INTRODUCTION
Coronaviruses (CoVs) are pleomorphic, enveloped positive-strand RNA viruses with unusually large genomes, comprising approximately 30 kilobases. Although they are endemic in the human population and account for 10–30 % of common colds, they were not considered a threat to human health since they mainly cause respiratory infections with mild symptoms (1). However, this view changed with the realisation that CoVs are maintained in an animal reservoir and that their transmission to humans is possible via intermediate hosts (2). At the end of 2019, Wuhan, China, became a hotspot for the uncontrollable spread of the SARS-CoV-2 virus (3). SARS-CoV-2 reached all parts of the world and caused COVID-19, the most severe pandemic of modern times (4). Both viral and host peptidases play important roles in key steps of coronaviral infection and replication processes (5). Peptidases encoded in the viral genome are essential for processing replicase polyproteins and evading the host immune response, while host peptidases are involved in various steps of viral uptake into the host cell. The viral genome encodes one or two cysteine peptidases, the papain-like peptidase (PLP) and chymotrypsin-like cysteine 3C-like peptidase – 3CLpro, both of which are pivotal for transcription of the viral genome and its replication (6). 3CLpro consists of three domains: domains 1 and 2 form the chymotrypsin-like fold, whereas domain 3 is required for dimer formation and affects catalytic activity through dynamically controlled allostery (7). Among viral peptidases, 3CLpro is an attractive target for the development of antiviral drugs against SARS-CoV-2 and other CoVs, because of its essential role in post-translational polyprotein processing. Numerous inhibitors of 3CLpro peptidase have been developed (8–11), and nirmatrelvir was the first-in-class inhibitor approved by regulatory agencies in combination with ritonavir under the trade name PaxlovidTM for the treatment of mild-to-moderate COVID-19 in adults (12). Nirmatrelvir, an orally available covalent inhibitor of 3CLpro, forms a reversible thioimidate adduct with the catalytic Cys145 (Fig. 1a) (13). Similarly, peptidomimetic and non-peptidic inhibitors of 3CLpro have been developed bearing a variety of warheads: aldehydes, α-acyloxy-, α-heteroaryl- and α-hydroxy-substituted ketones (14), α-haloacetamides, α-ketoamides, α,β-unsaturated ketones, activated esters and others (10, 15, 16). Acyloxymethylketones, which have been studied as cathepsin B inhibitors (17, 18) and as activity-based probes for cysteine protease profiling (19), are also extensively explored as 3CLpro inhibitors (20, 21) (Fig. 1b,c).
Fig. 1. 3CLpro inhibitors: a) nirmatrelvir and the resolved crystal structure (PDB code 7RFS) (13); b) hydroxymethyl ketone PF-00835231 and α-acetoxymethyl ketone A (20); c) benzothiazolyl ketone B (14). Covalent warheads are highlighted in yellow.
A rational and systematic computational approach reported by the Wolber group led to the identification of the covalently binding fragment F1, which inhibits the enteroviral cysteine 3C protease (Fig. 2a). Scaffold hopping subsequently yielded fragment C5, an α-phenylthiomethyl ketone (Fig. 2a), which covalently binds to Cys147 of the 3C protease, as confirmed by mass spectrometry (22). Recently, thiazolyl ketones have also been reported as inhibitors of cytosolic phospholipase A2α (23). A structurally related α-heteroarylthiomethylketo moiety is also present in the selective cathepsin X inhibitor Z9 developed by Pečar Fonović et al. (Fig. 2b) (24). In contrast to the α-phenylthiomethyl ketones and the analogues developed by the Wolber group, Z9 is a reversible inhibitor, as demonstrated by enzyme kinetics and reversibility assay (24). Further optimisation and structure-activity relationship (SAR) studies explored the relevant chemical space; however, the inhibitory potencies against cathepsin X and the biological activities in cellular models of the analogues remained comparable to those of Z9 (25). In addition to the inhibition of cathepsin X, α-heteroarylthiomethyl ketones are also described in the literature as covalent and noncovalent inhibitors for various biological applications. Fragment screening by native mass spectrometry identified 3-substituted 1,2,4-triazole (A) as a noncovalent, zinc-binding chemotype that inhibits carbonic anhydrase II (Fig. 2c), a validated target in the management of glaucoma and congestive heart failure (26). In addition, these compounds inhibit urease (A2–5, Fig. 2c) (27), fungal H+-ATPase (28) and are disclosed as antiviral and antibacterial agents (29).
Fig. 2. (Hetero)arylthiomethyl ketones: a) ketone fragment F1 and phenylthiomethyl ketone C5 (22); b) cathepsin X inhibitor Z9 (24); c) heteroarylthiomethy ketones are described as inhibitors of carbonic anhydrase II (A1) (26) and urease (A2–5) (27).
