Up to 3% of the world's population are estimated to be infected with hepatitis C virus (HCV). In 4 of 5 cases, the disease progresses into chronicity. Although escape from adaptive immunity is key to long-term persistence, evasion of innate antiviral responses is crucial in establishing a persistent infection in the first place.1 As a main player in innate immunity against viruses, the interferon (IFN) system has a key role in curtailing pathogens by putting infected and neighboring cells into an antiviral state.2 Also, HCV has been known for a long time to be highly sensitive to treatment with type I IFNs,3 and still today IFN-α is the major component of HCV therapy. The fact that HCV is sensitive to IFN yet it manages to establish persistent infections suggested that the virus may have evolved mechanisms to circumvent detection by the innate antiviral defense system. In fact, the viral protease, nonstructural protein (NS) 3/4A, has been described to suppress activation of IFN regulatory factor 3 (IRF-3), a central antiviral transcription factor promoting the production of IFN-β and a plethora of IFN-stimulated genes (ISGs), most prominently ISG56,4 in response to viral infection or treatment with double-stranded ribonucleic acid (dsRNA).1, 5 Interestingly, identification of cellular targets of NS3/4A indicated that HCV interferes with both, Toll-like receptor 3 (TLR3)–mediated signaling through cleavage of the essential adaptor TRIF (Toll/interleukin receptor domain-containing adapter-inducing interferon β),6 as well as with TLR-independent activation of IRF-3 by the cytosolic pathways of retinoic-acid inducible gene I (RIG-I) and MDA5 (melanoma differentiation associated protein 5) by inactivating the signaling adaptor Cardif (also known as interferon-beta promoter stimulator 1 [IPS-1], mitochondrial anti-viral signaling protein [MAVS], or virus-induced signaling adaptor [VISA]).7 This may indicate that both pathways play a role in controlling HCV infection, most likely, however, in different sets of cells, because TLR3 has been shown to be essential for IFN production in plasmacytoid dendritic cells, whereas RIG-I triggers IFN secretion on viral infection in conventional dendritic cells and other tissues.8 The human hepatoma cell line Huh-7, currently the only cell culture model robustly supporting the full life cycle of HCV,9 has been shown to be virtually devoid of TLR310 but does express RIG-I.10, 11 Consequently, Huh-7 cells do not respond to dsRNA, such as the synthetic oligonucleotide duplex poly(I:C), when added to the culture medium10; however, they do respond to transfected poly(I:C) or infection with the paramyxovirus Sendai virus (SeV),5 a known potent inducer of RIG-I. Moreover, a dominant negative (dn) mutation in the caspase recruitment domain (CARD) homology domain of the RIG-I protein (T55I) has been identified in a Huh-7 clone, termed Huh7.5,12 and has been implicated with the cells' outstanding permissiveness for HCV replication.5, 11 In the current study, we examined the general impact of host IRF-3 response onto permissiveness for HCV replication. We assessed the effects of functional inhibition of the RIG-I/IRF-3 pathway on HCV replication and, conversely, the impact of restored RIG-I signaling in Huh7.5 cells. Finally, we quantitatively analyzed the induction of RIG-I signaling by different ribonucleic acids (RNAs), including full-length HCV RNA, and came up with a refined model of how HCV manages to establish persistent replication in cells despite the presence of highly sensitive pathogen detection systems.
Hepatitis C virus (HCV) has been known to replicate with extremely varying efficiencies in different host cells, even within different populations of a single human hepatoma cell line, termed Huh-7. Several reports have implicated the retinoic-acid inducible gene I (RIG-I)/ interferon regulatory factor 3 (IRF-3) pathway of the innate antiviral response with differences in host cell permissiveness to HCV. To investigate the general impact of the IRF-3 response onto HCV replication in cell culture, we generated an ample array of stable Huh-7 cell lines with altered IRF-3 responsiveness. Neither blocking IRF-3 activation in various host cells by expression of dominant negative RIG-I or HCV NS3/4A protease nor reconstitution of RIG-I signaling in Huh7.5, a cell clone known to be defective in this pathway, had any impact on HCV replication. Only by overexpressing constitutively active RIG-I or the signaling adaptor Cardif (also known as interferon-beta promoter stimulator 1, mitochondrial anti-viral signaling protein, or virus-induced signaling adaptor), both leading to a stimulation of the IRF-3 pathway in the absence of inducers, was HCV replication significantly inhibited. We therefore assessed the extent of RIG-I– dependent IRF-3 activation by different species of RNA, including full-length HCV genomes and HCV RNA duplexes, and observed strong induction only in response to double-stranded RNAs. Conclusion: Based on these findings, we propose a refined model of innate immune escape by HCV involving limited initial induction and stringent subsequent control of the IRF-3 response. (HEPATOLOGY 2007.)
