Myelodysplastic Syndrome Patients Show Mutation-Specific DNA Methylation Patterns

Introduction: Myelodysplastic syndromes (MDS) are characterized by recurrent mutations, which contribute to the classification of patients into prognostic subgroups. Besides genetic lesions, MDS cells show aberrant DNA methylation, which is used as a therapeutic target for pharmacological DNA demethylation. Genetic and epigenetic aberrations are tightly linked: The most frequently mutated genes in MDS are key players in epigenetic pathways, including DNMT3A and TET2, which directly regulate DNA methylation. Additionally, other epigenetic master regulators (EZH2 and ASXL1) are recurrently mutated. Recent data in myeloid malignancies also suggested that cases with mutations in splicing factors (e.g. SRSF2) are characterized by a specific DNA methylation signature.

Aim: To describe the role of recurrent mutations in MDS as prognostic markers and elucidate the mutation-specific DNA methylation pattern.

Patients and Methods: By next-generation sequencing 786 bone marrow samples were analyzed for mutations in 14 genes. Patients (488 males, 298 females) had a median age of 73 (23- 91) years. The cohort was balanced according to WHO classification subtypes, karyotype risk groups and IPSS-R risk groups. For DNA methylation analysis, 196 samples were analyzed on the 450 K platform (Illumina). Differences between groups were calculated for individual probes by t-test with IMA script; differences with an adjusted p-value <0.01 were considered significant.

Results: We recently suggested four prognostic subgroups in MDS (Haferlach et al.; Leukemia; 2014), defined by combining the mutation status of 14 genes (including the aforementioned ASXL1, EZH2 and TET2) with clinical parameters. First, we re-analyzed the data set with an extended follow-up time (median 45 vs. 32 months) and confirmed separation of the subgroups regarding overall survival (OS, p<0.001) in 786 patients.

Of our initial cohort, we now selected 196 cases for a comparison of DNA methylation patterns in patients with and without different mutations. In 70/196 cases (36%) with at least one TET2 mutation we identified 2% aberrant DNA methylation. In line with the function of TET2 in DNA demethylation, 90% of aberrant DNA methylation was hypermethylation. Importantly, hypermethylation almost exclusively located outside of promoters (only 3% of fraction expected by array design), and was overrepresented by twofold in regions with enhancer function. Only 6% of hypermethylated CpGs were located in CpG Islands (expected background: 31%). In contrast, in 31/196 (16%) DNMT3A mutated cases, only 0.02% abnormal DNA methylation was observed and 94% was hypomethylation, as expected by the role of DNMT3A to regulated DNA de novo methylation.

In 42/196 (21%) of ASXL1 mutated cases, a methylation difference of 2% was identified with 67% hypermethylation, whereas in 13/196 (7%) EZH2 mutated cases only minute recurrent DNA methylation differences were observed (0.01%). Next, we analyzed the effect of mutated splicing factors on DNA methylation. In both 67/196 (34%) SF3B1 and 39/196 (20%) SRSF2 mutated cases, we identified strong DNA methylation differences (>3%), however with opposite direction (SF3B1: 99% hypomethylation, SRSF2: 91% hypermethylation). Finally, in 58/196 (30%) patients with an aberrant karyotype we could identify that 1% of probes showed DNA methylation changes, which was almost exclusively hypermethylation.

Conclusion: 1) We confirmed the prognostic capacity of our previously suggested scoring model including the mutation status of 14 genes in MDS. 2) The mutation differences resulted in a mutation specific epigenetic signature regarding the degree, direction and localization of aberrant DNA methylation. 3) For the most frequently mutated gene in MDS, TET2, DNAhypermethylation is found outside of the classically analyzed promoter regions and CpG Islands, and enriched in enhancer regions. 4) MDS is treated with drugs that alter DNA methylation. The understanding of mutation specific DNA methylation patterns would allow to choose the right genomic loci to monitor the DNA methylation reduction under treatment. This would be a key step toward a deeper understanding of drug function or response.