Hepatitis C Virus NS5B Polymerase Exhibits Distinct Nucleotide Requirements for Initiation and Elongation*

The hepatitis C virus (HCV) NS5B protein is an RNA-dependent RNA polymerase (RdRp) essential for replication of the viral RNA genome. Purified NS5B has been reported to exhibit multiple activities in vitro. Using a synthetic heteropolymeric RNA template with dideoxycytidine at its 3′-end, we examined de novo initiation and primer extension in a system devoid of self-priming and terminal nucleotide transferase activities. Products predominantly of template size and its multiples were detected. High concentrations of nucleoside triphosphates (Kappm ∼ 100–400 μm) corresponding to the first three incorporated nucleotides were found to be required for efficient de novo RNA synthesis. In the presence of initiating di- or trinucleotides, however, the amount of NTP needed to achieve maximal activity dropped 103- to 104-fold, revealing a much reduced nucleotide requirement for elongation (Kappm ∼ 0.03–0.09 μm). Accordingly, single round extension from an exogenous primer following preincubation of the enzyme with template and primer could also be supported by <0.1 μm levels of NTP. De novo synthesis at high NTP concentrations was shown to be preferred over primer extension. On a dideoxycytidine-blocked synthetic RNA template derived from the 3′-end of the HCV(–)UTR, the addition of the corresponding initiating trinucleotide also dramatically reduced the NTP levels needed to achieve efficient RNA synthesis. Thus, distinct nucleotide requirements exist for initiation and elongation steps catalyzed by the HCV NS5B polymerase.

The hepatitis C virus (HCV) NS5B protein is an RNA-dependent RNA polymerase (RdRp) essential for replication of the viral RNA genome. Purified NS5B has been reported to exhibit multiple activities in vitro. Using a synthetic heteropolymeric RNA template with dideoxycytidine at its 3-end, we examined de novo initiation and primer extension in a system devoid of selfpriming and terminal nucleotide transferase activities. Products predominantly of template size and its multiples were detected. High concentrations of nucleoside triphosphates (K m app ϳ 100 -400 M) corresponding to the first three incorporated nucleotides were found to be required for efficient de novo RNA synthesis. In the presence of initiating di-or trinucleotides, however, the amount of NTP needed to achieve maximal activity dropped 10 3 -to 10 4 -fold, revealing a much reduced nucleotide requirement for elongation (K m app ϳ 0.03-0.09 M). Accordingly, single round extension from an exogenous primer following preincubation of the enzyme with template and primer could also be supported by <0.1 M levels of NTP. De novo synthesis at high NTP concentrations was shown to be preferred over primer extension. On a dideoxycytidine-blocked synthetic RNA template derived from the 3-end of the HCV(؊)UTR, the addition of the corresponding initiating trinucleotide also dramatically reduced the NTP levels needed to achieve efficient RNA synthesis. Thus, distinct nucleotide requirements exist for initiation and elongation steps catalyzed by the HCV NS5B polymerase.
Hepatitis C virus (HCV), 4 a member of the Flaviviridae family, is the causative infectious agent of non-A non-B hepatitis affecting an estimated 3% of the world's population. About 80% of the infected people develop chronic hepatitis, which can further progress to liver cirrhosis and hepatocellular carcinoma (1). The best current treatment using interferon-␣ combined with the nucleoside analog ribavirin controls the disease in only a fraction of the patients infected with genotype 1 virus (2). Without a prophylactic vaccine for HCV, there is a critical need to develop new antiviral strategies. The ϳ9.5-kb-long positive strand RNA genome of HCV encodes a single polyprotein, which is processed by viral and host proteases into at least 10 individual structural and nonstructural proteins (3). NS5B (nonstructural protein 5B) possesses RNA-dependent RNA polymerase (RdRp) activity essential for viral RNA replication (4) and has therefore become an intensely pursued target for anti-HCV therapy (5,6).
Although cell culture-based infectious HCV particle systems recently became available (7)(8)(9), our understanding of HCV replication has been gained primarily from studies of cell-based subgenomic replicons (10,11), cell-free replicase complexes isolated from replicon-harboring cells (12)(13)(14), and from biochemical analyses of recombinant NS5B protein expressed and purified from bacteria or insect cells (15,16). The availability of purified NS5B has also permitted its structural determination by x-ray crystallography (17,18). Full-length NS5B is composed of 591 amino acids, the last 21 at the C terminus functioning as a cellular membrane anchor. Both full-length and various C-terminally truncated forms of NS5B have been purified, the former requiring detergent solubilization, and shown to be enzymatically active in vitro (16,19,20). Those with C-terminal truncations extending beyond the last 21 amino acids, such as NS5B-⌬55, exhibit enhanced RdRp activity but may have lost domains important for regulating initiation of RNA synthesis (21). The soluble 21-amino acid truncated form of the enzyme, NS5B-⌬21, has thus emerged as a preferred surrogate for the viral polymerase in in vitro studies.
