Sie sind auf Seite 1von 14

JOURNAL OF INTERFERON & CYTOKINE RESEARCH FULL MANUSCRIPT

Volume 00, Number 00, 2017


ª Mary Ann Liebert, Inc.
DOI: 10.1089/jir.2017.0002

Confounding by Single Nucleotide Polymorphism


rs117648444 (P70S) Affects the Association of Interferon
Lambda Locus Variants with Response
to Interferon-a-Ribavirin Therapy in Patients
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

with Chronic Genotype 3 Hepatitis C Virus Infection

Anand Bhushan, Sumona Ghosh, Samsiddhi Bhattacharjee, and Sreedhar Chinnaswamy

Genome-wide association studies discovered interferon lambda (IFNL or IFN-l) locus on chromosome 19 to be
involved in clearance of chronic hepatitis C virus (HCV) infection in patients following interferon-a-ribavirin
(IFN-RBV) therapy. Subsequent studies established a dinucleotide polymorphism rs368234815, as the prime
causal variant behind this association. The DG allele of this variant gives rise to a new IFNL gene, IFNL4, coding
for IFN-l4 whose activity paradoxically associates with lesser viral clearance rates. A low-frequency, non-
synonymous single nucleotide polymorphism (SNP) rs117648444 within the 2nd exon of IFNL4 changes the
70th amino acid from proline to serine resulting in lower activity of the functional IFN-l4 protein, thereby
increasing HCV clearance rates. In the present study, we used a cohort of genotype 3 HCV-infected patients,
drawn from different geographical regions of India who underwent IFN-RBV therapy, to examine the association
of several important IFNL locus SNPs/variants with sustained virological response (SVR). Intriguingly, the
causal variant rs368234815 did not show the best strength and significance of association with SVR, while
further analysis revealed that a negative confounding effect of rs117648444 was responsible for this phenom-
enon. Our results indicate that IFNL locus SNPs are subject to either a positive or a negative confounding effect
by rs117648444; the nature of confounding depends on the linkage of the IFNL SNPs with the low-activity IFN-
l4-generating minor allele of rs117648444. Thus, our work demonstrates that the linkage disequilibrium
structure of the IFNL region may confound the results of association studies. These results have implications for
the design and understanding of future case–control studies involving IFNL locus SNPs/variants.

Keywords: confounder SNPs, genotype 3 HCV, HCV, IFN-l4, IFN-RBV, IFNL SNP, rs368234815

Introduction emerged as the top hits in the initial GWAS (Ge and others
2009; Rauch and others 2010; Suppiah and others 2009;

H uman interferon lambda (IFNL) locus on chromo-


some 19 comprises 4 genes, IFNL1-4 encoding for
potent antiviral and immunomodulatory cytokines (IFN-l1-
Tanaka and others 2009; Chinnaswamy 2014) (Fig. 1);
later, fine-mapping studies led to identification of more
SNPs in the region (Ge and others 2009; Rauch and others
4) belonging to the type III IFN family (Kotenko and others 2010; di Lulio and others 2011; Smith and others 2011).
2003; Sheppard and others 2003; Prokunina-Olsson and Among them, SNPs rs28416813 and rs4803217 in 5¢ and 3¢
others 2013; Chinnaswamy 2016). This locus and its genes untranslated regions of IFNL3 gene were shown to influence
gained importance after several genome-wide association reporter gene expression thereby implicating expression dif-
studies (GWAS) in the year 2009 and showed involvement ferences of IFN-l3 protein as a likely causal mechanism re-
of single-nucleotide polymorphisms (SNPs) in clearance sponsible for HCV clearance (Chinnaswamy and others 2013;
of chronic hepatitis C virus (HCV) infections following Mcfarland and others 2014). Other SNPs such as rs7248668 and
interferon-a-ribavirin (IFN-RBV) therapy (Ge and others 2009; rs4803221 were identified to have best predictive values after
Suppiah and others 2009; Tanaka and others 2009; Rauch screening for all variants in a region spanning 100 kb at the locus
and others 2010). The SNPs rs12979860 and rs8099917 by DNA deep-sequencing methods (Smith and others 2011).

National Institute of Biomedical Genomics, Kalyani, India.

1
2 BHUSHAN ET AL.
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

FIG. 1. A schematic diagram of the genomic structure of IFNL locus highlighting the SNPs/variants that were genotyped
for this study. The SNP rs12979860 is referred to as the ‘‘tag SNP’’ due to its close correlation with the ‘‘functional
variant.’’ Position of each SNP/variant is marked by black vertical lines and approximate distance among rs8099917,
rs368234815, and rs117648444 is given by dotted lines; the alternate alleles for each SNP/variant are given in parenthesis.
Depending on the presence of the alternate alleles of the ‘‘activity’’ variant rs117648444, active IFN-l4 protein (shown as
3D structure, PDB ID: K9M1U5; visualized in UCSF CHIMERA) can be of 2 types: IFN-l4 P70 (DG-G, haplotype of
rs368234815-rs117648444) and IFN-l4 S70 (DG-A) and their properties are highlighted based on results from Terczyńska-
Dyla and others (2014). IFNL, interferon lambda; SNP, single-nucleotide polymorphism.

A combination of genetic and molecular methods identi- Even though the SNP rs8099917 is located *4 kb away
fied a dinucleotide variant rs368234815 as the prime causal from the functional variant rs368234815 (Fig. 1) and not in
variant (referred to as functional variant hereafter) at the high LD (linkage disequilibrium) with it in many world
IFNL locus in chronic HCV infections (Prokunina-Olsson and populations (Prokunina-Olsson and others 2013; Chinnas-
others 2013). The DG allele of the functional variant causes a wamy 2016), it has been consistently reported to be asso-
frameshift leading to the expression of a fully functional IFN- ciated with response to IFN-RBV therapy (referred to as
l4 protein (Fig. 1). The TT allele that obliterates the expression sustained virological response or SVR) in HCV infections; it
of a functional IFN-l4 protein seems to have been acquired is also the ‘‘top hit’’ associated with SVR in 3 GWAS
during evolution just before the out-of-Africa migration of (Suppiah and others 2009; Tanaka and others 2009; Rauch
modern humans (Key and others 2014). Present-day human and others 2010). In some reports, the association is even
populations harbor the DG allele in varying proportions (from stronger than the ‘‘tag SNP’’ rs12979860 (Fig. 1, referred as
0.05 to 0.78) with the African population having the maximum tag SNP hereafter) (Doehring and others 2010; Lindh and
frequency (0.78) (Prokunina-Olsson and others 2013). Another others 2011; Smith and others 2011; Suppiah and others
low-frequency variant rs117648444 (referred to as activity 2011; Antaki and others 2013; Khudayberganova and others
variant hereafter) changes the amino acid from Pro to Ser at the 2014; O’Connor and others 2016).
70th position in IFN-l4 protein (Fig. 1), altering its functional Some studies showed that rs8099917 is useful as a
properties (Prokunina-Olsson and others 2013; Terczyńska- predictor of SVR in those patients who were heterozygous
Dyla and others 2014). for the tag SNP (Fischer and others 2012, 2013; Palmieri
In HCV infections, expression of a functional IFN-l4 pro- and others 2014; Ragheb and others 2014; Palenzuela
tein by the DG allele of the functional variant is associated with Gardón and others 2016), suggesting that there may be an
nonclearance of the virus from the host either spontaneously underlying and yet unidentified gene structure that links
after an acute infection (Aka and others 2014) or following the 2 SNPs. A recent study, on a European HCV cohort
treatment with IFN-RBV in chronic cases (Prokunina-Olsson (Terczyńska-Dyla and others 2014), showed that the minor
and others 2013). This is referred to as the ‘‘IFN-l4-HCV allele of rs8099917 is strongly correlated with the DG-G
paradox’’ (O’Brien and others 2014), the reasons for which are haplotype involving the functional variant and the activity
unclear so far. variant rs117648444 (Fig. 1).
LINKAGE DISEQUILIBRIUM STRUCTURE AT THE IFNL LOCUS 3

