Evaluation of Potential Furin Protease Inhibitory Properties of Pioglitazone, Rosiglitazone, and Pirfenidone: An In Silico Docking and Molecular Dynamics Simulation Approach

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear authors,

The manuscript needs a deep evaluation of the method, results, and discussion, mainly of molecular dynamics. Below are some questions and suggestions for improving the text.

 

Abstract:

The authors cited 7QXZ, but other PDBs were used in the work. I suggest putting this information more generally.

The authors cite 100 and 50 ns simulations for FurAct and FurAll and standardize the simulations to at least 100 ns.

Introduction:

The authors could cite the state of the art of computational studies for Furin, emphasizing the study of docking and molecular dynamics.

Method:

Authors must indicate the resolution of the complexes used, which compounds are co-crystallized, and cite the articles referring to obtaining the crystallographic structure.

The authors must clarify the 3D structures of the compounds NPF, PFD, PGZ, and RGZ. Which 3-(3,5-Dichlorophenyl) pyridine derivatives were included in the study?

The authors should have carried out protonation studies at the pH in which the enzyme is located, in addition to possible tautomers.

Will the molecular formula data be used for any further discussion? If not, delete it from the table.

Insert the structures of the 3-(3,5-Dichlorophenyl) pyridine derivatives in table 1.

Topic 2.2 needs to be clarified. It generally begins by mentioning the programs that will be used. How did the authors insert the AM1-BCC payloads into the pdbqt files generated by AutodockTools?

What charges were added to the atoms of the enzyme? Were the hydrogens considered only polar or polar and non-polar hydrogens considered?

What was the pH at which the study was carried out? Conducting some protonation study of enzyme residues using servers such as H++ or propka is essential.

The authors should provide more details about the methodology.

As reported, Item 2.3 is not reproducible. In this topic, authors should provide information on how to perform the docking study, placing justifications and discussions appropriately.

Authors should use the PDB codes of the ligands instead of the class to which they belong.

Authors must start the topic by indicating which programs were used for the docking study. In lines 146-147, Autodock-Vina and Autodock4 are indicated. These programs have notable differences, such as the spacing of the grid box, making it impossible to use the Vina box in Autodock4.

Authors must indicate how they calculated the RMSD in the docking validation (redocking) stage. The RMSD value provided by Vina is not reliable since it uses the lowest energy pose instead of the PDB pose as a reference.

The allosteric site must be identified using appropriate software, such as FTMAP and DocksiteScorer, among others, preferably through a consensual analysis.

The algorithms implemented in Vina and autodock4 are not specialized for blind docking, so specific software is necessary for this approach.

In lines 152-153, covalent bonding is mentioned. Do any of these inhibitors co-crystallize binding covalently with Furin? If so, breaking this bond and rebuilding the portion that binds to the enzyme is necessary.

Reference 62, cited in lines 155-156, does not mention the distance cited by the authors. Authors must specify which interactions were considered and the distances and angles relating to each interaction.

To justify using only the 7QXZ, the authors must provide a detailed assessment of the difference between 7QXZ, 7LCU, 5MIM, and 7O1Y. Why did the authors not implement a cross-docking study?

In topic 2.4, authors need to review the information provided for molecular dynamics. Furthermore, the order in which some information is cited must follow the sequence in which it is carried out.

- As the ligand topologies were prepared, cgenff is generally used.

- The ionic concentration is considered in the simulation. Calcium is not expected to be added. Provide references that support this.

- algorithm used in the minimization step

- execution time of the NVT and NPT balancing steps. The information on the barostat and thermostat used must be close to this information. Did this stage have position restrictions?

Authors must provide a reference to the CARMA software.

  Line 179-184—Authors need to provide details of how the analyses were carried out. Were the RMSD, Rg, and RMSF performed for the entire trajectory? Which atoms were used as references? Authors should move explanations and justifications to the appropriate topics.

Authors must provide scripts created in Rstudio.

Line 191-198 – Authors must provide references that support the highlighted regions and illustrative figures to assist in understanding the text. Are the ligands co-crystallized in these two regions?

 

Authors must provide RMSD values and overlay figures of the co-crystallized pose with the pose obtained by docking using VINA and autodock4.

How did the authors calculate the energy values for the co-crystallized ligands before performing the redocking study?

How did the authors validate the pose obtained for FurAll?

Authors must standardize all simulations to at least 100ns.

