Hepatitis C virus can attenuate IFN induction at multiple levels in infected hepatocytes, such as through the NS3/4A-mediated MAVS cleavage 
and by using the eIF2α kinase PKR to control IFN and ISG expression at the translational level 
. 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 (). By transcriptome analysis, we showed that HCV induces a number of ISGs in the HCV-permissive 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 
and in HCV-infected patients 
. 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 
or in the control of HCV replication in cell cultures 
. Our data provide an explanation for the presence of ISGs at high expression levels in HCV-infected patients 
and in models of HCV-infected chimpanzees 
in the absence of, or with poor IFN expression.
Multiple levels of control of IFN induction during HCV infection.
The 15 Kda ISG15, or Interferon Stimulated Gene 15 
, also known as ubiquitin cross reactive protein (UCRP) 
, can be conjugated (ISGylation) to more than 150 cellular protein targets 
through the coordinated action of three E1, E2 and E3-conjugating 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 
. The major E3 ligase for human ISG15 is HERC5 
Interestingly, RIG-I was identified as a target for ISG15, among other IFN-induced proteins or proteins involved in IFN action 
. 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 
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 
as well as in the maintenance of HCV in livers and pinpoint ISG15 as among the predictor genes of non-response to IFN therapy 
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 
, 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 (). 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 
. 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) 
. PKR expression was previously reported to be elevated in HCC liver from chronically HBV infected patients 
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 
in addition to its ability to also bind 5′ triphosphorylated ss or dsRNA structures 
. 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-κB through interaction of its C terminus with members of the TRAF family, such as TRAF5 and TRAF6 
. PKR contains also TRAF interacting motif in its N terminus 
and an association between TRAF3 and PKR has been reported upon cotransfection in 293T cells 
. Intriguingly, PKR was previously reported to participate in the induction of IFNβ, in association with MAVS, through activation of NF-κB or ATF-2 but not or partially IRF3; however these studies were not performed in the absence of RIG-I 
. 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 IFNβ 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.