Somatic Mutations and DNA Hypermethylation at Enhancers and Promoters Identify Distinct Subtypes within Lower-Risk Myelodysplastic Syndromes

Myelodysplastic syndromes (MDS) are a group of heterogeneous disorders caused by the accumulation of somatic mutations in the hematopoietic stem and progenitor compartment. Besides del(5q), SF3B1 or TP53 mutations, referred to as defining genetic abnormalities, mutation patterning hardly structured the classification of MDS. Mutations in epigenetic factors occur early in the development of clonal hematopoiesis leading to precocious alterations of DNA methylation implicated in oncogenesis. Convergent DNA methylation patterns related to both mutations and microenvironment imprinting may induce specific gene expression profiles and contribute to phenotypic variations. To refine a pathophysiological classification of lower-risk MDS, we combined genetic profiling to DNA methylation and transcriptome data. Methods: We enrolled 543 lower-risk treatment-naïve MDS patients at diagnosis and performed DNA-sequencing of 24 genes. DNA methylation and RNA-sequencing data were generated in a representative subset of 175 cases with available material (75 MDS-SLD/MLD, 40 MDS-EB1, 53 MDS-RS-SLD/MLD, 5 MDS del(5q), 2 MDS-U) and 7 age-matched healthy controls. IPSS-R was very low, low or intermediate in 87.4%. During a median follow-up of 2 years, 16% of MDS cases progressed to AML (MDSp). Infinium EPIC 850K array allowed after filtering the examination of methylation of 723,612 CpG sites with enrichment at enhancers and promoters. Differentially methylated regions with differentially methylated CpG sites (DMR-CpG) between cases and controls were defined with a minimum of 2 probes, a Benjamini-Hochberg-adjusted P-value cut-off of 0.05, a false discovery rate of 0.001 and a mean Δβ-value >20% at CpG sites. Consensus motifs for transcription factors were identified using Homer. Results: Unsupervised clustering on genetic profiling of 543 patients identified distinct subsets of lower-risk MDS with different clinical features: one cluster (group A) was enriched in del(5q), three clusters were characterized by SF3B1 mutations, either isolated (group B) or associated with DNMT3A (group E) or TET2 (group F), while the two remaining clusters were enriched in either complex pattern of mutations (group C) or SRSF2 mutations (group D). Genetic clusters showed significantly different clinical outcomes, clusters C, D and E having lower overall survival and higher risk of progression to high risk MDS or AML compared to other clusters. In the subset of 175 cases, a total of 2,953 unique DMR