Distinct Poly(I-C) and Virus-activated Signaling Pathways Leading to Interferon-β Production in Hepatocytes*

Innate cellular antiviral defenses are likely to influence the outcome of infections by many human viruses, including hepatitis B and C viruses, agents that frequently establish persistent infection leading to chronic hepatitis, cirrhosis, and liver cancer. However, little is known of the pathways by which hepatocytes, the cell type within which these hepatitis agents replicate, sense infection, and initiate protective responses. We show that cultured hepatoma cells, including Huh7 cells, do not activate the interferon (IFN)-β promoter in response to extracellular poly(I-C). In contrast, the addition of poly(I-C) to culture media activates the IFN-β promoter and results in robust expression of IFN-stimulated genes (ISG) in PH5CH8 cells, which are derived from non-neoplastic hepatocytes transformed with large T antigen. Small interfering RNA knockdown of TLR3 or its adaptor, Toll-interleukin-1 receptor domain-containing adaptor inducing IFN-β (TRIF), blocked extracellular poly(I-C) signaling in PH5CH8 cells, whereas poly(I-C) responsiveness could be conferred on Huh7 hepatoma cells by ectopic expression of Toll-like receptor 3 (TLR3). In contrast to poly(I-C), both cell types signal the presence of Sendai virus infection through a TLR3-independent intracellular pathway requiring expression of retinoic acid-inducible gene I (RIG-I), a putative cellular RNA helicase. Silencing of RIG-I expression impaired only the response to Sendai virus and not extracellular poly(I-C). We conclude that hepatocytes contain two distinct antiviral signaling pathways leading to expression of type I IFNs, one dependent upon TLR3 and the other dependent on RIG-I, with little cross-talk between these pathways.

The innate immune system represents the first line of defense against viral infections in mammalian cells. It senses invading viral pathogens and initiates signaling pathways leading to the induction of protective cellular genes, including type I interferons (IFN-␣ and IFN-␤) 1 and proinflammatory cytokines that directly limit viral replication and also help to shape subsequent adaptive immune responses (1,2). Recognition of conserved molecular structures that are expressed by large groups of pathogens (pathogen-associated molecular patterns, PAMPs) is carried out by specific pattern recognition receptor (PRR) molecules (3,4). The Toll-like receptors (TLRs) are a class of PRRs that have been shown to detect infection by many types of pathogens, including viruses (5). TLR3 is engaged specifically by double-stranded (ds) RNA that is present either in viral genomes or generated during viral replication, and is involved in the cellular recognition of RNA viruses and induction of type I IFN responses (6). However, several recent studies indicate that viral infection can also activate host responses through TLR3-independent pathways (7)(8)(9). Mouse TLR7 and human TLR8, which are expressed within endosomal membranes, detect GU-rich viral single-stranded RNA (7,10,11). Whereas TLR3 signaling requires the adaptor protein, Toll-IL1 receptor (TIR) domain-containing adaptor inducing IFN-␤ (TRIF/TICAM-1) (12)(13)(14)(15), TLR7/8 engagement utilizes the MyD88 adaptor protein and requires endosomal acidification for activation of IFN responses (7,10). Finally, virally encoded proteins may engage TLR2 or TLR4, activating MyD88dependent pathways leading to expression of inflammatory cytokines and contributing to both viral clearance as well as pathogenesis (16 -18).
The TLRs are not the only class of PAMP receptors that contribute to the recognition of virus infection. Yoneyama and colleagues (19) have demonstrated recently that IFN-␤ production is induced in response to Newcastle disease virus infection through a pathway that is independent of TLR3 but requires the retinoic acid-inducible gene I (RIG-I, or DDX58), a cellular RNA helicase with homology to caspase-recruitment domain (CARD) proteins.
The ability of these pathways to induce the expression of type I IFNs and subsequently a wide array of IFN-stimulated genes (ISGs) is likely to influence the outcome of infection by many human viruses. Such responses may have direct antiviral effects within the infected cell (2). Type 1 IFNs also exert critically important immunoregulatory effects, including stimulation of natural killer cell cytotoxicity, activation of ␥␦ T cells, and stimulation of immunoglobulin synthesis. Cytokines and chemokines induced by the activation of these signaling pathways also contribute to the maturation of dendritic cells and influence the priming of T H 1 and cytotoxic T cells (20). These responses thus may play critical roles in shaping subsequent adaptive T cell responses that are required for the ultimate elimination of viruses (21,22).
