Hepatitis C Virus Reveals a Novel Early Control in Acute Immune Response

Recognition of viral RNA structures by the intracytosolic RNA helicase RIG-I triggers induction of innate immunity. Efficient induction requires RIG-I ubiquitination by the E3 ligase TRIM25, its interaction with the mitochondria-bound MAVS protein, recruitment of TRAF3, IRF3- and NF-κB-kinases and transcription of Interferon (IFN). In addition, IRF3 alone induces some of the Interferon-Stimulated Genes (ISGs), referred to as early ISGs. Infection of hepatocytes with Hepatitis C virus (HCV) results in poor production of IFN despite recognition of the viral RNA by RIG-I but can lead to induction of early ISGs. HCV was shown to inhibit IFN production by cleaving MAVS through its NS3/4A protease and by controlling cellular translation through activation of PKR, an eIF2α-kinase containing dsRNA-binding domains (DRBD). Here, we have identified a third mode of control of IFN induction by HCV. Using HCVcc and the Huh7.25.CD81 cells, we found that HCV controls RIG-I ubiquitination through the di-ubiquitine-like protein ISG15, one of the early ISGs. A transcriptome analysis performed on Huh7.25.CD81 cells silenced or not for PKR and infected with JFH1 revealed that HCV infection leads to induction of 49 PKR-dependent genes, including ISG15 and several early ISGs. Silencing experiments revealed that this novel PKR-dependent pathway involves MAVS, TRAF3 and IRF3 but not RIG-I, and that it does not induce IFN. Use of PKR inhibitors showed that this pathway requires the DRBD but not the kinase activity of PKR. We then demonstrated that PKR interacts with HCV RNA and MAVS prior to RIG-I. In conclusion, HCV recruits PKR early in infection as a sensor to trigger induction of several IRF3-dependent genes. Among those, ISG15 acts to negatively control the RIG-I/MAVS pathway, at the level of RIG-I ubiquitination.These data give novel insights in the machinery involved in the early events of innate immune response.


Introduction
IFN induction in response to several RNA viruses involves the intracytosolic pathogen recognition receptor (PRR) CARDcontaining DexD/H RNA helicase RIG-I. Following its binding to viral RNA, RIG-I undergoes a change in its conformation through Lys63-type ubiquitination by the E3 ligase TRIM25. This allows its N-terminal CARD domain to interact with the CARD domain of the mitochondria-bound adapter MAVS [1,2]. MAVS then interacts with TRAF3 to further recruit downstream IRF3 and NF-kB-activating kinases, that stimulate the IFNb promoter in a cooperative manner. In addition, IRF3 stimulates directly the promoters of some interferon-induced genes (early ISGs) while NF-kB stimulates that of inflammatory cytokines [3].
The RNA of Hepatitis C virus (HCV) has an intrinsic ability to trigger IFNb induction through RIG-I [4,5,6]. Yet HCV is a poor IFN inducer. One reason for this comes from the ability of its NS3 protease to cleave MAVS [7]. Another relates to the ability of HCV to trigger activation of the dsRNA-dependent eIF2a kinase PKR [8,9] which leads to inhibition of IFN expression through general control of translation while the viral genome can be translated from its eIF2a-insensitive IRES structure [8].
HCV infection can trigger important intrahepatic synthesis of several IFN-induced genes (ISGs) in patients [10,11] and in animal models of infection in chimpanzees [12]. Expression of ISGs can be explained at least in part by the ability of HCV to activate the IFN-producing pDCs in the liver through cell-to-cell contact with HCV-infected cells [13]. Intriguingly, despite the recognized antiviral activity of a number of these ISGs, their high expression paradoxically represents a negative predictive marker for the response of these patients to standard combination IFN/ribavirin therapy [14,15,16]. The ubiquitine-like protein ISG15 is among the ISGs which are the most highly induced by HCV [16] and was recently shown to act as a pro-HCV agent [17]. Interestingly, ISG15 was also shown to control RIG-I activity through ISGylation [18].
Here, we show that HCV controls IFN induction at the level of RIG-I ubiquitination through the ubiquitine-like protein ISG15, one of the early ISGs. Use of small interfering RNA (siRNA) targeting to compare the effect of ISG15 to that of PKR on IFN induction and HCV replication led to the unexpected finding that HCV infection triggers induction of ISG15 and other ISGs by using PKR as an adapter through its N terminal dsRNA binding domain. This recruits a signaling pathway which involves MAVS,TRAF3 and IRF3 but not RIG-I. Altogether, our results present a novel mechanism by which HCV uses PKR and ISG15 to attenuate the innate immune response.

HCV infection negatively controls RIG-I ubiquitination
We recently reported that the HCV permissive Huh7. 25.CD81 cells [19] that we used to identify the pro-HCV action of PKR, did not induce IFN in response to HCV infection, unless after ectopic expression of TRIM25 [8]. We started this study by investigating at which level this defect could occur. A P 358 L substitution in the endogenous TRIM25 of these cells, revealed by sequence analysis, proved to have no incidence of the ability of TRIM25 to participate in the IFN induction process. Indeed, ectopic expression of a TRIM25 P 358 L construct was as efficient as a TRIM25wt construct to increase IFN induction in the Huh7. 25.CD81 cells, after infection with Sendai virus (SeV) ( Figure 1A). Like some other members of the TRIM family, TRIM25 is localized in both the cytosol and nucleus and is induced upon IFN treatment [20]. No specific difference between the cellular localization of TRIM25 was observed in the Huh7. 25.CD81 cells when compared to Huh7 cells or Huh7.5 cells, which rules out a role for a cellular mislocalization in its inability to participate in IFN induction ( Figure 1B). TRIM25 was also efficiently induced by IFN ( Figure 1B and Figure S1). We assayed whether increasing TRIM25 upon IFN treatment could mimic the effect of its ectopic expression and restore IFN induction in response to HCV infection. However, this resulted only in a poor stimulation of an IFNb promoter (3 to 5-fold), in contrast to its effect upon SeV infection (230-fold) ( Figure 1C). Similarly, HCV infection at higher m.o.i, as an attempt to favour recognition of RIG-I by the viral RNA, only modestly increased IFN induction ( Figure 1D). TRIM25 plays an essential role in IFN induction through RIG-I ubiquitination [1]. We then analysed whether this step was affected by HCV infection in the Huh7. 25.CD81 cells. The results showed that, in contrast to SeV infection used as control, HCV infection could not trigger RIG-I ubiquitination, unless the cells are supplied with ectopic TRIM25 ( Figure 1E). Thus, HCV infection appears to mediate a control on IFN induction through regulation of RIG-I ubiquitination.

