Crystal structures of the RNA-dependent RNA polymerase genotype 2a of hepatitis C virus reveal two conformations and suggest mechanisms of inhibition by non-nucleoside inhibitors.

Crystal structures of the RNA-dependent RNA polymerase genotype 2a of hepatitis C virus (HCV) from two crystal forms have been determined. Similar to the three-dimensional structures of HCV polymerase genotype 1b and other known polymerases, the structures of the HCV polymerase genotype 2a in both crystal forms can be depicted in the classical right-hand arrangement with fingers, palm, and thumb domains. The main structural differences between the molecules in the two crystal forms lie at the interface of the fingers and thumb domains. The relative orientation of the thumb domain with respect to the fingers and palm domains and the beta-flap region is altered. Structural analysis reveals that the NS5B polymerase in crystal form I adopts a "closed" conformation that is believed to be the active form, whereas NS5B in crystal form II adopts an "open" conformation and is thus in the inactive form. In addition, we have determined the structures of two NS5B polymerase/non-nucleoside inhibitor complexes. Both inhibitors bind at a common binding site, which is nearly 35 A away from the polymerase active site and is located in the thumb domain. The binding pocket is predominantly hydrophobic in nature, and the enzyme inhibitor complexes are stabilized by hydrogen bonding and van der Waals interactions. Inhibitors can only be soaked in crystal form I and not in form II; examination of the enzyme-inhibitor complex reveals that the enzyme has undergone a dramatic conformational change from the form I (active) complex to the form II (inactive).

Hepatitis C virus (HCV) 1 is a debilitating human pathogen affecting an estimated 3% of the world's population (1). The virus establishes chronic infection in the majority of the cases, eventually leading to the development of liver diseases such as cirrhosis and hepatocellular carcinoma in almost 15-20% of those infected. Although a great deal of research has been focused on the development of anti-HCV agents, to date no vaccine is available and there is no effective therapy for all genotypes of HCV. The current therapies (a combination of polyethylene glycol-treated ␣-interferon and ribavirin) are associated with limited efficacy and severe adverse side effects. Therefore, the development of HCV-specific antiviral agents is needed urgently to alleviate this serious health problem.
HCV is a positive single-stranded-RNA virus and a member of the Flaviviridae family (2). Six major genotypes and 11 subtypes of HCV are known. The viral genome is comprised of a single open reading frame that codes a polyprotein of ϳ3000 amino acids (1). The polyprotein is subsequently processed into individual components by host and viral-encoded peptidases. The polyprotein consists of three structural proteins (C, E1, and E2) and six non-structural proteins (NS2, NS3, NS4A, NS4B, NS5A, and NS5B) (3)(4)(5). Among these, NS5B polymerase and the NS3 peptidase-helicase are the key enzymes involved in the genome replication and polyprotein processing of HCV. Therefore, these enzymes are potential drug targets, emphasizing the need for detailed studies of these enzymes in HCV.
NS5B has been characterized as an RNA-dependent RNA polymerase (RdRP) based on in vitro experiments (6,7). Several crystal structures of NS5B HCV polymerase (HCV-BK, genotype 1b) in several crystalline forms have been determined. The structure resembles a right hand with fingers, palm, and thumb domains (8 -10). More importantly, HCV polymerase has a fully encircled active site that is unique compared with other polymerases. The structure of HCV polymerase in complex with ribonucleotides has been analyzed (11). Recently substrate complexes of HCV RNA polymerase (HC-J4), describing nucleotide import and de novo initiation, have revealed that the polymerase does not undergo marked structural changes upon nucleotide binding (12). Structures of the RNA-dependent RNA polymerases from polio, bacteriophage ⌽6, and rabbit hemorrhagic disease viruses are also known (13)(14)(15). Structures of unliganded and ternary complexes of the polymerase from human immunodeficiency virus type 1 reverse transcriptase (HIV1-RT), which is both an RNA-and a DNA-dependent DNA polymerase, have been determined (16,17). The thumb domain of the HIV1-RT polymerase moves ϳ20°upon binding the template, primer nucleic acid.
