Design, synthesis and crystallographic analysis of nitrile-based broad-spectrum peptidomimetic inhibitors for coronavirus 3C-like proteases.

Coronaviral infection is associated with up to 5% of respiratory tract diseases. The 3C-like protease (3CL(pro)) of coronaviruses is required for proteolytic processing of polyproteins and viral replication, and is a promising target for the development of drugs against coronaviral infection. We designed and synthesized four nitrile-based peptidomimetic inhibitors with different N-terminal protective groups and different peptide length, and examined their inhibitory effect on the in-vitro enzymatic activity of 3CL(pro) of severe-acute-respiratory-syndrome-coronavirus. The IC(50) values of the inhibitors were in the range of 4.6-49 μM, demonstrating that the nitrile warhead can effectively inactivate the 3CL(pro) autocleavage process. The best inhibitor, Cbz-AVLQ-CN with an N-terminal carbobenzyloxy group, was ~10x more potent than the other inhibitors tested. Crystal structures of the enzyme-inhibitor complexes showed that the nitrile warhead inhibits 3CL(pro) by forming a covalent bond with the catalytic Cys145 residue, while the AVLQ peptide forms a number of favourable interactions with the S1-S4 substrate-binding pockets. We have further showed that the peptidomimetic inhibitor, Cbz-AVLQ-CN, has broad-spectrum inhibition against 3CL(pro) from human coronavirus strains 229E, NL63, OC43, HKU1, and infectious bronchitis virus, with IC(50) values ranging from 1.3 to 3.7 μM, but no detectable inhibition against caspase-3. In summary, we have shown that the nitrile-based peptidomimetic inhibitors are effective against 3CL(pro), and they inhibit 3CL(pro) from a broad range of coronaviruses. Our results provide further insights into the future design of drugs that could serve as a first line defence against coronaviral infection.

Introduction

Coronaviruses (CoVs) cause a number of respiratory tract and gastroenteritis diseases in human and animals [1]. Coronaviruses can be classified into three main groups: alphacoronavirus (group 1), betacoronavirus (group 2), and gammacoronavirus (group 3), while the group 2 is further sub-divided into group 2a and 2b [2]. In addition to the SARS-CoV (group 2b) that caused the outbreak of pneumonia in 2003, the four human group 1 (229E and NL63) and 2 (OC43 and HKU1) strains are associated with up to 5% of respiratory tract disease cases reported [3], [4]. The single-stranded RNA genome of CoVs encodes two polyproteins, which are processed into at least 15 mature proteins by papain-like protease and 3C-like protease (3CLpro) [5], [6]. Inhibition of the proteolytic processing can reduce viral replication and viral-induced cytotoxic effect [5], [6]. The ability of coronaviruses to generate novel strains with high virulence through high frequency of mutations and recombination remains a potential threat to human health. Therefore, it is desirable to develop a broad-spectrum drug that can combat against all kinds of coronavirus infection. Because 3CLpro is essential to viral replication and its active-site structure is conserved, the protease is an attractive drug target for the development of such a broad-spectrum inhibitor [7]. The SARS-CoV 3CLpro can be inactivated by peptidomimetic inhibitors that are composed of a substrate-like peptide and a reactive warhead such as an aldehyde [8], [9], a Michael acceptor [7], [10], [11], a halo-methyl ketone [10], [12], or an epoxide [13]. The peptide interacts with the substrate-binding cleft of the protease to form a non-covalent complex, and then the warhead covalently links with the catalytic cysteine residue and inhibits the enzyme.

In this study, we have shown that the nitrile group is an effective warhead against 3CLpro. Here, based on the auto-cleavage sequence of SARS-CoV 3CLpro (TSAVLQ↓), four nitrile-based inhibitors with different length of peptide sequences and protective groups were designed and synthesized. We have shown that these inhibitors were effective against 3CLpro from six different strains of CoVs with IC50 values in the μM range. The crystal structures of the four enzyme–inhibitor complexes were determined to reveal the mechanism of inhibition and the enzyme–inhibitor interactions.

