Rapid resistance profiling of SARS-CoV-2 protease inhibitors

Resistance to nirmatrelvir (Paxlovid) has been shown by multiple groups and may already exist in clinical SARS-CoV-2 isolates. Here a panel of SARS-CoV-2 main protease (Mpro) variants and a robust cell-based assay are used to compare the resistance profiles of nirmatrelvir, ensitrelvir, and FB2001. The results reveal distinct resistance mechanisms (“fingerprints”) and indicate that these next-generation drugs have the potential to be effective against nirmatrelvir-resistant variants and vice versa.

2 Thus, considerable urgency exists to develop next-generation M pro inhibitors with different resistance mechanisms and, in parallel, robust systems to rapidly assess the potential impact of candidate resistance mutations.
Ensitrelvir (Xocova) and FB2001 are being evaluated in clinical trials, and the former drug also recently received EUA in Japan 10,11 ( Figure 1A). We recently developed a gain-of-signal system for facile quantification of M pro inhibition 12 , and subsequently used it together with an evolutionand structure-guided approach to characterize candidate nirmatrelvir-and ensitrelvir-resistance mutations 2 . Here, an expanded panel of M pro single and double mutants based on recent studies by our group and others 2-9 is leveraged to determine resistance profiles of these two drugs, as well as FB2001, a potential next-generation therapy (heatmap of results in Figure 1B; quantification summary in Table 1; representative dose responses in Figure S1). In addition to providing a method to rapidly profile candidate resistance mutations in living cells, our gain-of-signal assay also provides a quantitative metric for M pro functionality 12 (Methods).
This system is based on the fact that overexpression of wildtype SARS-CoV-2 M pro results in the cleavage of multiple substrates in cells 13,14 including at least one required for RNA Polymerase IIdependent gene expression 12 . Therefore, expression of the Src-M pro -Tat-Luc reporter itself is rapidly shut down following transfection and can only be recovered by chemical or genetic inhibition of M pro . Thus, genetic mutations effectively phenocopy the chemical doseresponsiveness of the system, with some variations showing wildtype M pro activity (background luminescence) and others compromising activity weakly or strongly depending on the nature of the mutation (low to high luminescence). For example, in comparison to wildtype M pro , catalytic mutants such as C145A yields 50-to 100-fold higher luminescence 2,12 . The M pro variant constructs used here display a range of luminescence levels in the absence of drug indicative of near-normal M pro activity (notably, M49I and M49L), weakly compromised M pro activity (notably, A173V), and strongly compromised M pro activity (notably, E166V) ( Figure S2). These results suggest that several variants can confer at least partial drug resistance with little loss in M pro functionality (and accordingly high viral fitness), whereas others such as E166V require suppressor mutations such as L50F to restore M pro function to a level that enables virus replication (evidenced by recent resistance studies with pathogenic SARS-CoV-2 in culture and in vivo in animal models 3,5 ).
Regardless of the details of each molecular mechanism, the results here demonstrate that nirmatrelvir, ensitrelvir, and FB2001 have distinct resistance profiles and that the latter inhibitors (with appropriate formulations) may be effective in patients suffering from Paxlovid rebound 15 or bona fide resistance 2 . FB2001 may additionally have a higher resistance barrier given that no fully Brief Communication 4 functional single M pro variants tested to-date confer a strong resistance to this compound. Importantly, the gain-of-signal live cell assay recapitulates recent findings using replication competent viruses and provides a safe and rapid method for assessing resistance. As the SARS-CoV-2 variant pool deepens, this assay and variant panel can be expanded in lock-step to provide early resistance "fingerprints" of candidate next-generation M pro inhibitors. Such an early profiling strategy has the potential to minimize the risks of developing drugs prone to cross-resistance and, importantly, to help identify inhibitors with the highest barriers to resistance.

Competing Interests
The M pro gain-of-signal system is the subject of U.S. Provisional Application Serial No. 63/108,611, filed on November 2, 2020, with RSH and SAM as inventors. The authors declare that there are no additional competing interests.   Clear examples of single amino acid substitution mutations conferring selective resistance to nirmatrelvir and ensitrelvir are highlighted in gray; similar mutations have yet to be found for FB2001. The relative values in brackets are reflected in the heatmap in Figure 1B.

Cell culture
All M pro inhibition assays were done as described with the live cell gain-of-signal assay using the pcDNA5/TO-Src-M pro -Tat-fLuc reporter construct 12 . All M pro single and double mutants selected for analysis here were based on recent reports of candidate resistant mutatnts 2-9 generated by site-directed mutagenesis (primers available upon request) and verified by Sanger sequencing.
Transfections were done using 293T cells maintained at 37°C and 5% CO2 in DMEM (Gibco catalog number 11875093) supplemented with 10% fetal bovine serum (ThermoFisher catalog number 11965084) and penicillin-streptomycin (Gibco catalog number 15140122). % ℎ = %100 − (100/( ) Results were plotted using GraphPad Prism 9 and fit using a four-parameter non-linear regression to calculate IC50 values ( Figure S1; Table 1). Resistance of mutants was calculated by the fold change in IC50 of the mutant relative to WT M pro , and these values were used to generate a heatmap in GraphPad Prism9 ( Figure 1B).
As an increase in luminescence in the absence of any inhibitor treatment is indicative of decreased M pro catalytic activity, the relative activity of each mutant was calculated by the formula below using the relative luminescence of a mutant (RLm) to the WT enzyme in the absence of inhibitor ( Figure S2). Several single mutants such as T21I, M49I, M49L, L50F, and T304I show near WT activity. Other mutants such as A173V show modest 1.5 to 3-fold decreases in relative activity, and a few such as E166V are severely compromised. L50F partly restores the activity of E166A and E166V mutants consistent with prior reports 2-9 .