The Mac1 ADP-ribosylhydrolase is a therapeutic target for SARS-CoV-2

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) continues to pose a threat to public health. Current therapeutics remain limited to direct-acting antivirals that lack distinct mechanisms of action and are already showing signs of viral resistance. The virus encodes an ADP-ribosylhydrolase macrodomain (Mac1) that plays an important role in the coronaviral life cycle by suppressing host innate immune responses. Genetic inactivation of Mac1 abrogates viral replication in vivo by potentiating host innate immune responses. However, it is unknown whether this can be achieved by pharmacologic inhibition and can therefore be exploited therapeutically. Here, we report a potent and selective lead small molecule, AVI-4206, that is effective in an in vivo model of SARS-CoV-2 infection. Standard cellular models indicate that AVI-4206 has high target engagement and can weakly inhibit viral replication in a gamma interferon- and Mac1 catalytic activity-dependent manner. However, a stronger antiviral effect for AVI-4206 is observed in human airway organoids and peripheral blood monocyte-derived macrophages. In an animal model of severe SARS-CoV-2 infection, AVI-4206 reduces viral replication, potentiates innate immune responses, and leads to a survival benefit. Our results provide pharmacological proof of concept that Mac1 is a valid therapeutic target via a novel immune-restoring mechanism that could potentially synergize with existing therapies targeting distinct, essential aspects of the coronaviral life cycle. This approach could be more widely used to target other viral macrodomains to develop antiviral therapeutics beyond COVID-19.

Keywords: COVID; X-ray crystallography; coronavirus; infectious disease; macrodomain; microbiology; mouse; structure-based drug design; viruses.

Figure 1.. Iterative structure-based design and optimization of AVI-4206 activity against Mac1.

(A) Evolution of the early lead AVI-219 to AVI-4206 by introducing and optimizing urea functionality as found in AVI-92 to contact Asp22 and introducing geminal dimethyl substitution of the pyrrolidinone ring. Homogeneous time-resolved fluorescence (HTRF)-based IC₅₀ values from (B) and (C), and PDB codes from (E) are indicated. HTRF-based dose–response curves showing peptide displacement of an ADPr-conjugated peptide from Mac1 by compounds from the urea (B) and the pyrrolidinone ring (C) optimization paths. Data is plotted as % competition mean ± SD of three technical replicates. Data were fitted with a sigmoidal dose–response equation using non-linear regression and the IC₅₀ values are quoted with 95% confidence intervals. (D) Mac1 catalytic activity dose–response curve for indicated compounds. Data is plotted as % inhibition mean ± SD of four technical replicates. IC₅₀ values are quoted with 95% confidence intervals. (E) X-ray structures indicating conserved interactions during the optimization path from AVI-92 and AVI-219 (left) to AVI-4206 (right). Structures of compounds from the urea and the pyrrolidinone ring optimization paths are presented in the top and bottom middle panels, respectively. Multiple ligand conformations were observed for AVI-3367, AVI-3762, and AVI-4636 (labeled A and B). The F_O − F_C difference electron density map calculated prior to ligand modeling is shown for AVI-4206 (purple mesh contoured at 5 σ). Electron density maps used to model ligand other ligands are shown in Figure 1—figure supplement 1.

Figure 1—figure supplement 1.. X-ray density for ligand modeling.

Ligands were modeled using either traditional F_O − F_C electron density maps (AVI-1500, AVI-1501, and AVI-4206) or PanDDA event maps (AVI-4051, AVI-3367, AVI-3763, AVI-3762, AVI-3765, AVI-3764, and AVI-4636). The diffraction resolution and refined occupancy are indicated for each ligand. The occupancy is indicated for each confirmation when multiple ligand poses were modeled.

Figure 1—figure supplement 2.. AVI-4206 and AVI-219 inhibition of Mac1 determined using auto-mono-ADP-ribosylated PARP10 as a substrate.