Purified HCV NS5B has been shown to catalyze RNA synthesis from a wide range of homo-and heteropolymeric templates without a discernable preference for HCV viral RNA (16,19,20). A non-template-dependent terminal nucleotide transferase (TNTase) activity has also been described (15,22). RNA replication has been variously reported to occur through primer-dependent elongation, using either a self-priming ("copyback") mechanism in which the 3Ј terminus of the RNA template folds back intramolecularly (23) or exogenous primers (24), as well as through a de novo initiation mechanism (23)(24)(25)(26)(27), which is likely to be the mode of viral replication in vivo. Under certain de novo synthesis conditions, internal initiation has also been observed in vitro (28). A ␤-hairpin loop protruding into the active site of the enzyme has been implicated as part of the structural basis favoring de novo initiation from the 3Ј terminus of single-stranded RNA templates (29 -31). Such templates, especially those containing a 3Ј-terminal cytidylate moiety, have been found to direct de novo synthesis efficiently (24,25,28). Evidence has been presented also for utilization of di-or trinucleotides as short primers to initiate replication, in a pro-posed scenario in which the enzyme, template, and incoming nucleotides form an initiation complex that frequently releases abortive dinucleotide products before progressing to a processive elongation mode (24,28,29).
Although the precise conditions that regulate the transition from de novo initiation to chain elongation remain unresolved, one critical factor appears to be the nucleotide concentrations that support enzymatic activity. Studies of RdRps, including HCV NS5B, have revealed a requirement for a higher concentration of the initiating nucleotide than nucleotides needed only for elongation (32,33). However, there are conflicting results concerning the effect of the concentration of the initiating nucleotide. Using templates possessing a 3Ј-terminal cytidylate or uridylate residue, high concentrations of GTP or ATP, respectively, were required for de novo initiation by the HCV RdRp (23). In contrast, another report showed that GTP could stimulate NS5B activity independent of the specific initiating nucleotide (34). High levels of initiating GTP have also been suggested to favor de novo RNA synthesis by repressing primer extension activity (35). We undertook the present study of NS5B-⌬21 with an RNA template designed to preclude selfpriming and TNTase activities in order to examine more closely the nucleotide requirements for de novo initiation and elongation and to characterize the effect of initiating with di-and trinucleotides or with an exogenous primer on those nucleotide requirements.
Expression and Purification of HCV NS5B-Recombinant NS5B derived from the HCV-1b (Con-1) strain was expressed and purified from Escherichia coli JM109(DE3) cells as modified from Ferrari et al. (19). Briefly, an expression plasmid was constructed with the NS5B C-terminal 21-amino acid truncation and with a His 6 tag addition at the C terminus. Protein production was induced when cells reached A 600 ϭ 1 by 0.2 mM isopropyl-␤-D-thiogalactopyranoside at 24°C for 4 h. Soluble cell lysates were loaded onto an Ni 2ϩ -nitrilotriacetic acid column for affinity purification. Bound protein was eluted with a buffer containing 250 mM imidazole, pooled, and dialyzed against a buffer containing 1 M glycine before loading onto a RESOURCE S (Amersham Biosciences) column. Purified NS5B-⌬21 protein was eluted with a linear gradient of 0 -2 M NaCl, pooled, concentrated, and stored at Ϫ80°C before use.
End Labeling of RNA-5Ј-End labeling of RNA was carried out at 37°C for 1 h in a 10-l volume containing 20 pmol of RNA, 20 pmol of [␥-32 P]ATP (60 Ci), and 10 units of T4 polynucleotide kinase. 3Ј-End labeling of RNA was performed at 37°C for 1 h in a 10-l volume containing 5 pmol of RNA, 5 pmol of [ 32 P]pCp (15 Ci), and 20 units of T4 RNA ligase. Reaction products were extracted by a nucleotide removal kit (Qiagen) following the manufacturer's instructions.
