Selective Stimulation of Hepatitis C Virus and Pestivirus NS5B RNA Polymerase Activity by GTP*

NS5B of the hepatitis C virus is an RNA template-dependent RNA polymerase and therefore the key player of the viral replicase complex. Using a highly purified enzyme expressed with recombinant baculoviruses in insect cells, we demonstrate a stimulation of RNA synthesis up to 2 orders of magnitude by high concentrations of GTP but not with ATP, CTP, UTP, GDP, or GMP. Enhancement of RNA synthesis was found with various heteropolymeric RNA templates, with poly(C)-oligo(G)12 but not with poly(A)-oligo(U)12. Several amino acid substitutions in polymerase motifs B, C, and D previously shown to be crucial for RdRp activity were tested for GTP stimulation of RNA synthesis. Most of these mutations, in particular those affecting the GDD motif (motif C) strongly reduced or completely abolished activation by GTP, suggesting that the same NTP-binding site is used for stimulation and RNA synthesis. Since GTP did not affect the overall RNA binding properties or the elongation rate, high concentrations of GTP appear to accelerate a rate-limiting step at the level of initiation of RNA synthesis. Finally, enhancement of RNA synthesis by high GTP concentrations was also found with NS5B of the pestivirus classical swine fever virus, but not with the 3D polymerase of poliovirus. Thus, stimulation of RdRp activity by GTP is evolutionarily conserved between the closely related hepaciviruses and pestiviruses but not between these and the more distantly related picornaviruses.

The hepatitis C virus (HCV) 1 is an RNA virus that causes acute and chronic liver disease (for reviews see Refs. [1][2][3]. A high proportion of infected patients fail to clear the virus and contract chronic infection, which may lead to liver cirrhosis and, eventually, to hepatocellular carcinoma. Currently it is estimated that 100 -200 million people worldwide suffer from a chronic HCV infection. Thus, HCV became a focus of intensive research worldwide. Based on similarities of genome organization and virus particle structure with the flaviviruses like yellow fever virus and the animal pathogenic pestiviruses like the classical swine fever virus (CSFV), HCV has been classified as the separate genus Hepacivirus in the family Flaviviridae (4).
These viruses have in common an enveloped virus particle and a single-stranded RNA genome of positive polarity encoding a polyprotein that is cleaved by host cell signalases and viral proteinases. In the case of HCV, the genome has a length of ϳ9600 nucleotides. Its single long open reading frame is flanked at the 5Ј-and 3Ј-ends by nontranslated regions of about 340 and 230 nucleotides in length, respectively. The 5Ј nontranslated region is important for efficient translation of the viral polyprotein and functions as an internal ribosome entry site (5)(6)(7). The 3Ј nontranslated region most likely required for RNA replication carries a 3Ј-terminal highly conserved sequence forming stable secondary and probably also higher order structures (8 -11).
HCV polyprotein processing is accomplished by a combination of host and viral proteinases (for reviews, see Refs. 2, 12, and 13). At least 10 different cleavage products could be identified, which are aligned within the polyprotein as follows (from the amino to the carboxyl terminus): NH 2 -C-E1-E2-p7-NS2-NS3-NS4A-NS4B-NS5A-NS5B-COOH (14 -21). The structural proteins C-E2 are the major constituents of the virus particle, whereas the nonstructural proteins 2-5B most likely are required for RNA replication.
While the mechanisms of polyprotein processing have been studied in great detail, our knowledge about HCV replication is scarce. This is essentially due to the lack of robust cell culture systems allowing efficient virus propagation and the lack of an animal model other than the chimpanzee. By analogy to other plus strand RNA viruses, it is assumed that after infection of the host cell the viral RNA is liberated into the cytoplasm and used for synthesis of the viral proteins. These, most likely in conjunction with cellular proteins, assemble into a membraneassociated replicase complex responsible for the multiplication of the viral RNA via a minus strand RNA intermediate. At least two viral proteins most likely are directly involved in this reaction: NS3 carrying in the carboxyl-terminal domain an NTPase/helicase activity (22)(23)(24)(25) and NS5B, the RNA-dependent RNA polymerase (26 -30).
In an attempt to set up appropriate in vitro systems permitting a detailed analysis of HCV replication, we have recently developed a simple expression and purification system allowing the production of large quantities of an enzymatically active NS5B protein carrying a carboxyl-terminal hexahistidine affinity tag (27). In agreement with reports by other groups, we observed with the purified protein a primer-dependent RNA polymerase activity able to processively copy homo-and heteropolymeric templates without detectable requirement for additional cofactors and without template specificity (26 -30). On heteropolymeric templates, initiation of RNA synthesis is primed via a "copy-back" mechanism; i.e. 3Ј terminal sequences fold back intramolecularly, creating a template-bound primer, which is elongated to generate an RNA product twice the length of the input template. Only in the case of homopolymeric templates or templates with a 3Ј homopolymeric tract, which cannot form intramolecular base pairings, the enzyme requires the addition of complementary exogenous primers (26 -30).
Using an intensive mutation analysis, we identified four amino acid sequence motifs within NS5B crucial for enzymatic activity (27). These motifs, designated A-D (31), are located in the central NS5B domain and highly conserved between RdRps of plus strand RNA viruses. Motif A, most likely involved in NTP binding and catalysis, is most sensitive toward substitutions (27). Motif B probably is involved in template and/or primer positioning and characterized by an invariant glycine residue absolutely essential for enzymatic activity. The well known motif C, according to the primary sequence also designated the GDD motif, is important for NTP binding and catalysis, with the first aspartic acid residue being least tolerant toward substitutions. In contrast, alterations affecting the glycine and the second aspartic acid residue are much better tolerated, leading to enzymes with still well detectable RdRp activities (27). Motif D, probably also involved in NTP binding and catalysis, for most RdRps and reverse transcriptases is characterized by a highly conserved lysine residue, whereas an arginine residue is found for most HCV isolates. Interestingly, a lysine substitution for the arginine increases RdRp activity by about 50% (27).
