Identification of methylation biomarkers in colorectal carcinoma and their detection in circulating tumour cells

Circulating tumour cells (CTCs) provide a minimally invasive approach for cancer detection and disease monitoring in metastatic carcinoma. This study aimed to identify methylation biomarkers and assess their translational potential for CTC-based liquid biopsy in colorectal carcinoma (CRC). Predictive methylation sites were screened using the TCGA-Colon Adenocarcinoma dataset (430 cancer and 39 non-cancer colon mucosae) through differential methylation analysis, Lasso regression, and Cox modelling. Three representative sites— GPR125 , GDNF , and ADCY9 —were validated in 62 patients with colorectal adenocarcinoma using quantitative methylation-specific PCR, and comparative analyses between paired CTC and tissue samples were performed in 32 cases. The prediction model yielded a hazard ratio of 4.23 ( p < 0.001) with 1-, 3-, and 5-year area under the curve (AUC) values of 0.718, 0.765, and 0.787, respectively. GDNF showed the strongest CTC detection performance, with 81.2% sensitivity, 93.8% specificity, and an AUC of 0.887. Clinical association analyses revealed that GDNF methylation was significantly correlated with age ( p = 0.040), primary tumour site ( p = 0.004), tumour grade ( p = 0.017), T stage ( p = 0.027), and lymphovascular invasion (LVI) ( p = 0.041). GPR125 methylation was associated with tumour grade ( p = 0.017), LVI ( p = 0.041), and presence of other neoplasms ( p = 0.026). ADCY9 methylation showed limited clinical associations, correlating only with the presence of other neoplasms ( p = 0.029). Network analysis revealed that GDNF formed a distinct regulatory module with strong clinical relevance. These findings suggest that GDNF methylation holds strong potential as a CTC-based liquid biopsy biomarker in colorectal adenocarcinoma, demonstrating high detection accuracy despite systematic differences from tissue methylation patterns. The limited translational applicability of biomarkers derived from other tissues highlights the challenges of translating bioinformatic discoveries into effective liquid biopsy implementation.