1. Genetics and Genomics
  2. Microbiology and Infectious Disease
Download icon

Adaptation of hepatitis C virus to interferon lambda polymorphism across multiple viral genotypes

  1. Nimisha Chaturvedi  Is a corresponding author
  2. Evguenia S Svarovskaia
  3. Hongmei Mo
  4. Anu O Osinusi
  5. Diana M Brainard
  6. G Mani Subramanian
  7. John G McHutchison
  8. Stefan Zeuzem
  9. Jacques Fellay  Is a corresponding author
  1. École Polytechnique Fédérale de Lausanne, Switzerland
  2. Swiss Institute of Bioinformatics, Switzerland
  3. Gilead Sciences Inc, United States
  4. Goethe University Hospital, Germany
  5. Lausanne University Hospital, Switzerland
Research Article
  • Cited 1
  • Views 224
  • Annotations
Cite this article as: eLife 2019;8:e42542 doi: 10.7554/eLife.42542

Abstract

Genetic polymorphism in the interferon lambda (IFN-λ) region is associated with spontaneous clearance of hepatitis C virus (HCV) infection and response to interferon-based treatment. Here, we evaluate associations between IFN-λ polymorphism and HCV variation in 8729 patients (Europeans 77%, Asians 13%, Africans 8%) infected with various viral genotypes, predominantly 1a (41%), 1b (22%) and 3a (21%). We searched for associations between rs12979860 genotype and variants in the NS3, NS4A, NS5A and NS5B HCV proteins. We report multiple associations in all tested proteins, including in the interferon-sensitivity determining region of NS5A. We also assessed the combined impact of human and HCV variation on pretreatment viral load and report amino acids associated with both IFN-λ polymorphism and HCV load across multiple viral genotypes. By demonstrating that IFN-λ variation leaves a large footprint on the viral proteome, we provide evidence of pervasive viral adaptation to innate immune pressure during chronic HCV infection.

https://doi.org/10.7554/eLife.42542.001

Introduction

Infection with hepatitis C virus (HCV), a positive strand RNA virus of the Flaviviridae family, represents a major health problem, with an estimated 71 million chronically infected patients worldwide (WHO, 2017). In the absence of treatment, 15–30% of individuals with chronic HCV infection develop serious complications including cirrhosis, hepatocellular carcinoma and liver failure (Shepard et al., 2005; Alter and Seeff, 2000; Li et al., 2015; Drummer, 2014).

Seven major genotypes of HCV have been described, further divided into several subtypes (Simmonds, 2004; Smith et al., 2014). Moreover, within each infected individual, multiple distinct HCV variants co-exist as quasipecies (Farci et al., 2000). Inter-host and intra-host HCV evolution is shaped by multiple forces, including human immune pressure (Merani et al., 2011). To investigate the complex interactions between host and pathogen at the level of genetic variation, we proposed a genome-to-genome approach that allows the joint analysis of host and pathogen genomic data (Bartha et al., 2013). Using an unbiased association study framework, a genome-to-genome analysis aims at identifying the escape mutations that accumulate in the pathogen genome in response to host genetic variants. Ansari et al. (2017) used this approach to analyze a cohort of individuals of white ancestry predominantly infected with genotype 3a HCV; they identified associations between viral variants and human polymorphisms in the interferon lambda (IFN-λ) and HLA regions, demonstrating an impact of both innate and acquired immunity on HCV sequence variation during chronic infection.

The IFN-λ association is of particular interest considering the known impact of this polymorphic region on spontaneous clearance of HCV and on response to interferon-based treatment (Ge et al., 2009; Rauch et al., 2010; Thomas et al., 2009; Tanaka et al., 2009). The rs12979860 variant, which is located 3 kb upstream of IL28B (encoding IFN-λ3) and lies within intron 1 of IFNL4, showed the strongest correlation with treatment-induced clearance of infection in the first report (Ge et al., 2009). More recent studies have shown that rs12979860 is in fact a marker for a dinucleotide insertion/deletion polymorphism, IFNL4 rs368234815 [ΔG > TT], which causes a frameshift that abrogates IFN-λ4 protein production (Prokunina-Olsson et al., 2013). The two variants (rs12979860 and rs368234815) are in strong linkage disequilibrium in European and Asian populations (r2 = 0.98 in CEU and 1.00 in CHB and JPT): the rs12979860 C allele, associated with a higher rate of spontaneous HCV clearance and better response to interferon-based treatment, is found on the same haplotype as the rs368234815 TT allele and is thus tagging the absence of IFN-λ4 protein.

Here, we aim at characterizing the importance of innate immune response in modulating chronic HCV infection by describing the footprint of IFNL4 variation in the viral proteome. Using samples and data from a heterogeneous group of 8,729 HCV-infected individuals in a cross-sectional study design, we genotyped the single nucleotide polymorphism (SNP) rs12979860 and obtained partial sequences of the HCV genome (NS3, NS4A, NS5A and NS5B genes). We tested for associations between rs12979860, HCV amino acid variants and pre-treatment viral load. We show that the presence or absence of the IFN-λ4 protein has a pervasive impact on HCV, by describing multiple associations between host and pathogen variants in subgroups defined by viral genotype or human ancestry. We also present association analyses of human and viral variants with HCV viral load, which allows for a better understanding of the connections between genomic variation, biological mechanisms and clinical outcomes.

Results

Host and pathogen data

We obtained paired human and viral genetic data for 8,729 HCV-infected patients participating in various clinical trials of anti-HCV drugs. The samples were heterogeneous in terms of self-reported ancestry (85% Europeans, 13% Asians and 2% Africans) and HCV genotypes, with a majority of HCV genotype 1a, 2a and 3a (Table 1). We genotyped the human SNP rs12979860 and performed deep sequencing of the coding regions of the HCV non-structural proteins NS3, NS4A, NS5A and NS5B (Bartenschlager et al., 2004). A binary variable was generated for each alternate amino acid, indicating the presence or absence of that allele in a given sample (N = 10,681). For the analysis, we used only amino acids that were present in at least 0.3% of the samples (N = 4,022).

Table 1
Characteristics of study participants, by HCV genotype group.
https://doi.org/10.7554/eLife.42542.002
HCV genotypeAll1a1b2a2b3a4aOthers
N87293548 (41)1924 (22)304 (3)472 (5)1839 (21)193 (2)449 (5)
Europeans
Asians
Africans
Others
6704 (77)
1103 (13)
723 (8)
199 (2)
2987 (84)
59 (2)
421 (12)
81 (2)
1133 (59)
577 (30)
192 (10)
22 (1)
100 (33)
197 (65)
7 (2)
0 (0)
421 (89)
15 (3)
25 (5)
11 (2)
1635 (89)
111 (6)
19 (1)
74 (4)
178 (92)
2 (1)
8 (4)
5 (3)
250 (56)
142 (32)
51 (11)
6 (1)
Cirrhosis2410 (28)978 (28)536 (28)35 (12)77 (16)629 (34)60 (31)95 (21)
Male sex5605 (64)2434 (69)1096 (57)141 (46)301 (64)1230 (67)143 (74)260 (58)
SVR7702 (88)3240 (91)1773 (92)273 (90)426 (90)1452 (79)153 (79)385 (86)
  1. Data are indicated as number (percent); SVR: sustained virological response after treatment.

