Genome-Wide Analysis of Cell-Free DNA Methylation Profiling for the Early Diagnosis of Pancreatic Cancer

Abstract

As one of the most malicious cancers, pancreatic cancer is difficult to treat due to the lack of effective early diagnosis. Therefore, it is urgent to find reliable diagnostic and predictive markers for the early detection of pancreatic cancer. In recent years, the detection of circulating cell-free DNA (cfDNA) methylation in plasma has attracted global attention for non-invasive and early cancer diagnosis. Here, we carried out a genome-wide cfDNA methylation profiling study of pancreatic ductal adenocarcinoma (PDAC) patients by methylated DNA immunoprecipitation coupled with high-throughput sequencing (MeDIP-seq). Compared with healthy individuals, 775 differentially methylated regions (DMRs) located in promoter regions were identified in PDAC patients with 761 hypermethylated and 14 hypomethylated regions; meanwhile, 761 DMRs in CpG islands (CGIs) were identified in PDAC patients with 734 hypermethylated and 27 hypomethylated regions (p-value < 0.0001). Then, 143 hypermethylated DMRs were further selected which were located in promoter regions and completely overlapped with CGIs. After performing the least absolute shrinkage and selection operator (LASSO) method, a total of eight markers were found to fairly distinguish PDAC patients from healthy individuals, including TRIM73, FAM150A, EPB41L3, SIX3, MIR663, MAPT, LOC100128977, and LOC100130148. In conclusion, this work identified a set of eight differentially methylated markers that may be potentially applied in non-invasive diagnosis of pancreatic cancer.

Keywords: MeDIP-seq; biomarkers; cfDNA; methylation; pancreatic ductal adenocarcinoma.

FIGURE 1

The flow chart of screening cfDNA methylation biomarkers in pancreatic cancer.

FIGURE 2

The methylation patterns of pancreatic cancer patients and healthy controls after MeDIP-seq datasets analysis. (A) Principal component analysis (PCA) of the methylation profiles between patients and controls. (B) The unsupervised cluster analysis of the genome-wide methylation profiles in patients and controls.

FIGURE 3

Differentially methylated regions (DMRs) in pancreatic cancer patients and healthy controls. (A) Heat map of total 5,205 DMRs located in the whole genome of PDAC patients compared to healthy controls, including 5,117 hypermethylated and 88 hypomethylated regions. (B) Heat map of total 775 DMRs located in the promoter regions of patients compared to healthy controls, including 761 hypermethylated and 14 hypomethylated regions.

FIGURE 4

Differentially methylated regions (DMRs) of the CpG regions in pancreatic cancer patients and healthy controls. (A) Violin plots of DMRs located in CpG islands, CpG shores, and CpG shelfs of PDAC patients compared to controls. (B) Whole genomic and chromosomal location of DMRs in CGIs. (C) The different features of CGI distribution according to hypermethylated and hypomethylated regions.

FIGURE 5

Selection and definition of differentially methylated genes in both the CGIs and promoter regions. (A) Hypermethylated DMRs in the overlap of promoter regions and CGIs. (B) Top disease and bio functions by IPA analysis for genes derived from hypermethylated DMRs located in both the promoter regions and CGIs.

FIGURE 6

Identification of novel pancreatic cancer diagnostic markers from cfDNA methylation analysis. (A) Confusion tables of binary results of the diagnostic prediction model in the training and validation datasets. (B) ROC of the diagnostic prediction model with methylation markers in the training and validation datasets. (C) Unsupervised hierarchical clustering of the eight methylation markers selected for use in the diagnostic prediction model.

FIGURE 7

The comparison of the methylation level of the eight selected markers between pancreatic cancer patients and healthy controls.