Genome-Wide Association Analyses of HPV16 and HPV18 Seropositivity Identify Susceptibility Loci for Cervical Cancer

Abstract

Infection by high-risk human papillomavirus is known to exacerbate cervical cancer development. The host immune response is crucial in disease regression. Large-scale genetic association studies for cervical cancer have identified few susceptibility variants, mainly at the human leukocyte antigen locus on chromosome 6. We hypothesized that the host immune response modifies cervical cancer risk and performed three genome-wide association analyses for HPV16, HPV18 and HPV16/18 seropositivity in 7814, 7924, and 7924 samples from the UK Biobank, followed by validation genotyping in the German Cervigen case-control series of cervical cancer and dysplasia. In GWAS analyses, we identified two loci associated with HPV16 seropositivity (6p21.32 and 15q26.2), two loci associated with HPV18 seropositivity (5q31.2 and 14q24.3), and one locus for HPV16 and/or HPV18 seropositivity (at 6p21.32). MAGMA gene-based analysis identified HLA-DQA1 and HLA-DQB1 as genome-wide significant (GWS) genes. In validation genotyping, the genome-wide significant lead variant at 6p21.32, rs9272293 associated with overall cervical disease (OR = 0.86, p = 0.004, 95% CI = 0.78-0.95, n = 3710) and HPV16 positive invasive cancer (OR = 0.73, p = 0.005, 95% CI = 0.59-0.91, n = 1431). This variant was found to be a robust eQTL for HLA-DRB1, HLA-DQB1-AS1, C4B, HLA-DRB5, HLA-DRB6, HLA-DQB1, and HLA-DPB1 in a series of cervical epithelial tissue samples. We additionally genotyped twenty-four HPV seropositivity variants below the GWS threshold out of which eleven variants were found to be associated with cervical disease in our cohort, suggesting that further seropositivity variants may determine cervical disease outcome. Our study identifies novel genomic risk loci that associate with HPV type-specific cervical cancer and dysplasia risk and provides evidence for candidate genes at one of the risk loci.

Keywords: SNP; cervical carcinoma; eQTL; human papillomavirus; infection.

Figure 1

GWAS plots for HPV seropositivity GWASs. −log10 p values from the GWAS summary statistics are plotted on the y axis for (A) HPV16 seropositivity, (B) HPV18 seropositivity and (C) HPV16 and/or 18 seropositivity in Manhattan plots. Chromosomes are represented on the x axis. The dotted red line is set at the genome‐wide significance level of 5 × 10E‐8. Furthermore, −log10 p values after MAGMA gene‐based analysis are shown with Manhattan plots for (D) HPV16 seropositivity, (E) HPV18 seropositivity and (F) HPV16 and/or 18 seropositivity. Since SNPs were mapped to 19,380 protein coding genes, the red line indicates p = 0.05/19380 = 2.58 × 10E‐6.

Figure 2

eQTL analysis for rs9272293 in all cervical tissues. −log10 relative quantities (+/−SEM) are shown on the y axis, together with genotypes on the x axis (A1 = common homozygous genotype GG, het. = heterozygous genotype GA, A2 = rare homozygous genotype AA). Sample numbers per group are written below the respective bars on the x axis. p values are indicated after ANOVA was performed between three groups followed by a linear trend test, with the common genotype as the control.

Figure 3

eQTL analysis for rs9272293 in HPV negative and HPV positive cervical tissues. −log10 relative quantities (+/− SEM) are shown on the y axis, together with genotypes on the x axis (A1 = common homozygous genotype GG, het. = heterozygous genotype GA, A2 = rare homozygous genotype AA). Sample numbers per group are written below the respective bars on the x axis. p values are indicated after ANOVA was performed between three groups followed by a linear trend test, with the common genotype as the control.

Figure 4

Correlation analysis for genes that are eQTLs for rs9272293. Pearson correlation R values are shown between genes that showed evidence to be eQTLs for rs9272293. Panels from left to right indicate the genotype of rs9272293 (A1 = common homozygous genotype GG, het. = heterozygous genotype GA, A2 = rare homozygous genotype AA). Negative correlation (R = −1) is shown in deep red color, whereas positive correlation (R = 1) is shown in deep blue color. Missing values are represented with a black X through the white box.