t(8;21)(q22;q22)RUNX1-RUNX1T1 is a recurrent somatic lesion detected at diagnosis in approximately 12-15% of children with acute myeloid leukemia (AML). Children with this isolated translocation are usually considered at standard risk, but our last multicenter trial revealed a higher than expected cumulative incidence of relapse for these patients1. Genetic and epigenetic heterogeneity is emerging as a fundamental property of AML in the context of the clonal architecture dynamic evolution. In view of this observation, we hypothesized that within t(8;21) patients there may coexist a complex mosaic of cells containing combinations of the same genetic t(8,21) lesion together with different epigenetic variants, and that epigenetic complexity may play a crucial role in predisposing patients to relapse. The importance of the identification of molecular markers distinctive of t(8,21)-rearranged patients prone to develop relapse could be instrumental to improve their cure rate.

We performed high throughput DNA methylation profiling (RRBS-seq) and integrated results with gene expression profiling (Affymetrix HTA 2.0) of 16 isolated t(8;21) AML samples collected at diagnosis, and analyzed data by comparing patients who did or did not experience relapse. We applied a logistic regression algorithm to identify differentially methylated regions (DMRs) considering a minimum change in methylation level of 25%. We validated results in a proteome context by reverse phase protein array (RPPA) in an independent cohort of 35 t(8;21) AML patients.

DNA methylation profiling analysis identified 337 DMRs able to correctly predict t(8;21) patients who did relapse from those who did not. In particular, 23 DMRs (7%) were located at promoters, while most of them were equally distributed between intronic (48%) and exonic (45%) regions. Globally, we found hypomethylated DMRs being significantly enriched in relapsed patients, in particular in repetitive elements regions of the genome (LINE, SINE, DNA transposon: 38.9% vs 52.4%; p<0.01), supporting an enhanced transposable elements transcription contributing to cancer genomic instability. DMRs clustering analysis correctly divided t(8,21) patients according to their risk of experiencing relapse, independently of their different localization (at promoters, exons or introns), revealing that DNA methylation profiling has a predictive role for identifying patients with worse event-free survival.

We then considered the role of methylation over gene expression and found a weak correlation between DMRs (mostly at promoters) and their associated gene levels (14.5% of DMRs with an inverse correlation r <-0.4, p<0.05). To better understand the role of DMRs and transcriptional regulation, we searched for associations between DMRs and chromatin modification patterns. DMRs were enriched at regulatory regions; in particular, we found hypermethylation in promoter and enhancer regions, while hypomethylation was found in repressed chromatin regions (p<0.05). Looking at the transcription factors (TFs) binding sites within the DMRs, we identified that at hypermetylated DMRs the most represented TFs were E2F1 and HDAC1, suggesting they might be almost transcriptional silenced. By contrast, MAFK and FOXA2 binding sites were enriched at hypomethylated sites, suggesting their enhanced activity in relapsed patients as compared to the non-relapsed ones. Finally, we interrogated gene ontology for DMRs-associated genes and deregulated genes found by GEP, showing a significant enrichment for pathways involved in cell adhesion and cytoskeletal organization. Proteome analysis by RPPA validated these pathways being aberrant activated (global test p<0.01) in an independent cohort of t(8;21)-rearranged patients, and supported the ongoing in vitro experiments in t(8;21) cell lines to define candidates genes involved in the pathophysiology of t(8,21) relapse.

These data show that the methylation signature may be considered a novel, emerging diagnostic tool making possible to better stratifying t(8,21)-rearranged patients through the identification, already at diagnosis, of those who are prone to relapse . Preliminary data of functional analysis suggest that epigenome of t(8;21) blasts may control cell adhesion properties at bone marrow niche and treatment response, contributing to patients relapse.

1 Pession A, Blood. 2013;122(2):170-8.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.