Persistent infection with hepatitis C virus (HCV) is an important cause of chronic liver disease in the United States and many other countries. Infections with this hepatotropic flavivirus are associated with necroinflammatory liver injury, progressive fibrosis, and, in the most severely affected patients, cirrhosis and hepatocelluar carcinoma (Seeff 1997; Alter et al. 1998). Recombinant interferon-α, either alone or in combination with ribavirin, is widely used to treat HCV infections. However, these drugs eliminate the infection in less than 50% of persons with chronic hepatitis (McHutchison et al. 1998; Lindsay et al. 2001). The development of more effective antiviral agents for the treatment of chronic hepatitis C is urgently awaited, but has been slowed by the lack of cell culture systems that are fully permissive for viral replication as well as a lack of small animal models of the disease. Although the lack of permissive cell cultures has made difficult characterization of this virus and its replicative life cycle, much has been learned from the characterization of the viral genome and by in vitro expression of the viral proteins from recombinant cDNA (Major and Feinstone 1997; Blight et al. 1998; De Francesco 1999). More, recently, the development of subgenomic HCV replicons that are capable of replication in human hepatoma (Huh7) cells has provided powerful new approaches to characterizing the molecular processes involved in replication of the viral RNA (Lohmann et al. 1999).

HCV is presently classified as the sole member of the Hepacivirus genus in the family Flaviviridae. Its genome is a single-stranded RNA molecule ~9.6 kb in length with positive-sense polarity (Choo et al. 1989, 1991). This RNA has a 5′ nontranslated RNA (5′NTR) segment of ~342 nt and a 3′NTR of ~225 nt. These two regions are among the most highly conserved sequences in the HCV genome (Bukh et al. 1995). They flank a single large open reading frame that encodes at least 10 proteins, including (in a 5′ to 3′ direction) the viral structural proteins (core, E1, E2, and p7) and at least 6 nonstructural proteins (NS2, NS3, NS4A, NS4B, NS5A, NS5B; Major et al. 1997; Blight et al. 1998; De Francesco 1999). However, only the polypeptides spanning NS3 to NS5B are required for replication of the viral RNA (Lohmann et al. 1999). Given analogies with other positive-strand RNA viruses, and evidence for interactions of these proteins with each other as well as with intracellular membranes, these nonstructural proteins are likely to participate in formation of a large membrane-bound, macromolecular replicase complex containing multiple enzymatic activities (Egger et al. 2002; Mottola et al. 2002). The NS5B protein, which is located at the carboxy end of the viral polyprotein, is an RNA-dependent, RNA polymerase (RdRp), and forms the catalytic core of this complex. The structure of NS5B has been well characterized, and its enzymatic activities have been studied in detail using recombinant proteins, usually truncated at the carboxy terminus in order to remove a probable membrane-interaction domain and increase the solubility of the expression product (Lohmann et al. 1998; Ferrari et al. 1999; Oh et al. 1999).

Recombinant NS5B proteins are known to be capable of primer-independent initiation of RNA synthesis on a variety of virus-specific and nonspecific RNA templates (Lohmann et al. 1998; Ferrari et al. 1999; Oh et al. 1999). However, there are conflicting descriptions of the precise site of initiation of negative-strand transcription on virus-specific templates (Oh et al. 2000; Hong et al. 2001; Kim et al. 2002; Shim et al. 2002). Although the 3′NTR is both highly conserved and rich in structure (Yamada et al. 1996; Blight and Rice 1997; Ito and Lai 1997), purified, recombinant NS5B protein binds to the upstream NS5B coding sequence with significantly greater affinity than it does to this segment of the viral genomic RNA (Kim et al. 2002). Thus, the mechanisms by which the HCV RdRp recognizes the terminal sequences of the genomic RNA for proper and specific initiation of negative-strand RNA synthesis remain controversial.

Evidence derived from experiments done in vitro suggests that the 3′NTR may be required for transcription of genome-length RNA by a purified NS5B RNA dependent RNA polymerase (Oh et al. 1999, 2000). A surprising finding, however, was that transcription of the negative-strand RNA product was initiated 21 nt upstream of the 3′ terminus of the positive-strand template (Oh et al. 2000), within the single-stranded segment of a stem-loop (stem-loop 1, or SL1) that is putatively formed by the 3′ terminal 46 nt of the genomic RNA (Fig. 1 ; Blight and Rice 1997; Ito and Lai 1997). Other in vitro studies also suggest that primer-independent, de novo transcription of RNA from templates containing the 3′NTR sequences may involve the internal initiation of negative-strand RNA synthesis, but at other sites (Kim et al. 2002; Pellerin et al. 2002). However, other investigators have shown that transcription can in fact be initiated de novo by a recombinant NS5B polymerase in vitro at the 3′ end of short oligonucleotide templates representing the 3′ terminus of the positive-strand genomic RNA (Shim et al. 2002).

