Comparative Analysis of Translation Efficiencies of Hepatitis C Virus 5′ Untranslated Regions among Intraindividual Quasispecies Present in Chronic Infection: Opposite Behaviors Depending on Cell Type

Abstract

Hepatitis C virus (HCV) RNA translation initiation is dependent on the presence of an internal ribosome entry site (IRES) that is found mostly in its 5′ untranslated region (5′ UTR). While exhibiting the most highly conserved sequence within the genome, the 5′ UTR accumulates small differences, which may be of biological and clinical importance. In this study, using a bicistronic dual luciferase expression system, we have examined the sequence of 5′ UTRs from quasispecies characterized in the serum of a patient chronically infected with HCV genotype 1a and its corresponding translational activity. Sequence heterogeneity between IRES elements led to important changes in their translation efficiency both in vitro and in different cell cultures lines, implying that interactions of RNA with related transacting factors may vary according to cell type. These data suggest that variants occasionally carried by the serum prior to reinfection could be selected toward different compartments of the same infected organism, thus favoring the hypothesis of HCV multiple tropism.

The hepatitis C virus (HCV) 5′ untranslated region (5′ UTR), is 341 bases long and is the most highly conserved region of the virus genome among various genotypes (26), suggesting that it plays a key role in the viral cycle and may be a potent target for antiviral agents. It is now clear that initiation of translation of HCV RNA occurs by a cap-independent mechanism mediated by an internal ribosome entry site (IRES) (29, 30) that comprises most of the 5′ UTR and extends at least to the first 12 to 30 nucleotides (nt) of the coding sequence (14, 19). Most of the studies based on the model of the secondary structure of HCV 5′ UTR that was first proposed by Honda et al. (8) and recently refined (7) have demonstrated that the highly ordered structures within IRES elements are absolutely required for IRES activity (reviewed in reference 20).

Data have accumulated from mutation-deletion analyses (3, 7, 9, 18, 27) and in vitro reconstitution of IRES-mediated initiation complexes (17, 25) that were performed to gain insight into the control of viral translation initiation. Reports on comparisons between IRES efficiencies from different HCV genotypes are conflicting (2, 4, 11, 22). Like many other RNA viruses, HCV has a very high mutation rate and circulates as a population of closely related genomes, referred to as quasispecies (5). At present, little is known about 5′ UTR diversity in a viral population and its dynamics toward viral multiplication.

In this work, we studied authentic, biologically derived HCV 5′ UTRs isolated from human serum to assess whether sequence heterogeneity between IRES elements can be linked to changes in their function. Our data indicate that the HCV 5′ UTR, even if it is the most highly conserved part of the viral genome, has a quasispecies distribution with minor modifications in its sequence. These modifications result in dramatic changes of the IRES activity depending on the cell type used for transfection.

Construction of the pIRF bicistronic vector.

HCV IRES activity was monitored with the aid of a bicistronic reporter vector, pIRF, under the control of a T7 promoter, composed of firefly luciferase (FLuc) followed by the HCV 1a 5′ UTR sequence and then by Renilla luciferase (RLuc) (Fig. 1). In such a system, the upstream (control) reporter FLuc is translated in a cap-dependent fashion whereas the downstream (assay) reporter RLuc is under the control of HCV IRES. The primers used for constructions are detailed in Table 1. We designed three artificial mutants of the wild-type 5′ UTR aimed at testing the accuracy of our bicistronic system since they were expected to modify HCV IRES activity. These constructs, displayed in Fig. 1, were named pIRFΔ20 (lacking the first 20 nt of the HCV sequence), pIRFΔC (lacking the 30 nt coding for the capsid), and pIRF+8 (containing the additive 8-nt sequence recently found in an Australian isolate) (28).

FIG. 1.

Structures of the pIRF bicistronic reporter plasmid and its variant constructs. Hatched boxes indicate the two luciferase genes: RLuc, Renilla luciferase; FLuc, firefly luciferase. The reference HCV sequence is shown in black boxes, with the AUG initiator codon represented as a small empty box; lines indicate deletions, and the empty box upstream of nt +1 in the pIRF+8 construct denotes the extra 8-nt sequence (28). Nucleotide positions referring to HCV 1a (13) are shown below the constructs. T7, T7 promoter sequence. The relative in vitro translation efficiency of each bicistronic RNA variant, shown on the right, was expressed as the ratio of RLuc to FLuc enzymatic activities, normalized to that of pIRF reference, which is defined as 100%.

