Correspondence John G. McHutchison, MD, 333 Lakeside Blvd, Foster City, CA, USA Tel: +1 650-522-5302 Fax: +1 650-522-1975 e-mail: email@example.com
It has been understood for some time that the treatment outcome of hepatitis C virus (HCV) infection is influenced by host genetic factors. Three independent genome-wide association studies have recently identified that a genetic variation in the IL28B gene [interferon-λ3 (IFN-λ3)] determines the outcome of IFN-α-based therapy in patients with genotype 1 chronic hepatitis C infection. This genetic polymorphism is also strongly associated with a higher likelihood of spontaneous clearance following acute hepatitis C infection. These results confirm the importance of specific host genetic markers in predicting outcome and treatment response. They also provide the framework and potential for a clinically relevant and meaningful pharmacogenomic approach to personalizing anti-HCV treatment.
Host genetics have long been suspected to play a role in determining response to interferon-α (IFN-α)-based therapy for chronic hepatitis C virus (HCV). The current standard of care (SOC) therapy is pegylated IFN (PEG-IFN) and ribavirin (RBV) combination therapy. The sustained virological response (SVR) rate is approximately 40–45% in patients with genotype 1 HCV (1–4). Predictors of treatment response are poorly defined for any individual patient and do not allow personalization of therapy; both host (e.g. gender, age and liver fibrosis) and viral factors (e.g. genotype and viral levels) are recognized to be important, but do not adequately explain the variation in response that is observed. African-American ancestry is strongly associated with a poor response, and the SVR rate is half that observed in Caucasians (5, 6). These ethnic differences cannot be explained by the severity of disease, compliance, viral kinetics or more basic immunological or other parameters. Hispanic ethnicity has also been associated recently with lower response rates compared with Caucasians (7). In contrast, Asian populations appear to have the highest response rates (8, 9). Taken together, these ethnic or racial differences in the treatment response rates strongly suggest a genetic basis.
Technology that can screen the entire human genome for variants associated with human diseases has recently become widely available. Although they are not yet ready for large-scale clinical use, these tools could potentially define polymorphic sites that might translate into clinical diagnostics, as well as provide an insight into systems biology. Three published genome-wide association studies (GWAS) have been published recently that link a genetic variation in the IL28B gene region to response to PEG-INF-α plus RBV in patients with genotype 1 chronic HCV (10–12). Here, we explain these technologies and principles and summarize the publicly available data as well as discuss future directions and potential applications.
Genetic association studies
A case/control genetic association analysis involves a comparison of the frequency of a particular allele(s)/genotype(s) in a sample of affected patients compared with a control group of unaffected individuals. The existence of stretches of linkage disequilibrium (LD) in the human genome and the wealth of data that have been generated by the HapMap Project (http://www.hapmap.org) make it possible to assess common genetic variation (i.e. variants that are present in at least 5% of a given population) on a genome-wide scale using ‘tag’ SNPs. GWAS are therefore a hypothesis-free method for systematically testing the association between all common variants in the human genome and any polymorphic trait, notably clinical phenotypes (disease, drug response, drug toxicity and others). An important note about the use of tag SNPs is that GWAS are only able to identify a region within the genome that is associated with a phenotype. They do not usually identify the polymorphism itself that is directly responsible for the effect (the causal or the functional variant), although this may occur by chance. Follow-up of a discovery involves fine sequencing of the associated region, as well as more basic biological studies. As GWAS involve hundreds of thousands of association tests, the threshold for statistical significance must be stringent and correction for multiple testing must be performed (e.g. Bonferroni correction, Table 1). This requires large, well-characterized cohorts to have enough power to detect real associations. Even then, unless the P-value of the association is clearly beyond the genome-wide significance, replication of the study in an independent cohort should be undertaken. For further discussion, readers are directed to a recent overview of the interpretation of GWAS (13).
Table 1. Single nucleotide polymorphisms associated with the pegylated interferon/ribavirin treatment response in the three genome-wide association studies
Before genome-wide approaches became available, most studies used a candidate gene approach, in which polymorphisms affecting plausible biological pathways are chosen to be tested for association with a particular phenotype. A limited number of SNPs are selected, minimizing cost. The threshold for statistical significance is lower, although correction for multiple testing should still be performed. Unfortunately, such statistical vigilance has not always been the case. Another limitation of these targeted studies is the difficulty of correcting for population stratification: invalid associations can be created by systematic differences due to shared ancestry between subgroups of study subjects. Panels of ancestry markers, or informative markers extracted from genome-wide genotyping data, can be used to address this issue. Consequently, many candidate gene findings have not been replicated in follow-up studies, and this is also true for candidate gene studies of HCV treatment response (14–26). All have been limited by a small sample size and borderline significance in the setting of multiple testing. Results have been inconsistent and reproducibility limited.