(A) Standard curve of ADP-ribose detected using 100 nM NUDT5 and the AMP-Glo assay kit. Data are presented as mean ± SD for four technical replicates. Data were fitted with a power function in the form y = kx^a using non-linear regression (gray line). (B) Titration of Mac1 with auto-mono-ADP-ribosylated PARP10. The concentration of PARP10 was 10 μM based on absorbance at 280 nm, but the titration indicated that the concentration of ADP-ribose released by Mac1 was fivefold lower (~2 μM). Data are presented as mean ± SD for four technical replicates. (C) Counterscreen of compounds against 100 nM NudT5 with 2 μM ADP-ribose as a substrate. No inhibition was detected up to 1 mM compound. Data are presented as mean ± SD for four technical replicates.

Figure 2.. AVI-4206 engages Mac1 with high potency and selectivity in cells.

(A) CETSA-nLuc shows differential Mac1 stabilization after treatment of A549 cells with 10 μM of indicated compounds. Data are presented as mean ± SD of two technical replicates. Data were fitted with a sigmoidal dose–response equation using non-linear regression (gray line) and the T_agg values are quoted with 95% confidence intervals. (B) CETSA-nLuc shows a dose-dependent thermal stabilization of Mac1 after treatment of A549 cells with increasing concentrations of AVI-4206. Data are presented as mean ± SD of two technical replicates. Homogeneous time-resolved fluorescence (HTRF)-based dose–response curves showing displacement of an ADPr-conjugated peptide from indicated proteins by ADP-ribose (C) or AVI-4206 (D). ADP-ribose was used as a positive control. Data are presented as mean ± SD of three technical replicates. IC₅₀ values are quoted with 95% confidence intervals. (E) Structural modeling of MacroD2 (top, PDB code 4IQY) and Targ1 (bottom, PDB code 4J5S) showing design elements that prevent AVI-4206 cross-reactivity. The atoms of clashing residues (Cys140 in MacroD2, Arg122 in Targ1) are shown with a dot representation. The ADP-ribose present in both human macrodomain structures has been omitted for clarity.

Figure 2—figure supplement 1.. AVI-4206 increases thermal stability of Mac1 in cells.

(A) CETSA-WB shows thermal stabilization of FLAG-tagged Mac1 protein after treatment of A549 cells with 10 μM of AVI-4206. (B) Densitometry values were normalized to the lowest temperature for each treatment. Data are presented as a single densitometry measurement. (C) CETSA-nLuc shows differential stabilization of SARS-CoV2, SARS-CoV, and MERS macrodomain proteins in A549 cells treated with 10 μM of AVI-4206. Data are presented as mean ± SD of two technical replicates. Data were fitted with a sigmoidal dose–response equation using non-linear regression (gray line), and the T_agg values are quoted with 95% confidence intervals.

Figure 2—figure supplement 2.. Alignment of AVI-4206-bound SARS-CoV-2 Mac1 with diverse macrodomains suggests the origin of AVI-4206 selectivity.

(A) For human and (B) viral macrodomains, AVI-4206 is shown with teal spheres/sticks, SARS-CoV-2 Mac1 is shown with white sticks and the macrodomains being compared are shown with purple sticks. Selected Mac1 residues are labeled with black text, and the most important clashing human/viral residue is labeled and shown with red dots. For the clashing asparagine residue in the viral macrodomains (position equivalent to Phe156 in SARS-CoV-2 Mac1), rotation of the residue around the χ1 dihedral angle (shown with a red arrow) could relieve the clash and allow AVI-4206 binding. The human/viral macrodomain structures compared are bound to ADP-ribose, except for PARP9 MOD1, where no experimental structure is available and the Alphafold 2 model is shown. (C) The SARS-CoV-2 Mac1 sequence was aligned to human/viral macrodomains shown in (A) and (B) for SARS-CoV-2 Mac1 residues within 5 Å of AVI-4206. The SARS-CoV-2 residue numbering and the % identity for each sequence compared to SARS-CoV-2 Mac1 is indicated. (D) Surface view of Mac1 (PDB 9CY0) sliced through a plane containing AVI-4206 (teal sticks) and the site of the Val34Leu mutation (purple spheres).

