Integrated data analysis reveals potential drivers and pathways disrupted by DNA methylation in papillary thyroid carcinomas

Abstract

Background: Papillary thyroid carcinoma (PTC) is a common endocrine neoplasm with a recent increase in incidence in many countries. Although PTC has been explored by gene expression and DNA methylation studies, the regulatory mechanisms of the methylation on the gene expression was poorly clarified. In this study, DNA methylation profile (Illumina HumanMethylation 450K) of 41 PTC paired with non-neoplastic adjacent tissues (NT) was carried out to identify and contribute to the elucidation of the role of novel genic and intergenic regions beyond those described in the promoter and CpG islands (CGI). An integrative and cross-validation analysis were performed aiming to identify molecular drivers and pathways that are PTC-related.

Results: The comparisons between PTC and NT revealed 4995 methylated probes (88% hypomethylated in PTC) and 1446 differentially expressed transcripts cross-validated by the The Cancer Genome Atlas data. The majority of these probes was found in non-promoters regions, distant from CGI and enriched by enhancers. The integrative analysis between gene expression and DNA methylation revealed 185 and 38 genes (mainly in the promoter and body regions, respectively) with negative and positive correlation, respectively. Genes showing negative correlation underlined FGF and retinoic acid signaling as critical canonical pathways disrupted by DNA methylation in PTC. BRAF mutation was detected in 68% (28 of 41) of the tumors, which presented a higher level of demethylation (95% hypomethylated probes) compared with BRAF wild-type tumors. A similar integrative analysis uncovered 40 of 254 differentially expressed genes, which are potentially regulated by DNA methylation in BRAFV600E-positive tumors. The methylation and expression pattern of six selected genes (ERBB3, FGF1, FGFR2, GABRB2, HMGA2, and RDH5) were confirmed as altered by pyrosequencing and RT-qPCR.

Conclusions: DNA methylation loss in non-promoter, poor CGI and enhancer-enriched regions was a significant event in PTC, especially in tumors harboring BRAFV600E. In addition to the promoter region, gene body and 3'UTR methylation have also the potential to influence the gene expression levels (both, repressing and inducing). The integrative analysis revealed genes potentially regulated by DNA methylation pointing out potential drivers and biomarkers related to PTC development.

Keywords: BRAFV600E mutation; DNA methylation; FGF signaling pathway; Integrative analysis; Papillary thyroid cancer; Retinoic acid pathway.

Fig. 1

Workflow representative of the strategy used in the integrative analyses and in the cross-study validation. a Genome-wide methylation analysis revealed 6070 differentially methylated probes, and large-scale gene expression analysis identified 1657 differentially expressed genes in PTC (the last from a previous study). Corresponding probes/genes were submitted to a Pearson correlation test (34 PTC analyzed by both platforms) revealing 214 genes presenting probes with negative correlation and 49 genes with positive correlation. A total of 247 genes were classified as potentially regulated by DNA methylation in PTC. b A total of 4563 differentially methylated probes and 333 differentially expressed genes were identified in PTC according to BRAFV600E mutation. The Pearson correlation test revealed 69 and 17 genes with negative and positive correlation, respectively. Eighty three genes were classified as potentially regulated by DNA methylation in PTC BRAF mutated. *Tumor samples were initially corrected by NT samples (∆βPTC-∆βNT) and then BRAF positive and negative tumors were compared; ^§Some genes presented both methylation probes negatively and positively correlated. ^#Unadjusted p value

Fig. 2

Classification of the differentially methylated probes in PTC. a Supervised hierarchical clustering analysis showed 6070 differentially methylated probes in papillary thyroid carcinoma (PTC) versus normal thyroid (NT) tissues, mostly hypomethylated in PTC. The first cluster shows all normal samples (purple) and six PTC (orange), and the second is composed exclusively by tumor samples (orange). The beta values vary between zero (green) and one (red). b Methylation probes identified in PTC versus NT and those detected in the integrative analysis with negative (r−) and positive correlation (r+) according to the functional genomic distribution, CpG content, and neighborhood context and enhancer representation

Fig. 3

Methylation and gene expression profiling in PTC. a Genes identified in the integrative analysis with negative correlation and confirmed in the TCGA data. The outermost circle displays the human autosomal chromosomes, and the inner layers show both expression and methylation profiles. The figure was created following the parameters available in http://circos.ca. Unsupervised hierarchical clustering analysis revealed the b methylation and c gene expression profiles of 34 PTC evaluated with both platforms, and the relation with histological variant, genetic alteration, and follow-up. Two clusters were identified by both methodologies, and an overlapping between methylation and expression data was observed (dark and gray clustering). Gray cluster of methylation and gene expression was associated with a higher frequency of BRAF-mutated tumors (p = 0.034 and p = 0.013, respectively; Fisher’s exact test)

Fig. 4

Methylation (a) and expression levels (b) confirmation of the selected genes. a ERBB3, FGF1, GABRB2, HMGA2, and RDH5 hypomethylation and FGFR2 hypermethylation were confirmed in PTC samples by pyrosequencing after DNA modification by bisulfite. b. ERBB3, FGF1, GABRB2, HMGA2, and RDH5 overexpression and FGFR2 downexpression were confirmed in PTC by RT-qPCR. The boxplot indicates the interquartile range and median. ***p < 0.001 by comparing PTC to NT (Student’s t test)