus} 0.2 μM in line 100 have varying precision. Consider using integers for molecules.
      4. Address the discrepancy in expression levels of MinD and MinE between strain FW1541 and its parental strain W3110. Given the labeling effect, it is possible that MinD expression levels differ. However, MinC's expression level should be approximately the same. Conduct whole-genome sequencing of both strains to identify any additional mutations.
      5. Clarify the apparent discrepancy between lines 112 and 127. Line 112 suggests that the periodic regularity of interpolar oscillations increases with cell length, as demonstrated in Fig 1B-C, 1E, Fig S5. However, in the subsequent section (starting from line 127), the authors state that oscillation periods remain relatively stable across cells of different lengths. Provide clarification on this apparent discrepancy.
      6. Specify if the analysis was limited to non-constricted cells. If so, state this explicitly in the text, as it could impact the interpretation of results, especially in relation to the linear dependence of cell length on time before constriction, as shown in Fig S3C.
      7. Improve clarity in Fig 2A by using distinct colors (e.g., green and red) for differentiation on the Y-axis.
      8. Correct "of" to "from" in line 223 for improved clarity and accuracy.
      9. Include the missing "A" in Fig S6A for completeness and accuracy.
      10. Ensure consistency in referencing style (full names versus short names) throughout the manuscript.

      Significance

      While numerous commendable in vitro studies have explored the oscillatory behavior of the Min system, this work uniquely delves into the oscillation of MinD within live cells. It unveils the remarkable coordination between intracellular Min protein concentration gradients and cell growth, shedding light on the precise spatiotemporal regulation of cell division.

    3. 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 manuscript shows that the concentration of MinD does not change during the division cycle of E. coli. Due to the oscillation pattern the concentration of MinD decreases at the mid-cell which makes it favorable for the division. The mid-cell decrease in concentration of MinD is majorly length dependent. The oscillation pattern is not due to the change in concentration of MinD, but due to the plasticity arises from the spatial differences in molecular interactions between MinD and MinE. The manuscript is well written, the experiments are performed carefully and the results will be of interest to readers from variety of field. However, there are several concerns need explanation.

      Major concerns:

      One of my major concern is these interactions are not shown experimentally but explained using either the previously published literature or mathematical models. Further, the previous literatures are shown on in vitro models which does not mimic the in vivo system fully.

      The concentration of MinD does not change with the increasing length of the cell. Is the MinD concentration (or copy numbers) is different in the case of cells growing in low glucose and when compared to the cells growing at high glucose? As per the current study a particular I-ratio at the mid-cell is required to initiate the cell division. In the case of cells growing at low glucose, how this required I-ratio is achieved at the mid-cell? There is decrease in the MinD oscillation time observed in low glucose condition. As explained by the authors the MinD oscillation is mainly guided by the FtsE induced removal of MinD from the membrane, how the authors can explain this decrease? Further, it is explained that the concentration of cellular ATP is in much higher concentration compared to the required amount for this oscillation. As the Iratio is majorly dependent on the cell length, what could be the reason for the differential N in the case of low and high glucose condition? MinD is a highly insoluble protein. It also has an amphipathic helix and thus most of the time it binds to the membrane. The method used by the author to determine the cellular MinD concentration (mentioned in Fig S1) will only give the concentration of the soluble MinD and not of the total MinD. How the authors justify this as the total concentration. This is also the same in the case of MinE copy number calculation. Authors may need to perform the transcriptome analysis and compare both the data.

      One of the main question asked by the authors in the abstract is. "How the intracellular Min protein concentration gradients are coordinated with cell growth to achieve spatiotemporal accuracy of cell division is unknown". Although the authors have shown that there is a change in concentration gradient during cell growth, the mechanism for the same is not very well explained. Authors have not provided any specific explanation for the increase in the velocity of the MinD oscillation and the gradient formation. How the velocity of MinD is increasing although there is no increase in the MinD concentration. Figure 2B: shows the overall concentration of MinD in a single cell varies between 1180 - 1160 molecules/um2. In Fig 2C it is mentioned that mid-cell has a MinD concentration of 120-20 molecuels/ um2. Further, Fig3C and 3F shows I-ratio values varies between 0.6-0.4. Considering the values given the I-ratio (I min/ I max) should be between 0.1- 0.01. Authors need to explain the same. Figure 2C: The data in both the Y-axes are not matching and needs more clarification in the legend. Whether the number of molecules were counted only in the marked 200 nm area? If so, why the Y-axis 1 (molecules/um2) is decreasing 7 times, whereas, Y-axis 2 (molecules) is only by 2 times.

      Other comments:

      Line 84: Requires reference for this statement.

      Line 96: Can authors provide other evidence or validation for the determination of the copy numbers such as transcriptome analysis.

      Fig 1C: what is the units of time in Fig 1C? Is it equal for all the cell lengths?

      Page 6, line 136-138: what could be the possible mechanism for change in velocity at different cell cycle time?

      Page 7, line 155: Any evidence for claiming the same?

      Page 7, line 156: Is there any proof authors can show that burst MinD synthesis occurs during the division? If not in the case of MinD, is it shown in any other protein?

      Page 9, line 217: The Fig 4A is not explained clearly and all the terms mentioned needs to be explained. This figure is used to explain the differential concentration of MinD at the poles and the mid-cell, thus needs to be explain more clearly.

      Page 12, line 285: What is meaning of default speed of MinD oscillation in new-born cells? Do the authors observed any specific velocity in the new-born cells? What is the explanation for length dependent oscillation velocity for MinD?

      Significance

      General assessment: Major work of the manuscript is relying on the mathematical models, whereas the audience are majorly from the biology fields and thus simplified explanations are required in many places. Many of the legends in the figures require more explanation for better understanding. If possible more experimental data can be added, specifically to explain the model mentioned in figure 4A.

      Advance: The study is adding to the existing knowledge and will be helpful to fill the conceptual gaps in understanding the mid-cell MinD concentration and what may favor the initiation of bacterial division.

      Audience: Majorly the microbiology community will be interested in the study. This will also be interest to Physicists and mathematical persons working to understand bacterial division.

    4. 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 work by Parada et al showed that in the oscillatory Min System, MinD gradient was steeper in longer e.coli cells, while period was stable. This behavior was recapitulated in a mathematical model and it also revealed coordinated reaction rates in a wide range of parameter space.

      Major comments:

      1. There were some inconsistencies between experimental and modeling data. Wave slope (𝜆𝑁) plateaued at ~3um in the model but not shown in the experiment (Fig.3B). The period was much less in the model (Fig. S8) than in the experiment (Fig. 1B).
      2. Generally, I found that the data of starved condition added little to the major message. Unless the model can recapitulate the even steeper gradient in such condition by tuning starvation-related parameters, it may be removed.
      3. The authors need to compare what was different/novel between the model in this study and previous models such as Wu, et al 2015 and highlight the uniqueness of this work.
      4. The model explored parameter space of reaction rates and found 60 sets. The KdE, KD, KdD, KADP-ATP ranged 6 orders of magnitude. It is interesting data in itself, but cells were not likely to vary that much for reaction rates. The relevance should be discussed.

      Minor comments:

      1. Fig.1B colors were conflicting. The legend was different than diagram. Fig.1C no scale for x axis.
      2. Fig.S6A How the 638 oscillatory parameter sets were matched with experimental data and screened to 174 sets was not clear. Data of fitting error<0.12 and slope<2 were filtered. Authors should explain the criterion for data filtering.
      3. Significant digits were not used properly. For example, the period (table 1) was showed as 46.00 sec, but the imaging interval was 12 sec, the 2 decimal digits were thus meaningless. The same argument goes for length measurement at 2.84 um, while the optical resolution of the microscope used should be no good than 200nm.
      4. For scatter plot like Fig.1D-G, generally smaller dots would show trend more obvious.
      5. The molecular mechanism of why MinD gradient increases with length was not the scope of the current study, but better to be discussed.
      6. Fig.S8, why sudden jump in period in many of the sets of both groups?

      Significance

      Min system was well-studied oscillation mechanism to restrict FtsZ at cell center. Previous work has shown how the system work molecularly, simulated the behavior and reconstituted many different patterns in vitro. The major new information from this work was: 1. the rigorously measured endogenous level of MinD and MinE; 2. gradient increased with length; 3. a model recapitulated this relationship and explored parameter space of reaction rates.

      The paper was well presented, experiments and analysis were rigorous, and the conclusions were not overstated. It should interest specialized cell biologists studying cell size, oscillation pattern.

    5. 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:

      Parada et al. studied both experimentally and theoretically the MinD concentration distribution of Min waves during cell growth. The main finding was that (i) the gradient of MinD is steeper for longer cells and accordingly the MinD concentration at the middle of cell is lower, (ii) period of the oscillation is independent to the cell length, and (iii) those features are shared even under glucose starvation except the MinD gradient is steeper. (iv) Those results are supplemented by the analyses of the reaction-diffusion equations in which parameters that can reproduce the MinD concentration distribution are identified.

      I think the results are interesting; basically, as the cell grows, the contrast of the wave becomes clearer, such the MinD concentration at the cell centre decreases. The results may clarify the mechanism of FtsZ accumulation at the cell centre more quantitatively. The experiments were performed by measuring the fluorescent intensity of MinD during cell growth and analysing the intensity distribution along the long axis of the cell. The theoretical results were based on the analyses of the reaction-diffusion model. Both approaches are already well established and the results sound. Nevertheless, I do not think the novelty of this work is not well highlighted in the current manuscript; I think most of the results, except (iii) and (iv), have already been shown explicitly or implicitly in the previous studies. Min oscillations in a growing cell have been analysed both theoretically and experimentally in (Meacci 2005) and [1]. The concentration distribution and period of the oscillation were measured. The complete results were presented in [2], and I am not aware of those results in scientific journals (the thesis is available online). Nevertheless, I think it is fair to cite those studies and compare the current results with them. In fact, in [2], it was shown that the concentration of MinD near the cell centre decreases as the cell grows, the total MinD concentration is approximately constant during the growth (therefore, the number of the molecules increases), and that the variance of the period becomes smaller as the cell grows. I do not think those previous studies spoil this work, and this work deserves publication somewhere. Still, the authors should highlight the novelty of this study more clearly.

      Major comments:

      (i) In (Meacci 2005) and [1,2], it was claimed that the standard deviation of the period is comparable with the mean period, particularly for the shorter cell. Therefore, they did not claim the period is independent to the cell length. As far as I understood, the variance arises from the variance of the total protein concentration in the assemble of cells. I am wondering how the authors are able to conclude the constant period in different cell length. I also point out that in the theoretical part of (Meacci 2005), the period is, in fact, increasing as the cell grows and suddenly decreases at the length in which cell division occurs.

      (ii) I do not think the explanations of the reaction-diffusion model were well described. The authors mentioned that they studied a one-dimensional model and used the delta function to describe the membrane reaction. Did the authors study 1D cytosol and 0D membrane? Then, why the surface diffusion term exists in (4) and (5)? I believe the authors simply assumed that both the membrane and the cytosol are 1D (with larger diffusion constants for cytosolic Min concentrations). Then, the delta functions in (1)-(5) are not necessary. In (Wu 2015), the delta function was used in order to treat a 2D membrane embedded in 3D space.

      Besides that, there is no description of the initial conditions for the concentration fields to solve the reaction-diffusion equations. I think the description of the no-flux boundary condition is better put in the Methods rather than supplementary materials.

      (iii) As in the previous comment, the current model did not take into account the geometry of the system; namely, cytosol is in 3D and membrane is on 2D. Recent theoretical studies can handle the effect, and also the effect of confinement. I would appreciate it if the authors would make a comment on whether those issues are relevant or not for the conclusion of this work.

      (iv) I would appreciate it if the authors would describe the screening process more clearly. I did understand the first screening is a finite imaginary part and a positive real part at the first mode of spatial inhomogeneity in the eigenvalues. However, I did not understand the other processes clearly. The second screening is based on \lambda_N and I_Ratio, but its criteria is not clear. I think both quantities fluctuated in experimental results and I am not sure what to define numerical results match them.

      The third process is based on a fitting error using the fitting function of linear increase plus a constant. I am not sure why we need to exclude, for example, the bottom right example in Fig.S6 because it shows no oscillation until the cell length of 3um but then the gradient linearly increases. Please clarify how to justify the criteria. The same argument applies to the fourth screening process. It is not clear why the slope should be smaller than 2.

      (v) The authors claimed that the steeper gradient of MinD under glucose starvation results in cell division for shorter cells. I do not think the claim is convincing. It is necessary to measure the correlation between the length at the cell division and the gradient. It would also be nicer to show the correlation under other parameters. I think those studies truly support the authors' claim and the novelty of this work.

      (vi) The conclusion at Line 346 "This plasticity arises from spatial differences in molecular interactions between MinD and MinE, as demonstrated..." looks unclear to me. My understanding is that (i) by screening the randomly sampled parameters in the reaction-diffusion model, the authors found the parameters that "match" experimental results, and (ii) the parameters after screening show the correlation between them (k_dD-k_dE and k_D-k_ATP->ADP). The logic heavily relies on the reaction-diffusion model is quantitatively correct. First, I think it is better to explain the logic more explicitly, that is, the claim of the molecular interaction is not based on the experimental facts. Second, I personally think the reaction-diffusion model used in this work does not reproduce quantitatively the experimental results, as discussed in (iii) and also (iv). Please make some discussions on how to justify the comparison between the model and experiments.

      (vii) I did not capture the point why the authors can claim "... further distinguishing in vivo and in vitro observations. " at Line 350. I did not find the results comparing with vitro studies. I would appreciate a demonstration of vitro results and/or references.

      Minor comments:

      1. Line 214: It should be "Fange and Elf".
      2. I think it is better to show sampled points in Fig.4C and 4D to show how dense the authors sampled in the parameter space.

      REFERENCES:

      [1] Fischer-Friedrich, Elisabeth / Meacci, Giovanni / Lutkenhaus, Joe / Chaté, Hugues / Kruse, Karsten, "Intra- and intercellular fluctuations in Min-protein dynamics decrease with cell length", Proceedings of the National Academy of Sciences, 107, 6134-6139 (2010).

      [2] Meacci, Giovanni, "Physical Aspects of Min Oscillations in Escherichia Coli", PhD thesis (2006) available at https://www.pks.mpg.de/fileadmin/user_upload/MPIPKS/group_pages/BiologicalPhysics/dissertations/GiovanniMeacci2006.pdf

      Significance

      General assessment:

      I think the strength of this study is that it potentially shows the quantitative correlation between the MinD concentration gradient during the oscillation and the cell length when it divides. However, the current data of glucose starvation is not convincing enough. The model parts are interesting but their connection to the experiments is not clear in the current manuscript.

      Advance:

      The advance of this study is to measure the MinD concentration gradient under glucose starvation, and to compare the experimental results with the (simplified) model under a wide range of parameters. I do not think the advance in the current manuscript looks conceptual level because the conceptual conclusions are not really convincing from the results. In this respect, the advance of this work may be technical.

      Audience:

      As a theoretician working on biophysics, including the model of the Min system, I think a specialised audience would be interested in this study. People who are studying the mechanism of the Min oscillation and resulting cell division, particularly those who are interested in both experiments and models, would be interested in this work. For the broad audience, I do not think the novelty of this study is well described.

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

      Learn more at Review Commons


      Reply to the reviewers

      Manuscript number: RC-2024-02393

      Corresponding author(s): Katja Petzold

      1. General Statements [optional]

      We thank the reviewers for recognising the impact of our manuscript. The reviewers noted the novelty of the miRNA bulge structure, the importance of the three observed binding modes and their potential for use in future structure-based drug design, and the possible importance of the duplex release phenomenon. We are also thankful for the relevant and constructive feedback provided.

      Our responses to the comments are written point by point in blue, and any changes in the manuscript are shown in red.

      2. Description of the planned revisions

      In response to Reviewer 1 - major comment 2

      Some of the data is over-interpreted. For example, in Figure 3A, it is concluded that supplementary regions are more important for weaker seeds. Only two 8-mer seeds are present among the twelve target sites and thus it might be difficult to generalize.

      We found the relationship between seed type and the effect of supplementary pairing in our data intriguing. To further investigate this effect, we tested whether it exists in published microarray data from HCT116 cells transfected with six different miRNAs (Linsley et al., 2007; Argawal et al., 2015). Here we found that the for the two miRNAs (miR-103 and miR-106b) where we see an impact of supplementary pairing, the difference is primarily driven by 7mer-m8 seeds.

      Since the effect appears to be specific to the miRNA, we would like to test whether it can be observed for miR-34a in a larger dataset. Therefore, we plan to transfect HEK293T cells with miR-34a and analyse the mRNA response via RNAseq. We will repeat the analysis shown above, using the predicted number of supplementary pairs to categorise the dataset into groups with or without the effect of supplementary pairing. We will then compare the three seed types within these groups.

      In response to Reviewer 2 - minor comment 1, "why was the 34-nt 3'Cy3-labeled miR34a complementary probe shifted up in the presence of AGO?".

      We plan to investigate the upper band, which we hypothesise is a result of duplex release, using EMSA to ascertain whether the band height agrees with the size of the duplex.

      3. Description of the revisions that have already been incorporated in the transferred manuscript

      Reviewer #1

      Evidence, reproducibility and clarity

      Sweetapple et al. Biophysics of microRNA-34a targeting and its influence on down-regulation

      In this study, the authors have investigated binding of miR-34a to a panel of natural target sequences using EMSA, luciferase reporter systems and structural probing. The authors compared binding within a binary and a ternary complex that included Ago2 and find that Ago2 affects affinity and strengthens weak binders and weakens strong binders. The affinity is, however, generally determined by binary RNA-RNA interactions also in the ternary complex. Luciferase reporter assays containing 12 different target sites that belong to one of three seed-match types were tested. Generally, affinity is a strong contributor to repression efficiency. Duplex release, a phenomenon observed for specific miRNA-target complementarities, seems to be more pronounced when high affinity within the binary complex is observed. Furthermore, the authors use RABS for structural probing either in a construct in CIS or binding by the individual miRNA in TRANS or in a complex with Ago2. They find pronounced asymmetric target binding and Ago2 does not generally change the binding pattern. The authors observe one specific structural group that was unexpected, which was mRNA binding with bulged miRNAs, which was expected sterically problematic based on the known structures. MD simulations, however, revealed that such structures could indeed form.

      This is an interesting manuscript that contributes to our mechanistic understanding of the miRNA-target pairing rules. The combination of affinity measurements, structural probing and luciferase reporters allow for a broad correlation of target binding and repression strength, which is a well-thought and highly conclusive approach. However, there are a number of shortcomings that are summarized below.

      The manuscript is not easy to read and to follow for several reasons. First, many of the sub-Figures are not referenced in the text of the results section (1C, 1D, 2C, 4D), which is somewhat annoying. Figure 4A seems to be mis-labeled. Second, a lot of data is presented in suppl. Figures. It should be considered to move more data into the main text in order to make it easier for readers to evaluate and follow.

      Thank you for bringing this to our attention. We have now revised the figure references accordingly.

      We have relocated gel images of BCL2, WNT1, MTA2 and the control samples from Figure S3 and S4 to the main results (Figure 2A-B) to improve readability and provide controls and details that aid in clear understanding. Additionally, we have relocated panel C from Figure S6 to Figure 2C to enhance the clarity of our rationale for using polyuridine (pU) in our AGO2 binding assays.

      The updated figure is shown below, with changes to the legend marked in red.

      Figure 2. Binary and ternary____ complex binding affinities measured by EMSA. (A) Binary (mRNA:miR-34a) binding assays showing examples of BCL2, WNT1 and MTA2. (B) Ternary (mRNA:miR-34a-AGO2) binding assays showing examples of BCL2, WNT1, MTA2, and the three control targets PERFECT, SCRseed, and SCRall. The Cy5 labelled species is indicated with asterisk (*). F indicates the free labelled species (miR34a or mRNA), B indicates binary complex, and T indicates ternary complex. Adjacent titrations points differ two-fold in concentration, with maximum concentrations stated at the top right. Adjacent titration points for MTA2 differed three-fold to assess a wider concentration range. In theternary assay, miRNA duplex release from AGO2 was observed for amongst others BCL2, WNT1, PERFECT, and SCRseed (band indicated with B), while it was not observed for SCRall and MTA2. See Figures S3 and S4 for representative gel images for all targets. See Supplementary files 2 and 3 for all images and replicates. (C) Titrations with increasing miR-34a-AGO2 concentration against Cy5-labelled SCRall (left) or PNUTS (right) comparing the absence and presence of 20 μM polyuridine (pU) during equilibration. pU acted as a blocking agent, reducing nonspecific binding, as seen by the different KD,app values for SCRall and PNUTS after addition of 20 μM pU. Therefore, all final mRNA:miR-34a-AGO2 EMSAs were carried out in the presence of 20 μM pU. Labels are as stated above. (D) Individual binding profiles for each of the 12 mRNA targets assessed by electrophoretic mobility assay (EMSA). Each datapoint represents an individual experiment (n=3). Blue represents results for the binary complex, and green represents results for the ternary complex. Dotted horizontal lines represent the KD,app values, which are also stated in blue and green with standard deviations (units = nM). Note that the x-axis spans from 0.1 to 100,000 in CCND1, MTA2 and NOTCH2, whereas the remaining targets span 0.1 to 10,000.

      Some of the data is over-interpreted. For example, in Figure 3A, it is concluded that supplementary regions are more important for weaker seeds. Only two 8-mer seeds are present among the twelve target sites and thus it might be difficult to generalize.

      We have revised our wording to recognise that more 8-mer sites would be required to draw a stronger conclusion based on this hypothesis. This hypothesis would be interesting to confirm in a larger dataset but is unfortunately outside of the scope of this paper.

      Our hypothesis also aligns with recent data from Kosek et al. (NAR 2023; Figure 2D) where SIRT1 with an 8mer and 7mer-A1 seed was compared. Only the 7mer-A1 was sensitive to mutations in the central region or switching all mismatched to WC pairs.

      Page 21 now states:

      "This result indicates that the impact of supplementary binding may be greater for targets with weaker seeds, as has been observed earlier in a mutation study of miR-34a binding to SIRT1 (Kosek et al., 2023), although a larger sample size would be needed to confirm this observation."

      Furthermore, we found the relationship between seed type and the effect of supplementary pairing in our data intriguing. To further investigate this effect, we tested whether it exists in published microarray data from HCT116 cells transfected with six different miRNAs (Linsley et al., 2007; Argawal et al., 2015). Here we found that the for the two miRNAs (miR-103 and miR-106b) where we see an impact of supplementary pairing, the difference is primarily driven by 7mer-m8 seeds. We therefore plan to test whether the effect can be observed for miR-34a in a larger dataset. We have outlined our preliminary data and planned experiments in Section 2 - description of the planned revisions.

      I did not understand why the CIS system shown in 4A is a good test case for miR-34a-target binding. It appears very unnatural and artificial. This needs to be rationalized better. Otherwise it remains questionable, whether these data are meaningful at all.

      Thank you for pointing out the need for clearer rationalisation.

      The TRANS construct, where the scaffold carries the mRNA targeting sequence, provides reactivity information for the mRNA side only, while the microRNA is bound within RISC, with the backbone protected by AGO2. Therefore, to gain information on the miR-34a side of each complex we used the CIS construct, which provides reactivity information from both the miRNA and mRNA. We used the miRNA and mRNA reactivities to calculate all possible secondary structures for the binary complex, and then compared these structures to the mRNA reactivity in TRANS to find which structure fitted the reactivity patterns observed in the ternary complex.

      We have included an additional statement in the manuscript to clarify this point on pages 12-13:

      "Two RNA scaffolds were used for each mRNA target; i) a CIS-scaffold: RNA scaffold containing both mRNA target and miRNA sequence separated by a 10 nucleotide non-interacting closing loop, and ii) a TRANS-scaffold: RNA scaffold containing only the mRNA target sequence, to which free miR-34a or the miR-34a-AGO2 complex was bound (Figure 4A). The CIS constructs therefore provided reactivity information on the miRNA side, which is lacking in the TRANS construct, and was used to complement the TRANS data."

      It may be worthwhile noting that a non-interacting 10 nucleotide loop was inserted between then miRNA and mRNA of the CIS constructs, allowing the miRNA and mRNA strands to bind and release freely. The reactivity patterns of each mRNA:miRNA duplex were compared between CIS and TRANS, and showed similar base pairing (Figure 4D). Furthermore, we have previously compared the two scaffolds in our RABS methodology paper (Banijamali et al. 2022), where no differences were observed besides reduced end fraying in the CIS construct.

      For the TRANS experiments, only one specific scaffold structure is used. This structure might impact binding as well and thus at least one additional and independent scaffold should be selected for a generalized statement.

      For each construct, the potential of interaction with the scaffold was tested using the RNAstructure (Reuter & Mathews, 2010)package. Based on the results of this assessment, two different scaffolds were used for our TRANS experiments. The testing and use of scaffolds has now been clarified further on page 13:

      "The overall conformation of each scaffold with the inserted RNA was assessed using the RNAstructure (Reuter & Mathews, 2010) package to ensure that the sequence of interest did not interact with the scaffold. If any interaction was observed between the RNA of interest and the scaffold, then the scaffold was modified until no predicted interaction occurred. The different scaffolds and their sequence details are shown in supplementary information (Table S1)."

      We have previously examined the scaffold's effect on binding and structure during the development of the RABS method. We tested the same mRNA (SIRT1) in separate, independent scaffolds to verify the consistency of the results. An example of this can be found in the supplementary information (Figure S1a) of Banijamali et al. (2022).

      Generally, it would be nice to have some more information about the experiments also in the result section. Recombinant Ago2 is expressed in insect cells and re-loaded with miR-34a, luciferase reporters are transfected into tissue culture cells, I guess.

      We have now stated the cell types used for AGO2 expression and luciferase reporter assays in the results.

      On page 17 we have included:

      "Samples of each of the 12 mRNA targets, as well as miR-34a and AGO2, were synthesised in-house for biophysical and biological characterisation. Target mRNA constructs were produced via solid-phase synthesis while miR-34a was transcribed in vitro and cleaved from a tandem transcript (Feyrer et al., 2020), ensuring a 5' monophosphate group. AGO2 was produced in Sf9 insect cells."

      "To measure the affinity of each mRNA target binding to miR-34a, both within the binary complex (mRNA:miR-34a) and theternary complex (mRNA:miR-34a-AGO2), we optimised an RNA:RNA binding EMSA protocol to suit small RNA interactions. The protocol is loosely based on Bak et al. (2014)36, with major differences being use of a sodium phosphate buffering system so as not to disturb weaker interactions (James et al., 1996; Stellwagen et al., 2000), supplemented with Mg2+ as a counterion to reduce electrostatic repulsion between the two negatively charged RNAs (Misra & Draper, 1998), and fluorescently labelled probes."

      Page 19:

      " We successfully tested various RNA backgrounds, including polyuridine (pU) and total RNA extract (Figure S6B) to block any unspecific binding. Ultimately, we supplemented our binding buffer with pU at a fixed concentration of 20 µM for the ternary assays to achieve the greatest consistency."

      Page 20:

      "Repression efficacy for the 12 mRNA targets by miR-34a was assessed through a dual luciferase reporter assay6. Target mRNAs were cloned into reporter constructs and transfected into HEK293T cells."

      Page 22:

      "To infer base pairing patterns and secondary structure for each of the 12 mRNA:miR-34a pairs, we used the RABS technique (Banijamali et al., 2023) with 1M7 as a chemical probe. All individual reactivity traces are shown in Figure S9. Reactivity of each of the 22 miR-34a nucleotides was assessed upon binding to each of the 12 mRNA targets within a CIS construct, containing both miR-34a and the mRNA target site separated by a non-interacting 10-nucleotide loop. The two RNAs can therefore bind and release freely within the CIS construct and reactivity information is collected from both RNA strands."

      In the first sentence of the abstract, Argonaute 2 should be replaced by Argonaute only since other members bind to miRNAs as well.

      Thank you for recognising this. It has now been corrected.

      Significance

      This is an interesting manuscript that contributes to our mechanistic understanding of the miRNA-target pairing rules. The combination of affinity measurements, structural probing and luciferase reporters allow for a broad correlation of target binding and repression strength, which is a well-thought and highly conclusive approach. However, there are a number of shortcomings.

      We thank the reviewer for recognising the approach and impact of our work. In addition we thank the reviewer for identifying the need for further data to support our conclusions from the luciferase assays, which is something that we plan to address, as described in section 2.



      Reviewer #2

      Evidence, reproducibility and clarity

      Summary: Sweetapple et al. took the approaches of EMSA, SHAPE, and MD simulations to investigate target recognition by miR-34a in the presence and absence of AGO2. Surprisingly, their EMSA showed that guide unloading occurred even with seed-unpaired targets. Although previous studies reported guide unloading, they used perfectly complementary guide and target sets. The authors of this study concluded that the base-pairing pattern of miR-34a with target RNAs, even without AGO2, can be applicable to understanding target recognition by miR-34a-bound AGO2.

      Major comments:

      (Page 11 and Figure S4) The authors pre-loaded miR-34a into AGO2 and subsequently equilibrated the RISC with a 5' modified Cy5 target mRNA. Since properly loaded miR-34a is never released from AGO2, it is impossible for the miR-34a loaded into AGO2 to form the binary complex (mRNA:miR-34a) in the EMSA (guide unloading has been a long-standing controversy). However, they observed bands of the binary complex in Figure S4. The authors did not use ion-exchange chromatography. AGOs are known to bind RNAs nonspecifically on their positively charged surface. Is it possible that most miR-34a was actually bound to the surface of AGO2 instead of being loaded into the central cleft? This could explain why they observed the bands of the binary complex in EMSA.

      Thank you for mentioning this crucial point which has been a focus of our controls. We have addressed this point in four ways:

      Salt wash during reverse IMAC purification. Separation of unbound RNA and proteins via SEC. Blocking non-specific interactions using polyuridine. Observing both the presence and absence of duplex release among different targets using the same AGO2 preparation and conditions.

      Firstly, although we did not use a specific ion exchange column for purification, we believe the ionic strength used in our IMAC wash step was sufficient to remove non-specific interactions. We used A linear gradient with using buffer A (50 mM Tris-HCl, 300 mM NaCl, 10 mM Imidazole, 1 mM TCEP, 5% glycerol v/v) and buffer B (50 mM Tris-HCl, 500 mM NaCl, 300 mM Imidazole, 1 mM TCEP, 5% glycerol) at pH 8. The protocol followed recommendation by BioRad for their Profinity IMAC resins where it is stated that 300 mM NaCl should be included in buffers to deter nonspecific protein binding due to ionic interactions. The protein itself has a higher affinity for the resin than nucleic acids.

      A commonly used protocol for RISC purification follows the method by Flores-Jasso et al. (RNA 2013). Here, the authors use ion exchange chromatography to remove competitor oligonucleotides. After loading, they washed the column with lysis buffer (30 mM HEPES-KOH at pH 7.4, 100 mM potassium acetate, 2 mM magnesium acetate and 2 mM DTT). AGO was eluted with lysis buffer containing 500 mM potassium acetate. Competing oligonucleotides were eluted in the wash.

      As ionic strength is independent of ion identity or chemical nature of the ion involved (Jerermy M. Berg, John L. Tymoczko, Gregory J. Garret Jr., Biochemistry 2015), we reasoned that our Tris-HCl/NaCl/ imidazole buffer wash should have at comparable ionic strength to the Flores-Jasso protocol.

      Our total ionic contributions were: 500 mM Na+, 550 mM Cl-, 50 mM Tris and 300 mM imidazole. We recognise that Tris and imidazole are both partially ionized according the pH of the buffer (pH 8) and their respective pKa values, but even if only considering the sodium and chloride it should be comparable to the Flores-Jasso protocol.

      We have restated the buffer compositions below written the methods section more explicitly to describe this:

      "Following dialysis, any precipitate was removed by centrifugation, and the resulting supernatant was loaded onto a IMAC buffer A-equilibrated HisTrap-Ni2+ column to remove TEV protease, other proteins, and non-specifically bound RNA. A linear gradient was employed using IMAC buffers A and B."

      Secondly, after reverse HisTrap purification, AGO2 was run through size exclusion chromatography to remove any remaining impurities (shown Figure S2B).

      Thirdly, knowing that AGO2 has many positively charged surface patches and can bind nucleic acid nonspecifically (Nakanishi, 2022; O'Geen et al., 2018), we tested various blocking backgrounds to eliminate nonspecific binding effects in our EMSA ternary binding assays. We were able to address this issue by adding either non-homogenous RNA extract or homogenous polyuridine (pU) in our EMSA buffer during equilibration background experiments. This allowed us to eliminate non-specific binding of our target mRNAs, as shown previously in Supplementary Figure S6. We appreciate that the reviewer finds this technical detail important and have moved the panel C of figure S6 into the main results in Figure 2C, to highlight the novel conditions used and important controls needed to be performed. If miR-34a were non-specifically bound to the surface of AGO2 after washing, this blocking step would render any impact of surface-bound miR-34a negligible due to the excess of competing polyuridine (pU).

      Our EMSA results show that, using polyU, we can reduce non-specific interaction between AGO2 and RNAs that are present. And still, duplex release occurs despite the blocking step. It is therefore less likely that duplex release is caused by surface-bound miR-34a.

      Finally, the observation of distinct duplex release for certain targets, but not for others (e.g. MTA2, which bound tightly to miR-34a-AGO2 but did not exhibit duplex release; see Figure 2), argues against the possibility that the phenomenon was solely due to non-specifically bound RNA releasing from AGO2.

      In response to the reviewers statement "Since properly loaded miR-34a is never released from AGO2, it is impossible for the miR-34a loaded into AGO2 to form the binary complex (mRNA:miR-34a)" we would like to refer to the three papers, De et al. (2013) Jo MH et al. (2015), and Park JH et al. (2017), which have previously reported duplex release and collectively provide considerable evidence that miRNA can be unloaded from AGO in order to promote turnover and recycling of AGO. It is known that AGO recycling must occur, therefore there must be some mechanisms to enable release of miRNA from AGO2 to enable this. It is possible that AGO recycling proceeds via miRNA degradation (TDMD) in the cell, but in the absence of enzymes responsible for oligouridylation and degradation, the miRNA duplex may be released. As TDMD-competent mRNA targets have been observed to release the miRNA 3' tail from AGO2 (Sheu-Gruttadauria et al., 2019; Willkomm et al., 2022), there is a possible mechanistic similarity between the two processes, however, we do not have sufficient data to make any statement on this.

      (Page 18 and Figure S5) Previous studies (De et al., Jo MH et al., Park JH et al.) reported guide unloading when they incubated a RISC with a fully complementary target. However, neither MTA2, CCND1, CD44, nor NOTCH2 can be perfectly paired with miR-34a (Figure 1A). Therefore, the unloading reported in this study is quite different from the previously reported works and thus cannot be explained by the previously reported logic. The authors need to explain the guide unloading mechanism that they observed. Otherwise, they might misinterpret the results of their EMSA and RABS of the ternary complex.

      The three aforementioned studies have reported unloading/duplex release. However, they did not only report fully complementary targets in this process.

      De et al. (2013) reported that "highly complementary target RNAs promote release of guide RNAs from human Argonaute2".

      Subsequently, Park et al. (2017) reported: "Strikingly, we showed that miRNA destabilization is dramatically enhanced by an interaction with seedless, non-canonical targets."

      A figure extracted from Figure 5 of Park et al. is shown below illustrating the occurrence of unloading in the presence of seed mismatches in positions 2 and 3 (mm 2-3). Jo et al. (2015) also reported that binding lifetime was not affected by the number of base pairs in the RNA duplex.

      In addition to these three reports, a methodology paper focusing on miRNA duplex release was published recently titled "Detection of MicroRNAs Released from Argonautes" (Min et al., 2020).

      Therefore, we do believe that the previously observed microRNA release is similar to our observation. Here we also correlate it to structure and stability of the complex.

      (Page 20) The authors reported, "it is notable that the seed region binding does not appear to be necessary for duplex release." The crystal structures of AGO2 visualize that the seed of the guide RNA is recognized, whereas the rest is not, except for the 3' end captured by the PAZ domain. How do the authors explain the discrepancy?

      In this manuscript, we intend to present our observations of duplex release. There are many potential relationships between duplex release and AGO2 activity, which we do not have data to speculate upon. Previous studies, such as Park et al. (2017) have also observed non-canonical and seedless targets leading to duplex release, supporting our findings. Additionally, other publications including McGearly et al. (2019) report 3'-only miRNA targets, Lal et al. (2009) have documented seedless binding by miRNA and their downstream biological effects, and Duan et al. (2022) show that a large number of let-7a targets are regulated through 3′ non-seed pairing.

      It is also possible that duplex release is not coupled to classical repression outcomes, and does not need to proceed by the seed, but instead regulates AGO2 recycling before AGO2 enters the quality control mode of recognising the formed seed.

      (Pages 22) The authors mentioned, "It follows that the structure imparted via direct RNA:RNA interaction remains intact within AGO2, highlighting the role of RNA as the structural determinant." A free guide and a target can start their annealing from any nucleotide position. In contrast, a guide loaded into AGO needs to start annealing with targets through the seed region. Additionally, the Zamore group reported that the loaded guide RNA behaves quite differently from its free state (Wee et al., Cell 2012). How do the authors explain the discrepancy?

      The key point we would like to emphasise is that AGO does not seem to alter the underlying RNA:RNA interactions. The bound state in the ternary complex reflects the structure established in the binary complex. We do not aim to claim a specific sequence of events, as this interpretation is not possible from our equilibrium data. Our data indicates that the protein is flexible enough to accommodate the RNA structure that is favoured in the binary complex. This hypothesis is further supported by our MD simulation, which demonstrates the accommodation of a miRNA-bulge structure within AGO2.

      Targets lacking seeds have been identified previously (McGeary et al. 2019, Park et al. 2017, Lal et al. 2009) and can bind to miRNA within AGO. Therefore, there must be a mechanism by which these targets can anneal within AGO, such as via sequence-independent interactions (as discussed in question 3).

      With respect to Wee et al., (2012), which studied fly and mouse AGO2 and found considerable differences between the thermodynamic and kinetic properties of the two AGO2 species. Furthermore, they found different average affinities between the two species, with the fly AGO binding tighter the mouse. Following this logic, it is not unexpected that human AGO2 would have unique properties compared to those of fly and mouse.

      Below is an extract from Wee et al., (2012):

      "Our KM data and published Argonaute structures (Wang et al., 2009) suggest that 16-17 base pairs form between the guide and the target RNAs, yet the binding affinity of fly Ago2-RISC (KD = 3.7 {plus minus} 0.9 pM, mean {plus minus} S.D.) and mouse AGO2-RISC (KD = 20 {plus minus} 10 pM, mean {plus minus} S.D.) for a fully complementary target was comparable to that of a 10 bp RNA:RNA helix. Thus, Argonaute functions to weaken the binding of the 21 nt siRNA to its fully complementary target: without the protein, the siRNA, base paired from positions g2 to g17, is predicted to have a KD ∼3.0 × 10−11 pM (ΔG25{degree sign}C = −30.7 kcal mol−1). Argonaute raises the KD of the 16 bp RNA:RNA hybrid by a factor of > 1011."

      In the Wee et al. (2012) paper, affinity data on mouse and fly AGO2 was collected via filter binding assays, using a phosphorothioate linkage flanked by 2′-O-methyl ribose at positions 10 and 11 of the target to prevent cleavage. They then compared the experimentally determined mean KD and ΔG values for each species to predicted values of an RNA:RNA helix of 16-17 base-pairs. No comparison was made between individual targets, and no experimental data was collected for the RNA:RNA binding. The calculated energy values were made based on a simple helix without taking into account any possible secondary structure features. Considering the different AGO species, alternative experimental setup, modified nucleotides in the tested RNA, and the computationally predicted RNA values compared to the averaged experimental values, we believe there is considerable reason to observe differences compared to our findings.

      We have expanded our discussion on page 27 to the following:

      "An earlier examination of mRNA:miRNA binding thermodynamics by Wee and colleagues (2012) found that mouse and fly AGO2 reduce the affinity of a guide RNA for its target61. Our data indicate that the range of miR-34a binary complex affinities is instead constricted by human AGO2 in the ternary complex - strengthening weak binders while weakening strong binders. The 2012 study reported different average affinities between the two AGO2 species, with the fly protein binding tighter the mouse. Following this logic, it is not unexpected that human AGO2 would have unique properties compared to those of fly and mouse."

      The authors concluded that the range of binary complex affinities is constricted by human AGO2 in the ternary complex - strengthening weak binders while weakening strong binders. This may hold true for miR-34a, but it cannot be generalized. Other miRNAs need to be tested.

      That is true, we have now adjusted the wording to encompass this more clearly, shown below. Testing of further miRNAs is the likely content of future work from us and others.

      "Our data indicate that the range of miR-34a binary complex affinities is instead constricted by human AGO2 in the ternary complex - strengthening weak binders while weakening strong binders."

      Minor comments:

      (Figure S2) Why was the 34-nt 3'Cy3-labeled miR34a complementary probe shifted up in the presence of AGO?

      We believe this observation is also indicative of duplex release. At the time that these activity assays were collected, we were not as aware of the presence of duplex release so did not test it further, assuming it may be due to transient interactions. We plan to investigate this via EMSA and have included this in the planned revisions (section 2).

      2.(Page 17) Does the Cy3 affect the interaction of the 3' end of miR-34 with AGO2?

      miR-34a-3'Cy5 was used for binary experiments only and the reverse experiment was conducted as a control (where Cy5 was located on the mRNA) (Figure S3b), showing no change in affinity/interaction when the probe was switched to the target. For ternary experiments the mRNA target was labelled on the 5' terminus, to make sure there was no interference with loading miR-34a into AGO2.

      A Cy3 labelled RNA probe (fully complementary to miR-34a) was used to detect miR-34a in northern blots, but AGO2 interaction is not relevant here under denaturing conditions.

      Otherwise, the 34-nt slicing probe had Cy3 on the 5 nt 3' overhang and should therefore not interact with AGO.

      1. Several groups reported that overproduced AGOs loaded endogenous small RNAs. The authors should mention that their purified AGO2 was not as pure as a RISC with miR-34a. Otherwise, readers might think that the authors used a specific RISC.

      We have now improved our explanation of the loading efficiency to make it more clear to the reader that our AGO2 sample was not fully bound by miR-34a, and that all concentrations refer to the miR-34a-loaded portion of AGO2. The following text can be found in the results on page 18:

      "The mRNA:miR-34a-AGO2 assay had a limited titration range, reaching a maximum miR-34a-AGO2 concentration of 268 nM due to a 5% loading efficiency (see Figure S2D for loading efficiency quantification). The total AGO2 concentration was thus 20-fold higher than the miR-34a-loaded portion. Further increase in protein concentration was prevented by precipitation. Weaker mRNA targets (CD44, CCND1, and NOTCH2) did not reach a saturated binding plateau within this range, leading to larger errors in their estimated KD,app values. However, reasonable estimation of the KD,app was possible by monitoring the disappearance of the free mRNA probe. Note that we refer to the miR-34a-loaded portion of AGO2 when discussing concentration values for all titration ranges. To ensure AGO2 binding specificity despite low loading efficiency, a scrambled control was used (SCRall; lacking stable base pairing with miR-34a or other human miRNAs according to the miRBase database57). SCRall showed no interaction with miR-34a-AGO2 (Figure 2B)."

      (Figure legend of Figure S5) Binding was assessed "by."

      Thank you for pointing this out, it is now fixed.

      (Page 17) It would be great if the authors could even briefly describe the mechanism by which the sodium phosphate buffer with magnesium does not disturb weaker interactions by citing reference papers.

      We have now added a supplementary methods section to our manuscript and included the description below on page 10:

      "We found that a more traditional Tris-borate-EDTA (TBE) buffer disrupted weaker RNA:RNA binding interactions (Supplementary Methods Figure M1). Borate anions form stable adducts with carbohydrate hydroxyl groups (James et al., 1996) and can form complexes with nucleic acids, likely through amino groups in nucleic bases or oxygen in phosphate groups (Stellwagen et al., 2000). This makes TBE unsuitable for assessment of RNA binding, particularly involving small RNA molecules, which typically have weaker affinities. We therefore adapted our buffer system to a sodium phosphate buffer supplemented with magnesium. Magnesium acts as a counterion to reduce electrostatic repulsion between the two negatively charged backbones by neutralisation (Misra et al., 1998)."

      We have also clarified the buffer adaptions in our results section on page 17:

      The protocol is loosely based on Bak et al. (2014)36, with major differences being use of a sodium phosphate buffering system so as not to disturb weaker interactions(James et al., 1996; Stellwagen et al., 2000), supplemented with Mg2+ as a counterion to reduce electrostatic repulsion between the two negatively charged RNAs(Misra & Draper, 1998), and fluorescently labelled probes. Original gel images and quantification are shown in supplementary Figures S3 and S4. All KD,app values are shown in Supplementary Table 1, and represent the mean of three independent replicates.

      Figure M1. Comparison of Tris-borate EDTA (TBE) and sodium phosphate with magnesium (NaP-Mg2+) buffer systems for EMSA. Cy5-labelled miR-34a and unlabelled CD44 were equilibrated in the two different buffer systems, using the same titration range. No mobility shifts were observed in the TBE system, while clear binding shifts were observed in the NaP-Mg2+ system.

      6.(Page 22) The authors cited Figure 4C in the sentence, "Comparison between CIS and TRANS ..." Is this supposed to be Figure 4D?

      The reviewer was correct in their assumption, and this has now been corrected.

      7.(Figure 6) Readers would appreciate it if the guide and target were colored in red and blue. The color codes have been used in most papers reporting AGO structures. The current color codes are opposite.

      We have now adjusted the colour schemes throughout the manuscript, and Figure 6 has been modified to the following:

      __"Figure 6. The miRNA-bulge structure is readily accommodated by AGO2 as shown by molecular dynamics simulation. __Panel (A) displays a snapshot of the all-atom MD simulation of miR-34a (red) and NOTCH1 (blue) in AGO2. The NOTCH1:miR-34a duplex is shown with AGO2 removed for clarity and is rotated 90{degree sign} to show the miRNA bulge and bend in the duplex. This NOTCH1:miR-34a-AGO2 structure is compared with (B), which shows the crystal structure of miR-122 (orange) paired with its target (purple) via the seed and four nucleotides in the supplementary region (PDB-ID 6N4O17), and (C), which shows the crystal structure of miR-122 (orange) and its target (green) with extended 3' pairing, necessary for the TDMD-competent state (PDB-ID 6NIT19). AGO2 is depicted in grey, with the PAZ domain in green, and the N-terminal domain marked with N. The miRNA duplexes in (B) and (C) feature symmetrical 4-nucleotide internal loops, whereas the NOTCH1 structure in (A) has an asymmetrical miRNA bulge with five unpaired nucleotides on the miRNA side and a 3-nucleotide asymmetry."

      Significance

      This paper will have a significant impact on the field if seed-unpaired targets can indeed unload guide RNAs. The authors may want to validate their results very carefully.

      We thank the reviewer for recognising the significance of duplex release (or guide unloading) from AGO2. We agree that the observations should be tested rigorously and have outlined the actions we took to ensure validity in our AGO2 preparation.

      __Reviewer #3 __

      Evidence, reproducibility and clarity (Required):

      In this manuscript, the authors use a combination of biochemical, biophysical, and computational approaches to investigate the structure-function relationship of miRNA binding sites. Interestingly, they find that AGO2 weakens tight RNA:RNA binding interactions, and strengthens weaker interactions.

      Given this antagonistic role, I wonder: shouldn't there be an 'average' final binding affinity? Furthermore, if I understand correctly, not many trends were observed to correlate binding affinity with repression, etc.

      Overall, there was no 'average' final binding affinity observed, as the binary assays had a much higher maximum (NOTCH2binary affinity was within the micromolar range) skewing the mean average of the binary affinities to 657 nM, versus 111 nM for the ternary affinities. We also compare the variances of the binary and ternary affinity datasets using the F-test and found that F > F(critical one tail) and thus the variation of the two populations is unequal (binary variation is significantly larger than ternary).

      F-Test Two-Sample for Variances

      • *

      binary affinity

      ternary affinity

      Mean

      657.3

      110.971667

      Variance

      2971596.1

      24406.4012

      Observations

      12

      12

      df

      11

      11

      F

      121.754784

      P(F

      7.559E-10

      F(critical one-tail)

      2.81793047

      We agree that the overall correlation between affinity and repression was not strong, although we found a stronger correlation within the miRNA-bulge group (Figure 5C and S7C). A larger sample size of miRNA bulge-forming duplexes would be needed to test the generalizability of this observation.

      Given the context of the study - whereby structure is being investigated as a contributing factor to the interaction between the miRNA and mRNA, I find it interesting that the authors chose to use MC-fold to predict the structures of the mRNA, rather than using an experimental approach to assess / validate the structures. Thirty-seven RNAs were assessed; I think even for a subset (the 12 that were focused on in the study), the secondary structure should be validated experimentally (e.g., by chemical probing experiments, which the research group has demonstrated expertise in over the last several years). The validation should follow the in silico folding approach used to narrow down the region of interest. It is necessary to know whether an energy barrier (associated with the mRNA unfolding) has to occur prior to miRNA binding; this could help explain some of the unexplained results in the study. Indeed, the authors mention that there are many variables that influence miRNA regulation.

      Indeed, experimentally validated structures offer valuable insights that cannot be obtained solely through sequence-based predictions. This is why we opted to employ our RABS method to experimentally evaluate the binary and ternary complex binding of our 12 selected targets (as depicted in Figures 4 and S9 and discussed in the text on pages 23-24). While we (in silico) assessed all 37 RNA targets that were experimentally confirmed at the time, selecting 12 to represent both biological and predicted structural diversity, it would have been impractical to experimentally pre-assess all the targets not included in the final selection. Our in-silico assessment was designed to narrow down the regions of interest and evaluate predicted secondary structures present. The pipeline is shown in Figure 1. Details of the code used in the in-silico analysis are provided in Supplementary File 1.

      Regarding the energy of unfolding of mRNA, our constructs considered the isolated binding sites thus the effects of surrounding mRNA interactions were removed. We compared our affinities to dG as well as MFE and have now included this analysis in Figure S8A. Additionally, we have included the text on page 27-28 of the discussion:

      "Gibbs free energy (G), which is often included in targeting prediction models as a measure of stability of the miRNA:mRNA pair12,62, correlated with the log of our binary KD,app values, using ΔG values predicted by RNAcofold (R2 = 0.61). There was a weaker correlation with the free energy values derived from the minimum free energy (MFE) structures predicted by RNAcofold (R2 = 0.41) (Figure S8A). This result highlights the contribution of unfolding (in ΔG) as being an important in predicting KD. The differences between ΔG and KD,app are likely primarily due to inaccurately predicted structures used for energy calculations."

      Additionally, we assessed the free form of all mRNA targets via RABS (Figure S9) and observed that the seed of each free mRNA was available for miRNA binding (seeds of the free mRNA were not stably bound).

      Finally, when designing our luciferase plasmids we used RNAstructure (Reuter & Mathews, 2010) to check for self-folding effects which could interfere with target site binding and ensured that all plasmids were void of such effects.

      In the methods, T7 is italicized by accident in the T7 in vitro transcription section. Bacmid is sometimes written with a capital B and other times with a lower-cased b. The authors should be consistent. The concentration of TEV protease that was added (as opposed to the volume) should be described for reproducibility.

      Thank you for pointing out these overlooked points. They have now been corrected.

      In figure S2D, what is the second species in the gel on the right-hand side of the gel in the miR-34a:AGO lanes? The authors should mention this.

      We believe that the faint upper band corresponds to other longer RNA species loaded into AGO2. As AGO2 is loaded with a diversity of RNA species, it is likely that some of them may have a weak affinity for the miR-34a-complementary probe, and therefore show up on the northern blot.

      Figure S3B and S3A are referenced out of order in the text. In regard to S3A, what are the anticipated or hypothesized alternative conformations for NOTCH1, DLL1, and MTA2? There are really interesting things going on in the gels, also for HNF4a and NOTCH2. Can the authors offer some explanation for why the free RNA bands don't seem to disappear, but rather migrate slowly? Is this a new species?

      The order of the figure references have now been updated, thank you for alerting us to this.

      Figure S3A: For MTA2, the two alternative conformations are shown in Figure S9 and S10 (and shown below here, miR-34aseed marked in pink). It appears that a single conformation is favoured at high concentration (> 1 µM) while the two conformations are present at {less than or equal to} 1 µM. The RABS data for MTA2 also indicated multiple binding conformations, as the reactivity traces were inconsistent. We expect that the conformation shown on the left was most dominant within AGO2, based on the reactivity of the TRANS + AGO assays. However, we cannot exclude a possible G-quadruplex formation due to the high G content of MTA2 (shown below right).

      Regarding NOTCH1 and DLL1, a faint fluorescent shadow was observed beneath the miR-34a bound band. The RABS reactivity traces indicated a single dominant conformation for these targets, so it is possible that the lower shadow observed was due to more subtle differences in conformation, such as the opening/closing of one or a few base pairs at the terminus or bulge, (i.e. end fraying). HNF4α and NOTCH2 appear to never fully saturate the miR-34a, so a small un-bound population remains visible on the gel. For NOTCH2 this free miR-34a band appears to migrate upwards, possibly due to overloading the gel lane with excess NOTCH2 (which are not observed in the Cy5 fluorescence image).

      In the EMSA for Perfect, why does the band intensity for the bound complex increase then decrease? How many replicates were run for this? This needs to be reconciled.

      As for all EMSAs, three replicates were carried out for each mRNA target and all gels are shown in Supplementary Files 2 and 3, for the binary and ternary assays respectively.

      Uneven heat distribution across the gel can lead to bleaching of the Cy5 fluorophore. To address this, we we used a circulating cooler in our electrophoresis tank, as outlined in our methods (page 10). However, the aforementioned gel for one of thePERFECT sample replicates appears to have been evenly cooled. As the binding ratio (rather than total band volume) was used for quantification, the binding curve was unaffected, and this did not influence KD,app.

      We have now replaced the exemplary gel for PERFECT in Figure S3 with a more representative and evenly labelled gel from our replicates (Cy5 fluorescence image shown below). The binding curve for PERFECT is also shown here:

      The authors list that the RNA concentration was held constant at 10 nM; in EMSAs, the RNA concentration should be less than the binding affinity; what is the lowest concentration of protein used in the assays shown in S3A? Is this a serial dilution? It seems to me like the binding assays for MTA2, Perfect, and SRCseed might have too high of an RNA concentration. (Actually, now I see in the supplement the concentrations of proteins, and the RNA concentration is too high). Also, why is the intensity of bands for bound complex for SRCseed more intense than the free RNA?

      Why are the binding affinity error bars so large (e.g., for NOTCH2 with mir-34a) - 6 uM +/- 3 uM?

      No protein was used in the binary assays shown in Figure S3A. For the ternary assays in Figure S4, the maximum concentration of miR-34a-loaded AGO2 (miR-34a-AGO2) was 268 nM, with a serial dilution down to a minimum of 0.06 nM.

      Optimal EMSA conditions require a constant RNA concentration that is lower than the binding affinity to accurately estimate high-affinity interactions.

      For our tightest binders, such as SIRT1, we can confidently state that the KD,app is less than 10 nM, estimated at 0.4 {plus minus} 1.1 nM. Therefore, the accuracy of this estimation is reduced, and the standard deviation is larger than the estimated KD,app. As NOTCH2 bound miR-34a very weakly and did not reach a fully bound plateau, the resulting high error was expected. Consequently, we do not have the same level of certainty for extremely tight or weak binders. In this study, the relative affinities were of primary importance.

      We have included on page 18:

      As the Cy5-miR-34a concentration was fixed to 10 nM to give sufficient signal during detection, KD,app values below 10 nM have a lower confidence.

      Regarding the control samples PERFECT and SCRseed, our focus was not on determining the exact KD,app of these artificial constructs. Instead, we were primarily interested in whether they exhibited binding and under which conditions. For SCRseed, we neither adjusted the titration range nor calculated KD,app. For PERFECT, the concentration was adjusted to a lower range of 30 nM - 0.001 nM to give a relative comparison with the other tight binder SIRT1. However, further reduction in RNA concentration was not pursued, as it already fell well below the 10 nM sensitivity threshold.

      Regarding the intensity of the bound SCRseed band, we observed that the bound fluorophore often resulted in stronger intensity than for the free probe. This was observed for a number of the samples (PERFECT, BLC2, SCRseed). A previous publication reported that Cy5 is sequence dependent in DNA, that the effect is more sensitive to double-stranded DNA, and that the fluorophore is sensitive to the surrounding 5 base pairs (Kretschy, Sack and Somoza, 2016). It is likely that the same phenonenon exists in RNA.

      For MTA2, the two alternative conformations (shown in Figure S9 and S10) make assessment of KD,app more difficult. As the higher affinity conformation did not reach a fully-bound plateau before the weaker affinity conformation appeared, the binding curve plateau (where all miR-34a was bound) reflected the weaker conformation KD,app. We increased the range of titration tested by using a three-fold serial dilution, but further reduction in RNA concentration would not have been fruitful as it already dropped below well below the 10 nM sensitivity range. Therefore the MTA2 binary complex had a higher error at (944 {plus minus} 274 nM) and lower confidence.

      We then decided to run a competition assay to detect the weaker KD,app of MTA2. The assay was set up using the known binding affinity of CD44, which was labelled with Cy5 to track the reaction. MTA2 was titrated against a constant concentration of Cy5-CD44:miR-34a, and disruption of the CD44 and miR-34a binding was monitored. We fitted the data to a quadratic for competitive binding (Cheng and Prusoff., 1973) to calculate the KD,app for competitive binding, or KC,app.

      We validated our competition assay by comparing it with our direct binding assays, specifically assessing CD44 in a self-competition assay. The CD44 KC,app (168 {plus minus} 24 nM; mean and SD of three replicates) was found to be consistent with the KD,app obtained from the direct assay (165 {plus minus} 21 nM).

      As we wanted all affinity data to be directly comparable (using the same methodology), we compared the KD,app values obtained via direct assay in the manuscript. It appears that the competitive EMSA assay for MTA2 reflects the weaker affinity conformation observed in the direct assay.

      It would be very helpful if the authors wrote in the Kds in Figure 2A in green and blue (in the extra space in the plots). This would help the reader to better understand what's going on, and for me, as a reviewer, to better consider the analysis/conclusions presented by the authors.

      KD,app values are written in in green and blue in what is now Figure 2D (originally Figure 2A).

      The authors state on page 18 that 'Interestingly, however, we did not observe a correlation between binary or ternary complex affinity and seed type.' They should elaborate on why this is interesting.

      The prevailing view is that the miRNA seed type significantly influences affinity within AGO2. The largest biochemical studies of miRNA-target interactions to date, conducted by McGeary et al. (2019, 2022), used AGO-RBNS (RNA Bind-n-Seq) to reveal relative binding affinities. These studies demonstrated strong correlations between the canonical seed types and binding affinity. Therefore, we find it interesting that no such correlation was observed in our dataset (despite its small size).

      We have now added to the manuscript (page 20):

      "The largest biochemical studies of miRNA-target interactions to date (McGeary et al., 2019, 2022) used AGO-RBNS (RNA Bind-n-Seq) to extract relative binding affinities, demonstrating strong correlations between the canonical seed types and binding affinity. Therefore, it is intriguing that our dataset, despite its small size, showed no such correlation."

      Figure 2C is not referenced in the text (the authors should go back through the text to make sure everything is referenced and in order). The Kds should be listed alongside the gels in Figure 2C.

      Figure 2 has now been rearranged and updated, with KD,app values listed in what is now Figure 2D.

      Figure 3B is rather confusing to understand.

      We have now adapted Figure 3 to simplify readability. Panel B has now been moved to C, and we have introduced panel A (moved from Figure 2B). In Figure 3C (originally 3B) we have added arrows to indicate the direction of affinity change from binary to ternary complex, and moved the duplex release information to panel A. We thank the reviewer and think that the data is now much clearer.

      Figure 3. AGO2 moderates affinity by strengthening weak binders and weakening strong binders. (A) Correlation of relative mRNA:miR-34a with mRNA:miR-34aAGO2 binding affinities. No seed type correlation is observed, seeds coloured, where 8mer is pink, 7mer-m8 is turquoise, and 7-mer-A1 is mauve. The slope of the linear fit is 0.48, and intercept on the (log y)-axis is 7.11. The occurrence of miRNA duplex release from AGO2 is marked with diamonds. (B) miR-34a-mediated repression of dual luciferase reporters fused to the 12 mRNA targeting sites. Luciferase activity from HEK293T cells co-transfected with each reporter construct, miR-34a was measured 24 hours following transfection and normalised to the miR-34a-negative transfection control. Each datapoint represents the R/F ratio for an independent experiment (n=3) with standard deviations indicated. SCRseed is a scrambled seed control, SCRall is a fully scrambled control, and PERFECT is the perfect complement of miR-34a. Dotted horizontal lines represent the repression values for the 22-nucleotide seed-only controls6 for the respective seed types, in the absence of any other WC base pairing. (C) Comparison of relative target repression with relative affinity assessed by EMSA. Blue represents mRNA:miR-34a affinity (binary complex), while green represents mRNA:miR-34a-AGO2 affinity (ternary complex). Arrows indicate the direction of change in affinity upon binding within AGO2 compared to the binary complex. It is seen that AGO2 moderates affinity bi-directionally by strengthening weak binders and weakening strong binders.

      Page 20: Perfect should be italicized.

      Thank you for bringing this to our attention, this how now been adjusted.

      Have the authors considered using NMR to assess the base pair pattern formed between the miRNA:mRNA complexes (with / without AGO)? As a validation for results obtained by RABS? This could be helpful for the Asymmetric target binding section, the Ago increases flexibility section, and the three distinct structural groups section in the results. It is widely accepted that while chemical probing is insightful, results should be validated using alternative approaches. Distinguishing structural changes and protected reactivity in the presence of protein is challenging.

      NMR provides high-resolution information on RNA base-pairing patterns, allowing us to compare our RABS results for SIRT1with those obtained via NMR (Banijamali et al., 2022) for the binary complex. For SIRT1, the RNA:RNA structures identified were consistent between both methods. However, using NMR to measure RNA:RNA binding within AGO2 is challenging due to the protein's large size. Currently, there are no published complete NMR structures of RNA within AGO2. The largest solution-state NMR structures published that include AGO consist solely of the PAZ domain. Our group has been working on method development using DNP-enhanced solid-state NMR to obtain structural information within the complete AGO2 protein, but the current resolution does not allow us to fully reconstruct a complete NMR structure. We hope that in the coming years, this will be a method to evaluate RNA within AGO. This limitation highlights the advantage of RABS in providing RNA base-pairing information within the ternary complex in solution.

      Reviewer #3 (Significance (Required)):

      The work is helpful for understanding how microRNAs recognize and bind their mRNA targets, and the impact Ago has on this interaction. I think for therapeutic studies, this will be helpful for structure-based design. Especially given the three types of structures identified to be a part of the interaction.

      We thank the reviewer for their detailed remarks, especially concerning the importance of technical details the binding assays. We further thank the reviewer for recognising the potential impact of our work for rational design.

      4. Description of analyses that authors prefer not to carry out

      • *

      In response to Reviewer 2 - major comment 1, we prefer to not run an additional ion exchange purification on the AGO2 protein due to the reasoning discussed above, which is repeated here:

      We have addressed this point in three ways:

      Thank you for mentioning this crucial point which has been a focus of our controls. We have addressed this point in four ways:

      Salt wash during reverse IMAC purification. Separation of unbound RNA and proteins via SEC. Blocking non-specific interactions using polyuridine. Observing both the presence and absence of duplex release among different targets using the same AGO2 preparation and conditions.

      Firstly, although we did not use a specific ion exchange column for purification, we believe the ionic strength used in our IMAC wash step was sufficient to remove non-specific interactions. We used A linear gradient with using buffer A (50 mM Tris-HCl, 300 mM NaCl, 10 mM Imidazole, 1 mM TCEP, 5% glycerol v/v) and buffer B (50 mM Tris-HCl, 500 mM NaCl, 300 mM Imidazole, 1 mM TCEP, 5% glycerol) at pH 8. The protocol followed recommendation by BioRad for their Profinity IMAC resins where it is stated that 300 mM NaCl should be included in buffers to deter nonspecific protein binding due to ionic interactions. The protein itself has a higher affinity for the resin than nucleic acids.

      A commonly used protocol for RISC purification follows the method by Flores-Jasso et al. (RNA 2013). Here, the authors use ion exchange chromatography to remove competitor oligonucleotides. After loading, they washed the column with lysis buffer (30 mM HEPES-KOH at pH 7.4, 100 mM potassium acetate, 2 mM magnesium acetate and 2 mM DTT). AGO was eluted with lysis buffer containing 500 mM potassium acetate. Competing oligonucleotides were eluted in the wash.

      As ionic strength is independent of ion identity or chemical nature of the ion involved (Jerermy M. Berg, John L. Tymoczko, Gregory J. Garret Jr., Biochemistry 2015), we reasoned that our Tris-HCl/NaCl/ imidazole buffer wash should have at comparable ionic strength to the Flores-Jasso protocol.

      Our total ionic contributions were: 500 mM Na+, 550 mM Cl-, 50 mM Tris and 300 mM imidazole. We recognise that Tris and imidazole are both partially ionized according the pH of the buffer (pH 8) and their respective pKa values, but even if only considering the sodium and chloride it should be comparable to the Flores-Jasso protocol.

      Secondly, after reverse HisTrap purification, AGO2 was run through size exclusion chromatography to remove any remaining impurities (shown Figure S2B).

      Thirdly, knowing that AGO2 has many positively charged surface patches and can bind nucleic acid nonspecifically (Nakanishi, 2022; O'Geen et al., 2018), we tested various blocking backgrounds to eliminate nonspecific binding effects in our EMSA ternary binding assays. We were able to address this issue by adding either non-homogenous RNA extract or homogenous polyuridine (pU) in our EMSA buffer during equilibration background experiments. This allowed us to eliminate non-specific binding of our target mRNAs, as shown previously in Supplementary Figure S6. We appreciate that the reviewer finds this technical detail important and have moved the panel C of figure S6 into the main results in Figure 2C, to highlight the novel conditions used and important controls needed to be performed. If miR-34a were non-specifically bound to the surface of AGO2 after washing, this blocking step would render any impact of surface-bound miR-34a negligible due to the excess of competing polyuridine (pU).

      Our EMSA results show that, using polyU, we can reduce non-specific interaction between AGO2 and RNAs that are present. And still, duplex release occurs despite the blocking step. It is therefore less likely that duplex release is caused by surface-bound miR-34a.

      Finally, the observation of distinct duplex release for certain targets, but not for others (e.g. MTA2, which bound tightly to miR-34a-AGO2 but did not exhibit duplex release; see Figure 2), argues against the possibility that the phenomenon was solely due to non-specifically bound RNA releasing from AGO2.

      In response to the reviewers statement "Since properly loaded miR-34a is never released from AGO2, it is impossible for the miR-34a loaded into AGO2 to form the binary complex (mRNA:miR-34a)" we would like to refer to the three papers, De et al. (2013) Jo MH et al. (2015), and Park JH et al. (2017), which have previously reported duplex release and collectively provide considerable evidence that miRNA can be unloaded from AGO in order to promote turnover and recycling of AGO. It is known that AGO recycling must occur, therefore there must be some mechanisms to enable release of miRNA from AGO2 to enable this. It is possible that AGO recycling proceeds via miRNA degradation (TDMD) in the cell, but in the absence of enzymes responsible for oligouridylation and degradation, the miRNA duplex may be released. As TDMD-competent mRNA targets have been observed to release the miRNA 3' tail from AGO2 (Sheu-Gruttadauria et al., 2019; Willkomm et al., 2022), there is a possible mechanistic similarity between the two processes, however, we do not have sufficient data to make any statement on this.

