Epoxide based inhibitors of the hepatitis C virus non-structural 2 autoprotease

Highlights • Epoxide-containing small molecules inhibit hepatitis C virus NS2 autoprotease in vitro.• These molecules also show specific inhibition of NS2–NS3 autocleavage dependent HCV genome replication in a cell based assay.• NS2 protease is a valid target for further drug development.


Introduction
Hepatitis C virus (HCV) is a global health burden infecting 3% of the world's population. As a positive sense RNA virus within the Flaviviridae family HCV relies on proteolytic processing of a single polyprotein to generate mature proteins. The structural proteins Core and E1-E2, as well as p7, are processed by host proteases, while the non-structural (NS) proteins responsible for genome replication undergo maturation by virally encoded proteases. Autoproteolysis occurs at the NS2-NS3 boundary via a cysteine protease activity encoded principally within NS2 but enhanced by the presence of the NS3 N-terminus (Schregel et al., 2009). NS3 with its cofactor NS4A (NS3-4A) mediates subsequent downstream cleavages to generate NS4B, NS5A and NS5B (Scheel and Rice, 2013).
Inhibitors of the NS3-4A protease that disrupt polyprotein processing are now approved for the treatment of HCV infection. However, NS2 protease activity remains an unexplored target. NS2 plays no direct roles in genome replication, as demonstrated by the ability of a subgenomic replicon (SGR) to replicate in the absence of NS2 (Lohmann et al., 1999). However, the unprocessed NS2-NS3 precursor has reduced NS3 protease activity, potentially by reducing NS3 proteolysis kinetics or through reduced stability of NS3 (Welbourn et al., 2005). Hence where NS3 is derived from a NS2-NS3 precursor, as in the context of infectious virus, the activity of the NS2 autoprotease is essential (Jones et al., 2007;Kolykhalov et al., 2000). Mutational analysis and structural studies of the post-cleavage NS2 protease domain propose that NS2 acts as a cysteine protease, though the catalytic triad appears to adopt the geometry of a serine protease (Lorenz et al., 2006). Due to the essential nature of the NS2 autoprotease it has been proposed as an attractive target for antivirals that to date has not been explored (Rice, 2011).
A common route to develop a protease inhibitor is to incorporate an electrophilic 'warhead' so as to produce a mechanismbased inhibitor (Powers et al., 2002). Such reactive warheads form an irreversible covalent interaction with the active site residues, but often lack selectivity. In contrast, an epoxide warhead forms a covalent interaction with the nucleophilic catalytic residue only when the epoxide is held non-covalently in the optimal orientation. As such the rate limiting step in protease inhibition by epoxides is the formation of a non-covalent binding pose so as to optimally orientate the epoxide for nucleophilic attack (Bihovsky et al., 1993). This transient interaction is usually mediated by a conjugated substrate peptide derivative and can be tailored to the system, allowing epoxide-based protease inhibitors a greater degree of selectivity (Powers et al., 2002). However, unlike the HCV NS3-4A protease, which is inhibited by peptides corresponding to the N-terminus of the cleavage site (Llinas-Brunet et al., 1998), the NS2 autoprotease shows little or no sensitivity to substrate or proteolysis product peptides in vitro, nor to substrate peptides conjugated to mechanism-based protease inhibitors (Pallaoro et al., 2001;Thibeault et al., 2001). In addition, the cysteine protease inhibitor E64, an epoxysuccinyl peptide, has proven inactive against the NS2 autoprotease (Pieroni et al., 1997;Thibeault et al., 2001). We therefore sought to explore the activity of epoxide warheads conjugated to non-peptide derivatives, in particular a range of reported cis-bisamido epoxides conjugated to aromatic 'backbones' which show activity against the cysteine containing active site of factor XIIIa (Avery et al., manuscript submitted). Derivatives of this scaffold showed activity against the NS2 autoprotease in vitro. Importantly, the epoxide alone was insufficient for activity and variations to the aromatic backbone altered inhibitory activity, indicating the requirement for a transient interaction and suggesting epoxide based inhibitors can be tailored to the NS2 active site. Epoxide based inhibitors of the NS2 autoprotease exerted an effect in cell based assays in which HCV genome replication was NS2-dependent. To the best of our knowledge these findings represent the first example of direct pharmacological inhibition of the NS2 autoprotease to block HCV genome replication and exert an antiviral effect.