An in-house library of α-heteroarylthiomethyl ketones available at our Faculty, structurally related to previously reported α-hydroxymethyl ketone- and α-acetoxymethyl ketone-based 3CLpro inhibitors, was therefore screened against the recombinant SARS-CoV-2 main protease 3CLpro. The identified hit compounds inhibited 3CLpro in the micromolar range and are tentatively proposed to act as covalent inhibitors.
EXPERIMENTAL
Biochemical evaluation
Cloning, expression and purification of recombinant 3CLpro. – A codon optimised synthetic gene encoding the SARS-CoV-2 3CLpro protease (Integrated DNA Technologies, USA) was cloned into the pET-28c(+) plasmid, and used to transform E. coli NiCo21(DE3) (New England Biolabs, USA). Transformed cultures were cultivated in Lysogeny broth (LB) medium supplemented with 50 μg mL–1 kanamycin at 37 °C and 250 rpm until reaching an optical density at 600 nm (OD600) of approximately 1.8. Cultures were subsequently cooled on ice (0–4 °C) for 10 min, and 3CLpro protease expression was induced by the addition of 200 µmol L–1 isopropyl β-D-1-thiogalactopyranoside (IPTG). Expression proceeded for 24 h at 16 °C and 250 rpm. Then, cells were harvested by centrifugation (2 × 10 min, 3000 × g, 4 °C), and the resulting pellet resuspended in buffer A (20 mmol L–1 Tris-HCl, pH 7.5, 0.05 mmol L–1 EDTA, 2.5 mmol L–1 DTT, 10 % glycerol). Cell lysis was performed on ice via sonication, and the lysate was clarified by centrifugation (2 × 30 min, 16000 × g, 4 °C). The supernatant was filtered through a 100-kDa molecular weight cut-off (MWCO) centrifugal unit (Amicon Ultra-15; Merck, Germany). Ammonium sulfate was gradually added to the filtrate to a final concentration of 500 mmol L–1, and the solution was loaded onto a 1 mL HiTrap Phenyl HP column (Cytiva, USA) pre-equilibrated with buffer B (50 mmol L–1 Tris-HCl, pH 7.5, 0.5 mol L–1 (NH4)2SO4, 0.05 mmol L–1 EDTA, 2.5 mmol L–1 DTT, 10 % glycerol). After washing the column with 20 volumes of buffer B, the bound 3CLpro was eluted using a linear gradient into buffer A. Eluted fractions were concentrated using a 30-kDa MWCO centrifugal filter unit (Amicon Ultra-4; Merck), frozen in liquid N2, and stored at −80 °C. Protein concentration was determined by UV-absorbance at 280 nm, using the extinction coefficient of 34380 M−1 cm−1. Purity of 3CLpro was assessed by SDS-PAGE.
Enzyme activity assay. – 3CLpro protease activity was measured by kinetic assay using FRET fluorogenic substrates Dabcyl-KTSAVLQSGFRKME-Edans (Dabcyl-Lys-SARS-CoV2 Replicase pp1ab(3235-3246)-Glu-EDANS, CPC Scientific, USA) and Hilyte™Fluor488-ESATLQSGLRKAKQXL®520 (Hilyte™Fluor488, Anaspec, USA). Measurements were performed in 50 mmol L–1 Tris-HCl, pH 7.3, 1 mmol L–1 EDTA, 0.05 % Triton X-114. For the screening, compounds were pre-incubated at a concentration of 500 µmol L–1 with 3CLpro (final concentration, 50 nmol L–1) for 30 min at 30 °C. The reaction was started by adding Dabcyl-KTSAVLQSGFRKME-Edans (final concentration, 20 µmol L–1), and the increase in fluorescence intensity was measured using a microplate reader Synergy H4 (BioTek Instruments Inc., USA) at λex = 360 (bandwidth, 17 nm) and λem = 528 (bandwidth, 17 nm). Final concentration of DMSO was always 10 % (V/V). In control experiments, the compound was replaced by DMSO. For the blank determination (b), the enzyme was replaced with assay buffer. Initial velocities (v) were calculated from the linear trends obtained, with each measurement performed in duplicate. Inhibitory potencies were expressed as residual activities – RAs = (vi – b)/(v0 – b), where vi represents the velocity of the enzyme reaction in the presence of the test compound, and v0 the control velocity in the presence of DMSO. To confirm the activity of the compounds and to exclude assay spectral interference at 360 nm, the active compounds from the screening phase (RA at 500 µmol L–1 < 50 %) were evaluated using the above described procedure by replacing Dabcyl-KTSAVLQSGFRKME-Edans substrate with Hilyte™Fluor488-ESATLQSGLRKAKQXL®520 substrate (final concentration, 2 µmol L–1). For the active compound (RA at 500 µmol L–1 < 50 %, Hilyte™Fluor488-ESATLQSGLRKAKQXL®520 substrate), IC50 values using both substrates were determined by measuring RAs at seven to twelve concentrations of the compound. The IC50 values were calculated by fitting RAs at different concentrations to a 4-parameter logistic function [Y = Bottom + (Top – Bottom)/(1 + 10^((LogIC50 – X)×HillSlope), where Y represent RAs and X the log10 of compound concentration] using GraphPad Prism 10.4 (GraphPad Software Inc., USA). For progress curve analysis, the assays were performed by preincubating a serial dilution of compounds in the presence of the substrate Dabcyl-KTSAVLQSGFRKME-Edans (final concentration, 15 µmol L–1) for 15 min at 30 °C prior to the addition of 3CLpro (final concentration, 10 nmol L–1). The increase in fluorescence intensity was followed as described above. To determine kobs values, the progress curves obtained were fitted to the equation Y = v + v0 × [1– exp(–kobs × X)] / kobs. The first-order rate constants kobs for GC376 (control inhibitor) were then fitted to kobs = k + (kinact × [Inhibitor])/(KI + [Inhibitor]). Since compounds 31 and 33 are slow and inefficient inhibitors of 3CLpro, the kobs was fitted to simple linear regression, where the slope of the line equals kinact/KI (30). All fittings were performed in GraphPad Prism 10.4 (GraphPad Software Inc., USA).
