Discovery of M Protease Inhibitors Encoded by SARS-CoV-2

The coronavirus (CoV) disease 2019 (COVID-19) pandemic caused by severe acute respiratory syndrome CoV-2 (SARS-CoV-2) is a health threat worldwide. Viral main protease (Mpro, also called 3C‐like protease [3CLpro]) is a therapeutic target for drug discovery. Herein, we report that GC376, a broad-spectrum inhibitor targeting Mpro in the picornavirus-like supercluster, is a potent inhibitor for the Mpro encoded by SARS-CoV-2, with a half-maximum inhibitory concentration (IC50) of 26.

of an effective antiviral treatment (8). Among all the mature structural or nonstructural proteins in SARS-CoV-2, M pro is the most conserved target region within the whole viral genome (9). Due to the severity of SARS-CoV-2 infection, it is important to emphasize drug discovery for SARS-CoV-2 based on existing drugs for immediate uses or an expedited development timeline.
We previously discovered several small-molecule inhibitors for SARS-CoV during the SARS outbreak in 2003 (10). The M pro encoded by SARS-CoV-2 represents a key target for anti-SARS-CoV-2 strategies. However, to date, a promising SARS-CoV-2 M pro protease inhibitor has been lacking. Herein, we establish the SARS-CoV-2 M pro protease fluorescence-based assay to screen for potential inhibitors. Furthermore, molecular modeling studies were carried out to further demonstrate the interaction of M pro with GC376. GC376 or its optimized analogues hold great promise to be developed in humans with SARS-CoV-2 infection, alone or together with other antiviral drugs.

