Potential conflict of interest: Nothing to report.
Supported by the Deutsche Forschungsgemeinschaft (FOR1202, TP1 and TP3) and the Bundesministerium für Bildung und Forschung (01KI0786). L.K., M.B., and R.B. received funding from the European Union (SysPatho, grant no. 260429). E.D. was supported by the Viroquant project and P.M. received a Viroquant HBIGS-fellowship.
Persistent infection with hepatitis C virus (HCV) can lead to chronic hepatitis, liver cirrhosis, and hepatocellular carcinoma. All current therapies of hepatitis C include interferon-alpha (IFN-α). Moreover, IFN-gamma (IFN-γ), the only type II IFN, strongly inhibits HCV replication in vitro and is the primary mediator of HCV-specific antiviral T-cell responses. However, for both cytokines the precise set of effector protein(s) responsible for replication inhibition is not known. The aim of this study was the identification of IFN-α and IFN-γ stimulated genes (ISGs) responsible for controlling HCV replication. We devised an RNA interference (RNAi)-based “gain of function” screen and identified, in addition to known ISGs earlier reported to suppress HCV replication, several new ones with proven antiviral activity. These include IFIT3 (IFN-induced protein with tetratricopeptide repeats 3), TRIM14 (tripartite motif containing 14), PLSCR1 (phospholipid scramblase 1), and NOS2 (nitric oxide synthase 2, inducible). All ISGs identified in this study were up-regulated both by IFN-α and IFN-γ, demonstrating a substantial overlap of HCV-specific effectors induced by either cytokine. Nevertheless, some ISGs were more specific for IFN-α or IFN-γ, which was most pronounced in case of PLSCR1 and NOS2 that were identified as main effectors of IFN-γ-mediated anti-HCV activity. Combinatorial knockdowns of ISGs suggest additive or synergistic effects demonstrating that with either IFN, inhibition of HCV replication is caused by the combined action of multiple ISGs. Conclusion: Our study identifies a number of novel ISGs contributing to the suppression of HCV replication by type I and type II IFN. We demonstrate a substantial overlap of antiviral programs triggered by either cytokine and show that suppression of HCV replication is mediated by the concerted action of multiple effectors. (HEPATOLOGY 2012;56:2082–2093)
Interferon (IFN) sensitivity of hepatitis C virus (HCV) in cell culture is well established.1–3 Nevertheless, little is known about the specific IFN-induced factor(s) mediating this suppression, and even less about their molecular mode-of-action.4 More detailed knowledge of these defense mechanisms is crucial for understanding the immune response against HCV, which in the case of T cells is primarily mediated by IFN-γ.5 Identifying relevant ISGs may also help optimizing the current IFN-α-based therapy6 and allow for better predictions of therapy outcome.
More than 300 IFN-stimulated genes (ISGs) are described. Some of them modulate cell survival, differentiation, growth, and other activities to establish a general antiviral state.7 Other ISGs have evolved to counteract specific viruses. For instance, MX genes encode proteins restricting several RNA viruses by direct interaction with the viral nucleocapsid.8 However, for most ISG products, very little is known about their targets and their mode of action.
A number of IFN-α-induced ISGs were shown to suppress HCV replication.9–12 Those studies, however, were designed as loss-of-function assays based on overexpression of individual ISGs, thus neglecting altered cellular functions caused by the global IFN-induced antiviral state. Moreover, all studies conducted thus far have focused on the type I IFN response, although inhibition of HCV replication by CD8+ T cells is primarily mediated by IFN-γ.4 Because T-cell responses are crucial for controlling HCV replication, the important question of which type II ISGs contribute to virus suppression has so far not been addressed.
To overcome these limitations, we assessed the HCV-specific antiviral effect of ISGs in the context of a full innate response induced by IFN-α or IFN-γ. We devised an RNA interference (RNAi)-mediated gain-of-function approach and identified several ISGs contributing to the block of HCV replication by type I and II IFNs. We demonstrate that with either cytokine suppression of viral replication is mediated by the combined action of multiple ISGs.
