Context-dependent recruitment of co-factors directs RUNX1 activity in normal and leukemic cells

Abstract Aberrant transcription factor (TF) activity is a major driver of diseases such as leukemia. TFs bind to regulatory elements such as promoters and enhancers, where they can drive activation or repression of genes. Many TFs have dual activity, where their effect on gene expression depends on the context, but it is unclear what the underlying mechanisms are. RUNX1 is a key hematopoietic TF, essential for the emergence of hematopoietic stem and progenitor cells (HSPCs) in the embryo and for the differentiation and maturation of hematopoietic cells throughout life. The RUNX1 gene is also a frequent target of genetic alterations in acute myeloid (AML) and acute lymphoid (ALL) malignancies (Sood et al Blood 2017 doi: 10.1182/blood-2016-10-687830). Despite this importance, RUNX1 activity is still not completely understood. To date, identified interaction partners of RUNX1 include epigenetic modifying enzymes with an activating role such as the lysine acetyltransferases P300 and MOZ, Trithorax-group (TrxG) proteins such as KMT2A, and components of the nucleosome-remodelling complex SWI/SNF; as well as co-repressors such as histone deacetylases and Polycomb (PcG) Repressor Complex 1 (PRC1; Yu et al, Mol Cell 2012 doi: 10.1016/j.molcel.2011.11.032). Overall, this implicates RUNX1 in both activation and repression of gene targets, but it is unclear how RUNX1 mediates this dual role at specific loci and if this has any implications for RUNX1 mutant leukemias. To more precisely delineate the role of RUNX1 in gene regulation and complex assembly, we used an in vivo tethering system to observe its effects on a neutral chromatin context (devoid of other TFs and common histone modifications). We have previously used this system to show that MYB, another important hematopoietic TF, drives novel enhancer emergence and is primarily an activator of transcription. In contrast, we found that RUNX1 is sufficient to recruit both co-activators and co-repressors, including TrxG and PRC1 complexes, with no apparent preference. Thus, we next wanted to determine how this dual activation/repression recruitment role for RUNX1 functioned in normal gene regulation. To better define RUNX1 interactions at endogenous gene targets, we performed an unbiased correlation analysis at RUNX1 ChIP-seq peaks in the RUNX1-dependent KMT2A::AFF1 ALL cell line SEM. This showed a positive correlation between the binding of RUNX1 and co-activators and active histone modifications (e.g. H3K27ac, H3K4me3). However, although RUNX1 primarily associates with activation complexes in these cells, we also found a positive correlation with binding of the proteins RING1B and KDM2B, components of the repressive PRC1 complex. This suggests that RUNX1 may act as both an activator and repressor in ALL. To test the impact of RUNX1 recruitment on these complexes, we overexpressed RUNX1 in mouse embryonic stem cells (that have no endogenous RUNX1 expression) and performed RUNX1 siRNA knockdowns in KMT2A::AFF1 SEM cells. Overall, we found that RUNX1-mediated recruitment of co-activators and co-repressors is dependent on the factors and histone marks that are already present. Further, we show that C-terminally truncated RUNX1 mutant proteins, including the RUNX1-RUNX1T1 fusion protein in AML, fail to recruit PRC1, and that acquisition of PcG mutations in RUNX1-RUNX1T1 AML (associated with poor prognosis disease) results in de-repression of critical PcG target genes. This indicates that RUNX1 is required to stabilise and strengthen PRC1 binding. Our overall model is that, rather than directing transcription regulation, RUNX1 binding leads to multivalent interactions that “amplify” existing active or repressive signals. Taken with our other work, this suggests that some TFs such as MYB may be primarily activators, while a TF such as RUNX1 functions more agnostically, by stabilising pre-existing activation or repression states. Our study also provides a mechanistic understanding for how a TF is able to “switch” between co-factors, to selectively drive activation or repression at target sites; and how the disruption of interactions between a TF and its co-factors through disease-associated mutations can promote oncogenesis. This may provide an important paradigm for how the same master regulator has the potential to be either a tumour suppressor or an oncogene, depending on the chromatin context in different cell types and cell states.