Associations between IFN-λ polymorphism and HCV amino acids

We performed a separate analysis for each HCV genotype, using an additive logistic model with binary amino acid variables as traits of interest. To control for population stratification, we added host and viral covariates in the model and to control for multiple testing we used a Bonferroni threshold of 4.7 × 10−6, which was calculated based on the number of tests performed (more information in the Materials and methods section). We restricted the analysis to genotypes 1a, 1b, 2a, 2b, 3a and 4a, which were present in at least 100 participants.

We observed highly significant associations between rs12979860 and HCV amino acid variables for each HCV genotype that we examined (Figure 1, Table 2). The highest number of significant associations was detected in the largest group of patients, infected with genotype 1a, most likely reflecting an effect of sample size on statistical power. Most associations were specific to a single viral genotype; however, some associations were significant across genotypes. As an example, two strong associations were observed between rs12979860 and amino acid variables at position 2576 in viral protein NS5B, with the T allele associating with proline in genotypes 1a (p=1.5×10−10), 2b (p=5.4×10−15), 3a (p=8.3×10−12) and 4a (p=1.2×10−7), and the C allele associating with alanine in genotypes 1a (p=1.2×10−11), 2a (p=3.8×10−6), 2b (p=4.02×10−8) and 3a (p=1.04×10−14).

Figure 1 with 2 supplements see all
Per genotype integrated association analysis results.

Manhattan plot for associations between human SNP rs12979860 and HCV amino acid variants. The dotted line shows the Bonferroni-corrected significance threshold.

https://doi.org/10.7554/eLife.42542.003
Table 2
Genome-to-genome analysis results per genotype.

The table shows significant p-values (<4.7×10−6), NA representing non-significant associations. We also give odds ratio (OR) and 97% confidence interval for each significant association.

https://doi.org/10.7554/eLife.42542.006
HCV genePosition
(amino acid)
Genotype 1a
N = 3548
Genotype 1b
N = 1924
Genotype 2a
N = 304
Genotype 2b
N = 472
Genotype 3a
N = 1839
Genotype 4a
N = 193
 NS31332(A)1.02e-10
(OR 1.06; 1.04–1.08)
NANANANANA
 NS31355(I)3.14e-07
(OR 1.1; 1.06–1.14)
NANANANANA
 NS31370(I)NA1.09e-08
(OR 0.83; 0.78–0.88)
NANANANA
 NS31370(T)NA4.87e-08
(OR 1.2; 1.12–1.28)
NANANANA
 NS31473(D)3.82e-07
(OR 1.03 1.02–1.04)
NANANANANA
 NS31516(I)3.51e-07
(OR 1.06; 1.04–1.09)
NANANANANA
 NS31598(R)2.26e-07
(OR 1.04; 1.02–1.05)
NANANANANA
 NS31612(I)7.88e-16
(OR 0.86; 0.83–0.89)
NANANANANA
 NS31612(N)1.54e-11
(OR 1.09; 1.06–1.11)
NANANANANA
 NS31612(T)1.54e-08
(OR 1.11; 1.07–1.15)
NANANANANA
 NS31635(I)7e-07
(OR 1.1; 1.06–1.14)
NANANANANA
 NS4A1671(T)1.83e-07
(OR 1.03; 1.02–1.04)
NANANANANA
 NS4A1703(R)NA6.94e-07
(OR 1.19; 1.11–1.27)
NANANANA
 NS5A1996(R)7.87e-07
(OR 1.01; 1.01–1.02)
NANANANANA
 NS5A2009(F)NA1.04e-08
(OR 1.11; 1.07–1.15)
NANANANA
 NS5A2009(I)2.01e-06
(OR 1.02; 1.01–1.02)
NANANANANA
 NS5A2024(V)5.81e-09
(OR 1.04; 1.03–1.05)
NANANANANA
 NS5A2034(D)1.75e-07
(OR 1.03; 1.02–1.04)
NANANANANA
 NS5A2034(T)NANANANA1.61e-07
(OR 0.91; 0.87–0.94)
NA
 NS5A2040(K)3.05e-06
(OR 0.98; 0.97–0.99)
NANANANANA
 NS5A2040(R)2.54e-07
(OR 1.03; 1.02–1.04)
NANANANANA
 NS5A2047(A)9.8e-20
(OR 1.07; 1.06–1.09)
NANANANANA
 NS5A2065(H)9.81e-07
(OR 1.01; 1.01–1.02)
1.38e-07
(OR 1.06; 1.04–1.09)
NANANANA
 NS5A2080(K)NA2.9e-18
(OR 1.12; 1.09–1.14)
NANANANA
 NS5A2080(R)NA1.39e-06
(OR 0.95; 0.93–0.97)
NANANANA
 NS5A2187(R)NA1.07e-06
(OR 1.07; 1.04–1.09)
NANANANA
 NS5A2211(L)2.84e-06
(OR 0.99; 0.98–0.99)
NANANANANA
 NS5A2220(R)NA2.65e-06
(OR 1.03; 1.02–1.04)
NANANANA
 NS5A2224(L)NA1.6e-12
(OR 1.05; 1.04–1.07)
NANANANA
 NS5A2234(W)NA1.46e-07
(OR 1.06; 1.03–1.08)
NANANANA
 NS5A2237(K)NA2.6e-12
(OR 1.06; 1.04–1.08)
NANANANA
 NS5A2251(I)NA2.05e-11
(OR 1.07; 1.05–1.09)
NANANANA
 NS5A2252(I)1.29e-25
(OR 1.12; 1.1–1.15)
NANANA8.68e-07
(OR 1.05; 1.03–1.07)
NA
 NS5A2252(V)1.72e-22
(OR 0.89; 0.87–0.91)
NANANA5.5e-07
(OR 0.95; 0.92–0.97)
NA
 NS5A2287(I)1.54e-14
(OR 1.09; 1.07–1.12)
6.24e-07
(OR 1.08; 1.05–1.11)
NANANANA
 NS5A2287(V)1.82e-10
(OR 0.92; 0.90–0.95)
NANANANANA
 NS5A2298(I)1.56e-06
(OR 1.05; 1.03–1.08)
NANANANANA
 NS5A2298(V)1.66e-14
(OR 0.92; 0.90–0.94)
NANANANANA
 NS5A2300(P)NA2.7e-15
(OR 1.12; 1.09–1.15)
NANANANA
 NS5A2300(S)NA9.41e-08
(OR 0.94; 0.91–0.96)
NANANANA
 NS5A2320(Q)5.01e-09
(OR 1.08; 1.05–1.11)
NANANANANA
 NS5A2330(R)NA1.26e-06
(OR 1.03; 1.02–1.04)
NANANANA
 NS5A2360(A)NA1.46e-12
(OR 1.12; 1.09–1.16)
NANANANA
 NS5A2371(S)2.03e-07
(OR 1.03; 1.02–1.04)
NANANANANA
 NS5A2372(A)2.44e-06
(OR 0.96; 0.94–0.97)
NANANANANA
 NS5A2372(S)1.63e-14
(OR 1.06; 1.04–1.07)
NANANANANA
 NS5A2385(C)3.24e-14
(OR 1.09; 1.07–1.11)
4.35e-07
(OR 1.04; 1.03–1.06)
NANANANA
 NS5A2385(Y)2.7e-13
(OR 0.93; 0.91–0.94)
NANANANANA
 NS5A2411(G)NA4.61e-08
(OR 1.11; 1.07–1.15)
NANANANA
 NS5A2411(S)NA9.02e-07
(OR 0.92; 0.89–0.95)
NANANANA
 NS5A2412(K)5.74e-09
(OR 1.03; 1.02–1.05)
NANANANANA
 NS5A2412(T)7.87e-10
(OR 0.93; 0.91–0.95)
NANANANANA
 NS5A2414(D)2.43e-07
(OR 0.97; 0.96–0.98)
NANANANANA
 NS5A2416(G)NANANANA5.21e-07
(OR 1.06; 1.04–1.09)
NA
 NS5A2416(N)NANANANA2.5e-07
(OR 1.09; 1.05–1.12)
NA
 NS5A2416(S)NANANANA1.04e-11
(OR 0.89; 0.86–0.92)
NA
 NS5A2420(N)NANANANA3.39e-09
(OR 1.08; 1.05–1.11)
NA
 NS5A2420(S)NANANANA7.1e-07
(OR 0.95; 0.93–0.97)
NA
 NS5B2510(N)2.25e-06
(OR 1.02; 1.01–1.03)
NANANANANA
 NS5B2567(I)1.73e-13
(OR 1.02; 1.02–1.03)
5.73e-08
(OR 1.07; 1.04–1.09)
NANANANA
 NS5B2570(A)NANANANA2.63e-07
(OR 1.11; 1.06–1.15)
NA
 NS5B2570(T)NANANANA8.87e-15
(OR 1.11; 1.08–1.14)
NA
 NS5B2570(V)NANANANA5.57e-20
(OR 0.84; 0.81–0.87)
NA
 NS5B2576(A)1.21e-11
(OR 1.02; 1.01–1.02)
NA3.84e-06
(OR 1.27; 1.15–1.4)
4.02e-08
(OR 1.2; 1.13–1.28)
1.04e-14
OR 1.07; 1.05–1.08)
NA
 NS5B2576(P)1.53e-10
(OR 0.98; 0.98–0.99)
NANA5.41e-15
(OR 0.77; 0.72–0.82)
8.39e-12
(OR 0.95; 0.94–0.96)
1.13e-07
(OR 0.83; 0.77–0.88)
 NS5B2633(S)NA2.33e-09
(OR 1.08; 1.06–1.11)
NANANANA
 NS5B2729(Q)1.19e-12
(OR 0.91; 0.89–0.94)
1.38e-07
(OR 0.94; 0.92–0.96)
NANANANA
 NS5B2729(R)9.13e-12
(OR 1.09; 1.06–1.12)
2.22e-09
(OR 1.08; 1.05–1.11)
NANANANA
 NS5B2755(N)2.98e-06
(OR 1.04; 1.02–1.06)
NANANANANA
 NS5B2758(A)NA2.3e-06
(OR 1.05; 1.03–1.07)
NANANANA
 NS5B2794(Q)NANANANA3.56e-10
(OR 1.08; 1.05–1.1)
NA
 NS5B2860(G)NA4.63e-12
(OR 1.07; 1.05–1.09)
NANANANA
 NS5B2937(K)8.23e-07
(OR 0.95; 0.93–0.97)
NANANANANA
 NS5B2937(R)NANANANA4.4e-08
(OR 1.08; 1.05–1.11)
NA
 NS5B2986(H)NANANANA1.03e-06
(OR 0.95; 0.93–0.97)
NA
 NS5B2986(R)NANANANA2.9e-07
(OR 1.05; 1.03–1.07)
NA
 NS5B2991(H)NANANANA4.66e-12
(OR 0.88; 0.85–0.91)
NA
 NS5B2991(Y)NANANANA1.86e-17
(OR 1.17; 1.13–1.22)
NA
 NS5B3008(F)7.47e-08
(OR 1.01; 1.01–1.02)
NANANANANA