    2. 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, the authors use a combination of biochemical, biophysical, and computational approaches to investigate the structure-function relationship of miRNA binding sites. Interestingly, they find that AGO2 weakens tight RNA:RNA binding interactions, and strengthens weaker interactions.

      Given this antagonistic role, I wonder: shouldn't there be an 'average' final binding affinity? Furthermore, if I understand correctly, not many trends were observed to correlate binding affinity with repression, etc.

      Given the context of the study - whereby structure is being investigated as a contributing factor to the interaction between the miRNA and mRNA, I find it interesting that the authors chose to use MC-fold to predict the structures of the mRNA, rather than using an experimental approach to assess / validate the structures. Thirty-seven RNAs were assessed; I think even for a subset (the 12 that were focused on in the study), the secondary structure should be validated experimentally (e.g., by chemical probing experiments, which the research group has demonstrated expertise in over the last several years). The validation should follow the in silico folding approach used to narrow down the region of interest. It is necessary to know whether an energy barrier (associated with the mRNA unfolding) has to occur prior to miRNA binding; this could help explain some of the unexplained results in the study. Indeed, the authors mention that there are many variables that influence miRNA regulation.

      In the methods, T7 is italicized by accident in the T7 in vitro transcription section. Bacmid is sometimes written with a capital B and other times with a lower-cased b. The authors should be consistent. The concentration of TEV protease that was added (as opposed to the volume) should be described for reproducibility.

      In figure S2D, what is the second species in the gel on the right-hand side of the gel in the miR-34a:AGO lanes? The authors should mention this.

      Figure S3B and S3A are referenced out of order in the text. In regard to S3A, what are the anticipated or hypothesized alternative conformations for NOTCH1, DLL1, and MTA2? There are really interesting things going on in the gels, also for HNF4a and NOTCH2. Can the authors offer some explanation for why the free RNA bands don't seem to disappear, but rather migrate slowly? Is this a new species?

      In the EMSA for Perfect, why does the band intensity for the bound complex increase then decrease? How many replicates were run for this? This needs to be reconciled.

      The authors list that the RNA concentration was held constant at 10 nM; in EMSAs, the RNA concentration should be less than the binding affinity; what is the lowest concentration of protein used in the assays shown in S3A? Is this a serial dilution? It seems to me like the binding assays for MTA2, Perfect, and SRCseed might have too high of an RNA concentration. (Actually, now I see in the supplement the concentrations of proteins, and the RNA concentration is too high). Also, why is the intensity of bands for bound complex for SRCseed more intense than the free RNA?

      Why are the binding affinity error bars so large (e.g., for NOTCH2 with mir-34a) - 6 uM +/- 3 uM?

      It would be very helpful if the authors wrote in the Kds in Figure 2A in green and blue (in the extra space in the plots). This would help the reader to better understand what's going on, and for me, as a reviewer, to better consider the analysis/conclusions presented by the authors.

      The authors state on page 18 that 'Interestingly, however, we did not observe a correlation between binary or ternary complex affinity and seed type.' They should elaborate on why this is interesting.

      Figure 2C is not referenced in the text (the authors should go back through the text to make sure everything is referenced and in order). The Kds should be listed alongside the gels in Figure 2C.

      Figure 3B is rather confusing to understand.

      Page 20: Perfect should be italicized.

      Have the authors considered using NMR to assess the base pair pattern formed between the miRNA:mRNA complexes (with / without AGO)? As a validation for results obtained by RABS? This could be helpful for the Asymmetric target binding section, the Ago increases flexibility section, and the three distinct structural groups section in the results. It is widely accepted that while chemical probing is insightful, results should be validated using alternative approaches. Distinguishing structural changes and protected reactivity in the presence of protein is challenging.

      Significance

      The work is helpful for understanding how microRNAs recognize and bind their mRNA targets, and the impact Ago has on this interaction. I think for therapeutic studies, this will be helpful for structure-based design. Especially given the three types of structures identified to be a part of the interaction.

    3. 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:

      Sweetapple et al. took the approaches of EMSA, SHAPE, and MD simulations to investigate target recognition by miR-34a in the presence and absence of AGO2. Surprisingly, their EMSA showed that guide unloading occurred even with seed-unpaired targets. Although previous studies reported guide unloading, they used perfectly complementary guide and target sets. The authors of this study concluded that the base-pairing pattern of miR-34a with target RNAs, even without AGO2, can be applicable to understanding target recognition by miR-34a-bound AGO2.

      Major comments:

      1. (Page 11 and Figure S4) The authors pre-loaded miR-34a into AGO2 and subsequently equilibrated the RISC with a 5' modified Cy5 target mRNA. Since properly loaded miR-34a is never released from AGO2, it is impossible for the miR-34a loaded into AGO2 to form the binary complex (mRNA:miR-34a) in the EMSA (guide unloading has been a long-standing controversy). However, they observed bands of the binary complex in Figure S4. The authors did not use ion-exchange chromatography. AGOs are known to bind RNAs nonspecifically on their positively charged surface. Is it possible that most miR-34a was actually bound to the surface of AGO2 instead of being loaded into the central cleft? This could explain why they observed the bands of the binary complex in EMSA.
      2. (Page 18 and Figure S5) Previous studies (De et al., Jo MH et al., Park JH et al.) reported guide unloading when they incubated a RISC with a fully complementary target. However, neither MTA2, CCND1, CD44, nor NOTCH2 can be perfectly paired with miR-34a (Figure 1A). Therefore, the unloading reported in this study is quite different from the previously reported works and thus cannot be explained by the previously reported logic. The authors need to explain the guide unloading mechanism that they observed. Otherwise, they might misinterpret the results of their EMSA and RABS of the ternary complex.
      3. (Page 20) The authors reported, "it is notable that the seed region binding does not appear to be necessary for duplex release." The crystal structures of AGO2 visualize that the seed of the guide RNA is recognized, whereas the rest is not, except for the 3' end captured by the PAZ domain. How do the authors explain the discrepancy?
      4. (Pages 22) The authors mentioned, "It follows that the structure imparted via direct RNA:RNA interaction remains intact within AGO2, highlighting the role of RNA as the structural determinant." A free guide and a target can start their annealing from any nucleotide position. In contrast, a guide loaded into AGO needs to start annealing with targets through the seed region. Additionally, the Zamore group reported that the loaded guide RNA behaves quite differently from its free state (Wee et al., Cell 2012). How do the authors explain the discrepancy?
      5. The authors concluded that the range of binary complex affinities is constricted by human AGO2 in the ternary complex - strengthening weak binders while weakening strong binders. This may hold true for miR-34a, but it cannot be generalized. Other miRNAs need to be tested.

      Minor comments:

      1. (Figure S2) Why was the 34-nt 3'Cy3-labeled miR34a complementary probe shifted up in the presence of AGO? 2.(Page 17) Does the Cy3 affect the interaction of the 3' end of miR-34 with AGO2?
      2. Several groups reported that overproduced AGOs loaded endogenous small RNAs. The authors should mention that their purified AGO2 was not as pure as a RISC with miR-34a. Otherwise, readers might think that the authors used a specific RISC.
      3. (Figure legend of Figure S5) Binding was assessed "by."
      4. (Page 17) It would be great if the authors could even briefly describe the mechanism by which the sodium phosphate buffer with magnesium does not disturb weaker interactions by citing reference papers. 6.(Page 22) The authors cited Figure 4C in the sentence, "Comparison between CIS and TRNAS ..." Is this supposed to be Figure 4D? 7.(Figure 6) Readers would appreciate it if the guide and target were colored in red and blue. The color codes have been used in most papers reporting AGO structures. The current color codes are opposite.

      Significance

      This paper will have a significant impact on the field if seed-unpaired targets can indeed unload guide RNAs. The authors may want to validate their results very carefully.

    4. 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

      Sweetapple et al.

      Biophysics of microRNA-34a targeting and its influence on down-regulation

      In this study, the authors have investigated binding of miR-34a to a panel of natural target sequences using EMSA, luciferase reporter systems and structural probing. The authors compared binding within a binary and a ternary complex that included Ago2 and find that Ago2 affects affinity and strengthens weak binders and weakens strong binders. The affinity is, however, generally determined by binary RNA-RNA interactions also in the ternary complex. Luciferase reporter assays containing 12 different target sites that belong to one of three seed-match types were tested. Generally, affinity is a strong contributor to repression efficiency. Duplex release, a phenomenon observed for specific miRNA-target complementarities, seems to be more pronounced when high affinity within the binary complex is observed. Furthermore, the authors use RABS for structural probing either in a construct in CIS or binding by the individual miRNA in TRANS or in a complex with Ago2. They find pronounced asymmetric target binding and Ago2 does not generally change the binding pattern. The authors observe one specific structural group that was unexpected, which was mRNA binding with bulged miRNAs, which was expected sterically problematic based on the known structures. MD simulations, however, revealed that such structures could indeed form.

      This is an interesting manuscript that contributes to our mechanistic understanding of the miRNA-target pairing rules. The combination of affinity measurements, structural probing and luciferase reporters allow for a broad correlation of target binding and repression strength, which is a well-thought and highly conclusive approach. However, there are a number of shortcomings that are summarized below.

      1. The manuscript is not easy to read and to follow for several reasons. First, many of the sub-Figures are not referenced in the text of the results section (1C, 1D, 2C, 4D). Figure 4A seems to be mis-labeled. Second, a lot of data is presented in suppl. Figures. It should be considered to move more data into the main text in order to make it easier for readers to evaluate and follow.
      2. Some of the data is over-interpreted. For example, in Figure 3A, it is concluded that supplementary regions are more important for weaker seeds. Only two 8-mer seeds are present among the twelve target sites and thus it might be difficult to generalize.
      3. I did not understand why the CIS system shown in 4A is a good test case for miR-34a-target binding. It appears very unnatural and artificial. This needs to be rationalized better. Otherwise it remains questionable, whether these data are meaningful at all.
      4. For the TRANS experiments, only one specific scaffold structure is used. This structure might impact binding as well and thus at least one additional and independent scaffold should be selected for a generalized statement.
      5. Generally, it would be nice to have some more information about the experiments also in the result section. Recombinant Ago2 is expressed in insect cells and re-loaded with miR-34a, luciferase reporters are transfected into tissue culture cells, I guess.
      6. In the first sentence of the abstract, Argonaute 2 should be replaced by Argonaute only since other members bind to miRNAs as well.

      Significance

      This is an interesting manuscript that contributes to our mechanistic understanding of the miRNA-target pairing rules. The combination of affinity measurements, structural probing and luciferase reporters allow for a broad correlation of target binding and repression strength, which is a well-thought and highly conclusive approach. However, there are a number of shortcomings.

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

      Learn more at Review Commons


      Reply to the reviewers

      We would like to thank the reviewers for their thoughtful evaluation of our work. Our point-by-point responses to reviewer critiques follow below. Please note that any referenced changes to the manuscript are highlighted in yellow in the revised manuscript text.

      Response to Common Critiques

      1. Reviewers 1 and 2 state that some elements of this study confirm previously published results (many in murine systems). However, the reviewers also acknowledge that the mouse and human rDNA repeats may be subject to quite distinct regulation because of the much denser CG content of the human rDNA promoter (26 CpGs) vs. the mouse rDNA promoter (only 2 CpGs); these potential differences in regulation motivated this study in human cells. We evaluate the functions of rDNA methylation in human cells, which is directly relevant to understanding the regulation of rDNA function in human aging, and to understanding the functional implications of DNA methylation "aging clocks" more generally. We also apply a recently developed technology (dCas9-mediated epigenome editing) to directly test the function of rDNA methylation. Novel findings reported in this study include:
      2. Pol I - engaged rDNA repeats are hypomethylated at sites both in the promoter and the gene body; this contrasts with Pol II transcription, which is coincident with gene body methylation.
      3. rDNA copy number remains stable with age in mammals, in striking contrast to findings in other eukaryotes. rDNA copy number instability has been proposed to be a universal feature of the aging genome, and this finding refutes that possibility.
      4. Induction of DNA methylation by an average of ~20% along 7-11 of the 26 CpGs in the human rDNA repeat does not measurably inhibit rDNA transcription.
      5. Human Pol I and UBTF remain bound to rDNA promoters in the presence of elevated CpG methylation, in contrast to the murine Pol I machinery.

      Reviewers 1 and 2 questioned our strategy of mapping sequencing data to the consensus ribosomal DNA (rDNA) repeat alone. We followed the approach of Wang & Lemos Genome Research 2019, who initially described the rDNA methylation clock. Wang & Lemos also mapped genomic data to rDNA consensus sequences alone due to the computational efficiency of this approach, and describe a head-to-head comparison of mapping performance outcomes in their Methods section. Importantly, their analysis indicated that the vast majority (>98%) of sequencing reads can be mapped uniquely to the consensus human rDNA repeat (U13369.1). When we launched our study, we also initially compared the performance of mapping to the rDNA repeat consensus sequence alone versus to the whole human genome. We noted very similar performance in both cases, with the possible exception of a modest increase in simple repeat sequences being erroneously mapped to the intergenic spacer (IGS) region of the rDNA when we mapped to the rDNA repeat alone. As the reviewers pointed out, the IGS contains simple repeat sequences that are also found at numerous other non-rDNA sites in the genome. However, the minor mis-mapping of simple repeats to the IGS did not affect our analyses of non-IGS sequences, which were the focus of this study. We therefore proceeded with mapping to the rDNA consensus sequence only.

      Reviewers 1 and 2 pointed out that our dCas9-DNMT strategy induced only a 15-20% increase in rDNA methylation and questioned whether we could expect to detect downstream effects in rDNA transcription. While Reviewer 2 suggested that multiple sgRNAs could enhance methylation efficiency, it turns out that this has already been tested for other target genes and shown that multiple sgRNAs cannot increase efficiency of CpG methylation by dCas9-DNMTs (Stepper et al., Nucleic Acids Research 2017). Separately, the goal of this study was to model the effects of age-linked rDNA hypermethylation, which increases by 15-20% over mammalian lifespan (Wang & Lemos 2019; see also our Figure 1). Importantly for interpreting these data, induction of promoter methylation to a similar extent on the mouse rDNA repeat was able to direct detectable repression of rDNA transcription (Santoro et al., 2011). Further, dCas9-DNMT has been previously shown to induce a ~20% increase in CpG methylation of the Pol II target gene EpCAM and cause measurable transcriptional repression that was detectable by qPCR (Stepper et al., 2017). In contrast, we were able to induce rDNA methylation to a similar extent and observed no change in the levels of either pre-rRNA or mature rRNA. Because we see that UBF and Pol I remain bound to rDNA in spite of higher CpG methylation (Fig. 7 and Fig. S4), we interpret these data together to indicate that the human Pol I machinery can continue to engage with rDNA in the presence of intermediate levels of CpG methylation.

      Reviewer 1

      1. inactivation of rDNA transcription per se does not affect chromatin accessibility, to date only depletion or deletion of UBTF has been found to do this and even this does not enhance CpG methylation, these published findings should be referenced.

      Our analyses in Figure 2 focus on defining the relationships between chromatin accessibility, transcriptional activity, and CpG methylation throughout the human rDNA repeat. We cannot determine causation from this analysis - meaning whether chromatin accessibility influences CpG methylation or vice versa - and this point is beyond the scope of our study. Our major goal was to test whether induced CpG methylation affects transcription output.

      The authors overstate their results by writing "actively transcribed rDNA repeats are hypomethylated at their promoter" despite only one SmaI site but many CpG sites exist in the human promoter, the latter having not been assayed.

      We analyzed several pieces of data to come to this conclusion. First, ATAC-Me indicates that ATAC-accessible rDNA repeats are completely devoid of methylation both in their promoter and throughout the gene body; as UBTF binding controls rDNA accessibility (Sanij et al., JCB 2008; Hamdane et al., PLoS Genet 2014), we infer that ATAC-accessible repeats are engaged with the Pol I transcription machinery and hypomethylated. To more directly probe this question, we evaluated the methylation status of Pol I-bound rDNA repeats at five separate sites by ChIP-chop: two sites in the 5' regulatory region (5' ETS and core promoter, pooled together as "promoter" in Figure 2F) and three sites within the gene body (18S, 5.8S, and 28S, pooled together as "gene body" in Figure 2F). These data clearly indicate that Pol I preferentially binds to these regions when they are hypomethylated, as the extent of CpG methylation at these same sites is higher in input DNA and lower in Pol I-ChIPped DNA. While we do not comprehensively profile CpG methylation status of Pol I-bound DNA, these ChIP-chop analyses are consistent with our interpretation that "actively transcribed (that is, Pol I-engaged) rDNA repeats are hypomethylated at their promoter".

      Pol I's preference for binding hypomethylated promoters has been previously described in mouse cells (Santoro & Grummt 2001) and human cells (Brown & Szyf Mol Cell Biol 2007). We confirm this and also report the novel finding that rDNA gene bodies bound by Pol I are hypomethylated. This contrasts with known relationships between Pol II and CpG methylation, where genes actively transcribed by Pol II often have dense gene body CpG methylation.

      While we think it is reasonable to infer from ATAC-Me data and ChIP-chop data together that accessible and hypomethylated rDNA repeats reflect transcriptionally active repeats, we appreciate the reviewer's point that we analyzed only a select few CpG sites by Pol I ChIP-chop. We have adjusted the text to make our interpretation more parsimonious (see highlights).

      The human rDNA promoter contains many CpGs which may not affect transcription when methylated. RRBS and WGBS data can't tell us much if we don't understand which sites, when methylated, affect transcription*. *

      We agree, and this ambiguity is what motivated us to induce methylation and evaluate the consequences. In plasmid reporter experiments where the human rDNA promoter was fused to a luciferase reporter, it was shown that in vitro methylation of the plasmid potently inhibited transcription in human cells (Ghoshal et al., J Biol Chem 2004). In this study, methylation of 7/26 CpGs was sufficient to induce >75% inhibition of reporter plasmid transcription, while methylation at single sites could induce ~50% inhibition. We neglected to site this relevant study and have included a reference to it in the revised manuscript. Importantly, this plasmid reporter assay does not assess the effects of CpG methylation on the full rDNA repeat in its endogenous genomic context. We were able to induce significant CpG hypermethylation on 11/26 promoter CpGs with one guide (P+G) and on 7/26 CpGs with a second guide (P+A) (Figure 3D). This level of methylation did not induce detectable silencing of rRNA transcription. Instead, we found that both UBF (Fig. 7) and Pol I (Fig. S4) remained bound to rDNA in the presence of CpG hypermethylation.

      The argument that the mouse rDNA Pol I machinery is "exquisitely sensitive" to CpG methylation is a little misleading as there are only two CpGs in the mouse rDNA promoter. Which of the 26 human CpGs are the critical ones?

      Immediately following this statement in the Discussion, we state that "the human rDNA promoter is significantly more CG-rich than the mouse rDNA promoter". We have revised this section to emphasize the difference (26 CpGs in human vs. only 2 in the mouse) and discuss this point raised by the reviewer: which are the critical CpGs in the human rDNA? Here again it is relevant to cite the human rDNA promoter reporter assays performed by Ghoshal et al., J Biol Chem 2004. These data indicate that CpG methylation of 7/26 promoter CpGs interferes with transcription from an rDNA reporter plasmid. Notably, it is unclear how generalizable findings from reporter assays are to the genomic context of the endogenous full length rDNA sequence. Our data indicate that partial methylation of 7-11 CpGs in the human rDNA promoter causes no detectable rDNA inhibition, and indeed does not displace UBF or Pol I (Fig. 7; Fig. S4).

      Antibody SC13125 used for UBF ChIP sees nearly exclusively the shorter transcriptionally inactive UBF2 variant. These data need to be repeated with an antibody that detects both UBF forms.

      We thank the reviewer for raising the important issue of UBTF splice isoforms. Relevant citations demonstrating that the SC13125 antibody recognizes only UBF2 would have been very helpful. The human UBTF gene is alternatively spliced into full-length UBF1 (exon 8 retained) and UBF2 (exon 8 spliced out). The deletion of exon 8 results in a 37 amino acid deletion in UBF2 corresponding to residues 221-268 in HMG box 2 of UBF1 (see Ensembl entry ENSG00000108312.16). The truncation of HMG box 2 makes UBF2 a far less potent transcriptional activator than UBF1. Because of the small molecular weight difference between these two isoforms, preference of an antibody for one vs. another isoform is not readily apparent by Western blotting. However, according to the manufacturer of the UBTF antibody used in this study, the immunogen corresponds to residues 1-220 of UBTF1, which is immediately N-terminal to the residues deleted in UBF2 (AAs 221-268, encoded by exon 8). The antibody's immunogen is thus entirely sequence that is shared between UBF1 and UBF2. Further, a previous study performed immunoprecipitation followed by mass spectrometry using this antibody and reported detection of UBF1-specific peptides (Drakas et al., PNAS 2004). Therefore, absent our knowledge of any evidence to the contrary, we conclude that this antibody recognizes UBF1 and possibly also UBF2.

      We thank the reviewer for raising this point and have adjusted the text to avoid the misleading implication that we are unambiguously detecting only the UBF1 isoform; all mentions of "UBF1" in the revised text have been replaced with "UBTF".

      Setting aside the question about the UBTF antibody reagent used, we observe consistent results by evaluating both UBTF (Figure 7) and Pol I (Figure S4) binding to rDNA in spite of CpG methylation; therefore, we conclude that the human Pol I machinery is not displaced from the human rDNA promoter by intermediate levels of CpG methylation.

      Reviewer 2

      1. There is very little discussion concerning the methylation status of the IGS...the Kobayashi lab has convincingly demonstrated that rDNA repeats fall into 2 classes. Those in which the supposedly active repeats lack methylation on promoters and coding regions and those in which both promoters and coding regions are heavily methylated. In both cases the IGS is fully methylated.

      We cite this study in the Discussion (reference 18 in bibliography) and agree that this work is relevant to ours; we have adjusted the text to emphasize this point. Notably, this previous analysis of CpG methylation patterns by long-read sequencing implied that active repeats may be entirely hypomethylated along their coding sequence; our data more directly demonstrate this both by ATAC-Me and by Pol I ChIP-chop (Fig. 2).

      There is no description of how rRNA levels were assessed. I suggest this could be further complemented by in vivo incorporation studies such as EU labeling.

      We apologize for this lack of clarity. rRNA levels were assessed by qPCR of the 45S pre-rRNA (Fig. 3A) and of mature 28S rRNA (Fig. 3B), and these data are presented as a fold change in each rDNA-targeting sgRNA compared to a non-targeting control sgRNA. The primersets used are listed in Supplementary Table 1.

      While we agree that EU labeling could be useful for detecting nucleolar transcription, qPCR detection of the 45S rRNA also sensitively reports nascent transcription and we think is sufficient to address this question.

      Reviewer 3

      1. The study points to differences between mouse and human rDNA and the effect of DNA methylation on transcriptional output. Did the mouse rDNA dataset also measure transcription output to correlate with DNA methylation age differences?

      The original study that defined the rDNA methylation clock (Wang & Lemos Genome Research 2019) did not evaluate rDNA transcription in parallel. More generally, the relationship of age-linked "clock" CpG methylation sites to expression / function of CpG methylated loci is very unclear, and testing the potential relationship between age-linked rDNA methylation and function was the major goal of this study.

      Did the spacer promoter also get methylated and did that affect UBF and Pol I binding?

      While the existence and function of a spacer promoter has been more clearly defined in the mouse rDNA repeat, recent evidence indicates that the Pol I transcription machinery also binds a second location about 800 bp upstream of the core promoter in the human rDNA repeat (Mars et al G3 2018). The guides that we used to direct CpG methylation recognize single unique sites in the core rDNA promoter and do not recognize sequences in this putative spacer promoter, and we did not analyze methylation at the spacer promoter. Analysis of the spacer promoter is generally beyond the scope of this study, as it is unknown whether there is any relationship between spacer promoter methylation and aging progression.

    2. 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 manuscript Modeling the consequences of age-linked rDNA hypermethylation with dCas9-directed DNA methylation in human cells studies the DNA methylation during aging at the rDNA. The study is well performed and provides several new insights into rDNA transcriptional regulation. The main finding is that in human cells, rRNA methylation does not affect transcription output, UBF and RNA pol I binding, even though the bound gene copies are less methylated than the silent ones. The experimental approach is excellent; the data mining and experiments are appropriate and shows essential points. The results are very interesting and provides new aspects to the state of rDNA that will further the understanding of ribosomal transcription.

      Minor concerns

      The study points to differences between mouse and human rDNA, and the effect of DNA methylation on transcriptional output. Did the in the mouse rDNA data-set also measure transcription output to correlate with DNA methylation age-differences.

      Some rRNA genes, including the human gene repeat, has a second promoter 7-800 base pairs upstream of the promoter. This site also contains a CTCF binding site, upstream of which nucleosomal chromatin state is found. Downstream of the spacer promoter, a UBF associated chromatin state assembles, presumable on active copies. Did the spacer promoter also get methylated and did that affect UBF binding and pol I binding?

      Significance

      This is a very interesting and novel study which just needs to be extended to other feature of the rDNA to provide a full picture. The results presented in the manuscript are novel and contributes to the understanding of ribosomal transcription, in particular the outstanding question about the impact of DNA methylation on the transcriptional output and chromatin states. It provides important insight into how to think about rRNA transcription in different cell lines, states and diseases, such as cancer. The general aspects of the study suggest a broad broad.

    3. 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:

      Mammalian genomes typically contain between 150 and 250 copies of a ribosomal gene repeat (rDNA) that are transcribed by RNA polymerase I to yield pre-rRNAs that encode rRNAs. It is generally accepted, that in most cells as many as 50% of repeats are transcriptionally silent. It is now appreciated that the regulatory elements and transcribed regions of these "silent" repeats are heavily methylated. Thus rDNA hypermethylation correlates with silence. However, whether this is a driver of silencing or a consequence of silencing is open to debate. This manuscript weighs into this debate. Initial experiments remap existing bisulfite sequencing data from both the mouse and humans. These results confirm previous data that rDNA hypermethylation correlates with aging. Next, to strengthen links between hypermethylation and silencing, they remap methylation-resolved ATAC sequencing data. This confirms that hypomethylated rDNA is in a more open chromatin conformation, presumably the "active repeats". In mammals there have been competing claims regarding changes in rDNA copy number during aging. Notably it has been claimed previously that rDNA copy number drops during human aging. A potential flaw in that study is that it studies of rDNA copy number utilised genetically diverse human populations. Here, using digital PCR, they survey rDNA copy number in various tissues of an inbred mouse strain. Analysing young mice and old mice, they find no evidence for age related rDNA loss. While the above experiments are well performed the results are largely confirmatory in nature. The next set of experiments attempt to address a critical question, namely, is rDNA hyper methylation a 'cause' of a 'consequence' of silencing. They generated an inducible nuclease dead CAS9 fused with de novo methyltransferase function (dCas9-3A3L) and gRNAs targeting either the promoter of the 28S coding region. Experiments performed in transformed and non-transformed human cell lines demonstrated a 15-20% methylated rDNA. Analysis of pre and mature rRNAs as well as cell staining reveal that transcript levels and nucleolar morphology are unaltered. Furthermore, the finding that UBF 'chipped' rDNA is more heavily methylated argues that directed methylation of the human rDNA promoter does not displace UBF. These experiments suggest that rDNA hypermethylation is more of a consequence of silencing than a cause of silencing.

      Major comments:

      It is not clear from the methods how previous rDNA was mapped onto rDNA repeats. Did they generate a customised reference genome with rDNA added, or simply map reads to rDNA in isolation. This is of critical importance as only reads that uniquely map to rDNA should be considered. Mammalian genomes typically contain many rDNA pseudo genes. Furthermore, the rDNA intergenic spacer (IGS) contain many retro/repeated elements that are distributed throughout the genome.

      There is very little discussion concerning the methylation status of the IGS. Using nanopore sequencing the Kobayashi lab has convincingly demonstrated that rDNA repeats fall into 2 classes. Those in which the supposedly active repeats lack methylation on promoters and coding regions and those in which both promoters and coding regions are heavily methylated. In both cases the IGS is fully methylated.

      In the targeted methylation experiments the increase in rDNA methylation remains both local and modest 15-20% increase. Would it be possible to increase the number of gRNA so as to achieve a higher level and more distributed change in rDNA methylation.

      Minor comments:

      The older U13369 rDNA reference has many sequence errors and should be avoided.

      There is no description of how rRNA levels are assessed. I suggest this could be further complemented by in vivo incorporation studies such as EU/click-chemistry.

      Significance

      Around 50% of data presented in this manuscript (Figs 1-3) is confirmatory rather than novel. While the data regarding targeted methylation of "active rDNA repeats is interesting, and I think pointing us in the right direction, it is not comprehensive enough to overturn the pervasive notion that methylation causes silencing.

    4. 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 manuscript attempts to provide an answer to why methylation of the human rDNA correlates with aging. They conclude that this correlation is not connected with changes in rDNA activity of copy numbers.

      Major comments:

      The authors reanalyze public data from RRBS and WGBS that suggests a correlation between aging and rDNA methylation. They then use public ATAC-Me sequence data and show a good correlation between chromatin accessibility and lack of CpG methylation. This correlation has been known for some time, but the ATAC-Me approach is a nice confirmation that it extends through the coding region and probably the promoter and enhancer sequences. In referring to the correlation between open chromatin and hypomethylation the authors state that "these data imply that methylation of the rDNA promoter and gene body both occur exclusively on non-transcribed, silent repeats" However, it is known that inactivation of rDNA transcription per se does not affect chromatin accessibility, to date only depletion or deletion of UBTF has been found to do this and even this does not enhance CpG methylation, these published findings should be referenced. The authors do recognize this and use the so-called ChIP-chop (not ChIP-ChOP) method to analyze methylation of PolI ChIPped DNA at a single SmaI site in the 47S promoter and a site within the 28S (Table S1 showing primers was not available to me to define the exact regions, the ref to Santoro for the technique should be 2014 not 2013). The ChIP-chop assay repeats previous work but here is done on HEK293T, the cell line they use for later study. The authors also overstate their results by writing "actively transcribed rDNA repeats are hypomethylated at their promoter" despite only one SmaI site but many CpG sites exist in the human promoter, the latter having not been assayed.

      The authors do go on to convincingly show rDNA copy numbers are constant with age by assaying various mouse tissues from young and old mice, hence excluding this as an affector of aging. They then attempt to use targeted de novo methylation to ask if this has any effect on rDNA transcription. Such effects have been extensively claimed as a source of rDNA regulation, though there is little evidence that this occurs in vivo. The authors use dCas9 targeted DNMT to locally enhance methylation using two promoter and one 28S guide RNAs and are able to show mean increases of 15 to 20% by ChIP-chop (but 40 to 50% at other CpGs by WGB-seq (BSAS), not discussed). Measurement of pre-rRNA and 28S abundance (relative to what control is not stated), cell proliferation, PolI nucleolar distribution and UBF (incorrectly referred to UBF1, see comment below) occupancy at the promoter are all suggested to show no effects of this targeted methylation. Hence the authors conclude that "These data suggest that promoter methylation is not sufficient to impair transcription of the human rDNA and imply that the human rDNA transcription machinery may be resilient to age-linked rDNA hypermethylation" But in fact no more than a 20% change due to the targeted methylation should be expected in any of the parameters measured. It is not at all evident that such a small effect would be detected by the authors.

      Specific points:

      Mapping was to the rDNA repeat unit in the absence of the human genome. This may bias the mapping data since the human genome contains rDNA pseudogenes and intermediate repetitive elements that are also present in the rDNA unit. These will be present in all the RRBS and WGS datasets, may or may not change methylation levels with age and will be mapped onto the single copy of the rDNA used in the data alignment. These factors need to be controlled.

      The human rDNA promoters contain many 26 CpGs, most of which may have no effect on transcription when methylated. Thus, very little of significance can be gleaned from RRBS data and this goes for WBS data without understanding which sites when methylated affect transcription.

      The argument that the mouse rDNA is "mouse Pol I machinery is exquisitely sensitive to a single CpG methylation event in the UCE, which blocks UBF binding and prevents transcription". Here the reference is to one of only two CpGs in the mouse promoter and, in this reviewer's opinion, the effect of its methylation has never been convincingly shown in vivo on the endogenous genes. However, if true, it also opens the question of which of the 26 CpGs in the human promoter are critical ones.

      Antibody SC13125 used for UBF ChIP sees near exclusively the shorter transcriptionally inactive UBF2 variant. These data need to be repeated with an antibody that detects both UBF forms.

      Significance

      We believe that the authors are correct in their conclusion that rDNA activity is not significantly affected by the level of CpG methylation. This said, the data presented in the manuscript does not provide strong support for this notion and hence, does not significantly advance our understanding of the role of rRNA in aging.

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

      Learn more at Review Commons


      Reply to the reviewers

      We thank both reviewers for their reviews of our work and suggestions for improvement. Changes to the manuscript are captured with the Track Changes feature, and our point-by-point responses are included below in bold/italic text.


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

      Summary Bell et al. overexpress Prom1 or Ttyh1 and test its effect on EV formation from cell lines. They find that Ttyh1 expression leads to an increase in small EVs as well as tubulated EVs, while Prom1 expression leads to a milder increase in small EVs. EV induction by Prom1 is dependent on cholesterol and the authors show that Prom1 makes the cholesterol in EVs more resistant to detergent. The authors show no connection between Ttyh1 EV induction and cholesterol, although they claim it is important. They also show that a disease mutation in Prom1 decreases Prom1 trafficking to the plasma membrane and increases cholesterol resistance to detergent in EVs. The authors also find that the disease mutation decreases the size of the Prom1-induced EVs.

      Major Comments

      Results - line 99-106 - The EV isolation protocol would remove large EVs like the Prom1+ midbody remnants. It is important to explicitly specify that this study focused on small EVs.

      We agree with the reviewers and appreciate the suggestion to make this distinction. We have clarified the Results text (lines 104-105) to specify that our method specifically reconstitutes and isolates small EVs.

      Statistics - The t tests appear to have been performed without correction for multiple comparisons (Figure 2C-D, Fig. 4D). Given that >10 comparisons were made, this can alter the biological significance of p__We agree with the reviewers that multiple test correction is appropriate for these figures. We have applied Bonferroni correction to the t-tests in Figs 2C, 2D, and 4D by adjusting our significance thresholds (alpha), and included additional text in the figure legend to indicate how and why the correction was performed.__

      The DLS data does not appear to give any insight into EV size (unlike the EM data) and could be removed from the whole manuscript (or moved to supplemental). The authors should also remove any conclusions based on the DLS data.

      We appreciate the reviewers raising this point and agree that the DLS is less informative than our other measurements of EV size and morphology. We have moved all DLS figure panels where EV size is characterized by another method to the Supplement.

      Discussion - line 382-383 "Because Prom1 EVs arise directly from blebbing of the plasma membrane23, this finding suggests that Prom1 and Ttyh1 traffic to similar regions of the plasma membrane." The authors have not examined where Prom1 or Ttyh1 localize in the plasma membrane and can not draw this conclusion. That both proteins promote plasma membrane budding would only suggest that both proteins localize to the plasma membrane, not subregions of the plasma membrane. However, the authors have not demonstrated that Ttyh1 specifically induces plasma membrane budding. The different size of Ttyh1 EVs could be due to different biogenesis mechanisms (i.e. derived from intracellular organelles instead of the plasma membrane), making this statement an over-interpretation on both parts.

      This is a fair point. We have removed this sentence from the Discussion (lines 402-403) as the reviewer requests.

      Discussion - line 398-400 "Membrane cholesterol is necessary for Prom1-mediated remodeling20,21 and is present at similar levels in purified Prom1 and Ttyh1 EVs (Fig 5E), indicating that it is undoubtedly important for EV formation by both proteins." & line 415-417 "We find that conservative mutations in several of these adjacent aromatic residues impair EV formation by Prom1, but do not mimic the stable cholesterol binding of W795R (Figs 2C, 4D). " The author's data suggests that cholesterol is not important for Ttyh1 to induce EV formation. The authors show that cholesterol depletion does not alter Ttyh1 EV production. Similarly, they find separable effects on cholesterol binding and EV formation with Prom1 mutants, which suggest that there is more to Prom1-mediated EV formation than cholesterol. That cholesterol is present at similar levels can reflect that overexpression of these proteins does not alter the amount of cholesterol in the EV source membrane (i.e. plasma membrane). Also, wouldn't molecular crowding of a membrane protein be predicted to influence how easy it is to extract lipids?

      We thank the reviewer for highlighting this imprecisely phrased sentence. We only meant to indicate that cholesterol is present in both sets of EVs and contributes globally to membrane fluidity. We have removed this sentence from the Discussion (lines 419-421) to avoid over-interpretation or confusion.

      The reviewer is also correct to point out that molecular crowding could alter how extractable lipids are from EVs. We have included additional explanatory text in the Discussion (lines 421-426) addressing this point.

      Discussion - line 431-433 "Our findings suggest that the dynamic interaction of Prom1 with cholesterol may promote efficient maturation and trafficking of Prom1 between the endomembrane system and the plasma membrane. The authors did not investigate whether depleting cholesterol improved Prom1(W795R) trafficking to the plasma membrane, making this inference untested. Soften interpretation or test experimentally.

      We appreciate the reviewer raising this point. We have altered the text in this paragraph (lines449-459) to soften our interpretation of these results, as suggested by the reviewer.

      Minor Comments Abstract - "the EVs produced are biophysically similar" The authors don't perform any typical biophysical characterization (beyond size and perhaps density), so do they mean physically similar? Given the Prom1 and Ttyh1 EVs can have different shapes and are significantly different sizes, this statement feels misleading.

      We thank the reviewer for pointing out the ambiguity around this word. We agree that "physically similar" is a more precise and accurate term, and have revised all instances of this language in the manuscript.

      Intro - line 59-60 - "Large Prom1 EVs (500-700 nm in diameter) appear to form from bulk release of membrane from the cell midbody" Midbody remnants are well defined (if variously named, i.e. flemmingsome) large EVs derived from the spindle midbody, intercellular bridge, and cytokinetic ring. I'm not sure what the authors are trying to express by "bulk release of membrane". Midbody remnants are also a site of membrane tubulation.

      The reviewer is correct to point out that midbody remnant release is a well defined process. We originally included this statement to avoid indicating that we are studying the only known class of Prominin EVs, but now recognize that including this creates more confusion that it alleviates. To improve clarity concurrently with the changes referenced above emphasizing that we are specifically studying small EVs, we have removed this reference to the larger class of EVs from the introduction (lines 61-63).

      The effect on total numbers of EVs is buried in the y-axes of the EM graphs, making it difficult to distinguish where a higher n of images was examined vs. where there is an increase in EVs. This is especially hard to interpret given the high difference in n values.

      The reviewers raise a valid critique of these figure panels. To improve clarity, we have adjusted the y-axes to represent the fraction of EVs rather than the absolute value of EVs, and listed the n values in figure legends.

      Fig. 2C - Missing WT error bars

      We appreciate the reviewer's concern for the WT error bars in these figures. The measurements underlying these plots are derived from quantification of Western blots. Because the blots have a limited number of lanes, the WT sample was run as a normalization control on each of several sets of blots. By employing this approach, we could make quantitative comparisons within each blot without needing to make direct comparisons between blots, eliminating confounding variables such as blotting times, positions of blots on rotary shakers, developer incubation time, exposure times, etc. Because WT lanes were used for normalization, each "WT" blot condition has its own set of error bars that was used for t-test comparison with the samples that share a blot. For this purely technical reason, we can represent the data either normalized against WT values or with three separate WT measurements for each plot. In the interest of clarity and transparency, we elected to report the values normalized to WT and to include all raw blot images in Supplementary Fig. S4. We understand that we could have made this more transparent, so to clarify this decision for readers, we now explicitly reference the raw blot images in both the Results text (lines 185) and in the Figure 2 legend.

      Fig. 3H, 5C - Why not show raw numbers on the y-axes of the inset graphs like the main graph? Also, if it is only showing a subset of roundness ranges, then the x-axis should not go to 1 (i.e. axis range 0-0.8 would be clearer). I had a hard time figuring out what these insets were trying to show me, so please think about presenting this data more clearly (and larger).

      For clarity, we have moved the inset graphs to separate panels alongside the main panel and implemented the requested changes to the axes (see Figs. 3G, 5B).

      Discussion - line 377 - "Though we do not claim that Ttyh1 endogenously induces EV formation" This statement could be misinterpreted to say that you do not think endogenous Ttyh1 regulates EV formation. Rephrase as "although we have not examined whether..."

      We thank the reviewer for pointing out this unclear sentence and have applied the requested change (line 397).

      Discussion - line 400-402 "Our results do not indicate that Ttyh1 does not bind cholesterol, merely that it does not form an interaction that is sufficiently kinetically stable to be co-immunoprecipitated." The phrasing here is confusing with multiple "not". It is better to leave things open than to say what you have not shown. Rephrase suggestion: "Although Ttyh1 was not able to form a kinetically stable interaction for co-immunoprecipitation, it remains to be determined whether Ttyh1 is able to bind cholesterol."

      We thank the reviewer for their suggestion and have modified the sentence to avoid double-negative phrasing (lines 422-426).

      Movies - I'm not sure what the two videos add. It's difficult to convince myself that I see plasma membrane labeling in either movie, especially in comparison to the over-exposed WGA staining. Also, why are there ~5 sec of empty movie at the end of each?

      We appreciate the reviewer's feedback and have removed the movies from the manuscript.

      Reviewer #1 (Significance (Required)):

      The data is interesting and well presented, but over interpreted in the discussion. The data on Ttyh1 expression inducing EVs is novel, but limited to overexpression studies. This study will be of interest to the EV, membrane curvature, and Prmn1/Tthy1 fields My expertise is in basic research on membrane trafficking (including EV formation) and lipids

      We thank the reviewer for their favorable review and helpful suggestions.

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

      In this study, authors investigated the role of Prom1 and Ttyh1 proteins on EV formation. They showed that both proteins can induce EV formation, while the mechanisms by which they do it might differ slightly. Ttyh1 binding to cholesterol is not as pronounced as Prom1. Surprisingly, cholesterol binding efficiency inversely correlates with EV formation. Also, EVs induced by Tthy1 and Prom1 are structurally different.

      My suggestions to improve the manuscript are below.

      • Figure 2E is not very convincing. As the authors mentioned, the signal is too low to have a concrete conclusion. The line scans somehow show that WT is more membrane-localized than mutant, but colocalization of Prom1 and WGA seems very similar in both cases. Is it certain that the addition of fluorophore did not change the trafficking? Does endogenous Prom-1 staining look like this? Also, why is WGA staining brighter in mutant sample, just a usual variation or biologically important?

      We understand the reviewer's concern about low signal, but respectfully disagree that the signal is too low to draw a meaningful conclusion. The only point we conclusively make in Fig. 2E is that WT Prom1 is more efficiently trafficked to the plasma membrane than W795R Prom1. We feel that this effect is sufficiently well evidenced by the line scan analysis in Supp. Fig. S5, where Prom1 peaks are cleanly visible for WT but not for W795R protein.

      We observe somewhat variable WGA staining in our experiments, and the differences we show in this figure panel are representative of typical staining variation. We do not draw any biological conclusions from the level of WGA present, only from its localization. Because both the plasma membrane and late endosomes are WGA+, we suspect that the W795R Prom1 is failing to traffic from endosomes to the plasma membrane. However, given the limitations of our fluorescence assay, we have removed any claim beyond the change plasma membrane trafficking efficiency from discussion of this experiment.

      We cannot conclude whether the mStayGold fluorophore alters trafficking of Prom1 to the plasma membrane. In response to the reviewer's comment, we attempted to use immunofluorescence to measure membrane localization of untagged Prom1 with the AC133-1 antibody. Unfortunately, we were unable to optimize this protocol to achieve sufficient membrane staining for quantification. We have softened our interpretation of Fig. 2E in the Results and Discussion (lines 203-204, 450) to acknowledge that the effects we observe are only measured with fluorophore-tagged Prom1.

      • I also recommend showing the localization of Ttyh1 on cells.

      We appreciate the reviewer's suggestion here, and it is an experiment we considered. One of the challenges we faced in this assay was quantitatively measuring fluorescent signal along cell-boundary plasma membranes without saturating signal from the very bright WGA+ endosomes. Because Ttyh1 globally expresses at higher levels than Prom1 (see Figs. 3C, 3I), direct comparison of membrane-localized Prom1 and Ttyh1 is technically challenging in these cells. However, Ttyh membrane localization has been widely reported in other papers (Matthews et al., J. Neurochem, 2007; Jung et al., J. Neurosci., 2017; Sukalskaia et al., Nat. Commun., 2021; Melvin et al., Comm. Biol., 2022) that we now explicitly mention and cite for reader clarity in both the Introduction and Results (lines 69-71, 224-225).

      • A graph directly showing cholesterol binding vs EV formation efficiency would be very useful.

      We agree with the reviewer that this would be an interesting and useful addition to the paper. We now include this panel in the revised manuscript as Fig. 4F.

      • "Prominin and Tweety homology proteins are homologous and functionally analogous" involves speculation and authors should clearly mention this. Revealing that they are both contributing to EV formation does not make them definitely functionally analogous.

      We agree with the reviewer that this sentence is indeed ambiguous and somewhat speculative. We have revised the section heading to "Prominin and Tweety homology proteins are homologous proteins that both promote EV formation" (lines 461-462) to indicate the specific analogous function we observe.

      Reviewer #2 (Significance (Required)):

      Overall, it is a useful addition to the field of cell biology, particularly EV field. EV formation and efficiency are both important topics, and this manuscript might give insights.

      We thank the reviewer for their favorable review and helpful suggestions.

    2. 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, authors investigated the role of Prom1 and Ttyh1 proteins on EV formation. They showed that both proteins can induce EV formation, while the mechanisms by which they do it might differ slightly. Ttyh1 binding to cholesterol is not as pronounced as Prom1. Surprisingly, cholesterol binding efficiency inversely correlates with EV formation. Also, EVs induced by Tthy1 and Prom1 are structurally different.

      My suggestions to improve the manuscript are below.

      • Figure 2E is not very convincing. As the authors mentioned, the signal is too low to have a concrete conclusion. The line scans somehow show that WT is more membrane-localized than mutant, but colocalization of Prom1 and WGA seems very similar in both cases. Is it certain that the addition of fluorophore did not change the trafficking? Does endogenous Prom-1 staining look like this? Also, why is WGA staining brighter in mutant sample, just a usual variation or biologically important?
      • I also recommend showing the localization of Ttyh1 on cells.
      • A graph directly showing cholesterol binding vs EV formation efficiency would be very useful.
      • "Prominin and Tweety homology proteins are homologous and functionally analogous" involves speculation and authors should clearly mention this. Revealing that they are both contributing to EV formation does not make them definitely functionally analogous.

      Significance

      Overall, it is a useful addition to the field of cell biology, particularly EV field. EV formation and efficiency are both important topics, and this manuscript might give insights.

    3. 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

      Bell et al. overexpress Prom1 or Ttyh1 and test its effect on EV formation from cell lines. They find that Ttyh1 expression leads to an increase in small EVs as well as tubulated EVs, while Prom1 expression leads to a milder increase in small EVs. EV induction by Prom1 is dependent on cholesterol and the authors show that Prom1 makes the cholesterol in EVs more resistant to detergent. The authors show no connection between Ttyh1 EV induction and cholesterol, although they claim it is important. They also show that a disease mutation in Prom1 decreases Prom1 trafficking to the plasma membrane and increases cholesterol resistance to detergent in EVs. The authors also find that the disease mutation decreases the size of the Prom1-induced EVs.

      Major Comments

      Results - line 99-106 - The EV isolation protocol would remove large EVs like the Prom1+ midbody remnants. It is important to explicitly specify that this study focused on small EVs.

      Statistics - The t tests appear to have been performed without correction for multiple comparisons (Figure 2C-D, Fig. 4D). Given that >10 comparisons were made, this can alter the biological significance of p<0.05 (1 incorrect in 20 comparisons). Please reanalyze with a more appropriate statistical test for multiple comparisons (i.e. ANOVA) or apply a correction to the t test values (i.e. Bonferroni).

      The DLS data does not appear to give any insight into EV size (unlike the EM data) and could be removed from the whole manuscript (or moved to supplemental). The authors should also remove any conclusions based on the DLS data.

      Discussion - line 382-383 "Because Prom1 EVs arise directly from blebbing of the plasma membrane23, this finding suggests that Prom1 and Ttyh1 traffic to similar regions of the plasma membrane." The authors have not examined where Prom1 or Ttyh1 localize in the plasma membrane and can not draw this conclusion. That both proteins promote plasma membrane budding would only suggest that both proteins localize to the plasma membrane, not subregions of the plasma membrane. However, the authors have not demonstrated that Ttyh1 specifically induces plasma membrane budding. The different size of Ttyh1 EVs could be due to different biogenesis mechanisms (i.e. derived from intracellular organelles instead of the plasma membrane), making this statement an over-interpretation on both parts.

      Discussion - line 398-400 "Membrane cholesterol is necessary for Prom1-mediated remodeling20,21 and is present at similar levels in purified Prom1 and Ttyh1 EVs (Fig 5E), indicating that it is undoubtedly important for EV formation by both proteins." & line 415-417 "We find that conservative mutations in several of these adjacent aromatic residues impair EV formation by Prom1, but do not mimic the stable cholesterol binding of W795R (Figs 2C, 4D). " The author's data suggests that cholesterol is not important for Ttyh1 to induce EV formation. The authors show that cholesterol depletion does not alter Ttyh1 EV production. Similarly, they find separable effects on cholesterol binding and EV formation with Prom1 mutants, which suggest that there is more to Prom1-mediated EV formation than cholesterol. That cholesterol is present at similar levels can reflect that overexpression of these proteins does not alter the amount of cholesterol in the EV source membrane (i.e. plasma membrane). Also, wouldn't molecular crowding of a membrane protein be predicted to influence how easy it is to extract lipids?

      Discussion - line 431-433 "Our findings suggest that the dynamic interaction of Prom1 with cholesterol may promote efficient maturation and trafficking of Prom1 between the endomembrane system and the plasma membrane. The authors did not investigate whether depleting cholesterol improved Prom1(W795R) trafficking to the plasma membrane, making this inference untested. Soften interpretation or test experimentally.

      Minor Comments

      Abstract - "the EVs produced are biophysically similar" The authors don't perform any typical biophysical characterization (beyond size and perhaps density), so do they mean physically similar? Given the Prom1 and Ttyh1 EVs can have different shapes and are significantly different sizes, this statement feels misleading.

      Intro - line 59-60 - "Large Prom1 EVs (500-700 nm in diameter) appear to form from bulk release of membrane from the cell midbody" Midbody remnants are well defined (if variously named, i.e. flemmingsome) large EVs derived from the spindle midbody, intercellular bridge, and cytokinetic ring. I'm not sure what the authors are trying to express by "bulk release of membrane". Midbody remnants are also a site of membrane tubulation.

      The effect on total numbers of EVs is buried in the y-axes of the EM graphs, making it difficult to distinguish where a higher n of images was examined vs. where there is an increase in EVs. This is especially hard to interpret given the high difference in n values.

      Fig. 2C - Missing WT error bars

      Fig. 3H, 5C - Why not show raw numbers on the y-axes of the inset graphs like the main graph? Also, if it is only showing a subset of roundness ranges, then the x-axis should not go to 1 (i.e. axis range 0-0.8 would be clearer). I had a hard time figuring out what these insets were trying to show me, so please think about presenting this data more clearly (and larger).

      Discussion - line 377 - "Though we do not claim that Ttyh1 endogenously induces EV formation" This statement could be misinterpreted to say that you do not think endogenous Ttyh1 regulates EV formation. Rephrase as "although we have not examined whether..."

      Discussion - line 400-402 "Our results do not indicate that Ttyh1 does not bind cholesterol, merely that it does not form an interaction that is sufficiently kinetically stable to be co-immunoprecipitated." The phrasing here is confusing with multiple "not". It is better to leave things open than to say what you have not shown. Rephrase suggestion: "Although Ttyh1 was not able to form a kinetically stable interaction for co-immunoprecipitation, it remains to be determined whether Ttyh1 is able to bind cholesterol."

      Movies - I'm not sure what the two videos add. It's difficult to convince myself that I see plasma membrane labeling in either movie, especially in comparison to the over-exposed WGA staining. Also, why are there ~5 sec of empty movie at the end of each?

      Significance

      The data is interesting and well presented, but over interpreted in the discussion. The data on Ttyh1 expression inducing EVs is novel, but limited to overexpression studies. This study will be of interest to the EV, membrane curvature, and Prmn1/Tthy1 fields My expertise is in basic research on membrane trafficking (including EV formation) and lipids

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

      Learn more at Review Commons


      Reply to the reviewers

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

      The authors report a mass spectrometry (MS)-based interactomics technique, time-resolved interactome profiling (TRIP), which allows for tracking temporal changes in the interactome of protein of interest. To show that TRIP can successfully deconvolute interactomes over time, they pulsed thyroid cells with homopropargylglycine (Hpg), immunoprecipitated the Hpg incorporated thyroglobulin (Tg) and its interacting proteins at different time points, and subjected the samples to tandem mass tag (TMT)-based quantitative MS analysis. The MS results show that WT and variant Tg proteins indeed associate with different proteostasis network factors in a differential manner over the course of time. In addition, they utilized an siRNA-based luciferase fusion assay to evaluate whether silencing each proteostasis network component changes the levels of Tg in both lysate and media. From the combination of the TRIP and siRNA-based assays, they found many hits, including hits implicated in protein degradation, VCP and TEX264, which they validated with multiple experiments.

      I am overall quite positive and think this is an important study. But there are some meaningful points to consider.

      Our Response: We thank Reviewer #1 for their positive outlook on our manuscript and their constructive feedback. We have addressed the comments below.

      Significant comments:

      Reviewer #1, Comment #1: Oonly two replicates of the main data (the TRIP-MS experiments) for this paper is problematic. Especially since the manuscript is supposed to be demonstrating and validating the new technique. Consistent with this concern, the relative enrichment profiles for some of the results were surprising. For instance, interaction with CCDC47 was tapering off but then at 3 h it suddenly reaches the maximum level of engagement. Is this a real finding or the variability in the method? Impossible to tell with two replicates. Presenting heat maps based on biological duplicates is also very problematic. It masks the error, which is large as can be seen in some of the panels showing individual proteins. In my view, triplicates and a clear understanding of the error in the technique should be required.

      Our Response: The TRIP datasets for WT Tg contains 5 biological replicates, while the A2234D and C1264R Tg contains 6 biological replicates. Two replicates are typically included in a TMTpro 16plex mass spectrometry run, and each analysis consists of 3 MS runs. We apologize that the number of replicates and layout of the MS runs was not clearly explained. Data for individual replicates is found in Dataset EV1, Dataset EV3, and a newly added Table EV3 delineates the sample layout across the TMT channels and MS runs. We clarified the text as follows:

      "Subsequently, two sets of TRIP time course samples (0, 0.5, 1, 1.5, 2, and 3 hr) could be pooled using the 16plex TMTpro and analyzed by LC-MS/MS (Fig 2A). In total, 5 biological replicates were analyzed for WT and 6 biological replicates were analyzed for A2234D and C1264R, respectively (Table EV3)."

      Reviewer #1, Comment #2: The same concern arises for the high-throughput siRNA screen, which was performed only in duplicate for WT and A2234D.

      Our Response: While the initial screen was performed in duplicate for WT and A2234D, which is common for larger screens due to resource constraints, we would like to direct the reviewer to the fact that we followed up on observed hits using thyroid cell lines with many more replicates. Furthermore, most hits came from the C1264R Tg variant, which had three replicates in the initial screen. Hits were also extensively followed-up.

      Reviewer #1, Comment #3: *There are issues with some of the immunoprecipitation experiments: In Figure 1C, a negative control for FLAG IP is missing. *

      *-In Figure 2B, I am curious why the band (Hpg -, chase time 0 h) is so faint for the first WB (IB for FLAG) - is Hpg treatment indeed leading to much more Tg present at 0 h? If so, that is a concern. *

      -Also, a negative control must be included (either plain cells or cells expressing fluorescent protein or a different epitope-tagged WT Tg).

      -In this same figure, I am puzzled why the bands for 1.5-3 timepoints in Biotin PD elution, probed for Rhodamine, are very faint especially considering that in Figure 1D, the corresponding bands, which are 4 h after the pulse, look fine. It seems like the IP failed here?

      Our Response: In Fig 2B, we have updated this figure with higher-quality images that are more representative of the results found when performing this experiment. Furthermore, to address the missing negative controls in Fig. 1C, we have added a separate figure (Fig EV2) where (-) FLAG-tagged Tg is included in this panel. We updated the text as follows:

      "Furthermore, the C-terminal FLAG-tag and Hpg labeling are necessary for this two-stage enrichment strategy, and DSP crosslinking is necessary to capture these interactions after stringent wash steps (Fig 1D, Fig EV2)."

      Regarding the Biotin PD rhodamine/TAMRA signal in Fig 2B: The blots in this figure panel represent the time-resolved Tg fractions from cell lysate, corresponding only to intracellular thyroglobulin. The decrease in band intensity for 1.5-3 hr time points is expected due to continued secretion and/or degradation dynamics taking place that decrease the intracellular population of labeled thyroglobulin that is able to be captured. For comparison, please note the C1264R panel (Fig 2C), where the rhodamine/TAMRA signal in the Biotin PD elutions is more stable compared to WT, indicating the cellular retention of C1264R while WT Tg is efficiently secreted and the signal is lost more rapidly. Fig 1D contains samples derived from a 4 hr Hpg pulse (without chase), explaining why the overall fluorescent Tg signal is more intense.

      Suggestion to consider:

      Reviewer #1, Comment #4: This manuscript, supported by the title and abstract, mainly focuses on the presentation of the development and application of TRIP, which is highly significant. The story becomes less coherent and harder to follow as significant amounts of text/figures are dedicated to siRNA-based high throughput screening and follow-up. In addition, although the discovery of TEX264 as one of the hits is very interesting and exciting, TEX264 apparently was not a hit in the TRIP experiment and is pretty distracting from the main point of the paper highlighted in the abstract and title, therefore. The siRNA-based assay and follow-up studies could be a separate scientific story of their own. Especially considering my concerns on the number of replicates for both the TRIP and siRNA-based assay, it could be beneficial to actually split the manuscript into two and conduct more replicates of the -omic work, which should corroborate the exciting discoveries the authors have made.

      Our Response: We have edited the manuscript to hopefully provide a more cohesive presentation of all data, findings, and conclusions within the paper. Given the generally positive outlook on the manuscript from other reviewers and our responses to significant comments from Reviewer #1 we opted to keep the manuscript as a single piece and address all reviewer comments.

      Minor comments:

      Reviewer #1, Comment #5: Throughout the manuscript, the authors have not defined what FT is; presumably it means FLAG tag.

      Our Response: Reviewer #1 is correct in FT corresponding to FLAG tag. We have now edited the manuscript text to clarify this as follows:

      "Thyroglobulin was chosen as model secretory client protein, and we generated isogenic Fischer rat thyroid cells (FRT) cells that stably expressed FLAG-tagged Tg (Tg-FT), including WT or mutant variants (A2234D and C1264R)."

      Reviewer #1, Comment #6: The authors might discuss their rationale for choosing 0-3 hrs for their TRIP studies. That includes any relevant information about the half-life of WT versus variant Tg, whether the Hpg pulse time is short enough to avoid missing key features of the temporal interactome, and discussion of what would happen if the TRIP were performed at prolonged time points (e.g. 6-10 h).

      Our Response: Apologies that we omitted this important point, which is indeed related to the secretion and degradation half-life. We edited the manuscript text to discuss the rationale for 0-3 hr, length of the Hpg pulse and the impact on capturing interactions, and performing TRIP at prolonged time points as follows:

      "Our previous study indicated that ~70% of WT Tg-FT was secreted after 4 hours, while approximately 50% of A2234D and 15% of C1264R was degraded after the same time period (Wright et al, 2021). Therefore, we reasoned that a 3-hr chase period would be a enought time to capture the majority of Tg interactions throughout processing, secretion, cellular retention, and degradation, while still being able to capture an appreciable amount of sample for analysis."

      We explain the labeling timeline and limitations further in the discussion:

      "To address this, we utilized a labeling time of 1 hr which allows us to generate a large enough labeled population of Tg-FT for TRIP analysis, but some early interactions are likely missed within the TRIP workflow. In the case of mutant Tg, performing the TRIP analysis for much longer chase periods (6-8 hrs) may provide insightful details to the iterative binding process of PN components that is thought to facilitate protein retention within the secretory pathway."

      Reviewer #1, Comment #7: Lines 68-69: the two citations should probably come one sentence earlier (at least Coscia et al 2020 is a structure paper).

      Our Response: We agree. We have edited the manuscript as follows to correct this:

      "In earlier work, we mapped the interactome of the secreted thyroid prohormone thyroglobulin (Tg) comparing the WT protein to secretion-defective mutations implicated in congenital hypothyroidism (CH) (Wright et al, 2021). Tg is a heavily post-translationally modified, 330 kDa prohormone that is necessary to produce triiodothyronine (T3) and thyroxine (T4) thyroid specific hormones (Citterio et al, 2019; Coscia et al, 2020). Tg biogenesis relies extensively on distinct interactions with the PN to facilitate folding and eventual secretion."

      Reviewer #1, Comment #8: Line 91: "(Figure 1A)" should follow the sentence "To develop the time-resolved..." to help readers better understand the system.

      Our Response: __We agree. We have edited the manuscript to add the Fig 1A reference. Furthermore, we redesigned the schematic in Fig 1A to better explain the experimental system. (see also __Reviewer #2, comment 10)

      "To develop the time-resolved interactome profiling method, we envisioned a two-stage enrichment strategy utilizing epitope-tagged immunoprecipitation coupled with pulsed biorthogonal unnatural amino acid labeling and functionalization (Fig 1A). Cells can be pulse labeled with homopropargylglycine (Hpg) to synchronize newly synthesized populations of protein. After pulsed labeling with Hpg, samples can then be collected across time points throughout a chase period (Fig 1A, Box 1) (Kiick et al, 2001; Beatty et al, 2006). The Hpg alkyne incorporated into the newly synthesized population of protein can be conjugated to biotin using copper-catalyzed alkyne-azide cycloaddition (CuAAC) (Fig 1A, Box 2). Subsequently, the first stage of the enrichment strategy can take place where the client protein of interest is globally captured and enriched using epitope-tagged immunoprecipitation, followed by elution (Fig 1A, Box 3)."

      Reviewer #1, Comment #9: Line 101: Fisher should be Fischer

      Our Response: Thank you. We have edited the manuscript text to correct this.

      Reviewer #1, Comment #10: Line 131: Should be 1.5 hrs instead of 2 hrs.

      Our Response: We edited this point (see below in comment #11)

      Reviewer #1, Comment #11: Lines 135-136: I do not agree with the claim that HSPA5 profile looked similar for MS and WB. I do not see a peak for HSPA5 at 2 hrs in Figure 2D.

      Our Response: We replaced the mass spectrometry quantification in Fig 2D, E with the scaled, relative enrichments. This provides a more meaningful comparison, as all interactions are scaled in the same way. Unfortunately, it is still difficult to directly compare the Western blot results in Fig. 2B-C to the mass spectrometry quantifications in Fig 2D-E because the WB intensities are not normalized to the Tg bait protein amounts, which is changing over time. At 2-3hrs time points, little WT Tg is pulled down as most of it is secreted. Therefore, the HSPA5 interactions are no longer detectable by Western blot. On the other hand, MS is much more sensitive to capture the interactions. We modified the text as follows:

      "For C1264R, interactions with HSPA5 were highly abundant at the 0 hr time point and remained mostly steady throughout the first 1.5 hours (Fig 2C). A similar temporal profile was also observed for HSP90B1. Additionally, interactions with PDIA4 were detectable for C1264R and were found to gradually increase throughout the first 1.5 hr of the chase period, before rapidly declining (Fig 2C). We noticed similar temporal profiles for PDIA4 and HSPA5 to our western blot analysis, when measured via TMTpro LC-MS/MS as further outlined below (Fig 2D-E). In particular, the HSPA5 WT Tg interaction declined within the first hours, yet for C1264R Tg, the HSPA5 interactions remained mostly steady over the 3-hour chase period. (Fig 2E)."

      Reviewer #1, Comment #12: Line 186: The cited paper Shurtleff et al 2018 is missing in the reference list.

      Our Response: Thank you. We have corrected this in the citation management system and it is now available in the reference list.

      Reviewer #1, Comment #13: Line 188: I disagree with the authors' claim here because, at least for CCDC47, interactions with C1264R seem to come back at the 3 hr time point.

      Our Response: We have removed the discussion of EMC and PAT complex components from the text. The implications of these interactions for Tg biogenesis remain unclear and were therefore a distraction from the discussion of other core proteostasis network components pertinent to Tg processing. Nonetheless, the full dataset - including these interactions - remains available to readers in Appendix Fig S1 for further perusal.

      Reviewer #1, Comment #14: Line 203: I am not sure if P4HA1 can be included in the examples for showing distinct patterns for mutants compared to the WT according to their data in Figure 3H.

      Our Response: We agree. We have edited the text to remove the discussion of prolyl hydroxylation and isomerization family members and elected to discuss the new clustering analysis and the robustness of the TRIP method in more detail. The full TRIP data is nonetheless available to interested readers in Appendix Fig S1.

      Reviewer #1, Comment #15: Line 216: The authors should add citations about the functions of STT3A and STT3B proteins.

      Our Response: We've edited the manuscript text to include a reference to the primary literature for STT3A and STT3B functions, as follows:

      "Previously, we showed that A2234D and C1264R differ in interactions with N glycosylation components, particularly the oligosaccharyltransferase (OST) complex. Efficient A2234D degradation required both STT3A and STT3B isoforms of the OST, which mediate co-translational or post-translational N-glycosylation, respectively (Kelleher et al, 2003; Cherepanova & Gilmore, 2016)."

      Reviewer #1, Comment #16: Lines 248-251, "We found that interactions with these components...": this sentence should refer to Figure 3 - Figure Supplement 3 instead of Figure 3L and S4.

      Our Response: Thank you. This section of the manuscript was significantly rewritten and the figure references updated.

      Reviewer #1, Comment #17: Lines 258-260, "Another striking observation was that the temporal profile of EMC interactions for C1264R correlated with RTN3, PGRMC1, CTSB, and CTSD interactions.": Please provide more evidence to support the potential correlation between different interaction profiles. Or the authors should move this sentence to the discussion section as it sounds speculative. This highlights the issue of only having duplicates, as well.

      Our Response: We agree that this point was highly speculative and we removed discussion of the EMC interactions.

      To further investigate the correlation of interaction profiles across the dataset, we performed unbiased k-means clustering. This led to the identification of 7 and 6 unique clusters of interactors for WT and C1264R Tg-FT, respectively. These data are represented in Fig 3F and Fig EV5. Unique clusters highlight similar temporal interaction profiles for Tg-FT interactors, and provide a quantitative representation of correlative interactions that take place during Tg-FT processing.

      "To assess temporal interaction changes in an unbiased fashion and identify protein groups exhibiting comparative behavior, we carried out k-means clustering of the temporal profiles for WT and C1264R. This analysis revealed a large divergence in the interaction profiles. For WT Tg, only one cluster exhibited steadily decreasing interactions (cluster 4), while others increased with time, or showed peaks at intermediate times (Fig 3F, Fig EV5A). On the other hand, C1264R largely exhibited clusters with decreasing interactions over time (Fig 3F, Fig EV5B). Cluster 2 for WT with biomodal interactions at early and late time points contains many Hsp70/90 chaperoning components. For C1264R Tg, many Hsp70/90 chaperoning components and disulfide/redox-processing components are instead part of cluster 2', which exhibited an initial rise in interactions strength before plateauing (Fig 3F, Fig EV5A,B). This divergent temporal engagement between WT Tg and the destabilized C1264R mutant is aligned with the patterns observed in the manual grouping (Fig 3B,C), highlighting that the unbiased temporal clustering can reveal broader patterns in the reorganization of the proteostasis dynamics."