It is not clear from the text how the authors ensure that the ligands remain in the active site throughout the MD simulation. Give the distance between an atom of an important residue for the enzyme and a specific atom for the ligand.

Authors must ensure that the frame presented in the article is representative of the simulation. For this, clustering or PCA analyses are important to determine the predominant conformation.

Line 346-347—The authors must explain why only 10% of the simulation is used. Is this 10% concentrated in a specific time range? The authors should apply data reduction techniques such as clustering or PCA for the analysis.

Authors must make it clear whether the RMSF variation presented is due to the presence of the ligand or whether it is an intrinsic characteristic of the enzyme. To do this, the difference between Fur apo and Fur-ligand is applied.

The authors must analyze intermolecular interactions during the molecular dynamics simulation and evaluate their lifetime.

Authors must perform energy calculations, such as PBSA or FEP, to evaluate the influence of the enzyme's solvation and mobility toward ligands.

 

Comments on the Quality of English Language

The authors should provide a minor review of English spelling.

Author Response

The authors appreciate the reviewer's comments. All the responses are in the document attached.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Editor and Authors,

The manuscript "Evaluation of potential furin protease inhibitory properties of Pioglitazone, Rosiglitazone, and Pirfenidone: an in silico docking and molecular dynamics simulation approach" needs to be revised to be published in the Journal CIMB.
Below are some suggestions that could be good for future readers, and can improve the article quality.

1- line 25 and others. Please revise and improve the English grammar and correct some typos. Example: "The bests affinities of ligand/FurAct..."; "...co-cristalizated lignads" (Figure S6); "200ps" numbers and units must be separated; "Kcal/mol" k is always in lower case; "Zn+2/Cu+2".

2- line 89.  in silico and in vitro is written in italic. Please revise all text.

3- line 117. Why did the authors use four 3D structures of Furin? Please explain it in the text too.

4- line 131. Why did the authors do the energy minimization? What are the problems of this step?

5- line 132. It is not clear if the authors optimized the geometry of the ligands. Please, clarify it.

6- lines 130-132. Please explain why was used atomic charges AM1-BCC for the receptors and Gasteiger atomic charges for the ligands.

7- line 143. Table S1 was not found. Please provide it.

8- line 146. Why did the authors use Autodock-Vina and Autodock4 softwares? Explain it in the main text too.

9- line 135. Why the authors did not make the cross-docking? In addition, provide the RMSD for the redocking.

10- line 152. Please, comment about the ligand clusters. Did the authors analyze the most populated clusters? What is the best, the lowest energies or populated clusters?

11- line 135 and 157. Did the authors do a protocol validation for the molecular docking and dynamics? Please comment about it.

12- lines 229, 233 and 237. Please, improve the figure's quality. Some residues' names are not possible to read, especially in the 2D structures.

13- line 240. Table 2. What means "NA"?

14- line 240. Table 2. Please explain why the energies from Autodock-Vina and Autodock4 are different for the same ligand. In addition, is the conformation the same? Provide the RMSD and comment about it.

15-  line 252. Please highlight the source of the inhibition constant (Ki). In addition, make a discussion and comparison of these Ki values with standard drugs.

Comments on the Quality of English Language

A revision of the text is needed.

Author Response

The authors appreciate the reviewer's comments. All the responses are in the document attached.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This study explores the potential of Pioglitazone (PGZ) and Rosiglitazone (RGZ), drugs for type-2 diabetes, as inhibitors of the protease Furin (Fur), a key enzyme in protein maturation and a therapeutic target in cancer, autoimmune diseases, and viral infections. Using molecular docking and dynamics simulations, the researchers assessed the interactions of these drugs with Furin's active and allosteric sites, finding that PGZ, RGZ, and Naphthofluorescein (NPF) had the highest affinities. Key residues involved were Asp154 for the active site and Glu488 and Asn310 for the allosteric site. The findings suggest these drugs could be repurposed to modulate Furin activity in treating emerging diseases. 

Despite it is very interesting, I found the research a little bit confused in the presentation. I think that Authors should add a workflow, explaining which kind of structure are using or divide the manuscript according to the analysis are presenting. I think that MD simulations are short, since also RMSD of chemicals in the allosteric site are not equilibrated (PFD and PGZ).

Why the Authors stressed in the conclusion the Wars-Cov-2 topic, since in the manuscript seems not so pivotal?