This discovery certainly provides a convincing explana- therapy at 7 centers across India. Demographic characteristics,
tion for the association of rs8099917 with clearance of HCV clinical details, and viral load [including status of rapid viro-
and indirectly implicates the function/activity of IFN-l4 to logical response (RVR), SVR] of patients are shown in Table 1.
be the causal link behind IFNL SNP association with HCV Forty-five healthy controls were enrolled with similar demo-
clearance (Terczyńska-Dyla and others 2014). However, no graphic characteristics. Among them, 14 controls were from an
confirmation regarding this has been reported thus far, es- asthma clinic (based on immunological, X-ray, and spirometry
pecially from other populations. analysis) and 31 controls were from a gastroenterology clinic
There is no consensus on which SNP should be targeted (based on liver ultrasound grade and biochemical analysis).
for association studies involving the IFNL locus even Blood samples were drawn for SNP genotyping from patients
though numerous studies have been reported in this area and healthy controls. This study was approved by the Institu-
(Chinnaswamy 2014, 2016). There is agreement in the lit- tional Ethics Committee: Review Committee for Protection of
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

erature that rs368234815 is the causal variant (Chinnaswa- Research Risks to Humans, National Institute of Biomedical
my 2016), but less is known on how the activity variant Genomics, Kalyani. Informed written consent was obtained
rs117648444 influences association studies involving the from all patients.
functional variant. In view of this, we undertook the current
study in a cohort of genotype 3 HCV-infected patients in an Genotyping of IFNL SNPs/variants
Indian population who underwent IFN-RBV therapy.
Even though IFN-RBV therapy has become obsolete in Samples were genotyped using PCR-RFLP (polymerase
many parts of the world with the introduction of direct- chain reaction-restriction fragment length polymorphism) or
acting antivirals (DAAs) in 2011, they have become avail- by direct sequencing. For patient samples, SNPs rs117648444
able only recently in India (Puri and others 2016); also, and rs8099917 were genotyped by PCR-RFLP and confirmed
since genotype 3 HCV is no longer an ‘‘easy-to-treat’’ ge- by Sanger sequencing where 100% concordance was obtained.
notype with DAAs (Gondeau and others 2015; Kattakuzhy The remaining SNPs/variants were genotyped by Sanger se-
and others 2016), genotyping of IFNL SNPs/variants may quencing. Human genomic DNA was extracted from whole
still be important especially since response to DAA therapy blood using a QIAamp Blood DNA Midi Kit (Qiagen) as
is also influenced by IFNL locus SNPs (Akuta and others per the manufacturer’s protocol. DNA was quantified using
2010; Zeuzem and others 2013; Meissner and others 2014). NanoDrop device (Thermo Scientific). Two hundred to four
hundred nanograms of template DNA was used for each PCR.
SNPs/variants rs368234815, rs12979860, rs117648444,
Materials and Methods and rs4803221 were sequenced from an amplicon generated
Study participants by using the primers (1) IL28B4kbxhoRev: 5¢-GATATCCTC
GAGCCCGGATTTCAGGAC-3¢ and (2) IL28B4kbKpnFor:
The cohort consisted of 72 chronic HCV (genotype 3a/b)- 5¢-GATATCGGTACCGCCCTGGACGGGAAAG-3¢. SNPs
infected patients who had received Peg-IFN-RBV combination rs8099917 and rs8113007 were sequenced from an amplicon

Table 1. Clinical Characteristics of Hepatitis C Virus Patients Used in the Study


Parameters Values
Median age (IQR) 40 (15)
Males (%) 69.4
Mean BMI in kg/m2 (range) 25.17 (20.45)
Mean HCV RNA (copies/mL) at baseline (range) 19.7 · 105 (31.99 · 105)
Median ALT (IU/L) at baseline (range) 79.5 (394)
Median ALT (IU/L) at end of therapy (range) 38 (158)
Median AST (IU/L) at baseline (range) 77 (273)
Median AST (IU/L) at end of therapy (range) 39 (153)
Mean hemoglobin (g/dL) at baseline (range) 13.14 (12.8)
Median platelet count (per cumm) at baseline (range) 178,000 (349,998)
Median dose of ribavirin (mg/day) used (range) 800 (1,100)
Median dose of peginterferon-alpha used (mg) (range) 90 (130)
Median SAP (IU/L) at baseline (range) 156 (709)
Median SAP (IU/L) at end of therapy (range) 156 (319)
Median bilirubin total (mg/dL) at baseline (range) 0.8 (10.8)
Median bilirubin conjugated (mg/dL) at baseline (range) 0.2 (1.9)
Median fasting blood glucose level (mg/dL) at baseline (range) 92 (102)
Cirrhotic (%) 31
Diabetic (%) 7
Patients attaining RVR (%), n/N 80, 54/67
Patients attaining SVR (%), n/N 71, 51/72
Patients attaining EVR (%), n/N 91, 62/68
End of therapy response (%), n/N 90, 60/67
Data were missing for some patients for some parameters.
ALT, alanine transaminase; AST, aspartate transaminase; BMI, body mass index; EVR, early virological response; HCV, hepatitis C
virus; IQR, interquartile range; RVR, rapid virological response, SAP, serum alkaline phosphatase; SVR, sustained virological response.
4 BHUSHAN ET AL.

generated by the primers (1) 917For700bp: 5¢-CAAGTCAG Assay Kit (Promega) was used for the assay. The ratios of
GCAACCACATGC-3¢ and (2) 917REV700bp: 5¢-GCTCT firefly to Renilla luciferase values were used for analysis.
GTCTGTCTCAATCAATC-3¢. Sequencing was carried out
either at the National Institute of Biomedical Genomics, Statistical analysis
Kalyani, India (ABI-3500xL Genetic Analyzer) or at Eurofins
Genomics India Pvt. Ltd. (Bengaluru, India). Haplotypes were inferred using PHASE software, version
2.1 (University of Washington, http://stephenslab.uchicago
Correction for the altered activity of IFN-k4 .edu/software.html). For LD analysis, haplotypes were coded
as AA, AB, and BB; BB for the absence of haplotype, AB for
protein by restricted analysis
the presence of 1 copy of the haplotype, and AA for the
We used 2 models (referred to as restricted models) to presence of 2 copies of the haplotype. LD was calculated by
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

account for the altered activity of the IFN-l4 protein in ‘‘R programming’’ using ‘‘LD measure’’ function. Goodness
our association analysis of the tag SNP/functional variant of fit was measured by chi-square tests (with Yate’s correc-
with SVR. We had a total of 8 patients who were G/A at tion) and Fisher’s exact test (2-tailed). A P value of <0.05 was
rs117648444; 6 were also heterozygous for rs368234815 considered significant.
(TT/DG), all of them belonged to the SVR group; 2 patients We tested 4 models for association of IFNL variants with
were DG/DG at rs368234815, 1 belonged to the SVR group SVR: dominant, recessive, allelic, and additive models
and the other belonged to the nonresponder (NR) group. In (Supplementary Table S1; Supplementary Data are available
the first model, samples harboring TT/DG at rs368234815 online at www.liebertpub.com/jir). This being a fine-mapping
and G/A genotype at rs117648444 (ie, a total of 6 patients) study of an established associated locus rather than a dis-
were left out from the analysis (denoted as adjusted 1); and covery study, multiple testing correction was deemed to be
in the second model, all samples that are heterozygous for unnecessary and, to prevent loss of power, was not carried out
rs117648444 (a total of 8 patients) were left out from the for any analysis. GraphPad Prism, version 6.0, was used for
analysis (denoted as adjusted 2). The restricted models were odds ratio (OR) calculation, receiver operating characteristic
also applied for analysis of association of SVR with the (ROC) curve, and area under the curve (AUC) analysis.
remaining SNPs rs8099917, rs4803221, and rs8113007. Comparison of areas under the ROC for different models
were done by DeLong’s test using the function ‘‘roc.test’’
Plasmids and site-directed mutagenesis from ‘‘pROC’’ R package (R Core Team 2015).
Genotype data of IFNL variants for superpopulations (or
A 4 kb fragment of IFNL3 promoter was constructed in a subpopulation) of EAS, SAS, AMR, EUR, and AFR were
pGL3 basic vector (Invitrogen). The construct described as extracted from the 1000 genome browser; Ensembl GRCh37
‘‘p1.4kbIL28B’’ in Chinnaswamy and others (2013) was used release 87—December 2016. In the 1000 Genomes data, the
to clone the remaining portion of the IFNL3 promoter using genotype status of rs368234815 (TT/DG) was not available
the following primers: IL28B4kbKpnFor: 5¢-GATATC for all populations, and hence, we used proxy SNP data
GGTACCCAGTGGAATTCAGGGCAAATTAC-3¢ and IL [rs74597329 (T/G)].
28B1.1kbEcoR1Rev: 5¢-GTAATTTGCCCTGAATTCCAC Logistic regression analysis was performed to test the as-
TG-3¢. The clone was confirmed by Sanger dideoxy se- sociation of each SNP adjusting for rs117648444 as a po-
quencing of plasmid DNA. All the enzymes used for DNA tential confounder using the R function ‘‘glm’’ with ‘‘logit’’
cloning were from NEB. We used a site-directed mutagen- link (R Core Team 2015). We regressed the binary response
esis kit (NEB) to incorporate the alternate allele at rs4803221 variable (SVR) on additively coded SNP genotype 1 at a time
in the promoter construct and further confirmed the mutation with rs117648444 held as a binary covariate. Estimates of
by Sanger sequencing of the plasmid. coefficients and standard errors were used to construct 95%
confidence interval (CI) for the adjusted log odds ratio and
Methylation and luciferase assay then transformed to corresponding CIs for adjusted odds ratio.
One microgram of plasmid DNA (referred as 4 kb Log likelihood ratio test: We performed logistic regression
rs4803221 ‘‘C’’ and 4 kb rs4803221 ‘‘G’’) was methyl- of SVR on additively coded SNP genotype by using R func-
ated with CpG methyltransferase (M.SssI) and GpC me- tion ‘‘glm.’’ Two models were compared, 1 a marginal model
thyltransferase (M.CviPI) enzymes (NEB). Briefly, 4 U of (eg, rs368234815), nested within another model (eg, both
enzyme and 160 mM of S-adenosylmethionine (SAM) were rs368234815 and rs4803221). Log likelihood ratio test was
added to 1 mg of plasmid DNA to make a reaction volume performed to compare the models by using the ‘‘lrtest’’
of 20 mL and incubated overnight at 37C. Reaction mix command from ‘‘lmtest’’ (R Core Team 2015) R package.
without enzyme and substrate was used as negative controls.
The methylated plasmid DNA was transfected into Results
HEK293T cells as described previously (Chinnaswamy and LD map of IFNL SNPs/variants in an Indian
others 2013). Ten nanograms of the reporter construct, along HCV (genotype 3) cohort
with 30 ng of each of plasmids coding for p65 (NF-kB), IRF7
(interferon regulatory factor), HCV NS5B polymerase (ge- The HCV cohort used in this study is described previously
notype 2a), and 0.5 ng of RIG-I (retinoic acid-inducible gene I) (Chinnaswamy and others 2014) where we showed association
plasmids, was used per well. Renilla luciferase plasmid of rs8099917 with response to IFN-RBV therapy in patients
pRLTK was used as a transfection control at 0.01 ng/well. infected with genotype 3 (subtype a or b) HCV. The patients
Twenty-four hours after transfection, cells were lysed and (n = 72), who underwent standard IFN-RBV therapy for 6
luminescence was recorded by GloMax 96-well plate reader months between the year 2009 and 2012, were drawn from
(Promega) with the dual injection system. Dual Luciferase tertiary care centers in different parts of the country except from
LINKAGE DISEQUILIBRIUM STRUCTURE AT THE IFNL LOCUS 5

