Promoter/Origin Structure of the Complementary Strand of Hepatitis C Virus Genome*

Hepatitis C virus (HCV) NS5B protein encodes an RNA-dependent RNA polymerase (RdRp). Sequences in the 3′ termini of both the plus and minus strand of HCV genomic RNA harbor the activity of a replication origin and a transcription promoter. There are unique stem-loop structures in both termini of the viral RNA. We found that the complementary strand of the internal ribosome-binding site (IRES) showed strong template activity in vitro. The complementary strand RNA of the HCV genome works as a template for mRNA and viral genomic RNA. We analyzed the promoter/origin structure of the complementary sequence of IRES and found that the first and second stem-loops worked as negative and positive elements in RNA synthesis, respectively. The complementary strand of the second stem-loop of IRES was an important element also for binding to HCV RdRp.

NS5B shows RdRp activity in vitro (16, 18 -24). NS5B has a highly hydrophobic 21-amino acid sequence in its C terminus, and when it is removed NS5B becomes soluble (25)(26)(27). HCV RdRp exhibits de novo and copy-back initiation activities (18,24,28,29). It can utilize single-stranded RNA as a template but not double-stranded RNA (30,31). It prefers a cytidine at the 3Ј terminus of the template, and de novo initiation in vitro by HCV RdRp was selectively activated by a high GTP concentration (22,23,26,30,32). It can utilize both viral and non-viral RNA templates although a specific promoter has not been identified (18,22,24). NS5B bound to the poly(U) stretch (18) and the upstream conserved stem-loop structures at the 3Ј end of the genome (14). We have recently found that the complementary sequence of IRES showed very strong template activity (26,34).
In this study, we analyzed the promoter/origin structure of the complementary sequence of IRES, and we determined the role of the complementary strand of the first and second stemloops of IRES in RNA synthesis in vitro. This domain perfectly overlaps with those identified in the HCV replicon system in vivo (35).

EXPERIMENTAL PROCEDURES
Recombinant HCV RdRp-HCV NS5B protein truncated by 21 Cterminal amino acids with a His 6 tag was expressed in Spodoptera frugiperda (Sf)-21AE cells and purified as described previously (26). Purified HCV RdRp was stored at Ϫ25°C in the presence of 50% glycerol.
RNA Templates-RNA templates designed from the complementary sequences of HCV IRES (cIRES) were synthesized with a MEGAscript T7 RNA polymerase kit (Ambion) (Fig. 1). The DNA templates were produced by PCR using the primer pairs listed in Table I. The sequence UGGC was added to the 3Ј terminus of all the templates. The DNA templates were removed by digestion with DNase I after in vitro transcription, and all transcripts were purified by 6% PAGE, 7 M urea for use as the templates of in vitro transcription by HCV RdRp, followed by successive phenol-chloroform extraction and ethanol precipitation. These RNA templates were resuspended in RNase-free water and stored at Ϫ80°C until used.
Transcription in Vitro-Unless otherwise indicated, HCV RdRp activity was measured in 50 l of standard transcription buffer, TxG(ϩ) (20 mM Tris/HCl (pH 8.0), 100 mM KCl, 2.5 mM MnCl 2 , 50 M ATP, 50 M CTP, 5 M UTP, 0.5 mM GTP, 0.185 Mbq of [␣-32 P]UTP, 10 pmol of RNA template, 25 g/ml actinomycin D, 5 units of human placental RNase inhibitor (Nacalai Tesque, Japan), 1 mM DTT, and 10 pmol of NS5B). For the single-round transcription assay, the reaction mixture without nucleotides was preincubated with 50 or 500 M GTP at 29°C for 30 min. Then 0.2 mg/ml heparin (Wako Chemicals, Japan) was added to the mixture, followed by ATP, CTP, and UTP, respectively, and the reaction mixture was further incubated at 29°C for an additional 90 min. The reaction was stopped by extraction of 150 l of Sepasol RNA II (Nacalai Tesque, Japan) and 40 l of chloroform. [ 32 P]UMP-labeled RNA was precipitated with the equal volume of 2-propanol. The radiolabeled RNA was washed with 70% ethanol, dried, and resuspended in formamide dye loading buffer and analyzed by electrophoresis on a 6% PAGE containing 7 M urea. The radioactivity of the transcribed RNA was measured with a BAS-2000 image analyzer (Fuji Film), and the amount of transcribed RNA was calculated from the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. amount of UMP in the transcripts. Each value was calculated from the average of at least three independent assays.