Although the liver is a particularly important site of persistent viral infections in humans, very little is known about how these signaling pathways function specifically in hepatocytes. Despite this, there is strong, albeit indirect evidence, that type I IFN responses are important in the pathogenesis of chronic viral hepatitis. Both hepatitis B virus (HBV) and hepatitis C virus (HCV) cause persistent infections involving the hepatocyte, and both have evolved mechanisms to disrupt the induction of type I IFNs. Although not well understood, the core protein of HBV inhibits the transcription of IFN-␤ (23,24). Similarly, the NS3/4A protease of HCV blocks Sendai virus (SenV)-induced activation of IFN regulatory factor 3 (IRF-3), a cellular transcription factor that plays a critical role in the expression of IFN-␤ (25). IRF-3 blockade is dependent upon the protease activity of NS3/4A, suggesting that NS3/4A proteolytically targets one or more cellular proteins residing within the signaling pathways leading to IRF-3 activation and type I IFN production (25). These and other data (26,27) suggest that a more detailed understanding of the mechanisms by which virus infections trigger IRF-3 activation and IFN production in hepatocytes would be helpful in unraveling the pathogenesis of persistent HBV and HCV infections, and might possibly lead to the design of novel therapeutic interventions. Here, we describe efforts to better define the antiviral signaling pathways that are active in cultured hepatocyte-derived cell lines and that thus may be triggered by HBV or HCV infection in vivo.
Sendai Virus Infection-Cells were infected with 100 hemagglutinin units/ml of SenV (Charles River Laboratory) and harvested 16 h later for luciferase/␤-galactosidase reporter assays or immunoblot analysis as previously described (25).
Poly(I-C) Treatment-Poly(I-C) was added directly to the medium at 50 g/ml (M-pIC), or complexed with Lipofectin for transfection (T-pIC). Cells were assayed for poly(I-C) induced responses 6 h after exposure by either route.
Reporter Gene Assay-Cells (5 ϫ 10 4 cells per well in 24-well plates) were transfected with reporter plasmids (100 ng), pCMV-␤-galactosidase (100 ng), and the indicated amounts of an expression vector. Twenty-four hours later, cells were mock treated or treated with 20 ng/ml interleukin-1␤, or challenged with poly(I-C) or Sendai virus, then subsequently lysed and assayed for luciferase and ␤-galactosidase activities as indicated. For comparisons, luciferase activity was normalized to ␤-galactosidase activity. Data are expressed as mean relative luciferase activity Ϯ S.D. for one representative experiment carried out in triplicate, typically from a minimum of three separate experiments. The -fold induction of promoter activity was calculated by dividing the relative luciferase activity of stimulated cells with that of mock-treated cells.
RNA Interference-Transfections of siRNAs targeting TLR3, TRIF, or RIG-I were carried out using Oligofectamine (Invitrogen) with a final concentration of siRNA of 80 nM according to the manufacturer's instructions. The target sequences of siRNA used in this study were: TLR3, GGTATAGCCAGCTAACTAG; TRIF (13), GACCAGACGC-CACTCCAAC; RIG-I, GGAAGAGGTGCAGTATATT; and MDA5, GGT-GAAGGAGCAGATTCAG. siRNAs were purchased from Dharmacon and Ambion. The scrambled negative control siRNA was from Ambion.
RT-PCR-Total cellular RNA was extracted with TRIzol Reagent (Invitrogen), treated with DNase I to remove genomic DNA contamination, and reverse-transcribed using Advantage RT-4PCR kit (Clontech). The resulting cDNA was subjected to PCR using primers (Table I) specific for IFN-␤, TLR3, TRIF, ISG56, RIG-I, MDA5, ␤-actin, and GAPDH, respectively. The quantity of the cDNA template included in these reactions and the number of amplification cycles were optimized to ensure that reactions were stopped during the linear phase of product amplification, permitting semiquantitative comparisons of mRNA abundance between different RNA preparations. To exclude the possibility of contaminating DNA, control reactions were performed in parallel in the absence of reverse transcriptase. PCR products were visualized by agarose gel electrophoresis.

