Discovery of nirmatrelvir resistance mutations in SARS-CoV-2 3CLpro: A computational-experimental approach

Abstract

The COVID-19 pandemic has emphasized the urgency for effective antiviral therapies against SARS-CoV-2. Targeting the main protease (3CLpro) of the virus has emerged as a promising approach, and nirmatrelvir (PF-07321332), the active component of Pfizer’s oral drug Paxlovid, has demonstrated remarkable clinical efficacy. However, the emergence of resistance mutations poses a challenge to its continued success. In this study, we employed alchemical free energy perturbation (FEP) alanine scanning to identify nirmatrelvir-resistant mutations within the SARS-CoV-2 3CLpro. FEP identified several mutations, which were validated through in vitro IC₅₀ experiments and found to result in 8 and 72-fold increases in nirmatrelvir IC₅₀ values. Additionally, we constructed SARS-CoV-2 omicron replicons containing these mutations, and one of the mutants (S144A/E166A) displayed a 20-fold increase in EC₅₀, confirming the role of FEP in identifying drug-resistance mutations. Our findings suggest that FEP can be a valuable tool in proactively monitoring the emergence of resistant strains and guiding the design of future inhibitors with reduced susceptibility to drug resistance. As nirmatrelvir is currently widely used for treating COVID-19, this research has important implications for surveillance efforts and antiviral development.

Introduction

The disastrous COVID-19 pandemic, caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has highlighted the urgent need for effective antiviral therapies to mitigate the disease’s impact and complement vaccination strategies.¹ Targeting the main protease (M^pro) of SARS-CoV-2, also known as 3-chymotrypsin-like protease (3CLpro), has emerged as a promising approach developing antiviral treatments.^2–6 The 3CLpro is pivotal for viral replication as it cleaves viral polyproteins pp1a and pp1ab at 11 distinct sites, generating nonstructural proteins essential for the viral life cycle.^7,8 Given that human proteases do not recognize these cleavage sites, 3CLpro serves as an attractive and selective drug target.⁹

Nirmatrelvir (PF-07321332), the active component of Pfizer’s oral drug Paxlovid, is a peptidomimetic covalent inhibitor of 3CLpro that was granted Emergency Use Authorization by the FDA in December of 2021. Nirmatrelvir has exhibited remarkable clinical efficacy in reducing the risk of hospitalization or death by 89% when administered during the early stages of infection.¹⁰ Unlike other approved SARS-CoV-2 antivirals, such as remdesivir and molnupiravir, nirmatrelvir offers distinct advantages. Remdesivir, a nucleotide analog targeting the viral RNA-dependent RNA polymerase, has shown limited effectiveness in patients requiring ventilation and high-flow oxygen and also requires parenteral administration, restricting its use to hospital settings.¹¹ Molnupiravir, a nucleoside analog that induces mutations in the viral RNA genome, offers a 31% reduction in hospitalization or death, falling short of nirmatrelvir’s 89% reduction.¹² Nirmatrelvir’s superior efficacy and oral administration establish it as a more favorable option for patients at high risk of progressing to severe COVID-19.

Nevertheless, the emergence of resistance mutations in the 3CLpro poses a significant challenge to the continued success of nirmatrelvir as an antiviral treatment. A recent Pfizer report directed for healthcare providers disclosed a nirmatrelvir-resistant mutant, S144A, with remarkedly reduced binding to nirmatrelvir (i.e., >90-fold increase in K_i).¹³ Similarly, recent reports have reported the presence of nirmatrelvir resistance mutations in SARS-CoV-2 isolates, either generated de novo through in vitro passaging^14–18 or found in clinical samples.^19–21 SARS-CoV-2 drug-resistant variants have also been previously reported in patients treated with viral polymerase inhibitor remdesivir.^22,23 The development of antiviral resistance is not a novel phenomenon, as evidenced by treating chronic or persistent viral infections such as HIV, hepatitis B virus (HBV), hepatitis C virus (HCV), herpesviruses, and influenza.²⁴ In these cases, antiviral administration, particularly as monotherapy, has led to the emergence of viral escape mutations, which can compromise therapeutic effectiveness and contribute to treatment failure.^25,26 Thus, identifying nirmatrelvir resistance mutations and providing a detailed understanding of the mechanisms underlying resistance is crucial to minimize the risk of treatment failure, and can further be used to guide the design of inhibitors with reduced potential of evolving viral resistance.