RESULTS
Single-round Transcription-First of all, the concentration of heparin for the single-round transcription was determined ( Fig. 2A). HCV RdRp and cIRES template (10 pmol each) were treated with 3.1, 6.25, 12.5, 25, 50, 100, and 200 g/ml of heparin after preincubation with 50 M (G(Ϫ)) or 0.5 mM GTP (G(ϩ)) in 20 mM Tris/HCl (pH 8.0), 100 mM KCl, 2.5 mM MnCl 2 , 25 g/ml actinomycin D, and 5 units of human placental RNase inhibitor at 29°C for 30 min. Then 50 M ATP, 50 M CTP, 5 M UTP, and 0.185 Mbq of ␣-[ 32 P]UTP were added and incubated for an additional 90 min. The RdRp activity with more than 0.2 mg/ml heparin was about 20% that without heparin and did not drop below 20% even when more heparin was added. Next, the time course with 0.2 mg/ml heparin was examined (Fig. 1B). Under these conditions, the accumulation of transcribed RNA The secondary structure of cSL1 and cSL2 is predicted by mFold. That of cSL3 and cSL4 is drawn as a mirror of that of IRES because mFold predicted 17 patterns for their secondary structure and it has yet to be determined (36,37). The designation of the model templates indicates the name of the complementary sequences of the stem-loop structure of IRES. All the templates have UGGC (underlined) at their 3Ј terminus. The 3Ј termini of SL234 -1D, SL234, SL34-S, SL34, SL4, and SL0 start from the position marked by arrowtails (A). In SL34-SS, 6 Gs (bold) are substituted by 6 As (bold) of SL34-S (B). Templates carrying only cSL1 and cSL2 (SL12, SL12-1S, and SL12-1LD) are designed as an internal deletion of cIRES (C, D, and E). The 3Ј terminus of SL2 starts from the position marked by an arrowtail (C). In SL1234 -1S and SL12-1S, the GC stem sequences are substituted with AU (D). In SL1234 -1LD and SL12-1LD, the AAUC sequence of the loop structure is substituted with A (E). continued for 90 min and reached a plateau. Thus, 0.2 mg/ml heparin was used for the single-round transcription assay of HCV RdRp.
Effect of Stem-loop Structures on Transcription in Vitro-First, the activity of deletion mutant templates for stem-loop structures was measured by the single-round transcription assay because the initiation activity could be accurately measured with this assay (Figs. 1 and 3). Under de novo initiation by 0.5 mM GTP, the template activity of SL234 -1D, SL234, SL34, SL4, and SL0 was 74.8, 88.2, 34.2, 27.9, and 9.7% that of cIRES, respectively (Fig. 3D). There was no significant difference among cIRES, SL234 -1D, and SL234. However, there was a big decrease in activity between SL234 and SL34. Without the initiation, far less RNA was synthesized de novo, and there was little difference among them.
Several bands smaller than the template were found among the transcripts from cIRES, SL234 -1D, SL234, and SL34 (Fig.  3, A and B). Additional transcripts were also found in SL34-S, SL1234 -1S, SL1234 -1LD, and SL34-SS (Fig. 5A). However, no additional bands were found in SL4 and SL0. From the pattern and size of the transcripts, we concluded that they were produced by early termination. From the size of the two major additional transcripts estimated from PAGE (Fig. 3, *1 and *2), transcription terminated in bulge *1 and *2 of cSL3, respectively (Fig. 1). Until this experiment, we only calculated the transcripts of template size, excluding those derived from early termination.