In this report, we continue our biochemical characterization of HCV NS5B. We demonstrate that RdRp activity is selectively stimulated by up to 2 orders of magnitude by high concentrations of GTP. We present evidence for the specificity of this effect and show that GTP-mediated RdRp activation is not due to an increase of RNA binding or the elongation rate, suggesting that high concentrations of GTP accelerate a rate-limiting step during the early phase of RNA synthesis. Finally, we found that NS5B of CSFV can be stimulated by GTP with comparable efficiency but not the 3D polymerase of poliovirus.

EXPERIMENTAL PROCEDURES
Materials-All radiolabeled nucleotides (3000 Ci/mmol) were purchased from Amersham Pharmacia Biotech, and all unlabeled ribonucleotides were from Roche Molecular Biochemicals at the highest quality available. GDP and GMP (disodium salts) and heparin 6000 were obtained from Sigma, homopolymeric RNA templates (average length 400 nucleotides) from Amersham Pharmacia Biotech, and 12mer RNA oligonucleotides from MWG-Biotech. Mung bean nuclease was purchased from Biolabs, T7 RNA polymerase from Epicenter, and RNasin and RNase-free DNase from Promega.
Construction of Plasmids and Generation of Recombinant Baculoviruses-Construction of the plasmid used for expression of an NS5B (residues 2421-3010 of the HCV-BK polyprotein (32)) carrying a hexahistidine affinity tag at the carboxyl terminus has been described (27). Plasmid pBacCSFV-5B His allowing the expression of a full-length NS5B of the classical swine fever virus (33) under control of the polyhedrin promoter was obtained by insertion of an NcoI/XbaI polymerase chain reaction fragment covering the complete NS5B sequence into a modified baculovirus transfer vector (27) restricted with NcoI and SpeI. After sequence analysis of the transferred polymerase chain reaction fragment, the plasmid was used to generate a recombinant baculovirus as described recently (27). The resulting CSFV NS5B contained the hexahistidine epitope at the amino terminus.
Cell Culture and Protein Purification-Expression of NS5B proteins of HCV and CSFV in ovary cells of Trichoplusia ni (High 5 cells) and purification of proteins via differential extraction of cell lysates and subsequent affinity chromatography was performed as described recently (27). The purity of both proteins was higher than 90% as determined by densitometry scanning of Coomassie Blue-stained protein gels. Purified 3D pol of poliovirus was a generous gift from E. Wimmer and A. Paul (State University of New York, Stony Brook, NY) and was purified as described elsewhere (34).
Preparation of RNAs for in Vitro Assays-Full-length HCV RNA was transcribed from plasmid pAT1-9604 (27) containing a complete HCV genome downstream of the promoter for the RNA polymerase of bacteriophage T7. After linearization of the plasmid downstream of the HCV insert with XbaI, nucleotides corresponding to linker sequences were removed by treatment with mung bean nuclease as specified by the manufacturer. The DNA was purified by extraction with phenol and chloroform and concentrated by precipitation with ethanol, and 5 g were used for a 100-l in vitro transcription reaction containing 80 mM HEPES, pH 7.5, 12.5 mM MgCl 2 , 2 mM spermidine, 40 mM dithiothreitol, 3 mM NTPs, 160 units of RNasin, and 250 units of T7 RNA polymerase. After 2 h at 37°C, 125 units of T7 RNA polymerase were added; the reaction was incubated for a further 2 h, treated with 5 units of RNase-free DNase for 15 min, and extracted with phenol/chloroform; and RNA was precipitated with isopropyl alcohol at room temperature. The pellet was resolved in 50 l of diethyl pyrocarbonate-treated water, and remaining unincorporated nucleotides were removed by gel filtration using a G50 NICK column (Amersham Pharmacia Biotech). Integrity of the RNA was analyzed by denaturing formaldehyde-agarose gel electrophoresis, and concentration was determined by optical density measurement.
RdRp Assays-RdRp assays were performed essentially as described (27). In brief, a 25-l reaction was set up containing 20 mM Tris-HCl For analysis of reaction products by gel electrophoresis, the enzyme was inactivated by the addition of 50 l of PK buffer (300 mM NaCl, 100 mM Tris-HCl (pH 7.5), 1% (w/v) SDS, and 20 g of tRNA), 25 l of diethyl pyrocarbonate-treated water, and 50 g of proteinase K (Roche Molecular Biochemicals). After 30 min at 37°C, samples were extracted with phenol/chloroform, and RNAs were precipitated with isopropyl alcohol. Precipitates were resolved in 25 l of gel buffer (40 mM MOPS (pH 7.0), 10 mM sodium acetate, 1 mM EDTA, 50% (v/v) formamide, 2.2 M formaldehyde), heated to 55°C for 15 min, and chilled on ice, and 1 l of ethidium bromide (10 mg/ml) was added. After a 10-min incubation at room temperature, 5 l of loading buffer (50% glycerol, 0.25% bromphenol blue, 0.25% xylenecyanol, 1 mM EDTA) was added, and samples were loaded onto an agarose gel containing 2.2 M formaldehyde, 40 mM MOPS (pH 7.0), 10 mM sodium acetate, and 1 mM EDTA. Electrophoresis was performed at 5 V/cm. Gels were dried and analyzed using an BAS 2500 BioImage analyzer system (Fuji) or x-ray films (Eastman Kodak Co.). To determine incorporation of radioactivity, reactions were stopped by the addition of 100 g of calf thymus DNA and 1 ml of 10% trichloroacetic acid, 0.5% tetrasodium pyrophosphate. After a 30-min incubation on ice, samples were filtered through GF/C glass microfiber filters (Whatman); filters were washed five times with 1% trichloroacetic acid, 0.1% tetrasodium pyrophosphate; and bound radioactivity was determined by liquid scintillation counting. RdRp assays with 3D pol were done as described elsewhere (35). Fold stimulation of RNA synthesis was calculated from the ratio of cpm obtained with NS5B or 3D pol at a given NTP concentration and at 10 M NTP (cpm (X mM NTP):cpm (10 M NTP)).
