In Vitro Characterization of a Purified NS2/3 Protease Variant of Hepatitis C Virus

The cleavage of the Hepatitis C Virus polyprotein between the nonstructural proteins NS2 and NS3 is mediated by the NS2/3 protease, whereas the NS3 protease is responsible for the cleavage of the downstream proteins. Purification and in vitro characterization of the NS2/3 protease has been hampered by its hydrophobic nature. NS2/3 protease activity could only be detected in cells or in vitro translation assays with the addition of microsomal membranes or detergent. To facilitate purification of this poorly characterized protease, we truncated the N-terminal hydrophobic domain resulting in an active enzyme with improved biophysical properties. We define a minimal catalytic region of NS2/3 protease retaining autocleavage activity that spans residues 904-1206, and includes the C-terminal half of NS2 and the N-terminal NS3 protease domain. The NS2/3 (904-1206) variant was purified from Escherichia coli inclusion bodies and refolded by gel filtration chromatography. The purified inactive form of NS2/3 (904-1206) was activated by the addition of glycerol and detergent to induce autocleavage at the predicted site between Leu1026 and Ala1027. NS2/3 (904-1206) activity was dependent on zinc ions and could be inhibited by NS4A peptides, peptides that span the cleavage site or an N-terminal peptidic cleavage product. This NS2/3 variant will facilitate the development of an assay suitable for identifying inhibitors of HCV replication.

but no homology between NS2/3 protease and other proteases has been identified. Gorbalenya et al. (20) have suggested that the NS2/3 protease could be a cysteine protease.
However, the observation that the activity is stimulated by metal ions and inhibited by EDTA led to the suggestion that the NS2/3 protease is a metalloprotease (12,13). Studies with classical protease inhibitors in an in vitro translation assay (15) have not resulted in a definitive classification.
Additionally, NS2/3 auto-processing is inhibited by mutations that are presumed to perturb the local conformation of the polyprotein precursor and may indicate an extreme sensitivity of the NS2/3 cleavage reaction to the correct folding of the protease (12,13,16). It was also demonstrated that NS2/3 activity can be detected upon co-transfection of constructs harboring defects in either the NS2 or NS3 regions, but not both, in conjunction with constructs expressing intact versions of the defective region. Furthermore, trans-cleavage activity in cells was suggested by the co-expression of an active NS2/3 protease containing a mutated cleavage site with an inactive NS2/3 protease precursor (H952A or C993A) supplying the cleavage substrate (16).
The biochemical characterization of the NS2/3 protease, as well as mechanistic and structural studies have been limited due to the lack of a pure recombinant form of the enzyme. In this This construct was designated pET-11d-NS2/3. The DNA was transformed into E. coli XL-1 Blue cells, isolated and sequenced. The NS2/3 protease construct was translated in vitro using a rabbit reticulocyte lysate translation kit (Promega) and 35 S-methionine as a label (NEN-Dupont).
Translated [ 35 S]-labeled products were separated by SDS-PAGE (15%) and visualized with a PhosphorImager (Molecular Dynamics, Inc.). To evaluate protein expression, the DNA was transferred into E. coli BL21(DE3)pLysS cells followed by a 2h-induction at 37°C with 1 mM IPTG. The level of expression was verified by SDS-PAGE (15%) and immunoblot analysis using an anti-NS3 polyclonal antibody.

NS2/3 N-terminal deletion constructs-
The N-terminal deletion constructs 815-1206, 827-1206, 855-1206, 866-1206, 904-1206 and 915-1206 were derived from the pET-11d-NS2/3 template and obtained by PCR using the appropriate synthetic oligonucleotide primers. The DNA was then transformed into E. coli XL-Blue cells, isolated and sequenced. Protein production and expression of the different constructs were verified as described above.
NS2/3 (904-1206) bacterial expression constructs-Four lysine residues followed by a hexahistidine tag were added to the N-terminus and four lysine residues were added at the Cterminus of the NS2/3 protease. Constructs were obtained using PCR and the pET-11d-NS2/3 template for the wild-type and the H952A mutant using the appropriate synthetic oligonucleotide primers. The primers also introduced a NdeI and a BamHI sites at 5'and 3' end respectively.