Figure 2—figure supplement 3.. PARP14 macrodomain 1 activity is not inhibited by AVI-4206.

(A) Titration of AVI-4206 shows full inhibition of WT Mac1, but no effect against catalytically active PARP14:MD1-MD2. Negative controls of Mac1 N40D and PARP14: MD2 are included. We note that the IC₅₀ for WT Mac1 here is higher than in our other experiments, which we attribute to differences in the substrate preparation, the higher enzyme concentrations used (50 vs. 10 nM), and the exact timing of the end point measurements. (B) Enzyme activity measurements to verify that WT Mac1 and PARP14:MD1-MD2 are active and that PARP14:MD2 and Mac1 N40D are inactive.

Figure 2—figure supplement 4.. Thermal proteome profiling in A549 cellular lysates.

(A) Melting curve for Mac1 in A549 lysates treated in duplicate with either DMSO or 100 μM of AVI-4206. Data were normalized to the mean intensity at 37°C. Data were fitted with a sigmoidal dose–response equation using non-linear regression (gray line). (B) Volcano plot of the statistical significance and degree of melting temperature shift for all proteins with high-quality melting curves (n = 3446 proteins). Teal circles indicate proteins with a statistically significant shift in melting temperature (adjusted p value <0.05). The highest non-significant proteins are labeled and do not have obvious functional overlap with macrodomains.

Figure 3.. Vero-TMPRSS2 (A) or A549-ACE2^h (B) cells were pretreated with compounds and infected with mNeonGreen reporter SARS-CoV-2.

mNeonGreen expression was measured by the Incucyte system. Graphs represent mean ± SD of % replication normalized to the DMSO control 24 post-infection of three independent experiments performed in triplicate. Data were fitted with a sigmoidal dose–response equation using non-linear regression (gray line) and the EC₅₀ values are quoted with 95% confidence intervals. (C) Schematic of the replicon assay to test the efficacy of AVI-4206 in A549 ACE2^h cells. (D) Luciferase readout of A549 ACE2^h cells infected with WA1 or WA1 Mac1 N40D replicons and treated with or without AVI-4206 and IFN-ɣ at indicated concentrations; *p < 0.05 by two-tailed Student’s t-test relative to the no AVI-4206 and no IFN-ɣ control. Results are plotted as normalized mean ± standard deviation luciferase values of a representative biological experiment containing two technical replicates. (E) Representative images of A549 cells stably expressing Mac1 and Mac1-N40D treated with IFN-γ and/or RBN012759 or AVI-4206. DMSO-treated cells are shown as vehicle control. Poly/mono ADPr signal comes from Poly/Mono-ADP Ribose (D9P7Z) Rabbit mAb (CST, 89190S) staining. (F) Relative mean cytoplasmic poly/mono ADPr intensity of cells from (F). Data shown as mean values ± SD. At least 8000 cells were analyzed in each group, from triplicate wells. Two-tailed Student’s t-test was used to compare ADPr intensity levels of each treatment. *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 3—figure supplement 1.. AVI-4206 has limited antiviral efficacy and no cytotoxicity in cellular models of infection.

Drug cytotoxicity of AVI-4206 in Vero-TMPRSS2 (A) and A549 ACE2^h (B) was measured using the CellTiter-Glo viability assay. Graphs represent the mean ± SD of three biological replicates each conducted in triplicate. Luciferase readout of VAT (C) and A549 ACE2^h (D) cells infected with WA1 or WA1 Mac1 N40D replicons and treated with or without AVI-4206 and IFN-ɣ at indicated concentrations. Results are plotted as normalized mean ± SD luciferase values of a representative biological experiment containing three technical replicates.

