Dual Inhibitors of Main Protease (M^Pro) and Cathepsin L as Potent Antivirals against SARS-CoV2

Abstract

Given the current impact of SARS-CoV2 and COVID-19 on human health and the global economy, the development of direct acting antivirals is of paramount importance. Main protease (M^Pro), a cysteine protease that cleaves the viral polyprotein, is essential for viral replication. Therefore, M^Pro is a novel therapeutic target. We identified two novel M^Pro inhibitors, D-FFRCMKyne and D-FFCitCMKyne that covalently modify the active site cysteine (C145) and determined cocrystal structures. Medicinal chemistry efforts led to SM141 and SM142, which adopt a unique binding mode within the M^Pro active site. Notably, these inhibitors do not inhibit the other cysteine protease, papain-like protease (PL^Pro), involved in the life cycle of SARS-CoV2. SM141 and SM142 block SARS-CoV2 replication in hACE2 expressing A549 cells with IC₅₀ values of 8.2 and 14.7 nM. Detailed studies indicate that these compounds also inhibit cathepsin L (CatL), which cleaves the viral S protein to promote viral entry into host cells. Detailed biochemical, proteomic and knockdown studies indicate that the antiviral activity of SM141 and SM142 results from the dual inhibition of M^Pro and CatL. Notably, intranasal as well as intraperitoneal administration of SM141 and SM142 lead to reduced viral replication, viral loads in the lung, and enhanced survival in SARS-CoV2 infected K18-ACE2 transgenic mice. In total, these data indicate that SM141 and SM142 represent promising scaffolds on which to develop antiviral drugs against SARS-CoV2.

Graphical Abstract

Introduction

The worldwide impact of the coronavirus pandemic (COVID-19) on public health, safety, and economy has initiated significant research into the development of potent antivirals against severe acute respiratory syndrome coronavirus 2 (SARS-CoV2).^1–3 The dire need for direct acting antivirals (DAA) is highlighted by the emergence of several highly contagious SARS-CoV2 strains that can partially evade therapeutic antibodies and current vaccines.^4–9 Moreover, vaccines are not effective in those with compromised immune function or recommended for those who develop anaphylaxis. These issues demonstrate the pressing need for DAAs against SARS-CoV2 and other coronaviruses that may emerge in the future. Importantly, it is expected that a combination of vaccine and antiviral treatment will likely decrease morbidity and mortality more efficiently.

SARS-CoV2 is an enveloped, positive-sense, single-stranded RNA virus.^1–3 The genome sequence shares ~79% and 50% similarity to those of SARS-CoV and MERS-CoV; two other members of the betacoronavirus family that caused previous major outbreaks.^10–12 Following internalization of the virus into host cells via angiotensin converting enzyme 2 (hACE2) receptors, the S-protein on the viral surface is cleaved by host cell proteases, including cathepsins and transmembrane serine protease 2 (TMPRSS2).^13–15 Subsequent translation of the viral RNA leads to the formation of two polyproteins, pp1a and pp1ab. These polyproteins are cleaved into 16 non-structural proteins (nsp1–16) by two viral proteases, main protease (M^Pro) and papain-like protease (PL^Pro).^16–18 These proteases are essential for viral protein expression, viral genome replication, virion packaging, and viral genomic RNA processing. Therefore, M^Pro and PL^Pro are promising targets to develop DAAs against SARS-CoV2. Since M^Pro has a broader substrate profile, it is anticipated that M^Pro inhibitors will lead to better therapeutic outcomes.

M^Pro is a cysteine protease and its active site contains a catalytic dyad comprised of amino acids Cys145 and His41.^19–20 Homodimerization of M^Pro is crucial for enzymatic activity as it forms the S1 pocket of the substrate-binding site (Figure 1A). Like other betacoronaviruses, SARS-CoV2 M^Pro preferentially cleaves substrates with the consensus sequence: P2 (L/F/M/V), P1 (Gln), andP1′ (S/A) residues. Notably, human proteases with such substrate selectivity are rare. Since the onset of the pandemic, several M^Pro inhibitors have been reported and most of these covalently modify Cys145 with a variety of warheads, including α-ketoamides, α,β-unsaturated ketones, aldehydes, dihaloacetamides, and vinyl sulfones.^19–27 Recently, the US Food and Drug Administration issued an emergency use authorization for Pfizer’s Paxlovid, a combination of the M^Pro inhibitor, nirmatrelvir and the HIV protease inhibitor ritonavir; ritonavir blocks the rapid metabolism of nirmatrelvir by CYP3A (www.ClinicalTrials.gov identifier: NCT04756531; www.fda.gov/media/155050/download).²⁸ Despite its remarkable efficacy, this treatment regimen is not accessible to all patients as several other drugs are contraindicated with ritonavir due to impacts on their metabolism by CYP3A. Furthermore, Paxlovid is not recommended in patients with severe renal and/or hepatic impairment. These weaknesses along with the inevitable emergence of resistant variants necessitate the development of multiple M^Pro inhibitors with excellent cellular efficacy, metabolic stability, and pharmacokinetic properties.