southern India (Chinnaswamy and others 2014). SVR (defined SVR (Martinot-Peignoux and others 2009). On testing for
as HCV serum RNA test-negativity at 6 months after the ter- RVR using a dominant model, all the SNPs/variants except
mination of IFN-RBV therapy) was achieved by 51 patients, the activity variant rs117648444 showed significant associ-
while 21 were NRs to therapy as determined by their serum ation with RVR (Table 2).
HCV RNA status at 6 months after completion of therapy. When we tested the association of the 5 SNPs/variants with
The clinical characteristics of the patients are shown in Ta- SVR using a dominant model, only 2 SNPs rs8099917 and
ble 1. For the present study, we genotyped 6 additional SNPs/ rs4803221 showed significant association (Table 2, ORcr).
variants in this cohort: rs12979860, rs368234815, rs4803221, The allele and genotype frequencies of the different variants
rs8113007 (Tipu and others 2014), rs117648444, and rs7248668 in the SVR and NR group are shown in Supplementary
(Fig. 1). All the 7 SNPs/variants in the patient cohort were Table S1. Surprisingly, the IFN-l4-generating variant,
genotyped by DNA resequencing. The minor allele frequencies rs368234815, fell short of significance (P = 0.09). Among the
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

(MAFs) of the SNPs/variants are shown in Fig. 2A. Among the 2 significantly associated SNPs, rs4803221 had the better
7 SNPs/variants, rs7248668 was monomorphic in our study strength (OR = 5.9) and significance (P = 0.002) of association
population. LD between the remaining 6 SNPs/variants was with SVR. The activity variant rs117648444 failed to show
calculated and is shown in Fig. 2B (r2 values are depicted; for D¢ any significant association with SVR too. A previous study
values, refer Supplementary Table S2). The LD between reported from a northern Indian population has shown sig-
rs12979860 and rs368234815 was complete (LD, r2 = 1). We nificant association of rs12979860 with SVR in a larger co-
genotyped these 2 SNPs/variants in an additional control cohort hort of genotype 3 HCV-infected patients (Gupta and others
(n = 45; see the Materials and Methods section) and saw that the 2014), suggesting that our study may have lacked sufficient
LD was complete even in this control population. power to detect an association with this SNP.
In the rest of the article, we shall refer to only the func- However, we can conclude from these results that SNPs
tional variant while we are showing results of either that are not in high LD with the causal variant rs368234815
rs368234815 or rs12979860. (ie, rs8099917, r2 = 0.57 and rs4803221, r2 = 0.64) in our
study cohort are more strongly and significantly associated
with SVR than the ones which are in high (rs8113007,
Association of IFNL SNPs/variants r2 = 0.82) or complete (rs12979860, r2 = 1) LD with it (Fig. 2).
with RVR and SVR
We tested the association of the different IFNL SNPs/ Activity variant rs117648444 is a confounder
variants with RVR and SVR. We had previously recorded in the association of IFNL locus
from the same cohort that the covariates such as age, gender, SNPs/variants with SVR
HCV RNA copy numbers, and baseline ALT (alanine
transaminase) and AST (aspartate transaminase) levels were The S70 variant of IFN-l4 protein resulting from the minor
not confounders in the association of an IFNL SNP with A allele of the nonsynonymous variant rs117648444 leads to
RVR or SVR (Chinnaswamy and others 2014). RVR, de- decreased activity of the protein and has been previously
fined as HCV serum RNA test-negativity after 4 weeks of reported to be crucial in association analysis with SVR
IFN-RBV therapy, is known to be a predictor of subsequent (Fig. 1) (Galmozzi and Aghemo 2014; Terczyńska-Dyla and

FIG. 2. LD structure at the IFNL locus in the Indian HCV-infected patient cohort. (A) IFNL SNPs/variants and their
MAFs in our study population (N = 72; SVR, n = 51 and NR, n = 21). (B) The LD plot is based on the pairwise estimates of
disequilibrium coefficient (r2). Lighter regions are with low r2 values; while darker regions show higher r2 values. HCV,
hepatitis C virus; LD, linkage disequilibrium; MAFs, minor allele frequencies; NR, nonresponder; SVR, sustained viro-
logical response.
6 BHUSHAN ET AL.

number of samples in either RVR (N = 67) or SVR (N = 72, for crude and regression models; N = 66 for adjusted 1 and N = 64 for adjusted 2 models, also shown in Supplementary Table S1).
3.75 (0.72–19.41)

0.31 (0.036–2.72)

All the ORs shown are from univariate analysis (except ORreg); age and gender were also included in multivariate models but did not affect association of SNPs/variants with RVR/SVR. RVR
data were available only for 67 patients (54 achieved RVR, while 13 did not). SVR data were available for all 72 patients (n, SVR = 51; NR = 21). All SNPs/variants were tested for the same
others 2014). Therefore, the altered activity of the protein has

rs117648444
to be factored into the analysis presuming that IFN-l4-
generating polymorphism is the causal variant and increased

0.248

0.49






IFN-l4-mediated ISG activity is the causal mechanism be-
hind the association of IFNL locus SNPs with treatment re-
sponse to IFN-RBV therapy in chronic HCV infections.
We observed that we had 8 individuals who were het-
erozygous at rs117648444 and among them 6 were also

CI, confidence interval; NR, nonresponder; ORadj, odds ratio adjusted; ORcr, odds ratio crude; ORreg, odds ratio regression; SNPs, single-nucleotide polymorphisms.
heterozygous at rs368234815 (all of them in the SVR group)
6.66 (1.62–27.28)

3.66 (1.23–10.89)

3.66 (1.20–11.18)
and 2 were homozygous for the DG allele at rs368234815 (1
2.44 (0.86–6.89)

3.33 (1.25–8.84)
in SVR and the other in NR, NR group). We corrected for
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

rs8113007

the altered activity of the protein in our association analysis


0.033

0.039

0.015
0.01

0.14

of rs368234815, rs8099917, rs4803221, and rs8113007 with

Dominant model of inheritance for the minor alleles (eg TT/TT versus TT/DG +DG/DG for rs368234815) was used to compute OR in univariate models.
SVR, by using 2 models: adjusted 1 and adjusted 2 (refer to
the Materials and Methods section for a description of each
Association of Interferon Lambda Variants with Virological Response

model; analysis is shown in Supplementary Table S1). In


brief, in these models, we did a restricted analysis by ex-
cluding those patients who were able to express the low-
7.09 (1.87–26.92)

5.9 (1.95–17.89)

5.02 (1.65–15.28)

5.10 (1.63–15.94)

4.30 (1.56–11.83)

activity variant of the IFN-l4 protein.


