Repurposing Low-Molecular-Weight Drugs against the Main Protease of Severe Acute Respiratory Syndrome Coronavirus 2.

The coronavirus disease pandemic caused by infection with the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has affected the global healthcare system. As low-molecular-weight drugs have high potential to completely match interactions with essential SARS-CoV-2 targets, we propose a strategy to identify such drugs using the fragment-based approach. Herein, using ligand- and protein-observed fragment screening approaches, we identified niacin and hit 1 binding to the catalytic pocket of the main protease (Mpro) of SARS-CoV-2, thereby modestly inhibiting the enzymatic activity of Mpro. We further searched for low-molecular-weight drugs containing niacin or hit 1 pharmacophores with enhanced inhibiting activity, e.g., carmofur, bendamustine, triclabendazole, emedastine, and omeprazole, in which omeprazole is the only one binding to the C-terminal domain of SARS-CoV-2 Mpro. Our study demonstrates that the fragment-based approach is a feasible strategy for identifying low-molecular-weight drugs against the SARS-CoV-2 and other potential targets lacking specific drugs.

The coronavirus disease (COVID-19) pandemic caused by the severe acute respiratory syndrome coronavirus 2[1−3] (SARS-CoV-2) has so far affected >8 million people worldwide, with a mortality rate over 5%. Main protease (Mpro or 3CLpro) is one of the most extensively studied targets of coronaviruses.[4] Mpro plays an essential role in the cleavage of viral RNA-translated virus polypeptide[5] and recognizes at least 11 cleavage sites in replicase polyprotein 1ab, e.g., LQ↓SAG (↓denotes the cleavage site). Covalent inhibitors against the SARS-CoV-2 Mpro have recently demonstrated potency toward inhibiting viral replication in cellular assays;[6,7] this further underpins the druggability of Mpro. However, these compounds remain in the early stages of preclinical studies, and the development of new drugs usually takes years. The lack of drugs targeting SARS-CoV-2 currently poses a threat to numerous COVID-19 patients.

The COVID-19 pandemic has necessitated the repurposing of oral drugs.[8] As most recently approved drugs have been designed and optimized for specific targets, they are unlikely to completely match interactions with the SARS-CoV-2 targets. Compared with 13 550 potential drugs in the DrugBank database at various stages from preclinical studies through approval, the estimated number of druglike compounds (molecular weight of ∼500 Da) is reportedly approximately 1060. Therefore, the possibility of uncovering a highly potent and specific drug against SARS-CoV-2 is quite slim. Conversely, low-molecular-weight drugs with intermediate potency and high safety can be an alternative treatment against SARS-CoV-2. The toxicity of many low-molecular-weight drugs has been well understood owing to long clinical trials. Furthermore, their low structural complexity increases the odds of fully matching the interactions with anti-SARS-CoV-2 targets; for example, the chemical space of compounds with <11 non-hydrogen atoms is approximately 109. This is the cornerstone of fragment-based lead discovery,[9] and many of the compounds in the fragment library were indeed extracted from pharmacophores of approved drugs. We therefore hypothesize that it is highly possible to identify a low-molecular-weight drug containing pharmacophores using fragment-based screening (FBS).

We therefore compared 3508 compounds in our fragment library[10−15] with repurposed drugs from virtual screening, generously released by Prof. Hualiang Jiang at Chinese Academy of Sciences Shanghai Institute of Materia Medica. If these candidates did bind at a high affinity as predicted, their pharmacophores should bind as well, albeit at a weaker affinity. A total of 38 compounds (Table S1) as pharmacophores (substructures) of these repurposed drugs were thus screened against the SARS-CoV-2 Mpro (residue 4-306). These weak binders were readily identified using a nuclear magnetic resonance (NMR) fragment-based approach,[16−18] e.g., the ligand-observed spectra of saturation transfer difference (STD) and WaterLOGSY (Figure a). Three hits of the SARS-CoV-2 Mpro were identified: niacin, hit 1, and hit 2 (Figure b). The hit rate of 8% was slightly higher than that of our FBS against other targets,[19,20] which suggests that the success rate can be enhanced by virtual screening a priori. The remaining 35 fragments demonstrated no detectable binding, probably because of the distracting false positives in virtual screening. The potency of the three hits was then evaluated using enzymatic activity assay of the SARS-CoV-2 Mpro (Figure c). Niacin and hit 1 moderately inhibited the cleavage of fluorescent-labeled polypeptide (FITC-AVLQSGFR-Lys(Dnp)-Lys-NH2) by the SARS-CoV-2 Mpro.