      One of the clusters of the C1264R Tg interactions contained autophagy interactors along with glycosylation components. We therefore postulate that this could point to a coordination of these processes. We discuss this new point in the updated manuscript:

      "In the k-means clustered profiles, autophagy interactions largely group together in the same cluster, showing stronger interactions at earlier time points. In the same cluster are glycosylation components (UGGT1 and STT3B, MLEC), further supporting a possible coordination for C1264R Tg between lectin-dependent protein quality control and targeting to autophagy (Fig EV5B,C)."

      Reviewer #1, Comment #18: Line 340: As written, should cite more than one paper

      Our Response: Thank you. We reworded the manuscript to correct this, as follows:

      "The discovery of several protein degradation components as hits for rescuing mutant Tg secretion may suggest that the blockage of degradation pathways can broadly rescue the secretion of A2234D and C1264R mutant Tg, a phenomenon similarly found for destabilized CFTR implicated in the protein folding disease cystic fibrosis (Vij et al, 2006; Pankow et al, 2015; McDonald et al, 2022)."

      Reviewer #1, Comment #19: Line 371: Should be Figure 4 - figure supplement 2

      Our Response: We edited the manuscript to correct this error.

      Reviewer #1, Comment #20: Line 1231: "Zhang et al 2018" needs to be removed

      Our Response: We have removed this citation.

      Reviewer #1, Comment #21: Line 1286: FRTR should be FRT

      Our Response: Thank you. We have corrected this within the text.

      Reviewer #1, Comment #22: Figure 3E: Color used to highlight the three proteins (CCDC47, EMC1, EMC4) should match the color used in Figure 3 - Figure Supplement 3

      Our Response: __We have edited Figure 3 to remove the section related to membrane protein biogenesis. This data is still available in __Appendix Fig S1 with consistent color coding.

      Reviewer #1, Comment #23: Figure 4A: The bottom figure where lysate signal is inversely proportional to time is misleading because the authors are assessing steady-state level of proteins in this assay.

      __Our Response: __We agree. We updated the schematic in __Fig 4A __to better explain the workflow and differentiate the steady-state protein level being measured within the lysate.

      Reviewer #1, Comment #24: Figure 4 - Figure Supplement 1 caption: in (C), (F) should be (B). (K) should be (G) and I am not sure what the authors mean when they refer to (J) in caption of (G).

      Our Response: We have corrected this lettering mistake to match the figure properly. Please note that this figure is now Fig EV6, and it includes some new and reorganized panels.

      Reviewer #1, Comment #25: Figure 5 caption for (C and D): Need to specify the time that the samples were collected (8 hrs), as it seems different from A and B according to the main text.

      Our Response: We have specified the collection time within the caption for these data in Fig 5C __and __5D.

      Reviewer #1, Comment #26: Figure 5 - Figure Supplement 1: Data for HERPUD1 and P3H1 should be included.

      Our Response: We have now included data to confirm the knockdown for HERPUD1 and LEPRE1 (P3H1) in Fig EV7F-G.

      Reviewer #1, Comment #27: Figure 5 - Figure Supplement 2B: Please mention in the caption how degradation is defined.

      Our Response: We have updated the Fig EV7H caption to include how "degradation" is defined within these experiments:

      "% Degradation is defined as . Where is the fraction of Tg-FT detected in the lysate at a given timepoint n, and is the fraction of Tg-FT detected in the media at a given timepoint n."

      Reviewer #1 (Significance (Required)):

      Reviewer #1, Comment #28: This manuscript is highly significant because the authors (1) designed and validated a new methodology for time-resolved interactomics study, (2) presented the dynamic changes in Tg interactome for WT and variants, and (3) discovered how proteins implicated in degradation pathways (e.g. VCP, TEX264, RTN3) can change the secretion profile of WT and mutant Tg proteins. With TRIP, the authors demonstrated that they could obtain valuable data that were previously not captured from steady-state interactomics studies (Wright et al. 2021; Figure 3M and Figure 3 - Figure supplement 4D-4I). Furthermore, the authors treated cells with VCP inhibitors and performed both 35S pulse-chase analyses and TRIP. These experiments provide valuable information to the field by (1) presenting a new method to rescue Tg secretion defect, and (2) demonstrating a broader applicability of TRIP. If the major comments above can be addressed I believe this is a tremendous contribution to the field.

      Our Response: We thank Reviewer #1 for their review comments and praise for the work presented within this manuscript.

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

      Reviewer #2: In the manuscript 'Time-Resolved Interactome Profiling Deconvolutes Secretory Protein Quality Control Dynamics' Wright et al. developed an approach for time-resolved protein protein interaction mapping relying on pulsed unnatural amino acid incorporation, protein cross linking, sequential affinity purification, and quantitative mass spectrometry named time-resolved interactome profiling (TRIP). The authors applied the TRIP method to compare the interactions of the secreted thyroid prohormone thyroglobulin (Tg) comparing the WT protein to secretion-defective mutations implicated in congenital hypothyroidism. They further employed an RNA interference screening platform (1) to investigate if (1) interactors identified via TRIP are functionally relevant for Tg protein quality control and (2) to identify factors that can rescue mutant Tg secretion. The screen was initially performed in HEK293 cells, but selected hits with a phenotype in HEK cells were then followed up in Fisher rat thyroid cells. Further functional validation was performed by pharmacologic inhibition of VCP, a hit from the RNAi screen with an effect on Tg lysate abundance and Tg secretion. While the authors present a comprehensive study including identification of protein-protein interactions using proteomics followed up by an RNA interference screen for functional validation, major comments need to be addressed for both the proteomics as well as the functional genomics aspects of the study (see comments below).

      Our response: Thank you to reviewer 2 for their constructive feedback. We addressed all comments in detail below.

      Major comments:

      Reviewer #2, Comment #1: The authors describe a new method for quantitative, temporal interaction mapping. The protocol involves two enrichment steps as well as several reactions including cross-linking of the samples as well as functionalization of the unnatural amino acids. Given all these steps, the authors should rigorously characterize the quantitative reproducibility of the experiment when performed in independent biological replicates. This is important because in the final quantitative MS experiment, the authors only use two biological replicates, which is too low especially for such an involved sample preparation procedure, which would expect to have a high variability between replicates. Given the low number of replicates and the unknown reproducibility of the quantification for this protocol, it is questionable at this point how reliable the quantification over the time course is.

      __Our Response: __We apologize that the number of replicates and robustness of the analysis was not entirely clear in our manuscript. We thank the reviewer for the feedback, as this is important point to clarify. We included several additional analyses to further explain the robustness and quantitative reproducibility of our results:

      • We clarified the number of replicates For quantitative MS experiments five biological replicates were analyzed for WT, while six biological replicates were analyzed for A2234D and C1264R Tg-FT, respectively not two as mistakenly presumed by Reviewer #2. These data are available in Dataset EV1 and Table EV3. There is only one place where two biological replicates are included, C1264R Tg-FT FRT cells treated with ML-240 treatment for TRIP analysis. We have further clarified the number of biological replicates within the manuscript text as follows (see also reviewer #1, comment 1):

      "Subsequently, two sets of TRIP time course samples (0, 0.5, 1, 1.5, 2, and 3 hr) could be pooled using the 16plex TMTpro and analyzed by LC-MS/MS (Fig 2A). In total, 5 biological replicates were analyzed for WT and 6 biological replicates were analyzed for A2234D and C1264R, respectively (Table EV3)."

      • We displayed the reproducibility of TRIP time profiles for several individual proteins in Fig EV3 __and in __Fig 3K (VCP). We included shading to indicate the standard error of the mean (SEM) for the individual protein time courses to provide further assessment of the quantitative reproducibility. We updated the text as follows: "To benchmark the TRIP methodology, we chose to monitor a set of well-validated Tg interactors and compare the time-resolved PN interactome changes to our previously published steady-state interactomics dataset (Wright et al, 2021). Previously, we found that CALR, CANX, ERP29 (PDIA9), ERP44, and P4HB interactions with mutants A2234D or C1264R Tg exhibited little to no change when compared to WT under steady state conditions (Fig EV4A). However, in our TRIP dataset we were able to uncover distinct temporal changes in engagement that were previously masked within the steady-state data. Our time-resolved data deconvolutes these aggregate measurements, revealing prolonged CALR, ERP29, and P4HB engagements for both A2234D and C1264R Tg mutants compared to WT (Fig EV4B-F). We found that these measurements for key interactors and PN pathways exhibited robust reproducibility, as exemplified by the standard error of the mean for the TRIP data (Fig EV4B-I, Appendix Figure S1B)."

      • For full transparency, we also include the SEM of all TRIP profiles in the heatmap in Appendix Fig S1B.

      • Furthermore, we included 25-75% quartile ranges for the pathway aggregated time courses (Fig 3B,C,J,K) and the k-means hierarchical clustering analysis (Fig 3F, Fig EV5). Especially these clustering data allow for the visualization and analysis of temporal protein interactions that are correlated with one another, while the accompanying quartile ranges provide further context for the reproducibility of these measurements and cluster profiles (see __Reviewer #1, Comment 17 __above for further explanation about the k-means clustering).

        Reviewer #2, Comment #2: Compared to the previous dataset published last year, the authors discover an overlap in interactors, but also a huge discrepancy, with 96 previously identified interactors not detected in the current study, but 198 additional interactors identified. How do the authors explain the big differences between these datasets?

      __Our Response: __We can only speculate here but this difference in overlapping interactors may stem from several different factors, including but not limited to cell line, instrumentation, LC-MS/MS methodology, and sample processing workflows. Our previous dataset was published using transiently transfected HEK293 cell lines expressed FLAG-tagged constructs of Tg. The HEK293 cell line makes for a robust cell line used throughout several biological investigations, but it is not representative of the native cellular environment in which Tg is expressed. Moreover, transiently transfected cells can lead to high protein expression that may not always represent what is found within the native cellular environment and proteome. Here, we used Fischer rat thyroid (FRT) cells engineered to stably express FLAG-tagged constructs of Tg. This cell line model should more accurately represent the native cellular environment Tg is expressed as it is exclusively found within thyroid tissue. Our previous dataset was collected across two different instruments with similar LC-MS/MS methodology. Here, this dataset was collected on a single instrument after performing further method optimization from our methodology used to acquire the first dataset. In line with our LC-MS/MS methodology development, the sample processing workflows here are quite different. Our previous dataset utilized 6plex TMT labeling with globally immunoprecipitated samples from various Tg constructs. Global immunoprecipitation of Tg leads to much larger protein sample amounts than the TRIP methodology presented here, which we coupled with 16plex TMTpro labeling. This is also one of the reasons we chose to deploy a booster/carrier channel within our experimental labeling schemes.

      Reviewer #2, Comment #3: For the temporal interaction analysis the authors describe differences in the temporal profiles of selected interactions comparing wt and mutant, however no statistical analysis is performed comparing wt and mutant interaction profiles across the time course. Furthermore the variability between the replicates for the temporal profiles is not shown and some of the temporal profiles appear to be noisy. A more rigorous statistical analysis should be performed including additional biological replicates to evaluate the changes over the time course, especially as the temporal interaction analysis is the novelty of this study.

      Our Response: Please also see our response to Reviewer #2, comment 1 above. We previously presented an analysis of the variability of the TRIP measurements (SEM) (now in Appendix Fig S1B). We have since provided further statistical analysis found in the updated Fig 2B,C,J, which include 25-75% quartile ranges for respective proteostasis network pathways. We also included SEM for the time profiles of individual interactors in Fig EV4.

      To assess the divergence in time profiles in an unbiased way, we added a k-means hierarchical clustering analysis (Fig 3F, Fig. EV5). These clustering data allow for the visualization and analysis of temporal protein interaction profiles that are similar to one another and how groups of interactors shift between different clusters for WT Tg and the C1264R mutant.

      Reviewer #2, Comment #4: To functionally validate interactors derived from the TRIP analysis as well as to identify factors that can rescue mutant Tg secretion the authors developed an RNA interference screen. There are a number of aspects that need to be addressed/clarified for this part of the study.

      Our Response: We have added some clarifying changes to the text and the figure panels associated with the siRNA screening and follow-up experiments on the trafficking and degradation factors that rescue Tg secretion. We have addressed other comments from Reviewers #3 and #4 related to these portions of the paper and hope that Reviewer #2 finds them satisfactory.

      Reviewer #2, Comment #5: While the authors validate the stable cell lines expressing the nanoluciferase tagged Tg and the linearity of luminescence signal in lysate and media carefully, they do not validate their platform in combination with the RNAi knockdown strategy. The authors should select genes as positive controls that are expected to modulate Tg secretion and demonstrate that the knockout of these positive controls indeed results in changes in Tg secretion in their system.

      Our Response: This is an excellent suggestion and certainly something we would have done given any prior knowledge on known control genes that would positively or negatively regulate Tg secretion. The purpose for developing the siRNA screening platform was to investigate and hopefully discover genes that are able to positively or negatively regulate Tg processing. We have done so to the best of our ability, identifying for example NAPA which positively regulates WT Tg secretion, as seen by the decrease in WT Tg secretion when treated with NAPA siRNA. Conversely, we found that VCP may negatively regulate C1264R Tg secretion, as discovered by the increase in secretion with VCP siRNA or ML-240 treatment. We included a standard "TOX" siRNA control, which we knew would likely negatively affect WT Tg secretion and this was indeed the case. As we stated within the manuscript:

      "This is the first study to broadly investigate the functional implications of Tg in-teractors and other PQC network components on Tg processing."

      Reviewer #2, Comment #6: For the screen the authors select 167 Tg interactors and PN (Proteostasis network) related factors. This statement is very vague and the authors should clarify which genes were knocked down and which criteria were applied to narrow down the list of interactors and to select PN factors. The authors should therefore provide a supplementary table including all genes included in the screen, their source (were this derived from the initial study by Wright et al, from the current study or compiled from prior knowledge about PN), as well as their results from the screen based on luminescence in media and lysate. It is unclear how many of the selected factors are actually coming from the TRIP analysis.

      Our Response: The list of genes included within the siRNA screen, as well as the results were previously included, and are now included in Appendix Fig S2. We have further provided the information requested by Reviewer #2 within Dataset EV5 indicating whether a gene was included in the siRNA screen due to its identification within our previous proteomics dataset (Wright et al, 2021.), the proteomics dataset presented here, or based upon primary literature. We added a comment in the text:

      "Moreover, we were interested in identifying factors whose modulation may act to rescue mutant Tg secretion. HEK293 cells were engineered to stably express nanoluciferase-tagged Tg constructs (Tg-NLuc) and screened against 167 Tg interactors and related PN components (see Dataset EV5 for the list of genes)."

      Reviewer #2, Comment #7: Only a small number of the 167 selected genes shows an effect on Tg abundance/secretion. How do the authors explain this result? Would we not expect that Tg interactors, especially those from the TRIP method which interact with the newly synthesized are more enriched for functionally relevant genes.

      Our Response: The proteostasis network contains genes and proteins of high redundancy in structure and function, and many single-gene knockdowns are likely insufficient to have a large impact on Tg abundance or secretion. In fact, these results are in line with what we would have expected when designing these experiments. Our goal here was to identify the key players that control Tg protein quality control.

      We explain the proteostasis network redundancy in the manuscript:

      "The functional implications of protein-protein interactions can be difficult to deduce, especially in the case of PQC mechanisms containing several layers of redundancy across stress response pathways, paralogs, and multiple unique proteins sharing similar functions (Wright & Plate, 2021; Bludau & Aebersold, 2020; Karagöz et al, 2019; Braakman & Hebert, 2013)."

      Reviewer #2, Comment #8: The authors initially performed the screen in HEK293 cells and as a second step wanted to validate the hits from the HEK cells in more relevant Fisher rat thyroid cells. Indeed they could show that knockdown of NAPA increased WT TG in lysate and decreased WT Tg secretion. Furthermore, they further validated genes to modulate mutant Tg lysate and media abundance. The authors should perform a rescue experiment to demonstrate that the observed phenotype can be reversed through re-introduction of NAPA.

      Our Response: We have now performed the requested NAPA complementation experiments and provided the data within Fig EV 7I. Overexpression of a human, siRNA-resistant NAPA construct partially reversed the increase in WT Tg lysate retention. These results further support the identification of NAPA as a pro-trafficking factor for WT Tg. We updated the manuscript text to include these data as follows:

      "To understand if these results were directly attributable to NAPA function, we performed complementation experiments where FRT cells treated with NAPA siRNAs were co-transfected with a human NAPA plasmid. WT Tg lysate abundance decreased when NAPA expression was complemented, confirming that the observed retention phenotype could be attributed to NAPA silencing (Fig EV7I). These results established that NAPA acts as a pro-secretion factor for WT Tg."

      Reviewer #2, Comment #9: One hit from this analysis was the ER-phagy receptor TEX264, while TEX264 was not identified in the TRIP data, is selectively increased the C1264R secretion, but not wt and the other Tg mutant. Following Co-IP data however revealed some interaction between the C1264R and to a lesser extent the A2234D mutant. How do the authors explain that TEX264 was missed in the TRIP dataset?

      Our Response: The TRIP samples are of much lower protein abundance compared to globally purified samples used for the Co-IP analysis. While the interaction is seen with the globally purified Co-IP samples, this interaction is likely much more difficult to capture with the low abundance, time-resolved samples that are acquired through the TRIP workflow, especially if this interaction is transient or requires the coordination of other accessory proteins as has been detailed in the literature and discussed within the manuscript presented here:

      "While A2234D and C1264R Tg were preferentially enriched with TEX264 compared to WT, it remains unclear what other accessory proteins may be necessary for the recognition of TEX264 clients (Chino et al, 2019; An et al, 2019). Furthermore, TEX264 function in both protein degradation and DNA damage repair further complicates siRNA-based investigations (Fielden et al., 2022). Further investigation is needed to fully elucidate 1) if Tg degradation takes place via ER-phagy and 2) by which mechanisms this targeting is mediated."

      Minor comments:

      Reviewer #2, Comment #10: The workflow needs to be described clearer. For example, it should be better explained why the authors selected a two-stage enrichment strategy, I assume that the first based on the Flag affinity tag is to purify the protein of interest and the second step based on the incorporation and functionalization of the unnatural amino acids to enrich for the newly synthesized fraction at specific time points after protein synthesis? These are critical steps for the method but the rationals are not well explained, neither in the text nor the figures captures all these steps of the method very clearly, which makes it really difficult for the reader to understand the individual steps of the method. Moreover, the structures in Figure 1 workflow are not clearly labeled, so that it is confusing which part represents which protein/molecule.

      Our Response: Thank you for this feedback. We have updated Fig 1 to provide more detail to provide more clarity for the readers. Furthermore, we have edited the text to more clearly describe the workflow:

      "To develop the time-resolved interactome profiling method, we envisioned a two-stage enrichment strategy utilizing epitope-tagged immunoprecipitation coupled with pulsed biorthogonal unnatural amino acid labeling and functionalization (Fig 1A). Cells can be pulse labeled with homopropargylglycine (Hpg) to synchronize newly synthesized populations of protein. After pulsed labeling with Hpg, samples can then be collected across time points throughout a chase period (Fig 1A, Box 1) (Kiick et al, 2001; Beatty et al, 2006). The Hpg alkyne incorporated into the newly synthesized population of protein can be conjugated to biotin using copper-catalyzed alkyne-azide cycloaddition (CuAAC) (Fig 1A, Box 2). Subsequently, the first stage of the enrichment strategy can take place where the client protein of interest is globally captured and enriched using epitope-tagged immunoprecipitation, followed by elution (Fig 1A, Box 3). The second enrichment step can then utilize a biotin-streptavidin pulldown to capture the Hpg pulse-labeled, and CuAAC conjugated population, enriching samples into time-resolved fractions (Fig 1A, Box 4) (Li et al, 2020; Thompson et al, 2019)."

      Reviewer #2, Comment #11: Except for the general workflow shown in Figure 1, a more detailed workflow showing the experimental steps, such as the sample fractions with the following steps could be added so that the design of the method is clearer. Also the style of the workflows including Figure 1, Figure 2A, and Figure 3A are different. It would be helpful to make them the same style and make the Figure 2A as a zoom in or more detailed illustration on part of Figure 1.

      Our Response: Thank you for this feedback. In addition to updating Fig 1, we also expanded Fig 2A to more clearly outline the experimental steps in the TRIP workflow. Assuming the term "style" used here is in reference to color pallets and figure schematics used, these have been updated to ensure they are agreeable aesthetically across manuscript figures.

      Reviewer #2, Comment #12: A summary of proteomics results of time course labeling after all enrichment steps, including the total number of identified proteins at different conditions and control would be helpful for having an overview impression on the proteomics results

      Our Response: __We have included an updated __Dataset EV1 that provides a summary of proteomics data included which runs given proteins were identified in, % of TMT channels quantified, % of Hpg Pulse channels quantified, and generally number of proteins quantified across runs for each construct.

      Reviewer #2, Comment #13: In Figure 2B, the WB for PDIA4 in the Biotin PD elution is missing. Why was the PDIA4 interaction missing for the time course analysis, but the interaction was captured in the initial test for Wt Tg (Figure 1D). Additionally, in this panel the Rhodamine Probe Gel shows inconsistencies at the time points 1.5 - 3h. Does this mean that the labeling did not work well for these conditions? As we would expect a consistent Rhodamine Probe signal at every time point.

      Our Response: Please also see our response to Reviewer #1, comments 3 & 11. Fig 1D features continuous Hpg labeling for 4 hours to ensure that most intracellular Tg is labeled for this proof-of-concept experiment for the two-stage enrichment strategy. Fig 2B features a shorter 60 minute pulse of Hpg labeling, prior to the full chase period and two-stage enrichment strategy. PDIA4 interactions were detectable throughout Fig 1D because those measurements captured a larger population of labeled Tg, whereas in Fig 2B Tg bait protein amounts were much smaller after the two-stage enrichment procedure to capture the time-synchronized population.

      The Rhodamine/TAMRA Probe Gel in Fig 2B does not have inconsistencies in Tg abundance, but highlights the fact that pulse labeled WT Tg is being secreted or degraded in FRT cells. As you would expect as time continues during the chase period, intracellular WT Tg signal decreases as secretion and degradation take place. Constant Rhodamine/TAMRA probe signal would not be expected here. Consistent with this, the C1264R Tg signal remains more stable for the intial time course. This is expected as the C1264R Tg variant is retained intracellular undergoing increased interactions the proteostasis network. We have removed the PDIA4 panel for WT Tg because there was no signal above the detection limit. This is now explained as follows:

      "For WT Tg, interactions with HSPA5 peaked within the first 30 minutes of the chase period and rapidly declined, in line with previous observations, but PDIA4 interactions were not detectable by western blot analysis (Fig 2B) (Menon et al, 2007; Kim & Arvan, 1995)."

      Reviewer #2, Comment #14: In Figure 2, why was there no WB results for the A2234D? In Figure 2D and 2E, at which time point are the changes significant compared to WT?

      Our Response: We did not perform the WB experiments with A2234D. We used WT and C1264R Tg in our proof of concept experiments via WB and decided to move forward with analyzing A2234D Tg by LC-MS/MS. Please see our response above to Reviewer #2, comment 3 for information on the statistical analysis.

      Reviewer #2, Comment #15: All figure legends should indicate how many biological replicates were performed for each experiment represented in the figure.

      Our Response: We have updated the figure captions to include this information where applicable.

      Reviewer #2, Comment #16: The heatmaps shown in Figure 3, Figure 3 - Figure Supplement 3, and Figure 7 are in the current form incomprehensible. The heatmaps depict the relative enrichment vs the control sample, which was scaled between 1 and -1. The color coding with 5 different colors from 1 to -1 is very confusing and should be changed to just two colors, one for positive and one for negative relative enrichment. I would also suggest changing the visualization of the heatmap showing the wt and mutants side by side, instead of stacked on top of each other for each individual protein.

      Our Response: Thank you for this feedback, and we apologize for the confusion. We adjusted our data analysis approach by removing previous negative enrichment values. As these served only as "background" within the dataset, they did not carry much meaning. The TRIP enrichment is now scaled from 0 to 1, where a value of 1 represents the time point at which the enrichment is greatest, while 0 represents the background intensity in the (-) Hpg control sample. The associated figures have been updated accordingly, and we feel they are now more comprehensible and aesthetically pleasing.

      We opted to keep the Viridis color scheme in the heatmap to allow for more nuanced differentiation of the enrichment values.

      Reviewer #2, Comment #17: The data analysis method for generating relative enrichment shown in the heatmap is not explained. This should be described in the method section for a better understanding of the data analysis.

      Our Response: We have edited the methods section as follows to better explain the analysis:

      "For time resolved analysis, data were processed in R with custom scripts. Briefly, TMT abundances across chase samples were normalized to Tg TMT abundance as described previously and compared to (-) Hpg samples for enrichment analysis (Wright et al, 2021). For relative enrichment analysis, the means of log2 interaction differences were scaled to values from 0 to 1, where a value of 1 represented the time point at which the enrichment reached the maximum, and 0 represented the background intensity in the (-) Hpg channel. Negative log2 enrichment values were set to 0 as the enrichment fell below the background."

      Reviewer #2, Comment #18: There are no legends of flowcharts in Figure 2A and Figure 3A and it is difficult to understand which are the key components in the complex and what are the differences among different periods of labeling.

      Our Response: We have now consolidated Fig 2A and Fig 3A into a single panel found in Fig 2A, which is significantly reorganized to better explain the TRIP workflow. The caption has additionally been updated to highlight key steps within the workflow with numbering to allow readers to follow and visualize the steps more easily. The figure caption now reads as follows:

      "(A) Workflow for TRIP protocol utilizing western blot or mass spectrometric analysis of time-resolved interactomes. (1) Cells are pulse-labeled with Hpg (200μM final concentration) for 1 hr, chased in regular media for specified time points, and cross-linked with DSP (0.5mM) for 10 minutes to capture transient proteoastasis network interactions; (2) Lysates are functionalized with a TAMRA-Azide-PEG-Desthiobiotin probe using copper CuAAC Click reaction; (3) Lysates undergo the first stage of the enrichment strategy where the Tg-FT is globally captured and enriched using immunoprecipitation; (4) Eluted Tg-FT populations from the global immunoprecipitation undergo biotin-streptavidin pulldown to capture the pulse Hpg-labeled, and CuAAC conjugated population of Tg-FT, enriching samples into time-resolved fractions; (5) Time-resolved fraction may then undergo western blot analysis or (6) quantitative liquid chromatography - tandem mass spectrometry (LC-MS/MS) analysis with tandem mass tag (TMTpro) multiplexing or analysis. The (-) Hpg control channel is used to identify enriched interactors and a (-) Biotin pulldown channel to act as a booster (or carrier)."

      Reviewer #2, Comment #19: Why did only one of the VCP inhibitors (ML-240) exhibit a phenotype in Tg abundance and secretion, but not the other VCP inhibitors?

      Our Response: Please also see our response to Reviewer #3, comment 2 below. This could be due to a number of reasons, but we added a brief discussion on the mechanisms of action for the inhibitors that may at least partially explain the differences in phenotype seen with the VCP inhibitors. We updated the text as follows:

      "ML-240 and CB-5083 are ATP-competitive inhibitors that preferentially target the D2 domain of VCP subunits, whereas NMS-873 is a non-ATP-competitive allosteric inhibitor which binds at the D1-D2 interface of VCP subunits (Chou et al, 2013, 2014; Anderson et al, 2015; le Moigne et al, 2017; Tang et al, 2019). ML-240 and NMS-873 have been shown to decrease both proteasomal degradation and autophagy, in line with VCP playing a role in both processes (Chou et al, 2013, 2014; Her et al, 2016). Conversely, while CB-5083 is known to decrease proteasomal degradation it has been shown to increase autophagy. (Anderson et al, 2015; le Moigne et al, 2017; Tang et al, 2019)."

      Reviewer #2 (Significance (Required)):

      Reviewer #2, Comment #20: __The authors __describe a novel and elegant method to map time resolved protein interactions of newly synthesized proteins, which allows monitoring of proteins regulating protein quality control.

      Authors describe it as a general method, however, they only demonstrate the applicability to one protein and do not systematically evaluate the quantitative nature of their approach by determining quantitative reproducibility, which would be necessary to be able to claim that this is a method with broad applicability.

      Given my expertise in quantitative proteomics, I can mainly comment on the technological aspects of the proteomics part of the manuscript, but do not feel qualified to evaluate the significance of this study in terms of novel biology. Nevertheless, it feels that there is a stronger emphasis on the biology in the current form of the manuscript which will raise interest of scientists with a focus on protein quality control and Tg biology.

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

      Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate). Please place your comments about significance in section 2.

      In this manuscript, the authors describe their efforts to develop a methodology for determining time-resolved protein-protein interactions using quantitative mass spectrometry. With TRIP (time-resolved interactome profiling), they combine a pulsed bio-orthogonal unnatural amino acid labelling (homopropargylglycine, Hpg), CuAAC conjugation and biotin-streptavidin pulldowns to enrich at different timepoints and time-resolve by combining TMT labelling and LC-MS/MS (Figure 1). This technique is then applied to the maturation of the secreted WT and mutant thyroglobulin (Tg-WT, Tg-C1264R, Tg-A2234D) expressed in HEK293 and rat thyroid cells (FRT) and linked to hyperthyroidism. There, they identify a collection of ER resident proteins involved in protein folding/processing (e.g. chaperones, redox, glycans, hydroxylation) as well as degradation (e.g. autophagy, ERAD/proteasomes) (Fig. 2). Here the authors effectively use pulse-labelled form of TRIPs to highlight the different interactions formed with Tg-WT vs. Tg-mutants during biogenesis and secretion (or retention). The analysis found ~200 new interactions compared to previous studies along with about 40% of those identified previously. Differences in interactions were observed for mutants, which shown extended interaction with chaperones and redox processing pathways. While many interactions appeared as might be expected, the identification of membrane protein processing elements (e.g. EMC, PAT) was puzzling and raised some questions about the specificity within the protocol. Mutants enriched for CANX CALR and UGGT, suggesting prolonged association with glyco-processing factors. Interaction of C1264R with the ER-phagy factors CCPG1 and RTN3 was greater than WT. The authors note that their interaction correlated with that of EMC1 & 4, but it is not clear why that might be.

      With interactors in hand, the authors complemented the TRIP protocol with siRNA KD of identified factors, to investigate any changes to secreted vs intracellular Tg upon loss. KD of NAPA (a-SNAP) and LMAN1 increased WT lysate (intracellular) Tg but not mutants. NAPA also reduced Tg-WT secretion. In contrast, KD of NAPA increased A2234D secretion while LEPRE1 increased C1264R (but not A2234D or WT), suggesting mutants have differential processing paths and requirements. KD of VCP increased secretion of both mutants. Some ER-phagy receptors were found among interactors (e.g. RTN3 in Tg-C1264R only) but often their KD had no impact on secretion (CCPG1, SEC62, FAM134B). NAMA observations were recapitulated in thyroid derived cell line (FRT). KD of TEX264 and VCP increased Tg-C1264 secretion while RTN3 KD in FRTs decreased Tg-C1264 secretion. This was in contrast to data from HEK293s for reasons that are not clear. Co-IP with TEX264 enriched for all Tg forms but more so for C1264R and A2234D - motivating the authors to propose selective targeting of Tg to TEX264 and the consideration of ER-phagy as a "major" degradative pathway during Tg processing.

      Given the observations with siRNAs to VCP, the authors next use a selection of VCP inhibitors to ask whether secretion can be rescued upon pharmacological impairment of the AAA ATPase. They observed that ML-240, but interestingly not the more conventionally used CB-5083 or NMS-873, increased secretion of Tg-C1264R but not lysate. Inhibitors increased lysate but decreased the secreted fraction for Tg-WT (Fig 7). Finally, the authors used TRIP again in ML-240 treated Tg-C1264R expressing cells to look for changes to interactome with treatment - observed decreases to glycan and chaperone interactions, CANX and UGGT1, decreased interaction with DNAJB11 and C10, like that of WT. There was no apparent change to the UPR, although activation was not directly measured.

      Major comments:

      Reviewer #3, Comment #1: __Are the key conclusions convincing? __The TRIP methodology appears to be quite robust and should be a powerful strategy for this field and others going forward. The drawback will be the length of pulse required will limit the number/type of proteins to be monitored to ones with longer t1/2's. There were interesting interactions found with Tg and the mutants linked to hyperthyroidism, but cut and dry differences did not appear as obvious, even though strong "trends" appear to be present. The path from identifying interactors in a time-resolved manner to then following them up with targeted KD does provides some clarity, which is important.

      Our Response: We thank Reviewer #3 for their time in reviewing our manuscript and providing this positive feedback. We have enhanced our analysis of the TRIP data to more clearly highlight difference in time profiles between WT and mutant variants. Please see our response to Reviewer #2, comment 1 & 3. We also highlight the limitations of the time resolution in the discussion (see also Reviewer #2, comment 6):

      "To address this, we utilized a labeling time of 1 hr which allows us to generate a large enough labeled population of Tg-FT for TRIP analysis, but some early interactions are likely missed within the TRIP workflow. In the case of mutant Tg, performing the TRIP analysis for much longer chase periods (6-8 hrs) may provide insightful details to the iterative binding process of PN components that is thought to facilitate protein retention within the secretory pathway."

      We have addressed all further comments below.

      __Reviewer #3, Comment #2: __Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether? The data regarding VCP silencing and pharmacological impairment appear clear but leave some questions outstanding in this reviewer's opinion. The lack of effect with the 2 highly selective inhibitors suggests that the underlying mechanism for switching fate of intracellularly retained Tg-C1264R towards secreted forms is not at all clear. ML-240 is an early derivative of DBeQ and reportedly impairs both ERAD and autophagic pathways, similarly to DBeQ. The differences between the VCP inhibitors' mechanism of action were not discussed, but perhaps should be elaborated upon, particularly in the matter of how ERAD and ER-phagy pathways might be being differentially affected. At the risk of asking for too many additional experiments, this reviewer would just prefer to see this fleshed out in a bit more detail.

      Our response: We agree with Reviewer #3 that the underlying mechanism for switching fate of the intracellular retained Tg-C1264R towards secreted forms remains unclear. We have added additional text to discuss further the details surrounding the inhibitors used and the general manner in which ERAD and ER-phagy pathways can be affected. This added text reads as follows:

      "ML-240 and CB-5083 are ATP-competitive inhibitors that preferentially target the D2 domain of VCP subunits, whereas NMS-873 is a non-ATP-competitive allosteric inhibitor which binds at the D1-D2 interface of VCP subunits (Chou et al, 2013, 2014; Anderson et al, 2015; le Moigne et al, 2017; Tang et al, 2019). ML-240 and NMS-873 have been shown to decrease both proteasomal degradation and autophagy, in line with VCP playing a role in both processes (Chou et al, 2013, 2014; Her et al, 2016). Conversely, while CB-5083 is known to decrease proteasomal degradation it has been shown to increase autophagy. (Anderson et al, 2015; le Moigne et al, 2017; Tang et al, 2019)."

      "As we discovered that pharmacological VCP inhibition with ML-240 can rescue C1264R Tg secretion yet is detrimental for WT Tg processing, it is unclear whether VCP may exhibit distinct functions for WT and mutant Tg PQC. Finally, as ML-240 is shown to block both the proteasomal and autophagic functions of VCP it is unclear which of these pathways may be playing a role in the rescue of C1264R, or detrimental WT processing (Chou et al, 2013, 2014)."

      __Reviewer #3, Comment #3: __Q1. The degree (if any) of Tg-C1264 aggregation during and/or detergent solubility do not appear to have been considered as a potential source of the increase in released secreted material (Figure 4, 5). Do Tg mutants partition into RIPA-insoluble fractions at all? That is to say.. is the total population of synthesized Tg being considered? A full accounting? Could the authors address this and if biochemical extraction data (via urea or high SDS) is available, include it to answer this concern.

      Our response: The transient aggregation of Tg has been investigated in some detail previously (Kim et al, 1992, 1993). The transient aggregates have the ability to partition into RIPA-insoluble fractions. Of note, these aggregates are shown to be made up, at least in part, of mixed disulfide linkages requiring reducing agent to fully resolubilize. With that being said, these aggregates represent a minority of the overall Tg population. In our prior manuscript (Wright, et al. 2021), we quantified the RIPA-insoluble fraction found in the pellet (see Supplemental Info Fig. 5). As the majority of Tg remains soluble during processing it should be able to be captured via our TRIP methodology. That is to say, we are capturing most of the Tg that is available for analysis while understanding that some smaller population of Tg remains in RIPA-insoluble fractions.

      __Reviewer #3, Comment #4: __Q2. Along the same lines, what does Tg-WT and mutant expression look like by microscopy? Is Tg-WT uniformly distributed while Tg-mutants appear in puncta... more aggregated - perhaps reflecting the increased engagement of chaperones and redox machinery? Changes in the pattern of Tg-C1264R mutant (e.g. w/ VCP KD or inhibition) would add additional support for the authors interpretation of improved secretion. If this data is at hand, including it might be worth consideration.

      Our response: Thank you for this suggestion. The subcellular localization of Tg and any changes from proteostasis modulation is an ongoing area of follow up work in our lab. We have some preliminary results that the localization for WT and C1264R Tg indeed differs. However, given that this manuscript is already dense in information, we opted to reserve this data for a future manuscript where we plan to further elucidate the targeting mechanism of mutant Tg to VCP or TEX264. We direct the reviewer to work published by Zhang et al, 2022,(https://doi.org/10.1016/j.jbc.2022.102066) showing a staunch difference of WT vs mutant Tg in the localization from intracellular to a secreted population in rat tissue. While most all WT Tg is found in the follicular lumen (secreted), mutant Tg heavily co-localizes with the ER resident chaperone BiP. While this paper does not go into detail on the differences in subcellular localization, it further highlights the drastic changes in Tg processing and how these manifest in distinct differences in localization within tissue.

      __Reviewer #3, Comment #5: __Q3. Does the level of Tg mutant expression in the FRT clones impact the profiles obtained by TRIP? (Figure 3). This is a question of gauging the relative saturation of QC machinery and how that might impact profiles from TRIP. Were clones expressing at different levels tested? Perhaps a brief discussion of this.

      Our response: We do not foresee an impact from level of Tg expression on the profiles obtained by TRIP. We were able to identify distinct profiles because we processed the data and normalized it based on the relative Tg amount. For example, while WT and A2234D Tg are expressed at similar levels intracellularly, we were able to identify distinct differences in the interaction profiles across the two constructs. When developing FRT clones, we selected those that were expressed at similar levels and, therefore, did not have the capability to directly test differences, if any, in observed profiles that may be the result of different expression levels of the same Tg construct. Furthermore, Tg can make up 50% of all protein content within thyroid tissue (Di Jeso & Arvan, 2016). As such, thyroid cells are adept at maintaining the balance of QC machinery to process thyroid. Therefore, we do not anticipate that the amount of Tg expressed in TRIP experiments would have a significant impact on the profiles that we were able to observe.

      __Reviewer #3, Comment #6: __Q4. For Figure 3, the hour-long labelling period seems a bit long, compared with 3 hr of chase. Perhaps this reviewer missed this but how long does Tg take to mature and/or mutants to misfold and degrade? Is there any possibility to shorten this so that the profiles of labelled Tg could be more synchronized? If not, perhaps this could just be discussed.

      Our response: While the 1-hour labeling period may seem long, we had to balance the labeling time to 1) label a large enough population of Tg for it to remain detectible throughout the chase period, and 2) keep the chase period long enough to capture the large majority of Tg processing. In our hands we found that by 4 hours WT Tg was ~63% secreted, with ~25% retained intracellular (Fig EV7H). Conversely, we found that C1264R remains very stable over this period with most protein being retaining intracellularly and little degradation taking place (Fig EV9A). Hence, we opted for the overall ~4 hour total for sample processing (1 Hr pulse labeling + 3 hour chase period for time point collections). Literature suggest that WT Tg takes ~2 hours to be processed within the ER and reach the medial golgi. This is exemplified by the EndoH resistant population that appears at this ~2 hour time point (Menon et al. JBC. 2007). Please also see our response to Reviewer #1, comment 6. We updated the text as follows:

      "We pulse labeled WT Tg FRT cells with Hpg for 1 hr, followed by a 3 hr chase in regular media capturing time points in 30-minute intervals and analyzing via western blot or TMTpro LC-MS/MS (Fig 2A). Our previous study indicated that ~70% of WT Tg-FT was secreted after 4 hours, while approximately 50% of A2234D and 15% of C1264R was degraded after the same time period (Wright et al, 2021). Therefore, we reasoned that a 3-hr chase period would be a enought time to capture the majority of Tg interactions throughout processing, secretion, cellular retention, and degradation, while still being able to capture an appreciable amount of sample for analysis."

      We anticipate that this labeling period can be decreased with future iterations of this methodology. This will also be bolstered by the continued improvements that come about within quantitative proteomics in increased instrument sensitivity and improved sample preparation methods that have the ability to decrease sample loss.

      We explain the labeling timeline and limitations further in the discussion:

      "To address this, we utilized a labeling time of 1 hr which allows us to generate a large enough labeled population of Tg-FT for TRIP analysis, but some early interactions are likely missed within the TRIP workflow. In the case of mutant Tg, performing the TRIP analysis for much longer chase periods (6-8 hrs) may provide insightful details to the iterative binding process of PN components that is thought to facilitate protein retention within the secretory pathway."

      __Reviewer #3, Comment #7: __Q5. It is curious that only ML-240 and not other well characterized inhibitors of VCP/p97, has an effect, as both are used far more often than ML-240. The authors do not really address this in detail but does it suggest that the ML-240 effect on VCP/p97 could be affecting different pathways, given the nature of this compound. Is this compound acting on Tg-C1264R maturation at the level of translation or post-translationally? If the latter, through what means?

      Our Response: We thank Reviewer #3 for appreciating this surprising finding. We were similarly curious as to how, or why ML-240 was able to elicit this effect compared to other VCP inhibitors. We elaborated in the manuscript text on these compounds and on how the ERAD and ERphagy pathways, utilizing VCP, may be differentially regulated (See response to__ Reviewer #3, Comment 2__). While speculative, we believe that ML-240 acts on C1264R Tg maturation post-translationally. This is given by the fact that ML-240 does not seem to affect the translational velocity of C1264R Tg, as Fig EV9A shows similar levels of 35S-labeled C1264R in DMSO or ML-240 treated cells. It may be the case that acute treatment with ML-240 alters the folding vs degradation balance of the ER proteostasis network in such a way that some population of C1264R that is usually degraded is able to be secreted. Another Tg mutation G2320R was shown to be degraded via the proteasome in PLCCL3 thyrocytes, as MG-132 treatment slowed mutant Tg degradation (Menon et al. JBC. 2007), although G2320R degradation was not be exclusively proteasomal. The L2284P Tg mutation exemplified similar results to G2340R where MG-132 slowed degradation. Furthermore, L2284P Tg was not affected by autophagic/lysosomal inhibitors chloroquine and E64 (Tokunaga et al. JBC. 2000), suggesting ERAD more exclusively degrades L2284P. It is unclear which degradation pathway, ERAD or ER-phagy, may be the predominate pathway for C1264R Tg degradation. Furthermore, we do not exclude the possibility that both may be at play and affected by treatment with ML-240.

      We utilized our HEK293 Tg-NLuc cells and screened other proteasomal and lysosomal inhibitors bafilomycin and bortezomib. Neither of these compounds were able to rescue A2234D or C1264R secretion, highlighting that the effect is specific to ML-240 treatment. This new data is now shown in __Fig EV10A,B __and described in the text:

      "To understand whether this rescue in secretion was uniquely linked to VCP inhibition or could be more broadly attributed to blocking Tg degradation, we tested the proteasomal inhibitor bortezomib, and lysosomal inhibitor bafilomycin. Bafilomycin increased WT Tg lysate abundance, and bortezomib significantly increased A2234D lysate abundance, consistent with a role of these degradation processes in Tg PQC (Fig EV10A). When monitoring Tg-NLuc media abundance, neither bafilomycin nor bortezomib significantly altered WT, A2234D, or C1264R abundance (Fig. EV10B). confirming that general inhibition of proteasomal or lysosomal degradation does with rescue mutant Tg secretion."

      __Reviewer #3, Comment #8: __Q6. Continuing from Q5.. At what point and where is VCP/p97 able to affect mutant Tg processing? In line 317, the authors seem to correlate increased VCP association with mutants to their increased secretion. It is not clear how this would result, as engagement with VCP would be in a compartment different to that which supports trafficking and secretion. Could the authors expand on how this might come about. This is also relevant to the ML-240 data in Figure 7. Moreover, VCP is associated with ERAD (as is HerpUD1) rather than ER-phagy and at least in the siRNA raw data, there are also effects from Derlin3 and FAF2 KDs.. both ERAD factors. Some clarity here would be appreciated.

      Our Response: This line of discussion in the text was meant to suggest that, since VCP showed a higher enrichment for mutant Tg, particularly C1264R, it would make sense that inhibiting VCP would have a larger effect on mutant Tg processing as compared to WT Tg. As we saw with the siRNA screening data, suppression of VCP resulted in increased C1264R secretion, while not affecting WT Tg processing. This passage was not intended to suggest that increased VCP association with mutant Tg found within the TRIP dataset was the reason for rescued secretion. These are two different sets of experiments and environments in which these data are captured. We were simply looking for the opportunity to bridge the findings from the two sets of experiments to a single discussion point. Of note, we understand that VCP is associated with ERAD and acts to regulate autophagy. Given that core autophagy machinery is relevant for both bulk autophagy and ER-phagy, we did not want to rule out the fact that VCP inhibition via ML-240 could affect autophagic flux in these experiments (Chou et al. Chemmedchem. 2013; Khaminets et al. Nature. 2015; Hill et al. Nat. Chem. Bio. 2021.)

      It is great that the reviewer also noted that DERL3 and FAF2 knockdown increased C1264R Tg secretion. Since these ERAD factors did not reach the defined threshold in the screen, we did not include further discussion, but this data remains available in Appendix Fig S3. We have updated the manuscript text to clarify the previous points we aimed to make. The text now reads as follows:

      "VCP silencing exclusively affecting mutant Tg corroborates our TRIP dataset, and suggest a more prominent role for VCP in mutant Tg PQC compared to WT. VCP interactions were sparse for WT Tg while they remained more steady throughout the chase period for the mutants (Fig 3H,K)."

      __Reviewer #3, Comment #9: __Q7. There does not appear to be a direct demonstration of Tg-C1264R turnover by ER-phagy (via TEX264). Given the inconsistency with it not being detected by TRIP, while another receptor RTN3 was, but has not impact on Tg-C1264R secretion, perhaps including that data would go some way to demonstrating a fate of ER-phagy (at least partly) for this mutant.

      Our response: We performed follow-up experiments to test interactions with Tg and the wider panel of ER-phagy receptors. We transiently expressed FLAG-tagged CCPG1, RTN3L, and TEX264 in HEK293 cells stably expressing Tg-NLuc and performed FLAG IPs followed by western blot analysis. We found that WT and C1264R Tg were enriched, albeit modestly, in the RTN3L Co-IP compared to control samples expressing GFP. Additionally, we found that WT, A2234D, and C1264R Tg were all enriched with CCPG1 compared to control samples expressing GFP. CCPG1 was found to be a C1264R Tg interactor within our mass spectrometry datasets, along with RTN3. We have now integrated these data into the manuscript as Fig EV8, and updated the manuscript text as follows:

      "Additionally, we monitored Tg enrichment with ER-phagy receptors CCPG1 and RTN3 via Western blot as both were found to be C1264R Tg interactors within our TRIP dataset. RTN3L is found to be the only RTN3 isoform involved in ER turnover via ER-phagy (Grumati et al, 2017). WT and C1264R Tg-NLuc were modestly enriched with RTN3L compared to control samples expressing GFP. Conversely, we found that all Tg variants exhibited modest interactions with CCPG1 compared to control samples expressing GFP, although less than with TEX264 (Fig EV8).

      Together, these data suggest that TEX264, CCPG1, or RTN3L engage with Tg during processing, and CH-associated Tg mutants may be selectively targeted to TEX264. Furthermore, ER-phagy may be considered as a degradative pathway in Tg processing, as other studies have mainly focused on Tg degradation through ERAD (Tokunaga et al, 2000; Menon et al, 2007)."

      Whether the TEX246 recruitment of mutant Tg leads to degradation remains to be tested. When we monitored C1264R Tg degradation by pulse-chase assay (Fig. EV9A), only a small fraction (

      __Reviewer #3, Comment #10: __Q9. The authors provide data that the UPR was not induced by ML-240 at 3hrs (10µM) (Figure 7, supplemental 1). This is in stark contrast to the results of Chou et al (2013) which the authors reference, reporting that ML-240 induced ATF4 and CHOP by 2 hrs at concentrations lower than used here (albeit a different cell type). While not exclusively UPR, could the authors address the potential activation of the integrated stress response (eIF2a phosphorylation, ATF4 and CHOP) in the FRT cells due to ML-240 treatment? If present, is there some link that could this provide an explanation for increased Tg-C1264R secretion? [Basal PERK/UPR activation with mutants.]

      Our Response: Thank you for bringing up this important point. As the reviewer acknowledges, the difference in UPR activation could stem from the different cell lines. Additionally, we measured activation via qPCR, whereas Chou et al. measured via immunoblot. We would like to point out that while we did not observe the upregulation of HSPA5 or ASNS (markers of ATF6 and PERK/ISR activation, respectively) in the presence of short ML-240 treatment (2-3 hr), we did observe the upregulation of DNAJB9 (a marker of IRE1/XBP1s activation).

      To address Reviewer #3's point, we performed further experiments monitoring the potential activation of the ISR in FRT cells due to ML-240 treatment. We treated C1264R Tg-FT FRT cells with ML-240 (10μM) for 2 hours, and monitored eIF2a phosphorylation via immunoblot. Indeed, we observed that ML-240 induced eIF2a phosphorylation compared to cells treated with DMSO. Tunicamycin (1mg/mL) was used a positive control, and showed similar results to ML-240. We have integrated these results into the manuscript, available in Fig EV10C.

      However, we would like to point out that all of these markers represent signs of early UPR inductions. Importantly, our results that HSPA5 transcript levels are not induced suggest that there is only very modest upregulation of ER chaperone levels occurring. Typically, the ER proteostasis network remodeling requires a longer time than the acute 2-4 hr treatment with ML-240. We have updated the manuscript text as follows:

      "Finally, we monitored activation of the unfolded protein response (UPR) in the presence of ML-240 in FRT cells expressing C1264R Tg-FT. Phosphorylation of eIF2a, an activation marker for the PERK arm of the UPR, was induced within 2 hr of ML-240 treatment (Fig EV10C). We further investigated the induction of UPR targets via qRT-PC. HSPA5 and ASNS transcripts, markers of ATF6 and PERK UPR activation respectively, remained unchanged or slightly decreased after 3 hr treatment with ML-240 in C1264R Tg cells (Fig EV10D). Only DNAJB9 transcript expression showed a significant increase in both WT Tg and C2164R Tg FRT cells (Fig EV10D). Moreover, ML-240 did not significantly alter cell viability after 3 hr, as measured by propidium iodide staining (Fig EV10E). Overall, these results highlight that the short ML-240 treatment induces early UPR markers, but the selective rescue of C1264R Tg secretion via ML-240 treatment is unlikely the results of global remodeling of the ER PN due to UPR activation."

      __Reviewer #3, Comment #11: __Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments. Any of the suggested experiments above all use reagents reported in the manuscript and so would presumably incur minimal cost and hopefully time. This reviewer is sympathetic to time and financial constraints and so discussion of the issue could suffice.

      Our response: We have addressed follow-up experiments whenever possible or provided further discussion details where applicable. We are appreciative of Reviewer #3's sympathy for the time and financial constraints that go into this work and addressing manuscript revisions. Unfortunately, the 1st and 2nd authors both left the lab immediately after the reviews were received. Hence, many of the experiments had to be addressed by other lab members joining the project, which took considerably longer than anticipated. We apologize for the long delay with our revisions.

      __Reviewer #3, Comment #12: __Are the data and the methods presented in such a way that they can be reproduced? Yes. The methodology is explained in detail.

      Our Response: Thank you.

      __Reviewer #3, Comment #13: __Are the experiments adequately replicated and statistical analysis adequate? Yes. Relevant information is either in the figure legends or is provided in the source data.

      Our Response: Thank you.

      Minor comments:

      __Reviewer #3, Comment #14: __Are prior studies referenced appropriately? The references are generally appropriate, with a few exceptions of more general references used

      Our Response: Thank you.

      __Reviewer #3, Comment #15: __Are the text and figures clear and accurate? The text is clearly written, and the figures are clear.

      Our Response: Thank you.

      __Reviewer #3, Comment #16: __Do you have suggestions that would help the authors improve the presentation of their data and conclusions? A summary figure comparing the changing profiles of WT and C1264R and the factors implicated for them could be helpful.

      Our Response: We opted not to include a summary figure because the paper and figures area already dense in information.

      __Reviewer #3, Comment #17: __Perhaps include common nomenclature for proteins as well (e.g. HSP5A - BiP, HSP90B1 - Grp94, etc..)

      Our Response: We updated the manuscript throughout to reference common nomenclature or other protein names where applicable at their first mention.

      __Reviewer #3, Comment #18: __Line 317 - our is misspelled

      Our Response: Thank you. We have made this correction.

      __Reviewer #3, Comment #19: __Figure 4 - Supplemental Figure 1 - Legend has text referring to panels J and K, but Figure only goes up to F.

      Our Response: Thank you. This was an error in references to Figure panel lettering and we have since corrected this. Please note that this Figure is now Fig EV6.

      Reviewer #3 (Significance (Required)):

      __Reviewer #3, Comment #20: __

      • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

      • Place the work in the context of the existing literature (provide references, where appropriate).

      Protein-protein interactions are often used to illustrate complexes and functionality, but these provide only snapshots, rather than "movies". There are many datasets out there exploring P-P interactions, but most if not all lack any temporal resolution for the interactions they report. The TRIP method described approaches this from the dynamic perspective - identifying the transient interactions formed by folding nascent chains with proteins that aid in their maturation and trafficking, or degradation. This represents an important technical advance in our ability to dynamically monitor protein interactions. The use of Tg mutants is valuable and perhaps this will lead to new perspectives on how to rescue it or other pathophysiological mutants with loss of function phenotypes.

      • State what audience might be interested in and influenced by the reported findings.

      This work should appeal to a broad audience within cell biology, particularly as the TRIP technique is attempting to address a fundamental question - what interactions form during the biogenesis/lifetime of a protein. Moreover, the effort to try to understand the different interactions formed with pathologically relevant mutant proteins as a strategy to try to rescue functionality, is a valuable exercise of this approach.

      • 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.

      ER quality control

      Our Response: We thank reviewer #3 for this positive endorsement.

      Reviewer #4 (Evidence, reproducibility and clarity (Required)):

      Summary

      In this manuscript, Wright et al. developed an approach (termed TRIP) that allowed to map the temporal changes in the interaction landscape of a newly synthesized protein of interest. Using their TRIP approach, the authors found that the extensive interactions of thyroglobulin (Tg) with the proteostasis network (PN) during its passage through the secretory pathway were profoundly altered in response to disease-causing mutations (e.g. C1264R). The authors cross-validated their findings with a focus RNAi screen monitoring the cellular and secreted abundance of Tg variants upon deletion of PN components. In subsequent experiments the authors focused on two hits, VCP and TEX264, for which they confirmed their inhibitory effect on the secretion of Tg C1264R. Importantly, the authors found that TEX264 increasingly interacts with the Tg mutant and that pharmacological inhibition of VCP yielded the same phenotype than depletion of VCP. Overall, Wright and colleagues__ established an elegant method to map protein interaction in a time-resolved manner and demonstrated its value by the analysis of disease-related Tg mutants__. Hence, this work has the potential to serve as a rich resource for Tg-related research and as a powerful new tool to examine protein interactions. However, several concerns remain.

      Our response: Thank you to reviewer #4 for their valuable feedback and positive assessment. We addressed all comments in detail below.

      Major points:

      __Reviewer #4, Comment #1: __Overall, the TRIP workflow is quite difficult to understand at a first glance - even for a reader with a background in proteomics, biochemistry and cell biology. The authors may want to improve the description of the TRIP methodology and explain in more detail what the individual components and steps are good for. Along the same line, from the main text and the figure legend it was not clear that Tg was actually Flag-tagged. However, without this information it is difficult to follow the workflow. While Figure 1A is certainly helpful, the bulky graphics are deflecting the reader's attention. A more schematic version might be more informative.

      Our Response: Thank you for this feedback, which was also mirrored by Reviewer #2 (comment 10). We have made significant updates to clarify Fig 1 to provide more detail and eliminate some of unnecessary bulky graphics. We also expanded the schematic for the TRIP workflow in Fig 2A and we aligned all symbols used. Furthermore, we have edited the text to describe the workflow more clearly:

      "To develop the time-resolved interactome profiling method, we envisioned a two-stage enrichment strategy utilizing epitope-tagged immunoprecipitation coupled with pulsed biorthogonal unnatural amino acid labeling and functionalization (Fig 1A). Cells can be pulse labeled with homopropargylglycine (Hpg) to synchronize newly synthesized populations of protein. After pulsed labeling with Hpg, samples can then be collected across time points throughout a chase period (Fig 1A, Box 1) (Kiick et al, 2001; Beatty et al, 2006). The Hpg alkyne incorporated into the newly synthesized population of protein can be conjugated to biotin using copper-catalyzed alkyne-azide cycloaddition (CuAAC) (Fig 1A, Box 2). Subsequently, the first stage of the enrichment strategy can take place where the client protein of interest is globally captured and enriched using epitope-tagged immunoprecipitation, followed by elution (Fig 1A, Box 3). The second enrichment step can then utilize a biotin-streptavidin pulldown to capture the Hpg pulse-labeled, and CuAAC conjugated population, enriching samples into time-resolved fractions (Fig 1A, Box 4) (Li et al, 2020; Thompson et al, 2019)."

      Additionally, we have improved text to very clearly state that for the TRIP experiments Tg is FLAG-tagged and this epitope tag is required for the two-stage enrichment strategy. As one small example:

      "Thyroglobulin was chosen as the model secretory client protein. We generated isogenic Fischer rat thyroid cells (FRT) cells that stably expressed FLAG-tagged Tg (Tg-FT), including WT or mutant variants (A2234D and C1264R) (Fig EV1)"

      "Furthermore, the C-terminal FLAG-tag and Hpg labeling are necessary for this two-stage enrichment strategy, and DSP crosslinking is necessary to capture these interactions after stringent wash steps (Fig 1D, Fig EV2)."

      __Reviewer #4, Comment #2: __To what extend do the difference in protein abundance between Tg WT and Tg C1264R contribute to the increase binding of their interactors (e.g., HSP5 and PDIA4). The authors should perform a TRIP coupled immunoblot analysis where WT and Mutant are loaded side-by-side on the SDS-PAGE.

      Our Response: As Reviewer #3 (comment 5) had a similar inquiry, we provide the same response as listed above:

      We do not foresee an impact from level of Tg expression on the profiles obtained by TRIP. We were able to identify distinct profiles because we processed the data and normalized it based on the relative Tg amount. For example, while WT and A2234D Tg are expressed at similar levels intracellularly, we ere able to identify distinct differences in the interaction profiles across the two constructs. When developing FRT clones, we selected those that were expressed at similar levels and, therefore, did not have the capability to directly test differences, if any, in observed profiles that may be the result of different expression levels of the same Tg construct. Furthermore, Tg can make up 50% of all protein content within thyroid tissue (Di Jeso & Arvan, 2016). As such, thyroid cells are adept at maintaining the balance of QC machinery to process thyroid. Therefore, we do not anticipate that the amount of Tg expressed in TRIP experiments would have a significant impact on the profiles that we were able to observe.

      __Reviewer #4, Comment #3: __While the RNAi screen was done with pooled siRNA, it is not clear what was used for the RNAi validation experiments shown in Figure 5. This should be done by individual siRNA and not the same pooled reagents as used for the screen.

      Our Response: Similarly, pooled siRNAs were initially utilized for the data shown in Figure 5. The RNAi screen utilized siRNAs optimized for human cells, where as those found for Figure 5 were for rat cells. For the revisions, we performed control experiments with individual siRNAs, which are now shown in Fig EV7J,K. While we did not find that any one single siRNA recapitulated the full phenotype, we did find that several single siRNAs for VCP and TEX264 at least partially restored the observed phenotype of increased C1264R Tg secretion. This result is expected given that we reasoned the siRNAs are likely providing an additive effect contributing to the observed phenotypes. We provided these single siRNA control experiments in Fig EV7J,K, and updated the manuscript text as follows:

      "Several individual VCP and TEX264 siRNAs were able to partially recapitulate these increased secretion phenotype on C1264R Tg-FT, confirming that the effect is mediated by the respective gene silencing (Fig EV7J,K)."

      Reviewer #4, Comment #4: __In Figure 5A it is not clear which band was used to quantify the effect of NAPA reduction. Also, this analysis lacks normalization to an unrelated protein or loading control. Moreover, the authors should also examine the effect of the siRNA targets shown in Figure 5C for Tg WT and not only the mutant.__

      Our Response: The uppermost band in Fig 5A was used for quantification. We added a red asterisk similar to that found in Fig 5C to denote this lower back in the lysate panel(s) as a non-specific background band found within the Western blot. These data are the result of immunoprecipitations of both cell lysate and medium content, as such there is no applicable loading control that can be used within the western blots. For experiments, cell amounts were normalized by seeding and subsequently culturing the same amount of cells, as denoted within the Materials and Methods - FRT siRNA validation studies section of the manuscript. Furthermore, there are no loading controls that are easily utilized for analyzing cell culture medium. We have further clarified the Fig 5 caption to provide clearer experimental detail:

      "(A and B) Western blot analysis (A) and quantification (B) of WT Tg-FT secretion from FRT cells transfected with select siRNAs hits from initial screening data set. Red asterisk denotes a non-specific background band within the western blot. Cells were transfected with 25nM siRNAs for 36 hrs, media exchanged and conditions for 4 hrs, Tg-FT was immunoprecipitated from lysate and media samples, and Tg-FT amounts were analyzed via immunoblotting. N = 6.

      (C and D) Western blot analysis (C) and quantification (D) of C1264R Tg-FT secretion from FRT cells transfected with select siRNA hits from the initial screening data set. Red asterisk denotes a non-specific background band within the western blot. Cells were transfected with 25nM siRNAs for 36 hrs, media exchanged and conditions for 8 hrs, Tg-FT was immunoprecipitated from lysate and media samples, and Tg-FT amounts were analyzed via immunoblotting. All statistical testing performed using an unpaired student's t-test with Welch's correction. *pFinally, as the siRNA targets shown in Fig 5C were shown to be hits exclusively for C1264R Tg-FT we did not believe it was necessary to follow-up on these with WT Tg-FT. Similarly, we did not follow-up on hits that were exclusive to WT Tg-FT with C1264R and A2234D Tg-FT.

      __Reviewer #4, Comment #5: __The authors should also test for the binding of RTN3 to Tg WT and mutant - in particular in comparison to TEX264. This would be important in the context that only RTN3 but not TEX264 was detected in the TRIP approach. Do the authors also detect VCP and LC3B in their pulldowns?

      Our response: Please also see Reviewer #3, comment 9, who made a similar point.

      We performed follow-up experiments to test interactions with Tg and the wider panel of ER-phagy receptors. We transiently expressed FLAG-tagged CCPG1, RTN3L, and TEX264 in HEK293 cells stably expressing Tg-NLuc and performed FLAG IPs followed by western blot analysis. We found that WT and C1264R Tg were enriched, albeit modestly, in the RTN3L Co-IP compared to control samples expressing GFP. Additionally, we found that WT, A2234D, and C1264R Tg were all enriched with CCPG1 compared to control samples expressing GFP. CCPG1 was found to be a C1264R Tg interactor within our mass spectrometry datasets, along with RTN3. We have now integrated these data into the manuscript as Fig EV8, and updated the manuscript text as follows:

      "Additionally, we monitored Tg enrichment with ER-phagy receptors CCPG1 and RTN3 via Western blot as both were found to be C1264R Tg interactors within our TRIP dataset. RTN3L is found to be the only RTN3 isoform involved in ER turnover via ER-phagy (Grumati et al, 2017). WT and C1264R Tg-NLuc were modestly enriched with RTN3L compared to control samples expressing GFP. Conversely, we found that all Tg variants exhibited modest interactions with CCPG1 compared to control samples expressing GFP, although less than with TEX264 (Fig EV8).

      Together, these data suggest that TEX264, CCPG1, or RTN3L engage with Tg during processing, and CH-associated Tg mutants may be selectively targeted to TEX264. Furthermore, ER-phagy may be considered as a degradative pathway in Tg processing, as other studies have mainly focused on Tg degradation through ERAD (Tokunaga et al, 2000; Menon et al, 2007)."

      Regarding VCP, we can detect it routinely in our AP-MS experiment as presented previously (Wright et al. 2021), and here in Fig 3, Appendix Fig S1. However, we have not been able to detect interactions via western blot, which may be attributed to the increased sensitivity that LC-MS offers. We have not probed for LC3 interactions via western blot as we did not detect it by LC-MS either, but we identified several lysosomal and other autophagy-related components previously (Wright et al. 2021), and here shown in Appendix Fig S1 and Fig EV5C.

      __Reviewer #4, Comment #6: __The effect of TEX264 depletion on Tg secretion should be confirmed by TEX263 KO experiments. Do the authors observe a similar increase in secreted Tg C1264R in BafA1- or SAR405-treated cells? Moreover, the authors should show that Tg C1264R is actually delivered to lysosomes using biochemical assays such as LysoIP or colocalization experiments.

      Our response: To address this concern, we generated stable TEX264 knockout FRT cell lines by CRISPR, and probed several clones for their impact on Tg secretion. We found that TEX264 knockout did not recapitulate the increase in C1264R Tg secretion observed with transient siRNA knockout. While disappointing, these results are not necessarily surprising, considering that prolonged TEX264 knockout may lead the cell to adapt compensation mechanisms by modulating other proteostasis factors and/or autophagy machinery.

      We performed experiments utilizing the autophagy inhibitor Bafilomycin A1, and have now included these results with the manuscript available in Fig EV10A,B. We found that BafA1 treatment led to the accumulation of WT Tg in the lysate but not for the C1264R Tg. We updated the manuscript text to accompany these data as follows:

      "To understand whether this rescue in secretion was uniquely linked to VCP inhibition or could be more broadly attributed to blocking Tg degradation, we tested the proteasomal inhibitor bortezomib, and lysosomal inhibitor bafilomycin. Bafilomycin increased WT Tg lysate abundance, and bortezomib significantly increased A2234D lysate abundance, consistent with a role of these degradation processes in Tg PQC (Fig EV10A). When monitoring Tg-NLuc media abundance, neither bafilomycin nor bortezomib significantly altered WT, A2234D, or C1264R abundance (Fig. EV10B). confirming that general inhibition of proteasomal or lysosomal degradation does with rescue mutant Tg secretion."

      These results raise the possibility that the mutant Tg interaction with TEX264 may not lead to active autophagic degradation of mutant Tg. This is also consistent with the slow degradation of C1264R Tg observed in the pulse-chase experiment in Fig EV9A. Whether the TEX246 recruitment of mutant Tg leads to degradation or assumes an alternative function, for example, intracellular sequestration, remains to be tested. Importantly, we have refrained from making claims in the manuscript that C1264R Tg is delivered to the lysosome but have presented data showing that it interacts with ER-phagy-related components and have further speculated on the possibility how autophagy could play a role in Tg processing.

