The prevalence of respiratory illness caused by the novel SARS-CoV-2 virus associated with multiple organ failures is spreading rapidly because of its contagious human-to-human transmission and inadequate globalhealth care systems. Pharmaceutical repurposing, an effective drug development technique using existing drugs, could shorten development time and reduce costs compared to those of de novo drug discovery. We carried out virtual screening of antiviral compounds targeting the spike glycoprotein (S), main protease (Mpro), and the SARS-CoV-2 receptor binding domain (RBD)–angiotensin-converting enzyme 2 (ACE2) complex of SARS-CoV-2. PC786, an antiviral polymerase inhibitor, showed enhanced binding affinity to all the targets. Furthermore, the postfusion conformation of the trimeric S protein RBD with ACE2 revealed conformational changes associated with PC786 drug binding. Exploiting immunoinformatics to identify T cell and B cell epitopes could guide future experimental studies with a higher probability of discovering appropriate vaccine candidates with fewer experiments and higher reliability.
The prevalence of respiratory illness caused by the novel SARS-CoV-2 virus associated with multiple organ failures is spreading rapidly because of its contagious human-to-human transmission and inadequate globalhealth care systems. Pharmaceutical repurposing, an effective drug development technique using existing drugs, could shorten development time and reduce costs compared to those of de novo drug discovery. We carried out virtual screening of antiviral compounds targeting the spike glycoprotein (S), main protease (Mpro), and the SARS-CoV-2 receptor binding domain (RBD)–angiotensin-converting enzyme 2 (ACE2) complex of SARS-CoV-2. PC786, an antiviral polymerase inhibitor, showed enhanced binding affinity to all the targets. Furthermore, the postfusion conformation of the trimeric S protein RBD with ACE2 revealed conformational changes associated with PC786 drug binding. Exploiting immunoinformatics to identify T cell and B cell epitopes could guide future experimental studies with a higher probability of discovering appropriate vaccine candidates with fewer experiments and higher reliability.
Protein structure retrieval We have retrieved the crystal structures of prefusion SARS-CoV-2 spike glycoprotein with an RBD (S) (PDB ID: 6VSB) and the Mpro in complex with an inhibitor N3 (PDB-ID: 6 LU7) from PDB. For the complex interaction analyses with RBD of S protein, we have taken the structure of native human ACE–related carboxypeptidase (ACE2) (PDB ID: 1R42). We have used UCSF Chimera and Discovery Studio Visualizer to visualize and analyze the interactions.
Virtual screening and molecular docking The antiviral drug compounds were retrieved from the ChEMBL database with a search query term “antiviral drugs” and “coronavirus” that resulted in 640 chemical compounds with the corresponding filters for data availability, e.g., SMILES and Structure Data File formats. We have screened the antiviral drug molecules for RO5 violations and refined them to get 3D coordinates using the Open Babel command-line tool. All the structures that have passed the RO5 rule have been subjected for further refinement using MarvinSketch. The drug compounds and the corresponding target receptors, e.g., 6VSB and 6 LU7, were submitted for virtual screening using AutoDock Vina. The virtually screened best compounds were then docked with the target receptors again to ensure the conformation poses and binding affinities. We have performed blind docking as the location of binding site is unknown for both target receptors. The grid for the target receptors was set to 126 Å by 126 Å by 126 Å with a spacing of 1.000 Å. The interactions were visualized using Discovery Studio Visualizer. Furthermore, we have again docked the best-screened compounds to the complex of RBD-ACE2 in two conformations: (i) partial chain A–RBD of S protein with ACE2 complex and (ii) full trimeric S protein with ACE2 complex using AutoDock Vina.