Herein, we used alchemical free energy perturbation (FEP) alanine scanning to identify nirmatrelvir resistance mutations within the SARS-CoV-2 3CLpro. Potential resistance mutations, identified using FEP, were combined in two mutation combinations in recombinant 3CLpro proteins to verify our computational calculations. In vitro IC₅₀ experiments of the 3CLpro mutants revealed 8 and 72-fold increases in IC₅₀ values of nirmatrelvir. Furthermore, we constructed SARS-CoV-2 replicons on the omicron BA.1 strain background containing our 3CLpro mutations and found a 20-fold increase in EC_50, showcasing that FEP can be used successfully to identify drug resistance mutations within the 3CLpro.

Results and Discussion

Free-energy perturbation (FEP) alanine scanning to identify nirmatrelvir resistance mutations

The implementation of alanine scanning mutagenesis (ASM) as a computational technique for examining protein-protein and protein-ligand interactions has proven to be highly effective in elucidating the structural and energetic properties of hot spots and identifying drug-resistance mutations.^27–29 Combining alchemical free energy perturbation with all-atom molecular dynamics simulations (FEP/MD) is arguably the most precise computational strategy for determining the impact of mutations on molecular complexes.^30–34 Using alchemical FEP/MD, we performed alanine scanning on a subset of 3CLpro residues within 4 Å of the nirmatrelvir binding site (PDB ID: 7VH8) (Figure 1a and Figure 1c).³⁵ We refrained from conducting alchemical FEP/MD alanine scanning on residues H41, C145, and H163 within the 3CLpro, as these residues have been identified as essential for 3CLpro’s enzymatic activity.^21,36,37 H41 forms the catalytic dyad with C145. Consequently, it is improbable that these residues would naturally mutate to provide resistance against nirmatrelvir. Furthermore, we deliberately refrained from mutating P168 due to challenges in achieving optimal convergence and parameterization owing to the intricacies of the proline side chain and G143 to conserve computational resources, considering that glycine already possesses a minimal amino acid sidechain (Figure 1a). In total, ASM via FEP was performed for 16 residues in the 3CLpro (Figure 1b and Table 1).

Figure 1. Targeted alanine scanning mutagenesis in 3CLpro using FEP to identify nirmatrelvir resistance mutations.

a. Cartoon representation of nirmatrelvir (blue) in complex with 3CLpro (gray) (PDB:7VH8). Residues in 3CLpro within 4 Å of nirmatrelvir shown as sticks in orange. Yellow lines represent polar contacts between nirmatrelvir and 3CLpro. b. Heat map showcasing computed changes in free energy of binding (kcal/mol) for alanine mutations in 3CLpro and binding with nirmatrelvir. All binding values are rescaled from −1 (alanine mutations increase binding with nirmatrelvir shown in green) to +1 (alanine mutations decrease binding with nirmatrelvir shown in red). c. Nirmatrelvir in complex with 3CLpro. 3CLpro shown in electrostatics. Red and blue for negative and positive charges, respectively.

Table 1.

Alanine scanning mutagenesis with FEP (kcal/mol) identifies nirmatrelvir resistance mutations.^a

Mutation	3CLpro-nirmatrelvir ΔGBound statec	3CLpro ΔGFree statec	ΔΔGFEPb
M49A	2.35 ± 0.09	2.24 ± 0.10	0.11
Y54A	12.37 ± 0.17	9.41 ± 0.10	2.96
F140A	−9.13 ± 0.15	−7.62 ± 0.04	−1.52
L141A	12.17 ± 0.07	11.71 ± 0.07	0.46
N142A	77.32 ± 0.04	77.57 ± 0.03	−0.25
S144A	−2.85 ± 0.08	−5.77 ± 0.08	2.92
H164A	18.92 ± 0.10	21.10 ± 0.12	−2.18
M165A	2.32 ± 0.09	2.63 ± 0.06	−0.31
E166A	109.68 ± 0.10	108.38 ± 0.12	1.30
L167A	14.81 ± 0.10	14.81 ± 0.11	0.002
H172A	21.16 ± 0.13	21.11 ± 0.16	0.06
D187A	131.50 ± 0.11	133.39 ± 0.18	−1.88
R188A	267.10 ± 0.12	267.47 ± 0.18	−0.36
Q189A	56.59 ± 0.07	56.15 ± 0.07	0.44
T190A	18.08 ± 0.05	18.51 ± 0.04	−0.42
Q192A	56.53 ± 0.06	57.02 ± 0.06	−0.50

^aWhen ${\text{ΔΔG}}_{\text{FEP}}>0$ mutations to alanine are less favorable in the bound state than in the free state.

^b${\text{ΔΔG}}_{\text{FEP}}$ is calculated where ${\text{ΔΔG}}_{\text{FEP}}={\text{ΔG}}_{\text{Bound}\text{state}}-{\text{ΔG}}_{\text{free}\text{state}}$. All units are in kcal/mol.

^cStatistical uncertainties calculated for bidirectional (i.e., forward and backward transformation) FEP calculations using the BAR estimator are all between ± 0.03 to ± 0.18 kcal/mol.

We hypothesized that by introducing alanine mutations that diminish the binding affinity of nirmatrelvir to the 3CLpro, we would consequently reduce the effectiveness of nirmatrelvir. Additionally, we employed FEP/MD alanine scanning as a valuable technique to pinpoint “hot spot” residues within the 3CLpro that are crucial for nirmatrelvir’s binding. This approach could reveal targets for future nirmatrelvir-resistant mutations, beyond the scope of alanine mutations alone. In line with best practices³⁸, all FEP calculations were run bidirectionally (i.e., forward and backward transformations) and free-energy changes were calculated using the maximum-likelihood BAR estimator.³⁹

Numerous residues in the 3CLpro enzyme, including M49, N142, L167, and H172, exhibited considerable tolerance to alanine substitutions in our FEP calculations, as demonstrated by a ΔΔG_FEP value of ± 0.25 kcal/mol (Table 1). In a recent study, Flynn et al. conducted an exhaustive mutational scan of the 3CLpro to identify mutations that are compatible with the 3CLpro’s functionality.⁴⁰ Their findings were in line with our results, revealing that residues M49 and N142 are indeed highly tolerant to mutations, thereby suggesting their potential to facilitate the development of resistance. Although M49A and N142A mutations displayed minimal impact on the binding affinity to nirmatrelvir, it is plausible that non-alanine substitutions could have a more significant influence on binding. Consequently, nirmatrelvir may acquire resistance through alternative mechanisms, aside from the loss of binding affinity via these specific residues. Additional investigations have substantiated the findings of Flynn et al.⁴⁰, demonstrating that residues M49 and N142 can accommodate various mutations without substantially compromising enzymatic activity or nirmatrelvir sensitivity.^19,41 These observations suggest that although these residues exhibit high tolerance to mutations, alterations at these sites do not markedly alter the binding affinity of nirmatrelvir to the 3CLpro, as evidenced by our free energy perturbation (FEP) calculations (Table 1 and Figure 1b). Consequently, the sensitivity to nirmatrelvir is preserved despite the presence of mutations at these positions.

In FEP calculations, mutations Y54A, S144A, and E166A significantly reduced nirmatrelvir binding (i.e., ΔΔG_FEP > 1kcal/mol) (Table 1). Interestingly, S144A and E166A have been found in numerous studies to significantly decrease nirmatrelvir efficacy, including in vitro passaging studies^14,16,17 and studies investigating natural 3CLpro polymorphisms.⁴¹ In the course of nirmatrelvir’s clinical development, researchers utilized the murine CoV MHV-A59 to examine resistance. They discovered that the S144A mutation resulted in a more than 90-fold decrease in nirmatrelvir’s binding efficacy (Ki) to recombinant 3CLpro in vitro.¹³ Furthermore, the Y54A mutation was associated with a 24-fold increase in the Ki of nirmatrelvir to recombinant 3CLpro.¹³ Although the clinical implications of these substitutions remain uncertain, our findings clearly demonstrate that FEP can identify nirmatrelvir-resistance mutations consistent with experimental biochemical assays. Additionally, Heilmann and colleagues¹⁵ detected a similar Y54C mutation in an in vitro passaging study using a vesicular stomatitis virus (VSV)-based system, which exhibited some resistance to nirmatrelvir.

Taken together, these results effectively demonstrate the ability of FEP in accurately identifying residues essential for nirmatrelvir binding, as well as those that play a role in fostering resistance against nirmatrelvir, in accordance with existing experimental studies. In particular, S144 and E166 seem to be hot spots for drug resistance of nirmatrelvir.^14,41,42 Overall, these findings suggest that FEP can be useful to not only identify residues that may subsequently mutate to confer resistance to nirmatrelvir via loss in binding affinity, but also in identifying residues within proteins that may be highly tolerant to mutations as evidenced by M49 and N142.

IC₅₀ of nirmatrelvir against wild-type and mutant SARS-CoV-2 3CLpro

Next, we cloned and purified two double mutants, Y54A/S144A and S144A/E166A, along with wild-type SARS-CoV-2 3CLpro. Residues Y54 and S144 are located near the catalytic dyad, and all three mutation sites are involved in nirmatrelvir binding (Figure 2a). We designed these two double mutant 3CLpro based upon our FEP calculations (Table 1). Mutations Y54A, S144A, and E166A (ΔΔG > 1kcal/mol in FEP) were combined in two mutation combinations to model natural resistance to nirmatrelvir. Recombinant 3CLpro proteins with Y54A/S144A and S144A/E166A mutations had their enzymatic activities tested by a FRET assay using the peptide substrate 5-FAM-TSATLQ SGFRK(QXL520)-NH2 (Figure 2b). Compared with the wild-type 3CLpro, both double mutants displayed significantly reduced enzymatic activities. At equivalent 125 nM concentration of enzyme, 3CLpro mutants Y54A/S144A and S144A/E166A had 1 and 2% of the activity of the wild-type protein, respectively (Figure 2b). The loss in enzymatic activity may be due to loss of substrate binding and/or potential loss of protein stability and reduced dimerization of the 3CLpro.

Figure 2. Protease activity comparison of the wild-type and two 3CLpro mutants.

a. Dimer form of SARS-CoV-2 3CLpro structure complexed with nirmatrelvir (PDB code: 8DZ2). Residues of three mutation sites and catalytic dyad are shown in magenta and green, respectively, along with nirmatrelvir in gold color. b. 3CLpro mutants Y54A/S144A and S144A/E166A have severely diminished enzymatic activity compared to wild-type. The enzymatic activity of wild-type and mutant 3CLpro were measured and compared in a FRET assay using 125 nM, 250 nM, and 400 nM of 3CLpro enzyme. The FRET assay was conducted in triplicate (n = 3) and the mean with the standard deviation (SD) is shown.

All these mutations have been previously selected for in vitro passaging studies or identified as natural 3CLpro variants; however, the effect of these mutations combinations on nirmatrelvir resistance have not yet been tested.^13–15,17 Therefore, we hypothesized that these mutations might be plausible to combine to confer resistance to nirmatrelvir. Inhibitory activities (IC₅₀) of nirmatrelvir against wild-type and two double mutants were compared (Figure 3a). The 3CLpro mutants showed significantly decreased activity, so we used a 3.2 times higher 3CLpro enzyme concentration for the mutants. Our assay confirmed that nirmatrelvir is a nanomolar inhibitor of wild-type 3CLpro, with an IC₅₀ value of 0.050 ± 0.005 μM (Figure 3a), in agreement with previous reported IC₅₀ values of 0.022 and 0.023 μM.^17,19 Consistent with our FEP calculations, the IC₅₀ value of nirmatrelvir against the 3CLpro mutant Y54A/S144A was 0.400 ± 0.101 μM, which corresponds to an ~8-fold increase compared to the wild-type 3CLpro (Figure 3a). In addition, the IC₅₀ value of nirmatrelvir against the 3CLpro mutant S144A/E166A was 3.60 ± 0.64 μM, which is ~72-fold higher than the wild-type 3CLpro protein (Figure 3a). Therefore, we show that free energy perturbation calculations can reasonably predict nirmatrelvir resistance mutations that substantially decrease IC₅₀ efficacy as evidenced by our enzymatic FRET assay. The Ki value of nirmatrelvir against wild-type 3CLpro was determined at 0.012 μM (Figure 3b).

Figure 3. Nirmatrelvir resistance of 3CLpro mutants.

a. IC₅₀ values of nirmatrelvir to wild-type and Y54A/S144A and S144A/E166A 3CLpro mutants were determined using a FRET assay. The values represent the mean and standard deviation (SD) from n = 3 replicates. b. Kinetics of inhibition of 3CLpro activity by nirmatrelvir. Dixon plot for competitive inhibition is shown.

Although the 3CLpro mutants Y54A/S144A and S144A/E166A displayed 8-fold and 72-fold IC₅₀ resistance to nirmatrelvir, they also have severely reduced fitness overall, which may compromise their ability to combine and confer resistance in nature. Mutations L50F and T21I have previously been reported to restore replicative fitness.^14,18 Therefore, the addition of these mutations may help to restore fitness of the Y54A/S144A and S144A/E166A 3CLpro mutants.

SARS-CoV-2 omicron BA.1 replicon susceptibility to nirmatrelvir

Lastly, we constructed SARS-CoV-2 replicons on the omicron BA.1 strain background containing 3CLpro mutations Y54A/S144A and S144A/E166A (Table 2).

Table 2. Omicron BA.1 SARS-CoV-2 replicon with the S144A/E166A 3CLpro mutation has decreased susceptibility to nirmatrelvir.

Susceptibility of SARS-CoV-2 replicons to nirmatrelvir was assessed by Nanoluc activity. EC₅₀ mean and standard deviations (SD) are determined from using both technical (n = 2) and biological (n = 2) replicates.

Omicron BA.1 strain (3CLpro variant)	EC50 ± SD (μM) (Fold change from WT BA.1)
WTOmicron BA.1	0.11 ± 0.02 (1)
(S144A/E166A)Omicron BA.1	2.22 ± 0.41 (20)

Compared to wild-type the omicron BA.1 virus and replicon contain 13 mutations at various nonstructural protein genes (nsp), including a single mutation, P132H, in the nsp5 (3CLpro) gene. SARS-CoV-2 replicons allow for the evaluation of nirmatrelvir resistance without conducting heavily regulated gain-of-function experiments using infectious virus. The replicon systems provide a biologically safe alternative to infectious SARS-CoV-2 that recapitulate a large part of the viral replication cycle.^18,43 Consistent with previous studies^44,45, nirmatrelvir remained effective at neutralizing omicron BA.1 with an EC₅₀ of 0.11 ± 0.02 μM (Table 2). In comparison, our 3CLpro mutant discovered using FEP, S144A/E166A, constructed in the omicron BA.1 replicon had an EC₅₀ of 2.22 ± 0.41 μM (Table 2), indicating ~20-fold resistance to nirmatrelvir in agreement with our IC₅₀ experiments (Fig. 3a).

