Functional Analysis of RNA Binding by the Hepatitis C Virus RNA-dependent RNA Polymerase*

Protein-RNA interaction plays a critical role in regulating RNA synthesis by the hepatitis C virus (HCV) RNA-dependent RNA polymerase (RdRp). RNAs of 7 nucleotides (nt) or longer had affinities 5-fold better than an RNA of 5 nt, suggesting a minimal length required for binding. To identify RNA contact sites on the HCV RdRp, a biotinylated 7-nt RNA capable of directing de novo initiation was used in a process that coupled reversible formaldehyde cross-linking, RNA affinity chromatography, and mass spectrometry. By this process, we identified 18 peptides cross-linked to the 7-nt RNA. When these identified peptides were overlaid on the three-dimensional structures of NS5B, most mapped to the fingers subdomain, connecting loops between fingers and thumb subdomains and in the putative RNA binding channel. Two of the identified peptides resided in the active site cavity of the RdRp. Recombinant HCV RdRp with single residue changes in likely RNA contact sites were generated and characterized for effects on HCV RdRp activity. Mutant proteins had significant effects on cross-linking to 7-nt RNA and reduced RNA synthesis in vitro by 2- to 20-fold compared with wild type protein. When the mutations were tested for the replication of HCV RNA in the context of the cells transfected with the HCV subgenomic replicon, all except one prevented colony formation, indicating a defect in HCV RNA replication. These biochemical and functional analyses identified a number of residues in the HCV RdRp that are important for HCV RNA synthesis.

Approximately 3% of the world's population is chronically infected with the hepatitis C virus (HCV), 2 and a significant percentage of these individuals will progress to life-threatening liver cirrhosis and hepatocellular carcinomas (1,2). Current treatments for chronic HCV infection rely on the use of pegylated interferons (IFN-␣) in combination with the broad spectrum antiviral agent, ribavirin (3). This combination therapy has a positive response in ϳ50% of the patients infected with genotype 1 of HCV, the most prevalent form in North America and Europe (3,4). The lack of a response from the remaining genotype 1-infected individuals and the fact that some patients suffer severe side effects from the existing treatment (4) point to the need to develop additional treatment options. Improved drug development will benefit from a better understanding of the replication of HCV, the central enzyme of which is the nonstructural (NS) protein 5B, an RNA-dependent RNA polymerase (RdRp) (5).
A number of structures of the HCV NS5B have improved our understanding of HCV RNA-dependent RNA synthesis. The structures of the apoenzyme were determined by three independent groups (6 -8) and revealed a number of novel properties. Similar to other template-dependent polymerases, the HCV RdRp resembles a right hand and contains fingers, palm, and thumb subdomains. Unlike the more open structures of other template-dependent DNA polymerases such as the Klenow Fragment and the human immunodeficiency virus 1 reverse transcriptase, the HCV RdRp has a fully encircled active site through extensive interactions between the fingers and thumb subdomains, resulting in a protein that predominantly exists in a "closed" conformation (6 -8). A highly similar closed conformation is observed for the polymerase of the bacteriophage 6 (9). HCV RdRp also has an unusual ␤-hairpin loop that protrudes into the active site that helps position the 3Ј-end of the RNA template for proper initiation of RNA synthesis and decreases the ability to extend from a primed template (10,11). The C-terminal tail also lies in the catalytic pocket and modulates RNA synthesis by the HCV RdRp (11,12). The crystal structures of HCV RdRp complexed with RNA or NTP are also available. In these structures, the putative RNA binding channel, which would be formed between fingers and thumb subdomains, was predicted and several RNA contacting residues were determined, thus allowing models of the ternary complex to be built (7,13,14).
Despite the elucidation of a number of structures of HCV RdRp, our understanding of the mechanism of RNA-dependent RNA synthesis by the HCV RdRp remains incomplete. One area requiring additional characterization concerns how the HCV RdRp interacts with the RNA template to regulate RNA synthesis. Several lines of evidence suggest that the interaction between HCV RdRp and RNA could be more dynamic than was shown in the crystal structure. For example, it was recently demonstrated that the RdRp from an HCV 2a isolate could exist in two distinct conformations (15). Also, we have evidence that the HCV RdRp could alter its conformation to bind circular RNA. 3 Lastly, other polymerases, such as DNA-dependent RNA polymerases, are known to change their interactions with the template during different steps in transcription (17)(18)(19).