MD simulation MD simulation using GROMACS v.2019.2 has been performed for the complex molecules (drug bound proteins). We obtained the topologies for all the small antiviral molecules from the PRODRG database. We have optimized the parameters of the target receptor and the drug molecules using the GROMOS96 54a7 force field. The complex systems were placed in a periodic cubic box solvated with simple point charge solvent molecules. Periodic boundary conditions with a 15-Å cutoff for nonbonded interactions were applied, with the particle mesh Ewald method applied to account for the long-range electrostatic interactions. The system was neutralized with Na+ counterions to attain equilibration. Energy minimization and equilibration were carried out in three steps as follows: (i) We minimize the whole system containing ions, solvent, protein, and ligand for up to 50,000 steps using a steepest-descent algorithm. (ii) Constraints were added to protein and the ligand dimer for 100 ps during heating using NVT (number of atoms, volume, temperature) ensemble with leapfrog integrator and linear constraint solver holonomic constrains. (iii) NPT ensemble was used at constant pressure (1 bar) and temperature (300 K) for 100 ps using a time step of 2 fs for equilibration phase 2. The SHAKE algorithm was used to constraint hydrogen to heavy atom bonds. The MD production phase for all the systems has been simulated for 10 ns with a time step of 2 fs. Furthermore, after 10-ns simulation, the protein-ligand interaction energy was evaluated to compute the nonbonded interaction energy and short-range nonbonded energies, which were quantitatively reproduced with energy profiles generated by GROMACS tools. Furthermore, we used MM-PBSA to calculate the polar and nonpolar solvation energies with corresponding binding energy decomposition of the complexes. MM-PBSA calculates the free energy of the docked complex (the binding free energy of the protein with ligand in a solvent medium) where the general expression of the term can be depicted as ΔGbinding=Gcomplex−(Gprotein+Gligand) (1)where Gcomplex is the total free energy of the protein-ligand complex and Gprotein and Gligand are total free energies of the isolated protein and ligand in solvent, respectively.Gx=⟨EMM⟩−TS+⟨Gsolvation⟩(2)where x is the protein or ligand or protein-ligand complex. ⟨EMM⟩ is the average molecular mechanics potential energy in a vacuum. TS refers to the entropic contribution to the free energy in a vacuum where T and S denote the temperature and entropy, respectively. The l term ⟨Gsolvation⟩ is the free energy of solvation.
E=Ebonded+Enonbonded=Ebonded+( EvdW+Eelec)(3)
where Ebonded is bonded interactions consisting of bond, angle, dihedral, and improper interactions. The nonbonded interactions (Enonbonded) include both electrostatic (Eelec) and van der Waals (EvdW) interactions depicted using a Coulomb and Lennard-Jones potential function, respectively. Moreover, the free energy of solvation, which is the energy required to transfer a solute from a vacuum into the solvent, has been calculated including polar and nonpolar solvation energies that can be depicted as Gsolvation=Gpolar+Gnonpolar (4 )where Gpolar and Gnonpolar are the electrostatic and nonelectrostatic contributions to the solvation free energy, respectively.
Protein-protein interaction To predict the conformational changes upon binding of ACE2 to the trimeric S protein RBD, we have used PatchDock and FireDock for protein-protein interaction analysis. UCSF Chimera was used for the post protein-protein interaction analyses.
Antigenicity and T cell epitope identification We have retrieved the protein FASTA sequence of SARS-CoV-2 isolate Wuhan-Hu-1, complete genome sequence bearing ID NC_045512.2 for the epitope screening. The prediction of protective antigens and subunit vaccines was evaluated using VaxiJen 2.0 with default parameters. The NETCTL 1.2 server was used for the T cell epitope identification. The method integrates MHC-I binding, proteasomal C-terminal cleavage, TAP transport, and combinatorial scores for the prediction of epitopes. For MHC-I binding, IEDB tools have been used to get the best selected epitopes based on the stabilized matrix base method and inhibitory concentrations (IC50) values for peptide binding to MHC-class I molecules. Furthermore, the selected epitopes were further processed to obtain the specificity to TAP transport, proteasomal cleavage, TAP transport, and MHC-I. The web-based tool IEDB has been used for population coverage analysis.
Immunogenicity prediction The IEDB MHC class I immunogenicity tool and the European Molecular Biology Open Software Suite (EMBOSS) were used for immunogenicity prediction. The algorithm prediction was based on immunogenicity and antigenic scores.
T cell epitope structure prediction The selected T cell epitopes were subjected to the PEP-FOLD server to predict the 3D structure to be able to perform the protein-peptide interaction with HLA-A and HLA-B class I molecules.
B cell epitope prediction IEDB resources were used to classify B cell antigenicity such as Kolaskar and Tongaonkar antigenicity scale, Emini surface usability prediction, Karplus and Schulz versatility prediction, and BepiPred linear epitope prediction analysis. The Chou-Fasman beta-turn prediction tool is used as the antigenic sections of a protein belong to the β-turn areas. We have also used ElliPro and DiscoTope to predict linear and discontinuous peptides, respectively.
Molecular interaction of epitopes to HLA class I molecules The T cell epitopes were further processed for interaction analysis using HLA class I molecules using ClusPro 2.0. ClusPro 2.0 was based on ranking models by cluster size where the ligands were rotated 70,000 conformations. The server also predicts the cluster size and interacting members and gives best models with the lowest energies.
Statistical analysis All statistical data analyses were performed in Origin 2018. Linear curve fitting has been performed using independent and dependent variables with the goal of defining a “best fit” model of the relationship. Use of weighted least-square method to fit a linear model function to specified data has been performed. Box plots, scatter plots, and bar graphs have been depicted to represent the data.
Reference & Source information: https://advances.sciencemag.org/
Read More on: