A Dynamic Epigenetic Landscape in Genetic and Immunophenotypic Stable Pediatric AML

The majority of pediatric acute myeloid leukemia (AML) patients achieve complete remission. However, after this initial response to therapy, relapse occur in around forty per cent. The failure of therapy has been thought to be due to drug resistance resulting either from accumulating mutations, or already a characteristic of leukemic stem cells present at diagnosis. These mechanisms are shown to be true in part of the patients, however in quite a number of these relapses other mechanisms must play a role. In adult leukemias high epigenetic heterogeneity is associated with worse clinical outcomes. Remarkably, the DNMT3, TET and IDH mutations often found in adults are almost non-existent in children, despite the fact the leukemias are phenotypically very alike. We tested the hypothesis that, apart from genetic heterogeneity, epigenetic heterogeneity contributes to therapy resistance and thereby initiation of relapse.

Methods: We performed flow cytometric immunophenotypic analyses, whole exome sequencing and epigenome sequencing (by enhanced reduced representation bisulfite sequencing) on bone marrow (BM) samples taken at time of diagnoses, within two weeks after achievement of CR and at time of relapse of three selected patients (diagnosed between 2007-2014, treated according to the Dutch-Belgian DB-AML-01 protocol). Of each patient, apart from whole diagnosis sample and whole relapse sample (all with a minimum of 90% blast cells), the MRD sample was FACS sorted to select for blast cells.

Results: Leukemia associated immunophenotypes phenotype found at time of diagnosis remained the same over time with no emerging new immunophenotypes coinciding with emergence of relapses. We further evaluated whether changes in variant allele frequency (VAF) of mutations between diagnosis and subsequent time points could be linked to the development of relapse. Generally, VAF numbers decreased or remained stable over time, with no new mutations arising that characterized relapse. None of the patients had a mutation in DNTM3A, TET or IDH genes, accounting for any epigenetic dysregulation.

In contrast to the findings for immunophenotypes and genetics, extensive epigenetic changes were found at the different time points. In samples taken within two weeks after reaching complete remission, a median of 23,665 (range 19,303-28,808) differentially methylated regions (DMRs) was found compared with the diagnosis samples. When comparing the corresponding relapse sample with this CR/MRD sample, another set of DMRs was found: median 19,625 (range 14,680-30,410). In all three patients, the majority of exons, intergenic, intron and promoter regions are uniquely hyper or hypo methylated at time of relapse (median of 88%, range 81%-95%), with a small proportion already differentially methylated at time of remission (median 11%, range 4%-18%). We hypothesize that these changes in epigenetic status occurring after CR, possibly contributes to relapse. Pathway analysis will probably give clues for understanding the pathophysiology and hopefully new targets for treatment.

Conclusion: Our preliminary data suggest that epigenetic regulation may play a vital role in development of relapse. Ultimately, a better understanding of the role of epigenetics could open new therapeutic approaches for pediatric AML.