FIGURE 1. A schematic depicting the organization of the dicistronic Ntat2ANeo(SI) replicon RNA (Yi et al. 2002) in which mutations were created within the 46 nt that putatively form a 3′ terminal stem-loop structure, SL1. An expanded view below the map (more ...)

Despite substantial differences in these in vitro studies, there is convincing evidence that the 3′ terminal nucleotides of the positive-strand template influence both the efficiency of transcription as well as the site of transcriptional initiation (Kim et al. 2002; Shim et al. 2002). Such data are consistent with a role for the 3′ terminal sequences of the viral RNA in recognition of the template by the replicase complex. This hypothesis is further supported by the fact that deletion of the 3′ terminal 46 nt destroys the ability of otherwise infectious synthetic genome-length HCV RNA to initiate infection in intrahepatically inoculated chimpanzees (Yanagi et al. 1999). The deletion of this sequence also abrogates the ability of subgenomic replicon RNAs to replicate in cultured cells (Friebe and Bartenschlager 2002; Yi and Lemon 2002). Interestingly, this 46-nt sequence putatively forms a stable stem-loop structure at the extreme 3′ terminus of the genomic RNA (Fig. 1 ; Kolykhalov et al. 1996; Tanaka et al. 1996). The existence of SL1 is supported by results of studies mapping sites of ribonuclease cleavage or chemical modification within synthetic RNA transcripts (Blight and Rice 1997; Ito and Lai 1997). However, there is a paucity of data from mutational analyses or phylogenetic comparisons to support the computer predictions of the secondary structure assumed by this highly conserved segment of the genome. We have thus focused our attention on SL1 in an effort to determine whether the predicted structure is functionally required for replication of the viral RNA.

As mentioned above, several nonstructural proteins other than NS5B are likely to contribute to the functional replicase complex (Blight et al. 1998; De Francesco et al. 2000; Egger et al. 2002; Mottola et al. 2002). Thus, there are limitations on what can be learned about viral RNA replication from in vitro transcription reactions that contain only a single purified viral protein, NS5B, such as those that were characterized in the studies described above (Oh et al. 2000; Hong et al. 2001; Kim et al. 2002; Shim et al. 2002). We have studied the requirements for the SL1 structure during replication of subgenomic HCV RNAs in human hepatoma cells (Lohmann et al. 1999; Ikeda et al. 2002). Because the autonomous amplification of these selectable replicons appears to follow a process that is very similar if not identical to that mediating the replication of viral RNA, the results of this analysis are likely to be relevant to the process by which the viral genome is recognized by the replicase complex in virus-infected hepatocytes.

Figure 2A shows SEAP activity secreted over successive 24-h periods following transfection of En5-3 cells with the RNA. The results have been normalized to SEAP activity levels produced during the first 24 h after transfection, as during this period of time SEAP expression is due to translation of the transfected RNA and is not dependent on RNA replication (Yi and Lemon 2002; Yi et al. 2002). The figure shows that the secretion of SEAP increases gradually over the next several days, reaching a plateau within 7 days of transfection. Cells transfected with a replication-defective Ntat2ANeo(SI) mutant containing a deletion in the NS5B sequence (Fig. 2A , ΔGDD mutant) express a similar high level of SEAP activity during the first 48–72 h after transfection, but the secretion of SEAP by these cells subsequently drops dramatically and approaches that produced by normal En5-3 cells within 7 days. Also shown in Figure 2A are the results of SEAP assays carried out on media collected from cells transfected with ΔSL1, a mutant derived from Ntat2ANeo(SI) in which all of the sequence contributing to stem-loop SL1 has been deleted (Yi and Lemon 2002). The pattern of SEAP secreted by these cells parallels that produced by cells transfected with the ΔGDD mutant. These results confirm the essential nature of the SL1 sequence to viral RNA replication. As we and others have reported previously (Friebe and Bartenschlager 2002; Yi and Lemon 2002), the failure of subgenomic RNAs lacking this 3′ terminal sequence to replicate cannot be attributed either to decreased stability of the RNA, or to a reduction in translation mediated by the internal ribosome entry site of HCV. Although the Ntat2ANeo(SI) RNA directs the expression of neomycin phosphotransferase as a selectable marker (Fig. 1 ), G418 was not included in the culture media in transient transfections such as shown in this experiment.