Figure 1.

(A) Chemical structures of D-FF(R/Cit)-CMKyne and D-FF(R/Cit)-CMK. (B) M^Pro inhibition by D-FF(R/Cit)-CMKyne and D-FF(R/Cit)-CMK. Crystal structures of M^Pro in complex with D-FFR-CMKyne (C) and F-FFCit-CMKyne (D) (PDB ID: 7MAU and 7MAV). (E) Overlay of the crystal structures of M^Pro-D-FFR-CMKyne and M^Pro-D-FFR-CMK (PDB ID: 7MAT) complexes, indicating the importance of the N-terminal alkyne for inhibitor binding (Carbon atoms of D-FFR-CMKyne and D-FFR-CMK are highlighted in yellow and green, respectively).

In this paper, we report two novel M^Pro inhibitors, SM141 and SM142 that exhibit a unique binding mode in the active site of M^Pro. These two molecules also inhibit Cathepsin L (CatL) which is associated with viral entry into host cells. Dual inhibition of these two enzymes is attributed to the antiviral activity of these compounds both in SARS-CoV2 infected epithelial cells and K18-ACE2 transgenic mice.

Results and Discussion

Because several proteases that cleave after Gln (e.g., cathepsin B) also cleave after citrulline (Cit), we screened M^Pro against set of probes that were initially developed to identify novel proteases that cleave after Cit. Interestingly, these probes, D-FFR-CMKyne and D-FFCit-CMKyne, were highly potent M^Pro inhibitors. These compounds contain a C-terminal arginine or Cit followed by a cysteine-reactive chloromethyl ketone (CMK) warhead (Figure 1A). Additionally, they possess L-Phe and D-Phe residues at the P2 and P3 positions, respectively. The N-terminus of the probes contains an aliphatic alkyne handle to enable affinity enrichment of proteins via click chemistry. Gratifyingly, D-FFR-CMKyne and D-FFCit-CMKyne exhibited impressive biochemical IC₅₀ values of 20 ± 2 and 25 ± 5 nM, respectively, for the inhibition of M^Pro (Figure 1B, Table 1). Probes lacking the alkyne, D-FFR-CMK and D-FFCit-CMK, are 4–5-fold less potent inhibitors than their parents, indicating that the N-terminal capping group substantially increases the binding affinity of these molecules for M^Pro. Since these molecules contain a reactive electrophilic warhead, we performed time-dependent inhibition experiments and found that the second-order rate constants for enzyme inactivation (k_inact/K_I) for D-FFR-CMKyne and D-FFCit-CMKyne are (1.6 ± 0.1)×10⁶ and (1.9 ± 0.2)×10⁶ M⁻¹min⁻¹, respectively (Table 1, Figure S1).These rapid inactivation rates are consistent with the potent IC₅₀ values. While saturation was not observed in the k_obs versus [I] plots, likely due to a slow chemical step of inactivation, it was possible to estimate an apparent dissociation constant, K_i, from the initial portion of the progress curves. Notably, D-FFR-CMKyne and D-FFCit-CMKyne possess K_i values of 300 and 200 nM, respectively. Moreover, D-FFR-CMKyne and D-FFCit-CMKyne are more than 275-fold selective for M^Pro over PL^Pro (Table 1 and Figure S2).

Table 1.

Steady-state kinetic inhibition parameters of the inhibitors.