Enzyme Expression and Production-Full-length and N-terminal truncated variants of the NS2/3 protease were expressed in E. coli BL21(DE3)pLysS cells following induction with 1 mM IPTG for 2-3 h at 37°C to assess their level of expression. The highly expressed NS2/3 (904-1206) was selected for further characterization. A typical 4 L fermentation yielded approximately 20 g of wet cell paste. The cell paste may be stored at -80°C.
Inclusion bodies extraction-Following thawing at 23°C, the cells were homogenized in lysis buffer (5 mL/g) consisting of 100 mM Tris, pH 8.0, 0.1% Triton X-100, 5 mM EDTA, 20 mM MgCl 2 , 5 mM DTT followed by a DNase treatment (20 µg/mL) for 15 min at 4°C and a centrifugation at 22,000xg for 1h at 4°C. The resulting insoluble pellet was then washed twice by homogenization (5 mL/g) in 100 mM Tris, pH 8.0, 2% Triton X-100, 5 mM EDTA, 2 M urea, 5 mM DTT and centrifuged at 22,000xg for 30 min at 4°C. Finally, the insoluble material was washed in 100 mM Tris, pH 8.0, 5 mM EDTA, 5 mM DTT and inclusion bodies were recovered in the pellet by centrifugation at 22,000xg for 30 min at 4°C.
Protein purification from inclusion bodies-To solubilize the inclusion bodies, the pellet was suspended in the extraction buffer (4 mL/g) [100 mM Tris, pH 8.0, 6 M guanidine-HCl, 0.5 M NaCl] and maintained in that buffer for 1h at 23°C. The suspension was then centrifuged at 125,000xg for 30 min at 4°C. The resulting supernatant was filtered through a 0.22-µm filter.
The clarified filtrate was stored at -80°C. In order to purify the NS2/3 (904-1206), the filtrate was diluted 2-fold in 100 mM Tris, pH 8.0, 6 M guanidine-HCl, 0.5 M NaCl and applied to a Pharmacia Hi-Trap Ni +2 -chelating column. The NS2/3 protease was typically eluted with a 50 to 8 500 mM imidazole linear gradient with a 250 mM imidazole peak elution. Fractions from the major peak, containing the purified enzyme, were pooled.
Following a 15 min incubation at 23°C, the enzyme was loaded on a Pharmacia Superose 12 gel filtration column (HR 10/30 column, 24 mL bed volume) pre-equilibrated in refolding buffer (50 mM Tris, pH 8.0, 0.5 M arginine-HCl, 1% LDAO, 5 mM TCEP). The column was run with refolding buffer at 4°C at a flow rate of 0.4 mL/min. Fractions associated with the major peak were pooled. The inactive NS2/3 (904-1206) was stored at -80°C in the refolding buffer.   were purified per liter of E. coli cell culture. Immunoblot analyses showed that the NS2/3 (904-1206) did not auto-process upon refolding (Fig. 4B, lanes 6 and 9). The lack of autocleavage suggested that the cleavage assay conditions were either not optimal or that the enzyme was not refolded properly. In order to address the latter, the efficacy of refolding was assessed by monitoring the NS3 serine protease activity. As shown on Fig. 4A (dotted line), the NS3 serine protease activity co-eluted with the major NS2/3 protein peak. Our finding that NS2/3 (904-1206) possessed NS3 protease activity suggested proper refolding but sub-optimal reaction conditions permissive for autocleavage activity.

Activation of the NS2/3 (904-1206)-Earlier studies with NS2/3 (810-1615) in in vitro translation
systems suggested that detergents activate autocleavage (15). On the basis of this observation several detergents were evaluated for their ability to promote autocleavage (Fig. 5A). The detergents NP-40, Triton X-100, n-dodecyl-β-D-maltoside and CHAPS promoted autocleavage to a similar extent at concentrations varying from 0.125 to 1%. However, poor processing was observed in the presence of 0.125, 0.25 and 0.5% octyl-POE and LDAO and no processing was observed in the presence of 1% detergent.