TABLE 1.

Oligonucleotides used for PCR, RT-PCR, and sequencing

Primer	Sequence (5′→3′)a	Orientation	Location on HCV genome	Use
IRES 5′Δ20	gcaccggatccGACACTCCACCAT	Sense	20→33	PCR
IRES 5′	cgccggatccGCCAGCCCCCTGATG	Sense	1→15	PCR, seminested PCR
IRES 5′+8	cgccggatccCCCCCCCCAGCCAGCC	Sense	−8→7	PCR
IRES 3′	gcgccctgcagTTTTCTTTGAGGTTTAGG	Antisense	353→371	PCR, seminested PCR
IRES 3′ΔC	gcgccctgcagGTGCACGGTCTACG	Antisense	327→340	PCR
Quasi 5′	GCCAGCCCCTGTTGGGGG	Sense	1→18	PCR
Quasi 3′	AGTTCCCCGGGTGGCGGTC	Antisense	409→426	PCR, RT-PCR, sequence
SeqpIRF 5′	AAACTCGACGCAAGAAAA	Sense	—b	Sequence

^aExtra nucleotides from the HCV sequence are shown in lowercase letters. Nucleotides forming restriction sites used for cloning are underlined; start and stop codons are in boldface type. 

^b—, Nucleotide sequence located in FLuc,, upstream of HCV IRES in pJKF construct. 

We next examined in vitro IRES translational efficiency toward the RLuc gene in different constructs in the rabbit reticulocyte lysate (RRL) system, using the TNT coupled reticulocyte lysate system kit (Promega) for transcription-translation with T7 RNA polymerase. When analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, ³⁵S-labeled FLuc and HCV capsid-RLuc fusion proteins migrated to the expected sizes of 62 and 37 kDa, respectively (data not shown). The enzymatic activities of the luciferases were measured by using the dual luciferase kit assay (Promega). By means of this assay procedure, HCV IRES activity can be analyzed in vitro and in vivo, with both reporter enzymes being assessed in the same preparation. Another important advantage of such a bicistronic construct is that during the transfection procedure the upstream translation Fluc product appears as an internal control, which bypasses the possible differences in transfection efficiency. IRES relative translation efficiency was calculated as the ratio of two luciferase activities (RLuc/FLuc), and the relative activities of mutated constructs were compared to that of the parental HCV 5′ UTR, which was arbitrarily taken as 100% (Fig. 1). As expected, deletion of the first 30 nt from the HCV ORF profoundly impaired IRES activity (by 87%). Although conflicting, the hypothesis that IRES activity requires the extreme 5′ capsid coding sequence was verified in our system, in agreement with previous studies (14, 16, 19). Surprisingly, in contrast to some other reports (reviewed in reference 20) except for one earlier study (6), the pIRFΔ20 construct appeared less effective than the parental construct (it had 57% of the parental activity). Such a discrepancy could be explained by a possible intervening effect of the 3′ end of the Fluc coding sequence upstream of the HCV 5′ UTR. Interestingly, the pIRF+8 construct had a slightly higher efficiency than pIRF. Although it has not been shown to be a prerequisite for the full-length HCV RNA to be functional in order to initiate infection in the chimpanzee (13) and it is not considered part of the HCV IRES, the extra 8-nt sequence might be involved by means of structural interaction with the 20-nt stem-loop I in HCV translation modulation. We have observed identical stabilities of bicistronic RNA templates in RRL after a 30-min reaction by Northern blot analysis with the IRES 3′ oligonucleotide as a probe, ruling out a significant variability of these RNAs that could account for the observed differences in RLuc expression (data not shown).

Characterization of HCV 5′ UTR quasispecies.

