HPV Infection Leaves a DNA Methylation Signature in Oropharyngeal Cancer Affecting Both Coding Genes and Transposable Elements

Abstract

HPV oncoproteins can modulate DNMT1 expression and activity, and previous studies have reported both gene-specific and global DNA methylation alterations according to HPV status in head and neck cancer. However, validation of these findings and a more detailed analysis of the transposable elements (TEs) are still missing. Here we performed pyrosequencing to evaluate a 5-CpG methylation signature and Line1 methylation in an oropharyngeal squamous cell carcinoma (OPSCC) cohort. We further evaluated the methylation levels of the TEs, their correlation with gene expression and their impact on overall survival (OS) using the TCGA cohort. In our dataset, the 5-CpG signature distinguished HPV-positive and HPV-negative OPSCC with 66.67% sensitivity and 84.33% specificity. Line1 methylation levels were higher in HPV-positive cases. In the TCGA cohort, Line1, Alu and long terminal repeats (LTRs) showed hypermethylation in a frequency of 60.5%, 58.9% and 92.3%, respectively. ZNF541 and CCNL1 higher expression was observed in HPV-positive OPSCC, correlated with lower methylation levels of promoter-associated Alu and LTR, respectively, and independently associated with better OS. Based on our findings, we may conclude that a 5-CpG methylation signature can discriminate OPSCC according to HPV status with high accuracy and TEs are differentially methylated and may regulate gene expression in HPV-positive OPSCC.

Keywords: DNA methylation; HPV; gene expression; oropharyngeal squamous cell carcinoma; overall survival; transposable elements.

Figure 1

DNA methylation signatures of HPV-positive OPSCC. (A) Boxplot showing the mean methylation of 5 CpG sites in B3GALT6-SDF4, SYCP2-FAM217B, and HLTF-HLTF-AS1 loci assessed by pyrosequencing in HPV-positive (n = 21) and HPV-negative (n = 134) OPSCC. (B) Receiver operating characteristic (ROC) curve showing the best DNA methylation cut-off of the 5-CpG signature to distinguish HPV-positive and HPV-negative OPSCC. (C) Boxplot showing the methylation levels of Line1 transposable element assessed by pyrosequencing in HPV-positive (n = 21) and HPV-negative (n = 325) OPSCC.

Figure 2

Transposable elements are differentially methylated in OPSCC according to HPV status. (A) Volcano plots showing the Line1, Alu and LTR elements differentially methylated in HPV-positive relative to HPV-negative OPSCC. The X-axis shows the mean beta value differences (mean methylation delta) between groups and the Y-axis shows the -Log Benjamini-Hochberg (BH) adjusted p-values. Red dots represent elements with BH adjusted p-value ≤ 0.01. (B) Chromosomal distribution and (C) genomic region distribution of hypermethylated (red) and hypomethylated (blue) Line1, Alu and LTR elements in HPV-positive relative to HPV-negative OPSCC. The same TE may encompass more than one genomic region.

Figure 3

Differentially methylated transposable elements located at promoter regions show a correlation with gene expression and impact on prognosis. (A) Heatmap showing the unsupervised clustering of OPSCC TCGA samples according to the DNA methylation profile of transposable elements located up to 2000 bp of transcription start sites (promoter regions). Each line represents a transposable element and each column represents a sample. On the right, bar plots show the correlation rho (Spearman test) between the methylation levels of the element and the expression of the associated gene. Dark and light purple bars represent significant (BH adjusted p < 0.05) and nonsignificant correlations, respectively. (B) Heatmap showing the clustering of OPSCC TCGA samples according to the expression levels of genes correlated with the methylation levels of transposable elements in their promoter regions. Each line represents a transposable element and each column represents a sample. On the right, boxplots showing the expression of each gene in HPV-positive (dark pink) and HPV-negative (light pink) OPSCC. * Genes differentially expressed according to HPV status (BH adjusted p < 0.05). (C) Kaplan−Meier curves showing the overall survival of OPSCC patients from TCGA cohort according to the expression of ZNF541 and CCNL1. High and low expression were defined according to the median expression of each gene in all samples.