RESULTS
Fluorescence resonance energy transfer (FRET)-based screening assays. The M protease (M pro ) encoded by the SARS and SARS-CoV-2 coronaviruses differ in only 12 amino acid residues. According to our previous experience during the SARS outbreak, SARS-CoV-2 M pro is expressed as a glutathione S-transferase (GST) fusion protein in Escherichia coli (11,12). The GST fusion protein was purified by glutathione affinity chromatography. The fusion protein was cleaved by factor Xa, resulting in the generation of a prominent protein band with an apparent molecular mass of 35 kDa, the mature SARS-CoV-2 M pro (see Fig. S1c in the supplemental material).
Purified M pro proteins were checked for the proteolytic activity that cleaves the EDANS-KTSAVLQSGFRKME-DABCYL substrate, where EDANS is 5-((2-aminoethyl)amino) naphthalene-1-sulfonic acid and DABCYL is 4-(dimethylaminoazo)benzene-4-carboxylic acid (Fig. S2a). The purified enzyme was assayed using the FRET technique as described in Materials and Methods. The performance of the FRET assay was assessed, and the signal-to-noise ratio was determined to be Ͼ20 (Fig. S2b). The Z-factor value for the assay was 0.9, which corresponds to a valid screening system. The M pro of feline infectious peritonitis virus (FIPV) was prepared similarly. The same fluorogenic substrate, EDANS-KTSAVLQSGFRKME-DABCYL, was equally applicable in activity assessments for FIPV M pro .
Inhibitory activities of SARS-CoV-2 M pro by the Zinc ion, GC376, and lopinavir. In this study, based on the SARS-CoV-2 M pro activity assay, we screened a collection of protease inhibitors, such as reported M pro inhibitors for relevant viruses and clinically approved human immunodeficiency virus (HIV) protease inhibitors. At 10 M, GC376 (Fig. 1A), a broad-spectrum antiviral protease inhibitor used to treat cats with FIPV infection (13), showed complete inhibition of SARS-CoV-2 M pro activity. Since GC376 was well characterized for its inhibition of M pro encoded by FIPV, we conducted a head-to-head comparison for the inhibitory activity of GC376 on SARS-CoV-2 M pro and FIPV M pro . SARS-CoV-2 and SARS-CoV share high identity in amino acid sequence (Fig.  S3a), whereas SARS-CoV-2 M pro and FIPV M pro share 45% identity in amino acid sequence (Fig. S3b). These two viral proteases also share similar folding and crystal structures (14,15). In this study, we found that GC376 is an extremely potent inhibitor of the M pro encoded by SARS-CoV-2, with a half-maximum inhibitory concentration (IC 50 ) of 26.4 Ϯ 1.1 nM (Fig. 1B). Subsequent analysis showed that GC376 is a competitive inhibitor of the M pro from SARS-CoV-2, with a binding constant (K i ) of 12 Ϯ 1.4 nM (Fig. 1C). In contrast, the IC 50 and K i of GC376 toward FIPV M pro are 118.9 Ϯ 1.1 nM and 42.5 Ϯ 2.9 nM, respectively ( Fig. S4a and b).
To examine whether a covalent adduct is formed, SARS-CoV-2 M pro incubated with GC376 was subject to mass spectrometry (MS) analysis in accordance with a method described previously. Indeed, a gain of 403.2 Da in mass was observed in new peaks from GC376-incubated SARS-CoV-2 M pro , indicating the same mechanism for adduct formation as described previously (13; also results not shown). Through the mass spectrometry analysis, we observed a new MS peak with a mass value of 34,194.0 Da, which is equal to the dihydrogen molecular weight of M protease (33,790.8), reflecting conjugation with only one GC376 molecule. Even in an excess of GC376, only 30% of the M pro enzyme was conjugated based on the peak intensity. With the X-ray (NCBI Protein Data Bank accession number 7BRR; release date, 13 May 2020) and the MS analyses performed in this study, it is evident that GC376 forms a covalent bond with Cys145 of M pro . That only a small portion of SARS-CoV-2 M pro was covalently modified in a 25:1 molar excess of GC376 indicates that improved inhibitors are needed.
All the HIV protease inhibitors, including lopinavir, ritonavir, fosamprenavir, saquinavir, nelfinavir, atazanavir, darunavir, amprenavir, tipranavir, and indinavir, showed no inhibitory activity at 20 M, reflecting the fact that no benefit was observed with lopinavir-ritonavir treatment in patients with severe COVID-19 (16). Since Zn 2ϩ was shown to inhibit 3CL pro encoded by SARS-CoV (17), ZnCl 2 and ZnSO 4 were evaluated for their activity against SARS-CoV-2 M pro in this study. Zinc salts have been shown to To confirm that GC376 inhibited SARS-CoV-2 replication and cellular toxicity in cell culture, GC376 was tested for inhibition of SARS-CoV-2 infection in Vero E6 cells with 100 50% tissue culture infectious doses (TCID 50 ) per well in 96-well plates. SARS-CoV-2-infected cells were treated with increasing concentrations of GC376, and protection from cytopathic effects (CPE) was visually observed. GC376 dose dependently showed a reduction of the viral CPE (Fig. S5). After the cells were stained with crystal violet, their optical density at 570 nm (OD 570 ) was measured ( Fig. 2A). The results showed that GC376 inhibited SARS-CoV-2 infection, with an EC 50 of 0.91 Ϯ 0.03 M (Fig. 2B). GC376 exhibited a broad-spectrum antiviral activity against several coronaviruses in various cell lines (13). To evaluate whether GC376 was cytotoxic to cells, Vero E6 cells were treated with different concentrations of GC376 up to 100 M, and cell viability was determined using the 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assay. We found that GC376 did not show cytotoxicity in Vero E6 cells up to 100 M (Fig. 2B). Hence, we concluded that the selectivity index (SI) of GC376 was Ͼ114.
Molecular docking. To inform lead optimization efforts starting from GC376, in silico calculations to correlate IC 50 and K i into binding energy between GC376 and M pro were attempted in accordance with, in part, our previous work (18)(19)(20). By our calculation, the free binding energies of GC376 with SARS-CoV-2 M pro and FIPV M pro are Ϫ51.59 kcal/mol and Ϫ32.42 kcal/mol, respectively. As shown in Fig. 3A and B, upon removal of the bisulfite group, the compound is converted to an aldehyde form, giving rise to a covalent bond with catalytic Cys145. This result is in congruence with the cocrystal structures of GC376 and MERS M pro , where GC376 forms a covalent bond with Cys148 (31). In Fig. 3C, the amino acid residues on the inner surface of the substrate binding sites within FIPV M pro and SARS-CoV-2 M pro are well conserved. Only two sites of amino acid residues are different between the two M pro binding pockets. In SARS-CoV-2, the Gln189 on the surface of the binding pocket of SARS-CoV-2 M pro supports a H bond with the carbamate moiety of GC376. In contrast, this H bond cannot be formed because the counterpart residue in FIPV M pro is Pro188, rather than Gln. It appears that due to the covalent binding with Cys145 and hydrogen binding with Gln189 within the substrate binding pocket of SARS-CoV-2 M pro , GC376 was induced to bind more snugly into the pocket through a strong H bond network with Phe140, Gly143, Ser144, Cys145, His163, His164, Glu166, and Gln189 (Fig. 3A). In contrast, GC376 forms only a weaker H bond network with Gly142, His162, and Glu165 (Fig. 3B). The other different sites are Ser144 in SARS-CoV-2 M pro and Thr143 in FIPV M pro . These differences had little influence on the binding network. The root mean square deviation (RMSD) between the two docking conformations of GC376/SARS-CoV-2 and GC376/ FIPV is 1.16 Å. Importantly, our in silico prediction has informed our potential direction to improve GC376 with respect to its potency and further drug-like properties. With the docking analyses, GC376 may be improved by replacing the benzene group with H bond donors to interact with Glu166. The other alternative for improvement of binding potency is to replace the isobutyl group with moieties of a less bulky hydrophobic group so as to form interactions with Met49 (Fig. 3C). R. J. Hussey et al. have reported that the Michael acceptor inhibitor, acetyl-Glu-Phe-Gln-Leu-Gln-CHϭCHCOO-, forms a covalent bond with catalytic Cys139 in norovirus 3CL pro (22), suggesting an alternative avenue for optimization of GC376. After the submission of this paper, the crystal structure of the 3CL protease complexed with GC376 (accession no. 7BRR; release date, 13 May 2020) became available in the NCBI Protein Data Bank. When we compared the cocrystal X ray with our docked M pro -GC376 model based on 6LU7, the RMSD was 0.74, indicating that the in silico docking approach employed in this study adequately predicted the real complexed structure before its availability. In Fig. S6, the modeled conformation is shown in green and aligned to the X-ray result in pink.