NS5B RdRp Activity Assays-Polymerase activity was assayed in 25-or 50-l reaction mixtures typically containing 0.1 M NS5B-⌬21, 0.1 M RNA template, the indicated amounts of NTPs, and, unless otherwise specified, 1-5 Ci of [␣-32 P]CTP or [␣-33 P]CTP (3,000 Ci/mmol) supplemented with unlabeled CTP to the indicated concentrations, 20 mM HEPES, pH 7.3, 60 mM NaCl, 10 mM MgCl 2 , 7.5 mM dithiothreitol, 20 units/ml RNasin, and 100 g/ml bovine serum albumin at room temperature for 1 or 2 h. Where indicated, di-or trinucleotide initiators or an 11-mer primer was added at the concentrations shown. It should be noted that although the nominal input concentrations of both template and enzyme are 0.1 M, it has been established that only a small fraction (Ͻ1%) of NS5B-⌬21 enzyme purified from bacteria is catalytically competent in vitro (43); therefore, under our assay conditions, the active NS5B enzyme concentration was in fact much lower than that of the template and would allow multiple rounds of transcription. Reactions were stopped by the addition of an equal volume of 0.2 M EDTA. For denaturing PAGE analysis of products, the reaction mixture was adjusted to 150 l with H 2 O, and an equal volume of phenol/chloroform/isoamyl alcohol (25:24:1) was added. The aqueous phase was mixed with 3 l of GlycoBlue (Ambion) and 0.5 M ammonium acetate solution before an equal volume of isopropyl alcohol (ϳ150 l) was added. RNA precipitation was allowed to proceed for at least 30 min at Ϫ20°C. Precipitated pellet was washed once with 500 l of 70% EtOH and resuspended in 1ϫ TBE-Urea Sample Buffer (Invitrogen) to be analyzed by TBE-urea gel. For quantitation of radiolabel incorporation, the stopped reaction mixture was added to a DE81 filter plate (Millipore) prewet with wash buffer (0.5 M dibasic PO 4 buffer, pH 7.0), washed six times each with 200 l/well wash buffer, followed by drying for 30 min at 37°C before the addition of 40 l of Beckman Ready Safe scintillation mixture/well for counting on a TopCount (Packard) scintillation counter. A linear correlation was observed between quantitation of band intensity by a PhosphorImager (Amersham Biosciences) in PAGE analysis and filter plate read-out of radiolabel incorporation. To investigate single round replication in the presence of the enzyme-trapping agent heparin, a time course was performed on NS5B activity in the presence of an 11-mer primer by following [␣-33 P]CTP incorporation at 15 s, 30 s, 1 min, 5 min, 20 min, 60 min, and 120 min. RdRp reactions were initiated by the addition of nucleotides at the indicated concentrations in the presence or absence of heparin (0.2 mg/ml final concentration). Where indicated, NS5B was preincubated with primer and template (each at 0.1 M final concentration) at room temperature for 22 h prior to the reaction.
Nucleotide K m app Determination-For the K m app of GTP, ATP, or UTP in the presence or absence of di-or trinucleotide initiators, RdRp assays were performed as described above by following [␣-32 P]CTP incorporation (supplemented with unlabeled CTP to 0.1 M) with varying concentrations of the nucleotide in question, ranging from 2 to 5 mM, and a 500 M concentration of the remaining two nucleotides. For the K m app of CTP, the isotope dilution method was employed with 3 Ci of [␣-32 P]CTP (at 20 nM) diluted with unlabeled carrier CTP from 1 nM to 50 M, holding the remaining three nucleotides at 500 M in the absence or presence of 0.5 mM trinucleotides. Linearity of reaction up to 1 h was confirmed by radiolabeled CTP incorporation, from which the rate of RNA synthesis was obtained. K m app values were derived by curve-fitting of the rate versus NTP concentration using a hyperbolic equation for onesite binding in Prism (GraphPad). To determine the K m app of GTP under single-round elongation conditions, NS5B was preincubated with primer and template (each at 0.1 M final concentration) at room temperature for 22 h before RdRp reactions were initiated with 1.5 Ci of [␣-33 P]CTP (at 20 nM), 500 M each ATP and UTP, 0.2 mg/ml heparin, and GTP ranging from 10 nM to 500 M. NS5B activity in the presence of heparin was followed by [␣-33 P]CTP incorporation at 30 s, 1 min, 5 min, and 10 min, and the single round burst rate of RNA synthesis derived from the slope at each GTP concentration was obtained. K m app was derived by curve-fitting using Prism software as above.
RNase T1 Digest of RdRp Products-RNA products from RdRp assay were incubated at 30°C for 30 min in a 10-l volume containing 2.5 g of yeast RNA, 0.5 M NaCl, and 1 unit of RNase T1. The digest mixture was extracted by phenol/chloroform and precipitated with isopropyl alcohol in the presence of glycogen before analysis by 15% TBE-urea denaturing PAGE. To confirm that only single-stranded and not double-stranded RNA was cleaved by RNase T1 under the condition used, an end-labeled 30-mer RNA was shown to be completely digested but protected when it was preannealed to a complementary strand.

