Discovering peptide inhibitors of thrombin as a strategy for anticoagulation

Abstract

Unusual blood clots can cause serious health problems, such as lung embolism, stroke, and heart attack. Inhibiting thrombin activity was adopted as an effective strategy for preventing blood clots. In this study, we explored computational-based method for designing peptide inhibitors of human thrombin therapeutic peptides to prevent platelet aggregation. The random peptides and their 3-dimentional structures were generated to build a virtual peptide library. The generated peptides were docked into the binding pocket of human thrombin. The designed strong binding peptides were aligned with the native binder by comparative study, and we showed the top 5 peptide binders display strong binding affinity against human thrombin. The 5 peptides were synthesized and validated their inhibitory activity. Our result showed the 5-mer peptide AEGYA, EVVNQ, and FASRW with inhibitory activity against thrombin, range from 0.53 to 4.35 μM. In vitro anti-platelet aggregation assay was carried out, suggesting the 3 peptides can inhibit the platelet aggregation induced by thrombin. This study showed computer-aided peptide inhibitor design can be a robust method for finding potential binders for thrombin, which provided solutions for anticoagulation.

1. Introduction

Blood clot formation is a way to prevent bleeding from an injury. However, unusual blood clots in blood vessels can cause stroke, pulmonary embolism, and heart attack.^[1] Anticoagulation therapy is used to against blood clot formation.^[2] Using thrombin inhibitor was reported as a way for anticoagulation therapy. Previously, heparin and bivalirudin were successfully used as inhibitors of thrombin.^[3] The major role of heparin is to activate antithrombin to stop the clotting process.^[4] Bivalirudin was latterly reported as an alternative drug for anticoagulation therapy by directly inhibiting thrombin.^[5] However, using bivalirudin or heparin was expensive despite the 2 drugs being effective.^[5] Meanwhile, there is still lack of large prospective studies by using of bivalirudin. Currently, searching for effective drugs for effectively preventing blood clotting is still necessary.

Peptide drug is a subtype of protein drug but with a much lower molecular weight. Peptide drugs are now widely used for the treatment of various diseases.^[6] The key advantages of peptide drugs are their high affinity and specificity against targets, less immunogenicity, and lower production costs.^[7] The sales of peptide drugs reached more than US$70 billion in 2019, suggesting they are well-adopted worldwide.^[7] Rational design of peptide drugs has received growing attention as a result of the low experimental costs.^[8,9] The development of structural biology has led to the discovery of protein or peptide drugs entering a new era.^[10,11] Designing peptide drugs is a top-down process, the first step is to build a peptide library, followed by peptide docking to the receptor to filter strong binding peptides. Finally, active peptides can be recognized by in vitro tests.^[12–14]

Several studies were conducted for screening thrombin inhibitors.^[15–17] For screening chemical compound inhibitors of thrombin, 16 derivatives of (1R,3S)-2,3,4,9-tetrahydro-β-carboline-3-carboxylic acid with high docking score against bovine thrombin were synthesized and validated their activity.^[15] Their results suggested the 16 derivatives can inhibit the platelet aggregation induced by thrombin. In a separate study, Ru et al used soybean protein hydrolyzate to inhibit thrombin.^[16] The soybean protein hydrolyzate was analyzed in composition using LC-MS/MS, which resulted in 2176 identified peptides. Finally, a nonapeptide was successfully identified with high affinity against thrombin relying on in silico analysis. These studies supported that both chemical compounds and peptides were able to bind with thrombin to facilitate anticoagulation function.

Compared with harvesting peptide fragments from natural products, this study introduced a more concise and computer-aided method for screening peptide inhibitors. Here, we developed a combined protocol by integrating several different strategies for screening thrombin inhibitors (Fig. 1). Random peptides were generated using Rosetta BuildPeptide module, and docked into thrombin using Rosetta Flexpepdock.^[18] Previous study showed that the Rosetta score obtained from Rosetta Flexpepdock can be used to represent the correct binding between protein receptor and peptide ligand.^[19] The docking complex with lower total score is more likely to represent the real binding behavior and binding affinity.^[19] To find out peptides with strong binding affinity against thrombin, we selected docking complexes with low Rosetta docking score for virtual comparative study, followed by validating their affinity in vitro. In further, molecular dynamics (MD) simulation was carried out to gain an insight of the binding mechanism.