In the absence of detergent, autocleavage was not observed even in the presence of up to 50% (w/v) glycerol (Fig. 5B, lanes 1-6). Glycerol was found to enhance the ability of 0.1% CHAPS to promote autocleavage (Fig. 5B, lanes 7-12). However, the effect of glycerol reached a plateau at concentrations higher than 30% and under these conditions, ~50% autocleavage was observed ( Fig. 5B, lanes 10-12). The effect of temperature and incubation time on the NS2/3 (904-1206) activity was also examined. Autocleavage was observed at 15°C and 23°C (Fig. 6, lanes 1-14) Insert Fig.5 here although more NS3 protease product was detected at earlier time points from the 23°C reaction (compare lanes 2-4 to 9-11). At both temperatures, the processing reaction appears to reach a plateau following a 3-5h incubation. Little to no product was detected from reactions performed at 37°C (lanes [15][16][17][18][19][20][21]. All of our subsequent experiments were therefore conducted at 23°C.

Characterization of NS2/3 (904-1206)-
In an attempt to verify that the protease activity was NS2/3-dependent, the activity of the wild-type enzyme was compared to the corresponding H952A mutant (Fig. 7). The absence of cleavage products from the H952A mutant, when assayed under conditions that paralleled the wild-type, confirmed that the activity was indeed NS2/3 protease dependent (Fig. 7, compare lanes 2-3 with lanes 6-7). As a control for the NS2/3 wild-type, no significant cleavage was detected upon removal of 1% n-dodecyl-β-D-maltoside from the assay buffer following a 24h incubation (Fig. 7, compare lanes 7 and 8). Furthermore, no change in the activity was observed in the presence of potent substrate-based NS3 protease inhibitors, confirming that the NS2/3 protease activity was independent of the NS3 serine protease activity (data not shown). Finally, N-terminal sequencing of the NS3 product confirmed that the cleavage occurred between the residues Leu1026 and Ala1027 (data not shown).
As a further characterization of our purified NS2/3 (904-1206), we studied its stimulation by zinc, and inhibition by EDTA as previously reported for the in vitro translated enzyme (12,13,15). These studies are easily biased by trace amounts of zinc present in solutions, such that addition of 1 µM zinc chloride to the cleavage buffer only resulted in a slight increase in activity (Fig. 8, compare lanes 1 and 2) and addition of 100 µM EDTA partially inhibited the enzyme  (Fig. 8, compare lanes 1 and 3). However, upon stringent preparation and zinc depletion of both the refolding and the cleavage buffers (with Chelex-100 resin), autocleavage was observed only upon addition of zinc to the reaction mixture and EDTA addition prevented zinc activation completely (Fig. 8, lanes 4-6), emphasizing the importance of zinc for NS2/3 protease activity.
The effect of classical protease inhibitors on NS2/3 (904-1206) autocleavage was evaluated and is listed in Table I. Of the inhibitors tested, the serine/cysteine protease inhibitors TLCK and TPCK, known to react with active site histidine residues, as well as the thiol-reactive agents iodoacetamide and N-ethylmaleimide, were effective inhibitors of autocleavage. The metal chelators EDTA and 1,10-phenanthroline were also effective inhibitors. 1,10-phenanthroline acts via its chelating properties since 1,7-phenanthroline did not inhibit autocleavage. This inhibition profile was similar to the profile of the in vitro translated NS2/3 (810-1615) (15), but did not provide a definitive classification of this viral protease.
Remarkably, the ability of the NS4A-derived peptides to inhibit the NS2/3 protease appears to correlate with their ability to activate the NS3 serine protease (data not shown).

DISCUSSION
The initial characterizations of processing at the NS2-NS3 junction were based on expression of the NS2-NS3 region in cell-free translation systems or various cellular systems (12-16, 19, 22).
The expression of HCV polyprotein precursors, including the NS2/3 protease, was also reported in E. coli (12,23). However, no reports have thus far reconstituted auto-processing of a purified recombinant NS2/3 protease. In the present work, the production of the HCV NS2/3 (904-1206) protease in E. coli, its purification and initial biochemical characterization are described.
