Novel DNA methylation biomarkers in stool and blood for early detection of colorectal cancer and precancerous lesions

Abstract

Background: Early detection and prevention of precancerous lesions can significantly reduce the morbidity and mortality of colorectal cancer (CRC). Here, we developed new candidate CpG site biomarkers for CRC and evaluated the diagnostic value of their expression in blood and stool samples of CRC and precancerous lesions.

Methods: We analyzed 76 pairs of CRC and adjacent normal tissue samples, 348 stool samples, and 136 blood samples. Candidate biomarkers for CRC were screened using a bioinformatics database and identified using a quantitative methylation-specific PCR method. The methylation levels of the candidate biomarkers were validated using blood and stool samples. The divided stool samples were used to construct and validate a combined diagnostic model and to analyze the independent or combined diagnostic value of candidate biomarkers in stool samples of CRC and precancerous lesions.

Results: Two candidate CpG site biomarkers for CRC, cg13096260 and cg12993163, were identified. Although both biomarkers demonstrated diagnostic performance to a certain extent when using blood samples, they showed better diagnostic value for different stages of CRC and AA with stool samples.

Conclusions: cg13096260 and cg12993163 detection in stool samples could be a promising approach for screening and early diagnosis of CRC and precancerous lesions.

Keywords: Advanced adenoma; Colorectal cancer; DNA methylation biomarkers; Early detection.

Fig. 1

Flowchart of candidate CpG site screening. Initial screening was performed using the DNA methylation dataset of CRC in TCGA. The screening criteria were as follows: the methylation difference between CRC and normal tissues was Δβ ≥ 0.2 and was statistically significant (P < 0.01), and the methylation value of CpG sites in CRC tissues was β ≥ 0.2. Initially, 19,862 CpG sites with differences were screened. Next, the peripheral blood DNA methylation dataset of the non-tumor population in the GEO database was used for rescreening. The methylation levels of CpG sites in CRC tissues were compared with those in blood, and CpG sites with methylation values less than 0.25 were excluded. CpG sites (15,404) were identified. They were also ranked according to differences in their methylation values. Based on the comparison of some of the CRC-related methylation genes, it was found that the CpG sites of SDC2 and SHOX2 genes contained more and ranked higher among the 15,404 methylation sites, with a total of 17 CpG sites. Using the random forest algorithm and selecting CpG sites with low methylation values in normal tissues, the final two candidate CpG sites were obtained. CRC colorectal cancer

Fig. 2

Methylation levels of cg13096260 and cg12993163. Difference analysis of cg13096260 (a) and cg12993163 (b) in TCGA CRC methylation database. Differential methylation levels of cg13096260 (c) and cg12993163 (d) in cancerous and normal precancerous tissues of patients with CRC. Methylation levels of cg13096260 (e) and cg12993163 (f) in stool samples of patients with NED, AA, stage I and II CRC, and stage III&IV CRC. Methylation levels of cg13096260 (g) and cg12993163 (h) in blood samples of patients with NED, AA, stage I and II CRC, and stage III&IV CRC. NED no evidence of disease, AA advanced adenoma, CRC colorectal cancer

Fig. 3

ROC curves and AUC for cg13096260 and cg12993163 in the training set of stool samples. The predictive ability for AA and CRC (a), CRC (b), I and II CRC (c), and AA (d). ROC curve and AUC of cg13096260, cg12993163, and the combined diagnostic model in the validation set of stool samples. The predictive ability for AA and CRC (e), CRC (f), I&II CRC (g), and AA (h). AA advanced adenoma, CRC colorectal cancer

Fig. 4

Diagnostic performance in stool samples. Sensitivity and specificity of cg13096260 (a), cg12993163 (b), and the combined diagnostic model (c) for the stool sample training set, validation set, and total samples

Fig. 5

ROC curve and AUC of cg13096260 and cg12993163 in blood samples. The predictive ability for AA and CRC (a), CRC (b), I and II CRC (c), and AA (d). AA advanced adenoma; CRC colorectal cancer