Abstract
2019-nCov has caused more than 80 deaths as of 27 January 2020 in China, and infection cases have been reported in more than 10 countries. However, there is no approved drug to treat the disease. 2019-nCov Mpro is a potential drug target to combat the virus. We built homology models based on SARS Mpro structures, and docked 1903 small molecule drugs to the models. Based on the docking score and the 3D similarity of the binding mode to the known Mpro ligands, 4 drugs were selected for binding free energy calculations. Both MM/GBSA and SIE methods voted for nelfinavir, with the binding free energy of −24.69±0.52 kcal/mol and −9.42±0.04 kcal/mol, respectively. Therefore, we suggested that nelfinavir might be a potential inhibitor against 2019-nCov Mpro.
1. Introduction
In December 2019, cluster of patients with pneumonia were reported in Wuhan, Hubei Province, China.1, 2 Shortly, a new coronavirus, temporally named 2019-nCov, was identified to be the cause of the disease, which is the seventh member of the family betacoronavirus.1 More than 2,700 infection cases and 80 deaths were reported as of 27 January 2020 from China. In addition, infection cases have been reported from Thailand, Australia, Malaysia, Singapore, France, Japan, South Korea, the United States, Vietnam, Canada and Nepal, indicating that the disease is a potential threat to the global health.3 Sadly, the number is still growing rapidly and no drug has been approved to be effective. Therefore, it is urgent to discover and develop drugs to cure the disease.
Based on its function, the main protease (Mpro) or chymotrypsin-like protease (3CLpro)4 is suggested to be a potential drug target to combat 2019-nCov, which is highly conservable among coronaviruses. Sequence alignment revealed that the Mpro of 2019-nCov shares 96% similarity with that of SARS (severe acute respiratory syndrome) (Figure 1). Studies for identifying the inhibitors of 2019-nCov Mpro were quickly performed for discovering and developing drugs against the disease. For example, Hualiang Jiang and collaborators identified 30 drugs and compounds as the Mpro inhibitors via protein modelling and virtual screening, which is a rapid progress in the way to cope with the crisis. In addition, one of the 30 drugs/compounds, remdesivir, was also suggested to be potential inhibitor against 2019-nCov by Liu et al,5 who also suggested 3 other possible inhibitors.
Sequence alignment of 2019-nCov Mpro and SARS Mpro.
For finding more potential drugs as inhibitors of the protein, we modelled the 2019-nCov Mpro structures using SARS Mpro (PDB ID: 2GTB) as template and docked 1903 approved drugs against the model in this study. Fifteen drugs were selected based on the docking score and three dimensional (3D) similarity to available Mpro inhibitors against other coronavirus. For validation, we modelled 10 additional new models of 2019-nCov Mpro and docked the 15 drugs against the new models, which revealed that 6 drugs (nelfinavir, praziquantel, pitavastatin, perampanel, eszopiclone, and zopiclone) have good binding modes with the new models. Binding free energy calculation were then performed for 4 of the 6 drugs using MM/GBSA and SIE methods. Nelfinavir was identified as the best one with predicted binding free energies of −24.69±0.52 kcal/mol by MM/GBSA and −9.42±0.04 kcal/mol by SIE, respectively. Taking into account its high 3D similarity of binding mode to the known Mpro inhibitor, we proposed that nelfinavir should be a potential inhibitor against 2019-nCov Mpro.
