Potential Broad Spectrum Inhibitors of the Coronavirus 3CLpro: A Virtual Screening and Structure-Based Drug Design Study

Abstract

Human coronaviruses represent a significant disease burden; however, there is currently no antiviral strategy to combat infection. The outbreak of severe acute respiratory syndrome (SARS) in 2003 and Middle East respiratory syndrome (MERS) less than 10 years later demonstrates the potential of coronaviruses to cross species boundaries and further highlights the importance of identifying novel lead compounds with broad spectrum activity. The coronavirus 3CL(pro) provides a highly validated drug target and as there is a high degree of sequence homology and conservation in main chain architecture the design of broad spectrum inhibitors is viable. The ZINC drugs-now library was screened in a consensus high-throughput pharmacophore modeling and molecular docking approach by Vina, Glide, GOLD and MM-GBSA. Molecular dynamics further confirmed results obtained from structure-based techniques. A highly defined hit-list of 19 compounds was identified by the structure-based drug design methodologies. As these compounds were extensively validated by a consensus approach and by molecular dynamics, the likelihood that at least one of these compounds is bioactive is excellent. Additionally, the compounds segregate into 15 significantly dissimilar (p < 0.05) clusters based on shape and features, which represent valuable scaffolds that can be used as a basis for future anti-coronaviral inhibitor discovery experiments. Importantly though, the enriched subset of 19 compounds identified from the larger library has to be validated experimentally.

Keywords: 3CLpro; human coronaviruses; molecular docking; molecular dynamics; structure-based drug design; virtual screening.

Figure 1

The pose of an experimental ligand, ML188, as predicted by Vina and Glide. (a) Experimental orientation of ML188 (green) with overlaid pose predicted by Vina (blue); (b) Experimental orientation of ML188 (green) with overlaid pose predicted by Glide (blue). An root-mean-square deviation (RMSD) of 0.244 Å for Vina and 0.79 Å for Glide suggests that they are well calibrated for this class of enzyme and should achieve a high level of accuracy in predicting the pose of novel ligands.

Figure 2

Division of the nineteen candidate ligands into fifteen distinct 3D clusters, as determined by the PubChem (https://pubchem.ncbi.nlm.nih.gov) structure-clustering web-service. 3D Tanimoto similarity scores were calculated from alignments generated using shape + feature and ten conformers per ligand. The dendogram was generated using similarity score of 1.04, which is recommended to achieve statistical significance at the 95% confidence interval.

Figure 3

Pose of the ZINC27332786 ligand within the active site of the 3CL^pro. Representation of elements mediating hydrogen bonds with respective residues and the P₁ pyridine, hydrophobic P₂ and P₄ and a 1,2,4-triazole linker.

Figure 4

Pose of the ZINC09411012 ligand within the active site of the 3CL^pro. Representation of elements mediating hydrogen bonds with respective residues and the P₁ pyridine, hydrophobic P₂ and solvent exposed dioxolane.

Figure 5

Destabilization of ZINC09411012 from the 2ZU2 active site. (a) Starting conformation where the pyridine fills the S₁ pocket and forms a hydrogen bond with His162; (b) Initial destabilization resulting in the break of the hydrogen bond with His162 and subsequent formation with Ala143; (c) Further destabilization resulting a new hydrogen bond with Gly142; (d) Complete loss of orientation.