Genome-wide analyses of genotype 1 hepatitis C virus treatment response
Genetic variation in the IL28B gene region has recently been shown to be strongly associated with viral clearance following treatment with PEG-IFN and RBV in patients with chronic genotype 1 HCV infection.
Ge and colleagues: the first genome-wide association analysis to investigate genetic predictors of treatment response used patients from the IDEAL study, a large randomized-controlled trial that confirmed the similar efficacy of the two commercially available PEG-IFN preparations (PEG-IFN α-2b vs PEG-IFN α-2a) in North American patients with chronic genotype 1 HCV infection (1). One thousand six hundred and four patients consented to genetic testing. An additional 67 patients were enrolled from a second randomized clinical trial that compared the efficacy of PEG-IFN-2b/RBV in Caucasians vs African Americans (5). This study therefore had the advantage of an extremely well-characterized patient phenotype, defined within the context of clinical trials. Genome-wide analysis was performed using the Illumina Human 10-quad BeadChip (610 000 SNPs on chip/565 759 passed QC and were used for the association analysis). The primary analysis for treatment outcome compared SVR with true biological non-response in 1137 patients, in three independent ethnic groups – Caucasians, African Americans and Hispanics, defined by genetic ancestry (336 patients who did not attain SVR were excluded on the basis of <80% compliance to PEG-IFN or RBV; a further 198 were excluded for technical reasons).
Seven SNPs on chromosome 19 in the region of the IL28B gene (coding for IFN-λ3) met the threshold for Bonferroni-corrected significance in the GWAS, adjusting for other clinical factors known to affect the treatment response, including baseline HCV RNA level, hepatic fibrosis stage and steatosis grade, age, body mass index, baseline ALT level, blood sugar levels and weight-adjusted RBV dose (1). The P-value for association of the top discovery SNP, rs12979860, located 3 kb upstream of the IL28B gene, was 1.06 × 10−25 in Caucasians (P=1.37 × 10−28, combined across the three ethnic populations). The six other SNPs displayed different degrees of LD with rs12979860 (Table 1), and their effects were largely explained by rs12979860. Another two SNPs, not present on the genome-wide chip, were shown to be highly associated with rs12979860 by fine sequencing studies. These were both putative functional variants, rs28416813 lying in the promoter region 37 base pairs upstream of the IL28B start codon and rs8103142 a non-synonymous polymorphism in exon two (Lys70Arg). Given the high degree of correlation, it was not possible to resolve which of these three polymorphisms might be solely responsible for the association signal. A validation cohort was not required, given the strength of the association signal (exceeding genome-wide significance by a factor of 10−20).
In this compliant cohort, the IL28B polymorphism was associated with a two-fold increase in the SVR rate in all ethnic groups. In a multivariable logistic regression model, ‘IL28B-type’ was a stronger independent predictor of SVR than viral load (±600 000 IU/ml), hepatic fibrosis stage or ethnicity [adjusted odds ratio (OR) for SVR in Caucasians=7.3 (5.1–10.4)]. Furthermore, the favourable CC genotype was more common in Caucasians than African Americans, and it was estimated that this difference in genotype frequency was responsible for approximately half of the recognized discrepancy in the treatment response rates between the two populations. Random sampling of a healthy population then identified the rs12979860 CC genotype to be most common in Asian patients, suggesting that this genetic variant may also contribute to the high response rates that have been reported in this group (8). Finally, the frequency of the C allele in the study population was found be lower than that of an ethnically matched population of unknown HCV status, suggesting that the genetic variation in the IL28B gene region may be relevant to the natural clearance of HCV. This last point was subsequently confirmed in a separate study that examined the association between rs12979860 and the spontaneous clearance of HCV (28).
