Structure-Based Discovery of Novel Nonpeptide Inhibitors Targeting SARS-CoV‑2 Mpro

ABSTRACT: The continual spread of novel coronavirus disease 2019 (COVID-19) is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), posing a severe threat to the health worldwide. The main protease (Mpro, alias 3CLpro) of SARS-CoV-2 is a crucial enzyme for the maturation of viral particles and is a very attractive target for designing drugs to treat COVID-19. Here, we propose a multiple conformation-based virtual screening strategy to discover inhibitors that can target SARS-CoV-2 Mpro. Based on this strategy, nine Mpro structures and a protein mimetics library with 8960 commercially available compounds were prepared to carry out ensemble docking for the first time. Five of the nine structures are apo forms presented in different conformations, whereas the other four structures are holo forms complexed with different ligands. The surface plasmon resonance assay revealed that 6 out of 49 compounds had the ability to bind to SARS-CoV-2 Mpro. The fluorescence resonance energy transfer experiment showed that the biochemical half-maximal inhibitory concentration (IC50) values of the six compounds could hamper Mpro activities ranged from 0.69 ± 0.05 to 2.05 ± 0.92 μM. Evaluation of antiviral activity using the cell- based assay indicated that two compounds (Z1244904919 and Z1759961356) could strongly inhibit the cytopathic effect and reduce replication of the living virus in Vero E6 cells with the half-maximal effective concentrations (EC50) of 4.98 ± 1.83 and 8.52
± 0.92 μM, respectively. The mechanism of the action for the two inhibitors were further elucidated at the molecular level by molecular dynamics simulation and subsequent binding free energy analysis. As a result, the discovered noncovalent reversible inhibitors with novel scaffolds are promising antiviral drug candidates, which may be used to develop the treatment of COVID-19.

Infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) will cause novel coronavirus disease 2019 (COVID-19),1 and the pandemic of the disease has rapidly become a global health concern2 and led to 160,074,167 confirmed cases and 3,325,260 deaths worldwide as of May 13, 2021.1 To cope with the severe crisis, great efforts have been paid to developing therapeutic approaches and vaccines against SARS-CoV-2.3,4 Discovering inhibitors of key proteins involved in the viral life cycle is an often-used and efficient approach to disrupt the replication of virus.5 Like SARS-CoV, the encoded 4 structural and 16 nonstructural proteins (NSPs) of SARS-CoV-2 provide multiple avenues to identify potential drug targets.6,7 Among the encoded proteins, the main protease (Mpro, alias 3CLpro), which has no human homolog, has become an attractive therapeutical target for the drug discovery and development of anti-COVID-19.8,9 Mpro belongs to the 16 NSPs of coronavirus (CoV) and is a vital enzyme that has an essential role in mediating the replication and transcription of CoVs.8 Together with papain- like proteases (PLPs), the enzyme processes the polyproteins that are translated from CoV RNA.10 Mpro is a highly conservative protein existing in all CoVs consisting of three domains (domains I to III).8 Crystal structures of SARS-CoV-2 Mpro (Figure 1)9,11 show that they are the chymotrypsin-like domain (domain I, residues 10 to 99), picornavirus 3C protease-like domain (domain II, residues 100 to 182), and a globular cluster formed by five helices (domain III, residues 198 to 303). The substrate-binding site (active site) of Mpro composed of four subsites (S1, S2, S3, and S4) is located at the six-stranded antiparallel β barrels between domains I and II.9 Based on the crystal structures of SARS-CoV or SARS-CoV-2 Mpro, computer-aided drug design techniques have been successfully used in anti-COVID-19 studies regarding the rapid discovery of potential inhibitors,12−16 drug repurposing,14,16−20 and making the action mechanism of the active compound against SARS-CoV-2 more understandable.21