The results are shown in Table 2. Intriguingly, by restricting
rs4803221

0.005

0.002

0.007

0.008

0.004

our analysis to those patients where only the high-activity IFN-


l4 could be expressed, functional variant rs368234815 now
showed significant association with SVR in both the models,
suggesting that the activity variant rs117648444 was a negative
confounder in the initial crude analysis (compare ORcr with
ORadj1, ORadj2 in Table 2). Similar results were seen for
rs8113007 (Table 2), which is in high LD (r2 = 0.82; Fig. 2) with
6.42 (1.70–24.20)

4.33 (1.47–12.75)

3.66 (1.22–10.88)

3.66 (1.20–11.17)

the functional variant. The activity variant rs117648444, unlike


3.33 (1.25–8.84)
rs8099917

its effect on the functional variant and rs8113007, does not seem
0.008

0.013

0.033

0.039

0.015

to negatively affect the association of rs8099917 and rs4803221


with SVR (Table 2). In fact, we saw evidence that the activity
variant had a positive confounding effect on rs8099917 and
rs4803221 (compare ORcr with ORadj1, ORadj2 in Table 2).
We used logistic regression analysis to test for the effect of
the activity variant on different IFNL SNPs/variants for their
association with SVR (Table 2, ORreg). After regressing out
6.14 (1.50–25.04)

4.00 (1.39–11.91)

4.00 (1.31–12.18)
2.73 (0.95–7.80)

3.68 (1.42–9.48)

the effect of the activity variant, rs368234815 showed a better


rs368234815

strength and significance of association than rs8099917, sim-


0. 015

0.098

0.022

0.025

0.006

ilar to our observations above with the restricted models. Si-


milarly, rs4803221 retained its strongest association among all
SNPs/variants with SVR even after adjusting for the effect of
the activity variant (Table 2) in the regression analysis model.
A recent study showed that rs8099917, even though is
*4 kb (Fig. 1) away from the functional variant rs368234815,
shows strong association with SVR, due to its high correlation
ORadj1 (95% CI)

ORadj2 (95% CI)

ORreg (95% CI)

(r2 = 0.74) with the DG-G haplotype giving rise to P70 variant
Table 2.

ORcr (95% CI)


OR (95% CI)

of IFN-l4 (Terczyńska-Dyla and others 2014). We did a


similar analysis by deducing the haplotypes involving
rs368234815 and rs117648444 in each of our patients by
haplotype phasing. We found 3 haplotypes to be present in
our cohort (rs368234815-rs117648444): 74.3% of TT-G;
P

20.1% DG-G; and 5.5% DG-A. We did not find any TT-A
haplotype in our cohort. We examined the correlation of the
Regression

SNPs in our study cohort with TT-G, DG-G, and DG-A hap-
Restricted
Model

lotypes (Table 3; for LD, r2 values are shown; for D¢ values


Crude

Crude

refer Supplementary Table S3). We saw that rs4803221, which


P < 0.05 is in bold.

shows the strongest association with SVR in our cohort, had


the highest correlation with DG-G haplotype (r2 = 0.87) and the
least correlation (r2 = 0.0007) with the DG-A haplotype.
To statistically validate the hypothesis that rs4803221
Endpoint

was the best predictor SNP in our cohort, we used log


RVR

SVR

likelihood ratio tests (Table 4). Comparing the different


variants for predicting SVR in nested models confirmed that
LINKAGE DISEQUILIBRIUM STRUCTURE AT THE IFNL LOCUS 7

Table 3. Correlation of the Single-Nucleotide Polymorphisms/Variants in Our Study


Cohort with Haplotypes Formed by the Functional and Activity Variants
and Their Predictive Values of Sustained Virological Response
LD (r2)
Haplotypea rs8099917 (T/G) rs4803221 (C/G) rs368234815 (TT/DG) rs8113007 (A/T)
TT-G/non TT-G 0.537 0.636 0.998 0.810
DG-G/non-DG-G 0.748 0.873 0.727 0.531
DG-A/non-DG-A 0.013 0.0007 0.191 0.214
TT-A/non-TT-A 0 0 0 0
PPV (95% CI) 0.81 (0.69–0.90) 0.83 (0.69–0.92) 0.80 (0.63–0.90) 0.78 (0.63–0.89)
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

NPV (95% CI) 0.50 (0.29–0.70) 0.54 (0.33–0.73) 0.41 (0.24–0.59) 0.40 (0.22–0.59)
AUC 0.67 0.7 0.64 0.61
95% CI 0.52–0.81 0.56–0.84 0.49–0.78 0.46–0.76
P 0.025 0.007 0.067 0.127
a
Haplotypes between rs368234815 (TT/DG) and rs117648444 (G/A).
AUC, area under the curve; LD, linkage disequilibrium; NPV, negative predictive value; PPV, positive predictive value.

rs4803221 is significantly a better predictor of SVR than Protective haplotypes TT-G and DG-A are linked
rs368234815, rs8113007, and rs8099917. We calculated the to the major alleles of rs4803221 and rs8099917
predictive values of the different SNPs/variants on SVR by
using AUC analysis under 2 cutoff levels (dominant and re- We tested the association of SVR among the 3 haplotypes
cessive models of inheritance). The best predictive SNP involving the functional and activity variants (Fig. 3A).
rs4803221 had the best AUC value of 0.7 and also had the best We saw that the DG-A haplotype showed a trend toward
positive and negative predictive values (Table 3). being protective in our cohort in line with previous stud-
However, when we statistically compared the AUC values ies (Galmozzi and Aghemo 2014; Terczyńska-Dyla and
of each of the other variants with that of the best predictor others 2014). (TT-G versus DG-G: OR = 3.33, 95% CI =
SNP rs4803221, we saw that rs4803221 was significantly 1.42–7.82, P = 0.006; DG-A versus DG-G: OR = 7.5, 95%
a better predictor of SVR than the functional variants CI = 0.81–68.93, P = 0.1 using 2-tailed Fisher’s exact test;
rs368234815 and rs8113007 but did not reach significance Fig. 3A.) By using log likelihood ratio tests, we obtained a
to be a better predictor than rs8099917 (Table 4). The significant P value (P = 0.026) by comparing the nested model
above results and those from regression models (Table 2) of TT-G versus TT-G+DG-A for association with SVR, sug-
suggested that rs4803221 may have an effect independent gesting a significant effect of the minor A allele of the activity
of both the functional and activity variants. To further test variant rs117648444 on attaining SVR in our cohort.
this hypothesis, we used the log likelihood ratio test and We sought to examine how the 2 protective haplo-
saw that the effect was not statistically significant (Table 4; types are linked to the major alleles of different IFNL SNPs.
compare rs368234815+rs117648444+rs4803221 versus First, we deduced haplotypes involving the 3 variants
rs368234815+rs117648444). rs368234815, rs4803221, and rs117648444 to examine how
the protective TT-G and DG-A haplotypes are linked to the
Table 4. Single-Nucleotide Polymorphism best predictive SNP rs4803221. Phasing at the 3 variant
rs4803221 Is a Better Predictor of Sustained positions showed that 100% of both the protective haplo-
Virological Response than rs368234815 types TT-G and DG-A were linked to the major allele C of
rs4803221 (Fig. 3B). Similar phasing by including the tag
Model 1 Model 2 Pa
SNP rs12979860, in stark contrast, showed that 100% of the
Log likelihood ratio testa protective DG-A haplotype was linked to its minor allele T
rs368234815+rs4803221 rs368234815 0. 021 (Fig. 3B). The protective DG-A haplotype was 100% linked
rs8099917+rs4803221 rs8099917 0.032 to the major allele T of rs8099917 and to the minor allele T
rs8113007+rs4803221 rs8113007 0.003 of rs8113007 (Fig. 3B) similar to the distribution at rs4803221
rs8099917+rs4803221 rs4803221 0.171 and the tag SNP, respectively.
rs368234815+rs4803221 rs4803221 0.466 However, the distribution of the DG-G and TT-G haplo-
rs8113007+rs4803221 rs4803221 0.139 types among the 2 alleles of rs8099917 was proportionally
½ 
½ 
rs368234815+ rs368234815+ different when compared to rs4803221 (86% of DG-G
rs117648444+ rs117648444 0.347
rs4803221 rs4803221 0.593 linked to the minor allele G and 99% of TT-G linked to
major allele T of rs8099917 versus 90% of DG-G linked to
Variant 1 Variant 2 P the minor allele G and 100% of TT-G linked to major allele
C of rs4803221). These results suggest that rs4803221 op-
DeLong’s test for comparing AUC values timally incorporates information from both the primary
rs368234815 rs4803221 0.015 functional variant (rs368234815) and a secondary modifying
rs8099917 0.194 functional variant (ie, the activity variant rs117648444) and
rs8113007 0.011 hence could be acting as the best ‘‘tagging-SNP’’ to report
a
Log likelihood ratio test compares model 1 versus model 2 (one on IFN-l4 function.
is nested in the other). Pa and P value <0.05 is significant and is We next sought to examine the nature of linkage of
shown in bold. IFNL SNPs with the functional and activity variant alleles
8 BHUSHAN ET AL.
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

FIG. 3. SNPs rs8099917 and rs4803221 are better predictors of SVR due to their linkage with the protective haplotypes
(of functional and activity variants) TT-G and DG-A. (A) DG-A haplotype shows a protective phenotype in our cohort (n,
SVR = 51; NR = 21) (left). Haplotypes were obtained for each patient after phasing at the 2 positions: rs368234815 and
rs117648444. Haplotypes were grouped as No IFN-l4 (TT-G), IFN-l4 P70 (DG-G), and IFN-l4 S70 (DG-A). The 3
haplotype groups were distributed into SVR and NR groups based on the status of the respective samples/patients they were
derived from and compared. ORs shown in the left panel are from univariate analysis. Logistic regression was also used to
calculate OR and 95% CI (TT-G: OR = 3.68, 95% CI = 1.42–9.48, P = 0.006; DG-A: OR = 9.82, 95% CI = 0.90–107.01,
P = 0.06). The right panel depicts the number of haplotypes in each of the 3 types in SVR and NR groups. (B) Distribution
of haplotypes involving rs368234815 and rs117648444 in linkage between the 2 alleles of rs4803221, rs12979860,
rs8099917, and rs8113007. Haplotypes were obtained for each patient after phasing at all the 6 variant positions. Proportion
of haplotypes involving rs368234815 and rs117648444 (TT-G, DG-A, and DG-G) that are linked with the 2 alleles of the
remaining SNPs are plotted. CI, confidence interval; OR, odds ratio.

