Global DNA Methylation Profiling Reveals Differentially Methylated CpGs between Salivary Gland Pleomorphic Adenomas with Distinct Clinical Course

Abstract

Pleomorphic adenomas (PAs) are the most frequently diagnosed benign salivary gland tumors. Although the majority of PAs are characterized by slow growth, some develop very fast and are more prone to recur. The reason for such differences remains unidentified. In this study, we performed global DNA methylation profiling using the Infinium Human Methylation EPIC 850k BeadChip Array (Illumina) to search for epigenetic biomarkers that could distinguish both groups of tumors. The analysis was performed in four fast-growing tumors (FGTs) and four slow-growing tumors (SGTs). In all, 85 CpG dinucleotides differentiating both groups were identified. Six CpG tags (cg06748470, cg18413218, cg10121788, cg08249296, cg18455472, and cg19930657) were selected for bisulfite pyrosequencing in the extended group of samples. We confirmed differences in DNA methylation between both groups of samples. To evaluate the potential diagnostic accuracy of the selected markers, ROC curves were constructed. We indicated that CpGs included in two assays showed an area under the curve with an acceptable prognostic value (AUC > 0.7). However, logistic regression analysis allowed us to indicate a more optimal model consisting of five CpGs ((1) cg06748470, (2) cg00600454, (3) CpG located in chr14: 77,371,501−77,371,502 (not annotated in GRCh37/hg19), (4) CpG2 located in chr16: 77,469,589−77,469,590 (not annotated GRCh37/hg19), and (5) cg19930657) with AUC > 0.8. This set of epigenetic biomarkers may be considered as differentiating factors between FGT and SGT during salivary gland tumor diagnosis. However, this data should be confirmed in a larger cohort of samples.

Keywords: CpG; DNA methylation; bisulfite pyrosequencing; fast-growing tumors; salivary gland pleomorphic adenoma; slow-growing tumors.

Figure 1

Visualization of MDS analysis using the “Manhattan” method of distance calculation. Two analyzed cohorts are shown: fast-growing salivary gland tumors (red) and slow-growing salivary gland tumors (green).

Figure 2

Volcano plot of DNA methylation data derived from microarray analysis of FGT (n = 4) and SGT (n = 4) samples. Red and blue dots represent hyper- and hypomethylated CpG sites in FGTs, respectively, gray dots represent nonsignificant CpG sites. In all, 85 CpGs hypermethylated in FGT, selected based on MMD, are shown as yellow dots in the background of all significantly hypermethylated tags.

Figure 3

Mean methylation levels between fast-growing (dark gray) and slow-growing (light gray) tumors for the six CG tags selected based on the results of the Illumina Infinium Human Methylation EPIC BeadChip Array.

Figure 4

Mean methylation differences between fast-growing (dark gray) and slow-growing (light gray) tumors for each assay performed by DNA bisulfite pyrosequencing. * p < 0.05, ** p < 0.001, *** p < 0.0001, **** p < 0.00001.

Figure 5

Visualization of ROC curves representing two logistic regression models: (A) full model (all explanatory variables, i.e., all CpG dinucleotides included) and (B) final model obtained by stepwise approach (AIC) including only selected CpG dinucleotides.