      Thank you for the LysoIP suggestion. Ongoing work in the lab is addressing this question and experiments suggested by the reviewer, but this is better reserved for a follow-up manuscript.

      __Reviewer #4, Comment #7: __Figure 7A and 7C lack loading controls. The quantification shown in Figure 7B and 7D should be normalized to this control. Since VCP activity is often coupled to the of the proteasome, the authors should check whether blocking the proteasome yields a similar effect than ML-240.

      Our Response: Like Fig 5A discussed above (Reviewer #4, comment 4), these data are the result of immunoprecipitations from cell lysate and medium. As a result, there is not applicable loading control that can be used within the western blots. For experiments, cell amounts were normalized by seeding and subsequently culturing the same amount of cells, as denoted within the Materials and Methods - FRT siRNA validation studies section of the manuscript and Material and Methods - VCP pharmacological inhibition studies.

      Regarding the effect of proteasome inhibition, we tested whether bortezomib treatment can increase C1264R Tg secretion. We found that bortezomib led to a small but significant increase in A2234D Tg accumulation in the lysate, but did not increase secretion of Tg for WT or any of the mutant variants. This new data is shown in Fig EV10A,B. We updated the text as follow:

      "To understand whether this rescue in secretion was uniquely linked to VCP inhibition or could be more broadly attributed to blocking Tg degradation, we tested the proteasomal inhibitor bortezomib, and lysosomal inhibitor bafilomycin. Bafilomycin increased WT Tg lysate abundance, and bortezomib significantly increased A2234D lysate abundance, consistent with a role of these degradation processes in Tg PQC (Fig EV10A). When monitoring Tg-NLuc media abundance, neither bafilomycin nor bortezomib significantly altered WT, A2234D, or C1264R abundance (Fig. EV10B). confirming that general inhibition of proteasomal or lysosomal degradation does with rescue mutant Tg secretion."

      __Reviewer #4, Comment #8: __With regard to Figure 7 - Figure supplement 1: The authors should monitor the effect of ML-240 on Tg secretion such that WT and C1264R mutants are directly compared (side-by-side on the same immunoblot). Otherwise, it is difficult to claim that ML-240 rescues the secretion of the mutant.

      Our response: The reviewer is referring to the S35 pulse-chase experiments now shown in Fig EV9. We would like to clarify that these images are not immunoblots but autoradiographs. Even though the samples for WT and C1264R Tg were loaded onto separate gels, the gels were imaged at the same time and are therefore directly comparable. Regardless, the more meaningful information that can be gleaned from these experiments are the absolute rates of protein secretion and degradation and how they change in response to ML-240 treatment. The scale in the quantifications (0 - 100%) is the same and corresponds to the total amount of WT or C1264R Tg that is labeled with 35S during the 30 min pulse. Importantly, we found that C1264R Tg-FT secretion is significantly increased in the presence of ML-240, changing from

      __Reviewer #4, Comment #9: __How did ML-240 affect the ER-phagy components (in particular RTN3) in the TRIP analysis of Tg C1264R (Figure 7G-L)?

      Our response: This is a great discussion point raised by reviewer #4. We have updated the manuscript text to discuss in more detail changes in interactions with degradation components, especially with proteasomal degradation machinery (Fig 7M,N). The manuscript text now reads as follows:

      "The most striking interaction changes occurred with proteasomal degradation components, which remained steady until 1.5 hr, but then abruptly declined with ML-240 treatment at later time points (Fig 7M,N). This decline tracks with changes to the glycan processing machinery, highlighting that the coordination between N-glycosylation and diverting Tg away from ERAD may be a key to the rescue mechanism."

      Minor points:

      __Reviewer #4, Comment #10: __The candidate labeling in Figure 3 - Figure supplement 2 and 3 is too small und unreadable. The authors should provide a higher resolution of these figures or increase the font.

      Our response: These figures are now in the Appendix and we have edited this figure to provide higher resolution.

      Reviewer #4 (Significance (Required)):

      Please see above

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

      Learn more at Review Commons


      Referee #4

      Evidence, reproducibility and clarity

      Summary

      In this manuscript, Wright et al. developed an approach (termed TRIP) that allowed to map the temporal changes in the interaction landscape of a newly synthesized protein of interest. Using their TRIP approach, the authors found that the extensive interactions of thyroglobulin (Tg) with the proteostasis network (PN) during its passage through the secretory pathway were profoundly altered in response to disease-causing mutations (e.g. C1264R). The authors cross-validated their findings with a focus RNAi screen monitoring the cellular and secreted abundance of Tg variants upon deletion of PN components. In subsequent experiments the authors focused on two hits, VCP and TEX264, for which they confirmed their inhibitory effect on the secretion of Tg C1264R. Importantly, the authors found that TEX264 increasingly interacts with the Tg mutant and that pharmacological inhibition of VCP yielded the same phenotype than depletion of VCP. Overall, Wright and colleagues established an elegant method to map protein interaction in a time-resolved manner and demonstrated its value by the analysis of disease-related Tg mutants. Hence, this work has the potential to serve as a rich resource for Tg-related research and as a powerful new tool to examine protein interactions. However, several concerns remain.

      Major points

      1. Overall, the TRIP workflow is quite difficult to understand at a first glance - even for a reader with a background in proteomics, biochemistry and cell biology. The authors may want to improve the description of the TRIP methodology and explain in more detail what the individual components and steps are good for. Along the same line, from the main text and the figure legend it was not clear that Tg was actually Flag-tagged. However, without this information it is difficult to follow the workflow. While Figure 1A is certainly helpful, the bulky graphics are deflecting the reader's attention. A more schematic version might be more informative.
      2. To what extend do the difference in protein abundance between Tg WT and Tg C1264R contribute to the increase binding of their interactors (e.g., HSP5 and PDIA4). The authors should perform a TRIP coupled immunoblot analysis where WT and Mutant are loaded side-by-side on the SDS-PAGE.
      3. While the RNAi screen was done with pooled siRNA, it is not clear what was used for the RNAi validation experiments shown in Figure 5. This should be done by individual siRNA and not the same pooled reagents as used for the screen.
      4. In Figure 5A it is not clear which band was used to quantify the effect of NAPA reduction. Also, this analysis lacks normalization to an unrelated protein or loading control. Moreover, the authors should also examine the effect of the siRNA targets shown in Figure 5C for Tg WT and not only the mutant.
      5. The authors should also test for the binding of RTN3 to Tg WT and mutant - in particular in comparison to TEX264. This would be important in the context that only RTN3 but not TEX264 was detected in the TRIP approach. Do the authors also detect VCP and LC3B in their pulldowns?
      6. The effect of TEX264 depletion on Tg secretion should be confirmed by TEX263 KO experiments. Do the authors observe a similar increase in secreted Tg C1264R in BafA1- or SAR405-treated cells? Moreover, the authors should show that Tg C1264R is actually delivered to lysosomes using biochemical assays such as LysoIP or colocalization experiments.
      7. Figure 7A and 7C lack loading controls. The quantification shown in Figure 7B and 7D should be normalized to this control. Since VCP activity is often coupled to the of the proteasome, the authors should check whether blocking the proteasome yields a similar effect than ML-240.
      8. With regard to Figure 7 - Figure supplement 1: The authors should monitor the effect of ML-240 on Tg secretion such that WT and C1264R mutants are directly compared (side-by-side on the same immunoblot). Otherwise, it is difficult to claim that ML-240 rescues the secretion of the mutant.
      9. How did ML-240 affect the ER-phagy components (in particular RTN3) in the TRIP analysis of Tg C1264R (Figure 7G-L)?

      Minor points

      1. The candidate labeling in Figure 3 - Figure supplement 2 and 3 is too small und unreadable. The authors should provide a higher resolution of these figures or increase the font.

      Significance

      Please see above

    3. 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

      Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate). Please place your comments about significance in section 2.

      In this manuscript, the authors describe their efforts to develop a methodology for determining time-resolved protein-protein interactions using quantitative mass spectrometry. With TRIP (time-resolved interactome profiling), they combine a pulsed bio-orthogonal unnatural amino acid labelling (homopropargylglycine, Hpg), CuAAC conjugation and biotin-streptavidin pulldowns to enrich at different timepoints and time-resolve by combining TMT labelling and LC-MS/MS (Figure 1). This technique is then applied to the maturation of the secreted WT and mutant thyroglobulin (Tg-WT, Tg-C1264R, Tg-A2234D) expressed in HEK293 and rat thyroid cells (FRT) and linked to hyperthyroidism. There, they identify a collection of ER resident proteins involved in protein folding/processing (e.g. chaperones, redox, glycans, hydroxylation) as well as degradation (e.g. autophagy, ERAD/proteasomes) (Fig. 2). Here the authors effectively use pulse-labelled form of TRIPs to highlight the different interactions formed with Tg-WT vs. Tg-mutants during biogenesis and secretion (or retention). The analysis found ~200 new interactions compared to previous studies along with about 40% of those identified previously. Differences in interactions were observed for mutants, which shown extended interaction with chaperones and redox processing pathways. While many interactions appeared as might be expected, the identification of membrane protein processing elements (e.g. EMC, PAT) was puzzling and raised some questions about the specificity within the protocol. Mutants enriched for CANX CALR and UGGT, suggesting prolonged association with glyco-processing factors. Interaction of C1264R with the ER-phagy factors CCPG1 and RTN3 was greater than WT. The authors note that their interaction correlated with that of EMC1 & 4, but it is not clear why that might be.

      With interactors in hand, the authors complemented the TRIP protocol with siRNA KD of identified factors, to investigate any changes to secreted vs intracellular Tg upon loss. KD of NAPA (a-SNAP) and LMAN1 increased WT lysate (intracellular) Tg but not mutants. NAPA also reduced Tg-WT secretion. In contrast, KD of NAPA increased A2234D secretion while LEPRE1 increased C1264R (but not A2234D or WT), suggesting mutants have differential processing paths and requirements. KD of VCP increased secretion of both mutants. Some ER-phagy receptors were found among interactors (e.g. RTN3 in Tg-C1264R only) but often their KD had no impact on secretion (CCPG1, SEC62, FAM134B). NAMA observations were recapitulated in thyroid derived cell line (FRT). KD of TEX264 and VCP increased Tg-C1264 secretion while RTN3 KD in FRTs decreased Tg-C1264 secretion. This was in contrast to data from HEK293s for reasons that are not clear. Co-IP with TEX264 enriched for all Tg forms but more so for C1264R and A2234D - motivating the authors to propose selective targeting of Tg to TEX264 and the consideration of ER-phagy as a "major" degradative pathway during Tg processing.

      Given the observations with siRNAs to VCP, the authors next use a selection of VCP inhibitors to ask whether secretion can be rescued upon pharmacological impairment of the AAA ATPase. They observed that ML-240, but interestingly not the more conventionally used CB-5083 or NMS-873, increased secretion of Tg-C1264R but not lysate. Inhibitors increased lysate but decreased the secreted fraction for Tg-WT (Fig 7). Finally, the authors used TRIP again in ML-240 treated Tg-C1264R expressing cells to look for changes to interactome with treatment - observed decreases to glycan and chaperone interactions, CANX and UGGT1, decreased interaction with DNAJB11 and C10, like that of WT. There was no apparent change to the UPR, although activation was not directly measured.

      Major comments:

      • Are the key conclusions convincing?

      The TRIP methodology appears to be quite robust and should be a powerful strategy for this field and others going forward. The drawback will be the length of pulse required will limit the number/type of proteins to be monitored to ones with longer t1/2's. There were interesting interactions found with Tg and the mutants linked to hyperthyroidism, but cut and dry differences did not appear as obvious, even though strong "trends" appear to be present. The path from identifying interactors in a time-resolved manner to then following them up with targeted KD does provides some clarity, which is important. - Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

      The data regarding VCP silencing and pharmacological impairment appear clear but leave some questions outstanding in this reviewer's opinion. The lack of effect with the 2 highly selective inhibitors suggests that the underlying mechanism for switching fate of intracellularly retained Tg-C1264R towards secreted forms is not at all clear. ML-240 is an early derivative of DBeQ and reportedly impairs both ERAD and autophagic pathways, similarly to DBeQ. The differences between the VCP inhibitors' mechanism of action were not discussed, but perhaps should be elaborated upon, particularly in the matter of how ERAD and ER-phagy pathways might be being differentially affected. At the risk of asking for too many additional experiments, this reviewer would just prefer to see this fleshed out in a bit more detail. - Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

      Q1. The degree (if any) of Tg-C1264 aggregation during and/or detergent solubility do not appear to have been considered as a potential source of the increase in released secreted material (Figure 4, 5). Do Tg mutants partition into RIPA-insoluble fractions at all? That is to say.. is the total population of synthesized Tg being considered? A full accounting? Could the authors address this and if biochemical extraction data (via urea or high SDS) is available, include it to answer this concern.

      Q2. Along the same lines, what does Tg-WT and mutant expression look like by microscopy? Is Tg-WT uniformly distributed while Tg-mutants appear in puncta... more aggregated - perhaps reflecting the increased engagement of chaperones and redox machinery? Changes in the pattern of Tg-C1264R mutant (e.g. w/ VCP KD or inhibition) would add additional support for the authors interpretation of improved secretion. If this data is at hand, including it might be worth consideration.

      Q3. Does the level of Tg mutant expression in the FRT clones impact the profiles obtained by TRIP? (Figure 3). This is a question of gauging the relative saturation of QC machinery and how that might impact profiles from TRIP. Were clones expressing at different levels tested? Perhaps a brief discussion of this.

      Q4. For Figure 3, the hour-long labelling period seems a bit long, compared with 3 hr of chase. Perhaps this reviewer missed this but how long does Tg take to mature and/or mutants to misfold and degrade? Is there any possibility to shorten this so that the profiles of labelled Tg could be more synchronized? If not, perhaps this could just be discussed.

      Q5. It is curious that only ML-240 and not other well characterized inhibitors of VCP/p97, has an effect, as both are used far more often than ML-240. The authors do not really address this in detail but does it suggest that the ML-240 effect on VCP/p97 could be affecting different pathways, given the nature of this compound. Is this compound acting on Tg-C1264R maturation at the level of translation or post-translationally? If the latter, through what means?

      Q6. Continuing from Q5.. At what point and where is VCP/p97 able to affect mutant Tg processing? In line 317, the authors seem to correlate increased VCP association with mutants to their increased secretion. It is not clear how this would result, as engagement with VCP would be in a compartment different to that which supports trafficking and secretion. Could the authors expand on how this might come about. This is also relevant to the ML-240 data in Figure 7. Moreover, VCP is associated with ERAD (as is HerpUD1) rather than ER-phagy and at least in the siRNA raw data, there are also effects from Derlin3 and FAF2 KDs.. both ERAD factors. Some clarity here would be appreciated.

      Q7. There does not appear to be a direct demonstration of Tg-C1264R turnover by ER-phagy (via TEX264). Given the inconsistency with it not being detected by TRIP, while another receptor RTN3 was, but has not impact on Tg-C1264R secretion, perhaps including that data would go some way to demonstrating a fate of ER-phagy (at least partly) for this mutant.

      Q9. The authors provide data that the UPR was not induced by ML-240 at 3hrs (10µM) (Figure 7, supplemental 1). This is in stark contrast to the results of Chou et al (2013) which the authors reference, reporting that ML-240 induced ATF4 and CHOP by 2 hrs at concentrations lower than used here (albeit a different cell type). While not exclusively UPR, could the authors address the potential activation of the integrated stress response (eIF2a phosphorylation, ATF4 and CHOP) in the FRT cells due to ML-240 treatment? If present, is there some link that could this provide an explanation for increased Tg-C1264R secretion? - Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

      Any of the suggested experiments above all use reagents reported in the manuscript and so would presumably incur minimal cost and hopefully time. This reviewer is sympathetic to time and financial constraints and so discussion of the issue could suffice. - Are the data and the methods presented in such a way that they can be reproduced?

      Yes. The methodology is explained in detail. - Are the experiments adequately replicated and statistical analysis adequate?

      Yes. Relevant information is either in the figure legends or is provided in the source data.

      Minor comments:

      • Specific experimental issues that are easily addressable.
      • Are prior studies referenced appropriately?

      The references are generally appropriate, with a few exceptions of more general references used - Are the text and figures clear and accurate?

      The text is clearly written, and the figures are clear. - Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

      A summary figure comparing the changing profiles of WT and C1264R and the factors implicated for them could be helpful.

      Perhaps include common nomenclature for proteins as well (e.g. HSP5A - BiP, HSP90B1 - Grp94, etc..)

      Line 317 - our is misspelled

      Figure 4 - Supplemental Figure 1 - Legend has text referring to panels J and K, but Figure only goes up to F.

      Significance

      • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.
      • Place the work in the context of the existing literature (provide references, where appropriate).

      Protein-protein interactions are often used to illustrate complexes and functionality, but these provide only snapshots, rather than "movies". There are many datasets out there exploring P-P interactions, but most if not all lack any temporal resolution for the interactions they report. The TRIP method described approaches this from the dynamic perspective - identifying the transient interactions formed by folding nascent chains with proteins that aid in their maturation and trafficking, or degradation. This represents an important technical advance in our ability to dynamically monitor protein interactions. The use of Tg mutants is valuable and perhaps this will lead to new perspectives on how to rescue it or other pathophysiological mutants with loss of function phenotypes.<br /> - State what audience might be interested in and influenced by the reported findings.

      This work should appeal to a broad audience within cell biology, particularly as the TRIP technique is attempting to address a fundamental question - what interactions form during the biogenesis/lifetime of a protein. Moreover, the effort to try to understand the different interactions formed with pathologically relevant mutant proteins as a strategy to try to rescue functionality, is a valuable exercise of this approach. - 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.

      ER quality control

    4. 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 'Time-Resolved Interactome Profiling Deconvolutes Secretory Protein Quality Control Dynamics' Wright et al. developed an approach for time-resolved protein protein interaction mapping relying on pulsed unnatural amino acid incorporation, protein cross linking, sequential affinity purification, and quantitative mass spectrometry named time-resolved interactome profiling (TRIP). The authors applied the TRIP method to compare the interactions of the secreted thyroid prohormone thyroglobulin (Tg) comparing the WT protein to secretion-defective mutations implicated in congenital hypothyroidism. They further employed an RNA interference screening platform (1) to investigate if (1) interactors identified via TRIP are functionally relevant for Tg protein quality control and (2) to identify factors that can rescue mutant Tg secretion. The screen was initially performed in HEK293 cells, but selected hits with a phenotype in HEK cells were then followed up in Fisher rat thyroid cells. Further functional validation was performed by pharmacologic inhibition of VCP, a hit from the RNAi screen with an effect on Tg lysate abundance and Tg secretion. While the authors present a comprehensive study including identification of protein-protein interactions using proteomics followed up by an RNA interference screen for functional validation, major comments need to be addressed for both the proteomics as well as the functional genomics aspects of the study (see comments below).

      Major comments:

      • The authors describe a new method for quantitative, temporal interaction mapping. The protocol involves two enrichment steps as well as several reactions including cross-linking of the samples as well as functionalization of the unnatural amino acids. Given all these steps, the authors should rigorously characterize the quantitative reproducibility of the experiment when performed in independent biological replicates. This is important because in the final quantitative MS experiment, the authors only use two biological replicates, which is too low especially for such an involved sample preparation procedure, which would expect to have a high variability between replicates. Given the low number of replicates and the unknown reproducibility of the quantification for this protocol, it is questionable at this point how reliable the quantification over the time course is.
      • Compared to the previous dataset published last year, the authors discover an overlap in interactors, but also a huge discrepancy, with 96 previously identified interactors not detected in the current study, but 198 additional interactors identified. How do the authors explain the big differences between these datasets?
      • For the temporal interaction analysis the authors describe differences in the temporal profiles of selected interactions comparing wt and mutant, however no statistical analysis is performed comparing wt and mutant interaction profiles across the time course. Furthermore the variability between the replicates for the temporal profiles is not shown and some of the temporal profiles appear to be noisy. A more rigorous statistical analysis should be performed including additional biological replicates to evaluate the changes over the time course, especially as the temporal interaction analysis is the novelty of this study.
      • To functionally validate interactors derived from the TRIP analysis as well as to identify factors that can rescue mutant Tg secretion the authors developed an RNA interference screen. There are a number of aspects that need to be addressed/clarified for this part of the study.
      • While the authors validate the stable cell lines expressing the nanoluciferase tagged Tg and the linearity of luminescence signal in lysate and media carefully, they do not validate their platform in combination with the RNAi knockdown strategy. The authors should select genes as positive controls that are expected to modulate Tg secretion and demonstrate that the knockout of these positive controls indeed results in changes in Tg secretion in their system.
      • For the screen the authors select 167 Tg interactors and PN (Proteostasis network) related factors. This statement is very vague and the authors should clarify which genes were knocked down and which criteria were applied to narrow down the list of interactors and to select PN factors. The authors should therefore provide a supplementary table including all genes included in the screen, their source (were this derived from the initial study by Wright et al, from the current study or compiled from prior knowledge about PN), as well as their results from the screen based on luminescence in media and lysate. It is unclear how many of the selected factors are actually coming from the TRIP analysis.
      • Only a small number of the 167 selected genes shows an effect on Tg abundance/secretion. How do the authors explain this result? Would we not expect that Tg interactors, especially those from the TRIP method which interact with the newly synthesized are more enriched for functionally relevant genes.
      • The authors initially performed the screen in HEK293 cells and as a second step wanted to validate the hits from the HEK cells in more relevant Fisher rat thyroid cells. Indeed they could show that knockdown of NAPA increased WT TG in lysate and decreased WT Tg secretion. Furthermore, they further validated genes to modulate mutant Tg lysate and media abundance. The authors should perform a rescue experiment to demonstrate that the observed phenotype can be reversed through re-introduction of NAPA.
      • One hit from this analysis was the ER-phagy receptor TEX264, while TEX264 was not identified in the TRIP data, is selectively increased the C1264R secretion, but not wt and the other Tg mutant. Following Co-IP data however revealed some interaction between the C1264R and to a lesser extent the A2234D mutant. How do the authors explain that TEX264 was missed in the TRIP dataset?

      Minor comments:

      • The workflow needs to be described clearer. For example, it should be better explained why the authors selected a two stage enrichment strategy, I assume that the first based on the Flag affinity tag is to purify the protein of interest and the second step based on the incorporation and functionalization of the unnatural amino acids to enrich for the newly synthesized fraction at specific time points after protein synthesis? These are critical steps for the method but the rationals are not well explained, neither in the text nor the figures captures all these steps of the method very clearly, which makes it really difficult for the reader to understand the individual steps of the method. Moreover, the structures in Figure 1 workflow are not clearly labeled, so that it is confusing which part represents which protein/molecule.
      • Except for the general workflow shown in Figure 1, a more detailed workflow showing the experimental steps, such as the sample fractions with the following steps could be added so that the design of the method is clearer. Also the style of the workflows including Figure 1, Figure 2A, and Figure 3A are different. It would be helpful to make them the same style and make the Figure 2A as a zoom in or more detailed illustration on part of Figure 1.
      • A summary of proteomics results of time course labeling after all enrichment steps, including the total number of identified proteins at different conditions and control would be helpful for having an overview impression on the proteomics results
      • In Figure 2B, the WB for PDIA4 in the Biotin PD elution is missing. Why was the PDIA4 interaction missing for the time course analysis, but the interaction was captured in the initial test for Wt Tg (Figure 1D). Additionally, in this panel the Rhodamine Probe Gel shows inconsistencies at the time points 1.5 - 3h. Does this mean that the labeling did not work well for these conditions? As we would expect a consistent Rhodamine Probe signal at every time point.
      • In Figure 2, why was there no WB results for the A2234D? In Figure 2D and 2E, at which time point are the changes significant compared to WT?
      • All figure legends should indicate how many biological replicates were performed for each experiment represented in the figure.
      • The heatmaps shown in Figure 3, Figure 3 - Figure Supplement 3, and Figure 7 are in the current form incomprehensible. The heatmaps depict the relative enrichment vs the control sample, which was scaled between 1 and -1. The color coding with 5 different colors from 1 to -1 is very confusing and should be changed to just two colors, one for positive and one for negative relative enrichment. I would also suggest changing the visualization of the heatmap showing the wt and mutants side by side, instead of stacked on top of each other for each individual protein.
      • The data analysis method for generating relative enrichment shown in the heatmap is not explained. This should be described in the method section for a better understanding of the data analysis. There are no legends of flowcharts in Figure 2A and Figure 3A and it is difficult to understand which are the key components in the complex and what are the differences among different periods of labeling.
      • Why did only one of the VCP inhibitors (ML-240) exhibit a phenotype in Tg abundance and secretion, but not the other VCP inhibitors?

      Significance

      The authors describe a novel and elegant method to map time resolved protein interactions of newly synthesized proteins, which allows monitoring of proteins regulating protein quality control. Authors describe it as a general method, however, they only demonstrate the applicability to one protein and do not systematically evaluate the quantitative nature of their approach by determining quantitative reproducibility, which would be necessary to be able to claim that this is a method with broad applicability.

      Given my expertise in quantitative proteomics, I can mainly comment on the technological aspects of the proteomics part of the manuscript, but do not feel qualified to evaluate the significance of this study in terms of novel biology. Nevertheless, it feels that there is a stronger emphasis on the biology in the current form of the manuscript which will raise interest of scientists with a focus on protein quality control and Tg biology.

    5. 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 report a mass spectrometry (MS)-based interactomics technique, time-resolved interactome profiling (TRIP), which allows for tracking temporal changes in the interactome of protein of interest. To show that TRIP can successfully deconvolute interactomes over time, they pulsed thyroid cells with homopropargylglycine (Hpg), immunoprecipitated the Hpg incorporated thyroglobulin (Tg) and its interacting proteins at different time points, and subjected the samples to tandem mass tag (TMT)-based quantitative MS analysis. The MS results show that WT and variant Tg proteins indeed associate with different proteostasis network factors in a differential manner over the course of time. In addition, they utilized an siRNA-based luciferase fusion assay to evaluate whether silencing each proteostasis network component changes the levels of Tg in both lysate and media. From the combination of the TRIP and siRNA-based assays, they found many hits, including hits implicated in protein degradation, VCP and TEX264, which they validated with multiple experiments.

      I am overall quite positive and think this is an important study. But there are some meaningful points to consider.

      Significant comments:

      1. Only two replicates of the main data (the TRIP-MS experiments) for this paper is problematic. Especially since the manuscript is supposed to be demonstrating and validating the new technique. Consistent with this concern, the relative enrichment profiles for some of the results were surprising. For instance, interaction with CCDC47 was tapering off but then at 3 h it suddenly reaches the maximum level of engagement. Is this a real finding or the variability in the method? Impossible to tell with two replicates. Presenting heat maps based on biological duplicates is also very problematic. It masks the error, which is large as can be seen in some of the panels showing individual proteins. In my view, triplicates and a clear understanding of the error in the technique should be required.
      2. The same concern arises for the high-throughput siRNA screen, which was performed only in duplicate for WT and A2234D.
      3. There are issues with some of the immunoprecipitation experiments: In Figure 1C, a negative control for FLAG IP is missing. In Figure 2B, I am curious why the band (Hpg -, chase time 0 h) is so faint for the first WB (IB for FLAG) - is Hpg treatment indeed leading to much more Tg present at 0 h? If so, that is a concern. Also, a negative control must be included (either plain cells or cells expressing fluorescent protein or a different epitope-tagged WT Tg). In this same figure, I am puzzled why the bands for 1.5-3 timepoints in Biotin PD elution, probed for Rhodamine, are very faint especially considering that in Figure 1D, the corresponding bands, which are 4 h after the pulse, look fine. It seems like the IP failed here?

      Suggestion to consider:

      This manuscript, supported by the title and abstract, mainly focuses on the presentation of the development and application of TRIP, which is highly significant. The story becomes less coherent and harder to follow as significant amounts of text/figures are dedicated to siRNA-based high throughput screening and follow-up. In addition, although the discovery of TEX264 as one of the hits is very interesting and exciting, TEX264 apparently was not a hit in the TRIP experiment and is pretty distracting from the main point of the paper highlighted in the abstract and title, therefore. The siRNA-based assay and follow-up studies could be a separate scientific story of their own. Especially considering my concerns on the number of replicates for both the TRIP and siRNA-based assay, it could be beneficial to actually split the manuscript into two and conduct more replicates of the -omic work, which should corroborate the exciting discoveries the authors have made.

      Minor comments:

      Throughout the manuscript, the authors have not defined what FT is; presumably it means FLAG tag.

      The authors might discuss their rationale for choosing 0-3 hrs for their TRIP studies. That includes any relevant information about the half-life of WT versus variant Tg, whether the Hpg pulse time is short enough to avoid missing key features of the temporal interactome, and discussion of what would happen if the TRIP were performed at prolonged time points (e.g. 6-10 h).

      Lines 68-69: the two citations should probably come one sentence earlier (at least Coscia et al 2020 is a structure paper).

      Line 91: "(Figure 1A)" should follow the sentence "To develop the time-resolved..." to help readers better understand the system.

      Line 101: Fisher should be Fischer

      Line 131: Should be 1.5 hrs instead of 2 hrs.

      Lines 135-136: I do not agree with the claim that HSPA5 profile looked similar for MS and WB. I do not see a peak for HSPA5 at 2 hrs in Figure 2D.

      Line 186: The cited paper Shurtleff et al 2018 is missing in the reference list.

      Line 188: I disagree with the authors' claim here because, at least for CCDC47, interactions with C1264R seem to come back at the 3 hr time point.

      Line 203: I am not sure if P4HA1 can be included in the examples for showing distinct patterns for mutants compared to the WT according to their data in Figure 3H.

      Line 216: The authors should add citations about the functions of STT3A and STT3B proteins.

      Lines 248-251, "We found that interactions with these components...": this sentence should refer to Figure 3 - Figure Supplement 3 instead of Figure 3L and S4.

      Lines 258-260, "Another striking observation was that the temporal profile of EMC interactions for C1264R correlated with RTN3, PGRMC1, CTSB, and CTSD interactions.": Please provide more evidence to support the potential correlation between different interaction profiles. Or the authors should move this sentence to the discussion section as it sounds speculative. This highlights the issue of only having duplicates, as well.

      Line 340: As written, should cite more than one paper

      Line 371: Should be Figure 4 - figure supplement 2

      Line 1231: "Zhang et al 2018" needs to be removed

      Line 1286: FRTR should be FRT

      Figure 3E: Color used to highlight the three proteins (CCDC47, EMC1, EMC4) should match the color used in Figure 3 - Figure Supplement 3

      Figure 4A: The bottom figure where lysate signal is inversely proportional to time is misleading because the authors are assessing steady-state level of proteins in this assay.

      Figure 4 - Figure Supplement 1 caption: in (C), (F) should be (B). (K) should be (G) and I am not sure what the authors mean when they refer to (J) in caption of (G).

      Figure 5 caption for (C and D): Need to specify the time that the samples were collected (8 hrs), as it seems different from A and B according to the main text.

      Figure 5 - Figure Supplement 1: Data for HERPUD1 and P3H1 should be included.

      Figure 5 - Figure Supplement 2B: Please mention in the caption how degradation is defined.

      Significance

      This manuscript is highly significant because the authors (1) designed and validated a new methodology for time-resolved interactomics study, (2) presented the dynamic changes in Tg interactome for WT and variants, and (3) discovered how proteins implicated in degradation pathways (e.g. VCP, TEX264, RTN3) can change the secretion profile of WT and mutant Tg proteins. With TRIP, the authors demonstrated that they could obtain valuable data that were previously not captured from steady-state interactomics studies (Wright et al. 2021; Figure 3M and Figure 3 - Figure supplement 4D-4I). Furthermore, the authors treated cells with VCP inhibitors and performed both 35S pulse-chase analyses and TRIP. These experiments provide valuable information to the field by (1) presenting a new method to rescue Tg secretion defect, and (2) demonstrating a broader applicability of TRIP. If the major comments above can be addressed I believe this is a tremendous contribution to the field.

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

      Learn more at Review Commons


      Reply to the reviewers

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

      This manuscript investigates the dynamics of GC-content patterns in the 5'end of the transcription start sites (TSS) of protein-coding genes (pc-genes). The manuscript introduces a quite careful and comprehensive analysis of GC content in pc-genes in humans and other vertebrates, specially around the TSS. The result of this investigation states that "GC-content surrounding the TSS is largely influenced by patterns of recombination." (from end of Introduction)

      My main concern with this manuscript is one of causal reasoning, whether intended or not. I hope the authors can follow my reasoning bellow on how the logic sometimes seems to fail, and that they introduce changes to clarify their suggested mechanisms of action.

      The above quoted sentence form the end of the Intro is in conflict with this other sentence that appears at the end of the Abstract "the dynamics of GC-content in mammals are largely shaped by patterns of recombination". The sentence in the Intro seems to indicate that the effect is specific to TSSs, but the one in the abstract seem to indicate the opposite, that is, that the effect is ubiquitous.

      We are sorry about the lack of clarity. We have now rewritten the abstract and intro to emphasize that our results are restricted to the 5' end of genes, and that by "patterns of recombination" we mean "historic patterns of recombination".

      The observations as stated in the abstract are: "We observe that in primates and rodents, where recombination is directed away from TSSs by PRDM9, GC-content at protein-coding gene TSSs is currently undergoing mutational decay."

      If I understand the measurements described in the manuscript correctly, and the arguments around them, you seem to show that the mutational decay of GC-content in humans is independent of location (TSSS or not), as noted here (also from the abstract) "These patterns extend into the open reading frame affecting protein-coding regions, and we show that changes in GC-content due to recombination affect synonymous codon position choices at the start of the open reading frame."

      Again, we have rewritten this section to clarify these points.

      There is one more result described in the manuscript, that in my mind is very important, but it is not given the relevance that it appears to me that it has. That is presented in Figure S3G. "we concluded that GC-content at the TSS of protein-coding genes is not at equilibrium, but in decay in primates and rodents. This decay rate is similar to the decay seen in intergenic regions that have the same GC-content (Figure S3G)"

      Thus, if the decaying effect happens everywhere, how can it be related to "recombination being directed away from TSSs by PRDM9" as it is stated in the abstract and in the model described in Figure 7?

      We make the argument that the GC-peak as likely caused by past recombination events. This is based on:

      1) The change in GC-content at the TSS in Dogs and Fox, coupled to the fact that they perform recombination at the TSS

      2) That the TSS can act as a default recombination site in mice when PRDM9 is knocked out

      3) That some forms of PRDM9 allow for recombination at TSS (see Schield et al., 2020, Hoge et al. 2023, and Joseph et al., 2023) and that this is expected to cause an increase in GC-content

      We thus speculate that the GC-peak in humans and rodents was caused by past recombination at TSSs that were permitted by ancient variants of PRDM9. We further point out that PRDM9 is undergoing rapid evolution, and some of the past versions of the protein may have had this property.

      We have tried to clarify these points in the latest version of the text.

      The fact that the decay rate is similar to any other region with similar GC-content should be an indication that the effect is not related to anything having to do with TSS or recombination being directed away from TSSs by PRDM9.

      We are sorry about the lack of clarity. TSSs in humans, chimpanzees, mouse and rats are are experiencing GC-decay at the same rate as in non-functional DNA regions with high GC-content. Thus the GC-peak is not being maintained by selection. This is surprising, given the role that GC-content plays in gene expression. This is a critical point, and we added it to the "conclusion" section of the abstract.

      I hope these paragraphs show my confusion about the relationship between the results presented which I think are very comprehensive and their interpretation and suggested model for GC-content dynamics around TSSs in human.

      On another note, can you provided a bit more background on recombination and its mechanisms?

      We have done our best to clarify these issues.

      You seem to have confident sets of genes under high/low/med recombination. How are those determined.

      We used the recombination rates per gene provided in Pouyet et al 2017 to identify the sets of genes under low/med/high recombination. Those rates were estimated from the HapMap genetic map (Frazer et al., 2007). This is now all specified in the methods section.

      You also seem to concentrate the cause of recombination on PRDM9, please explain. Is PRDM9 the unique indicator of recombination?

      PRDM9 has been shown to be the primary determinant of where recombination occurs in the genome (Grey et al., 2011, Brick et al., 2012). This is very well established. We now reword some of the introduction to make this clear.

      specific comments


      Figure 1, it is very hard to understand the differences between the three rows. Please explain more clearly in the legend, and add more information to the figure itself.

      We altered the axis titles to make this clearer. We also label "Upsream", "Exon 1" and "Part of Intron 1" in Figure 1C, F and I, and in Figure 2C. We now spell this out in the Figure Legend.

      Figure 7, express somewhere in the figure that the y axis measures GC content.

      We now added "GC Content" to the left of the first "graph" in Figure 7.

      Figure seems to introduce a 'causal' model of GC-content dismissing (diminishing?) based on recombination being directed away from TSSs. How about the diminishing of GC-content on any other genomic regions as you have shown in Figure S3G?

      Our focus in this model, and manuscript, is on TSSs. I think that to add the dynamics of other GC-rich regions is distracting. We do not know what caused these intergenic genomic regions to be high in GC-content prior to decay. After excluding known recombination sites and TSSs, these regions are very rare in the human genome. They may be ancient recombination sites that are decaying in GC-content. However, unlike TSSs, which have some connection to recombination (i.e. data from PRDM9 knockout mice and dogs and fox), we do not have any direct or indirect evidence that these other sites were used for recombination in the past. Alternatively, there could have been some other pressure on these sites in the past to increase GC-content that we are not aware of.

      -- The title is too selective, as to the results, and it has the implication that the decay is exclusive to the surrounding of the TSSs.

      Decay of GC-content towards equilibrium is the default state for non-functional DNA. That it is occurring at the TSS is surprising, as it indicates that the GC-peak is not maintained by selection. We now state this in the paper and include this in the "conclusion" portion of the abstract.

      Reviewer #1 (Significance (Required)):

      The statistical analysis is comprehensive and robust.

      We thank the reviewer for this.

      Their model interpretation as is describe induces confusion and needs to be clarified.

      We are sorry about this. Hopefully our revised text will clear up the confusion.

      I am an expert computational biologist, I do not have a deep knowledge of sequence implications of recombination, and it would be good if the manuscript could add some more background on that.

      We thank the reviewer for their perspective, and we hope that our text changes better explain to the non-expert why our findings are so surprising. We further clarify how recombination affects DNA sequence by gBGC and some of these changes are detailed in our response to the other reviewers.

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

      In this work, the author present various analyses suggesting that GC-content in TSS of coding genes is affected by recombination. The article findings are interesting and novel and are important to our understanding of how various non-adaptive evolutionary forces shape vertebrate genome evolutionary history.

      We thank the reviewer for these kind words.

      The Methods section includes most needed details (see comments below for missing information), and the scripts and data provided online help in transparency and usability of these analyses.

      I have several comments, mostly regarding clarifications in the text and several suggestions:

      1. In introduction: CpG islands, have been shown to activate transcription (Fenouil et al., 2012) - what is known about CpG Islands is somewhat inaccurately described. It should be rephrased more accurately, e.g. - CpG Islands found near TSS are associated with robust and high expression level of genes, including genes expressed in many tissues, such as housekeeping genes.

      We thank the reviewer for that. We have rewrote this part of the introduction.

      1. The following claim (in Introduction), regarding retrogenes and their GC content is not in agreement recent analyses: "Indeed, it has been observed that these genes have elevated GC-content at their 5' ends in comparison to their intron-containing counterparts, suggesting that elevation of GC-content can be driven by positive selection to drive their efficient export (Mordstein et al., 2020). Moreover, retrogenes tend to arise from parental genes that have high GC-content at their 5'ends (Kaessmann et al.,2009)." Recent work showed that retrogenes in mouse and human are significantly depleted of CpG islands in their promoters (PMID: 37055747). This follows the notion that young genes, such as these retrogenes, have simple promoters (PMID: 30395322) with few TF binding sites and without CpGs. The two reported trends should be both mentioned with some suggestions regarding why they seem to be contrasting each other and how they can be reconciled.

      We thank the reviewer for this information. The previous report (Mordstein et al., 2020) indicated that the increase in GC-content occurs downstream of the TSS in retrogenes. Since sequences upstream of the TSS are not part of the retro-insertion, it is not surprising that GC-content may differ between the retrogene and the parental gene. That retrogenes have lower numbers of CpGs upstream of the TSS, bolsters the idea that GC-content is not required for transcription and that the GC-peak is not being maintained in most genes by purging selection.

      1. In "Thus GC-content is expected, and is indeed observed to be higher near recombination hotspots due to gBGC (REF)." I think you forgot the reference...

      We thank the reviewer for catching this.

      1. In Results, regarding average GC content (Fig 2X): "Interestingly, this pattern is different in the nonamniotes examined, including anole lizard, coelacanth, shark and lamprey." - in lizard, it seems that the genomic average is lower (and lizards are amniotes)

      You are absolutely right. We now fix this.

      1. In Discussion, the statement: "This model is supported by findings in a recent preprint, which documents the equilibrium state of GC-content in TSS regions from numerous organisms" seems to contrast with the findings of the mentioned preprint. If "most mammals have a high GC-content equilibrium state" but still have a functional PRDM9, in the lack of evidence for functional differences between ortholog PRDM9 proteins (such as signatures for positive selection or functional assays), the authors' findings regarding the relationship between a lack of PRDM9 in canids and the trends observed in their TSS, are weakened.

      We are sorry about the confusion. We were not exactly sure what points were being commented on. 1) whether GC-content is at equilibrium for most mammals or 2) that the equilibrium state is high for most mammals despite containing PRDM9. We rewrote this sentence to clarify both issues (especially given that these concepts may not be clear to non-experts, such as the first reviewer). To answer the first potential concern, the paper in question (Joseph et al., 2023), does not show that GC-content at the TSS in mammals is at equilibrium, rather, it calculates what the equilibrium state is given the nucleotide substitution rates. In most organisms, the TSS is not at equilibrium. To answer both 1 and 2, Joseph et al., show that the equilibrium GC-content at the TSS for canids is much higher than for other mammals. They and others infer that the diversity between other mammals (where the equilibrium state is higher than humans and rodents but lower than canids) has to do with the variation between PRDM9 orthologues, however this has yet to be tested. Although the action of PRDM9 has not been evaluated in most mammals, we do point out that in snakes PRDM9 allows for some recombination at the TSS.

      1. In Methods, the ENSEMBL version (in addition of the per-species genome version) should be mentioned.

      This has been fixed.

      1. In Fig 1, it is worth clarifying in the legend that the differences between the first and second rows of panels is in the length of the plotted region.

      We have now indicated this in the figure legend.

      Reviewer #2 (Significance (Required)):

      The manuscript provides a rigorous analysis of the possible processes that have impacted the TSS GC-content during evolution. It should be of interest to a diverse set of investigators in the genomics community, since it touches on different topics including genome evolution, transcription and gene structures.

      Thank you.

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

      This study analyzes the distribution of GC-content along genes in humans and vertebrates, and particularly the higher GC-content in the 5'-end than in the 3'-end of genes. The results suggest that this pattern is ancient in vertebrates, currently decaying in mouse and humans, and probably driven by recombination and GC-biased gene conversion. It is proposed that the 5'-3' gradient was generated during evolution when PRDM9 was less active (in which case recombination occurs mostly near transcription start sites), and decays when PRDM9 is very active, as it is currently in humans and mouse. This is a very interesting hypothesis, also corroborated by a recent, similar analysis in mammals (Joseph et al. 2023). These two preprints, which appeared around the same time, are, I think, quite novel and important. The analyses performed here are thorough and convincing. Source code and raw data sets are openly distributed. I only have a couple of minor comments and suggestions, which I hope might help improve the manuscript.

      Thank you very much for the kind words.

      A1. There has been quite some work on the 5'-3' GC-content gradient in plants (e.g. Clément et al. 2014 GBE, Ressayre et al. 2015 GBE, Brazier & Glemin 2023 biorxiv), which you might like to cite.

      Thank you for pointing out these very interesting papers, we have incorporated them into the latest version.

      A2. CpG-content and GC-content are related in various ways (e.g. see Galtier & Duret 2000 MBE, Fryxell & Moon 2005 MBE) that you might like to discuss; currently the manuscript discusses the CpG hypermutation rate as a driver of GC-content but the picture might be a bit more complex.

      Thank you for this, we have incorporated these citations.

      A3. The model introduced by this manuscript (figure 7) is dependent on the evolution of recombination determination in vertebrates and the role of PRDM9. A recent preprint by Raynaud et al (biorxiv) seems relevant to this issue.

      Thank you for pointing out this pre-print. We have added a paragraph to the discussion that mentions this work. This also initiated a conversation with the authors, and we include some "personal communications" that illuminate what is going on in teleost fish.

      Line-by-line comments

      B1. "First, highly spliced mRNAs tend to have high GC-content at their 5' ends despite the fact that it is not required for export and does not affect expression levels (Mordstein et al., 2020)" -> I do not totally understand this sentence, which seems to imply some link between splicing and export/expression, could you please clarify?

      We rewrote that sentence to make it clearer.

      B2. "mismatches will form in the heteroduplex which are typically corrected in favor of Gs and Cs over As and Ts by about 70%" -> This 70% figure is human-specific, and varies a lot among species; I know in this introduction you're mainly reviewing the human literature but since this part of the text introduces gBGC as a process maybe clarify by adding "in humans" or refrain from giving this figure?

      Thank you. This is a good point. We fixed this.

      B3. "Thus GC-content is expected, and is indeed observed to be higher near recombination hotspots due to gBGC (REF)." -> reference missing here; actually I'm not sure you will find a good reference for this because PRDM9-dependent hotspots are so short-lived that GC-content would only respond weakly; mayber rather refer to the equilibrium GC-content (and cite, for instance, Pratto et al 2014 Science), or to high-recombining regions instead of hotspots (and you have plenty of papers to cite)?

      Thanks for this.

      B4. Paragraph starting: "PRDM9 and recombination hotspots also experience accelerated rates of evolution..." -> I would suggest removing the word "also" and moving this paragraph up, just before the sentence I'm commenting above (the one starting "Thus GC-content..."). This will justify my suggestion in comment B3 of mentioning high-recombining regions instead of hotspots, while also avoiding to have the important paragraph on recombination at TSS (the one starting "There are interesting connections...") being sandwiched between two sections on PRDM9.

      We did not move this paragraph, although we did adjust the wording slightly.

      B5. Paragraph starting "There are interesting connections..." is crucial to your discussion and might be emphasized a bit more in introduction, in my opinion. For instance, what about adding a sentence like "Also not directly relevant to humans, these observations suggest that gBGC might have played a role in shaping the observed 5'-3' GC-content gradient."

      We did not alter the structure of this paragraph but we did reword sections of it.

      1. "Interestingly, this pattern is different in the non-amniotes examined, including anole lizard, coelacanth, shark and lamprey. These organisms had clear differences in GC-content between their first exon and surrounding sequences (upstream and intronic sequences), which came close to the overall genomic GC-content." -> I'm not sure I got the point the authors are intending to make here. Also please note that lizards are amniotes.

      We thank the reviewer for catching this error, we have fixed this.

      Reviewer #3 (Significance (Required)):

      This is one of two preprints having appeared ~at the same time (the other one being the cited Joseph et al 2023), which I think are quite important and convincing regarding the role of PRDM9-dependent and PRDM9-independent recombination on GC-content evolution in vertebrates. I support publication of this preprint in a molecular evolutionary journal.

      We thank the reviewer for their kind assessment!

    2. 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 analyzes the distribution of GC-content along genes in humans and vertebrates, and particularly the higher GC-content in the 5'-end than in the 3'-end of genes. The results suggest that this pattern is ancient in vertebrates, currently decaying in mouse and humans, and probably driven by recombination and GC-biased gene conversion. It is proposed that the 5'-3' gradient hass generated during evolution when PRDM9 was less active (in which case recombination occurs mostly near transcription start sites), and decays when PRDM9 is very active, as it is currently in humans and mouse. This is a very interesting hypothesis, also corroborated by a recent, similar analysis in mammals (Joseph et al. 2023). These two preprints, which appeared around the same time, are, I think, quite novel and important. The analyses performed here are thorough and convincing. Source code and raw data sets are openly distributed. I only have a couple of minor comments and suggestions, which I hope might help improve the manuscript.

      A1. There has been quite some work on the 5'-3' GC-content gradient in plants (e.g. Clément et al. 2014 GBE, Ressayre et al. 2015 GBE, Brazier & Glemin 2023 biorxiv), which you might like to cite.

      A2. CpG-content and GC-content are related in various ways (e.g. see Galtier & Duret 2000 MBE, Fryxell & Moon 2005 MBE) that you might like to discuss; currently the manuscript discusses the CpG hypermutation rate as a driver of GC-content but the picture might be a bit more complex.

      A3. The model introduced by this manuscript (figure 7) is dependent on the evolution of recombination determination in vertebrates and the role of PRDM9. A recent preprint by Raynaud et al (biorxiv) seems relevant to this issue.

      Line-by-line comments

      B1. "First, highly spliced mRNAs tend to have high GC-content at their 5' ends despite the fact that it is not required for export and does not affect expression levels (Mordstein et al., 2020)" -> I do not totally understand this sentence, which seems to imply some link between splicing and export/expression, could you please clarify?

      B2. "mismatches will form in the heteroduplex which are typically corrected in favor of Gs and Cs over As and Ts by about 70%" -> This 70% figure is human-specific, and varies a lot among species; I know in this introduction you're mainly reviewing the human literature but since since this part of the text introduces gBGC as a process maybe clarify by adding "in humans" or refrain from giving this figure?

      B3. "Thus GC-content is expected, and is indeed observed to be higher near recombination hotspots due to gBGC (REF)." -> reference missing here; actually I'm not sure you will find a good reference for this because PRDM9-dependent hotspots are so short-lived that GC-content would only respond weakly; mayber rather refer to the equilibrium GC-content (and cite, for instance, Pratto et al 2014 Science), or to high-recombining regions instead of hotspots (and you have plenty of papers to cite)?

      B4. Paragraph starting: "PRDM9 and recombination hotspots also experience accelerated rates of evolution..." -> I would suggest removing the word "also" and moving this paragraph up, just before the sentence I'm commenting above (the one starting "Thus GC-content..."). This will justify my suggestion in comment B3 of mentioning high-recombining regions instead of hotspots, while also avoiding to have the important paragraph on recombination at TSS (the one starting "There are interesting connections...") being sandwiched between two sections on PRDM9.

      B5. Paragraph starting "There are interesting connections..." is crucial to your discussion and might be emphasized a bit more in introduction, in my opinion. For instance, what about adding a sentence like "Also not directly relevant to humans, these observations suggest that gBGC might have played a role in shaping the observed 5'-3' GC-content gradient."

      1. "Interestingly, this pattern is different in the non-amniotes examined, including anole lizard, coelacanth, shark and lamprey. These organisms had clear differences in GC-content between their first exon and surrounding sequences (upstream and intronic sequences), which came close to the overall genomic GC-content." -> I'm not sure I got the point the authors are intending to make here. Also please note that lizards are amniotes.

      Significance

      This is one of two preprints having appeared ~at the same time (the other one being the cited Joseph et al 2023), which I think are quite important and convincing regarding the role of PRDM9-dependent and PRDM9-independent recombination on GC-content evolution in vertebrates. I support publication of this preprint in a molecular evolutionary journal.

    3. 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 work, the author present various analyses suggesting that GC-content in TSS of coding genes is affected by recombination. The article findings are interesting and novel and are important to our understanding of how various non-adaptive evolutionary forces shape vertebrate genome evolutionary history.

      The Methods section includes most needed details (see comments below for missing information), and the scripts and data provided online help in transparency and usability of these analyses.

      I have several comments, mostly regarding clarifications in the text and several suggestions:

      1. In introduction: CpG islands, have been shown to activate transcription (Fenouil et al., 2012) - what is known about CpG Islands is somewhat inaccurately described. It should be rephrased more accurately, e.g. - CpG Islands found near TSS are associated with robust and high expression level of genes, including genes expressed in many tissues, such as housekeeping genes.
      2. The following claim (in Introduction), regarding retrogenes and their GC content is not in agreement recent analyses: "Indeed, it has been observed that these genes have elevated GC-content at their 5' ends in comparison to their intron-containing counterparts, suggesting that elevation of GC-content can be driven by positive selection to drive their efficient export (Mordstein et al., 2020). Moreover, retrogenes tend to arise from parental genes that have high GC-content at their 5'ends (Kaessmann et al.,2009)." Recent work showed that retrogenes in mouse and human are significantly depleted of CpG islands in their promoters (PMID: 37055747). This follows the notion that young genes, such as these retrogenes, have simple promoters (PMID: 30395322) with few TF binding sites and without CpGs. <br /> The two reported trends should be both mentioned with some suggestions regarding why they seem to be contrasting each other and how they can be reconciled.
      3. In "Thus GC-content is expected, and is indeed observed to be higher near recombination hotspots due to gBGC (REF)." I think you forgot the reference...
      4. In Results, regarding average GC content (Fig 2X): "Interestingly, this pattern is different in the nonamniotes examined, including anole lizard, coelacanth, shark and lamprey." - in lizard, it seems that the genomic average is lower (and lizards are amniotes)
      5. In Discussion, the statement: "This model is supported by findings in a recent preprint, which documents the equilibrium state of GC-content in TSS regions from numerous organisms" seems to contrast with the findings of the mentioned preprint. If "most mammals have a high GC-content equilibrium state" but still have a functional PRDM9, in the lack of evidence for functional differences between ortholog PRDM9 proteins (such as signatures for positive selection or functional assays), the authors' findings regarding the relationship between a lack of PRDM9 in canids and the trends observed in their TSS, are weakened.
      6. In Methods, the ENSEMBL version (in addition of the per-species genome version) should be mentioned.
      7. In Fig 1, it is worth clarifying in the legend that the differences between the first and second rows of panels is in the length of the plotted region.

      Significance

      The manuscript provides a rigorous analysis of the possible processes that have impacted the TSS GC-content during evolution. It should be of interest to a diverse set of investigators in the genomics community, since it touches on different topics including genome evolution,transcription and gene structures.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      This manuscript investigates the dynamics of GC-content patterns in the 5'end of the transcription start sites (TSS) of protein-coding genes (pc-genes). The manuscript introduces a quite careful and comprehensive analysis of GC content in pc-genes in humans and other vertebrates, specially around the TSS. The result of this investigation states that "GC-content surrounding the TSS is largely influenced by patterns of recombination." (from end of Introduction)

      My main concern with this manuscript is one of causal reasoning, whether intended or not. I hope the authors can follow my reasoning bellow on how the logic sometimes seems to fail, and that they introduce changes to clarify their suggested mechanisms of action.

      The above quoted sentence form the end of the Intro is in conflict with this other sentence that appears at the end of the Abstract "the dynamics of GC-content in mammals are largely shaped by patterns of recombination". The sentence in the Intro seems to indicate that the effect is specific to TSSs, but the one in the abstract seem to indicate the opposite, that is, that the effect is ubiquitous.

      The observations as stated in the abstract are: "We observe that in primates and rodents, where recombination is directed away from TSSs by PRDM9, GC-content at protein-coding gene TSSs is currently undergoing mutational decay."

      If I understand the measurements described in the manuscript correctly, and the arguments around them, you seem to show that the mutational decay of GC-content in humans is independent of location (TSSS or not), as noted here

      (also from the abstract) "These patterns extend into the open reading frame affecting protein-coding regions, and we show that changes in GC-content due to recombination affect synonymous codon position choices at the start of the open reading frame."

      There is one more result described in the manuscript, that in my mind is very important, but it is not given the relevance that it appears to me that it has. That is presented in Figure S3G. "we concluded that GC-content at the TSS of protein-coding genes is not at equilibrium, but in decay in primates and rodents. This decay rate is similar to the decay seen in intergenic regions that have the same GC-content (Figure S3G)"

      Thus, if the decaying effect happens everywhere, how can it be related to "recombination being directed away from TSSs by PRDM9" as it is stated in the abstract and in the model described in Figure 7?

      The fact that the decay rate is similar to any other region with similar GC-content should be an indication that the effect is not related to anything having to do with TSS or recombination being directed away from TSSs by PRDM9.

      I hope these paragraphs show my confusion about the relationship between the results presented which I think are very comprehensive and their interpretation and suggested model for GC-content dynamics around TSSs in human.

      On another note, can you provided a bit more background on recombination and its mechanisms? You seem to have confident sets of genes under high/low/med recombination. How are those determined. You also seem to concentrate the cause of recombination on PRDM9, please explain. Is PRDM9 the unique indicator of recombination?

      Specific comments

      Figure 1, it is very hard to understand the differences between the three rows. Please explain more clearly in the legend, and add more information to the figure itself.

      Figure 7, express somewhere in the figure that the y axis measures GC content. Figure seems to introduce a 'causal' model of GC-content dismissing based on recombination being directed away from TSSs. How about the diminishing of GC-content on any other genomic regions as you have shown in Figure S3G?

      The title is too selective, as to the results, and it has the implication that the decay is exclusive to the surrounding of the TSSs.

      Significance

      The statistical analysis is comprehensive and robust. Their model interpretation as is describe induces confusion and needs to be clarified.

      I am an expert computational biologist, I do not have a deep knowledge of sequence implications of recombination, and it would be good if the manuscript could add some more background on that.

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

      Learn more at Review Commons


      Reply to the reviewers

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

      In "BDNF signaling requires matrix metalloproteinase-9 during structural synaptic plasticity", Legutko et al. used two-photon microscopy and glutamate uncaging to show that rapid release (seconds) of MMP-9 from dendritic spines following synaptic stimulation as well as MMP-9 dependent activation of TrkB. The authors also show and MMP-dependent increase on dendritic spine volume. These data support the possibility that MMP-9 rapidly activates BDNF to promote the spine maturation required for LTP. All is all the manuscript is well written, and the data is convincing and important.

      Answer: We thank the reviewer for that comment.

      Questions/Concerns:

      • The authors show cell free cleavage of BDNF by recombinant MMP-9. It would be more convincing to show that MMP-9 cleaved BDNF using concentrated supernatants following synaptic stimulation in control versus inhibitor treated slices. Answer: In the present study we focus on a single-spine approach; thus, we did not include general stimulation techniques and biochemical analyses. To our knowledge, there is no method to show BDNF cleavage by MMP-9 directly at a single synapse. We agree with the reviewer that the general stimulation is important; however, at the synapse, there is potentially a whole array of proteases such as plasmin, tissue plasminogen activator (tPA) that might not only create catalytic cascade and proteolytically activate MMP-9 but also directly cleave proBDNF. When stimulating neurons and analysing supernatants, it is therefore impossible to determine if MMP-9 directly digests proBDNF to mBDNF or, alternatively, whether it is just a part of a proteolytic cascade leading to BDNF maturation. Therefore, our result where we use recombinant proteins provide an important piece of evidence that MMP-9 can indeed cleave proBDNF directly. Of note, experiments using brain extracts have been published previously, for example in a paper of Mizoguchi et al., J.Neurosci. (2011); DOI:10.1523/JNEUROSCI.3118-11.2011, where the authors showed increased cleavage of BDNF after pentylenetetrazole kindling and the kindling induced proBDNF cleavage was decreased in MMP-9 KO mice.

      • The concentration of the MMP-9/13 inhibitor used was quite high and would also inhibit MMP-1, -3 and -7. This concern is, however, abrogated by the use of the MMP-9 KO. But it might be important to mention that the inhibitor is not MMP-9 specific at higher concentrations. Answer: To comply with this remark, we have stressed the notion in the Discussion of the revised ms.:

      "There are over twenty MMPs with overlapping substrate specificity (Fields, 2015; Cieplak & Strongin, 2017) and there are no fully specific, commercially available inhibitors for MMP-9. Since Inhibitor I might affect also other MMPs, to further test the involvement of the protease in sLTP, we have used hippocampal slice cultures prepared from MMP-9 KO mice and their WT littermates (Fig. 1E, 1F)."

      • In figure 1C vs E, as well as Fig 3C vs E, it appears that the DMSO to inhibitor (1C and 3C) change is larger than the WT vs MMP-9 KO (1E and 3E). Is this possibly because DMSO has a potentiating effect and/or because the inhibitor is getting other MMPs or the MMP-9 KO has compensatory increases in other MMPs? __Answer: __At the concentrations used in the study (not exceeding 0.08%), we do not consider DMSO having any potentiating effect. As we discuss in the manuscript, the difference between DMSO control and MMP-9 WT is most likely due to differences between genetic lines of the mice. This is also a reason why each set of experiments has its own control. Of note, in the paper preceding this study, Harvard et al., Nature (2016); doi:10.1038/nature19766, spine volume change induced by uncaging, vary between 200 and 300% depending on mice strain used in the experiment.

      • The idea that MMP-9 and pro-BDNF are in the same vesicular stores is an interesting and very plausible one. Perhaps the authors could discuss what is known about the types of vesicles thought to harbor these two proteins. Answer: To follow on this remark, we added information about the vesicles containing BDNF and MMP-9 in the Discussion:

      "Given that the release kinetics of BDNF and MMP-9 are similar, one could speculate that the effect of MMP-9 inhibition on early TrkB activation can be achieved because both, MMP-9 and BDNF are co-localized and co-released from the same release vesicles. BDNF is widely considered to be stored and released from Large Dense-Core Vesicles (Dieni et al, 2012; Kojima et al, 2020), and MMP-9 release although not studied in neurons but in cell lines, also points to the same type of vesicles (Stephens et al, 2019)."

      • It might be useful to add to the discussion pathological conditions such as major depression and post-stroke plasticity in which MMP-9 dependent BDNF activation could be important. Answer: We thank the reviewer for that suggestion. We have added the information about MMP-9 and BDNF link in the brain pathologies in the Discussion:

      "Additionally, our data may provide a functional link between the involvement of MMP-9 and BDNF in various brain pathologies, in which such a link has previously been implicated, for example in addiction (Cheng et al, 2019), schizophrenia (Pan et al, 2022; Yamamori et al, 2013), ischemic stroke (Li et al, 2022) or even following cochlear implantation (Matusiak et al, 2023)."

      Reviewer #1 (Significance (Required)):

      The results are significant to understanding synaptic plasticity in health and disease.

      __Answer: __We thank the reviewer for that comment stressing the importance of our study.

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

      The study addresses the molecular mechanisms of activity-dependent morphological plasticity of dendritic spines, focusing on the role of MMP-9 and BDNF-TrkB in the signalling and biochemical activities that lead to and maintain spine enlargement ('structural LTP', sLTP) induced experimentally.

      It is based on a combination of 2-photon imaging of spine morphology, 2-photon imaging of MMP-9-SEP fluorescence, 2-photon FLIM of a biosensor for TrkB activity and 2-photon glutamate uncaging in organotypic hippocampal brain slices. In addition, it includes an assay of protein digestion based on Western immuno blots.

      As results, the study reports that diminishing MMP-9 activity (pharmacologically or genetically) in the slices reduces sLTP, that repetitive glutamate uncaging evokes the release of MMP-9 from the spines that undergo sLTP, and that this effect can be blocked by pharmacological blockade of NMDA or exocytosis, that repetitive glutamate uncaging on a spine increases TrkB activation in the spine, and that this effect is diminished in slices from MMP-9 KO animals or treated by an MMP-9 blocker, and that MMP-9 can cleave proto-BDNF into mature BDNF in a cell-free medium.

      The experiments are technically challenging but they are well conceived, designed and executed. The conclusions are well supported by the results, which are clearly discussed in light of the substantial and somewhat contradictory literature.

      Reviewer #2 (Significance (Required)):

      The study provides a finer view of the dynamic role of MMP-9 in activity-dependent spine plasticity, reinforcing and expanding existing knowledge on this timely topic.

      The study is well executed and the conclusions are warranted. The study is an experimental tour de force, even if the biological results and insights are rather incremental and don't force us to revise our main assumptions or expectations.

      Answer: We thank the reviewer for that comment and the appreciation of our work.

      I only have a few questions and suggestions:

      • 2: Do TeTx and AP5 treatments also block spine enlargement? The MMP9-SEP and mCherry signals in the spines are going up, what about their ratio F/R? __Answer: __Yes, we do have results showing that TeTx and AP5 block spine enlargement, however we did not present them in the original manuscript. The AP5 application on spine enlargement was previously demonstrated for example by Tanaka and co-workers (2008); DOI: 10.1126/science.1152864, and the effect of TeTx on LTP and insertion of AMPA receptors has also been demonstrated multiple times for example by Penn et al., Nature (2017); doi:10.1038/nature23658. To comply with the reviewer's request we have included the data in the revised version of the manuscript (Figure 2C).

      As far as the F/R ratio is concerned we shall stress that the aim of our experiments was to show the release of MMP-9 into extracellular space upon uncaging. We have initially tried to analyse the ratio of F/R, however the green signal that comes from MMP9-SEP does not accumulate at the spine, apparently being rapidly diffused. Therefore, the overall red signal for mCherry increases much faster (mCherry fills the cytoplasm in the spine) than the MMP9-SEP; therefore, the F/R ratio is decreasing over time. Figure 2G shows that increases in MMP9-SEP fluorescence are only transient (around 0.5 s) after uncaging pulses.

      • 3B shows increased TrkB activation after glutamate uncaging, but is it possible to see the spine enlargement in the FRET-FLIM signal/images? Answer: Yes, it is possible to observe spine enlargement during FRET-FLIM experiments by counting photons from the red channel (RFP) as well as from the green one (GFP), however due to technical difficulties spine volume change was measured in separate experiments.

      • Fix: mW and Chameleon in the Methods section - corrected

      • Consider streamlining the Discussion a bit - we have reviewed the discussion
      • Consider adding a schematic to summarise the new and existing findings Answer: We thank the reviewer for the suggestion, we have added a schematic summarising the paper as a separate figure (Fig.4).

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

      In this short report, Legutko et al address the role of MMP9 in BDNF signaling in the context of structural long-term potentiation (sLTP). In particular, they assess whether MMP9 is secreted fast enough to mediate the cleavage of proBDNF in mBDNF during sLTP. The study uses 2-photon imaging of hippocampal organotypic slices, glutamate uncaging and FRET-based sensors of TrkB activity. The authors demonstrate that MMP9 is secreted within seconds upon 2-photon glutamate uncaging and that MMP9 secretion precedes spine enlargement. They also show that MMP9 can cleave proBDNF in vitro. However, the role of MMP9 in sLTP and associated TrkB signaling remains speculative at the end of the manuscript.

      Major comments

      • The title of the first result section "spine head enlargement during structural plasticity depends on MMP9 activity" is an overstatement. The authors provide evidence that MMP inhibition and MMP9 KO decrease spine enlargement during the early phase of sLTP. However, after the first few minutes, spines still display long-term enlargement, and no difference between WT and MMP9 KO mice can be detected. These data suggest that MMP9 is only involved in the initial phase of sLTP, and that other MMPs are involved in sLTP.

      __Answer: __We thank the reviewer for that comment. We have change the wording in the revised manuscript to accommodate the suggestion.

      • The authors cannot conclude that "spine head enlargement during sLTP depends on MMP9 activity".

      __Answer: __We thank the reviewer for that comment. We have changed the title and wording in the revised manuscript to accommodate the suggestion.

      • The authors should apply Inhibitor I on MMP9 KO slices to determine if MMPs other than MMP9 are involved in spine enlargement.

      __Answer: __We thank the reviewer for the suggestion, and indeed we agree that other MMPs might be involved in spine enlargement induced by glutamate uncaging. Furthermore, applying Inhibitor I will not resolve the question which MMPs or other proteases are involved in the spine enlargement. Applying Inhibitor I on MMP-9 KO slices would only eliminate one of the proteases. To deal with this difficult issue, we have used slices from MMP-9 KO mice and showed the influence of MMP-9 on the transient phase of spine enlargement induced by glutamate uncaging.

      • If Inhibitor I still impacts sLTP in MMP9 KO slice, it would greatly benefit this study to determine which MMPs are involved (for example by analyzing the expression patterns of MMPs in their neurons and selectively inactivating those expressed with shRNAs).