Figure 1.

The protocol for designing peptide inhibitors.

2. Materials and methods

2.1. Generating random peptides

Peptide library is built via 2 steps. Firstly, the sequences of random peptides were generated using Python script, which enabling peptides to be generated at manual length. The peptides generated in this study at a length of 5 based on previous reported peptide binders.^[16,20,21] The 3-dimentional (3-D) conformation of the random peptides were generated using Rosetta BuildPeptide module based on the peptide sequences.^[22] For each sequence, only one 3-D peptide conformation was generated, and refined using Rosetta Monte Carlo simulation. Initial peptide library contained 9944 samples. The Python scripts for controllable generating peptides and building their structures were provided in Table S1, Supplemental Digital Content 2, http://links.lww.com/MD/L279.

2.2. Molecular docking

Human thrombin (PDB: 2BVR) was selected as the docking receptor according to previous study.^[23] The structure of thrombin was refined using Rosetta Relax.^[24] Two rounds of Rosetta Relax were carried out in order to fully refine the input structure.^[25] The first round aims to refine the side chains of protein, and the second round aims to refine the whole structure. We used Rosetta ref2015 for scoring throughout this study.^[26] For each round of relax, we selected structures with the lowest score (indicating the lowest folding energy) for the next step.^[27] The total score obtained by Rosetta relax were provided in Table S2, Supplemental Digital Content 3, http://links.lww.com/MD/L280.

Molecular docking was carried out based on Rosetta FlexPepDock,^[18] and we used the refinement protocol to obtain the best binding pose between peptides and human thrombin. The refinement protocol of Rosetta FlexPepDock aims to optimize the backbone and rigid-body orientation of the peptides. By using this module, a proper relative position was required between receptor and docking ligands. To do this, we used the Editconf of Gromacs-2020^[28] to accommodate ligands to the binding area of the receptor. The binding pocket of thrombin was determined according to previous studies.^[15–17,23] Next, we combined the PDB file of the receptor and the ligand for molecular docking. Noticed that the side chain of both protein and peptides were optimized, and peptide conformation was separately refined during docking.

2.3. Molecular dynamics simulation

To understand the binding mechanism between designed peptides and thrombin, molecular dynamics simulation was carried out. In this study, the designed peptide binder in docking complex with human thrombin, and human thrombin carried its native binder (2-[2-(4-chloro-phenylsulfanyl)-acetylamino]-3-(4-guanidino-phenyl)-propionamide) were used for MD simulation. The docking complex was independently subjected to MD simulation using Gromacs-2020,^[28] and the simulation system was filled with SPC/E water in a cubic box. The system was neutralized using Na⁺ and Cl⁻, and the distance between the edge of water box and docking complex was 15 Å. The system undergoes energy minimization using the steepest descent method, followed by equilibration using the isochoric–isothermal ensemble and isothermal–isovolumetric ensemble under 300 K for 100 and 200 ps, respectively. The production of MD simulation was carried out for 100 ns under 300 K, and we used the last 10 ns of simulation for gmxMMPBSA^[29] analysis.

2.4. Determination of half-maximal inhibitory concentration of thrombin

Half-maximal inhibitory concentration (IC₅₀) of the peptides against thrombin was determined as previously reported.^[16,30] Human thrombin and its chromogenic substrate S-2238 (beta-Ala-Gly-Arg para-nitroanilide) were purchased from Sigma-Aldrich (Shanghai, China). The peptides used in this study was synthesized by Sangon (Shanghai, China). For activity assay, S-2238 was prepared at a concentrations of 1 mM, human thrombin was prepared at 2 U/mL, and the peptide solution varied its concentration from 0 to 30 mM. All the 3 components were dissolved in Tris-HCl (50 mM, pH 7.5) buffer. 10 μL of thrombin was premixed with 40 μL of the peptide solution at different concentrations separately. For the blank and test sample, 10 μL of Tris-HCl (50 mM, pH 7.5) buffer or S-2238 solution was added, respectively. The reaction was conducted for 5 minutes at 37°C, and the absorbance was measured at 375 nm using the Cytation 3 microplate reader (BIOTEK, USA). The IC₅₀ value indicated the peptide concentration needed for obtaining 50% of thrombin inhibitory activity.