Cultured Hepatocyte Cell Lines Differ in Their Ability to
Activate IFN-␤ Transcription in Response to dsRNA or Virus Infection-The induction of type I IFNs represents an early protective response to many viral infections in mammalian cells. IFN-␤ induction represents the immediate response of cells to viral infection, and precedes the transcription of most IFN-␣ species, the induction of which depends on autocrine/ paracrine feedback of IFN-␤ and activation of IRF-7 (31). We thus focused on characterizing the IFN-␤ response. The molecular basis for induction of IFN-␤ expression has been exten- sively studied and shown to be induced by dsRNA or other products of virus infection through coordinate activation of transcription factors IRF-3, NF-B, and ATF-2/c-Jun (2). Most of these studies were conducted in human embryonic kidney 293 or epithelial cells, and relatively little is known about these events in hepatocytes, which constitute ϳ80% of the liver cell population (32), and are the primary cells within which both HBV and HCV replicate. Because primary differentiated hepatocytes are difficult to maintain in culture, we studied 3 different continuous cell lines derived from human hepatocellular carcinomas: Huh7 cells, which appear to be particularly permissive for HCV RNA replication (33,34); HepG2 cells and Hep3B cells, the latter of which contains integrated HBV DNA and express the HBV envelope protein, HBsAg (35). Huh7 cells are not available from any standard repository, but are carried in many different laboratories and different laboratory variants may show significant variation in morphology and perhaps other characteristics. We thus studied 3 different Huh7 sublines: Huh7 SL, cells carried in this laboratory for many years; Huh7 2-3c, cells derived from Huh7 SL that were selected for their ability to support replication of genome-length HCV RNA and subsequently "cured" of the replicating RNA with IFN-␣2b treatment (36), and Huh7 MG cells, Huh7 cells carried in the Gale laboratory at the University of Texas Southwestern Medical Center. We also studied the PH5CH8 cell line, which is a clonal variant of a non-neoplastic hepatocyte cell line, PH5CH, immortalized with the simian virus 40 large T antigen (28,37,38). We characterized virus-induced activation of the IFN-␤ promoter in these various hepatocyte-derived cell lines after transient transfection with a reporter plasmid expressing luciferase under control of the IFN-␤ promoter. Although the PH5CH8 cell line has been reported to support replication of HCV (28), no cultured cell has been reported to be fully for HCV replication. Thus, to challenge these cells, we exposed them to the dsRNA analog, poly(I-C), or infected them with SenV. Interestingly, when poly(I-C) was added to the culture medium, there was no induction of IFN-␤ promoter activity in the Huh7, HepG2, or Hep3B hepatoma cells (M-pIC, Fig. 1A, upper panel). In contrast, the PH5CH8 cells demonstrated a 6-fold up-regulation of IFN-␤ promoter activity upon exposure to poly(I-C) (Fig. 1A, upper panel). Consistent with the reporter data, M-pIC treatment significantly up-regulated the expression of ISG15 and ISG56, which are responsive to either IRF-3 or IFN, in PH5CH8 cells, but not in hepatoma cell lines Huh7, HepG2, or Hep3B (Fig. 1B). To mimic intracellular dsRNA generated during viral replication, we also transfected poly(I-C) into cells using a liposome-mediated procedure. This resulted in much more potent stimulation of the IFN-␤ promoter in PH5CH8 cells, leading to a 17-fold increase in activity over basal promoter levels (T-pIC, Fig. 1A, middle panel). It also resulted in significant activation of the IFN-␤ promoter in HepG2 (8-fold induction) and Hep3B (4-fold induction) cells. However, under the same conditions, there was little if any activation of the promoter in any of the Huh7 cell sublines. These promoter assay results were confirmed by RT-PCR analysis of endogenous IFN-␤ mRNA synthesis in cells transfected with increasing concentrations of poly(I-C) (Fig. 1C). Whereas transfection of as little as 1 g/ml poly(I-C) induced IFN-␤ transcription in HepG2, Hep3B, and PH5CH8 cells, there was no detectable IFN-␤ mRNA in Huh7 SL cells transfected with up to 100 g/ml poly(I-C).
A similar pattern of IFN-␤ induction was observed following infection of these different cell lines with SenV. The greatest induction of IFN-␤ promoter activity occurred in PH5CH8 cells (96-fold increase over basal promoter activity), followed by HepG2 (35-fold), and Hep3B (10-fold) (Fig. 1A, lower panel). However, there were significant differences in the magnitude of the IFN-␤ promoter activation induced by SenV infection in the different Huh7 cell lines. SenV infection induced IFN-␤ promoter activity by ϳ10-fold in the Huh7 MG cells, but only 1.5-fold in Huh7 SL cells and 3-fold in the cured Huh7 2-3c cells (Fig. 1A, lower panel).
We conclude from these results that cultured hepatoma cells generally have impaired poly(I-C)-and virus-activated IFN responses, compared with the PH5CH8 cells that were established from non-neoplastic hepatocytes. These latter cells retain robust IFN-␤ responses to extracellular and intracellular poly(I-C) as well as SenV infection. They are likely to more closely resemble normal hepatocytes in vivo, and thus may represent a superior cell model for investigation of antiviral responses in hepatocytes. The considerable variation we observed in the IFN response in different Huh7 cell lines mandates caution in comparing studies carried out with these cells in different laboratories, an important point because Huh7 cells are widely used for cell culture studies with HCV.
