In vitro blood cell responsiveness to IFN-α predicts clinical response independently of IL28B in hepatitis C virus genotype 1 infected patients

The standard of care for patients infected with hepatitis C virus (HCV) for the past decade has consisted of dual therapy with the anti-viral cytokine, interferon-alpha (IFN-α) and the nucleoside analogue, ribavirin. IFN-α induces anti-viral immunity by upregulating hundreds of IFN stimulated genes (ISGs) [1], including many with potent direct and indirect anti-viral activity [2]. Response rates to therapy are highly variable, with patients infected with genotype 1 (G1) having sustained virological response (SVR) rates of less than 50%, whereas genotype 3 (G3) infected patients can achieve SVRs of up to 82% [3].

Increased insight into the viral life cycle of HCV has led to development of several new directly acting anti-viral agents (DAAs), including NS3-4A protease inhibitors (PIs). Several PIs, including telaprevir and boceprevir, are administered in combination with standard IFN-α/ribavirin treatment. Triple therapy increases response rates from less than 50% to 75% in some G1 infected cohorts [4, 5]. However, PIs are expensive and associated with significant additional side effects, such as anaemia and rash, and emergence of drug resistant variants, a major challenge in cases of non-compliance with therapy. Identification of patients with a high probability of obtaining an SVR to dual therapy would obviate the need for additional PIs and alleviate these issues.

Multiple efforts have been made to accurately predict response to therapy using viral and host characteristics. Viral predictive markers include viral load and genotype, while host predictive markers include age, sex, race, and liver fibrosis stage [6]. Elevated serum levels of the chemokine CXCL10 have also been reported to be associated with non-response to IFN-α therapy [7]. Additionally, a single nucleotide polymorphism (SNP), rs12979860, in the recently described IFNλ4 gene [8] is highly predictive of response [9, 10]. The major C allele correlates strongly with viral clearance particularly in patients infected with HCV G1. Nevertheless, no single marker or combination of markers accurately predicts patient response in individual cases.

We hypothesised that in vitro responsiveness to IFN-α would predict clinical responsiveness to dual therapy. Hepatic ISG expression is elevated pre-treatment in patients who fail to achieve SVR [11] and has been shown to be a stronger predictor of response than IL28B genotype [12]. However, liver biopsy is an invasive procedure with associated risks and has limited value as a prognostic tool. Leukocytes are sensitive responders to IFN-α and provide a more accessible alternative, requiring just a peripheral blood sample. In fact, in vivo upregulation of ISGs in PBMCs following therapeutic IFN-α is similar to ISG upregulation in vitro following IFN-α stimulation [13], suggesting in vitro PBMC responsiveness may indeed be an accurate reflection of in vivo clinical response.

IFN-α activates the JAK-STAT signalling pathway, leading to upregulation of over 500 ISGs [14]. PKR, OAS and MxA are three well-characterised ISGs, strongly induced by IFN-α in PBMCs, which have direct anti-viral action. Activation of PKR by virus results in inhibition of protein translation, including inhibition of viral mRNA translation through phosphorylation of the alpha subunit of eukaryotic protein synthesis initiation factor 2 (eIF2α) [15]. Indeed, HCV has evolved several mechanisms to block the action of this important regulator of translation [16, 17]. OAS is an IFN regulated activator of latent ribonuclease RNase L, which is triggered by activated OAS to directly cleave RNA, including HCV RNA, thus destroying viral RNA products and producing pathogen associated molecular patterns (PAMPs) that further stimulate innate immune activity [18, 19]. MxA recognises viral nucleocapsids and renders them redundant by wrapping around the viral structure and forming MxA/nucleocapsid oligomers [20]. MxA may also direct nucleocapsids to alternative sites in the cytoplasm, where they are not functional for RNA synthesis and likely to be immobilised and subsequently degraded [2]. Interestingly, we have demonstrated that the core protein of HCV co-localises with MxA in a granular pattern in the cytoplasm of cells, a phenomenon that is potentiated with the co-treatment of IFN-α and ribavirin [21]. Hepatic expression of MxA is a known predictor of response to IFN-α therapy [22, 23]. Because of their role as key mediators of IFN-α-induced antiviral activity, PKR, OAS and MxA were chosen as potential indicators of IFN-α treatment responsiveness in this prospective study.

Approximately 50% of HCV infected patients infected with G1 achieve an SVR with IFN-α and ribavirin therapy and do not actually need additional PIs in their treatment regime. However, it has it has not hitherto been possible to identify responsive patients prior to commencing treatment. Here, we hypothesised that in vitro analysis of immune cell responsiveness to IFN-α stimulation prior to treatment would predict clinically responsive patients. In this prospective study, we found that G1 infected patients who achieved SVR could indeed be identified prior to treatment by the significantly greater ISG induction in their PBMCs following in vitro IFN-α stimulation than in PBMCs from G1 infected patients who did not achieve SVR. Interestingly, this finding seems to be exclusive to patients infected with GI as the clinical response of G3 infected patients could not be predicted pre-treatment. IL28B genotype did not influence ISG expression in PBMCs nor in vitro IFN-α stimulated responses and there was no difference in ISG expression between patients who achieved SVR or did not achieve SVR among the different IL28B genotypes.