Though these timely research studies have led to the design of several first-in-class SARS-CoV-2 Mpro inhibitors as promising drug candidates,8,9,11 currently no Mpro-based therapeutics have been officially approved for COVID-19.3 The need to develop novel as well as more effective antiviral drugs to inhibit SARS- CoV-2 has become more urgent.3 However, larger flexibility and figurability of active sites on SARS-CoV-2 Mpro proved to be a challenge for the rational design of small molecule inhibitors.22,23 For addressing this problem, the crystal structures of Mpro could be complemented by the all-atom molecular dynamics (MD) trajectory data released publicly in the spirit of open science.In the present work, based on nine different conformations of the SARS-CoV-2 Mpro substrate-binding site, a multiple conformational-based virtual screening strategy in combination with experimental validation was proposed to identify the enzyme inhibitors from a protein mimetics library with 8960 commercially available compounds (Figure 1). Considering the docking pose and scaffold diversity, 49 selected candidates were purchased for testing their binding profiles to SARS-CoV- 2 Mpro using the surface plasmon resonance (SPR) assay. The identified six compounds were further evaluated by the fluorescence resonance energy transfer assay (enzyme kinetics study) and bioluminescence resonance energy transfer (BRET) assay. All six compounds showed inhibition activities against the cell lines of SARS-CoV-2 Mpro. The live virus assay indicated that two out of the six inhibitors had the activity to interdict the viral infection of SARS-CoV-2. In addition, computational absorption-distribution-metabolism-excretion (ADME) analysis showed that the two inhibitors had good pharmacokinetic properties and low toxicity.

Compounds Selected through Ensemble Docking. To account for protein flexibility, nine Mpro ensembles including five MD-sampled apo structures and four holo structures (one homology model and three crystal structures in complex with different ligands) were collected. Meanwhile, a protein mimetics library (with 8960 compounds) from Enamine was prepared and resulted in 28,727 conformations. As an initial step, ensemble docking of the prepared library against the active site of the defined nine Mpro ensembles was performed to identify potential lead candidates. By considering the GlideScore, binding pose, and scaffold diversity profiles, the nine sets of hits from the ensemble docking were then used for selecting 50 top-ranked compounds (with 49 purchased) for experimental testing using SPR assays. As shown in Table S1, the selected compounds as potential Mpro inhibitors have GlideScore ≤−5.651 kcal/mol, and each of the compound
forms at least two hydrogen bonds with the residues located at the protease active site. The hierarchical clustering of the fingerprints using Tanimoto similarity and Ward’s cluster linkage method26 of the selected compounds shown in Figure S1 indicated the high diversity of the scaffolds. HPLC chromatograms and mass spectrograms were applied to verify the chemical structures and purity of the 49 compounds, and the data of the six active compounds (see the next section) are provided in the Supporting Information.Evaluation of Compounds as Inhibitors of SARS-CoV- 2 Mpro In Vitro. By using dipyridamole (DIP) as a positive control, the binding toward SARS-CoV-2 Mpro of the purchased 49 compounds was tested using the SPR assay at 100 μM concentrations (Figure S2). In addition, 6 out of the 49 compounds that have the abilities of binding to Mpro (Table 1) were selected to investigate whether their binding alters the biochemical function of the enzyme. The ranking of Glide- Score for the six compounds Z236230776, Z1244904919, Z225729516, Z1759961356, Z108564100, and Z106460362 was 42/50, 33/50, 19/50, 25/50, 11/50, and 1/50,respectively. There is only one compound (Z106460362) that was at the top 10 of the list (Table S1). Furthermore, we determined the biochemical half-maximal inhibitory concen- tration (IC50) values of the six chemical compounds, ranging from 0.69 to 2.05 μM (Figure 2).

All compounds presented a aThe resonance units (RU) of the SPR assay in the presence of each compound at a concentration of 100 μM. bThe nine SARS-CoV-2 Mpro structures including five apo forms (extracted per 2 μs from 10 μs MD simulation of 6LU711) and the four holo forms (one homology model using 3ATW33 as a template and three crystal structures 6LU7,11 6Y2F,9 and 6Y2G9 in complex with different ligands. cThe docking scores (kcal/mol) were calculated by the Glide extra precision algorithm strong inhibitory effect on Mpro activity, among which Z1759961356 (IC50 = 0.69 ± 0.05 μM) had the strongest effect (Figure 2). Consistent with the IC50 results, the BRET ratio showed that all compounds had a good inhibitory effect on Mpro in HEK293T cells (Figure 3). However, in this structure-based virtual screening study, although multiple conformation strategy was employed, the success rate was still very low (only 12% cases were correctly predicted by Glide docking). It is hypothesized that this is because those compounds were selected from specific conformations of the SARS-CoV-2 Mpro. However, according to the experimental test, the specific structure may not occupy the preferred conformation of the protease. Therefore, to increase the success rate of virtual screening, enhanced conformational sampling of the protease by state-of-the-art MD simulation is needed. In addition, the flexibility of the protease active site was not considered during each docking process, which was
crucial for the protein−ligand recognition. Therefore, the induced fit docking method may be used to address this
problem even if the calculation is time consuming.