2. Materials and methods
2.1 Homology modelling
43 Mpro complexes with ligands were downloaded from protein data bank6 (PDB IDs: 1WOF, 2A5I, 2A5K, 2ALV, 2AMD, 2GTB, 2GX4, 2GZ7, 2GZ8, 2OP9, 2QIQ, 2V6N, 2ZU4, 2ZU5, 3SN8, 3SND, 3SZN, 3TIT, 3TIU, 3TNS, 3TNT, 3V3M, 4F49, 4MDS, 4TWW, 4TWY, 4WY3, 4YLU, 4YOG, 4YOI, 4YOJ, 4ZRO, 5C5N, 5C5O, 5EU8, 5N5O, 5N19, 5NH0, 5WKJ, 5WKK, 5WKL, 5WKM,6FV1) and aligned to 2GTB in PyMOL.7 11 complexes (PDB IDs: 2A5K, 2GTB, 2GX4, 3SND, 3TNS, 3V3M, 4F49, 4YLU, 5NH0, 5WKJ, 5WKM) were served as templates to build 11 2019-nCov Mpro models in SWISS-MODEL server by “user template” mode.8
2.2 Approved Drugs
1905 approved small molecule drugs with 3D coordinates were downloaded from DrugBank release version 5.1.5,9 while 1903 drugs could be converted to pdbqt format by prepare_ligand4.py script in MGLToos version 1.5.6.10
2.3 Molecular Docking
1903 approved drugs in pdbqt format were docked to 2019-nCov Mpro model (template: 2GTB) by smina,11 which is a fork of AutoDock Vina12 with improving scoring and minimization. The hydrogens were added to 2019-nCov Mpro model by pdb2pqr (--ff=amber --ffout=amber --chain --with-ph=7).13 Then the model was converted to pdbqt format by prepare_receptor4.py script in MGLToos version 1.5.6.10 The ligand in 2GTB was used to define the grid and the buffer space was set to 6.0 Å (autobox_add). The random seed was explicitly set to 0 (seed). The exhaustiveness of the global search was set to 32 (exhaustiveness) and at most 1 binding mode was generated for each drug (num_modes). MolShaCS, which utilized Gaussian-based description of molecular shape and charge distribution, was used to calculate the 3D similarities between approved drugs and available Mpro inhibitors.14
2.4 Molecular dynamics simulation
Each simulation system was immersed in a cubic box of TIP3P water that was extended by 9 Å from the solute, with a rational number of counter ions of Na+ or Cl− to neutralize the system. General Amber force field (GAFF)15 and Amber ff03 force field16 were used to parameterize the ligand and protein, respectively. 10,000 steps of minimization with constraints (10 kcal/mol/Å2) on heavy atoms of complex, including 5,000 steps of steepest descent minimization and 5,000 steps of conjugate gradient minimization, was used to optimize each system. Then each system was heated to 300 K within 0.2 ns followed by 0.1 ns equilibration in NPT ensemble. Finally, 5 ns MD simulation on each system at 300 K was performed. The minimization, heating and equilibrium are performed with sander program in Amber16. The 5 ns production run was performed with pmemd.cuda.
2.5 Binding free energy calculation
Based on the 5 ns MD simulation trajectory, bind-free energy (ΔG) was calculated with MM/GBSA17, 18 and SIE19 approaches. In the MM/GBSA, the ΔG was calculated according to equation (1),
where ΔEele and ΔEVDW refer to electrostatic and van der Waals energy terms, respectively. ΔGgb and ΔGnp refer to polar and non-polar solvation free energies, respectively. Conformational entropy (TΔS) was calculated by nmode module in Amber16. The dielectric constants for solvent and solute were set to 80.0 and 1.0, respectively, and OBC solvation model (igb = 5 and PBradii = 5)20 was used in this study. Other parameters are set to default values.
In the SIE, the ΔG was calculated based on equation (2),
where EC and EVDW refer to the sum of intermolecular Coulomb and van der Waals interaction energies, respectively. 
and ΔMSA refer to the changes of reaction field energy and molecular surface area upon ligand binding, respectively. Default values of the global proportionality coefficient (α = 0.1048), the solute interior dielectric constant (Din = 2.25), the van der Waals radii linear scaling coefficient (ρ = 1.1), the molecular surface area coefficient (γ = 0.0129 kcal/mol/Å2), and the constant (C = −2.89 kcal/mol) are used in this study.
3. Results and discussion
3.1 Docking results of 1903 approved drugs against Mpro models
1903 approved small molecule drugs were docked to a Mpro homology model built by SWISS-MODEL using 2GTB as template. There are 672 drugs with the docking score better than −7.0 kcal/mol. The binding modes are essential to the activities. If a compound shares a similar binding mode to a known ligand, it is more likely to have a similar activity to the ligand. Therefore, we calculated the 3D similarities of the binding mode of the docked drugs to the available binders against Mpro. 39 binders in 44 complexes were chosen as the references to calculate the 3D similarities (Table 1). 159 drugs have a 3D similarity larger than 65.0% with at least one known Mpro binder.
Known Mpro binders as references to calculate the 3D similarity of binding modes.
After visualizing the docked complexes carefully, we selected 15 drugs (Table 2) for further analysis. There are some differences between the conformations of the protein in the 3D similarity reference and 2GTB model. Therefore, we modelled 10 new homology models using the proteins in 3D similarity reference as templates and we re-docked the 15 drugs to the 10 new homology models. 6 drugs (nelfinavir, pitavastatin, perampanel, praziquantel, zopiclone, and eszopiclone) show good docking scores and binding modes (Table 3). Because eszopiclone and zopiclone was used to treat insomnia in a low dosage which may not be suitable to treat pneumonia, we carried out further binding free energy calculation for the rest 4 drugs.