LucUbiNeoET cells13 (3 × 105) were seeded into 6-cm diameter dishes and transfected with 10 nM predesigned small interfering RNA (siRNA) (primary screen: Qiagen, Hilden, Germany; validation: Life Technologies, Darmstadt, Germany) using Hiperfect transfection reagent (Qiagen). After 96 hours, 6 × 104 siRNA-transfected cells were reseeded into 24-well plates and transfected a second time with the respective siRNAs. After another 10 hours, cells were either left untreated, treated with 20 IU/mL IFN-α (PBL Laboratories, Piscataway, NJ) or 0.4 IU/mL IFN-γ (R&D Systems, Minneapolis, MN). Thirty-six hours after IFN treatment, medium was replaced with plain Dulbecco's modified Eagle's medium (DMEM) for 24 hours and luciferase activity was determined as described.14 Raw data were processed statistically as described in the Supporting Methods.
Combination RNAi-Based Reporter Assay.
Huh7-Lunet cells (1 × 106) were coelectroporated with 5 μg of HCV in vitro transcripts derived from replicon construct pFKI389Luc-EI/NS3-3′_ET (Fig. 6) and 10 nM predesigned siRNA (Life Technologies). After 24 hours, cells were treated with 2.5 IU/mL IFN-α (PBL Laboratories) or 0.5 ng/mL IFN-γ (R&D Systems) and 72 hours later cells were lysed and luciferase activity was determined. Further details are provided in the Supporting Materials and Methods.
Identification of ISGs Responsible for Inhibition of HCV Replication Using an RNAi-Based Screening Approach.
In order to assemble a list of ISGs that potentially contribute to the IFN-induced suppression of HCV replication, we first identified genes that were up-regulated by IFNs in the presence of replicating HCV. We analyzed the transcriptional profile of several subgenomic and full-genomic HCV replicon cell lines.15 Cells were treated with 100 IU/mL IFN-α or 1 μg/mL IFN-γ or mock-treated for 18 hours and total RNA was analyzed by complementary DNA (cDNA) microarray technology. Transcripts of 90 genes were up-regulated by at least 2-fold with a significant overlap of ISGs being stimulated by IFN-α and IFN-γ. We excluded known signal transducers and transcription factors, but included prominent ISGs from published transcriptome datasets,16 thus creating a final list of 60 candidate genes (Supporting Table S1; Fig. 1A).
We next devised an siRNA-based rescue assay by individually knocking down each candidate gene in IFN-treated cells and screening for a restoration of HCV replication therein (Fig. 1B). Briefly, a subgenomic HCV luciferase reporter replicon cell line was serially transfected twice with an ISG-specific siRNA to achieve the highest possible knockdown efficiency. This was important, given the long half-life of many ISGs and the fact that their expression was induced by IFN treatment. Cells were then pulse-treated with IFN-α or IFN-γ for 24 hours and lysed after an additional 24-hour incubation in plain medium. HCV replication was assessed by measuring firefly luciferase activity. As positive controls, we used siRNAs targeting the expression of STAT2 or the IFN-γ receptor subunit 2 (IFNγR2), two central components of the IFN-α and IFN-γ signaling pathways, respectively. In both cases HCV replication could be restored ∼10-fold, demonstrating high assay sensitivity and pathway specificity (Fig. 1C). By using this approach, in our primary screen we tested two independent siRNAs for each selected candidate and used a stringent statistical approach for hit selection (Supporting Materials and Methods). In this way, 18 ISGs were selected: 14 IFN-α and 13 IFN-γ-stimulated genes, of which 9 were common to both cytokines (Fig. 1D; Fig. S1).
As the screen was focused on identifying effector gene products directly inhibiting viral replication, four of the hits were excluded from further validation: IRF1 and IRF9, transcription factors crucial for ISG expression, and PSME2 and PSMB9, two components of the proteasomal degradation machinery. The final hit list of 14 candidates was subjected to an RNAi-based validation screen. To exclude pleiotropic effects, siRNAs (sets of three per gene) with seed regions different from the ones used in the primary screen were purchased from a new provider. Otherwise, the experimental set-up remained identical to the primary screen. By again using stringent selection criteria, 7 of the 14 primary candidates could be confirmed (Fig. 2; Table S2); of these, five were stimulated by both IFN-α and IFN-γ. Three of the high confidence hits, RNase L17 and the IFN-induced transmembrane proteins (IFITM)1 and IFITM3,12, 18 were previously identified as restriction factors of HCV replication, confirming the power of our screening approach and the significance of the identified factors.