      Answer: The proposed experiment is an excellent suggestion for a future project however it is not an easy experiment to perform. MMPs expression pattern could be assessed by single cell RNA sequencing to distinguish it from for example astrocytic expression, however it often fails to detect mRNAs which are expressed at low level. For example mRNA coding MMP-9 belongs to this group as its mRNA is kept at very low level, see, e.e.g, Konopacki et al., Neuroscience (2007); https://doi.org/10.1016/j.neuroscience.2007.08.026, Dziembowska et al. J.Neurosci (2012); https://doi.org/10.1523/JNEUROSCI.6028-11.2012. There is also quite low correlation between mRNA levels and protein levels at a global scale, see e.g., Reimegård et al., Comm. Biol. (2021); https://doi.org/10.1038/s42003-021-02142-w, therefore predictive power of mRNA sequencing for the importance of a particular protein might not be sufficiently informative. Moreover, the situation is even more complex in neurons which are strongly compartmentalized, and where local translation plays a significant role. We have previously studied this particular aspect for MMP-9, Dziembowska et al. J.Neurosci. (2012); DOI:10.1523/JNEUROSCI.6028-11.2012..

      • The title of the third/last result section "TrkB signaling depends on MMP9 activity" is also an overstatement. In Figure 3, the authors show that the pharmacological inhibition of MMPs slightly inhibits TrkB signaling in the early phase of sLTP, and almost abolishes TrkB signaling in the second phase (> 3 min after uncaging). However, the data suggesting a specific role for MMP9 in TrkB signaling are not convincing (Figure 3E-F). The activation of trkB during sLTP is weak even in WT, the peak of trkB activation upon glutamate uncaging in not disrupted in MMP9 KO mice, and the data are noisy. It is a major concern that the authors cannot convincingly show that TrkB signaling is altered in MMP9-deficient neurons. Answer: To the best of our knowledge, using FRET-FLIM sensors is the best and state-of-the-art method to track biochemical changes (such as receptor activation) in real time using live preparations. The method is very sensitive and published previously by one of the authors of the current study where TrkB sensor is activated in the same magnitude (Harward et al., Nature, 2016; doi: 10.1038/nature19766). Moreover similar magnitude of sensor activation was reported previously in single dendritic spines for other sensors using FLIM-FRET method: Rho GTPases (Hedrick et al., Nature 2016; doi: 10.1038/nature19784), IGF1R (Tu et al., Sci Advanc. 2023; doi: 10.1126/sciadv.adg0666) or CaMKII (Chang et al., Nat. Commun. 2019; https://doi.org/10.1038/s41467-019-10694-z). The noise is to be expected, as we are imaging small compartments in a short time where collecting enough number of photons is challenging. Similarly to previous studies using FRET-FLIM sensors, we bin experimental points to reduce noise for statistical analysis. Notably, the biological effect we observe, namely sensor activation, is well above the experimental noise that in inevitable in this experimental approach. For statistical analyses we have used repeated measures ANOVA, which is very sensitive to noise and signal fluctuation. The differences we measure are statistically significant.

      • The authors discuss that the problem might stem from mouse genetic backgrounds. However, if the MMP9 KO mouse model is not appropriate to answer the question, the authors should use another one (i.e. MMP9 knockdown using sh/siRNAs).

      Answer: We believe that the effect of MMP-9 KO in this experiment is evident, as supported by Fig. 3 E,F and statistical analysis. Furthermore, the experiment with the inhibitor further supports our reasoning.

      • In addition to the graphs, the authors should mention in the text the percentage of inhibition compared to WT). This would make the results easier to read.

      Answer: To comply with this request the appropriate information has been added to the revised manuscript.

      • The change in TrkB activation following glutamate uncaging is low (max 5-7 % at the peak, compared to 200% for spine volume). This raises the question of the physiological relevance of TrkB activation in this model. The authors should include experiments with a trkB inhibitor to assess whether it prevents sLTP in WT and MMP9 KO mice. They should also discuss other potential targets of MMP9. This would strengthen the rationale of the experiments. Answer: Previously published results using the same TrkB sensor (Harward et al., Nature, 2016; doi: 10.1038/nature19766), show exactly the same change in binding fraction calculated from a change in GFP fluorescence lifetime. These data are also in agreement with well-established standard in the field, see, e.g., Rho GTPases (Hedrick et al., Nature 2016; doi: 10.1038/nature19784), IGF1R (Tu et al., Sci Advanc. 2023; doi: 10.1126/sciadv.adg0666) or CaMKII (Chang et al., Nat. Commun. 2019; https://doi.org/10.1038/s41467-019-10694-z). In response to the comment we have addressed this issue in the Discussion in the revised ms.

      Minor comments

      • In the introduction, the authors should provide more context. Could the authors develop the "long standing debate on which enzymes process proBDNF to mBDNF"? Answer: We have removed the sentence as we realized it was too confusing and the paper does not compare between different proteases which may process proBDNF to mBDNF.

      • In the result section:

      • First paragraph, the last sentence should be moved from the end of the paragraph to before "During sLTP induction...".

      Answer: Following the reviewer suggestion, we have moved the sentence.

      • Several paragraphs in the result section lack a proper conclusion/interpretation, which makes it difficult to read. Examples: after (Fig. 2E), after (Fig. 2F). The authors should explicit what their results mean.

      Answer: We have changed the paragraphs and tried to explain the results better.

      • Clarify when and for how long the MMP inhibitor was applied. Answer: The inhibitor was applied 30 min. before stimulation. We have added the information in the Methods section.

      • In figure 1, The authors observe a specific alteration of the early, transient, sustained increase in spine head volume in MMP9 KO mice. The later phase of sLTP is not impacted, which means that sLTP is induced and maintained in the KO. Could the authors discuss the role/importance of this transient peak in spine head volume? Answer: In response to this comment, we have discussed this issue in the revised ms. as follows:

      " The transient spine expansion might be important for the remodeling of the synapse (Lang et al, 2004) and is associated with NMDAR-dependent formation of "memory gel" created by enlargement pool of actin (Honkura et al, 2008; Kasai et al, 2010; Bonilla-Quintana & Rangamani, 2024). It has also been reported that TrkB activity can influence actin dynamics (Woo et al, 2019; Hedrick et al, 2016), in some instances in concert with integrin 1 (Wang et al, 2016), which is also activated by MMP-9 (Wang et al, 2008; Michaluk et al, 2009, 2011) and further supports our observations."

      Reviewer #3 (Significance (Required)):

      The manuscript aims to bring conceptual advance in our understanding of structural synaptic plasticity by investigating the role and timing of MMP9 secretion in TrkB signaling. Previous work from the Yasuda lab and others have shown that trkB is activated early on by BDNF during sLTP. However, how, when and where BDNF is cleaved from proBDNF in mBDNF is poorly understood. The authors demonstrate that the pharmacological inhibition of metalloproteases attenuates structural long-term plasticity (sLTP) and that MMP9 is secreted early enough to cleave proBDNF. They also show that MMP9 can cleave proBDNF in BDNF in vitro. Whether MMP9 specifically cleaves BDNF during sLTP and whether this cleavage is physiologically relevant for sLTP remain an open question.

      This report will be of interest to neurobiologists interested in the molecular mechanisms of synaptic plasticity.

    2. 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 short report, Legutko et al address the role of MMP9 in BDNF signaling in the context of structural long-term potentiation (sLTP). In particular, they assess whether MMP9 is secreted fast enough to mediate the cleavage of proBDNF in mBDNF during sLTP. The study uses 2-photon imaging of hippocampal organotypic slices, glutamate uncaging and FRET-based sensors of TrkB activity. The authors demonstrate that MMP9 is secreted within seconds upon 2-photon glutamate uncaging and that MMP9 secretion precedes spine enlargement. They also show that MMP9 can cleave proBDNF in vitro. However, the role of MMP9 in sLTP and associated TrkB signaling remains speculative at the end of the manuscript.

      Major comments

      1. The title of the first result section "spine head enlargement during sLTP depends on MMP9 activity" is an overstatement. The authors provide evidence that MMP inhibition and MMP9 KO decrease spine enlargement during the early phase of sLTP. However, after the first few minutes, spines still display long-term enlargement, and no difference between WT and MMP9 KO mice can be detected. These data suggest that MMP9 is only involved in the initial phase of sLTP, and that other MMPs are involved in sLTP.
        • The authors cannot conclude that "spine head enlargement during sLTP depends on MMP9 activity".
        • The authors should apply Inhibitor I on MMP9 KO slices to determine if MMPs other than MMP9 are involved in spine enlargement.
        • If Inhibitor I still impacts sLTP in MMP9 KO slice, it would greatly benefit this study to determine which MMPs are involved (for example by analyzing the expression patterns of MMPs in their neurons and selectively inactivating those expressed with shRNAs).
      2. The title of the third/last result section "TrkB signaling depends on MMP9 activity" is also an overstatement. In Figure 3, the authors show that the pharmacological inhibition of MMPs slightly inhibits TrkB signaling in the early phase of sLTP, and almost abolishes TrkB signaling in the second phase (> 3 min after uncaging). However, the data suggesting a specific role for MMP9 in TrkB signaling are not convincing (Figure 3E-F). The activation of trkB during sLTP is weak even in WT, the peak of trkB activation upon glutamate uncaging in not disrupted in MMP9 KO mice, and the data are noisy. It is a major concern that the authors cannot convincingly show that TrkB signaling is altered in MMP9-deficient neurons.
        • The authors discuss that the problem might stem from mouse genetic backgrounds. However, if the MMP9 KO mouse model is not appropriate to answer the question, the authors should use another one (i.e. MMP9 knockdown using sh/siRNAs).
        • In addition to the graphs, the authors should mention in the text the percentage of inhibition compared to WT). This would make the results easier to read.
      3. The change in TrkB activation following glutamate uncaging is low (max 5-7 % at the peak, compared to 200% for spine volume). This raises the question of the physiological relevance of TrkB activation in this model. The authors should include experiments with a trkB inhibitor to assess whether it prevents sLTP in WT and MMP9 KO mice. They should also discuss other potential targets of MMP9. This would strengthen the rationale of the experiments.

      Minor comments

      1. In the introduction, the authors should provide more context. Could the authors develop the "long standing debate on which enzymes process proBDNF to mBDNF"?
      2. In the result section:
        • First paragraph, the last sentence should be moved from the end of the paragraph to before "During sLTP induction...".
        • Several paragraphs in the result section lack a proper conclusion/interpretation, which makes it difficult to read. Examples: after (Fig. 2E), after (Fig. 2F). The authors should explicit what their results mean.
      3. Clarify when and for how long the MMP inhibitor was applied.
      4. In figure 1, The authors observe a specific alteration of the early, transient, sustained increase in spine head volume in MMP9 KO mice. The later phase of sLTP is not impacted, which means that sLTP is induced and maintained in the KO. Could the authors discuss the role/importance of this transient peak in spine head volume?

      Significance

      The manuscript aims to bring conceptual advance in our understanding of structural synaptic plasticity by investigating the role and timing of MMP9 secretion in TrkB signaling. Previous work from the Yasuda lab and others have shown that trkB is activated early on by BDNF during sLTP. However, how, when and where BDNF is cleaved from proBDNF in mBDNF is poorly understood. The authors demonstrate that the pharmacological inhibition of metalloproteases attenuates structural long-term plasticity (sLTP) and that MMP9 is secreted early enough to cleave proBDNF. They also show that MMP9 can cleave proBDNF in BDNF in vitro. Whether MMP9 specifically cleaves BDNF during sLTP and whether this cleavage is physiologically relevant for sLTP remain an open question.

      This report will be of interest to neurobiologists interested in the molecular mechanisms of synaptic plasticity.

    3. 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 study addresses the molecular mechanisms of activity-dependent morphological plasticity of dendritic spines, focusing on the role of MMP-9 and BDNF-TrkB in the signalling and biochemical activities that lead to and maintain spine enlargement ('structural LTP', sLTP) induced experimentally.

      It is based on a combination of 2-photon imaging of spine morphology, 2-photon imaging of MMP-9-SEP fluorescence, 2-photon FLIM of a biosensor for TrkB activity and 2-photon glutamate uncaging in organotypic hippocampal brain slices. In addition, it includes an assay of protein digestion based on Western immuno blots.

      As results, the study reports that diminishing MMP-9 activity (pharmacologically or genetically) in the slices reduces sLTP, that repetitive glutamate uncaging evokes the release of MMP-9 from the spines that undergo sLTP, and that this effect can be blocked by pharmacological blockade of NMDA or exocytosis, that repetitive glutamate uncaging on a spine increases TrkB activation in the spine, and that this effect is diminished in slices from MMP-9 KO animals or treated by an MMP-9 blocker, and that MMP-9 can cleave proto-BDNF into mature BDNF in a cell-free medium.

      The experiments are technically challenging but they are well conceived, designed and executed. The conclusions are well supported by the results, which are clearly discussed in light of the substantial and somewhat contradictory literature.

      Significance

      The study provides a finer view of the dynamic role of MMP-9 in activity-dependent spine plasticity, reinforcing and expanding existing knowledge on this timely topic.

      The study is well executed and the conclusions are warranted. The study is an experimental tour de force, even if the biological results and insights are rather incremental and don't force us to revise our main assumptions or expectations.

      I only have a few questions and suggestions:

      • Fig. 2: Do TeTx and AP5 treatments also block spine enlargement? The MMP9-SEP and mCherry signals in the spines are going up, what about their ratio F/R?
      • Fig. 3B shows increased TrkB activation after glutamate uncaging, but is it possible to see the spine enlargement in the FRET-FLIM signal/images?
      • Fix: mW and Chameleon in the Methods section
      • Consider streamlining the Discussion a bit
      • Consider adding a schematic to summarise the new and existing findings
    4. 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 "BDNF signaling requires matrix metalloproteinase-9 during structural synaptic plasticity", Legutko et al. used two-photon microscopy and glutamate uncaging to show that rapid release (seconds) of MMP-9 from dendritic spines following synaptic stimulation as well as MMP-9 dependent activation of TrkB. The authors also show and MMP-dependent increase on dendritic spine volume. These data support the possibility that MMP-9 rapidly activates BDNF to promote the spine maturation required for LTP. All is all the manuscript is well written, and the data is convincing and important.

      Questions/Concerns:

      1. The authors show cell free cleavage of BDNF by recombinant MMP-9. It would be more convincing to show that MMP-9 cleaved BDNF using concentrated supernatants following synaptic stimulation in control versus inhibitor treated slices.
      2. The concentration of the MMP-9/13 inhibitor used was quite high and would also inhibit MMP-1, -3 and -7. This concern is, however, abrogated by the use of the MMP-9 KO. But it might be important to mention that the inhibitor is not MMP-9 specific at higher concentrations.
        1. In figure 1C vs E, as well as Fig 3C vs E, it appears that the DMSO to inhibitor (1C and 3C) change is larger than the WT vs MMP-9 KO (1E and 3E). Is this possibly because DMSO has a potentiating effect and/or because the inhibitor is getting other MMPs or the MMP-9 KO has compensatory increases in other MMPs?
      3. The idea that MMP-9 and pro-BDNF are in the same vesicular stores is an an interesting and very plausible one. Perhaps the authors could discuss what is known about the types of vesicles thought to harbor these two proteins.
      4. It might be useful to add to the discussion pathological conditions such as major depression and post-stroke plasticity in which MMP-9 dependent BDNF activation could be important.

      Significance

      The results are significant to understanding synaptic plasticity in health and disease.

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

      Learn more at Review Commons


      Reply to the reviewers

      Compared to our initial submission to Review Commons, we have addressed all the reviewers' comments. We have extensively re-written the manuscript to make it clearer to a larger audience. In particular, we have transferred Figure EV1 to Figure 1 with more complete panels and included a scheme (Figure EV3) on the steps of D2R internalization which we measure with live cell imaging. We have added a new paragraph to the start of the Discussion to summarize our main conclusions and reordered the discussion on the possible mechanisms of membrane PUFA enrichment on D2R endocytosis. All the changes in the text are in red for easier comparison with the previous version.

      As suggested by reviewer 1, we have performed additional experiments to test the specificity of the effects of PUFA treatments on D2R endocytosis, reinforcing the results shown in Figure 4 using feeding assays. We show with live cell TIRF imaging and the ppH assay that TfR-SEP endocytosis is not affected (Figure EV5) and that SEP-β2AR endocytosis and βarr2-mCherry recruitment to the plasma membrane are not affected (Figure EV6).

      Reviewer #1

      Evidence, reproducibility and clarity

      *The manuscript, using different live and fixed cell trafficking assays, demonstrates that incorporation of poly-unsaturated, but not saturated, free fatty acids in the membrane phospholipids reduce agonist induced internalization of the D2 dopamine receptor but not the adrenergic beta2 receptors or the transferrin receptor. Pulsed pH (ppH) live microscopy further demonstrated that the reduced internalization by incorporation of free fatty acid was accompanied by a blunted recruitment of Beta-arrestin for the D2R.

      I believe said claims put forward in the manuscript are overall well supported by the data and as such I do not believe that further experiments are necessarily needed to uphold these key claims. Also, the methodology is satisfactorily reported, and statistics are robust, although two-way Anova like used in Fig 1 seems appropriate for Fig 2 and 3*

      We thank the reviewer for his/her positive assessment of our work. We have checked the statistical tests used for all our measures. For Figure 2 and 3 (now 3 and 4) we test for only one factor (PUFA treatment or not) so we ran ordinary one-way ANOVA using Graphpad Prism.

      That said, I suggest that the fixed cell internalization experiments (Fig 2 and 3), which relate the effect on the D2R to B2AR and transferrin are revised. This is important since this is relevant to judge whether the effect is a general or a selective molecular mechanism since this is the one of the three assay which this comparison relies on. Alternatively, I suggest omitting this data and include the B2AR in the Live DERET assay and both B2AR and TfR in the ppH assay. Specifically, my concerns with the fixed cell internalization are: • The analysis is based on counting the number of endosomes, which is not necessarily equivalent to the number of receptors internalized

      The number of puncta, as well as their fluorescence, is reported by the analysis program (written in Matlab2021 and available upon request). We chose to show number of puncta because they reflect more directly the number of labelled endosomes (in Figures 3 and 4). As shown in the figure below, we found slight but significant differences between groups for FLAG-D2R (88.6 % and 87.6 % of average fluorescence in DHA and DPA treated cells compared to control cells), (panel A), and no differences for FLAG-β2AR (panel B). We find a significant decrease in puncta fluorescence for transferrin uptake in cells incubated with DHA (but not DPA) relative to control cells (panel C). However, because we did not detect differences in the number of puncta or in the frequency and amplitude of endocytic vesicle creation events (see below), we still conclude that enrichment with exogenous PUFAs does not affect clathrin mediated endocytosis.

      In conclusion, the most robust measure of endocytosis for this assay is the number of detected puncta per cell rather than their fluorescence.

      • The analysis relies on fully effective stripping of the surface pool of receptors - i.e clustered surface receptors not stripped by the protocol will be assessed as internalized. It is often very difficult to obtain full efficiency of the Flag-tag stripping and this is somewhat expression dependent. • The protocol for the constitutive and agonist induced internalization is different and yet shown on the same absolute graph. Although I take it the microscope gain setting are unaltered between the constitutive and agonist induced internalization I don't believe the quantification can be directly related. This is confusing at the very least. More critically however, the membrane signal from the non-stripped condition of constitutive internalization will likely fully shield internalized receptors in the Rab4 membrane proximal recycling pathway leading to under-estimation of the in the constitutive endocytosis. I believe this methodological limitation underlies the massive relative difference in the constitutive endocytosis between panel 2A,B and 2C,D. For comparison, by a quantitative dual color FACS endocytosis assay, we have previously demonstrated the ligand endocytosis a ~4 fold increased over constitutive (in concert with Fig 2A,B here) (Schmidt et al 20XX). Importantly, high relative variability by this methodology could well shield an actual effect of incorporation of FFAs on the constitutive endocytosis. We thank the reviewer for pointing this difference in the protocol. As a matter of fact, we have not used acid stripping in all the conditions used for the uptake assays (Figures 3 and 4). We apologize for the confusion and we have clarified this point in the Methods section. In early experiments we compared conditions with or without stripping but we concluded from these experiments that indeed, the stripping was not complete. Moreover, we noticed early on that many cells treated with DHA or DPA did not have any detectable cluster (13 cells out of 58 quantified cells treated with DHA after addition of QPL, 12/56 cells treated with DPA, 0/68 for cells treated with vehicle). Stripping the antibody would have made these cells undetectable, biasing the analysis. Therefore, to make our results more consistent we decided to use non-stripping conditions. To detect endosomes specifically, we used a segmentation tool developed earlier (see Rosendale et al.* 2019). This tool is based on wavelet transforms which recognizes dot-like structures. In addition, we excluded from the cell mask the labelled plasma membrane by a mask erosion.

      We agree the design of experiments was not aimed at comparing the effect of PUFA treatment on low levels of constitutive D2R endocytosis. This would require more sensitive assays and be addressed in subsequent studies.

      'Optional' Also, it would be informative to see the ppH Beta-arrestin experiments with the B2AR to assess, whether the putative discrepancy between D2R and B2AR is upstream or downstream of the blunted Beta-arrestin recruitment. To the same point, it would be very informative to assess how the incorporation of the free fatty acids affect receptor signalling, which would also help relate the effect of incorporation of the FFA's in the phospholipids to previous experiment using short term incubation with FFA's

      We have now performed live imaging experiments in HEK293 cells expressing SEP-β2AR, GRK2 and βarr2-mCherry and stimulated with isoproterenol (Figure EV6). We show that the clustering of SEP-β2AR, of βarr2-mCherry, as well as endocytosis, are not affected by treatments with DHA or DPA. In this study, we focused on the early trafficking steps of D2R internalization. It will be interesting in a future study to address its consequences on G protein dependent and independent signaling. Moreover, and for good measure, we performed experiments to assess TfR-SEP endocytosis with the ppH assay. Again, we found no difference between cells treated or not with PUFAs (Figure EV5)

      *References overall seem appropriate although Schmidt et al would be relevant for reference of the constitutive vs agonist induced endocytosis of D2R and B2AR. *

      We have now cited Schmidt et al. 2020 doi 10.1111/bcpt.13274 in the discussion with the following sentences: "D2R also shows constitutive endocytosis (Schmidt et al, 2020) which may be modulated by PUFAs although we did not detect any significant difference in our measures (see Figure 3) which were aimed at detecting high levels of internalization induced by agonists. Further work will be required to specifically examine the effect of PUFAs on constitutive GPCR internalization."

      Overall, the figures are well composed and convey the messages fairly well. Specific point that would strengthen the rigor include: • Chosing actual representative pictures of the quantitative data in Fig 2 and 3 (e.g. hard to see 25 endocytic events in Fig 2A constitutive endo, EtOH)

      We apologize for the confusion. We employ a normalization procedure to account for cell size. In addition, all numbers have been normalized to the condition stimulated with agonist with no PUFA treatment). In fact, we detect in unstimulated cells very few puncta (on average 0.6, range 0-5) compared to 27.3 clusters (range 2-87) in cells stimulated with QPL.

      • Showing actual p values for the statistical comparisons* For easier reading, we have kept the stars convention for the figures but added two tables with all statistical tests and the p values for both main figures and EV figures.

      Moreover, for ease of reading the figures (without consulting the legend repeatedly) it would be very helpful to headline individual panel with what the experiments assesses. Figure 1a and 1b for example can't be distinguished at all before reading the figure legend. Also, y-axis could be more informative on what I measured rather than just giving the unit.

      We have added titles to panels (in particular for Figure 2A,B which correspond to former Figure 1A,B) and we have given new titles to Y axes to make them clearer. We hope that the reading of our figures will now be easier.

      Finally, the figure presentation and description of S1 is very hard to follow. I cannot really make out what is assessed in the different panels.

      We have changed substantially Figure EV1 (now Figure 1) with new presentation of data: all 4 conditions (control, treated with DHA, DPA or BA) systematically presented in the same graph, and clearer titles for the parameter displayed on the Y axes. We hope that this figure is now easier to follow.

      Significance

      *The strength of the manuscript is the use and validation of incorporation of FFA's in the plasma membrane, which more closely mimics the physiological situation than brief application of FFAs as often done. Is addition, the blunted recruitment of beta-arrestin as assessed by the ppH protocol is quite intriguing mechanistically. The limitation are the relative narrow focus on the D2 receptor (and not multiple GPCRs) that does not really speak to as or assess the physiological, pathophysiological or therapeutic role of the observations (except from referring the relation between FFAs and disease). Also, despite the putative role of Beta-arrestin recruitment in the process, the actual causation in the process is not clear. This shortcoming is underscored by the putative effect on the constitutive internalization described above.

      My specific expertise for assessing the paper is within general trafficking processes (including the trafficking methodology applied), trafficking of GPCRs and function of the dopamine system including the role of D2 receptors.*

      • *

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

      • *

      The only conclusion that I was able to understand from the study was that enrichment of cell membranes with polyunsaturated fatty acids specifically inhibited agonist-induced internalization of D2 receptors. However, I think that the experiments used to conclude that PUFAs do not alter D2R clustering but reduce the recruitment of β-arrestin2 and D2R endocytosis need some clarification (i.e. data depicted in Fig. 2-5). This lack of clarity might be due to the fact I am not familiar enough with the employed technologies or to the unclear writing style of the paper. There was an overuse of acronyms, initialisms and abbreviations, which are difficult to understand for researchers outside of the specific lipid field. I think that the manuscript should be written in a way to be legible also for researchers not working in the immediate filed.

      The paper was not written in a manner that a general audience of cell biologists or those interested in GPCR biology could understand and judge. It is indeed interesting that polyunsaturated fatty acids specifically inhibit D2R internalization in HEK293 cells, and it could be significant. But, it is difficult to judge the significance of the observation without more in vivo data.

      I would suggest the following. Remove all acronyms and abbreviations. Significantly, expand the Materials and Methods section, either in the manuscript or in the Supplemental section. I suggest clearly explaining each construct used, and the function of each module in the construct, with diagrams. In addition, provide a comprehensive step by step description of each experimental protocol, providing the reader with the rationale for each step in the protocol with explanatory diagrams. The authors should also more clearly explain the rationale and logic that was utilized to make the conclusions that they did from the depicted observations. Only then can a broader audience determine if the authors' conclusions are justified.

      We thank the reviewer for his/her comments. Indeed, our main message was that two types of PUFAs (DHA and DPA) specifically alter D2R endocytosis by reducing the recruitment of β-arrestin2 without changing D2R clustering at the plasma membrane. We are sorry that our writing was not clear enough. We also found out that in the last steps of the submission to Review Commons, the first paragraph of the Discussion was inadvertently erased. This made our main conclusions, summarized in this first paragraph, less clear. We have now put back this important paragraph. Moreover, we have extensively rewritten the manuscript thriving to make it as clear as possible to a large audience. We have reduced the use of acronyms to keep only the most used ones [e.g. PUFA (used 99 times), DHA (37 times), GPCR (34 times), D2R (126 times), GRK (17 times)] and made them consistent throughout the manuscript. Following the reviewer's suggestion, we have also added a scheme of the steps following D2R activation by agonist leading to its internalization (Figure EV3).

      We understand that the reviewer implies by "in vivo data" results obtained in the brain of animals. As written in the Introduction and in the Discussion, the current work follows up on a recently published manuscripts by a subset of the authors, namely (i) Ducrocq et al. 2020 (doi 10.1016/j.cmet.2020.02.012) in which we show that deficits in motivation in animals deprived in ω3-PUFAs can be restored specifically by conditional expression of a fatty acid desaturase from c. elegans (FAT1) that allows restoring PUFA levels specifically in D2R-expressing striatal projection neurons (which mediate the so-called indirect pathway), and (ii) Jobin et al. 2023 (doi: 10.1038/s41380-022-01928-6) which combines in cellulo (HEK 293 cells) and in vivo data to show that PUFAs affects the ligand binding of the dopamine D2 receptor and its signaling in a lipid context that reflects patient lipid profiles regarding poly-unsaturation levels.

      Reviewer #2 (Significance (Required)):

      • *

      In summary, I will reiterate that the reported experiments need to be much better explained to make the study understandable to a broader audience and for that audience to determine whether the conclusions are justified.

      • *

      • *

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

      • *

      Summary:

      The authors investigate the role of lipid polyunsaturation in endocytic uptake of the dopamine D2 receptor (D2R). To modulate the degree of unsaturation in live cell plasma membranes, the authors incubate cell lines with pure fatty acid that is metabolized and incorporated into the cellular membranes. To quantify the internalization of D2R in these live cells, the authors utilized quantitative fluorescence assays such as DERET and endosome analysis to determine the degree and rate of D2R internalization in the presence of two model agonists - dopamine and quinpirole. The authors conclude that when the PUFA content of the plasma membrane is increased (i.e., via ω3 or ω6 fatty acids), both the quantity and rate of D2R internalization decrease substantially. The authors confirmed that these phenomena are specific to D2R as caveolar endocytosis and clathrin-mediated endocytosis were unaffected when these same experimental techniques were utilized for β2 adrenergic receptor and transferrin. Additionally, the authors conclude that the clustering ability of D2R is unaffected by lipid unsaturation but that the ability of D2R clusters to interact with β-arrestin2 is inhibited in the presence of excess PUFA. Based on these findings, the authors propose several hypothetical mechanisms for lipid-D2R interactions on the plasma membrane, which will likely be the scope of future work.

      Overall, this is a highly thorough and rigorous body of work that convincingly illustrates the connection between PUFA levels and D2R activity. However, I do not agree with the authors' conclusions pertaining to how their results should be interpreted in the context of fatty acid-related disorders. Additionally, this manuscript could benefit from some reorganization which would present the work more clearly. Please see the comments below.

      We thank the reviewer for the positive appreciation of our work, qualified as a "thorough and rigorous body of work that convincingly illustrates the connection between PUFA levels and D2R activity". We will address the specific points raised by the reviewer with our answers below.

      Comments:

        • A recurring motivation for this study that is brought up by the authors is that dietary deficiency of ω3 fatty acids is tied to D2R dysfunction. This would indicate that PUFA reduction in the plasma membrane results in D2R dysfunction. However, the experiments emphasized in this manuscript investigate the condition where PUFA content is INCREASED in the plasma membrane and D2R function is compromised. It seems inappropriate for the authors to cite dietary deficiency of ω3 as a motivation when they experimentally test a condition that is tied to ω3 surplus.* Regarding the general comment of the reviewer, we agree that direct conclusion cannot be drawn on the etiology of psychiatric disorders by looking at the effect of membrane fatty acid levels on D2R in HEK 293 cells. Nevertheless, we mention in the Introduction the intriguing occurrence of low PUFA levels in psychiatric disorders as starting point to look at D2R as an important target for psychoactive drugs prescribed for these disorders. In the Discussion, we propose that manipulating fatty acid levels might potentiate the efficacy of D2R ligands used as treatments. We felt raising these aspects was not putting too much emphasis on psychiatric disorders. However, in accordance with the reviewer's comment, we toned down these descriptions in the revised manuscript.

      The goal of increasing the levels of fatty acids at the membrane in HEK 293, the most widely used cellular system to study GPCR trafficking, was to try to emulate the levels of lipids in brain cells. Indeed, the levels of PUFAs in our culture conditions are much lower (~8 %, Figure 1B) than in brain extracts (~30 %). Therefore, the "control" condition in HEK 293 cells would correspond to PUFA deficiency while after our enrichment protocol these levels are closer to those found in brain cells. Our results could therefore be interpreted as endocytosis of D2R being augmented under membrane PUFA decrease. Importantly, increased receptor internalization often correlates with decreased signaling. Therefore, membrane PUFA enrichment in our conditions would rather potentiate D2R signaling.

      • Following up on the first comment, the authors' results seem to indicate that excess ω3's are detrimental to D2R function. This result would be at odds with the conventional view that ω3's are essential and that excessive ω3 may not be harmful. The authors should rationalize their findings in the context of what is known about excess dietary ω3.*

      The Reviewer is right that the conventional view is that excessive ω3 PUFA may not be harmful. However, this rather applies to dietary consumption, which might have limited effect to brain fatty acid contents since their accretion is highly regulated. Moreover, the majority of studies looking at ω3 supplementation have been performed in young adults and the effects on the developing brain - as it might be happening in pathological conditions in which D2R is involved - remain poorly understood. Furthermore, as mentioned above, blunted internalization of D2R under membrane PUFA enrichment is not an indication of "detrimental" to D2R function. Nor do we argue that membrane enrichment corresponds to excess PUFAs.

      • I would argue that the control experiments with saturated fatty acids (i.e., Behenic Acid in figure 1), represent a scenario mimicking ω3 deficiency as the enrichment of Behenic Acid causes an overall reduction in PUFAs (Figure EV1C - an increase in SFA must correspond to a decrease in PUFA). These Behenic acid results are the only experiments presented by the authors that mimic a scenario resembling ω3 deficiency and the results show that the D2R internalization is unaffected (Figure 1G-H). Therefore, I would further argue that if anything, the authors results suggest that ω3 deficiency is NOT correlated to D2R internalization. Again, the authors must rationalize these findings in the context of what is known about dietary intake of ω3's.*

      The Reviewer must refer to the fact that nutrients rich in SFAs are usually poor in PUFAs and vice-versa. Based on our lipidomic analysis, we now present in Figure 1B the effect of treatments (DHA, DPA, BA) on the levels of PUFAs (Figure 1B) and saturated fatty acids (Figure 1C). In cells treated with behenic acid (BA), PUFA levels are not significantly changed relative to control, untreated cells, while saturated fatty acid levels are increased. BA was used here to determine whether the effects observed with PUFAs was related to the enrichment in unsaturations or due to carbon chain length (C22). It is not the case because BA treatment, unlike DHA or DPA treatment, does not affect D2R endocytosis (Figure 2G,H).

      • It's not clear why the authors decided to include an ω6 fatty acid in this study. The authors built up a detailed rationale for investigating ω3's as they are dietarily essential and tied to disease when deficient. To my knowledge, ω6's are considered much less beneficial than ω3's in a dietary sense. The inclusion of an ω6 almost seems coerced as the ω6-related results don't provide any interesting additional insights. It would benefit the manuscript if the authors provided some additional discussion explaining why ω6's are being investigated in addition to ω3's. *

      We agree that we could have made the rationale clearer. The goal in comparing ω3-DHA and ω6-DPA was to assess whether the position of the first unsaturation (n-3 vs n-6), with the same carbon chain length (C22) might differentially impact D2R endocytosis.

      • In Figure EV1D, the AHA and DPA percentages each increase by ~6%. The corresponding Figure EV1B indicates that the overall PUFA% in the plasma membrane also increases by 6%. This makes sense as the total change in PUFA content is consistent with the amount of AHA or DPA being internalized to cells. However, this consistency was not observed with BA and SFAs. In Figure EV1E, the BA percentage increases only ~1% while the total SFA percentage in Figure EV1C increases by ~6%. How can something undergoing a 1% change (relative to total lipid content) result in a 6% overall change in SFA content?*

      The reviewer is correct: the level of SFAs is increased by 5.2% (34.5 % of total FAs in control cells to 39.7 % in BA treated cells), more than the increase in BA alone (1.18% from 0.35 % to 1.53 %). A close look at our lipidomics data showed that many of the 10 saturated fatty acids quantified are enhanced. In particular, the two most abundant ones, palmitic acid (16:0) and stearic acid (18:0) are increased, from 21.37 % to 22.28 % and 8.47 % to 11.17%, respectively. The reasons for these apparent discrepancies may involve lipid metabolic pathways which convert the rare and long BA into more common and shorter SFAs to preserve lipid contents and thus membrane properties.

      • In Figure 4, the discussion of kinetics does not make sense. How exactly are kinetics being monitored in this figure? (Recruitment kinetics are discussed in panels D and G)*

      We wanted to convey the impression that the time to reach the peak βarr2-mCherry recruitment was shorter in PUFA-treated cells than in control cells. However, after analyzing the kinetics in individual cells, we did not find a statistically significant difference in the time to maximum fluorescence. Therefore, we removed this reference to the kinetics of recruitment.

      We now write: " However, treatment with DHA or DPA significantly decreased peak βarr2-mCherry fluorescence (Figure 5F-G).."

      • In Figure 5, What is the purpose of panel D? Would it be more helpful to include additional, overlaid "cumulative N" plots for scenarios in which PUFAs were enriched? This would work well in conjunction with panel F.*

      The purpose of this panel is to show the kinetics of increase in the frequency of endocytic vesicle formation upon agonist addition, and the decrease in frequency when the agonist is removed. We have now added examples of cells treated with DHA and DPA of similar surface for direct comparison with control (EtOH) cells.

      • For the readers who are new to this area or unfamiliar with the assays used, Figure 1 is not intuitive and initially difficult to interpret. It would greatly benefit the flow of the manuscript if Figures EV1A-C and EV2A were included in the main text and "Normalized R" was clearly defined in the main text, prior to discussion of Figure 1.*

      We have now transferred Figure EV1 as Figure 1. We have adapted the scheme of the DERET assay and its legend (now in Figure EV1A) to make it clearer. We did not put in Figure 2 because this figure is already very big. We have changed "Normalized R" to "Ratio 620/520) (% max)" to be clearer and more consistent with the scheme.

      Reviewer #3 (Significance (Required)):

      • *

      General assessment: The work, for the most part, is rigorous and scientifically sound. The authors utilize impressive, quantitative assays to expand our understanding of protein-lipid interactions. However, the authors need to improve their discussion of the actual physiological conditions that correspond to their experimental results.

      • *

      Advance: This work may fill a gap in our understanding of disorders related to the dopamine D2 receptor. However, some of the results may be at odds with what is currently known/understood about dietary ω3 fatty acids.

      • *

      Audience: This work will be of broad interest to researchers in the biophysics field, with particular emphasis on researchers who study protein and membrane biophysics. This work will also be of interest to researchers who study membrane molecular biology.

      • *

      Reviewer Expertise: quantitative fluorescence spectroscopy and microscopy; membrane biophysics; protein-lipid interactions

      • *
    2. 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 authors investigate the role of lipid polyunsaturation in endocytic uptake of the dopamine D2 receptor (D2R). To modulate the degree of unsaturation in live cell plasma membranes, the authors incubate cell lines with pure fatty acid that is metabolized and incorporated into the cellular membranes. To quantify the internalization of D2R in these live cells, the authors utilized quantitative fluorescence assays such as DERET and endosome analysis to determine the degree and rate of D2R internalization in the presence of two model agonists - dopamine and quinpirole. The authors conclude that when the PUFA content of the plasma membrane is increased (i.e., via ω3 or ω6 fatty acids), both the quantity and rate of D2R internalization decrease substantially. The authors confirmed that these phenomena are specific to D2R as caveolar endocytosis and clathrin-mediated endocytosis were unaffected when these same experimental techniques were utilized for β2 adrenergic receptor and transferrin. Additionally, the authors conclude that the clustering ability of D2R is unaffected by lipid unsaturation but that the ability of D2R clusters to interact with β-arrestin2 is inhibited in the presence of excess PUFA. Based on these findings, the authors propose several hypothetical mechanisms for lipid-D2R interactions on the plasma membrane, which will likely be the scope of future work.

      Overall, this is a highly thorough and rigorous body of work that convincingly illustrates the connection between PUFA levels and D2R activity. However, I do not agree with the authors' conclusions pertaining to how their results should be interpreted in the context of fatty acid-related disorders. Additionally, this manuscript could benefit from some reorganization which would present the work more clearly. Please see the comments below.

      Comments:

      1. A recurring motivation for this study that is brought up by the authors is that dietary deficiency of ω3 fatty acids is tied to D2R dysfunction. This would indicate that PUFA reduction in the plasma membrane results in D2R dysfunction. However, the experiments emphasized in this manuscript investigate the condition where PUFA content is INCREASED in the plasma membrane and D2R function is compromised. It seems inappropriate for the authors to cite dietary deficiency of ω3 as a motivation when they experimentally test a condition that is tied to ω3 surplus.
      2. Following up on the first comment, the authors' results seem to indicate that excess ω3's are detrimental to D2R function. This result would be at odds with the conventional view that ω3's are essential and that excessive ω3 may not be harmful. The authors should rationalize their findings in the context of what is known about excess dietary ω3.
      3. I would argue that the control experiments with saturated fatty acids (i.e., Behenic Acid in figure 1), represent a scenario mimicking ω3 deficiency as the enrichment of Behenic Acid causes an overall reduction in PUFAs (Figure EV1C - an increase in SFA must correspond to a decrease in PUFA). These Behenic acid results are the only experiments presented by the authors that mimic a scenario resembling ω3 deficiency and the results show that the D2R internalization is unaffected (Figure 1G-H). Therefore, I would further argue that if anything, the authors results suggest that ω3 deficiency is NOT correlated to D2R internalization. Again, the authors must rationalize these findings in the context of what is known about dietary intake of ω3's.
      4. It's not clear why the authors decided to include an ω6 fatty acid in this study. The authors built up a detailed rationale for investigating ω3's as they are dietarily essential and tied to disease when deficient. To my knowledge, ω6's are considered much less beneficial than ω3's in a dietary sense. The inclusion of an ω6 almost seems coerced as the ω6-related results don't provide any interesting additional insights. It would benefit the manuscript if the authors provided some additional discussion explaining why ω6's are being investigated in addition to ω3's.
      5. In Figure EV1D, the AHA and DPA percentages each increase by ~6%. The corresponding Figure EV1B indicates that the overall PUFA% in the plasma membrane also increases by 6%. This makes sense as the total change in PUFA content is consistent with the amount of AHA or DPA being internalized to cells. However, this consistency was not observed with BA and SFAs. In Figure EV1E, the BA percentage increases only ~1% while the total SFA percentage in Figure EV1C increases by ~6%. How can something undergoing a 1% change (relative to total lipid content) result in a 6% overall change in SFA content?
      6. In Figure 4, the discussion of kinetics does not make sense. How exactly are kinetics being monitored in this figure? (Recruitment kinetics are discussed in panels D and G)
      7. In Figure 5, What is the purpose of panel D? Would it be more helpful to include additional, overlaid "cumulative N" plots for scenarios in which PUFAs were enriched? This would work well in conjunction with panel F.
      8. For the readers who are new to this area or unfamiliar with the assays used, Figure 1 is not intuitive and initially difficult to interpret. It would greatly benefit the flow of the manuscript if Figures EV1A-C and EV2A were included in the main text and "Normalized R" was clearly defined in the main text, prior to discussion of Figure 1.

      Significance

      General assessment: The work, for the most part, is rigorous and scientifically sound. The authors utilize impressive, quantitative assays to expand our understanding of protein-lipid interactions. However, the authors need to improve their discussion of the actual physiological conditions that correspond to their experimental results.

      Advance: This work may fill a gap in our understanding of disorders related to the dopamine D2 receptor. However, some of the results may be at odds with what is currently known/understood about dietary ω3 fatty acids.

      Audience: This work will be of broad interest to researchers in the biophysics field, with particular emphasis on researchers who study protein and membrane biophysics. This work will also be of interest to researchers who study membrane molecular biology.

      Reviewer Expertise: quantitative fluorescence spectroscopy and microscopy; membrane biophysics; protein-lipid interactions

    3. 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 only conclusion that I was able to understand from the study was that enrichment of cell membranes with polyunsaturated fatty acids specifically inhibited agonist-induced internalization of D2 receptors. However, I think that the experiments used to conclude that PUFAs do not alter D2R clustering but reduce the recruitment of β-arrestin2 and D2R endocytosis need some clarification (i.e. data depicted in Fig. 2-5). This lack of clarity might be due to the fact I am not familiar enough with the employed technologies or to the unclear writing style of the paper . There was an overuse of acronyms, initialisms and abbreviations, which are difficult to understand for researchers outside of the specific lipid field. I think that the manuscript should be written in a way to be legible also for researchers not working in the immediate filed.

      The paper was not written in a manner that a general audience of cell biologists or those interested in GPCR biology could understand and judge. It is indeed interesting that polyunsaturated fatty acids specifically inhibit D2R internalization in HEK293 cells, and it could be significant. But, it is difficult to judge the significance of the observation without more in vivo data.

      I would suggest the following. Remove all acronyms and abbreviations. Significantly, expand the Materials and Methods section, either in the manuscript or in the Supplemental section. I suggest clearly explaining each construct used, and the function of each module in the construct, with diagrams. In addition, provide a comprehensive step by step description of each experimental protocol, providing the reader with the rationale for each step in the protocol with explanatory diagrams. The authors should also more clearly explain the rationale and logic that was utilized to make the conclusions that they did from the depicted observations. Only then can a broader audience determine if the authors' conclusions are justified.

      Significance

      In summary, I will reiterate that the reported experiments need to be much better explained to make the study understandable to a broader audience and for that audience to determine whether the conclusions are justified.

    4. 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, using different live and fixed cell trafficking assays, demonstrates that incorporation of poly-unsaturated, but not saturated, free fatty acids in the membrane phospholipids reduce agonist induced internalization of the D2 dopamine receptor but not the adrenergic beta2 receptors or the transferrin receptor. Pulsed pH (ppH) live microscopy further demonstrated that the reduced internalization by incorporation of free fatty acid was accompanied by a blunted recruitment of Beta-arrestin for the D2R.

      I believe said claims put forward in the manuscript are overall well supported by the data and as such I do not believe that further experiments are necessarily needed to uphold these key claims. Also, the methodology is satisfactorily reported, and statistics are robust, although two-way Anova like used in Fig 1 seems appropriate for Fig 2 and 3

      That said, I suggest that the fixed cell internalization experiments (Fig 2 and 3), which relate the effect on the D2R to B2AR and transferrin are revised. This is important since this is relevant to judge whether the effect is a general or a selective molecular mechanism since this is the one of the three assay which this comparison relies on. Alternatively, I suggest omitting this data and include the B2AR in the Live DERET assay and both B2AR and TfR in the ppH assay. Specifically, my concerns with the fixed cell internalization are:

      • The analysis is based on counting the number of endosomes, which is not necessarily equivalent to the number of receptors internalized
      • The analysis relies on fully effective stripping of the surface pool of receptors - i.e clustered surface receptors not stripped by the protocol will be assessed as internalized. It is often very difficult to obtain full efficiency of the Flag-tag stripping and this is somewhat expression dependent.
      • The protocol for the constitutive and agonist induced internalization is different and yet shown on the same absolute graph. Although I take it the microscope gain setting are unaltered between the constitutive and agonist induced internalization I don't believe the quantification can be directly related. This is confusing at the very least. More critically however, the membrane signal from the non-stripped condition of constitutive internalization will likely fully shield internalized receptors in the Rab4 membrane proximal recycling pathway leading to under-estimation of the in the constitutive endocytosis. I believe this methodological limitation underlies the massive relative difference in the constitutive endocytosis between panel 2A,B and 2C,D. For comparison, by a quantitative dual color FACS endocytosis assay, we have previously demonstrated the ligand endocytosis a ~4 fold increased over constitutive (in concert with Fig 2A,B here) (Schmidt et al 20XX). Importantly, high relative variability by this methodology could well shield an actual effect of incorporation of FFAs on the constitutive endocytosis.

      'Optional' Also, it would be informative to see the ppH Beta-arrestin experiments with the B2AR to assess, whether the putative discrepancy between D2R and B2AR is upstream or downstream of the blunted Beta-arrestin recruitment. To the same point, it would be very informative to assess how the incorporation of the free fatty acids affect receptor signalling, which would also help relate the effect of incorporation of the FFA's in the phospholipids to previous experiment using short term incubation with FFA's

      References overall seem appropriate although Schmidt et al would be relevant for reference of the constitutive vs agonist induced endocytosis of D2R and B2AR. Overall, the figures are well composed and convey the messages fairly well. Specific point that would strengthen the rigor include:

      • Chosing actual representative pictures of the qunatiative data in Fig 2 and 3 (e.g. har to see 25 endocytic events in Fig 2A constitutive endo, EtOH)
      • Showing actual p values for the statistical comparisions

      Moreover, for ease of reading the figures (without consulting the legend repeatedly) it would be very helpful to headline individual panel with what the experiments assesses. Figure 1a and 1b for example can't be distinguished at all before reading the figure legend. Also, y-axis could be more informative on what I measured rather than just giving the unit.

      Finally, the figure presentation and description of S1 is very hard to follow. I cannot really make out what is assessed in the different panels.

      Significance

      The strength of the manuscript is the use and validation of incorporation of FFA's in the plasma membrane, which more closely mimics the physiological situation than brief application of FFAs as often done. Is addition, the blunted recruitment of beta-arrestin as assessed by the ppH protocol is quite intriguing mechanistically. The limitation are the relative narrow focus on the D2 receptor (and not multiple GPCRs) that does not really speak to as or assess the physiological, pathophysiological or therapeutic role of the observations (except from referring the relation between FFAs and disease). Also, despite the putative role of Beta-arrestin recruitment in the process, the actual causation in the process is not clear. This shortcoming is underscored by the putative effect on the constitutive internalization described above.

      My specific expertise for assessing the paper is within general trafficking processes (including the trafficking methodology applied), trafficking of GPCRs and function of the dopamine system including the role of D2 receptors.

    1. 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 manuscript, Széliová et al. used a simple self-replicating cell model to study why the ribosome consists of both RNA and protein from an economic point of view. Their base model predicts an RNA-only ribosome, which is not surprising since the smaller RNAP has a higher turnover number compared to the larger ribosome. When rRNA instability is included, the model predicts an "RNA+Protein" ribosome. In particular, the predicted ribosome composition is comparable to the measured ribosome composition when strong cooperative binding of ribosomal proteins to rRNA is considered. The authors conclude that the maximal growth rate is achieved by the real ribosome composition when rRNA instability is taken into account.

      Major comments:

      1. The authors modeled the rRNA degradation rate as a function of the concentration of fully assembled ribosomes (equation 5). However, only partially assembled ribosomes are susceptible to RNase, and they make up only a small fraction of total ribosomes. The majority of ribosomes are fully assembled. In addition, the turnover number obtained from Fazal et al. (2015) and used here is the degradation rate of double-stranded RNA, not the fully assembled ribosomes, which have a stable tertiary structure. In my opinion, the rRNA degradation rate should be modeled as a function of the concentration of partially assembled ribosomes (i.e., pre-R in Figure 7) rather than the concentration of fully assembled ribosomes.
      2. Compared to the work by Kostinski and Reuveni (2020), the authors have made an improvement by avoiding the use of constant ribosome allocation to ribosomal protein (Φ_rP^R) and RNAP (Φ_RNAP^R), allowing these parameters to vary with predicted growth rates (by changing 𝑥_rP). This is indeed important, as bacteria are very likely to adjust these parameters in response to different growth conditions. However, certain other growth rate-dependent parameters are still treated as constants (or treated as nutrient-specific parameters) across predicted growth rates under given conditions. For example, experiments have shown that the fraction of active RNAP (f_RNAP^act) and the ribosome elongation rate (k_R^el) are growth rate-dependent (Bremer and Dennis, 1996). In contrast, when the authors predict the maximum growth rate by changing 𝑥_rP, f_RNAP^act and k_R^el are held constant regardless of the predicted growth rates.
      3. If amino acids or nucleotides are provided in the media, the cell does not have to synthesize all of them de novo. However, the model assumes that the cell always synthesizes all amino acids or nucleotides de novo for growth on growth on amino acid-supplemented media or on LB. This problem could in principle be solved by assuming very fast kinetics of the metabolic reactions in these media, but that should be discussed in the manuscript. Furthermore, why does the turnover number for EAA depend on the growth rate while that of ENT is constant?
      4. All parameters related to transcription (RNAP) and translation (ribosome) used in this manuscript are adopted from Kostinski and Reuveni (2020), which are slightly modified based on Bremer and Dennis' research (1996, 2008). However, the authors changed some of the original parameters or data points, but did not provide explanations for these changes:

      (a) The original data depicted a growth rate-dependent translation elongation rate, but Table 2 presents it as a constant value.

      (b) Figure 2b displays five experimental data points, as opposed to the six data points in the original dataset and other figures in this manuscript.

      (c) The model does not consider the fraction of RNAP transcribing rRNA (Φ_rRNA^RNAP), except in Appendix Figure 4. In the original data (Bremer and Dennis 1996), the fraction of RNAP transcribing rRNA increases dramatically with growth rate; however, in this study, it remains constant at 1. Furthermore, Φ_rRNA^RNAP was first introduced in line 205 but was not explained until line 337. The value(s) of Φ_rRNA^RNAP for Appendix Figure 4 are also missing from this manuscript. 5. How, exactly, is the unit of flux converted to mmol g-1 h-1? 6. What is the (dry) mass constraint and how is it defined? In the manuscript, both the second equation in line 101 and the bottom row of Table 1 are dry mass constraint(s). Why are they different? Furthermore, why is the right-hand side of the second equation in line 101 a dimensionless 1, and how does the last row of Table 1 result in the unit of growth rate, time^(-1)? 7. The concentrations of all components that serve as "substrates" will be zero when growth rate is maximized, as these molecules do not catalyze any reactions, nor do they influence reaction kinetics in the model. These "0" concentration components are C, AA, NT, rP, and rRNA. Why are these concentrations even included in the model?

      Minor comments:

      1. Questions regarding Figure 2:

      (a) The explanation of Figure 2a is unclear. Intuitively, I assumed that it was a comparison between model predictions and experimental data, with the points representing experimental data and the line representing predictions; and the authors wrote in the figure legend "The points represent maximum growth rates in six experimental conditions". However, the growth rates shown in the figure do not match the original experimental data. Are all the data in the figure predictions?

      (b) Figure 2b is difficult to understand. This figure shows the non-optimal solutions of the model. It is unclear how these solutions are achieved and why there are three lines in the figure. 2. Table 1 is also difficult to understand. While the stoichiometric constraints can be easily derived, the capacity constraints and the dry mass constraint cannot be easily derived from related equations from the text.<br /> 3. As the authors ask a question in the title, they should provide an explicit answer in the abstract. 4. The authors should cite a seminal modeling paper, which was the first to examine resource allocation in simplified self-replicating cell systems (Molenaar et al. 2009, Molecular Systems Biology 5:323). 5. The meaning of v is not consistently defined throughout the manuscript. It refers to the fluxes of enzymatic reactions in some instances, but in other contexts, it refers to the fluxes of the entire network of enzymatic reactions and protein synthesis reactions (Figure 1, Equation 1, and Line 386). 6. Line 85, it might be difficult to interpret "RNAP fluxes" as the flux of rRNA synthesis without reading the subsequent text. 7. Typo in line 102-103. "...protein fluxes 𝒘" → "...protein synthesis fluxes 𝒘". 8. Line 104, f_RNAP^act and f_R^act are not explained in the text; and their biological significance cannot be understood from their names in Table 2 ("RNAP activity" and "Ribosome activity"). 9. Notion "**" in Table 2. The coupling between transcription and translation means the coupling of "mRNA" transcription and translation, not rRNA. At least in E. coli, the transcription rate of rRNA is faster than that of mRNA. 10. Is the citation correct in line 136? I didn't find related information in Bremer and Dennis' paper after a quick scan.<br /> 11. Lines 136-138. The statement is not accurate, as the fraction of inactive ribosomes increases with decreasing growth rate in E. coli (Dai et al. 2016, Nat Microbiol 2, 16231). If the studied growth rates are relatively high, it is acceptable to use a constant active ribosome fraction as an approximation, but this approximation should be made explicit. 12. The citation in line 142 is not accurate. It should be (Bremer and Dennis, 1996). 13. Lines 192-193: "six" different growth media, not five. 14. Line 287: The statement "... translation rate does not increase in ribosomes with a higher protein content" could be misinterpreted as discussing translation elongation rate changes with different protein content in ribosomal protein mutant strains in a given species. It should be rephrased to remove ambiguity. 15. Parameters for the three panels in Figure 8 are missing.

      Significance

      Strengths: Why the ribosome is composed of RNA and protein parts is a fundamental biological question. This manuscript proposes a very interesting hypothesis, arguing that the mixed ribosome composition results from rRNA instability. To test their hypothesis, the authors parameterize a simplified self-replicating cell model with realistic parameters. The model is first developed/parameterized for E. coli, and it could be easily adapted to other organisms with higher ribosomal protein content.

      Limitations: The main limitations of this manuscript lie in the development of the model, especially the modeling of rRNA degradation and the use of constant values for growth rate-dependent parameters.

      Advances: (1) This manuscript proposes a new hypothesis that rRNA instability is a universal factor that influences the ribosome composition across living organisms. (2) Compared to Kostinski and Reuveni's work, the authors have made certain improvements by including adjustable ribosome allocation to RNA and ribosomal protein when maximizing growth rate, which may lead to more realistic conclusions.

      Audience: This work will be of interest to people in the field of theoretical biology, computational biology, and evolution, as well as to researchers studying ribosome structure and function.

      Areas of expertise: Microbial systems biology, computational biology, and prokaryotic genomics.

  2. May 2024
    1. Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

      Learn more at Review Commons


      Reply to the reviewers

      General Statements

      We thank all three reviewers for their time and care in reviewing our manuscript, in particular Reviewer 3 for providing a detailed critique that was very useful for planning revisions. We are grateful that all three reviewers indicate that the new genome resources presented in this work are of high-quality and address an existing knowledge gap. We are also grateful for general assessments that the manuscript is 'well-written', and the analyses 'well performed' and 'thorough'.

      We acknowledge Reviewer 3’s legitimate criticism that the assembly and annotation data is not already publicly available and would like to assure the reviewing team that we have been pressing NCBI to progress the submission status since before the preprint was submitted. We regret the delay but hope that we can resolve this issue promptly. Furthermore, as some additional fields in the REAT genome annotation are lost during the NCBI submission process, we will ensure that comprehensive annotation files are also added to Zenodo.

      Reviewer 3 also made the general comment that 'the manuscript could greatly benefit from merging the result and discussion sections' and we would naturally be happy to make this adjustment if the journal in question uses that format.

      Description of the planned revisions

      • We will follow suggestions by Reviewer 3 to improve clarity of two figures:

      Figure S9: Please use a more appropriate colour palette. It is difficult to know the copy number based on the colour gradient.

      Figure 5: Consider changing panel B for a similar version of Fig S12. I think it gives a cleaner and more general perspective of the presence of starship elements.

      • We will address the choice of LOESS versus linear regression for investigating the relationship between candidate secreted effector protein (CSEP) density and transposable element (TE) density, as queried by Reviewer 3:

      Lines 140-144: LOESS smoothing functions are based on local regressions and usually find correlations when there are very weak associations. The authors have to justify the use of this model versus a simpler and more straightforward linear regression. My suspicion is that the latter would fail to find an association. Also, there is no significance of Kendall's Tau estimate (p-value).

      We agree with the reviewer, that as we did not find an association with the more sensitive LOESS, we expect that linear regression would also not find an association, supporting our current conclusions. We will add this negative result into the text.

      • We will check for other features associated with the distribution of CSEPs, as queried by Reviewer 3:

      Lines 157-163: Was there any other feature associated with the CSEP enrichment? GC content? Repetitive content? Centromere likely localisation?

      • We will integrate TE variation into the PERMANOVA lifestyle testing, as suggested by Reviewer 3:

      Line 186: Why not to test the variation content of TEs as a factor for the PERMANOVA?

      In reviewing this suggestion, we also spotted an error in our data plotting code, and the PERMANOVA lifestyle result for all genes will be corrected from 17% to 15% in Fig. 4a. Correcting this error does not impact our ultimate results or interpretation.

      • To complement the current graphical-based assessment of approximate data normality, we will include additional tests (Shapiro-Wilk for sample sizes

      Line 743: Q-Q plots are not a formal statistical test for normality.

      • One of the main critiques from Reviewer 3 was that, although we already acknowledged low sample sizes being a limitation of this work, the manuscript could benefit from reframing with greater consideration of this factor. They also highlighted a few specific places in the text that could be rephrased in consideration of this:

      Line 267: "Multiple strains" can be misleading about the magnitude.

      Lines 305-307: The fact that there is significant copy number variation between the two GtA strains suggests that the variation in the GtA lineage has not been fully captured and that there may be an unsampled substructure. Although the authors acknowledge the need for pangenomic references, they should recognize this limitation in the sample size of their own study, especially when expressing its size as "multiple strains" (line 267).

      Lines 314-317: Again, the sample size is still very small and likely not representative. It suggests UNSAMPLED substructure even for the UK populations.

      Line 164 (and whole section): I would invite the authors to cautiously revisit the use of the terms "core", "soft core". The sample size is very low, as they themselves acknowledge, and probably not representative of the diversity of Gaeumannomyces.

      We intend to edit the text to address this, including removal of both text and figure references to ‘soft-core’ genes, as we agree the term is likely not meaningful in this case, and removing it has no bearing on the results or interpretation.

      Description of the revisions that have already been incorporated in the transferred manuscript

      • We have amended the text in a number of places for clarity/fluency as suggested by Reviewer 3:

      ii) There need to be an explicit conclusion about the differences between pathogenic Gt and non-pathogenic Gh. Somehow, this is not entirely clear and is probably only a matter of rephrasing.

      Please see new lines 477-478: ‘Regarding differences between pathogenic Gt and non-pathogenic Gh, we found that Gh has a larger overall genome size and greater number of genes.’

      Lines 309-314: The message seems a bit out of context in the paragraph.

      This is valid, these lines have now been removed.

      Lines 392-395: The idea that crop pathogenic fungi are under pressure that favours heterothallism does not take into account the multiple cases of successful pathogenic clonal lineages in which sexual reproduction is absent. This paragraph seems very speculative to me. Please rephrase it.

      Our intention here was the exact reverse, that crop pathogens are under pressure to favour homothallism (as Reviewer 3 points out, anecdotally this often seems to play out in nature). We have rephrased lines 386-390 to hopefully make our stance more explicit: 'Together, this could suggest a selective pressure towards homothallism for crop fungal pathogens, and a switch from heterothallism in Gh to homothallism in Gt and Ga may, therefore, have been a key innovation underlying lifestyle divergence between non-pathogenic Gh and pathogenic Gt and Ga.'

      Lines 463-464: Please refer to the analyses when discussing the genetic divergence.

      We have rephrased this sentence to make our intended point clearer, please see new lines 459-461: ‘If we compare Ga and Gt in terms of synteny, genome size and gene content, the magnitude of differences does not appear to be more pronounced than those between GtA and GtB.’

      • We have also fixed the following typographic errors highlighted by Reviewer 3:

      Line 399: You mean, Fig 4C?

      Line 722: You missed "trimAI"

      Lines 723-727: Missing citations for "AMAS" and RAxML-NG, "AHDR" and "OrthoFinder"

      • We have added genome-wide RIP estimates to Supplementary Table S1 as requested by Reviewer 3:

      Lines 416-422: Please provide the data related to the genome-wide estimates of RIP.

      • We have added a note clarifying that differences in overall genome size between lineages are not fully explained by differences in gene copy-number (lines 406-408: 'We should note that the total length of HCN genes was not sufficiently large to account for the overall greater genome size of GtB compared to GtA (Supplemental Table S1).') in response to a comment from Reviewer 3:

      Line 396: The difference in duplicated genes raises the question of whether there are differences in overall genome size between lineages and, if so, whether they can be explained by the presence of genes.

      • We have made an alteration to the author order and added equal second-author contributions.

      Description of analyses that authors prefer not to carry out

      • In response to our analysis regarding the absence of TE-effector compartmentalisation in this system, Reviewer 1 requested additional analyses:

      While TE enrichment is typically associated with accessory compartments, it is not a defining feature. To bolster the authors' claim, it is essential to demonstrate that there is no bias in the ratio of conserved and non-conserved genes across the genomes.

      We believe that there are two slightly different compartmentalisation concepts being somewhat conflated here – (1) the idea of compartments where TEs and virulence proteins such as effectors are significantly colocalised in comparison with the rest of the genome, and (2) the idea of compartments containing gene content that is not shared in all strains (i.e. accessory). The two may overlap – as Reviewer 2 states, accessory compartments may also be enriched with TEs – but not necessarily. We specifically address the first concept in our text, and we appreciate Reviewer 3’s response on this subject:

      There is a clear answer for the compartmentalisation question. The authors favour the idea of "one-compartment" with compelling analyses.

      We believe that the second concept of accessory compartments is shown to be irrelevant in this case from our GENESPACE results (see Fig. 2), which demonstrate that gene content is conserved, broadly syntenic even, across strains, with no clear evidence of accessory compartments or chromosomes regarding gene content. We have already acknowledged that other mechanisms of compartmentalisation beyond TE-effector colocalisation may be at play (as seen from our exploration of effector distributions biased towards telomeres, see section from line 156: ‘Although CSEPs were not broadly colocalised with TEs, we did observe that they appeared to be non-randomly distributed in some pseudochromosomes (Fig. 3a)…’).

      • Reviewer 1 questioned the statement that higher level of genome-wide RIP is consistent with lower levels of gene duplication:

      L422: Is the highest RIP rate in GtA consistent with its low levels of gene duplication? Does this suggest that duplicated sequences in GtA are no longer recognizable due to RIP mutations? This seems counterintuitive, as RIP is primarily triggered by gene duplication.

      Our understanding is that, while RIP can directly mutate coding regions, it predominantly acts on duplicated sequences within repetitive regions such as TEs (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-020-02060-w), which has a knock-on effect of reducing TE-mediated gene duplication. In Neurospora crassa, where RIP was first discovered and thus the model species for much of our understanding of the process, a low number of gene duplicates has been linked to the activity of RIP (https://www.nature.com/articles/nature01554). We therefore believe the current text is reasonable.

      • Reviewer 2 stated that experimental validation of gene function is required to make clear links to lifestyle or pathogenicity:

      In my eyes, the study has two main limitations. First of all, the research only concerns genomics analyses, and therefore is rather descriptive and observational, and as such does not provide further mechanistic details into the pathogen biology and/or into pathogenesis. This is further enhanced by the lack of clear observations that discriminate particular species/lineages or life styles from others in the study. Some observations are made with respect to variations in candidate secreted effector proteins and biosynthetic gene clusters, but clear links to life style or pathogenicity are missing. To further substantiate such links, lab-based experimental work would be required.

      We agree that in an ideal world supportive wet biology gene function experimental evidence would be included. Unfortunately, transformation has not been successfully developed yet in this system (see lines 33-35: ‘There have also been considerable difficulties in producing a reliable transformation system for Gt, preventing gene disruption experiments to elucidate function (Freeman and Ward 2004).’) not for lack of trying – after 18 months of effort using all available transformation techniques and selectable markers neither Gt or Gh was transformable. Undertaking that challenge has proven to be far beyond the scope of this paper, the purpose of which was to generate and analyse high-quality genomic data, a major task in itself. We again appreciate Reviewer 3’s response to this point, agreeing that it is out of scope for this work:

      I just want to respectfully disagree with reviewer #2 about the need for more experimental laboratory work, as in my opinion it clearly goes beyond the intention and scope of the submitted work. This could be a limitation that would depend on the chosen journal and its specific format and requirements. Finally, I think it would suffice for the authors to discuss on the lack of in-depth experimental work as part of the limitations of their overall approach.

      As per the suggestion by Reviewer 3, we will add text to address the absence of in-depth experimental work within the scope of this study.

      • Reviewer 3 suggested we might 'consider including formal population differentiation estimators', however, as they previously highlighted above, our sample sizes are too small to produce reliable population-level statistics.

      • Reviewer 3 raised the disparity in the appearance of branches at the root of phylogenetic trees in various figures:

      Figure 4a (and Figs S5, S13): The depicted tree has a trichotomy at the basal node. Please correct it so Magnaporthiopsis poae is resolved as an outgroup, as in Fig. S17.

      All the trees were rooted with M. poae as the outgroup, and although it may seem counterintuitive, a trifurcation at the root is the correct outcome in the case of rerooting a bifurcating tree, please see this discussion including the developers of both leading phylogeny visualisation tools ggtree and phytools (https://www.biostars.org/p/332030/). Although it is possible to force a bifurcating tree after rooting by positioning the root along an edge, the resulting branch lengths in the tree can be misleading, and so in cases where we wanted to include meaningful branch lengths in the figure (i.e. estimated from DNA substitute rates, in Figures 4a, S5 and S13) we have not circumvented the trifurcation. In Fig S17 meaningful branch lengths have not been included and the tree only represents the topology, resulting in the appearance of bifurcation at the root.