Initial expression of a NS2/3 construct in E. coli encompassing the entire NS2 and NS3 protease domains, NS2/3 (810-1206), resulted in low level of expression that was probably due to the hydrophobicity of the NS2 protein. A series of N-terminal truncations identified the region spanning residues 904-1206 as a functional NS2/3 protease in which the putative transmembrane domain of NS2 was deleted. When expressed in E.coli, high levels of NS2/3 (904-1206) were obtained as an insoluble protein and therefore required denaturation and refolding.
Insert Table II  here Various buffer conditions were investigated in the refolding of NS2/3 (904-1206) in order to keep it in solution and promote its activation after refolding. The presence of 0.5 M arginine-HCl in the refolding buffer was necessary to keep the enzyme in solution. Arginine is a polar additive known to slightly destabilize proteins in a manner comparable to low concentrations of chaotrophs, and is likely to increase the solubilization of folding intermediates (24). A reducing agent was required to avoid formation of intermolecular disulfide bonds. Several detergents were evaluated. Detergents are required for refolding of hydrophobic proteins such as β-barrel membrane proteins and probably reduce aggregation during renaturation (25). LDAO was selected since it allows refolding and reconstitution of the NS3 protease activity without promoting autocleavage.
The refolded NS2/3 (904-1206) was activated by the non-ionic detergents NP-40, Triton X-100 and n-dodecyl-β-D-maltoside at concentrations above their respective CMC. CHAPS also promoted autocleavage at concentrations higher and lower than its CMC. The effect of CHAPS was enhanced by glycerol, while glycerol alone had no effect on NS2/3 activity; yet glycerol did potentiate the effect of other detergents (data not shown). The results suggest that glycerol or detergent alone is not sufficient to confer an optimal active conformation. Temperatures greater than 23°C were detrimental to the processing efficiency. Using an optimized cleavage buffer, a plateau of ~50% autocleavage was observed for the NS2/3 (904-1206) following a 5h incubation at 23°C. Further significant processing was not observed with prolonged incubation times. We cannot exclude the possibility that only a fraction of the refolded enzyme adopts an active conformation as the amount of active NS2/3 protease enzyme cannot be assessed directly by active site titration. In this respect, our estimation of the extent of autocleavage may be an underestimate as the efficiency of processing was calculated with the assumption that all of the refolded enzyme was active. Alternatively, the plateau in processing may reflect instability of the enzyme over time since no further improvement in autocleavage was observed following a 5h-incubation.
The following observations established the protease activity of the NS2/3 (904-1206) as NS2/3dependent: 1) autocleavage was precisely mapped to the Leu1026-Ala1027 junction; 2) autocleavage was activated by zinc and inhibited by EDTA; 3) autocleavage was not reconstituted with the refolded H952A mutant; 4) the activity was unperturbed by a potent substrate-based NS3 protease inhibitor. The NS2/3 (904-1206) demonstrated a protease inhibitor profile consistent with the in vitro translated NS2/3 (810-1615) (15). Yet the purified enzyme did not clarify the ambiguous inhibitor profile for this unique protease. Sequence comparison with other proteases have not facilitated the NS2/3 protease classification and no consensus motif for zinc binding is evident. Its classification as either a cysteine or a metalloprotease will require more in-depth biochemical characterization.
Inhibition of the NS2/3 processing by NS4A-derived peptides was previously reported in a cellfree translation system (22). In our study, NS4A-derived peptides inhibited the NS2/3 (904-1206) with IC50's as low as 0.6 µM and their potency correlated with their ability to activate the NS3 serine protease (data not shown). Crystallographic studies of the NS3 protease domain complexed with the central hydrophobic domain of NS4A reveal an extensive interaction between the N-terminus of NS3 and NS4A, leading to NS3 protease conformational changes (27,28). Consequently, inhibition by NS4A-derived peptides may be a result of an overall NS2/3 conformational change. Alternatively, interaction of NS4A-derived peptides with NS2/3 may induce local conformational changes at the cleavage site and thereby impair NS2/3 processing.
Toward this end, autocleavage at the NS2/3 junction appears to be determined primarily by polyprotein folding, similarly to cleavage at the NS3/4A junction by the NS3 protease (16,29).