TLR3 and TRIF Are Not Required for SenV Activation of the IFN-␤ Promoter in PH5CH8 Cells-TLR3 is the only known TLR that recognizes viral dsRNA and it is expressed both on the cell surface and within intracellular vesicles (39,40). We were thus interested in determining whether the virus-induced IFN response involves TLR3 in PH5CH8 cells. To investigate this, we transfected the cells with a synthetic siRNA duplex targeting human TLR3, or a scrambled negative-control siRNA, then challenged the cells by adding poly(I-C) to the media or infecting them with SenV prior to measuring IFN-␤ promoter activity (Fig. 2). PH5CH8 cells express a readily detectable abundance of TLR3 mRNA, which was reduced to almost undetectable levels upon transfection of TLR3-specific siRNA (Fig. 2B). The induction of IFN-␤ promoter activity by extracellular poly(I-C) was substantially reduced by this siRNA knockdown of TLR3 expression ( Fig. 2A, left panel). In contrast, there was no appreciable effect on SenV-induced IFN-␤ promoter activation ( Fig. 2A, right panel).
TRIF is an essential adaptor protein that links TLR3 to downstream kinases responsible for IRF-3 activation and IFN-␤ production (13)(14)(15). It may well play additional roles in intracellular signaling events, given its relatively large size (712 amino acids) compared with other TLR adaptors. Mice that are deficient for TRIF function have an impaired response to murine cytomegalovirus infection (12). The liver contains a higher abundance of TRIF than any other organ (14), suggesting that it may play a particularly important role in intrahepatic signaling. To determine whether TRIF is required to mount a type I IFN response against SenV in hepatocytes, we carried out a TRIF knockdown experiment in PH5CH8 cells. As shown in Fig. 2D, semiquantitative RT-PCR demonstrated that transfection of the TRIF-specific siRNA efficiently reduced TRIF mRNA abundance. As expected from the TLR3 knockdown ( Fig. 2A), TRIF knockdown also significantly inhibited the induction of IFN-␤ promoter activity in response to extracellular poly(I-C) (Fig. 2C, left panel). It also significantly reduced extracellular poly(I-C) induction of ISG56 mRNA transcription (Fig. 2D, compare lanes 2 versus 4). In contrast, there was no effect on either IFN-␤ promoter activity (Fig. 2C, right panel) or ISG56 mRNA transcription (Fig. 2D, compare lanes 6 versus 8) triggered by SenV infection. Similar results were obtained in an ISG54 ISRE promoter assay (data not shown). We conclude from these data that exposure to extracellular poly(I-C) triggers activation of the IFN-␤ promoter in PH5CH8 cells through a TLR3-TRIF dependent pathway, whereas SenV induction of IFN-␤ transcription is TLR3-and TRIF-independent.
SenV Induction of IFN-␤ in Hepatocytes Does Not Utilize MyD88 nor Require Endosomal Acidification-In addition to TLR3, recent data indicate that TLR7, TLR8, and TLR9 may sense viral components and lead to type I IFN production. Murine TLR7 and human TLR8 recognize viral ssRNA, whereas TLR9 senses unmethylated CpG DNA present in the murine cytomegalovirus genome (41). A common feature of these three TLRs is that they all localize within endosomes and signal through MyD88 and IRF-7 (42,43). In addition, the responses induced through these TLRs require intact endocytic pathways and are thus sensitive to endosomal acidification inhibitors, such as chloroquine and bafilomycin A1 (7, 10). Because SenV induction of IFN-␤ occurs via a TLR3-TRIF independent pathway in PH5CH8 cells, we considered the possibility that TLR7/8 may initiate this response. To determine whether MyD88 is required for SenV activation in the PH5CH8 cells, as would be the case were it mediated by TLR7, TLR8, or TLR9, we measured promoter activation in PH5CH8 cells transfected with a vector expressing a dominant-negative MyD88 mutant that lacks the amino-terminal death domain that is required for interaction with IRF-7 (42) (MyD88DN). Expression of MyD88DN significantly inhibited the activation of the NF-B-dependent PRDII promoter (44) by interleukin-1␤, which is known to signal through MyD88 (29). However, we found that the activity of both PRDII and IFN-␤ promoters was induced to similar levels by SenV with or without MyD88DN co-expression (Fig. 3A). Consistent with these reporter data, SenV-induced hyperphosphorylation of IRF-3 was not affected by overexpression of MyD88DN (Fig. 3B). This was also true for SenV-induced expression of ISG56, which is responsive to either IRF-3 or IFN (45), and MxA, which responds only to IFN (46).