There are some typos in the manuscript (e.g. Kcal/mol -> kcal/mo; in vitro and in silico should be in italic)

Author Response

The authors appreciate the reviewer's comments. All the responses are in the document attached.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

Carranza-Aranda et al. present an in-silico drug discovery study investigating potential small molecule inhibitors for the therapeutic target Furin. The authors utilized molecular docking and molecular dynamics simulations to evaluate the binding site preference and affinity of two known Furin inhibitors, NPF and PFD, as well as two additional compounds, PGZ and RGZ, which may be repurposed as Furin inhibitors. Their results suggest that all four compounds can bind to Furin, either at the catalytic active site or an allosteric site near the interface of the catalytic domain and P domain. The predicted binding site preferences and poses could inform future experimental validation studies and efforts in Furin inhibitor design and optimization.

This manuscript aligns well with the focus of CIMB and the Special Issue on “Synthesis and Theoretical Study of Bioactive Molecules.” The majority of its content is scientifically sound and innovative. However, some conclusions regarding binding affinity, stability, and inhibition mechanisms require further support through additional data and analysis. Therefore, I recommend reconsidering this manuscript after addressing these major concerns, along with some minor issues. Detailed major and minor concerns are provided below.

Major Concerns:

1.    Binding affinity estimation: In Section 3.1, the authors calculated inhibition constant (Ki) based on the molecular docking scores using Autodock 4. It is widely acknowledged that molecular docking scores provide rough estimations of the binding free energy (BFE) due to their limited precision, as they lack protein dynamics and explicit solvent model. Therefore, docking scores are typically used for large-scale virtual screening and ranking compounds rather than for quantitatively predicting binding affinity. The Ki values that the authors calculated from docking scores are significantly different from experimental enzymatic or cellular IC50 values (based on previous studies on 3DPP-D, NPF, PFD) by several orders of magnitude. Instead, MM-GBSA is a more accurate method for estimating protein-ligand BFE, at least for ranking different compounds. Given that the authors have already performed MD simulations, it would be beneficial to calculate MM-GBSA BFE using the existing MD simulation trajectories. This analysis will provide a more accurate estimation of binding site preference for each compound.

2.    Local stability of the binding site: In Section 3.2, the authors assessed the stability of the protein-ligand complex by computing some global properties of the entire protein from MD simulations. These metrics, such as Rg of the entire protein, are relatively insensitive to the local movement of the ligand in the binding site. Additional analyses, such as RMSD of ligand and nearby binding residues, and the residence time/probability of residues in contact with the ligand, will be more helpful to demonstrate the local binding behaviors.

3.    Allosteric inhibition mechanism: The authors proposed that some compounds prefer to bind at an allosteric site near the interface of two domains of Furin. This, however, cannot lead to a conclusion that they are “allosteric inhibitors”, as there is no data to demonstrate that they inhibit the function of Furin while binding at this allosteric site. The authors suggested that “non-competitive inhibitor through destabilization of the interdomain region” (on page 18 lines 503-505), but did not provide any data or analysis to support this assumption. More insights can be gained from further analysis of the MD simulation – upon binding with these compounds, does the interdomain region become more flexible compared with the apo?

Minor Comments:

·   Figure 2: The current figure cannot distinguish whether these compounds all bind at the same allosteric site. Provide a superposed figure with all four compounds binding with Furin (of panels A, C, E, G). The pi-pi stacking and hydrophobic contacts are also crucial for ligand stable binding and should be shown in panels B, D, F, H.

·   Table 2: Explain why Autodock 4 failed to calculate the docking score of NPF at active site and PFD at allosteric site, while Autodock VINA could.

·   Figure 6: Clarify the meaning of the black virtual boxes. Panel A’s legend should be “FurAct” instead of “FurAll”.

·   Page 17 lines 458-460: The statement “The high affinity of the TDZs to Furin may be through its sulfur-heterocycle functional group interactions” is not supported by the docking and MD results for PGZ and RGZ, as the sulfur group does not form direct interactions with Furin residues. How was this conclusion reached? Can these sulfur groups be replaced by other heterocycles?

Overall, the manuscript presents promising results, but addressing these concerns will significantly strengthen the validity and impact of the study.

Author Response

The authors appreciate the reviewer's comments. All the responses are in the document attached.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear authors,

I want to clarify that my previous comments were made to improve the communication of your article.

Thank you for all the information; it was arduous work. I think that this work has great potential within your area of ​​study. Therefore, I believe many of the answers should have been incorporated into the main text.

