Establishment of Myelodysplastic Syndrome-Specific Methylome: Implications for the Pathogenesis of Clonal Evolution and Association with Survival

Epigenetic silencing of genes, such as tumor suppressor genes (TSG), throughaberrant promoter hypermethylation has been implicated in the pathogenesis of MDS. Epigenetic silencing may cooperate with chromosomal abnormalities to completelysilence a TSG or duplication of methylated alleles through UPD could potentially leadto complete TSG silencing. Hypomethylating agents can reverse aberrant silencing;however targeted application of epigenetic therapy is not possible, as ypermethylatedsites relevant to MDS pathogenesis remain mostly unknown. Traditional echnologieslimited methylation analysis to a small number of individual loci; as a result, there has been a lack of systematic studies on the methylation pattern in MDS. New methylationarray techniques allow for rapid polygenic analysis of methylation and the stablishmentof tissue- and disease-specific methylomes. We hypothesized that using ethylationarrays (Illumina®) aberrantly hypermethylated CpG sites and whole methylation patternspathognomic for MDS can be identified. First, we compared patients with MDS andAML (n=240) to controls (n=64) using low-density methylation arrays (1,505 CpG sites)to explore the general applicability of whole genome methylation arrays. We hen usedhigh density arrays (27,578 CpG sites) to fully explore a disease-specific ethylome in a representative sub-cohort of MDS/AML patients. We developed ananalytic algorithm that included establishment of the methylome of normal marrow as a reference and analysis of concordantly hypermethylated genes in patients, using methylation status as either a continuous or dichotomized variable. Global methylation analysis demonstrated that there was concordant hypermethylation in 25% and 50% of MDS patients in 1,199 CpG and 93 CpG sites, respectively, and in 25% and 50% of high-risk MDS patients at 1,816 and 288 CpG sites, respectively. The average methylation level was significantly higher in MDS than in controls and was associated with IPSS score (p=.06). The methylation pattern in MDS include aberrant hypermethylation of tumor suppressor genes (DCC, HIC1), and genes involved in DNA repair (OGG1, MGMT), cell cycle control (DBC1), development and differentiation (HOXA5, HOXB6) and apoptosis (ALOX12). Analysis of the most frequently aberrantly methylated genes identified several genes and we used FZD9 as a candidate TSG on chromosome 7 as an illustrative example for further analysis. Methylation level at this site was significantly predictive of survival in proportional hazards regression analysis (p=.002) and inversely correlated with expression of FZD9 mRNA. Using high density arrays to examine methylation status at locations most commonly associated with chromosomal lesions in MDS (chromosome 5, 7, 11, 13 and 20), we discovered 8 genes with functions and tissue expression patterns suggestive of involvement with MDS. Of these genes, 7 have previously been reported to be aberrantly hypermethylated in malignancy. Patients who did not exhibit hypermethylation at any of these sites were less likely to have developed AML (OR = 4.2, p = 0.074) and showed prolonged survival (p =.09). The absence of hypermethylation at one of the 8 pathognomonic sites was significantly predictive of survival (p=.05). In conclusion, genetic silencing by hypermethylation can produce molecular phenotypes identical to loss of function mutations and deletion of genetic information; however, unlike genetic lesions, epigenetic lesions are more common and are reversible by hypomethylating therapy. Development of a predictive algorithm based on methylation data will allow targeted therapy with epigenetic therapies.