Despite the low sequence homology among polymerases, conserved domain organization persists among RNA-dependent RNA polymerases, DNA-dependent DNA polymerases, DNA-* This work was supported in part by funding from the Alberta Heritage Foundation for Medical Research for the purchase of area detector facilities and Canadian Institutes of Health Research for operating expenses. 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.
The  (18,19), and we have recently reported on the crystal structures of HCV polymerase genotype 1b/inhibitor complexes (20). As part of ongoing efforts to study protein/inhibitor complexes for the development of antiviral drugs, we have determined the structure of HCV polymerase genotype 2a complexed with two thiophene 2-carboxylic acid non-nucleoside inhibitors (Fig. 1). These inhibitors were synthesized as part of ongoing structure-activity relationship optimization efforts that have been described elsewhere (21)(22)(23). The details of the inhibitor binding site, protein-inhibitor interactions, and plausible mechanisms of inhibition will be discussed in the second part of the paper. The HCV polymerase genotype 2a used in the present study contains a Cterminal deletion of 21 amino acid residues (⌬21) and an Nterminal hexahistidine tag.

MATERIALS AND METHODS
Purification and Biochemical Studies-For crystallographic studies and measurement of the inhibitory effects of compounds, a soluble C-terminal truncated form of the polymerase 2a enzyme was obtained by the following approach. A full-length synthetic gene was produced initially by a PCR-mediated gene assembly based on the procedure described by Stemmer et al. (24). HCV polymerase 2a full-length sequences were identified from the NCBI data bank, and a consensus sequence based on codon usage in bacteria was produced. From this consensus sequence, a series of overlapping oligonucleotides (Invitrogen) spanning the complete gene were synthesized. The gene assembly procedure was performed by fusion PCR in two steps. Briefly, the first step involved the production of four sub-fragments of the gene, each containing overlapping sequences for the second gene assembly step. Each of the four reactions underwent one round of annealing/extension (40 cycles of 94°C 20 s, 51°C for 20 s, and 72°C for 45 s) using Vent polymerase (New England Biolabs). Following the annealing/extension of the primers, the product was diluted 100-fold, and the outer primers of each fragment were added for a round of amplification (40 cycles of 94°C for 30 s, 62°C for 30 s, and 72°C for 90 s). For the second step, the product of each reaction was then diluted 25-fold and annealed together (40 cycles at 94°C for 20 s, 62°C for 20 s, and 72°C for 45 s). A second round of PCR (40 cycles at 94°C for 30 s, 62°C for 30 s, and 72°C for 120 s) was performed by diluting the annealed fragments 100-fold and using the outer primers to create the complete full-length gene. Then, the final PCR product was cloned and used as a PCR template for the production of the HCV polymerase 2a ⌬21 using a 5Ј-primer containing a His tag and a 3Ј-primer 63 bases upstream from the stop codon. This PCR product was cloned into a pET-21b vector (Novagen Inc., Madison, WI), and its sequence was confirmed by DNA sequencing and then expressed in Escherichia coli BL21 (DE3). Soluble polymerase was subsequently obtained as described previously (24). Briefly, the polymerase was initially purified using Hi-Trap nickel-nitrilotriacetic affin-ity chromatography with a 10 -500 mM imidazole gradient. The polymerase fractions were pooled, and the imidazole was removed using PD10 desalting columns (Amersham Biosciences). Further purification of the polymerase was performed by passing the nickel-nitrilotriacetic fractions through a Hi-Trap Mono S cation exchange column. Positive fractions were exchanged into a buffer containing 10 mM Tris, pH 7.5, 10% glycerol, 5 mM dithiothreitol, and 600 mM NaCl, and glycerol was added to a final concentration of 40% for storage at Ϫ80°C and for subsequent activity and kinetic assays. Protein for crystal structure studies was concentrated using an Ultra-15 centrifugal filter unit (Millipore) to ϳ10 mg/ml and stored at 4°C.
