Molecular mechanism of ensitrelvir inhibiting SARS-CoV-2 main protease and its variants

SARS-CoV-2 poses an unprecedented threat to the world as the causative agent of the COVID-19 pandemic. Among a handful of therapeutics developed for the prevention and treatment of SARS-CoV-2 infection, ensitrelvir is the first noncovalent and nonpeptide oral inhibitor targeting the main protease (Mpro) of SARS-CoV-2, which recently received emergency regulatory approval in Japan. Here we determined a 1.8-Å structure of Mpro in complex with ensitrelvir, which revealed that ensitrelvir targets the substrate-binding pocket of Mpro, specifically recognizing its S1, S2, and S1' subsites. Further, our comprehensive biochemical and structural data have demonstrated that even though ensitrelvir and nirmatrelvir (an FDA-approved drug) belong to different types of Mpro inhibitors, both of them remain to be effective against Mpros from all five SARS-CoV-2 variants of concern, suggesting Mpro is a bona fide broad-spectrum target. The molecular mechanisms uncovered in this study provide basis for future inhibitor design.


Results
Ensitrelvir is a potent inhibitor against WT SARS-CoV-2 M pro . Based on the fluorescence-resonance-energy transfer (FRET)based assay we previously established, the inhibition of ensitrelvir on SARS-CoV-2 M pro was performed to examine the enzyme kinetics of WT. The half-maximal inhibitory concentrations (IC 50 ) value of ensitrelvir against SARS-CoV-2 M pro is 0.049 μM (Fig. 1a), while the IC 50 of nirmatrelvir against WT is 0.044 μM, indicating that ensitrelvir exhibited as potent inhibition on viral protease as nirmatrelvir did in vitro.
The crystal structure of SARS-CoV-2 M pro in complex with ensitrelvir. In order to elucidate the molecular inhibition mechanism of ensitrelvir, we determined a 1.8-Å M pro -ensitrelvir complex structure (PDB: 8HBK). In the crystals, there is merely one protomer in an asymmetric unit, and all 1-301 residues could be traced in the electron density map. Two protomers form a functional dimer by a crystallographic twofold axis of symmetry. The overall protomer of M pro comprises three domains (Fig. 1b). Domain I (residues 8-101) and domain II (residues 102-184) possess an antiparallel β-barrel structure and take on a chymotrypsin-like fold, harboring the catalytic pocket between them. Domain III (residues 201-303) is a globular domain composed of five antiparallel α-helixes unique to M pro and is crucial for the dimerization of M pro , which is a prerequisite for its catalytic activity. The catalytic pocket containing the Cys-His dyad is located in the cleft between domain I and domain II (Fig. 1b). An ensitrelvir molecule could be identified in the substrate-binding pocket of each protomer, occupying S1, S2, and S1' subsites of M pro . In the crystal structure of dimeric M pro , the N-terminus of one promoter deeply inserts into the S1 subsite of the neighboring promoter and participates in the stabilization of the substrate pocket (Fig. 1c). In the S1 subsite, the Ser1 from the neighboring protomer stabilizes the pocket by forming four hydrogen bonds with E166 and F140. The 1-methyl-1H-1,2,4triazole group forms a hydrogen bond with the sidechain imidazole group of H163. In the S2 subsite, the 2,4,5-trifluoromethyl forms a π-π stack with the sidechain of H41. In the S1' subsite, the 6-chloro-2-methyl-2H-indazole part interacts with the NH of the mainchain of T26 through a hydrogen bond. In addition, H163, C145, G143, and Q189 are also involved in the hydrogen-bond network to stabilize the binding of ensitrelvir (Fig. 1d). Nirmatrelvir forms a covalent bond with C145 of M pro in the S1' subsite, whereas ensitrelvir forms no covalent bond in the substrate pocket. Instead, it interacts through hydrogen bonding with C145 and T26 in the S1' pocket. Furthermore, nirmatrelvir stabilizes in the S4 pocket through extensive hydrophobic interactions, whereas ensitrelvir has few interactions in this pocket. The interaction network in the S1 subsite is highly conserved for both inhibitors, as they both stabilize through hydrogen bonds with E166, F140, Ser1, and H172 in the S1 pocket ( Fig. 1e, f).