Materials and Methods
Cells and Cell Culture.
All cells were grown as described by Quinkert et al.13 Huh-7 HD refers to the naïve Huh-7 cells from our laboratory, hp or lp indicating a high (>80) or low (<30) passage number, respectively. Huh7-Lunet14 and Huh-7/5-2 (unpublished) are highly permissive cell lines generated by curing stable replicon cell clones by treatment with a specific inhibitor. The highly permissive clone Huh7.512 and its parental cell line Huh-7 WU were provided by Charles Rice. Huh-7 TY are naïve Huh-7 cells provided by Takaji Wakita. HuH6 are unrelated human hepatoblastoma cells supporting lower-level HCV replication.15
HCV Constructs, In Vitro Transcription and Generation of Duplex RNA.
The subgenomic construct Con1/luc refers to pFK-I341-PI-Luc-EI-NS3-3′/ET/δg,16 genome length constructs Jc1 and Jc1/luc refer to pFK-J6/C317 and pFK-Luc-Jc1.14 Nonreplicating construct Con1/lucΔGDD has a deletion of the GDD motif in the polymerase.18 The construct Con1ΔGDD refers to the full-length Con1 genome (accession AJ238799) bearing the GDD deletion. This construct was linearized for in vitro transcription using SpeI and HindIII for positive and negative strand transcripts, respectively. In vitro transcription was performed as described previously.16 Briefly, linearized plasmid deoxyribonucleic acid (DNA) was purified by phenol and chloroform extraction, followed by ethanol precipitation. DNA was then transcribed using T7 (positive strands) or T3 (negative strands) RNA polymerase (Promega, Mannheim, Germany) under the exact reaction conditions given in Binder et al.16 Transcription was terminated by 1 U DNase (Promega) per microgram plasmid DNA, and transcribed RNA was extracted using acidic phenol and chloroform, followed by isopropanol precipitation. HCV duplex RNA was generated by combining equimolar amounts of Con1ΔGDD-positive and Con1ΔGDD-negative strand transcripts, adding 50 mM sodium chloride, heating to 98°C, and cooling down to 10°C at 0.1°C/second.
Electroporation and Determination of HCV Replication Efficiency.
Cells were electroporated with Con1/luc in vitro transcripts, seeded in single or duplicate wells and lysed after 4, 24, 48, and 72 hours as described previously.16 Luciferase reporter activity was determined as a correlate of HCV RNA abundance.18 Luciferase activity at 4 hours represents initial translation of the RNAs in the absence of replication and was used to normalize the data for transfection efficiency. Normalized relative light units (RLU) at 48 hours (48 hours/4 hours) are referred to as “replication efficiency.” Values are given as the mean of duplicate measurements per well in 1 representative experiment. Error bars represent the standard deviation of the mean. All replication data are displayed on a logarithmic scale.
Generation of Virus Stocks and HCV Infection Assays.
Generation of virus stock and reporter virus assays were done as in Koutsoudakis et al.14 Reporter virus infections were harvested 24, 48, and 72 hours after infection. Values are given as the mean of duplicate measurements of 3 independent wells in 1 representative experiment. Error bars indicate the standard deviation of the mean. All infection data are displayed on a logarithmic scale.