In Vitro NS5B Assay-The inhibitory effects of compounds on HCV NS5B genotype 2a polymerization activity were measured by evaluating the amount of radiolabeled UTP incorporated by the enzyme on a homopolymeric RNA template/primer (24,25). The 50% inhibitory concentrations (IC 50 ) of the compounds were determined in a final volume of 50 l of reaction mixture consisting of 20 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 0.5 mM MnCl 2, 1 mM dithiothreitol, 50 mM NaCl, 400 ng of purified NS5B enzyme, 500 ng of poly(rA)⅐oligo(dT) 15 (Invitrogen), 30 M UTP, and 1.5 Ci of ␣-32 P-labeled UTP (3000 Ci/mmol; Amersham Biosciences). RNA-dependent RNA polymerase reactions were allowed to proceed for 120 min. at 22°C. The reactions were stopped by the addition of 10 l of 0.5 mM EDTA. Thereafter, a volume of 50 l (25 g) of salmon sperm DNA (Invitrogen) and 100 l of a solution of 20% trichloroacetic acid at 4°C were added to the mixture followed by incubation on ice for 30 min to ensure complete precipitation of nucleic acids. The samples were then transferred onto 96-well MultiScreen filter plates (Millipore). The filter plates were washed with 600 l of 1% trichloroacetic acid per well and dried for 20 min at 37°C. 50-l of liquid scintillation mixture (Wallac Oy, Turku, Finland) was added, and the incorporated radioactivity was quantified using a liquid scintillation counter (Wallac MicroBeta Trilux; PerkinElmer Life Sciences). The IC 50 values were calculated using the computer software GraphPad Prism (version 2.0; GraphPad Software Inc., San Diego, CA).
Crystallization and Data Collection-Crystals of HCV 2a NS5B were grown by the hanging drop method at room temperature. 3 l of reservoir solution (15% polyethylene glycol 8000, 0.2 M ammonium sulfate, 80 mM sodium citrate, pH 6.0, 7% glycerol, 4% 1,6-hexanediol, and 1% benzamidine) were mixed with 1.5 l of protein solution (10 mg/ml protein concentration in 50 mM citrate buffer, pH 6.0, 5% glycerol, and 5 mM ␤-mercaptoethanol), and the resultant drops were equilibrated against 1 ml of reservoir solution. Needle-shaped crystals grew to a maximum size (0.04 ϫ 0.04 ϫ 0.5 mm) within 2-4 days. Two crystal forms of the enzyme were observed from different protein preparations. Both crystal forms belong to the space group C222 1 but differ markedly in the a-axis unit cell dimension (Table I). Throughout this paper the crystal form with the larger a cell dimension will be referred to as form I, whereas that with the smaller a cell dimension will be referred to as form II. Protein/inhibitor complexes were prepared by soaking experiments. Crystals were soaked in 2 mM inhibitor solution for ϳ12 h.
Intensity data from both crystal forms and from the protein-inhibitor B complex were collected at the beam line 8.3.1 of the Advanced Light Source in Berkeley, CA, whereas data from the protein-inhibitor A complex were collected using an R-AXIS IV ϩϩ image plate detector with copper K␣ radiation generated by a Rigaku RU-300 rotating anode x-ray generator. Both crystal forms contain one NS5B molecule in the asymmetric unit with solvent contents of ϳ64 and 60% in form I and form II, respectively (26). The data sets were indexed, integrated, and scaled using the programs DENZO and SCALEPACK (27). Data collection statistics of both crystal forms and the protein-inhibitor complexes are given in Table I.