HCV controls RIG-I ubiquitination through ISG15
Inhibition of the function of TRIM25 or RIG-I ubiquitination has been suggested to occur via the small ubiquitin-like protein ISG15 and the process of ISGylation [18,21]. We then analysed whether ISG15 was involved in the control of RIG-I ubiquitination upon HCV infection. For this, we chose a transient transfection approach using siRNAs targeting ISG15 in the Huh7.25.CD81 cells. Indeed, this resulted in a strong ubiquitination of RIG-I at 9 hrs and 12 hrs post-HCV infection, which was equivalent to that observed in cells supplied with ectopic TRIM25 (Figure 2A). A similar result was obtained after JFH1 infection in the Huh7 cells, used as another HCV-permissive cell line ( Figure  S2). Thus, ISG15 can control RIG-I ubiquitination in different cells infected by HCV. We next investigated whether ISGylation was involved in this process. Absence of detection of RIG-I ubiquitination after HCV infection of the Huh7.25.CD81 cells precludes direct analysis of the effect of ISG15 on RIG-I. We used an IFNb-luc reporter assay instead, as it proved to be sensitive enough to detect some IFN induction in response to JFH1 infection in those cells (see Figure 1D). We found that IFN induction increased when cells were transfected with siRNAs targeting ISG15 while it decreased in cells overexpressing IGS15 ( Figure 2B). Expression of ISG15 in the presence of the E1, E2 and E3 ligases involved in ISGylation (respectively Ube1L, UbcH8 and HERC5) [22] further inhibits IFNb induction ( Figure 2B). Similar results were observed upon infection with Sendai virus ( Figure S3). The ISGylation process is strictly dependent on the presence of the E1 ligase Ube1L [23]. Indeed, enhanced IFN promoter activity has been observed in Ube1L2/2 cells in response to NDV [18]. In accord with this, depletion of endogenous Ube1L from the Huh7.25.CD81 cells ( Figure S4), as such or after ectopic expression of ISG15, UbcH8 and HERC5, resulted in an increase in IFNb induction after infection with HCV ( Figure 2B). We then analysed the effect of siISG15 on IFNb induction after infecting the cells with HCV up to 72 hours, in order to pass through the 24 hr time-point where the signaling pathway leading to the transcription of this gene is expected to stop because of the NS3/4A-mediated cleavage of MAVS [8]. The results show that, whereas IFNb transcription was indeed strongly inhibited after 24 hr in the control cells, it still occurred significantly in the cells expressing siRNA ISG15 ( Figure 2C). Previous data have shown a positive role for ISG15 on HCV production [24,25]. In accord with this, silencing of ISG15 resulted in clear inhibition of HCV RNA expression with however no significant consequence on the ability of the virions produced to re-infect fresh cells ( Figure 2D). Analysis of expression of MAVS and NS3, as well as the expression of the core protein as another example of viral protein, then showed that the depletion of ISG15 both decreased and delayed the expression of the viral proteins as compared to the siRNA control cells and that this was correlated by a delay in the NS3/4A-mediated cleavage of MAVS ( Figure 2E). These results show that ISG15 controls the process of IFN induction during HCV infection by interfering with RIG-I ubiquitination through an ISGylation process and by boosting efficient accumulation of NS3, among other viral proteins, thus favouring its negative control on IFN induction by cleavage of MAVS.
ISG15 strengthens the pro-HCV activity of PKR ISG15 ( [24,25] and this study) and PKR [8,9] emerge as two ISGs with pro-HCV activities, instead of playing an antiviral role. We then assayed the effect of a combined depletion of PKR and ISG15 on HCV replication and IFN expression in the Huh7.25.CD81 cells. As shown in Figure 2 D and B, siRNAs targeting ISG15 were sufficient both to inhibit HCV replication ( Figure 3A) and to increase IFNb expression, either measured by RTqPCR ( Figure 3B) or by using an IFNb-luciferase reporter

Author Summary
Hepatitis C Virus (HCV) is a poor interferon (IFN) inducer, despite recognition of its RNA by the cytosolic RNA helicase RIG-I. This is due in part through cleavage of MAVS, a downstream adapter of RIG-I, by the HCV NS3/4A protease and through activation of the eIF2a-kinase PKR to control IFN translation. Here, we show that HCV also inhibits RIG-I activation through the ubiquitin-like protein ISG15 and that HCV triggers rapid induction of 49 genes, including ISG15, through a novel signaling pathway that precedes RIG-I and involves PKR as an adapter to recruit MAVS. Hence, we propose to divide the acute response to HCV infection into one early (PKR) and one late (RIG-I) phase, with the former controlling the latter. Furthermore, these data emphazise the need to check compounds designed as immune adjuvants for activation of the early acute phase before using them to sustain innate immunity.  Figure 3C). Very limited additional effect was observed in the concomitant presence of siRNAs targeting PKR. ( Figure 3B). Interestingly, we noticed that expression of luciferase from the IFNb promoter increased throughout the first 18 hours of HCV infection in the siISG15 cells ( Figure 3C). This was intriguing as it should have been inhibited after 12 hours of HCV infection through the eIF2a kinase activity of PKR and its control on translation [8]. We therefore analysed whether the state of PKR activation (phosphorylation) was dependent on the expression of ISG15. For this, the Huh7.25.CD81 cells were transfected either with siRNAs targeting ISG15 or with a plasmid expressing an HA-ISG15 construct and PKR phosphorylation was analysed as described previously [8]. The results showed that depletion of ISG15 inhibits PKR activation in the HCV-infected cells, while its overexpression stimulates it ( Figure 3D and Figure S5). Therefore these data reveal that, in addition to negatively controlling RIG-I ubiquitination, ISG15 can also positively control PKR activity. The conjugation of both effects results in an efficient control of IFN induction during HCV infection.

