JACS | How to get the “COVID-19 killer” – SARS-CoV-2 protease inhibitor through virtual screening

During the COVID-19 pandemic, drugs targeting SARS-CoV-2 may have saved millions of lives, and developing coronavirus inhibitors is now crucial. This article explores two virtual screening strategies for searching for inhibitors of SARS-CoV-2 major proteases in a large-scale chemical molecular library.
Firstly, using structure based docking, 235 million virtual compound libraries were screened for active sites. 100 top ranked compounds were detected in combination and enzyme analysis;
Secondly, under the guidance of millions of fine molecule docking and experimental testing of 93 compounds, the fragments discovered through crystallographic screening were optimized;
The crystal structure of the target inhibitor complex confirms the docking prediction and guided hit to read optimization, resulting in non covalent major protease inhibitors with nanomolar affinity, good in vitro pharmacokinetic characteristics, and broad-spectrum antiviral effects in infected cells.
01Research background
Since the beginning of 2020, drugs for treating coronavirus infections have been developed, and most of these compounds (such as remdesivir and hydroxychloroquine) have little effect on mortality or hospital stay. The major protease (M pro) has become a promising target in the proteins encoded by the SARS-CoV-2 genome. Inhibiting Mpro can block the processing of polymeric proteins produced by viral RNA translation, which is a necessary step in SARS-CoV-2 replication.
Targeting proteases is a successful strategy against human immunodeficiency and infection caused by hepatitis C virus, but due to the different structures and mechanisms of Mpro, it is necessary to develop new inhibitors targeting the coronavirus. Prior to the COVID-19 pandemic, several compounds targeting the coronavirus Mpro were identified through covalent (such as GC376) or non covalent mechanisms (such as ML188). However, non covalent scaffolds are peptide like drugs with poor pharmacokinetic properties, and covalent modulators typically require extensive optimization to regulate activity and selectivity.
This article explores two different strategies for searching for Mpro inhibitors based on structure docking in a super large chemical library. In the first screening, a library containing hundreds of millions of diverse lead like molecules was docked with the active sites of Mpro, and the top ranked compounds were selected for experimental evaluation. The second screening focus is to optimize the identified fragments in crystallographic screening by creating a focused library containing millions of compounds. Compare the most promising lead compounds with previously identified Mpro inhibitors (including clinical candidate drug PF-07321332).
Result (1): Screening of Mpro inhibitor super large molecular library
In the first virtual screening, 235 million compounds were docked with the crystal structure of Mpro, and Mpro was identified as a complex with substrate based inhibitor X77. X77 is an Mpro inhibitor that occupies all four major pockets of the active site (S1, S1 ‘, S2, and S3). Cluster based on topological similarity among 300000 top ranked compounds to identify a diverse set of candidate molecules. From 5000 clusters ranked high, 82 compounds were selected for experimental evaluation based on their complementarity with active sites. Select another group of 18 molecules from the top 3000 compounds, which form the same hydrogen bond with X77, namely with His163, Glu166, and Gly143. In the compound selection step, we also considered the contribution of ligand binding that was not included in the docking score function.
Figure 1. Overview of two virtual filtering methods
Table 1. Virtual screening hits, inhibitory potency, equilibrium constant, and structure of Mpro complex
Using a biosensor based on surface plasmon resonance (SPR), 100 selected compounds were detected in Mpro enzyme inhibition assay and direct binding interaction assay at three concentrations (5, 15-20, and 50 μ M).
19 compounds showed measurable dose-dependent binding in SPR biosensor screening, and 3 of them (compounds 1-3) were also found to inhibit Mpro in enzyme inhibition assays. Compound 1 has the highest inhibitory effect, with an IC50 value of approximately 40 μ M. The crystallographic binding patterns of the two inhibitors based on the hydantoin framework are very consistent with the predicted complexes obtained through molecular docking, with root mean square deviations (RMSDs) of 0.6 and 1.4 Å, respectively. The carbonyl group of hydantoin forms hydrogen bonds with the skeleton of residues Gly143 and Glu166, and the substituents on the core of hydantoin extend into the pockets of S2 and S1.
According to the docking score, compound 1 ranked in the top 0.002% of the chemical library, and the pyridine hydantoin scaffold was strongly enriched through virtual screening.
