In vitro resistance studies of hepatitis C virus serine protease inhibitors, VX-950 and BILN 2061: structural analysis indicates different resistance mechanisms.

We have used a structure-based drug design approach to identify small molecule inhibitors of the hepatitis C virus (HCV) NS3.4A protease as potential candidates for new anti-HCV therapies. VX-950 is a potent NS3.4A protease inhibitor that was recently selected as a clinical development candidate for hepatitis C treatment. In this report, we describe in vitro resistance studies using a subgenomic replicon system to compare VX-950 with another HCV NS3.4A protease inhibitor, BILN 2061, for which the Phase I clinical trial results were reported recently. Distinct drug-resistant substitutions of a single amino acid were identified in the HCV NS3 serine protease domain for both inhibitors. The resistance conferred by these mutations was confirmed by characterization of the mutant enzymes and replicon cells that contain the single amino acid substitutions. The major BILN 2061-resistant mutations at Asp(168) are fully susceptible to VX-950, and the dominant resistant mutation against VX-950 at Ala(156) remains sensitive to BILN 2061. Modeling analysis suggests that there are different mechanisms of resistance to VX-950 and BILN 2061.


From Vertex Pharmaceuticals Inc., Cambridge, Massachusetts 02139
We have used a structure-based drug design approach to identify small molecule inhibitors of the hepatitis C virus (HCV) NS3⅐4A protease as potential candidates for new anti-HCV therapies. VX-950 is a potent NS3⅐4A protease inhibitor that was recently selected as a clinical development candidate for hepatitis C treatment. In this report, we describe in vitro resistance studies using a subgenomic replicon system to compare VX-950 with another HCV NS3⅐4A protease inhibitor, BILN  It is estimated that 170 million patients worldwide and about 1% of the population in developed countries are chronically infected with hepatitis C virus (HCV) 1 (1). The majority of acute HCV infections become chronic, some of which progress toward liver cirrhosis or hepatocellular carcinoma (2,3). The current standard of care is pegylated interferon ␣ in combination with ribavirin, which has a sustained viral response rate of 40 -50% in genotype 1 HCV-infected patients, which accounts for the majority of the hepatitis C population in the United States and Japan, and of 80 -90% in patients infected with genotype 2 or 3 HCV (4, 5) (for a review, see Ref. 6). Thus, more effective therapeutic drugs with fewer side effects and shorter treatment durations are needed for patients infected with HCV.
HCV is an enveloped, single-stranded RNA virus with a 9.6-kb positive-polarity genome, which encodes a polyprotein precursor of about 3,000 amino acids. The HCV polyprotein is proteolytically processed by cellular and HCV proteases into at least 10 distinct products, in the order of NH 2 -C-E1-E2-p7-NS2-NS3-NS4A-NS4B-NS5A-NS5B-COOH (for a review, see Ref. 7). NS3 serine protease and helicase as well as NS5B RNA-dependent RNA polymerase are believed to be components of a replication complex responsible for viral RNA replication and have been shown to be essential for the HCV replication in chimpanzees (8). These HCV enzymes have been the major targets for the development of HCV-specific therapeutics during the past decade (for a review, see Ref. 9). However, successful discovery of a new HCV-specific drug candidate has been hampered by the lack of a robust, reproducible infectious virus cell culture system. The development of a HCV replicon system by Lohmann et al. (10) and subsequent optimization by several laboratories (11,12) has enabled quantitative evaluation of the antiviral potency of HCV inhibitors.
The HCV NS3⅐4A protease is responsible for cleavage at four sites within the HCV polyprotein to generate the N termini of the NS4A, NS4B, NS5A, and NS5B proteins (13)(14)(15)(16)(17). It has been shown that the central region (amino acids [21][22][23][24][25][26][27][28][29][30] of the 54-residue NS4A protein is essential and sufficient for the enhancement of proteolytic activity of the NS3 serine protease (18 -22). The central region of NS4A forms a tight heterodimer with the NS3 protein (21), for which the first x-ray crystal structure was solved in 1996 (23). BILN 2061 is the first HCV serine protease inhibitor (PI) in clinical trials for hepatitis C (24). In phase I trials, a 2-3-log reduction of HCV viral load was observed after a 2-day treatment, which provided the first proof-of-concept evidence that HCV NS3⅐4A protease inhibitors could be a new therapeutic option for hepatitis C patients (24). Recently, another HCV NS3⅐4A protease inhibitor, VX-950 (25), was selected as a clinical candidate for hepatitis C (26).
Resistance to specific antiviral drugs is a major factor limiting the efficacy of therapies against many retroviruses or RNA viruses, due to the error-prone nature of the viral reverse transcriptases or RNA-dependent RNA polymerases. As new HCV-specific inhibitors enter clinical trials, resistance could become a major problem in patients treated with drugs targeting HCV NS3⅐4A serine protease or NS5B RNA polymerase. In this report, we used the HCV subgenomic replicon system to identify resistance mutations against two HCV protease inhibitor clinical candidates, BILN 2061 and VX-950. The in vitro resistance mutations selected against either inhibitor resulted in a significant reduction in susceptibility to the inhibitor itself. However, the primary resistance mutations against BILN 2061 were fully susceptible to VX-950, and the major resistance mutation against VX-950 remained sensitive to BILN 2061. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Plasmid Construction-A DNA fragment encoding residues Ala 1 -Ser 181 of the HCV NS3 protease (GenBank TM CAB46913) was obtained by PCR from the HCV Con1 replicon plasmid, I 377 neo/NS3-3Ј/wt (renamed as pBR322-HCV-Neo in this study) (10) and inserted into pBEV11 2 for expression of the HCV proteins with a C-terminal hexahistidine tag in Escherichia coli. Resistance mutations against the HCV NS3⅐4A PI were introduced into this construct by PCR-based, sitedirected mutagenesis. To generate the HCV replicon containing the PI-resistant mutations, a 1.2-kb HindIII/BstXI fragment derived from the HCV Con1 replicon was subcloned into a TA cloning vector, pCR2.1 (Invitrogen). The PI-resistant mutations in the NS3 serine protease domain were introduced into the pCR2.1 vector containing the HindIII/ BstXI HCV fragment by PCR, and a 579-bp BsrGI/BstXI fragment containing the mutated residue was subcloned back into a second generation Con1 replicon plasmid containing three adaptive mutations, pBR322-HCV-Neo-mADE (see below). All constructs were confirmed by sequencing.
Generation of HCV Replicon Cells-The Con1 subgenomic replicon plasmid, pBR322-HCV-Neo, was digested with ScaI (New England Biolabs). Full-length HCV subgenomic replicon RNA was generated from the linearized DNA template using a T7 Mega-script kit (Ambion) and treated with DNase to remove the template DNA. The run-off RNA transcripts were electroporated into Huh-7 cells, and stable HCV replicon cell lines were selected with 0.25 or 1 mg/ml G418 (Geneticin) in Dulbecco's modified minimal essential medium (DMEM) containing 10% fetal bovine serum (FBS). HCV replicon-stable cells were maintained in DMEM, 10% FBS, and 0.25 mg/ml G418.
During the course of generation of the HCV subgenomic replicon stable cell lines, several different patterns of adaptive mutations were identified. One of these patterns contains three substitutions in the HCV nonstructural proteins, 3 which were introduced into the original pBR322-HCV-Neo plasmid by site-directed mutagenesis to generate the second generation subgenomic replicon plasmid, pBR322-HCV-Neo-mADE. When the T7 run-off RNA transcripts from the ScaI-linearized pBR322-HCV-Neo-mADE plasmid were electroporated into Huh7 cells, stable replicon cell colonies were formed at a much higher efficiency than the original Con1 replicon RNA. The resistance mutations identified in this study were introduced into the pBR322-HCV-Neo-mADE replicon plasmid by site-directed mutagenesis. Stable replicon cell lines were generated using the T7 transcripts derived from either wild type pBR322-HCV-Neo-mADE or the ones with the resistance mutations.
IC 50 Determination of HCV PIs in the HCV Replicon Cell Assay-HCV Con1 subgenomic replicon cells were maintained in DMEM containing 10% FBS and 0.25 mg/ml G418. On the day prior to the assay, 10,000 HCV replicon cells/well were plated in a 96-well plate in DMEM plus 10% FBS. The next day, the medium was removed, and a compound serially diluted in DMEM, 2% FBS, and 0.5% Me 2 SO was added. The replicon cells were incubated with the compounds for 48 h. Total cellular RNA was extracted using RNeasy-96 (Qiagen), and the copy number of the HCV RNA was determined by a quantitative, real time RT-PCR (Taqman) assay. The cytotoxicity of the compounds was measured using a mitochondrial enzyme-based cell viability assay, CellTiter 96 AQ ueous One Solution Cell Proliferation Assay (Promega). The IC 50 and CC 50 values of the compounds were calculated using four-parameter curve fitting (SoftMax Pro).
Selection of HCV PI-resistant Replicon Cells-The HCV Con1 subgenomic replicon stable cells were serially passed in the presence of 0.25 mg/ml G418 and slowly increasing concentrations of VX-950 (series A) or BILN 2061 (series B). The concentrations of VX-950 ranged from 3.5 M (or 10ϫ IC 50 ) in the 48-h assay (see above), to 28 M (80ϫ IC 50 ). For BILN 2061, the starting concentration was 80 nM (80ϫ IC 50 ), and the final concentration was 12.5 M (12,500ϫ IC 50 ). During the course of selection, replicon cells were split twice per week when a 70 -90% confluence was reached. Fresh HCV PI was added every 3-4 days regardless of whether the cell culture was split.
Identification of HCV PI Resistance Mutations-During the selection of HCV PI-resistant replicon cells, cell pellets were collected every time the cell culture was split. Total cellular RNA was extracted using the RNeasy miniprep kit (Qiagen). A 1.7-kb-long cDNA fragment encompassing the HCV NS3 serine protease region was amplified with a pair of HCV-specific oligonucleotides (5Ј-CCTTCTATCGCCTTCTTG-3Ј and 5Ј-CTTGATGGTCTCGATGG-3Ј) using the Titan One-Step RT-PCR kit (Roche Applied Science). The amplified products were purified using the QIA-quick PCR purification kit (Qiagen). To monitor the emergence of the HCV PI-related mutations in the HCV NS3 serine protease domain during the selection, the purified 1.7-kb RT-PCR products of PI-treated replicons from several different culture time points were subjected to sequence determination. To determine the frequency of PI-resistant mutations, the 1.7-kb RT-PCR products of HCV RNA of the VX-950 or BILN 2061-resistant replicon cells were ligated into the TA cloning vector pCR2.1 (Invitrogen). For each time point, multiple individual bacterial colonies were isolated, and the HCV NS3 protease coding region of the purified plasmid DNA was sequenced.
Expression and Purification of the HCV NS3 Serine Protease Domain-Each of the expression constructs for the HCV NS3 serine protease domain containing the wild type sequence or the resistance mutations (A156S, D168V, or D168A) were transformed into BL21/DE3 pLysS E. coli cells (Stratagene). Freshly transformed cells were grown at 37°C in a BHI medium (Difco) supplemented with 100 g/ml carbenicillin and 35 g/ml chloramphenicol to an optical density of 0.75 at 600 nM. Induction with 1 mM isopropyl-1-thio-␤-D-galactopyranoside was performed for 4 h at 24°C. Cell pastes were harvested by centrifugation and flash frozen at Ϫ80°C prior to protein purification. All purification steps were performed at 4°C. For each of the HCV NS3 proteases, 100 g of cell paste was lysed in 1.5 liters of buffer A (50 mM HEPES (pH 8.0), 300 mM NaCl, 0.1% n-octyl-␤-D-glucopyranoside, 5 mM ␤-mercaptoethanol, 10% (v/v) glycerol) and stirred for 30 min. The lysates were homogenized using a microfluidizer (Microfluidics, Newton, MA), followed by ultracentrifugation at 54,000 ϫ g for 45 min. Imidazole was added to the supernatants to a final concentration of 5 mM along with 2 ml of Ni 2ϩ -nitrilotriacetic acid resin pre-equilibrated with buffer A containing 5 mM imidazole. The mixtures were rocked for 3 h and washed with 20 column volumes of buffer A plus 5 mM imidazole. The HCV NS3 proteins were eluted in buffer A containing 300 mM imidazole. The eluates were concentrated and loaded onto a Hi-Load 16/60 Superdex 200 column, pre-equilibrated with buffer A. The appropriate fractions of the purified HCV proteins were pooled and stored at Ϫ80°C.
Enzymatic Assays for the HCV NS3 Serine Protease Domain-Enzymatic activity was determined using a modification of the assay described by Taliani et al. (27). An internally quenched fluorogenic depsipeptide (FRET substrate), Ac-DED(EDANS)EE␣Abu[COO]ASK (DABCYL)-NH 2 , was purchased from AnaSpec Inc. (San Jose, CA). The assay was run in a continuous mode in a 96-well microtiter plate format. The buffer was composed of 50 mM HEPES (pH 7.8), 100 mM NaCl, 20% glycerol, 5 mM dithiothreitol, and 25 M KK4A peptide (KKGSVVIVGRIVLSGK). The KK4A peptide represents the central region of the NS4A cofactor from genotype 1a with lysine residues added for improved solubility (28). The reaction was initiated by the addition of the FRET substrate after a 10-min preincubation of the buffer components with a 2 nM concentration of the NS3 protease at room temperature. The reaction was monitored at 30°C for 20 min using a Molecular Devices fmax fluorometric plate reader. The filters for excitation and emission wavelengths were 355 and 495 nm, respectively. For determination of substrate kinetic parameters, concentrations of the FRET peptide were varied from 0.5 to 7.0 M. Intermolecular quenching was not observed in this range. The substrate kinetic parameters, K m and V max , were determined by fitting the data to the Michaelis-Menten equation. Inhibition constants (K i ) were determined by titration of enzyme activity using the assay described above, except that compound dissolved in Me 2 SO (no greater than 2% (v/v) Me 2 SO; solvent only was used as control) was added to the buffer components and enzyme after the initial 10-min preincubation as described above. This mixture was incubated for an additional 15 min at room temperature prior to an incubation with the FRET substrate for another 20 min at 30°C. Seven or eight concentrations of compound were assayed, and the resulting data were fitted to the integrated form of Morrison's equation for tight binding inhibition (29). All substrate and inhibitor data were fitted using Marquardt-Levenberg nonlinear regression with GraphPad Prism software.
Modeling-VX-950 and BILN 2061 were modeled into the active site of the NS3 serine protease domain using the crystal structure of a full-length HCV NS3 protein fused with a NS4A polypeptide, which was published by Yao et al. (30) (Protein Data Bank code 1CU1). The coordinates of the protease domain of the A segment in this structure showed that the C-terminal strand of the NS3 protein binds in the substrate-binding site of the protease. The terminal carboxyl group of this strand is located near active site residues (His 57 , Asp 81 , and Ser 139 ) such that it forms hydrogen bonds with the side chains of His 57 and Ser 139 as well as the backbone amides of residues 137 and 139, which form the oxyanion hole. Additionally, the last six residues (residues 626 -631) of the NS3 protein form an extended, antiparallel ␤ strand along the edge of the E2 strand of the protease ␤ barrel (31) and makes 12 backbone-to-backbone hydrogen bonds. A product-based inhibitor like BILN 2061 is expected to bind to the NS3 protease in a similar fashion. Therefore, we utilized the coordinates of this crystal structure to build our models of inhibitor-protease co-complexes. BILN 2061 molecule was built using QUANTA molecular modeling software (Accelrys Inc., San Diego, CA), and manually docked into the active site such that its carboxyl group overlays with the C-terminal carboxylate of the full-length NS3 protein. The inhibitor molecule was then rotated such that it makes all of the following backbone hydrogen bonds: P1 NH with Arg 155 carbonyl, P3 carbonyl with Ala 157 NH, and P3 NH with Ala 157 carbonyl. This mode of binding placed the large P2 group of the BILN 2061 in direct clash with the Arg 155 side chain. To avoid the clash, the Arg 155 side chain was modeled in an extended conformation, which was observed in a crystal structure of NS3 protease complexed with a close analogue of BILN 2061 (32). The inhibitor was energy-minimized in two stages. In the first stage, only the inhibitor and the side-chain atoms of Arg 123 , Arg 155 , and Asp 168 of the protease were allowed to move during energy minimization for 1000 steps. In the second stage, all of the side-chain atoms of the protease were allowed to move along with the inhibitor for 1000 additional steps. This modeled structure closely mimics the published structure of the BILN 2061 analog (32). A similar procedure was adopted for modeling VX-950 into the NS3 protease active site. VX-950 was modeled as a covalent adduct with si-face attachment of the Ser 139 side chain to the keto carbonyl of the inhibitor. This binding mode was observed for analogous ketoamide inhibitors (33) and ketoacid inhibitors (34). The main chain of the inhibitor was overlaid with residues 626 -631 of the C-terminal strand of the fulllength NS3 protein such that it makes all of the following backbone hydrogen bonds: P1 NH with Arg 155 carbonyl, P3 carbonyl with Ala 157 NH, P3 NH with Ala 157 carbonyl, and P4 cap carbonyl with NH of Cys 159 . In this binding mode, the P2 group of VX-950 was placed in the S2 pocket without any need to move the Arg 155 side chain. The t-butyl and the cyclohexyl groups were placed in S3 and S4 pockets, respectively. To be consistent, we used the same two-stage energy minimization protocol used for the BILN 2061 model.
The side chain of Asp 168 is exposed to solvent. The valine side chain of the D168V mutant can adopt three canonical conformations with 1 ϭ 60, Ϫ60, or 180°. All three orientations of the Val 168 side chain were modeled. The interaction energy of the D168V mutant enzyme and the inhibitor was minimized by allowing the inhibitor and Val 168 atoms to move while fixing positions of all of the other atoms of the protein molecule. In all cases, the Val 168 side chain does not cause any steric clash with the inhibitor atoms. The serine mutation at Ala 156 was modeled by the following procedure. The Ala 156 side chain is in van der Waals contact with the P2 group of both of the inhibitors. The serine side chain of the A156S mutant was modeled at three canonical conformations of 1 ϭ 60, Ϫ60, and 180°, and the energy was minimized by holding the conformation of the rest of the protein fixed. These models were used to examine the effects of this mutation on inhibitor binding. The Ϫ60°conformation was found to have the lowest energy as it forms a hydrogen bond with the neighboring Arg 155 carbonyl, but it causes the maximal number of unfavorable contacts with both inhibitors. The 60 and 180°conformations are energetically equivalent, but the 60°conformation has fewer unfavorable contacts and was used in our analysis.

Development of Resistance to VX-950 in HCV Replicon
Cells-VX-950 (Fig. 1) (25) was recently selected as a clinical candidate for hepatitis C treatment (26). VX-950 is a reversible, covalent inhibitor of the HCV NS3⅐4A serine protease. Although competitive with the peptide substrate in the active site, it exhibits apparent noncompetitive inhibition as a result of its tight binding properties and time-dependent inhibition mechanism. 4 Incubation of the HCV Con1 subgenomic replicon cells with VX-950 resulted in a concentration-dependent decline of the HCV RNA level, as measured by the real time RT-PCR (Taqman) method (Fig. 2B). The IC 50 value of VX-950 is 354 nM in the 48-h assay.
To identify VX-950 resistance mutations, the Con1 subgenomic replicon cells were serially passed in the presence of 0.25 mg/ml G418 and slowly increasing concentrations of VX-950 (series A) (Fig. 2A). The starting concentration of VX-950 was 3.5 M, or 10 times the IC 50 , and the highest concentration was 28 M, or 80 times the IC 50 . Replicon cells were split, or the medium was replenished every 3 or 4 days, and fresh VX-950 was added. Since a HCV NS3 serine protease inhibitor, such as VX-950, inhibits the HCV polyprotein processing and consequently blocks replication of HCV RNA, the steady state level of HCV proteins and neomycin transferase protein gradually declined and eventually became undetectable in the presence of high concentrations of VX-950 (data not shown). Cells with low or no neomycin transferase protein proliferate at a gradually decreasing rate and eventually die in the presence of G418. Only HCV RNA with mutations that are resistant to VX-950 can replicate in the presence of high concentrations of VX-950 and support the growth of the replicon cells harboring them.  Replicon cells in series A grew normally for the first 10 days in the presence of 3.5 M VX-950. After 10 days, the series A cells grew significantly more slowly, and massive cell death was observed between days 10 and 17 ( Fig. 2A). Normal growth did not resume until day 21. The IC 50 of VX-950 against the series A replicon cells at day 56 was determined to be 8.1-12.0 M, which is 23-34-fold higher than the IC 50 (354 nM) against wild-type replicon cells (Fig. 2B).
Total cellular RNA from the series A cells at days 10, 21, and 56, was extracted and subjected to RT-PCR to amplify the coding region of the NS3 serine protease domain. The RT-PCR product was bulk-sequenced to identify the position(s) of potential mutations that could be responsible for the observed reduction in sensitivity to VX-950. No VX-950-related mutation was observed in the NS3 serine protease domain of the series A replicon cells at day 10 when compared with the wild type Con1 replicon cells cultured in the absence of VX-950. At days 21 and 56 in series A, substitutions at Ala 156 in the protease domain were observed, suggesting that mutations at residue 156 might be critical for the reduced sensitivity to VX-950. No mutation was found at any of the four proteolytic sites in the HCV nonstructural protein region that are cleaved by the NS3⅐4A serine protease. To delineate the identity and frequency of the substitutions, a 1.7-kb RT-PCR product of the series A replicon cells at day 7 or 98 was subcloned into the TA vector, and multiple clones were sequenced for both samples. All clones derived from the day 7 samples contained the wild type Ala 156 . In the day 98 sample of the series A replicon cells, which had been cultured in the presence of 28 M VX-950 for 63 days, 79% or 60 out of 76 clones had an alanine to serine (A156S) substitution.
Development of Resistance to BILN 2061 in HCV Replicon Cells-Another HCV NS3⅐4A protease inhibitor, BILN 2061 ( Fig. 1) is the first PI to demonstrate efficacy in hepatitis C patients (24). HCV replicon cells resistant to BILN 2061 (series B) were selected in a similar manner as for VX-950. Again, wild-type Con1 subgenomic HCV replicon cells were serially passed in the presence of 0.25 mg/ml G418 and slowly increasing concentrations of BILN 2061 (Fig. 3A). Series B replicon cells grew normally for the first 7 days in the presence of 80 nM BILN 2061 or 80-fold above the IC 50 . However, the proliferation of series B cells slowed down significantly after day 7, and massive cell death was observed between days 7 and 17. As before, normal growth did not resume until day 21. BILN 2061 had an IC 50 value of 1.0 -1.8 M against the series B cells at day 59, which is 1,000 -1,800-fold higher than the IC 50 (1 nM) against wild-type replicon cells (Fig. 3B).
No BILN 2061-related mutation was observed in the NS3 serine protease domain at day 7. By day 24, a variety of substitutions were observed at amino acid 168 of the NS3 protein, suggesting that substitutions at residue 168 may account for the resistance against BILN 2061. No mutation at the four sites in the HCV nonstructural protein region that are cleaved by the NS3⅐4A serine protease was observed. To determine the frequency of various substitutions at the NS3 residue 168, the HCV serine protease of the series B replicon at day 98, which was cultured in the presence of 3.2 M BILN 2061, was sequenced. 60 of 94 clones or 64% had an Asp 168  The kinetic parameters for the FRET substrate for the wild type NS3 protease domains from genotype 1a and 1b were identical (Table I) under our assay conditions. Although the NS4A peptide co-factor was from HCV genotype 1a, no discernible difference in the kinetic parameters was observed. This is consistent with molecular modeling, which suggests that the conservative variations in the central region of NS4A between genotypes 1a and 1b do not affect the interaction between the NS4A core peptide and the NS3 protease domain. K i values of VX-950 and BILN 2061 were determined using genotypes 1a and 1b wild type protease, and there were no statistically significant differences between the two wild type proteases (Table II).
The kinetic parameters of the FRET substrate for the A156S mutant protease were virtually the same as those of the wild type protease (Table I). However, the K i value of VX-950 was 2.9 M against the A156S mutant protease, which is 29-fold higher than that against the wild type protease (0.1 M) (Table  II). BILN 2061 had a K i value of 112 nM against the A156S mutant, which was 6-fold higher than that against the wild type protease, 19 nM (Table II).
The HCV RNA level in the replicon cells containing the A156S substitution was similar to that of wild type replicon cells (data not shown), which is consistent with the similar enzymatic catalytic efficiency of the A156S mutant and the wild type NS3 serine proteases. The IC 50 value of VX-950 against the A156S replicon cells was 4.65 M, which is 12 times higher than that against the wild type replicon cells (0.40 M) ( Table III). The difference between the IC 50 values of BILN 2061 against the A156S (7 nM) and the wild type replicon (4 nM) cells was not significant (Table III).