Tanaka and colleagues: a second GWAS, conducted in a Japanese population, also identified an important role for the IL28B gene region in HCV treatment response (11). A two-stage design was used with a GWAS discovery stage in 154 patients, using Affymetrix SNP 6.0 genome-wide SNP typing array (900 000 SNPs/621 220 SNPs passed QC), followed by a subsequent validation stage in 172 patients. The association analysis used null virological response as a phenotype (NVR, defined as a <2log10 IU/ml reduction in serum HCV RNA by week 12 of therapy and detectable HCV RNA at week 24) rather than SVR. Adherence >80% during the first 12 weeks was required for inclusion. Two SNPs in the IL28B gene region showed strong associations in the GWAS phase I, rs8099917 and rs12980275 (P=3.11 × 10−15 and P=1.93 × 10−13 respectively). These two SNPs were validated in the replication phase [combined P=2.68 × 10−32, OR 27.1; (14.6–50.3) and P=2.84 × 10−27, OR 17.7 (10.0–31.3) respectively]. Six other SNPs were associated beyond the genome-wide threshold, found either on the basis of LD and haplotype structure (HapMap release 23a, Japanese ancestry), or by fine resequencing studies. All were in the IL28B gene region, in strong LD, and it was concluded that the association signal was driven by one of the identified SNPs, although again it was not possible to conclude which, if any, was the causal variant. Logistic regression modelling confirmed a strong independent role for SNP rs8099917 in predicting NVR. In a secondary analysis, Tanaka and colleagues also observed that SNPs rs8099917 and rs12980275 were significantly associated with SVR vs no SVR [unadjusted OR 12.1 (6.5–22.4), P=1.18 × 10−18 and 8.8 (5.1–15.4), P=1.17 × 10−16 respectively]. This study therefore confirmed that genetic variation in the IL28B gene region was associated with viral clearance, as well as null response (an expected finding, given the recognized role of NVR in strongly predicting overall non-response).
Suppiah and colleagues: a third GWAS was conducted in genotype 1 HCV-infected patients of European ancestry (Australian/western Europe) (12). The investigators tested for polymorphisms associated with SVR. Adherence to therapy was not defined. A two-stage approach was adopted. The initial GWAS was performed in an Australian cohort (n=293) using the Illumina Infinium HumanHapMap300 or the CNV 370-Quad genotyping BeadChip (311 159 SNPs used for the association analysis). One hundred and seventy-two SNPs were taken forward to the second stage, based on a P-value for association <10−5, or P<10−3 and immune-regulatory/antiviral function. The replication cohort included patients from the UK, Germany, Italy and Australia. The polymorphism rs8099917 exceeded the threshold for genome-wide significance in the GWAS stage and replication phases [OR 1.98 (1.57–2.52), P=9.25 × 10−9]. Seven other SNPs in the IL28B gene region were noted to be associated with SVR in a subsequent haplotype analysis using tagging SNPs identified within the distinct IL28B haplotype block (Table 1). Although only one of these seven SNPs reached genome-wide significance, it was notable that six of the seven were common to the Ge/Tanaka papers (Table 1). Several other SNPs were described as being of interest in this study, on the basis of suggestive but not genome-wide significance (P<10−4)±biological plausibility. These included SNPs related to the genes IL21R and CASP1. However, given the absence of these association signals in the other two studies, it is reasonable to conclude that they probably represent false-positive associations.
Three independent original GWAS have now identified variants within the IL28B gene region that are strongly associated with response to PEG-IFN plus RBV combination therapy in patients chronically infected with genotype 1 HCV. Furthermore, a strong association of rs12979860 with both an early virological response and an SVR in IFN-naïve patients treated with PEG-IFN and RBV was also reported. In particular, the association of rs12979860 with virological response appeared to be stronger in the naïve patients' population compared with prior non-responders (27).
There are now more than 10 independent validations of the importance of this genetic variant related to HCV treatment response in different populations of chronically infected individuals (other HCV genotypes, HIV coinfected patients and in the setting of transplantation, among others). Another study has also shown that the SNP rs12979860 is strongly associated with the spontaneous clearance of HCV. Most probably, studies have detected the same genetic signal. The fact that multiple SNPs have been identified, with differing effect sizes, almost certainly reflects the differences between the studies with respect to ethnicity, phenotype definition and SNP genotyping platform, and also suggests the existence of a strong LD in the region. The causal variant(s) have yet to be defined.
