Reversible Inhibitors of SARS CoV-1 3CLpro Peptidyl HMK inhibitors and their corresponding ethers have been designed as potent reversible cysteine protease inhibitors of cathepsin K34 and cruzain.43 High-resolution crystal structures of two such inhibitors complexed with cruzain highlighted the presence of a strong hydrogen bond between the catalytic His and the α-hydroxyl moiety of the inhibitor but interestingly no covalent bond between the active-site Cys and the ketone pharmacophore. Utilizing the crystal structure of 28 in complex with SARS CoV-1 3CLpro, molecular modeling of an HMK-based design indicated the potential for establishing a covalent bond between the sulfur of the active-site cysteine (Cys145) and the HMK carbonyl as well as hydrogen bond interactions between catalytic His41 and the terminal hydroxy moiety. Additionally, the structure of 28 suggested the possibility to increase the lipophilic filling of the deep hydrophobic S2 pocket as a strategy to increase binding affinity. To evaluate these structure-based observations, a series of reversible HMK inhibitors containing P2 diversity were prepared as depicted in Table 3. Table 3 SARS 3CLpro Inhibition Data for P2-Modified HMK Inhibitors a&b See the Experimental Section for details on assay methods; the values were calculated from at least eight data points with at least two independent determinations. c This value is a Ki measurement. d These values are from a single determination. As suggested by molecular modeling, the S2 site appears to accommodate a variety of linear, branched, and cyclic alkyl moieties (entries 36 and 39–41). Interestingly, the P2 Phe derivative 42 displayed less potent inhibition of the SARS CoV-1 3CLpro than the corresponding saturated analogue 41. Also noteworthy is the large attenuation in enzyme inhibition seen with P2-N-methyl-Leu inhibitor 38 (IC50 = 83 nM) as compared to 4 (Ki = 4 nM). This loss of potency is consistent with the inhibitor ligand interactions observed with 2 and 4 that show the P2 NH involved in a hydrogen bond with the side-chain amide of Gln189. Additionally, the methyl substitution in 38 would be expected to alter the conformation of the 4-methoxy indole cap and perturb the observed ligand–enzyme hydrogen bond network present in 2 and 4. The results for 4 and the other examples in Table 3 reveal a high antiviral EC50/enzyme IC50 ratio, which may arise from poor cell permeability. Indeed, these HMK inhibitors exhibit very low permeability and high levels of efflux beyond the sensitivity of the Caco-2 in vitro assay. However, the impact of high efflux on antiviral potencies from Vero 76 cells (derived from monkey kidney) versus disease-relevant human lung epithelial cells is unknown. To better understand the observed high antiviral EC50/enzyme IC50 ratio for 4, we evaluated the role of efflux in the Vero 76 cell line by the in vitro experimental design discussed further below. Concurrently, a strategy to design molecules reducing efflux by active transporters, such as P-glycoprotein, was pursued to decrease high antiviral EC50/enzyme IC50 ratios. An analysis of the physicochemical properties of 4 suggested that increasing logP, reducing polar surface area (PSA), and reducing the number of hydrogen bond donors/acceptors were design strategies with the potential to improve cellular permeability and reduce efflux. An examination of the crystal structure of 2 in complex with SARS CoV-1 3CLpro suggested a limited functional role for the 4-methoxy substituent contained in the P3 indole cap. Additionally, the terminal α-hydroxyl moiety located at P1′ appeared to provide a potential opportunity to remove a hydrogen bond donor. Although it was anticipated that the α-hydroxyl moiety would form a hydrogen bond to the catalytic His41, the directionality of this interaction was uncertain. As noted earlier, His41 presents a hydrogen bond donor to the interior of the protease, leaving an acceptor to participate in the proton transfer cycling from neutral to cationic in the putative catalytic mechanism44 of proteolysis. If His41 is protonated in the inhibitor-bound complex, a lone pair from the α-hydroxyl moiety would function as a hydrogen bond acceptor from the proton on the His residue. On the other hand, if His41 is deprotonated, the hydrogen of the α-hydroxyl