Figure 3—figure supplement 2.. Relative mean cytoplasmic poly/mono ADPr intensity of A549 cells stably expressing Mac1 and Mac1-N40D treated with IFN-γ and/or AVI-4206.

DMSO-treated cells are shown as vehicle control. Poly/mono ADPr signal comes from Poly/Mono-ADP Ribose (E6F6A) Rabbit mAb (CST, 83732S) staining. Data shown as mean values ± SD. At least 10,000 cells were analyzed each group, from triplicate wells. Two-tailed Student’s t-test was used to compare ADPr intensity levels of each treatment. *p < 0.05; **p < 0.01.

Figure 3—figure supplement 3.. ADP-ribosylation profiling during infection by Western Blot.

(A) Immunoblot showing pan-ADP-ribose (panADPr) levels in Calu-3 cells under indicated infection conditions. UI = uninfected cells, WT = cells infected with WA1, N40D = cells infected with WA1 NSP3 Mac1 N40D, with or without 100 μM AVI-4206 treatment. HeLa cells treated with H₂O₂ were included as a positive control for the ADPr signal. Actin serves as a loading control. (B) Densitometric analysis of panADPr levels normalized to actin. Data are presented as mean ± SD (n = 3), with p- values indicated. (C) Immunoblot showing mono-ADP-ribose (monoADPr) levels under the same conditions as (A). (D) Densitometric analysis of monoADPr levels normalized to actin. No statistically significant differences (ns) were observed between conditions. Data are shown as mean ± SD (n = 3).

Figure 4.. AVI-4206 displays efficacy in organoids and primary cell models.

(A) Schematic of the human airway organoid (HAO) experiment. (B) Viral particle production was measured by plaque assay at indicated time points and AVI-4206 concentrations. Error bars indicate SEM. **p < 0.01; *p < 0.05 by two-tailed Student’s t-test relative to the vehicle control. (C) Monocyte-derived macrophages (MDMs) were exposed to SARS-CoV-2, and virus particle production was assessed 24 hr later using a plaque assay. **p < 0.01 by two-tailed Student’s t-test relative to WA1. (D) Plaque assay of MDMs infected with WA1 or WA1 Mac1 N40D viruses and treated with AVI-4206 at indicated concentrations and IFN-γ at 50 ng/ml. *p < 0.05 by two-tailed Student’s t-test compared to the untreated control. (E) MDMs were incubated with AVI-4206 for 24 hr, after which cytotoxicity was assessed using an ATP-based cytotoxicity assay.

Figure 5.. AVI-4206 has a favorable pharmacological profile.

(A) Pharmacokinetic properties of AVI-4206. (B) Unbound plasma exposure time course of AVI-4206, corrected for plasma protein binding, following administration by IV, PO, or IP routes in male CD-1 mice at the indicated doses. (C) Free plasma exposure of AVI-4206 and total exposure in lung homogenate following an IP dose of 10 mg/kg in female C57BL/6 mice. (D) Inhibition of CYP isoforms by AVI-4206 at a fixed concentration of 10 μM. Two experiments were performed with CYP3A4 using different positive controls. (E) Heatmap of AVI-4206 activity in an off-target safety panel including receptors, ion channels, and proteases, showing no antagonist response >15% at 10 μM.

Figure 6.. AVI-4206 reduces viral replication and increases survival and cytokine abundance in vivo.