In this work, we have characterized the binding of different lengths of RNAs to the HCV RdRp in solution to determine the minimal length of RNA for stable binding of the HCV RdRp. We employed reversible formaldehyde cross-linking coupled with RNA affinity chromatography and mass spectrometry (MS) to identify RNA contact sites on the HCV RdRp. Mutational analysis of the HCV RdRp at several residues presumed to be important in making contact with RNA had negative effects on RNA synthesis in vitro. In addition, all but one mutation prevented the replication of an HCV subgenomic replicon in Huh7 cells, indicating that these residues are important for replication.
Expression and Purification of WT and Mutant HCV NS5B ⌬21 Proteins-The HCV NS5B from the type 1b BK strain was expressed without the hydrophobic C-terminal 21 residues in Escherichia coli BL21 (DE3) using the plasmid pET21b. The protein contained a C-terminal hexahistidine tag to allow immobilized metal affinity purification as previously described (27). The protein was further purified by poly(U)-Sepharose 4B (Amersham Biosciences) ion exchange column. After washing with 10 column volumes of binding buffer A (20 mM Tris (pH 7.6), 150 mM NaCl, 10% glycerol, and 3 mM dithiothreitol), bound protein was eluted with elution buffer (buffer A ϩ 250 mM NaCl). The purified protein was concentrated using a spin concentrator (30 K MWCO; Pall), and the purity of the protein as judged by SDS-PAGE was Ͼ95%.
Site-directed mutagenesis was carried out using the QuikChange kit (Stratagene) according to the manufacturer's protocol. The entire cDNA clone was then sequenced to confirm that the specified mutation was made and no additional changes had inadvertently taken place.
Fluorescence Spectroscopy-Fluorescence measurements were made at 22-23°C with a PerkinElmer luminescence spectrometer LS55 and cuvette with an optical path length of 0.4 cm. Equilibrium anisotropy measurements were taken of fluorescein-labeled RNAs (F-RNAs) in binding buffer (50 mM Hepes (pH 7.5), 5 mM MgCl 2 , 0.002% Tween 20, and varying concentrations of NaCl) with excitation and emission wavelengths of 490 and 520 nm, respectively. The integration time of 1 s and a slit width of 5 nm were used throughout the measurements. Titrations were performed by successive addition of ⌬21 with constant stirring. Anisotropy values were recorded from eight measurements after each sample had equilibrated for 60 s. Throughout the titration, no significant changes in quantum yield of fluorescein-labeled RNAs were observed.
Binding data were analyzed by non-linear least square fitting using KaleidaGraph software (Synergy software, Reading, PA). The equilibrium dissociation constants (K d ) were determined by nonlinear regression analysis using a Hill equation (⌬A ϭ B max x n /[x n ϩ K d n ]) as a binding model. In this equation, ⌬A is the value of anisotropy change by the ligand binding, B max is the value of maximum anisotropy change, x is the total concentration of the input ligand, and the exponential term (n) is the Hill coefficient, which can be used to estimate the degree of cooperativity in the binding of protein and RNA.
Reversible Cross-linking of ⌬21 and RNA, Affinity Purification of Cross-linked RNA Peptides, and Retrieval of Cross-linked Peptides-For analytical cross-linking of protein and RNA, 1 M each of ⌬21 and 5Ј-radiolabeled 7-nt RNA (3Ј-CAUAUGC-P*-5Ј, named 7CP) were incubated in 20 mM Hepes (pH 7.5), 4 mM MgCl 2 , 1 mM dithiothreitol at room temperature for 5 min by the addition of a final concentration of 0.1% formaldehyde in a 20-l reaction. The cross-linking reaction was quenched by the addition of final concentration of 0.2 M glycine, and aliquots (10 l) of the reaction products were mixed with SDS sample buffer and loaded onto 4 -12% gradient NuPAGE gel (Invitrogen). Quantification of the cross-linked products used a PhosphorImager and Amersham Biosciences software.
Preparative cross-linking was performed in a 100-l reaction with a final concentration of 2 M ⌬21 and 4 M 7CB RNA in 20 mM Hepes (pH 7.5), 4 mM MgCl 2 , 1 mM dithiothreitol. Following the addition of formaldehyde to 0.1% final concentration, the reaction was incubated at room temperature for 5 min before the addition of 0.2 M glycine to quench the reaction. Subsequent trypsin treatment used sequencing grade trypsin (Trypsin Gold; Promega, Madison, WI). The reaction contained a protease:substrate ratio 1:50 (w/w) and was performed in 100 mM NH 4 HCO 3 (pH 7.8) at 37°C overnight.