Results and discussion

SARS-CoV 3CLpro is inhibited by nitrile-based peptidomimetic inhibitors

Based on the autocleavage tetrapeptide sequence (AVLQ) of SARS-CoV 3CLpro, three nitrile-based inhibitors with different protective groups, 5-methylisoxazole-3-carboxyl (Mic), tert-butyloxycarbonyl (Boc) and carboxybenzyl (Cbz), were synthesized (Scheme 1 ). The protease activity in the presence of inhibitors was measured by using a fluorescent-protein substrate as described previously [14], [15]. The IC50 values of ‘Mic-AVLQ-CN’, ‘Boc-AVLQ-CN’, and ‘Cbz-AVLQ-CN’ were 49 ± 2, 49 ± 2, and 4.6 ± 0.2 μM, respectively (Fig. 1 ). Interestingly, the inhibitor with the Cbz protective group was ∼10x more potent than the other inhibitors tested. To test the effect of increasing the length of the peptidomimetic inhibitor, we synthesized a hexapeptide inhibitor ‘Cbz-TSAVLQ-CN’. The IC50 value for this inhibitor was 39 ± 1 μM (Fig. 1), suggesting that increasing the length of the inhibitor did not improve inhibition and the Cbz group at the P4 position may be responsible for the higher potency of ‘Cbz-AVLQ-CN’.

Scheme 1

Reaction scheme of nitrile-based peptidomimetic inhibitor synthesis. Step (a) was the coupling of peptides by using N,N′-dicyclohexylcarbodiimide and N-hydroxysuccinimide. Step (b) was the coupling of peptides by the mixed anhydride method. Step (c) was the deprotection of benzyl carbamates by using palladium on charcoal. Step (d) was the deprotection of Boc, Trt, Pbf groups and tertiary butyl esters by using trifluoroacetic acid. Step (e) was the transformation of primary amide to nitrile by using trifluoroacetic anhydride. Final products were highlighted with dashed box.

Fig. 1

IC50 values of the nitrile-based inhibitors against SARS-CoV 3CLpro. (A) A schematic diagram shows chemical structures and IC50 values of the inhibitors. (B) SARS-CoV 3CLpro was pre-mixed with ‘Miu-AVLQ-CN’ (circle), ‘Boc-AVLQ-CN’ (triangle), ‘Cbz-AVLQ-CN’ (square) and ‘Cbz-TSAVLQ-CN’ (diamond) at 0–256 μM, followed by measuring the protease activity to determine the inhibitory effect. The IC50 values were determined by fitting the curve to a four-parameter logistics equation.

Crystal structures of enzyme–inhibitor complexes

To better understand the enzyme–inhibitor interaction, we determined the crystal structures of SARS-CoV 3CLpro in complex with the four inhibitors at 1.95–2.5 Å resolution (Table S1, Supplementary data). The 3CLpro formed a dimer in the crystal structure, and the inhibitors occupied the substrate binding site in both protomers. Electron density was observed for the tetrapeptide (Ala–Val–Leu–Gln) at P1 to P4 positions for all inhibitor complexes (Fig. S1, Supplementary material). Overlaying these inhibitor structures demonstrated that the tetrapeptides adopted the same conformation when bound to the 3CLpro. On the other hand, the protective group of ‘Mic-AVLQ-CN’, the P5-Ser and P6-Thr of ‘Cbz-TSAVLQ-CN’ were disordered and not observable in the electron density.

In SARS-CoV, the catalytic dyad consists of Cys145 and His41 [16]. The inhibitor was covalently bonded with the thiol group of Cys145 via the carbon atom of the nitrile warhead (Fig. 2 A). This mechanism is consistent with that observed in other cysteine proteases [17]. The side chains of P1-Gln, P2-Leu and P4-Ala occupied the S1, S2 and S4 subsites, respectively, of SARS-CoV 3CLpro, while the side-chain of P3-Val pointed towards the solvent (Fig. 2A). A number of hydrogen bonds are involved in enzyme–inhibitor interactions. The P1-Gln side chain was hydrogen-bonded with the backbone O atom of Phe140 and the side-chain of His163 and Glu166. The backbone amide groups of P2 to P4 formed hydrogen bonds with Glu166, Gln189 and Thr190, and these hydrogen bonds help the P3 and P4 residues to adopt a β-strand like conformation (Fig. 2B).