in other world populations. For this purpose, we used the No effect of rs4803221 variants
genotype data of all the representative world populations on IFNL3 promoter activity
comprising around 2,500 individuals deposited in the 1000
genomes project database and deduced the haplotypes in- Our results described above allowed us to speculate that
volving all the SNPs/variants studied in our cohort (Sup- rs4803221 could be the best predictor of therapy response
plementary Table S4). because of the local LD structure that defines its correla-
The results showed that the DG-A haplotype (rs368234815- tion with the 2 important IFNL4 variants rs368234815 and
rs117648444) was linked to the SNPs rs12979860, rs4803221, rs117648444. However, another possible mechanism by
and rs8099917 as a single haplotype in all the populations in which IFNL SNPs can affect HCV clearance is by affecting
the 1000 genomes database (Supplementary Table S4). Similar the expression of IFNL3 (Chinnaswamy and others 2013;
to the linkage in our cohort, both the DG and A alleles (of the Chinnaswamy 2014; McFarland and others 2014; Lu and
functional and activity variants, respectively) were always others 2015). Therefore, it is important to examine alternate
linked to the T and C alleles of rs8099917 and rs4803221, mechanisms by which rs4803221 affects HCV clearance,
respectively; and always to the T allele of rs12979860 in all including that of regulation of IFNL3 promoter activity.
world populations, including the African population. This In fact, a previous study (Smith and others 2011), which
shows that, due to this linkage structure, all world popu- identified rs4803221 by deep sequencing-based fine-mapping
lations may be affected by the confounding effect of the analysis of the IFNL locus, found that it was 1 of the 2 SNPs
activity variant in genetic association studies similar to that had the best predictive value of IFN-RBV therapy response
the one observed in our study, involving SNPs/variants in an Australian cohort, and further speculated that rs4803221
rs368234815, rs12979860, rs4803221, and rs8099917. may be affecting the DNA methylation-based transcriptional
LINKAGE DISEQUILIBRIUM STRUCTURE AT THE IFNL LOCUS 9
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

FIG. 4. SNP rs4803221 does not affect transcription activity of the IFNL3 promoter. (A) Schematic representation of the
4 kb fragment of IFNL3 promoter (upper part). The rectangular box shows the CpG Island (position: chr19q13.2:
39737689–39739288; CpG count-137, UCSC genome browser, https://genome.ucsc.edu). The horizontal arrow denotes
transcription start site. This fragment is cloned into pGL3basic vector (Invitrogen) in front of the firefly luciferase gene for
reporter assay (lower part). (B, C) The wild-type and mutant constructs were methylated by CpG methyltransferase (M.SssI)
(B) and GpC methyltransferase (M.CviPI) (C) enzymes. The constructs were transfected along with a TK promoter-driven
Renilla luciferase construct and genes encoding p65, IRF7, HCV NS5B, and RIG-I in a 96-well plate containing HEK293T
cells. The ratios of firefly to Renilla luciferase values are plotted with error bars showing standard deviation from triplicates.
The data are representative of at least 2 independent experiments. Plus (‘‘+’’) and minus (‘‘-’’) represent the presence and
absence of different components (E, enzyme; S, substrate; TFs, transcription factors). IRF, interferon regulatory factor;
RIG-I, retinoic acid-inducible gene.

regulation of IFNL3 expression. This is because the allelic rs4803221 either before or after methylation by either en-
variants of rs4803221 disrupt a CpG motif in the CpG island zyme. We do note a minor increase in the activity from the
present upstream of the IFNL3 gene that also overlaps most of construct that carried the minor G allele at rs4803221
the IFNL4 genes (Chinnaswamy 2016) (Fig. 4A). compared to the one with the major C allele, after methyl-
To test this possibility, we cloned a 4 kb promoter DNA ation by both enzymes (Fig. 4B, C); however, this difference
region upstream of IFNL3 gene encompassing the CpG island was within the margin of error. These results argue against a
into a pGL3 basic vector (Invitrogen). DNA sequencing major effect of rs4803221 alone on IFNL3 expression and
showed the presence of C allele at rs4803221 in this promoter the possibility that the SNP singularly affects HCV clear-
clone; we used site-directed mutagenesis to generate the al- ance by this mechanism is therefore less likely.
ternate G allele at rs4803221 in the promoter construct. The 2
constructs were subject to in vitro methylation by the enzymes Discussion
M.Sss I and M.CviPI. Constructs subject to in vitro methyl-
ation or not were transfected into HEK293T cells as described A true causal SNP or its closely correlated SNP(s) will have
before (Chinnaswamy and others 2013) along with genes for the strongest association with the phenotype in genetic asso-
transcription factors (TFs) NF-kB and IRF7 that were also co- ciation studies (O’Brien and others 2015). In this study, we
overexpressed along with the HCV polymerase gene and show that the activity variant rs117648444 has a confounding
RIG-I gene to activate the RLR (RIG-I-like receptor) pathway effect on the association of IFNL locus SNPs/variants with
as described previously (Chinnaswamy and others 2016). response to IFN-RBV therapy in chronic HCV infections. The
The 4 kb promoter construct showed promoter activity activity variant rs117648444 is a nonsynonymous variant
only when TFs were also co-overexpressed and the RLR whose minor allele has been shown in vitro to decrease the
pathway activated (Fig. 4B, C). Methylation of the plasmids activity of IFN-l4 protein (Terczyńska-Dyla and others 2014)
by both enzymes decreased the promoter activity by 3- to thereby having a direct effect on its function. Two previous
4-folds. However, we did not see any major difference in studies have also shown the influence of rs117648444 on
the luciferase activity arising from either of the alleles of genetic association of the functional variant rs368234815
10 BHUSHAN ET AL.

(Galmozzi and Aghemo 2014; Terczyńska-Dyla and others founding effect of the activity variant (Fig. 5A). Since the
2014) with response to therapy in HCV infections, elucidating minor A allele of the activity variant is always linked to the
the protective role of the low-activity IFN-l4 protein com- DG risk allele of the functional variant (Fig. 3B and Supple-
pared to the high-activity variant. mentary Table S4), the latter is always subject to the negative
Our study explains how the activity variant modulates the confounding effect of the former; as a consequence, any SNP/
association test results involving certain important IFNL variant that has the minor A allele of the activity variant linked
SNPs/variants due to the confounding effect and how the to its protective allele will always be subject to a positive
nature of confounding is dependent on whether the minor A confounding effect (eg, rs4803221, rs8099917) (Fig. 5A).
allele of the activity variant is linked to the major or minor The most widely genotyped SNPs at the IFNL locus both
alleles of IFNL SNPs (Fig. 3B). Even though the restricted in HCV studies and others are rs12979860 and rs8099917
models were artificial and involved ‘‘neglecting’’ some sam- (Chinnaswamy 2016); since our results show the linkage of
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

ples from analysis, our logistic regression analysis gave results the minor A allele of the activity variant being conserved
that were consistent with our models (Table 2). across all human populations (Supplementary Table S4),
We further show that the linkage structure involving some these 2 ‘‘popular’’ SNPs should be corrected for the con-
of the most important IFNL SNPs/variants (Chinnaswamy founding effect of the activity variant in future studies to
2014, 2016) is highly conserved across several human investigate causal mechanism (s) behind IFNL SNP asso-
populations (Supplementary Table S4), and therefore, our ciation with human diseases.
findings should be relevant in all other populations, in- In our study, we saw that 7 out of 8 patients carrying the
cluding the African populations or those admixed with it. DG-A haplotype attained SVR, suggesting a predominant
However, the extent of confounding would vary based on effect of the IFN-l4S70 variant in our cohort on viral
(1) the relative protective effect of DG-A haplotype com- clearance. However, whether this can be replicated in larger
pared to the TT-G haplotype (the higher the effect of DG-A cohorts of genotype 3 HCV patients remains to be tested.
haplotype, the larger will be the confounding effect) and (2) The small sample size is indeed a major limitation of our
the LD (r2) values of the different IFNL SNPs with the study, and therefore, the results from this study pertaining to
functional variant rs368234815. the predictive abilities of various SNPs/variants and also the
A strong LD of any given IFNL SNP with the functional large observed effect of the activity variant have to be
variant also makes it more susceptible to the negative con- cautiously interpreted. However, our results offer an