Inhibitor	MPro			PLPro	Cathepsin L		Cathepsin B
	IC50 (μM)	kinact/KI (M−1min−1)	Ki(app) (μM)	IC50 (μM)	IC50 (μM)	kinact/KI (M−1min−1)	IC50 (μM)
D-FFR-CMKyne	0.020±0.002a	(1.6 ± 0.1)×106	0.3 ± 0.04	5.5 ± 0.7	-	-	-
D-FFCit-CMKyne	0.025 ±0.005a	(1.9 ± 0.2)×106	0.2 ± 0.05	40 ± 5	-	-	-
D-FFR-CMK	0.10±0.010	-	-	5 ± 0.3	-	-	-
D-FFCit-CMK	0.13 ±0.015	-	-	>50	-	-	-
SM136	20 ± 3	(7.7 ± 0.3)×102	73 ± 4	-	-	-	-
SM138	23 ± 2	(6.5 ± 0.2)×102	92 ± 14	-	-	-	-
SM140	2.2 ± 0.2	(9.0 ± 0.2)×103	16 ± 2	>50	-	-	>50
SM142	1.8 ± 0.1	(1.1 ± 0.3)×104	10 ± 1	>50	0.145 ± 0.014	(1.1 ± 0.01)×105	40 ± 9
SM137	10 ± 2	(1.8 ± 0.2)×103	50 ± 6	-	-	-	-
SM139	10 ± 1	(1.5 ± 0.1)×103	65 ± 5	-	-	-	-
SM141	0.9 ± 0.1	(2.0 ± 0.2)×104	9 ± 1	>50	0.060 ± 0.001	(1.9 ± 0.5)×105	>50
SM143	0.9 ± 0.1	(2.1 ± 0.1)×104	7 ± 1	>50	0.073 ± 0.003	(1.4 ± 0.4)×105	>50
SM144	0.7 ± 0.1	(3.4 ± 0.1)×104	4.5 ± 0.5	>50	0.220 ± 0.026	(6 ± 0.2)×104	>50
SM145	0.8 ± 0.1	(2.4 ± 0.1)×104	5.3 ± 0.4	>50	0.200 ± 0.004	(2.5 ± 0.01)×104	>50

^aBecause the IC₅₀ value is close to the enzyme concentration used in the assay, these values should be considered apparent IC₅₀ values.

Crystal structures of these inhibitors in complex with M^Pro (Table S1) confirm that covalent bond formation occurred between Cys145 and the CMK warhead. The carbonyl on the CMK warhead H-bonds with the backbone amide of Cys145 and Gly143, likely positioning the warhead for nucleophilic attack (Figures 1C, D). Notably, the Arg and Cit at the P1 position of D-FFR-CMKyne and D-FFCit-CMKyne occupy the S1′ pocket of the substrate-binding site, make extensive van der Waals packing with Gly143 and form H-bonds with the backbone carbonyl on Thr26 (Figures 1C, D). These inhibitors do not occupy the S1 site, which stands in stark contrast to most inhibitors that utilize the γ-lactam Gln mimic that occupies the S1 site in M^Pro.^{19–20, 22, 29–31} Owing to the D-configuration of the P3 Phe, the P2 and P3 phenylalanines are both directed towards the same side of the inhibitor. In this orientation, the two phenyl rings form π-π stacking interactions and further form van der Walls contacts with the 187–192 loop which covers the S2-S4 pockets. This binding mode is also supported by several H-bonds between the P3 Phe and the backbone of Glu166 and between the P1 amide and the side chain of the catalytic His41. Notably, removal of the alkyne handle would impact H-bonds between the P3 Phe and Glu166 (Figure 1E), consistent with the 4–5-fold loss in potency.

Since the CMK warhead is highly reactive and is associated with off-target toxicity, we considered developing inhibitors with a less reactive warhead. Given the unique binding mode of our D-Phe-Phe motif, we wondered whether we could combine this functionality with the γ-lactam mimic of Gln which has been commonly exploited in other M^Pro inhibitors and binds in the S1 pocket.^19–28 To test this hypothesis, we conjugated the D-Phe-Phe motif to a γ-lactam-derived acrylate war-head to generate a series of hybrid inhibitors (Figure 2A and S3–22). Since the S2 and S4 pockets of M^Pro accommodate hydrophobic groups, we synthesized compounds containing 4-fluorophenylalanine and/or 4-fluoro-D-phenylalanine at the P2 and P3 positions. We also synthesized compounds that have an N-terminal Boc group as well as aliphatic alkyne functionality like the parent compounds. The 4-fluoro substitution on the P2 residue is not well accommodated in the active site of M^Pro, while the same substitution on the P3 residue affords reasonably potent inhibitors (Table 1). Interestingly, the presence of the N-terminal Boc group significantly diminishes the potency, while the most potent inhibitors in this series, SM144 (IC₅₀ = 0.7 μM; k_inact/K_I = (3.4 ± 0.1) × 10⁴ M⁻¹min⁻¹) and SM145 (IC₅₀ = 0.8 μM; kinact/K_I = (2.4 ± 0.1) 10⁴ M⁻¹min⁻¹) possess the aliphatic alkyne handle at the N-terminus (Table 1 and Figures S23–28). Notably, none of these compounds inhibit PL^Pro at up to 50 μM concentration, highlighting their excellent selectivity to M^Pro.