Compounds
Synthesis and analysis of compounds 1-10 has been reported elsewhere (Avery et al., manuscript submitted). Tosyl phenylalanine chloromethyl ketone (TPCK) was from VWR International. E64 and Telaprevir were from Sigma Aldrich and MedChem Express respectively. All compounds were confirmed by mass spectrometry analysis using a VG Autospec mass spectrometer with electron spray ionisation (ES) at 70 eV.

SGR compound treatments
Stable cell lines harbouring SGR (Supplementary material) were maintained in Dulbecco's modified Eagle's medium (DMEM; Sigma) supplemented with 10% foetal bovine serum (FBS), 100 IU penicillin ml À1 , 100 lg streptomycin ml À1 , 1% non-essential amino acids and 300 lg/ml G418 (Sigma) in a humidified incubator at 37°C in 5% CO 2 . Cells were seeded at a density of 2 Â 10 4 cells/well in a 96 well plate. After 16 h media was removed and replaced with media supplemented with the indicated concentrations of compounds with standard DMSO at 0.25%. Plates were incubated in the absence of G418 for a further 48 h.

Firefly luciferase endpoint assay
Luciferase activity was quantified by addition of 30 ll Passive Lysis Buffer (PLB: Promega) and addition of 40 ll LarI reagent (Promega) using a BMG Labtech plate reader with light emission recorded over 6 s. Data was normalised to DMSO control. Statistical significance was determined using the Student's t test.

Cell viability endpoint assay
Cellular metabolism was quantified by 2 h incubation in 1 mM Thiazolyl Blue Tetrazolium Bromide (Sigma Aldrich) before crystals were suspended in 100 ll DMSO and absorbance at 570 nm measured using an infinite F50 platereader (Tecan). Data was normalised to DMSO control. CC 50 was calculated using Prism 6 (GraphPad).

HCVcc compound treatments
Transcripts (5 lg) of a Jc1 derivative expressing Nanoluciferase (JC1-NLuc) (Amako et al., 2015) were electroporated into Huh7 cells (see Supplementary). Compound was added as in Section 2.3. Cell viability was performed as in Section 2.5 following 4% paraformaldehyde fixation of cells. NanoLuc was measured using a BMG Labtech plate reader following addition of 50 ll PLB and addition of equal volume of NanoGlo Luciferase Assay Substrate (Promega).

The NS2 autoprotease is inhibited by halomethyl ketones but not the epoxide-based inhibitor E64
To assess the ability of a small molecule to inhibit NS2 protease activity, an in vitro auto-processing assay was employed. A NS2-NS3 precursor protein (NS2-3) comprising the catalytic C-terminal domain of NS2 and the N-terminal protease domain of NS3 (JFH1 polyprotein residues 906-1209, J4 residues 904-1206) flanked by an N-terminal His tag and C-terminal FLAG tag was bacterially expressed and purified from inclusion bodies under denaturing conditions by immobilised metal ion affinity chromatography (IMAC). Upon dilution into Refolding buffer, NS2-3 forms significant secondary structure (Foster et al., 2010) allowing the autoprotease to become active. This can be monitored by western blot analysis of NS2-3 refolding reactions with an anti-FLAG antibody to reveal 35 kDa precursor NS2-3-FLAG and 20 kDa NS3-FLAG, one of the proteolysis products. Quantification of the proteolysis product was used as a relative measure of NS2 autoprotease activity.