Inhibitors Suppress SARS-CoV-2 Infection In Vitro. For examining whether these two lead candidates could prevent viral replication, further qRT-PCR and plaque-reduction assays were carried out in Vero E6 cells infected by SARS-CoV-2. As can be seen from Figure 4, quantitative qRT-PCR results showed that Z1244904919 and Z1759961356 exhibited a stronger effect on anti-SARS-CoV- 2 (Figure 4A,B). The plaque-reduction assay indicated that Z1244904919 and Z1759961356 displayed inhibitory effect on SARS-CoV-2, and the individual EC50 values were 4.98 ± and
8.52 ± μM, respectively (Figure 4C,D). Furthermore, the SPR assay showed that Z1244904919 and Z1759961356 bound to SARS-CoV-2 Mpro with Kd values of 465 and 133 μM, respectively (Figure 5A,B). In conclusion, these data suggest that the inhibition of Z1244904919 and Z1759961356 on Mpro is mainly achieved through direct binding to the enzyme active site. MD Simulation of the Inhibitor−Mpro Complex.
Though the two lead candidates were recognized by ensemble docking, we thought that their predicted binding modes in Mpro were not enough because the protease flexibility was not considered in each independent docking. To investigate inhibitor−Mpro interaction flexibility, 1 μs MD simulation was executed for sampling enough conformations of the two complexes. The root-mean-square deviation (RMSD) of the backbone atoms on protein and heavy atoms on the ligand referred to the starting structure was computed to reflect the stabilities of the studied systems during the period of simulation (Figure S3). The RMSD value variation suggested that the two complexes had small changes of conformation on the process of simulation. The average RMSD values of the binding site residues for Z1244904919 and Z1759961356 bound Mpro were 0.88 and 1.86 Å, respectively. The values for Z1244904919 and Z1759961356 were 1.09 and 1.96 Å. The trends of RMSD variation in Figure S3 indicated that the poses of ligands predicted were consistent with the active site of Mpro. In addition, we have compared the predicted poses of inhibitors Z1244904919 and Z1759961356 with positive control DIP in Mpro (Figure S4A,B). The results showed that the three ligands occupied the same binding site of the protease.

Binding Free Energy and Interaction Mode of Inhibitors in Mpro. On the account of the MD trajectories, the binding free energy of the two inhibitors bound to Mpro(ΔGcalc) was estimated using the MM/GBSA method.27 As shown in Table 2, the ΔGcalc for Z1244904919 and Z1759961356 bound to Mpro was −45.72 and −48.01 kcal/ mol, respectively. The variation trend of ΔGcalc values is compatible with the order of the experimental binding free energies (ΔGexp). The energy terms of ΔGcalc are listed in Table 2, indicating that the electrostatic (ΔEele) and hydrophobic (ΔEvdW + ΔGnonpol) interactions were of great importance for the binding of the four anticoagulants; however, polar solvent energies (ΔGpolar) were not conducive to the binding of inhibitors. In order to acquire a more particular understanding of the protein−ligand interaction, we decomposed the binding free energies into each residue. Residues with an absolute energy contribution of ≥0.5 kcal/ mol would be identified as key residues, which were conducive to the binding of inhibitors to the pocket; these key residues are displayed in Table S2. Meanwhile, the recognized key residues of the two complexes suggested that there was a certain degree of similar interactions between them. As shown in Table S2, a total of 14 and 13 residues in SARS-CoV-2 Mpro were identified to play an important role in Z1244904919 and Z1759961356 binding, respectively. Compared with the characterized interactions between the protease with the substrate28 and N3,29 10, 6, 7, and 5 common residues were found for Z1244904919- and Z1759961356-bound complexes (Figure S5), indicating that the key interactions between the protease pocket and ligands were maintained for the identified new nonpeptide inhibitors. Meanwhile, the superposition between SARS-CoV-2 Mpro in complex with N3 and Z1244904919 (Figure S4C) and ZN1759961356 (Figure 4D) indicates the overlap between the occupied pockets of these inhibitors, especially N3 and ZN1759961356.