Streptavidin magnetic beads (New England Biolab, Beverly, MA) were used to capture the biotinylated RNA and RNA-peptide conjugate. The trypsin-digested, cross-linked reaction products were added to 50 l of beads washed previously and equilibrated into 20 mM Hepes (pH 7.5) buffer. The samples were washed three times with 20 mM Hepes, 1 M NaCl, 1 mM EDTA, 1 mM dithiothreitol buffer and twice with 25 mM ammonium bicarbonate (pH 7.8). The RNA-peptide conjugates were reversed by incubating the samples at 70°C for 1 h. The samples were then centrifuged at 3000 ϫ g for 3 min, and the supernatants containing the peptides were desalted and concentrated using a Ziptip (Millipore, Bedford, MA). The peptide samples were eluted in 2.5 l of 70% acetonitrile/0.1% trifluoroacetic acid solution and sent for MS analyses.
Mass Spectrometry-All MALDI MS analyses were performed on an ABI 4700 proteomics analyzer (Applied Biosystems, Framingham, MA). The instrument is equipped with a 200 Hz Nd:YAG laser (PowerChip; JDS Uniphase, San Jose, CA) and controlled by the Applied Biosystems 4000 series Explorer V3.0 software package. All MS analyses were performed using the dried droplet method and 5 mg/ml ␣-cyano-4-hydroxycinnamic acid (Sigma) in 60% acetonitrile as the matrix. The laser intensity was set just above the threshold required to ionize the peptides. For the tandem MS experiments, the acceleration was 1 kV in all cases, the collision gas was air, and the laser intensity was increased by 10% over the MS mode experiment. The number of laser shots used to obtain a spectrum varied from 500 -5000, depending on signal quality. The fragmentation data obtained in these experiments were handled using the Applied Biosystems Data Explorer software package. In some cases internal calibrants were used to maximize mass accuracy in MS mode (Ϯ10 ppm). Accuracy in MS/MS mode is routinely better than 0.2 Da. Peaks selected had a signal to noise ratio greater than 5 and the appropriate carbon and hydrogen isotope ratios.
RdRp Activity Assay-Standard in vitro RdRp assay mixtures consisted of 0.125 M template with 0.08 M ⌬21 or its derivatives in a 20-l reaction mixture containing 20 mM sodium glutamate (pH 8.2), 12 mM dithiothreitol, 4 mM MgCl 2 , 0.5% (v/v) Triton X-100, 1 mM MnCl 2 , 200 M GTP, 200 M ATP, 200 M UTP, and 250 nM [␣-32 P]CTP (400 Ci/mmol, 10 mCi/ml; Amersham Biosciences). MnCl 2 was used to increase the level of RNA synthesis as described in a previous study (27). The reaction mixtures were incubated for 60 min at 25°C, and the reactions were stopped by phenol-chloroform (1:1 v/v) extraction. The products were precipitated in 6 volumes of ethanol, 5 g of glycogen, and 0.4 M ammonium acetate. Products were separated by electrophoresis on denaturing 7.5 M urea-20% polyacrylamide gels. Gels were wrapped in plastic, and quantification of radiolabeled bands was performed using a PhosphorImager (Amersham Biosciences).
HCV Subgenomic RNA Replicon Assay-A subgenomic HCV replicon cDNA (pFK/I 389 /NS3-3Ј/wt) (28) was used as a template for mutational analysis. To insert a point mutation in NS5B in this subgenomic HCV replicon, BglII fragments of mutant clones of pET21bNS5B were exchanged with those of the wild type subgenomic HCV replicon. Each mutant HCV subgenomic replicon was confirmed by DNA sequencing.
Subgenomic HCV replicon RNAs were transcribed in vitro using a T7 Ampliscibe kit (Epicenter, Madison, WI) from the WT and mutant DNA constructs after linearizing with ScaI. The subgenomic HCV replicon RNAs were transfected into Huh7 cells by electroporation with a GenePulser system (Bio-Rad, Hercules, CA). Briefly, 0.2 g of the in vitro transcribed RNAs along with total RNA from naïve Huh7 cells to a final amount of 10 g were electroporated into ϳ1 ϫ 10 6 Huh7 cells in 0.4 ml of ice-cold Cytomix (29). Electroporation conditions were 960 F and 270 V. Cells were immediately transferred to 4 ml of complete Dulbecco's modified Eagle's medium containing 1.25% Me 2 SO and seeded in a 5-cm dish. At 24 h after transfection, the cell culture medium was replaced with complete Dulbecco's modified Eagle's medium supplemented with 0.5 mg/ml of G418. The medium was changed every week. After 3 weeks of selection with G418 sulfate, cell colonies in culture dishes were stained with 0.01% Coomassie Blue.