Sequences of the primers used in this study
The T7 polymerase promoter sequence is shown in bold (TK-1 and TK-2). The mutated nucleotides are shown in italics (TK-10 to TK-12). An extra GCCA sequence was added to the 3Ј terminus of the RNA templates (TK-4 to TK-13). respectively. One pmol of cIRES and SL234 -1D inhibited the binding of cIRES with HCV RdRp. One pmol of 3NTR almost inhibited the binding as well. Ten pmol of SL234, SL34, SL34-S, and XREG inhibited the gel shift with cIRES. From these results, cSL2 is concluded important for binding with HCV RdRp especially in 0.5 mM GTP. DISCUSSION The HCV RNA genome contains conserved 5Ј-and 3Ј-UTRs (8 -12). As in the case of Flaviviridae family viruses, the 3Јterminal X-region is expected to play an important role in the synthesis of the minus strand, and the complementary strand of IRES is expected to serve as the origin for plus strand synthesis in genome replication (4). The complementary strand of IRES may also work as a promoter of transcription. Mutations in IRES also affected the replication of genomic RNA (35). The reason for this may be that the mutation in the 3Ј terminus of the complementary strand (cIRES) affected the replication and transcription. Both termini of the viral genomic RNA have stem-loop secondary structures. The 3Ј termini of both the plus and minus strands of genomic RNA are able to serve as templates for RdRp in vitro (16) The complementary sequences of IRES had the highest template activity for de novo RNA synthesis in vitro (26,34). However, activity for the de novo synthesis of RNA in vitro by RdRp in the X region was very weak (26). Therefore, we determined the promoter/origin structure of  In the multiround transcription system, the products from short templates were sometimes larger than those from long ones. To compare the initiation activity, we established a de novo single-round transcription system using 0.2 mg/ml heparin to treat HCV RdRp and templates followed by preincubation with 0.5 mM GTP (Fig. 2). Because HCV RdRp prefers a cytidine at the 3Ј terminus and interacts with GTP (28,30,32,33), all the templates are designed to have UGGC at the 3Ј terminus ( Figs. 1 and 5). We have temporally used the secondary structure of cIRES as a mirror image of that of IRES in Fig.  1 until it is determined experimentally (36,37).
Because the transcription activity following the treatment with 0.5 mM GTP decreased markedly when cSL2 was deleted, cSL2 was important for de novo initiation (Fig. 3D). However, the activity of SL1234 -1S was also half of that of cIRES. Because the T m of the UA stem was 12°C, a stem structure might not form in the reaction at 29°C. A comparison of the activity of cIRES, SL1234 -1D, SL1234 -1S, and SL1234 -1LD indicated the complicated secondary structure of the template, although cSL2 and the stem of cSL1 may affect the structure of the promoter/origin. The sequence between cSL2 and cSL3 could also affect the activity. The seven Gs, which exist between cSL2 and cSL3, did not affect the activity. From a comparison of the activity of SL32-S, SL34-SS, and SL34, the length of the single stranded sequence at the 3Ј terminus was confirmed important as reported previously (26).
In this series of experiments, we calculated the product amount from only the template-sized bands. Additional products were transcribed from the templates containing cSL3 (Figs. 2 and 5). We measured the size of the products of template size and smaller (Fig. 3, *1 and *2). By comparing their size with that of the templates, the products *1 and *2 were identified as early termination products from the templates. The positions of possible termination sites are mapped in the bulges of cSL3 (Fig. 1). There is a triple helical structure in IRES corresponding to bulge *1, and a complex stem-loop in the stem structure of stem-loop 3 corresponding to bulge *2. We predict a strong secondary structure for these sequences in cIRES. The results and the prediction of secondary structure by mFold (bioinfo.math.rpi.edu/ϳzukerm/) (38, At this point, the 3Ј end of the templates did not reach to the catalytic center of the RdRp when cSL1 formed because a 5-nucleotide-long single-stranded RNA end is required to reach the catalytic center (31,42). NS3 helicase binds to cSL1 (44), and open its stem and the 3Ј terminus extend to reach the catalytic center (45). HCV RdRp continues to synthesize RNA and moves to the 5Ј end of the RNA template. 39), which predicted too many to be shown here, suggest a complicated secondary structure of cIRES. The results obtained with deletion mutants of stem-loop structures of cIRES were difficult to interpret. The cIRES sequence may make complicated stem-loop structures which interact with each other. Therefore, we decided to design a simpler template.