FIGURE 2. Expression of SEAP by En5-3 cells following transient transfection with synthetic (filled circles) Ntat2ANeo(SI) RNA, and related replication-defective mutants, (filled diamonds) ΔGDD and (open triangle) ΔSL1, in the absence of G418 selection. (more ...)

Figure 2B presents these results differently, relating the SEAP activity expressed by cells transfected with the parent Ntat2ANeo(SI) RNA and its derivative, the ΔSL1 mutant, to that expressed by cells transfected with the ΔGDD mutant. This figure demonstrates the exponential increase in SEAP secretion induced by the Ntat2ANeo(SI) RNA relative to that induced by the related ΔGDD mutant. Figure 2B also demonstrates the complete failure of the ΔSL1 mutant to replicate. This was confirmed by a lack of surviving G418-resistant cells when attempts were made to transduce G418 resistance with this RNA. In contrast, the parental Ntat2ANeo(SI) RNA gave rise to G418-resistant cells at high efficiency (>10⁴ colonies/10⁶ transfected cells).

FIGURE 3. Transient transfection of En5-3 cells with Ntat2ANeo(SI) and related mutants containing nucleotide substitutions within the putative loop segment of the SL1 stem-loop located at the 3′ end of the 3′NTR. (A) Schematic showing the substitution (more ...)

We next investigated the effect of single nucleotide substitutions within the loop sequence (Fig. 3A ). Surprisingly, individual base substitutions to the Watson–Crick complement were well tolerated at some positions, whereas other single base substitutions were lethal for replication in this assay (Fig. 3B ). No intermediate phenotypes were detected in these experiments (Fig. 3B ). Transversion substitutions at the base of the loop, involving the substitution of 21U, 22C, or 26G (nucleotide positions numbered from the 3′ end of the genome) to 21A, 22G, and 26C, respectively, completely abrogated replication of the RNA. In contrast, similar substitutions at the internal loop positions, with substitutions at 23A, 24U, and 25A to 23U, 24A, and 25U, respectively, had no discernible effect on SEAP expression and thus replication of the RNA. These results suggest that interactions between the bases in the lower half of the loop may be important for defining the conformation of the loop and the recognition of the 3′ terminus of the RNA by the replicase complex. Alternatively, this part of the stem-loop structure might interact directly with a component of the replicase complex. In contrast, the top of the loop does not appear to be involved in critical RNA–RNA or RNA–protein interactions given the apparent plasticity of the apical 23A, 24U, and 25A residues.

Because 20G and 27U are predicted by MFOLD to form a G-U closing pair at the base of the loop (Fig. 3A ), it is unlikely that 21U and 26G form a similar base pair. However, to gain further insight into the requirements for these two bases and potential interactions between them, we evaluated a second series of mutants with additional substitutions at 26G, or both 26G and 21U (Fig. 4 ). The substitution of 26G with either C, A, or U led to a loss of detectable replication in the transient transfection assay (data not shown). This confirms the importance of this particular nucleotide for replication of the RNA, as suggested by the data shown in Figure 3 . Furthermore, replication competence was not restored to the 26U mutant by a second, potentially compensatory 21G substitution (Fig. 4 , mutant 21G-26U). This indicates that a 21G-26U pair cannot functionally replace 21U and 26G.

FIGURE 4. Summary of results obtained in transient transfection experiments and attempts to transduce the selection of G418-resistant En5-3 cell lines following transfection of Ntat2ANeo(SI) and related mutants with different nucleotide substitutions at nucleotide (more ...)

Despite the absence of detectable replication in transient transfections with the 21G-26U mutant, it was possible to select G418-resistant cell colonies following transfection with this RNA. Although time consuming and tedious, the selection of G418-resistant cells is a more suitable approach for detection of a small population of cells that support the replication of transfected RNAs below the limit of detection in the transient assays depicted in Figures 2 and 3 (Friebe and Bartenschlager 2002; Ikeda et al. 2002; Pietschmann et al. 2002). The ability to select antibiotic-resistant cells following transfection of mutated replicons expressing neomycin phosphotransferase also allows for the identification of primary or second-site revertants that are capable of efficient RNA replication. RNAs like the 21G-26U mutant that fail to generate levels of SEAP expression that are indicative of replication of the RNA in transient transfection assays, but that remain capable of giving rise to G418-resistant cell clones following antibiotic selection, can be considered to contain mutations that are ‘‘pseudo-lethal’’ for replication. 21G-26U was the only mutant among those shown in Figure 4 with a pseudo-lethal phenotype. Because G418-resistant cells could not be selected with the other mutants, including the single mutant 26U, these data suggest that the 21G mutation was able to rescue the replication competence of 26U, albeit only at a low level.