Aromatic bis-amido based epoxides as inhibitors of the NS2 autoprotease
To test the ability of non-peptide derived epoxides to inhibit the NS2 autoprotease, a range of previously reported compounds containing an epoxide conjugated to a functionalised benzene ring (Avery et al., manuscript submitted) were tested in the in vitro auto-processing assay. Structures for these compounds are shown in Table 1.
Compound 1 was identified as capable of blocking in vitro NS2-NS3 processing in a dose responsive manner for both JFH1 and J4 isolates, with J4 showing enhanced sensitivity (Fig. 2). Interestingly, a derivative lacking only the benzene 3-chloro group (2) appeared inactive up to 300 lM (Fig. 2B). This finding indicated that elements other than the epoxide alone contributed to the inhibitory effect observed.
To further explore the contributions of the benzene substituents to activity, a range of compounds containing the same epoxide as 1 with alternate R 1 groups (Table 1) were tested against the JFH1 NS2-3 auto-processing assay. These molecules are all active to varying degrees against factor XIIIa (Avery et al., manuscript submitted), and are derived from the anti-factor XIIIa activity of the epoxide-based natural product cerulenin, which contains an extended hydrocarbon chain. Cerulenin showed no activity against the NS2 autoprotease (Table 1). Activities of these compounds (3-9) are shown in Table 1.
A number of chemical modifications reduced activity of 1 to below the level of detection -these included removal of the 3chloro substituent from the benzene (2), introduction of a 4-bromo substituent without (3) or with the 3-chloro group (4) and presence of 3,4-dimethyl groups (5). Compounds 7 and 8, with a 4amino-phenyloxy or 4-nitro-phenyloxy substituent respectively, showed some activity but to a lesser extent to that shown by 1, though a 3-phenyloxy substituent did enhance activity (6). The most potent inhibitor was a 4-benzyloxy, 3-chloro substituted derivative (9) Table 1 Activity of a range of epoxide based inhibitors of NS2-3 autoproteolysis. EC 50 calculated from 10-point dose response curve against JFH1 NS2-3 refolding reaction ( Supplementary Fig. 1). CC 50 calculated from 11-point dose response curve against Huh7 cells (Supplementary Fig. 2 1-9 demonstrates the ability of the compound backbone to modulate activity. In agreement with this, the epoxide alone, where the functionalised amido-benzene backbone is replaced by a primary alcohol (10), was inactive (Table 1, Supplementary Fig. 1).

Activity of epoxide based inhibitors in a cell based assay
In other systems, the requirement of epoxide-based inhibitors to form a transient interaction with the protease before becoming covalently linked allows them greater selectivity and makes them more amenable to in vivo applications (Powers et al., 2002). To explore the potential for bis-amido based epoxide warheads to exert an inhibitory effect within a HCV model system, compounds 1-10 and cerulenin were tested for cytotoxic effects by calculating CC 50 from 72 h incubation in a hepatoma cell line capable of supporting HCV replication ( Supplementary Fig. 2). Of the compounds showing significant activity in vitro only 1 and 7 exerted no significant cytotoxicity up to 100 lM (Table 1). The activity of 1 in cell based assays was further explored due to the availability of 2 which, as an inactive derivative without significant cytotoxic effects, allows a comparative tool.
In order to test for specific activity against NS2-dependent HCV replication, stable cell lines were selected that harbour a HCV SGR comprising either NS2-5B or NS3-5B of the JFH1 or Con1 isolates (Fig. 3A). These replicons contain a fused firefly luciferase neomycin phosphotransferase-reporter enabling both stable selection and luciferase production as a measure of HCV replication (Wyles et al., 2009). All replicon cell lines displayed reported sensitivity to cyclosporine A (data not shown). Replication kinetics as measured by luciferase activity were comparable in the absence or presence of NS2 for both genotypes (Fig. 3B). Western blot confirmed the presence of mature NS2 (JFH1 strain only due to genotype-specific antibody reactivity) while sequence analysis of SGR RNA extracted from the stable cell lines did not reveal any additional culture adaptive mutations.
To assess activity against HCV genome replication, SGR-JFH1-NS2-5B was treated with 1 (at in vitro EC 50 90 lM) and the same concentration of the inactive derivative 2. Quantification of luciferase activity revealed reduction to 48.1 ± 10.7% of vector control upon treatment with 1 (Fig. 4A), whereas 2 had no effect. Cell viability was unaffected by 1, indicative that the decreased luciferase activity represented a specific inhibitory effect against genome replication. The same treatment was performed in parallel against SGR-JFH1-NS3-5B, which lacks the NS2 coding sequence and can therefore support HCV replication independent of NS2 autoprotease activity. Treatment of this replicon cell line with 1 (90 lM) produced a slight decrease in luciferase activity which correlated with a decrease in cellular metabolism and was not statistically significant, indicating that 1 exerts no inhibitory activity against NS2-independent HCV replication (Fig. 4A). To confirm there are no significant differences between the response to inhibitor treatment across the two cell lines, both were treated with an intermediate concentration of the NS3-4A protease inhibitor Telaprevir (Kwong et al., 2011). For both SGR-JFH1-NS2-5B and SGR-JFH1-NS3-5B, addition of Telaprevir (150 nM) produced a comparable reduction in luciferase activity with no effect on cellular metabolism (Fig. 4A).  Chemical structures are shown in Table 1.
Finally, epoxides 1 and 2 were tested against a chimeric infectious clone derived from genotype 2a isolates; Jc1 and containing  a Nanoluc reporter gene (Jc1-NLuc). Following electroporation of Jc1-NLuc RNA into Huh7 cells treatment with 1 reduced reporter activity to 14.8 ± 5.4% of control (Fig. 4C). Hence in the context of a transient assay with infectious HCV, 1 exhibited enhanced antiviral activity compared to a stable SGR system. This effect most likely reflects the differences between transient and stable systems for monitoring HCV replication, as it was also observed upon treatment with Telaprevir (Fig. 4).