Figure 2. Confirmation of predicted binding modes through high-resolution crystal structure
Result (2): Fragment guided virtual screening of M pro inhibitors
The second screening was based on fragment guided virtual screening. In the crystal structure (PDB code: 5RF7), compound 4 occupied the S1 and S2 pockets compared to the larger inhibitor X77, but did not extend to S1 ‘or S3. In the SPR biosensor experiment, the fragment showed a super stoichiometric binding with Mpro (KD>200 μ M), and had no effect on enzyme activity at 50 μ M. Based on visual analysis of the complex, we have designed chemical patterns that can include key features of fragment polarity interactions, as well as the possibility of placing growth carriers in S1 ‘and S2 pockets.
In the SPR biosensor experiment, 21 fragments showed dose-dependent binding, of which 5 also inhibited Mpro activity at 50 μ M. For the 5 confirmed matches in the two experiments, the KD values measured by SPR biosensors ranged from 7.2 to 79 μ M. The IC50 of compound 5 in the enzyme assay is 20 μ M, and the KD in the SPR assay is 7.2 μ M.
Result (3): Structural guided inhibitor optimization
The optimization of scaffold hits represented by compounds 1 and 3 is guided by the crystal structure of these inhibitors that bind to Mpro. Due to the excellent hydrogen bonding complementarity between their common hydantoin scaffold and active sites, this chemical type is maintained.
We systematically explored the changes in the pockets of S1 and S2 in several cycles to determine the optimal combination of substituents. In the S2 pocket, the optimal substituents (compounds 14 and 15) were identified from the spirocyclic analogues of compound 3. Compared with simulated screening, the binding of the spiro indole moiety of compound 3 to pyridine (compound 13) in the S1 pocket did not enhance its potency. But crystallography supports the predicted binding mode of this inhibitor. The optimal substituents in the pockets of S1 and S2 are integrated into a single chemical series through internal synthesis, resulting in a synergistic improvement in inhibitor efficacy and affinity.
Compounds 16 and 17 have IC50 values of 0.46 and 0.33 μ M, respectively, and exhibit high affinity for Mpro (KD values of 0.14 and 0.15 μ M, respectively).
Antiviral effects in cell experiments and in vitro pharmacokinetic analysis
The antiviral effects of compounds 16, 17, and 19 were evaluated in SARS-CoV-2 infected cells. The three compounds exhibited dose-dependent inhibition of the cytopathic effect (CPE) with EC50 values of 1.7, 1.6, and 0.077 μ M, and showed no significant cytotoxicity at the highest test concentration in Vero E6 cells (50% cytotoxicity concentration, CC50>20 μ M). The antiviral activity was also confirmed in a yield reduction assay, which evaluated the inhibitory activity of the compounds on virus replication using RT qPCR. In these experiments, compounds 16, 17, and 19 inhibited SARS-CoV-2 virus replication with EC50 values of 1.0, 0.9, and 0.044 μ M, respectively.
To further evaluate the potential of compound 19 as a starting point for the development of broad-spectrum coronavirus drugs, we performed computational docking with homologous models of 29 reported Mpro active site mutants and SARS-CoV-1 and MERS-CoV variants in terms of crystal structure. The binding mode of compound 19 was maintained in all predicted complexes, indicating the activity of our inhibitor against other coronaviruses. Consistent with our computational modeling, compound 19 also inhibits the cytopathic effects of SARS-CoV-1 (EC50=0.39 μ M, Vero E6 cells) and MERS CoV (EC50=0.20 μ M, Huh7 cells).
Compound 19 also exhibits good in vitro ADME properties and good metabolic stability in the presence of human liver microsomes and binding to plasma proteins in human plasma. Consistent with its strong antiviral effect in cells, compound 19 exhibited high permeability in Caco-2 cell experiments and no efflux was observed.
02 summary
Firstly, molecular docking was performed on diverse and concentrated screening libraries to identify 8 M pro inhibitors, of which 5 predicted binding patterns were confirmed by crystallography;
Secondly, effective lead by lead optimization was achieved through crystal structure and search in billions of virtual compounds, resulting in three series of non covalent inhibitors;
Finally, the most promising lead compound has a nanomolar IC value of 50 in the Mpro enzyme assay, excellent in vitro ADME properties, and demonstrated strong antiviral effects against several coronaviruses in cell assays.