Genome-wide analysis of cell-Free DNA methylation profiling with MeDIP-seq identified potential biomarkers for colorectal cancer

Abstract

Background: Colorectal cancer is the most common malignancy and the third leading cause of cancer-related death worldwide. This study aimed to identify potential diagnostic biomarkers for colorectal cancer by genome-wide plasma cell-free DNA (cfDNA) methylation analysis.

Methods: Peripheral blood from colorectal cancer patients and healthy controls was collected for cfDNA extraction. Genome-wide cfDNA methylation profiling, especially differential methylation profiling between colorectal cancer patients and healthy controls, was performed by methylated DNA immunoprecipitation coupled with high-throughput sequencing (MeDIP-seq). Logistic regression models were established, and the accuracy of this diagnostic model for colorectal cancer was verified using tissue-sourced data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) due to the lack of cfDNA methylation data in public datasets.

Results: Compared with the control group, 939 differentially methylated regions (DMRs) located in promoter regions were found in colorectal cancer patients; 16 of these DMRs were hypermethylated, and the remaining 923 were hypomethylated. In addition, these hypermethylated genes, mainly PRDM14, RALYL, ELMOD1, and TMEM132E, were validated and confirmed in colorectal cancer by using publicly available DNA methylation data.

Conclusions: MeDIP-seq can be used as an optimal approach for analyzing cfDNA methylomes, and 12 probes of four differentially methylated genes identified by MeDIP-seq (PRDM14, RALYL, ELMOD1, and TMEM132E) could serve as potential biomarkers for clinical application in patients with colorectal cancer.

Keywords: Biomarkers; Colorectal cancer; MeDIP-seq; cfDNA.

Fig. 1

The cfDNA methylation patterns derived from MeDIP-seq datasets between colorectal cancer patients and controls. a Heuristic cluster analysis of methylation profiling between patients and controls. b Unsupervised cluster analysis of the genome-wide methylation profiling in patients and controls

Fig. 2

Differentially methylated regions in patients and controls. a The genomic distributions of hypomethylated and hypermethylated DMRs in introns, intergenomic, exons, non-coding, promoters and other regions. b The distribution of DMRs mapped to the whole genome on different chromosomes in patients. c Heat map of total 8398 DMRs, including 1875 hypermethylated and 6523 hypomethylated. d Heat map of DMRs located in promoter regions in patients and controls, including 16 hypermethylated and 923 hypomethylated

Fig. 3

Diagnostic predictive models and receiver operating characteristic (ROC) curves for colorectal cancer. a, b Confusion matrix built from the diagnostic predictive models in training (a) and validation (b) dataset. COAD, colon adenocarcinoma. c ROC curves and the associated area under the curve (AUCs) of the training and validation dataset

Fig. 4

Validation of hypermethylated genes by using publicly available DNA methylation data. a Unsupervised cluster analysis of these 12 probes extracted from the 488 cases of 450K methylation array dataset. b The comparison of methylation level between tumor and normal tissue of the 12 selected probes. All p values < 0.05