Gain-of-function studies using ectopic expression of RUNX1/ETO in t(8;21)-negative leukemic and non-leukemic cells have yielded insights into the molecular mechanisms of initiating leukemogenesis,6, 46, 52, 53, 54
whereas loss of function studies in actual leukemic t(8;21) cells (including the Kasumi-1 cells described here) inform about the role of RUNX1/ETO in maintaining leukemia.24, 25, 36, 39
Our experiments and novel integrative data analyses demonstrate a substantial concordance between these two approaches, indicating that integral parts of the t(8;21)-specific leukemia-initiating program are also required for maintaining the leukemic phenotype. In both primary cells and cell lines, RUNX1/ETO binds genes associated with the control of the cell cycle and cell structure. Concordantly, siRNA-mediated depletion of RUNX1/ETO affects transcriptional programs associated with myeloid differentiation, proliferation and self-renewal, in addition to those promoting cell cycle progression and DNA synthesis.
Our siRNA knockdown shows that RUNX1/ETO binding leads to large-scale, but reversible, alterations throughout the epigenome. H3K9Ac at RUNX1/ETO-binding sites was mostly increased after knockdown, which is consistent with a repressive role of the fusion protein at these sites. However, RUNX1/ETO depletion was also associated with upregulation of gene expression. In a recent study, RUNX1/ETO was shown to also regulate gene activation in a p300-dependent manner.20
Concurrent with this observation, when combined with gene expression data, RUNX1/ETO depletion concurred with strikingly complex regulatory patterns, with an increase of H3K9Ac and RNA Pol II binding alterations being associated with both upregulated and downregulated genes. An illustrative example for this notion is the CD34
locus (Supplementary Figure 6C
, lower panel), which contains an intragenic RUNX1/ETO site and is downregulated by RUNX1/ETO depletion, but shows increased histone acetylation at the main promoter. At the WT1
locus, RUNX1/ETO binds to an element downstream of the main start site that contains a bidirectional promoter driving an alternative WT1
transcript and an inhibitory antisense intronic transcript.55
RUNX1/ETO may therefore maintain WT1
expression by repressing the expression of non-coding RNAs.55
This is of significant clinical interest because the level of WT1
transcription is a prognostic factor in leukemia diagnosis.56
Moreover, we also found widespread siRNA-mediated changes in H3K9Ac at RUNX1/ETO peaks that were not associated with alterations of steady-state mRNA levels. These elements may represent sites that respond to the upregulation of myeloid regulators such as C/EBPα and prime genes for the onset of myeloid differentiation. Follow-up of these observations is outside of the scope of this study, but our datasets provide a wealth of resources for experiments unravelling the mechanistic details of such changes.
Another important result from this study is our finding that the depletion of RUNX1/ETO and subsequent cell differentiation is associated with a redistribution of RUNX1-binding activity throughout the genome. We observed a large number of new binding sites distinct from those previously bound by RUNX1/ETO. The induction of myeloid differentiation after RUNX1/ETO depletion therefore involves not only loss of repression, but also increased recruitment of transcriptional activators to additional sites. Currently, we do not know how RUNX1 is directed to new sites, but it is likely that this involves the interaction of RUNX1 with other transcription factors whose activity is altered by RUNX1/ETO, such as C/EBPα and PU.1.15, 57
Our observation of a differential binding of RUNX1 is also important in the context of leukemogenesis in general because it implies that mutations of RUNX1 that are widespread in leukemogenesis and cause specific disease phenotypes may differentially affect alternate subsets of genes depending on how the interaction with cooperating transcription factors is altered at each gene.58, 59
Therefore, one of the future challenges in leukemia research will be to unravel the differential activity of transcription factors in a system-wide manner and to model the regulatory consequences of such differences for each specific type of leukemia.
In conclusion, our study demonstrates that epigenetic alterations mediated by RUNX1/ETO are reversible at a global scale by solely interfering with its function, emphasizing the feasibility of targeted therapeutic approaches either using siRNA or small molecules.