Figure 2.

(A) Chemical structures of SM136-145. Crystal structures of M^Pro in complex with SM141 (B, PDB ID: 7MB0), SM143 (D, PDB ID:7MB1) and SM144 (E, PDB ID: 7MB2). (C) Overlay of crystal structures of M^Pro-SM141 and M^Pro-SM137 (PDB ID:7MAX) complexes (Carbon atoms of SM137 and SM141 are highlighted in yellow and green, respectively. Carbon atoms of the protein residues for M^Pro-SM137andM^Pro-SM141 complexes are displayed in pink and white, respectively).

Co-crystal structures (Table S1) showed that all inhibitors form a thioether bond with the active site cysteine (Cys145), confirming the covalent nature of their inhibition. Similar to other known M^Pro covalent inhibitors, the γ-lactam glutamine mimetic occupies the S1 pocket of the substrate-binding site (Figures 2B–E and S29). As observed for D-FF(R/Cit)-CMKynes, the amide and carbonyl groups of the P3 residue H-bond with the backbone carbonyl and amide groups of Glu166. Also, the P1 amide moiety forms an H-bond with His164 (Figures 2B–E). Interestingly, the two aromatic groups at the P2 and P3 positions show “T” edge-to-face stacking rather than face-to-face stacking as observed for D-FFR-CMKyne and D-FFCit-CMKyne. Overlays of SM141 and SM137 show that the 4-fluoro-Phe at the P2 position clashes with the loop between residues D186-Q188 and M49, accounting for the poor activity of SM137 and SM139 (Figure 2C). However, the fluorine on 4-fluoro-Phe at the P3 of SM143 interacts with Gln192, likely explaining why it retains potency (Figure 2D). The N-terminal alkyne handle forms additional favorable interactions (Figure 2E), although it packs in a different conformation than in the structures of M^pro bound to the D-FF(R/Cit)CMKynes (Figures1C, D).

Given the nanomolar potency of our inhibitors, we next tested their antiviral efficacy in mammalian cells infected with SARS-CoV2. We independently treated human adenocarcinoma-derived alveolar basal epithelial A549 cells constitutively expressing hACE2 (A549-hACE2) and human hepatocellular carcinoma Huh7.5 cells with SARS-CoV2 at a multiplicity of infection (MOI) 0.05 for 1 h to allow for viral entry.³² Note that Huh7.5 cells are derived from Huh7 cells, which are an immortal cell line composed of epithelial-like, tumorigenic cells. Huh7.5 cells have been widely used for SARS-CoV-2 infection because of the expression of ACE2, however, compared to A549-hACE2 cells, Huh7.5 cells express lower levels of ACE2 and are consequently less susceptible to SARS-CoV-2 infection, showing lower RNA abundance and lower viral titers. Cells were then treated with an M^Pro inhibitor (2 μM final) for 24 h. Total RNA in the cells was then extracted, reverse-transcribed using the iScript cDNA synthesis kit (Bio-Rad) and diluted cDNAs were subjected to qPCR analysis using iQ SYBR Green Supermix reagent (Bio-Rad). Gene expression levels of SARS-CoV2 mRNA were normalized to GAPDH, a housekeeping gene. In both cell lines, D-FFR-CMKyne and D-FFCit-CMKyne exhibit poor antiviral activity (Figure 3A and B). This lack of efficacy may be due to poor cell permeability.

Figure 3.

Antiviral activity in A549-hACE2 (A) and Huh7.5 (B) cells evaluated at a concentration of 2 μM. (C) Dose-response curves for the inhibition of SARS-CoV2 infection of A549-hACE2 cells. The table indicates the corresponding EC₅₀ values.