Determination of Elongation Rates-Elongation rate was determined as described recently with minor modifications (36). In brief, RdRp assays were performed as described above with the difference that NS5B and RNA template were incubated for 5 min at room temperature in 20 mM Tris-HCl (pH 7.0), 12.5 mM MgCl 2 , 10 mM KCl, and 1 mM EDTA prior to the addition of the reaction mixture. To inhibit further initiation of RNA synthesis during the reaction, heparin 6000 was included at a final concentration of 1 mg/ml. Reactions were terminated at various times and processed for denaturing formaldehyde-agarose gel electrophoresis as described above.
RNA Binding Assay-75 ng of purified NS5B were incubated with 160 ng in vitro transcribed full-length HCV RNA supplemented with the analogous radiolabeled RNA (200,000 cpm; specific activity ϳ2.5 ϫ 10 9 cpm/g) in 50 mM HEPES (pH 7.5), 7 mM MgCl 2 , 5 mM 2-mercaptoethanol, 20 units of RNasin, and various concentrations of GTP in a total volume of 100 l. Under these conditions, ϳ30% of the input RNA was bound. After a 1-h incubation at 22°C, samples were diluted to 1 ml with TE (10 mM Tris-HCl (pH 7.5), 1 mM EDTA) and filtered through 0.45-m pore size nitrocellulose filter disks (BA 85; Schleicher & Schü ll), which had been boiled for 30 min in TE prior to use. Bound radioactivity was determined by liquid scintillation counting. Back-ground as determined by an analogous titration with bovine serum albumin was subtracted.

Expression and Purification of HCV and CSFV NS5B
Polypeptides-The method we developed for purification of HCV NS5B carrying a carboxyl-terminal hexahistidine affinity tag is based on the observation that in contrast to most cellular and baculovirus proteins, NS5B is poorly soluble under physiological conditions. However, it can be efficiently extracted from cell lysates with buffers containing high concentrations of salt, glycerol, and detergent (27). Therefore, lysates of infected insect cells were repeatedly extracted with buffers containing increasing concentrations of these ingredients, and soluble proteins contained in the supernatants were discarded (Fig. 1,  lanes 2 and 3). After three extractions, NS5B was strongly enriched in supernatant 3 (S3; lane 4) and used for further purification by affinity chromatography. To compare the biochemical properties of HCV NS5B with the analogous protein of the closely related pestivirus CSFV, we constructed two recombinant baculoviruses directing the expression of NS5B of the CSFV Alfort/187 strain (33) carrying a hexahistidine affinity tag either at the amino or the carboxyl terminus. Both proteins could be purified in an enzymatically active form using the method developed for HCV-5B. 2 As exemplified for the amino-terminal fusion protein, the differential extraction of the cell lysate allowed a strong enrichment of the protein in super-natant 3, and it could be further purified by affinity chromatography (lanes 10 and 12, respectively). Because the expression level of this protein was much higher than the one of the carboxyl-terminal fusion protein, the amino-terminal fusion protein was used for the analyses described in this report. Purity of this protein was Ͼ90% as was the case for HCV-5B.
Stimulation of NS5B RdRp Activity by High Concentrations of GTP-We have recently described a detailed determination of the kinetic constants of HCV-5B (36). During the course of these experiments, we observed a biphasic titration curve for GTP. At low concentrations of GTP, the enzyme displays a standard Michaelis-Menten kinetic with a K m for GTP of 0.5 M on heteropolymeric templates and ϳ3.0 M on the poly(C)oligo(G) 12 homopolymer (36) ( Fig. 2A). Surprisingly, a strong stimulation of RdRp activity was found when GTP concentrations were raised to beyond 100 M. Using an in vitro transcribed HCV full-length RNA ϳ9600 nucleotides in length as template, a linear increase of RNA synthesis was found up to 5 mM GTP ( Fig. 2A). Stimulation of RNA synthesis was maximal at 10 mM GTP, whereas higher concentrations led to an inhibition of the enzyme. Taking the amount of radioactive incorporation at 10 M GTP as reference (6 ϫ 10 3 cpm), an ϳ40and 55-fold stimulation of RdRp activity was found at 5 and 10 mM GTP, respectively. It should be noted that radioactive incorpo-2 V. Lohmann, and R. Bartenschlager, unpublished results.

FIG. 1. Expression of HCV-and CSFV-5B proteins in insect cells and purification.