HCV triggers a PKR-dependent pathway early in infection to induce ISG15 and other genes
The Huh7.25.CD81 cells express ISG15 at significant basal levels. This situation was not surprising as various cellular systems can also express some of the ISGs at basal level. Expression of ISG15 was approximately 2-and 5-fold higher in the Huh7.25.CD81 cells than in the Huh7.5 or Huh7 cells (data not shown). Intriguingly however, we noticed that ISG15 expression was increased in response to HCV infection (see Figure 2E). To investigate this further, we simply re-used the RNAs prepared for the experiment shown in Figure 3B and performed a quantitative kinetics analysis. The results confirmed that HCV can trigger induction of ISG15 ( Figure 4A). Unexpectedly, analysis of the RNA extracted from the cells treated with siRNAs targeting PKR, revealed that ISG15 RNA expression was strongly repressed when PKR was silenced ( Figure 4A). This surprising result was confirmed by analysing induction of ISG56, another early ISG [26], both at the level of its endogenous RNA ( Figure 4B) or by using an ISG56-luciferase vector ( Figure 4C). In the latter case, a strong increase of the reporter expression in the cells treated with siRNAs targeting ISG15, was similar to the situation observed for IFNb RNA ( Figure 2B and 2C). This can be related to activation of the RIG-I pathway, which can function when ISG15 is absent. These data suggest that HCV may use PKR to activate gene transcription. Importantly, this phenomenon was specific to HCV as infection with Sendai virus resulted in a similar induction of ISG15 and ISG56, regardless of PKR ( Figure 4D and Figure  S6). We then examined whether overexpression of PKR could boost induction of ISG15 during HCV infection and how this would affect HCV replication and IFN induction, in relation to the pro-HCV action of ISG15. Huh7.25.CD81 cells were transfected with a plasmid expressing PKR alone or in presence of siRNAs targeting ISG15, before being infected with HCV over 48 hours. Overexpression of PKR increased the ability of HCV to induce ISG15 and concomitantly, led to an increase in HCV RNA expression. The latter increase was abolished when ISG15 was silenced, thus showing that the PKR-dependent increase in HCV expression is mediated by ISG15 ( Figure 4E). However, while the cells silenced for ISG15 are able to induce IFN in response to HCV infection, as shown in Figure 3B, they are unable to do so when PKR is overexpressed. This suggests that PKR may also interfere with the process of IFN induction, independently of ISG15, a possibility that remains to be explored.
A role for PKR in gene induction in response to HCV infection has not been described before. Additional information was therefore obtained through a transcriptome analysis of 2165 genes in the Huh7.25.CD81 cells treated with control siRNAs or siRNAs targeting PKR and infected with HCV for 12 hrs. Out of the most significant 422 genes that were identified, 99 were unmodified or barely modified and 33 were down-regulated, while 290 genes were found to be up-regulated by HCV infection (data not shown). Among those, HCV infection triggered up-regulation of 49 genes which are directly dependent on PKR expression ( Table 1). Forty percent of these genes (20) belong to the family of the ISGs, with ISG15 among the most induced genes ( Table 1). In the reciprocal situation, only 17 genes depended on PKR for their downregulation by HCV infection, with no link to a particular family of genes and limited variation both in number and intensity (Table  S1). Thus, induction of ISGs upon HCV infection may occur through a novel signaling pathway that involves PKR.
Induction of ISG15 by HCV is independent of RIG-I, involves MAVS/TRAF3 association with PKR and involves the DRBD region but not the catalytic activity of PKR Infection with RNA viruses or transient transfection with dsRNA can directly and rapidly induce early ISGs, such as ISG15, through IRF3, after activation of the RIG-I/MAVS pathway and recruitment of TRAF3, an essential adapter which recruits the downstream IRF3 kinases TBK1/IKKe. We have shown that the RIG-I pathway was not operative during HCV infection in the Huh7.25.CD81 cells, precisely due to the presence of ISG15. To determine how ISG15 induction through PKR relates to or differs from the RIG-I/MAVS pathway, the Huh7.25.CD81 cells were treated with siRNAs aimed at targeting separately PKR, RIG-I, MAVS, TRAF3 and IRF3 ( Figure S7) and infected with HCV. The results clearly showed that induction of ISG15 in response to HCV infection depends on PKR, MAVS, TRAF3 and IRF3 but not on RIG-I ( Figure 5A). The participation of IRF3 was further confirmed by immunofluorescence studies which showed its nuclear translocation at 6 hours post-infection ( Figure S8). ISG15, as well as ISG56, was also clearly induced in response to HCV infection in two other HCV permissive cell lines, such as Huh7 and Huh7.5 cells, and this induction was abrogated in presence of siRNAs targeting PKR ( Figure 5B and Figure S9). Importantly, since Huh7.5 cells express a non-functional RIG-I/ MAVS pathway due to a mutation in RIG-I, result with these cells supports the notion that the ability of HCV to trigger induction of ISGs through PKR is independent of RIG-I. To have more insights on this novel PKR signaling pathway, PKR was immunoprecipitated at early times points following infection of Huh7.25.CD81 cells with HCV and the immunocomplexes were analysed for the presence of MAVS, TRAF3 and RIG-I. Both MAVS and TRAF3, but not RIG-I, associate with PKR in a time dependent manner, beginning at 2 hrs post-infection ( Figure 5C). Strikingly, these associations were abrogated by the cell-permeable peptide PRI which is analogous to the first dsRNA binding TRIM25 after IFN treatment was shown by immunoblot (C). (E) The Huh7.25.CD81 cells were transfected for 48 hrs with 5 mg of His-Myc-Ubiquitin expression plasmid in absence or presence of a plasmid expressing HA-TRIM25 and infected with SeV (40 HAU/ml) or HCV (m.o.i = 6). At the times indicated, 10% of the lysate was precipitated with TCA and the remaining lysate subjected to nickel pulldown under denaturing conditions. Total and ubiquitin (Ub)-modified proteins were separated by SDS-PAGE and revealed by immunoblot. doi:10.1371/journal.ppat.1002289.g001 domain (DRBD) of PKR [8], while unaffected by C16, a chemical compound which inhibits the catalytic activity of PKR ( Figure 5D). In line with this, PRI but not C16, abrogated the ability of HCV to induce ISG15 ( Figure 5E). The same result was obtained for induction of ISG56 ( Figure S10). We then used human primary hepatocytes (HHP) to determine whether HCV was also able to induce ISGs through PKR in a more physiological cellular model. A follow-up of the infection over a period of 96 hours showed that JFH1 was replicating correctly in those cells as well as leading to induction of ISG15 (10-fold) and to some induction of IFNb (2.5-fold). These cells were infected with JFH1 for 8 hours in the absence or presence of PRI, making convenient copies (siISG15). Supernatants collected at different times post-infection were used to infect fresh cells. After 24 hours, the RNAs were extracted from the cells and expression of HCV RNA was determined by RTqPCR. (E) Huh7.25.CD81 cells, transfected with 25 nM of siRNA (Control or ISG15) for 48 hrs, were infected with JFH1 for the times indicated. Cell extracts were analysed by immunoblot with Abs directed against ISG15, MAVS, the HCV NS3 and core proteins and Actin as loading control. doi:10.1371/journal.ppat.1002289.g002 2) for the times indicated. In each case, IFN expression was expressed as fold-induction over control cells that were simply transfected with pGL2-IFNb-FLUC/pRL-TK-RLUC. The graph represents the level of firefly luciferase activity normalized to the ratio R-luc RNA/GAPDH RNA. Such normalization is required because of the negative control of general translation through PKR after 12 hrs post-infection [8]. Error bars represent the mean 6S.D for triplicates. (D) Huh7.25.CD81 cells, in 100 cm 2 plates, were transfected with siRNA Control or siRNA ISG15 or transfected with a plasmid expressing HA-ISG15 for 48 hrs and infected with JFH1 (m.o.i = 6). At the indicated times post-infection, cell extracts (2.2 mg) were processed for immunoprecipitation of PKR or for incubation with mouse IgG as a control of specificity (asterisk). The immunoprecipitated complexes were run on two different NuPAGE gels and blotted using Mab 71/10 or anti-phosphorylated PKR antibodies (PKR-P). The presence of PKR and PKR-P was revealed using the Odyssey procedure. The ratio PKR-P/PKR in the absence or in the presence of ISG15, either endogenous or endogenous and ectopic, is shown in Figure S5  use of the cell-penetrating ability of this peptide. Longer period of treatment with PRI were not investigated for practical reasons (see Materials and Methods). The results showed that PRI was significantly inhibiting the induction of ISG15 while it had no effect on that of IFNb ( Figure 5F). Altogether, these data demonstrate that HCV triggers induction of early ISGs through MAVS and TRAF3 by using PKR as an adapter protein.