In patients infected with genotype 3a, we replicated the previously reported associations (Ansari et al., 2017) between IFNL4 variation and valine at position 2570 in NS5B (p=5.5×10−20), histidine at position 2991 in NS5B (p=4.6×10−12) and asparagine at position 2414 in NS5A (p=2.4×10−7). We also observed novel associations with alanine (p=2.6×10−7) and threonine (p=8.8×10−15) at position 2570 in NS5B, as well as with glycine (p=5.2×10−7) and serine (p=1.04×10−11) at position 2414 in NS5A. All these associations were only detected in the 3a subgroup. In concordance with a previous study (Peiffer et al., 2016), we also observed a significant association with histidine at position 2065 of NS5A in patients infected with HCV genotypes 1a (p=9.8×10−7) and 1b (p=1.3×10−7).

We also observed multiple significant associations in the interferon-sensitivity determining region (ISDR, amino acid positions 2209 to 2248 in NS5A) in patients infected with genotype 1b, the strongest one being with the presence of leucine at position 2224 (p=1.5×10−12). For genotype 1a, we observed a single significant association in the ISDR region with the presence of leucine at position 2211 (p=2.8×10−6).

To check whether the association of IFNL4 genotype with HCV amino acid variables could be dependent of the effect of IFNL4 genotype on viral replication rates, we also compared the results from two sets of logistic regression models: one that does and one that does not include HCV viral load as an additional covariate. We did not observe any significant difference in the results of the two models (Figure 1—figure supplement 1).

Viral load association analyses

To further understand the clinical implications of viral mutations associated with IFN-λ polymorphism, we searched for associations between rs12979860, HCV amino acid variants and viral load. For this, we first searched for associations between rs12979860 and Box-Cox transformed pre-treatment HCV viral load, in subgroups defined by HCV genotypes. Pre-treatment viral load was found to be significantly associated (p<0.05) with rs12979860 for all HCV genotypes, with the rs12979860 T allele consistently associated with lower viral load (Figure 1—figure supplement 2). The strength of the association p-values varied between genotypes due to sample size, but the effect size associated with the T allele was comparable across genotype groups.

We then searched for associations between viral load and HCV amino acid variables. These analyses identified significant associations in all viral genotype groups except 4a (Figure 2). Amongst the viral amino acids that associated with viral load, a number also associated with rs12979860 genotype (genotype 1a, 9 of 18 amino acids; 1b, 5 of 17 amino acids; 2a, 0 of 2 amino acids; 2b, 0 of 6 amino acids; 3a, 2 of 3 amino acids). As an example of such a complex association pattern, we looked at position 2224 of NS5A (in the ISDR) in genotype 1b. Mean viral load was higher in patients infected with a virus harboring a leucine in comparison to the most common amino acid alanine (t-test p-value: 5.6 x10−9, with Halternative =μvlL μvlA>0) (Figure 3A). This was true for both CC and non CC genotypes of SNP rs12979860 (t-test p-value: 6.2 x10−6 for CC,L vs. CC,non-L; t-test p-value: 4.1 x10−2 for CT,L vs. CT,non-L), indicating a possible impact of that leucine residue on viral replication (Figure 3B).