All genes were cloned into pWPI-GUN, a derivative of the bicistronic lentiviral vector pWPI (gift from Didier Trono) replacing green fluorescent protein in the reporter cistron by a green fluorescent protein–ubiquitin-neomycin phosphotransferase fusion gene. NS3/4A refers to amino acids (aa) 1027 through 1711 of the Con1 genome. Myc-RIG-I wild-type (wt) and constitutively active myc-RIG-I ca (aa 1-229) were subcloned from pcDNA3.1/Zeo/Myc-RIG-I and pcDNA3.1/Zeo/Myc-ΔRIG-I (provided by John Hiscott). Myc-RIG-I dn was cloned by polymerase chain reaction amplification of the last 570 codons (aa 735-925)19 of RIG-I wt, introducing the myc-sequence via the upstream primer. FLAG-Cardif was cloned from pCR3.V64-Met-FLAG-Cardif,7 cleavage-resistant FLAG-CardifC508R was generated by site-directed mutagenesis. Cloning details are available on request.
Stable Transduction of Cells.
Stable cell lines were generated by lentiviral transduction. Retroviral particles were prepared in 293T cells as described previously,14 using pWPI-GUN vectors and packaging constructs pCMVΔR8.91 and pMD.G (provided by Didier Trono). Target cells were consecutively transduced thrice with filtered supernatants and immediately selected for high transgene expression by addition of 750 μg/mL G418 for 3 days (1,000 μg/mL thereafter).
Immunoblot Analysis of Proteins.
Gel electrophoresis and Western blotting was performed exactly as described previously.13 NS3, Cardif, and myc-RIG-I wt were detected on sodium dodecyl sulfate 10% polyacrylamide gel electrophoresis using an NS3 rabbit polyclonal serum,13 Cardif rabbit polyclonal (ALX-210-929, Axxora, Grünberg, Germany) and mouse monoclonal α-myc (9E10, Santa Cruz, Heidelberg, Germany), respectively. Myc-RIG-I dn and myc-RIG-I ca were detected on sodium dodecyl sulfate 12% polyacrylamide gel electrophoresis.
IRF-3 Activity Reporter Assays.
Cells were seeded in 24-well plates (1 × 105 cells/well), 24 hours later they were cotransfected with the ISG56 promoter firefly luciferase construct pGL3B/5614 (gift from Ganes Sen) and Renilla luciferase pRL-SV40 (Promega) (ratio 3:1) using Effectene (Qiagen, Hilden, Germany). After 6 hours, cells were infected with 100 hemagglutination units per milliliter SeV in phosphate-buffered saline containing 0.3% bovine serum albumin for 1 hour at room temperature; mock control was phosphate-buffered saline/bovine serum albumin without virus. Alternatively, cells were stimulated by transfection of defined RNAs using 0.1 μg sample RNA plus 0.9 μg poly(C) (Sigma, Munich, Germany) per well (or no RNA in mock controls) with Lipofectamine2000 (Invitrogen). Sixteen hours after transfection, cells were lysed on plate in 200 μL luciferase lysis buffer14 and frozen. Firefly luciferase was measured as described,14 Renilla luciferase was measured in 20 μL lysate by addition of 100 μL 1.5 μM coelenterazine (PJK, Kleinblittersdorf, Germany) in luciferase assay buffer.14 Firefly RLU were normalized to Renilla RLU. Values are expressed as fold-induction over mock-treated cells and are given as the mean of duplicate measurements of triplicate wells. Error bars represent the standard deviation of the mean.
Synthetic RNA Homopolymeres.
Polyinosinic acid and polycytidylic acid (poly(I) and poly(C)) were ordered from Sigma (Munich, Germany). Duplex polyinosinic-polycytidylic acid [poly(I:C)] was generated by combining poly(I) and poly(C).
No Correlation of IRF-3 Response and Permissiveness to HCV in Different Host Cells.