PKR interacts both with MAVS and TRAF3 and binds HCV RNA ahead of RIG-I
The ability of HCV to control activation of the RIG-I/MAVS pathway after induction of ISG15 through a novel PKR/MAVS pathway suggests that PKR has the possibility to bind MAVS prior to RIG-I. To determine this, we established the kinetics of these interactions, after treating the Huh7.25.CD81 cells with siRNAs targeting ISG15 prior to HCV infection. This was necessary in view of the negative control of ISG15 on RIG-I. MAVS was immunoprecipitated from the cell extracts at different times postinfection and the presence of PKR and RIG-I was examined in the immunocomplexes, as well as that of TRAF3, used as marker of activation of the MAVS signaling pathway. As expected, only PKR was able to associate with MAVS and TRAF3 in the control cells ( Figure 6A) whereas both PKR, RIG-I and TRAF3 were found in the immunocomplexes in the absence of ISG15 ( Figure 6B). The PKR/MAVS association took place at 4 hrs post-infection in the control cells but was observed 2 hrs earlier in the ISG15-depleted cells. Whether ISG15 plays a role in the regulation of the PKR/MAVS association remains to be determined. However, the presence of TRAF3 in association with MAVS at 2 hrs post-infection in the control cells ( Figure 6A) correlates with its association with PKR ( Figure 5C) which indicates that the MAVS pathway can be activated through PKR as soon as 2 hrs post infection. In ISG15 knock-down cells, the RIG-I/MAVS association occurred later at 6 hrs post-infection with an increase in TRAF3 association at 9-12 hrs post infection. Altogether, these data revealed that HCV infection triggers an earlier interaction of MAVS with PKR than with RIG-I.
Finally, we asked whether PKR was able to associate with HCV RNA and how this association can be compared to that of RIG-I. PKR and RIG-I were immunoprecipitated at 2, 4 and 6 hrs postinfection and the presence of HCV RNA was analysed in the complexes. The results showed that PKR associates with HCV RNA with best efficiency at 2 hrs post-infection. Importantly, this association was strongly inhibited in presence of PRI, thus confirming the importance of PKR DRBD in the process. In contrast, the association of HCV RNA with RIG-I was detected only at 6 hrs post-infection. Interestingly, the association between RIG-I and HCV RNA was not affected by PRI, which rules out the possibility that the initial formation of a complex between PKR and HCV RNA was a pre-requisite for the subsequent binding of RIG-I to HCV RNA. Immunoprecipitation of PKR at 1, 2, 4 and 6 hrs post-infection, in presence of an inhibitor of ribonucleases also did not lead to detection of RIG-I in the complexes ( Figure S11). Association of HCV RNA with eIF2a, used as negative control, was not significant, thus showing the specificity of the assay ( Figure 6C). Whether a direct interaction of PKR with HCV RNA represents the initial event leading to the MAVS-dependent induction of early ISGs remains now to be characterized. Altogether, these data reveal an earlier mobilization of PKR than RIG-I in response to HCV infection which leads to activation of a MAVS-dependent signaling pathway.