RESULTS
Minimal RNA Length for Stable Binding of ⌬21-We wanted to determine the minimal length of RNA required for stable binding of the ⌬21. Therefore, we prepared chemically synthesized RNAs of 5, 7, 9, 11, and 14 nucleotides that all have the same initiation cytidylate (ϩ1C) and a 5Ј-fluorescein tag (Fig. 1A). Fluorescence anisotropy changes of these RNAs by ⌬21 binding were measured for each F-RNA, and the K d s were derived by fitting the binding isotherms into the Hill equation (Fig. 1A). Representative binding isotherms and curve fittings are presented in Fig. 1, B and C. At 50 mM NaCl, the K d of ⌬21 to the RNAs of F-14, F-11, F-9, and F-7 were all between 0.29 and 0.42 M. However, the K d of ⌬21 to the F-5 was ϳ5-fold higher ( Fig. 1), suggesting that a 7-nt RNA is minimally required for efficient binding under our conditions.
To address whether the binding of ⌬21 to RNAs of different lengths was qualitatively different, we examined the K d s for the F-5, F-7, and F-14 in the presence of NaCl concentrations up to 125 mM. The K d of F-5 increased from 3.5 to 12.1 M (5.8-fold) when NaCl was at 125 mM; the K d s with RNAs of 7 and 14 nt increased by 3.8-and 2.3-fold, respectively (Fig. 1A). These results indicate that nonspecific ionic interaction contributed more to the binding affinity with the 5-nt RNA than to longer RNAs. Therefore we used a 7-nt RNA for subsequent investigations of HCV RdRp-RNA interaction.
Cross-linking of the ⌬21 to the 7-nt RNA (7CB) using Formaldehyde-The overall scheme used to identify the RNA contact sites in ⌬21 is shown in Fig. 2A. The key procedure in this analysis is the reversible formaldehyde cross-linking of ⌬21 to a biotinylated RNA of 7 nt. Reversible formaldehyde cross-linking is widely used to study protein-DNA and protein-RNA interactions, especially in vivo (21,22). Crosslinking of RNA to protein using formaldehyde occurs through the linkage between the side-chains of Lys, Arg, His, Cys, aromatic residues, and bases (Ala, Gly, and Cys) of RNA by the formation of methylene bridges via dehydration and Schiff base formation (23). These cross-links can be reversed by rehydration and heating the cross-linked sample to release formaldehyde (21). After this formaldehyde cross-linking of ⌬21 to a biotinylated RNA of 7 nt, we treated the protein-RNA complex with trypsin to generate peptide-RNA cross-links, which were then affinity purified with streptavidin, followed by reversing the cross-links by heating the cross-linked sample to release formaldehyde. The purified peptides were analyzed by MS.
Prior to analyzing RNA interaction with peptides from ⌬21 in a preparative scale, we tested whether a minimal length of 7-nt RNA, named 7CB, chemically synthesized with a 5Ј-biotin and spacer atom between 5Ј-nucleotide and biotin group (Fig. 2B), could direct RNA synthesis by the HCV RdRp in vitro. 7CB directed the synthesis of 7-nt RNA and longer template switch products at a level at least as well as the unmodified 7C RNA (Fig. 2C), indicating that 7CB is competent to direct de FIGURE 1. Binding of ⌬21 to RNAs of different lengths. A, the names and sequences of the RNA and summary of the analyzed K d , R 2 values, and Hill coefficient (n) values. The affinities for interaction with ⌬21 with all of the F-RNAs (F-5, F-7, F-9, F-11, and F- 14) were all determined at 50 mM NaCl. Selected ones were analyzed at increasingly higher concentrations to determine whether the differences in affinities could be affected. The anisotropy change of F-RNA-⌬21 complex was plotted as a function of ⌬21 concentration, and the K d and Hill coefficients (n) were derived by fitting the binding isotherm into the Hill equation. B, the representative binding isotherm of F-7 and ⌬21 at 50 mM NaCl concentration. C, the representative binding isotherm of F-5 and ⌬21 at 50 mM NaCl concentration. NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45 novo initiation. Next, we examined the cross-linking conditions in analytical scale using radiolabeled 7-nt RNA (7C). Several different concentrations of formaldehyde needed to cross-link a 7-nt RNA to ⌬21 were tested, and a 0.1% solution of formaldehyde was found to be optimal to produce cross-linked products between 7C and ⌬21 (data not shown). This concentration was used in subsequent studies. Under these conditions, the yield of protein-RNA conjugate was estimated to be ϳ1% by phosphorimage analysis of 32 P-labeled RNA 7C (Fig. 2D). A small amount of dimer-sized cross-linked product was observed, suggesting that protein dimers were also cross-linked to the 7-nt RNA. This RNAinduced HCV RdRp oligomerization was also observed in a previous study (24).

