Purpose: This study aimed to identify novel biological markers for the prediction of colorectal cancer liver metastasis.
Experimental design: We established two models that mimicked the interactions between colorectal tumor cells and the liver microenvironment. From these models we established subcell lines that had an enhanced ability to metastasize to the liver. Genes that related to hepatic metastasis were screened by microarray. The candidate markers were tested by immunohistochemistry, and their predictive accuracy was assessed by the cross-validation method and an independent test set.
Results: Highly metastatic colon cancer cell sublines SW1116p21 and SW1116v3 were established from the tumor cell-microenvironment interaction models. Seven of the up-regulated genes in the sublines were selected as candidate markers for predicting metastatic potential. A total of 245 colorectal cancer samples were divided into a training set containing 117 cases and a test set containing 128 cases. In the training set, immunohistochemical analysis showed CCL2 and SNCG expression was higher in the hepatic metastasis group than in the nonmetastasis group, and was correlated with poor survival. Logistic regression analysis revealed that CCL2 and SNCG levels in primary tumors, serum carcinoembryonic antigen level, and lymph node metastasis status were the only significant (P < 0.05) parameters for detecting liver metastasis. In leave-one-out-cross-validation, the two markers, when combined with clinicopathologic features, resulted in 90.5% sensitivity and 90.7% specificity for hepatic metastasis detection. In an independent test set, the combination achieved 87.5% sensitivity and 82% specificity for predicting the future hepatic metastasis of colorectal cancer.
Conclusion: Our results suggest that these models are able to mimic the interactions between colorectal cancer cells and the liver microenvironment, and may represent a promising strategy to identify metastasis-related genes. CCL2 and SNCG, combined with clinicopathologic features, may be used as accurate predictors of liver metastasis in colorectal cancer.