Functional genomic approaches allow for the unbiased identification of HCV host cofactors independent of any preconceived models of the HCV lifecycle or any assumptions about gene function. This technology has been successfully applied to the identification of novel host dependency factors for HIV (Brass et al., 2008
). We have conducted a whole-genome RNAi screen for host proteins that support HCV replication and were able to functionally confirm the significance of several of these cofactors in the fully infectious JFH1 HCV strain.
HCV, like other positive-strand RNA viruses, replicates on altered host membranes, which in the case of HCV have been called “membranous webs” (Egger et al., 2002
; Gosert et al., 2003
). The mechanisms by which these membranous webs form are largely unknown. Here, we identified 27 host genes that play a role in lipid metabolism, membrane biogenesis, kinetics, and trafficking; all are likely to be key processes in the formation and maintenance of the membranous web structure.
PI4KA is an attractive “druggable” target for anti-HCV therapeutics. Both genotype 1 and 2 HCV models exhibit a very strong dependency on PI4KA function; furthermore, PI4KA
silencing is well tolerated in cultured cells. Further evidence that PI4KA has a direct function in HCV replication comes from a study (Ahn et al., 2004
) that identified an interaction between HCV NS5A and PI4KA in a yeast two-hybrid screen. The development of selective PI4KA inhibitors will facilitate the study of this protein’s cellular function as well as its role in HCV replication, and it may also yield therapies for chronic HCV infection.
The dependence of HCV replication on the COPI coatomer complex is shared by poliovirus and Drosophila
C virus (Cherry et al., 2006
; Maynell et al., 1992
). All three of these RNA viruses replicate on host membrane-derived compartments, and we speculate that the assembly of these viral replication compartments is somehow dependent on COPI. Further support for this hypothesis comes from the finding that γ-COP and its interacting partner, CDC42, are significantly enriched on detergent-insoluble membranes from HCV replicon-expressing cells (Mannová et al., 2006
). These data would suggest that COPI is directly involved in membranous web formation and that the block of HCV replication by COPI inhibition is not an indirect effect of disrupting Golgi-ER vesicle trafficking.
Finally, we identified hepcidin as a cofactor for HCV replication. Chronic hepatitis C infection has long been recognized to be associated with increased serum iron and transferrin saturation as well as hepatic iron accumulation; moreover, hepatic hepcidin mRNA expression is increased in patients with chronic HCV infection (Aoki et al., 2005
). Hepcidin transcription is stimulated by iron overload as well as by inflammation through IL-6 (Nemeth et al., 2004a
), which is elevated in patients with chronic HCV. The identification of hepcidin as a HCV replication cofactor points to a molecular basis for the well-known clinical association between chronic HCV infection and dysregulation of iron homeostasis. Moreover, the potent upregulation of hepcidin transcription by IL-6 potentially creates a positive feedback loop between chronic inflammation and HCV replication.
An important limitation of our study is that the replicon model permits identification of host cofactors of replication but not of other stages of the viral lifecycle. Future studies using fully infectious, cell-culture-adapted HCV strains will be able to study other aspects of the HCV lifecycle, such as viral entry, uncoating, virion assembly, and secretion.
The host cofactors identified in this screen had little overlap with those identified in three previous limited siRNA screens for HCV host cofactors (Ng et al., 2007
; Randall et al., 2007
; Supekova et al., 2008
). Similarly, relatively little overlap has been observed among the hits in three genome-wide RNAi screens for HIV host cofactors (Brass et al., 2008
; König et al., 2008
; Zhou et al., 2008
). There are many factors that must be considered when comparing our studies with prior limited siRNA screens. First, the siRNA sequences used in our screen were the same as for the Ng et al. screen, but were different from those used in the Supekova et al. and Randall et al. screens, likely resulting in different knockdown efficiencies and off-target effects for each gene screened.
Second, the HCV models were different in each screen. Randall et al. used the J6/JFH1 genotype 2a infectious virus, while the other screens used different subgenomic genotype 1b replicons. Silencing of the host protein DDX3X, which binds to the HCV core protein, suppresses subgenomic HCV replication much less efficiently than full-length HCV replication (Ariumi et al., 2007
). Furthermore, genotype-specific differences in host cofactor dependency could also potentially account for differences in screens using different HCV genotypes.
Third, the methods of siRNA transfection and siRNA concentration and the duration of silencing varied among the screens, thus biasing the screens toward protein with shorter half-lives (in the Supekova et al. screen) or longer half-lives (Ng et al.). Randall et al. tested 62 genes, which is a sufficiently small number that siRNA could be introduced by electroporation, and the duration of silencing could be optimized for each gene tested; both methods are impractical for whole-genome screens.
Fourth and finally, the threshold for identifying significant hits for secondary validation varied among the different screens. For example, TBXA2R would have been selected as a hit in our primary screen had we used the Ng et al. threshold of 30% inhibition. However, using this lower threshold in our screen would have resulted in over 1800 pools for secondary screening, which would have been an impractical number of hits for deconvolution.
The strength of whole-genome siRNA screens is their ability to identify roles for genes and biological pathways in biological processes free of a priori hypotheses regarding their function. Our screen has identified many candidate host cofactors for HCV replication, some of which may offer new therapeutic avenues for the treatment of chronic infection. Further studies of these cofactors may yield important mechanistic insights into the formation and maintenance of the membrane-associated HCV replication complex. Furthermore, we have identified a possible molecular link between iron homeostasis, chronic inflammation, and chronic HCV infection. However, because of the very significant false-negative (variability in protein depletion owing to variation in siRNA efficacy and protein half-life) and false-positive (off-target effects) rates inherent to high-throughput siRNA screens, no single screen can be expected to yield a truly comprehensive map of all of the host genes that are involved in a particular process. Such maps will require the integration of multiple RNAi screens with complementary technologies.