RESULTS
De Novo Initiation of RNA Synthesis by HCV NS5B on a Synthetic Heteropolymer RNA Template-An in vitro transcribed heteropolymeric host-derived liver-specific co-factor RNA of 399 nucleotides known as D-RNA has been shown previously to serve as an efficient template for either a full-length or an 18amino acid C-terminally truncated form of the purified HCV NS5B polymerase (15). A dimer size RNA product from a selfpriming (copy-back) mechanism and a template monomerlength product attributed to either de novo initiation or TNTase activity were observed. To ensure template homogeneity and to facilitate its modification, we studied a chemically synthesized 75-mer version, DCoH75 (Fig. 1A), whose sequence matches the 3Ј-end of D-RNA. As shown in Fig. 1B, in the presence of all four ribonucleotides (lane 1), HCV NS5B synthesized prominent template-size products (1ϫ) as well as discrete products that are multiples of monomer in length (detectable at least up to 3ϫ). No products were observed in the absence of the enzyme or the template (data not shown). Withdrawal of either ATP (lane 2), UTP (lane 3), or GTP (lane 4) effectively eliminated the 2ϫ and 3ϫ products and reduced but did not completely abolish the synthesis of the 1ϫ product, suggesting that it may arise at least in part through a non-template-directed TNTase activity. Indeed, even when CTP was the only nucleotide present in the assay (lane 5), the amount of monomer-length product was dramatically reduced but not entirely eliminated.
To facilitate our investigation on de novo initiation, we modified the DCoH75 template by replacing its 3Ј-terminal cytidine with dideoxycytidine (ddC), which precludes extension at the 3Ј terminus, thus abrogating both copy-back replication and TNTase reaction. Absence of a free 3Ј-ribose hydroxyl on the modified template was confirmed by parallel treatment of DCoH75 and DCoH75ddC templates with T4 RNA ligase in the presence of radiolabeled pCp, which showed no 3Ј-end labeling of the ddC template (Fig. 1C). This template was able to direct robust RNA synthesis and yielded a product pattern indistinguishable from that generated by the unblocked template in the complete reaction containing all four nucleotides (Fig. 1B, lanes  1 and 6). In contrast to the unblocked template, however, no monomer-length product was visible using the DCoH75ddC template when one of the nucleotides was withdrawn or when CTP was the only nucleotide present (lanes 7-10), confirming that this template is not susceptible to TNTase activity and indicating that only products from de novo initiation were synthesized.
To ascertain whether the labeled products derived from the unblocked or ddC-blocked template were single-or doublestranded RNA, RNase T1 treatment was performed under conditions that allowed cleavage of single-stranded RNA only. As expected, the 1ϫ products from the DCoH75ddC template were protected from the RNase T1 treatment (Fig. 1D, lanes 4  and 5), confirming de novo initiation as the mechanism of RNA synthesis on this template. From the unblocked template, a fraction of the 1ϫ products was digested by RNase T1 (lane 3), consistent with the presence of some single-stranded RNA molecules produced by TNTase activity. Interestingly, the 2ϫ products from both templates were completely protected from RNase T1 digest (Fig. 1D, lanes 3 and 5), indicating that they are not hairpin loops from a copy-back mechanism, as previously described for D-RNA (15). Furthermore, because the ddC template lacks both 5Ј-phosphate and 3Ј-ribose hydroxyl groups, the 2ϫ products generated by this template could not have arisen from template ligation, but probably through an end-toend template switching mechanism, as described for other RdRps (36,37).
Nucleotide Requirements for de Novo RNA Synthesis-Earlier reports on the nucleotide requirements for HCV NS5B-catalyzed RNA synthesis have been based on homopolymeric or heteropolymeric templates that are capable of supporting multiple enzymatic activities including de novo initiation, extension from a primer, and nontemplated nucleotide addition. Apparent K m values from submicromolar to double digit micromolar have been reported for each nucleotide, whereas high concentrations of initiating GTP were found to stimulate de novo initiation (32-34, 38, 39). Using the DCoH75ddC template, which contains four unique nucleotides at its 3Ј-end (CUAG at ϩ1to ϩ4-positions), we sought to examine the individual nucleotide requirements where only de novo synthesis is permitted. Significant [␣-32 P]CTP incorporation was observed in the presence of a 500 M concentration of the other three nucleotides ( Fig.  2A). However, NS5B activity was reduced by Ͼ90% when one of the NTPs was lowered from 500 to 5 M and became virtually undetectable when concentrations of two or more nucleotides were reduced to 5 M. We then determined the requirement for each nucleotide, varying the concentration of the substrate in question while holding the remaining unlabeled nucleotides at 500 M. As summarized in Table 1, under de novo conditions, apparent K m values of 230, 360, and 90 M were obtained for GTP, ATP, and UTP, respectively, whereas that for CTP was far lower at ϳ0.09 M.
