Structure-based discovery of highly bioavailable, covalent, broad-spectrum coronavirus MPro inhibitors with potent in vivo efficacy

Abstract

The main protease (M^Pro) of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a validated drug target. Starting with a lead-like dihydrouracil chemotype identified in a large-library docking campaign, we improved M^Pro inhibition >1000-fold by engaging additional M^Pro subsites and using a latent electrophile to engage Cys¹⁴⁵. Advanced leads from this series show pan-coronavirus antiviral activity, low clearance in mice, and for AVI-4773, a rapid reduction in viral titers >1,000,000 after just three doses. Both compounds are well distributed in mouse tissues, including brain, where concentrations >1000× the 90% effective concentration are observed 8 hours after oral dosing for AVI-4773. AVI-4516 shows minimal inhibition of major cytochrome P450s and human proteases. AVI-4516 also exhibits synergy with the RNA-dependent RNA polymerase inhibitor, molnupiravir, in cellular infection models. Related analogs strongly inhibit nirmatrelvir-resistant M^Pro mutant virus. The properties of this chemotype are differentiated from existing clinical and preclinical M^Pro inhibitors and will advance therapeutic development against emerging SARS-CoV-2 variants and other coronaviruses.

Fig. 1.. Initial discovery and early structure-based optimization of the AVI-4673 series.

(A) Large-library docking led to 17 diverse inhibitors, of which Z7212 (AVI-1084) is shown (10). (B) Structure-based optimization of one of them Z7212, explored side chains modeled to bind in the S2 and S1 pockets, leading to submicromolar inhibitors. (C) Surface representation of the homodimer of M^Pro. Protomer A is in pink and protomer B is in light brown. (D) Superposition between the docking predicted (green carbons) and the crystal structure (gray carbons, PDB: 9MVM) of the inhibitor AVI-3318 [from (A)]. (E) Docked pose of AVI-3779, the most potent inhibitor based on docking.

Fig. 2.. Medicinal chemistry optimization of the uracil scaffold and identification of latent electrophilic warhead.

The arms of the compounds are highlighted according to the corresponding M^Pro subsite. Pink, S1; green, S1′; blue, S2. (A) Structures of notable compounds during optimization of noncovalent series with uracil core; AVI-4673 is the most potent noncovalent. (B) Structure of the most potent C6-aryl covalent compounds. (C) The structures of the C6-methyl version of the covalent alkyne compounds, AVI-4516 and AVI-4773. (D to F) Structures of M^Pro bound to inhibitors; the ligands were modeled using F_o − F_c electron density maps, with subsites annotated; AVI-4303 and M^Pro costructure solved at a maximum resolution of 1.58 Å PDB: 9MVQ (D) AVI-4692 and M^Pro costructure solved at a maximum resolution of 1.85 Å PDB: 9MVO (E) AVI-4516 and M^Pro costructure solved at a maximum resolution of 2.35 Å PDB: 9MVP (F). (G) Overall comparison of 116 compounds with measured biochemical IC₅₀ values. AVI-4303 is red, AVI-4516 is blue, AVI-4692 is yellow, AVI-4694 is green, and AVI-4773 is purple. (H) Deconvoluted whole protein denaturing mass spectrum of M^Pro alone, and M^Pro treated with AVI-4516 indicating one modification. (I) Concentration of AVI-4516 plotted against apparent inaction of M^Pro (k_app) with calculated inhibitor kinetic parameters denoted. (J) Dialysis experiment demonstrating that AVI-4516, AVI-4773, AVI-4692, and AVI-4694 are irreversible covalent inhibitors. (K) Uracil ring atom numbering nomenclature. (L) MS2 spectra of chymotrypic peptide of AVI-4516 and the structure of y₆ ion observed with proposed adduct bound to Cys¹⁴⁵.

Fig. 3.. In cellulo efficacy of M^Pro inhibitors against SARS-CoV-2 strains, related coronaviruses, and nirmatrelvir-resistant mutations.