Discussion
The NS2 autoprotease is essential in the HCV lifecycle and with no homology to any human proteases represents an attractive drug target. Recent evidence suggests processing of the NS2-NS3 junction may represent a rate limiting step in the onset of HCV genome replication (Madan et al., 2014), hence the virus lifecycle may prove particularly susceptible to inhibition of the NS2 autoprotease. Despite concerted efforts, identification of NS2 inhibitors has been hampered by weak activities of classical protease inhibitors and substrate peptides. Mechanism-based protease inhibitors conjugated to peptide derivatives, such as E64, or to NS2 substrate peptides have shown no inhibitory activity, suggesting the modification of peptide substrates may not be a viable route for the development of an NS2 autoprotease inhibitor. Here we report the use of bis-amido aromatic scaffolds conjugated to an epoxide warhead as inhibitors of the NS2 autoprotease. Treatment of NS2-3 auto-processing reactions with epoxides conjugated to small, non-peptide backbones reduced production of the NS3-FLAG proteolysis product in a dose responsive manner. Importantly, these inhibitory effects did not appear to be imparted solely by the epoxide, as modifications to the 'backbone' of epoxy-bis-amido derivatives altered activity. Such scaffolds might be modified to introduce selectivity to the NS2 autoprotease active site. While a degree of structurerelationship analysis has already been performed in vitro for this series, future work should attempt to increase the potency of inhibition before consideration as a clinical candidate.
Epoxide-derived mechanism-based inhibitors have been reported to show selectivity to cysteine proteases over serine or aspartyl proteases (Powers et al., 2002) and to specific target proteases within this class (Schiefer et al., 2013). In agreement, cerulenin and other epoxysuccinyl peptides display little reactivity to free thiols in the form of glutathione (Powers et al., 2002) (Avery et al., manuscript submitted). The observation of selectivity with these molecules suggest that epoxide warheads could be targeted to the NS2 autoprotease to increase both activity and selectivity through the use of a backbone scaffold not derived from peptide substrates. The inactivity of peptide conjugated mechanism-based inhibitors could be attributed to size, in line with the inactivity of cerulenin.
Observations made in the in vitro NS2-3 auto-processing assay were also made in the context of HCV genome replication. A concentration of 1 active against the NS2 autoprotease in vitro exhibited an inhibitory effect specific to NS2-dependent HCV genome replication whereas 2, inactive in vitro, also showed no activity in the same SGR cell lines. Furthermore, a genotype difference in sensitivity to 1 was observed both in vitro and in the context of NS2-dependent HCV replication, where enhanced sensitivity was observed against genotype 1b compared to 2a. Compound 1 therefore appears more potent against genotype 1b than genotype 2a both in vitro and in HCV model systems and it would be of interest to extend this analysis to other genotypes. Encouragingly 1 showed increased activity against full length infectious HCV compared to that observed against stable SGR harbouring cells. This suggests that, as expected, preexisting replication complexes present in stable SGR-harbouring cells are resistant to NS2 inhibitors.
Although the compounds described here are unlikely to be taken forward to the clinic, the current data represent the first reported example of direct pharmacological inhibition of the NS2 autoprotease to block HCV genome replication and exert an antiviral effect and thus provide proof of principle that the NS2 autoprotease is a valid target for future development.