Fig. 2

Crystal structure of 3CLpro-inhibitor complex. (A) The 3CLpro was shown in surface representation, while the inhibitor was shown in stick representation. The side-chains of P1-Gln, P2-Leu and P4-Ala occupy the S1, S2, S4 subsites of 3CLpro, while the side-chain of P3-Val points towards the solvent. (B) The hydrogen bonds (dotted lines) between the inhibitor (orange) and the 3CLpro (grey) were shown. (C) The Cbz protective group of the ‘Cbz-AVLQ-CN’ docks into a pocket formed by Glu166, Leu167, and Pro168 of SARS-CoV 3CLpro. Surface was shown for atoms of 3CLpro and P1-P4 residues of the inhibitors. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Noteworthy, the aromatic ring of the Cbz group is docked into a pocket where the ring can form favourable hydrophobic interactions with Pro168 and P3-Val (Fig. 2C). These extra interactions, which are absent in other enzyme–inhibitor complexes, may enhance the binding affinity between 3CLpro and increase the potency of the ‘Cbz-AVLQ-CN’ inhibitor.

Nitrile-based peptidomimetic inhibitor exhibits broad-spectrum inhibition against 3CLpro from different groups of coronaviruses

Next, we tested if the ‘Cbz-AVLQ-CN’ inhibitor has broad-spectrum inhibition against 3CLpro from different groups of coronaviruses. We measured and compared the IC50 values of ‘Cbz-AVLQ-CN’ against 3CLpro from human coronavirus (HCoV) strains 229E (group 1), NL63 (group 1), OC43 (group 2a), HKU1 (group 2a), SARS-CoV (group 2b), and infectious bronchitis virus (IBV) (group 3). Our results showed that ‘Cbz-AVLQ-CN’ can inhibit all 3CLpro tested, with values of IC50 ranging from 1.3 to 4.6 μM (Fig. 3 ). In contrast, the ‘Cbz-AVLQ-CN’ had no observable inhibition effect on caspase-3 at 0.5–256 μM (Fig. S2, Supplementary material), suggesting that the inhibitor is specific for 3CLpro from coronaviruses. As structures of 3CLpro from group 1, 2a, 2b, and 3 are similar, our modelling study suggests that the ‘Cbz-AVLQ-CN’ inhibitor can fit nicely to the active site of 3CLpro from all groups of coronaviruses and inhibit the protease via the same mechanism (Fig. S3, Supplementary material).

Fig. 3

Broad-spectrum inhibition of ‘Cbz-AVLQ-CN’ against 3CLpro from (A) HCoV-NL63, (B) HCoV-299E, (C), HCoV-OC43, (D) HCoV-HKU1, (E) SARS-CoV, and (F) IBV. The proteases were pre-mixed with ‘Cbz-AVLQ-CN’ at 0–32 μM. The IC50 values were determined by fitting the curve to a four-parameter logistics equation.