Replication of a given HCV RNA has been reported to vary drastically depending on the type of Huh-7 cells used.12, 18 The underlying mechanisms for these significant differences in replication efficiency, however, were still enigmatic. To approach this question, we first determined the relative permissiveness of a broad panel of cells available in our laboratory by transiently transfecting luciferase reporter replicons (Con1/luc, Fig. 1A) via electroporation. Abundance of viral RNA at different time points was assessed by measuring luciferase activity in cell lysates, which has been shown to correlate well with RNA replication.18 Relative replication efficiencies among defined passages of cells were highly reproducible among different experiments, whereas absolute values might fluctuate because of varying quality of transcripts or other technical parameters. We found Huh-7 cells with a low passage number to be generally least permissive whereas highly permissive cells were found among the higher passages and Huh-7 clones that originally carried a selectable replicon and were cured from HCV using specific drugs or interferon (Fig. 1B). The best clone in our hand was Huh7-Lunet,14 which replicated HCV RNA up to 1,000-fold more efficiently than a low passage of naïve Huh-7 cells (Fig. 1C). Because the high permissiveness of another cell clone, Huh7.5, has been attributed to a defect in RIG-I–mediated IRF-3 activation in those cells,11 we wanted to investigate the general role of the innate antiviral IRF-3 response in permissiveness to HCV. Therefore, we first compared Huh7.5 cells in transient Con1/luc replication assays with their parental cell line, Huh-7 WU, and found a 50- to 100-fold increase in permissiveness (Fig. 2A). Conversely, in ISG56 promoter reporter assays with SeV as an inducer, we saw profound induction of IRF-3 activity only in Huh-7 WU cells, but not in Huh7.5 (Fig. 2A). We obtained very similar results with IFN-β promoter and repeated PRDII reporter constructs; however, the ISG56 promoter turned out to be the most sensitive one. Because the inverse correlation of IRF-3 response and permissiveness was striking in this pair of cells, we assayed a broader set of Huh-7 passages and clones for their capacity to support HCV RNA replication as well as for their IRF-3 activation on SeV infection (Fig. 2B). We observed no correlation of the 2 parameters whatsoever. Taken together, we confirmed remarkably varying capacities of different Huh-7 cells to replicate HCV RNA but could not generally correlate permissiveness with inducibility of the RIG-I/IRF-3 pathway.
Blocking of IRF-3 Signaling Does Not Increase Permissiveness for HCV Replication.
Although we found no overall correlation between IRF-3 responsiveness and host cell permissiveness, this finding did not rule out control of HCV replication by the RIG-I/IRF-3 pathway. To investigate this, we wanted to shut off RIG-I signaling in different host cells and assess the impact on permissiveness. From an array of differently permissive cells, including Huh-7 from various sources (Fig. 3A), we selected a highly permissive clone (Huh7-Lunet), 3 naïve lineages of Huh-7 cells (HD, TY, and WU), as well as HuH6 as an unrelated hepatoma cell line that has been reported to support a low level of HCV replication.15 These cells were stably transduced with a selectable lentiviral vector, either without insert, carrying the NS3/4A protease of HCV (genotype 1b) or a myc-tagged dominant negative variant of RIG-I (RIG-I dn, aa 735-925). In Western blot analyses, we confirmed expression of the transgenes (Fig. 3B). Additionally, in cells expressing NS3/4A, we checked for cleavage of the essential signaling adaptor Cardif and found it to be completely degraded to a truncated form that has been shown to be inactive in RIG-I signaling.7, 20, 21 To functionally confirm the block in IRF-3 activation, we assayed the stable cell lines for ISG56 promoter activity on SeV infection (Fig. 3C). Whereas all cells transduced with the empty vector exhibited significant induction of IRF-3 activity, their respective counterparts expressing RIG-I dn or NS3/4A were virtually unresponsive to SeV, closely resembling the phenotype of Huh7.5 cells. Reasoning that deficient RIG-I signaling might increase a cell's capacity to replicate HCV, we performed transient Con1/luc replication assays (Fig. 3D). However, for none of the 5 different parental cell lines did the block of RIG-I/IRF-3 signaling have any significant impact on the efficiency of HCV RNA replication. Repetition of the experiment in Huh7-Lunet and Huh-7 WU cells using a weakly adapted replicon yielded similar results (Supplementary Fig. 1). If any, only a diminishing increase in permissiveness was observed for Huh-7 WU RIG-I dn, which are derived from the same parental cell line as Huh7.5 and phenotypically resemble them in terms of IRF-3 responsiveness. Speculating that differences in cellular permissiveness, particularly if caused by antiviral responses, might be more pronounced in infection rather than transfection, we also infected the cell lines with a chimeric genotype 2a reporter virus (Jc1/luc). Also in this setting, no rescue in permissiveness was observed on knock-down of RIG-I/IRF-3 signaling (data not shown). These data indicate that the presence of an intact RIG-I signaling pathway does not limit HCV replication in Huh-7 and HuH6 cells.