Discussion
Hepatitis C virus can attenuate IFN induction at multiple levels in infected hepatocytes, such as through the NS3/4A-mediated MAVS cleavage [7,27] and by using the eIF2a kinase PKR to control IFN and ISG expression at the translational level [8,9]. Here, we have identified another process by which HCV controls IFN induction at the level of RIG-I ubiquitination through ISG15 and an ISGylation process. Importantly, we have shown that ISG15 is rapidly induced, among other ISGs, in response to HCV infection, through a novel signaling pathway that involves PKR, MAVS, TRAF3 and IRF3 but not RIG-I. In this pathway, PKR is not used for its kinase function but rather as an adapter protein with its dsRNA binding domain (DRBD) playing an essential role in this mechanism (Figure 7). By transcriptome analysis, we showed that HCV induces a number of ISGs in the HCVpermissive Huh7.25.CD81 cells and we confirmed the induction of two of these, ISG15 and ISG56, in other HCV-permissive cells, such as Huh7.5 and Huh7 cells. In addition, induction of ISG15 by HCV in a PKR-dependent manner was confirmed in human primary hepatocytes. The ability of HCV to trigger high expression levels of ISG15 and ISG56, as well as other ISGs, has previously been reported in models of HCV-infected chimpanzees [10,12,28] and in HCV-infected patients [14,15,16]. Induction of ISGs thus represents a general propriety of the response of the cells to HCV. In addition to this, natural variations in intra-hepatic levels of ISG15 in vivo may increase the susceptibility of some patients to HCV infection. The ability of HCV to control RIG-I activity through ISG15 is important to note in view of several reports which highlight the importance of a role for ISG15 in the maintenance of HCV in livers [15,16] or in the control of HCV replication in cell cultures [17,25]. Our data provide an explanation for the presence of ISGs at high expression levels in HCV-infected patients [14,15,16] and in models of HCVinfected chimpanzees [10,12,28] in the absence of, or with poor IFN expression.
The 15 Kda ISG15, or Interferon Stimulated Gene 15 [29], also known as ubiquitin cross reactive protein (UCRP) [30], can be conjugated (ISGylation) to more than 150 cellular protein targets [31] through the coordinated action of three E1, E2 and E3conjugating enzymes, in a process similar but not identical to ubiquitination. While both ubiquitin and ISG15 can use the same E2 enzyme UbcH8, Ube1L functions as a specific E1 enzyme for ISG15, in spite of its 45% identity with Ube1, the E1 enzyme for ubiquitin [32]. The major E3 ligase for human ISG15 is HERC5 [33].
Interestingly, RIG-I was identified as a target for ISG15, among other IFN-induced proteins or proteins involved in IFN action [31]. However, its activity appears to be negatively controlled by ISG15 and the ISGylation process, either as shown previously after cotransfection with the ISG15 and the ISG15-conjugating enzymes [18] or as shown here, in a model of infection with HCV. Indeed, ISG15 is now emerging as playing a proviral role in case of HCV infection. Several reports now highlight the importance of a role for ISG15 in the control of HCV replication in cell cultures [17,25] as well as in the maintenance of HCV in livers and siISG15 and with 1 mg of a plasmid expressing PKR where indicated. After 48 hrs, the cells were infected with HCV (m.o.i = 6) for the times indicated. Expression of HCV, ISG15 and IFNb RNA was determined by RTqPCR. Cell lysates prepared from cells treated in the same conditions but not infected were used to control expression of PKR and ISG15 by immunoblot. doi:10.1371/journal.ppat.1002289.g004 pinpoint ISG15 as among the predictor genes of non-response to IFN therapy [14,15,16]. At present, we do not know at which level ISGylation regulates IFN induction in response to HCV infection. An HCV-mediated increase of ISG15 would favour preferential binding of ISG15 over that of ubiquitin to the E2 enzyme UbcH8 and hence enhance the spatio-temporal availability of UbcH8-ISG15 for HERC5 over that of UbcH8-ubiquitin for TRIM25. It may also lead to inhibition of TRIM25, through autoISGylation [21,34], which would decrease its ability to ubiquitinate RIG-I. We showed that overexpression of HERC5 together with Ube1L, UbcH8 and ISG15 was increasing the ability of ISG15 to inhibit IFN induction by HCV ( Figure 2B). All three enzymes Ube1L, UbcH8 and HERC5 belong to the family of genes induced by IFN and it has been reported that ISGylation is optimum in a context of IFN treatment [18,35]. Therefore, it is tempting to speculate that elevated levels of ISG15 in some HCV-infected patients would bring the most favourable context for the virus when those patients are under IFN therapy. This would be in accord with the clinical data showing that HCV-induced high expression of ISG act as a negative predictive marker for response to IFN therapy.
It is doubtful that viruses with high IFN-inducing efficiency, such as Sendai virus may control RIG-I through ISG15 and PKR. However, viruses that avoid inducing IFN may have use of the PKR pathway. A good example might be that of Hepatitis B Virus (HBV) [36,37,38]. PKR expression was previously reported to be elevated in HCC liver from chronically HBV infected patients [39] and a relationship between PKR and IFN induction during HBV infection would be important to evaluate.
At present, we have established that HCV RNA interacts with PKR as soon as 2 hours post-infection. This interaction occurs prior the interaction of HCV RNA with RIG-I, which suggests that PKR may rapidly detect structures containing the incoming HCV RNA genome. Indeed, PKR has been reported to bind the dsRNA domains III and IV of HCV IRES [40] in addition to its ability to also bind 59 triphosphorylated ss or dsRNA structures [41]. Whether PKR behaves as a pathogen recognition receptor for HCV RNA, like RIG-I, remains to be clarified. It is however clear that, in contrast to RIG-I, PKR acts here in favour of the pathogen rather than in favour of the host defense. We have established that the HCV RNA/PKR interaction depends on the first DRBD present at the N terminus of PKR and is independent on its kinase activity. The ability of PKR to serve as adapter in signaling pathways is not a total surprise since it has been previously shown to activate NF-kB through interaction of its C terminus with members of the TRAF family, such as TRAF5 and TRAF6 [42]. PKR contains also TRAF interacting motif in its N terminus [42] and an association between TRAF3 and PKR has been reported upon cotransfection in 293T cells [43]. Intriguingly, PKR was previously reported to participate in the induction of IFNb, in association with MAVS, through activation of NF-kB or ATF-2 but not or partially IRF3; however these studies were not performed in the absence of RIG-I [44,45,46]. The mode of interaction between PKR, TRAF3 and MAVS, independently of RIG-I, and how it leads to a preferential induction of ISGs and not of IFNb in response to HCV infection in contrast with the RIG-I/ MAVS pathway remains to be determined. Based on our data, we propose now to divide the innate response to acute HCV infection into two phases: an early acute phase in which PKR is activated and a late acute phase that depends on RIG-I, the early phase controlling activation of the late phase. It is now essential to progress towards the generation of specific pharmaceutical inhibitors targeting PKR in order to abrogate the early acute phase to the benefit of the RIG-I-driven late phase. In a more general view, care should now be taken in the choice of compounds designed to be used as immune adjuvants, such as to be devoid of activation of the early acute PKR phase. This will ensure their efficiency as to activate properly the innate immune response through the late acute RIG-I phase.