Per genotype viral load GWAS analysis results.

Manhattan plot for associations between human Box-Cox transformed pre-treatment viral load and HCV amino acid variants. The dotted line shows the Bonferroni-corrected significance threshold.

https://doi.org/10.7554/eLife.42542.007
Figure 3 with 5 supplements see all
Associations between amino acid variables at position 2224 of NS5A, rs12979860 genotypes and HCV viral load in the group of patients infected with HCV genotype 1b.

(A) Boxplot of transformed viral load stratified by amino acids present at position 2224 of NS5A. (B): Boxplot of transformed viral load stratified by rs12979860 genotypes (CC, CT, TT) and by presence or absence of leucine at position 2224 of NS5A.

https://doi.org/10.7554/eLife.42542.008

We also replicated the previously shown (Ansari et al., 2017) association between viral load and the change from a serine to an asparagine at position 2414 in NS5A protein (p=4.5×10−7) in genotype 3a and observed a lower mean viral load for patients with non-CC genotype and presence of serine at position 2414 (Figure 3—figure supplement 1).

To further understand these associations, we performed a residual regression analysis. We searched for associations between the amino acid variables and viral load residuals, obtained after regressing the transformed viral load on rs12979860. The objective of this analysis was to identify amino acids associated with changes in viral load that cannot be entirely explained by rs12979860 genotype. We observed multiple significantly associated amino acids with residual viral load across genotypes (Figure 3—figure supplement 2). A total of 7 amino acids in genotype 1a (supplementary file 1) and six amino acids in genotype 1b (supplementary file 2) associated with rs12979860 genotype, viral load and viral load residuals, including again leucine at position 2224 of NS5A in genotype 1b (presidual = 4.9×10−8).

Ancestry-specific sub-analyses

We also ran association analyses between IFN-λ variations and the variations in the HCV genome in subgroups defined by self-reported ancestry: European, Asian, and African. The association results are broadly similar to per genotype analysis and are presented in supplementary file 3.

We further dissected the association signals within the largest ancestry group, Europeans, by running a per genotype analysis within this sample (Figure 3—figure supplement 3). The strongest association was observed with the presence of isoleucine at position 2252 of viral protein NS5A in patients infected with HCV genotype 1a (p=1.2×10−24). All the significant results from this study are presented in supplementary file 4.

Results of the ancestry-specific sub-analyses of associations with HCV viral load are comparable to the results obtained in the whole study population and are presented in Figure 3—figure supplement 4, Figure 3—figure supplement 5 and supplementary file 5.

Discussion

We used an integrated association analysis approach to explore the impact of human genetic variation in the IFN-λ region on part of the HCV proteome during chronic infection. Our results reveal a strong footprint of innate immune pressure on the non-structural regions of the HCV genome and provide strong evidence for pervasive HCV adaptation to innate immunity. We performed analyses in different sub-groups, which showed an impact of IFNL4 variation on HCV across genotypes and ancestry categories. Finally, we report viral amino acids significantly associated with both IFNL4 variation and HCV viral load, indicating that some of the HCV clinical and biological outcomes could be explained by traceable host–pathogen interactions.

Because we genotyped the human SNP rs12979860, a reliable marker for the dinucleotide insertion/deletion polymorphism rs368234815, our analyses exclusively focus on the effects of the presence or absence of the IFN-λ4 protein on HCV amino acids and viral load. Therefore, one clear limitation of our study is the impossibility to distinguish between the two haplotypes encoding the IFN-λ4 P70 and S70 isoforms, which have been shown to have distinctive influences on HCV pathogenesis (Ansari, 2018).

Our analysis detected multiple associations in all tested proteins, including NS5A. This protein is required for HCV RNA replication and virus assembly and has been shown to associate with interferon signaling and hepatocarcinogenesis (Nakamoto et al., 2014). Previous studies have also shown strong associations between variants in the ISDR of NS5A and HCV viral load as well as response to IFN-based therapy (Enomoto et al., 1995; Frangeul et al., 1998). Some of the strongest associations that we observed were in and around this highly variable region, suggesting a possible role of these variants in determining the response to IFN-based antiviral treatment. The strongest association in the ISDR was with leucine at position 2224 in patients infected with 1b genotype, with higher mean viral load observed in presence of leucine for patients with the rs12979860 CC genotype. We also confirmed previously reported findings in the region, including associations with histidine at position 206518 (also known as the NS5A Y93H variant) and with asparagine at position 241411. Using a genotype three replicon assay, Ansari et al. showed that this later variant - a change from a serine to asparagine at site 2414 - is associated with an increase in RNA replication, which is concordant with our results.

This is the first comprehensive analysis of IFN-λ-driven HCV adaptation across different viral genotypes and ancestry groups. In addition to identifying genotype or ancestry-specific associations, we observed sites of interaction that were consistent across HCV genotypes and ethnicities; for example, the NS5A variant Y2065H, which was found to be associated with rs12979860 in individuals infected with HCV genotypes 1a and 1b. These results indicate that IFN-λ-driven viral adaptation is a part of evolution across HCV genotypes.

In an attempt to delineate the biological impact of these associations, we evaluated the associations between HCV amino acid variants and pre-treatment viral load. We were able to detect a subset of amino acids that associated with both IFN-λ variation and HCV viral load across different viral genotypes, supporting the clinical relevance of host and pathogen interactions. Furthermore, we also performed a similar analysis with residual viral load, that is the fraction of the viral load variance that that is not explained by IFN-λ variation. We detected a group of viral amino acid variants that associated with SNP variations as well as residual viral load, indicating a stronger role of host–pathogen interactions in explaining the variations in HCV viral load.

Interestingly, only a fraction of the host-driven HCV amino acid variants was found to be associated with viral load, indicating that an integrated association analysis between host and pathogen genome variations can reveal correlations that would go unnoticed in association studies that use more downstream laboratory measurements or clinical outcomes as phenotypes.

IFN-λ polymorphism is the strongest human genetic predictor of spontaneous HCV clearance and response to IFN-based therapy. By integrating IFN-λ and HCV amino acid variation in a joint analysis, we here contribute to a better understanding of the genomic mechanisms involved in inter-individual differences in HCV disease outcomes. Our results confirm that IFN-λ4 is a functional gene that plays a pivotal role in HCV pathogenesis. The large footprint left by IFNL4 variation on the HCV proteome is indeed a clear indicator of the importance of innate immunity in viral control and of the remarkable capacity of HCV to evolve escape strategies.

Materials and methods

Clinical samples

Request a detailed protocol

Across 82 studies involving >100 sites in many countries, appropriate informed consent was obtained from study participants allowing the current analysis to be performed (Welzel et al., 2017). The studies were run by Gilead Sciences (Foster City, CA) and Pharmasset (formerly Princeton, NJ). Study protocols followed the ethical guidelines set in place by the 1975 Declaration of Helsinki and were approved by the relevant institutional review board committees. All samples included in this analysis are baseline samples collected from treatment naive and experienced patients from >25 countries in North America, Europe, Asia, Oceania, and Africa between years 2010 and 2015.