IRF-3 Activity Is Capable of Controlling HCV Replication.
The inability to rescue HCV replication by blocking signaling to IRF-3 might imply either that HCV prevents activation of the pathway or that IRF-3 activity does not have the capability to control HCV. Potentially, however, the latter could be an artifact attributable to a lack of effector expression in Huh-7 cells. Huh-7 cells are very poor producers of IFN-β,10 and culture supernatant of various Huh-7 lineages infected with SeV did not inhibit replication of highly IFN-sensitive subgenomic HCV replicons (data not shown). To investigate whether IRF-3 promotes the expression of direct effectors in Huh-7 that can control HCV replication, we generated cell lines with increased IRF-3 signaling by reconstituting RIG-I signaling in Huh7.5 cells. Therefore, we stably introduced myc-tagged wild-type RIG-I (RIG-I wt) or a constitutively signaling variant of RIG-I (RIG-I ca, aa 2-229) and confirmed the expression of the transgenes in Western blots (Fig. 4A) and immunofluorescence analysis (Supplementary Fig. 2B, red channel). Expression of RIG-I wt in Huh7.5 cells resulted in a super-inducible cell line: ISG56 promoter activity in absence of stimulation remained at basal levels, whereas SeV infection lead to a 500-fold to 1,000-fold increase in reporter activity (Fig. 4B), which is significantly higher than in the parental Huh-7 WU cells. When we tested the cell line for HCV replication in either Con1/luc subgenomic reporter replicon assays or Jc1/luc infection experiments, we observed no difference to Huh7.5 vector control cells (Fig. 4C, D). The same was true for replication assays using weakly adapted replicons (Supplementary Fig. 2A). A different picture emerged when we expressed myc-RIG-I ca. In these cells, IRF-3 activity was substantially increased even before stimulation (approximately 30-fold) with no significant further induction on SeV infection (Fig. 4B). This pre-induction of the pathway resulted in a reproducible reduction of HCV replication, 2-fold in Con1/luc assays to greater than 5-fold in Jc1/luc infections (Fig. 4C, D).
This demonstrated that IRF-3 activity has the potential to control HCV replication in Huh-7 cells. However, in immunofluorescence microscopy of Jc1-infected Huh75 RIG-I ca, a substantial fraction of cells stained doubly positive for HCV and RIG-I ca (Supplementary Fig. 2B, arrowheads), indicating that NS3/4A protease mediated cleavage of Cardif also might have rendered such pre-induced cells permissive for HCV. This raised the question as to which extent this pathway is capable of controlling HCV replication in the absence of NS3/4A interference. We therefore overexpressed Cardif wt or mutated Cardif C508R (CR), the latter being noncleavable by the viral protease (data not shown7, 20, 21). As a highly permissive cell line with intact RIG-I signaling, we selected the clone Huh-7/5-2 for stable transduction with FLAG-tagged versions of both Cardif variants (Fig. 5A). Overexpression of Cardif wt or CR significantly induced ISG56 promoter activity, which was further increased by SeV infection (Fig. 5B). HCV replication was severely affected by expression of either Cardif variant, inhibiting subgenomic Con1/luc replication up to 100-fold and Jc1/luc infections 10-fold (Fig. 5C, D). Very similar results were obtained in Huh-7 HD hp cells (data not shown).
To examine at a single-cell level the impact of increased Cardif or Cardif CR expression onto permissiveness for HCV, we analyzed Jc1-infected Huh-7/5-2 cells bearing the respective transgene by immunofluorescence microscopy (Fig. 5E). Numbers of infected cells were significantly decreased in cell lines expressing either of the Cardif variants. In the case of Cardif wt, no FLAG signal was detectable in HCV-positive cells. However, judging whether Cardif was cleaved on HCV infection or whether infection only occurred in cells with diminishing transgene levels (Fig. 5E) was not possible. In the case of non-cleavable Cardif CR, only a minority of infected cells with substantial levels of Cardif could be identified (Fig. 5E, yellow arrowheads), whereas in most HCV-positive cells Cardif CR could not be detected (Fig. 5E, white arrowheads), highlighting the inverse correlation between Cardif expression (and therefore IRF-3 activity) and HCV replication.