To determine whether SenV activation of the IFN-␤ promoter requires endosomal acidification, PH5CH8 cells were pretreated with the endosomal inhibitor bafilomycin A1 prior to infection with SenV. Although bafilomycin A1 treatment completely ablated NF-B activation in RAW264.7 cells induced by a TLR7/8 ligand, R-848 (Fig. 3C), SenV-stimulated IFN-␤ and PRDII promoter activity was minimally reduced in cells treated with the compound (Fig. 3D). Moreover, immunoblot analyses indicated that there were no differences in SenVinduced hyperphosphorylation of IRF-3, or ISG15 and MxA expression in these cells in the presence or absence of bafilomycin A1 (Fig. 3E). Bafilomycin A1 did not alter the expression of SenV proteins.
These data suggest that SenV induction of IFN responses is MyD88-independent in PH5CH8 cells, and does not require active endocytic pathways. Both lines of evidence argue strongly against the involvement of TLR7/8. In addition, because MyD88 is thought to be used by all TLRs other than TLR3 (and TLR4 for signaling to IRF-3), these data, taken in the context of the absence of any inhibition of the SenV response by TRIF knockdown (Fig. 2, C and D), suggest that the recognition of SenV by PH5CH8 cells is not dependent upon any known TLR.

SenV Activates Type I IFN Responses through a RIG-I-dependent Pathway in Hepatocytes-Studies with knock-out mice have indicated that both virus infection and dsRNA can trigger type I IFN responses via TLR3-independent mechanisms (7, 8).
Recently, Yoneyama and colleagues (19) demonstrated that RIG-I, a putative DExD/H box RNA helicase containing an N-terminal sequence with CARD-like homology domains, is essential for IFN-␤ production induced in response to infection with Newcastle disease virus. To determine whether SenV induces IFN-␤ promoter activity in PH5CH8 cells through a RIG-I-dependent pathway, we utilized RNA interference to knockdown RIG-I expression prior to virus challenge. Transfection of a RIG-I-specific siRNA, but not a scrambled control siRNA, reproducibly caused a ϳ50% reduction in SenV-induced activation of both the IFN-␤ (Fig. 4A, right panel) and PRDII promoters (Fig. 4B, right panel). Semiquantitative RT-PCR confirmed that the transfection of RIG-I siRNA efficiently knocked down the basal expression of RIG-I (Fig. 4C, lanes 2  and 4 in the left panel and lanes 7 and 11, in right panel), and also significantly blunted the up-regulation of this IFN-induced protein upon SenV infection (Fig. 4C, right panel, compare  lanes 9 and 10 with lanes 11 and 12). SenV-induced transcription of ISG56 mRNA was also significantly reduced in cells transfected with RIG-I siRNA (Fig. 4C, lanes 9 -12). Therefore, RIG-I is an essential component in the TLR-independent pathway by which PH5CH8 cells sense SenV infection and initiate a type I IFN response.
In contrast to these results, PH5CH8 cells transfected with the RIG-I siRNA responded normally when poly(I-C) were added to the media, in terms of activation of both the IFN-␤ (Fig. 4A, left panel) and NF-B-dependent PRDII (Fig. 4B, left  panel) promoters. In contrast to SenV-induced signaling, the poly(I-C) induction of RIG-I transcription was not dramatically reduced by transfection of the RIG-I siRNA (Fig. 4C, compare  lanes 5 and 6 with lanes 7 and 8). Similarly, there was no inhibition of poly(I-C) induction of ISG56 transcription (Fig.  4C, lanes 5-8). We also confirmed the RT-PCR results by immunoblot analysis of ISG15 expression under these conditions (Fig. 4D). These data indicate that TLR3-and RIG-I-mediated signaling function independently of each other in PH5CH8 cells, similar to what has been reported previously for epithelial HeLa cells (19).