A rate-limiting step in NS5B-catalyzed RNA synthesis has been suggested to be the formation of an initiation complex of the enzyme with template and initiating nucleotides before progressing to elongation mode (28,29). For the DCoH75ddC template, because C is the fourth nucleotide incorporated, following the initiating G (ϩ1), A (ϩ2), and U (ϩ3), we asked whether the dramatically lower requirement for CTP concentration that was observed might be due to its utilization only after the transition to processive elongation. In Fig. 2B, enzymatic activity was assayed in parallel experiments following the incorporation of each of the four radiolabeled NTPs at a concentration of 20 nM with the other three nucleotides held at 500 M each. Efficient label incorporation should be observed when that nucleotide is present near or above its K m app value. Accordingly, only radiolabeled CTP yielded significant incorporation at 20 nM on the DCoH75ddC template. However, when we tested an alternative template identical to DCoH75ddC except for an A to G swap at the ϩ3and ϩ4-positions (such that C and U would now be the third and fourth nucleotides incorporated,  respectively), UTP label incorporation was now dominant, whereas CTP at 20 nM could only weakly support activity. UTP labeling on the alternative template was eliminated when the CTP concentration in the reaction was lowered from 500 to 5 M (data not shown). Thus, efficient de novo synthesis appeared to require high concentrations of the first three incorporated nucleotides, with a reduced requirement for nucleotides utilized subsequently.

Effect of Initiation with Di-and Trinucleotides-Previous
biochemical studies have shown that di-or trinucleotides could serve as initiators for NS5B-directed RNA synthesis (29,33). We asked whether the addition of such initiators to facilitate transition to elongation might overcome the high nucleotide requirement we observed for de novo initiation. As shown in Fig. 3, although virtually no de novo synthesis on the DCoH75ddC template was detected when GTP and ATP were present at 5 M each, the addition of a 1 mM concentration of the dinucleotide GpA resulted in an ϳ30-fold enhancement of activity. Thus, in the presence of GpA, high concentrations of the first two incorporated nucleotides were no longer necessary to support RdRp activity. Furthermore, of the 11 dinucleotides tested, only GpA produced such dramatic activation, consistent with this dinucleotide serving to initiate specifically from the 3Ј terminus of the template.
To probe further the effect of dinucleotide addition on nucleotide requirements for de novo initiation, we determined the NTP concentration dependence in the presence or absence of 0.5 mM GpA. Remarkably, the addition of GpA reduced the concentrations of ATP needed to achieve maximal activity by 4 orders of magnitude, from K m app of 360 M to ϳ0.03 M (Fig. 4A and Table 1). A similarly dramatic drop in GTP requirement in the presence of GpA was also observed, resulting in K m app of ϳ0.03 M for GTP as well (Table 1). No significant change in K m app for CTP or UTP was seen with the addition of the dinucleotide (data not shown).
We next investigated the effect of initiating with the trinucleotide GAU. RNA synthesis was stimulated with increasing concentrations of the trinucleotide (K m app ϳ100 M for GAU, comparable with that for GpA). As with GpA, K m app values for GTP and ATP were lowered by several orders of magnitude to ϳ0.03 M in the presence of saturating concentrations of GAU (data not shown). In addition, the trinucleotide GAU also led to a striking reduction in the UTP concentrations required for activity, bringing its K m app value to ϳ0.03 M (Fig. 4B and Table  1). Accordingly, the drop in RdRp activity when GTP, ATP, and UTP concentrations were lowered from 500 to 0.1 M was reversed by the addition of a 500 M concentration of the trinucleotide GAU. Incorporation of 5Ј-end-labeled GAU into the template-length product was also observed at 0.1 M NTPs by denaturing PAGE analysis (data not shown). Taken together, these findings indicate that high concentrations of the first three incorporated nucleotides required for de novo initiation (K m app ϳ100 -400 M) could be surmounted by the pres-  ence of initiating trinucleotides, which presumably served to facilitate the transition from initiation to elongation, where the nucleotide requirement is significantly reduced (K m app ϳ 0.03-0.09 M).