2.5. Evaluating peptide properties

The physicochemical properties of peptides were calculated using Python peptides module. The calculated properties including solubility, hydrophobicity, and net charge. The bioactivity of these peptides were calculated using PeptideRanker (http://distilldeep.ucd.ie/PeptideRanker/).^[31]

3. Results and discussion

3.1. Virtual screening peptide inhibitors by molecular docking

In this study, we aim to design peptide inhibitors of human thrombin. A proper length of the designed peptide was selected to 5-mer, since peptides at this length were enough to bind to thrombin according to previous studies.^[16,20,21] Meanwhile, we considered short peptides may be less able to form helix or strands, and with less possible 3-D conformations compared with long peptides.^[32] Therefore, short peptides may contribute to obtain highly accurate docking poses. Additionally, the short peptides face big challenging for degradation in blood circulation, current techniques such as peptide N- or C-terminus modification, conjugating with macromolecule, or using drug delivery vehicle have shown promised effect for anti-degradation.^[33] Random peptide sequences were generated firstly, and the 3-D structures of the random 5-mer peptides were generated using Rosetta BuildPeptide. The 5-mer length random peptides have 5²⁰ possible compositions. Here, we generated 9944 random peptides at a length of 5-mer for molecular docking (Supplementary File 1, Supplemental Digital Content 1, http://links.lww.com/MD/L278).

Several studies showed the designed or isolated peptides and compounds can be used as thrombin inhibitor.^[15,16,23] These inhibitors can against human or bovine thrombin.^[15,16,23] Structural alignment between human^[34] and bovine^[23] thrombin showed they are in high similarity with only 0.229 Å RMSD differences (Fig. 2A), indicating similar inhibitory mechanism. The 2 binding complexes of thrombin carried their native binders were visualized, and the steric position of the inhibitors were shown as Figure 2B. We selected the centroid of the native inhibitors (shown as Fig. 2B) for accommodating peptides. The structure of human thrombin (PDB: 2BVR) was refined using Rosetta relax^[24] prioritize to molecular docking. The refined structure displayed 0.885 Å difference from the initial structure (Fig. 2C). Molecular docking was carried out using our developed Rosetta scripted integrated FlexPepDock refine protocol.

Figure 2.

Alignment of human and bovine thrombin. (A) Structural alignment of human thrombin (PDB: 2BVR, colored cyan) and bovine thrombin (PDB: 1KTS, colored green). (B) Binding mode of human thrombin (cyan) and bovine thrombin (green) with their inhibitors while the inhibitors for human and bovine thrombin colored in blue and red, respectively. The coordinates for the center of the ligand was shown. (C) Structural alignment of human thrombin before (green) and after (cyan) refinement. (D) The docking score and total energy of peptide docked into thrombin receptor, where x-axis indicates total energy, and y-axis indicates dG_cross/SASA.

For each docking complex, we choose the docking pose with the lowest docking score (represented by dG_cross/SASA) for further analysis. The obtained docking score of these peptides ranged from −4.2 to −0.3, while the total energy for the docking complex varied from −1017.6 to −779.5 (Fig. 2D). The total energy showed weak correlation with the docking score, indicating the docking complex is fully minimized regarding the docking affinity. These results indicated that this protocol can be used to search the minimized docking structure. In detailed, we inspected the peptide and protein conformational change before and after molecular docking to illustrate the protocol introduced in this study is able to perform whole structural molecular docking (Figure S1, Supplemental Digital Content 4, http://links.lww.com/MD/L281).