Cell cultures and viruses
The culture of Huh7, Huh7.5, Huh7.25.CD81 cells, the preparation of Sendai virus stocks (<2000 HAU/ml) and of HCV JFH1 stocks (<6.10 4 FFU/mL and <6.10 6 FFU/mL) was as described [8,47]. Preparation and cultures of human primary hepatocytes was as described [48]. Of note, the ability of the Huh7.25.CD81 cells to induce IFN in response to SeV without prior IFN treatment (40-fold) was not observed in our previous study [8].The ability of Sendai virus to induce IFN is related to the presence of copyback DI (Defective Interfering) genomes [49]. The higher IFN inducing ability of the novel Sendai virus stock may have come from an important accumulation of these copyback DI genomes, during its growth in chicken eggs.

PKR inhibitors
The C16 compound [50] and the cell-permeable PRI peptide [51] were provided by Jacques Hugon. These drugs were applied (200 nM for C16 and 30 mM for PRI) one hour before the end of the 2 hr-incubation time with JFH1 and re-added to the medium after washing the cells with phosphate buffered saline (PBS). Note that PRI loses its effect very rapidly, probably through degradation in the cells, and requires to be added every hour to the cells until the end of treatment.

Expression vectors
TRIM25 was cloned from the IFN-treated Huh7.25CD81 cells (500 U/ml IFN-a2a; Cellsciences) after RT-PCR using the forward: 59-ATGGCAGAGCTGTGCCCCCT-39 and reverse 59-CTACTTGGGGGAGCAGATGG-39 primers. The pcDNA3.1(+) vector expressing 59HA tagged-TRIM25 (provided by D. Garcin; University of Geneva, Switzerland) was used to generate the TRIM25 P 358 L construct by site-directed mutagenesis. The The Huh7.25.CD81 cells, seeded at 3.10 6 cells in 10-cm plates, were transfected after 24 hrs with 25 nM of siRNA Control or siRNA PKR using Fugene HD. 24 hrs post transfection, they were either mock-infected or infected for 2 hrs at 37uC with JFH1 (moi = 0.2) (three independent plates/sample). The medium was then removed and cells were incubated with complete DMEM for 12 hrs at 37uC. The cells were washed twice with TBS containing phosphatase and protease inhibitors, harvested by scraping, the cell pellets were centrifuged, the supernatants were removed and the pellets were frozen and stored at 280uC before being processed for micro-array.The list shows genes that were affected no more than twice by the depletion of PKR in the control cells (0.5, siPKR mock/siCt ,1.6). The dependence of each of these genes in regards with PKR for their induction by HCV is expressed as log2 (ratio (siPKR HCV/siCt Mock)2(siCt HCV/siCt Mock) (indicated by log2*) with a cut-off of <2.0 fold. doi:10.1371/journal.ppat.1002289.t001   were incubated with anti-MAVS antibodies. In addition, cell extracts prepared at 0 hr post-infection were incubated with mouse IgG as a control of specificity (asterisk). The immunoprecipitated complexes were run on three different NuPAGE gels and blotted using Mab 71/10, anti-MAVS, anti-RIG-I or anti-TRAF3 antibodies. The expression level of each protein was controlled in the total cell extracts. (C)-Huh7.25.CD81 cells were incubated with JFH1 (m.o.i = 6) for 2 hrs at 37uC or at 4uC in the absence or presence of 30 mM of PRI. This drug was applied one hour before the end of the incubation time. After washing the cells twice with PBS, the cells were further incubated for 2, 4 or 6 hrs at 37uC or at 4uC in the absence or presence of PRI (added every hour). The cell extracts were processed for crosslinking of RNA to proteins before lysis, as described in Materials and Methods and different immunoprecipitations were performed with antibodies directed against PKR, RIG-I or eIF2a. After extensive washing, the presence of HCV RNA linked to the immunocomplexes was analysed by RTqPCR and the presence of the proteins was verified by Western blot. Measure of HCV RNA in IFNb-firefly luciferase (pGL2-IFNb) and pRL-TK Renillaluciferase reporter plasmids were described previously [8]. The pGL3 luciferase reporter construct containing the 23 to 2654 nucleotides of the ISG56 promoter was provided by N.Grandvaux [52]. The Myc-HIS-Ubiquitin construct was provided by R.Kopito (Stanford University, CA). ISG15 was cloned from IFN-treated Huh7 cells using the forward: 59-GGATCCCATGGGCT-GGGACCTGACGGTG-39 and reverse 59-CTCGAGCTCC-GCCCGCCAGGCTCTGT-39 primers and inserted into the pcDNA3.1(+)HA vector. The Ube1L, UbcH8 and HERC5 constructs were kindly provided by Jon M. Huibregtse [35]. The pcDNA1/AMP vector expressing PKR has been described previously [53].