Heterogeneity within HCV 5′ UTR was investigated by using a pretreatment serum sample from a 46 year-old man with chronic hepatitis C related to HCV 1a infection. RNA was extracted from 140 μl of serum using the QIAamp viral RNA kit (Qiagen). Then reverse transcription-PCR (RT-PCR) was performed with the Access RT-PCR kit (Promega). Briefly, HCV RNA was first reverse transcribed at 46°C for 60 min with antisense primer Quasi 3′, and cDNA fragments were further amplified by seminested PCR, including the high-fidelity Arrow Taq DNA polymerase (Stratagene), using primers Quasi 5′ and Quasi 3′ for the first round and primers IRES 5′ and IRES 3′ for the second round. The PCR amplifications involved 30 cycles of 94°C for 20 s, 50°C for 1 min, and 72°C for 1 min, followed by a final elongation step at 72°C for 7 min. PCR products were cloned in place of the parental 5′ UTR into the pIRF vector. A total of 43 clones were generated and sequenced in both directions on an ABI 377 automated DNA sequencer (Applied Biosystems, Foster City, Calif.) using the Dye Dideoxy Terminator cycle-sequencing kit (Applied Biosystems). Despite the well-known genetic stability of this region, 15 different variants were found to coexist in the serum sample studied (Fig. 2). When comparing nucleotide sequences to that of pIRF (HCV 1a), we found two common differences. First, GA residues at positions 34 and 35 (specific to genotype 1a sequence) were observed whereas AG residues (specific to genotype 1b sequence) were found in the pIRF reference. Second, a change of A (present at position 204 in the pIRF IRES sequence obtained from chimpanzee liver biopsy specimens) (C. Daener, C. Wychowski, and S. Feinstone, unpublished results) to C (detected in quasispecies isolated from a human serum sample) was observed, in agreement with a previous report (10) which suggested that this position was representative of the tissue distribution. Since Q7 was the prevalent quasispecies detected, it was chosen as the reference variant throughout our study. Most of the sequences presented in Fig. 2 differed from Q7 by 1 nt, although changes of 2 nt (Q2, Q20, Q27, and Q31), 3 nt (Q8, and Q24), and up to 4 nt (Q1) changes can also be noted. Of the 22 nt changes, most were substitutions, except for 2 deletions (nt 55, Q1; nt 126, Q15) and a common C insertion (between nt 126 and 127) for Q24 and Q27. Moreover, all mutations were located within the 5′ UTR, except for the mutation at nt 348 (Q32), which occurred in the coding sequence known to be part of the IRES (19) (Fig. 3). The diversity observed among 43 clones was assumed to reflect the quasispecies distribution of 5′ UTR in the serum sample studied. The hypothesis that this variability might be a consequence of nucleotide misincorporation during RT-PCR has been considered but did not seem valid for several reasons. First, unlike many studies on quasispecies, we took advantage during seminested PCR of a proofreading activity of Taq polymerase, which is expected to reduce the misincorporation rate. Second, if incorporation errors from polymerase were implicated, we should have observed an increment in mutations from a single mutant to multiple mutants, which was not the case. Finally, we can state that in another study in progress (data not shown) with the same system, almost the same mutations have been detected in two independent experiments, each one using a new RNA template. All these points argue in favor of the reliability of our sequence data and naturally occurring variability and quasispecies balance coevolving in the patient at the sampling time point.

FIG. 2.

Alignment of nucleotide sequences of the IRES region in the quasispecies characterized in the serum of a patient infected with the HCV 1a genotype relative to plasmid pIRF, whose complete sequence is given in the first line; only differences in nucleotide composition are indicated. Nucleotide numbers refer to HCV 1a (13); X, deletion from 1 nt; , C insertion; brackets, frequency of the clone among the whole population of 43 clones sequenced.

FIG. 3.

Scheme of the secondary structure of the HCV IRES, showing the locations of the nucleotide mutations in IRES elements of sequenced quasispecies studied. The secondary structure prediction and loop numbering are based on those of Honda et al. (7); the initiator AUG codon in stem-loop IV is circled, and the coding sequence is represented by a dotted line. All quasispecies mutations are depicted by arrows preceded by ● for substitution, ∖○ for deletion, and ○ for insertion.

In vitro translational efficiency of 5′ UTR quasispecies.