The Major BILN 2061-resistant Mutants, D168V and D168A, Remain Fully Susceptible to VX-950 -The substrate kinetic parameters were not affected by the D168V mutation and showed only minor changes (less than 10-fold) for the D168A mutant as indicated by the comparison of the k cat and k cat /K m values of the wild type and the two mutant NS3 serine proteases (Table I). Similarly, no significant effect of either substitution at Asp 168 was observed on the K i value of VX-950 (Table II). However, the substitution of valine or alanine for aspartic acid at position 168 resulted in a mutant NS3 protease that was not inhibited by up to 1.2 M BILN 2061 (Table II). These data indicate that either mutant protease is at least 63-fold less susceptible to BILN 2061 as compared with the wild type protease. The actual magnitude of resistance cannot be determined, since BILN 2061 was not fully soluble at concentrations greater than 1.2 M in the assay buffer, as measured by the absorbance at 650 nm. The D168V or D168A mutation was also introduced into the wild type HCV replicon by site-directed mutagenesis, and a stable replicon cell line carrying either substitution was generated. BILN 2061 had an IC 50 of 5.09 M against the D168V replicon cells, which is more than 1,300 times higher than that against wild type replicon cells (4 nM) ( Table  III). The IC 50 of BILN 2061 was 1.86 M against the D168A mutant replicon. There was little change in IC 50 values of VX-950 against the D168V, the D168A, and the wild type replicon cells (Table III). DISCUSSION In this study, we identified in vitro resistance mutations in HCV replicons against two clinical candidates for the hepatitis C treatment, VX-950 and BILN 2061, both peptidomimetic inhibitors of the HCV NS3⅐4A protease. The dominant resistance mutation observed against VX-950 was a substitution of Ala 156 in the HCV NS3 protease domain with a serine. The major mutations, which conferred resistance to BILN 2061, were substitutions at the residue 168 in the NS3 serine protease domain. Substitutions at Asp 168 have been identified in a previous study as the resistance mutations against a less potent HCV protease inhibitor, which has an IC 50 of about 1 M in the replicon cell assay (35).
Ala 156 is located on the E2 strand in the HCV NS3⅐4A protease structure (31). Several backbone atoms of this strand (mainly the carbonyl of Arg 155 and both the main-chain nitrogen and carbonyl of Ala 157 ) make hydrogen bonds with the backbone atoms of substrates or substrate-based inhibitors. In our structural model of the VX-950-NS3 protease co-complex (Fig. 4), three hydrogen bonds are formed between P1 NH and Arg 155 carbonyl, P3 carbonyl and Ala 157 NH, and P3 NH and Ala 157 carbonyl. The same hydrogen bonds are also formed in     (Fig. 4) (31). It is also part of the S4 binding pocket. The aliphatic part of this side chain is in van der Waals contact with the terminal cyclopentyl group of BILN 2061, which is not expected to be affected by the D168V mutation, since a valine side chain at this position does not cause any steric clash with the inhibitor. However, this D168V substitution results in the loss of salt bridge interaction with the Arg 155 side chain on the neighboring E2 strand, which in turn makes multiple contacts with the large P2 group of BILN 2061 in our model. The conformation of the Arg 155 (Fig. 4, colorcoded in cyan) in the model of the BILN 2061-wild type NS3 protease complex is no longer energetically favored in the D168V mutant for two reasons. First, it cannot remain close to the backbones of the E2 strand in the absence of the salt bridge interaction between Arg 155 and Asp 168 . Second, an uncompensated and solvent-exposed positive charge of Arg 155 side chain will seek a larger solvation shell, as observed in the crystal structures of the apoprotease and the two protease-inhibitor complexes that are available in the Protein Data Bank (codes 1DY8 and 1DY9) (34). These conformations of Arg 155 are in direct clash with the P2 quinoline group of BILN 2061 and destabilize its binding. Therefore, substitution of Asp 168 with any amino acid, other than glutamate, will disrupt the salt bridge interactions with Arg 155 and result in reduction of BILN 2061 binding. On the other hand, the conformation of Arg 155 in the two published crystal structures of the NS3 protease-inhibitor complex is similar to that in our VX-950-protease complex model (color-coded in orange in Fig. 4). In addition, this conformation of Arg 155 confers stabilization of VX-950 binding as it allows the maximal number of van der Waals contacts between the Arg 155 side chain and the inhibitor. Therefore, VX-950 is not expected to be affected by the substitutions at Asp 168 as compared with BILN 2061.