NS3, NS5A, and NS5B sequencing

Request a detailed protocol

The genotype assignment from Siemens VERSANT HCV Genotype INNO-LiPA 2.0 Assay (Innogenetics, Ghent, Belgium) was used to select genotype-specific primers located outside of the gene target(s) that amplify the entire NS3/4A, NS5A, or NS5B regions of HCV. Standard reverse transcription polymerase chain reaction (RT-PCR) was performed on patient plasma with HCV RNA >1000 IU/mL at DDL Diagnostic Laboratory (Rijswijk, The Netherlands). For deep sequencing, amplicons encoding the subject-derived NS3/4A, NS5A and NS5B were run using Illumina MiSeq v2 150 paired-end deep sequencing at DDL or WuXi AppTec (Shanghai, China). FASTQ files were split based on 100% matched barcodes. Contigs were generated from paired-end FASTQ files using VICUNA (Yang et al., 2012) and merged to create a de novo assembly sequence. All paired-end reads were merged using PEAR (Zhang et al., 2014), chopped at the 3’ end when MAPQ <15, and filtered to remove reads <50 bases. The filtered reads were aligned to the de novo assembly sequence using MOSAIK (Lee et al., 2014) (v1.1.0017) to create a final assembly sequence. The average coverage of >5000 reads per position was obtained for most of the samples. The aligned reads were translated in-frame and the resulting tabulated summary of variants from the final assembly was utilized to generate a consensus sequence. Mixtures were reported when present in ≥15% of the viral population. NS3/4A, NS5A and NS5B consensus nucleotide and amino acid sequences were compared by the NCBI alignment tool BLAST to a set of reference sequences to assign HCV genotype and subtype. Amino acid variation between the samples that were assigned to genotype 1a, 1b, 2a, 2b, 3a and 4a were tabulated and analyzed. The raw HCV sequences are available in the zenodo repository, https://doi.org/10.5281/zenodo.1476713.

Host genotyping

Request a detailed protocol

Human genotype was determined by PCR amplification and sequencing of the rs12979860 SNP region. Possible genotypes were CC, CT or TT.

Association analyses

Request a detailed protocol

To run the integrated association analysis between genotyped host SNP and viral amino acids, we used logistic regression where the traits of interest were the presence or absence of each amino acid at the variable sites of the virus proteome. We assumed an additive model and corrected for host population stratification by adding sex, country of origin, self-reported ethnicity, cirrhosis status and prior treatment experience as covariates. To account for residual viral stratification within each HCV genotype, the first five phylogenetic principal components (Revell, 2009), calculated per HCV gene to account for recombination, were also added as covariates.

For the viral load GWAS analysis, we used linear regression where the trait of interest was Box-Cox transformed pre-treatment viral load. We used Box-Cox transformation to transform the positively skewed viral load distribution into a normally distributed dependent variable. We corrected for host and viral population stratification by adding sex, country of origin, self-reported ethnicity, cirrhosis status and prior treatment experience, as well as the first five viral phylogenetic principal components as covariates.

To correct for multiple testing we calculated the Bonferroni threshold as 0.05nA, where nA represents the number of tests performed. For the analyses described in the paper, we performed a total of 10,681 tests. Given the heterogeneity of the dataset with multiple genotypes and ethnicities, we performed the integrated association analysis as well as viral load GWAS analyses on different sample subsets, created per genotype as well as per ethnic group.

Software used

Request a detailed protocol

We used muscle (Edgar, 2004) to align the pathogen sequences, RaXML (Stamatakis, 2014) to obtain the phylogenetic trees and R (R Development Core Team, 2013) for all other analyses.

References

  1. 1
  2. 2
  3. 3
    Evidence for a widespread effect of interferon lambda 4 on hepatitis C virus diversity
    1. AM Ansari
    (2018)
    Journal of Pharmaceutical Sciences & Emerging Drugs, 10.4172/2380-9477-C4-015.
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
    R: A language and environment for statistical computing
    1. R Development Core Team
    (2013)
    R Foundation for Statistical Computing, Vienna, Austria; URL, Vienna, Austria.
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
    Global Hepatitis Report, 2017
    1. WHO
    (2017)
    World Health Organization.
  29. 29
  30. 30

Decision letter

  1. Wendy S Garrett
    Senior Editor; Harvard TH Chan School of Public Health, United States
  2. Thomas O'Brien
    Reviewing Editor; NIH, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Pervasive adaptation of hepatitis C virus to interferon lambda polymorphism across multiple genotypes" for consideration by eLife. Your article has been reviewed by four peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Wenhui Li as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

Two papers by Ansari et al. are highly relevant to this submission. Previously, Ansari et al. reported results of a genome-to-genome study of 542 individuals who were chronically infected with HCV viral genotype (VGT) 3 (predominantly) or VGT 2. (Nature Genetics, 2017) In a joint submission with the present paper, Ansari et al. now report analyses restricted to the subset of those subjects who were infected with HCV genotype 3a. Results from both of those papers should be referenced in the present paper, as appropriate.

The present paper examines associations between genotype for IFNL4 rs12979860 (a marker for rs368234815) and variation in four HCV proteins (NS3, NS4A, NS5A and NS5B) among 8,729 HCV-infected patients enrolled in clinical trials conducted by Gilead Sciences. These individuals were infected with a range of VGTs (1a – 3,548; 1b – 1,924; 2a- 304; 2b- 472; 3a- 1,839; 4a-193). Thus, the authors have data to replicate findings of Ansari et al. regarding VGT3, examine associations with other VGTs and compare viral associations with IFNL4 genotype across VGTs. Given the richness of this resource, the analysis is disappointing; the paper needs substantial additional analyses and a new set of figures. In addition, the study population and the statistical methods must be described more clearly. With significant revisions, this paper could make an important contribution to the HCV literature.

Essential revisions:

The lack of line numbers or page numbers makes it difficult to convey comments. Please add line numbers (or at least page numbers) to the manuscript.

Study population and design:

Given that baseline samples are analyzed in relation to baseline outcome measures, this is a cross sectional study rather than a cohort analysis. The terminology should be clarified, as otherwise it is misleading.

The authors only genotype the SNP rs12979860 and, therefore, cannot distinguish between the two haplotypes encoding the IFNL4 P70 and IFNL4 S70 proteins per the co-submission. This limitation should be acknowledged in the Discussion.

Nomenclature:

This paper uses protein-specific numbering of amino acid positions, whereas Ansari et al. employ virome wide numbering. For the sake of comparison, the present paper should include virome wide numbering.

"a genetic polymorphism in the interferon region" is ambiguous as there are multiple polymorphisms in this region. "rs12979860 genotype, a tag for IFN-λ haplotypes" or "rs12979860, a known marker of IFN λ haplotypes" is also ambiguous and not relevant as haplotypes are not explained or used in this paper. The most relevant point re: rs12979860 that it is a marker for rs368234815.

Statistical analyses and results:

Consensus viral sequences and the propensity for specific viral variants vary markedly by VGT, therefore, associations between IFNL4 genotype and viral variants vary by VGT. This is demonstrated in Supplementary Table 3, which provides key data. A revised version of this table (or a derived figure) should be included in the body of the paper. The organizing principle for presentation of the HCV variants in Supplementary Table 3 is unclear. The table should be simplified by restricting it to results of either additive or recessive genetic models, based on an evaluation of which model is more appropriate.