Inhibition by NS4A-derived peptides via a local or global NS2/3 conformational change cannot be distinguished at this point and will require further investigation.
Mixing of wild-type NS2/3 (904-1206) and the H952A mutant under optimized assay conditions, did not inhibit autocleavage (data not shown). The cleavage site derived-peptide substrates P10-P10' and P6-P6' were then evaluated as potentially competing substrates. In a well defined assay system using purified NS2/3 (904-1206) and an optimized cleavage buffer (containing 50% glycerol and 0.5% n-dodecyl-β-D-maltoside), the P10-P10' and P6-P6' peptides inhibited NS2/3 processing with IC50's of 270 and 630 µM respectively; yet under identical assay conditions, no trans-cleavage of the peptides was observed (data not shown). The results suggest non-productive binding of the peptide substrate at the active site. Notably, the shorter P10-P1 Nterminal cleavage product peptide was the best inhibitor with an IC50 of 90 µM, whereas the corresponding C-terminal product was devoid of inhibitory activity. Product inhibition of the NS2/3 (904-1206) was reminiscent of the NS3 protease inhibition by peptides corresponding to the N-terminal cleavage product of the three intermolecular sites processed by the NS3 enzyme (30, 31). The observation that potent competitive hexapeptide NS3 protease inhibitors, while inhibiting the NS3 serine protease, failed to inhibit the NS2/3 (904-1206) activity suggests the absence of cross-talk between the two viral protease active sites. Experiments with the in vitro translated NS2/3 (810-1615) showed no significant inhibition with 500 µM peptide substrates that correspond to the P7-P7', P5-P5' residues, or the N-terminal cleavage product peptides (22).
This discrepancy with our results may reflect differences in the intrinsic binding affinity of these peptides for NS2/3 (810-1615) and NS2/3 (904-1206) and/or reflect differences in the assay systems used. Lastly, our results with the purified NS2/3 protease are inconsistent with the observation that certain defective NS2/3 variants partially inhibited cleavage of HCV precursor polyproteins containing a NS2/3 cleavage site (16). This difference may again be due to our use of a defined assay system. The NS2/3 protease appears to be unique among viral proteases in that its sole role in viral maturation is its auto-inactivation. NS2/3 protease auto-processing results in the separation of the NS2 and the NS3 protease domain with the subsequent translocation of NS2 into the ER membrane (19). HCV NS2/3 protease can be viewed as a positive-stranded RNA virus accessory protease, which is defined as a protease not involved directly in the proteolytic processing of key replicative proteins. Accessory proteases fall predominantly within the papain family, are found mostly in the N-terminal region of positive-stranded RNA virus polyproteins and are wide spread among positive-stranded RNA viruses. Though not directly involved in genome replication, accessory proteases appear to be indispensable for virus reproduction (20,32,33). Remarkably, functional HCV sub-genomic RNAs replicate in the absence of the structural proteins and NS2 in cell culture, and suggest that the NS2/3 protease activity is not essential for RNA replication (34). However, the NS2/3 protease activity appears to be essential for virion production in chimpanzees as no signs of HCV infection can be detected upon inoculation with an HCV infectious clone devoid of NS2/3 protease activity (17).
NS2/3 protease also seems to share some features with proteases encoded by other positivestranded RNA viruses. The Rubella Virus protease appears to be the most functionally related to the NS2/3 protease. The RV protease: (i) mediates a single cis-cleavage at its C-terminus, (ii) has a Cys/His catalytic dyad, and (iii) requires divalent cations for its catalytic activity (35). Liu et al. (36) recently proposed that the RV protease is a novel virus metalloprotease rather than a papain-like cysteine protease as originally thought. It remains to be seen whether the NS2/3 protease and the RV protease define a new class of viral metalloproteases.
The availability of an active purified recombinant NS2/3 protease, obtained by refolding the Nterminal truncated form 904-1206, will facilitate the detailed biochemical characterization of the enzyme and the development of in vitro assays using defined components for drug discovery purposes. Our results from inhibition studies suggest two approaches for drug design: identification of molecules that induce a conformational change and optimization of substratederived peptides.