FIG. 5. Activity SNP rs117648444 acts as a negative or positive confounder in association of IFNL SNPs/variants with
human diseases dictated by an underlying linkage structure. (A) Activity variant minor A allele when linked to risk allele of an
IFNL SNP (eg, rs12979860) exerts a negative confounding effect on that SNP; the same when linked to the protective allele of
an IFNL SNP (eg, rs8099917) exerts a positive confounding effect of that SNP in genetic association studies. Arrows indicate
haplotypes formed between the alleles of the activity and functional variants, which in turn are linked to any other IFNL SNP.
Dotted line denotes moderate to high LD between an IFNL SNP with the functional variant rs368234815. P, protective; R, risk;
P2P, extent of positive confounding; P2N, extent of negative confounding. (B) Confounding by the activity variant likely arises
due to early evolutionary events that shaped the nucleotide changes at the 2 key variant positions. Both the changes (at
functional and activity variant positions) seem to have happened on distinct allelic backgrounds of IFNL SNPs/variants.
Concordance denotes concurrence of the given IFNL SNP with the phenotype derived from the haplotype involving the
functional and activity variants. For example, ‘‘C’’ allele at rs4803221 is predicted to have a protective phenotype (pred.),
while the ‘‘G’’ allele is predicted to show a risk phenotype due to the nature of LD with the functional variant rs368234815 (in
dominant models of inheritance). These predicted phenotypes of rs4803221 perfectly concur with the predicted phenotypes of
3 haplotypes TT-G, DG-G, and DG-A (involving the functional and activity variants) linked to alternate alleles, and therefore,
score 3 out of 3 on the concordance score. The numbers in parenthesis show the MAFs of the 3 variants rs4803221 (0.18),
rs12979860/rs368234815 (0.25), and rs117648444 (0.05) in our HCV patient cohort.
LINKAGE DISEQUILIBRIUM STRUCTURE AT THE IFNL LOCUS 11

explanation of why certain SNPs such as rs8099917 and now been associated with several other human diseases and/or
rs4803221, which are not in strong LD with the functional their clinical characteristics (Gaudieri and others 2012; An-
variant unlike the tag SNP rs12979860, show stronger as- gulo and others 2015; Eslam and others 2015; Arpaci and
sociation with the phenotype in several previous studies others 2016; Corrales and others 2016; Dominguez-Molina
involving different populations (Doehring and others 2010; and others 2016; Morán-Auth and others 2016; Price and
Lindh and others 2011; Smith and others 2011; Suppiah and others 2016). Our results should be useful in designing and
others 2011; Antaki and others 2013; Khudayberganova and interpreting future case–control studies with IFNL SNPs and
others 2014; O’Connor and others 2016). It would also mean different human diseases.
that the best predictive SNP would not be the same in dif-
ferent populations but would vary according to the local LD Acknowledgments
structure at the IFNL locus.
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

The DG allele is thought to be the ancient allele and the TT The funding for this work was provided by the Department
allele was acquired at a later stage in human evolution (Key of Biotechnology, Government of India, as grant No. BT/
and others 2014). Such information is not available for the PR7035/MED/12/579/2012 to S.C. We thank Prof. Partha
activity variant. We can speculate that the events leading to Majumder, Founder Director, NIBMG (Kalyani, India) for
both these changes affecting the expression and activity of the his support and encouragement for this work. We thank the
IFN-l4 protein seem to have occurred on the protective allele following clinicians who provided us the patients for the
background of some SNPs such as rs8099917 and rs4803221 study: Drs. Kaushik Das (IPGME&R, Kolkata, India);
(Fig. 5B). For other SNPs such as rs12979860 (and others Shalimar (AIIMS, New Delhi, India); Ajay Duseja (PGI,
in high LD with it), events leading to the generation of the Chandigarh, India); Deepak Amarapurkar (Bombay Hos-
activity variant seem to have occurred on the risk allele pital, Mumbai, India); Rosangluaia (Civil Hospital, Ai-
background; however, since this change leads to a protective zawl, India); Sujit Chowdhury (AMRI Hospital, Kolkata,
phenotype, it becomes the source of confounding that we India); Ashokanand Konar (Peerless Hospital, Kolkata,
observed in our study (Fig. 5B). Population genetics studies India). Our special thanks to Prof. Abhijit Chowdhury
will reveal on how these events have led to the current (SDLD, IPGME&R, Kolkata, India) for coordinating with
structure of the IFNL locus in different world populations. all clinicians for this study. We thank Dr. Analabha Basu,
In our cohort, rs4803221 even though was the best predictor NIBMG, for helpful discussions. We thank Dr. Priyodarshi
SNP for SVR based on results from some tests, other statistical Basu and Ankita Chatterjee for some of the control sam-
tests did not agree that it had an effect independent of both the ples and Dr. Parthasarathi Bhattacharyya, IPCR (Kolkata,
functional and activity variants (Tables 2 and 4). However, the India), for the asthma control samples. Finally, we thank
low sample size of our study may have not had the power to all the study participants who consented for the study.
unravel any independent effect of the SNP rs4803221 in at-
taining SVR. Therefore, to examine an alternate mechanism, Author Disclosure Statement
we carried out reporter assays to examine if rs4803221 regu-
No competing financial interests exist.
lates IFNL3 promoter activity as speculated by Smith and
others (2011). Our reporter assays do not show evidence for
rs4803221 to modulate IFNL3 promoter activity either before References
or after in vitro methylation of promoter constructs. Aka PV, Kuniholm MH, Pfeiffer RM, Wang AS, Tang W, Chen
However, we have not tested if rs4803221 in conjunction S, Astemborski J, Plankey M, Villacres MC, Peters MG,
with other SNPs (such as rs12979860) present on the CpG Desai S, Seaberg EC, Edlin BR, Strickler HD, Thomas DL,
island could affect promoter activity before or after meth- Prokunina-Olsson L, Sharp GB, O’Brien TR. 2014. Asso-
ylation of the CpG island. Since there is evidence from a ciation of theIFNL4-DG allele with impaired spontaneous
recent article (Waring and others 2016) that the minor allele clearance of hepatitis C virus. J Infect Dis 209(3):350–354.
of rs12979860 associates in a dominant model of inheritance Akuta N, Suzuki F, Hirakawa M, Kawamura Y, Yatsuji H,
with the methylation levels of the CpG island, such studies Sezaki H, Suzuki Y, Hosaka T, Kobayashi M, Saitoh S, Ikeda
need to be carried out to test if alternate mechanisms are K, Chayama K, Nakamura Y, Kumada H. 2010. Amino acid
also behind the association of IFNL locus SNPs with HCV substitution in hepatitis C virus core region and genetic
infections and other human diseases. variation near the interleukin 28B gene predict viral response
In summary, we show that a confounding effect is manifest to telaprevir with peginterferon and ribavirin. Hepatology
due to the activity variant rs117648444 in our association 52(2):421–429.
study between IFNL SNPs/variants and viral clearance in- Angulo J, Pino K, Echeverrı́a-Chagas N, Marco C, Martı́nez-
Valdebenito C, Galeno H, Villagra E, Vera L, Lagos N,
volving genotype 3 HCV-infected patients. The confounding
Becerra N, Mora J, Bermúdez A, Cárcamo M, Dı́az J, Miquel
arises due to the linkage of the activity variant with different
JF, Ferrés M, López-Lastra M. 2015. Association of single
IFNL SNPs and likely due to its direct effect on the phenotype nucleotide polymorphisms in IL28B, but not TNF-a, with
as a result of modulation of the activity of the IFN-l4 protein. severity of disease caused by Andes virus. Clin Infect Dis
The nature and level of confounding vary with the level of LD 61(12):e62–e69.
that the activity variant (or its haplotype with the functional Antaki N, Bibert S, Kebbewar K, Asaad F, Baroudi O, Alideeb
variant) has with the particular IFNL SNP and depend on S, Hadad M, Abboud D, Sabah H, Bochud PY, Negro F.
whether the minor allele A of the activity variant is linked to 2013. IL28B polymorphisms predict response to therapy
the major or minor alleles of the concerned SNP(s). Under- among chronic hepatitis C patients with HCV genotype 4. J
standing the confounding effect is critical to identify the best Viral Hepat 20(1):59–64.
predictive SNP in a population and to adjust for its effect in Arpaci D, Karakas Celik S, Can M, Cakmak Genc G, Kuzu F,
association studies, especially since IFNL locus SNPs have Unal M, Bayraktaroglu T. 2016. Increased serum levels of
12 BHUSHAN ET AL.