(A) K18-hACE2 mice were intranasally infected and dosed as indicated with either AVI-4206 (n = 15, intraperitoneally), nirmatrelvir (n = 5, per os), or vehicle (n = 10 for the AVI-4206 group or n = 5 for the nirmatrelvir group). Mice infected with the WA1 N40D mutant, which lacks Mac1 catalytic activity, served as a positive control (n = 10). Lungs were harvested at indicated time points for virus titration by plaque assay. (B) The percent body weight loss for all animals treated with AVI-4206 (100 mg/kg IP) (C) or nirmatrelvir (300 mg/kg PO). The data are presented as mean ± SD. *p < 0.05; **p < 0.01; ***p < 0.001 by two-tailed Student’s t-test relative to the vehicle control at each time point. (D) Survival curve plotted based on the percent weight loss humane endpoint (20%) for AVI-4206 and (E) nirmatrelvir. (F) Viral load in the lungs and brain of infected mice at the indicated time points. The data are shown as mean ± SEM. *p < 0.05; **p < 0.01 by Mann–Whitney’s test relative to the vehicle control. (G) Schematics and graphs demonstrating the abundance of indicated cytokines at 4 and 7 days post-infection in the lungs of infected mice. The data are presented as mean ± SEM. *p < 0.05; **p < 0.01 by two-tailed Student’s t-test relative to the vehicle control at each time point. None of the mice reached the humane endpoint at day 4 post-infection. For mice that reached the humane endpoint before day 7 post-infection, the tissues were collected and analyzed with mice at the 7-day time point.

Figure 6—figure supplement 1.. Lung histology.

We examined determinants of tissue damage by either (A) caspase 3 staining for apoptosis (**", signifies a p value of 0.0013 in a two tailed unpaired T-test analysis) or (B) Masson’s Trichrome stain for collagen deposition and pulmonary fibrosis. In the lungs of AVI-4206 treated animals, apoptosis is significantly reduced compared to the lungs of the vehicle cohort. While collagen deposition in the lungs of AVI-4206 treated animals is trending lower, the result is not significant. There is no difference in pathology between the N40D cohort and vehicle-treated cohort with these markers. This could suggest that AVI-4206 provides an additional mechanism that results in protection.

Figure 6—figure supplement 2.. Lower dose AVI-4206 reduces viral replication and increases survival in vivo.

(A) K18-hACE2 mice were intranasally infected with SARS-CoV-2 WA1 or SARS-CoV-2 WA1 Mac1 N40D mutant. Mice were treated as indicated with AVI-4206 (BID, 30 mg/kg) or vehicle. Each group was composed of n = 10 mice (5 mice per time point). (B) The percent body weight loss is presented as mean ± SD. **, p < 0.01; ***, p < 0.001 by two-tailed Student’s t-test relative to the vehicle control at each time point. (C) Survival curve based on the percent body weight loss humane endpoint. (D) Viral load in the lung at indicated time points is presented as mean ± SEM **, p < 0.01 by Mann –Whitney’s test relative to the vehicle control.

Figure 6—figure supplement 3.. AVI-4206 suppresses replication of mouse-adapted SARS-CoV-2 in wild-type mice.

(A) Wild-type mice were intranasally infected with SARS-CoV-2 and treated with AVI-4206, AVI-6451, or vehicle (n = 10 per group). Mice infected with the WA1 N40D mutant served as a positive control (n = 10). Lung tissues were collected at designated time points for viral titer analysis using a plaque assay. Viral load in the lungs of wild-type mice treated with AVI-4206 (B, day 2; C, day 5). **, p < 0.01 by Mann –Whitney’s test relative to the vehicle control.

Chemical structure 1.. 1-((8-Amino-9H-pyrimido[4,5-b]indol-4-yl)amino)pyrrolidin-2-one.

Chemical structure 2.. 1-Amino-5,5-dimethylpyrrolidin-2-one hydrochloride.

Chemical structure 3.. AVI-4051.

Chemical structure 4.. AVI-1501.

Chemical structure 5.. AVI-3367.

Chemical structure 6.. AVI-1500.

Chemical structure 7.. AVI-3762 and AVI-3763.

Chemical structure 8.. AVI-3764 and AVI-3765.

Chemical structure 9.. AVI-4636.

Chemical structure 10.. AVI-4206.