Three clonally distinct G418-resistant cell lines were established from En5-3 cells transfected with the 21G-26U double mutant. The 3′ terminal nucleotide sequence of the replicon RNAs present in each of these cell lines was determined by sequencing cDNA obtained following amplification by RT-PCR (see Materials and Methods). In each case, the engineered 21G and 26U substitutions had reverted to the original 21U and 26G nucleotides that are present in Ntat2ANeo(SI) RNA. Interestingly, whereas in one of these revertants (Fig. 4 , R1) the 3′NTR sequence was identical to that of the NtatNeo2A(SI) parent, two revertants (Fig. 4 , R2 and R3) contained mutations in addition to the reversions at 21U and 26G. In R2, 25A was substituted with G within the loop of SL1 (note that the results in Fig. 3B demonstrated that an A to U substitution is also tolerated at this base position). In R3, an additional A to G substitution was present within the loop sequence of a stem-loop structure, SL2, that has been proposed to exist immediately upstream of the SL1 (Fig. 1 ), at the 60th base position from the 3′ end of the genome.

When we reintroduced cDNA fragments amplified from R2 and R3 by RT-PCR into the background of Ntat2ANeo(SI), we found that RNA bearing the R2 substitution replicated as efficiently as the parental RNA in a transient transfection assay, and similar to the 25U mutant (Fig. 3B ). RNA bearing the R3 mutation also replicated, but at a reduced level compared to the parental RNA, based on the quantity of SEAP produced by the transfected cells (data not shown). These results suggest the possibility that repair of the replication defect in 21G-26U may require a complex, multistep process involving sequential base changes. If so, the substitution found in the loop of SL2 suggests the possibility of tertiary interactions between the loop sequences of SL1 and SL2. However, further studies will be needed to properly assess this hypothesis.

FIGURE 5. Transient transfection of En5-3 cells with Ntat2ANeo(SI) and related mutants containing block nucleotide substitutions within the putative duplex stem of the SL1 stem-loop. (A) Schematic showing the highlighted block mutations within the loop segment. (more ...)

When these mutant RNAs were transfected into En5-3 cells, none gave rise to sustained levels of SEAP secretion, indicating that each of these Ntat2ANeo(SI) mutants was either nonviable or otherwise severely handicapped in its replication capacity. These results suggest that the specific sequence of the duplex stem in SL1 is important for efficient replication of HCV RNA. Structures possessing duplex stems of comparable stability but different primary nucleotide sequence cannot functionally substitute for SL1, even when the nucleotide sequence is kept constant within either the apical or basal half of the stem as well as within the loop. Despite this, we found that it was possible to select G418-resistant cells following transfection with the SL1-A/AA mutant RNA at a very low frequency. This indicates that this particular mutant retains a very low level of replication competence. Interestingly, none of the other mutants shown in Figure 5 possessed such a pseudo-lethal phenotype, including the single block substitution SL1-A and SL1-AA mutants. This provides support for the existence of the duplex stem, and for a specific requirement for the structure in replication of the RNA. We discuss the characterization of replicon RNAs recovered from G418-resistant cell lines selected following transfection with the pseudo-lethal SL1-A/AA mutant in a subsequent section.

FIGURE 6. Summary of results obtained following transfection of En5-3 cells with Ntat2ANeo(SI) and related mutants containing single or double nucleotide substitutions within the duplex stem of SL1. (A) Schematic showing the predicted SL1 structure and location (more ...)

Interestingly, however, a very different result was obtained with point mutations introduced at either end of the duplex stem. The 19C and 28G mutations are located at the penultimate positions in the stem, adjacent to the predicted G-U closing base pair at the bottom of the loop in SL1 (Fig. 6A ). Either mutation completely abrogated replication of the subgenomic RNA in En5-3 cells (Fig. 6B ). Importantly, both substitutions also had a significant effect on the structure of the apical end of the hairpin as predicted by MFOLD, as well as the structure of the loop (Fig. 6C ). Strikingly, a combination of these two substitutions, which restored the predicted SL1 structure, also restored replication to near normal levels. These latter results provide strong evidence for the existence of the predicted 19G-28C base pair. They suggest that base pairing in this region of the predicted stem is important for replication of the viral RNA, presumably by facilitating recognition of the 3′ terminus of the RNA by the viral replicase complex. Taken together with the results shown in Figure 3 , the results suggest that structure in the region of the junction of the loop and the duplex stem is critically important in these processes.