Structure Determination and Refinement-Structure solutions of both crystal forms were achieved by molecular replacement with the CNS package (28) using the unliganded 1b genotype polymerase structure (8) (Protein Data Bank code 1C2P) as the search model. The inhibitor complex structures were solved by the difference Fourier method. A difference Fourier map, ͉F PI ͉ Ϫ ͉F P ͉ (␣ calc (͉F PI ͉ values of the protein-inhibitor complex and ͉F P ͉ values from the apoprotein), permitted an initial positioning of the inhibitor molecule into the difference density. Structure refinement was carried out with the CNS package using a maximum likelihood target (28). All of the structures reported here were refined in the same manner. Initially, the structure was refined by treating the whole molecule as a rigid body. Subsequently, the model was subjected to iterations of positional refinement, simulated annealing, torsion angle dynamics, and individual B-factor refinement. Electron density maps (2͉F o ͉ Ϫ ͉F c ͉ and ͉F o ͉ Ϫ ͉F c ͉) were calculated at this stage of refinement, and model building was performed wherever necessary using XtalView (29). Extensive model building was done in

Structure of HCV RNA-dependent RNA Polymerase Genotype 2a
the thumb domain of the molecule in form II. The refinement was continued using CNS. The R work and R free values were monitored closely throughout the refinement. Once the refinement had converged to an R work value of 0.24, identification of bound water molecules in the model was carried out. This was achieved in several stages based on electron density peaks of at least 3 in ͉F o ͉ Ϫ ͉F c ͉ and 1 in 2͉F o ͉ Ϫ ͉F c ͉ maps. Cycles of position and B-factor refinement, correction of the model using Fourier maps, and the identification of water molecules continued until no significant peaks were left in the electron density maps. Bulk solvent correction and anisotropic B-factor scaling were incorporated during the entire refinement. The molecule in form I was well defined in the electron density maps except for the region Pro 149 -Gly 153 . The electron density maps for the form II crystals were very clear for all the residues from Ser 1 to Lys 548 . The electron density map corresponding to a portion of the ⌬1 loop, Pro 25 -Asn 35 of the form II crystals is shown in Fig. 2. The stereochemical validity of the structure was examined using PRO-CHECK (30). In all of the structures Ͼ88% of the total number of residues lie in the most allowed regions of the Ramachandran plot. The refinement parameters are given in Table I. The refined atomic coordinates have been deposited in the Protein Data Bank (accession codes 1YUY, 1YV2, 1YVZ, and 1YVX for the form I, form II, and the NS5Binhibitors A and B complexes, respectively).

RESULTS AND DISCUSSION
Overall Features of NS5B Polymerase in Form I and Form II-The three-dimensional structures of the NS5B HCV polymerase genotype 2a in both crystal forms have the same righthand disposition of fingers, palm, and thumb domains (Fig. 3) as seen in HCV polymerase genotype 1b and also in other polymerases. However, the detailed structure of the polymerase in form I is substantially different from that in form II. The loop (⌬1) protruding from the fingers domain and comprising residues Ile 11 -Ser 46 (Fig. 4) exhibits significant structural variability in both forms. The root mean square deviation of a part (Lys 20 -Thr 40 ) of this loop between the two molecules is 1.82 Å (16 C␣ atoms) as compared with the overall root mean square deviation of 1.08 Å (531 C␣ atoms). The region Asn 24 -Leu 31 is a helix in form I, whereas in form II it is part of a small ␤-hairpin (Fig. 4). The average B-factor of the protein atoms of NS5B molecule in form II is 27.8 Å 2 , whereas the average B-factor of the atoms of the Lys 20 -Thr 40 residues is 47.7 Å 2 , indicating a high degree of flexibility in this region. The molecule in form I is relatively rigid, as is evident from its relatively low B-factors. The average B-factors are 26.5 and 30.0 Å 2 for protein atoms of the entire molecule and atoms in the region Lys 20 -Thr 40 , respectively. Extensive interactions between the extension (⌬1 loop) from the fingers domain and the thumb domain of the molecule in form I maintain the polymerase in a more rigid arrangement. The number of van der Waals interactions at the fingers-thumb domains interface listed in Table  II shows that the molecule in form I has 28% more intramolecular interactions than the molecule in form II. The hydrogen bonding networks in the region Lys 20 -Thr 40 are listed in Table  III, and the observed differences clearly demonstrate that the molecules adopt different secondary structures in this region. An analysis of the crystal packing interactions indicates that the neighboring molecules in the crystal lattice have very little effect on the conformation of the region Lys 20 -Thr 40 in both crystal forms. Recent studies on HCV polymerase genotype 1b involving the mutation of Leu 30 to polar serine or arginine amino acids resulted in a non-functional polymerase, presumably due to a local perturbation in the ⌬1 loop (32). Our studies therefore provide structural evidence that the ⌬1 loop is, to a major extent, responsible for determining the active state of HCV polymerase genotype 2a and, by analogy, for other RNA polymerases including the polymerase of HCV genotype 1b.