IL-28 and IL-29 and the protective effect of IL28B GJ, van Bömmel F, Brodzinski A, Fülöp B, Berg T. 2012.
rs8099917 polymorphism in patients with Hashimoto’s thy- Combined effects of different interleukin-28B gene variants
roiditis. Immunol Invest 45(7):668–678. on the outcome of dual combination therapy in chronic hep-
Chinnaswamy S. 2014. Genetic variants at the IFNL3 locus and atitis C virus type 1 infection. Hepatology 55(6):1700–1710.
their association with hepatitis C virus infections reveal novel Galmozzi E, Aghemo A. 2014. Nonsynonymous variant Pro70Ser
insights into host-virus interactions. J Interferon Cytokine (rs117648444) in IFNL4 gene identifies carriers of the
Res 34(7):479–497. rs368234815 DG allele with higher HCV RNA decline during
Chinnaswamy S. 2016. Gene-disease association with human the first 4 weeks of pegylated interferon and ribavirin therapy
IFNL locus polymorphisms extends beyond hepatitis C virus in HCV-1 patients. J Clin Virol 59(4):274–275.
infections. Genes Immun 17(5):265–275. Gaudieri S, Lucas M, Lucas A, McKinnon E, Albloushi H,
Chinnaswamy S, Bhushan A, Behera AK, Ghosh S, Rampurkar Rauch A, di Lulio J, Martino D, Prescott SL, Tulic MK.
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

V, Chandra V, Pandit B, Kundu TK. 2016. Roles for tran- 2012. Genetic variations in IL28B and allergic disease in
scription factors Sp1, NF-kB, IRF3, and IRF7 in expression children. PLoS One 7(1):e30607.
of the human IFNL4 gene. Viral Immunol 29(1):49–63. Ge D, Fellay J, Thompson AJ, Simon JS, Shianna KV, Urban
Chinnaswamy S, Chatterjee S, Boopathi R, Mukherjee S, TJ, Heinzen EL, Qiu P, Bertelsen AH, Muir AJ, Sulkowski
Bhattacharjee S, Kundu TK. 2013. A single nucleotide poly- M, McHutchison JG, Goldstein DB. 2009. Genetic variation
morphism associated with hepatitis C virus infections located in IL28B predicts hepatitis C treatment induced viral clear-
in the distal region of the IL28B promoter influences NF-kB- ance. Nature 461(7262):399–401.
mediated gene transcription. PLoS One 8(10):e75495. Gondeau C, Pageaux GP, Larrey D. 2015. Hepatitis C virus
Chinnaswamy S, Das K, Bairagya BB, Bhattacharyya C, Sha- infection: are there still specific problems with genotype 3?
limar, Duseja A, Amarapurkar D, Rosangluaia, Chowdhury World J Gastroenterol 21(42):12101–12113.
S, Konar A, Majumder PP. 2014. Association of IL28B single Gupta AC, Trehanpati N, Sukriti S, Hissar S, Midha V, Sood A,
nucleotide polymorphism rs8099917 with response to treat- Sarin SK. 2014. Interleukin-28b CC genotype predicts early
ment in genotype 3 HCV-infected patients from India. Trop treatment response and CT/TT genotypes predicts non-
Gastroenterol 35(2):96–102. response in patients infected with HCV genotype 3. J Med
Corrales I, Solano C, Amat, P, Giménez E, de la Cámara R, Virol 86(4):707–712.
Nieto J, López, J, Garcı́a-Noblejas A, Piñana JL, Navarro D. Kattakuzhy S, Levy R, Rosenthal E, Tang L, Wilson E, Kottilil S.
2016. IL28B genetic variation and cytomegalovirus-specific 2016. Hepatitis C genotype 3 disease. Hepatol Int 10(6):861.
T-cell immunity in allogeneic stem cell transplant recipients. Key FM, Peter B, Dennis MY, Huerta-Sánchez E, Tang W,
J Med Virol 89(4):685–695. Prokunina-Olsson L, Nielsen R, Andrés AM. 2014. Selection on
di Lulio J, Ciuffi A, Fitzmaurice K, Kelleher D, Rotger M, a variant associated with improved viral clearance drives local,
Fellay J, Martinez R, Pulit S, Furrer H, Günthard HF, Bat- adaptive pseudogenization of interferon lambda 4 (IFNL4).
tegay M, Bernasconi E, Schmid P, Hirschel B, Barnes E, PLoS Genet 10(10):e1004681.
Klenerman P, Telenti A, Rauch A, Swiss HIV Cohort Study. Khudayberganova D, Sugiyama M, Masaki N, Nishida N,
2011. Estimating the net contribution of interleukin-28B Mukaide M, Sekler D, Lalipov R, Nataliya K, Dildora S,
variation to spontaneous hepatitis C virus clearance. Hepa- Sharapov S, Usmanova G, Raxmanov M, Musabaev E, Mi-
tology 53(5):1446–1454. zokami M. 2014. IL28B polymorphisms and clinical impli-
Doehring A, Hofmann WP, Schlecker C, Zeuzem S, Susser S, cations for hepatitis C virus infection in Uzbekistan. PLoS
Geisslinger G, Sarrazin C, Lötsch J. 2010. Screening for One 9(3):e93011.
IL28B gene variants identifies predictors of hepatitis C Kotenko SV, Gallagher G, Baurin VV, Lewis-Antes A, Shen M,
therapy success. Antivir Ther 15(8):1099–1106. Shah NK, Langer JA, Sheikh F, Dickensheets H, Donnelly
Dominguez-Molina B, Tarancon-Diez L, Hua S, Abad-Molina RP. 2003. IFN-lambdas mediate antiviral protection through
C, Rodriquez-Gallego E, Machmach K, Vidal F, Tural C, a distinct class II cytokine receptor complex. Nat Immunol
Moreno S, Goñi JM, Ramı́rez de Arellano E, Del Val M, 4(1):69–77.
Gonzalez-Escribano MF, Del Romero J, Rodriquez C, Capa Lindh M, Lagging M, Arnholm B, Eilard A, Nilsson S, Nork-
L, Viciana P, Alcamı́ J, Yu XG, Walker BD, M L, Lichterfeld rans G, Söderholm J, Wahlberg T, Wejstål R, Westin J,
M, Ruiz-Mateos E, ECRIS integrated in the Spanish AIDS Hellstrand K. 2011. IL28B polymorphisms determine early
Research Network. 2016. HLA-B*57 and IFNL4-related viral kinetics and treatment outcome in patients receiving
polymorphisms are associated with protection against HIV-1 peginterferon/ribavirin for chronic hepatitis C genotype 1. J
disease progression in controllers. Clin Infect Dis pii:ciw833. Viral Hepat 18(7):e325–e331.
Eslam M, Hashem AM, Leung R, Romero-Gomez M, Berg T, Lu YF, Mauger DM, Goldstein DB, Urban TJ, Weeks KM,
Dore GJ, Chan HL, Irving WL, Sheridan D, Abate ML, Bradrick SS. 2015. IFNL3 mRNA structure is remodeled by a
Adams LA, Mangia A, Weltman M, Bugianesi E, Spengler functional non-coding polymorphism associated with hepa-
U, Shaker O, Fisher J, Mollison L, Cheng W, Powell E, titis C virus clearance. Sci Rep 5:16037.
Nattermann J, Riordan S, McLeod D, Armstrong NJ, Douglas Martinot-Peignoux M, Maylin S, Moucari R, Ripault MP,
MW, Liddle C, Booth DR, George J, Ahlenstiel G, Interna- Boyer N, Cardoso AC, Giuily N, Castelnau C, Pouteau M,
tional Hepatitis C Genetics Consortium (IHCGC). 2015. Stern C, Aupérin A, Bedossa P, Asselah T, Marcellin P. 2009.
Interferon-l rs12979860 genotype and liver fibrosis in viral Virological response at 4 weeks to predict outcome of hep-
and non-viral chronic liver disease. Nat Commun 6:6422. atitis C treatment with pegylated interferon and ribavirin.
Fischer J, Böhm S, George J, Sarrazin C, Berg T. 2013. Impact Antivir Ther 14(4):501–511.
of cohort size and host factors on combined analysis of In- McFarland AP, Horner SM, Jarret A, Joslyn RC, Bindewald E,
terleukin 28B rs12979860 and rs8099917 in hepatitis C virus Shapiro BA, Delker DA, Hagedorn CH, Carrington M, Gale
infection. Hepatology 57(1):416–417. M Jr, Savan R. 2014. The favorable IFNL3 genotype escapes
Fischer J, Böhm S, Scholz M, Müller T, Witt H, George J, mRNA decay mediated by AU-rich elements and hepatitis C
Sarrazin C, Susser S, Schott E, Suppiah V, Booth DR, Stewart virus-induced microRNAs. Nat Immunol 15(1):72–79.
LINKAGE DISEQUILIBRIUM STRUCTURE AT THE IFNL LOCUS 13

Meissner EG, Bon D, Prokunina-Olsson L, Tang W, Masur H, gene variants rs12979860 and rs8099917 in chronic hepatitis
O’Brien TR, Herrmann E, Kottilil S, Osinusi A. 2014. C genotype 4. Liver Int 34(6):890–895.
IFNL4-DG genotype is associated with slower viral clearance Rauch A, Kutalik Z, Descombes P, Cai T, Di Lulio J, Mueller
in hepatitis C, genotype-1 patients treated with sofosbuvir T, Bochud M, Battegay M, Bernasconi E, Borovicka J, Co-
and ribavirin. J Infect Dis 209(11):1700–1704. lombo S, Cerny A, Dufour JF, Furrer H, Günthard HF, Heim
Morán-Auth Y, Penna-Martinez M, Perner D, Susser S, Wicker M, Hirschel B, Malinverni R, Moradpour D, Müllhaupt B,
S, Zeuzem S, Sarrazin C, Badenhoop K. 2016. IL28B gene Witteck A, Beckmann JS, Berg T, Bergmann S, Negro F,
variants and glucose metabolism in Type 2 Diabetes. Hum Telenti A, Bochud PY, Swiss Hepatitis C Cohort Study,
Immunol 77(12):1280–1283. Swiss HIV Cohort Study. 2010. Genetic variation in IL28B
O’Brien TR, Pfeiffer RM, Paquin A, Lang kuhs KA, Chen S, is associated with chronic hepatitis C and treatment failure:
Bonkovsky HL, Edlin BR, Howell CD, Kirk GD, Kuniholm a genome-wide association study. Gastroenterology 138(4):
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

MH, Morgan TR, Strickler HD, Thomas DL, Prokunina- 1338–1345.