The development of DAAs is an exciting advance in treating HCV infection and offers many genotype 1 infected patients who would not respond to standard IFN-α therapy, additional options for treatment. However, currently licensed PIs are given in combination with dual IFN-α therapy, thus significantly increasing costs and introducing additional significant side effects. Trials involving DAAs in “IFN free” regimes are extremely promising for certain groups of patients, however IFN-α therapy still boasts a lack of drug-drug interactions, relatively low cost, lack of effect of viral resistant mutants and a long clinical track record [28]. Crucially, the cost of DAAs will prevent their large scale distribution in low and middle income countries, which have the highest rates of HCV infection, for the foreseeable future. Therefore, identification of subgroups of patients who would not require additional PIs would make significant economic savings.

Several predictors of response to dual therapy have previously been described including demographic, genetic and intrahepatic markers [3]. Studies have shown hepatic ISG expression to be a strong predictor of treatment response [11, 29]. Using microarrays, others have shown that global transcriptional responses in PBMCs following in vivo and ex vivo IFN-α stimulation are different between responders and non-responders following treatment [30, 31]. We found that G1 infected patients who achieved an SVR had significantly higher ISG expression in PBMCs prior to treatment. Pre-treatment, endogenous expression of PKR and MxA was lower in PBMCs from G1 infected patients who failed to achieve an SVR, contrasting with previous reports from the liver showing elevated pre-treatment ISG expression predictive of therapeutic non-responsive [11]. A range of IFN-sensitive mechanisms may already be activated in HCV-infected liver, which harbours mixed populations of cells including epithelial and endothelial cells as well as hepatocytes and immune cells. In comparison, PBMCs provide a more accessible homogenous population for use as a resource for identifying patients who are responsive to IFN-α.

Interestingly, pre-treatment ISG expression in PBMCs from G3 infected patients could not predict SVR, yet ISG expression was strongly predictive of therapeutic response in G1 infected patients, suggesting that HCV G1 may differ significantly from G3 in its ability to target IFN-α signalling. Several HCV proteins, including the NS3-4A protease, are known inhibitors of the IFN response [32, 33]. We recently found that HCV G1 degrades signal transducers and activator of transcription 1 (STAT1) and STAT3, key transcription factors activated by IFN-α signalling, and showed that this suppression was consistent across all PBMC subsets [34]. However, it is possible that HCV genotypes differentially target IFN-α signalling perhaps explaining the differences in ISG expression we observed in G1 and G3 infected patients.

Patients who have rapid decreases in viral load and achieve an RVR have a 90% chance of achieving overall SVR [35]. However, we found that pre-treatment ISG expression was not strongly associated with RVR, suggesting that ISG expression in PBMCs is independent of RVR, in accordance with a study by Sarasin-Filipowicz et al. [29]. That study did not examine SVR. Unusually, RVR was not predictive of overall SVR in our G1 infected cohort. Although ISG expression tended to be higher in patients who achieved EVR or were PCR –ve at the end of treatment compared to RVR in G1 infected patients, the strongest association seen was that between pre-treatment ISG expression and SVR.

In this study, we analysed expression of three well-defined anti-viral genes (PKR, OAS and MxA) that have been shown using microarray analysis to be upregulated in PBMCs and specific immune cell subsets, including macrophages, DCs and NK cells, following in vitro IFN-α stimulation [2, 36‐39]. IFN-α is a strong regulator of gene expression in PBMCs and Zimmerer et al. demonstrated that in vitro stimulation of PBMCs mimics in vivo effects [40]. This study also demonstrated that T cells, monocytes and NK cells all respond to in vitro IFN-α stimulation. Recently, a retrospective analysis of G1 infected HCV patients of mixed race demonstrated that SVR and monocyte activation in response to 24 h in vitro stimulation with IFN-α were negatively correlated but failed to show any correlation between IFN-α induced myeloid and plasmacytoid dendritic cell activation and therapeutic response. In our study, we found enhanced responses to 4 h in vitro stimulation in unseparated PBMCs from G1 infected patients was associated with SVR, indicating that duration of IFN-α stimulation is critical.

The rs12979860 SNP, located upstream of the IL28B gene and present in the novel IFNλ4 gene [8], is a robust predictor of spontaneous viral clearance and response to treatment in some cohorts [9, 10]. Of note, hepatic ISG expression has also been shown to be a stronger predictor of response than IL28B genotype [12, 22]. The predictive value of IL28B is increased when combined with the predictive power of other known indicators of response, such as serum CXCL10 levels [41]. Amongst our G1 infected patients, ISG expression in PBMCs was more predictive of therapeutic response than IL28B genotype which was a surprising result given the strong predictive power of this SNP in other studies [10, 27]. However, it is possible that the IL28B genotype is not a strong predictor in all cohorts; indeed MxA staining in hepatic macrophages has been found to be a better predictor of SVR than IL28B genotype in a Canadian cohort [22]. Endogenous ISG expression has been shown to be higher in livers from patients carrying the minor T allele for rs12979860 [12], yet we found that in PBMCs, ISG expression was similar between patients of any genotype. Microarray studies have shown that ISG expression in PBMCs from G1 infected patients was independent of IL28B genotype [42]. In a HCV/HIV co-infected cohort of patients, induction of ISGs following 12 hours of in vivo stimulation with IFN-α, was independent of IL28B genotype, echoing our findings in PBMCs from HCV infected patients.