HCV RdRp-RNA Interactions
Identification of Isolated Peptides from ⌬21 using MALDI-TOF MS, Peptide Mass Fingerprinting, and MS/MS-The MS spectrum of peptides bound to 7CB RNA is shown in Fig. 3A. The experiment was repeated three times, and identical results were observed each time. As a control, a sample was treated identically to the cross-linked sample with the exception that no formaldehyde was used (Fig. 3B). Peptides unique to the cross-linked sample are indicated with an asterisk (Fig. 3A). Each of the identified peptide peaks showed unique isotope distributions expected of peptides (data not shown). A number of low intensity peptide peaks that were not labeled in Fig. 3B were also observed. These corresponded to derivatives of higher intensity peptide peaks except that they had missed one or two potential trypsin cleavage sites (TABLE ONE). The presence of overlapping peptides was actually helpful in confirming the assignment of the peaks. Despite the stringent washing steps included in the protocol, several peptides were observed in the control sample, although in low abundance compared with corresponding peaks from the cross-linked sample, indicating that some uncross-linked peptides were not completely eliminated by this highly sensitive analysis. There also were low intensity peaks in both cross-linked and control samples that could not be assigned. These peaks could be attributed to several factors, including the matrix, multiply cross-linked peptides, other possible modified peptides, or peptides that interacted with the streptavidin resin. Nonetheless, by comparing the peaks from the noncross-linked control sample, the unassignable peaks did not affect the identification of the cross-linked peptides from the spectra.
Tandem MS was used to confirm the identity of several of the peptides. Representative fragment ion spectra of the parent ions 1238.69 and 1128.67 are shown in Fig. 4. The expected b-and y-type daughter ions were detected and indicated in these spectra. We performed tandem MS for several additional ions but only obtained incomplete fragmentation with them (data not shown). Nonetheless, the partial sequences obtained were consistent with the sequences expected of the parent ions. The amino acid position, observed and calculated masses, and sequences of the identified peptides by peptide mass fingerprinting results are summarized in Fig. 5A.
Location of the Identified Peptides in ⌬21-When the identified peptides were located in the three-dimensional structures of HCV NS5B ⌬21 (Protein Data Bank code 1QUV, Ref. 6), an asymmetric distribution of the cross-linked peptides was observed, with no peptides from the thumb subdomain and only one from the palm subdomain of the HCV RdRp (Fig. 5B). Instead, the peptides mostly mapped in the fingers subdomain and the connecting loops between fingers and thumb subdomains. Overall, the cross-linked peptides are located along the putative RNA binding channel predicted from the crystal structure of Bressanelli et al. (7) (Fig. 5B). Two of the identified peptides (amino acids 142-154 colored green and amino acids 155-168 colored yellow in Fig. 5B) were mapped within the active site cavity. Notably, the Arg-158 in peptide 155-168 (colored yellow in Fig. 5B) was previously determined to interact with the ␣-phosphate of GTP and shown to be important for RdRp activity (13,25). We did not observe residues within the novel ␤-loop to be among the regions cross-linked to RNA.
Analyses of Mutant HCV RdRp for RNA Binding-The identified peptides might contain a large number of residues that could potentially interact with RNA, and it was not practical to analyze all of the possibilities. Therefore, we used the following four criteria to select a subset of amino acids for mutational analysis: 1) amino acids that are preferentially reactive to formaldehyde (23), 2) basic and aromatic residues that are more likely to interact with RNA, 3) the peptide having a relatively higher MS intensity in which the residue resides, and 4) proximity to the template channel and the active site cavity. Based on these criteria, 8 residues were selected for mutations, 5 that lie near the template channel (Lys-20, Lys-141, Phe-162, Arg-168, and Lys-172), and 3 that are within the active site cavity (Phe-145, Arg-154, and Arg-158). In each case, the original residue was mutated to an alanine (Fig. 6A). All mutant proteins were purified to a similar homogeneity and final concentration as WT ⌬21 (Fig. 6B).