TABLE II List of residues making van der Waals contacts (distance cut-off 4 Å) between the extension from the fingers (⌬1 loop) and thumb domains of the molecules in crystal forms I and II
The number of contacts between the residues are given in parentheses.
Between the form I and form II, the fingers domain also undergoes noticeable structural changes as each C␣ atom in the region Ala 80 -Lys 120 moves by Ͼ1 Å. Superimpositions were done using the program Align (33). The palm domain (residues Gly 188 -Asp 225 and Thr 287 -Val 370 ), which includes catalytic residues (Asp 220 and Asp318), maintains the same geometry in both molecules. The root mean square deviation of the palm domain between the two molecular forms is 0.26 Å. The conformations of the structural motifs A-E (motif A, residues Asp 213 -Glu 230 ; motif B, residues Arg 277 -Cys 303 ; motif C, residues Val 309 -Ser 326 ; motif D, residues Leu 336 -Asp 352 ; and motif E, residues Leu 362 -Pro 376 ) remain essentially the same between these two molecules.
The largest structural changes that are observed are in the thumb domain; the main difference between form I and form II is a rigid body rotation of 7.5°of the thumb domain relative to the fingers and palm domains. It is known that the thumb domain moves during the formation of the ternary complex in human immunodeficiency virus type 1 reverse transcriptase (17) and also between the two independent molecules in the asymmetric unit, representing closed and open conformations of the RNA-dependent RNA polymerase of the rabbit hemorrhagic virus (15). In addition, a part of the thumb domain comprising residues Leu 443 -Val 454 , the ␤-flap region, moves as shown in Fig. 4. This region has been proposed to move and/or interact with RNA during elongation (8 -10). This finding demonstrates that HCV polymerase genotype 2a can have an en-semble of conformations. The solvent-accessible surface area of the molecule in form II (Table IV) is 6% greater than that in form I, which further suggests that the molecule in form II is relatively open.
On the basis of the observed conformational variability in the thumb domain, the interface between the fingers and thumb domains, and the changes in the ␤-flap region, we provide structural evidence of the existence of closed and open conformations of the NS5B HCV polymerase genotype 2a. The molecule in crystal form I is in the closed (active) conformation, and that in crystal form II is in the open (inactive) conformation. This is the first structural evidence of the existence of an open conformation of NS5B HCV polymerase genotype 2a. More importantly, the conformations of the NS5B-inhibitor-bound structures, described later, resemble the conformation of the molecule in crystal form II. Hence, this finding would indicate that the molecule in crystal form II likely resembles the inactive conformation of the polymerase, although the conformation of catalytic aspartic residues in both forms remains the same.
Comparison of HCV NS5B Polymerases of Genotypes 2a and 1b-To date, crystal structure studies of HCV polymerase genotype 1b have revealed the molecular structure in several crystal forms (8 -10). The structure determination of the HCV polymerase of genotype 1b led to the understanding of the overall three-dimensional structure of the enzyme and the architecture of its active site. Unlike other polymerase structures  determined by x-ray diffraction methods, the active site of HCV NS5B polymerase genotype 1b is completely encircled (8). Although the overall structure of the HCV polymerase genotype 2a from both crystal forms is similar to the structure of HCV polymerase genotype 1b, consistent with the high amino acids sequence identity of 75% (Fig. 5) over the entire polypeptide   chain, there are marked structural differences between the structures of the polymerases from these two genotypes. To understand the three-dimensional structure of HCV polymerase and its variability among different genotypes or in the same genotype in a different crystal environment, we have analyzed the available structures as listed in Table IV.
To elucidate the structural differences, the C␣ atoms of pairs of molecules were superimposed. The resulting root mean square deviations are listed in Table V. From these results, it is apparent that, the structure of polymerase of genotype 2a in form I crystal is similar to the HCV polymerase genotype 1b structures. However, the molecule in crystal form II shows greater variation from other 1b polymerase structures and also from the molecule in crystal form I. Among the three domains, the thumb domain exhibits the greatest variation. The fingers domain, however, agrees much better. The palm domain preserves a relatively rigid structure across all of the molecules.