Poly(I-C) Activation of the IFN-␤ Promoter Is Partially MDA5dependent-Because siRNA knockdown of RIG-I did not ablate but only partially reduced the IFN-␤ promoter response to SenV infection in PH5CH8 cells (Fig. 4A, right panel), we carried out similar experiments to determine whether the human melanoma differentiation associated gene-5 product (MDA5) participates in this signaling pathway, possibly in a redundant role with respect to RIG-I. MDA5, another DExD/H box RNA helicase, is an IFN-inducible protein that shares a subdomain architecture and considerable sequence homology with RIG-I (47). It has been associated with the induction of apoptosis, but not clearly identified as playing a role in activation of IRF-3 similar to that of RIG-I. siRNA-mediated knockdown of MDA5 expression had only a minor, but reproducible, suppressive effect on SenV induction of the IFN-␤ promoter. The magnitude of the reduction in promoter activity was substantially less than that observed with RIG-I knockdown (Fig.  5A, left panel), but it was associated nonetheless within a modest reduction in SenV-induced expression of ISG56 (Fig.  5B, compare lanes 6 versus 8). The suppressive effect of MDA5 knockdown was not additive with RIG-I knockdown, as the degree of suppression of promoter activity was not increased when both RIG-I and MDA5 siRNAs were cotransfected into PH5CH8 cells. The MDA5 knockdown had no effect on SenV protein expression (Fig. 5B).
Interestingly, MDA5 knockdown caused ϳ30% reduction in the induction of IFN-␤ promoter activity by extracellular poly(I-C) (Fig. 5A, right panel). This effect was less than that observed with TLR3 or TRIF knockdown, yet nonetheless reproducible and significant. Consistent with this, MDA5 knockdown also caused a moderate reduction in poly(I-C)-induced ISG56 expression (Fig. 5B, compare lanes 2 and 4). We conclude from these results that MDA5, like RIG-I, may contribute to viral activation of the IFN-␤ promoter but, unlike RIG-I, plays a greater role in poly(I-C) activation of the promoter than in SenV-induced responses. Whether this occurs in association with, or independent of, the TLR3-TRIF pathway remains to be determined. Fig. 1, all three Huh7 cell sublines studied were defective in poly(I-C) signaling. This lack of poly(I-C) responsiveness may be explained by the fact that Huh7 cells express a negligible abundance of TLR3 mRNA, in contrast to PH5CH8 cells that demonstrate robust TLR3 expression (Fig. 6A). In contrast, TRIF mRNA abundance was approximately equal in these cell lines. Overexpression of FLAG-tagged TLR3 effectively restored the ability of Huh7 cells to respond to poly(I-C) when added to the culture medium (50 g/ml), as indicated by expression of ISG15, ISG56, and MxA (Fig. 6B, compare lanes 2 and 4). Thus, Huh7 cells are normally deficient in their ability to respond to externally applied poly(I-C) because of a lack of sufficient expression of TLR3.

TLR3 Overexpression Reconstitutes Poly(I-C)-induced ISG Expression in Huh7 Cells-As shown in
TLR3-and RIG-I-mediated Signaling Function Independently in Huh7.5 Cells-TLR3 signaling has been reported not to be adversely affected by overexpression of a dominant-negative form of RIG-I in HeLa cells (19). Consistent with this, the siRNA knockdown experiments shown in Figs. 2 and 4 suggest that these pathways function largely independently of each other in PH5CH8 cells. To confirm this in another hepatocytederived cell line, we utilized a Huh7 subline, Huh7.5, which is highly permissive for replication of HCV RNA replicons (48) and which has recently been shown to be defective for RIG-I  1, 2, and 5, 6) or MDA5 siRNA (lanes 3, 4, and 7, 8). Where indicated, cells were treated with 50 g/ml M-pIC for 8 h or infected with 100 HAU/ml SenV for 16 h. In lane 9, cells were treated with 500 units/ml of IFN-␣2b as a control for ISG56 expression. C, semiquantitative RT-PCR detection of MDA5 and GAPDH mRNAs in control (Ctrl)/MDA5 siRNA-transfected PH5CH8 cells. signaling because of a point mutation within its CARD-like homology domain (49). Thus, in contrast to PH5CH8 cells in which both TLR3 and RIG-I pathways are intact, neither pathway is functional in Huh7.5 cells. By reconstituting expression of either TLR3 or a functional RIG-I molecule in Huh7.5 cells, we were thus able to assess the role played by each signaling pathway independently in recognition of viral infection while determining the extent to which these pathways act independently of each other.