By contrast, the hybrid inhibitors exhibited significantly higher antiviral efficacy with SM141 being the most potent. This compound almost completely blocks viral infection in both cell types at a concentration of 2 μM (Figure 3A and B). Intrigued by these results, we conducted a dose-response study in A549-hACE2 cells. SM141 exhibits EC₅₀ and EC₉₀ values of 8.2 and 22.1 nM, respectively, for the inhibition of SARS-CoV-2 infection (Figures 3C). These values are better than nirmatrelvir/PF-07331332, a reversible covalent M^Pro inhibitor that blocks SARS-CoV2 infection in A549-hACE2 cells with EC₅₀ and EC₉₀ values of 77.9 and 215 nM, respectively.²⁸ Importantly, SM141 does not cause any notable cytotoxicity up to 50 μM, whereas this concentration is only 3 μM for nirmatrelvir (Figure S30A). SM142 also significantly blocks SARS-CoV2 infection in A549-hACE2 cells with an EC₅₀ of 14.7 nM (Figure 3C). Notably, the somewhat more potent hybrid inhibitors, SM144 and SM145, exhibited significantly lower antiviral activity than SM141 and SM142. These results indicate that the N-terminal aliphatic alkyne functionality in SM144 and SM145 impairs their antiviral activity; the reasons for this are unclear. Although to a lesser extent than SARS-CoV2, SM141 and SM142 also inhibits OC-43, another member of the betacoronavirus family in A549-hACE2 cells (Figure S30B).

Although SM141 and SM142 exhibit potent antiviral activity in cells, the corresponding EC₅₀ values are more than 100-fold lower than their potency for the M^Pro inhibition. These observations suggested that SM141 and SM142 may target multiple proteins in A549-hACE2 cells that synergize with the inhibition of M^Pro. Since SM144 possesses an alkyne functionality and covalently modifies M^Pro, it represents an activity-based probe that can be coupled to reporter tags (e.g., TAMRA-azide or biotin-azide) using copper-catalyzed azide-alkyne click chemistry. Therefore, SM144 can be utilized to enrich its cellular targets on agarose beads. Enriched proteins are subsequently identified by tandem mass spectrometry. Treatment of M^Pro with SM144 followed by copper-catalyzed click chemistry in the presence of TAMRA-azide and visualization by in-gel fluorescence indicated that SM144 can dose-dependently label recombinant M^Pro (Figure 4A). Notably, the limit of detection is only 5 pmol of M^Pro (Figure 4B). Moreover, labeling of recombinant M^Pro by SM144 is competitively inhibited by SM141, confirming the engagement of M^Pro active site by both these molecules (Figure 4C). Next, we used SM144 to evaluate its selectivity for M^Pro in the presence of a complex proteome. For safety concerns, we chose to identify putative targets by spiking the A549-hACE2 lysates with recombinant M^Pro (250 nM) rather than using SARS-CoV2 infected cells. Importantly, SM144 selectively labels M^Pro under these conditions in a dose-dependent manner as was observed with recombinant protein. Fluorescence labeling was also dose dependently inhibited in the presence of the parent inhibitors, SM141 and SM142 (Figure 4D), consistent with M^Pro being the primary target of these inhibitors. In total, these results confirm the engagement of M^Pro by our M^Pro inhibitors in the presence of a complex proteome. As a further control, we also treated unspiked A549-hACE2 cell lysates with SM144 alone and in the presence of SM141 and SM142. Notably, no proteins were significantly labeled, indicating the high proteome-wide selectivity of these inhibitors (Figure 4E). Note that A549-hACE2 cells were used because this cell line was used for our antiviral assays.

Figure 4.

Dose-dependent labeling (A) and the limit of detection (B) of M^Pro by SM144. (C) Inhibition of M^Pro labeling by SM141. (D) Labeling of A549-hACE2 lysate spiked with recombinant M^Pro by SM144 and competitive inhibition with SM141andSM142. The triangle indicates M^Pro. (E) Labeling of A549-hACE2 lysate by SM144 in the presence and absence of SM141andSM142. Volcano plot indicating the proteins in the A549-hACE2 lysate enriched by SM144 in the absence (F) and presence (G) of SM141.