The G15S, K90R, and P132H substitutions have a limited impact on the structural and enzymatic properties of M pro . To elaborate on the impact of the G15S, K90R, and P132H mutations on the structure of M pro , the apo-form structures of G15S, K90R, and P132H were determined at 1.77, 1.66, and 1.82 Å resolution, respectively (Supplementary Table S1). There is only one protomer in an asymmetric unit in each crystal structure, and all 1-301 residues could be traced in the electron density map.
The mutant residues 15 and 90 are located in domain I and residue 132 is located in domain II. All these mutations are far away from the catalytic pocket ( Fig. 2a and Supplementary Fig. S3); thus, it is reasonable to speculate that the G15S, K90R, and P132H mutations do not directly affect the architecture of the catalytic pocket. This was further confirmed when comparing the G15S, K90R, and P132H structures with the M pro apo-form structure (PDB ID: 6Y2E). As shown in Fig. 2a, G15S, K90R, and P132H showed merely a slight difference from WT, with average rootmean-square deviation (RMSD) values of only 0.54 Å, 0.52 Å, and 0.58 Å, respectively, indicating that all these three mutations had a limited impact on the apo structure of M pro . It is worth mentioning that the mutation of the amino acid at position 15 from glycine to serine did not cause any serious structural changes to the backbone, except that the substitute of serine formed an additional hydrogen bond with a surrounding water molecule (Fig. 2b). The mutation of the amino acid at position 90 from lysine to arginine results in the breaking of a hydrogen bond from lysine to the water molecule but does not significantly alter the conformations of surrounding residues (Fig. 2c). Similarly, the P132H substitution did not change the architecture of the mainchain at this site or the residues in the vicinity except that the sidechain of E240 was slightly "pushed" away from residue 132. It is presumably caused by the steric hindrance of the histidine, and it is noticed that the sidechain of H132 formed an extra hydrogen bond with a water molecule (Fig. 2d).
To further validate whether these slight structural variations affect enzymatic efficiency, a previously established fluorescenceresonance-energy transfer (FRET)-based assay was used to compare the enzyme kinetic parameters of WT, G15S, K90R, and P132H. As shown in Fig. 2e, f, the catalytic efficiency (k cat /K m ) values were determined to be 2.66 × 10 4 M −1 s −1 for G15S, 3.05 × 10 4 M −1 s −1 for K90R, 2.64 × 10 4 M −1 s −1 for P132H, respectively, which are comparable to 2.71 × 10 4 M −1 s −1 for WT. It is indicated that the G15S, K90R, P132H, and WT have similar enzymatic kinetic parameters and the impact of the three substitutions on M pro could be neglected.
Ensitrelvir remains strong inhibition on SARS-CoV-2 M pro variants. Next, we tested the half-maximal inhibitory concentrations (IC 50 ) of ensitrelvir against SARS-CoV-2 M pro and variants. The results showed that ensitrelvir exhibited similar inhibition on WT and its variants with an IC 50 of approximately 0.04 μM (Fig. 3a). To obtain the structural basis for the inhibition of SARS-CoV-2 M pro and variants of the ensitrelvir, we eventually determined the structures of three variants individually with ensitrelvir for a total of three complex structures (Supplementary Table S1). Superimposition of the structures of all three complexes has shown that all three variant mutant sites (G15, K90, P132) are more than 20 Å away from the binding sites of ensitrelvir ( Fig. 2a and Supplementary Fig. S3), indicating that these changes in M pro between different SARS-CoV-2 variants may not affect the architecture of the substrate-binding pocket and thus would not impair the efficacy of the current compounds targeting M pro . Among the structures of mutant complexes of ensitrelvir, H163, E166, C145, G143, and T26 are involved in hydrogen bonding (Fig. 3b), and certain residues in the substratebinding pocket participate in stabilizing ensitrelvir, which has demonstrated a conservative binding mode among M pro from the WT and other variants. To summarize, all the data obtained above strongly support that M pro from SARS-CoV-2 WT and its variants have similar structural features and kinetic characters, and the G15S, K90R, and P132H substitutions do not impair the inhibition of ensitrelvir in vitro.