Mutants K141A, F162A, and K172A were tested for effects on crosslinking to 7CB using MS/peptide mass fingerprinting analysis. The interpretation is somewhat complex, as detailed below. A mutation of a residue that is not a Lys or Arg to Ala will cause the cognate peptide mass to decrease in mass by the difference between the targeted residue and Ala. A mutation of a Lys or Arg to Ala will also affect the site for trypsin cleavage; the cleavage site will thus shift to the next available Lys or Arg, with a corresponding increase in the mass of the expected peptide. Lastly, additional partial cleavage product would also be seen. After analyzing the MS spectra of mutant proteins, we did not observe tryptic peptide peaks corresponding to the mass of the peptide that contained A, outline of our strategy to identify the RNA contact sites on HCV RdRp. B, the sequence of the biotinylated 7-nt RNA. C, a comparison of RNA synthesis by ⌬21 from 7CB, the 7-nt template used for RNA cross-linking studies, and 7C, the 7-nt template without biotin label. D, autoradiogram of an SDS-PAGE containing formaldehyde cross-linking of ⌬21 and 7-nt RNA (radiolabeled 7C). The crosslinking reactions were performed with 1 M ⌬21 and 7-nt RNA, respectively. The presence of 0.1% formaldehyde is indicated by the arrow, and the cross-linked products are highlighted by an asterisk. The lane identified with "" was a control in which no formaldehyde was added to the crosslinking reaction.
the mutated Ala residues, indicating that each amino acid change did alter the ability to form a cross-link to RNA (TABLE ONE). For example, mutant K141A should produce a peak of 3021.5 m/z (amino acids 125-151) if the change did not affect RNA binding and cross-linking. However, this peptide peak was not detected in MS spectra, and neither were any other peptides that could be derivatives of this peptide fragment due to missed cleavage by trypsin. The expected fully cleaved peptides from three mutant proteins, K141A, F162A, and K172A, are shown at the bottom panel of TABLE ONE. These results indicate that the single amino acid changes did affect cross-linking of the HCV RdRp to the 7-nt RNA.
Interestingly, we observed that some peptides that should not be directly affected by the amino acid substitution were also missing in the peptide mass fingerprint profile (TABLE ONE). For example, the peptide corresponding to amino acids 91-98 was not detected with the mutant K141A. In F162A, seven peptides, including overlapping ones (amino acids 57-72, 82-98, 80 -98, 91-98, 101-109, 101-114, and 255-270), were also not detected. This result suggests that RNA contact by one region of ⌬21 could influence RNA contact by another region of the protein. This more complex interaction would be difficult to predict in a crystal structure of the RdRp in complex with RNA.
We tested binding affinity of mutant proteins, including K141A, F162A, and K172A, in comparison to ⌬21 using fluorescence anisotropy of F-7. Mutant proteins, including K20A, F162A, and K172A, showed binding affinities from 530 to 730 nM in comparison with the 340 nM ⌬21. This Ͻ2-fold decrease in the affinity of the mutants for RNA suggests that there are multiple RNA binding residues within ⌬21 and no single residue is primarily responsible for contacting the template RNA. These results are also consistent with a previous study (26).

Effects of Mutations on RNA Synthesis in Vitro-
We also tested the mutant proteins for RNA synthesis in vitro using the biotinylated RNA 7CB as template. Mutant proteins K20A, K141A, F162A, R168A, K172A, F145A, R154A, and R158A all directed the de novo initiated 7-nt RNA product at less than 20% of the level seen with ⌬21. Several of the mutations, including K20A, K141A, F162A, K172A, R154A, and R158A, were also defective for nontemplated nucleotide addition, characterized by the presence of the 8-and 9-nt products (Fig. 7A). We previously determined that mutations in the HCV RdRp differentially affected de novo initiation and nontemplated nucleotide addition (25).