We further evaluated the proteome-wide selectivity via an orthogonal chemoproteomic assay. A549-hACE2 lysates spiked with recombinant M^Pro were treated with SM144 in the presence and absence of a competitive inhibitor, SM141. Subsequently, the labeled proteins were biotinylated with copper-catalyzed click chemistry in the presence of Biotinazide, and the biotinylated proteins were selectively enriched on Streptavidin-agarose beads, proteolyzed with trypsin and analyzed by tandem mass spectrometry. These studies indicate that M^Pro is the most statistically significantly enriched protein, again confirming the proteome-wide selectivity of SM144 (Figure 4F). There were, however, a few off-targets, including glutathione reductase (GSHR), S-adenosylhomocysteinase (SAHH), aldose reductase (ALDR), S-adenosylmethionine synthase isoform type-2 (METK2), gamma-actin (ACTG), tubulin alpha-4A chain (TBA4A), ubiquitin-like modifier-activating enzyme 1 (UBA1) and junction plakoglobin (PLAK). Notably, all these proteins are negligibly enriched in the presence of SM141, confirming that these proteins are bona fide cellular targets of our M^Pro inhibitors (Figure 4G). However, none of these proteins are related to the entry, replication, and/or exocytosis of SARS-CoV2 in host cells, indicating that their modification is unlikely to contribute to the antiviral activity of our M^Pro inhibitors and may simply be labeled because they are highly abundant proteins with reactive cysteines.

Cathepsins L and B (CatL/B) play important roles by cleaving the Spike protein and thereby facilitating the release of SARS-CoV2 genomic RNA into the cytosol of a host cell.^{13, 33–35} CatL/B are mainly expressed in the lysosome and therefore, cleave the S protein only after the virus is internalized into host cell through endocytosis. Recent studies suggest that the circulating levels of cathepsin L (CatL) are significantly elevated in COVID-19 patients after SARS-CoV2 infection. SARS-CoV2 pseudovirus infection also elevates CatL expression in human cells as well as in hACE2 transgenic mice.³⁶ Notably, both these cathepsins are cysteine proteases that can be targeted with small molecule inhibitors containing cysteine-reactive warheads. Several reports indicated that the inhibition of either of these Cathepsins, or both, markedly reduces SARS-CoV2 infection.^37–40 Bogyo and coworkers showed that most of the known M^Pro inhibitors also potently inhibit CatL and B.⁴¹ However, we did not find significant enrichment of CatL/B in our chemoproteomic analyses (Figure 4F and G), possibly because CatL and CatB are inactivated under standard lysis conditions. As such, we next determined whether our M^Pro inhibitors inhibit CatL/B using commercially available kits. Interestingly, SM141-145 are excellent inhibitors of CatL with SM141 being the most potent (Table 1, and Figures S31A and S32). The IC₅₀ and k_inact/K_I values indicate that SM141 is 10-fold more potent for CatL inhibition than M^Pro. However, none of these compounds appreciably inhibit CatB (IC₅₀>50 μM), indicating that SM141 likely exerts its antiviral activity by selectively targeting both M^Pro and CatL (Figure S31B). Since M^Pro acts downstream of the action of CatL, treatment with SM141 can simultaneously block viral replication as well as further entry of the virus into the host cell. SM141 may also inhibit excess CatL that is expressed post-infection, further contributing to its antiviral activity.

To help confirm that the observed antiviral activity is at least partially due to the inhibition of CatL, we evaluated SARS-CoV2 pseudovirus entry into A549-hACE2 and A549-hACE2/TMPRSS2 cells. Since SARS-CoV2 does not explicitly require CatL for entering host cells when TMPRSS2 is overexpressed, the efficacy of SM141-145 is expected to be reduced if their antiviral activity is driven by CatL inhibition. Consistent with that notion, we do observe a significant reduction in efficacy for SM141-145 when TMPRSS2 is overexpressed (Figure S33A, B). Thus, SM141–145 inhibit SARS-CoV-2 pseudovirus entry by inhibiting CatL.

Next, we evaluated the effect of TMPRSS2 overexpression using live native virus (USA-WA1/2020/NR-52281; BEI Resources). For these experiments, A549-hACE2 and A549-hACE2/TMPRSS2 cells were incubated with SARS-CoV-2 for 1 h and then treated with SM141 and SM142 (Figure S33C). We observed only a partial loss in efficacy upon overexpression of TMPRSS2. These results suggest that SM141 and SM142 inhibit SARS-CoV2 infection by inhibiting both M^Pro and CatL.

To further validate our data suggesting that SM141 and SM142 inhibit both CatL and M^pro, we used siRNAs to knockdown CatL expression in A549-hACE2 cells (Figure S34A and B). Knockdown of CatL led to a significant reduction in pseudovirus cell entry (Figure S34C) and live virus infection (Figure S34D) confirming that CatL is important for viral entry. Notably, in CatL knockdown cells infected by SARS-CoV-2 pseudovirus, SM141 or SM142 did not show any significant additional inhibitory activity on SARS-CoV-2 entry. These data are consistent with CatL being targeted by SM141 or SM142 (Figure S34E). However, SM141 or SM142 do inhibit live virus infection, relative to the DMSO control, in the knockdown cells indicating that SM141 and SM142 inhibit SARS-CoV2- infection independent of CatL (Figure S34F). Altogether, these results demonstrate that SM141 or SM142 function as dual inhibitors of CatL and M^pro.