RNA-mediated interference
The siRNAs directed against PKR, MAVS, RIG-I, TRAF3 and IRF3 which were used for the experiment described in figure 5A correspond to pools of siRNA (Smartpool) obtained from Dharmacon Research, Inc. (Lafayette, CO), as well as siRNAs directed against Ube1L used in Figure 2B. Control (scrambled) siRNA and siRNA directed against PKR or ISG15, used in all other experiments, were chemically synthesized by Dharmacon the cell extracts allowed to estimate its percentage of binding to PKR as 1.09%, 0.47% and 0.25% at 2, 4 and 6 hrs post-infection respectively, and its percentage of binding to RIG-I as 0.34% at 6 hrs post-infection. doi:10.1371/journal.ppat.1002289.g006 Figure 7. Multiple levels of control of IFN induction during HCV infection. Soon after infection, the HCV RNA is detected by the dsRNA binding domains (DRBD) of PKR ahead (2 hr) of its recognition by the RNA helicase RIG-I (6 hr). Recruitment of PKR by HCV triggers a signaling pathway that involves PKR as an adapter protein to recruit MAVS and TRAF3. This leads to a strong induction of the di-ubiquitine like protein ISG15 as well as other IRF3-dependent ISGs (Interferon-Stimulated Genes). ISG15 negatively controls the TRIM25-mediated ubiquitination (Ub) of RIG-I through an ISGylation process and thus interferes with the ability of RIG-I to recruit its downstream partners, including MAVS and TRAF3, and to induce IFNb and ISGs. As the infection proceeds, HCV activates the eIF2a kinase function of PKR (12 hr). This leads to a transient (few hours) inhibition of general translation, including that of IFN [8] and ISGs [9] while the eIF2a-independent translation of the viral proteins proceeds unabated. At later times in the infection (18 hr), additional control of IFN induction occurs through cleavage of MAVS by the HCV NS3/4A protease, once the viral proteins have sufficiently accumulated in the cytosol [7,27]. doi:10.1371/journal.ppat.1002289.g007 (scrambled and PKR) and by EUROFINS MWG Operon (ISG15) (Text S1). The siRNAs (final concentration 25 nM or 50 nM) were transfected for 48 h using jetPRIME reagent according to the manufacturer's instructions (PolyPlus transfection TM) before transfection with other plasmids or before infection.

Real-time RT-PCR analysis
Total cellular RNA was extracted using the TRIZOL reagent (Invitrogen). HCV RNA was quantified by one-step RTqPCR. Reverse-transcription, amplification and real-time detection of PCR products were performed with 5 ml total RNA samples, using the SuperScript III Platinum one-step RTqPCR kit (Invitrogen) and an AbiPrism 7700 machine. For the sequence of the different primers, see Text S1. The results were normalized to the amount of cellular endogenous GAPDH RNA using the GAPDH control kit from EuroGentec. Copies number of HCV RNA may vary due to internal calibration and depending on the preparation of the viral stocks. All m.o.i were calculated using the titers expressed in FFU/ml. The IFNb, ISG15, ISG56, Ube1L and GAPDH amplicons were quantified by a two-step RTqPCR assay as described [8].

Transcriptome analysis
Cellular RNA was extracted and purified from the cells using RNAeasy mini kit (QIAGEN K.K., Tokyo, Japan). Comprehensive DNA microarray analysis was performed with 3D-Gene Human Oligo chip25k with 2-color fluorescence method by New Frontiers Research Laboratories, Toray Industries Inc, Kamakura, Japan as previously described [56]. In brief, each sample was hybridized with 3D-Gene chip. Hybridization signals were scanned using Scan Array Express (PerkinElmer, Waltham, MA). The scanned image was analyzed using GenePix Pro (MDS Analytical Technologies, Sunnyvale, CA). All the analyzed data were scaled by global normalization.

Immunoprecipitation and immunoblot analysis
Cells were washed once with PBS and scraped into lysis buffer 1 (50 mM TRIS-HCl [pH 7.5], 140 mM NaCl, 5% glycerol, 1% CHAPS) that contained phosphatase and protease inhibitors (Complete, Roche Applied Science). The protein concentration was determined by the Bradford method. For immunoprecipitation, lysates were incubated at 4uC overnight with the primary antibodies as indicated and then in the presence of A/G-agarose beads (Santa Cruz Biotechnology) for 60 minutes. The beads were washed three times, and the precipitated proteins were extracted at 70uC using NuPAGE LDS sample buffer. Protein electrophoresis was performed on NuPAGE 4-12% Bis TRIS gels (Invitrogen). Proteins were transferred onto nitrocellulose membranes (Biorad), and probed with specific antibodies. Fluorescent immunoblot images were acquired and quantified by using an Odyssey scanner and the Odyssey 3.1 software (Li-Cor Biosciences) as described previously [8]. For detection of ISG15, cells were lysed in RIPA buffer (50 mM TRIS-HCl [pH 8.0]; 200 mM NaCl; 1% NP-40; 0.5% Sodium Deoxycholate; 0.05% SDS; 2 mM EDTA) and protein electrophoresis was performed on 4-20% polyacrylamide gels (PIERCE).

Nuclear/cytoplasmic extract
Pellets from cells washed in ice-cold phosphate-buffered saline (PBS) were lysed in ice-cold cytoplasmic buffer (10 mM TRIS [pH 8.0], 5 mM EDTA, 0.5 mM EGTA, 0.25% Triton X-100) containing phosphatase and protease inhibitors. The suspension was centrifuged for 30 seconds at 14,000 g and the supernatant (cytoplasmic fraction) was transferred into microcentrifuge tubes. The nuclear pellet was resuspended in Urea buffer (8 M Urea, 10 mM TRIS [pH 7,4], 1 mM EDTA, 1 mM dithiothreitol) containing phosphatase and protease inhibitors, homogenized by vortex and boiled for 10 minutes. The protein concentration was determined by the Bradford method.