3.2. Comparative study

Comparative study was used to evaluate the capacity of the selected compound or peptides against target proteins.^[35,36] In this study, we aligned the docking score of the designed peptides and human thrombin with its native binder (2-[2-(4-chloro-phenylsulfanyl)-acetylamino]-3-(4-guanidino-phenyl)-propionamide, 4CP). The native binder displayed a docking score of -38 against human thrombin, which was higher than that of several generated peptides. It is obvious that the generated peptides were steric larger than that of the native binder. As shown in Figure 3A, the interactions between human thrombin and its native binder relied on G216, G219, E192, W60, W215, A190, and Y228. However, the area dramatically enlarged while peptides docked into the pocket of human thrombin (Fig. 3B). Analyzing the interactions of the docked peptides revealed several key residues for binding, including the above mentioned 7 residues and D189, G193, D221, and F227 (Fig. 3B). These results suggested the generated peptides with higher chances to bind to the surface of thrombin pocket.

Figure 3.

Interactions of thrombin carried different binders. (A) Thrombin docked with its native binder 4CP. (B) Thrombin docked with peptide AEGYA, CPFVM, EVVNQ, and VEICM.

In further, the docking complex of human thrombin carried its native binder was minimized and scored using Rosetta. The calculated total energy for the docking complex was −898, within the range of generated peptides docked into human thrombin (−1017.6 to −779.5), suggesting the overall energy of the achieved protein-peptide docking complex is acceptable. These results indicated that the designed peptides potentially have stronger affinity against human thrombin than the native binder.

3.3. Analyzing peptide properties

To evaluate the designed peptides as potential therapeutic drugs, peptide properties were analyzed. The solubility and hydrophobicity determined the solution for dissolving peptides, and the net charge of peptides determined its potential immunogenicity. This study, the solubility, hydrophobicity, net charge, and bioactivity of designed peptide were analyzed. Our results showed that half of the top 10 peptide binders with the lowest docking score were soluble (Fig. 4). All of these peptides were negatively charged varied from −0.003 to −1.0459 (Fig. 4). Meanwhile, the strong binders tend to be high bioactivity (Fig. 4).

Figure 4.

The properties of selected peptides. Peptide bioactivity is calculated using PeptideRanker (http://distilldeep.ucd.ie/PeptideRanker/).^[31] Peptide solubility, charge, and hydrophobicity were calculated using Python Peptides module.

3.4. In vitro validating peptide inhibitors against thrombin

To test the inhibitory activity of selected peptides against thrombin, we carried out inhibitory assay. Thrombin and its native substrate S-2238 were incubated, followed by adding the selected peptide inhibitors. Strong binding peptides are suggested to competitively bind to thrombin to prevent the reaction triggered by thrombin, which disrupted the substrate consumption. According to peptide property calculation, we synthesized 5 peptides which were predicted to be soluble with the highest bioactivity, including AEGYA (P1), EVVNQ (P2), TVGAV (P3), SEGRM (P4), and FASRW (P5). The peptides were used for in vitro inhibitory assay, our result indicated that 3 out of 5 validated peptides can bind with thrombin, which are P1, P2, and P5 (Table 1). The IC₅₀ for P1, P2, and P5 are 0.53 μM, 4.35 μM, and 2.67 μM (Fig. 5). The peptides showed binding affinity against thrombin are more hydrophobic (Table 1), which in agree with previous study that hydrophobic residues can contribute to the inhibitory activity.^[37]

Table 1.

The inhibitory activity of each peptides and their properties.

	IC50 (mM)	Predicted total score	GRAVY
FVEIQAL	0.53	−1019.931633	1.443
SLVTKVS	4.35	−1019.750547	0.857
FFQQSGG	2.67	−1018.225218	−0.429
AEVSNPV	–	−1014.049728	0.114
STYINSF	–	−1012.798991	0.029
NFMMRQM	–	−1012.5918	−0.429

GRAVY was calculated using Expasy ProtParam (https://web.expasy.org/protparam/) to represent the hydrophobicity of peptides. Higher GRAVY values indicated high hydrophobicity.

Figure 5.

Thrombin inhibitory activity of different peptides. The peptide sequences were listed in figure.

Previous studies attempted to isolate peptide inhibitors of thrombin from plants, the identified functional peptide length varied from 5 to 13-mers.^[16,20,21] In this study, the length of the generated peptide was 5-mer, but the IC₅₀ of these functional peptides ranged from 0.53 to 4.35μM, which is comparable stronger than the isolated peptides.^[16] Several studies highlighted the significance of using computer-aided ways for identification chemical compounds or designing peptide inhibitors of protein or enzyme,^[38] suggesting computer-aided methods were gradually widely adopted. Compared with extraction from plants or mammals, the computer-aided method for designing peptides transferred abundant of laboratory work to computational work which avoids of time consumption and was labor saving.