In our experiments, IRF-3 activation led to induction of antiviral effectors that were clearly capable of controlling HCV replication. However, this was only seen when the pathway was pre-activated as in case of Cardif or RIG-I ca expression, but not when HCV established replication in nonstimulated cells. The latter indicated that either HCV tightly controlled the pathway by cleavage of Cardif very early on, or HCV genomic RNA was not potently enough inducing RIG-I signaling, or both.
Induction of the RIG-I Pathway by Double-Stranded RNA.
In light of the second hypothesis, we were interested in how potently different species of HCV RNA could trigger RIG-I signaling. The unmatched inducibility of our Huh7.5 RIG-I wt cells (Fig. 4) enabled us to quantitatively analyze the extent of IRF-3 activation by different kinds of HCV RNA in Huh-7–derived cells. In ISG56 promoter reporter assays, we first compared Huh7.5 RIG-I wt with vector control cells in their response to transfected RNA. We used no RNA (mock), poly(C), poly(I), and hybridized poly(I:C) as inducers and measured luciferase activity 16 hours posttransfection. Concordant with the SeV experiments (Fig. 4B), vector control cells were not inducible by any of these RNAs whereas RIG-I wt transduced cells responded to poly(I:C) (Fig. 6A). To compare different species of HCV RNA in terms of their propensity to stimulate RIG-I signaling, we transcribed replication-incompetent full-length genomes (Con1/ΔGDD) in positive- or negative-sense orientation and hybridized them to yield double-stranded HCV duplexes (Fig. 6B). In ISG56 reporter assays, we observed 10- to 20-fold induction for both single-stranded HCV variants comparable to induction by a linear, nonviral transcript (beta-actin) (Fig. 6C). In contrast, equimolar amounts of HCV duplex RNA stimulated IRF-3 activation approximately 100-fold (Fig. 6C). Corroborating these findings, we failed to detect any induction of IRF-3 activity on transfection of replication-competent HCV RNA (Supplementary Fig. 3A). However, cotransfection of poly(I:C) together with HCV did induce a strong response (Supplementary Fig. 3A) and had a clear impact on HCV replication (approximately 10-fold reduction after 48 hours; Supplementary Fig. 3B).
Our results suggest that HCV genomes entering a cell only moderately induce the innate antiviral IRF-3 response. In contrast, RNA duplexes that may form during replication have a higher potential to stimulate RIG-I signaling, by which time, however, NS3/4A would have disrupted the pathway (Supplementary Fig. 4). Importantly, because Huh7.5 vector control cells were unresponsive to all tested RNAs and transcriptomic comparison with Huh7.5 RIG-I wt indicated no significant upregulation of any other factor (data not shown), the observed inductions are likely to exclusively depend on RIG-I and suggest a direct involvement of RIG-I in recognition of double-stranded RNA.
Factors determining host cell permissiveness for HCV in vivo and in vitro are still largely elusive. In this study, we assessed the contribution of RIG-I–mediated innate antiviral responses. Although interferon production by infected hepatocytes still is under debate, the only profound model cell line, Huh-7, has been demonstrated to be a poor producer of IFN-β.10 Nevertheless, we have shown that antiviral effectors induced by RIG-I signaling are capable of controlling HCV replication, as seen in cells overexpressing Cardif or, to a lesser extent, RIG-I ca. This is in line with reports that IRF-3 activation can inhibit HCV independently of IFN-β22 and suggests that Huh-7 cells are suitable as a model for the interplay between antiviral responses and HCV replication. Remarkably, however, experimentally altering inducibility of RIG-I signaling did not impact host cell permissiveness at all: neither could we inhibit HCV replication in Huh7.5 cells by reconstituting functional RIG-I nor could we increase permissiveness in any of the tested cell lines by blocking IRF-3 activation. This is in seeming discordance with previous studies. Most evident might be the difference between our findings and the reports of Sumpter and colleagues11 attributing the high permissiveness of Huh7.5 to their defect in RIG-I signaling. Also after careful repetitions of the experiments, using both highly and weakly adapted replicons, we failed to detect a significant decrease in HCV replication in Huh7.5 cells on functional reconstitution of RIG-I. However, the discrepancy with the study of Sumpter and colleagues11 might be attributable to different technical approaches such as transient