Insect cells were infected with recombinant baculoviruses directing the expression of the given NS5B proteins. 72 h postinfection, cells were lysed with a hypotonic buffer. After centrifugation, the supernatant (S1) was discarded, and the pellet was extracted with a buffer containing intermediate concentrations of salt, glycerol, and detergent. The supernatant (S2) obtained after centrifugation was discarded, and the pellet was extracted with a buffer containing high concentrations of these components. The resulting supernatant (S3) was applied to a nickel nitrilotriacetic acid spin column, and the flow-through (Ft) was discarded. After several washing steps and a nuclease treatment, proteins were eluted with 250 mM imidazole (El). 1 ⁄100 of each fraction, corresponding to 4 ϫ 10 5 cells, was analyzed by SDS-11% polyacrylamide gel electrophoresis for HCV-5B or SDS-10% polyacrylamide gel electrophoresis for CSFV-5B, and gels were stained with Coomassie Brilliant Blue R-250. In the case of total cell lysate (T) only 1 ⁄200 was loaded onto the gels. Numbers between the panels correspond to the sizes of marker proteins (kDa).

FIG. 2. Activation of NS5B RdRp activity by high concentrations of GTP.
A, a standard RdRp assay was performed using an in vitro transcribed genomic HCV RNA and increasing concentrations of GTP. After a 2-h incubation, 1 ⁄5 of the reaction was withdrawn, and, after trichloroacetic acid precipitation on glas fiber filters, radioactive incorporation was determined by liquid scintillation counting. For comparison, the determination of the K m value for GTP (36) is given in the insert in the upper right. B, 4 ⁄5 of the reaction was processed for denaturing formaldehyde-agarose gel electrophoresis, and radiolabeled RNAs were visualized by autoradiography. Numbers to the left refer to the sizes of RNA marker (nucleotides). ration at 10 M GTP was rather low and somewhat variable. Therefore, stimulation rates obtained in different experiments varied between 50 and 100 (see below).
In agreement with the copy-back priming mechanism, analysis of the reaction products on denaturing formaldehyde-agarose gels revealed that up to 1 mM GTP the size of the majority of radiolabeled RNAs was larger than the input template (Fig.  2B, lane 4). With higher GTP concentrations in addition to an increase of these products, RNAs smaller than the input became prominent. At GTP concentrations beyond 12.5 mM, the sizes of all radiolabeled RNAs were significantly reduced, and only RNA products about the size of the input template and RNAs ϳ2500 -500 nucleotides in length were found (lanes 9 -11). This pattern indicated an inhibition of elongation or processivity at very high concentrations of GTP with templatesized RNA products being generated by "abortive" copy-back priming from the input template and the small RNAs either from RNA primers present in the NS5B or the RNA template preparation or, less likely, by de novo initiation. Only under conditions of lower GTP concentrations would the enzyme have been processive enough to copy the template completely during the 2-h incubation period.
Several possibilities could be envisaged for the RdRp inhibition at very high GTP concentrations: (i) a competitive inhibition because of the high excess of GTP over the other NTPs; (ii) impurities present in low amounts in the GTP preparation, or (iii) an insufficient concentration of Mg 2ϩ due to the excess of GTP over Mg 2ϩ . Because the latter possibility was the most likely, the GTP titration was repeated, but this time for every GTP concentration a 12.5 mM excess of MgCl 2 was used. Under these conditions, an almost identical stimulation of RdRp activity with a maximum at 12.5 mM GTP was found (not shown). At higher GTP concentrations, RNA synthesis still was reduced but to a lower extent, suggesting that inhibition of RdRp activity at high GTP concentrations in part was due to the excess of GTP over Mg 2ϩ . Attempts to overcome this inhibition by further increasing the Mg 2ϩ concentration were unsuccessful, be-cause the enzyme is strongly inhibited by high concentrations of MgCl 2 (36).
To exclude the possibility that the observed effects were due to contaminants in the GTP preparation used, several batches as well as GTP preparations from other suppliers were tested including lithium and sodium salts. In all cases, comparable stimulation rates were found (not shown). In summary, these results show that RdRp activity of HCV-5B is stimulated by high concentrations of GTP and that the inhibition found at GTP concentrations Ͼ10 mM in part was due to an imbalance of the GTP:Mg 2ϩ concentrations. Since a linear increase of RNA synthesis was found with GTP only up to 5 mM, most subsequent titrations were performed up to this value using a constant concentration of MgCl 2 (12.5 mM) shown to be optimal for our purified enzyme (36).
Selectivity of RdRp Stimulation-To analyze whether stimulation of RNA polymerase activity could also be mediated by other NTPs, RdRp reactions were carried out with increasing concentrations of ATP, CTP, and UTP. As shown in Fig. 3A, none of these nucleotides could enhance RNA synthesis to the extent of GTP. Maximum stimulation was found with ATP (ϳ4-fold), whereas only a 3-fold increase of enzymatic activity was observed with CTP and UTP (Fig. 3B). Analysis of the reaction products by agarose gel electrophoresis revealed that comparable with the results obtained with GTP, under conditions of high NTP concentrations products about the size of the input template and short RNAs (500 -1500 nucleotides) accumulated during the 2-h incubation period (Fig. 3C, lanes 4, 8,  and 12). Thus, efficient stimulation of RNA synthesis can only be achieved with GTP, whereas the inhibition of the enzyme is found with high concentrations of either NTP, further supporting the conclusion that it is due primarily to an imbalance of the GTP:MgCl 2 concentrations.