NS5B Activity in the Presence of a Primer-An earlier study has presented evidence that de novo initiation with high levels of GTP could repress primer extension activity (35). Using the ddC-blocked template that prohibits a self-priming mechanism, we studied the effect of an exogenous 11-mer primer complementing the 3Ј terminus of the template on NS5B activity. Under high nucleotide concentrations (500 M) that support de novo initiation, the presence of the primer did not result in enhanced RdRp activity but instead led to significant reduction, which was partially restored by preincubation of the enzyme with template and primer at room temperature for 18 h (Fig. 5A). The activity increase correlated with the time of preincubation of enzyme-template-primer, achieving a maximum ϳ2-4-fold increase by 24 h; longer preincubation did not lead to further activation (data not shown). Contribution of primer extension activity to total RNA synthesis was assessed by comparing the amount of a 5Ј-end-labeled primer incorporated into the template-length product with that of labeled CTP incorporation in an RdRp reaction after preincubation of enzyme with template and primer for 18 h. As shown in Fig. 5B, of the ϳ80 pM full-length size products synthesized, ϳ10 pM arose from primer extension. Thus, given sufficient amounts of nucleotides to support initiation, de novo synthesis was found to be preferred over primer-initiated synthesis.
Nucleotide Requirements for Single Round Elongation in the Presence of Heparin-To characterize in detail the enhanced NS5B activity after preincubation of the enzyme with primertemplate, we undertook a reaction time course analysis. For a reaction initiated after preincubation of enzyme-templateprimer for 22 h at room temperature, a biphasic time course was observed, with an initial burst phase followed by a slower linear phase (Fig. 6A); only the linear phase was seen for the reaction without preincubation. When reactions were initiated in the presence of heparin, a protein-binding reagent that traps unbound polymerase, activity of the linear phase was abolished, suggesting that very little NS5B was in the heparin-resistant elongation complex under normal assay conditions. In contrast, product formation was completely unaffected in the initial burst phase observed after preincubation (Fig. 6A), suggesting that the burst phase represents single-round elongation of primer-template-enzyme complexes formed during preincubation. The amplitude of the burst phase, as extrapolated from the linear phase to zero time, was estimated to be ϳ0.25 nM CTP incorporated, or equivalent to ϳ14 pM full-length products containing 18 cytidines. Thus, assuming a 1:1 stoichiometry of enzyme and template in a single round of polymerization, only ϳ0.014% of the input enzyme (nominal concentration of 0.1 M) appeared to be participating in productive elongation under our assay conditions. We then measured the nucleotide requirements under single-round elongation conditions in the presence of heparin. As shown in Fig. 6B, a K m app of ϳ0.02 M was obtained for GTP in single round elongation during burst phase synthesis, comparable with the K m app value derived in the presence of a dinucle-otide initiator (Table 1). Similar low requirements (Ͻ0.1 M) were found as well for the other nucleotides in single round elongation (data not shown). These findings further support the conclusions on the distinct nucleotide requirements for initiation versus elongation obtained under conditions permitting multiple rounds of initiation and elongation cycles. A Heteropolymer RNA Template Derived from the HCV(Ϫ)UTR-We have focused the study of the nucleotide requirements for initiation on DCoH75ddC, a heterologous RNA template containing four unique nucleotides at its 3Ј-end. To determine whether our findings could be extended to an HCV-derived template, we synthesized a 67-mer RNA matching the 3Ј-end of the negative strand of the HCV genome, replacing the 3Ј-terminal C with ddC to ensure that only de novo synthesis would be permitted (HCV(Ϫ)UTR67ddC; FIGURE 5. Effect of primer addition on NS5B activity. An 11-mer primer complementing the 3Ј terminus of the DCoH75ddC template was used to probe primer extension activity. A, RdRp activity was measured by [␣-33 P]CTP incorporation with or without primer in the presence of a 500 M concentration each of GTP, ATP, and UTP at 30°C for 2 h. Where indicated, enzyme, template, and primer (each at 0.1 M final concentration) were preincubated at room temperature for 18 h prior to the addition of nucleotides to start the reaction. B, primer was 5Ј-end-labeled with T4 polynucleotide kinase and [␥-32 P]ATP. A 500 M concentration each of GTP, ATP, and UTP and 1 M CTP were present in the RdRp reaction with radiolabeled primer, template, and enzyme (at 0.1 M each) at 30°C for 2 h, with (lane 1) or without (lane 2) preincubation of enzyme, template, and primer at room temperature for 18 h. Primer extension activity was evaluated by comparing the amount of 32 Plabeled primer incorporated into the template-length product with that of total [ 32 P]CTP incorporation in a preincubated reaction performed as in A. Fig. 7A). Using this template, the initiating G (ϩ1) and C (ϩ2, ϩ3) are the first nucleotides incorporated. Unlike the DCoH75ddC template, no significant incorporation of radiolabeled CTP occurred (data not shown), consistent with our previous results showing inefficient labeling when one of the first three incorporated nucleotides was used (Fig. 2B). However, efficient label incorporation was observed on this template using radiolabeled UTP (first incorporated at ϩ13) at 20 nM and a 500 M concentration of each of the other three nucleotides (Fig. 7B). Activity was greatly diminished when the NTPs were lowered from 500 to 0.1 M each. The addition of the initiating dinucleotide GpC was not sufficient to restore activity as long as the CTP concentration was kept low, but when the initiating trinucleotide GCC was provided, RNA synthesis was activated even with 0.1 M CTP. Thus, as with the DCoH75ddC template, the presence of the 3Ј-terminal trinucleotide corresponding to the HCV-derived template overcame the high nucleotide requirements needed to support de novo initiation.