Effect of Non-nucleoside Inhibitors on NS5B Activity-Both thiophene 2-carboxylic acid inhibitors, namely compounds A and B (Fig. 1), were tested for anti-HCV polymerase genotype 2a activity using the C-terminal truncated form (⌬21) of the enzyme. Both compound A and B were found to be active against polymerase 2a in a dose-dependent manner with IC 50 values of 4.4 and 8.0 M, respectively (Fig. 6).
Inhibitor Binding Site and NS5B-Inhibitor Interaction-Thiophene-2-carboxylic acids A and B, found previously to be inhibitors of HCV polymerase genotype 1b, were also found to inhibit polymerase genotype 2a (⌬21 C-terminal truncated). Our observations from soaking experiments suggest that both inhibitors can only bind to crystal form I, because they cannot be soaked into the enzyme that crystallizes in form II. Unexpectedly, our analysis of the inhibitor/polymerase complex revealed that the enzyme has now adopted the form II crystal form (Table I). Based upon our crystallization experiments, for a given batch of protein purification only one of the two crystal forms can be obtained. Relatively few form I crystals were seen, whereas form II produced a large number of needle-shaped crystals.
As in other known polymerases, the HCV NS5B polymerase active site is situated in the palm domain. Two conserved aspartic acid residues (Asp 220 and Asp 318 ) located in the palm domain along with two Mg 2ϩ ions are essential for the polymerization reaction. Both inhibitors bind NS5B molecule in a shallow cavity on thumb domain (Fig. 7), and the inhibitor binding site is ϳ35 Å away from the polymerase active site. The simulated annealing omit electron density maps clearly revealed the orientation and conformation of all substituents of both inhibitors (Fig. 8, a and b).  Fig. 9, a and b.
In an attempt to provide a rationale as to why the NS5B molecule in form II does not allow inhibitor complex formation, the inhibitors were docked to the form II molecular structures, and we observed that the generated NS5B-inhibitor interactions were virtually identical to the one obtained experimentally with form I NS5B. The only visible difference was the length of the hydrogen-bonding distance between the two carboxylate oxygen atoms of the inhibitor and the backbone amide nitrogens of the residues Thr 476 and Tyr 477 . The hydrogen bonding distance between the inhibitor's carboxylate O22 atom and the main chain amide nitrogen atom of residue Thr 476 is 2.7 Å in the inhibitor bound structure, whereas in the inhibitordocked structure the corresponding distance is 2.2 Å. A similar situation was observed for the other hydrogen bond between the inhibitor's carboxylate O21 atom and Tyr 477 . It is therefore possible that the increased steric hindrance precludes inhibitors from binding. The other possibility could be that as the thumb domain of the molecule in form II moves by 7.5°with respect to the thumb domain of the molecule in form I, the conformation of the former may not be conducive for the inhibitors to be bound.
Conformational Changes upon Inhibitor Binding and Plausible Mechanisms of Inhibition-As mentioned earlier, only the NS5B polymerase in the closed conformation (form I), which is similar to that of the polymerase of genotype 1b conformation, binds the inhibitor. Upon inhibitor binding, NS5B undergoes major conformational changes as shown in Fig. 10. Three regions of the molecule undergo major structural changes. The thumb domain moves ϳ7.5°relative to the fingers and palm domains, the fingers-thumb domains interface, and the ␤-flap region. The nature of the conformational changes resulting upon inhibitor binding is similar for both inhibitors (Fig. 10). The inhibitor-bound structures of NS5B are very similar to the unbound structure of NS5B molecule in form II. As presented in Table III, the average B-factors of the protein atoms of the inhibitor-bound structures are larger than those of the native structures, indicating that, the former structures are relatively flexible.
