Acute myeloid leukemia (AML) is a clinically and biologically heterogeneous hematologic malignancy characterized by various genetic alterations. Currently, DNA methylation patterns were reported to be associated with molecular subtypes, chromosomal abnormalities, gene fusion, and prognosis in AML. Furthermore, previous study reported that aberrant cancer-associated DNA hypermethylation targets CpG islands characterized by bivalent chromatin in human embryonic stem cells (hESCs), and the bivalent chromatin signature in hESCs was a key determinant of the instructive program for aberrant DNA methylation. Thus, we analyzed genome-wide DNA methylation in 64 pediatric patients with AML to reveal its association with clinical features, genetic alterations, and prognostic impact.
Between 2006 and 2010, 443 pediatric patients with de novo AML (0-17 years) participated in the Japanese AML-05 trial conducted by the Japanese Pediatric Leukemia/Lymphoma Study Group. Of these, 64 patients were enrolled in this study. The cytogenetic features of 64 patients were as follows: normal karyotype, 28; RUNX1-RUNX1T1, 8; KMT2A rearrangement, 15; complex karyotype, 6; and other cytogenetics, 7. This cohort included 15 patients with FLT3-internal tandem duplication (ITD), 8 with CEBPA biallelic mutations, 5 with high MECOM (EVI1) expression, and 17 with high PRDM16 (MEL1) expression. We performed genome-wide DNA methylation analysis using Infinium MethylationEPIC BeadChip (Illumina) in 64 pediatric patients.
Results and Discussion
824,848 methylation sites per sample were analyzed in 64 pediatric patients with AML. To capture DNA methylation differences across samples, we selected 567 CpG sites which showed most variable methylation values between 64 individuals such as standard deviations across samples were more than 0.3. The unsupervised hierarchical clustering of DNA methylation data from 567 CpG sites generated 4 clusters (clusters 1-4) with distinct molecular and clinical characteristics. Cluster 1 or 2 was the lowest or highest methylation level, respectively. Clusters 3 and 4 showed intermediate methylation level. Cluster 1 was characterized by RUNX1-RUNX1T1 and KMT2A rearrangement with low MECOM expression, which are known as favorable prognostic factors. Clusters 2 and 4 were composed of patients with the molecular features showing adverse outcome such as FLT3-ITD, KMT2A-PTD and/or normal karyotype with high PRDM16 expression. Interestingly, KMT2A rearrangement with high MECOM expression, considered as the adverse prognostic factor, were included in clusters 2 or 4. As for KMT2A rearrangement, nine of 15 patients with KMT2A rearrangement harbored KMT2A-MLLT3. Of these, five of nine classified into the hypomethylation group, and all five patients had no event. On the other hand, remaining four patients with KMT2A-MLLT3 all relapsed. All patients with normal karyotype with CEBPA biallelic mutations considered as the favorable factor were found in cluster 3. When we focused on CpG sites with significant difference in their methylation values between patients with and without FLT3-ITD, 15 FLT3-ITD patients were divided into two clusters (clusters A and B) by the hierarchical clustering. Remarkably, 8 FLT3-ITD positive patients in cluster A showed significantly worse overall survival (OS) and event-free survival (EFS) when compared with those in cluster B (5-year OS, 13% vs. 100%, P = 0.002; 5-year EFS 0% vs. 86%, P < 0.001). Next, 244 CpG sites significantly associated with PRDM16 expression were extracted to investigate the relationship between PRDM16 expression and DNA methylation profiles. Interestingly, patients with high and low PRDM16 expression showed distinct methylation pattern, respectively. Furthermore, most of hypermethylated sites gene were PRDM16 gene body in patients with high PRDM16 expression and located at important regions which were the targets of repressed polycomb in reference cells. As for 567 CpG sites which were used for the unsupervised hierarchical clustering, 168 of 567 (30%) CpG sites colocalized at bivalent promoter regions in reference leukemic blast cells, and the hypermethylation of bivalent promoter regions tended to be related to worse outcome. These results indicate DNA methylation plays key role for leukemogenesis and is remarked as a novel biomarker to predict prognosis.
Ogawa:ChordiaTherapeutics, Inc.: Consultancy, Equity Ownership; RegCell Corporation: Equity Ownership; Kan Research Laboratory, Inc.: Consultancy; Asahi Genomics: Equity Ownership; Qiagen Corporation: Patents & Royalties; Dainippon-Sumitomo Pharmaceutical, Inc.: Research Funding.
Asterisk with author names denotes non-ASH members.