      • Reviewer 3 suggested that the discussion on giant Starship TEs resembled more of a review:

      Lines 434-451: This section resembles more a review than a discussion of the results of the present work. This also highlights the lack of analysis on the genetic composition and putative function of the identified starship-like elements.

      The reviewer has a valid point. However, Starships are a recently discovered and thus underexplored genetic feature that readers from the wider mycology/plant pathology community may not yet be aware of. We believe it is warranted to include some additional exposition to give context for why their discovery here is novel, interesting and unexpected. We are naturally keen to investigate the make-up of the elements we have found in this lineage, however that will require a substantial amount of further work. Analysis of Starships is not trivial, for example the starfish tool is still under development and a limited number of species have been used to train it. How best to compare elements is also an active area of investigation – they are dynamic in their structure and may include genes originating from the host genome or a previous host – and for this reason we believe is out of scope to interrogate alongside the other foundational genomic data presented in this paper.

    2. 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 "Evolutionary genomics reveals variation in structure and genetic content implicated in virulence and lifestyle in the genus Gaeumannomyces" by Rowena Hill and collaborators is a thorough, well-planned and designed work. They have described 9 almost complete new assemblages, from their most general characteristics to their genetic content and implications. I am very pleased with the quality and completeness of this work and agree that it provides a very useful resource and framework for further research on this important organism.

      The three main motivations of the present study were:

      1) Are there genomic signatures distinguishing Gt A/B virulence lineages?;

      2) How do gene repertoires differ between pathogenic Gt and non-pathogenic Gh? And, iii) Is there evidence of genome compartmentalisation in Gaeumannomyces?

      a) The authors themselves recognise the low number of samples in their work (Lines 453-454) and this limitation hampers the establishment of a clear association between lineage-specific virulence and genomic signatures. I would argue that the present work needs to be reframed factoring this main limitation.

      b) There need to be an explicit conclusion about the differences between pathogenic Gt and non-pathogenic Gh. Somehow, this is not entirely clear and is probably only a matter of rephrasing.

      c) There is a clear answer for the compartmentalisation question. The authors favour the idea of "one-compartment" with compelling analyses.

      Major comments:

      The authors have not published the genomic data. Therefore, it is impossible to audit the quality of the assemblies and impedes its reproducibility. It is also bad practice by current scientific standards.

      I strongly believe that the manuscript could greatly benefit from merging the result and discussion sections. It would be easier for the reader to follow their entire logic. This is of course something optional and contingent on the journal format.

      Minor and specific comments:

      RESULTS

      • Lines 140-144: LOESS smoothing functions are based on local regressions and usually find correlations when there are very weak associations. The authors have to justify the use of this model versus a simpler and more straightforward linear regression. My suspicion is that the latter would fail to find an association. Also, there is no significance of Kendall's Tau estimate (p-value).

      • Lines 157-163: Was there any other feature associated with the CSEP enrichment? GC content? Repetitive content? Centromere likely localisation?

      • Line 164 (and whole section): I would invite the authors to cautiously revisit the use of the terms "core", "soft core". The sample size is very low, as they themselves acknowledge, and probably not representative of the diversity of Gaeumannomyces.

      • Figure 4a (and Figs S5, S13): The depicted tree has a trichotomy at the basal node. Please correct it so Magnaporthiopsis poae is resolved as an outgroup, as in Fig. S17.

      • Line 186: Why not to test the variation content of TEs as a factor for the PERMANOVA?

      • Figure S9: Please use a more appropriate colour palette. It is difficult to know the copy number based on the colour gradient.

      • Figure 5: Consider changing panel B for a similar version of Fig S12. I think it gives a cleaner and more general perspective of the presence of starship elements.

      DISCUSSION

      • Line 267: "Multiple strains" can be misleading about the magnitude.

      • Lines 305-307: The fact that there is significant copy number variation between the two GtA strains suggests that the variation in the GtA lineage has not been fully captured and that there may be an unsampled substructure. Although the authors acknowledge the need for pangenomic references, they should recognize this limitation in the sample size of their own study, especially when expressing its size as "multiple strains" (line 267).

      • Lines 309-314: The message seems a bit out of context in the paragraph.

      • Lines 314-317: Again, the sample size is still very small and likely not representative. It suggests UNSAMPLED substructure even for the UK populations.

      • Lines 392-395: The idea that crop pathogenic fungi are under pressure that favours heterothallism does not take into account the multiple cases of successful pathogenic clonal lineages in which sexual reproduction is absent. This paragraph seems very speculative to me. Please rephrase it.

      • Line 396: The difference in duplicated genes raises the question of whether there are differences in overall genome size between lineages and, if so, whether they can be explained by the presence of genes.

      • Line 399: You mean, Fig 4C?

      • Lines 416-422: Please provide the data related to the genome-wide estimates of RIP.

      • Lines 434-451: This section resembles more a review than a discussion of the results of the present work. This also highlights the lack of analysis on the genetic composition and putative function of the identified starship-like elements.

      • Lines 463-464: Please refer to the analyses when discussing the genetic divergence. Consider including formal population differentiation estimators.

      METHODS

      • Line 722: You missed "trimAI"

      • Lines 723-727: Missing citations for "AMAS" and RAxML-NG, "AHDR" and "OrthoFinder" Line 743: Q-Q plots are not a formal statistical test for normality.

      Referees cross-commenting

      I agree with my peer reviewers and appreciate that we have shared common concerns and suggestions. I also agree with their comments.

      I just want to respectfully disagree with reviewer #2 about the need for more experimental laboratory work, as in my opinion it clearly goes beyond the intention and scope of the submitted work. This could be a limitation that would depend on the chosen journal and its specific format and requirements. Finally, I think it would suffice for the authors to discuss on the lack of in-depth experimental work as part of the limitations of their overall approach.

      Significance

      The work presented by Hill and co-workers contributes to the understanding of the genetic basis of host-pathogen interactions and evolutionary dynamics in the important fungus responsible for wheat "take-all-disease", Gaeumannomyces tritici. By providing 9 new near-complete assemblages, this work will provide a valuable resource for research on this agronomically important organism. This work sets the stage for developing a global pangenome of G. tritici that can expand analyses of its population structure and specific genetic elements that are associated with its virulence.

    3. 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, the authors present genome assemblies for nine strains of the genus Gaeumannomyces, including 5 strains that belong to two different virulence lineages of the wheat take-all decline pathogen G. tritici, 2 strains of the antagonist G. hyphopodioides and 2 of the oat take-all decline pathogen G. avenae. The authors assess gene catalogs, CAZyme repertoires, effector catalogs, TE abundance, compartmentalisation and the occurrence of Starship giant transposable elements. Overall, there are no striking differences that discriminate the genomes and that can be linked to differential life styles. Weak correlations were found for some of the different lineages, but no functional analyses have been performed to further solidify such differences.

      Significance

      • Overall, the study fills a knowledge gap, given that no-few high quality genomes for the soil-borne fungi of the Gaeumannomyces genus are available. The genome assemblies are of high quality, and the work that is presented is mainly solid and robust. The analyses are well performed, sound and informative.

      • In my eyes, the study has two main limitations. First of all, the research only concerns genomics analyses, and therefore is rather descriptive and observational, and as such does not provide further mechanistic details into the pathogen biology and/or into pathogenesis. This is further enhanced by the lack of clear observations that discriminate particular species/lineages or life styles from others in the study. Some observations are made with respect to variations in candidate secreted effector proteins and biosynthetic gene clusters, but clear links to life style or pathogenicity are missing. To further substantiate such links, lab-based experimental work would be required.

    4. 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 Hill et al. presents nearly complete genomes of nine Gaeumannomyces strains, including both phytopathogenic and non-pathogenic (symbiotic) fungi. The manuscript is well-written, and the data it presents are of high quality, offering implications for understanding the evolution and diversification of Magnaporthales fungi, which encompass economically important phytopathogenic species such as Gaeumannomyces graminis and Pyricularia oryzae. I believe that the determination of these nearly complete genomes alone justifies publication. However, I have some concerns as described below.

      Major concern:

      One potential criticism pertains to whether the authors' assertion that Gaeumannomyces taxa have one-compartment genomes is fully supported by the data. The authors demonstrate in this manuscript that transposable elements (TE) and putative effector genes (CSEPs) are not co-localized in the Gaeumannomyces genomes. However, this evidence may not be robust enough to substantiate their claim. The concept of two- or multi-speed genomes suggests that fungal genomes consist of compartments that differ in the rate of evolution but not necessarily in TE content. While TE enrichment is typically associated with accessory compartments, it is not a defining feature. To bolster the authors' claim, it is essential to demonstrate that there is no bias in the ratio of conserved and non-conserved genes across the genomes.

      Minor concern:

      L422: Is the highest RIP rate in GtA consistent with its low levels of gene duplication? Does this suggest that duplicated sequences in GtA are no longer recognizable due to RIP mutations? This seems counterintuitive, as RIP is primarily triggered by gene duplication.

      In my opinion, the analysis of the genomic differences facilitating parasitic and symbiotic lifestyles seems somewhat weak.

      Significance

      This manuscript offers new genomic insights into economically important phytopathogenic fungal species, and sheds light on the diversification of parasitic and symbiotic fungi during evolution. While the analyses conducted are mostly appropriate and reasonable, they do not yield particularly surprising findings.

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

      Learn more at Review Commons


      Reply to the reviewers

      Manuscript number: RC- 2023-02122

      Corresponding author(s): Andrew Graham Cox and Juan Manuel González-Rosa

      1. General Statements

      We thank the reviewers for taking the time to assess our work and for their considered and constructive comments. We are glad that they appreciate the value of the methodology we have developed. In addressing the points raised by the reviewers, we have significantly strengthened the conclusions reached in our study. Below is a point-by-point response (in regular type, blue) to the specific reviewer comments (in italics, black).

      1. Point-by-point description of the revisions

      Experiment 1: Perform lineage tracing of hepatocytes following cryoinjury.

      Reviewer #1 would like us to have a better understanding of the origin of the regenerative hepatocytes following cryoinjury. There are two potential sources of regenerating hepatocytes. In many cases, hepatocytes proliferate giving rise to regenerative hepatocytes. However, during severe injury, the liver can undergo a ductular reaction in which biliary epithelial cells (BECs) can expand and transdifferentiate to give rise to regenerating hepatocytes.

      ● To address this query we have now used a new transgenic line created in laboratory that can indelibly label hepatocytes for lineage tracing Tg(fabp10a:Tet-ON-Cre). We have crossed this line to floxed reporters (Ubb:Switch) and collect livers at 7 dpci. The healthy parenchyma surrounding the injured area was predominantly labelled in the tracing experiment suggesting that pre-existing hepatocytes are driving the regenerative response.

      Experiment 2: Examine BEC and EC proliferation in the ventral and contralateral lobes following cryoinjury.

      Reviewers #1 and #2 would like us to better characterise the temporal dynamics of proliferation in BECs and ECs following cryoinjury. Specifically, the reviewers would like to know whether then compensatory hyperplasia in the contralateral lobe also leads to increased BEC and EC proliferation. Moreover, the reviewers would like us to better quantify the extent of EC and BEC proliferation at different stages of regeneration after cryoinjury.

      ● We have now performed extensive BrdU pulse-chase cryoinjury experiments using Tg(fli1a:nEGFP) zebrafish to visualise ECs. We have also conducted multiplexed immunostaining of the regenerating livers with the BEC marker (Anxa4) in conjunction with immunodetection of proliferation (BrdU and PCNA). These studies outline the kinetics of the regenerative response and provide evidence to support epimorphic regeneration around the site of injury as well as a compensatory hyperplasia on the contralateral lobe.

      Experiment 3: Quantification of the temporal dynamics of fibrosis upon cryoinjury.

      Reviewer #1 suggested we better characterise the extent of fibrosis in our model.

      ● We have now performed extensive studies quantifying the extent of collagen deposition at the regenerative margin over the time course (SHAM, 1, 3, 5, and 7 dpci) using immunohistochemical detection.

      Experiment 4: Examine the role of Macrophage depletion in liver regeneration.

      Reviewer #1 suggested we examine regeneration following cryoinjury in immunodeficient zebrafish in order to understand the role of macrophages in the model.

      ● To address this question, we have now performed studies involving macrophage depletion, using the well established IP injection of clodronate liposomes. We have now performed cryoinjury comparing untreated and chlodronate-treated Tg(fabp10a:NLSmCherry) or Tg(fabp10a: GreenLantern-H2B) zebrafish and examined the extent of regeneration at 3 and 7 dpci.

      Experiment 5: Examine the impact of age and gender on liver regeneration following cryoinjury.

      Reviewer #3 wanted to know if the regenerative response to cryoinjury was different depending on age and gender.

      ● To address this query, we have now performed cryoinjuries on young (4 month) and aged (9 month) males and females in a Tg(fabp10a:NLS-mCherry) or Tg(fabp10a: GreenLantern-H2B) background and examined regeneration at 7 dpci.

      Experiment 6: Characterization of the dynamics of Hepatoblasts, Hepatic Stellate Cells, Macrophages and Neutrophils following cryoinjury.

      Reviewer #3 suggested that it would be good to have a better cellular characterization of regeneration in the cryoinjury model.

      ● To address this question, we have now examined distinct cell types over the cryoinjury timecourse including SHAM, 1, 3, 5, and 7 dpci livers to provide a temporal landscape of the cellular response. In addition to BECs and ECs as discussed above, we have also performed immunofluorescence to detect macrophages (mfap4) neutrophils (mpx) during liver regeneration.

      Specific Reviewer comments

      Reviewer #1

      Major points:

      Full Revision

      1) In this cryoinjury model, the authors found cell proliferation in hepatocytes, BECs, and other cell types near the injury site. The proliferating hepatocytes exclusively provide hepatocytes, and BECs provide BECs, or some transdifferentiation is involved? Like other extreme ablation models, BECs can contribute to some hepatocytes in this model.

      We thank the Reviewer #1 for the interesting suggestion. We have addressed this by performing lineage tracing analysis as explained in Experiment 1 (above). For this approach, we have used Tg(fabp10a:Tet-ON-Cre; Ubb:Switch) to indelibly label and trace hepatocytes. These experiments reveal that the new regenerated tissue is derived from pre-existing hepatocytes (see Supplementary Figure 2 Q, R, S, T).

      2) In this model, the authors observed the long-range effect of the cryoinjury as they identified increased cell proliferation in the contralateral liver lobes. Is this long-range effect specific to hepatocytes? BECs or endothelial cells also undergo increased cell proliferation in the contralateral lobes?

      We thank the Reviewer #1 for this question. We have addressed this query by performing Experiment 2 (above). Briefly, cryoinjuries were performed and markers of proliferating HCs and BECs (PCNA or BrdU stained) were quantified in the ventral and contralateral lobes (see Supplementary Figure 6). The data clearly demonstrates that proliferation is higher at the site of injury, however lower rates of compensatory hyperplasia are still evident on the contralateral lobe. A strong epimorphic hyperplasia and weaker compensatory growth response, has been previously observed in the cardiac cryoinjury model (Pauline Sallin et al. Developmental Biology 2015).

      3) This model is a unique liver regeneration model as it induces transient focal fibrosis. Is the fibrosis beneficial for liver regeneration? What happens if you reduce fibrosis pharmacologically? Will it interfere with the rate of regeneration?

      We thank the Reviewer #1 for the comments. Although pharmacological interventions of fibrosis are beyond the scope of the current manuscript, we have better quantified the extent of fibrosis in the first week following cryoinjury in Experiment 3 (above; Figure 3I).

      4) Do Lcp1+ leucocytes contribute to liver regeneration in this model? In immunodeficiency models such as irf8 mutant, liver regeneration after cryoinjury changed?

      We thank the Reviewer #1 for the suggestion of using an immunodeficiency model. We addressed this question by performing Experiment 4 (above). Briefly, we have IP injected clodronate liposomes, which are a well-established method for macrophage depletion, and examined the effect on liver regeneration (Supplementary Figure 5). These extensive experiments showed that macrophage depletion had no significant effect on liver regeneration at 3 and 7 dpci.

      5) The CUBIC-clearing procedure is beneficial in the field. The quantitative benefit of the CUBICbased method should be added. Supplement figures 1C and D need scale bars, especially for X Z and Z-Y planes. Can you quantify the Z-plane depth that you can scan with or without CUBIC treatment?

      We thank the Reviewer #1 for the input and apologise if we did not present the current data clearly. We have now included the scale bars on the reviewed manuscript in Supplementary Figure 1C, 1D, and 1G. We have quantified the Z-plane depth on our current acquisitions and modified our current panels to make clear the difference in depth (z-stack) that CUBIC-imaging enables during liver acquisitions in Supplementary Figure 1D-I.

      6) In the manuscript, the authors measured the injured area after the cryoinjury. But how about the depth of the injury? Does the procedure induce a relatively constant injury depth, or can it not be controlled? The total volume of injured tissue would be more important than the surface injured area.

      We thank the Reviewer #1 for the comments. The hepatic cryoinjury approach was developed to injure the liver and avoid deeper tissue lesions to the gastrointestinal tract. Our existing CUBIC data suggests that injury depth remains constant.

      Minor points:

      7) The sham procedure means exposing the liver by removing the scale and cutting the skin, right? What is the survival rate of the sham procedure? Is the survival rate of sham group significantly lower than cryoinjury-induced group?

      The Reviewer #1 is correct about the cryoinjury procedure in SHAM samples. SHAM survival is 95% while the injured animal survival is 92.97% (Figure below; n= 444). This analysis shows no significant difference between the groups (unpaired Student's t-test; p-value: 0.5843)

      8) The original RNA-seq data, including FASTQ files, should be deposited to NCBI (Gene Expression Omnibus) or other public databases.

      We apologize for not submitting our Bulk RNA-seq data to NCBI GEO during the initial submission. The Bulk RNA-seq data can be found under the accession number GSE245878.

      Full Revision

      Reviewer #2

      Major points:

      1) While the authors assayed changes in major cell types during liver regeneration in this model, the selection of varying timepoints for analysis and incomplete quantification for all timepoints precludes detailed comparisons that may lead to mechanistic insights. For example, closure of injury area is assayed at 1,3,7,14 dpci but hepatocyte proliferation is measured at 1,3,5,7, 18, 30 dpi. Fibrosis was only assayed at 5 dpi (assume dpi is the same as dpci). Cholangiocytes and endothelial cells are imaged at 1, 3, 7, 30 dpci but no quantification was provided only a single image. Since most changes are occurring at 1-7 dpci, the authors should at least measure the same timepoints from 1-7 dpci for the different cell types so comparisons can be made and conclusions can be drawn. For example, does hepatocyte proliferation, which seem to peak at 5 dpci, happen before endothelial proliferation, which is measured at 3 and 5 dpci but not measured at 5 dpci?

      We thank the Reviewer #2 for the comments regarding temporal dynamics of regeneration. In response we have performed Experiment 2 (above). Briefly, this included examination of BECs and ECs at different time points during regeneration (SHAM, 1, 3, 5, and 7 dpci; Figure 6, Supplementary Figure 6L-P).

      2) Fibrosis level seems to be highly variable at 5 dpci, which is the only time point measured. If this level of variability is found across all timepoints then this might not be a good model to study the intersection of fibrosis and regeneration. Since the authors have collected animals at all timepoints, it should be fairly straight forward to carry out collagen staining and quantification across different timepoints without the need of additional fish experiments.

      We thank the Reviewer #2 for the comments regarding the fibrotic response. In response we have undertaken Experiment 3 (above). This experiment involves quantifying collagen deposition at the different timepoints (SHAM, 1, 3, 5, and 7 dpci; Figure 3I).

      3) The lack of quantification of cholangiocytes and endothelial cells makes it difficult to gauge the reproducibility of this model across different animals and experiments.

      We thank the Reviewer #2 for the comments regarding the need to quantify ECs and BECs during regeneration. In response we will undertake Experiment 2 (above). Briefly, this included examination of BECs and ECs at different time points during regeneration (SHAM, 1, 3, and 7 dpci; Figure 6 and Supplementary Figure 7).

      4) Transcriptomic data analysis/presentation in Figure 7 can be improved. Cannot read any of the gene labels in Figure 7B. Figure 7H should use at least a few different gene markers from each cell type to approximate cell abundance.

      We apologise for the inconvenience and have addressed the issue of legibility. We have increased font size on the volcano plots in Figure 7 and incorporate a new analysis with more markers for each cell type in Figure 7H. In addition, we have included the comparison between Bulk RNA-seq ventral samples and contralateral lobe samples, together with further GOenrichment of the samples in Supplementary Figure 8.

      Full Revision

      Minor:

      5) Is "dpi" the same as "dpci"? Please use the same nomenclature throughout manuscript.

      We apologise. Dpi means days post-injury and dpci means days post-cryoinjury. Nomenclature has been corrected in revised version of the manuscript.

      6) In the mouse PHx model, hepatocytes reach max proliferation (as measured with Ki67/PCNA staining) at 40-48hrs across different labs and experiments, not at 24rs.

      We thank the Reviewer #2, we have changed this reference.

      7) Zebrafish references are used when the author is talking about mouse PHx model on page 12.

      We thank the Reviewer #2, we have changed this reference 7 and 8 to reference the right papers.

      Reviewer #3

      Major points:

      1) It is not clear whether both male and female fish were used in the analyses and whether there is any gender difference in regeneration responses at cellular and molecular levels. The method mentioned that 4-9 month old fish were used in the study. Was there any difference between young and old fish?

      We thank the Reviewer #3 for the comments regarding the need to consider age and gender in regeneration studies. Our experiments have been performed on adult male zebrafish. To examine the impact of age and gender on regeneration we have performed Experiment 5 (above). In brief, we have undertaken cryoinjuries in 4 month or 9 month old females and males in the Tg(fabp10a:NLS-mCherry) or Tg(fabp10a: GreenLantern-H2B) background and examine regeneration at 7 dpci (Supplementary Figure 2 J-N and P. We could not detect a significant difference among any of these comparisons. However, we observed a subtle trend with female adult zebrafish showing smaller insult area compared to adult male zebrafish, both at 3 and 7 dpci (Supplementary Figure 2P).

      2) The authors detected increased hepatocyte proliferation following cryoinjury. It will be interesting to investigate if activation of stem cells and transdifferentiation of cholangiocytes also contribute to regeneration in this particular model.

      We thank the Reviewer #3 for the comments regarding the need to examine the potential involvement of hepatoblasts and transdifferentiating BECs in regeneration following cryoinjury. We have addressed these aspects with Experiment 6 (above). Briefly, we have performed cryoinjuries in adult zebrafish and utilised Anxa4 staining for detection of BECs at SHAM, 1, 3, 5, and 7 dpci (Figure 6A-F). This analysis showed that the were no detectable signs of transdifferentiation between hepatocytes and cholangiocytes (ie: there were no double positive cells (Anxa+/fabp10a:H2B-GreenLantern+ or fabp10a:H2B-mCherry+). Moreover, we performed lineage tracing experiments and found evidence that pre-existing hepatocytes give rise to the regenerating tissue (Supplementary Figure 2 Q-T). Together, these experiments indicate that hepatocytes are responsible for the regeneration of the liver upon cryoinjury without the necessity of BEC transdifferentiation.

      3) It will be important to characterize hepatic stellate cells, macrophages, and neutrophils in this model, given their critical and complex roles in liver regeneration. Transgenic reporter lines marking these cell types are available.

      We thank the Reviewer #3 for the comments regarding the need to examine hepatic stellate cells (HSCs), macrophages and neutrophils in regeneration following cryoinjury. We have addressed these aspects with Experiment 6 (above). Briefly, we have studied the temporal dynamics of neutrophils upon cryoinjury by immunofluorescent detection of myeloperoxidase (mpx) (Supplementary Figure 4). We have also explored the role of macrophage depletion in response to cryoinjury by performing clodronate injections. We found no significant changes in liver regeneration following clodronate injections (Supplementary Figure 5). To examine the temporal dynamics of HSCs we attempted to use two approaches, namely imaging transgenic lines labelling HSCs (Tg(BAC-pdgfrb:EGFP) and HCR for HSCs (pdgfrb), but unfortunately we were not able to detect HSCs with these approaches.

      4) It is not appropriate to call Fli1a + cells liver sinusoidal cells. As far as I know, there is no specific marker for LSEC in zebrafish. Fli1a transgene labels all vascular cells.

      We acknowledge this mistaken nomenclature and have made the necessary amendment to use the term endothelial cells (ECs).

    2. 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, Sande-Melon et al described a new model for studying liver regeneration in zebrafish that is induced by cryoinjury. They showed that this model induced hepatocyte proliferation, transient fibrosis and inflammation, and regeneration of the biliary and vascular network. Compared to the other established models, such as partial hepatectomy, drug-induced liver injury, the cryoinjury model is easy to perform, consistent, and involves shorter recovery time. Overall, it is a useful tool that complements existing liver regeneration models. The tissue clearing methodology is highly effective.

      Main critiques:

      1. It is not clear whether both male and female fish were used in the analyses and whether there is any gender difference in regeneration responses at cellular and molecular levels. The method mentioned that 4-9 month old fish were used in the study. Was there any difference between young and old fish?
      2. The authors detected increased hepatocyte proliferation following cryoinjury. It will be interesting to investigate if activation of stem cells and transdifferentiation of cholangiocytes also contribute to regeneration in this particular model.
      3. It will be important to characterize hepatic stellate cells, macrophages, and neutrophils in this model, given their critical and complex roles in liver regeneration. Transgenic reporter lines marking these cells types are available.
      4. It is not appropriate to call Fli1a + cells liver sinusoidal cells. As far as I know, there is no specific marker for LSEC in zebrafish. Fli1a transgene labels all vascular cells.

      Significance

      In this manuscript, Sande-Melon et al described a new model for studying liver regeneration in zebrafish that is induced by cryoinjury. They showed that this model induced hepatocyte proliferation, transient fibrosis and inflammation, and regeneration of the biliary and vascular network. Compared to the other established models, such as partial hepatectomy, drug-induced liver injury, the cryoinjury model is easy to perform, consistent, and involves shorter recovery time. Overall, it is a useful tool that complements existing liver regeneration models. The tissue clearing methodology is highly effective.

    3. 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 titled "Development of a hepatic cryoinjury model to study liver regeneration" by Sande-Melon et al., the authors developed a novel model to study liver regeneration, namely a cryoinjury model in adult zebrafish. The authors described the methodology in detail and extensively characterized the kinetics of liver regeneration in this model, including hepatocyte necrosis/apoptosis, the proliferation of hepatocytes, cholangiocytes, endothelial cells, and infiltration of leukocytes. Most of the characterization were performed by immunostaining for various cell markers, which the authors corroborated with transcriptomic analysis by bulk RNAseq.

      Major comments:

      • While the authors assayed changes in major cell types during liver regeneration in this model, the selection of varying timepoints for analysis and incomplete quantification for all timepoints precludes detailed comparisons that may lead to mechanistic insights. For example, closure of injury area is assayed at 1,3,7,14 dpci but hepatocyte proliferation is measured at 1,3,5,7, 18, 30 dpi. Fibrosis was only assayed at 5 dpi (assume dpi is the same as dpci). Cholangiocytes and endothelial cells are imaged at 1, 3, 7, 30 dpci but no quantification was provided only a single image. Since most changes are occurring at 1-7 dpci, the authors should at least measure the same timepoints from 1-7 dpci for the different cell types so comparisons can be made and conclusions can be drawn. For example, does hepatocyte proliferation, which seem to peak at 5 dpci, happen before endothelial proliferation, which is measured at 3 and 5 dpci but not measured at 5 dpci?
      • Fibrosis level seems to be highly variable at 5dpci, which is the only time point measured. If this level of variability is found across all timepoints then this might not be a good model to study the intersection of fibrosis and regeneration. Since the authors have collected animals at all timepoints, it should be fairly straight forward to carry out collagen staining and quantification across different timepoints without the need of additional fish experiments.
      • The lack of quantification of cholangiocytes and endothelial cells makes it difficult to gauge the reproducibility of this model across different animals and experiments.
      • Transcriptomic data analysis/presentation in Figure 7 can be improved. Cannot read any of the gene labels in Figure 7B. Figure 7H should use at least a few different gene markers from each cell type to approximate cell abundance.
      • OPTIONAL: Sheets of DAPI staining are observed in Figure 6G'. Is this DNA from necrotic cells? Could they make up a neutrophil extracellular trap (NET)-scaffold like structure that covers/protects the injury site from infection? This is purely speculative but might represent an interesting area of study.
      • OPTIONAL: To demonstrate this a useful model that complements existing models of liver regeneration, the authors can try to capitalize on the proposed strength of the model to provide some novel insights into liver regeneration. A notable feature of this model that is missing from the PHx and APAP rodent models is the development of robust fibrosis that rapidly resolves within a short time frame, providing an unique opportunity to investigate the potential crosstalk between fibrosis and regeneration that often co-occur in chronic liver disease patients.

      Minor comments:

      • Is "dpi" the same as "dpci"? Please use the same nomenclature throughout manuscript
      • In the mouse PHx model, hepatocytes reach max proliferation (as measured with Ki67/PCNA staining) at 40-48hrs across different labs and experiments, not at 24rs
      • Zebrafish references are used when the author is talking about mouse PHx model on page 12

      Significance

      Mouse 2/3 partial hepatectomy surgery (PHx) is the most frequently used model to study liver regeneration and much has been learnt from this model. However, mouse PHx involving tying off certain lobes of the liver and the inducing a sterile injury, where hepatocyte proliferation and liver regeneration occurs in the absence of significant inflammation and fibrosis. To understand the full complexity of the liver regeneration response, especially against the backdrop of a necroinflammatory environment that characterize chronic liver disease in patients, alternative models to study liver regeneration have been used such as the rodent APAP model of chemically induced injury. Here, Sande-Melon et al. aims to establish such a liver regeneration model in adult zebrafish that would harness the power of the zebrafish model, such as availability of various transgenic lines that label different cell populations, ease of accessibility to imaging techniques, large N number, and the convenience of working with lower complexity model organisms. While such a zebrafish liver regeneration model will be welcomed by the greater research community interested in studying liver regeneration, this paper in its current forms falls short of demonstrating the robustness and reproducibility of this model that would make it a useful research tool.

    4. 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 study, the authors presented a novel cryoinjury model of liver damage and regeneration that reflects essential features of liver disease, including local fibrosis. Because of its rapid and consistent method, this model will be helpful and provide opportunities to delve into the molecular basis of liver regeneration. This manuscript also contains a high technique of visualization of the regenerating liver. The manuscript is well-written, and the points are clear. However, this form of manuscript might be overly descriptive, and adding functional, mechanical, or lineage tracing-based fate decision insights would make this manuscript significantly better.

      Major points:

      1. In this cryoinjury model, the authors found cell proliferation in hepatocytes, BECs, and other cell types near the injury site. The proliferating hepatocytes exclusively provide hepatocytes, and BECs provide BECs, or some transdifferentiation is involved? Like other extreme ablation models, BECs can contribute to some hepatocytes in this model.
      2. In this model, the authors observed the long-range effect of the cryoinjury as they identified increased cell proliferation in the contralateral liver lobes. Is this long-range effect specific to hepatocytes? BECs or endothelial cells also undergo increased cell proliferation in the contralateral lobes?
      3. This model is a unique liver regeneration model as it induces transient focal fibrosis. Is the fibrosis beneficial for liver regeneration? What happens if you reduce fibrosis pharmacologically? Will it interfere with the rate of regeneration?
      4. Do Lcp1+ leucocytes contribute to liver regeneration in this mode? In immunodeficiency models such as irf8 mutant, liver regeneration after cryoinjury changed?
      5. The CUBIC-clearing procedure is beneficial in the field. The quantitative benefit of the CUBIC-based method should be added. Supplement figures 1C and D need scale bars, especially for X-Z and Z-Y planes. Can you quantify the Z-plane depth that you can scan with or without CUBIC treatment?
      6. In the manuscript, the authors measured the injured area after the cryoinjury. But how about the depth of the injury? Does the procedure induce a relatively constant injury depth, or can it not be controlled? The total volume of injured tissue would be more important than the surface injured area.

      Minor points:

      1. The sham procedure means exposing the liver by removing the scale and cutting the skin, right? What is the survival rate of the sham procedure? Is the survival rate of sham group significantly lower than cryoinjury-induced group?
      2. The original RNA-seq data, including FASTQ files, should be deposited to NCBI (Gene Expression Omnibus) or other public databases.

      Significance

      The strength of this manuscript is that the authors established the new cryoinjury liver regeneration model. Compared to other models, this model introduced local fibrosis and relatively quick resolution of the fibrosis, which is unique to this model. Fibrosis is like a double-edged sword, as it can be a severe problem, but it may also enhance healing and regeneration. This useful model would advance our understanding of the role of fibrosis in liver regeneration. Also, this manuscript contains important new technologies, such as CUBIC-clearing, and will be helpful for the research field.

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

      Learn more at Review Commons


      Reply to the reviewers

      Reply to the Reviewers

      We thank the referees for their careful reading of the manuscript and their valuable suggestions for improvements.

      General Statements:

      Existing SMC-based loop extrusion models successfully predict and characterize mesoscale genome spatial organization in vertebrate organisms, providing a valuable computational tool to the genome organization and chromatin biology fields. However, to date this approach is highly limited in its application beyond vertebrate organisms. This limitation arises because existing models require knowledge of CTCF binding sites, which act as effective boundary elements, blocking loop-extruding SMC complexes and thus defining TAD boundaries. However, CTCF is the predominant boundary element only in vertebrates. On the other hand, vertebrates only contain a small proportion of species in the tree of life, while TADs are nearly universal and SMC complexes are largely conserved. Thus, there is a pressing need for loop extrusion models capable of predicting Hi-C maps in organisms beyond vertebrates.

      The conserved-current loop extrusion (CCLE) model, introduced in this manuscript, extends the quantitative application of loop extrusion models in principle to any organism by liberating the model from the lack of knowledge regarding the identities and functions of specific boundary elements. By converting the genomic distribution of loop extruding cohesin into an ensemble of dynamic loop configurations via a physics-based approach, CCLE outputs three-dimensional (3D) chromatin spatial configurations that can be manifested in simulated Hi-C maps. We demonstrate that CCLE-generated maps well describe experimental Hi-C data at the TAD-scale. Importantly, CCLE achieves high accuracy by considering cohesin-dependent loop extrusion alone, consequently both validating the loop extrusion model in general (as opposed to diffusion-capture-like models proposed as alternatives to loop extrusion) and providing evidence that cohesin-dependent loop extrusion plays a dominant role in shaping chromatin organization beyond vertebrates.

      The success of CCLE unambiguously demonstrates that knowledge of the cohesin distribution is sufficient to reconstruct TAD-scale 3D chromatin organization. Further, CCLE signifies a shifted paradigm from the concept of localized, well-defined boundary elements, manifested in the existing CTCF-based loop extrusion models, to a concept also encompassing a continuous distribution of position-dependent loop extrusion rates. This new paradigm offers greater flexibility in recapitulating diverse features in Hi-C data than strictly localized loop extrusion barriers.

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

      This manuscript presents a mathematical model for loop extrusion called the conserved-current loop extrusion model (CCLE). The model uses cohesin ChIP-Seq data to predict the Hi-C map and shows broad agreement between experimental Hi-C maps and simulated Hi-C maps. They test the model on Hi-C data from interphase fission yeast and meiotic budding yeast. The conclusion drawn by the authors is that peaks of cohesin represent loop boundaries in these situations, which they also propose extends to other organism/situations where Ctcf is absent.

      __Response: __

      We would like to point out that the referee's interpretation of our results, namely that, "The conclusion drawn by the authors is that peaks of cohesin represent loop boundaries in these situations, ...", is an oversimplification, that we do not subscribe to. The referee's interpretation of our model is correct when there are strong, localized barriers to loop extrusion; however, the CCLE model allows for loop extrusion rates that are position-dependent and take on a range of values. The CCLE model also allows the loop extrusion model to be applied to organisms without known boundary elements. Thus, the strict interpretation of the positions of cohesin peaks to be loop boundaries overlooks a key idea to emerge from the CCLE model.

      __ Major comments:__

      1. More recent micro-C/Hi-C maps, particularly for budding yeast mitotic cells and meiotic cells show clear puncta, representative of anchored loops, which are not well recapitulated in the simulated data from this study. However, such punta are cohesin-dependent as they disappear in the absence of cohesin and are enhanced in the absence of the cohesin release factor, Wapl. For example - see the two studies below. The model is therefore missing some key elements of the loop organisation. How do the authors explain this discrepency? It would also be very useful to test whether the model can predict the increased strength of loop anchors when Wapl1 is removed and cohesin levels increase.

      Costantino L, Hsieh TS, Lamothe R, Darzacq X, Koshland D. Cohesin residency determines chromatin loop patterns. Elife. 2020 Nov 10;9:e59889. doi: 10.7554/eLife.59889. PMID: 33170773; PMCID: PMC7655110. Barton RE, Massari LF, Robertson D, Marston AL. Eco1-dependent cohesin acetylation anchors chromatin loops and cohesion to define functional meiotic chromosome domains. Elife. 2022 Feb 1;11:e74447. doi: 10.7554/eLife.74447. Epub ahead of print. PMID: 35103590; PMCID: PMC8856730.

      __Response: __

      We are perplexed by this referee comment. While we agree that puncta representing loop anchors are a feature of Hi-C maps, as noted by the referee, we would reinforce that our CCLE simulations of meiotic budding yeast (Figs. 5A and 5B of the original manuscript) demonstrate an overall excellent description of the experimental meiotic budding yeast Hi-C map, including puncta arising from loop anchors. This CCLE model-experiment agreement for meiotic budding yeast is described and discussed in detail in the original manuscript and the revised manuscript (lines 336-401).

      To further emphasize and extend this point we now also address the Hi-C of mitotic budding yeast, which was not included the original manuscript. We have now added an entire new section of the revised manuscript entitled "CCLE Describes TADs and Loop Configurations in Mitotic S. cerevisiae" including the new Figure 6, which presents a comparison between a portion of the mitotic budding yeast Hi-C map from Costantino et al. and the corresponding CCLE simulation at 500 bp-resolution. In this case too, the CCLE model well-describes the data, including the puncta, further addressing the referee's concern that the CCLE model is missing some key elements of loop organization.

      Concerning the referee's specific comment about the role of Wapl, we note that in order to apply CCLE when Wapl is removed, the corresponding cohesin ChIP-seq in the absence of Wapl should be available. To our knowledge, such data is not currently available and therefore we have not pursued this explicitly. However, we would reinforce that as Wapl is a factor that promotes cohesin unloading, its role is already effectively represented in the optimized value for LEF processivity, which encompasses LEF lifetime. In other words, if Wapl has a substantial effect it will be captured already in this model parameter.

      1. Related to the point above, the simulated data has much higher resolution than the experimental data (1kb vs 10kb in the fission yeast dataset). Given that loop size is in the 20-30kb range, a good resolution is important to see the structural features of the chromosomes. Can the model observe these details that are averaged out when the resolution is increased?

      __Response: __

      We agree with the referee that higher resolution is preferable to low resolution. In practice, however, there is a trade-off between resolution and noise. The first experimental interphase fission yeast Hi-C data of Mizuguchi et al 2014 corresponds to 10 kb resolution. To compare our CCLE simulations to these published experimental data, as described in the original manuscript, we bin our 1-kb-resolution simulations to match the 10 kb experimental measurements. Nevertheless, CCLE can readily predict the interphase fission yeast Hi-C map at higher resolution by reducing the bin size (or, if necessary, reducing the lattice site size of the simulations themselves). In the revised manuscript, we have added comparisons between CCLE's predicted Hi-C maps and newer Micro-C data for S. pombe from Hsieh et al. (Ref. [50]) in the new Supplementary Figures 5-9. We have chosen to present these comparisons at 2 kb resolution, which is the same resolution for our meiotic budding yeast comparisons. Also included in Supplementary Figures 5-9 are comparisons between the original Hi-C maps of Mizuguchi et al. and the newer maps of Hsieh et al., binned to 10 kb resolution. Inspection of these figures shows that CCLE provides a good description of Hsieh et al.'s experimental Hi-C maps and does not reveal any major new features in the interphase fission yeast Hi-C map on the 10-100 kb scale, that were not already apparent from the Hi-C maps of Mizuguchi et al 2014. Thus, the CCLE model performs well across this range of effective resolutions.

      3. Transcription, particularly convergent has been proposed to confer boundaries to loop extrusion. Can the authors recapitulate this in their model?

      __Response: __

      In response to the suggestion of the reviewer we have now calculated the correlation between cohesin ChIP-seq and the locations of convergent gene pairs, which is now presented in Supplementary Figures 17 and 18. Accordingly, in the revised manuscript, we have added the following text to the Discussion (lines 482-498):

      "In vertebrates, CTCF defines the locations of most TAD boundaries. It is interesting to ask what might play that role in interphase S. pombe as well as in meiotic and mitotic S. cerevisiae. A number of papers have suggested that convergent gene pairs are correlated with cohesin ChIP-seq in both S. pombe [65, 66] and S. cerevisiae [66-71]. Because CCLE ties TADs to cohesin ChIP-seq, a strong correlation between cohesin ChIP-seq and convergent gene pairs would be an important clue to the mechanism of TAD formation in yeasts. To investigate this correlation, we introduce a convergent-gene variable that has a nonzero value between convergent genes and an integrated weight of unity for each convergent gene pair. Supplementary Figure 17A shows the convergent gene variable, so-defined, alongside the corresponding cohesin ChIP-seq for meiotic and mitotic S. cerevisiae. It is apparent from this figure that a peak in the ChIP-seq data is accompanied by a non-zero value of the convergent-gene variable in about 80% of cases, suggesting that chromatin looping in meiotic and mitotic S. cerevisiae may indeed be tied to convergent genes. Conversely, about 50% of convergent genes match peaks in cohesin ChIP-seq. The cross-correlation between the convergent-gene variable and the ChIP-seq of meiotic and mitotic S. cerevisiae is quantified in Supplementary Figures 17B and C. By contrast, in interphase S. pombe, cross-correlation between convergent genes and cohesin ChIP-seq in each of five considered regions is unobservably small (Supplementary Figure 18A), suggesting that convergent genes per se do not have a role in defining TAD boundaries in interphase S. pombe."

      Minor comments:

      1. In the discussion, the authors cite the fact that Mis4 binding sites do not give good prediction of the HI-C maps as evidence that Mis4 is not important for loop extrusion. This can only be true if the position of Mis4 measured by ChIP is a true reflection of Mis4 position. However, Mis4 binding to cohesin/chromatin is very dynamic and it is likely that this is too short a time scale to be efficiently cross-linked for ChIP. Conversely, extensive experimental data in vivo and in vitro suggest that stimulation of cohesin's ATPase by Mis4-Ssl3 is important for loop extrusion activity.

      __Response: __

      We apologize for the confusion on this point. We actually intended to convey that the absence of Mis4-Psc3 correlations in S. pombe suggests, from the point of view of CCLE, that Mis4 is not an integral component of loop-extruding cohesin, during the loop extrusion process itself. We agree completely that Mis4/Ssl3 is surely important for cohesin loading, and (given that cohesin is required for loop extrusion) Mis4/Ssl3 is therefore important for loop extrusion. Evidently, this part of our Discussion was lacking sufficient clarity. In response to both referees' comments, we have re-written the discussion of Mis4 and Pds5 to more carefully explain our reasoning and be more circumspect in our inferences. The re-written discussion is described below in response to Referee #2's comments.

      Nevertheless, on the topic of whether Nipbl-cohesin binding is too transient to be detected in ChIP-seq, the FRAP analysis presented by Rhodes et al. eLife 6:e30000 "Scc2/Nipbl hops between chromosomal cohesin rings after loading" indicates that, in HeLa cells, Nipbl has a residence time bound to cohesin of about 50 seconds. As shown in the bottom panel of Supplementary Fig. 7 in the original manuscript (and the bottom panel of Supplementary Fig. 20 in the revised manuscript), there is a significant cross-correlation (~0.2) between the Nipbl ChIP-seq and Smc1 ChIP-seq in humans, indicating that a transient association between Nipbl and cohesin can be (and in fact is) detected by ChIP-seq.

      1. *Inclusion of a comparison of this model compared to previous models (for example bottom up models) would be extremely useful. What is the improvement of this model over existing models? *

      __Response: __

      As stated in the original manuscript, as far as we are aware, "bottom up" models, that quantitatively describe the Hi-C maps of interphase fission yeast or meiotic budding yeast or, indeed, of eukaryotes other than vertebrates, do not exist. Bottom-up models would require knowledge of the relevant boundary elements (e.g. CTCF sites), which, as stated in the submitted manuscript, are generally unknown for fission yeast, budding yeast, and other non-vertebrate eukaryotes. The absence of such models is the reason that CCLE fills an important need. Since bottom-up models for cohesin loop extrusion in yeast do not exist, we cannot compare CCLE to the results of such models.

      In the revised manuscript we now explicitly compare the CCLE model to the only bottom-up type of model describing the Hi-C maps of non-vertebrate eukaryotes by Schalbetter et al. Nat. Commun. 10:4795 2019, which we did cite extensively in our original manuscript. Schalbetter et al. use cohesin ChIP-seq peaks to define the positions of loop extrusion barriers in meiotic S. cerevisiae, for which the relevant boundary elements are unknown. In their model, specifically, when a loop-extruding cohesin anchor encounters such a boundary element, it either passes through with a certain probability, as if no boundary element is present, or stops extruding completely until the cohesin unbinds and rebinds.

      In the revised manuscript we refer to this model as the "explicit barrier" model and have applied it to interphase S. pombe, using cohesin ChIP-seq peaks to define the positions of loop extrusion barriers. The corresponding simulated Hi-C map is presented in Supplementary Fig. 19 in comparison with the experimental Hi-C. It is evident that the explicit barrier model provides a poorer description of the Hi-C data of interphase S. pombe compared to the CCLE model, as indicated by the MPR and Pearson correlation scores. While the explicit barrier model appears capable of accurately reproducing Hi-C data with punctate patterns, typically accompanied by strong peaks in the corresponding cohesin ChIP-seq, it seems less effective in several conditions including interphase S. pombe, where the Hi-C data lacks punctate patterns and sharp TAD boundaries, and the corresponding cohesin ChIP-seq shows low-contrast peaks. The success of the CCLE model in describing the Hi-C data of both S. pombe and S. cerevisiae, which exhibit very different features, suggests that the current paradigm of localized, well-defined boundary elements may not be the only approach to understanding loop extrusion. By contrast, CCLE allows for a concept of continuous distribution of position-dependent loop extrusion rates, arising from the aggregate effect of multiple interactions between loop extrusion complexes and chromatin. This paradigm offers greater flexibility in recapitulating diverse features in Hi-C data than strictly localized loop extrusion barriers.

      We have also added the following paragraph in the Discussion section of the manuscript to elaborate this point (lines 499-521):

      "Although 'bottom-up' models which incorporate explicit boundary elements do not exist for non-vertebrate eukaryotes, one may wonder how well such LEF models, if properly modified and applied, would perform in describing Hi-C maps with diverse features. To this end, we examined the performance of the model described in Ref. [49] in describing the Hi-C map of interphase S. cerevisiae. Reference [49] uses cohesin ChIP-seq peaks in meiotic S. cerevisiae to define the positions of loop extrusion barriers which either completely stall an encountering LEF anchor with a certain probability or let it pass. We apply this 'explicit barrier' model to interphase S. pombe, using its cohesin ChIP-seq peaks to define the positions of loop extrusion barriers, and using Ref. [49]'s best-fit value of 0.05 for the pass-through probability. Supplementary Figure 19A presents the corresponding simulated Hi-C map the 0.3-1.3 kb region of Chr 2 of interphase S. pombe in comparison with the corresponding Hi-C data. It is evident that the explicit barrier model provides a poorer description of the Hi-C data of interphase S. pombe compared to the CCLE model, as indicated by the MPR and Pearson correlation scores of 1.6489 and 0.2267, respectively. While the explicit barrier model appears capable of accurately reproducing Hi-C data with punctate patterns, typically accompanied by strong peaks in the corresponding cohesin ChIP-seq, it seems less effective in cases such as in interphase S. pombe, where the Hi-C data lacks punctate patterns and sharp TAD boundaries, and the corresponding cohesin ChIP-seq shows low-contrast peaks. The success of the CCLE model in describing the Hi-C data of both S. pombe and S. cerevisiae, which exhibit very different features, suggests that the current paradigm of localized, well-defined boundary elements may not be the only approach to understanding loop extrusion. By contrast, CCLE allows for a concept of continuous distribution of position-dependent loop extrusion rates, arising from the aggregate effect of multiple interactions between loop extrusion complexes and chromatin. This paradigm offers greater flexibility in recapitulating diverse features in Hi-C data than strictly localized loop extrusion barriers."

      Reviewer #1 (Significance (Required)):

      This simple model is useful to confirm that cohesin positions dictate the position of loops, which was predicted already and proposed in many studies. However, it should be considered a starting point as it does not faithfully predict all the features of chromatin organisation, particularly at better resolution.

      Response:

      As described in more detail above, we do not agree with the assertion of the referee that the CCLE model "does not faithfully predict all the features of chromatin organization, particularly at better resolution" and provide additional new data to support the conclusion that the CCLE model provides a much needed approach to model non-vertebrate contact maps and outperforms the single prior attempt to predict budding yeast Hi-C data using information from cohesin ChIP-seq.

      *It will mostly be of interest to those in the chromosome organisation field, working in organisms or systems that do not have ctcf. *

      __Response: __

      We agree that this work will be of special interest to researchers working on chromatin organization of non-vertebrate organisms. We would reinforce that yeast are frequently used models for the study of cohesin, condensin, and chromatin folding more generally. Indeed, in the last two months alone there are two Molecular Cell papers, one Nature Genetics paper, and one Cell Reports paper where loop extrusion in yeast models is directly relevant. We also believe, however, that the model will be of interest for the field in general as it simultaneously encompasses various scenarios that may lead to slowing down or stalling of LEFs.

      This reviewer is a cell biologist working in the chromosome organisation field, but does not have modelling experience and therefore does not have the expertise to determine if the modelling part is mathematically sound and has assumed that it is.

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

      Summary: Yuan et al. report on their development of an analytical model ("CCLE") for loop extrusion with genomic-position-dependent speed, with the idea of accounting for barriers to loop extrusion. They write down master equations for the probabilities of cohesin occupancy at each genomic site and obtain approximate steady-state solutions. Probabilities are governed by cohesin translocation, loading, and unloading. Using ChIP-seq data as an experimental measurement of these probabilities, they numerically fit the model parameters, among which are extruder density and processivity. Gillespie simulations with these parameters combined with a 3D Gaussian polymer model were integrated to generate simulated Hi-C maps and cohesin ChIP-seq tracks, which show generally good agreement with the experimental data. The authors argue that their modeling provides evidence that loop extrusion is the primary mechanism of chromatin organization on ~10-100 kb scales in S. pombe and S. cerevisiae.

      Major comments:

      1. I am unconvinced that this analysis specifically is sufficient to demonstrate that extrusion is the primary organizer of chromatin on these scales; moreover, the need to demonstrate this is questionable, as extrusion is widely accepted, even if not universally so. How is the agreement of CCLE with experiments more demonstrative of loop extrusion than previous modeling?

      __Response: __

      We agree with the referee's statement that "loop extrusion is extrusion is widely accepted, even if not universally so". We disagree with the referee that this state of affairs means that "the need to demonstrate this (i.e. loop extrusion) is questionable". On the contrary, studies that provide further compelling evidence that cohesin-based loop extrusion is the primary organizer of chromatin, such as ours, must surely be welcomed, first, in order to persuade those who remain unconvinced by the loop extrusion mechanism in general, and, secondly, because, until the present work, quantitative models of loop extrusion, capable of reproducing Hi-C maps quantitatively, in yeasts and other non-vertebrate eukaryotes have been lacking, leaving open the question of whether loop extrusion can describe Hi-C maps beyond vertebrates. CCLE has now answered that question in the affirmative. Moreover, the existence of a robust model to predict contact maps in non-vertebrate models, which are extensively used in the pursuit of research questions in chromatin biology, will be broadly enabling to the field.

      It is fundamental that if a simple, physically-plausible model/hypothesis is able to describe experimental data quantitatively, it is indeed appropriate to ascribe considerable weight to that model/hypothesis (until additional data become available to refute the model).

      How is the agreement of CCLE with experiments more demonstrative of loop extrusion than previous modeling?

      Response:

      As noted above and in the original manuscript, we are unaware of previous quantitative modeling of cohesin-based loop extrusion and the resultant Hi-C maps in organisms that lack CTCF, namely non-vertebrate eukaryotic models such as fission yeast or budding yeast, as we apply here. As noted in the original manuscript, previous quantitative modeling of Hi-C maps based on cohesin loop extrusion and CTCF boundary elements has been convincing that loop extrusion is indeed relevant in vertebrates, but the restriction to vertebrates excludes most of the tree of life.

      Below, the referee cites two examples of loop extrusion outside of vertebrates. The one that is suggested to correspond to yeast cells (Dequeker et al. Nature 606:197 2022) actually corresponds to mouse cells, which are vertebrate cells. The other one models the Hi-C map of the prokaryote, Bacillus subtilis, based on loop extrusion of the bacterial SMC complex thought to most resemble condensin (not cohesin), subject to barriers to loop extrusion that are related to genes or involving prokaryote-specific Par proteins (Brandao et al. PNAS 116:20489 2019). We have referenced this work in the revised manuscript but would reinforce that it lacks utility in predicting the contact maps for non-vertebrate eukaryotes.

      Relatedly, similar best fit values for S. pombe and S. cerevisiae might not point to a mechanistic conclusion (same "underlying mechanism" of loop extrusion), but rather to similar properties for loop-extruding cohesins in the two species.

      Response:

      In the revised manuscript, we have replaced "suggesting that the underlying mechanism that governs loop extrusion by cohesin is identical in both species" with "suggesting loop-extruding cohesins possess similar properties in both species" (lines 367-368).

      As an alternative, could a model with variable binding probability given by ChIP-seq and an exponential loop-size distribution work equally well? The stated lack of a dependence on extrusion timescale suggests that a static looping model might succeed. If not, why not?

      Response:

      A hypothetical mechanism that generates the same instantaneous loop distributions and correlations as loop extrusion would lead to the same Hi-C map as does loop extrusion. This circumstance is not confined to CCLE, but is equally applicable to previous CTCF-based loop extrusion models. It holds because Hi-C and ChIP-seq, and therefore models that seek to describe these measurements, provide a snapshot of the chromatin configuration at one instant of time.

      We would reinforce that there is no physical basis for a diffusion capture model with an approximately-exponential loop size distributions. Nevertheless, one can reasonably ask whether a physically-sensible diffusion capture model can simultaneously match cohesin ChIP-seq and Hi-C. Motivated by the referee's comment we have addressed this question and, accordingly, in the revised manuscript, we have added (1) an entire subsection entitled "Diffusion capture does not reproduce experimental interphase S. pombe Hi-C maps" (lines 303-335) and (2) Supplementary Figure 15. As we now demonstrate, the CCLE model vastly outperforms an equilibrium binding model in reproducing the experimental Hi-C maps and measured P(s).

      *2. I do not understand how the loop extrusion residence time drops out. As I understand it, Eq 9 converts ChIP-seq to lattice site probability (involving N_{LEF}, which is related to \rho, and \rho_c). Then, Eqs. 3-4 derive site velocities V_n and U_n if we choose rho, L, and \tau, with the latter being the residence time. This parameter is not specified anywhere and is claimed to be unimportant. It may be true that the choice of timescale is arbitrary in this procedure, but can the authors please clarify? *

      __Response: __

      As noted above, Hi-C and ChIP-seq both capture chromatin configuration at one instant in time. Therefore, such measurements cannot and do not provide any time-scale information, such as the loop extrusion residence time (LEF lifetime) or the mean loop extrusion rate. For this reason, neither our CCLE simulations, nor other researchers' previous simulations of loop extrusion in vertebrates with CTCF boundary elements, provide any time-scale information, because the experiments they seek to describe do not contain time-scale information. The Hi-C map simulations can and do provide information concerning the loop size, which is the product of the loop lifetime and the loop extrusion rate. Lines 304-305 of the revised manuscript include the text: "Because Hi-C and ChIP-seq both characterize chromatin configuration at a single instant of time, and do not provide any direct time-scale information, ..."

      In practice, we set the LEF lifetime to be some explicit value with arbitrary time-unit. We have added a sentence in the Methods that reads, "In practice, however, we set the LEF dissociation rate to 5e-4 time-unit-1 (equivalent to a lifetime of 2000 time-units), and the nominal LEF extrusion rate (aka \rho*L/\tau, see Supplementary Methods) can be determined from the given processivity" (lines 599-602), to clarify this point. We have also changed the terminology from "timesteps" to "LEF events" in the manuscript as the latter is more accurate for our purpose.

      1. The assumptions in the solution and application of the CCLE model are potentially constraining to a limited number of scenarios. In particular the authors specify that current due to binding/unbinding, A_n - D_n, is small. This assumption could be problematic near loading sites (centromeres, enhancers in higher eukaryotes, etc.) (where current might be dominated by A_n and V_n), unloading sites (D_n and V_{n-1}), or strong boundaries (D_n and V_{n-1}). The latter scenario is particularly concerning because the manuscript seems to be concerned with the presence of unidentified boundaries. This is partially mitigated by the fact that the model seems to work well in the chosen examples, but the authors should discuss the limitations due to their assumptions and/or possible methods to get around these limitations.

      4. Related to the above concern, low cohesin occupancy is interpreted as a fast extrusion region and high cohesin occupancy is interpreted as a slow region. But this might not be true near cohesin loading and unloading sites.

      __Response: __

      Our response to Referee 2's Comments 3. and 4. is that both in the original manuscript and in the revised manuscript we clearly delineate the assumptions underlying CCLE and we carefully assess the extent to which these assumptions are violated (lines 123-126 and 263-279 in the revised manuscript). For example, Supplementary Figure 12 shows that across the S. pombe genome as a whole, violations of the CCLE assumptions are small. Supplementary Figure 13 shows that violations are similarly small for meiotic S. cerevisiae. However, to explicitly address the concern of the referee, we have added the following sentences to the revised manuscript:

      Lines 277-279:

      "While loop extrusion in interphase S. pombe seems to well satisfy the assumptions underlying CCLE, this may not always be the case in other organisms."

      Lines 359-361:

      "In addition, the three quantities, given by Eqs. 6, 7, and 8, are distributed around zero with relatively small fluctuations (Supplementary Fig. 13), indicating that CCLE model is self-consistent in this case also."

      In the case of mitotic S. cerevisiae, Supplementary Figure 14 shows that these quantities are small for most of genomic locations, except near the cohesin ChIP-seq peaks. We ascribe these greater violations of CCLE's assumptions at the locations of cohesin peaks in part to the low processivity of mitotic cohesin in S. cerevisiae, compared to that of meiotic S. cerevisiae and interphase S. pombe, and in part to the low CCLE loop extrusion rate at the cohesin peaks. We have added a paragraph at the end of the Section "CCLE Describes TADs and Loop Configurations in Mitotic S. cerevisiae" to reflect these observations (lines 447-461).

      1. *The mechanistic insight attempted in the discussion, specifically with regard to Mis4/Scc2/NIPBL and Pds5, is problematic. First, it is not clear how the discussion of Nipbl and Pds5 is connected to the CCLE method; the justification is that CCLE shows cohesin distribution is linked to cohesin looping, which is already a questionable statement (point 1) and doesn't really explain how the model offers new insight into existing Nipbl and Pds5 data. *

      Furthermore, I believe that the conclusions drawn on this point are flawed, or at least, stated with too much confidence. The authors raise the curious point that Nipbl ChIP-seq does not correlate well with cohesin ChIP-seq, and use this as evidence that Nipbl is not a part of the loop-extruding complex in S. pombe, and it is not essential in humans. Aside from the molecular evidence in human Nipbl/cohesin (acknowledged by authors), there are other reasons to doubt this conclusion. First, depletion of Nipbl (rather than binding partner Mau2 as in ref 55) in mouse cells strongly inhibits TAD formation (Schwarzer et al. Nature 551:51 2017). Second, at least two studies have raised concerns about Nibpl ChIP-seq results: 1) Hu et al. Nucleic Acids Res 43:e132 2015, which shows that uncalibrated ChIP-seq can obscure the signal of protein localization throughout the genome due to the inability to distinguish from background * and 2) Rhodes et al. eLife 6:e30000, which uses FRAP to show that Nipbl binds and unbinds to cohesin rapidly in human cells, which could go undetected in ChIP-seq, especially when uncalibrated. It has not been shown that these dynamics are present in yeast, but there is no reason to rule it out yet.*

      Similar types of critiques could be applied to the discussion of Pds5. There is cross-correlation between Psc3 and Pds5 in S. pombe, but the authors are unable to account for whether Pds5 binding is transient and/or necessary to loop extrusion itself or, more importantly, whether Pds5 ChIP is associated with extrusive or cohesive cohesins; cross-correlation peaks at about 0.6, but note that by the authors own estimates, cohesive cohesins are approximately half of all cohesins in S. pombe (Table 3).

      *Due to the above issues, I suggest that the authors heavily revise this discussion to better reflect the current experimental understanding and the limited ability to draw such conclusions based on the current CCLE model. *

      __Response: __

      As stated above, our study demonstrates that the CCLE approach is able to take as input cohesin (Psc3) ChIP-seq data and produce as output simulated Hi-C maps that well reproduce the experimental Hi-C maps of interphase S. pombe and meiotic S. cerevisiae. This result is evident from the multiple Hi-C comparison figures in both the original and the revised manuscripts. In light of this circumstance, the referee's statement that it is "questionable", that CCLE shows that cohesin distribution (as quantified by cohesin ChIP-seq) is linked to cohesin looping (as quantified by Hi-C), is demonstrably incorrect.

      However, we did not intend to suggest that Nipbl and Pds5 are not crucial for cohesin loading, as the reviewer states. Rather, our inquiries relate to a more nuanced question of whether these factors only reside at loading sites or, instead, remain as a more long-lived constituent component of the loop extrusion complex. We regret any confusion and have endeavored to clarify this point in the revised manuscript in response to Referee 2's Comment 5. as well as Referee 1's Minor Comment 1. We have now better explained how the CCLE model may offer new insight from existing ChIP-seq data in general and from Mis4/Nipbl and Pds5 ChIP-seq, in particular. Accordingly, we have followed Referee 2's advice to heavily revise the relevant section of the Discussion.

      To this end, we have removed the following text from the original manuscript:

      "The fact that the cohesin distribution along the chromatin is strongly linked to chromatin looping, as evident by the success of the CCLE model, allows for new insights into in vivo LEF composition and function. For example, recently, two single-molecule studies [37, 38] independently found that Nipbl, which is the mammalian analogue of Mis4, is an obligate component of the loop-extruding human cohesin complex. Ref. [37] also found that cohesin complexes containing Pds5, instead of Nipbl, are unable to extrude loops. On this basis, Ref. [32] proposed that, while Nipbl-containing cohesin is responsible for loop extrusion, Pds5-containing cohesin is responsible for sister chromatid cohesion, neatly separating cohesin's two functions according to composition. However, the success of CCLE in interphase S. pombe, together with the observation that the Mis4 ChIP-seq signal is uncorrelated with the Psc3 ChIP-seq signal (Supplementary Fig. 7) allows us to infer that Mis4 cannot be a component of loop-extruding cohesin in S. pombe. On the other hand, Pds5 is correlated with Psc3 in S. pombe (Supplementary Fig. 7) suggesting that both proteins are involved in loop-extruding cohesin, contradicting a hypothesis that Pds5 is a marker for cohesive cohesin in S. pombe. In contrast to the absence of Mis4-Psc3 correlation in S. pombe, in humans, Nipbl ChIP-seq and Smc1 ChIP-seq are correlated (Supplementary Fig. 7), consistent with Ref. [32]'s hypothesis that Nipbl can be involved in loop-extruding cohesin in humans. However, Ref. [55] showed that human Hi-C contact maps in the absence of Nipbl's binding partner, Mau2 (Ssl3 in S. pombe [56]) show clear TADs, consistent with loop extrusion, albeit with reduced long-range contacts in comparison to wild-type maps, indicating that significant loop extrusion continues in live human cells in the absence of Nipbl-Mau2 complexes. These collected observations suggest the existence of two populations of loop-extruding cohesin complexes in vivo, one that involves Nipbl-Mau2 and one that does not. Both types are present in mammals, but only Mis4-Ssl3-independent loop-extruding cohesin is present in S. pombe."

      And we have replaced it by the following text in the revised manuscript (lines 533-568):

      "As noted above, the input for our CCLE simulations of chromatin organization in S. pombe, was the ChIP-seq of Psc3, which is a component of the cohesin core complex [75]. Accordingly, Psc3 ChIP-seq represents how the cohesin core complex is distributed along the genome. In S. pombe, the other components of the cohesin core complex are Psm1, Psm3, and Rad21. Because these proteins are components of the cohesin core complex, we expect that the ChIP-seq of any of these proteins would closely match the ChIP-seq of Psc3, and would equally well serve as input for CCLE simulations of S. pombe genome organization. Supplementary Figure 20C confirms significant correlations between Psc3 and Rad21. In light of this observation, we then reason that the CCLE approach offers the opportunity to investigate whether other proteins beyond the cohesin core are constitutive components of the loop extrusion complex during the extrusion process (as opposed to cohesin loading or unloading). To elaborate, if the ChIP-seq of a non-cohesin-core protein is highly correlated with the ChIP-seq of a cohesin core protein, we can infer that the protein in question is associated with the cohesin core and therefore is a likely participant in loop-extruding cohesin, alongside the cohesin core. Conversely, if the ChIP-seq of a putative component of the loop-extruding cohesin complex is uncorrelated with the ChIP-seq of a cohesin core protein, then we can infer that the protein in question is unlikely to be a component of loop-extruding cohesin, or at most is transiently associated with it.

      For example, in S. pombe, the ChIP-seq of the cohesin regulatory protein, Pds5 [74], is correlated with the ChIP-seq of Psc3 (Supplementary Fig. 20B) and with that of Rad21 (Supplementary Fig. 20D), suggesting that Pds5 can be involved in loop-extruding cohesin in S. pombe, alongside the cohesin core proteins. Interestingly, this inference concerning fission yeast cohesin subunit, Pds5, stands in contrast to the conclusion from a recent single-molecule study [38] concerning cohesin in vertebrates. Specifically, Reference [38] found that cohesin complexes containing Pds5, instead of Nipbl, are unable to extrude loops.

      Additionally, as noted above, in S. pombe the ChIP-seq signal of the cohesin loader, Mis4, is uncorrelated with the Psc3 ChIP-seq signal (Supplementary Fig. 20A), suggesting that Mis4 is, at most, a very transient component of loop-extruding cohesin in S. pombe, consistent with its designation as a "cohesin loader". However, both References [38] and [39] found that Nipbl (counterpart of S. pombe's Mis4) is an obligate component of the loop-extruding human cohesin complex, more than just a mere cohesin loader. Although CCLE has not yet been applied to vertebrates, from a CCLE perspective, the possibility that Nipbl may be required for the loop extrusion process in humans is bolstered by the observation that in humans Nipbl ChIP-seq and Smc1 ChIP-seq show significant correlations (Supplementary Fig. 20G), consistent with Ref. [32]'s hypothesis that Nipbl is involved in loop-extruding cohesin in vertebrates. A recent theoretical model of the molecular mechanism of loop extrusion by cohesin hypothesizes that transient binding by Mis4/Nipbl is essential for permitting directional reversals and therefore for two-sided loop extrusion [41]. Surprisingly, there are significant correlations between Mis4 and Pds5 in S. pombe (Supplementary Fig. 20E), indicating Pds5-Mis4 association, outside of the cohesin core complex."