DISCUSSION
RNA synthesis catalyzed by the HCV NS5B polymerase is a multistep process involving initiation, elongation, and product dissociation, followed by reinitiation for another round of polymerization. The initiation step can be further divided into binding of RNA to polymerase and subsequent transition into productive complexes capable of the addition of nucleotides. It has been shown that at least under in vitro conditions, the transition to productive elongation complexes is the rate-limiting step. Although extensive analysis has been carried out to characterize various aspects of NS5B biochemical properties, such as the activity of different enzyme forms, template preference, metal ion dependence, etc., the requirement for nucleotides in initiation and elongation steps remains unclear. Higher concentrations of initiating NTPs have been shown to be required by other RNA polymerases, such as T7 RNA polymerase (40), yeast mitochondrial RNA polymerase (41), and more recently, influenza virus polymerase (42). For the HCV NS5B RdRp, the K m values reported for NTPs vary from submicromolar to double digit micromolar (32,33,38,39).
In this study, we used heteropolymeric RNA templates with modified 3Ј-ends to eliminate copy-back and TNTase activities and showed that distinct nucleotide concentrations were required during initiation and elongation steps. High concentrations of NTP (100 -400 M) corresponding to the first three incorporated nucleotides were shown to be needed for efficient de novo RNA synthesis. Such a requirement for high concentration of initiating nucleotides can be overcome by providing the corresponding trinucleotides. In the presence of initiating di-or trinucleotides, the amount of NTPs needed for elongation was 10 3 -to 10 4 -fold lower than that for initiation. The observed differences in NTP binding affinity appear not to be dependent on specific template features but rather an intrinsic characteristic of the HCV NS5B-catalyzed RNA synthesis, since similar results were obtained using templates based on a heterologous synthetic sequence or derived from the 3Ј(Ϫ)UTR of the HCV genome. Since our assay conditions generally allow for multiple rounds of replication, the NTP K m app we obtained is a composite kinetic constant of multiple initiation and elongation cycles. Nevertheless, the respective K m values for initiation and elongation steps can be deduced from our  experimental measurements. High di-or trinucleotide initiator concentrations greatly facilitated the initiation of RNA synthesis; the K m app values obtained for the four nucleotides in their presence were therefore largely a measurement of the NTP binding affinity during elongation. Indeed, the K m app for GTP obtained under such conditions was very similar to that under single round elongation conditions with heparin treatment. Under de novo conditions in the absence of initiators, the high K m app values (Ͼ100 M) obtained for the first three nucleotides largely reflect their binding affinity to NS5B during initiation, since the contribution of elongation K m app (Ͻ0.1 M) to the overall nucleotide requirement is minimal. Our results are consistent with the observations made in another system, although the reported NTP K m values in the presence of di-or trinucleotides were much higher (submicromolar to double digit micromolar) (33). The discrepancy is probably attributable to the suboptimal concentrations (Ͻ ϽK m ) of initiating nucleotide used in that study. Similar to results obtained using short initiators, another group has also reported that the initiation K m for GTP is 10 3fold higher than the elongation K m under de novo synthesis conditions (32).
A previous study has shown that di-or trinucleotides can be used efficiently by NS5B to initiate RNA synthesis, whereas preannealed double-stranded RNAs greater than 4 bp in length are much less preferred (28). These results are consistent with our observations that di-or trinucleotides can dramatically reduce the nucleotide requirement for de novo synthesis, in contrast to the longer 11-mer primer, which was less effective than GAU in supporting RdRp activity and indeed repressed de novo synthesis in the presence of high concentrations of nucleotides. Taken together, these experimental findings support the structural model proposed by Zhong et al. (29) of an initiation complex comprising HCV NS5B, a single-stranded RNA template, and initiating nucleotides. According to this model, the ␤-hairpin loop unique to HCV NS5B protrudes into the active site and sterically hinders the binding of double-stranded RNA, thereby ensuring terminal initiation from single-stranded RNA templates. However, a short duplex RNA of up to 3 bp is readily accommodated at the active site and can serve as the initiator for RNA synthesis. Infrequently, the ␤-loop undergoes a conformational change and moves away from the active site to allow binding of a duplex RNA, resulting in primer-mediated RNA synthesis. The finding that the formation of a productive elongation complex with enzyme-template-primer is a slow process (this study and Refs. 32 and 43) suggests that the ␤-loop conformational change is energetically unfavorable. Nevertheless, a primer can compete with initiating nucleotides for occupancy at the active site, which probably accounts for its repression of de novo synthesis supported by high nucleotide concentrations. During the elongation phase, the template RNA must be translocated processively through the catalytic site to allow incorporation of each incoming nucleotide. We did not observe strand separation of the nascent product-template duplex, which would have resulted in RNase T1 digestion of the labeled strand. In the authentic replication complex, a helicase activity such as that encoded by the HCV NS3 helicase may be required for generating single-stranded viral RNA in vivo.