The panel of recombinant proteins was also tested for effects on RNA synthesis from a previously characterized 21-nt RNA template, LE21, which is capable of directing both de novo initiation and primer extension. The de novo initiated RNA product is 21 nt, whereas the primer extension product is 34 nt (27). Mutant proteins F162A, R168A, K172A, and R158A produced a severely reduced amount of the de novo and primer-extended RNA products in comparison to ⌬21 (Fig. 7A). Mutants K20A, K141A, F145A, and R154A yielded reduced but detectable amounts of RNA products in comparison with the WT protein (Fig.  7A). The quantitative results of de novo RNA synthesis are summarized in Fig. 7B. All of the mutations decreased RNA synthesis from both 7CB and LE21 to Ͻ59% of ⌬21, but the severity of the defect was dependent on the template used, possibly because of differences in the length or the structure of the two RNAs. In general, the defects were less severe with LE21 than with 7CB, suggesting that LE21 may allow the mutant polymerase to compensate better for a specific mutation. NS5B on RNA binding in a more biologically relevant context, the eight mutations were engineered into the replication-competent subgenomic HCV RNA. In vitro transcripts of the subgenomic replicon were used to transfect Huh7 cells by electroporation and selected for G418 resistance. Replication of the subgenomic HCV replicon RNA in Huh7 cells results in expression of neomycin phosphotransferase, which renders Huh7 cells resistant to the antibiotic G418 (28,29). Mutations K20A, K141A, F162A, R168A, K172A, F145A, R154A, and R158A were tested. The cells were stained with Coomassie Blue to visualize the G418-resistant colonies. Only the wild type subgenomic replicon and the one with the K20A mutation produced colonies (Fig. 8). The results were reproducible in two independent experiments, indicating that the a "V" indicates that the peptide was detected in the mass spectra. b The residues and the m/z ratio of the peptide fragment expected to contain the mutated residue in the three mutant proteins are shown. None of these peptides was observed in the respective spectra, indicating that the mutation affected cross-linking to RNA.

DISCUSSION
In this study, the minimal length of RNA required for stable binding of HCV RdRp was determined by fluorescence anisotropy analyses of different lengths of RNAs, and the HCV RdRp regions that could be cross-linked to the 7-nt RNA template were identified by reversible cross-linking and MS analyses. We also demonstrated that mutation of the recombinant HCV RdRp with one of the putative RNA-contacting residues within the identified peptides had significant effects on RNA cross-linking and severe negative effects on RNA synthesis in vitro. All but one of the mutations tested in the context of the HCV subgenomic replicon prevented replication of HCV RNA in Huh7 cells.
We have found that the minimal size of RNA template (7 nt) for the HCV RdRp determined in this study is quite comparable with results from previous biochemical studies of poliovirus 3D pol (30), the bovine viral diarrhea virus RdRp (31), and HCV RdRp (32)(33)(34). Presumably six to eight nucleotides of template RNA can interact with the RdRp and form a stable initiation complex.

Identification of RNA Contact Sites on the HCV RdRp-Reversible
cross-linking of protein to RNA using formaldehyde has key advantages over other methods such as UV cross-linking or irreversible chemical (e.g. psoralen) cross-linking, in that cross-linked peptides can be retrieved from the RNA-peptide conjugates and analyzed by standard MaLDl-TOF protocols. A caveat of this approach is that this highly sensitive detection of peptides by MS requires that controls be included to facilitate the assignment of the relevant peptide ions.
The location of the RNA binding peptides we mapped agrees well with the subdomains of the crystal structure seen to interact with RNA. Furthermore, most of the peptides we identified are located around the central template channel of the HCV RdRp, within the fingers subdomain and connecting loops between fingers and thumb subdomains in the three-dimensional structure of HCV NS5B. These identified contact sites include all of the predicted amino acid residues in the putative RNA binding channel from the crystal structure (7). For example, the predicted RNA binding residues in the crystal structure (HC-BK) were Cys-14, Lys-90, His-95, Lys-98, Lys-106, Arg-109, Lys-141, Phe-162, Arg-168, and Lys-172 (7). These predicted residues were all included in our identified peptides, 1-32, 82-98, 101-109, 121-141, and 159 -172. Furthermore, the predicted RNA binding residues in the crystal structure of a different strain of HCV (HC-J4) were Cys-14, Pro-93, His-95, Arg-97, Lys-98, Met-139, Lys-141 and Ile-160 (14). These predicted residues also were all included in our identified peptides, 1-32, 82-98, 121-141, and 159 -172. These data indicate the usefulness and validity of our approaches in this study to identify the RNA contact sites on HCV RdRp.
