Identification of DNA methylation biomarkers for risk of liver metastasis in early-stage colorectal cancer

Background Liver metastases can occur even in CRC patients who underwent curative surgery. While evidence suggested that adjuvant chemotherapy can help to reduce the occurrence of liver metastases for certain patients, it is not a recommended routine as the side effects outweigh the potential benefits, especially in Stage II CRC patients. This study aims to construct a model for predicting liver metastasis risk using differential methylation signals in primary CRC tumors, which can facilitate the decision for adjuvant chemotherapy. Methods Fifty-nine stage I/II and IV CRC patients were enrolled. Primary tumor, adjacent normal tissue, and metastatic tumor tissues were subject to targeted bisulfite sequencing for DNA methylation. The Least Absolute Shrinkage and Selection Operator (LASSO) algorithm was used to identify potential DMRs for predicting liver metastasis of CRC. Results We identified a total of 241,573 DMRs by comparing the DNA methylation profile of primary tumors of stage II patients who developed metastasis to those who were metastasis-free during the follow up period. 213 DMRs were associated with poor prognosis, among which 182 DMRS were found to be hypermethylated in the primary tumor of patients with metastases. Furthermore, by using the LASSO regression model, we identified 23 DMRs that contributed to a high probability of liver metastasis of CRC. The leave-one-out cross validation (LOOCV) was used to evaluate model predictive performance at an AUC of 0.701. In particular, 7 out of those 23 DMRs were found to be in the promoter region of genes that were previously reported prognostic biomarkers in diverse tumor types, including TNNI2, PAX8, GUF1, KLF4, EVI2B, CEP112, and long non-coding RNA AC011298. In addition, the model was also able to distinguish metastases of different sites (liver or lung) at an AUC of 0.933. Conclusion We have identified DNA methylation biomarkers associated with the risk of cancer liver metastasis in early-stage CRC patients. A risk prediction model based on those epigenetic markers was proposed for outcome assessment.

Automated summary

This study aimed to construct a model for predicting liver metastasis risk in CRC patients by utilizing differential methylation signals in primary tumors. A total of 241,573 DMRs were identified, with 182 hypermethylated DMRs associated with metastasis in stage II CRC patients. The predictive model showed promising performance with an AUC of 0.701 and was able to distinguish metastases in different sites with an AUC of 0.933. Identified DMRs in the promoter region of known prognostic biomarkers suggest their potential role in predicting liver metastasis in early-stage CRC patients.