      In response to Referee 2's specific comment that "at least two studies have raised concerns about Nibpl ChIP-seq results", we note (1) that, while Hu et al. Nucleic Acids Res 43:e132 2015 present a general method for calibrating ChIP-seq results, they do not measure Mis4/Nibpl ChIP-seq, nor do they raise any specific concerns about Mis4/Nipbl ChIP-seq, and (2) that (as noted above, in response to Referee 1's comment) while the FRAP analysis presented by Rhodes et al. eLife 6:e30000 indicates that, in HeLa cells, Nipbl has a residence time bound to cohesin of about 50 seconds, nevertheless, as shown in Supplementary Fig. 20G in the revised manuscript, there is a significant cross-correlation between the Nipbl ChIP-seq and Smc1 ChIP-seq in humans, indicating that a transient association between Nipbl and cohesin is detected by ChIP-seq, the referees' concerns notwithstanding.

      We thank the referee for pointing out Schwarzer et al. Nature 551:51 2017. However, our interpretation of these data is different than the referee's. As noted in our original manuscript, Nipbl has traditionally been considered to be a cohesin loading factor. If the role of Nipbl was solely to load cohesin, then we would expect that depleting Nipbl would have a major effect on the Hi-C map, because fewer cohesins are loaded onto the chromatin. Figure 2 of Schwarzer et al. Nature 551:51 2017, shows the effect of depleting Nibpl on a vertebrate Hi-C map. Even in this case when Nibpl is absent, this figure (Figure 2 of Schwarzer et al. Nature 551:51 2017) shows that TADs persist, albeit considerably attenuated. According to the authors' own analysis associated with Fig. 2 of their paper, these attenuated TADs correspond to a smaller number of loop-extruding cohesin complexes than in the presence of Nipbl. Since Nipbl is depleted, these loop-extruding cohesins necessarily cannot contain Nipbl. Thus, the data and analysis of Schwarzer et al. Nature 551:51 2017 actually seem consistent with the existence of a population of loop-extruding cohesin complexes that do not contain Nibpl.

      Concerning the referee's comment that we cannot be sure whether Pds5 ChIP is associated with extrusive or cohesive cohesin, we note that, as explained in the manuscript, we assume that the cohesive cohesins are uniformly distributed across the genome, and therefore that peaks in the cohesin ChIP-seq are associated with loop-extruding cohesins. The success of CCLE in describing Hi-C maps justifies this assumption a posteriori. Supplementary Figure 20B shows that the ChIP-seq of Pds5 is correlated with the ChIP-seq of Psc3 in S. pombe, that is, that peaks in the ChIP-seq of Psc3, assumed to derive from loop-extruding cohesin, are accompanied by peaks in the ChIP-seq of Pds5. This is the reasoning allowing us to associate Pds5 with loop-extruding cohesin in S. pombe.

      1. I suggest that the authors recalculate correlations for Hi-C maps using maps that are rescaled by the P(s) curves. As currently computed, most of the correlation between maps could arise from the characteristic decay of P(s) rather than smaller scale features of the contact maps. This could reduce the surprising observed correlation between distinct genomic regions in pombe (which, problematically, is higher than the observed correlation between simulation and experiment in cervisiae).

      Response:

      We thank the referee for this advice. Following this advice, throughout the revised manuscript, we have replaced our original calculation of the Pearson correlation coefficient of unscaled Hi-C maps with a calculation of the Pearson correlation coefficient of rescaled Hi-C maps. Since the MPR is formed from ratios of simulated to experimental Hi-C maps, this metric is unchanged by the proposed rescaling.

      As explained in the original manuscript, we attribute the lower experiment-simulation correlation in the meiotic budding yeast Hi-C maps to the larger statistical errors of the meiotic budding yeast dataset, which arises because of its higher genomic resolution - all else being equal we can expect 25 times the counts in a 10 kb x10 kb bin as in a 2 kb x 2 kb bin. For the same reason, we expect larger statistical errors in the mitotic budding yeast dataset as well. Lower correlations for noisier data are to be expected in general.

      *7. Please explain why the difference between right and left currents at any particular site, (R_n-L_n) / Rn+Ln, should be small. It seems easy to imagine scenarios where this might not be true, such as directional barriers like CTCF or transcribed genes. *

      __Response: __

      For simplicity, the present version of CCLE sets the site-dependent loop extrusion rates by assuming that the cohesin ChIP-seq signal has equal contributions from left and right anchors. Then, we carry out our simulations which subsequently allow us to examine the simulated left and right currents and their difference at every site. The distributions of normalized left-right difference currents are shown in Supplementary Figures 12B, 13B, and 14D, for interphase S. pombe, meiotic S. cerevisiae, and mitotic S. cerevisiae, respectively. They are all centered at zero with standard deviations of 0.12, 0.16, and 0.33. Thus, it emerges from our simulations that the difference current is indeed generally small.

      8. Optional, but I think would greatly improve the manuscript, but can the authors: a) analyze regions of high cohesin occupancy (assumed to be slow extrusion regions) to determine if there's anything special in these regions, such as more transcriptional activity

      __Response: __

      In response to Referee 1's similar comment, we have calculated the correlation between the locations of convergent genes and cohesin ChIP-seq. Supplementary Figure 18A in the revised manuscript shows that for interphase S. pombe no correlations are evident, whereas for both of meiotic and mitotic S. cerevisiae, there are significant correlations between these two quantities (Supplementary Fig. 17).

      *b) apply this methodology to vertebrate cell data *

      __Response: __

      The application of CCLE to vertebrate data is outside the scope of this paper which, as we have emphasized, has the goal of developing a model that can be robustly applied to non-vertebrate eukaryotic genomes. Nevertheless, CCLE is, in principle, applicable to all organisms in which loop extrusion by SMC complexes is the primary mechanism for chromatin spatial organization.

      1. *A Github link is provided but the code is not currently available. *

      __Response: __

      The code is now available.

      Minor Comments:

      1. Please state the simulated LEF lifetime, since the statement in the methods that 15000 timesteps are needed for equilibration of the LEF model is otherwise not meaningful. Additionally, please note that backbone length is not necessarily a good measure of steady state, since the backbone can be compacted to its steady-state value while the loop distribution continues to evolve toward its steady state.

      __Response: __

      The terminology "timesteps" used in the original manuscript in fact should mean "the number of LEF events performed" in the simulation. Therefore, we have changed the terminology from "timesteps" to "LEF events".

      The choice of 15000 LEF events is empirically determined to ensure that loop extrusion steady state is achieved, for the range of parameters considered. To address the referee's concern regarding the uncertainty of achieving steady state after 15000 LEF events, we compared two loop size distributions: each distribution encompasses 1000 data points, equally separated in time, one between LEF event 15000 and 35000, and the other between LEF event 80000 and 100000. The two distributions are within-errors identical, suggesting that the loop extrusion steady state is well achieved within 15000 LEF events.

      2. How important is the cohesive cohesin parameter in the model, e.g., how good are fits with \rho_c = 0?

      __Response: __

      As stated in the original manuscript, the errors on \rho_c on the order of 10%-20% (for S. pombe). Thus, fits with \rho_c=0 are significantly poorer than with the best-fit values of \rho_c.

      *3. A nice (but non-essential) supplemental visualization might be to show a scatter of sim cohesin occupancy vs. experiment ChIP. *

      __Response: __

      We have chosen not to do this, because we judge that the manuscript is already long enough. Figures 3A, 5D, and 6C already compare the experimental and simulated ChIP-seq, and these figures already contain more information than the figures proposed by the referee.

      1. *A similar calculation of Hi-C contacts based on simulated loop extruder positions using the Gaussian chain model was previously presented in Banigan et al. eLife 9:e53558 2020, which should be cited. *

      __Response: __

      We thank the referee for pointing out this citation. We have added it to the revised manuscript.

      1. It is stated that simulation agreement with experiments for cerevisiae is worse in part due to variability in the experiments, with MPR and Pearson numbers for cerevisiae replicates computed for reference. But these numbers are difficult to interpret without, for example, similar numbers for duplicate pombe experiments. Again, these numbers should be generated using Hi-C maps scaled by P(s), especially in case there are systematic errors in one replicate vs. another.

      __Response: __

      As noted above, throughout the revised manuscript, we now give the Pearson correlation coefficients of scaled-by-P(s) Hi-C maps.

      1. *In the model section, it is stated that LEF binding probabilities are uniformly distributed. Did the authors mean the probability is uniform across the genome or that the probability at each site is a uniformly distributed random number? Please clarify, and if the latter, explain why this unconventional assumption was made. *

      __Response: __

      It is the former. We have modified the manuscript to clarify that LEFs "initially bind to empty, adjacent chromatin lattice sites with a binding probability, that is uniformly distributed across the genome." (lines 587-588).

      *7. Supplement p4 line 86 - what is meant by "processivity of loops extruded by isolated LEFs"? "size of loops extruded by..." or "processivity of isolated LEFs"? *

      __Response: __

      Here "processivity of isolated LEFs" is defined as the processivity of one LEF without the interference (blocking) from other LEFs. We have changed "processivity of loops extruded by isolated LEFs" to "processivity of isolated LEFs" for clarity.

      1. The use of parentheticals in the caption to Table 2 is a little confusing; adding a few extra words would help.

      __Response: __

      In the revised manuscript, we have added an additional sentence, and have removed the offending parentheses.

      1. *Page 12 sentence line 315-318 is difficult to understand. The barrier parameter is apparently something from ref 47 not previously described in the manuscript. *

      __Response: __

      In the revised manuscript, we have removed mention of the "barrier parameter" from the discussion.

      1. *Statement on p14 line 393-4 is false: prior LEF models have not been limited to vertebrates, and the authors have cited some of them here. There are also non-vertebrate examples with extrusion barriers: genes as boundaries to condensin in bacteria (Brandao et al. PNAS 116:20489 2019) and MCM complexes as boundaries to cohesin in yeast (Dequeker et al. Nature 606:197 2022). *

      __Response: __

      In fact, Dequeker et al. Nature 606:197 2022 concerns the role of MCM complexes in blocking cohesin loop extrusion in mouse zygotes. Mouse is a vertebrate. The sole aspect of this paper, that is associated with yeast, is the observation of cohesin blocking by the yeast MCM bound to the ARS1 replication origin site, which is inserted on a piece of lambda phage DNA. No yeast genome is used in the experiment. Therefore, the referee is mistaken to suggest that this paper models yeast genome organization.

      We thank the referee for pointing out Brandao et al. PNAS 116:20489 2019, which includes the development of a tour-de-force model of condensin-based loop extrusion in the prokaryote, Bacillus subtilis, in the presence of gene barriers to loop extrusion. To acknowledge this paper, we have changed the objectionable sentence to now read (lines 571-575):

      "... prior LEF models have been overwhelmingly limited to vertebrates, which express CTCF and where CTCF is the principal boundary element. Two exceptions, in which the LEF model was applied to non-vertebrates, are Ref. [49], discussed above, and Ref. [76] (Brandao et al.), which models the Hi-C map of the prokaryote, Bacillus subtilis, on the basis of condensin loop extrusion with gene-dependent barriers."

      *Referees cross-commenting *

      I agree with the comments of Reviewer 1, which are interesting and important points that should be addressed.

      *Reviewer #2 (Significance (Required)):

      Analytically approaching extrusion by treating cohesin translocation as a conserved current is an interesting approach to modeling and analysis of extrusion-based chromatin organization. It appears to work well as a descriptive model. But I think there are major questions concerning the mechanistic value of this model, possible applications of the model, the provided interpretations of the model and experiments, and the limitations of the model under the current assumptions. I am unconvinced that this analysis specifically is sufficient to demonstrate that extrusion is the primary organizer of chromatin on these scales; moreover, the need to demonstrate this is questionable, as extrusion is widely accepted, even if not universally so. It is also unclear that the minimal approach of the CCLE necessarily offers an improved physical basis for modeling extrusion, as compared to previous efforts such as ref 47, as claimed by the authors. There are also questions about significance due to possible limitations of the model (detailed above). Applying the CCLE model to identify barriers would be interesting, but is not attempted. Overall, the work presents a reasonable analytical model and numerical method, but until the major comments above are addressed and some reasonable application or mechanistic value or interpretation is presented, the overall significance is somewhat limited.*

      __Response: __

      We agree with the referee that analytically approaching extrusion by treating cohesin translocation as a conserved current is an interesting approach to modeling and analysis of extrusion-based chromatin organization. We also agree with the referee that it works well as a descriptive model (of Hi-C maps in S. pombe and S. cerevisiae). Obviously, we disagree with the referee's other comments. For us, being able to describe the different-appearing Hi-C maps of interphase S. pombe (Fig. 1 and Supplementary Figures 1-9), meiotic S. cerevisiae (Fig. 5) and mitotic S. cerevisiae (Fig. 6), all with a common model with just a few fitting parameters that differ between these examples, is significant and novel. The reviewer prematurely ignores the fact that there are still debates about whether "diffusion-capture"-like model is the more dominant mechanism that shape chromatin spatial organization at the TAD-scale. Many works have argued that such models could describe TAD-scale chromatin organization, as cited in the revised manuscript (Refs. [11, 14, 15, 17, 20, 22-24, 55]). However, in contrast to the poor description of the Hi-C map using diffusion capture model (as demonstrated in the revised manuscript and Supplementary Fig. 15), the excellent experiment-simulation agreement achieved by CCLE provides compelling evidence that cohesin-based loop extrusion is indeed the primary organizer of TAD-scale chromatin.

      Importantly, CCLE provides a theoretical base for how loop extrusion models can be generalized and applied to organisms without known loop extrusion barriers. Our model also highlights that (and provides means to account for) distributed barriers that impede but do not strictly block LEFs could also impact chromatin configurations. This case might be of importance to organisms with CTCF motifs that infrequently coincide with TAD boundaries, for instance, in the case of Drosophila melanogaster. Moreover, CCLE promises theoretical descriptions of the Hi-C maps of other non-vertebrates in the future, extending the quantitative application of the LEF model across the tree of life. This too would be highly significant if successful.

    2. 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:

      Yuan et al. report on their development of an analytical model ("CCLE") for loop extrusion with genomic-position-dependent speed, with the idea of accounting for barriers to loop extrusion. They write down master equations for the probabilities of cohesin occupancy at each genomic site and obtain approximate steady-state solutions. Probabilities are governed by cohesin translocation, loading, and unloading. Using ChIP-seq data as an experimental measurement of these probabilities, they numerically fit the model parameters, among which are extruder density and processivity. Gillespie simulations with these parameters combined with a 3D Gaussian polymer model were integrated to generate simulated Hi-C maps and cohesin ChIP-seq tracks, which show generally good agreement with the experimental data. The authors argue that their modeling provides evidence that loop extrusion is the primary mechanism of chromatin organization on ~10-100 kb scales in S. pombe and S. cerevisiae.

      Major comments:

      1. I am unconvinced that this analysis specifically is sufficient to demonstrate that extrusion is the primary organizer of chromatin on these scales; moreover, the need to demonstrate this is questionable, as extrusion is widely accepted, even if not universally so. How is the agreement of CCLE with experiments more demonstrative of loop extrusion than previous modeling? Relatedly, similar best fit values for S. pombe and S. cerevisiae might not point to a mechanistic conclusion (same "underlying mechanism" of loop extrusion), but rather to similar properties for loop-extruding cohesins in the two species. As an alternative, could a model with variable binding probability given by ChIP-seq and an exponential loop-size distribution work equally well? The stated lack of a dependence on extrusion timescale suggests that a static looping model might succeed. If not, why not?
      2. I do not understand how the loop extrusion residence time drops out. As I understand it, Eq 9 converts ChIP-seq to lattice site probability (involving N_{LEF}, which is related to \rho, and \rho_c). Then, Eqs. 3-4 derive site velocities V_n and U_n if we choose rho, L, and \tau, with the latter being the residence time. This parameter is not specified anywhere and is claimed to be unimportant. It may be true that the choice of timescale is arbitrary in this procedure, but can the authors please clarify?
      3. The assumptions in the solution and application of the CCLE model are potentially constraining to a limited number of scenarios. In particular the authors specify that current due to binding/unbinding, A_n - D_n, is small. This assumption could be problematic near loading sites (centromeres, enhancers in higher eukaryotes, etc.) (where current might be dominated by A_n and V_n), unloading sites (D_n and V_{n-1}), or strong boundaries (D_n and V_{n-1}). The latter scenario is particularly concerning because the manuscript seems to be concerned with the presence of unidentified boundaries. This is partially mitigated by the fact that the model seems to work well in the chosen examples, but the authors should discuss the limitations due to their assumptions and/or possible methods to get around these limitations.
      4. Related to the above concern, low cohesin occupancy is interpreted as a fast extrusion region and high cohesin occupancy is interpreted as a slow region. But this might not be true near cohesin loading and unloading sites.
      5. The mechanistic insight attempted in the discussion, specifically with regard to Mis4/Scc2/NIPBL and Pds5, is problematic. First, it is not clear how the discussion of Nipbl and Pds5 is connected to the CCLE method; the justification is that CCLE shows cohesin distribution is linked to cohesin looping, which is already a questionable statement (point 1) and doesn't really explain how the model offers new insight into existing Nipbl and Pds5 data.

      Furthermore, I believe that the conclusions drawn on this point are flawed, or at least, stated with too much confidence. The authors raise the curious point that Nipbl ChIP-seq does not correlate well with cohesin ChIP-seq, and use this as evidence that Nipbl is not a part of the loop-extruding complex in S. pombe, and it is not essential in humans. Aside from the molecular evidence in human Nipbl/cohesin (acknowledged by authors), there are other reasons to doubt this conclusion. First, depletion of Nipbl (rather than binding partner Mau2 as in ref 55) in mouse cells strongly inhibits TAD formation (Schwarzer et al. Nature 551:51 2017). Second, at least two studies have raised concerns about Nibpl ChIP-seq results: 1) Hu et al. Nucleic Acids Res 43:e132 2015, which shows that uncalibrated ChIP-seq can obscure the signal of protein localization throughout the genome due to the inability to distinguish from background and 2) Rhodes et al. eLife 6:e30000, which uses FRAP to show that Nipbl binds and unbinds to cohesin rapidly in human cells, which could go undetected in ChIP-seq, especially when uncalibrated. It has not been shown that these dynamics are present in yeast, but there is no reason to rule it out yet.

      Similar types of critiques could be applied to the discussion of Pds5. There is cross-correlation between Psc3 and Pds5 in S. pombe, but the authors are unable to account for whether Pds5 binding is transient and/or necessary to loop extrusion itself or, more importantly, whether Pds5 ChIP is associated with extrusive or cohesive cohesins; cross-correlation peaks at about 0.6, but note that by the authors own estimates, cohesive cohesins are approximately half of all cohesins in S. pombe (Table 3).

      Due to the above issues, I suggest that the authors heavily revise this discussion to better reflect the current experimental understanding and the limited ability to draw such conclusions based on the current CCLE model. 6. I suggest that the authors recalculate correlations for Hi-C maps using maps that are rescaled by the P(s) curves. As currently computed, most of the correlation between maps could arise from the characteristic decay of P(s) rather than smaller scale features of the contact maps. This could reduce the surprising observed correlation between distinct genomic regions in pombe (which, problematically, is higher than the observed correlation between simulation and experiment in cervisiae). 7. Please explain why the difference between right and left currents at any particular site, (R_n-L_n) / Rn+Ln, should be small. It seems easy to imagine scenarios where this might not be true, such as directional barriers like CTCF or transcribed genes. 8. Optional, but I think would greatly improve the manuscript, but can the authors: a) analyze regions of high cohesin occupancy (assumed to be slow extrusion regions) to determine if there's anything special in these regions, such as more transcriptional activity

      b) apply this methodology to vertebrate cell data 9. A Github link is provided but the code is not currently available.

      Minor Comments:

      1. Please state the simulated LEF lifetime, since the statement in the methods that 15000 timesteps are needed for equilibration of the LEF model is otherwise not meaningful. Additionally, please note that backbone length is not necessarily a good measure of steady state, since the backbone can be compacted to its steady-state value while the loop distribution continues to evolve toward its steady state.
      2. How important is the cohesive cohesin parameter in the model, e.g., how good are fits with \rho_c = 0?
      3. A nice (but non-essential) supplemental visualization might be to show a scatter of sim cohesin occupancy vs. experiment ChIP.
      4. A similar calculation of Hi-C contacts based on simulated loop extruder positions using the Gaussian chain model was previously presented in Banigan et al. eLife 9:e53558 2020, which should be cited.
      5. It is stated that simulation agreement with experiments for cerevisiae is worse in part due to variability in the experiments, with MPR and Pearson numbers for cerevisiae replicates computed for reference. But these numbers are difficult to interpret without, for example, similar numbers for duplicate pombe experiments. Again, these numbers should be generated using Hi-C maps scaled by P(s), especially in case there are systematic errors in one replicate vs. another.
      6. In the model section, it is stated that LEF binding probabilities are uniformly distributed. Did the authors mean the probability is uniform across the genome or that the probability at each site is a uniformly distributed random number? Please clarify, and if the latter, explain why this unconventional assumption was made.
      7. Supplement p4 line 86 - what is meant by "processivity of loops extruded by isolated LEFs"? "size of loops extruded by..." or "processivity of isolated LEFs"?
      8. The use of parentheticals in the caption to Table 2 is a little confusing; adding a few extra words would help.
      9. Page 12 sentence line 315-318 is difficult to understand. The barrier parameter is apparently something from ref 47 not previously described in the manuscript.
      10. Statement on p14 line 393-4 is false: prior LEF models have not been limited to vertebrates, and the authors have cited some of them here. There are also non-vertebrate examples with extrusion barriers: genes as boundaries to condensin in bacteria (Brandao et al. PNAS 116:20489 2019) and MCM complexes as boundaries to cohesin in yeast (Dequeker et al. Nature 606:197 2022).

      Referees cross-commenting

      I agree with the comments of Reviewer 1, which are interesting and important points that should be addressed.

      Significance

      Analytically approaching extrusion by treating cohesin translocation as a conserved current is an interesting approach to modeling and analysis of extrusion-based chromatin organization. It appears to work well as a descriptive model. But I think there are major questions concerning the mechanistic value of this model, possible applications of the model, the provided interpretations of the model and experiments, and the limitations of the model under the current assumptions. I am unconvinced that this analysis specifically is sufficient to demonstrate that extrusion is the primary organizer of chromatin on these scales; moreover, the need to demonstrate this is questionable, as extrusion is widely accepted, even if not universally so. It is also unclear that the minimal approach of the CCLE necessarily offers an improved physical basis for modeling extrusion, as compared to previous efforts such as ref 47, as claimed by the authors. There are also questions about significance due to possible limitations of the model (detailed above). Applying the CCLE model to identify barriers would be interesting, but is not attempted. Overall, the work presents a reasonable analytical model and numerical method, but until the major comments above are addressed and some reasonable application or mechanistic value or interpretation is presented, the overall significance is somewhat limited.

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

      Learn more at Review Commons


      Referee #1

      Evidence, reproducibility and clarity

      This manuscript presents a mathematical model for loop extrusion called the conserved-current loop extrusion model (CCLE). The model uses cohesin ChIP-Seq data to predict the Hi-C map and shows broad agreement between experimental Hi-C maps and simulated Hi-C maps. They test the model on Hi-C data from interphase fission yeast and meiotic budding yeast. The conclusion drawn by the authors is that peaks of cohesin represent loop boundaries in these situations, which they also propose extends to other organism/situations where Ctcf is absent.

      Major comments

      1. More recent micro-C/Hi-C maps, particularly for budding yeast mitotic cells and meiotic cells show clear puncta, representative of anchored loops, which are not well recapitulated in the simulated data from this study. However, such punta are cohesin-dependent as they disappear in the absence of cohesin and are enhanced in the absence of the cohesin release factor, Wapl. For example - see the two studies below. The model is therefore missing some key elements of the loop organisation. How do the authors explain this discrepency? It would also be very useful to test whether the model can predict the increased strength of loop anchors when Wapl1 is removed and cohesin levels increase.

      Costantino L, Hsieh TS, Lamothe R, Darzacq X, Koshland D. Cohesin residency determines chromatin loop patterns. Elife. 2020 Nov 10;9:e59889. doi: 10.7554/eLife.59889. PMID: 33170773; PMCID: PMC7655110. Barton RE, Massari LF, Robertson D, Marston AL. Eco1-dependent cohesin acetylation anchors chromatin loops and cohesion to define functional meiotic chromosome domains. Elife. 2022 Feb 1;11:e74447. doi: 10.7554/eLife.74447. Epub ahead of print. PMID: 35103590; PMCID: PMC8856730. 2. Related to the point above, the simulated data has much higher resolution than the experimental data (1kb vs 10kb in the fission yeast dataset). Given that loop size is in the 20-30kb range, a good resolution is important to see the structural features of the chromosomes. Can the model observe these details that are averaged out when the resolution is increased? 3. Transcription, particularly convergent has been proposed to confer boundaries to loop extrusion. Can the authors recapitulate this in their model?

      Minor comments

      1. In the discussion, the authors cite the fact that Mis4 binding sites do not give good prediction of the HI-C maps as evidence that Mis4 is not important for loop extrusion. This can only be true if the position of Mis4 measured by ChIP is a true reflection of Mis4 position. However, Mis4 binding to cohesin/chromatin is very dynamic and it is likely that this is too short a time scale to be efficiently cross-linked for ChIP. Conversely, extensive experimental data in vivo and in vitro suggest that stimulation of cohesin's ATPase by Mis4-Ssl3 is important for loop extrusion activity.
      2. Inclusion of a comparison of this model compared to previous models (for example bottom up models) would be extremely useful. What is the improvement of this model over existing models?

      Significance

      This simple model is useful to confirm that cohesin positions dictate the position of loops, which was predicted already and proposed in many studies. However, it should be considered a starting point as it does not faithfully predict all the features of chromatin organisation, particularly at better resolution. It will mostly be of interest to those in the chromosome organisation field, working in organisms or systems that do not have ctcf.

      This reviewer is a cell biologist working in the chromosome organisation field, but does not have modelling experience and therefore does not have the expertise to determine if the modelling part is mathematically sound and has assumed that it is.

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

      Learn more at Review Commons


      Reply to the reviewers

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

      I have trialled the package on my lab's data and it works as advertised. It was straightforward to use and did not require any special training. I am confident this is a tool that will be approachable even to users with limited computational experience. The use of artificial data to validate the approach - and to provide clear limits on applicability - is particularly helpful.

      The main limitation of the tool is that it requires the user to manually select regions. This somewhat limits the generalisability and is also more subjective - users can easily choose "nice" regions that better match with their hypothesis, rather than quantifying the data in an unbiased manner. However, given the inherent challenges in quantifying biological data, such problems are not easily circumventable.

      *

      * I have some comments to clarify the manuscript:

      1. A "straightforward installation" is mentioned. Given this is a Method paper, the means of installation should be clearly laid out.*

      __This sentence is now modified. In the revised manuscript we now describe how to install the toolset and we give the link to the toolset website if further information is needed. __On this website, we provide a full video tutorial and a user manual. The user manual is provided as a supplementary material of the manuscript.

      * It would be helpful if there was an option to generate an output with the regions analysed (i.e., a JPG image with the data and the drawn line(s) on top). There are two reasons for this: i) A major problem with user-driven quantification is accidental double counting of regions (e.g., a user quantifies a part of an image and then later quantifies the same region). ii) Allows other users to independently verify measurements at a later time.*

      We agree that it is helpful to save the analyzed regions. To answer this comment and the other two reviewers' comments pointing at a similar feature, we have now included an automatic saving of the regions of interest. The user will be able to reopen saved regions of interest using a new function we included in the new version of PatternJ.

      * 3. Related to the above point, it is highlighted that each time point would need to be analysed separately (line 361-362). It seems like it should be relatively straightforward to allow a function where the analysis line can be mapped onto the next time point. The user could then adjust slightly for changes in position, but still be starting from near the previous timepoint. Given how prevalent timelapse imaging is, this seems like (or something similar) a clear benefit to add to the software.*

      We agree that the analysis of time series images can be a useful addition. We have added the analysis of time-lapse series in the new version of PatternJ. The principles behind the analysis of time-lapse series and an example of such analysis are provided in Figure 1 - figure supplement 3 and Figure 5, with accompanying text lines 140-153 and 360-372. The analysis includes a semi-automated selection of regions of interest, which will make the analysis of such sequences more straightforward than having to draw a selection on each image of the series. The user is required to draw at least two regions of interest in two different frames, and the algorithm will automatically generate regions of interest in frames in which selections were not drawn. The algorithm generates the analysis immediately after selections are drawn by the user, which includes the tracking of the reference channel.

      * Line 134-135. The level of accuracy of the searching should be clarified here. This is discussed later in the manuscript, but it would be helpful to give readers an idea at this point what level of tolerance the software has to noise and aperiodicity.

      *

      We agree with the reviewer that a clarification of this part of the algorithm will help the user better understand the manuscript.__ We have modified the sentence to clarify the range of search used and the resulting limits in aperiodicity (now lines 176-181). __Regarding the tolerance to noise, it is difficult to estimate it a priori from the choice made at the algorithm stage, so we prefer to leave it to the validation part of the manuscript. We hope this solution satisfies the reviewer and future users.

      *

      **Referees cross-commenting**

      I think the other reviewer comments are very pertinent. The authors have a fair bit to do, but they are reasonable requests. So, they should be encouraged to do the revisions fully so that the final software tool is as useful as possible.

      Reviewer #1 (Significance (Required)):

      Developing software tools for quantifying biological data that are approachable for a wide range of users remains a longstanding challenge. This challenge is due to: (1) the inherent problem of variability in biological systems; (2) the complexity of defining clearly quantifiable measurables; and (3) the broad spread of computational skills amongst likely users of such software.

      In this work, Blin et al., develop a simple plugin for ImageJ designed to quickly and easily quantify regular repeating units within biological systems - e.g., muscle fibre structure. They clearly and fairly discuss existing tools, with their pros and cons. The motivation for PatternJ is properly justified (which is sadly not always the case with such software tools).

      Overall, the paper is well written and accessible. The tool has limitations but it is clearly useful and easy to use. Therefore, this work is publishable with only minor corrections.

      *We thank the reviewer for the positive evaluation of PatternJ and for pointing out its accessibility to the users.

      *

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

      # Summary

      The authors present an ImageJ Macro GUI tool set for the quantification of one-dimensional repeated patterns that are commonly occurring in microscopy images of muscles.

      # Major comments

      In our view the article and also software could be improved in terms of defining the scope of its applicability and user-ship. In many parts the article and software suggest that general biological patterns can be analysed, but then in other parts very specific muscle actin wordings are used. We are pointing this out in the "Minor comments" sections below. We feel that the authors could improve their work by making a clear choice here. One option would be to clearly limit the scope of the tool to the analysis of actin structures in muscles. In this case we would recommend to also rename the tool, e.g. MusclePatternJ. The other option would be to make the tool about the generic analysis of one-dimensional patterns, maybe calling the tool LinePatternJ. In the latter case we would recommend to remove all actin specific wordings from the macro tool set and also the article should be in parts slightly re-written.

      *

      We agree with the reviewer that our initial manuscript used a mix of general and muscle-oriented vocabulary, which could make the use of PatternJ confusing especially outside of the muscle field. To make PatternJ useful for the largest community, we corrected the manuscript and the PatternJ toolset to provide the general vocabulary needed to make it understandable for every biologist. We modified the manuscript accordingly.

      * # Minor/detailed comments

      # Software

      We recommend considering the following suggestions for improving the software.

      ## File and folder selection dialogs

      In general, clicking on many of the buttons just opens up a file-browser dialog without any further information. For novel users it is not clear what the tool expects one to select here. It would be very good if the software could be rewritten such that there are always clear instructions displayed about which file or folder one should open for the different buttons.*

      We experienced with the current version of macOS that the file-browser dialog does not display any message; we suspect this is the issue raised by the reviewer. This is a known issue of Fiji on Mac and all applications on Mac since 2016. We provided guidelines in the user manual and on the tutorial video to correct this issue by changing a parameter in Fiji. Given the issues the reviewer had accessing the material on the PatternJ website, which we apologize for, we understand the issue raised. We added an extra warning on the PatternJ website to point at this problem and its solution. Additionally, we have limited the file-browser dialog appearance to what we thought was strictly necessary. Thus, the user will experience fewer prompts, speeding up the analysis.

      *

      ## Extract button

      The tool asks one to specify things like whether selections are drawn "M-line-to-M-line"; for users that are not experts in muscle morphology this is not understandable. It would be great to find more generally applicable formulations. *

      We agree that this muscle-oriented vocabulary can make the use of PatternJ confusing. We have now corrected the user interface to provide both general and muscle-specific vocabulary ("center-to-center or edge-to-edge (M-line-to-M-line or Z-disc-to-Z-disc)").*

      ## Manual selection accuracy

      The 1st step of the analysis is always to start from a user hand-drawn profile across intensity patterns in the image. However, this step can cause inaccuracy that varies with the shape and curve of the line profile drawn. If not strictly perpendicular to for example the M line patterns, the distance between intensity peaks will be different. This will be more problematic when dealing with non-straight and parallelly poised features in the image. If the structure is bended with a curve, the line drawn over it also needs to reproduce this curve, to precisely capture the intensity pattern. I found this limits the reproducibility and easy-usability of the software.*

      We understand the concern of the reviewer. On curved selections this will be an issue that is difficult to solve, especially on "S" curved or more complex selections. The user will have to be very careful in these situations. On non-curved samples, the issue may be concerning at first sight, but the errors go with the inverse of cosine and are therefore rather low. For example, if the user creates a selection off by 5 degrees, which is visually obvious, lengths will be affected by an increase of only 0.38%. The point raised by the reviewer is important to discuss, and we therefore added a paragraph to comment on the choice of selection (lines 94-98) and a supplementary figure to help make it clear (Figure 1 - figure supplement 1).*

      ### Reproducibility

      Since the line profile drawn on the image is the first step and very essential to the entire process, it should be considered to save together with the analysis result. For example, as ImageJ ROI or ROIset files that can be re-imported, correctly positioned, and visualized in the measured images. This would greatly improve the reproducibility of the proposed workflow. In the manuscript, only the extracted features are being saved (because the save button is also just asking for a folder containing images, so I cannot verify its functionality). *

      We agree that this is a very useful and important feature. We have added ROI automatic saving. Additionally, we now provide a simplified import function of all ROIs generated with PatternJ and the automated extraction and analysis of the list of ROIs. This can be done from ROIs generated previously in PatternJ or with ROIs generated from other ImageJ/Fiji algorithms. These new features are described in the manuscript in lines 120-121 and 130-132.

      *

      ## ? button

      It would be great if that button would open up some usage instructions.

      *

      We agree with the reviewer that the "?" button can be used in a better way. We have replaced this button with a Help menu, including a simple tutorial showing a series of images detailing the steps to follow by the user, a link to the user website, and a link to our video tutorial.

      * ## Easy improvement of workflow

      I would suggest a reasonable expansion of the current workflow, by fitting and displaying 2D lines to the band or line structure in the image, that form the "patterns" the author aims to address. Thus, it extracts geometry models from the image, and the inter-line distance, and even the curve formed by these sets of lines can be further analyzed and studied. These fitted 2D lines can be also well integrated into ImageJ as Line ROI, and thus be saved, imported back, and checked or being further modified. I think this can largely increase the usefulness and reproducibility of the software.

      *

      We hope that we understood this comment correctly. We had sent a clarification request to the editor, but unfortunately did not receive an answer within the requested 4 weeks of this revision. We understood the following: instead of using our 1D approach, in which we extract positions from a profile, the reviewer suggests extracting the positions of features not as a single point, but as a series of coordinates defining its shape. If this is the case, this is a major modification of the tool that is beyond the scope of PatternJ. We believe that keeping our tool simple, makes it robust. This is the major strength of PatternJ. Local fitting will not use line average for instance, which would make the tool less reliable.

      * # Manuscript

      We recommend considering the following suggestions for improving the manuscript. Abstract: The abstract suggests that general patterns can be quantified, however the actual tool quantifies specific subtypes of one-dimensional patterns. We recommend adapting the abstract accordingly.

      *

      We modified the abstract to make this point clearer.

      * Line 58: Gray-level co-occurrence matrix (GLCM) based feature extraction and analysis approach is not mentioned nor compared. At least there's a relatively recent study on Sarcomeres structure based on GLCM feature extraction: https://github.com/steinjm/SotaTool with publication: *https://doi.org/10.1002/cpz1.462

      • *

      We thank the reviewer for making us aware of this publication. We cite it now and have added it to our comparison of available approaches.

      * Line 75: "...these simple geometrical features will address most quantitative needs..." We feel that this may be an overstatement, e.g. we can imagine that there should be many relevant two-dimensional patterns in biology?!*

      We have modified this sentence to avoid potential confusion (lines 76-77).

      • *

      • Line 83: "After a straightforward installation by the user, ...". We think it would be convenient to add the installation steps at this place into the manuscript. *

      __This sentence is now modified. We now mention how to install the toolset and we provide the link to the toolset website, if further information is needed (lines 86-88). __On the website, we provide a full video tutorial and a user manual.

      * Line 87: "Multicolor images will give a graph with one profile per color." The 'Multicolor images' here should be more precisely stated as "multi-channel" images. Multi-color images could be confused with RGB images which will be treated as 8-bit gray value (type conversion first) images by profile plot in ImageJ. *

      We agree with the reviewer that this could create some confusion. We modified "multicolor" to "multi-channel".

      * Line 92: "...such as individual bands, blocks, or sarcomeric actin...". While bands and blocks are generic pattern terms, the biological term "sarcomeric actin" does not seem to fit in this list. Could a more generic wording be found, such as "block with spike"? *

      We agree with the reviewer that "sarcomeric actin" alone will not be clear to all readers. We modified the text to "block with a central band, as often observed in the muscle field for sarcomeric actin" (lines 103-104). The toolset was modified accordingly.

      * Line 95: "the algorithm defines one pattern by having the features of highest intensity in its centre". Could this be rephrased? We did not understand what that exactly means.*

      We agree with the reviewer that this was not clear. We rewrote this paragraph (lines 101-114) and provided a supplementary figure to illustrate these definitions (Figure 1 - figure supplement 2).

      * Line 124 - 147: This part the only description of the algorithm behind the feature extraction and analysis, but not clearly stated. Many details are missing or assumed known by the reader. For example, how it achieved sub-pixel resolution results is not clear. One can only assume that by fitting Gaussian to the band, the center position (peak) thus can be calculated from continuous curves other than pixels. *

      Note that the two sentences introducing this description are "Automated feature extraction is the core of the tool. The algorithm takes multiple steps to achieve this (Fig. S2):". We were hoping this statement was clear, but the reviewer may refer to something else. We agree that the description of some of the details of the steps was too quick. We have now expanded the description where needed.

      * Line 407: We think the availability of both the tool and the code could be improved. For Fiji tools it is common practice to create an Update Site and to make the code available on GitHub. In addition, downloading the example file (https://drive.google.com/file/d/1eMazyQJlisWPwmozvyb8VPVbfAgaH7Hz/view?usp=drive_link) required a Google login and access request, which is not very convenient; in fact, we asked for access but it was denied. It would be important for the download to be easier, e.g. from GitHub or Zenodo.

      *

      We are sorry for issues encountered when downloading the tool and additional material. We thank the reviewer for pointing out these issues that limited the accessibility of our tool. We simplified the downloading procedure on the website, which does not go through the google drive interface nor requires a google account. Additionally, for the coder community the code, user manual and examples are now available from GitHub at github.com/PierreMangeol/PatternJ, and are provided as supplementary material with the manuscript. To our knowledge, update sites work for plugins but not for macro toolsets. Having experience sharing our codes with non-specialists, a classical website with a tutorial video is more accessible than more coder-oriented websites, which deter many users.

      * Reviewer #2 (Significance (Required)):

      The strength of this study is that a tool for the analysis of one-dimensional repeated patterns occurring in muscle fibres is made available in the accessible open-source platform ImageJ/Fiji. In the introduction to the article the authors provide an extensive review of comparable existing tools. Their new tool fills a gap in terms of providing an easy-to-use software for users without computational skills that enables the analysis of muscle sarcomere patterns. We feel that if the below mentioned limitations could be addressed the tool could indeed be valuable to life scientists interested in muscle patterning without computational skills.

      In our view there are a few limitations, including the accessibility of example data and tutorials at sites.google.com/view/patternj, which we had trouble to access. In addition, we think that the workflow in Fiji, which currently requires pressing several buttons in the correct order, could be further simplified and streamlined by adopting some "wizard" approach, where the user is guided through the steps.

      *As answered above, the links on the PatternJ website are now corrected. Regarding the workflow, we now provide a Help menu with:

      1. __a basic set of instructions to use the tool, __
      2. a direct link to the tutorial video in the PatternJ toolset
      3. a direct link to the website on which both the tutorial video and a detailed user manual can be found. We hope this addresses the issues raised by this reviewer.

      *Another limitation is the reproducibility of the analysis; here we recommend enabling IJ Macro recording as well as saving of the drawn line ROIs. For more detailed suggestions for improvements please see the above sections of our review. *

      We agree that saving ROIs is very useful. It is now implemented in PatternJ.

      We are not sure what this reviewer means by "enabling IJ Macro recording". The ImageJ Macro Recorder is indeed very useful, but to our knowledge, it is limited to built-in functions. Our code is open and we hope this will be sufficient for advanced users to modify the code and make it fit their needs.*

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

      Summary In this manuscript, the authors present a new toolset for the analysis of repetitive patterns in biological images named PatternJ. One of the main advantages of this new tool over existing ones is that it is simple to install and run and does not require any coding skills whatsoever, since it runs on the ImageJ GUI. Another advantage is that it does not only provide the mean length of the pattern unit but also the subpixel localization of each unit and the distributions of lengths and that it does not require GPU processing to run, unlike other existing tools. The major disadvantage of the PatternJ is that it requires heavy, although very simple, user input in both the selection of the region to be analyzed and in the analysis steps. Another limitation is that, at least in its current version, PatternJ is not suitable for time-lapse imaging. The authors clearly explain the algorithm used by the tool to find the localization of pattern features and they thoroughly test the limits of their tool in conditions of varying SNR, periodicity and band intensity. Finally, they also show the performance of PatternJ across several biological models such as different kinds of muscle cells, neurons and fish embryonic somites, as well as different imaging modalities such as brightfield, fluorescence confocal microscopy, STORM and even electron microscopy.

      This manuscript is clearly written, and both the section and the figures are well organized and tell a cohesive story. By testing PatternJ, I can attest to its ease of installation and use. Overall, I consider that PatternJ is a useful tool for the analysis of patterned microscopy images and this article is fit for publication. However, i do have some minor suggestions and questions that I would like the authors to address, as I consider they could improve this manuscript and the tool:

      *We are grateful to this reviewer for this very positive assessment of PatternJ and of our manuscript.

      * Minor Suggestions: In the methodology section is missing a more detailed description about how the metric plotted was obtained: as normalized intensity or precision in pixels. *

      We agree with the reviewer that a more detailed description of the metric plotted was missing. We added this information in the method part and added information in the Figure captions where more details could help to clarify the value displayed.

      * The validation is based mostly on the SNR and patterns. They should include a dataset of real data to validate the algorithm in three of the standard patterns tested. *

      We validated our tool using computer-generated images, in which we know with certainty the localization of patterns. This allowed us to automatically analyze 30 000 images, and with varying settings, we sometimes analyzed 10 times the same image, leading to about 150 000 selections analyzed. From these analyses, we can provide with confidence an unbiased assessment of the tool precision and the tool capacity to extract patterns. We already provided examples of various biological data images in Figures 4-6, showing all possible features that can be extracted with PatternJ. In these examples, we can claim by eye that PatternJ extracts patterns efficiently, but we cannot know how precise these extractions are because of the nature of biological data: "real" positions of features are unknown in biological data. Such validation will be limited to assessing whether a pattern was found or not, which we believe we already provided with the examples in Figures 4-6.

      * The video tutorial available in the PatternJ website is very useful, maybe it would be worth it to include it as supplemental material for this manuscript, if the journal allows it. *

      As the video tutorial may have been missed by other reviewers, we agree it is important to make it more prominent to users. We have now added a Help menu in the toolset that opens the tutorial video. Having the video as supplementary material could indeed be a useful addition if the size of the video is compatible with the journal limits.

      * An example image is provided to test the macro. However, it would be useful to provide further example images for each of the three possible standard patterns suggested: Block, actin sarcomere or individual band.*

      We agree this can help users. We now provide another multi-channel example image on the PatternJ website including blocks and a pattern made of a linear intensity gradient that can be extracted with our simpler "single pattern" algorithm, which were missing in the first example. Additionally, we provide an example to be used with our new time-lapse analysis.

      * Access to both the manual and the sample images in the PatternJ website should be made publicly available. Right now they both sit in a private Drive account. *

      As mentioned above, we apologize for access issues that occurred during the review process. These files can now be downloaded directly on the website without any sort of authentication. Additionally, these files are now also available on GitHub.

      * Some common errors are not properly handled by the macro and could be confusing for the user: When there is no selection and one tries to run a Check or Extraction: "Selection required in line 307 (called from line 14). profile=getProfile( ;". A simple "a line selection is required" message would be useful there. When "band" or "block" is selected for a channel in the "Set parameters" window, yet a 0 value is entered into the corresponding "Number of bands or blocks" section, one gets this error when trying to Extract: "Empty array in line 842 (called from line 113). if ( ( subloc . length == 1 ) & ( subloc [ 0 == 0) ) {". This error is not too rare, since the "Number of bands or blocks" section is populated with a 0 after choosing "sarcomeric actin" (after accepting the settings) and stays that way when one changes back to "blocks" or "bands".*

      We thank the reviewer for pointing out these bugs. These bugs are now corrected in the revised version.

      * The fact that every time one clicks on the most used buttons, the getDirectory window appears is not only quite annoying but also, ultimately a waste of time. Isn't it possible to choose the directory in which to store the files only once, from the "Set parameters" window?*

      We have now found a solution to avoid this step. The user is only prompted to provide the image folder when pressing the "Set parameter" button. We kept the prompt for directory only when the user selects the time-lapse analysis or the analysis of multiple ROIs. The main reason is that it is very easy for the analysis to end up in the wrong folder otherwise.

      * The authors state that the outputs of the workflow are "user friendly text files". However, some of them lack descriptive headers (like the localisations and profiles) or even file names (like colors.txt). If there is something lacking in the manuscript, it is a brief description of all the output files generated during the workflow.*

      PatternJ generates multiple files, several of which are internal to the toolset. They are needed to keep track of which analyses were done, and which colors were used in the images, amongst others. From the user part, only the files obtained after the analysis All_localizations.channel_X.txt and sarcomere_lengths.txt are useful. To improve the user experience, we now moved all internal files to a folder named "internal", which we think will clarify which outputs are useful for further analysis, and which ones are not. We thank the reviewer for raising this point and we now mention it in our Tutorial.

      I don't really see the point in saving the localizations from the "Extraction" step, they are even named "temp".

      We thank the reviewer for this comment, this was indeed not necessary. We modified PatternJ to delete these files after they are used.

      * In the same line, I DO see the point of saving the profiles and localizations from the "Extract & Save" step, but I think they should be deleted during the "Analysis" step, since all their information is then grouped in a single file, with descriptive headers. This deleting could be optional and set in the "Set parameters" window.*

      We understand the point raised by the reviewer. However, the analysis depends on the reference channel picked, which is asked for when starting an analysis, and can be augmented with additional selections. If a user chooses to modify the reference channel or to add a new profile to the analysis, deleting all these files would mean that the user will have to start over again, which we believe will create frustration. An optional deletion at the analysis step is simple to implement, but it could create problems for users who do not understand what it means practically.

      * Moreover, I think it would be useful to also save the linear roi used for the "Extract & Save" step, and eventually combine them during the "Analysis step" into a single roi set file so that future re-analysis could be made on the same regions. This could be an optional feature set from the "Set parameters" window. *

      We agree with the reviewer that saving ROIs is very useful. ROIs are now saved into a single file each time the user extracts and saves positions from a selection. Additionally, the user can re-use previous ROIs and analyze an image or image series in a single step.

      * In the "PatternJ workflow" section of the manuscript, the authors state that after the "Extract & Save" step "(...) steps 1, 2, 4, and 5 can be repeated on other selections (...)". However, technically, only steps 1 and 5 are really necessary (alternatively 1, 4 and 5 if the user is unsure of the quality of the patterning). If a user follows this to the letter, I think it can lead to wasted time.

      *

      We agree with the reviewer and have corrected the manuscript accordingly (line 119-120).

      • *

      *I believe that the "Version Information" button, although important, has potential to be more useful if used as a "Help" button for the toolset. There could be links to useful sources like the manuscript or the PatternJ website but also some tips like "whenever possible, use a higher linewidth for your line selection" *

      We agree with the reviewer as pointed out in our previous answers to the other reviewers. This button is now replaced by a Help menu, including a simple tutorial in a series of images detailing the steps to follow, a link to the user website, and a link to our video tutorial.

      * It would be interesting to mention to what extent does the orientation of the line selection in relation to the patterned structure (i.e. perfectly parallel vs more diagonal) affect pattern length variability?*

      As answered to reviewer 1, we understand this concern, which needs to be clarified for readers. The issue may be concerning at first sight, but the errors grow only with the inverse of cosine and are therefore rather low. For example, if the user creates a selection off by 3 degrees, which is visually obvious, lengths will be affected by an increase of only 0.14%. The point raised by the reviewer is important to discuss, and we therefore have added a comment on the choice of selection (lines 94-98) as well as a supplementary figure (Figure 1 - figure supplement 1).

      * When "the algorithm uses the peak of highest intensity as a starting point and then searches for peak intensity values one spatial period away on each side of this starting point" (line 133-135), does that search have a range? If so, what is the range? *

      We agree that this information is useful to share with the reader. The range is one pattern size. We have modified the sentence to clarify the range of search used and the resulting limits in aperiodicity (now lines 176-181).

      * Line 144 states that the parameters of the fit are saved and given to the user, yet I could not find such information in the outputs. *

      The parameters of the fits are saved for blocks. We have now clarified this point by modifying the manuscript (lines 186-198) and modifying Figure 1 - figure supplement 5. We realized we made an error in the description of how edges of "block with middle band" are extracted. This is now corrected.

      * In line 286, authors finish by saying "More complex patterns from electron microscopy images may also be used with PatternJ.". Since this statement is not backed by evidence in the manuscript, I suggest deleting it (or at the very least, providing some examples of what more complex patterns the authors refer to). *

      This sentence is now deleted.

      * In the TEM image of the fly wing muscle in fig. 4 there is a subtle but clearly visible white stripe pattern in the original image. Since that pattern consists of 'dips', rather than 'peaks' in the profile of the inverted image, they do not get analyzed. I think it is worth mentioning that if the image of interest contains both "bright" and "dark" patterns, then the analysis should be performed in both the original and the inverted images because the nature of the algorithm does not allow it to detect "dark" patterns. *

      We agree with the reviewer's comment. We now mention this point in lines 337-339.

      * In line 283, the authors mention using background correction. They should explicit what method of background correction they used. If they used ImageJ's "subtract background' tool, then specify the radius.*

      We now describe this step in the method section.

      *

      Reviewer #3 (Significance (Required)):

      • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. Being a software paper, the advance proposed by the authors is technical in nature. The novelty and significance of this tool is that it offers quick and simple pattern analysis at the single unit level to a broad audience, since it runs on the ImageJ GUI and does not require any programming knowledge. Moreover, all the modules and steps are well described in the paper, which allows easy going through the analysis.
      • Place the work in the context of the existing literature (provide references, where appropriate). The authors themselves provide a good and thorough comparison of their tool with other existing ones, both in terms of ease of use and on the type of information extracted by each method. While PatternJ is not necessarily superior in all aspects, it succeeds at providing precise single pattern unit measurements in a user-friendly manner.
      • State what audience might be interested in and influenced by the reported findings. Most researchers working with microscopy images of muscle cells or fibers or any other patterned sample and interested in analyzing changes in that pattern in response to perturbations, time, development, etc. could use this tool to obtain useful, and otherwise laborious, information. *

      We thank the reviewer for these enthusiastic comments about how straightforward for biologists it is to use PatternJ and its broad applicability in the bio community.

    2. 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 manuscript, the authors present a new toolset for the analysis of repetitive patterns in biological images named PatternJ. One of the main advantages of this new tool over existing ones is that it is simple to install and run and does not require any coding skills whatsoever, since it runs on the ImageJ GUI. Another advantage is that it does not only provide the mean length of the pattern unit but also the subpixel localization of each unit and the distributions of lengths and that it does not require GPU processing to run, unlike other existing tools. The major disadvantage of the PatternJ is that it requires heavy, although very simple, user input in both the selection of the region to be analyzed and in the analysis steps. Another limitation is that, at least in its current version, PatternJ is not suitable for time-lapse imaging.

      The authors clearly explain the algorithm used by the tool to find the localization of pattern features and they thoroughly test the limits of their tool in conditions of varying SNR, periodicity and band intensity. Finally, they also show the performance of PatternJ across several biological models such as different kinds of muscle cells, neurons and fish embryonic somites, as well as different imaging modalities such as brightfield, fluorescence confocal microscopy, STORM and even electron microscopy.

      This manuscript is clearly written, and both the section and the figures are well organized and tell a cohesive story. By testing PatternJ, I can attest to its ease of installation and use. Overall, I consider that PatternJ is a useful tool for the analysis of patterned microscopy images and this article is fit for publication. However, i do have some minor suggestions and questions that I would like the authors to address, as I consider they could improve this manuscript and the tool:

      Minor Suggestions:

      In the methodology section is missing a more detailed description about how the metric plotted was obtained: as normalized intensity or precision in pixels. The validation is based mostly on the SNR and patterns. They should include a dataset of real data to validate the algorithm in three of the standard patterns tested. The video tutorial available in the PatternJ website is very useful, maybe it would be worth it to include it as supplemental material for this manuscript, if the journal allows it. An example image is provided to test the macro. However, it would be useful to provide further example images for each of the three possible standard patterns suggested: Block, actin sarcomere or individual band. Access to both the manual and the sample images in the PatternJ website should be made publicly available. Right now they both sit in a private Drive account. Some common errors are not properly handled by the macro and could be confusing for the user: When there is no selection and one tries to run a Check or Extraction: "Selection required in line 307 (called from line 14). profile=getProfile( <)>;". A simple "a line selection is required" message would be useful there. When "band" or "block" is selected for a channel in the "Set parameters" window, yet a 0 value is entered into the corresponding "Number of bands or blocks" section, one gets this error when trying to Extract: "Empty array in line 842 (called from line 113). if ( ( subloc . length == 1 ) & ( subloc [ 0 <]> == 0) ) {". This error is not too rare, since the "Number of bands or blocks" section is populated with a 0 after choosing "sarcomeric actin" (after accepting the settings) and stays that way when one changes back to "blocks" or "bands".<br /> The fact that every time one clicks on the most used buttons, the getDirectory window appears is not only quite annoying but also, ultimately a waste of time. Isn't it possible to choose the directory in which to store the files only once, from the "Set parameters" window? The authors state that the outputs of the workflow are "user friendly text files". However, some of them lack descriptive headers (like the localisations and profiles) or even file names (like colors.txt). If there is something lacking in the manuscript, it is a brief description of all the output files generated during the workflow. I don't really see the point in saving the localizations from the "Extraction" step, they are even named "temp". In the same line, I DO see the point of saving the profiles and localizations from the "Extract & Save" step, but I think they should be deleted during the "Analysis" step, since all their information is then grouped in a single file, with descriptive headers. This deleting could be optional and set in the "Set parameters" window. Moreover, I think it would be useful to also save the linear roi used for the "Extract & Save" step, and eventually combine them during the "Analysis step" into a single roi set file so that future re-analysis could be made on the same regions. This could be an optional feature set from the "Set parameters" window. In the "PatternJ workflow" section of the manuscript, the authors state that after the "Extract & Save" step "(...) steps 1, 2, 4, and 5 can be repeated on other selections (...)". However, technically, only steps 1 and 5 are really necessary (alternatively 1, 4 and 5 if the user is unsure of the quality of the patterning). If a user follows this to the letter, I think it can lead to wasted time. I believe that the "Version Information" button, although important, has potential to be more useful if used as a "Help" button for the toolset. There could be links to useful sources like the manuscript or the PatternJ website but also some tips like "whenever possible, use a higher linewidth for your line selection" It would be interesting to mention to what extent does the orientation of the line selection in relation to the patterned structure (i.e. perfectly parallel vs more diagonal) affect pattern length variability? When "the algorithm uses the peak of highest intensity as a starting point and then searches for peak intensity values one spatial period away on each side of this starting point" (line 133-135), does that search have a range? If so, what is the range? Line 144 states that the parameters of the fit are saved and given to the user, yet I could not find such information in the outputs. In line 286, authors finish by saying "More complex patterns from electron microscopy images may also be used with PatternJ.". Since this statement is not backed by evidence in the manuscript, I suggest deleting it (or at the very least, providing some examples of what more complex patterns the authors refer to). In the TEM image of the fly wing muscle in fig. 4 there is a subtle but clearly visible white stripe pattern in the original image. Since that pattern consists of 'dips', rather than 'peaks' in the profile of the inverted image, they do not get analyzed. I think it is worth mentioning that if the image of interest contains both "bright" and "dark" patterns, then the analysis should be performed in both the original and the inverted images because the nature of the algorithm does not allow it to detect "dark" patterns. In line 283, the authors mention using background correction. They should explicit what method of background correction they used. If they used ImageJ's "subtract background' tool, then specify the radius.

      Significance

      • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. Being a software paper, the advance proposed by the authors is technical in nature. The novelty and significance of this tool is that it offers quick and simple pattern analysis at the single unit level to a broad audience, since it runs on the ImageJ GUI and does not require any programming knowledge. Moreover, all the modules and steps are well described in the paper, which allows easy going through the analysis.
      • Place the work in the context of the existing literature (provide references, where appropriate). The authors themselves provide a good and thorough comparison of their tool with other existing ones, both in terms of ease of use and on the type of information extracted by each method. While PatternJ is not necessarily superior in all aspects, it succeeds at providing precise single pattern unit measurements in a user-friendly manner.
      • State what audience might be interested in and influenced by the reported findings. Most researchers working with microscopy images of muscle cells or fibers or any other patterned sample and interested in analyzing changes in that pattern in response to perturbations, time, development, etc. could use this tool to obtain useful, and otherwise laborious, information.
      • 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 am a biologist with extensive experience in confocal microscopy and image analysis using classical machine vision tools, particularly using ImageJ and CellProfiler.
    3. Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

      Learn more at Review Commons


      Referee #2

      Evidence, reproducibility and clarity

      Summary

      The authors present an ImageJ Macro GUI tool set for the quantification of one-dimensional repeated patterns that are commonly occurring in microscopy images of muscles.

      Major comments

      In our view the article and also software could be improved in terms of defining the scope of its applicability and user-ship. In many parts the article and software suggest that general biological patterns can be analysed, but then in other parts very specific muscle actin wordings are used. We are pointing this out in the "Minor comments" sections below. We feel that the authors could improve their work by making a clear choice here. One option would be to clearly limit the scope of the tool to the analysis of actin structures in muscles. In this case we would recommend to also rename the tool, e.g. MusclePatternJ. The other option would be to make the tool about the generic analysis of one-dimensional patterns, maybe calling the tool LinePatternJ. In the latter case we would recommend to remove all actin specific wordings from the macro tool set and also the article should be in parts slightly re-written.

      Minor/detailed comments

      Software

      We recommend considering the following suggestions for improving the software.

      File and folder selection dialogs

      In general, clicking on many of the buttons just opens up a file-browser dialog without any further information. For novel users it is not clear what the tool expects one to select here. It would be very good if the software could be rewritten such that there are always clear instructions displayed about which file or folder one should open for the different buttons.

      Extract button

      The tool asks one to specify things like whether selections are drawn "M-line-to-M-line"; for users that are not experts in muscle morphology this is not understandable. It would be great to find more generally applicable formulations.

      Manual selection accuracy

      The 1st step of the analysis is always to start from a user hand-drawn profile across intensity patterns in the image. However, this step can cause inaccuracy that varies with the shape and curve of the line profile drawn. If not strictly perpendicular to for example the M line patterns, the distance between intensity peaks will be different. This will be more problematic when dealing with non-straight and parallelly poised features in the image. If the structure is bended with a curve, the line drawn over it also needs to reproduce this curve, to precisely capture the intensity pattern. I found this limits the reproducibility and easy-usability of the software.

      Reproducibility

      Since the line profile drawn on the image is the first step and very essential to the entire process, it should be considered to save together with the analysis result. For example, as ImageJ ROI or ROIset files that can be re-imported, correctly positioned, and visualized in the measured images. This would greatly improve the reproducibility of the proposed workflow. In the manuscript, only the extracted features are being saved (because the save button is also just asking for a folder containing images, so I cannot verify its functionality).

      ? button

      It would be great if that button would open up some usage instructions.

      Easy improvement of workflow

      I would suggest a reasonable expansion of the current workflow, by fitting and displaying 2D lines to the band or line structure in the image, that form the "patterns" the author aims to address. Thus, it extracts geometry models from the image, and the inter-line distance, and even the curve formed by these sets of lines can be further analyzed and studied. These fitted 2D lines can be also well integrated into ImageJ as Line ROI, and thus be saved, imported back, and checked or being further modified. I think this can largely increase the usefulness and reproducibility of the software.

      Manuscript

      We recommend considering the following suggestions for improving the manuscript. Abstract: The abstract suggests that general patterns can be quantified, however the actual tool quantifies specific subtypes of one-dimensional patterns. We recommend adapting the abstract accordingly.

      Line 58: Gray-level co-occurrence matrix (GLCM) based feature extraction and analysis approach is not mentioned nor compared. At least there's a relatively recent study on Sarcomeres structure based on GLCM feature extraction: https://github.com/steinjm/SotaTool with publication: https://doi.org/10.1002/cpz1.462

      Line 75: "...these simple geometrical features will address most quantitative needs..." We feel that this may be an overstatement, e.g. we can imagine that there should be many relevant two-dimensional patterns in biology?!

      Line 83: "After a straightforward installation by the user, ...". We think it would be convenient to add the installation steps at this place into the manuscript.

      Line 87: "Multicolor images will give a graph with one profile per color." The 'Multicolor images' here should be more precisely stated as "multi-channel" images. Multi-color images could be confused with RGB images which will be treated as 8-bit gray value (type conversion first) images by profile plot in ImageJ.

      Line 92: "...such as individual bands, blocks, or sarcomeric actin...". While bands and blocks are generic pattern terms, the biological term "sarcomeric actin" does not seem to fit in this list. Could a more generic wording be found, such as "block with spike"?

      Line 95: "the algorithm defines one pattern by having the features of highest intensity in its centre". Could this be rephrased? We did not understand what that exactly means.

      Line 124 - 147: This part the only description of the algorithm behind the feature extraction and analysis, but not clearly stated. Many details are missing or assumed known by the reader. For example, how it achieved sub-pixel resolution results is not clear. One can only assume that by fitting Gaussian to the band, the center position (peak) thus can be calculated from continuous curves other than pixels.

      Line 407: We think the availability of both the tool and the code could be improved. For Fiji tools it is common practice to create an Update Site and to make the code available on GitHub. In addition, downloading the example file (https://drive.google.com/file/d/1eMazyQJlisWPwmozvyb8VPVbfAgaH7Hz/view?usp=drive_link) required a Google login and access request, which is not very convenient; in fact, we asked for access but it was denied. It would be important for the download to be easier, e.g. from GitHub or Zenodo.

      Significance

      The strength of this study is that a tool for the analysis of one-dimensional repeated patterns occurring in muscle fibres is made available in the accessible open-source platform ImageJ/Fiji. In the introduction to the article the authors provide an extensive review of comparable existing tools. Their new tool fills a gap in terms of providing an easy-to-use software for users without computational skills that enables the analysis of muscle sarcomere patterns. We feel that if the below mentioned limitations could be addressed the tool could indeed be valuable to life scientists interested in muscle patterning without computational skills.

      In our view there are a few limitations, including the accessibility of example data and tutorials at sites.google.com/view/patternj, which we had trouble to access. In addition, we think that the workflow in Fiji, which currently requires pressing several buttons in the correct order, could be further simplified and streamlined by adopting some "wizard" approach, where the user is guided through the steps. Another limitation is the reproducibility of the analysis; here we recommend enabling IJ Macro recording as well as saving of the drawn line ROIs. For more detailed suggestions for improvements please see the above sections of our review.

    4. 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

      I have trialled the package on my lab's data and it works as advertised. It was straightforward to use and did not require any special training. I am confident this is a tool that will be approachable even to users with limited computational experience. The use of artificial data to validate the approach - and to provide clear limits on applicability - is particularly helpful.

      The main limitation of the tool is that it requires the user to manually select regions. This somewhat limits the generalisability and is also more subjective - users can easily choose "nice" regions that better match with their hypothesis, rather than quantifying the data in an unbiased manner. However, given the inherent challenges in quantifying biological data, such problems are not easily circumventable.

      I have some comments to clarify the manuscript:

      1. A "straightforward installation" is mentioned. Given this is a Method paper, the means of installation should be clearly laid out.
      2. It would be helpful if there was an option to generate an output with the regions analysed (i.e., a JPG image with the data and the drawn line(s) on top). There are two reasons for this: i) A major problem with user-driven quantification is accidental double counting of regions (e.g., a user quantifies a part of an image and then later quantifies the same region). ii) Allows other users to independently verify measurements at a later time.
      3. Related to the above point, it is highlighted that each time point would need to be analysed separately (line 361-362). It seems like it should be relatively straightforward to allow a function where the analysis line can be mapped onto the next time point. The user could then adjust slightly for changes in position, but still be starting from near the previous timepoint. Given how prevalent timelapse imaging is, this seems like (or something similar) a clear benefit to add to the software.
      4. Line 134-135. The level of accuracy of the searching should be clarified here. This is discussed later in the manuscript, but it would be helpful to give readers an idea at this point what level of tolerance the software has to noise and aperiodicity.

      Referees cross-commenting

      I think the other reviewer comments are very pertinent. The authors have a fair bit to do, but they are reasonable requests. So, they should be encouraged to do the revisions fully so that the final software tool is as useful as possible.

      Significance

      Developing software tools for quantifying biological data that are approachable for a wide range of users remains a longstanding challenge. This challenge is due to: (1) the inherent problem of variability in biological systems; (2) the complexity of defining clearly quantifiable measurables; and (3) the broad spread of computational skills amongst likely users of such software.

      In this work, Blin et al., develop a simple plugin for ImageJ designed to quickly and easily quantify regular repeating units within biological systems - e.g., muscle fibre structure. They clearly and fairly discuss existing tools, with their pros and cons. The motivation for PatternJ is properly justified (which is sadly not always the case with such software tools).

      Overall, the paper is well written and accessible. The tool has limitations but it is clearly useful and easy to use. Therefore, this work is publishable with only minor corrections.