Olsson L. 2015. Comparison of functional variants in IFNL4 Sheppard P, Kindsvogel W, Xu W, Henderson K, Schlutsmeyer
and IFNL3 for association with HCV clearance. J Hepatol S, Whitmore TE, Kuestner R, Garrigues U, Birks C, Rora-
63(5):1103–1110. back J, Ostrander C, Dong D, Shin J, Presnell S, Fox B,
O’Brien TR, Prokunina-Olsson L, Donnelly RP. 2014. IFN-l4: Haldeman B, Cooper E, Taft D, Gilbert T, Grant FJ, Tackett
the paradoxical new member of the interferon lambda family. M, Krivan W, McKnight G, Clegg C, Foster D, Klucher KM.
J Interferon Cytokine Res 34(11):829–838. 2003. IL-28, IL-29 and their class II cytokine receptor IL-
O’Connor KS, Read SA, Wang M, Schibeci S, Eslam M, Ong 28R. Nat Immunol 4(1):63–68.
A, Weltman MD, Douglas MW, Mazzola A, Craxi A, Petta S, Smith KR, Suppiah V, O’Connor K, Berg T, Weltman M, Abate
Stewart GJ, Liddle C, George J, Ahlenstiel G, Booth DR. ML, Spengler U, Bassendine M, Matthews G, Irving WL,
2016. IFNL3/4 genotype is associated with altered immune Powell E, Riordan S, Ahlenstiel G, Stewart GJ, Bahlo M,
cell populations in peripheral blood in chronic hepatitis C George J, Booth DR, International Hepatitis C Genetics
infection. Genes Immun 17(6):328–334. Consortium. 2011. Identification of improved IL28B SNPs
Palenzuela Gardón D, Guillen IA, Fernández JR, Camacho H, and haplotypes for prediction of drug response in treatment of
Estevez ZC, Dueñas S, Alvares-Lajonchere L, Amador Y, hepatitis C using massively parallel sequencing in a cross-
Martinez-Donato G, Han J, Zhang Z, Zhang X, Gao Y, sectional European cohort. Genome Med 3(8):57.
Campaña JR, Novoa LI. 2016. Assessment of IL-28: Suppiah V, Gaudieri S, Armstrong NJ, O’Connor KS, Berg T,
rs12979860 and rs8099917 Polymorphisms in a Cohort of Weltman M, Abate ML, Spengler U, Bassendine M, Dore GJ,
Cuban Chronic HCV Genotype 1b Patients. J Biomol Tech. Irving WL, Powell E, Hellard M, Riordan S, Matthews G,
28(2):80–86. Sheridan D, Nattermann J, Smedile A, Müller T, Hammond
Palmieri O, Ippolito AM, Margaglione M, Valvano MR, Giof- E, Dunn D, Negro F, Bochud PY, Mallal S, Ahlenstiel G,
freda D, Fasano M, D’Andrea G, Corritore G, Milella M, Stewart GJ, George J, Booth DR, International Hepatitis C
Andriulli N, Morisco F, Giannitrapani L, Latiano A, Fontana Genetics Consortium (IHCGC). 2011. IL-28B, HLA-C, and
R, Gatti P, Tundo P, Barone M, Cozzolongo R, Santantonio KIR variants additively predict response to therapy in chronic
T, Andriulli A. 2014. Variation in genes encoding for inter- hepatitis C virus infection in a European Cohort: a cross-
feron l-3 and l-4 in the prediction of HCV-1 treatment- sectional study. PLoS Med 8(9):e1001092.
induced viral clearance. Liver Int 34(9):1369–1377. Suppiah V, Moldovan M, Ahlenstiel G, Berg T, Weltman M,
Price AA, Tedesco D, Prasad MR, Workowski KA, Walker Abate ML, Bassendine M, Spengler U, Dore GJ, Powell E,
CM, Suthar MS, Honegger JR, Grakoui A. 2016. Prolonged Riordan S, Sheridan D, Smedile A, Fragomeli V, Müller T,
activation of innate antiviral gene signature after childbirth is Bahlo M, Stewart GJ, Booth DR, George J. 2009. IL28B is
determined by IFNL3 genotype. Proc Natl Acad Sci U S A associated with response to chronic hepatitis C interferon-
113(38):10678–10683. alpha and ribavirin therapy. Nat Genet 41(10):1100–1104.
Prokunina-Olsson L, Muchmore B, Tang W, Pfeiffer RM, Park Tanaka Y, Nishida N, Sugiyama M, Kurosaki M, Matsuura K,
H, Dickensheets H, Hergott D, Porter-Gill P, Mumy A, Ko- Sakamoto N, Nakagawa M, Korenaga M, Hino K, Hige S, Ito
haar I, Chen S, Brand N, Tarway M, Liu L, Sheikh F, As- Y, Mita E, Tanaka E, Mochida S, Murawaki Y, Honda M,
temborski J, Bonkovsky HL, Edlin BR, Howell CD, Morgan Sakai A, Hiasa Y, Nishiguchi S, Koike A, Sakaida I, Im-
TR, Thomas DL, Rehermann B, Donnelly RP, O’Brien TR. amura M, Ito K, Yano K, Masaki N, Sugauchi F, Izumi N,
2013. A variant upstream of IFNL3 (IL28B) creating a new Tokunaga K, Mizokami M. 2009. Genome-wide association
interferon gene IFNL4 is associated with impaired clearance of IL28B with response to pegylated interferon-alpha and
of hepatitis C virus. Nat Genet 45(2):164–171. ribavirin therapy for chronic hepatitis C. Nat Genet 41(10):
Puri P, Saraswat VA, Dhiman RK, Anand AC, Acharya SK, 1105–1109.
Singh SP, Chawla YK, Amarapurkar DN, Kumar A, Arora A, Terczyńska-Dyla E, Bibert S, Duong FH, Krol I, Jørgensen S,
Dixit VK, Koshy A, Shood A, Duseja A, Madan K, Kapoor Collinet E, Kutalik Z, Aubert V, Cerny A, Kaiser L, Mal-
D, Srivastava A, Kumar A, Wadhawan M, Goel A, Verma A, inverni R, Mangia A, Moradpour D, Müllhaupt B, Negro F,
Shalimar,, Pande G, Malik R, Agrawal S. 2016. Indian Na- Santoro R, Semela D, Semmo N, Swiss Hepatitis C Cohort
tional Association for Study of the Liver (INASL) guidance Study Group, Heim MH, Bochud PY, Hartmann R. 2014.
for antiviral therapy against HCV infection. J Clin Exp He- Reduced IFNl4 activity is associated with improved HCV
patol 6(2):119–145. clearance and reduced expression of interferon-stimulated
R Core Team. 2015. R: a language and environment for sta- genes. Nat Commun 5:5699.
tistical computing. Vienna: R Foundation for Statistical Tipu I, Marriage F, Farooqi ZR, Platt H, Athar MA, Day
Computing. Available at www.R-project.org PJ, Short A. 2014. The IFN-lambda Genetic polymor-
Ragheb MM, Nemr NA, Kishk RM, Mandour MF, Abdou MM, phism association with the viral clearance induced by hep-
Matsuura K, Watanabe T, Tanaka Y. 2014. Strong prediction atitis C virus treatment in Pakistani patients. Hepat Mon
of virological response to combination therapy by IL28B 14(3):e15076.
14 BHUSHAN ET AL.

Waring JF, Davis JW, Dumas E, Cohen D, Idler K, Abel S, Address correspondence to:
Georgantas R, Podsadecki T, Dutta S. 2017. Epigenetic Dr. Sreedhar Chinnaswamy
analysis of the IFNl3 gene identifies a novel marker for re- National Institute of Biomedical Genomics
sponse to therapy in HCV-infected subjects. J Viral Hepat P.O.:N.S.S.
24(5):397–403. Kalyani
Zeuzem S, Soriano V, Asselah T, Bronowicki JP, Lohse AW, West Bengal 741251
Mullihaupt B, Schuchmann M, Bourliere M, Buti M, Ro- India
berts SK, Gane EJ, Stern JO, Vinisko R, Kukoji G, Gal-
livan JP, Bocher W, Mensa FJ. 2013. Faldaprevir and E-mail: sc2@nibmg.ac.in
deleobuvir for HCV genotype 1 infection. N Engl J Med 369:
630–639. Received 11 January 2017/Accepted 7 June 2017
Downloaded by Tufts University package NERL from online.liebertpub.com at 07/20/17. For personal use only.

Das könnte Ihnen auch gefallen