We also attempted to characterize the Y54A/S144A 3CLpro mutant in omicron BA.1 replicon. However, the SARS-CoV-2 BA.1 replicon with the Y54A/S144A mutations in 3CLpro was not viable due to a significant loss in replicon fitness, which is consistent with our enzymatic activity FRET assay (Figure 2b). Similar to our discussion on the reduced enzymatic activity of our mutants (see above), the addition of mutations L50F and T21I in the 3CLpro have previously shown to restore the replicative fitness of SARS-CoV-2.^14,18 As the potential of drug resistance mutations also depends on the fitness of the virus, additional compensatory mutations that increase virus fitness would likely need to accompany Y54A/S144A and S144A/E166A mutations in the SARS-CoV-2 3CLpro.

Conclusion

As the utilization of nirmatrelvir continues, it brings forth the potential for mutations to arise in the 3CLpro enzyme, granting resistance to nirmatrelvir—a phenomenon that has been observed in relation to other viruses.⁴⁶ Moreover, recent developments have shown that antibody therapeutics and other treatment methods against SARS-CoV-2 have proven ineffective against new omicron subvariants.⁴⁷ Considering that nirmatrelvir is currently the most widely used antiviral drug for treating COVID-19, it is important to proactively identify and characterize nirmatrelvir-resistant strains before they become prevalent in the population. To address this concern, we employed alchemical free energy perturbation (FEP) techniques to identify mutations within the 3CLpro that confer resistance to nirmatrelvir. The mutations identified through FEP were then validated using an IC₅₀ FRET assay, which demonstrated significant decreases—by 8-fold and 72-fold, respectively—in nirmatrelvir’s inhibitory capacity against the Y54A/S144A and S144A/E166A 3CLpro mutants. Furthermore, we characterized the S144A/E166A 3CLpro mutant within a SARS-CoV-2 omicron BA.1 replicon. Our findings revealed that this mutant exhibited a 20-fold resistance to nirmatrelvir. Collectively, our IC₅₀ and EC₅₀ experiments successfully showcase the efficacy of FEP in identifying 3CLpro mutations that confer resistance to nirmatrelvir. We show that combining FEP with biochemical experiments provides a valuable tool for actively monitoring the emergence of 3CLpro mutants in SARS-CoV-2. This research holds significant implications for surveillance efforts and can guide the strategic design of future 3CLpro inhibitors with reduced susceptibility to drug resistance.

Methods

Alchemical free-energy perturbation (FEP) alanine scanning

All FEP/MD simulations were carried out for both the bound (3CLpro-nirmatrelvir) and the free state (3CLpro) using GPU accelerated NAMD 3.0.^48,49 We used PDB: 7VH8 which contains the X-ray crystal structure of SARS-CoV-2 3CLpro in complex with nirmatrelvir solved at 1.59 Å resolution.³⁵ A series of in silico alanine scanning mutations and necessary input files for FEP calculations were generated for 3CLpro residues within 4 Å of the nirmatrelvir binding site using the alanine scanning plugin of VMD.⁵⁰ Before FEP calculations, the corresponding bound (3CLpro-nirmatrelvir) and the free state (3CLpro) systems underwent 2 ns of equilibration MD simulations. The systems force fields were prepared using the CHARMM-GUI webserver⁵¹ utilizing the CHARMM36m additive force field for proteins⁵² and CGenFF⁵³ for ligands. TIP3P was used to model water.⁵⁴ In order to mimic physiological conditions, 0.15 M of NaCl ions were added using the Monte-Carlo ion placing method. Periodic boundary conditions (PBC) were applied to all simulations and the particle-mesh-Ewald (PME) method⁵⁵ was used to treat all long-range interactions between atoms and the cut-off for non-bonded atom interactions was set to 12 Å. The temperature was kept constant at 300 K using the Langevin thermostat^56,57 and pressure was kept constant at 1.0315 bar using the Nose-Hoover Langevin piston method.⁵⁸ All bonds involving hydrogen were constrained using the SHAKE algorithm⁵⁹ and water was restrained using the SETTLE algorithm.⁶⁰ The integration time step of dynamics was set to 2-fs.