DISCUSSION
To date, no proven effective therapy has been shown to be effective for SARS-CoV-2 infection (9). As of the submission date of this paper, the once-promising medicines, including remdesivir and hydroxychloroquine, are facing challenges after more stringently controlled observations and trials (23,24). When coronaviruses replicated inside cells, cellular innate immunity was shown to be compromised by M pro (25). We have previously shown that the compromised interferon (IFN)-mediated antiviral mechanism of viral 3C pro of enterovirus 71 can be rescued by effective protease inhibitor (26). Thus, effective inhibition of viral protease may not only restrict virus replication but also prevent interruption of the antiviral IFN pathway.
We also found that GC376 is a promising M pro inhibitor for SARS-CoV-2. GC376 is a dipeptidyl bisulfite adduct salt with excellent inhibitory activity against several picornaviruses and coronaviruses (13,27,28). Administration of GC376 leads to a full recovery in laboratory cats with FIPV infection, a highly fatal feline disease (29). Y. Kim in 2016 also studied the pharmacokinetic properties and the safety of GC376 in laboratory cats. In their safety study of GC376, no adverse effects were observed and no Antimicrobial Agents and Chemotherapy changes in clinical lab parameters were reported in cats subcutaneously given GC376 at 10 mg/kg of body weight/dose twice a day for 4 weeks (29). In this safety study, the plasma drug concentrations were shown to remain slightly above 1,000 ng/ml (i.e., ϳ2,000 nM, as the molecular weight of GC376 is 507.53), which was well above the concentrations needed for effective inhibition of SARS-CoV-2 as observed in this study. Therefore, the existing pharmacology and efficacy data for GC376 as an investigational drug in cats with FIPV infection encourage a proof-of-principle study of COVID-19 patients and then of the in vitro and in vivo antiviral activities of GC376 or furtheroptimized analogues.
(BIOVIA, Inc., San Diego, CA). The detailed method of LigandFit has been described previously (30). To illustrate the binding interactions, GC376 was docked to the binding site. The binding pocket was identified from the MERS and GC376 cocrystal structures (NCBI Protein Data Bank accession no. 5WKJ) (31). The force field for calculating ligand-receptor interaction energies employed the piecewise linear potential 1 (PLP1). The rectangular grid was set at 0.5 Å spacing, and the extension from the site was set as 8 Å. The number of docking poses was set as 50 with default parameters. The docking root mean square (RMS) threshold for ligand site matching was set as 5 Å. The method of steepest descent for rigid-body minimization during pose docking was used. The covalent docking calculation was performed using the two-point attractor method by the AutoDock Tools (version 1.5.6) as described previously (17). The decision of the best pose was based on the similar conformations of the MERS complex cocrystal structure. Molecular weight analysis using MS. The premixed GC376 compound (1 l, 10 mM in DMSO) and SARS-CoV-2 M pro (5 l, 2.5 mg/ml) were incubated at 25°C for 30 min. Subsequently, 10 l of the reaction mixture was transferred into 40 l of the infusion solution (50% acetonitrile in 0.1% formic acid) for measuring the molecular weight using a quadrupole time of flight (QTOF) mass spectrometer (G1; Waters) through the direct infusion model. The ion signal (m/z) was acquired in positive-ion mode with a capillary temperature of 100°C and an electrospray voltage of 2,800 V in the scan range from 800 to 2,500 m/z. Mass deconvolution was performed using Waters MassLynx (v4.1) software using the MaxEnt 1 program with a half-height of 0.1 Da and a maximum number of interactions of 100.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 1.3 MB.

ACKNOWLEDGMENTS
This work was financially supported by a special government fund to tackle the COVID-19 pandemic from the Ministry of Health and Welfare, Taiwan, and a 5-year grant for the Research Center for Emerging Viral Infections (Chang Gung University) from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) and the Ministry of Science and Technology (MOST) in Taiwan.
The team also thanks Huai-Tzu Chang for coordination and facilitation on the project.
We declare that we have no competing interests.