Thiol reactivity assay – DTNB assay. – The assay was performed according to a previously reported procedure (31). Briefly, experiments were performed in duplicate in 96-well microplates in assay buffer (20 mmol L–1 sodium phosphate, 150 mmol L–1 NaCl, pH 7.4). Reagent solutions were prepared freshly prior to the experiments. 2-Chloro-N-(3-chlorophenyl)acetamide was used as a control compound.
Thiol reactivity assay – TNB2– assay. – The assay was performed according to a previously reported procedure (31). Briefly, 100 μmol L–1 of compound was incubated in a mixture of 50 μmol L–1 TNB2− in assay buffer containing 5 % final DMSO concentration at 37 °C. Absorbance at 412 nm was measured at 5-minute intervals for 14–21 h using a microplate reader Synergy H4 (BioTek Instruments, Inc., USA) to monitor TNB2− depletion. To determine the baseline drift due to the oxidation of TNB2− to DTNB, a blank experiment was performed, where 100 % DMSO replaced the compound. Baseline drift due to TNB2− oxidation and compound background absorbances were subtracted from each measurement.
RESULTS AND DISCUSSION
The in-house library of fully characterised α-heteroarylthiomethyl ketones, synthetic intermediates and close analogues (compounds 1–38, Table I) from the cathepsin X campaign (25) was screened for inhibition of the recombinant SARS-CoV-2 main protease 3Clpro, which was cloned, expressed and purified as previously described (31). The initial screening was conducted at a compound concentration of 500 µmol L–1 with a 30-minute preincubation and using the fluorogenic substrate Dabcyl-KTSAVLQSGFRKME-EDANS (Dabcyl-EDANS). Compounds with residual activities (RAs) below 50 % were considered hits (Table I).
Given the relatively low excitation wavelength of the Dabcyl-EDANS substrate, i.e. at 360 nm, potential interference with assay readout due to the inner filter effect was considered (32). To address this, the absorbance spectra of active compounds were recorded at 500 µmol L–1. Spectral interferences (i.e. absorbance > 0.1 AU at 360 nm) were observed for compounds 22, 32 and 37. All active compounds were subsequently retested under identical conditions using the HiLyteTMFluor488-QXL520 substrate, which has a higher excitation wavelength (~490 nm) and is less prone to spectral interference, even with lightly colored yellow compounds. For all active compounds, IC50 values were determined using both substrates.
Of the 38 compounds tested, eleven inhibited 3CLpro with IC50 values below 500 µmol L–1 (Table I). Among the 1-(2,3-dihydrobenzo[b][1,4]dioxin-6-yl)ethan-1-ones 1–21, only the (4H-1,2,4-triazol-3-yl)thio analogues bearing 4-isopropyl and 4-ethyl substitutions inhibited 3CLpro. In contrast, the analogues with smaller, unsubstituted triazoles, imidazoles, or 4-aryl-substituted 4H-1,2,4-triazololes were inactive. Substitution at position 3 of the 4H-1,2,4-triazole, such as cyclohexyl (18) or phenyl (19), was also not tolerated, indicating a narrow structure-activity window for modifications at the heteroarylthio moiety. Notably, reduction of the ketone group in compound 1 to the corresponding racemic secondary alcohol 2 abolished inhibitory activity, underscoring the essential role of the ketone functionality for inhibition of 3CLpro. In contrast, neither the α-hydroxymethyl ketone 21 nor the thiol 22 inhibited the enzyme, suggesting that the presence of either the thiol or ketone alone is insufficient for 3CLpro inhibition, although such functionalities have been described in the literature as effective covalent inhibitors (33). Replacement of the 2,3-dihydrobenzo[b][1,4]dioxine moiety by smaller fragments such as substituted phenyl groups (compounds 23–38) was tolerated when the substituents were smaller (e.g. methyl, methoxy, hydroxy, nitro) and on para or meta position relative to the ketone. Among the compounds tested, p-tolyl and phenyl derivatives 31 and 33, respectively, were the most potent inhibitors, with IC50 values of 95.4 ± 3.1 and 95.0 ± 6.9 µmol L–1, respectively.