Because we could not predict the secondary structure of cIRES but wanted to elucidate the role of cSL1 and cSL2 in the template activity, we constructed templates carrying only cSL1 and cSL2 (SL2) to exclude early termination (Fig. 5B). In this experiment, the activity of SL12-1S was more than twice that of SL12 but that of SL12-1LD was similar to the activity of SL2. Because the T m of the UA stem in Fig. 1D was 12°C, a stem structure might not form in the reaction at 29°C. The stem structure of cSL1 could inhibit the activity. The sequence of the cSL1 loop did not affect the activity. Because the results from the templates carrying mutant sequences of cSL1 were not conclusive, we made additional templates carrying simple structures. When cSL1 was added to SL0 (SL0ϩSL1), the amount of product decreased, confirming that cSL1 was a negative element (Fig. 5C). cSL1 was always identified as shown in Fig. 1 when the secondary structure of the templates was predicted by mFold (data not shown). When cSL1 exists, the length of the 3Ј singlestranded RNA is only four nucleotides, too short to initiate the reaction because at least a five-nucleotide singlestranded RNA is required for initiation (31, 40 -43).
We demonstrated that HCV RdRp bound to the 3Ј terminus of the complementary strand RNA as well as that of the positive strand RNA (14). Gel shift assay indicates that cSL2 is also important for the binding of HCV RdRp especially with 0.5 mM GTP (Fig. 4A). Without 0.5 mM GTP, the template-RdRp interaction was enhanced, and SL34 also bound to RdRp effectively. The RdRp binding activity of cSL2 was similar to that of the poly(U/C) tract in the 3Ј terminus of HCV genome RNA (Fig.  4B). RdRp could also bind to other stems or sequences although the binding was weaker than that of cIRES, SL234 -1D, and SL234. Considering the results of transcription and RdRp binding, cSL2 would be a positive element of the promoter/origin structure because it binds specifically to RdRp with 0.5 mM GTP. 0.5 mM GTP may give specificity to the binding of RdRp to cSL2.
Although we did not show the results, copy-back products became apparent without 0.5 mM GTP preincubation even in single-round transcription. The mechanism of switching from de novo initiation with 0.5 mM GTP preincubation to copy-back initiation remains to be resolved. Fig. 6 shows the proposed scheme of initiation from the 3Ј terminus of the complementary strand of IRES. HCV RdRp binds to cSL2. HCV RdRp can start with single-stranded 3Ј termini (17,31). However, the 3Ј single-stranded sequences of cIRES are only four nucleotides long and are too short to reach the active site of RdRp because of the ␤-hairpin structure (30,31,42), so they cannot initiate transcription efficiently. NS3 helicase binds to cSL1 (44) and may relax the stem-loop structure to bring the 3Ј terminus to the RdRp active site so that RdRp interacts with GTP. NS3 interacts with RdRp (45). In both cases, NS3 and HCV RdRp are expected to form a transcription and replication initiation complex. Recently, HS5A was found to bind RdRp and modulate its activity (46). We have demonstrated that cSL2 played an important role in the initiation of transcription and replication of the HCV genome by interacting with RdRp. High concentrations of GTP may give specificity to the interaction between RdRp and template. The HCV initiation complex needs to be reconstituted in vitro using purified RdRp and NS3.