Given these observed differences in viral associations in different VGTs, it is inappropriate to combine all 8,729 patients in analyses. These subjects comprise a convenience sample of patients enrolled in the clinical trials, which is not a population of intrinsic biological interest. (A different set of trials amongst patients with a different distribution of VGTs would yield different summary data.) Figures 1, 2B and 3 should be deleted; Figures 1 and 3 should be replaced by sets of VGT-specific figures.

Table 1 provides little information. The paper should include a table that describes the characteristics of the subjects (e.g., sex, age, race, country of enrollment and HCV disease status), by VGT.

The presentation of results of the statistical analysis is difficult to follow. The authors should specify the associations. For example, the presence of a functional IFNL4 variant associates with a proline at position XXX in viral protein Y. The data presented on the correlation of specific HCV amino acids is interesting but equally underdeveloped.

The authors largely present results as p-values, however, to evaluate the effect size, the amplitude of the signals should be presented as well.

The data describing HCV viral load should be analyzed in more detail and presented more clearly. It should be possible for the reader to determine if a given amino acid change associates with higher or lower viral load, how higher or lower viral load associates with the three major IFNL genotypes, and the effect of viral load of being heterozygous.

Describe the effect of being heterozygous more clearly to the reader, often two correlation coefficients are listed (recessive and additive) but without explanation and only the p-value is listed. The viral load data should be analyzed more carefully, including using unadjusted and adjusted regression models, to inform the reader about what the effect of this particular genotype is on viral load.

In association of host and viral genotype, how was the Bonferroni threshold determined?

In the logistic models used to assess associations, what was the outcome? Host or viral genotype? Were the models adjusted for any covariates? Please specify.

Were models considered that contained more than one viral amino acid, i.e. were viral results mutually adjusted to better pinpoint where the signal is coming from?

The association of IFNL4 genotype with the frequency of HCV polymorphisms could reflect an effect of IFNL4 on viral replication rates. To assess that possibility, the investigators should compare the results of two logistic regression models: one that does and one that does not include HCV RNA as an additional covariate to IFNL4. Otherwise these paired models should include identical adjustments.

Viral load is a quantitative trait and it is important that proper methods be used in those analyses. Residual regression analysis (subsection “HCV amino acid variants and viral load”): what are the assumptions made here, e.g. normal distribution of viral load? Was viral load transformed?

Supplementary Table 6 – There is no mention of VL GWAS in the main text or Materials and methods. Please explain this approach in Materials and methods and table footnotes; comment on differences in p-values between VL GWAS and G2G analysis.

Discussion:

The results should be discussed in the context of current knowledge from the literature regarding the function of rs368234815. HCV does not adapt to a genetic polymorphism, but to its functional effect, which in this case is likely the production of IFN-λ4.

Other:

Throughout the paper, clearly state what is a new finding in this paper and what confirms or contradicts the two papers by Ansari et al.

Supplementary tables should include footnotes explaining the analyses with regard to the statistical model employed, adjustments and significance thresholds. Also, the number of subjects in each subgroup.

Re: title, suggest: Adaptation of hepatitis C virus to interferon lambda polymorphism across multiple viral genotypes.

Additional data files:

Current data sharing covers only HCV variation, not the human rs12979860 genotypes and relevant demographic and clinical variables. It would be important to make this unique dataset available to the scientific community through controlled dbGAP access. Otherwise, this analysis cannot be validated.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Adaptation of hepatitis C virus to interferon lambda polymorphism across multiple viral genotypes" for further consideration at eLife. Your revised article has been favorably evaluated by Wendy Garrett as the Senior Editor and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

Summary:

The authors were very responsive to the reviewer comments. The revised paper, which includes substantial additional analyses and a new set of figures, is much improved. The new analyses demonstrate that most of the associations between IFNL4 genotype and viral variants are viral genotype (VGT) specific, as would be expected given that consensus viral sequences and the propensity for specific viral variants vary markedly by VGT. This paper will be an important contribution to the HCV literature.

Essential revisions:

Previous comments requiring further attention.

1) In association of host and viral genotype, how was the Bonferroni threshold determined?

More details on the regression model as well as significance threshold are given in the Materials and methods section (subsection “Association analyses”).

Unfortunately, that information is not enough. Specify the number of tests performed. If that number varies meaningfully for the different analyses, then the number should be specified each time the Bonferroni correction is used.

2) Viral load is a quantitative trait and it is important that proper methods be used in those analyses. Residual regression analysis (subsection “HCV amino acid variants and viral load”): what are the assumptions made here, e.g. normal distribution of viral load? Was viral load transformed?

We did use Box-Cox transformation for viral load and then used the viral load residuals obtained from transformed viral load. We have added more information on this is in the Materials and methods section (subsection “Association analyses”).

The description in the subsection “Association analyses” states "we used logistic regression where the trait of interest was Box-Cox transformed pre-treatment viral load." However, logistic regression outcomes are usually binary variables. This statement,"Viral load GWAS analysis was performed using linear regression, between transformed viral load as trait of interest and viral amino acid variations." Please clarify what was done.

New comments arising from the extensive revisions.

1) The Abstract should be more specific. Describe the distribution of viral genotypes examined and some key findings.

2) The term 'genome-to-genome' seems inappropriate for these analyses. 'Genome' implies a complete set of genetic information; however, these analyses involve a single human variant and variants from a selected set of HCV proteins.

3) Subsection “Viral load association analyses”, second paragraph – I found this text confusing. Suggest the following revision (if correct): 'We then searched for associations between viral load and HCV amino acid variables. These analyses identified significant associations in all viral genotype groups except 4a (Figure 2). Amongst the viral amino acids that associated with viral load, a number also associated with rs12979860 genotype (genotype 1a, 9 of x amino acids; 1b, 5 of x amino acids; 2a, 0 of x amino acids; 2b, 0 of x amino acids; 3a, 2 of x amino acids).'

4) Subsection “Viral load association analyses”, third paragraph – Please clarify if this analysis is restricted to individuals infected with genotype 1b. Please verify that Figure 3B is consistent with this text and that the y axis is not mislabeled.

5) Table 2 presents '97% CI', rather than 95% CI, which are more usual. If 97% is in fact correct, provide a justification for using an interval of that width.

https://doi.org/10.7554/eLife.42542.023

Author response

Essential revisions:

The lack of line numbers or page numbers makes it difficult to convey comments. Please add line numbers (or at least page numbers) to the manuscript.

We have added the page and line numbers in the revised version.

Study population and design:

Given that baseline samples are analyzed in relation to baseline outcome measures, this is a cross sectional study rather than a cohort analysis. The terminology should be clarified, as otherwise it is misleading.

We have changed the terminology in the revised version, as suggested by the reviewer. Changes are in the last paragraph of the Introduction.

The authors only genotype the SNP rs12979860 and, therefore, cannot distinguish between the two haplotypes encoding the IFNL4 P70 and IFNL4 S70 proteins per the co-submission. This limitation should be acknowledged in the Discussion.

We have added a part about this limitation in the Discussion (second paragraph).

Nomenclature:

This paper uses protein-specific numbering of amino acid positions, whereas Ansari et al. employ virome wide numbering. For the sake of comparison, the present paper should include virome wide numbering.