moiety would act as a hydrogen bond donor to the nitrogen lone pair of the His residue. Consistent with the strategy to decrease PSA and hydrogen bond donors, a series of inhibitors were prepared that contained alterations to the P3 cap and substitution of the α-hydroxyl group. Table 4 summarizes this effort where two of the optimal P2 residues (Leu and β-tert-butyl-Ala) were conserved. Removal of the methoxy group from the indole generally led to slightly weaker potency in both the enzymatic and antiviral assays across pairs (43 vs 36, 44 vs 37, 50 vs 48 and 49 vs 4) and therefore no improvement in the antiviral/protease inhibition ratio. The impact of more significant changes to the P3 cap is illustrated by tetrahydrofuranyl derivatives 46 and 47. These derivatives lack the lipophilic aryl ring and NH hydrogen bond donor present in the indole P3 cap and exhibit a pronounced reduction in enzymatic and antiviral potency. A noteworthy reduction in biochemical and antiviral potency was observed in comparing each hydroxymethylketone derivative with its corresponding ether counterpart. These results suggest that modest structural and physicochemical alterations of 4 fail to significantly decrease the high EC50/IC50 ratios for this class of peptidomimetics. However, in vitro analysis of 4 revealed this derivative to possess high levels of metabolic stability in human liver microsomes (t1/2 = 107 min).45 Further evaluation of this breakthrough derivative in several aqueous-based formulations suitable for IV administration indicated high levels of solubility.46 Table 4 SARS CoV-1 3CLpro Inhibition Data for P3- and P2-Modified Hydroxymethyl and Alkoxymethylketone Inhibitors a&b See the Experimental Section for details on assay methods; the values were calculated from at least eight data points with at least two independent determinations. c This value is a Ki measurement. d These values are from a single determination. Cocrystal structures of 4 bound to the 3CLpro of both SARS CoV-1 and CoV-2 were solved at 1.47 and 1.26 Å resolutions, respectively (PDB codes 6XHL and 6XHM).47 As expected, the ligand binding sites of the main protease from SARS CoV-1 and SARS CoV-2 are conserved in sequence and are nearly identical structurally. The schematic diagram in Figure 5 is representative of the covalent adduct between 4 and the 3CLpro from both CoV-1 and CoV-2. The warhead HMK carbonyl carbon of 4 forms a covalent bond to the 3CLpro active-site cysteine (Cys145) sulfur generating a tetrahedral carbinol complex (1.8 Å C–S bond length). This carbinol hydroxyl forms hydrogen bonds with the backbone NH of Cys145 and with the amide NH of Gly143 via a bridging water molecule. Another key active-site interaction is the hydrogen bond between the primary alcohol moiety of 4 and the catalytic His41. Similar interactions between catalytic His residues and the OH moiety of HMK in other protease–inhibitor complexes have been reported.43 Figure 5 Cocrystal structure of the covalent adduct of 4 bound to SARS CoV-2 3CLpro (6XHM). The Connolly surface for the inhibitor binding pocket is shown in gray. The bonds are represented as dashed lines, with the bond length between heavy atoms depicted. The schematic rendering of the active site with dashed lines represented as hydrogen bonds with key residues and curved lines to show S1 and S2 binding pockets is depicted. In S1, the lactam carbonyl of 4 forms a strong hydrogen bond with the side chain of His163, while the lactam NH is within the hydrogen-bonding distance but poorly aligned with the side-chain oxygen of Glu166 and the backbone oxygen of Phe140. These three residues partially define the bottom and edge of the S1 pocket. Additionally, the backbone carbonyl and NH of Glu166 form β-sheet-like hydrogen-bonding interactions with the NH and C2-carbonyl of the indole fragment of 4. The NH of the P2 Leu of inhibitor 4 and the side chain of Gln189 form a hydrogen bond, while the carbonyl of P2 Leu is exposed to the solvent. The inhibitor P1 NH forms a strong hydrogen bond with the backbone carbonyl of His164. The inhibitor’s lipophilic leucine residue is bound to the SARS 3CLpro in the hydrophobic S2 pocket formed by residues Asp187, Arg188, Gln189, Met49, and His41. The indole portion of 4 does not protrude into the S3 pocket but rather rests