The Interaction of Factor Xa and IXa with Non-Activated Antithrombin in Michaelis Complex: Insights from Enhanced-Sampling Molecular Dynamics Simulations
Round 1
Reviewer 1 Report
Balogh and Bereczky report in their manuscript on enhanced molecular dynamics simulations of coagulation factors XA and IXA, respectively, and their interaction with the inhibitor antithrombin with and without a pentasaccharide bound. The authors find a large conformational variety for both factors. Based on the analysis of several important distances and positional fluctuations, the authors claim to have provided an " atomic level model for understanding the interactions in the AT-FXa and AT- FIXa complexes in different activation states of AT".
While this is true to some extent, there is a problem with drawing such conclusions from the present simulation data: There are significant differences between separate simulation runs, which leads one to think the simulations are far from being equilibrated, or simply too short, even with accelerated MD. With differences between simulation runs of the same setup ("group of simulations") being as large as those between different groups, how useful are the observations made? How can the authors draw conclusion about the effect of e.g. ligand binding, or the starting structure (Xray or docked) when e.g. in one two of four simulations important contacts exist but in the other not, and similarly so in the system that should be compared to? Many of the distances analyses seem to have a rather random behaviour (acros the different simulation runs) rather than showing a trend. The correlation matrices are not any better in that regard, i.e. at places they differ more between simuation runs of the syme setup than between different models.
The system appears to be one that is difficult to capture by MD simulations, haveing arge conformational freedom and thus ssampling different conformations. It may be worth investing in more computer time and generating further data (significantly longer simulations, also in three or four replicas like done in the present manuscript). If that does not help, or this i not feasible, then the authors should limit the discussion, and even more so the conclusion, to quantities which have an acceptable error (read variance between several simuations of the same system).
To this end, a different representation of some of the data is desired: While time series are often a valuable representation to show transition events, comparison of the different groups of simulations can be facilitated by also providing plots of e.g. distance distributions. In principle, these could be averaged over the different simulation runs of the same system. With the present simulation data, which show large dfferences between such simulations of the same system, there will be large errors on many such averages. But these errors show which quantities cannot do not, and which do, have siginificant (read: larger than the errors) differences between the different systems that are worth being discussed.
There are some details missing in the methods description:
What boost potential was used in the GAMD simualtions?
What distance metric has been used in the clusters analysis? Have all simulation data been included simultaneously? That is, cluster 1 of one "group of simulations" is similar to cluster 1 of another group? If not, how different are the clusters from different groups of simulations?
On which ground was 6 Angstrom chosen as a threshold for a contact?
An obsolete question, if the authors choose to prolong their simulations, but nevertheless: why were the simulation lengths chosen differently, i.e 400, 300, and 200ns?
Simulations without ligand are labelled "no ligand", but it may be useful to also label wimulations with ligand by "with ligand" or something similar.
An overwiew figure showing where the ligand is located in the complex, e.g. from the crystal structures, would be helpful. In the same figure the authors should also clearly show which part is one of the coagulation factors and which is the antithrombin.
There are some sentences which are a bit more difficult to nderstand because they have a word too many, slightly wrong or similar. It is nothin, careful proof reading cannot cure, though.
Author Response
Response to Reviewer 1
First of all, we would like to thank the reviewer for the valuable comments and suggestions.
We agree with the reviewer that quantitative or semi-quantitative comparisons between the systems would require parameters with lower variance between the simulations. We also agree that significantly longer simulations would provide an even more detailed picture of the conformational behavior. However, our current simulations suggest a very large conformational flexibility of the systems. Thus, the differences between individual runs based on a single model systems could be still significant even when the trajectories are much longer.
The main reason for simulating both docking-based and X-ray diffraction based systems was to explore the conformational behavior of the systems from different starting conformations. The simulations could be considered as models for different steps of the processes. For example, we have discussed in the manuscript that we interpret the data from the X-ray diffraction based simulations without a ligand as possible early steps of a conformational change towards the non activated state. So we still think that even “divergent” trajectories for these systems can provide a better picture on the systems than the currently available X-ray data.
Regarding the differences between the two factors, the current data is clearly not sufficient to compare the stability, kinetics, etc. quantitatively. However we think that our simulations provide a better qualitative picture for the behavior of the systems than the currently available X-ray structures (showing only one conformation for both complexes) or mutagenesis data. The information provided in the manuscript, such as conformations not reported before, could facilitate the design of experiments such as mutagenesis that could help to understand these systems even better.
As for the pentasaccharide complex systems, we agree with the Reviewer that conformational changes not occurring in all simulations is a limitation of the study. We have modified the discussion accordingly. We have also modified the conclusions according to the comments of the reviewer and added further comments on the limitations of the current methodology on the comparison of the systems.
Hereby we give a point-to-point answer for your questions.
Q1. What boost potential was used in the GAMD simualtions?