Nirmatrelvir is also a potent inhibitor against SARS-CoV-2 M pro variants. Nirmatrelvir is also an inhibitor targeting The crystal structure of dimeric M pro complex with ensitrelvir. c The interaction network of ensitrelvir in the substrate-binding pocket of M pro . Ensitrelvir is shown in green, M pro is shown in silver, and the residues of the neighboring protomer are labeled in yellow. Blue dotted lines represent hydrogen bonds, red dashed lines represent π-π stack, and red spheres represent water molecules. The substrate pocket is indicated separately. d Topology of the hydrogen-bond network between ensitrelvir and M pro . Hydrogen bonds are indicated by green dashed lines, and dashed boxes indicate interaction with the stable S1 pocket. e Band diagram of the WT and ensitrelvir complex crystal structure superimposed on WT bond nirmatrelvir. f Analysis of the residues involved in the interactions between ensitrelvir and nirmatrelvir in the substrate-binding pocket. Ensitrelvir is highlighted in green, while nirmatrelvir is highlighted in yellow.  SARS-CoV-2 M pro , which has just been fully approved to treat mild to moderate COVID-19 in adults at risk of severe infections. Next, we tested IC 50 values of nirmatrelvir against SARS-CoV-2 M pro and variants to see whether these mutations affect nirmatrelvir binding. The results showed that IC 50 of nirmatrelvir against WT was 0.044 μM while IC 50 of ensitrelvir against WT was 0.049 μM (Fig. 3c), suggesting that nirmatrelvir strongly inhibits WT M pro like ensitrelvir. In addition, the IC 50 values of nirmatrelvir against all variants are approximately 0.04 μM, indicating that these mutations did not obviously affect nirmatrelvir binding.
To obtain the structural basis for the inhibition of SARS-CoV-2 M pro and variants of nirmatrelvir, we eventually determined the structures of three variants individually with nirmatrelvir for a total of three complex structures (Supplementary Table S1). Superimposition of the structures of all three complexes has shown all three variant mutant sites (G15, K90, P132) were more than 20 Å away from the binding site of nirmatrelvir ( Fig. 2a and Supplementary Fig. S3), indicating that these changes in M pro would not impair the in vitro efficacy of nirmatrelvir. All three mutant M pro structures show little difference from that of WT, with an average RMSD of only 0.25-0.35 Å, and neither the binding pose of the compound nor the conformation of residues participating in drug binding exhibits significant differences. As shown in our previous analysis of the complex structure of WT with nirmatrelvir (PDB ID: 7VH8) 12 , the nitrile group of nirmatrelvir is attached to the Sγ atom of C145 through a standard 1.8 Å C-S covalent bond, the classical (S)-γ-lactam ring at the P1 position fits into the S1' subsite, and a hydrogen bond is formed between the oxygen atom of the lactam ring and the Nε2 atom of H163. In addition, the Oε1 atom of E166 interacts with the NH group to stabilize nirmatrelvir. The rigid dimethylcyclopropyl proline (DMCP) located at the S2 subsite is surrounded by extensive hydrophobic interactions. Most of the amino acids used to stabilize nirmatrelvir near the substrate pocket described above show great similarity in the complex structures of M pro (Fig. 3d), which may imply that nirmatrelvir may exhibit similar inhibition for WT M pro and other variants, which is consistent with the in vitro enzyme activity inhibitory of nirmatrelvir (Fig. 3a, c).