Conclusion

In this study, we have shown that nitrile-based peptidomimetic inhibitors display broad-spectrum inhibition against 3CLpro from all groups of coronaviruses. Crystal structures of enzyme–inhibitor complexes provide interesting insights in the design of peptidomimetic inhibitors. First, the enzyme–inhibitor binding is mainly determined by the tetrapeptide sequence, AVLQ. This suggestion is consistent with the observation that extending the length of the inhibitor to hexapeptide (‘Cbz-TSAVLQ-CN’) did not improve inhibition (Fig. 1). Second, we found that the Cbz protective group at the N-terminus of the tetrapeptide can result in ∼10 fold increases in potency in the ‘Cbz-AVLQ-CN’ inhibitor than the other inhibitors tested. This observation suggests that modification of the peptidomimetics, e.g. cross-linking the P3-Val with the Cbz groups, may further improve the efficacy of inhibition. Because of the high recombination frequency among the RNA genome of CoVs, development of a broad-spectrum inhibitor is an attractive strategy to combat against infection of all kinds of CoVs [7]. Here, we have shown that the nitrile-based inhibitor ‘Cbz-AVLQ-CN’ can inhibit 3CLpro from all groups of CoVs tested with IC50 values in the μM range. As nitrile-based protease inhibitors (e.g. odanacatib [18], vildagliptin [19]) have been successful in clinical applications [20], our work demonstrated that the inhibitors described in this study may serve as a lead for the developmentof antiviral drugs that target a broad-spectrum of coronaviruses.

Experimental section

Inhibitor synthesis

The reaction scheme for inhibitor synthesis is summarized in Scheme 1. 5-Methylisoxazole-3-carboxylic acid (Mic-OH), N,N′-dicyclohexylcarbodiimide (DCC), N-hydroxysuccinimide (HOSu) were purchased from Sigma–Aldrich. Identities of the final products were verified by 1H nuclear magnetic resonance and mass spectrometry analysis.

Peptide coupling using N,N′-dicyclohexylcarbodiimide (DCC) and N-hydroxysuccinimide (HOSu)

In part 1, to a stirred suspension of N-protected-amino acid-OH or N-protected-peptide-OH (1.0 equiv) and HOSu (1.1 equiv) in anhydrous tetrahydrofuran (THF) (20 mL/mmol), DCC (1.1 equiv) in THF was added to the solution at 0 °C. The reaction mixture was stirred for 30 min at the same temperature and 6 h at 25 °C. N,N′-dicyclohexylurea was then removed by filtration and the filtrate was collected. In part 2, the amino acid was dissolved in THF/H2O (v/v = 1:1), then the pH of the solution was adjusted to 9 with 2 M K2CO3. The filtrate from part 1 was added to the solution at 0 °C. This mixture was stirred for 30 min at 0 °C and overnight at 25 °C. Solvents were then evaporated in vacuo and the pH of the residue was adjusted to acidic with citric acid at 0 °C and extracted with ethyl acetate (EtOAc). The combined extracts were washed with brine, dried (MgSO4), filtered and evaporated in vacuo. The residue was purified by flash chromatography on silica gel to give the desired compound.

Peptide coupling using mixed anhydride method

N-protected-amino acid-OH or N-protected-peptide-OH (1 equiv) was dissolved in THF and cooled to −15 °C. To the stirred solution, N-methylmorpholine (NMM) (1 equiv) and isobutylchloroformate (1 equiv) were added consecutively. After precipitation of N-methylmorpholine hydrochloride, the amine (1 equiv) was added and the mixture was allowed to warm to 25 °C within 30 min. It was stirred for additional 90 min, and the solvent was removed under reduced pressure. The residue was dissolved in EtOAc and the solution was washed with H2O, saturated NaHCO3, 5% citric acid and brine. The solvent was dried (MgSO4) and evaporated. The crude residue was purified by flash column chromatography to give the product.

Deprotection of benzyl carbamates using palladium on charcoal (Pd/C)

To a stirred solution of the benzyl carbamate (1.0 equiv) in MeOH, palladium on charcoal (benzyl carbamate/palladium on charcoal = 1 g/100 mg) was added. The solution was evacuated, placed under H2 (1 atm), and stirred for 1 h. The reaction was monitored by thin layer chromatography (TLC). The suspension was then filtered through Celite and washed with MeOH. The filtrate was concentrated under reduced pressure to afford the desired amine.

Deprotection of Boc, Trt groups and tertiary butyl esters using trifluoroacetic acid (TFA)

To a stirred solution of the starting materials (1.0 equiv) in CH2Cl2, TFA (16.0 equiv) was added. The solution was stirred at 25 °C and monitored by TLC. The CH2Cl2/TFA solvent was removed under reduced pressure. The residue was diluted with toluene and the solvent was removed once again under reduced pressure.