Ubiquitination assay
Huh7.25.CD81 cells were transfected for 48 hrs with 5 mg of Myc-His-Ubiquitin expression plasmid using jetPRIME reagent. The cells were then washed in ice-cold PBS containing 20 mM N-ethylmaleimide (Sigma-Aldrich), harvested directly in Gua8 buffer (6 M guanidine-HCl, 300 mM NaCl, 50 mM Na 2 HPO 4 , 50 mM NaH 2 PO 4 [pH 8.0]), briefly sonicated, and centrifuged at 14,000 g for 15 min at 4uC. 1/10th of the lysate was subjected to precipitation with 10% trichloroacetic acid for protein analysis in whole cell extracts. The rest of the lysate was incubated for 2 hrs with 20 ml (packed volume) of Talon resin Ni-affinity beads (Clontech) on a rotating wheel. Bound proteins were washed four times in Gua8 buffer, three times in Urea 6.3 buffer (8 M Urea, 10 mM TRIS, 0.1 M Na 2 HPO 4 , 20 mM Imidazole [pH 6.3]), and three times in cold PBS, after which they were eluted by boiling in NuPAGE LDS sample buffer. Electrophoresis was performed on 4-12% of acrylamide NuPAGE gels (Invitrogen). The three sets were centrifuged for 30 seconds at 14,000 g and 4uC and the cell pellets were lysed into lysis buffer 1 containing phosphatase/protease and RNAse inhibitors (Promega) for sets 1 and 2 or into TRIZOL reagent for set 3. Cell lysates from sets 1 and 2 were then incubated at 4uC, first overnight with the appropriate primary antibodies and for 60 minutes in the presence of A/G-agarose beads (Santa Cruz Biotechnology). After the incubation period, the beads were washed four times with buffer 1. Set 1 (HCV RNA bound to immunocomplexes) and set 3 (input HCV RNA) were submitted to TRIZOL treatment and HCV RNA was quantified by one-step RTqPCR as described previously. The immunoprecipitated proteins from set 2 were extracted at 70uC using NuPAGE LDS sample buffer and analysed by immunoblot after electrophoresis on 4-12% of acrylamide NuPAGE gels (Invitrogen).

Supporting Information
Figure S1 Efficient induction of TRIM25 by IFN in the Huh7.25.CD81 cells. Huh7.25.CD81 cells, seeded at 8610 4 cells in 24-well plates containing coverslips, were treated with 500 U/ml of IFNa for 24 hrs (IFN) or left untreated (Cont). Cells were fixed with 4% PFA and TRIM25 was detected using anti-TRIM25 antibodies (red). Nuclei are shown in blue after DAPI labelling. Microscope magnification was 663. Huh7.25.CD81 cells, in 100 cm 2 plates, were transfected with siRNA Control or siRNA ISG15 or transfected with a plasmid expressing HA-ISG15 for 48 hrs and infected with JFH1 (m.o.i = 6). At the indicated times post-infection, cell extracts (2.2 mg) were processed for immunoprecipitation of PKR. The immunoprecipitated complexes were run on two different NuPAGE gels and blotted using Mab 71/10 or anti-phosphorylated PKR antibodies (PKR-P). The presence of PKR and PKR-P was revealed using the Odyssey procedure. The bands corresponding to total PKR and phosphorylated PKR were quantified using the Odyssey software and expressed as the ratio PKR-P/ PKR in the absence (siISG15) and in the presence of ISG15 in the control cells (Control) or after transfection of the ISG15 expressing plasmid (HA-ISG15). (PDF) Huh7.25.CD81 cells, seeded at 10 5 cells in 24-well plates containing coverslips, were infected for different times (0, 4 and 6 hours) at 37uC with JFH1 (moi = 6) or with SeV (40 HAU/ml) in the absence or in the presence of 10 ng/ml Leptomycine B (LB; Sigma), which was used here as a convenient mean to enhance the nuclear detection of IRF3 since it can interfere with nuclear export [57]. Cells were fixed with 4% PFA and IRF3 was detected using anti-IRF3 antibodies (red). Nuclei are shown in blue after DAPI labelling. The arrows show the presence of IRF3 in the nucleus. Microscope magnification was 663. (PDF) Figure S9 Induction of ISG56 by HCV in the Huh7.5 and Huh7 cells depends on PKR. Huh7.5 or Huh7 cells were transfected with siRNA Control or siPKR (50 nM) for 48 hrs and infected with JFH1 (m.o.i = 0.2 for Huh7.5 or 10 for Huh7). At the times indicated, expression of endogenous ISG56 was determined by RTqPCR and expressed as fold induction. Error bars represent the mean 6S.D for triplicates. The expression levels of ISG56 RNA at the start of infection in the siControl cells was 1.37610 4 copies (Huh7.5 cells) and 1.28610 4 copies (Huh7 cells). (PDF) Figure S10 Induction of ISG56 by HCV is specifically inhibed by the PKR inhibitor PRI. The Huh7.25.CD81 cells were incubated with PRI or C16 and infected with JFH1 (m.o.i = 0.2) for the times indicated. RTqPCR analysis of endogenous ISG56 was determined by RTqPCR and expressed as fold induction. The expression levels of ISG56 RNA at the start of infection in the control cells was 1.97610 4 copies. (TIF) Figure S11 The RNAse inhibitor RNAsin does not favour the formation of a RIG-I/PKR complex upon HCV infection. Two sets of Huh7.25.CD81 cells were plated into 100 cm 2 plates and infected with JFH1. At the times indicated, cell extracts (3.5 mg) from the two sets were processed similarly for immunoprecipitation of PKR or for incubation with mouse IgG as a control of specificity (asterisk), except that care was taken to add the RNAse inhibitor RNAsin (1000 U/ml) at all steps for the second set (+RNAsin). Detection of RIG-I, MAVS, and PKR in the complexes and in the whole cell extracts (WCE) was revealed by immunoblot using the Odyssey procedure. Detection of Actin in WCE served as loading control. (PDF) Table S1 Transcriptome analysis of PKR-dependent downpregulated gene upon 12 hrs of HCV infection. Preparation of samples was as described under Table 1. The list shows genes that were affected no more than twice by the depletion of PKR in the control cells (0.5, siPKR mock/siCt ,1.6). The dependence of each of these genes in regards with PKR for their inhibition by HCV is expressed as log2 (ratio (siPKR HCV/siCt Mock)2(siCt HCV/siCt Mock) (indicated by log2*) with a cut-off of <2.0 fold. (DOC) Text S1 Supplementary methods. (DOC)