A Pre-Leukemic DNA Methylation Signature in Healthy Individuals at Higher Risk for Developing Myeloid Malignancy

Abstract

Purpose: DNA methylation alterations are widespread in acute myelogenous leukemia (AML) and myelodysplastic syndrome (MDS), some of which appear to have evolved independently of somatic mutations in epigenetic regulators. Although the presence of somatic mutations in peripheral blood can predict the risk of development of AML and MDS, its accuracy remains unsatisfactory.

Experimental design: We performed global DNA methylation profiling in a case control study nested within the Singapore Chinese Health Study to evaluate whether DNA methylation alterations were associated with AML/MDS development. Targeted deep sequencing and methylated DNA immunoprecipitation sequencing (MeDIP-seq) were performed on peripheral blood collected a median of 9.9 years before diagnosis of AML or MDS, together with age-matched still-healthy individuals as controls.

Results: Sixty-six individuals who developed AML or MDS displayed significant DNA methylation changes in the peripheral blood compared with 167 age- and gender-matched controls who did not develop AML/MDS during the follow-up period. Alterations in methylation in the differentially methylation regions were associated with increased odds of developing AML/MDS.

Conclusions: The epigenetic changes may be acquired independently and before somatic mutations that are relevant for AML/MDS development. The association between methylation changes and the risk of pre-AML/MDS in these individuals was considerably stronger than somatic mutations, suggesting that methylation changes could be used as biomarkers for pre-AML/MDS screening.

Fig.1.. Differences in methylation profile between pre-AML/MDS cases and controls.

(a) PCA plot showing visible separation of global methylation profiling between controls and pre-AML/MDS. (b) Heat map of top 292 DMRs identified as methylation signature in the peripheral blood between pre-AML/MDS and healthy control individuals. Significantly hypomethylated regions in pre-AML/MDS cases compared to controls are in blue and significantly hypermethylated regions in red. (c) Volcano plot displaying a bias of DMRs towards hypomethylation of pre-AML/MDS cases compared controls. (d) Volcano plot displaying no significant differences in methylation between pre-MDS and pre-AML samples.

Fig.2.. Association of DMRs with AML/MDS development in healthy individuals.

(a) Forest plot of 20 representative DMRs displaying significant odds of developing AML/MDS. (b) Representative DMRs (2 hypomethylated, left panel; 2 hypermethylated, right panel) showing significant difference in methylation expression between pre-AML/MDS cases and controls, clear group separation of pre-AML/MDS from controls indicated by high AUC values, and significant difference in risk of AML/MDS development.

Fig.3.. Risk association of CH mutations with AML/MDS development in healthy individuals.

(a) Frequencies of recurrent mutations detected in both controls and pre-AML/MDS cases. Left panel - mutational heat maps; right panel - comparing the mutation frequency of each gene between pre-AML/MDS and the control group. (b) Violin plot depicting significantly higher mutation burden in pre-AML/MDS (red) compared to controls (blue). (c) Forest plots of mutations displaying the odds ratios of developing AML/MDS. (d) Kaplan-Meier curves for AML/MDS-free probability based on mutation status of DNMT3A, TET2, TP53 or any mutation(s) versus their corresponding wildtypes.

Fig.4.. Associations between methylation profile and mutations.

(a-c) Methylation profile alternations between (a) samples with and without any CH mutations, samples with (b) DNMT3A mutations or (c) TET2 mutations and their corresponding wildtype samples in controls (left panel) or pre-AML/MDS cases (right panel), respectively. In controls, cutoff values for fold change and adj p were set to be 2 and 0.01, respectively, while reduced cutoff values (fold change: 1.5 & adj p: 0.05) were applied in pre-AML/MDS cases to get reasonable numbers of differential methylation sites. (d) Differential methylation between pre-AML/MDS cases and controls for wildtype (left panel) or mutated (right panel) samples. The cutoff values for differential methylation (fold change: 2.0 & adj p: 0.01) were applied. (e) Distributions of the relative contributions of age, mutations and DMRs across all possible model combinations, showing generally significantly higher contributions of DMRs in modelling risk of AML/MDS development. (f) Two examples of contributions of the three factors in two combinations. (g) Distributions of hazard ratios obtained from multivariate Cox regression models across all possible model combinations of age, mutated genes and DMRs in log₂ scale, and (h) Distributions of the corresponding p values in −log₁₀ scale. In (g,h), hypomethylated DMRs in pre-AML/MDS cases were colored in red, and hypermethylated ones in pink.