Learning a Structural and Functional Representation for Gene Expressions: To Systematically Dissect Complex Cancer Phenotypes

Genome-wide methylation profiling identified novel differentially hypermethylated biomarker MPPED2 in colorectal cancer

BACKGROUND

Epigenetic alternation is a common contributing factor to neoplastic transformation. Although previous studies have reported a cluster of aberrant promoter methylation changes associated with silencing of tumor suppressor genes, little is known concerning their sequential DNA methylation changes during the carcinogenetic process. The aim of the present study was to address a genome-wide search for identifying potentially important methylated changes and investigate the onset and pattern of methylation changes during the progression of colorectal neoplasia.

METHODS

A three-phase design was employed in this study. In the screening phase, DNA methylation profile of 12 pairs of colorectal cancer (CRC) and adjacent normal tissues was analyzed by using the Illumina MethylationEPIC BeadChip. Significant CpG sites were selected based on a cross-validation analysis from The Cancer Genome Atlas (TCGA) database. Methylation levels of candidate CpGs were assessed using pyrosequencing in the training dataset (tumor lesions and adjacent normal tissues from 46 CRCs) and the validation dataset (tumor lesions and paired normal tissues from 13 hyperplastic polyps, 129 adenomas, and 256 CRCs). A linear mixed-effects model was used to examine the incremental changes of DNA methylation during the progression of colorectal neoplasia.

RESULTS

The comparisons between normal and tumor samples in the screening phase revealed an extensive CRC-specific methylomic pattern with 174,006 (21%) methylated CpG sites, of which 22,232 (13%) were hyermethylated and 151,774 (87%) were hypomethylated. Hypermethylation mostly occurred in CpG islands with an overlap of gene promoters, while hypomethylation tended to be mapped far away from functional regions. Further cross validation analysis from TCGA dataset confirmed 265 hypermethylated promoters coupling with downregulated gene expression. Among which, hypermethylated changes in MEEPD2 promoter was successfully replicated in both training and validation phase. Significant hypermethylation appeared since precursor lesions with an extensive modification in CRCs. The linear mixed-effects modeling analysis found that a cumulative pattern of MPPED2 methylation changes from normal mucosa to hyperplastic polyp to adenoma, and to carcinoma (P < 0.001).

CONCLUSIONS

Our findings indicate that epigenetic alterations of MPPED2 promoter region appear sequentially during the colorectal neoplastic progression. It might be able to serve as a promising biomarker for early diagnosis and stage surveillance of colorectal tumorigenesis.