15 drugs selected from 2GTB model.
6 drugs selected from 10 new homology models.
3.2 Binding Free Energy calculated by MM/GBSA and SIE
The 4 docked complex structures, i.e., nelfinavir-2GX4, pitavastatin-2GTB, perampanel-5NH0 and praziquantel-3V3M, were subjected to 5 ns molecular dynamics simulations using Amber 16. To provide insight into their binding mechanisms, the binding free energies were calculated by MM/GBSA and SIE approaches. In results of the MM/GBSA approach, the calculated binding free energies of nelfinavir-2GX4, pitavastatin-2GTB, perampanel-5NH0, and praziquantel-3V3M, are −24.69±0.52, −12.70±0.38, −14.98±0.34, and −6.51±0.21 kcal/mol, respectively (Table 4), which highlight nelfinavir as the most active one. In pitavastatin-2GTB, perampanel-5NH0, and praziquantel-3V3M, the van der Waals interaction (Evdw) makes a more significant contribution than the electrostatic interaction (Eele) (Table 4), indicating that van der Waals interaction is the main driving force for the 3 drugs binding. However, the Eele interaction of nelfinavir-2GX4 is very strong, suggesting that the electrostatic interaction also play an important role in the binding of nelfinavir. Furthermore, in results of the SIE approach, the rank of the calculated binding free energies for the 4 docking complexes is consistent with that of the MM/GBSA approach (Table 5) that the nelfinavir has the strongest binding free energy (−9.42±0.04 kcal/mol), indicating the reliability of the current binding free energy analysis.
Components of the Binding Free Energy (kcal/mol) Calculated by MM/GBSA Approach*.
Components of the Binding Free Energy (kcal/mol) Calculated by SIE Approach*.
3.3 Binding modes of nelfinavir against 2019-nCov Mpro
As shown in Figure 2, the binding model of nelfinavir in its docking complex turned out to be very similar with that of the original ligand (TG-0205221)25 of 2GX4 (Figure 2a&b), which is an inhibitor of SARS Mpro with in vitro Ki of 0.053 μM and IC50 of 0.6 μM in the Vero-E6 cells.25 In its crystal structure, TG-0205221 is able to form hydrogen bonds with HIS-163, GLU-166 and GLN-189, and a very week hydrogen bond with PHE-140 (Figure 2c). Our docking results showed that three of the hydrogen bonds involving GLU-166 and GLN-189 maintained upon the binding of nelfinavir and 2019-nCov Mpro, with additionally the possible formation of π-π stacking interaction with HIS-41 (Figure 2d). These observations further demonstrated that nelfinavir would interact with key residues of 2019-nCov Mpro in a similar way to that of the existing inhibitors against coronavirus Mpro.
The binding model of nelfinavir against 2019-CoV Mpro. (a) Binding models of the original ligand (TG-0205221, white) in 2GX4 and nelfinavir (cyan) in the 2019-nCov Mpro protein pocket (white surface); (b) Superposition of TG-0205221 (white) and nelfinavir (cyan) in their binding conformations; (c) Interactions between TG-0205221 and associated residues in the crystal structure (2GX4) of SARS Mpro; (d) Interactions between nelfinavir and associated residues in the homology model of 2019-nCov Mpro. The data in red is the interaction distance (Å).
4. Conclusions
2019-nCov caused more than 80 deaths in China as of 27 January 2020 and is a potential threat to the global health. However, there is no approved drug to treat the disease. 2019-nCov Mpro is a potential drug target to combat the virus, which shares 96% sequence similarity with the corresponding one in SARS. We built 11 homology models of 2019-nCov Mpro and docked 1903 approved small molecule drugs to the 2GTB model. Based on the docking score and the 3D similarity of binding mode to 39 known Mpro binders, 15 drugs were selected for further evaluation. The 15 drugs were then docked to all the 11 homology models, leading to 4 drugs for binding energy calculations. Both MM/GBSA and SIE calculations voted for nelfinavir, a HIV-1 protease inhibitor to treat HIV. Therefore, we suggested that nelfinavir might be active against 2019-nCov Mpro. In addition, pitavastatin, perampanel, and praziquantel might also have moderate activities against 2019-nCov.
Acknowledgments
This work was supported by the National Key R&D Program of China (2017YFB0202601). The simulations were partially run at TianHe 1 supercomputer in Tianjin. We thank Prof. Tingjun Hou in Zhejiang University, China for helpful discussions on the MM/GBSA calculations.
References