Point mutations at the base of the hairpin also had a dramatic effect on replication of the RNA. The 2C and 45G substitutions, located at penultimate positions near the base of the predicted stem (Fig. 6A ), both completely blocked RNA replication in the transient transfection assay (Fig. 6B ). However, in this case, the combination of these compensatory mutations failed to restore an efficient replication phenotype to the RNA, even though the combination of the mutations restored the predicted structure at the base of the hairpin. These data suggest that the specific nucleotide sequence at the base of the hairpin is also critical for recognition by the replicase and/or initiation of negative-strand, viral RNA synthesis. This was a surprising finding, given the pseudo-lethal nature of the SL1-A/AA mutant (Fig. 5A ) described above, and the fact that it retained at least a low level capacity for replication.

The SL1-A/AA mutant (Fig. 5A ) also had a pseudo-lethal phenotype, as G418-resistant colonies could be selected following transfection of this RNA despite little apparent difference in the levels of SEAP expression induced by this RNA and the replication-incompetent ΔGDD mutant in the transient replication assay (Fig. 5B ). These G418-resistant cells were of particular interest because the entire basal half of the SL1 stem had been inverted in sequence by large block substitutions in this mutant (Fig. 5A ). Northern analysis confirmed the presence of the replicon RNA in clonally isolated G418-resistant cell lines obtained following transfection with the SL1-A/AA RNA (Fig. 7A , lanes 3,4,5), although the abundance of the replicon was substantially lower than in either of two G418-resistant cell lines selected following transfection with RNA containing the intact 3′NTR, Ntat2ANeo(RG) (Fig. 7A , lanes 2,6; Yi and Lemon 2002; Yi et al. 2002). In one of three independent G418-resistant SL1-A/AA cell lines studied, the abundance of the replicon RNA was at the limits of detection by northern analysis (Fig. 7A , lane 4).

FIGURE 7. Analysis of replicating RNAs within stable, G418-resistant cell lines. (A) Northern analysis of HCV RNA present in G418-resistant cells lines selected following transfection with replicon RNAs: (lane 1) normal En53 cells, (lanes 2, 6) cell lines selected (more ...)

Sequencing of the 5′ and 3′ termini of the replicon RNAs present in cell lines transduced with SL1-A/AA demonstrated the presence of the block substitutions in the stem of SL1. There were no changes in the mutated nontranslated RNA segments except for, remarkably, the presence of an additional 3′ terminal U (Fig. 7B ). This additional nucleotide was present in RNAs recovered from each of the three independent SL1-A/AA replicon-bearing cell lines shown in Figure 7A . The SL1-A/AA transcripts, like all the RNAs transfected into En5-3 cells in the preceding experiments, were derived by runoff transcription from plasmid DNAs that had been linearized by digestion with Xba1 (see Materials and Methods; Ikeda et al. 2002). Because of the protruding 5′ end generated by this restriction enzyme, these RNAs contain four exogenous nucleotides downstream of the 3′ terminus of the HCV sequence. We have previously shown that these exogenous nucleotides are removed following transfection and replication of the RNA, and that the nucleotide sequence of the extreme 3′ end of replicon RNAs present in stable G418-resistant cell clones is identical to that of the virus (Yi and Lemon 2002). In constructing the SL1-AA and SL1-A/AA plasmids (Fig. 5A ), it was necessary to add an additional T downstream from the 3′ terminus of the mutated viral sequence to maintain the XbaI restriction site required for linearization of the DNA prior to transcription of the RNA. This increased the number of non-HCV bases included at the 3′ end of these RNA transcripts from four to five (Fig. 7B ). The additional U at the 3′ terminus of the SL1-A/AA replicons recovered from the G418-resistant cells is likely to have been derived by retention of the extra U that was present in these transcripts due to the need to maintain the restriction endonuclease recognition site in the plasmid DNA.

The retention of this U in the replicating RNAs suggests the possibility that the 3′ terminal U residue may also play an important role in recognition of the RNA by the replicase. It is important to note that the retention of this exogenous U preserved the native 3′ terminal GU dinucleotide sequence in HCV RNA, and that the G residue in the second base position from the 3′ end is base paired in these replication-competent RNAs (see above).

FIGURE 8. Importance of the 3′ terminal U nucleotide to replication of subgenomic HCV RNA. (A) Schematic showing location of the XbaI and BciVI restriction endonuclease recognition sites placed downstream from the HCV sequence in the plasmid version of (more ...)