Transformation from primary amide to nitrile using trifluoroacetic anhydride

To a stirred cold solution (0 °C) of the primary amide (1.0 equiv) in dry THF and Et3N (2.2 equiv), trifluoroacetic anhydride (1.1 equiv) was added over 5 min. The solution was allowed to warm to 25 °C. After 30 min, the reaction was quenched with H2O. The THF was removed under reduced pressure, and the product was extracted with EtOAc. The combined extracts were washed with saturated NaHCO3, 5% citric acid and brine, dried (MgSO4), filtered and evaporated in vacuo. The residue was purified by flash chromatography on silica gel to give the desired compound.

Identification of products by NMR and mass spectrometry

Mic-AVLQ-CN

After this compound was synthesized by following the procedure as described in Section 4.1.2, the product was purified by flash chromatography on silica gel to afford the target compound (yield = 64%). R f: 0.7 (EtOAc/MeOH = 4/1).

1H NMR (300 MHz, d 6-DMSO) δ (ppm): 0.50–1.08 (12H, m, CH(CH 3)2), 1.30 (3H, d, J = 6.8 Hz, CH3), 1.46–1.49 (2H, m, CH2), 1.52–1.67 (1H, m, CH), 1.74–1.86 (1H, m, CH), 1.88–2.15 (2H, m, CH2), 2.23–2.42 (2H, m, CH 2CONH2), 2.46 (3H, s, CH 3C = CH), 3.96–4.40 (3H, m, NHCH), 4.42–4.65 (1H, m, NHCH), 6.56 (1H, s, CCH), 7.03 (1H, s, NH), 7.26 (1H, s, NH) 7.94 (2H, d, J = 8.4 Hz, NH), 8.08 (1H, d, J = 7.2 Hz, NH), 8.62 (1H, d, J = 7.2 Hz, NH); ESI-HRMS (m/z): calculated for C24H37N7O6 + Na+: 542.2703, found: 542.2706.

Boc-AVLQ-CN

After this compound was synthesized by following the procedure as described in Section 4.1.2, the product was purified by flash chromatography on silica gel to afford the target compound (yield = 67%). R f: 0.1 (EtOAc).

1H NMR (300 MHz, d 6-DMSO) δ (ppm): 0.68–0.96 (12H, m, CH(CH 3)2), 1.15 (3H, d, J = 6.9 Hz, CH3), 1.37 (9H, s, C(CH3)3), 1.41–1.52 (2H, m, CH2), 1.52–1.66 (1H, m, CH), 1.67–1.87 (1H, m, CH), 1.87–2.09 (2H, m, CH2), 2.33–2.49 (2H, m, CH 2CO), 3.86–4.04 (1H, m, NHCH), 4.06–4.33 (3H, m, NHCH), 7.09 (1H, d, J = 7.5 Hz, NH), 7.17 (1H, s, NH), 7.32 (1H, s, NH), 7.60 (1H, d, J = 8.7 Hz, NH), 7.93 (1H, d, J = 8.1 Hz, NH). 8.06 (1H, d, J = 7.8 Hz, NH). ESI-HRMS (m/z): calculated for C24H42N6O6 + Na+: 533.3058, found: 533.2850.

Cbz-AVLQ-CN

After this compound was synthesized by following the procedure as described in Section 4.1.2, the product was purified by flash chromatography on silica gel to afford the target compound (yield = 73%). R f: 0.65 (CH2Cl2/MeOH = 1/4).