Discussion
In the 21st century, three previously unknown coronaviruses have spread globally, including severe acute respiratory syndrome (SARS) caused by SARS-CoV in 2003, Middle East respiratory syndrome (MERS) caused by MERS-CoV in 2012, and the current COVID-19 caused by SARS-CoV-2. During the first two rounds of the coronavirus outbreak, no approved targeted therapies, vaccines, or compounds were available for treatment 13 . However, for the COVID-19 pandemic, it is the first time newly developed numerous approved vaccines, targeted compounds, etc., that we can use to battle against the coronavirus 14,15 . Unfortunately, due to the long-term prevalence of the virus in the population worldwide and the error-prone nature of RNA viruses, an increasing number of SARS-CoV-2 variants have been reported worldwide, raising concerns about the effect of current vaccines and compounds. M pro has received much attention in the last three years as an ideal target for drugs against SARS-CoV-2. As the first noncovalent and non-peptidomimetic drug candidate 16 , ensitrelvir has just received Phase 3 clinical approval in the United States. It is critical to understand its molecular mechanism of inhibition. In this work, a 1.8-Å M pro -ensitrelvir complex structure (PDB: 8HBK) was determined to elucidate the precise molecular inhibition mechanism of ensitrelvir. We found that ensitrelvir mainly recognizes the S1, S2, and S1' subsites of M pro and relies on the stability of the hydrogen bonding network in the substrate-binding pocket, unlike the covalent inhibition mechanism of nirmatrelvir. In addition, the functional M pro exists in a homo-dimer form in the physiological state, and the N-terminus residues 1-7 penetrate deeply into the substrate-binding pocket of the neighboring promoter and the N-terminus serine residue stabilizes the S1 subsite by hydrogen bonding, contributing to the stability of the substrate pocket and ensitrelvir binding (Fig. 2c, d). This differs from the previously reported model (PDB ID:7VU6) 9 , in which the N-terminal of M pro lacks two residues (Ser1 and Gly2), while these two N-terminal residues can be traced in our complex structure (PDB ID:8HBK) based on the clear electron density (Supplementary Fig. S1). Moreover, the electron density of the Serl and Gly2 at the N-terminus is also clearly visible in all the complex structures of nirmatrelvir we solved (Supplementary Fig. S2).
The disruption of the N-terminus of M pro may result from their construction of in vitro preparations of the protein. In previous research, we identified that the N-terminus of M pro of SARS had excess amino acids, which will significantly impact on in vitro enzyme activity. The enzyme activity of M pro with two extra residues at the N-terminus (GS-M pro ) decreased by about 24 times compared to the clean N-terminus M pro 17 . The additional amino acids (Gly-2 and Ser-1) may have led to a difference in M pro activity compared to the physiological state, with the protein being less active in vitro. This may explain why ensitrelvir exhibits stronger inhibition of enzyme activity than nirmatrelvir in vitro in that work, while our results show that the enzyme activity inhibition of ensitrelvir is similar to that of nirmatrelvir. Furthermore, there is a big gap in binding affinity between the two types of proteins, and the Kd of GS-M pro was 418 and 8.43 nM compared to the clean N-terminus M pro (Supplementary Fig. S4). Under physiological conditions, residues in the N-terminus of M pro penetrate deep into the substrate pocket of the neighboring protomer and are involved in stabilizing the S1 subsite 18 . The extra residues may spatially block ensitrelvir from entering the substrate pocket.
Overall, the apo-form structures and in vitro enzyme activity assays of several variants suggest that minor changes in M pro have little effect on the overall structure, particularly the substrate pocket and the active center, which is consistent with our earlier findings. In addition, mutations in variants do not alter the properties of M pro , so they may not impair the in vitro enzymatic inhibition of nirmatrelvir and ensitrelvir, which further suggests that M pro is an ideal drug target because it is one of the least variable viral components. Finally, we determined the in vitro inhibitory effects of ensitrelvir and nirmatrelvir on M pro enzymatic activities, and the results showed that ensitrelvir and nirmatrelvir exhibit consistent in vitro enzymatic inhibition against SARS-CoV-2 M pro and its variants, which is consistent with their excellent clinical efficacy. Although these two inhibitors belong to different types, both of them remain to be effective against M pro s from all 5 SARS-CoV-2 variants of concern, suggesting M pro is a bona fide broad-spectrum target.