PKR Does Not Contribute to IFN-α-Induced Suppression of HCV Replication.
In spite of the high reliability of our screen results, we were surprised that protein kinase R (PKR) was not identified even though this protein has repeatedly been implicated with the innate antiviral response against HCV19, 20 (Fig. S1). To resolve this seeming contradiction, we assessed more closely the effects of PKR on HCV. First, we generated HCV replicon cells stably expressing short hairpin RNAs (shRNAs) either targeting PKR or the unrelated control gene p53 that is mutated in Huh7 cells and thus is nonfunctional.21 Although PKR expression was strongly reduced even upon treatment with 100 IU/mL IFN-α (Fig. 3A), PKR silencing did not rescue HCV replication (Fig. 3B). Second, we transiently overexpressed dominant-negative (dn) PKR in HCV replicon cells (Fig. 3C,D). Immunofluorescence analysis revealed that HCV replication was broadly reduced upon IFN-α treatment, even in cells that expressed large amounts of dnPKR (Fig. 3E,F). Third, we transfected Huh7-Lunet cells with PKR-specific siRNA or a nonsilencing control and then infected cells with the HCV reporter virus JcR2a.22 Reduced PKR amount did not affect HCV replication, thus confirming results obtained with the replicon (Fig. S2). Taken together, these data corroborate the siRNA screening results and confirm that PKR is not a main contributor to IFN-induced suppression of HCV replication.
Regulation of HCV Restriction Factors in Huh-7 Cells and Primary Human Hepatocytes.
To confirm that expression of the newly identified HCV restriction factors is indeed regulated by IFNs, we analyzed expression levels in Huh7-Lunet cells, a standard cell culture system for HCV, which had been treated with different IFN concentrations and for different durations. Indeed, IFIT3 (IFN-induced protein with tetratricopeptide repeats 3), IFITM1, IFITM3, PLSCR1 (phospholipid scramblase 1), TRIM14 (tripartite motif-containing 14), and NOS2 (nitric oxide synthase 2, inducible) were significantly upregulated upon IFN-α and/or IFN-γ treatment. In contrast, RNase L was constitutively expressed and mRNA levels did not significantly increase in response to either IFN; NOS2 expression was significantly elevated only after IFN-γ treatment (Fig. S3).
Induction of identified ISGs was then assessed in a more physiological setting resembling an untreated HCV infection in the liver, primary human hepatocytes (PHHs). Cells were infected with HCV and RNA replication and HCV-stimulated ISG expression were quantified. In agreement with Huh7-Lunet data, apart from RNase L and NOS2, expression of all other ISGs was indeed up-regulated in response to HCV infection. A more detailed analysis revealed three distinct induction kinetics: (1) PKR and IFIT3 showing rapid induction kinetics like IFN-β and IFIT1 (also called ISG56) and peaking even before HCV RNA reached its maximum; (2) PLSCR1 and TRIM14 with intermediate induction kinetics, reaching peak levels slightly after the peak in HCV RNA; and (3) IFITM1 and IFITM3 with slow induction kinetics (Fig. 4; Figs. S4, S5). We noticed significant donor-to-donor variation of PHHs as deduced from the variability of the magnitude of ISG induction, which correlated with the magnitude of HCV replication. Nevertheless, the timepoint at which most ISGs reached maximal expression levels (∼48 hours postinfection) coincided with the onset of a sharp decline in intracellular HCV RNA levels (Fig. 4; Figs. S4, S5) arguing for potent suppression of HCV replication by the induced antiviral response.
Overexpression of ISGs Decreases HCV Replication.