Figure 1

Fragment-based screening identified three hits of the SARS-CoV-2 main protease. (A) NMR ligand-observed spectra of three representative hits in the presence of 10 μM full-length SARS-CoV-2 Mpro. (B) Chemical structures of the three hits. (C) Inhibition of the enzymatic activity of the SARS-CoV-2 Mpro (5 μM) by the three hits (4 mM). The negative control was treated using fluorescence-labeled peptide (16 μM) in the absence of Mpro.

To further map the binding sites of niacin and hit 1, we determined the chemical shift perturbations (CSPs) of 15N-labeled Mpro induced by the titration of these two compounds. However, the severe signal overlap was observed in the heteronuclear single-quantum correlation (HSQC) spectrum of the 15N labeled full-length SARS-CoV-2 Mpro. The N-terminus (residues 1-199)[21] and the C-terminus (187-306)[22] of SARS-CoV Mpro were separately studied by NMR spectroscopy, and the free-form crystal structures (PDB codes: 2QCY, 4HI3, and 3VB3) reveal remarkable interdomain plasticity of the wild-type or R298A mutant of SARS-CoV Mpro.[23−25] Considering the high sequence identity of 96% between the main proteases of SARS-CoV and SARS-CoV-2, the N-terminal domain (Mpro-N) with the catalytic core included (residues 4-199) was used with a well-dispersed HSQC spectrum. It thus enabled many 1H–15N amide chemical shift assignments transferred directly from SARS-CoV Mpro-N. Key residues proximal to the catalytic site, including H41, V42, D48, N51, G143, H163, and V186, were thus assigned (Figure S1). Both niacin and hit 1 perturb a common residue V42, suggesting that these two hits bind to the catalytic core of SARS-CoV-2 Mpro. Interestingly, these two hits also recognize different sets of residues in the catalytic core; for example, niacin perturbs H41, G143, N51, and V186 (Figure a), whereas hit 1 induces CSPs of residues M165, E166, and L167 (Figure b). Mapping of these residues to the surface representation of the crystal structure of the SARS-CoV-2 Mpro (PDB code: 6LU7)[26] suggested that these two hits adopted different orientations in the catalytic site, with a shared anchor point near V42 (Figure c). Considering the molecular size of niacin and the spatial distribution of the perturbed residues, niacin probably binds more than one site. Nevertheless, the CSP pattern suggests that the catalytic core of the SARS-CoV-2 Mpro accommodates compounds larger than niacin and hit 1.

Figure 2

Binding topology of niacin and hit 1 determined from NMR chemical shift perturbations. (A) Superimposition of 2D 1H–15N HSQC spectra of the SARS-CoV-2 Mpro-N in the absence and presence of niacin. The ligand/protein molar ratios are shown. (B) Chemical shift perturbations induced by hit 1. (C) Chemical shift perturbations induced by niacin (red), hit 1 (blue), or both (green) were mapped to the surface of the crystal structure of the SARS-CoV-2 Mpro (PDB code: 6LU7).

We therefore searched for low-molecular-weight (<400 Da) drugs containing the pharmacophores of niacin and hit 1. As a niacin derivative, carmofur induced an extra set of cross peaks at a ligand/protein molar ratio of 2:1, and some original signals completely disappeared at a molar ratio of 4:1 (Figure S2a). This indicated a strong binding between carmofur and SARS-CoV-2 Mpro as validated by the IC50 of 2.8 ± 0.2 μM determined using enzymatic assay at an Mpro concentration of 0.5 μM (Figure S2b). However, the original NMR signals of SARS-CoV-2 Mpro-N completely disappeared at a ligand/protein molar ratio that significantly deviated from a stoichiometry of 1:1. It has been recently demonstrated that carmofur is a covalent inhibitor of the main protease of the SARS-CoV-2, with an IC50 value of 1.82 μM in the presence of 0.2 μM enzyme;[26] this was consistent with our measurement as a higher Mpro concentration was used in our case. Collectively, these data suggest that covalent linking to the SARS-CoV-2 Mpro is driven by excess carmofur. Nevertheless, using this fragment-based approach is a feasible strategy for repurposing low-molecular-weight drugs targeting SARS-CoV-2 Mpro.