Most Amino Acid Substitutions in NS5B Reducing RdRp Activity Also Affect GTP-mediated Stimulation of RNA Synthesis-The experiments described thus far suggest a direct stimulation of RdRp activity by high concentrations of GTP. How- . After a 2-h incubation, incorporation of radioactivity was determined by trichloroacetic acid precipitation and liquid scintillation counting. B, RdRp assays were performed as in A with given concentrations of ATP (q), CTP (), or UTP (E). 1 ⁄5 of the reaction mixture was subjected to trichloroacetic acid precipitation and liquid scintillation counting, and 4 ⁄5 was processed for denaturing formaldehyde-agarose gel electrophoresis (C). Radiolabeled products were visualized by autoradiography. In the case of ATP and UTP titrations, reactions were performed with [␣-32 P]CTP, while for the CTP titration [␣-32 P]UTP was used as the radiolabeled nucleotide. Numbers to the left refer to the sizes of RNA markers (nucleotides). Note the different scales of the ordinates in A and B. ever, since the NS5B preparations used in this study still contained very low amounts of contaminating cellular proteins, we could not exclude the possibility that the activation effect was related to a copurified contaminant. For example, a GTPase in the enzyme preparation could lower the actual GTP concentration, although in this case we would have been unable to determine the K m for GTP (0.5 M on heteropolymeric templates; Fig. 2A; Ref. 36). Alternatively, a terminal transferase activity, present in low amounts in the enzyme preparation (27), might be stimulated by GTP, generating primer-templates that might be used efficiently by NS5B. To exclude these possibilities, we tested a panel of mutations in polymerase motifs B and C with reduced RdRp activity and an arginine substitution in motif D that increased enzymatic activity ϳ1.5-fold (Table I and Fig. 4A; Ref. 27). A significant reduction of GTPmediated stimulation was found with all substitutions affecting the GDD motif, whereas the mutation in motif D had no effect (Table I). Two opposite phenotypes were found with the mutations in motif B. While the valine substitution for the nonconserved threonine at position 286 did not affect stimulation by GTP, the cysteine substitution for the highly conserved threonine at position 287 completely abolished this property.
To further characterize the latter mutant, RdRp assays were performed in parallel with the parental enzyme and an NS5B, which was enzymatically inactive due to an asparagine substitution for the invariant 318 aspartic acid. Analysis of the reaction products by agarose gel electrophoresis revealed that RdRp activity of 5B 287C was severely impaired and not stimulatable by GTP (Fig. 4B). Only RNA products corresponding to the input template and ϳ700 nucleotides in length were found, indicating a strong processivity or elongation defect. This pattern was similar to the one obtained with the parental NS5B at GTP concentrations of Ն20 mM GTP (Fig. 2B), and it suggested that the ϳ10-kilobase RNA corresponds to the input RNA template labeled at the 3Ј-end by abortive copy-back priming. The small RNAs probably represent short copy-back primed molecules or molecules initiated from exogenous primers or by de novo synthesis. In summary, the results obtained with these altered NS5B proteins strongly argue against the contribution of a cellular contaminant in the enzyme preparation, and they suggest that high concentrations of GTP directly stimulate RNA synthesis by NS5B. Furthermore, the fact that all mutations affecting highly conserved residues assumed to contribute to the active site of the enzyme can no longer be or can only poorly be stimulated by GTP suggests that the same NTPbinding site is used by GTP for stimulation and incorporation.
Evidence for Incorporation of the Stimulating GTP-In the experiments described so far, we always used an RNA template corresponding to the HCV genome. To analyze whether GTPmediated stimulation was dependent on this particular template, a panel of hetero-and homopolymeric RNAs was characterized in the same way. Comparable stimulation rates were found with an ϳ400-nucleotide-long HCV RNA corresponding to the 3Ј-end of the genome and with an in vitro transcribed lacZ RNA (data not shown) showing that GTP enhances RNA synthesis on very different heteropolymeric templates. In the case of homopolymeric RNAs, a stimulation was only found with poly(C)-oligo(G) 12 , whereas no activation was observed with poly(A)-oligo(U) 12 even at 5 mM GTP (Fig. 5A) or UTP (data not shown), suggesting that the stimulating GTP has to be incorporated into the RNA chain.
To further substantiate this assumption, we analyzed whether stimulation of RNA synthesis also could be achieved with GMP or GDP. Therefore, RdRp assays were performed with increasing concentrations of GMP or GDP. As shown in Fig. 5B, instead of a stimulation of RNA synthesis a strong reduction was found with GMP and, in particular, with GDP, corroborating that the stimulating nucleotide also has to be incorporated and that GMP and GDP compete with GTP for the same nucleotide binding site.

GTP Most Likely Accelerates an Early
Step of RNA Synthesis-Template-dependent synthesis of a nucleic acid by an RNA polymerase can be divided into at least four consecutive steps: binding of the enzyme to the template, initiation of RNA synthesis, elongation, and termination. To find out which of these steps might be affected by GTP, the following experiments were carried out.
To analyze whether GTP has an effect on RNA binding, a radiolabeled RNA corresponding to the HCV genome was incubated with purified NS5B in the presence of increasing GTP concentrations, and bound RNA was quantitated by liquid scintillation counting (Fig. 6A). Using the assay conditions given under "Experimental Procedures," no influence of GTP on RNA binding was found. This result corroborates the finding that GTP does not stimulate RNA synthesis from the poly(A)-oligo(U) 12 template and that the stimulating nucleotide has to be incorporated into the RNA product.
