Methylation of HPV18, HPV31, and HPV45 genomes and cervical intraepithelial neoplasia grade 3

Abstract

Background: Persistent infections with carcinogenic human papillomavirus (HPV) types are the necessary cause of cervical cancer. We recently demonstrated that the HPV16 genome is strongly methylated in cervical precancer compared with transient infections. However, the extent of methylation in other HPV types and its role in progression to cancer is poorly understood.

Methods: We analyzed whole-genome methylation patterns of the three next most carcinogenic HPV genotypes: HPV31 (closely related to HPV16), and two other closely related types, HPV18 and HPV45. DNA was extracted from cervical cytology specimens from 92 women with precancer and 96 women infected with HPV31, HPV18, or HPV45, but who had no cytological or histological abnormalities. After bisulfite modification, genome-wide pyrosequencing was performed covering 80-106 sites. We calculated differences in median methylation, odds ratios, areas under the curve, and Spearman rank correlation coefficients for methylation levels between different sites. All statistical tests were two-sided.

Results: For all three HPV types, we observed strongly elevated methylation levels at multiple CpG sites in the E2, L2, and L1 regions among women with cervical intraepithelial neoplasia grade 3 compared with women with transient infections. We observed high correlation of methylation patterns between phylogenetically related types. The highest areas under the curve were 0.81 for HPV31, 0.85 for HPV18, and 0.98 for HPV45. Differential methylation patterns in cervical intraepithelial neoplasia grade 3 patients with multiple infections suggest that methylation can clarify which of the infections is causal.

Conclusions: Carcinogenic HPV DNA methylation indicates transforming HPV infections. Our findings show that methylation of carcinogenic HPV types is a general phenomenon that warrants development of diagnostic assays.

Figures

Figure 1.

HPV31 methylation levels in case patients and control subjects. For each of the analyzed 80 CpG sites, the median percent methylation in case patients (red bars) and control subjects (grey bars) is shown (secondary y-axis). The black dots indicate Mann-Whitney U-test P values for median methylation differences between case patients and control subjects on an inverse logarithmic scale (primary y-axis). The HPV31 genomic region is shown on the x-axis. The correlation matrix in the bottom part of the graph indicates the correlation of methylation levels at all sites across the HPV31 genome. Black lines indicate transitions between adjacent HPV genes. For each combination of two sites, the Spearman rho correlation coefficient is shown color-coded with a range from perfect correlation (1, dark red) to inverse correlation (-1, dark blue) as indicated by the color bar. All statistical tests were two-sided.

Figure 2.

HPV18 methylation levels in case patients and control subjects. For each of the analyzed 105 CpG sites, the median percent methylation in case patients (red bars) and control subjects (grey bars) is shown (secondary y-axis). The black dots indicate Mann-Whitney-U test P values for median methylation differences between case patients and control subjects on an inverse logarithmic scale (primary y-axis). The HPV18 genomic region is shown on the x-axis. The correlation matrix in the bottom part of the graph indicates the correlation of methylation levels at all sites across the HPV18 genome. Black lines indicate transitions between adjacent HPV genes. For each combination of two sites, the Spearman rho correlation coefficient is shown color-coded with a range from perfect correlation (1, dark red) to inverse correlation (-1, dark blue) as indicated by the color bar. All statistical tests were two-sided.

Figure 3.

HPV45 methylation levels in case patients and control subjects. For each of the analyzed 106 CpG sites, the median percent methylation in case patients (red bars) and control subjects (grey bars) is shown (secondary y-axis). The black dots indicate Mann-Whitney U-test P values for median methylation differences between case patients and control subjects on an inverse logarithmic scale (primary y-axis). The HPV45 genomic region is shown on the x-axis. The correlation matrix in the bottom part of the graph indicates the correlation of methylation levels at all sites across the HPV45 genome. Black lines indicate transitions between adjacent HPV genes. For each combination of two sites, the Spearman rho correlation coefficient is shown color-coded with a range from perfect correlation (1, dark red) to inverse correlation (-1, dark blue) as indicated by the color bar. All statistical tests were two-sided.

Figure 4.

Median methylation levels in case patients with single and multiple infections. For HPV31, median methylation levels are shown for the complete genome stratified by single infections (red line), multiple infections with HPV16 (green line), and multiple infections without HPV16 (orange line). Percent methylation is indicated on the y-axis, and CpG site positions are indicated on the x-axis. The panel below the graph shows the Mann-Whitney U-test P values for median differences between case patients and control subjects in the three strata. The color scale ranges from low (dark green) to high (dark red) P values. The distribution of statistically significant P values across the HPV31 genome was similar for case patients with single HPV31 infections, and case patients with multiple infections without HPV16. All statistical tests were two-sided.