Having shown that an RNA with the precise 3′ terminal sequence of HCV had a moderately improved capacity for replication, we next constructed a series of Ntat2ANeo(SI) mutants with the BCiVi restriction site in which the 3′ terminal U of the HCV sequence was replaced with A, C, or G. The RNAs produced from these plasmids have the exact 3′ terminus of HCV RNA with the exception that the nucleotide composition is varied at the extreme 3′ terminal position. Figure 8C depicts an experiment in which En5-3 cells were transfected with these RNAs. These results demonstrated a clear difference in the efficiency of replication of these RNAs that was reproducible in other experiments. RNA transcripts with the native 3′ terminal U replicated more efficiently than the other RNAs. The order of replication efficiency was U first, followed by C, A, and then G, with the G mutant inducing only about 10% of the SEAP activity produced by the wild-type 3′ terminal U (Fig. 8C ).

To show that the variation in the replication efficiency of these mutants was not due to the 3′ terminal nucleotide influencing the stability of RNA, we created a deletion mutation within the NS5B sequence in each of these constructs. Because these ΔGDD mutants are incapable of RNA replication, any induction of SEAP secretion is due entirely to translation of the input transfected RNA. Therefore, under these conditions, the kinetics of SEAP expression represent a measure of the stability of the transfected RNA (Yi et al. 2002). We compared the levels of SEAP expressed by cells transfected with these double mutants with that expressed by cells transfected with RNAs containing only the 3′ nucleotide substitution. There was a clear difference in the levels of SEAP induced by the replication-competent single mutants with different 3′ terminal nucleotides that was evident within 3 days of transfection (Fig. 9A ). However, the 3′ nucleotide substitution had no influence on the kinetics of expression of SEAP following transfection of the ΔGDD double mutants (Fig. 9B ). This indicates that the difference in replication efficiency was not due to variation in the stability of the RNA. We were also unable to detect consistent differences in the stabilities of these RNAs by measuring RNA abundances with a real-time RT-PCR assay following transfection of the mutants (data not shown). Because tat is expressed from these replicon RNAs under the translational control of the HCV IRES, the fact that the levels of SEAP expression were indistinguishable 24 h after transfection of the different RNAs (Fig. 9 ) also indicates that the variance in replication efficiency was not associated with significant differences in the translational activity of the HCV IRES.

FIGURE 9. (A) SEAP activity present in supernatant culture fluids collected at 24-h intervals following transfection of En5-3 cells with BciVI runoff transcripts containing the precise 3′ terminus of the genomic RNA of HCV with the exception of the first (more ...)

Figure 8C shows that the rate of increase in SEAP activity was similar with all four terminal nucleotide variants 7 days after transfection, following passage of the transfected cells, although the absolute levels of SEAP were quite different. This contrasts with the different rates of increase in SEAP activity that were evident in the first few days following transfection (Figs. 8C, 9A ). Because we have previously shown that the levels of SEAP produced by transfected cells are highly correlated with intracellular replicon abundance (Yi et al. 2002), this suggests that the mutants had regained a level of replication efficiency similar to that of the parental RNA with a 3′ terminal U by 8 days post-transfection. To determine whether this apparent change in replication capacity was associated with reversion of the 3′ terminal nucleotide substitutions to U, we isolated RNA 24 and 96 h following transfection of the mutants into cells and determined the 3′ terminal nucleotide sequence of the RNA as described in Materials and Methods. These results indicated that there was rapid reversion of the mutated 3′ terminal nucleotides to U. RNAs containing a 3′ terminal U were present as mixtures with the transfected RNA sequence as early as 24 h after transfection, with U being the dominant sequence at 96 h after transfection of each RNA (results summarized in Fig. 8A ). These results confirm that the HCV RNA replicase possesses a strong preference for U as the 3′ terminal residue of the genomic RNA.

Although previous studies have shown that the 3′ nontranslated sequence of the genomic RNA of HCV is essential for replication of the virus (Yanagi et al. 1999; Kolykhalov et al. 2000; Friebe and Bartenschlager 2002; Yi and Lemon 2002), there has been little understanding of how these sequences contribute specifically to the viral life cycle. Here, we present an extensive mutational analysis of the 46-nt sequence that has been suggested to form a hairpin structure, SL1, at the extreme 3′ end of the RNA (Kolykhalov et al. 1996; Tanaka et al. 1996; Blight and Rice 1997; Ito and Lai 1997). The results of these studies provide strong functional evidence for the existence of this structure and for an essential role of the structure in replication of the viral RNA. The two compensatory mutations present in the mutant 19C-28G were compatible with continued replication of a subgenomic replicon, whereas either of these mutations was lethal to replication when introduced individually (Fig. 6 ). In addition, the dual SL1-A/AA mutant, which contains compensatory block substitutions in the lower half of the duplex stem of SL1, was able to transduce the selection of G418-resistant colonies, whereas individual SL1-A or SL1-AA mutants did not have this capacity (Figs. 5, 7 ). These data provide strong functional evidence for a structure that up until now has been supported only by computer modeling of the RNA sequence, and by physical mapping of ribonuclease cleavage sites and sites of chemical modification within synthetic RNA.