To demonstrate that CatL is a direct target of our inhibitors, A549-hACE2 cells were preincubated with DMSO or SM141 (5 μM) for one h. SM144 (5 μM) was then added, and the cells were incubated for 24 h. Labelled proteins were then coupled to biotin-azide using copper-catalyzed azide-alkyne click chemistry. SM144 labeled proteins were visualized before and after pulldown on streptavidin-agarose by Western blotting. Notably, we see predominant labeling and isolation of an ~30 kDa protein, a molecular weight that is consistent with mature CatL (Figure S35A). To confirm that SM144 can isolate CatL, we performed western blotting using an anti-CatL antibody. Indeed, SM144 treatment leads to the enrichment of CatL (Figure S35B). By contrast, SM141 completely blocks the enrichment of CatL consistent with both SM141 and SM144 being able to covalently modify CatL in cells (Figure S35B). We further evaluated the proteome-wide selectivity of SM144 in live A549-hACE2 cells by performing quantitative proteomics on the enriched proteomes. Notably, CatL was the most highly enriched protein (>128-fold) (Figure S35B) and SM141 could effectively prevent the enrichment of this protein. In addition to CatL, we observed significant enrichment of cathepsins B and S as well as several cytosolic and mitochondrial proteins including HSP74 and aconitase. These data suggest that SM144 can label a combination of lysosomal and cytosolic proteins.

Given the appreciable cellular potency of SM141 and SM142, we next investigated their in vivo antiviral efficacy against SARS-CoV-2 infection. For these studies, we used K18-hACE2 transgenic mice (K18-ACE2) that express human ACE2 under the control of the epithelial cell cytokeratin-18 (K18) promoter; this model provides robust SARS-CoV-2 infection.³² Notably, both compounds are significantly nontoxic as they cause no significant weight loss in mock infected mice (Figure S36A). Intranasal infection of K18-ACE2 mice with SARS-CoV2 results in significant weight loss and lethality 7 days post-infection. However, intranasal treatment with a once daily dose of SM141 or SM142 for 3 days starting 2 h prior to infection protects mice from SARS-CoV2-induced weight loss and lethality (Figures 5A and B). Moreover, the post-infection administration of SM141 or SM142 via intraperitoneal injection twice daily for 5 days also protects K18-ACE2 mice from SARS-CoV-2-induced weight loss and lethality (Figures 5C and D). Notably, the survival of mice is significantly improved in the latter treatment, indicating that these compounds exhibit better antiviral efficacy when administered intraperitoneally. Of the two compounds, SM141 exhibits better antiviral efficacy (Figures 5B and D) likely due to the higher potency for M^Pro and CatL inhibition. We also investigated the viral RNA loads and titers of SARS-CoV-2 in the lung tissue after treatment with SM141 and SM142. As shown in Figures 5E–G, the viral RNA loads and titers in lung tissues are remarkably lower in the inhibitor-treated mice than in those treated with vehicle control. We detected reduced inflammatory cytokines and chemokines, including IFN-β, TNF-α, IL-1β and IL-6 in the lung tissues of inhibitor-treated mice consistent with lower viral loads (Figure S36B–E). In agreement with these results, histopathological analysis of the lung tissues also shows a significant reduction in lung inflammation and improved overall lung pathology upon SM141 or SM142 treatment for 5 days (Figures 5H–I), suggesting that these compounds not only reduce viral loads but also ameliorate SARS-CoV2-triggered inflammatory cytokine storm and lung damage.

Figure 5.