across a partially collapsed S3, making extended van der Waals interactions across the 189–191 residue backbone atoms. The closed S3 is the only significant conformational difference compared to a recently reported structure of SARS CoV-2 3CLpro inhibited by α-ketoamides.8 The energetics of this extended interaction, the associated protein conformation stabilities, or warhead difference may contribute to the 3-order-of-magnitude increased SARS CoV-2 3CLpro potency of HMK 4 (see below) compared to that of the referenced ketoamides. Hydroxymethylketone 4 was evaluated against a panel of other viral and human proteases (Table 5). In general, 4 is highly selective for 3CLpro inhibition, displaying IC50 values of >10 μM against many of the other proteases and possessing modest levels of inhibition of cathepsin B (IC50 = 1.3 μM) and rhinovirus 3Cpro (IC50 = 1.7 μM). High levels of similarity between the catalytic sites of 3CLpro proteases suggest the potential of 4 as a pan-coronavirus inhibitor. Testing revealed that 4 demonstrates potent inhibition of the related hCoV 229E protease (IC50 = 0.004 μM) but more importantly, given the current global pandemic, potently inhibits SARS CoV-2 3CLpro with a Ki of 0.27 nM. Table 5 Protease Inhibition by 4 protease source % inhibition @ 10 μMa IC50 (μM)a 3CLpro SARS CoV-1   0.004 ± 0.0003b 3CLpro SARS CoV-2 0.00027 ± 0.0001b cathepsin B human 93 1.3 ± 0.1 cathepsin D human 3 >10 leukocyte elastase human –31 >10 chymotrypsin human 8 >10 thrombin human –28 >10 pepsin human –2 >10 caspase 2 human –20 >10 HIV-1 HIV-1 0 >10 HCV HCV 16 >10 HRV HRV   1.7 ± 1 hCoV 229E human   0.004 ± 0.0004 a See the Experimental Section for details on assay methods; the values were calculated from at least eight data points with at least two independent determinations. b For SARS CoV-1 and CoV-2 3CLpro, Ki values are reported here. The antiviral specificity of 4 was evaluated against a panel of viruses (Table 6). No antiviral activity was observed when 4 was tested against two human rhinovirus strains (HRV-14 and HRV-16), human immunodeficiency virus-1 (HIV-1), or human cytomegalovirus (HCMV) in cell culture. This compound also does not inhibit the HCV replicon. As anticipated by the potent hCoV 229E protease activity, 4 is a potent inhibitor of the human coronavirus hCoV 229E in MRC-5 cells (human lung-derived) with an EC50 of 0.090 μM. Table 6 Viral Inhibition by 4 virus EC50 (μM)a CC50 (μM)a SARS CoV-1 4.8 ± 2.2 452 hCoV 229E 0.09 ± 0.04 >40 HRV-14 >100 >1,260 HRV-16 >100 >1,260 HIV-1 RF >32 >32 HCV replicon >320 >320 HCMV RC256 >100 >100 a See the Experimental Section for details on assay methods; the values were calculated from at least eight data points with at least two independent determinations. The hCoV 229E antiviral potency is more than 50-fold greater than that for SARS CoV-1 in the Vero 76 line, even though 4 possesses nearly equivalent 3CLpro biochemical potency for the hCoV 229E and CoV-1 viruses. While undefined differences between hCoV 229E and SARS CoV-1 could account for these large differences in antiviral EC50/enzyme IC50 ratios, different active transporter expression levels between MRC-5 and Vero 76 cell lines may also be a contributing factor. Consistent with the previously stated goal to characterize the impact of efflux on potency measured in the SARS CoV-1 antiviral assay, we co-dosed a known P-glycoprotein transport inhibitor at a fixed concentration (0.5 μM CP-100356)48 in a full dose–response of inhibitor 4. This resulted in a pronounced potency shift of SARS CoV-1 antiviral activity (EC50 = 0.11 μM) that now more closely matched the antiviral potency seen for hCoV 229E. As a control, the same co-dosing of efflux inhibitor led to no such shift in the hCoV 229E antiviral potency of 4, demonstrating that the human lung-derived MRC-5 cells show less P-glycoprotein-based transporter under similar assay conditions. These results suggest that the high antiviral EC50/enzyme IC50 ratio observed in Vero 76 cells is an artifact of the high efflux potential of that assay cell line and may underestimate the antiviral potency in human lung cells, the relevant tissue for SARS and COVID-19. Assays to directly measure the antiviral potency of 4 against SARS CoV-2 in clinically relevant human lung cells are currently under investigation.