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

      Learn more at Review Commons


      Reply to the reviewers

      Response to Reviewer 1


      __Glycosaminoglycan (GAG)-binding proteins regulating essential processes such as cell growth and migration are essential for cell homeostasis. It is reported that the GAG has the ability to bind to Herpin sulfate. As both GAGs and the LPS lipid A disaccharide core of gram-negative bacteria contain negatively charged disaccharide units, the researchers proposed that heparin-binding peptides might have cryptic antimicrobial peptide motifs. To prove the hypothesis, they have synthesized five candidates [HBP1-5], which showed a binding affinity towards heparin and LPS binding. By using various methods, they showed that these molecules have antimicrobial activity. The key finding in this study is the finding of the CPC domain, where C is a cationic amino acid and P is a polar amino acid. __

      Major comments

      1. __ Even though the Authors propose here that CPC' clip motif is needed for antimicrobial activity. However, various studies have demonstrated that the mere presence of cationic amino or hydrophobic amino acids does not give the activity, the location of these amino acids at the strategic position is critically needed. The major issue in this work, the authors have not presented, whether there was a single CPC motif or multiple in the 5 peptides they have synthesized. Further, they need to demonstrate how are the charged and hydrophobic amino acids distributed in the peptides. these things will clearly explain the difference in the activity as well spectrum of the peptides. The authors should make an extra figure or add information highlighting this unique characteristic for better understanding to the reader.__

      We thank the reviewer for his/her comments and suggestions. We concur that the distribution of amino acids is crucial for the antimicrobial activity of the peptides and their ability to bind heparin. We also agree with the suggestion of illustrating the location of the CPC' motifs of HBPs in the context of the parental proteins and have accordingly done so in the new Supplementary Figure 1. In all cases, only one CPC' motif was identified in the antimicrobial region, as highlighted in the figure, and the inter-residue distances measured are consistent with the CPC' motif definition. Thus, we demonstrate that a CPC' motif exists in all five HBPs, which explains how they recognize and bind heparin.

      To illustrate the distribution of charged and hydrophobic amino acids in HBPs, we have also prepared new Supplementary Figure 2, displaying electrostatic potentials in the predicted HBP structures, and showing how the distribution of charged residues creates hydrophobic and cationic patches on the surface of the peptides. Our analysis reveals cationic patches to be surrounded by hydrophobic residues, which may explain the ability of the peptides to disrupt membranes and exert antimicrobial activity.

      __ It is strange to observe that there are quite a number of reports showing that the peptides derived from the Herprin binding proteins have antimicrobial activity, but no one has reported their efficacy in the in vivo mouse model. if possible, the authors could add their observations if in vivo studies were done. or as a future line of study.__

      We thank the reviewer for his/her comment on the observation of antimicrobial activity in peptides derived from heparin-binding proteins. Indeed, a few such studies have appeared in the literature, some with moderate success [1]. It is possible that a lack of understanding on how to identify heparin-binding regions in proteins and AMPs underlies their relative paucity. In this context, we believe our results will spur further efforts, specifically by providing a rationale on how to identify CPC' motifs hence heparin-binding regions in protein sequences.

      Regarding the suggestion of assessing the in vivo efficacy of HBPs, we would agree that it would be helpful for better understanding their potential therapeutic applications. However, we feel that such experiments are beyond the scope of our manuscript, which offers ample, compelling in vitro and in silico evidence of how heparin-binding proteins can be a source of AMPs. We have done this by showing that CPC' motifs embedded in such proteins can be unveiled, accurately defined in structural terms, and experimentally shown to possess antimicrobial activity. Furthermore, we have shown that heparin binding correlates with LPS binding, allowing us to propose a mechanistic explanation for how heparin binding can be related to antimicrobial activity.

      Translating these results to animal models is possibly premature at this stage as, from a classical medicinal chemistry perspective, it would require previous structural elaboration in terms of, e.g., optimized serum half-life or serum protein binding, both of which can modulate activity in in vivo studies regardless of heparin affinity or bactericidal activity per se. Ongoing work in our laboratories is focused in these directions and will be reported in due time.

      *Referees cross-commenting**

      Minor comments

      1. __ The presence of Cryptic antimicrobial domain in various heparin-binding proteins like laminin isoforms, von Willebrand factor, vitronectin, protein C inhibitor, matrix glycoproteins thrombospondin, proline arginine-rich end leucine-rich repeat protein and fibronectin, have been reported previous. It is not clear why the authors did not refer to that work. The authors should refer to the works. (same as reviewer 3)__

      We were aware of other prior studies on heparin-binding proteins and did indeed cite some of them, though not exhaustively for conciseness' sake. However, as encouraged by reviewers 1 and 3 we have cited the following studies:

      Malmström E, Mörgelin M, Malmsten M, Johansson L, Norrby-Teglund A, Shannon O, Schmidtchen A, Meijers JC, Herwald H. Protein C inhibitor--a novel antimicrobial agent. PLoS Pathog. 2009 Dec;5(12):e1000698. doi: 10.1371/journal.ppat.1000698. Epub 2009 Dec 18. PMID: 20019810; PMCID: PMC2788422.

      Ishihara, J., Ishihara, A., Fukunaga, K. et al. Laminin heparin-binding peptides bind to several growth factors and enhance diabetic wound healing. Nat Commun 9, 2163 (2018). https://doi.org/10.1038/s41467-018-04525-w

      Chillakuri Chandramouli R, Jones Céline and Mardon Helen J(2010), Heparin binding domain in vitronectin is required for oligomerization and thus enhances integrin mediated cell adhesion and spreading, FEBS Letters, 584, doi: 10.1016/j.febslet.2010.06.023

      Papareddy P, Kasetty G, Kalle M, Bhongir RK, Mörgelin M, Schmidtchen A, Malmsten M. NLF20: an antimicrobial peptide with therapeutic potential against invasive Pseudomonas aeruginosa infection. J Antimicrob Chemother. 2016 Jan;71(1):170-80. doi: 10.1093/jac/dkv322. Epub 2015 Oct 26. PMID: 26503666.

      All the earlier studies related to the antimicrobial activity of the peptides derived from the Heparin-binding protein reported a consensus Cardin and Weintraub motifs i.e, XBBBXXBX or XBBXBX, where X represents hydrophobic or uncharged amino acids, and B represents basic amino acids. However, in this work, the researchers report about the presence of the new CPC motif. So, this is unique and a novelty in the study.

      We thank the reviewers for these observations. Indeed, our quest to unveil CPC' motifs in antimicrobial regions of heparin-binding proteins is the key point of our investigation, and what distinguishes it from previous studies on consensus motifs such as XBBBXXBX or XBBXBX. We believe our definition of CPC' motifs in simple, structure-based, and experimentally verifiable terms is not only a significant departure but also a step forward from earlier views, highlighting the importance of a structural perspective in defining heparin-binding regions. In point of fact, we show that our peptides, even without consensus Cardin-Weintraub motifs, bind heparin with high affinity. The presence of the CPC' motif is crucial for such binding, as well as for LPS binding, and the new experiments performed at editor/reviewer's request, where the CPC motif in HBP5 is abolished, with predictable impact, fully support our view, see new section "Insights into the CPC' motif of HBP-5 and its implication on the antibacterial mechanism" and new Table 3 in the revised manuscript.

      __ Even though the researchers report on the role of the CPC motif in the antimicrobial activity and binding to the heprin, the authors did not show any data or draw the conclusions related to the CPC domain when it comes to differences in the activity. this is the weakness of the manuscript. (same as reviewer 2)__

      We welcome the reviewer's observation. To address it, we made and tested three HBP-5 mutants aimed at showing how alterations in the CPC' motif might influence interaction with heparin and LPS, as well as antimicrobial properties. The first two mutants involved replacing positively charged R10 and R14 residues with glutamine, similar in size and polarity but uncharged. As shown in the new section "Insights into the CPC' motif of HBP-5 and its implication on the antibacterial mechanism" and on the new Table 3 of the revised manuscript, the changes reduced heparin binding, i.e., shorter retention times on affinity chromatography, as well as LPS binding, i.e., a decrease in EC50 in the cadaverine assay (Table 3). The modifications had a lesser impact on antimicrobial activity, most likely due to the low resolution of MIC assays.

      In a further step to assess the effect of the CPC' motif on antimicrobial activity, we deleted it in full by replacing residues H9, R10 and R14 of HBP-5 by alanine. As expected, this DCPC' peptide showed a sharp reduction in both heparin and LPS binding (Table 3) and, most importantly, a significant and asymmetric change in antimicrobial activity, with substantial impact on Gram-negatives yet practically no effect on Gram-positives, suggesting that LPS plays a key role in this selective response. Altogether, these observations align with our hypothesis that heparin-binding proteins might exploit their intrinsic affinity for heparin as an opportunity to developing antimicrobial properties by leveraging structural similarities between glycosaminoglycans and LPS.

      __ It is strange to observe that there are quite a number of reports showing that the peptides derived from the Herprin (sic) binding proteins have antimicrobial activity, but no one has reported their efficacy in the in vivo mouse model. if possible, the authors could add their observations if in vivo studies were done. or as a future line of study. (Same as reviewer 2)__

      We would kindly direct attention to #2 in the response to reviewer 1 above.

      __ There are more than 20 different AMP databases or prediction software. however, not all of them are 100 % current, their success rate varies from 30-50% only. It needs to be investigated if adding this search in the hit peptides might increase the success rate of the extra in silico-based AMPs prediction software.__

      If we understand the question correctly, the reviewer wonders whether including a CPC' motif predictor would increase the accuracy of AMP search algorithms. In our view, this strategy has two main limitations to be considered: (i) locating a CPC' motif in a peptide sequence typically requires a known 3D structure. Unfortunately, this is not always the case, and for proteins lacking reliable 3D data it can be a challenging and resource-intensive process; (ii) while CPC' motifs may predispose proteins to evolve antimicrobial properties, it is unclear if this is a required feature for all AMPs. Imposing the presence of a CPC' motif may not be applicable to all AMPs, although it might help identifying peptides with specific activity against gram-negative strains.

      In summary, while the query of including a CPC' motif search tool in AMP predictors is intriguing and worthy of exploration for its potential bearing on antimicrobial research, it is technically complicated and beyond the scope of our manuscript.

      __Reviewer #1 (Significance (Required)): __

      __All the earlier studies related to the antimicrobial activity of the peptides derived from the Heparin-binding protein reported a consensus Cardin and Weintraub motifs i.e, XBBBXXBX or XBBXBX, where X represents hydrophobic or uncharged amino acids, and B represents basic amino acids. However, in this work, the researchers report about the presence of the new CPC motif. So this is unique and a novelty in the study. __

      Even though the researchers report on the role of the CPC motif in the antimicrobial activity and binding to the heparin, the authors did not show any data or draw conclusions related to the CPC domain when it comes to differences in the activity. This is the weakness of the manuscript.

      We would direct reviewer's attention to #1 in the Referee's cross-commenting section above.


      Response to Reviewer 2


      This is a very nice paper by the Andreu and Torrent groups that report the antimicrobial and heparin-binding of several encrypted peptides. Overall, this study presents an intriguing exploration into the potential dual functionality of glycosaminoglycan (GAG)-binding proteins, specifically heparin-binding proteins (HBPs), in recognizing lipopolysaccharide (LPS) and exhibiting antimicrobial properties. The findings, particularly the identification and characterization of novel encrypted peptides, such as HBP-5, are promising and contribute to our understanding of the intricate interplay between GAG-binding proteins and immunity. The data provided and methodology are thorough and well described. In sum, this is a very nice work. Please see below my minor comments.


      Minor comments:

      1. __ Fig. 1 legend does not show antimicrobial activity. Please remove from the figure legend title.__

      As pointed out by the reviewer, the legend was incorrect and has been corrected accordingly and now reads "Figure 1. Structural and bioinformatics analysis of HBPs".

      __ Discussion section: the authors should expand this section a bit to discuss recent work in the encrypted/cryptic peptide area. There are some recent relevant papers published in the past 3 years that should be discussed.__

      We agree with the reviewer's suggestion to expand the discussion section to address recent work in the field of encrypted/cryptic peptides. We have carefully reviewed the recent literature and added several references in this topic:

      Torres MDT, Melo MCR, Flowers L, Crescenzi O, Notomista E, de la Fuente-Nunez C. Mining for encrypted peptide antibiotics in the human proteome. Nat Biomed Eng. 2022 Jan;6(1):67-75. doi: 10.1038/s41551-021-00801-1. Epub 2021 Nov 4. Erratum in: Nat Biomed Eng. 2022 Dec;6(12):1451. PMID: 34737399.

      • *

      Santos MFDS, Freitas CS, Verissimo da Costa GC, Pereira PR, Paschoalin VMF. Identification of Antibacterial Peptide Candidates Encrypted in Stress-Related and Metabolic Saccharomyces cerevisiae Proteins. Pharmaceuticals (Basel). 2022 Jan 28;15(2):163. doi: 10.3390/ph15020163. PMID: 35215278; PMCID: PMC8877035.

      • *

      Boaro A, Ageitos L, Torres MT, Blasco EB, Oztekin S, de la Fuente-Nunez C. Structure-function-guided design of synthetic peptides with anti-infective activity derived from wasp venom. Cell Rep Phys Sci. 2023 Jul 19;4(7):101459. doi: 10.1016/j.xcrp.2023.101459. PMID: 38239869; PMCID: PMC10795512.

      __ References provided are a bit outdated and do not accurately reflect the latest in the field (see comment above).__

      We thank the reviewer for this comment. Older references were updated as suggested.

      __ Gram should be capitalized throughout the text.__

      Gram has been capitalized as suggested by the reviewer.

      __ Can the authors comment on the potential translatability of HBP-5? Please also comment on the potential advantages of having peptides that 1) bind to heparin; and 2) kill bacteria.__

      We appreciate the reviewer's interest in the potential of HBP-5. Indeed, we believe it has promise for clinical applications due to its unique attributes, but further studies, including in vivo experiments and pharmacokinetic assessments, are needed to fully evaluate its potential. The advantages of peptides that bind to heparin and kill bacteria include targeted delivery or localization of therapeutic agents, enhanced efficacy, and minimized off-target effects. HBP-5's ability to perturb outer membrane LPS, a crucial aspect of its antibacterial activity, makes it a promising approach to combat Gram-negative bacterial infections, which are often challenging to treat. By disrupting the outer membrane integrity, HBP-5 may also enhance the susceptibility of Gram-negative bacteria to other antimicrobial agents or host immune responses, underscoring its translational potential for treating bacterial infections.

      __ More details on the computational tools and methods used to mine the peptides are needed.__

      We have updated the Methods section to provide more details on the computational tools used for defining AMPs. Briefly, from the library of heparin-binding proteins obtained from previous studies [2] and AMP scanning for all these proteins was performed using the AMPA tool. The predicted antibacterial segments were located in the 3D structure of their respective proteins. Then, the CPC' motifs were searched in each segment following the criteria previously reported in [3, 4]. The motif involves two cationic residues (Arg or Lys) and a polar residue (preferentially Asn, Gln, Thr, Tyr or Ser), with fairly conserved distances between the carbons and the side chain center of gravity, defining a clip-like structure where heparin would be lodged. This structural motif is highly conserved and can be found in many proteins with reported heparin binding capacity. Finally, for all these regions, docking with a heparin disaccharide was performed using AutoDock Vina to evaluate the potential binding energy.



      Response to Reviewer 3


      __Summary: This manuscript has identified and investigated antimicrobial peptides from GAG binding proteins. Authors hypothesized that due to physiochemical similarity between GAG and LPS, fragments of GAG binding proteins might exert antimicrobial activity particularly against G- bacteria. Authors have identified few such AMPs that demonstrate LPS binding and displayed antibacterial activity. They have also solved NMR structure of the potent peptide and mode of action. __

      Major comments: AMPs are promising molecules that can serve as lead for the development of therapeutics against MDR bacteria. In particular, currently therapeutic options to treat MDR Gram negative pathogens are limited. The current study is interesting and provides new non-toxic AMPs. Conclusions drawn from the works are largely valid. However, authors should address following comments:

      1. __ The design and characterization of the peptide YI12WF is not described. Previous studies had shown design of β-boomerang peptides (Bhattacharjya and coworkers) that target LPS.__

      We thank the reviewer for this comment. YI12WF (YVLWKRKRFIFI-amide) has been previously reported [4, 5] and shown to bind LPS with high affinity. YI12WF also contains a CPC' motif that, if deleted, reduces heparin binding [4]. References have been added in the text.

      __ Mutations or substitution of the key residues peptide 5 might improve the novelty of the work.__

      We thank the reviewer for this comment and agree that targeted substitutions in HBP-5 might shed light on the importance of the CPC' motif. As this point was also raised by reviewer 1, we would direct the reviewer's attention to #2 in the *Referees cross-commenting** section above.

      __ How these peptides disrupt LPS permeability is not investigated. As LPS is the major target.__

      We thank the reviewer for this suggestion and have accordingly evaluated the outer membrane (OM) permeability of the peptides by the 1-N-phenyl-naphthylamine (NPN) assay, a widely used method to assess OM integrity in Gram-negative bacteria. NPN is typically unable to cross the intact outer membrane; however, when the membrane is damaged or disrupted, it can penetrate and interact with lipids and proteins inside the cell, leading to an increase in fluorescence which is directly correlated with the degree of OM permeability and serves as an indicator of membrane damage.

      Our results, illustrated in the new Figure 2D, show that all peptides are able to disrupt the OM of Gram-negative bacteria comparably to the LL-37 positive control, except for HBP2. Notably, HBP-5 exhibits the highest activity against OM, consistent with findings elsewhere in the manuscript and altogether confirming the ability of HBPs to bind to and disrupt the LPS structure.

      __ Are the D-enantiomers of the peptides active against bacteria?__

      We tested the antibacterial activity of the D-enantiomer of HBP5 (dHBP-and 5) and found it to be even higher than that of all-L HBP-5 against both Gram-negative and -positive bacteria, probably due to increased proteolytic stability as found in many AMP studies [6, 7]. As for LPS and heparin affinity, L- and D-HBP-5 behaved similarly (Table R1). As expected, the CD signatures of L- and D-HBP-5 were mirror images (Figure R1). These results suggest that the conformation of the CPC' motif is preserved in dHBP5, in tune with all previous results.

      Antibacterial Activity

      ID

      E. Coli

      P. Aeruginosa

      A. Baumannii

      S. Aureus

      E. Faecium

      L. monocytognes

      HPB-5

      0.4

      0.8

      0.2

      6.3

      25

      1.6

      dHBP-5

      0.1

      0.2

      0.2

      1.6

      0.4

      0.2



      Binding Affinity


      LPS (EC50, µM)

      Heparin (% Elution buffer)

      HPB-5

      0.9 {plus minus} 0.7

      98.0

      dHBP-5

      1.1 {plus minus} 0.8

      97.2

      Table R1. Antimicrobial activity of HBP-5 and dHBP-5









      Figure R1. CD spectra of HBP-5 (red line) and dHBP-5 (green line) in LPS (left panel) and heparin (right panel).


      __ 3D structure of peptide 5 is solved in DPC micelle which is a mimic for eukaryotic cells. Authors should attempt to determine structure in LPS as shown in several recent studies with potent AMPs thanatin, MSI etc.__

      We appreciate the suggestion and have indeed attempted to obtain NMR spectra of HBP-5 in LPS micelles. However, we've been hindered by peptide precipitation and, despite considerable efforts, have not been able to obtain satisfactory results thus far. In contrast, we have succeeded in obtaining CD spectra of HBP5 in LPS micelles, showing an a-helix conformation similar to the one in SDS micelles, hence suggesting similar conformation in both environments.

      Minor comments: There are examples of AMPs derived from human proteins. Authors should highlight such works.

      Other studies have been cited according to the reviewers' comments:

      Malmström E, Mörgelin M, Malmsten M, Johansson L, Norrby-Teglund A, Shannon O, Schmidtchen A, Meijers JC, Herwald H. Protein C inhibitor--a novel antimicrobial agent. PLoS Pathog. 2009 Dec;5(12):e1000698. doi: 10.1371/journal.ppat.1000698. Epub 2009 Dec 18. PMID: 20019810; PMCID: PMC2788422.

      Ishihara, J., Ishihara, A., Fukunaga, K. et al. Laminin heparin-binding peptides bind to several growth factors and enhance diabetic wound healing. Nat Commun 9, 2163 (2018). https://doi.org/10.1038/s41467-018-04525-w

      Chillakuri Chandramouli R, Jones Céline and Mardon Helen J(2010), Heparin binding domain in vitronectin is required for oligomerization and thus enhances integrin mediated cell adhesion and spreading, FEBS Letters, 584, doi: 10.1016/j.febslet.2010.06.023

      Papareddy P, Kasetty G, Kalle M, Bhongir RK, Mörgelin M, Schmidtchen A, Malmsten M. NLF20: an antimicrobial peptide with therapeutic potential against invasive Pseudomonas aeruginosa infection. J Antimicrob Chemother. 2016 Jan;71(1):170-80. doi: 10.1093/jac/dkv322. Epub 2015 Oct 26. PMID: 26503666.



      References

      1. Papareddy, P., et al., An antimicrobial helix A-derived peptide of heparin cofactor II blocks endotoxin responses in vivo. Biochimica et Biophysica Acta (BBA) - Biomembranes, 2014. 1838(5): p. 1225-1234.
      2. Ori, A., M.C. Wilkinson, and D.G. Fernig, A systems biology approach for the investigation of the heparin/heparan sulfate interactome. J Biol Chem, 2011. 286(22): p. 19892-904.
      3. Torrent, M., et al., The "CPC Clip Motif": A Conserved Structural Signature for Heparin-Binding Proteins.PLOS ONE, 2012. 7(8): p. e42692.
      4. Pulido, D., et al., Structural similarities in the CPC clip motif explain peptide-binding promiscuity between glycosaminoglycans and lipopolysaccharides. J R Soc Interface, 2017. 14(136).
      5. Bhunia, A., et al., Designed beta-boomerang antiendotoxic and antimicrobial peptides: structures and activities in lipopolysaccharide. J Biol Chem, 2009. 284(33): p. 21991-22004.
      6. Varponi, I., et al., Fighting Pseudomonas aeruginosa Infections: Antibacterial and Antibiofilm Activity of D-Q53 CecB, a Synthetic Analog of a Silkworm Natural Cecropin B Variant. Int J Mol Sci, 2023. 24(15).
      7. Chen, Y., et al., Comparison of Biophysical and Biologic Properties of α-Helical Enantiomeric Antimicrobial Peptides. Chemical Biology & Drug Design, 2006. 67(2): p. 162-173.
    2. 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 has identified and investigated antimicrobial peptides from GAG binding proteins. Authors hypothesized that due to physiochemical similarity between GAG and LPS, fragments of GAG binding proteins might exert antimicrobial activity particularly against G- bacteria. Authors have identified few such AMPs that demonstrate LPS binding and displayed antibacterial activity. They have also solved NMR structure of the potent peptide and mode of action.

      Major comments: AMPs are promising molecules that can serve as lead for the development of therapeutics against MDR bacteria. In particular, currently therapeutic options to treat MDR Gram negative pathogens are limited. The current study is interesting and provides new non-toxic AMPs. Conclusions drawn from the works are largely valid. However, authors should address following comments

      1. The design and characterization of the peptide YI12WF is not described. Previous studies had shown design of b-boomerang peptides (Bhattacharjya and coworkers) that target LPS.
      2. Mutations or substitution of the key residues peptide 5 might improve the novelty of the work.
      3. How these peptides disrupt LPS permeability is not investigated. As LPS is the major target.
      4. Are the D-enantiomers of the peptides active against bacteria?
      5. 3-D structure of peptide 5 is solved in DPC micelle which is a mimic for eukaryotic cells. Authors should attempt to determine structure in LPS as shown in several recent studies with potent AMPs thanatin, MSI etc,

      Minor comments: There are examples of AMPs derived from human proteins. Authors should highlight such works.

      Significance

      The work described in the manuscript is novel and hold promises to develop antimicrobials in future.

    3. 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 is a very nice paper by the Andreu and Torrent groups that report the antimicrobial and heparin-binding of several encrypted peptides. Overall, this study presents an intriguing exploration into the potential dual functionality of glycosaminoglycan (GAG)-binding proteins, specifically heparin-binding proteins (HBPs), in recognizing lipopolysaccharide (LPS) and exhibiting antimicrobial properties. The findings, particularly the identification and characterization of novel encrypted peptides, such as HBP-5, are promising and contribute to our understanding of the intricate interplay between GAG-binding proteins and immunity. The data provided and methodology are thorough and well described. In sum, this is a very nice work. Please see below my minor comments.

      Minor comments:

      • Fig. 1 legend does not show antimicrobial activity. Please remove from the figure legend title.
      • Discussion section: the authors should expand this section a bit to discuss recent work in the encrypted/cryptic peptide area. There are some recent relevant papers published in the past 3 years that should be discussed.
      • References provided are a bit outdated and do not accurately reflect the latest in the field (see comment above).
      • Gram should be capitalized throughout the text.
      • Can the authors comment on the potential translatability of HBP-5? Please also comment on the potential advantages of having peptides that 1) bind to heparin; and 2) kill bacteria.
      • More details on the computational tools and methods used to mine the peptides are needed.

      Significance

      The data provided and methodology are thorough and well described. In sum, this is a very nice work.

    4. 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

      Glycosaminoglycan (GAG)-binding proteins regulating essential processes such as cell growth and migration are essential for cell homeostasis. It is reported that the GAG has the ability to bind to Herpin sulfate. As both GAGs and the LPS lipid A disaccharide core of gram-negative bacteria contain negatively charged disaccharide units, the researchers proposed that heparin-binding peptides might have cryptic antimicrobial peptide motifs. To prove the hypothesis, they have synthesized five candidates [HBP1-5], which showed a binding affinity towards heparin and LPS binding. By using various methods, they showed that these molecules have antimicrobial activity. The key finding in this study is the finding of the CPC domain, where C is a cationic amino acid and P is a polar amino acid.

      Major comments

      1. Even though the Authors propose here that CPC' clip motif is needed for antimicrobial activity. However, various studies have demonstrated that the mere presence of cationic amino or hydrophobic amino acids does not give the activity, the location of these amino acids at the strategic position is critically needed. The major issue in this work, the authors have not presented, whether there was a single CPC motif or multiple in the 5 peptides they have synthesised. Further, they need to demonstrate how are the charged and hydrophobic amino acids distributed in the peptides. these things will clearly explain the difference in the activity as well spectrum of the peptides. The authors should make an extra figure or add information highlighting this unique characteristic for better understanding to the reader.
      2. It is strange to observe that there are quite a number of reports showing that the peptides derived from the Herprin binding proteins have antimicrobial activity, but no one has reported their efficacy in the in vivo mouse model. if possible, the authors could add their observations if in vivo studies were done. or as a future line of study.

      Minor comments

      1. The presence of Cryptic antimicrobial domain in various heparin-binding proteins like laminin isoforms, von Willebrand factor, vitronectin, pro-tein C inhibitor, matrix glycoproteins thrombospondin, proline arginine-rich end leucine-rich repeat protein and fibronectin, have been reported previous. It is not clear why the authors did not refer to that work. the authors should refer to the works.

      Referees cross-commenting

      Minor comments

      1. The presence of Cryptic antimicrobial domain in various heparin-binding proteins like laminin isoforms, von Willebrand factor, vitronectin, pro-tein C inhibitor, matrix glycoproteins thrombospondin, proline arginine-rich end leucine-rich repeat protein and fibronectin, have been reported previous. It is not clear why the authors did not refer to that work. the authors should refer to the works. (same as reviewer 3)

      All the earlier studies related to the antimicrobial activity of the peptides derived from the Heparin-binding protein reported a consensus Cardin and Weintraub motifs i.e, XBBBXXBX or XBBXBX, where X represents hydrophobic or uncharged amino acids, and B represents basic amino acids. However, in this work, the researchers report about the presence of the new CPC motif. So this is unique and a novelty in the study.

      Even though the researchers report on the role of the CPC motif in the antimicrobial activity and binding to the heprin, the authors did not show any data or draw the conclusions related to the CPC domain when it comes to differences in the activity. this is the weakness of the manuscript. (same as reviwer 2) 2. It is strange to observe that there are quite a number of reports showing that the peptides derived from the Herprin binding proteins have antimicrobial activity, but no one has reported their efficacy in the in vivo mouse model. if possible, the authors could add their observations if in vivo studies were done. or as a future line of study.(Same as reviewer 2)

      Significance

      All the earlier studies related to the antimicrobial activity of the peptides derived from the Heparin-binding protein reported a consensus Cardin and Weintraub motifs i.e, XBBBXXBX or XBBXBX, where X represents hydrophobic or uncharged amino acids, and B represents basic amino acids. However, in this work, the researchers report about the presence of the new CPC motif. So this is unique and a novelty in the study.

      Even though the researchers report on the role of the CPC motif in the antimicrobial activity and binding to the heprin, the authors did not show any data or draw the conclusions related to the CPC domain when it comes to differences in the activity. this is the weakness of the manuscript.

      There are more than 20 different AMP databases or prediction software. however, not all of them are 100 % current, their success rate varies from 30-50% only. It needs to be investigated if adding this search in the hit peptides might increase the success rate of the extra in silico-based AMPs prediction software

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

      Learn more at Review Commons


      Reply to the reviewers

      Description of the planned revisions

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

      • Again, in Figure 5, were FoxP3/CD4+ cells enumerated? Author Response: Fig 5 showed that the inflammatory score, and activation of CD4 and CD8 cells, were lower in the intestine of DSS-treated mice transplanted with Jag1Ndr/Ndr lymphocytes than in those transplanted with Jag1+/+ lymphocytes. However, in Figure 5 we had not quantified the number of FoxP3/CD4+ cells (Tregs). We agree that it would be interesting to know whether the dampened intestinal inflammation (in response to a classical inflammatory disease model (DSS-treatment)) is also mediated by excess Tregs. We will therefore now quantify Foxp3+ cells on the intestinal sections of experimental animals used for acquisition of data in Fig 5.

      • *

      Description of the revisions that have already been incorporated in the transferred manuscript.

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

      Reviewer 1 comment: This is an interesting study that examines defects in the Jag1ndr/ndr mouse model of Alagille syndrome. The novel aspects of this manuscript are the comparisons, at many levels, between the mouse model and ALG patient samples, including an examination of immune profiles. The conclusions that the Jag1ndr/ndr mouse model is an accurate representation of the human ALG syndrome appear valid. However the reported differences in immune profiles, particularly in the Jag1ndr/ndr mouse model are difficult to understand. The data presented indicate a reduction in CD4+ cells in the Jag1ndr/ndr mouse at day P3 in both liver and spleen. Additionally, the authors report differences between the the Jag1ndr/ndr mouse and controls at day P30 in the relative percentages of DN, DP and SP CD4 and CD8 cells in the thymus. When examining the peripheral lymphoid system, CD4+ numbers are the same in both the Jag1ndr/ndr animals and controls however CD8+ numbers are reduced and FoxP3/CD4+ cells are increased in both the spleen and the thymus. FoxP3/CD4+ T cells are usually assumed to be regulatory T cells that dampen the inflammatory responses of T cells. Therefore, the increase in this population in an animal model of what is assumed to be an inflammatory disease is confusing and confounding. The authors do not present a clear analysis of how they feel an increase of Tregs would lead to this disease. One possibility is that this population is not functioning as conventional Tregs and rather are promoting inflammation but this conclusion would require a functional analysis of this population of cells, at the very least in an in vitro analysis of T cell suppression. From an immunologist's point of view, their data are antithetical to what one would expect to find in an inflammatory disease. Perhaps this reviewer is missing an important point but if I am missing it, then other who read this manusgcript also may be confused.

      Author Response: *We thank the reviewer for carefully assessing our work, and for noting which aspects of the immune analyses should be more thoroughly explained. We apologize for any confusion, which a clearer introduction will help to avoid. *

      *Alagille syndrome is not thought of as an inflammatory disorder, it is a congenital disorder affecting bile duct development (Kohut et al 2021, Semin Liver Dis). During normal bile duct development, JAG1+ portal fibroblasts signal to NOTCH2+ hepatoblasts to instruct bile duct development. In the context of low JAG1 signaling, hepatoblasts either fail to adopt a cholangiocyte fate, or fail to undergo bile duct morphogenesis, resulting in bile duct paucity and cholestasis. This cholestasis should activate inflammatory processes leading to fibrosis, which is the subject of this study. *

      • *

      We agree with the reviewer that Tregs would be expected to suppress inflammation, and our data are consistent with Treg suppression of inflammation. We show, for the first time, that Tregs are enriched in Jag1Ndr/Ndr mice (Fig 4) and present evidence that they suppress inflammation (Fig 5) and fibrosis (Fig 6), which could explain the atypical fibrosis seen in patients with ALGS.

      • *

      *To clarify that ALGS is a genetic liver disease affecting bile duct formation, we: *

      1. Modified and extended the following text in the Introduction (Page 2, lines 14-17): “ALGS is mainly caused by mutations in the Notch ligand JAGGED1 (JAG1, 94%) (Mašek & Andersson, 2017; Oda et al, 1997), affecting bile duct development and morphogenesis, resulting in bile duct paucity and cholestasis. Immune dysregulation has also been described (Tilib Shamoun et al, 2015), but how this might interact with liver disease in ALGS to affect fibrosis is not known.
      2. *Introduce the disease, the animal model, and the scientific question in a schematic in new Fig 1A. *
      3. * Reviewer 1 comment: Minor points that should be addressed include: • The source cells used in the transfer experiments reported in Figure 5 is unclear. Are they using total spleen cells with T, B and myeloid cells or are they using purified T cells. And if it is the latter, have they assessed the ratio of CD4+ versus FoxP3/CD4+ cells in the transferred cells?

      Author Response: *Total spleen cells including all lymphocytes were transplanted, as described in Materials and Methods. The constituent T-cell populations are characterized and shown in Fig 4F. To clarify this, we: *

      1. *added the text “Adoptive transfer of lymphocytes” to the schematic in Fig 5A, FigS5A, and Fig 6A, and *
      2. modified the opening paragraph related to results presented in Fig.5 and FigS5 in the following way (page 8, line 209): “To investigate Jag1Ndr/Ndr T cell function, we performed adoptive transfer of the splenic lymphocytes into Rag1-/- mice, which lack mature B- and T cell populations, but provide a host environment with normal Jag1 (Mombaerts et al, 1992).
      3. *

      *To acknowledge that B-cells and innate lymphoid cells might contribute to the observed results, we include a following sentence in the Discussion: *

      (page 12, lines 369-371) “Finally, our experimental setup does not exclude an additional contribution of other lymphocytes (B-cells or innate lymphoid cells) to the BDL-induced fibrosis, and selective testing of the individual subpopulations would be an intriguing follow up to this study.”

      Reviewer 1 comment: In the DSS experiments in Figure 5, there does not appear to be a no DSS control. What does the architecture look like without DSS?

      Author Response: The intestinal architecture and phenotype of mice transplanted with Jag1+/+ or Jag1Ndr/Ndr lymphocytes, not treated with DSS, are presented in Supplementary Figure 5. In the absence of DSS, Jag1+/+- or Jag1Ndr/Ndr -transplanted mice exhibit no overt differences in survival or weight gain/loss. The intestinal inflammatory score was not different in the two conditions and was *2.29 +/-0.44 and 2.03 +/-0.92 for Jag1+/+- or Jag1Ndr/Ndr -transplanted mice, respectively. *

      To compare the results with and without DSS, we added the following text to the results section, when describing the DSS results (Page 9, lines 223-226):

      As expected, histological scoring of intestinal and colonic inflammation revealed elevated inflammation in Jag1+/+→Rag1-/- mice treated with DSS (Fig. 5C,D) compared to Jag1+/+→Rag1-/- mice not treated with DSS (Fig. S5). However, there was significantly less inflammation in Jag1Ndr/Ndr→Rag1-/- mice than in Jag1+/+→Rag1-/- mice (Fig. 5C,D)."

      Reviewer 1 comment: The authors noted that splenomegaly was observed in the Jag1ndr/ndr mouse model. Again this is antithetical to what one would expect when one sees an increase in FoxP3/CD4+ T regs.

      Author Response: *We thank the reviewer for pointing at a possible discrepancy, related to Fig1 in which we report the presence of splenomegaly. Although there can be multiple causes of splenomegaly, it is one of the hallmarks of portal hypertension (as also corroborated by Reviewer 2), tightly connected with liver fibrosis, present in patients with ALGS and we report it as such in the manuscript. To clarify this, we added the following text sections: *

      1. Results (page 2, lines 37,38) “Liver fibrosis compresses blood vessels and reduces their blood flow, leading to portal hypertension, a serious consequence of liver disease which can manifest as splenomegaly.
      2. Discussion (page 13, line 394-401): “Splenomegaly has been described as a consequence of portal hypertension in ALGS (Kamath et al, 2020), but could also be attributed to immune-related pathology. Jag1Ndr/Ndr mice exhibit splenomegaly as early as P10, and is exacerbated at P30 ( 1E,F). Patients with other liver diseases display portal hypertension and cirrhosis, with both splenomegaly and hypersplenism associated with a high CD4+/CD8+ ratio, but a low Treg+/CD4+ ratio (Nomura et al, 2014). However, Jag1Ndr/Ndr mice present with splenomegaly but not hypersplenism. An overactive spleen (hypersplenism) would remove red blood cells which are instead enriched in Jag1Ndr/Ndr mice, and Tregs were enriched in Jag1Ndr/Ndr mice, not depleted as seen in cirrhosis/hypersplenism. These data are thus consistent with portal hypertension-induced splenomegaly rather than hypersplenism.*” *

      Reviewer #1 (Significance (Required)):

      Reviewer 1 comment: The strengths of this paper are the careful comparisons between the mouse model and the human ALG syndrome. These comparisons are valuable and worth publication.

      Author Response: We thank the reviewer for these comments.

      Reviewer 1 comment: Weaknesses are stated above. Needs a clearer explanation for their immune analysis.

      Author Response: *We thank the reviewers for highlighting points requiring clarification and hope the proposed text changes and additional data presented in response to the comments of all three reviewers lead to a significant clarification of the immunological aspect of our study. *

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

      Reviewer 2 comment:

      Summary: Masek and colleagues use multi-pronged studies on the Jag1[Ndr/Ndr] mouse model of Alagille syndrome (ALGS) combined with transcriptomic analysis on livers from patients with ALGS to elucidate the potential mechanisms regulating liver fibrosis in this disease. The authors first show that Jag1[Ndr/Ndr] animals develop pericellular and perisinusoidal fibrosis and exhibit evidence for portal hypertension, similar to patients with ALGS. Single-cell RNA-sequencing indicated more hepatoblasts and less hepatocytes, relatively speaking, in Jag1[Ndr/Ndr] P3 livers, which suggested hampering of hepatoblast differentiation to hepatocytes. Deconvolution of previously generated bulk RNA-seq data from Jag1[Ndr/Ndr] P10 livers and GESA on RNAseq data from livers of these mice and patients with ALGS confirmed the P3 scRNA-seq observations and indicated mild pro-inflammatory activation of immature hepatocytes in ALGS livers. GESA also suggested an inability of Jag1[Ndr/Ndr] livers to attract T cells upon cholestatic injury. Indeed, 25-color flow cytometry on liver and spleen from mutant and control mice indicated a defect in T cell response to cholestasis in this model. The authors then examined the effects of the Ndr mutation on T-cell development and function. They found that the Ndr/Ndr thymi were significantly smaller than control thymi. Moreover, Ndr/Ndr thymi showed an increase in CD4+ T-cells and Tregs at the expense of double-positive T-cells. The authors then performed lymphocyte transplantation studies and concluded that Ndr/Ndr T-cells fail to mount an adequate response to inflammation in a DSS model of ulcerative colitis. The authors tested the contribution of Ndr/Ndr immune cells to liver fibrosis in a model of experimentally induced cholestasis (bile duct ligation; BDL). Ndr/Ndr T-cells did not show any defects in migrating into the liver upon BDL. However, the periportal fibrosis observed in BDL model was reduced in animals receiving Ndr/Ndr immune cells compared to those receiving Jag1+/+ immune cells. This was accompanied by significantly less aSMA staining in these livers. Finally, reanalysis of bulk RNAseq data from liver samples from ALGS and other liver diseases suggested that the presence of FOXP3+ T-reg cells in the liver is associated with higher liver fibrosis in non-ALGS liver diseases but lower liver fibrosis in ALGS livers. The authors have used an impressive combination of single-cell RNA-sequencing, reanalysis of previous bulk RNA-sequencing data from their group and others, 25-color FACS analysis, and adoptive immune transfer experiments in this manuscript, and systematically provide quantification and statistical analysis for their data. Overall, this is an interesting and important study. Prior studies are referenced appropriately. The text and figures are clear and accurate. I don't think any additional experiments are essential. However, the issues listed under Major comments should be discussed and clarified in the manuscript, especially the first item.

      Author Response: *We sincerely thank the reviewer for the comprehensive and insightful assessment of our manuscript. We are particularly gratified to note your acknowledgment of the thoroughness of our experimental approach and the clarity of our presentation. We are pleased that no further experiments would be required, and will address the points raised under Major comments which enhance our study's quality and accessibility. *

      Reviewer 2 comment:

      Major comments:

      • Only a small fraction of the cells in scRNA-seq experiments have been assigned to hepatocytes/hepatoblast clusters, with the majority of these cells allocated to Hepato-Ery cluster. This suggests that many hepatocytes and potentially hepatoblasts have been lost during sample preparation. The authors should discuss this issue and its potential implications on the interpretation of the cell ratios and gene expression conclusions of scRNA-seq data. Author Response: We agree with the reviewer regarding this aspect of our study. We mentioned this limitation in the supplementary methods section: ”Liver parenchymal cells constituted ~6.5% of cells at E16.5, and ~7.5% of cells at P3 and included mesenchymal cells, endothelial cells, hepatoblasts and hepatocytes (Fig. S1D), this parenchymal proportion is lower than in vivo, but consistent with ex vivo liver digest (Guilliams et al, 2022).” We recognize it may be too inaccessible there, and we thus added the following text to the Discussion section of the manuscript: (Pages 11-12, lines 330-337) “A limitation of this study is the underrepresentation of the hepatoblast/cyte parenchymal cells in the scRNA-seq dataset (Fig. 2A-D), which constituted ~6.5% of analyzed cells at E16.5, and ~7.5% of cells at P3 (Fig. S1D). This parenchymal proportion is lower than in vivo, but is consistent with scRNA seq datasets obtained with ex vivo liver digest (Guilliams et al, 2022). One risk is that cell stress as a result of dissociation could result in further loss of injured Jag1Ndr/Ndr hepatocytes, impacting the interpretation of cell type abundance. Nuclear scRNAseq can overcome cell type-dependent dissociation sensitivity bias (Guilliams et al, 2022), and could provide further insights into Jag1Ndr/Ndr livers at the single cell level. Nonetheless, both bulk RNA seq deconvolution and histological analyses confirmed that patients and Jag1Ndr/Ndr mice exhibit hepatoblast enrichment and less differentiated hepatocytes.

      Reviewer 2 comment: The Jag1[Ndr/Ndr] strain is an excellent model for various aspects of ALGS phenotypes. However, when it comes to linking the effects of this mutation to the function of a specific cell type, it is worth considering that Jag1[Ndr/Ndr] might not recapitulate the effects of loss of one copy of JAG1 observed in most patients with ALGS. This is especially important given the sensitivity of various cellular and organ-level processes to the degree of Notch pathway activation. In the context of the present manuscript, it is possible that what the authors have observed in Jag1[Ndr/Ndr] lymphocytes does not mirror how a JAG1-heterozygous human lymphocyte behaves. This is not a major concern, but it is worth considering.

      Author Response: We agree and thus added the following discussion paragraph (page 11, lines 315-321) “In patients with ALGS, who have a single mutation in either JAG1 or NOTCH2, the remnant healthy allele(s) could be expected to mediate signaling. However, some JAG1 mutations exhibit dominant negative effects (Ponio et al, 2007; Xiao et al, 2013; Guan et al, 2023), which could entail further repression of JAG1/NOTCH2 signaling. In this context, it is important to note that the Jag1Ndr/Ndr mice are homozygous for the missense mutation, but retain some JAG1 activity, and it is not clear to which degree this mimics JAG1 heterozygosity in humans. It would be of interest to test whether Jag1 potency affects hepatoblast differentiation or injury-induced reversion of hepatocytes in patients as a function of their genotype.

      Reviewer 2 comment: •The basis for the opposite type of correlation between COL1A1 expression and POXP3 level in ALGS versus non-ALGS liver disease is not clear.

      Author Response: We thank the reviewer for pointing out the unclear interpretation of the patient data. In patients with ALGS, the extent of fibrosis is likely to be highly multifactorial, involving (as we show) hepatocyte immaturity, dampened inflammation, and immune system dysregulation (possibly involving more than T-cells). Since human patients ARE so heterogeneous, teasing apart the relative contribution of each is currently outside the scope of our study, but will be an important area of future research. Nonetheless we thought it was important and interesting to show these patterns in supplementary Fig 6, now extended with further data, and analyses, and described in the following manner:

      • *

      Results section: (page 10, lines 267-275) “Liver damage in non-ALGS liver disease (using liver injury marker LGALS3BP) (Yang et al, 2021), was positively correlated with recruitment of lymphocytes (including CD8A+,and FOXP3+ populations of T cells), as well as the extent of fibrosis (COL1A1 abundance) (Fig. S6G). However, in ALGS, the extent of liver damage, lymphocyte recruitment and fibrosis were unlinked (Fig. S6G). These data are in line with the observation that liver stiffness (a proxy for fibrosis) in ALGS is independent of biomarkers of liver disease (Leung et al, 2023). While Treg infiltration in ALGS was independent of liver damage, it exhibited a tendency towards a negative correlation with fibrosis (Fig. S6G), corroborating that elevated levels of Tregs may limit fibrosis in ALGS. Altogether, these data suggest that the liver and lymphocytes may be differentially affected in different patients with ALGS, a disorder that is well known for its heterogenous presentation.

      Minor comments:

      • Page 2, last paragraph of Introduction, Page 12 last sentence, and Supplementary Methods: Please use "adoptive immune transfer" instead of "adaptive immune transfer". • Pages 3 and 4: Reference is made to Figures 3E-O, which appears to be Figure 2E-O. • Figure 3 legend: "Analysis in (E) is one-way ANOVA with Dunnett's multiple comparison test". Panel E compares two means, so ANOVA is not the appropriate statistical analysis for these data. Is this sentence related to panel D? • Page 9: Please correct misspelling: "response to intestinal insult (Fig. 5). W therefore". • The Science Translation Medicine references lack page number. Author Response: *We thank the reviewer deeply for taking the time to meticulously note and convey these errors, helping us to correct these. The suggested corrections have been implemented. Science Transl Med is an online journal and does not have page numbers – we have added an issue number to facilitate retrieval of these references. *

      • *

      Additionally, we noticed that the image of a consecutive liver section with CYP1A2 staining from Jag1Ndr/Ndr liver in Fig 2 L was accidentally flipped along the horizontal axis, which we have now corrected. We also changed the scRNAseq cell cluster naming from Hepatoblasts/cytes, Hepato_Ery, and Kupffer cells, Kuffer cells_Ery to Hepatoblasts/cytes I, and II, and Kupffer cells I and II, respectively, to match the Neutrophil progenitors I and II naming convention. Names were subsequently also changed in Fig S1 and methods.

      **Referees cross-commenting**

      To my knowledge, ALGS is not considered to be an inflammatory disorder. Furthermore, the splenomagaly observed in the mouse model could be due to portal hypertension rather than a primary immune disturbance. Having said that, I agree with the other reviewers that the manuscript will benefit from further discussion and clarification on the immune-related observations.

      Author Response: We thank Reviewer 2 for indicating to Reviewer 1 that ALGS is not considered an inflammatory disorder, which we agree with. It was not our intention to convey this idea. To avoid confusion, we now:

      1. *Added a schematic in Fig 1A. *
      2. Modified and extended the following text in the Introduction: (Page 2, lines 14-17): “ALGS is mainly caused by mutations in the Notch ligand JAGGED1 (JAG1, 94%) (Mašek & Andersson, 2017; Oda et al, 1997), affecting bile duct development and morphogenesis, resulting in bile duct paucity and cholestasis. Immune dysregulation has also been described (Tilib Shamoun et al, 2015), but how this might interact with liver disease in ALGS to affect fibrosis is not known. *Furthermore, we have addressed or will address all comments from reviewer 1 to clarify the immune-related observations. *

      Reviewer #2 (Significance (Required)):

      Despite severe cholestasis, ALGS patients do not show as much fibrosis as other cholestatic diseases, including biliary atresia (BA). A previous study had suggested that this phenomenon could be due to the difference in the nature of reactive hepatobiliary cells in ALGS compared to BA (Fabris et al, 2007). Moreover, a number of studies have suggested a role for Notch pathway activation in several cell types in the liver in the development of liver fibrosis (for example, Sawitza et al, Hepatology, 2009; Chen et al, Plos One, 2012; Duan et al, Hepatology, 2018; Yu et al, Science Translational Medicine, 2021). However, although a role for Notch signaling in T-cells is well established, it was not known whether impaired T-cell development/function contributes to reduced fibrosis in ALGS liver disease. Accordingly, the current manuscript provides novel insight into the mechanism of fibrosis in this disease. Moreover, the observation that Jag1-mutant T-cells do not confer as much protection as control T-cells to immunodeficient mice subjected to DSS-induced ulcerative colitis provides strong evidence for impaired T-cell immunity in this ALGS model and might help explain other aspects of ALGS phenotypes.

      The manuscript will be of interest to broad audience (Notch signaling, cholestatic liver disease, mechanisms of liver fibrosis, T-cell development).

      I have expertise in Notch signaling and in using animal models of human developmental disorders.

      __Author Response: __We thank the reviewer for the balanced assessment of our manuscript in light of the current knowledge, and for highlighting its importance in the context of not only Notch and ALGS, but also other cholestatic and fibrotic liver diseases.

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

      The article entitled "Jag1 Insufficiency Disrupts Neonatal T Cell Differentiation and Impairs Hepatocyte Maturation, Leading to Altered Liver Fibrosis" by Mašek et al described the role of Notch ligand JAGGED1 (JAG1) in the T-cell differentiation contributing to liver fibrosis and immune system development in ALGS. This article is well written and has important preliminary findings that could establish Jag1 and its downstream signaling pathways as potential therapeutic targets to attenuate liver fibrosis.

      Author Response: We thank the reviewer for recognizing our work and pointing out the therapeutical implications of our findings.

      Reviewer 3 comment 1: Minor comments: In page 4, they mentioned that "the hepatoblast marker alpha fetoprotein (AFP) was 3.1-fold enriched (Fig. 3J,K), while the mature hepatocyte marker CYP1A2 protein was 1.7-fold less expressed (Fig. 3L-M)", the figure numbers should be changed to 2J, K, L-M etc.

      Author Response:* We thank the reviewer for identifying these errors. The suggested corrections have been implemented. *

      Reviewer 3 comment 2: In liver fibrosis the Th17 cells play crucial roles. Please show the level of IL17A mRNA level in the liver in the Jag1Ndr/Ndr mice compared to the Jag1+/+ mice.

      Author Response: We thank the reviewer for the insightful comments. We indeed investigated the Th17 vs Treg immune response, however we detect neither Th17-expressed Il17, Il17a, Il17f, nor Il21 and Il22 mRNA in the bulk RNA data, suggesting their expression is either masked or they are not present in significant numbers within the liver tissue at P10, preventing us from drawing any conclusions about this cell population.

      Reviewer ____3 comment 3: Also, please show the expression level of pro-inflammatory molecules, for example, TNFα, IL1β, MCP1 etc and the level of MMPs (especially MMP2, MMP8, MMP9) in the livers of the mice models used.

      Author Response: *The expression of Il10, Il1b, Mcp1(Ccl2), was presented in the manuscript Fig. 2O, and we attach in the response to reviewers *

      *a full list together with the expression levels of Mmp2/8/9, Tnfa, Ifng, Il17 receptor family and Tgfb1-3. Out of these, Mmp8 (0.9 Log2fold change = 1.9-fold), Ccl2 (2.2 Log2fold change = 4.7-fold), and Tl17rb (1.1 Log2fold change = 2.1-fold) were significantly upregulated, but do not indicate any specific leukocyte population’s response. This is in line with data in Fig S2E, demonstrating a dominance of myeloid over adaptive immune response in the GSEA of the immune KEGGs. *

      *Since lymphocytes are underrepresented in the bulk transcriptomics, and individual genes might report activity of many different cell types, we chose to focus on the list of genes shown to be markers of activated hepatocytes, to avoid over interpretation of the RNA sequencing data. Instead, the immune analyses were based on flow cytometry data, which we expect should accurately report cell type abundance across organ systems. *

      Reviewer 3 comment____ 4. Authors have shown significant alterations in the Treg population in their Jag1Ndr/Ndr mice of ALGS. Please also show the expression of IL10 and TGFβ in the liver and whether they are correlated with the level of Treg populations.

      Author response:* IL10 and Tgfb mRNA levels in liver are shown in the heatmap in the response to reviewers, and were not significantly different between genotypes at P10. They were also not correlated with Foxp3 levels, as shown in the correlation matrices below (Pearson’s R values in top row, significance values in bottom row). *

      Reviewer 3 comment 5. It would be interesting to know whether the IFNγ mRNA expression in the livers were altered in the Jag1Ndr/Ndr mice with altered populations of CD8 T cells.

      Author Response: There was no significant difference in IFNγ mRNA expression levels between Jag1+/+ and Jag1Ndr/Ndr *livers at P10 (please see the heatmap in response to comment no.3, above). *

      Reviewer #3 (Significance (Required)): Strength: This article is well written and has important preliminary findings that could establish Jag1 and its downstream signaling pathways as potential therapeutic targets to attenuate liver fibrosis.

      Author Response: Thank you for these comments and pointing out the wider implications of our findings.


      Reviewer 3____ Limitations: This study lacked the detailed molecular pathways which could explain how the Jag1 altered the T-cell recruitment, development and hepatocyte maturation in the development of liver fibrosis in the ALGS model.

      Author Response: We agree that this study does not focus on molecular pathways. The intention of this study was to identify which cell populations contribute to atypical neonatal fibrosis in ALGS. Because we expected this process to be multifactorial, Jag1Ndr/Ndr mice, carrying a systemic mutation, present both advantages (Jag1 abrogation in all cells --> ALGS-like organ interactions) and limitations (inability to identify contributions of individual cell types). However, by identifying maturing hepatocytes and Tregs as dysregulated, and demonstrating that Jag1Ndr/Ndr lymphocytes behave abnormally and suppress inflammation and fibrosis in Rag1-/- mice (with normal Jag1 expression), we establish a biological framework that can now be further investigated with conditional genetic tools and in vitro systems, to elucidate specific molecular pathways, that were beyond the scope of the current study.

    2. 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 article entitled "Jag1 Insufficiency Disrupts Neonatal T Cell Differentiation and Impairs Hepatocyte Maturation, Leading to Altered Liver Fibrosis" by Mašek et al described the role of Notch ligand JAGGED1 (JAG1) in the T-cell differentiation contributing to liver fibrosis and immune system development in ALGS. This article is well written and has important preliminary findings that could establish Jag1 and its downstream signaling pathways as potential therapeutic targets to attenuate liver fibrosis.

      1. Minor comments: In page 4, they mentioned that "the hepatoblast marker alpha fetoprotein (AFP) was 3.1-fold enriched (Fig. 3J,K), while the mature hepatocyte marker CYP1A2 protein was 1.7-fold less expressed (Fig. 3L-M)", the figure numbers should be changed to 2J, K, L-M etc.
      2. In liver fibrosis the Th17 cells play crucial roles. Please show the level of IL17A mRNA level in the liver in the Jag1Ndr/Ndr mice compared to the Jag1+/+ mice.
      3. Also, please show the expression level of pro-inflammatory molecules, for example, TNFα, IL1β, MCP1 etc and the level of MMPs (especially MMP2, MMP8, MMP9) in the livers of the mice models used.
      4. Authors have shown significant alterations in the Treg population in their Jag1Ndr/Ndr mice of ALGS. Please also show the expression of IL10 and TGFβ in the liver and whether they are correlated with the level of Treg populations.
      5. It would be interesting to know whether the IFNγ mRNA expression in the livers were altered in the Jag1Ndr/Ndr mice with altered populations of CD8 T cells.

      Significance

      Strength: This article is well written and has important preliminary findings that could establish Jag1 and its downstream signaling pathways as potential therapeutic targets to attenuate liver fibrosis.

      Limitations: This study lacked the detailed molecular pathways which could explain how the Jag1 altered the T-cell recruitment, development and hepatocyte maturation in the development of liver fibrosis in the ALGS model.

    3. 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:

      Masek and colleagues use multi-pronged studies on the Jag1[Ndr/Ndr] mouse model of Alagille syndrome (ALGS) combined with transcriptomic analysis on livers from patients with ALGS to elucidate the potential mechanisms regulating liver fibrosis in this disease. The authors first show that Jag1[Ndr/Ndr] animals develop pericellular and perisinusoidal fibrosis and exhibit evidence for portal hypertension, similar to patients with ALGS. Single-cell RNA-sequencing indicated more hepatoblasts and less hepatocytes, relatively speaking, in Jag1[Ndr/Ndr] P3 livers, which suggested hampering of hepatoblast differentiation to hepatocytes. Deconvolution of previously generated bulk RNA-seq data from Jag1[Ndr/Ndr] P10 livers and GESA on RNAseq data from livers of these mice and patients with ALGS confirmed the P3 scRNA-seq observations and indicated mild pro-inflammatory activation of immature hepatocytes in ALGS livers. GESA also suggested an inability of Jag1[Ndr/Ndr] livers to attract T cells upon cholestatic injury. Indeed, 25-color flow cytometry on liver and spleen from mutant and control mice indicated a defect in T cell response to cholestasis in this model. The authors then examined the effects of the Ndr mutation on T-cell development and function. They found that the Ndr/Ndr thymi were significantly smaller than control thymi. Moreover, Ndr/Ndr thymi showed an increase in CD4+ T-cells and Tregs at the expense of double-positive T-cells. The authors then performed lymphocyte transplantation studies and concluded that Ndr/Ndr T-cells fail to mount an adequate response to inflammation in a DSS model of ulcerative colitis. The authors tested the contribution of Ndr/Ndr immune cells to liver fibrosis in a model of experimentally induced cholestasis (bile duct ligation; BDL). Ndr/Ndr T-cells did not show any defects in migrating into the liver upon BDL. However, the periportal fibrosis observed in BDL model was reduced in animals receiving Ndr/Ndr immune cells compared to those receiving Jag1+/+ immune cells. This was accompanied by significantly less aSMA staining in these livers. Finally, reanalysis of bulk RNAseq data from liver samples from ALGS and other liver diseases suggested that the presence of FOXP3+ T-reg cells in the liver is associated with higher liver fibrosis in non-ALGS liver diseases but lower liver fibrosis in ALGS livers. The authors have used an impressive combination of single-cell RNA-sequencing, reanalysis of previous bulk RNA-sequencing data from their group and others, 25-color FACS analysis, and adoptive immune transfer experiments in this manuscript, and systematically provide quantification and statistical analysis for their data. Overall, this is an interesting and important study. Prior studies are referenced appropriately. The text and figures are clear and accurate. I don't think any additional experiments are essential. However, the issues listed under Major comments should be discussed and clarified in the manuscript, especially the first item.

      Major comments:

      • Only a small fraction of the cells in scRNA-seq experiments have been assigned to hepatocytes/hepatoblast clusters, with the majority of these cells allocated to Hepato-Ery cluster. This suggests that many hepatocytes and potentially hepatoblasts have been lost during sample preparation. The authors should discuss this issue and its potential implications on the interpretation of the cell ratios and gene expression conclusions of scRNA-seq data.
      • The Jag1[Ndr/Ndr] strain is an excellent model for various aspects of ALGS phenotypes. However, when it comes to linking the effects of this mutation to the function of a specific cell type, it is worth considering that Jag1[Ndr/Ndr] might not recapitulate the effects of loss of one copy of JAG1 observed in most patients with ALGS. This is especially important given the sensitivity of various cellular and organ-level processes to the degree of Notch pathway activation. In the context of the present manuscript, it is possible that what the authors have observed in Jag1[Ndr/Ndr] lymphocytes does not mirror how a JAG1-heterozygous human lymphocyte behaves. This is not a major concern, but it is worth considering.
      • The basis for the opposite type of correlation between COL1A1 expression and POXP3 level in ALGS versus non-ALGS liver disease is not clear.

      Minor comments:

      • Page 2, last paragraph of Introduction, Page 12 last sentence, and Supplementary Methods: Please use "adoptive immune transfer" instead of "adaptive immune transfer".
      • Pages 3 and 4: Reference is made to Figures 3E-O, which appears to be Figure 2E-O.
      • Figure 3 legend: "Analysis in (E) is one-way ANOVA with Dunnett's multiple comparison test". Panel E compares two means, so ANOVA is not the appropriate statistical analysis for these data. Is this sentence related to panel D?
      • Page 9: Please correct misspelling: "response to intestinal insult (Fig. 5). W therefore".
      • The Science Translation Medicine references lack page number.

      Referees cross-commenting

      To my knowledge, ALGS is not considered to be an inflammatory disorder. Furthermore, the splenomagaly observed in the mouse model could be due to portal hypertension rather than a primary immune disturbance. Having said that, I agree with the other reviewers that the manuscript will benefit from further discussion and clarification on the immune-related observations.

      Significance

      Despite severe cholestasis, ALGS patients do not show as much fibrosis as other cholestatic diseases, including biliary atresia (BA). A previous study had suggested that this phenomenon could be due to the difference in the nature of reactive hepatobiliary cells in ALGS compared to BA (Fabris et al, 2007). Moreover, a number of studies have suggested a role for Notch pathway activation in several cell types in the liver in the development of liver fibrosis (for example, Sawitza et al, Hepatology, 2009; Chen et al, Plos One, 2012; Duan et al, Hepatology, 2018; Yu et al, Science Translational Medicine, 2021). However, although a role for Notch signaling in T-cells is well established, it was not known whether impaired T-cell development/function contributes to reduced fibrosis in ALGS liver disease. Accordingly, the current manuscript provides novel insight into the mechanism of fibrosis in this disease. Moreover, the observation that Jag1-mutant T-cells do not confer as much protection as control T-cells to immunodeficient mice subjected to DSS-induced ulcerative colitis provides strong evidence for impaired T-cell immunity in this ALGS model and might help explain other aspects of ALGS phenotypes.

      The manuscript will be of interest to broad audience (Notch signaling, cholestatic liver disease, mechanisms of liver fibrosis, T-cell development).

      I have expertise in Notch signaling and in using animal models of human developmental disorders.

    4. 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 is an interesting study that examines defects in the Jag1ndr/ndr mouse model of Alagille syndrome. The novel aspects of this manuscript are the comparisons, at many levels, between the mouse model and ALG patient samples, including an examination of immune profiles. The conclusions that the Jag1ndr/ndr mouse model is an accurate representation of the human ALG syndrome appear valid. However the reported differences in immune profiles, particularly in the Jag1ndr/ndr mouse model are difficult to understand. The data presented indicate a reduction in CD4+ cells in the Jag1ndr/ndr mouse at day P3 in both liver and spleen. Additionally, the authors report differences between the the Jag1ndr/ndr mouse and controls at day P30 in the relative percentages of DN, DP and SP CD4 and CD8 cells in the thymus. When examining the peripheral lymphoid system, CD4+ numbers are the same in both the Jag1ndr/ndr animals and controls however CD8+ numbers are reduced and FoxP3/CD4+ cells are increased in both the spleen and the thymus. FoxP3/CD4+ T cells are usually assumed to be regulatory T cells that dampen the inflammatory responses of T cells. Therefore, the increase in this population in an animal model of what is assumed to be an inflammatory disease is confusing and confounding. The authors do not present a clear analysis of how they feel an increase of Tregs would lead to this disease. One possibility is that this population is not functioning as conventional Tregs and rather are promoting inflammation but this conclusion would require a functional analysis of this population of cells, at the very least in an in vitro analysis of T cell suppression. From an immunologist's point of view, their data are antithetical to what one would expect to find in an inflammatory disease. Perhaps this reviewer is missing an important point but if I am missing it, then other who read this manuscript also may be confused.

      Minor points that should be addressed include:

      • The source cells used in the transfer experiments reported in Figure 5 is unclear. Are they using total spleen cells with T, B and myeloid cells or are they using purified T cells. And if it is the latter, have they assessed the ratio of CD4+ versus FoxP3/CD4+ cells in the transferred cells?
      • In the DSS experiments in Figure 5, there does not appear to be a no DSS control. What does the architecture look like without DSS?
      • Again, in Figure 5, were FoxP3/CD4+ cells enumerated?
      • The authors noted that splenomegaly was observed in the Jag1ndr/ndr mouse model. Again this is antithetical to what one would expect when one sees an increase in FoxP3/CD4+ T regs.

      Significance

      The strengths of this paper are the careful comparisons between the mouse model and the human ALG syndrome. These comparisons are valuable and worth publication.

      Weaknesses are stated above. Needs a clearer explanation for their immune analysis.

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

      Learn more at Review Commons


      Reply to the reviewers

      Reviewer #1

      Evidence, reproducibility and clarity (Required):

      Au et al. used two fly models to study how mitochondrial defects are implicated in C9ALS, the most common familial ALS type. They found that in these flies, mitochondrial, but not cytosolic, ROS is upregulated, accompanied by locomotion defects agreeing with previous publications. Consistent with these data, sod2, but not sod1, rescues the behavioral defects in these flies. Also, manipulating mitochondrial dynamics or mitophagy does not rescue these defects. Furthermore, the authors showed that the Nrf2 activity is upregulated, likely due to oxidative stress, and genetically or pharmacologically suppressing the Keap1 function, which activates Nrf2 and thereby its downstream antioxidative genes, suppresses behavior defects in these flies. This part is generally solid and convincing, with minor issues that need some revision. Finally, the authors showed that mitochondrial ROS and nuclear Nrf2 are both upregulated in C9 iPS neurons, both of which are suppressed by the Keap1 inhibitor DMF, or a known antioxidant. For this part, the data are convincing but insufficient to support a good translation of their fly data.

      __Major concerns: __

      1a. The authors really need a phenotypic readout for their iPS experiments, either cell death or some sort of toxicity, to support the translatability of their fly data.

      • We agree and appreciate the value of having such as phenotypic readout for the iPSC experiments but, unfortunately, within the context of the current work we did not obvious any clear phenotype of toxicity or diminished viability under basal, unchallenged conditions. To support this, we have added our analysis of cell viability at the time of imaging, shown in new Supplementary Figure 3C and mentioned in the text (line 620-621).

      1b. The authors also need to test the toxicity of DMF in iPS neurons.

      • As above, we found that treatment with DMF conferred no overt toxicity within the time-course of our experiments. These data are shown in new Supplementary Figure 3D and mentioned in the text (line 626-628).

      The authors should use genetic ways, e.g., knocking down Keap1, to activate Nrf2 and test whether this suppresses ROS and neurodegeneration phenotype in iPS neurons, as they did in flies.

      They need to better characterize the Nrf2 activity in iPS neurons (see Minor Concern #1).

      • Regarding these two points, we agree that it would be interesting to further investigate the Keap1/Nrf2 pathway in these cells, but time, personnel and resource constraints preclude additional investigations on this occasion. It is important to note that the cell models were used specifically to validate that elevated mitochondrial oxidative stress and increased nuclear Nrf2 localisation also occurred in patient-derived neurons, and whether DMF treatment could reverse the oxidative stress. This was the extent to which the cell models were used in this instance and the current data are sufficient to support the conclusions made based on this. We regret that it was not possible to delve deeper into this at the current time but will be possible in future work.

      __Minor concerns: __

      1a. Fig 4A and B are hard to comprehend. Can the authors show images with more obvious differences?

      • We have now revised these figure panels replacing with alternative images. We hope that the new images show more appreciable differences. We understand that the differences can sometimes be subtle which is why we rely on the quantification for unbiased interpretation.

      1b. Also, Gst-D1 is the only Nrf2 downstream gene tested. Can the authors use RT-PCR to test multiple genes? These will strengthen the point that Nrf2 is activated. Similar things should be done in iPS neurons.

      • Thanks for this suggestion. To complement the immunoblots of the genomic GstD1-GFP reporter, we have now performed qRT-PCR on flies treated with or without DMF f