Weight loss (A) and survival (B) of K18-ACE2 transgenic mice infected with SARS-CoV2 (3 × 10⁴ PFU/mouse) after intranasal treatment with SM141 or SM142 (10 mg/kg, once 2 h prior to the infection and two more doses on the consecutive days after infection). Weight loss (C) and survival (D) of K18-ACE2 transgenic mice infected with SARS-CoV-2 (3 × 10⁴ PFU/mouse) after intraperitoneal treatment with 25 mg/kg SM141 or SM142 twice daily for 5 days. SM141 and SM142 were first dissolved in 100% DMSO at 0.1 mg/μL, then diluted with PBS (1:20) to the final concentration 0.005 mg/ μL. The final concentration of DMSO is 5% with PBS as a working stock. qPCR analysis of SARS-CoV2-N (E), Nsp14 (F) and viral titer (G) in the lung tissue of K18-ACE2 transgenic mice infected with SARS-CoV2 after 72 h treatment with vehicle, SM141, or SM142. (H) Representative images of H&E-stained lung sections from K18-ACE2 transgenic mice infected intranasally with SARS-CoV-2 with intraperitoneal injection of SM141, SM142 or vehicle control for 5 days. A scale bar is shown in the mock sample and is the same for all other groups: 2×, 0.5 mm; 20×, 50 μm. (I) Pathology evaluations were performed by board certified pathologists and the pathology score was determined based on the inflammation, fibrosis and endothelial changes of each samples.

We next evaluated the pharmacokinetic properties of SM141 and SM142 after administration into BALB-6c mice (Table 2). Importantly, no adverse events or signs of toxicity were observed for either compound at the tested doses. As would be anticipated based on its structure, SM141 is more soluble which allows for easy formulation in 5% DMSO : 95% saline. SM141 had a relatively short half-life of 0.8 hours and high observed clearance of 72 mL/min/kg. A comparison of the AUC from the oral and IV dose was used to calculate an oral bioavailability of 5%. The low bioavailability is believed to be due to high first-pass metabolism and not due to poor absorption due to solubility or permeability limitations. This is supported by the rapid absorption of the intraperitoneal dose and a calculated volume of distribution of 1.6 L/kg. Whereas suspension formulations can be used for oral doses, intravenous dosing requires a solution. For this reason, the nonionic surfactant/emulsifier, Tween 80 was added at a final concentration of 5% to obtain a clear solution. The more lipophilic SM142 had a high volume of distribution of 13 L/kg indicating extensive tissue distribution of the compound. Tissue distribution limits hepatic exposure and likely contributes to the improved half-life of 2.1 h. The improved oral bioavailability of SM142 was encouraging with C_max = 0.6 μM and a fraction bioavailable = 38%, both being superior to oral SM141. The clearance of SM142 was higher than the hepatic blood flow in a mouse, 90–110 ml/min/kg, indicating that there may be extrahepatic clearance mechanisms that contribute to the overall clearance of SM142. This will be explored during the design of future analogs.

Table 2.

Pharmacokinetic parameters of SM141 and SM142.

Parameters		SM141		SM142
	IV	PO	IP	IV	PO
T1/2 (h)	0.8	ND	0.8	2.1	2.1
Tmax (h)	0.08	0.8	0.14	0.08	0.08
Cmax (μM)	3.9	0.2	2.7	0.4	0.6
AUClast (μM.h)	1.3	0.14	1.6	0.4	0.5
AUCINF_obs (min.ng/mL)	42350	7325	49620	18490	20843
CI_Obs (mL/min/kg)	71.5		-	169	-
MRTINF_obs (h)	0.4		-	1.9	-
VSS_obs (l/kg)	1.6		-	13.2	-
% Bioavailability	-	5	35.2	-	37.5

IV: Intravenous dosing; IP: Intraperitoneal dosing; PO: Oral dosing. For SM141 the final formulation was 5/95 DMSO/saline and for SM142 the final formulation was 5% DMSO, 5% Tween-80, 90% saline (v:v:v).

Conclusion

In conclusion, we identified a novel D-FF motif that exhibits a unique binding mode in the active site of M^Pro when conjugated with an electrophilic cysteine-targeted warhead at the C-terminus. Medicinal chemistry studies on this scaffold afforded SM141 and SM142 which inhibit M^Pro with nanomolar potency. Interestingly, SM141 and SM142 inhibit the replication of SARS-CoV2 in A549 cells expressing human ACE2 receptor with IC₅₀ values of 8.2 and 14.7 nM, respectively. We found that this cellular antiviral efficacy results from the combined inhibition of M^Pro and CatL by these inhibitors. Furthermore, intraperitoneal administration of SM141 and SM142 into SARS-CoV2 infected K18-hACE2 mice markedly inhibits viral replication as well as reduces lung damage and inflammation, indicating that these compounds are potential candidates for the development of antivirals against SARS-CoV2.

ABBREVIATIONS

M^Pro

Main Protease

PL^Pro

Papain-like protease

Cat

Cathepsin

IFN

Interferon

TNF

Tumor necrosis factor

IL

Interleukin