On the basis of the major conformational changes in the NS5B molecule observed upon inhibitor binding, we propose the following mechanisms of inhibition. Upon inhibitor binding, the ⌬1 loop, an extension from the fingers domain, moves away from the thumb domain to reduce the inter-fingersthumb domain interactions, thereby resulting in a perturbation of the integrity of the structure. Part of the ⌬1 loop (residues Asn 24 -Leu 31 ), which adopts a helical conformation in the native structure (form I), changes to a small ␤-hairpin-like structure in the inhibitor-bound structures. It is therefore possible that inhibitor binding triggers the unwinding of this helix, thus inducing the 7.5 o shift in the thumb domain relative to the fingers and palm domains. In the presence of an inhibitor, the polymerase is "locked" into form II and is incapable of polymerization or reverting back to the active form I. Recent studies on HCV polymerase genotype 1b have provided some insight into the importance of the ⌬1 loop in coordinating the motion between the thumb and finger domains, hence giving rise to the "closed" or "open" conformation. Labonté et al. have shown by analytical ultracentrifugation experiments that substitution of Leu 30 by polar serine or arginine results in a nonfunctional polymerase due to a local perturbation in the ⌬1 loop that impairs the ability of the thumb domain to assume the closed conformation (32). Furthermore, this displacement of the ⌬1 loop may also impair its ability to bind the allosteric modulator rGTP at the outer thumb region. The ability of the polymerase to oligomerize to the functional enzyme may also be prevented, as one of the key amino acids involved in that process (Glu 18 ) forms part of the ⌬1 loop. Our studies therefore provide, in addition to the mode of action of the thiophene-2carboxylic acid inhibitors, structural evidence that the ⌬1 loop is indeed responsible for determining the active state of HCV polymerase genotype 2a and, by analogy, of other RNA-depend- FIG. 8. Stereo views of simulated annealed omit ͦF o ͦ ؊ ͦF c ͦ electron density maps contoured at the 3 level for the inhibitor A complex (a) and the inhibitor B complex (b). The final refined inhibitor models are superimposed on the electron density maps. Inhibitors are shown as stick models with carbon, nitrogen, oxygen, sulfur, and chloride atoms in green, blue, red, orange, and magenta, respectively. The inhibitor-binding site is shown by surface representation with carbon, nitrogen, and oxygen in gray, blue, and red, respectively. The figure was prepared by the program PyMol (pymol.sourceforge.net) (35).

FIG. 9. Stereo views of the NS5B polymerase-inhibitor A (a) and NS5B polymerase-inhibitor B (b) interactions.
The carbon, nitrogen, oxygen, and sulfur atoms of the protein are shown in gray, blue, red, and yellow, respectively. The color codes used to represent the inhibitors are same as those described in the Fig. 8 legend. Hydrogen bonding and van der Waals interactions are shown in green and blue, respectively. Single letter amino acid abbreviations are used with position numbers. ent RNA polymerases, including the polymerase of HCV genotype 1b. Although these experiments have provided structural evidence on the mechanism of action of the thiophen-2-caroxylic acid inhibitors, further investigations are needed to assess the influence of RNA templates, substrate nucleotides, and the allosteric rGTP on the inhibition mechanism.
Structure-based mutations of residue Leu 30 to either serine or arginine reduces the activity of the polymerase (32). Hence, the ⌬1 loop and thumb domain interface is critical for the polymerase activity. Perturbation of this region would ultimately affect the activity. Second, the substantial movement of the thumb domain (7.5°) relative to the fingers and palm domains upon inhibitor binding may inhibit the function of polymerase, as it is known that the thumb domain moves by a similar magnitude between the proposed active and inactive structures of the RNA-dependent RNA polymerase of rabbit hemorrhagic virus (15).
In summary, the structures of two crystal forms of HCV polymerase genotype 2a have been determined. These two forms correspond to a closed and an open conformation of the NS5B polymerase. Structure analysis has provided insights into our understanding of the structural variability among different genotypes and different crystal environments of the same genotype. Enzyme-inhibitor complexes could only be generated with the crystal form I, which is the closed form and is believed to be the active entity. The presence of the inhibitor was found to induce conformational changes that result in the open or inactive form. FIG. 10. Stereo view of conformational changes that occur upon inhibitor binding. Inhibitor bound and unbound structures are superimposed. Because similar conformational changes were observed in the case of both inhibitors, only that induced by inhibitor A is shown. Green and red colors correspond to the inhibitor bound and unbound structures, respectively.