In general, in studies involving docking and molecular dynamics simulations, I expect to find a discussion of the validation of the docking protocol, analysis of the docking result, the complex stability, and the determination of the binding energy by molecular dynamics simulation.

In this sense, autodock and autodock vina are not ideal programs for blind docking because once they identify a local minimum, they cannot evaluate other possible local minima except in special conditions as mentioned in the comment answer 16 (population conditions, evaluation energy, number of runs and grid spacing).

Considering that you indicate in the letter that the allosteric site is based on the article by Feng et al. (2023) (doi: 10.3390/molecules28041883) and Gross (2014)

(http://hdl.lib.byu.edu/1877/etd7209), could have focused on a specific study in this region, selecting through a cluster analysis available in autodock tools.

In the case of redocking, only the study with 7QXZ and 7LCU must have given RMSD values ​​below 2 angstroms. At no point in the discussion is the reason for studying these structures deposited in the PDB mentioned. If there are no differences in the active site, leave 7QXZ; if not, it is necessary to discuss this data.

The analysis of intermolecular interactions could be concentrated on those that will be addressed in dynamics, such as hydrogen bonding. Proteases are known to follow Schechter and Berger notation, which can help in this discussion.

The energy value obtained by docking is not a valid criterion for the dynamics simulation time. The article by Swegat and collaborators (10.1016/S0006-3495(03)74962-5) carries out a mechanistic study where runs of 50 ns were carried out using 21 different configurations. Alturki and collaborators (10.3390/molecules27144391) do not describe how they analyzed the stability of the ligand in the active site, but the figure indicates that they calculated the RMSD of the ligand fitted to the protein. Haider, Barakat, and Ui-Haq (10.3390/molecules25204829) do not describe how they analyzed the stability of the ligand in the active site, and it is not clear whether the ligand was fitted to the protein or whether the fit was made to ligand itself.

You must make at least 100 ns to the allosteric site. Review comments on DM and incorporate them into the main text. Note that you have a change in the initial position of the MD, and this is very common, so an RMSD for the ligand fitted to the protein or an analysis of the distance from an essential residue to the center of mass of the ligand will tell you at what moment he leaves the starting position. Even if the ligand leaves, it may indicate an affinity for a specific region if it remains in a particular position longer.

It is crucial to perform an energy analysis for the ligands using MM-PBSA or another MD-based technique.

 

Best regards

Author Response

Dear authors,

I want to clarify that my previous comments were made to improve the communication of your article.

Thank you for all the information; it was arduous work. I think that this work has great potential within your area of ​​study. Therefore, I believe many of the answers should have been incorporated into the main text.

In general, in studies involving docking and molecular dynamics simulations, I expect to find a discussion of the validation of the docking protocol, analysis of the docking result, the complex stability, and the determination of the binding energy by molecular dynamics simulation.

Comment 1: In this sense, autodock and autodock vina are not ideal programs for blind docking because once they identify a local minimum, they cannot evaluate other possible local minima except in special conditions as mentioned in the comment answer 16 (population conditions, evaluation energy, number of runs and grid spacing).

Response 1: We appreciate the reviewer's suggestions. In this regard, an analysis using the BINDSURF server (https://bio-hpc.eu/software/bindsurf/) was included in the main manuscript; the results obtained by this blind docking agree with the previous analysis obtained with Autodock Vina and Autodock4. The information related to these actualizations can be found in the Method and Result sections (Lines 145-172 and 221 -285, respectively)

 

Comment 2: Considering that you indicate in the letter that the allosteric site is based on the article by Feng et al. (2023) (doi: 10.3390/molecules28041883) and Gross (2014)

(http://hdl.lib.byu.edu/1877/etd7209), could have focused on a specific study in this region, selecting through a cluster analysis available in autodock tools.

Response 2: The reviewer is in the right when suggesting that the docking could be directed to that specific site; nevertheless, we consider that since this pocket is well predicted by Feng and Gross, this information can still be confirmed by whet-lab methods. The authors consider that our results add to the knowledge about these sites' relevance in the regulation of furin.

 

Comment 3: In the case of redocking, only the study with 7QXZ and 7LCU must have given RMSD values ​​below 2 angstroms. At no point in the discussion is the reason for studying these structures deposited in the PDB mentioned. If there are no differences in the active site, leave 7QXZ; if not, it is necessary to discuss this data.

Response 3: We appreciate the reviewer's comment. We add information in the method section to clarify the reason for this procedure, lines 150-152.

