Mutations in IDH1, IDH2, and TET2 occur in more than 30% of AML patients, and are thought to contribute to leukemogenesis by disrupting TET2-mediated DNA demethylation, either directly or indirectly. While IDH1/2 and TET2 mutations are both predicted to result in hypermethylation, the specific methylation changes associated with these mutations are poorly understood. Additionally, we recently showed that CpG island (CGI) hypermethylation is common across virtually all AML subtypes, and it has been yet unclear whether IDH or TET2 hypermethylation is distinct from these 'generic', AML-associated changes. To investigate genome-wide methylation in IDH and TET2 mutated AML samples, we performed whole genome bisulfite sequencing (WBGS) on 15 primary AML samples with canonical IDH mutations (IDH1R132C, N=6; IDH1R132G, N=1; IDH2R140Q, N=7; and IDH2R172K, N=1) and 9 samples with biallelic loss-of-function mutations in TET2. These data were analyzed with WGBS data from primary CD34+ hematopoietic stem/progenitor cells (HSPCs) from 6 normal adult bone marrow donors, and 39 additional primary AML samples (representing 9 AML mutational categories) to define the methylation phenotypes specifically associated with IDH and/or TET2 mutations. Differential methylation analysis of each AML subtype vs. HSPCs identified a mean of 8,622 (53,73-13,773) AML-associated Differentially Methylated Regions (DMRs) in each mutational category (mean size: 986 bp), of which 6% to 97% were hypermethylated in the AML samples vs. HSPCs. IDH-mutant AML samples were the most hypermethylated subtype, with 97% of the 6,392 identified DMRs having increased methylation compared to HSPCs. The TET2-mutant samples were hypermethylated at fewer loci, representing 71% of the 5,892 identified DMRs vs HSPCs. Analysis of all AML subtypes demonstrated that 55% (16,275/29,534) of identified DMRs were shared across multiple subtypes, with more pronounced overlap among hypermethylated DMRs (>70%). To distinguish mutation-specific changes in IDH and TET2 mutated AMLs from commonly hypermethylated loci across AML subtypes, we conducted a joint analysis of either IDH or TET2 mutant AMLs with data from all other AML samples and normal HSPCs. We performed hierarchical clustering of the methylation values at each DMR, and used the clustering topology to identify DMRs where the single outlier branch represented either the IDH or TET2 AML samples. We identified 2,024 DMRs (mean size: 998 bp) specific to IDH mutant AML samples, of which 100% were hypermethylated with respect to normal HSPCs (mean DMR b value of 0.60 vs 0.35 in HSPCs). IDH outlier DMRs were enriched for functional genomic elements relative to the set of commonly hypermethylated loci, including CGIs (65% of DMRs; 2.4-fold enrichment) and potential enhancers marked by H3K27ac in HSPCs (82% of DMRs; 1.4-fold enrichment); overlap with promoter regions was reduced (20% of DMRs; 0.77-fold decrease). Outlier analysis of TET2 mutated samples demonstrated that most changes were also present in other AML types, with only 93 unique DMRs (mean size 552 bp), and a subtler hypermethylation phenotype (83% of DMRs were hypermethylated, with a mean b value of 0.51 vs. 0.32 in HSPCs). TET2 DMRs were also enriched for overlap with CGIs with respect to background, but to a lesser extent than IDH outliers (52% of DMRs; 1.9-fold). Intersection of IDH and TET2-associated DMRs showed that TET2-specific DMRs represented a subset of IDH hypermethylated regions, with 60% of TET2 DMRs overlapping a DMR independently identified in the IDH samples. Conversely, the vast majority of IDH outliers (97%) were identified only in IDH mutated AMLs. This comprehensive, genome-wide analysis demonstrates that IDH mutations result in a clear hypermethylation phenotype in primary AML samples that is far more expansive than the observed TET2-associated methylation changes. IDH-specific methylation changes were enriched for CGIs located outside of promoters and were distinct from the AML-associated CGI hypermethylation that occurs in other AML subtypes. TET2 mutations were associated with far fewer hypermethylated DMRs, and only a small fraction of mutation-specific changes occurred in bothIDH and TET2 samples. Our results demonstrate that the methylation phenotypes of IDH and TET2 mutant AML samples do not phenocopy, and suggest that IDH-associated hypermethylation must have a TET2-independent component. Disclosures No relevant conflicts of interest to declare.
Read full abstract