A Blast search of the GenBank TM data base was conducted using the amino acid sequences of the HCV NS3 protease domain from the Con1 replicon. A total of 437 HCV isolates from all six major genotypes were identified, and Ala 156 is absolutely conserved in all of the isolates. The lack of polymorphism at amino acid 156 of the NS3 serine protease suggests that substitution at this position might be unfavorable for viral replication. It remains to be examined if the substitution at Ala 156 has a deleterious effect on the virus life cycle. Three naturally occurring variants were observed at amino acid 168 of the HCV NS3 serine protease. The vast majority (over 96%) of the 437 HCV isolates reported in GenBank TM have aspartic acid at position 168. Glutamate was found at the residue 168 in 10 isolates of genotypes 1b or 5. Since glutamate 168 is expected to be able to maintain the salt bridge with Arg 155 and Arg 123 , these genotype 1b or 5 isolates with Glu 168 would remain susceptible to both HCV protease inhibitors, BILN 2061 and VX-950. However, six isolates of genotype 3 reported in GenBank TM have glutamine at position 168. For these geno-type 3 HCV strains, no salt bridge between Gln 168 and Arg 155 is expected to form, and the Arg 155 side chain would cause interference with binding of the large P2 group of BILN 2061 as was suggested by our modeling analysis of the D168V mutant. The Gln 168 NS3 serine protease is expected to have a reduced susceptibility against BILN 2061 but not against VX-950. However, these six isolates of genotype 3 with a Gln 168 have a threonine instead of arginine at position 123. It is not obvious whether such a double mutant will have the same differential effect on the binding of the protease inhibitors, VX-950 and BILN 2061.
One of the major factors limiting the efficacy of virus-specific therapies against many retroviruses and RNA viruses is the development of resistance to antiviral drugs. Resistance to inhibitors of HIV reverse transcriptase or protease is caused by specific mutations in the viral enzymes (for a review, see Ref. 36). Due to the error-prone nature of the HIV reverse transcriptase, resistance mutations were selected under the pressure of HIV inhibitors in patients who were on monotherapy. It is estimated that all possible single mutations can be randomly generated within 1 day in an HIV-infected patient. Although elimination or cure of HIV infection in patients remains an elusive task, multidrug combination or "cocktail" therapies are much more effective than monotherapy to reduce HIV viral load and to suppress the emergence of resistance mutations. Drug-resistant strains of hepatitis B virus containing specific mutations in the viral polymerase are the primary cause of treatment failure of lamivudine or 3TC, the first approved hepatitis B virus-specific drug. It was reported that the frequency of resistance mutations against 3TC increased from 24% in the first year to 67% in the fourth year in the hepatitis B patients treated with lamivudine (37).
From these lessons, it is clear that as new HCV-specific inhibitors enter clinical trials, resistance could become a major problem in patients treated with drugs targeting the HCV enzymes. The replication rate of HCV in patients was reported to be in the range of 10 10 to 10 12 viral particles per day, higher than the viral replication rate in HIV-infected patients (38). In this study, we demonstrate that the major in vitro resistant mutations against BILN 2061 remain fully susceptible to VX-950, and the dominant VX-950 resistance mutation is still sensitive to BILN 2061. In vitro resistance mutations against the HCV polymerase inhibitors have also been identified in the replicon system (39,40). These studies suggest that future hepatitis C therapy involving small molecule inhibitors of HCV enzymes might require multidrug combination, as in the case of the current HIV treatments. Clearly, combinations of small molecule, HCV-specific inhibitors with either interferon ␣ or other HCV-specific inhibitors will represent an important strategy to suppress the emerging of resistance and increase the efficacy of HCV therapy. Our current findings should prove to be useful in optimization of future protease inhibitor-based therapies against hepatitis C.