3.5. Molecular dynamics simulation

To understand the binding mechanism behind thrombin and the designed peptide, MD simulation was carried out using the strongest binding peptide P1 and the native binder (2-[2-(4-chloro-phenylsulfanyl)-acetylamino]-3-(4-guanidino-phenyl)-propionamide, 4CP) from binding complex of PDB: 2BVR. The fluctuation of human thrombin carried its native binder and the designed peptide were stabilized after 5 ns. The average RMSD for thrombin-4CP and thrombin-P1 were 2.2 Å and 1.8 Å, respectively (Fig. 6A), suggesting that the peptide P1 binding to human thrombin was more stable than that of thrombin-4CP complex. The RMSF analysis indicated that the region of 190 to 195 and 220 to 225 surrounded the P1 pocket of thrombin was very flexible in the 2 docking complex (Fig. 6B). These results indicated that the 2 binders can stimulate the conformational changes of the P1 pocket to stabilize the binding form of the 2 complexes.

Figure 6.

Molecular dynamics simulation of thrombin in bound with FVEIQAL (P1). (A) RMSD changes during the simulation. (B) RMSF changes during the simulation. (C) The binding structure of the top 1 cluster from the simulation, left: the stereo view of 3-D and 2-D binding of thrombin and peptide P1, right: the 2-D view of thrombin and 4CP interaction. (D) The binding energy calculated using gmxMMPBSA adopting the last 10 ns of the simulation.

To investigate the key residues that dominate the interaction between thrombin and binders, gmxMMPBSA was carried out based on the last 10 ns of simulation.^[29] The key binding sites between thrombin and peptide P1 were recognized, suggesting G191, R226, D225, G221, and G223 exhibit the highest binding affinity (Fig. 6C). The protein-ligand interaction was further carried out based on the top 1 cluster of each trajectory. The obtained docking structure showed that the 5 key binding sites recognized from gmxMMPBSA critically contributed to thrombin and the inhibitors binding, while hydrogen bonds were mainly occupied for binding (Fig. 6C). In comparison, thrombin bond against 4CP mainly relied on G221 and G223 as that against peptide P1, while the other residues showed less contributions to thrombin-4CP binding (Fig. 6D). The overall binding energy for thrombin-4CP and thrombin-P1 were −49.2 and −56.3, respectively. These results indicated that thrombin-P1 was more tightly binding than that of thrombin-4CP.

4. Conclusions

In this study, we explored in silico methods for designing peptide inhibitors of thrombin, and validated the inhibitory activity of our designed peptides. This design was 3-D structure and molecular docking-based. We generated random peptides using Python script and Rosetta BuildPeptide module. The generated 9944 peptides were accommodated in the thrombin P pocket for molecular docking. We showed the generated peptides were in high diversity which display distinct docking score against thrombin. To extract peptides with strong inhibitory activity, we selected peptides with the lowest docking score for comparative study. Our result indicated that the designed peptides were able to block the activity of thrombin. The strongest binding peptide showed remarkably inhibitory activity. We also characterized the key sites for thrombin and peptide binding using MD simulation. Here, a large peptide library was organized and finally, only a small portion was selected for experimental validation. Although we demonstrated our protocol can be used for peptide design, large size benchmark tests were still needed.

Author contributions

Conceptualization: Shuxin Zhen.

Investigation: Guiping Wang, Xiaoli Li.

Methodology: Shuxin Zhen, Jing Yang.

Writing – original draft: Jiaxin Yu.

Writing – review & editing: Yucong Wang.

Supplementary Material

medi-103-e36849-s001.docx ^{(28.1KB, docx)}

medi-103-e36849-s002.docx ^{(26.6KB, docx)}

medi-103-e36849-s003.xlsx ^{(440.7KB, xlsx)}

medi-103-e36849-s004.docx ^{(719.7KB, docx)}