versus stable RIG-I expression or different HCV replicons. Also, the reciprocal approach, blocking IRF-3 activation in Huh-7 cells, did not alter replication efficiency in our system but has been reported by Saito and co-workers to result in increased permissiveness.19 To account for greatly differing properties of Huh-7 cells from different origins, we have used a selection of cells from various laboratories, including the direct ancestor of Huh7.5, but not the very same cells as Saito and colleagues. Although this may or may not explain the diverging results, both our data and theirs indicate that blocking RIG-I signaling is not sufficient to increase permissiveness of naive Huh-7 to the level of Huh7.5. Also a very recent report corroborates the notion that inactivation of IRF-3 does not detectably increase permissiveness of HCV host cells, even in primary human hepatocytes.23
Given the observed insensitivity of HCV replication toward absence or presence of a functional IRF-3 activating pathway, we have hypothesized 2 conceivable mechanisms to account for these observations: First, translation of the viral protease from incoming HCV RNA might quickly and comprehensively enough counteract the induction of the antiviral response, not allowing effectors to be expressed to adequate levels. This, however, is argued against by our observation that introducing a potent inducer together with HCV RNA did result in substantial stimulation of IRF-3 activity. As a second model, sufficient stimulation of RIG-I by HCV might require RNA duplexes that can form only during replication. At that time, however, a vast excess of NS-proteins, including NS3/4A, has already been produced,13 giving the protease ample opportunity to deplete the cell of Cardif (Supplementary Fig. 4B). This hypothesis is supported by the observation that introduction of replication-competent HCV RNA at no point triggered detectable IRF-3 activity. This fits to our data indicating that HCV genomes, as well as HCV-negative strands, only moderately stimulate RIG-I signaling, whereas duplexes of HCV RNA turned out to strongly trigger IRF-3 activity. That active replication might give rise to potently inducing RNA duplexes is suggested by a very recent publication showing that inhibition of NS3/4A in replicon cell lines markedly triggered expression of ISG56.24 These findings are in line with reports showing RIG-I binding to and being activated by dsRNA11, 19, 25; however, 2 independent publications recently demonstrated 5′-triphosphorylated RNA to be the actual trigger for RIG-I.26, 27 Because in vitro transcripts always bear terminal triphosphates, this explains the induction observed with single-stranded HCV and beta-actin transcripts in our experiments. However, poly(I:C), according to the manufacturer's information, is synthesized from nucleotide diphosphates using polynucleotide-phosphorylase and therefore should not activate RIG-I.27 Because Huh7.5 vector control and Huh7.5 RIG-I wt cells differ solely in RIG-I expression levels, whereas only the latter responded to transfection of HCV duplex RNA and poly(I:C) but not poly(I) or poly(C), our data suggest a direct involvement of RIG-I in recognition of dsRNA, irrespective of 5′-triphosphorylation. This issue very clearly is worth further investigation.
Summarizing the presented data, we can conclusively demonstrate HCV host cell permissiveness, at least in cultured cells, to be independent of the IRF-3–mediated innate antiviral response. According to our model (Supplementary Fig. 4), this is attributable to an elaborate evasion strategy of the virus. First, HCV silently enters cells with only very limited triggering of RIG-I. Then, substantial translation of the viral genome leads to quantitative depletion of the essential signaling adaptor Cardif by the viral protease NS3/4A before the onset of RNA replication. Liberation of duplex RNA in the course of active replication will therefore not be able to activate IRF-3. Based on this model, future therapeutical approaches might consider agonists of IRF-3 activating pathways not only as immunomodulators but also for their direct effect on HCV replication in newly infected cells.
The authors thank Charles M. Rice (New York), Takaji Wakita (Tokyo), Etienne Meylan and Jürg Tschopp (Lausanne), Didier Trono (Lausanne), John Hiscott (Montreal), and Ganes Sen (Cleveland) for providing cells, plasmids, and reagents. We thank Ulrike Herian, Nina Kezmic, and Rahel Klein for excellent assistance. We apologize to all colleagues whose work could not be cited appropriately due to space limitation.