After equilibrating the structures in the bound and free states, we performed FEP calculations.^31,61 For each system, a decoupling (forward) and coupling (backward) transformation were performed in line with best practices.³⁸ In the alchemical perturbation method, many intermediate stages (denoted by λ) whose Hamiltonian H(λ) = λH_f + (1 – λ)H_i are inserted between the initial and final states. For each system, λ varies from 0 to 1.0 (for forward transformations) and 1.0 to 0 (for backward transformations) using 20 perturbation windows. Each window consisted of a 200 ps equilibration trajectory followed by a 1 ns production run. Each system (i.e., bound and the free state) underwent 20 ns of production simulation for a total of 40 ns FEP simulations for each alanine mutation calculation. The aggregate simulation time for 16 FEP calculations was 640 ns. The ParseFEP toolkit⁶² implemented in VMD⁶³ was used to test the convergence of the FEP simulations and to compute the free-energy changes and their statistical uncertainties using the maximum-likelihood Bennet Acceptance Ratio (BAR) estimator.³⁹ The level of stratification between the backward and forward transformations all exhibited a hysteresis smaller than 0.18 kcal/mol, indicating the FEP calculations are well converged.

Construction of the SARS-CoV-2 subgenomic replicon (SARS-2R)

SARS-CoV-2 replicon (SARS-2R) NSP5 mutants (Y54A/S144A, S144A/E166A) of Omicron BA.1 were constructed in the wild-type background as previously described.⁴³ Briefly, NSP5 mutations were introduced into PCR fragments with primers [SEM957 GTTAATAATTGGTTGAAGCAGTTAATTAAAGTTACACTTGTGTTCCTTTTTG (PacI), SEM624 CGTTCACCTAAGTTGGCGTATACGCG (MluI), SEM1079 GTAAATCTTCAgcATTAGGGTTAAGCATGTCTTCAGAGG (Y54A as), SEM1080 GCTTAACCCTAATgcTGAAGATTTACTCATTCGTAAGTCTAATC (Y54A s), SEM1081 CCAACACTACCACATGcACCATTAAGGAATGAACCCTTAATAG (S144A as), SEM1082 CCTTAATGGTgCATGTGGTAGTGTTGGTTTTAACATAG (S144A s), SEM1083 CCAGTTGGTAATgCCATATGGTGCATGTAACAAAAAGAG (E166A as), SEM1084 CATGCACCATATGGcATTACCAACTGGAGTTCATGCTGGC (E166A s)] and Q5 Hi-Fidelity 2X master mix (NEB, Cat# M0492S). Gel purified PCR fragments were assembled into wild type Omicron BA.1 replicon vector between restriction sites PacI and MluI with NEBuilder^® HiFi DNA Assembly Master Mix (Cat# E2621L). All sequences were validated by long-read sequencing (Plasmidsaurus, Eugene, OR, USA). Sequencing results were analyzed with Lasergene/DNASTAR software (Madison, WI, USA).