We first confirmed previous observations (49) that SenVinduced signaling through RIG-I is deficient in Huh7.5 cells, by demonstrating that SenV-induced expression of ISG15 was dependent upon ectopic expression of RIG-I in these cells (Fig. 7A). In contrast, ectopic expression of TLR3 did not rescue SenV-induced ISG15 expression in Huh7.5 cells, suggesting that the products of SenV infection do not engage TLR3, or that optimal TLR3 signaling requires functional RIG-I. The latter is unlikely, however, as RIG-I knockdown had no effect on IFN induction by extracellular poly(I-C) in PH5CH8 cells (Fig. 4). Moreover, ectopic expression of TLR3 alone was able to restore responsiveness to extracellular poly(I-C) in Huh7.5 cells (Fig. 7B, lanes 8 versus 10), to a degree similar to that in normal Huh7 cells (Fig. 7C, lanes 15  versus 21). The rescue of poly(I-C) responsiveness was specific for TLR3, as overexpression of a TIR domain-deleted TLR3 mutant failed to restore expression of ISG15 and ISG56 in response to poly(I-C) (Fig. 7B, lanes 10 versus 12), nor did RIG-I (Fig. 7C, lanes 15 versus 17). Dual ectopic expression of TLR3 and RIG-I only slightly increased the ISG15 induction in response to poly(I-C) (Fig. 7C, lanes 18 and 19). Taken together, the data suggest that there is little if any cross-talk between the TLR3-and RIG-I pathways in hepatocytes. DISCUSSION We have shown here that cultured hepatoma cells generally have impaired poly(I-C) and viral-activated IFN responses, compared with PH5CH8 cells that are derived from normal hepatocytes (Fig. 1). These data are thus in agreement with a previous report (50) showing that hepatoma cells have impaired antiviral responses. HepG2 and Hep3B cells are deficient in signaling in response to extracellular poly(I-C), and appear to have diminished responses to either intracellularly delivered poly(I-C) or SenV infection. As described by others (51,52), we found that Huh7 cells also failed to respond to poly(I-C), irregardless of whether it was added to the culture medium or introduced into cells via transfection. In addition, only a weak response was observed in some Huh7 cell sublines (Huh7 2-3c and MG) after infection with SenV, a more potent IFN inducer. Poly(I-C) signaling could be restored to Huh7 cells by ectopic expression of TLR3 (Fig. 6), indicating that the lesion in Huh7 cells is at the level of the PRR molecule, and that the downstream pathway involving TRIF is intact. In contrast, the non-neoplastic PH5CH8 cells retain robust IFN-␤ responses to extracellular and intracellular poly(I-C) as well as SenV infection, and thus may be more representative of the antiviral signaling pathways present within hepatocytes in vivo.
dsRNA is commonly expressed during the replication of most RNA viruses, and has long been considered a candidate viral PAMP. Treatment of mammalian cells with the synthetic dsRNA analog, poly(I-C), induces type I IFN production through activation of transcription factors IRF-3 and NF-B (53). TLR3 is well characterized as a PRR that is engaged specifically by dsRNA in many cell types. We have shown here that it is expressed in non-neoplastic hepatocytes, and that it plays an important role in these cells in activation of the IFN-␤ promoter following exposure to extracellular poly(I-C) ( Fig.  2A). Furthermore, we have shown that poly(I-C) signaling through TLR3 leading to the induction of ISG expression is dependent upon the adaptor protein TRIF in these cells, as might be expected (Fig. 2C). Thus, the data presented demonstrate clearly that the TLR3-TRIF pathway is functional in the non-neoplastic hepatocyte-derived PH5CH8 cells, although it is FIG. 7. TLR3 and RIG-I signaling pathways function independently in Huh7.5 cells. A, the defect of SenV-induced ISG15 expression can be restored by overexpression of RIG-I, but not TLR3 in Huh7.5 cells. B, the defect of extracellular dsRNA-induced expression of ISG15 and ISG56 can be restored by overexpression of TLR3, but not a mutant TLR3 with the TIR domain deleted (⌬TIR) in Huh7.5 cells. C, ectopic expression of TLR3, but not RIG-I, restored the response to M-pIC in Huh7.5 cells. All panels shown were immunoblot analysis. Where indicated, cells were treated with 50 g/ml M-pIC for 12 h, or 100 HAU/ml SenV for 16 h, or 500 units/ml of IFN-␣2b. Please note that the TLR3 antibody detects transfected FLAG-RIG-I for unknown reasons. The superintense bands of RIG-I detected by anti-FLAG antibody is likely because of the 2ϫ FLAG tag versus 1ϫ FLAG tag in TLR3.
generally absent in cultured cells derived from hepatocellular carcinomas (Fig. 1).