GTP has been shown to stimulate RNA synthesis by HCV NS5B (34), and it has been postulated that the stimulation may be due to allosteric regulation via a low affinity GTP-binding site located between the finger and thumb region distal to the active site (44). However, more recent mutagenesis analysis showed that activity of NS5B with mutations in the putative GTPbinding site was also enhanced by increasing concentrations of GTP (45). Our findings that activation of RNA synthesis cannot be achieved by high amounts of GTP alone but requires three initiating nucleotides or di-/trinucleotide initiators lend support to the hypothesis that nucleotide binding to the active initiation site is responsible for the observed stimulation.
During HCV NS5B-catalyzed elongation, all four NTPs exhibit very similar apparent K m values, as would be expected from their binding to the same nucleotide pocket. It is noteworthy that these elongation K m values are also in accord with the steady-state K m (double digit nanomolar) reported for the HIV reverse transcriptase (46). The low steady-state K m values have been proposed to be a property of processive polymerases due to the slow rate of dissociation of the enzyme from the template once elongation begins. There is evidence that HCV NS5B does not dissociate appreciably from the template during each round of synthesis, resulting in predominantly full-length products (this study and Refs. 32 and 43). Although short RNA templates have been widely used to characterize HCV RdRp for ease of synthesis and analysis, it would be of interest to investigate elongation kinetics using longer templates (e.g. the full-length HCV genome) where the dissociation rate may be different.
Using the DCoH75ddC template where self-priming or copy-back synthesis was blocked, we nevertheless observed RdRp products that were multimers in length of the input template. They probably arose through an end-to-end template switching mechanism, although confirmation will await more detailed characterization. Evidence for such a mechanism has been presented for poliovirus, carmovirus, tombusvirus, bovine viral diarrhea virus, brome mosaic virus, cytomegalovirus, and other RdRps (36,37,47). Although many in vitro studies on the HCV RdRp have described products longer than input templates, most have attributed them to copy-back synthesis. Our finding lends support to a more critical investigation into the ability of the HCV NS5B to direct the template switch, given the importance of this mechanism in potentially contributing to copy choice RNA recombination in vivo, a process vital to the evolution and pathogenesis of RNA viruses (48).
Intracellular HCV replication occurs in a membrane-associated complex (replicase), which not only consists of the NS5B polymerase but also NS3 helicase, NS4B, and NS5A as well as host factors (49). In particular, NS5B and NS3 helicase have been shown to regulate each other's activities (50,51). Although NS5A has been demonstrated to bind to RNA in vitro (52), its precise function in replication remains unclear. The initiation conditions and nucleotide requirement of NS5B in the context of an authentic replication complex remain to be determined. Interestingly, the nucleotide concentrations required for initiation are similar to the intracellular nucleotide concentrations (ranging from ϳ300 M for CTP to 3,000 M for ATP) (53). HCV RNA replication has been shown to be suppressed with reduced intracellular pools of nucleosides, and the supply of exogenous nucleosides can rescue the inhibition (54). It is possible that one mechanism for the virus to regulate its replication in response to changes in cellular environment is to couple the initiation of new RNA synthesis to the availability of intracellular nucleotides.
The NS5B polymerase has been a major target for the development of antiviral drugs against chronic hepatitis C infection, and compounds inhibiting various steps of viral RNA synthesis have been discovered (5,6). The elucidation of different nucleotide requirements for initiation and elongation by NS5B will facilitate establishment of proper assay conditions to identify inhibitors targeting distinct aspects of NS5B activity. For example, the use of initiating di-or trinucleotides allows the assay to be performed at relatively low NTP concentrations, which should favor the identification of inhibitors competitive with nucleotides. Further mechanistic studies of the HCV NS5B polymerase and its interaction with other viral and host factors will be of great importance to the understanding and treatment of chronic HCV infection.