To analyze a possible correlation between the 5′ UTR sequence diversity observed in the patient's serum and IRES activity, each of the isolated quasispecies was first subjected to in vitro translation as described above. As shown in Fig. 4, a great heterogeneity was observed in translation efficiencies, varying from 3.4 to 93.6% relative to pIRF. Q7 was the most efficient, but on the whole, the activity was independent of the number of additional mutations detected. Indeed, Q1 (four mutations) was more effective than Q12 and Q22 (one mutation), suggesting a more critical role for the nucleotide location (nt 301 and 266) for IRES activity than for the number of mutations. Base-pairing disruption resulting from a substitution at nt 301 within the pseudo-knot organization of domain IIIe might explain the severe loss of IRES function. Interestingly, the G-to-A substitution at nt 266 (Q22), located within the loop of domain IIId, had the most drastic effect on RLuc expression. This observation fits the results of a recent study highlighting the importance of nucleotides contained in the hairpin structure of stem-loop IIId in modulating the HCV 5′ UTR tertiary folding structure required for functioning (12). Despite the presence of three changes, Q7 exhibited a marginal reduction compared to pIRF in its capacity of translation (93.6%), indicating that mutation from AG to GA at positions 34 and 35 had no influence on translation efficiency. Other changes resulting in intermediate RLuc expression either were located in loops or did not strongly affect base-pairings within the corresponding stem structures, as was the case for deletion at nt 126 (Q15) and insertion between nt 126 and 127 (Q24 and Q27) in the pyrimidine tract (nt 120 to 130) with relative efficiencies of 79.8, 41, and 62.5%, respectively. In agreement with a previous report (31), these mutations, while inducing a slight change in the surrounding sequences, were not sufficient to impair IRES function. This could imply, as mentioned above, that the secondary structure of IRES is not the only parameter for optimal activity but that tertiary folding must also be considered. The only mutation observed in the capsid coding sequence, located at nt 348 (Q32), resulted in an A-to-G substitution, which was predicted to induce a less stable base-pairing from UA to UG within domain IV (Fig. 3). Rather than the expected enhancement of IRES efficiency (8), a significant decrease was observed, which emphasizes the requirement of the first capsid coding nucleotides to ensure HCV IRES function (19).

FIG. 4.

Relative efficiencies of IRES elements of different quasispecies in vitro. Bicistronic plasmids corresponding to the indicated quasispecies were used to program in vitro transcription-translation in the TNT reticulocyte lysate system. IRES relative activity was then assessed by measuring the ratio of the expressed RLuc to FLuc by using the dual luciferase kit assay. Relative activities were normalized to the pIRF reference construct, whose ratio was arbitrarily taken as 100%. Data shown represent means from two independent experiments with a variation among the different experiments of less than 10%.

In vivo translational efficiency of 5′ UTR quasispecies.

Our constructs were then tested in vivo using three different cell lines to confirm our in vitro findings in a system having conditions closer to those existing during viral replication. Indeed, the HCV IRES has been extensively reported to bind a variety of cellular factors that might be absent from the RRL system. For that purpose, we selected the following: pIRF as a reference; Q7 representing the consensus quasispecies; and Q2, Q12, Q22, and Q31 representative of IRES quasispecies impaired in in vitro activity. Three cell lines were used: Vero cells (kidney cells derived from the African green monkey), HepG2 cells (human cells of liver carcinoma origin), and Jurkat cells (human cells of lymphocyte origin). For transfections, an appropriate number of cells were seeded in 24-well plates. The next day, the cells were infected with the vTF7-3 recombinant vaccinia virus expressing T7 RNA polymerase (kindly provided by B. Moss, National Institutes of Health, Bethesda, Md.) at 5 PFU/cell in 300 μl of serum-free Opti-MEM medium (Gibco-BRL) for 1 h at 37°C. The cells were then transfected with 1 μg of plasmid DNA mixed with 5 μg of DAC-30 (Eurogentec) in 300 μl of Opti-MEM medium. At 16 to 18 h posttransfection, the cells were harvested and lysates were assayed for luciferase activities as described above. The results of these experiments are summarized in Fig. 5A. As in the RRL assay, the IRES activity of Q12 and Q22 was dramatically reduced in the three types of cells tested, confirming the crucial role of nt 266 and 301 for this function. Surprisingly, a considerable disparity of IRES efficiency was found for the four other clones tested with regard to the cell line used. These observations were highly reproducible and argue in favor of a biological difference linked to the nature of the IRES sequence and its capacity to promote internal initiation of translation in vivo, depending on the cell line used for transfection (1, 4, 22). To verify that variations in RLuc expression observed with IRES variants in transfected cells actually reflect different translational capacities, we investigated the stability of corresponding transcripts in different cell lines by Northern blot analysis (Fig. 5B). The results indicate a comparable amount of RNA in transfected HepG2 and Jurkat cells, ruling out the notion that the observed variations in RLuc expression could be due to differences in transcription or stability of the RNA transcripts. Nor could they be attributed to differences in transfectability of cells, this factor being abrogated by the use of a bicistronic system. Our results support the emerging view that HCV translation might be dependent on the interaction with cellular factors distributed differently among cell types. A noteworthy observation was the evidence of opposite patterns of IRES activity depending on the cell type used for the RNA transfection assay. Although showing a high efficiency in HepG2 and Vero cells, pIRF and Q7 5′ UTRs were much less active in Jurkat cells. The opposite was observed for Q2, Q22, and Q31 (the low level of activity of Q12 did not permit any interpretation in that context). In light of these data, it is tempting to distinguish between nonlymphoid (pIRF and Q7) and lymphoid (Q2, Q22, and Q31) optimal IRESs. These observations strongly indicate that the contribution of the nucleotide sequence of the HCV 5′ UTR relative to IRES function differs according to the cellular system used, suggesting that the interactions between the highly ordered HCV IRES structure and related host factors are cell type specific. Moreover, we have noted that for the former, the activity ratio between Vero and HepG2 cells was always low (ratio, <1), whereas for the latter, the activity was always higher in Vero cells (ratio, ca. 2).