Importantly, neither RNA interference directed silencing of TLR3 nor TRIF resulted in any impairment of IFN-␤ promoter activation or ISG expression in SenV-infected PH5CH8 cells (Fig. 2). This indicates that dsRNA, or possibly another PAMP produced during SenV replication, triggers activation of the IFN-␤ promoter in these hepatocytes through a distinctly different signaling pathway. This is consistent with several recent reports that dsRNA and/or virus infection can initiate antiviral signaling through TLR3-independent pathways (7,8). SenV could potentially activate signaling through recognition of viral ssRNA by endosomally located TLR7/8, as reported for murine plasmacytoid dendritic cells and B lymphocytes (7). However, neither endosomal acidification inhibitors nor expression of a dominant-negative MyD88 mutant affected the response in SenV-infected PH5CH8 cells (Fig. 3). Thus, SenV appears to activate a TLR-independent intracellular signaling pathway in hepatocytes that is distinct from the TLR3-TRIF pathway activated by extracellular poly(I-C).
Consistent with a previous study carried out in mouse fibroblasts and human epithelial cells (19), we found that RIG-I is an essential component of this TLR-independent antiviral signaling pathway in cells derived from non-neoplastic hepatocytes (Fig. 4). Similarly, we have recently shown that SenV activation of IRF-3 is dependent upon RIG-I expression in hepatoma cells as well (49). RIG-I is a cytoplasmic RNA helicase that contains tandem motifs near its N terminus with limited homology to CARD domains and a downstream DExD/ H-box helicase domain. It putatively binds viral dsRNA within its helicase domain, resulting in activation of IRF-3 and NF-B through signaling involving the N-terminal CARD-like homology domains (19). MDA5 is a closely related DExD/H box helicase with similar domain architecture, and has been associated with apoptotic signaling in terminally differentiating melanoma cells (47). Its expression is induced by IFN-␣ as well as infection with human immunodeficiency virus (47,54). We found that it contributes to the induction of IFN signaling by extracellular poly(I-C) (Fig. 5), but has only a minor, yet clearly discernible, contribution to SenV induction of IFN-␤ promoter activity and downstream ISG expression. Whether it functions in a fashion similar to that proposed for RIG-I, as a PRR molecule binding dsRNA through its helicase domain, seems likely but remains to be demonstrated. It is also not known whether RIG-I and MDA5 are evolved to preferentially recognize dsRNAs derived from viruses. The relationship of MDA5 to the TLR3-TRIF pathway also remains to be defined.
Although HCV replicon RNAs expressing selectable markers are capable of replicating in some alternative cell types (55), Huh7 cells have proven to be nearly unique in their ability to support the autonomous replication of these viral RNAs (34). In addition, we previously demonstrated that Huh7 cells are uniquely permissive for self-amplification of subgenomic replicons derived from hepatitis A virus, another positive-strand RNA virus (56). Huh7.5, a Huh7 subline that is highly permissive for HCV RNA replication (48), has a lethal mutation in the RIG-I CARD-like homology domain that renders it unresponsive to structured HCV RNA or SenV induced signaling (49). Whereas this accounts for the highly permissive phenotype of Huh7.5 cells, it does not fully explain why parental Huh7 cells are already more permissive than other cell types. Although relatively little is known regarding the cellular factors associated with permissiveness for HCV (34,(57)(58)(59), it is intriguing to speculate that the enhanced permissiveness for positivestrand virus RNA replication in normal Huh7 cells may relate to the absence of significant TLR3-mediated antiviral re-sponses in these cells. It remains to be determined, however, whether HCV RNA replication results in specific engagement of TLR3 and activation of the downstream signaling pathway, and whether the TLR3 signaling cascade acts to limit HCV replication. Recent data indicate that expression of the HCV NS3/4A protease inhibits poly(I-C)-induced, TLR3-dependent signaling by directing the proteolytic cleavage of TRIF in osteosarcoma cells as well as in HeLa cells supporting replication of subgenomic HCV RNA replicons (26). However, for reasons that remain unclear, the ectopic expression of NS3/4A does not appear to block TLR3 signaling in PH5CH8 cells. 2 In summary, hepatocytes contain two distinct antiviral signaling pathways leading to expression of type I IFNs, one dependent upon TLR3 and the other on RIG-I, and with little evidence of significant cross-talk between them. However, although the RIG-I and TLR3 pathways function independently in hepatocytes (Figs. 4 and 7), it is very likely that during viral infection in vivo both pathways are activated in a coordinate manner to enhance innate immune responses. Activation of the RIG-I pathway by the intracellular presentation of dsRNA or another viral PAMP produced by replicating viruses may contribute to the initial induction of type I IFNs. Subsequently, viral dsRNA released by lysis of cells in later stages of the infection may engage TLR3 and induce TRIF-dependent signaling, resulting in further amplification of the antiviral response (53). Importantly, type I IFNs induced through either pathway would have a positive feedback on signaling through both pathways, as both RIG-I and TLR3 are ISGs (19,60).