(A) A table of EC₅₀ values for dose-response inhibition of viral replication in replicon-based assay. The BA.2.86.1 curve is presented below (C) and the rest are presented in fig. S9 (A and B). (B) A schematic cartoon of the replicon assay for measurement of antiviral potency. (C) Dose-response curves for selected compounds in the BA.2.86.1 replicon assay. (D) A table of EC₅₀ values for dose-response inhibition of viral replication in the live virus–based assay. The EG.5.1 curve is presented below (E) and the other strains are presented in fig. S9 (C to E). (E) Dose-response curves for selected compounds against EG.5.1 live virus SARS-CoV-2 variant in A549-ACE2^h cells. (F) AVI-4516 EC₅₀ values for pan-coronavirus antiviral efficacy screen determined through CPE. All error bars are plotted as ±SD.

Fig. 4.. ADME, PK, and safety properties of M^Pro inhibitors.

(A) ADME and PK properties of AVI-4516, AVI-4773, AVI-4692, and AVI-4694. (B) Unbound concentration in mouse plasma of AVI-4516, AVI-4773, AVI-4692, and AVI-4694 after 50 mg/kg po dosing. (C) Comparison of mouse plasma concentration of AVI-4516 when dosed 10 mg/kg iv or 50 mg/kg po. The concentration at 24 hours was below the LOQ. (D) Percent inhibition of mammalian peptidase panel when treated with 10 μM AVI-4516, AVI-4673, and AVI-4694. AVI-4516 has low inhibition across the panel. Proteases that inhibited >50% have dose-response curves that generated IC₅₀ values in fig. S13. (E) Percent inhibition of human CYP panel of AVI-4516, AVI-4773, AVI-4692, and AVI-4694 compared to nirmatrelvir and ensitrelvir. (F) Percent inhibition of mammalian receptor panel when treated with 10 μM AVI-4516, AVI-4673, and AVI-4694. AVI-4516 has low inhibition across the panel. Proteins that were inhibited >50% have dose-response curves that generated IC₅₀ values in fig. S13 (A and B). (G) Total brain and plasma concentration of AVI-4516, AVI-4773, and ensitrelvir after the 100 mg/kg po dose. All error bars are plotted as ± SD. (H) Unbound concentration of AVI-4516, AVI-4773, and ensitrelvir in mouse brain. (I) Unbound brain to plasma partitioning coefficient for AVI-4516, AVI-4773, and ensitrelvir.

Fig. 5.. Oral administration of AVI-4516 limits virus replication.

(A) Schematic of antiviral efficacy experiment. WT mice were intranasally infected with 10³ PFU of the SARS-CoV-2 Beta variant. Infected animals were orally dosed (BID) with either vehicle, nirmatrelvir, or AVI-4516. The lung tissues were harvested and processed for further analysis at 2, 4, and 7 days postinfection (dpi) (n = 5 per group per time point). (B) Graphs presenting mature virus particles from the lungs of infected mice at the indicated time points. Each dot represents the infectious virus titer in an individual mouse. (C) A separate group of mice were infected with the SARS-CoV-2 Beta variant and treated with various concentrations of AVI-4516. Virus titers were measured at 2 dpi. (D) Representative images of immunohistochemistry for the SARS-CoV-2 N protein in the left lung lobe of mice from different treatment groups at the specified time points. (E) Hematoxylin and eosin (H-E) staining of lung tissue. Significant mononuclear cell infiltrations were marked by red arrows, and severe injury in respiratory epithelia, characterized by epithelial cell debris in the lumen and incomplete epithelial regeneration, was highlighted by green arrows. (F) Schematic of antiviral efficacy experiment. WT mice were intranasally infected with 103 PFU of the SARS-CoV-2 Beta variant. Infected animals were orally dosed (BID) with either vehicle, ensitrelvir, or AVI-4773. The lung tissues were harvested and processed for further analysis at 2 and 4 dpi (n = 5 per group per time point). (G) Graphs presenting mature virus particles from the lungs of infected mice at the indicated time points are presented. Each dot represents the infectious virus titer in an individual mouse. All data points are presented as mean ± SEM for each time point and were analyzed using a two-tailed unpaired Student’s t test. Scale bars, 100 μm.