Alternatively, GTP might increase the elongation rate. To validate this assumption, RdRp assays were carried out with the HCV genomic RNA and increasing concentrations of GTP. After an initial 5-min incubation of NS5B and template, the reaction was started by the addition of NTPs together with an excess of heparin. The latter was included to prevent further initiation of RNA synthesis during the incubation allowing the measurement of the elongation rate of synchronously primed molecules. The determination of the sizes of the reaction products was complicated by the fact that most RNAs were primed by a copy-back mechanism and therefore covalently linked to the template. Given the lengths of these molecules, the sizes of newly synthesized RNAs could not be accurately determined. Therefore, RNA products taken at different time points during the reaction were treated with RNases under high salt conditions, allowing the cleavage of only single-stranded but not double-stranded RNA. In this way, single-stranded loops at the 3Ј-end of the template connecting the input RNA and product are cleaved allowing a separation of template and newly synthesized RNA by denaturing formaldehyde-agarose gel electrophoresis. As shown in Fig. 6B, no significant difference of the lengths of RNA products was found when assays were performed with 50, 250, or 1000 M GTP (lanes 1-3 and 7-9), b Nomenclature of the polymerase motifs according to (31). c Percentage of RdRp activity of each altered NS5B protein compared with the parental NS5B (5B C-His ) using synthetic full-length HCV RNA and 10 M GTP. d Ratio of cpm obtained with NS5B at 5 mM and 10 M GTP (cpm (5 mM):cpm (10 M)). e not applicable.
whereas a strong reduction of their sizes was found with higher GTP concentrations (lanes 4 -6 and 10 -12). Thus, high concentrations of GTP have two opposite effects on RdRp activity: a strong stimulation of RNA synthesis and a strong reduction of the elongation rate, the latter being due to the excess of GTP over Mg 2ϩ . In summary, these data demonstrate that stimulation of RdRp activity by GTP is not due to an enhancement of RNA binding or an increase of the elongation rate. Based on these exclusion criteria, high concentrations of GTP (up to ϳ10 mM) appear to stimulate initiation of RNA synthesis. GTP-mediated Stimulation of RNA Synthesis by NS5B of a Pestivirus but Not by 3D pol of Poliovirus-As described in the Introduction, HCV belongs to the genus Hepacivirus and is most closely related to the pestiviruses, a group of animal pathogenic viruses to which CSFV belongs. This close evolutionary relationship led us to speculate that GTP-mediated stimulation of RNA synthesis might be a biochemical trait conserved between these virus groups and raised the question of whether it would also be found with the more distantly related poliovirus. To address these questions, we analyzed CSFV-5B and 3D pol of poliovirus for their ability to be stimulated by high concentrations of GTP. As described above, the CSFV enzyme was purified from insect cells (Fig. 1), whereas 3D pol of poliovirus was expressed in Escherichia coli and purified as described elsewhere (34). GTP-mediated stimulation of RNA synthesis was measured for these two enzymes in parallel with NS5B using the HCV genome length in vitro transcript as described under "Experimental Procedures." A strong stimulation was found both with NS5B of HCV and the analogous enzyme of CSFV (Fig. 7A). Enhancement of RNA synthesis for the pestiviral polymerase was linear between 0.1 and 5 mM GTP, and the stimulation rate was well comparable to the one of HCV NS5B. Only at low GTP concentrations did these enzymes behave somewhat differently in that RdRp activity of CSFV NS5B was lower but showed a stronger stimulation between 10 and 100 M GTP. In contrast, no stimulation was found with 3D pol of poliovirus, both with the full-length synthetic HCV RNA template (Fig. 7A) and with poly(C)-oligo(G) 12 (data not shown). Attempts to detect a significant stimulation by varying reaction conditions (in particular MgCl 2 or MnCl 2 concentrations) were not successful (data not shown).
An analysis of the reaction products by agarose gel electrophoresis revealed that the RNAs generated by all three enzymes were larger than the input suggesting that copy-back priming is a common mechanism (Fig. 7B). However, the overall sizes of the products were somewhat different. Compared with HCV-5B, RNAs generated by CSFV-5B were shorter and products obtained with 3D pol were longer, indicating different elongation rates or processivities. Furthermore, in contrast to the GTP-mediated stimulation observed with the NS5Bs, the poliovirus enzyme was instead inhibited by high concentrations of GTP. At Ն2.5 mM, much lower amounts of RNA products were found, and these RNAs were shorter in size than the ones obtained at lower GTP concentrations ( Fig. 7B; compare, for example, lane 14 with lane 18). This pattern was reminiscent of the one found with HCV-5B at GTP concentrations of Ն15 mM (Fig. 2B) and could similarly be explained at least in part by the excess of GTP over Mg 2ϩ . In summary, these results clearly show that RdRp activities of HCV and CSFV NS5B but not of poliovirus 3D pol are stimulated by high concentrations of GTP. DISCUSSION Studies aimed at elucidating the mechanism of HCV replication are limited by the lack of robust and reliable cell culture systems and animal models other than the chimpanzee. Therefore, we have begun to characterize the biochemical properties of NS5B RdRp, most likely the key player of the viral replicase complex. During the course of the determination of the K m value for GTP, we observed a strong stimulation of RNA synthesis by high concentrations of GTP. It is toward a detailed analysis of this effect that the present study was undertaken.