The fact that point mutations within the central region of the duplex stem (at nucleotides 10, 13, and 15 and 33, 35, and 38, from the 3′ end of the RNA) had only small effects on replication (Fig. 8 ), coupled with the ability of the SL1-A/AA revertants to replicate in En5-3 cells (Fig. 7 ), suggests that there is relatively little requirement for specific nucleotide sequence within this part of the hairpin structure, as long as the structure is maintained. The lack of effect of the single base mutations also suggests that it is not necessary to have a perfect duplex structure in this part of the stem. In contrast, at either end of the stem, specific base sequences and/or structure appear to be much more important. It was not possible to recover G418-resistant cells following transfection of the double SL1-B/BB (Fig. 5A ) mutant. RNA replication was also abrogated by individual 19C and 28G mutations at the penultimate base positions near the apex of the stem (Fig. 6 ), or 21A, 22G, and 26C mutations adjacent to the stem but within the loop sequence (Fig. 3 ). In contrast, the distal nucleotides within the loop (Fig. 3 , 23A, 24U, 25A) could be altered without loss of replication competence.

These results are consistent with the location of natural nucleotide substitutions that have been found to be present within the SL1 sequence of other strains of HCV. Although there is only limited diversity within this highly conserved sequence, an analysis of GenBank submissions shows that certain bases are absolutely conserved, including the 21U, 22C, and 26G bases, as well as 19G and 28C (Fig. 10 ). Base substitutions that are present in other strains are generally located within the central part of the duplex stem, particularly near 7A and 30U, which are looped out of the duplex in the dominant sequence, or at the apex of the loop. It is noteworthy that the sequence producing the most stable predicted SL1 structure is that which is present in the replicon we studied (Fig. 10 ), and which constitutes the large majority of sequences present in GenBank.

FIGURE 10. Natural genetic variation within the SL1 sequence. (A) MFOLD-predicted structure of the SL1 sequence present in Ntat2ANeo(SI). Shaded regions indicate regions of critical sequences/structure as determined in this study. A BLASTN search of this 46-nt sequence (more ...)

These observations lead us to suggest that the structure present at the junction of the stem and the loop is critically important for replication, possibly because it plays an essential role in recognition of the 3′ end of the RNA by the viral replicase complex. The critical nature of this structure is further supported by reversion of the 21G and 26U mutations at the base of the loop in replicating RNAs that were present in the G418-resistant cell lines selected following transfection with the 21G-26U mutant (Fig. 4 ). It is interesting that there were second-site mutations in some of these revertant RNAs, including a base change within the putative loop sequence of a second stem-loop (SL2) that has been proposed to form just upstream of SL1 (Fig. 4 ; Kolykhalov et al. 1996; Blight and Rice 1997). The presence of these additional mutations raises the possibility that the repair process involved the evolution of intermediate mutants, with replication capacity greater than the 21G-26U mutant but less than the parental sequence, and that these sequences may thus be involved in complex tertiary interactions that have yet to be deduced, but that are important for RNA replication.

We also found that the bases present at the 3′ terminus of the RNA had an important influence on the replication capacity of subgenomic RNAs. The wild-type U was strongly preferred at the 3′ end of the RNA (Fig. 8 ). This preference cannot be attributed simply to the need to maintain base pairing in the stem. Although a C at this position would be predicted to form a more stable base pair with G46, RNA containing a 3′ terminal C substitution replicated significantly less efficiently than RNA possessing a 3′ terminal U (Fig. 8C ). Moreover, the 3′ terminal C was rapidly replaced with U as the RNA replicated in the transient assay (Fig. 8A ). The data also suggest a preference for G at the second base position from the 3′ end. As shown in Figure 6 , a mutant in which 2G was substituted with C failed to replicate in the transient assay, even when a compensatory G mutation at 45C was introduced to maintain base pairing. The inability of the dual 2C-45G mutant to replicate cannot be attributed to a need to maintain a C at the 45th nucleotide from the 3′ end, however, because revertant RNAs selected from the pseudo-lethal SL1-A/AA mutant were capable of relatively efficient replication despite substitutions throughout this part of the stem, including the 45G substitution itself (Fig. 7 ). These data thus support the notion that there is a strong preference for the 5′ … GU-3′ dinucleotide sequence that normally resides at the 3′ end of HCV RNA. It is interesting to note that M. Nomaguchi, M. Ackermann, S. You, and R. Padmanbhan (in prep.) have recently shown that the dengue virus replicase also has a strong preference for a 3′ terminal U in the RNA template, suggesting that this preference may be broadly conserved among the Flaviviridae.