The average GaMD boost potential values in the production simulations were between 13.8 and 15.6 kcal/mol. We have added this information to the Methods section.
Q2. What distance metric has been used in the clusters analysis? Have all simulation data been included simultaneously? That is, cluster 1 of one "group of simulations" is similar to cluster 1 of another group? If not, how different are the clusters from different groups of simulations?
We have added discussion about the settings used in the cluster analysis in the Methods section. The clusters were numbered by the algorithm according to the number of structures in each, so the same cluster number in two groups of simulations (e.g. Cluster 1) does not imply similar conformations.
Q3. On which ground was 6 Angstrom chosen as a threshold for a contact?
We have chosen 6 Å as a threshold for a contact based on the observation that dissociation of aminoacids from the surface of the interaction partner protein tended to occur at that distance.
Q4. An obsolete question, if the authors choose to prolong their simulations, but nevertheless: why were the simulation lengths chosen differently, i.e 400, 300, and 200ns?
The largest conformational changes were observed in the docking-based simulations. As these systems required the most extensive conformational sampling, we simulated these systems for a longer period of time. On the other hand, the conformation changes in the X-ray diffraction based systems were more limited, and in some cases, mainly in the pentasaccharide containing systems, “convergent” behavior was observed after some simulation time. Therefore we have chosen a shorter length for these systems due to the significant computational cost of these simulations.
Q5. Simulations without ligand are labelled "no ligand", but it may be useful to also label wimulations with ligand by "with ligand" or something similar.
We have modified the captions on figures 2 and 5 according to the request of the Reviewer.
Q6. An overwiew figure showing where the ligand is located in the complex, e.g. from the crystal structures, would be helpful. In the same figure the authors should also clearly show which part is one of the coagulation factors and which is the antithrombin.
Figure 1 now provides an overview of the entire complex for the docked structures and the corresponding X-ray structure. We labelled antithrombin and the two factors on the figure. The pentasaccharide ligand in the X-ray structures is shown in a color different from the proteins.
Reviewer 2 Report
Thank you for submitting your manuscript entitled " The interaction of factor Xa and IXa with non-activated antithrombin in Michaelis complex, insights from enhanced sampling molecular dynamics simulations" . The authors studied the conformational behavior of non-activated antithrombin (AT) when not binding a pentasaccharide, by studying the interaction between coagulation factors Xa and IXa and the activated state of AT. They employ HADDOCK 2.4 for the initial structure of non-activated AT-FXa and AT-FIXa complexes, and Gaussian Accelerated Molecular Dynamics simulations to study the conformations of the complexes. Comparing the simulations with and without the pentasaccharide, the authors gain insights into conformational activation effects on the Michaelis complexes. I would like to request the following minor revisions to further improve the manuscript: Once these revisions are made, this could be published.
Comments
In the section : Material and Methods Line 114 , Authors mentioned they reduced the size of simulated systems by omitting domains (GLA, EGF1) and amino acids .
While this reduction in system size may have allowed for more manageable simulations, it is important to consider the potential impact of these exclusions on the accuracy and validity of the results. The missing domains and amino acids may play a crucial role in the interactions and behavior of the molecules under investigation. Consequently, their absence in the simulations could lead to an incomplete or distorted understanding of the molecular mechanisms at play.
Please Justify how this won’t effect the outcome of the results ?
Correct to X-tay to X-Ray on line 259
In figure the fonts are not legible enough to read , Also I suggest to improve the font size to make it clear for the readers .
Why authors haven’t carried out simulations for more than 300 ns for X-ray structure based ? like docking based model ?
Author Response
Response to Reviewer 2.
We thank the reviewer for the positive comments.
Hereby we are providing answers to the questions.
Q1. In the section : Material and Methods Line 114 , Authors mentioned they reduced the size of simulated systems by omitting domains (GLA, EGF1) and amino acids .
While this reduction in system size may have allowed for more manageable simulations, it is important to consider the potential impact of these exclusions on the accuracy and validity of the results. The missing domains and amino acids may play a crucial role in the interactions and behavior of the molecules under investigation. Consequently, their absence in the simulations could lead to an incomplete or distorted understanding of the molecular mechanisms at play.
Please Justify how this won’t effect the outcome of the results ?
Answer 1. The region removed from our model systems contained the GLA and the EGF1 domains. From our simulations it was clear that all direct amino acid interactions in the complexes were between the catalytic domain of the two coagulation factors and antithrombin. The EGF2 domain, which was not omitted from the models, was too far from the involved regions for such interactions in all conformations observed.
Based on X-ray diffraction structures of proteins containing the same domains in the same order (porcine factor IX, RCSB PDB: 1PFX, human factor VII, RCSB PDB: 1W0Y), we expect that the missing domains would be located at an even higher distance from antithrombin than the EGF2 domain. Direct interaction of amino acids in these domains with residues in antithrombin is therefore unlikely. We cannot rule out a role of long-range electrostatic interactions on the conformations of the complexes but we expect that these effects would be small due to the large distance.
Q2. Correct to X-tay to X-Ray on line 259
Answer 2. We have corrected the spelling error mentioned by the reviewer, as well as a few others.
Q3. In figure the fonts are not legible enough to read , Also I suggest to improve the font size to make it clear for the readers .
Answer 3. We have replaced Figures 2 and 3 to increase the font size on the axes of the plots.
Q4. Why authors haven’t carried out simulations for more than 300 ns for X-ray structure based ? like docking based model ?
Answer 4. In the X-ray diffraction based systems, the conformational changes were significantly smaller compared to the simulations based on the docked structure. We do not expect that expansion of these simulations to 400 ns would change the results significantly. The docking based simulations, on the other hand, had a very high conformational variability, so longer simulations for these systems were necessary.
We thank the reviewer for the valuable comments and we hope that our manuscript is now acceptable for publication.
Reviewer 3 Report
The manuscript "The interaction of factor Xa and IXa with non-activated antithrombin in Michaelis complex, insights from enhanced sampling molecular dynamics simulations" addresses an important issue: the study of the interaction between coagulation factors Xa and IXa and the activated state of their inhibitor. The authors propose a model based on docking and advanced sampling molecular dynamics simulations that can reveal the conformational behavior of the systems when AT is not binding a pentasaccharide.
The objectives were clearly stated and explained in the manuscript, however the experimental strategy raises some major concerns and so the experimental information from which the conclusions were drawn. The manuscript is overall well written and has good organization with minor English language and style spell check required. The authors have done a great job on analyzing the experimental data and on discussing the results and their limitations, considering always different alternative explanations/considerations for interpreting the results.
The paper is interesting but there is a need for more experimental detail in order to critically review the data. Specifically, they should provide information for the following questions and comments:
Major points:
1. The authors should include more recent update on this topic and compare how this study further advances the current knowledge in the “Introduction section”.
2. What is the advantage of the techniques used compared to other techniques currently used in molecular modeling?
3. Unify the style of the references in the References Section and add DOI in the cases it is possible. And use the same reference and citation (follow MDPI’s guidelines) style in the main text.
Minor points:
1. Captions in several Figures are scarce, a more detailed description is needed specially for Figure 5.
2. The resolution and quality of some Figures is low, the authors should provide higher quality Figures specially for Figure 3 and 6.
I would like to comment on the quality of English language used in the manuscript. While the content of the manuscript appears to be well-researched and informative, the English language used in the manuscript needs some improvement.
I recommend that the authors work with a professional editor or native English speaker to improve the language used in the manuscript. This will help to ensure that the manuscript is clear, concise, and easy to understand for readers.
I believe that improving the quality of English language used in the manuscript will significantly enhance the readability and impact of the research presented.
Author Response
Response to Reviewer 3
We would like to thank the reviewer for the comments, which help to improve our manuscript. Specifically, we give a point-to-point answer for the questions and issues.
Q1. The authors should include more recent update on this topic and compare how this study further advances the current knowledge in the “Introduction section”.
Answer 1. We have added additional discussion into the Introduction section about the significance of our paper in the field. We have added one relevant new reference to the Introduction, with a short discussion of its importance.
Q2. What is the advantage of the techniques used compared to other techniques currently used in molecular modeling?
Answer 2. Gaussian Accelerated Molecular Dynamics allows significantly enhanced conformational sampling without requiring pre-defined reaction coordinates. This is particularly advantageous for systems with a complex conformational space such as protein-protein complexes. On the other hand, their computational costs are significantly smaller than other techniques such as Replica Exchange Molecular Dynamics. This is discussed in the Methods section of the manuscript.
Q3. Unify the style of the references in the References Section and add DOI in the cases it is possible. And use the same reference and citation (follow MDPI’s guidelines) style in the main text.
Answer 3. We have added the missing DOIs for all references except for one reference (reference 35, AMBER 20 software) which, to the best of our knowledge, has no DOI. We have unified the style of the references. In case of electronic journals no page number is provided, because of this the reference list cannot be totally uniform.
Minor points:
Q1. Captions in several Figures are scarce, a more detailed description is needed specially for Figure 5
Answer 1. The captions of figures 3, 4, 5 and 7 were updated to provide a more detailed description.
Q2. The resolution and quality of some Figures is low, the authors should provide higher quality Figures specially for Figure 3 and 6.
Answer 2. We have replaced figures 2, 3 and 6 to make the numbers on the axes easier to read. The new version of Figures 3 and 6 now have a better resolution than in the original submission.
We thank again for the reviewer for these helpful comments and we hope that the modified manuscript is now acceptable for publication.
Round 2
Reviewer 1 Report
The authors have addressed all concerns raised by this reviewer and the manuscript is now acceptable for publication.
only spell checking