Using synthetic full-length HCV RNA and purified NS5B, we found a 50 -100-fold stimulation of RNA synthesis by GTP. This stimulation was selective and was not obtained with ATP, CTP, UTP, GMP, or GDP. As shown by the reduced stimulation FIG. 4. A single amino acid substitution in polymerase motif B of NS5B abolishes GTP-mediated stimulation of RNA synthesis. A, a schematic presentation of NS5B indicating the positions of polymerase motifs A-D is given at the top. The amino acid sequences of individual motifs highly conserved among HCV isolates are shown below. Capital letters indicate amino acid residues that are highly conserved among RdRps of plus strand RNA viruses, and underlined and boldface capital letters refer to residues that are highly conserved among all viral RdRps. Numbers above the letters indicate the amino acid positions of NS5B, arrows below the letters refer to substituting residues (see Table I). B, standard RdRp assays were performed with synthetic HCV genomic RNA and GTP concentrations given above the lanes (mM). After 2 h at 22°C, samples were processed for denaturing agarose gel electrophoresis. Numbers between the gels refer to the sizes of RNA markers (nucleotides). 5B C-His refers to the parental NS5B with the carboxyl-terminal hexahistidine affinity tag.
of several NS5B mutants, GTP most likely directly acts on NS5B and occupies the same binding site that is used for RNA synthesis. The stimulation was found with every heteropolymeric template tested and with poly(C)-oligo(G) 12 but not with the poly(A)-oligo(U) 12 homopolymer. All of these data taken together demonstrate a selective GTP-mediated stimulation of RNA synthesis by NS5B.
The way this stimulation is accomplished currently is not known. GTP has no effect on the overall RNA binding properties or the elongation rate. Based on these observations the acceleration of a rate-limiting step during initiation of RNA synthesis is the most likely explanation. In a simple model, GTP would bind to NS5B and induce an "initiation-competent state," e.g. by the induction or stabilization of a conformational change required for efficient initiation. The "activated" GTP-NS5B complex would then bind to the template and initiate RNA synthesis. Alternatively, NS5B might first bind to the template and then to GTP. In either case, according to this model template binding in the absence of GTP would be nonproductive, i.e. NS5B binds to the RNA but in most cases falls off without initiation of RNA synthesis. Once the initiation is completed, the enzyme processively copies the template and is no longer stimulated by high concentrations of GTP.
In this context, it should be noted that an alteration or stabilization of the enzyme conformation by high concentrations of GTP has been suggested for 3D pol of poliovirus (37).
Using cross-linking of oxidized GTP to purified 3D pol , it was found that NTP binding protects RdRp activity from heat denaturation (37). Interestingly, GTP protected the enzyme to a greater extent than the other three NTPs indicating a tighter binding. Although K m for GTP of 3D pol is very low, high concentrations (ϳ2 mM) were required for stoichiometric crosslinking and maximum protection from heat denaturation (37). Two NTP binding sites were mapped, one near the amino terminus (residues 57-74) outside of the polymerase core region and the other in the central region overlapping with motif B (residues 266 -286) (38). Only the amino-terminal NTP binding site was shown to be crucial for RNA replication in that a leucine substitution for the highly conserved lysine residue at position 61 completely abolished RdRp activity (39,40). For NS5B, the results described here with the various mutations suggest that an NTP binding site in the core of the enzyme is required. As deduced from the x-ray crystal structure of 3D pol (41), the invariant carboxylates of motifs A and C of NS5B would be clustered in the "palm" (for a review see Ref. 42). Since substitutions of these residues both reduce RdRp activity FIG. 5. Evidence that the stimulating nucleotide has to be incorporated into the RNA. A, RdRp assays were performed as described under "Experimental Procedures" using poly(C)-oligo(G) 12 (E) or poly(A)-oligo(U) 12 (q) and given concentrations of GTP. Incorporation of radioactivity was determined by trichloroacetic acid precipitation and liquid scintillation counting. Because of the isotope dilution, GTP concentrations higher than 1 mM were not tested with the poly(C)oligo(G) 12 template. B, effect of increasing concentrations of GMP (q) and GDP (E) on RdRp activity using synthetic full-length HCV RNA and 1 M GTP (see "Experimental Procedures").
FIG. 6. Stimulation of RNA synthesis by GTP is not due to an enhancement of RNA binding or an increase of the elongation rate. A, RNA binding assays were performed with a radiolabeled HCV genomic RNA in the presence of increasing concentrations of GTP. RNA-NS5B complexes were immobilized on nitrocellulose filters, and bound radioactivity was determined by liquid scintillation counting. B, influence of increasing GTP concentrations on the elongation rate. RdRp assays were performed as described under "Experimental Procedures" using the in vitro transcribed HCV genomic RNA. After the given incubation times, RNAs contained in the reaction mixtures were purified by SDS/proteinase K digestion and subjected to an RNase treatment under conditions allowing only cleavage of single-stranded RNA. Protected RNAs were analyzed by denaturing formaldehydeagarose gel electrophoresis and radiolabeled products were visualized by autoradiography. Numbers to the right refer to the sizes of RNA markers (nucleotides). and stimulation by GTP, the NTP binding site in the core appears to be involved in both reactions. However, currently we cannot completely rule out the possibility that these amino acid substitutions caused structural alterations affecting an NTP binding site somewhere else in NS5B. Furthermore, we also found a mutant in motif B (5B 286V ) with a very low RdRp activity that could be stimulated with an efficiency comparable with the wild type. However, this mutation is the only one affecting a nonconserved residue, suggesting that this particular substitution affected a property of the enzyme not directly involved in NTP binding and/or catalysis. It remains to be established whether NS5B has more than one NTP binding site and, if so, which of these is required for enzymatic activity and/or stimulation by GTP.