Table I. Structures of compounds and inhibition of 3CLpro expressed as residual activities (RAs) and IC50 values
a RAs are means of a single experiment performed in duplicate; the standard deviation for RAs was < 10 %. b IC50s are means ± standard error of the mean (SEM) for two independent experiments, each performed in duplicate. Aldehyde bisulfite GC376 and α-ketoamide boceprevir (34) were used as positive controls (IC50(Dabcyl-EDANS substrate) = 0.05081 ± 0.0041 and 3.977 ± 0.1444 µmol L–1, respectively). c n.t. – not tested. d Assay spectral interference at 360 nm.
Based on the literature reports, the expected mechanism of action for the compounds investigated involves either covalent modification via the ketone moiety or a nucleophilic substitution followed by elimination of the nucleofuge, the heteroarylthiole (20, 22, 35). To evaluate the intrinsic reactivity of the compounds in a non-proteinaceous environment in vitro, a thiol-containing colourimetric probe, 5-mercapto-2-nitrobenzoic acid (TNB2–), was employed as a cysteine surrogate. TNB2– was generated in situ by reduction of 5,5'-dithio-bis(2-nitrobenzoic acid) (DTNB) with tris(2-carboxyethyl)phosphine (TCEP). However, since TCEP itself is a phosphine nucleophile and could potentially react with the electrophilic compounds under investigation, an alternative assay was conducted using commercially available TNB2– to avoid interference (36). Under the experimental conditions applied, none of the compounds, with the exception of fragments 21 and 22, exhibited reactivity towards TNB2–. Nonetheless, it should be noted that cysteine reactivity in a protein environment is influenced by local electronic effects in the active site of the enzyme (37). Consequently, the results obtained using thiol surrogate compounds should be interpreted with caution.
Although IC50 values are commonly used in medicinal chemistry to compare the inhibitory potency of compounds under standardised conditions, a detailed kinetic evaluation is more appropriate for covalent inhibitors (38). The progress curves for the hydrolysis of Dabcyl-KTSAVLQSGFRKME-Edans substrate by 3CLpro indicated that covalent inactivation by inhibitors 31 and 33 is rather slow and inefficient, particularly when compared to the reference inhibitor CG376 (Fig. 3) (39). Nevertheless, further optimisation and comprehensive characterisation are required before the proposed mechanism of action can be conclusively confirmed. Covalent mode should be confirmed by mass spectrometry to verify modification of the catalytic Cys145. Secondly, establishing the pre-reaction binding pose of the intact inhibitor through molecular modelling would further aid in understanding key interactions in the 3CLpro’s active site. These insights could then guide the rational optimisation of α-heteroarylthiomethyl ketones to enhance their potency and selectivity toward the targets. Key findings regarding structure-activity relationship and plausible further steps to improve inhibitory activities based on the data presented herein are presented in Fig. 4.
Fig. 3. Progress curve analysis. Left: progress curves of 3CLpro reaction in the absence or presence of indicated concentrations (in µmol L–1) of inhibitors; Right: Secondary plot of kobs as a function of inhibitor concentration.
Fig. 4. Key structural-activity relationship findings for 3CLpro inhibition and proposed directions for further optimisation.
CONCLUSIONS
Small, academic in-house compound libraries, often compiled from previous medicinal chemistry projects, represent a valuable resource for the identification of novel hits against disease-relevant targets. Here, we present an example of a cathepsin X-focused compound library, which was screened to identify structurally novel inhibitors of the SARS-CoV-2 main protease 3CLpro. Preliminary in vitro evaluation, together with supporting literature data, suggests a covalent mode of action. This hypothesis warrants further experimental confirmation by native and before proceeding with SARs optimisation and broader biological evaluation.
Acknowledgements and funding. – This work was funded by the Slovenian Research and Innovation Agency (research core funding No. P1-0208).
Conflict of interest. – The authors declare no conflict of interest.
Authors contributions. – Conceptualisation, D.K. and S.G.; biochemical experiments, D.K., M.P., and K.B.; writing, original draft preparation, D.K.; writing, review and editing, D.K., M.P, K.B., and S.G.; supervision, S.G. and D.K; funding, S.G. All authors have read and agreed to the published version of the manuscript.