1H NMR (300 MHz, d 6-DMSO) δ (ppm): 0.62–1.00 (12H, m, CH(CH 3)2), 1.18 (3H, d, J = 6.9 Hz, CHCH 3), 1.20–1.30 (1H, m, CH2), 1.32–1.52 (2H, m, CH2), 1.52–1.66 (1H, m, CH), 1.66–1.86 (1H, m, CH(CH3)2), 1.86–2.14 (2H, m, CH 2CONH2), 2.30–2.46 (1H, m, CH(CH 3)2), 4.00–4.18 (2H, m, NHCH), 4.18–4.36 (2H, m, NHCH), 5.01 (2H, s, ArCH 2), 7.18 (1H, s, NH), 7.20–7.44 (6H, ArH, NH) 7.53 (1H, d, J = 7.5 Hz, NH), 7.77 (1H, d, J = 8.7 Hz, NH), 7.94 (1H, d, J = 8.1 Hz, NH). 8.05 (1H, d, J = 7.5 Hz, NH). ESI-HRMS (m/z): calculated for C27H40N6O6 + Na+: 567.2907, found: 567.2902.

Cbz-TSAVLQ-CN

After this compound was synthesized by following the procedure as described in Section 4.1.2., the product was purified by flash chromatography on silica gel to afford the target compound (yield = 48%). R f: 0.75 (MeOH/EtOAc = 3/1).

1H NMR (300 MHz, d 6-DMSO) δ (ppm): 0.68–0.96 (12H, m, CH(CH3)2), 1.06 (3H, d, J = 6.3 Hz, CHCH 3), 1.21–1.32 (4H, m, CH3, CH), 1.36–1.53 (2H, m, CH2), 1.53–1.68 (1H, m, CHH), 1.68–1.88 (1H, m, CHH), 1.88–2.12 (2H, m, CH 2CONH2), 2.33–2.46 (1H, m, CH), 3.44–3.70 (2H, m, CH 2OH), 3.84–4.00 (1H, m, CH3CHOH), 4.00–4.14 (2H, m, NHCH), 4.14–4.40 (4H, m, NHCH), 4.92–4.94 (1H, m, OH), 4.96–5.13 (3H, m, ArCH 2, OH), 7.02 (1H, d, J = 8.1 Hz, NH), 7.16 (1H, s, NH), 7.26 (1H, s, NH), 7.28–7.44 (5H, m, ArH), 7.73 (1H, d, J = 8.1 Hz, NH), 7.85 (1H, d, J = 8.1 Hz, NH), 7.89–8.02 (2H, m, NH), 8.13 (1H, d, J = 6.9 Hz, NH). ESI-HRMS (m/z): calculated for C34H52N8O10 + Na+: 755.3698, found: 755.3669.

Protease assay and determination of IC50 values

The detailed procedures for the production of 3CLpro, fluorescent substrate, and protease assay for 3CLpro were described previously [14], [15]. 3CLpro at 0.5–1 μM were pre-mixed with 0.5–256 μM of inhibitors for 5 min, followed by addition of 35 μM substrate. The decay of fluorescence at 530 nm was fitted to a single-exponential decay to yield the observed rate constant (k obs) and k obs/[3CLpro]. The values of k obs/[3CLpro] at different concentration of inhibitors were normalized by those without inhibitors to determine the relative protease activities, which were fitted to a four-parameter logistics equation to obtain the IC50 values.

Structure determination of protease-inhibitor complexes

SARS-CoV 3CLpro was crystallized at 16 °C using the hanging-drop-vapour-diffusion method by mixing 5 mg/mL 3CLpro in 1:1 ratio with 50 mM 2-(N-morpholino)ethanesulfonic acid, pH 5.5, 8.5% (w/v) polyethylene glycol 6000, 10% (v/v) glycerol, 3% (v/v) DMSO, 1 mM ethylenediaminetetraacetic acid and 1 mM dithiothreitol. Single crystals were transferred to 5 μL of the mother liquor with 600 μM inhibitors. After incubation at 16 °C overnight, the crystals were cryoprotected by 20% (v/v) glycerol, and diffraction data were collected at 110 K in an in-house Rigaku FRE+ X-ray source. The phase was solved by molecular replacement, the models were built interactively by the program COOT [21] and refined by PHENIX [22].