Why do results conflict regarding the prognostic value of the methylation status in colon cancers? the role of the preservation method

BACKGROUND: In colorectal carcinoma, extensive gene promoter hypermethylation is called the CpG island methylator phenotype (CIMP). Explaining why studies on CIMP and survival yield conflicting results is essential. Most experiments to measure DNA methylation rely on the sodium bisulfite conversion of unmethylated cytosines into uracils. No study has evaluated the performance of bisulfite conversion and methylation levels from matched cryo-preserved and Formalin-Fixed Paraffin Embedded (FFPE) samples using pyrosequencing. METHODS: Couples of matched cryo-preserved and FFPE samples from 40 colon adenocarcinomas were analyzed. Rates of bisulfite conversion and levels of methylation of LINE-1, MLH1 and MGMT markers were measured. RESULTS: For the reproducibility of bisulfite conversion, the mean of bisulfite-to-bisulfite standard deviation (SD) was 1.3%. The mean of run-to-run SD of PCR/pyrosequencing was 0.9%. Of the 40 DNA couples, only 67.5%, 55.0%, and 57.5% of FFPE DNA were interpretable for LINE-1, MLH1, and MGMT markers, respectively, after the first analysis. On frozen samples the proportion of well converted samples was 95.0%, 97.4% and 87.2% respectively. For DNA showing a total bisulfite conversion, 8 couples (27.6%) for LINE-1, 4 couples (15.4%) for MLH1 and 8 couples (25.8%) for MGMT displayed significant differences in methylation levels. CONCLUSIONS: Frozen samples gave reproducible results for bisulfite conversion and reliable methylation levels. FFPE samples gave unsatisfactory and non reproducible bisulfite conversions leading to random results for methylation levels. The use of FFPE collections to assess DNA methylation by bisulfite methods must not be recommended. This can partly explain the conflicting results on the prognosis of CIMP colon cancers.