Evaluation of a large, population-based sample supports a CpG island methylator phenotype in colon cancer

BACKGROUND & AIMS: The concept of a CpG island methylator phenotype (CIMP), especially in microsatellite stable colon cancer, is not accepted universally. We therefore evaluated a large population-based sample of individuals with colon cancer and used univariate and multivariate analyses of CIMP with clinicopathologic variables and tumor mutations to determine the biologic relevance of this phenotype. METHODS: A total of 864 tumors from individuals with colon cancer from Utah and Northern California were evaluated by methylation-specific polymerase chain reaction of CpG islands in hMLH1, methylated in tumors (MINT) 1, MINT 2, MINT 31, and CDKN2A (p16). CIMP high was defined as methylation at 2 or more of these loci. The BRAF V600E mutation was determined by sequencing. Microsatellite instability had been determined previously. RESULTS: In a multivariate analysis of microsatellite stable tumors, CIMP high was related significantly to the V600E BRAF mutation (odds ratio, 39.52; 95% confidence interval, 11.44-136.56), KRAS2 mutations (odds ratio, 2.22; 95% confidence interval, 1.48-3.34), older age (P trend = .03), and increased stage (P trend = .03), and these tumors were less likely to be located in the distal colon (odds ratio, .42; 95% confidence interval, .27-.65). CIMP-high unstable tumors also were more likely to have the V600E BRAF mutation, be located proximally, and occur in older individuals (in univariate analyses). However, CIMP-high unstable tumors were significantly more likely than their stable counterparts to be KRAS2 wild type, TP53 wild type, poorly differentiated, proximally located, occur at lower stages, and have the BRAF V600E mutation (64.1% vs 17.6%). CONCLUSIONS: The evaluation of a large, population-based sample strongly supports the biologic relevance of CIMP in colon cancer. However, the presence or absence of microsatellite instability has a major effect on the expression of this phenotype.