Recently, a 2.2 Å resolution complex structure of M pro -ensitrelvir was reported 19 . In their functional dimer, one protomer has clear electron density for its N-terminus, but its neighboring protomer does not. In the 1.8 Å resolution structure in this work, clear electron density can be observed for M pro N-terminus, and we provide an accurate model for elucidating the ensitrelvir binding mode. The molecular mechanisms uncovered in this study provide the basis for future inhibitor design.

Methods
Cloning, protein expression, and purification of SARS-CoV-2 variant M pro . The expression plasmid used to produce full-length M pro was obtained by using site-directed mutagenesis to introduce G15S, K90R, and P132H substitutions into the expression plasmid of full-length WT M pro . The expression plasmid was transformed into E. coli BL21 (DE3) cells and then cultured in a 2-L shaking flask with 1 L Luria broth medium containing 0.1 g/L ampicillin at 37°C. When the optical density at 600 nm of the bacteria reaches 0.6-0.8, a final concentration of IPTG was added to the culture to induce protein expression at 16°C. After 10 h, the bacteria were pelleted by centrifugation at 3000×g for 15 min. Then the bacteria were resuspended in lysis buffer (50 mM HEPES, pH 8.0, 300 mM NaCl), and the supernatant was incubated with Ni-NTA agarose gravity column (GE) after centrifuging at 18,000 rpm for 30 min. The fusion proteins were washed with wash buffer (50 mM HEPES, pH 8.0, 300 mM NaCl, 20 mM imidazole). After washing 10-20 column volume, the PreScission protease was added to remove the His-tag. Then the samples of the protein were loaded onto a Hitrap Q HP column and then purified by size-exclusion chromatography with a Superdex 75 increase column with storage buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, and 4 mM DTT). The fractions were concentrated to 10 mg/mL for the next test 20 .
Crystallization, data collection, and structure determination. The crystals of M pro apo forms were directly screened by crystallization kits, and the crystals of the complexes with inhibitors were screened through co-crystallization of the variants M pro at a concentration of 5 mg/ml with nirmatrelvir or ensitrelvir at 0.5 mM. The crystals were obtained at 18°C by the sitting method. Crystals were cryoprotected using the reservoir solution with 20% ethylene glycol and flash-frozen in liquid nitrogen. Diffraction data were collected at 100 K at a wavelength of 0.9785 Å. The structures were solved by molecular replacement with the program CCP4 using the complex structure of WT M pro and N3 (PDB ID: 7VH8) as a search model 12 . The model was refined using PHENIX 21 and manually constructed using Coot 22 .
IC 50 measurement. The method of IC 50 measurement has been previously demonstrated 20 . In brief, the fluorescent substrate was applied to measure the hydrolytic activity of M pro . M pro (0.1 μM) was incubated with different concentrations of nirmatrelvir or ensitrelvir for 90 s before 10 μM substrate was added. Fluorescence intensity was monitored by an EnVision multimode plate reader (Perkin Elmer) using wavelengths of 320 nm for excitation and 405 nm for emission. The changes in initial rates when adding different concentrations of nirmatrelvir or ensitrelvir were calculated to evaluate the inhibitory effect. The dose-response curve for IC 50 values was determined by nonlinear regression using GraphPad Prism.
ITC assay. The ITC experiment was performed on a MicroCal PEAQ-ITC. After cleaning the sample cell and needle, the OM pro (20 μM) protein was carefully injected into the sample cell using a micro-syringe without any bubbles, and ensitrelvir (200 μM) was filled into a 40 μL titration syringe. OM pro and ensitrelvir were diluted with TB (1% DMSO, pH = 8.0), and deionized water was injected into the reference cell as a heat balance control. After 35 times titrations of ensitrelvir into the sample cell at a constant rate of 150 s, the One Set of Sites is selected as the fitting model.
Statistics and reproducibility. Statistical analysis was carried out using Prism software. The amount of enzyme was determined by the initial experiment, and the enzyme kinetic and half-inhibition tests consisted of three replicate trials.
Reporting summary. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Materials availability
Materials used in this study will be made available under an appropriate Materials Transfer Agreement.