Genome-wide analysis of spontaneous clearance of hepatitis C virus
Data supporting a role for IL28B genetic variations in spontaneous clearance of HCV followed rapidly after the treatment response discovery. Thomas and colleagues used a candidate gene approach to investigate the potential role of rs12979860 variation in determining the natural clearance rate following acute HCV, in cohorts of individuals who spontaneously cleared the virus (n=388) or had persistent infection (n=620) (28). Patients with the CT/TT genotypes at rs12979860 were three times less likely to spontaneously clear HCV [clearance rate: CT/TT=28% vs CC=53% vs OR 0.33 (0.25–0.45), P=10−13, overall cohort]. Similar effects were found in individuals of European or African American ethnicity. Nineteen per cent of the cohort was co-infected with HIV, and 10% were HBsAg-positive; neither infection altered the effect of this locus on the outcome of an acute HCV infection. These data were extended by a second group that performed a GWAS for spontaneous clearance in a Swiss/German cohort (chronic hepatitis C, n=1015; spontaneous clearance, n=347; n=448 were co-infected with HIV). They confirmed that variants of the IL28B gene region are the only common genetic variants associated with the spontaneous clearance of HCV infection, using a genome-wide approach [top association SNP=rs8099917, OR 2.31 (1.74–3.06), P=6.7 × 10−9] (29).
The mechanism by which genetic variation in the IL28B gene region influences the treatment response is unknown. None of the identified variants has an obvious effect on gene function (e.g. non-sense mutation). But even without a dramatic impact on protein production or structure, polymorphisms may exert a functional influence through gene expression, mRNA splicing, protein stability/half-life or, for cytokines like IFNs, receptor binding and activation. The data available to date are limited to studies of gene expression, with conflicting results. Ge and colleagues did not identify any association between the polymorphism rs12980275 and IL28B RNA expression in peripheral blood mononuclear cells from 80 HCV-negative individuals. In contrast, Suppiah and colleagues observed a weak association between rs8099917 and IL28B/IL28A expression in 49 healthy volunteers, where a higher expression was noted in the responder genotype (P=0.044); this effect was largely driven by lower expression in three patients homozygous for the non-responder genotype. Tanaka and colleagues were the only group to examine expression in HCV-infected patients. In a small cohort (n=20), higher expression levels of IL28B mRNA in patients with the responder genotype were observed. Further studies are required. Finally, although the association signals in all three studies were mainly due to IL28B gene variants, an effect from the nearby IL28A gene cannot be excluded.
We have to recall that in non-responders, some IFN-stimulated genes were highly expressed before treatment; thus, pre-activation of the IFN system in patients appears to limit the effect of IFN antiviral therapy (30, 31). Further ongoing studies evaluating the inter-relationships between IL28B polymorphisms and intrahepatic gene expression (in particular, the IFN pathway and immune response) may also shed some additional light.
Nevertheless, the finding that polymorphisms in genes of the IFN-λ family, also known as type III IFNs, are important for both spontaneous and treatment-related clearance of genotype 1 HCV has significant biological plausibility. Three members of the type III IFN family have been described, IFN-λ1/2/3 (IL29, IL28A, IL28B). IFN-λs appear to be ubiquitously expressed in response to the same viral stimuli that induce type 1 IFN (TLR ligands, double-stranded RNA) (34, 35). The IFN-LR is made up of two subunits – IFN-λ receptor 1 (IFN-LR1=IL28RA) and interleukin 10 receptor 2 (IL10R2=IL10RB); all three members of the IFN-λ signal via this receptor. The IFN-LR is believed to share a common downstream signalling pathway with the type 1 IFN receptor, activating the complex IFN-stimulated gene factor 3 (ISGF3=STAT-1/2·IRF-9), which binds to the IFN-stimulated response element to induce the transcription of IFN-stimulated genes. The major distinction recognized between type 1 and type III IFN appears to be the tissue distribution of the receptor. Whereas the type 1 IFN receptor is expressed by all cells, the type III IFN receptor (specifically, the IFN-LR chain) is more restricted. It is expressed on epithelial cells, including hepatocytes, but not on haemopoietic cells that therefore have an impaired response to IFN-λ. IFN-λs have been shown to inhibit HCV in cell culture models: it was less potent than IFN-α, but also at least additive with IFN-α. More recently, a subcutaneous preparation of IFN-λ1 has shown antiviral activity in HCV patients. Importantly, it remains unclear how IFN-α and IFN-λ interact in vivo, and the relevance of genetic variation in the IL28B gene region to IFN-λ1 therapy is not known.