Expression and purification of SARS-CoV-2 3CL^pro

The full-length SARS-CoV-2 3CL^pro gene was codon-optimized, synthesized, and cloned into pGEX6p-1 expression vector with HRV cleavable his6-tag at the C-terminus. The overall overexpression and purification procedures were done similarly as described.⁶⁴ In short, 2 L of BL21(DE3) cells containing plasmid with SARS-CoV-2 3CL^pro gene were grown to an OD₆₀₀ of 0.7 at 37 °C in LB medium followed by 0.5 mM IPTG induction for 15 hours at 25 °C. Harvested cells were lysed by sonication and purified by a two-step (1 mL HisTrap affinity-16/60 Superdex 75 SEC) purification protocol using AKTA Pure FPLC. The his6-tag was cleaved by HRV protease, producing native 3CL^pro proteins. Two double mutants were generated using the wild-type as a template, and overexpression and purification were done similar to the wild-type.

IC₅₀ value determination by dose response curve

Nirmatrelvir was initially prepared as 10 mM stocks in 100% DMSO and prepared as 50X of a series of increasing concentrations (0.0051 to 100 μM final concentration at 3-fold serial dilution) in 100% DMSO in a 384-well PCR plate. 50X of nirmatrelvir solutions were further diluted to 3X final concentrations with an assay buffer (50 mM Tris, pH 7.3, 2 mM DTT, and 1 mM EDTA). 3X of the 3CLpro enzyme (125 nM final concentration) was prepared in the same assay buffer. 7 μL each of 375 nM (3X) enzyme solution was distributed into a black low volume 384-well plate wells, and 7 μL of varying concentration of 3X nirmatrelvir were added and incubated for 10 minutes at room temperature. A fluorogenic peptide substrate, 5-FAM-TSATLQSGFRK (QXL520)-NH₂, was initially prepared as 5 mM stock in 100% DMSO and diluted to 3X (15 μM) final concentration in assay buffer. The enzyme reaction was initiated by adding 7 μL of the (3X) substrate, and its activity was continuously monitored for 20 minutes by measuring fluorescence intensity at 490 nm/520 nm (excitation/emission) wave lengths with a BMG Labtech Optima microplate reader. The IC₅₀ values were calculated by fitting with the 3 parameter Hill equation (1), using Sigmaplot v14.5 where y is percent inhibition, $\text{x}$ is inhibitor concentration, n is the slope of the concentration–response curve (Hill slope), and ${\text{V}}_{\text{max}}$ is maximal inhibition from two to four independent assays.

$$y=V_{max}\left(\frac{x^n}{I{C}_{50}^n+x^n}\right)$$ (1)

K_i determination

The 3CLpro enzyme was prepared in the same way as IC₅₀ determination. 7 μL each of 375 nM (3X) enzyme solution was distributed into a black 384-well plate wells, and 7 μL of varying concentration of 3X nirmatrelvir were added and incubated for 10 minutes. A fluorogenic peptide substrate was prepared at 3X final concentration (1.63 –13.0 μM at 2-fold dilution) in assay buffer. The enzyme reaction was initiated by adding 7 μL of the (3X) substrate, and its activity was continuously monitored for 20 minutes by measuring fluorescence intensity at 490 nm/520 nm (excitation/emission) wave lengths with a BMG Labtech Optima microplate reader. The initial velocity of the proteolytic activity of the 3CLpro was calculated by linear regression for the first 6 min of the linear kinetic progress curves. To calculate Ki value, Dixon plot was used by plotting nirmatrelvir concentration vs 1/rate using SigmaPlot v14.5.

Replicon dose response

HEK293T cells (ATCC #CRL-11268) seeded in a 6-well plate were transfected with 1 μg replicon plasmid (SARS-2R) using jetPRIME transfection reagent (Polyplus transfection, Illkirch-Graffenstaden, France). At 16 h post transfection, cells were trypsinized then seeded into 96-well plates and treated with serial dilutions of nirmatrelvir (Medkoo Biosciences, Cat#555985). NanoLuc luciferase assays were performed 48 h post treatment (Nano-Glo^® Luciferase Assay System, Promega Cat# N1150). Dose response curves were calculated with Prism software (Graphpad, San Diego, CA, USA).