We have changed the protein specific numbering of amino acids to virome wide numbering as suggested by the reviewer.

"a genetic polymorphism in the interferon region" is ambiguous as there are multiple polymorphisms in this region. "rs12979860 genotype, a tag for IFN-λ haplotypes" or "rs12979860, a known marker of IFN λ haplotypes" is also ambiguous and not relevant as haplotypes are not explained or used in this paper. The most relevant point re: rs12979860 that it is a marker for rs368234815.

We have incorporated the suggested changes in the paper and removed the mentions of rs12979860 genotype being a tag for IFN-λ haplotypes.

Statistical analyses and results:

Consensus viral sequences and the propensity for specific viral variants vary markedly by VGT, therefore, associations between IFNL4 genotype and viral variants vary by VGT. This is demonstrated in Supplementary Table 3, which provides key data. A revised version of this table (or a derived figure) should be included in the body of the paper. The organizing principle for presentation of the HCV variants in Supplementary Table 3 is unclear. The table should be simplified by restricting it to results of either additive or recessive genetic models, based on an evaluation of which model is more appropriate.

We have moved the supplementary table in the main paper (Table 2) and added more details on VGT specific analysis in the Results section. Table 2 (previously Supplementary Table 3) now contains association results from the additive model only.

Given these observed differences in viral associations in different VGTs, it is inappropriate to combine all 8,729 patients in analyses. These subjects comprise a convenience sample of patients enrolled in the clinical trials, which is not a population of intrinsic biological interest. (A different set of trials amongst patients with a different distribution of VGTs would yield different summary data.) Figures 1, 2B and 3 should be deleted; Figures 1 and 3 should be replaced by sets of VGT-specific figures.

We have removed the results from the global analysis of our study population (including all figures and tables). The Results section now focuses on VGT specific analysis. Ancestry specific sub-analyses are mentioned in the Results section.

Table 1 provides little information. The paper should include a table that describes the characteristics of the subjects (e.g., sex, age, race, country of enrollment and HCV disease status), by VGT.

We have added more information in Table 1, per VGT, including ethnicity, sex, cirrhosis status and SVR status.

The presentation of results of the statistical analysis is difficult to follow. The authors should specify the associations. For example, the presence of a functional IFNL4 variant associates with a proline at position XXX in viral protein Y. The data presented on the correlation of specific HCV amino acids is interesting but equally underdeveloped.

We have changed the presentation of the results and incorporated the suggested modifications. We have also performed additional analyses to dissect specific association results and added the results to the revised version.

The authors largely present results as p-values, however, to evaluate the effect size, the amplitude of the signals should be presented as well.

We have added odds ratios and confidence intervals to the p-values presented in Table 2 (main results) as well as in the supplementary files that present analysis results.

The data describing HCV viral load should be analyzed in more detail and presented more clearly. It should be possible for the reader to determine if a given amino acid change associates with higher or lower viral load, how higher or lower viral load associates with the three major IFNL genotypes, and the effect of viral load of being heterozygous.

We have added more analysis results showing the correlation between HCV amino acid changes and viral load as well as host genotype and viral load (subsection “Viral load association analyses”).

Describe the effect of being heterozygous more clearly to the reader, often two correlation coefficients are listed (recessive and additive) but without explanation and only the p-value is listed. The viral load data should be analyzed more carefully, including using unadjusted and adjusted regression models, to inform the reader about what the effect of this particular genotype is on viral load.

We have now removed the results from the recessive analysis from the paper and provided more details on additive regression model in the Materials and methods section (subsection “Association analyses”).

In association of host and viral genotype, how was the Bonferroni threshold determined?

More details on the regression model as well as significance threshold are given in the Materials and methods section (subsection “Association analyses”).

In the logistic models used to assess associations, what was the outcome? Host or viral genotype? Were the models adjusted for any covariates? Please specify.

For the genome-to-genome analysis, we used a logistic regression based association model, where the binary HCV amino acids were the outcomes. We added multiple host as well as pathogen population covariates to account for population stratification and thus reduce the risk of obtaining false positives association results. We have added more details on the regression analyses in the Materials and methods section (subsection “Association analyses”).

The association of IFNL4 genotype with the frequency of HCV polymorphisms could reflect an effect of IFNL4 on viral replication rates. To assess that possibility, the investigators should compare the results of two logistic regression models: one that does and one that does not include HCV RNA as an additional covariate to IFNL4. Otherwise these paired models should include identical adjustments.

As suggested by the reviewer, we ran two series of logistic regression models, one with HCV RNA as an additional covariate and one without. The association results from the two models were very similar. We have added the results of these analyses in the last paragraph of the subsection “Genome-to-genome analyses” (Figure 1—figure supplement 1).

Viral load is a quantitative trait and it is important that proper methods be used in those analyses. Residual regression analysis (subsection “HCV amino acid variants and viral load”): what are the assumptions made here, e.g. normal distribution of viral load? Was viral load transformed?

We did use Box-Cox transformation for viral load and then used the viral load residuals obtained from transformed viral load. We have added more information on this is in the Materials and methods section (subsection “Association analyses”).

Supplementary Table 6 – There is no mention of VL GWAS in the main text or Materials and methods. Please explain this approach in Materials and methods and table footnotes; comment on differences in p-values between VL GWAS and G2G analysis.

We have added more information on the viral load GWAS (VL GWAS) in the Materials and methods subsection “Association analyses”.

Discussion:

The results should be discussed in the context of current knowledge from the literature regarding the function of rs368234815. HCV does not adapt to a genetic polymorphism, but to its functional effect, which in this case is likely the production of IFN-λ4.

We have modified the Introduction and the Discussion accordingly.

Other:

Throughout the paper, clearly state what is a new finding in this paper and what confirms or contradicts the two papers by Ansari et al.

We have added more comparisons between the results from our analysis and the results from Ansari et al.’s papers (subsection “Genome-to-genome analyses”, third paragraph and subsection “Viral load association analyses”, fourth paragraph), including replications as well as new results from our analysis that were not detected in the analysis by Ansari et al.

Supplementary tables should include footnotes explaining the analyses with regard to the statistical model employed, adjustments and significance thresholds. Also, the number of subjects in each subgroup.

We added this information in the supplementary files.

Re: title, suggest: Adaptation of hepatitis C virus to interferon lambda polymorphism across multiple viral genotypes.

We have changed the title to the one suggested by the reviewer.

Additional data files:

Current data sharing covers only HCV variation, not the human rs12979860 genotypes and relevant demographic and clinical variables. It would be important to make this unique dataset available to the scientific community through controlled dbGAP access. Otherwise, this analysis cannot be validated.

Unfortunately, we are not able to provide individual-level demographic and clinical data. The consent form signed by study participants does not allow for the sharing of that information.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

Summary:

The authors were very responsive to the reviewer comments. The revised paper, which includes substantial additional analyses and a new set of figures, is much improved. The new analyses demonstrate that most of the associations between IFNL4 genotype and viral variants are viral genotype (VGT) specific, as would be expected given that consensus viral sequences and the propensity for specific viral variants vary markedly by VGT. This paper will be an important contribution to the HCV literature.

