Discovery of SARS-CoV-2 papain-like protease (PLpro) inhibitors with efficacy in a murine infection model

Abstract

Vaccines and first-generation antiviral therapeutics have provided important protection against COVID-19 caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). However, there remains a need for additional therapeutic options that provide enhanced efficacy and protection against potential viral resistance. The SARS-CoV-2 papain-like protease (PL^pro) is one of the two essential cysteine proteases involved in viral replication. While inhibitors of the SARS-CoV-2 main protease have demonstrated clinical efficacy, known PL^pro inhibitors have, to date, lacked the inhibitory potency and requisite pharmacokinetics to demonstrate that targeting PL^pro translates to in vivo efficacy in a preclinical setting. Here, we report the machine learning-driven discovery of potent, selective, and orally available SARS-CoV-2 PL^pro inhibitors, with lead compound PF-07957472 (4) providing robust efficacy in a mouse-adapted model of COVID-19 infection.

Fig. 1.. Discovery of PL^pro inhibitors guided by machine learning.

(A) Summary of key compounds in our discovery campaign with their biochemical potency, cell antiviral potency in Vero E6 cells containing the P-glycoprotein (Pgp) inhibitor CP-100356 (2 μM), passive permeability determined by Ralph Russ canine kidney (RRCK) cells (23) and apparent intrinsic clearance (CL_int,app) obtained from metabolic stability studies in cryopreserved human hepatocytes (24). *All values provided are the geometric mean of at least three replicates. (B and C) The potency of compounds made in the Suzuki and C-N coupling libraries; n denotes the number of compounds made in each library. The red line denotes the potency of the compound that inspired the library. (D) Biochemical (blue) and cell (green) potency over time, and (E) product between potency and CL_int,app in human hepatocytes (24) of synthesized compounds plotted over time. In (D) and (F), each data point is a compound where we track the date of synthesis completion, and the solid lines show the running minimum. As our synthetic chemistry efforts were based in China, COVID lockdown there precipitated a marked disruption to campaign progress. (F) PL^pro inhibition translates to antiviral activity. K_i and inhibition are correlated on a cellular level [in the presence of 2 μM of Pgp inhibitor CP-100356 to counteract the high levels of Pgp efflux in Vero E6 cells (12)]. The dashed lines show EC₅₀ = K_i and 100 × K_i, respectively.

Fig. 2.. X-ray crystal structure of compound 4 bound to SARS-CoV-2 PL^pro and structural comparison with GRL0617 binding.

In each panel, SARS-CoV-2 PL^pro bound to 4 (sticks, salmon) is shown in cartoon representation, color-coded by subdomain, and with light gray surface representation. (A) Structural superimposition of SARS-CoV-2 PL^pro bound to 4 and SARS-CoV-2 PL^pro bound to mouse ISG15 [yellow; Protein Data Bank (PDB) ID 6YVA]. The PL^pro protein structure in 6YVA is not shown for clarity. The C terminus of ISG15 (His¹⁴⁹-Gly¹⁵⁵) overlaps with the binding site of 4. The catalytic triad is shown in magenta; a catalytically dead construct of PL^pro (C111S) was used to generate the crystal structure. (B) Detailed view of 4 binding in the BL2 loop of SARS-CoV-2 PL^pro, highlighting critical protein-ligand interactions. Hydrogen bonds are represented as black dashed lines. Tyr²⁶⁴ is not labeled but is in the background, forming a hydrogen bond with the amide of 4 (C to E) Structural overlays of SARS-CoV-2 PL^pro bound 4 and SARS-CoV-2 bound to GRL0617 (wheat; PDB ID 7CMD). The orientation of the binding pocket varies in each panel, with the BL2 loop labeled throughout. Surface representation is shown for the structure of SARS-CoV-2 PL^pro bound to 4 only.

Fig. 3.. Compound 4 is an antiviral lead and a suitable in vivo tool compound.

(A) Dose-response curve of 4 assayed in a human airway epithelial cell model of SARS-CoV-2 infection (12). The fraction unbound of 4 in the dNHBE medium was measured to be 0.916, assumed to be 1 for modeling purposes. (B) Profiling 4 through safety panels and CYP panels reveal no clear off-target liabilities. See table S1 for full Cerep profile. (C) In vitro and in vivo absorption, distribution, and metabolism properties across multiple species. MDR, multi-drug resistance protein 1 (P-glycoprotein) efflux ratio; BCRP, murine Breast Cancer Resistance Protein BA/AB ratio; PPB, Plasma Protein Binding.

Fig. 4.. PL^pro inhibitor 4 is efficacious in a murine SARS-CoV-2 infection model.

(A) Dose-range finding studies showed that the selected doses cover a range of trough concentrations from subtherapeutic to therapeutic levels. (B) Representative immunohistochemistry images from lung histopathology of mouse-adapted SARS-CoV-2 infection 4 days after dose of 4 at 20, 50, and 150 mg/kg; Nirmatrelvir at 1000 mg/kg; vehicle; or uninfected, untreated animals. (C) Compound 4 protected mice from weight loss, with roughly 10% weight loss seen in the vehicle arm consistent with data in (12). PO, by mouth/oral administration; ***P < 0.001; ****P < 0.0001; n.s., not significant. (D) Compound 4 led to a statistically significant and dose-dependent reduction in day 4 lung viral titers. Titers were plotted as mean log₁₀ 50% cell culture infectious dose (CCID₅₀)/ml ± SEM, with data analysis and significance levels matching those in (12).

Fig. 5.. High-frequency mutations in Omicron variants (6,891,950 reads) on Nsp3 and resistance selection mutants of PL^pro.

(A) The location of the cysteine protease, PL^pro in SARS-CoV-2 genome, and the cleavage sites that it processes to generate functional viral proteins. (B) The Nsp3 protein (residues 1 to 1944) showing all domains identifying clinically observed Omicron mutations with frequency > 1%. (C) PL^pro domain (residues using PL^pro numbering 3 to 315) with location of mutation with frequency > 0.5% in green, which are distal from residues within 5.5 Å for 4 (pink spheres). Binding surface is shown in as blue and cyan, where cyan indicates the three highest frequency mutations (A246, P347, and K157) within the ligand binding site (table S8). (D) Structural superposition of ISG15 binding site and 4 showing interaction with similar residues.