Our results also extend and complement the analysis of RNA binding by the HCV RdRp previously reported by three groups (26,35,36). The first group (26) used a filter binding assay to analyze a panel of recombinant HCV RdRp proteins with mutations in specified motifs for binding to homopolymeric RNAs or in vitro transcribed 3Ј-untranslated sequence of HCV genomic RNA. They found that mutations in conserved motifs A, B, C, and D did not obviously affect RNA binding. The second group (35) used a 15-nt uridylate sequence and targeted mutations in conserved RdRp motifs to demonstrate that RNA binding by the recombinant HCV RdRp was affected by a double mutation at Thr-287 and Asn-291 and by the deletion of 70 amino acids at the N terminus of the HCV NS5B protein. The third group (36) made a large panel of mutant proteins that are primarily centered around conserved motifs in the HCV RdRp and identified three stretches of amino acids in residues 149 -155, 220 -226, and 274 -280 that contributed to the binding of homoguanylates and homouridylates. Our strategy to identify RNA binding regions was performed without a priori assumptions of specific motifs being involved in RNA binding. Nonetheless, compared with the previous result (36), we have observed similar RNA binding regions such as the N-terminal portion of the HCV NS5B protein (residues 1-70) and the RNA binding channel region (residues 145-155). One discrepancy between the results of Qin et al. (36) and ours is that we did not detect the peptide containing residues 274 -280 in palm subdomains to be cross-linked to our RNA. This discrepancy may be a consequence of using a homouridylate rather than a heteropolymeric RNA. The lengths of the RNA used in the analysis, and possibly differences in the HCV strains used, may also be distinguishing factors. Another difference in our results is that we identified a number of additional regions not found by others. For instance, peptide residues 80 -98, 99 -109, 101-120, 121-141, 142-158, and 159 -172 were identified by this study. It is not surprising that these regions could contribute to RNA binding because they help form the RNA binding channel and part of the active site cavity in HCV NS5B (7,14).
Effects of Mutation in the RNA Binding Regions on Biochemical and Cell-based Assays-In addition to good agreement with the mapping of the RNA binding results from other groups, the putative binding sites we identified appear to be functionally relevant. Although only a limited   number of the residues that could contact RNA were examined, all 8 that we analyzed (Lys-20, Lys-141, Phe-162, Arg-168, Lys-172, Phe-145, Arg-154, and Arg-158) had reduced RNA synthesis in vitro from ϳ2to 20-fold. In addition, except for K20A, the same point mutants were unable to support HCV subgenomic RNA replication. The defects observed in vitro are thus generally amplified in the more complex in vivo setting. Currently we cannot adequately explain the discrepancy of results of K20A in vitro and in vivo. In the subgenomic replicon assay, K20A produced transductants at a level indistinguishable from wild type. We note that K20A was not predicted to contact RNA from the crystal structure of NS5B (7). All these in vitro and in vivo characterizations of mutations suggest that multiple RNA binding residues play a synergistic role in RNA binding and that single residues that make contact with RNA play a critical role in regulating catalytic activity of RdRp.
Potential Relevance to Polymerase Mechanism-RNA binding by the HCV RdRp could be compared with results from other template-dependent polymerases. A UV cross-linking study of T7 RNA polymerase and DNA promoter using psoralen revealed that the four identified cross-linked peptides reside on a template binding cleft composed of fingers, palm, and thumb domains, which matched well to the region from the crystallographic model (37). As previously determined, the different classes of polymerases may thus bind the template in a generally similar manner (38).
We note that identified RNA contact regions may only represent a subset of the RNA binding sites and some of the identified peptides may represent regions of HCV RdRp that have different functions, because contact between the HCV RdRp and the RNA could change when the polymerase exists in different steps in RNA-dependent RNA synthesis (17,18,39). For example, parts of the identified peptides might come from the putative nascent RNA exit tunnel. Other unidentified regions on the HCV RdRp in this study may be detected and mapped to different regions of HCV RdRp, such as the thumb subdomain or catalytic motifs in the palm subdomain, when tested in different cross-linking conditions of different RNAs and proteins.
Our MS analysis of some of the mutant proteins for cross-linking to the 7-nt RNA already suggests that RNA contact by one region of ⌬21 can affect RNA contact by another region of the protein (TABLE ONE), indicating flexible conformation of HCV RdRp. Recently, structural analyses of two different crystal forms of HCV RdRp from genotype 2a revealed two different conformational states, closed (active) and open (inactive) forms of NS5B (15). Such open and closed forms have been reported for rabbit hemorrhagic disease virus RdRp as well (40). In that report, the authors also postulated that a significant movement of the thumb subdomain regulates RdRp activity by converting it from an inactive to an active state. Furthermore, we have found that circular RNAs can serve as templates for the HCV RdRp for initiation of RNA synthesis and elongative synthesis of multimeric lengths of the product RNA. 3 All of these results suggest that the interaction between the HCV RdRp and the template RNA is a dynamic process that will be affected by flexible conformations of the protein and perhaps the properties of the RNA.