Further pharmacophore identification and molecular docking nominated several low-molecular-weight analogues of hit 1, for example, triclabendazole, emedastine, and bendamustine (Figure S3a). The single-dose enzymatic assay showed that these three drugs had significantly higher potency than hit 1 (Figure S3b). We further determined the dose-dependent response of bendamustine and emedastine in the inhibition of the SARS-CoV-2 Mpro activity, with IC50 values of 26 ± 1 and 82 ± 7 μM (Figure a). The IC50 value of triclabendazole was roughly estimated to be 70 μM from the two-dose inhibition rates (31% and 72% inhibition at 50 μM and 100 μM triclabendazole, respectively), as limited by the low aqueous solubility of triclabendazole. Further, bendamustine and emedastine induced significantly larger CSPs in a dose-dependent manner than hit 1 (Figure b).

Figure 3

Potency and binding topology of low-molecular-weight drugs as derivatives of hit 1. (A) Dose-dependent inhibition of the enzymatic activity of the SARS-CoV-2 Mpro (0.5 μM) by bendamustine and emedastine in the presence of 16 μM fluorescent-labeled substrate. (B) Chemical shift perturbations of 15N-labeled SARS-CoV-2 Mpro-N induced by bendamustine and emedastine at the annotated ligand/protein molar ratio.

In the process of searching for effective low-molecular-weight drugs targeting SARS-CoV-2 Mpro, omeprazole as a hit 1 analogue was uncovered capable of inhibiting the protease activity with IC50 values of 283 ± 24 μM (Figure A). However, omeprazole did not induce any detectable CSPs of the SARS-CoV-2 Mpro-N (Figure S4).We hence titrated omeprazole to the 15N-labeled C-terminal protease (Mpro-C, residues 187-306) of SARS-CoV-2. Benefiting from the well-dispersed HSQC spectrum of the SARS-CoV-2 Mpro-C and high sequence identity with the SARS Mpro, we may transfer many 1H–15N amide chemical shift assignments directly (Figure S5)20. The omeprazole-induced CSPs (Figure B) suggests that omeprazole binds to the C-terminus instead of the catalytic N-terminus of SARS-CoV-2. The docking pose of omeprazole (Figure S6) suggests that it deviates over 17.6 Å from the known N3 covalent inhibitor in the N-terminus of SARS-CoV-2 Mpro, which impedes the cross-linking of these two inhibitors. Conversely, omeprazole can be in combinational use with other hit 1 analogues, as the protease activity was mediated by binders in the N- and C-terminus of SARS-CoV-2 Mpro. We hence carried out the enzymatic assay of SARS-CoV-2 Mpro in the presence of omeprazole or emedastine individually, or both (Figure C). The data show that the inhibition of the enzymatic activity of SARS-CoV-2 Mpro is additive; that is, cocktails can in principle be used at a lower dose of each component. Thus, less toxicity was expected.

Figure 4

Omprazole suppressed the enzymatic activity via binding to the noncatalytic C-terminus of SARS-CoV-2Mpro. (A) Dose-dependent inhibition of the enzymatic activity of the SARS-CoV-2 Mpro (0.5 μM) by omeprazole. (B) Chemical shift perturbations of 15N-labeled SARS-CoV-2 Mpro-C induced by omeprazoleat the annotated ligand/protein molar ratio. (C) Enzymatic activities of the SARS-CoV-2 Mpro (0.5 μM) inhibited by omeprzole or emedastine individually, or by both.

Taken together, our fragment-based strategy facilitates the identification of low-molecular-weight drugs against the SARS-CoV-2 Mpro. First, this approach can be readily applied to identify low-molecular-weight drugs against other SARS-CoV-2 targets (e.g., RNA-dependent RNA polymerase or the receptor-binding domain of the spike protein). Second, a combination of these low-molecular-weight drugs may be used to gain higher potency than that achieved via a single compound if their binding topologies show no evidence of steric repulsion. Finally, although carmofur and bendamustine show higher potency than the fragment screening hits, the toxicity of anticancer drugs remains a challenge in their clinical applications. Conversely, triclabendazole, emedastine, or omeprazole could be valuable in inhibiting the SARS-CoV-2 replication at the early stage. These compounds may serve as a new starting point for the next round of drug discovery, as they contain pharmacophores distinct from the published covalent or peptidomimetic inhibitors.[6,7,26−28] In general, our study provides new insights toward the repurposing of low-molecular-weight drugs against the SARS-CoV-2 and other potential targets lacking specific drugs.