The binding modes of Z1244904919 (Figure 5C) and Z1759961356 (Figure 5D) were investigated by the representative conformations extracted from the MD trajecto- ries. As is known, the active site of Mpro consists of four subpockets, which are S1, S2, S3, and S4.9 Residues Leu27, His41, Met49, His164, Met165, and Gln189 identified as key residues were of great importance for both Z1244904919-Mpro (Figure 5E) and Z1759961356-Mpro (Figure 5F) complexes. All the identified key residues uniformly distributed in the four subpockets of the Mpro active site (Figure 5C,D). Taking Z1244904919 as an example (Figure 5C), the backbone atoms of Gly143, Ser144, Cys145, and Asn166 interact with the compound via hydrogen bonds. The fluorophenol moiety of Z1244904919 embedded into the S1 site consisted of residues Phe140 and Asn166, and the piperidine moiety took up the S4 site containing residues M165 and Gln189, while the indole analogue moiety and linkages in contact with residues Met49, Thr25, and His41 located at the S2 and S3 sites. The piperidine moiety acts like a linker to connect fluorophenol and indole analogue motifs. Compared to Z1244904919, the higher binding abilities of the Z1759961356 may come from the energy contributions of residues His164 and Met165 in the S1 pocket and residue Asn47 in the S2 pocket of the Mpro active site. In this study, histidine (His41, His163, and His164) and cysteine (Cys145) located at the binding site of inhibitors were treated as neutral states during docking and MD simulation. However, it is important to note that the altering protonation states of titratable groups in histidine and cysteine in SARS-CoV-2 Mpro, which can modulate protein dynamics and stability, is important in virtual screening studies. This has been well studied in the recently published work by Pavlova et al.

In Silico Pharmacokinetic Analysis. The pharmacoki- netic properties of new lead candidates are essential for the development of an effective druggable molecule. Herein, the ADME properties of Z1244904919 and Z1759961356 were calculated in QikProp (v. 4.5) (Table 3). The QikProp method is based on 1700 known oral drugs, and the rms errors of its predictions are 0.5−0.6 log unit.31 The ADME properties of the recently reported two SARS-CoV-2 Mpro inhibitors (13a and 13b)9 and FDA approved drug DIP,32 which have favorable pharmacokinetic properties, were also calculated and are included in Table 3. The predicted ADME values of Z1244904919 and Z1759961356 compare favorably with the drug leads 13a and 13b or FDA approved drug DIP (Table 3). Moreover, some properties, such as QPPCacod and PercentHu- manOralAbsorption, are better than those of 13a, 13b, and DIP. Therefore, we are optimistic about the application of the two compounds Z1244904919 and Z1759961356 as new drug leads targeting the Mpro protein.

We report that the IC values of the six identified inhibitors then, all compounds were processed through generating tautomers, stereoisomers, and ionization states by Epik (Version 3.2).38 All the ligand preparations were under the condition of 7.0 ± 2.0 pH value with the OPLS3 force field.35 QikProp (Version 4.5) was used for calculating the five compounds’ ADME properties summarized in Table 3, and the library was prefiltered using druglike properties.Ensemble Docking. Screening the prepared library via docking them into the generated grids using Glide (Version 6.8).36,39 High-throughput virtual screening was first carried out for maintaining 10% top-ranked structures, and those molecules were redocked on the scoring algorithm of standard precision, reserving the 10% top-scored molecules. The resulting set was further filtered at the extra precision level, and a database involving 245 compounds was retained ultimately. From the retained sub-database, 50 compounds were selected by considering the docking scores, binding mode, and scaffold diversity. Finally, 49 compounds available commercially were purchased from TargetMol for further biological PF-00835231 evaluation.