Gene expression profiling-based prediction of response of colon carcinoma cells to 5-fluorouracil and camptothecin.

5-Fluorouracil (5-FU) is the most common chemotherapeutic agent used in the treatment of colorectal cancer, yet objective response rates are low. Recently, camptothecin (CPT) has emerged as an effective alternative therapy. Decisive means to determine treatment, based on the likelihood of response to each of these agents, could greatly enhance the management of this disease. Here, the ability of cDNA microarray-generated basal gene expression profiles to predict apoptotic response to 5-FU and CPT was determined in a panel of 30 colon carcinoma cell lines. Genes whose basal level of expression correlated significantly with 5-FU- and CPT-induced apoptosis were selected, and their predictive power was assessed using a "leave one out" jackknife cross-validation strategy. Selection of the 50 genes best correlated with 5-FU-induced apoptosis, but not 50 randomly selected genes, significantly predicted response to this agent. Importantly, this gene expression profiling approach predicted response more effectively than four previously established determinants of 5-FU response: thymidylate synthase and thymidine phosphorylase activity; and p53 and mismatch repair status. Furthermore, reanalysis of the database demonstrated that selection of the 149 genes best correlated with CPT-induced apoptosis maximally and significantly predicted response to this agent. These studies demonstrate that the basal gene expression profile of colon cancer cells can be used to predict and distinguish response to multiple chemotherapeutic agents and establish the potential of this methodology as a means by which rational decisions regarding choice of therapy can be approached.