Another unresolved question is why the stimulation is selectively exerted by GTP. Perhaps the initiating nucleotide has to be a guanylate as is the case for many RNA polymerases like the one from bacteriophage T7 or the plus strand RNA virus BMV (43,44). For example, the DNA-dependent RNA polymerases of E. coli and T7 possess at least two distinct NTPbinding sites: one that recognizes only the initiating GTP in a Mg 2ϩ -independent manner and the other used for elongation, binding to all four NTPs in a reaction requiring Mg 2ϩ (43). In case of BMV, initiation of RNA synthesis can be stimulated up to ϳ10-fold by GMP or the dinucleotide GpG, corresponding to the penultimate and antepenultimate nucleotide of the viral template (45). The fact that this stimulation is only found with low concentrations of GTP suggests the existence of two NTPbinding sites: one for the initiation nucleotide, binding preferentially GTP or GMP, and one for the elongation nucleotide, binding each of the four NTPs with a comparable K m . For NS5B of HCV we did not detect a stimulation by GMP, and the K m for GTP was ϳ100-fold lower than the one reported for BMV (50 M;Ref. 45), suggesting that a different mechanism is responsible for the stimulation of RNA synthesis by high concentrations of GTP.
Another possibility of how GTP stimulates RNA synthesis by NS5B would be an enhancement of the transition from initiation to elongation. For the well studied DNA-dependent RNA polymerases, it was shown that transcription can be functionally separated into several distinct steps: template binding, promoter localization, melting of the DNA to form a transcriptionally open complex, nucleotide substrate binding, formation of the first phosphodiester bond, abortive RNA synthesis, promoter clearance, and processive elongation and termination (for a review, see Ref. 46). A characteristic feature of these enzymes is the inefficient transition from initiation to elongation resulting in the synthesis of abortive products. These oligonucleotide products dissociate from the initiated complex when they are not elongated. A similar phenomenon has been described for the RdRp of BMV, resulting in the synthesis of an 8-mer oligonucleotide in a 10-fold molar excess over the fulllength RNA product (47,48). Interestingly, GTP appears to stabilize the RdRp-RNA interaction and in this way increases efficiency of RNA synthesis (49). Although a similar mechanism might be responsible for stimulation of RNA synthesis with NS5B, it should be kept in mind that BMV RdRp initiates RNA synthesis de novo, whereas NS5B, at least under the in vitro conditions used here, is primer-dependent. If abortive RNA products would be made, then they probably will not be released but remain covalently bound to the 3Ј-end of the primer. On the other hand, currently we cannot exclude the possibility that in addition to a stimulation of primer-dependent RNA synthesis, a certain fraction of the RNA molecules synthesized at high GTP concentrations are initiated de novo. Further studies will be required to clarify these issues.
The relevance of RNA synthesis stimulation by GTP for the in vivo situation currently is not known. The fact that unphysiologically high concentrations of GTP are required for maximal stimulation clearly suggests that a different mechanism may operate in the infected cell. One simple explanation would be that the effect we observe here is due to the enzyme preparation containing largely misfolded or thermosensitive proteins and that GTP binds to and activates or stabilizes these molecules and in this way increases the number of enzymatically active NS5B proteins. However, the results obtained with the panel of NS5B mutants and with the NS5B of CSFV suggest that this is not the case. Furthermore, the findings that GTP does not enhance poly(U) synthesis from the poly(A) template and that GMP inhibits RNA synthesis indicate that GTP does not act allosterically on RdRp but rather has to be incorporated into the RNA. An alternative explanation would be that high concentrations of GTP substitute for a function normally executed by a cellular or viral protein. For example, in the case of poliovirus it was shown that VPg and in particular the VPg precursor 3AB enhance RNA synthesis by 3D pol 50 -100-fold (50,51). Similar to what we describe here for NS5B, RNA synthesis stimulated by 3AB is primer-dependent. 3AB has a dual function; it can interact with 3D pol (52), and it is an RNA-binding protein able to interact with the template-primer (53). Mutational ablation of RNA binding also reduces 3D pol stimulation, suggesting that 3AB enhances interaction between the template and 3D pol (51,53). An additional function of VPg is to act as an acceptor of UMP added by 3D pol to the OH group of a tyrosine residue within VPg in a template-dependent reaction (54). Uridylylated VPg then is used as the primer for transcription of viral RNA. Whether RNA synthesis by NS5B FIG. 7. GTP stimulates RNA synthesis by HCV and CSFV NS5B but not by 3D pol of poliovirus. A, RdRp assays were performed with the in vitro transcribed HCV genomic RNA and HCV-5B (q), CSFV-5B (E), or 3D pol (). After 2 h at 22°C, incorporation of radioactivity was determined by trichloroacetic acid precipitation and liquid scintillation counting. B, analysis of the reaction products by denaturing formaldehyde agarose gel electrophoresis. Numbers between the gels refer to the size of RNA markers (nucleotides). GTP concentrations are given above the lanes (mM).
The finding that RNA synthesis by NS5B of HCV and CSFV but not by 3D pol is stimulated by high concentrations of GTP demonstrates the conservation of this biochemical trait between hepaciviruses and pestiviruses and probably reflects the evolutionary distance between this virus group and the picornaviruses. It also might correlate with different mechanisms responsible for initiation of RNA synthesis. All of these RdRps are strictly primer-dependent. As described above, in the case of poliovirus it was shown that the primer for RNA synthesis by 3D pol is the VPg (or its precursor 3AB), which is uridylylated by 3D pol in a template-dependent reaction (54). For pestiviruses and hepaciviruses, it is not known whether these enzymes also use a protein as a primer. The difference from 3D pol with respect to the GTP-mediated stimulation, together with the observation that high concentrations of GTP probably enhance initiation of RNA synthesis, indicates that a different mechanism is responsible. Possibilities are de novo initiation of RNA synthesis or priming by a nucleic acid (which is more difficult to imagine with linear templates). In any case, the results presented here describe an interesting biochemical property of NS5B, and they should facilitate further functional analyses of this enzyme. Since NS5B also is an attractive target for antiviral therapy, our findings may have important consequences for the development of potent inhibitors.