Structural insights into the substrate‐binding site of main protease for the structure‐based COVID‐19 drug discovery

Abstract

An attractive drug target to combat COVID‐19 is the main protease (Mpro) because of its key role in the viral life cycle by processing the polyproteins translated from the viral RNA. Studying the crystal structures of the protease is important to enhance our understanding of its mechanism of action at the atomic‐level resolution, and consequently may provide crucial structural insights for structure‐based drug discovery. In the current study, we report a comparative structural analysis of the Mpro substrate binding site for both apo and holo forms to identify key interacting residues and conserved water molecules during the ligand‐binding process. It is shown that in addition to the catalytic dyad residues (His41 and Cys145), the oxyanion hole residues (Asn142–Ser144) and residues His164–Glu166 form essential parts of the substrate‐binding pocket of the protease in the binding process. Furthermore, we address the issue of the substrate‐binding pocket flexibility and show that two adjacent loops in the Mpro structures (residues Thr45–Met49 and Arg188–Ala191) with high flexibility can regulate the binding cavity’ accessibility for different ligand sizes. Moreover, we discuss in detail the various structural and functional roles of several important conserved and mobile water molecules within and around the binding site in the proper enzymatic function of Mpro. We also present a new docking protocol in the framework of the ensemble docking strategy. The performance of the docking protocol has been evaluated in predicting ligand binding pose and affinity ranking for two popular docking programs; AutoDock4 and AutoDock Vina. Our docking results suggest that the top‐ranked poses of the most populated clusters obtained by AutoDock Vina are the most important representative docking runs that show a very good performance in estimating experimental binding poses and affinity ranking.

Identification of potential anti-COVID-19 drug leads from Medicinal Plants through Virtual High-Throughput Screening

“…As mentioned above, the computational pipeline employed in this research (including the structural clustering strategy to construct the protein ensemble for performing docking calculations and the proposed manner to choose the representative docking poses from the ensemble docking results) has already been designed for correctly predicting experimental binding poses and affinity ranking of M pro -ligand complexes (32) and thus can be utilized to properly produce a rank-ordered list of the phytochemicals of the PHCD Database, according to the Autodock Vina scoring function of the representative docking poses. Consequently, two rank-ordered lists of the docked phytochemicals have been produced for each protein target, one results from the representative poses of the first clusters and the other from the representative poses of the most populated clusters.…”

“…In this work, the antiviral activity of the phytochemical library from the PHCD database (https://persianherb.com) (90) against four key protein targets in the SARS-CoV-2 life cyclethe M pro and PL pro proteases and the Spike and human ACE2 proteinshave been studied and considered using a new successful computational pipeline in the framework of the ensemble docking strategy. The computational pipeline employed in this research, including the structural clustering strategy to construct the protein ensemble for performing docking calculations and the proposed manner to choose the representative docking poses from the ensemble docking results, has recently been designed for correctly predicting experimental binding poses and affinity ranking of M pro -ligand complexes (32) and has now been utilized to properly rank about 5,000 phytochemical compounds of the PHCD database during their screening against different protein targets.…”

“…Membership in clusters depended on the simultaneous fulfillment of two conditions: first that the number of pairwise RMSD values more than 2.0 Å for each pair of residues between members of one cluster should not be more than a predefined number of residues and, second, that average pairwise RMSD values of each residue between all pairs of members in one cluster should be less than a chosen cutoff value. The success of this strategy in protein clustering and the selection of the representative protein structures with the maximum conformational diversity of the binding site has recently been shown in different molecular docking studies (32,110). The 291 monomeric M pro chains were classified into eight clusters, when the values of the two thresholds were set to seven numbers for the first criterion (the number of residues with RMSD > 2.0 Å between members of each cluster) and 1.3 Å for the second criterion (average RMSD value over the residues of each member in a cluster).…”

“…In this research, all the 4892 phytochemicals were docked individually to each of the selected representative structures of the protein targets (8, 8, 10, and 5 representative structures for the M pro , PL pro , Spike, and ACE2 proteins, respectively, as described in sections 2.3). It is noteworthy to point out that the use of an ensemble of multiple rigid receptor conformations for one target protein in docking simulations, often referred to as ensemble docking, is the most common strategy to incorporate the receptor flexibility in the docking that achieves better enrichment than rigid receptor docking to any of the individual members of the ensemble (17,20,32,117,118). However, in the absence of a common strategy to choose the representative docking poses from ensemble docking results, we have very recently presented a new strategy to pick up the most appropriate docking poses from the ensemble docking results (32).…”

“…Each Vina run generates 20 poses. Next, all predicted docking poses for each phytochemical -160 (=8×20) poses for the M pro and 160 (=8×20), 200 (=10×20), and 100 (=5×20) poses for the PL pro , Spike, and ACE2 proteins, respectivelywere collected separately for each protein targets and re-clustered based on the symmetry-corrected heavy-atom root mean square deviation (RMSD) algorithm implemented in AutoDock4 with an RMSD cutoff of 2.0 Å (32,110,123,124). Then, the topranked poses of the first and the most populated clusters were selected as the representative docking poses of each phytochemical in complex with each of four protein targets for further analyses.…”