 

Comment 4: The energy value obtained by docking is not a valid criterion for the dynamics simulation time. The article by Swegat and collaborators (10.1016/S0006-3495(03)74962-5) carries out a mechanistic study where runs of 50 ns were carried out using 21 different configurations. Alturki and collaborators (10.3390/molecules27144391) do not describe how they analyzed the stability of the ligand in the active site, but the figure indicates that they calculated the RMSD of the ligand fitted to the protein. Haider, Barakat, and Ui-Haq (10.3390/molecules25204829) do not describe how they analyzed the stability of the ligand in the active site, and it is not clear whether the ligand was fitted to the protein or whether the fit was made to ligand itself.

You must make at least 100 ns to the allosteric site. Review comments on DM and incorporate them into the main text. Note that you have a change in the initial position of the MD, and this is very common, so an RMSD for the ligand fitted to the protein or an analysis of the distance from an essential residue to the center of mass of the ligand will tell you at what moment he leaves the starting position. Even if the ligand leaves, it may indicate an affinity for a specific region if it remains in a particular position longer.

Response 4: We appreciate the reviewer's observation. In this sense, a molecular dynamic extension was performed through 100 ns. All the newly added information is marked in blue.

 

Comment 5: It is crucial to perform an energy analysis for the ligands using MM-PBSA or another MD-based technique.

Response 5: We appreciate the reviewer's suggestion for additional analysis. However, due to limitations in our current computational resources, we were unable to perform it at this time. We plan to evaluate this affinity through enzymatic and cellular assays to determine drug inhibitory capacity in future studies. We believe the current in silico data provides valuable groundwork for these wet lab experiments. Additionally, a limitations section addressing this analysis has been incorporated into the study (lines 535-542).

 

The authors appreciate the critical analysis by the reviewer.

Reviewer 3 Report

Comments and Suggestions for Authors

 

Minor:

Table 2: Kcal/mol -> kcal/mol.

Lines 356-257: 24.87 (µM) -> (24.87 µM); (41.75pM) -> (41.75 pM); (1.28µM) -> (1.28 µM); (3.33µM) -> (3.33 µM).

Table 3: Problem with the caption that is not correctly formatted.

Figure 5 and 6: in panel B, the x-axes should be extended to 100 ns as in panel A that is presented above

 

Author Response

Minor:

Table 2: Kcal/mol -> kcal/mol.

Lines 356-257: 24.87 (µM) -> (24.87 µM); (41.75pM) -> (41.75 pM); (1.28µM) -> (1.28 µM); (3.33µM) -> (3.33 µM).

Table 3: Problem with the caption that is not correctly formatted.

Figure 5 and 6: in panel B, the x-axes should be extended to 100 ns as in panel A that is presented above

 

Comment: The authors appreciate the critical analysis by the reviewer. All the changes suggested by the reviewer were included in the final version of the manuscript.

Reviewer 4 Report

Comments and Suggestions for Authors

All my previous concerns have been adequately addressed in this revised version. Therefore, I recommend its publication.

Author Response

All my previous concerns have been adequately addressed in this revised version. Therefore, I recommend its publication.

 

Comment: The authors appreciate the critical analysis by the reviewer.

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Author,

Thank you for your responses and for indicating where the changes were made. I insist that it is necessary to calculate the free energy for the molecular dynamics simulation.

I suggest applying the PCA method or cluster analysis and selecting at least 20 frames from a region where the RMSD has an average standard deviation below 1 Angstrom.

Best regards

Author Response

Comment:

Dear Author,

Thank you for your responses and for indicating where the changes were made. I insist that it is necessary to calculate the free energy for the molecular dynamics simulation.

I suggest applying the PCA method or cluster analysis and selecting at least 20 frames from a region where the RMSD has an average standard deviation below 1 Angstrom.

Best regards

 

Response:

We appreciate the reviewer's suggestions in advance. The authors consider that although the analyses proposed by the reviewer provide interesting information regarding the interaction, this information can be obtained using wet lab tools. These tools are part of further analyses to be performed by the working group, with which three of the four reviewers to whom the present work was assigned agree.

We reiterate that the objective of the present work is not the discovery of new drugs but a starting point for the repositioning of drugs evaluated here. Two have already shown interaction with furin, although without information on the mechanism of interaction, either in the active site or in the allosteric site, which was the main objective of the present study.

We appreciate the reviewer's remarks and his dedication to reviewing our present study, as he is always looking for the best tools.

Best regards!