To confirm the antiviral activity of individual ISGs in a different setting, we generated Huh7-Lunet cell pools stably overexpressing single candidate ISGs expressing 10- to 10,000-fold increased mRNA levels as compared to nontransduced control cells (Fig. 5A). We then transfected subgenomic HCV luciferase reporter replicon RNA into these cells and 72 hours later viral replication was assessed (Fig. 5B). Overexpression of NOS2, RNase L, and Viperin (an ISG that was recently described to inhibit HCV replication and included as additional control) reduced viral replication by ∼70%; TRIM14, PLSCR1, and IFITM3 inhibited viral replication by ∼50%; IFITM1 and IFIT3 did not affect HCV replication (Fig. 5B). The lack of replication inhibition observed with the latter two factors argues that they might need additional IFN-dependent factors in order to act antiviral. Indeed, IFIT3 is part of a signaling complex that requires other IFIT proteins for antiviral activity.23
Combinatorial Knockdown of ISGs Reveals an Additive and Synergistic Mode of Action.
The findings that knocking-down individual ISGs did not fully rescue HCV replication during IFN treatment, and that the overexpression of individual candidate genes suppressed HCV replication less efficiently than IFN, suggested that more than one factor mediated the IFN-induced antiviral effect against HCV.
We therefore addressed possible additive or synergistic effects of our candidate ISGs by a combinatorial double-silencing approach, in which we tested all possible pairwise combinations. For this purpose we employed a different cell line, Huh7-Lunet cells, to exclude cell line-specific effects and conducted transient knock-down, which facilitated the analysis and reflected the transient nature of an infection more closely (Fig. 6A). Knock-down efficiencies and cell viability were determined for each silenced ISG (Fig. 6B,C). For IFN-α, all candidates partially rescued HCV replication when singly silenced. Intriguingly, rescue efficiencies were enhanced in a combinatorial setting, revealing additive and even synergistic effects of ISGs. Combined silencing of IFITM1 and IFITM3, for example, exhibited a synergistic effect resulting in a stronger rescue than one would expect from the sum of the individual effects (Fig. 7). For IFN-γ, save for IFITM1, all singly silenced factors showed effects on HCV replication. Surprisingly, the individual knock-down of NOS2 rescued replication to ∼50% of the non-IFN treated control in this experimental set-up (Fig. 8), arguing that NOS2 is a major restriction factor of HCV replication. Given the overall comparatively low additive effects of double knock-downs when stimulated with IFN-γ as compared to IFN-α, we propose that the mode of action of IFN-regulated factors and particularly their mutual interplay is distinct for different types of IFNs. In conclusion, these results corroborate the notion that only the concerted action of several ISGs efficiently suppresses HCV replication.
As an unbiased method to identify IFN-α and IFN-γ stimulated effectors targeting HCV, we performed an RNAi-based screening assay assessing ISGs in a gain of function approach. In contrast, previously published studies were designed as loss-of-function assays using protein overexpression.10, 11 Although yielding profound effects, the most potent ISGs identified in those studies were regulatory factors such as RIG-I, IRF1, or IRF7, which upon overexpression trigger the expression of other ISGs. As a consequence, the actual effectors inhibiting HCV remained rather elusive. Moreover, those studies neglected the possibility that in the antiviral state, function and activity of proteins might by altered by posttranslational modifications (e.g., ISGylation or phosphorylation), changes in half-lives of proteins (e.g., by alteration of the proteasome), or in their sequence resulting, e.g., from RNA editing or alternative splicing as in the case of ADAR1.24 Therefore, an ectopically expressed ISG might function differently as compared to an ISG produced during the antiviral state. In fact, we found that although similar NOS2 levels were reached during IFN-α and IFN-γ treatment, its knock-down in IFN-γ treated cells was substantially more effective in rescuing transient HCV replication than in IFN-α treated cells. Finally, a gain-of-function screen has the advantage that possible off-target effects exerted by siRNAs are excluded by definition, because pleiotropic effects impairing host cell homeostasis will reduce HCV replication further rather than rescue it from the antiviral activity of IFN. However, pleiotropic effects exerted by (high level) overexpression of an ISG would reduce HCV replication and thus score positive as a restriction factor. Their exclusion thus requires detailed follow-up studies.