Essential revisions:

Previous comments requiring further attention.

1) In association of host and viral genotype, how was the Bonferroni threshold determined?

More details on the regression model as well as significance threshold are given in the Materials and methods section (subsection “Association analyses”).

Unfortunately, that information is not enough. Specify the number of tests performed. If that number varies meaningfully for the different analyses, then the number should be specified each time the Bonferroni correction is used.

We have now added the number of tests performed in the Materials and methods section (subsection “Association analyses”, last paragraph).

2) Viral load is a quantitative trait and it is important that proper methods be used in those analyses. Residual regression analysis (subsection “HCV amino acid variants and viral load”): what are the assumptions made here, e.g. normal distribution of viral load? Was viral load transformed?

We did use Box-Cox transformation for viral load and then used the viral load residuals obtained from transformed viral load. We have added more information on this is in the Materials and methods section (subsection “Association analyses”).

The description in the subsection “Association analyses” states "we used logistic regression where the trait of interest was Box-Cox transformed pre-treatment viral load." However, logistic regression outcomes are usually binary variables. This statement,"Viral load GWAS analysis was performed using linear regression, between transformed viral load as trait of interest and viral amino acid variations." Please clarify what was done.

We have fixed the mistake in the second paragraph of the subsection “Association analyses”, and replaced logistic regression with linear regression.

New comments arising from the extensive revisions.

1) The Abstract should be more specific. Describe the distribution of viral genotypes examined and some key findings.

We have added more information in the Abstract.

2) The term 'genome-to-genome' seems inappropriate for these analyses. 'Genome' implies a complete set of genetic information; however, these analyses involve a single human variant and variants from a selected set of HCV proteins.

We have replaced the term “genome-to-genome analysis” with “integrated association analysis” throughout the article.

3) Subsection “Viral load association analyses”, second paragraph – I found this text confusing. Suggest the following revision (if correct): 'We then searched for associations between viral load and HCV amino acid variables. These analyses identified significant associations in all viral genotype groups except 4a (Figure 2). Amongst the viral amino acids that associated with viral load, a number also associated with rs12979860 genotype (genotype 1a, 9 of x amino acids; 1b, 5 of x amino acids; 2a, 0 of x amino acids; 2b, 0 of x amino acids; 3a, 2 of x amino acids).'

We have replaced the previous text with the one suggested by the reviewer (subsection “Viral load association analyses”, second paragraph).

4) Subsection “Viral load association analyses”, third paragraph – Please clarify if this analysis is restricted to individuals infected with genotype 1b. Please verify that Figure 3B is consistent with this text and that the y axis is not mislabeled.

Done (subsection “Viral load association analyses”, third paragraph). We checked the Figure 3B. It is consistent with the text and the y axis is not mislabeled.

5) Table 2 presents '97% CI', rather than 95% CI, which are more usual. If 97% is in fact correct, provide a justification for using an interval of that width.

Both 95% as well as 97% CI are correct. Since we had a large data size we could choose 97% CI to have more confidence in our results, without making the width between the lower and the upper bound too large. It can be seen from Table 2 that in spite of having 97% CI we have a narrow CI for the OR values and the lower and the upper range values of the CI are very close to the estimated OR values.

https://doi.org/10.7554/eLife.42542.024

Article and author information

Author details

  1. Nimisha Chaturvedi

    1. School of Life Sciences, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
    2. Swiss Institute of Bioinformatics, Lausanne, Switzerland
    Contribution
    Conceptualization, Data curation, Formal analysis, Methodology, Writing—original draft, Writing—review and editing
    For correspondence
    chaturvedi.nimisha20@gmail.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3065-0202
  2. Evguenia S Svarovskaia

    Gilead Sciences Inc, Foster City, United States
    Contribution
    Resources, Data curation, Writing—review and editing
    Competing interests
    This study was partially funded by Gilead Sciences and the author is an employee of Gilead Sciences
  3. Hongmei Mo

    Gilead Sciences Inc, Foster City, United States
    Contribution
    Resources, Writing—review and editing
    Competing interests
    This study was partially funded by Gilead Sciences and the author is an employee of Gilead Sciences
  4. Anu O Osinusi

    Gilead Sciences Inc, Foster City, United States
    Contribution
    Resources, Writing—review and editing
    Competing interests
    This study was partially funded by Gilead Sciences and the author is an employee of Gilead Sciences
  5. Diana M Brainard

    Gilead Sciences Inc, Foster City, United States
    Contribution
    Resources, Writing—review and editing
    Competing interests
    This study was partially funded by Gilead Sciences and the author is an employee of Gilead Sciences
  6. G Mani Subramanian

    Gilead Sciences Inc, Foster City, United States
    Contribution
    Resources, Writing—review and editing
    Competing interests
    This study was partially funded by Gilead Sciences and the author is an employee of Gilead Sciences
  7. John G McHutchison

    Gilead Sciences Inc, Foster City, United States
    Contribution
    Resources, Writing—review and editing
    Competing interests
    This study was partially funded by Gilead Sciences and the author is an employee of Gilead Sciences
  8. Stefan Zeuzem

    Goethe University Hospital, Frankfurt, Germany
    Contribution
    Writing—review and editing
    Competing interests
    has been a consultant for Abbvie, Gilead, Janssen, Merck/MSD
  9. Jacques Fellay

    1. School of Life Sciences, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
    2. Swiss Institute of Bioinformatics, Lausanne, Switzerland
    3. Precision Medicine Unit, Lausanne University Hospital, Lausanne, Switzerland
    Contribution
    Conceptualization, Funding acquisition, Writing—review and editing
    For correspondence
    jacques.fellay@epfl.ch
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8240-939X

Funding

Gilead Sciences

  • Jacques Fellay

Swiss National Science Foundation (PP00P3_157529)

  • Jacques Fellay

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Ethics

Human subjects: Across 82 studies involving <100 sites in many countries, appropriate informed consent was obtained from study participants allowing the current analysis to be performed. The studies were run by Gilead Sciences (Foster City, CA) and Pharmasset (formerly Princeton, NJ). Study protocols followed the ethical guidelines set in place by the 1975 Declaration of Helsinki and were approved by the relevant institutional review board committees (further details for the studies can be found in Supplementary Table 1 in Welzel et al. [Journal of Hepatology, 2017]). All samples included in this analysis are baseline samples collected from treatment naive and experienced patients from <25 countries in North America, Europe, Asia, Oceania, and Africa between years 2010 and 2015.

Senior Editor

  1. Wendy S Garrett, Harvard TH Chan School of Public Health, United States

Reviewing Editor

  1. Thomas O'Brien, NIH, United States

Publication history

  1. Received: October 3, 2018
  2. Accepted: May 31, 2019
  3. Version of Record published: September 3, 2019 (version 1)

Copyright

© 2019, Chaturvedi et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 224
    Page views
  • 26
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Genetics and Genomics
    2. Microbiology and Infectious Disease
    Thomas R O'Brien et al.
    Insight

    Polymorphisms in the IFNL4 gene that affect both the presence and the form of the coded protein are associated with changes in the hepatitis C virus.

    1. Genetics and Genomics
    2. Plant Biology
    Mark Zander et al.
    Research Article Updated