FIG. 5.

Relative translational efficiencies of IRES elements of different quasispecies in vivo. (A) Relative levels of RLuc expression from various quasispecies tested. Three different cell lines were used for transfection: HepG2, Vero, and Jurkat. Cell lysates were prepared 16 to 18 h posttransfection and assayed for FLuc and RLuc activities as described for the in vitro experiment. Relative efficiencies of different IRES quasispecies were measured by the RLuc/FLuc ratio. For each construct, experiments were performed in triplicate wells, and standard deviations were calculated from the data obtained for these wells. (B) Northern blot analysis of bicistronic RNAs extracted from HepG2 or Jurkat cells transfected for 18 h with the indicated constructs: pIRF, Q7, and Q12. IRES 3′ oligonucleotide was used as a probe. As controls, untransfected Jurkat cells (MOCK) and pIRF in vitro transcript (WT) (indicated by an arrow) were used.

In conclusion, we have shown that HCV 5′ UTR has a quasispecies distribution in a given infected individual and we have been able, to our knowledge for the first time, to demonstrate that the naturally occurring sequence diversity of IRES elements leads to important changes in their ability to direct cap-independent translation. Such differences observed in vitro were confirmed although differently modulated in in vivo assays, depending on the transfected cell type. Because viral particles in serum are thought to be released from the liver but also from other compartments of the organism such as peripheral blood mononuclear cells (21, 23, 24), the observed diversity within 5′ UTR might reflect the existence of various HCV IRES sequences targeting RNA translation specifically to the liver or extrahepatic compartments. In accordance with that view, initiation of protein translation may appear as one rate-limiting factor for viral replication. It will be of interest to assess HCV 5′UTR polymorphism due to viral tropism in different parts of the organism and its impact on the selection of replicative variants. Moreover, the bicistronic vector designed in this work presents a unique feature in addition to its accuracy and reproducibility. Indeed, in contrast to several reports on IRES efficiency, we have conserved the entire HCV 5′UTR in our construct in order to preserve a possible influence of cis non-IRES elements on the final secondary and/or tertiary structures of IRES. In the perspective of correlating differences in the activity of IRES sequences with pathogenesis of HCV infection, the system used in our study provides a useful tool for structure-function analyses of HCV IRES. So far, most of the studies on the dynamics of HCV quasispecies have been conducted on variable regions of the HCV genome such as E2 hypervariable region 1 or parts of the NS5A protein and thus have been limited to sequence comparison without any functional investigation. To date, although little work has been undertaken in that direction, the HCV 5′UTR appears to be an element of choice for such an approach. Contradictory conclusions have been proposed concerning a possible correlation between the sequence variability found for HCV IRESs of different genotypes and a response to interferon therapy (15, 22, 32). Currently, work is under way to study HCV 5′ UTR diversity displayed in a viral quasispecies population and its dynamics toward the viral life cycle under different biological pressures.