We identified seven ISGs as suppressors of HCV replication. To confirm the same ISGs in an infection-based HCV cell culture model (HCVcc), we conducted analogous rescue experiments by using the highly sensitive reporter virus JcR2a.22 However, in this system rescue efficiency was low, which has to be ascribed primarily to the exceptional replication competence of the genotype 2a isolate JFH1 that is the basis of JcR2a (Fig. S6). In fact, we had to use much higher IFN concentrations as compared to Con1 replicons and, even then, suppression of JcR2a was rather low. Because similar results were obtained when using a subgenomic JFH1-derived replicon (Fig. S7), we assume that this particular HCV isolate, rather than the route of RNA delivery, accounts for the low rescue efficiency. Nevertheless, in spite of the lower robustness of the JFH1-based assay, the results obtained suggest that the ISGs identified with Con1 replicons apply to multiple HCV isolates and genotypes.
We were surprised that PKR was not identified in our screen. Although earlier reports described PKR as an inhibitor of HCV-replication,19, 25 transient or stable knock-down of PKR as well as overexpression of a dominant-negative mutant neither affected replication of subgenomic replicons nor HCVcc production (Fig. S2).26 Thus, our results suggest that PKR does not execute IFN-induced antiviral activities against HCV.
Six of the seven validated ISGs were induced by both IFNs. The exception was PLSCR1. This enzyme is localized in the nucleus, but upon palmitoylation redistributes to membranes throughout the cell, where it “scrambles” phospholipids.27 Because HCV replication is highly dependent on intracellular membrane rearrangements,28 scrambling of phospholipids could be a possible mechanism by which PLSCR1 interferes with viral replication.
In primary and secondary screening, NOS2 was identified as an only moderately effective, IFN-α specific inhibitor of HCV. However, in the transient replication assay and in overexpression experiments, NOS2 exhibited a much stronger IFN-γ-specific antiviral effect. We assume that NOS2 targets a very early event in the viral replication cycle and is therefore significantly less effective in a persistent steady-state. Alternatively, HCV-induced membrane rearrangements established in the persistent state28 might protect viral replication sites from the antiviral effect exerted by NOS2. This enzyme is believed to execute its antimicrobial activities by way of synthesis of nitric oxide (NO).29 We have previously reported that neither depletion nor exogenous administration of NO affected stable HCV replicons.2 The impact of NOS2 on transient HCV replication observed in the present study could imply that the early steps of viral replication might be most sensitive toward NO. Alternatively, the antiviral effect of NOS2 on HCV might not be mediated by NO, but by another function of the protein.
Apart from NOS2, all other screening hits were similarly active under IFN-α and IFN-γ and regardless of the HCV replication system (i.e., transient and stable HCV replication). One of these candidates is IFIT3, a member of the IFN-induced protein family with tetratricopeptide repeats. IFIT3 was described to interact with TBK1, IRF3, and other IFIT family members, leading to enhanced signaling,23, 30 raising the question whether it is a direct or indirect inhibitor of HCV replication. Another candidate was TRIM14, a member of the tripartite motif (TRIM) protein family. Some TRIM family members are described to have important antiviral functions against HIV and flaviviruses31 through their E3 ubiquitin ligase activity. Because this domain is not present in TRIM14 its antiviral mode of action differs from that of other TRIMs and remains to be determined.
We also identified IFITM1 and IFITM3, recently described to target a broad range of RNA viruses32 at the level of entry or intracellular trafficking. In line with two recent studies,12, 18 we found that both candidates inhibited HCV replication. A combined knock-down of both IFITMs showed clearly synergistic effects, arguing for heteromeric complexes with increased activity. In support of this notion, the induction kinetics of the two molecules upon infection of PHHs with HCV were virtually identical, suggesting strict coregulation.
Our results suggest that IFN-induced inhibition of HCV is not mediated by a single factor, but rather by the concerted action of a plethora of ISGs. This was also implied by a recent study on HCV-specific IFN effector genes10 and is in contrast to antiviral response against other viruses, most notably orthomyxoviruses, for which the Mx protein has been shown to dominate the IFN-induced response.8 However, this “death by a thousand cuts” strategy against HCV might be significantly more effective in keeping a highly variable pathogen in check, as even complete resistance to a single effector would not profoundly affect the overall efficacy of the IFN response.
We thank Ulrike Herian and Stephanie Kallis for excellent technical assistance and Rick Randall for providing expression constructs.