Shim et al. (2002) reported recently that purified recombinant HCV NS5B polymerase could initiate RNA transcription de novo in an in vitro reaction at the first, second, or third bases from the 3′ end of short oligoribonucleotide templates. Using template oligoribonucleotides that were identical to the 3′ terminus of HCV RNA except at the most 3′ base, they found that the efficiency with which the polymerase initiated transcription at the 3′ terminus was greatest when the terminal nucleotide was U, and progressively less when it was G, A, or C. These results agree with the preference of the HCV polymerase for a 3′ terminal U, but they differ from our results in the order of the relative efficiencies of other bases at this location. In the more complex in vivo replicon system that we studied, the order was U, C, A, and G. It is noteworthy that the 3′ terminal bases are 5′- … GU-3′ in all HCV genomic RNA sequences reported to date, whereas the predicted 3′ terminal bases of the negative strand intermediate RNA are 5′- … GC-3′.

The difference between the order of the efficiency of transcriptional initiation with various 3′ terminating bases in vitro (Shim et al. 2002) and in vivo, as reported here, may reflect the incomplete nature of the in vitro reaction studied by the former group. Nonstructural proteins other than NS5B were not present in the reaction mix, although they are likely to be part of the native replicase complex (Egger et al. 2002). Moreover, the C-terminal residues had been removed from the recombinant NS5B polymerase used in the in vitro reaction to enhance its solubility (Shim et al. 2002), and RNA transcription did not take place at the surface of intracellular membranes as it is likely to do in vivo (Egger et al. 2002). The incomplete nature of the in vitro reaction may also explain why transcription was abortive in this system, leading to few reaction products beyond dinucleotides (Shim et al. 2002). Despite these differences, it is important to note that results from both the in vivo and in vitro systems indicate that the 3′ terminal nucleotide sequence of the template is very important to the viral replicase.

The results presented here provide strong evidence that the SL1 stem-loop plays an important role in replication of HCV RNA, most likely by facilitating the recognition of the 3′ end of the genomic RNA by the viral replicase complex. The data suggest that two subdomains located at opposing ends of the structure are particularly critical for this process (the region at the junction of the stem and the loop, and the 3′ terminal nucleotides), whereas the requirements for the intervening duplex stem are less stringent (Fig. 10 ). The molecular basis for these requirements remains unknown. It is possible that the 3′ terminal bases interact in a specific fashion with the active site of the replicase in order to achieve the initiation of transcription of negative-strand RNA at the 3′ end of the genome. This requires further study, but it is an attractive hypothesis. If it is correct, then the apical part of the stem-loop, specifically the structure at the junction of the duplex stem and the loop, may be involved in RNA–RNA or RNA–protein interactions that facilitate this process.

The plasmid pU-BciVI is a modified version of pNtat2ANeo(SI) that contains a BciVI restriction recognition sequence downstream from the HCV sequence, permitting the runoff transcription of synthetic RNAs with 3′ termini that are identical to the authentic HCV 3′ terminus. It was constructed using Quick-Change (Stratagene) mutagenesis, first introducing the BciVI site downstream from the HCV sequence in pHCV-NΔCla. Two BciVI sites that were located in the Neo and 4B coding regions (nt 5657 G to C, HCV-N) of pNtat2ANeo(SI) were similarly removed by site-directed mutagenesis without changing the amino acid sequence of the polyprotein. pU-BciVI was subsequently assembled by a three-way ligation of fragments containing each of these changes. To construct pC-BciVI, pA-BciVI, and pG-BciVI, PCR was carried out with primer pairs in which the forward primer contained the ClaI site located within the NS5B-coding sequence and the reverse primer contained different 3′ terminal nucleotides followed by the XbaI site. The PCR amplimers were subsequently ligated into pU-BciVI at the ClaI and XbaI sites. The DNA sequences of the manipulated regions were verified in the final construct before its use in transcription reactions for replication assays. Where indicated, additional ΔGDD mutations that are lethal for replication were constructed in these genetic backgrounds by QuickChange mutagenesis using primer pairs that remove 10 amino acids from the NS5B sequence (Yi and Lemon 2002).

This work was supported in part by grants from the National Institute of Allergy and Infectious Diseases (U19-AI40035) and the Texas Advanced Technology Program (004952-0027-2001). We thank Frank Scholle and Kui Li for their critical review of the manuscript.

The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked ‘‘advertisement’’ in accordance with 18 USC section 1734 solely to indicate this fact.