Genotyping of one of the IL28B polymorphisms has great potential as a clinical diagnostic tool to aid in assessing the probability of response to current therapy. To test for a single polymorphism is not technically difficult or time consuming (the PCR-based assay uses the same technology as the current test used for haemochromatosis). A licensed assay has been available in the USA since July 2010. The SNP rs12979860 appears to be the most suitable simple genetic predictor of treatment outcome, because it has the strongest association signal in the largest cohort studied to date, thus explaining the lower observed response rates in AA patients. Knowledge of host IL28B type could help in the clinical decision-making process. The use of this genetic predictor might allow genotype 1 HCV-infected patients to be divided into 2 groups in the future: (i) those with the good response genotype, who might be managed in the same way as patients with genotypes 2/3 HCV, and (ii) those with the less favourable genetic response genotypes, in whom the decision about starting therapy must be evaluated in light of the expected availability of direct antivirals in the near future.
A number of important clinical and more translational questions now must be addressed. These include whether IL28B type can be used to personalize the duration of therapy with SOC. That is, will rs12979860 CC ‘good responder’ patients only require 24 weeks of PEG-IFN and RBV therapy? Will direct antivirals overcome or attenuate the IL28B-type effect? It is likely that they will, at least to some degree, with fixed-duration regimens. However, perhaps IL28B type will allow a shortened duration of therapy, minimizing toxicity and drug resistance. Is IL28B type relevant to non-genotype 1 HCV infection and treatment outcomes? And is this relevant for those with HIV/HCV co-infection or for the rapidity and severity of post-transplant HCV recurrence [which might also depend on IL28B type of both the donor and the recipient liver as reported recently (40)]? What is the future for IFN-λ as a therapeutic, and will the IL28B type be relevant? And how will the geographical differences in the favourable IL28B allele frequency relate to treatment response, as well as to newer therapeutic regimens and durations? Is this a marker of IFN response above and beyond HCV infection, and if so, do other diseases with IFN-based regimens (such as hepatitis B, some malignancies, multiple sclerosis and others) interact differentially with IL28B types? And finally, the functional underpinnings of this genetic marker should, once unravelled, provide greater biological insight into treatment response in this disease and potential novel therapeutic approaches.
Should IL28B typing be included in hepatitis C virus clinical trials?
Because of the effect of the IL28B on treatment outcomes and early viral kinetics (Thompson A. J., personal communication), it seems very relevant that this factor, like HCV genotype, race, viral load and ethnicity, be incorporated into clinical trial designs. This is particularly relevant for new small-molecule studies, where an imbalance in IL28B type could be associated with significant numbers of patients, with the favourable IL28B type being randomized unequally into different arms of a particular study. This is especially true in small proof-of-concept studies (with less than 100 patients) and to a lesser degree in studies with approximately 250 subjects (Fig. 1). If this factor is not taken into account in trials of this size, there is an approximately 25–50% chance of a mismatch in terms of the proportion of patients with a favourable IL28B type randomized into each arm of any trial. Because of these discrepancies, the results of these trials could be incorrectly interpreted as favouring one treatment arm or dose of a drug, while in fact it is due to this factor alone.
What important clinical trials should now be performed?
It seems clear that many opportunities now exist to segment our HCV patients into groups of patients who are highly and less likely to respond to PEG-IFN and RBV therapy. Being able to determine the IL28B type before treatment makes it possible to position and perform certain trials:
1A trial comparing PEG-IFN and RBV for 24 and 48 weeks in IL28B C/C favourable patients. Similar to genotype 2- and 3-infected patients, who only require 24 weeks of therapy, IL28B favourable patients might also only require a shorter duration of therapy. This is particularly relevant to East Asia, where the higher prevalence of the favourable C/C alleles and higher response rates might make a shorter duration of therapy possible for most patients.
2A trial comparing PEG-IFN and RBV for 48 and 72 weeks (or longer) in IL28B unfavourable patients.
3Evaluation of shorter triple therapy direct antiviral combination regimens (12 and 24 weeks) in IL28B C/C patients.
In conclusion, the identification of the genetic variation in the IL28B gene region as a predictor of genotype 1 chronic HCV treatment outcome is an exciting discovery. It sheds new light on virus–host interaction, and appears to have immediate clinical use as a diagnostic tool, providing potentially helpful information to both patients and clinicians considering therapy with PEG-IFN plus RBV. It is therefore a step towards a personalized approach to anti-HCV therapy. However, the treatment paradigm for genotype 1 HCV is about to change with the introduction of direct antivirals. The role of the IL28B variation in this setting as well as determining the duration of therapy required in these favourably genetically determined individuals with both current two-drug and future three-drug regimens requires semi-urgent investigation.
Conflicts of interest
John McHutchison is an employee of Gilead. He has received funding for research and acted as a consultant and advisor for Schering Plough and Merck. He is co-inventor of patents relating to the IL 28B and ITPA discoveres.