RUNX1 colludes with NOTCH1 to reprogram chromatin in T cell acute lymphoblastic leukemia

Summary Runt-related transcription factor 1 (RUNX1) is oncogenic in diverse types of leukemia and epithelial cancers where its expression is associated with poor prognosis. Current models suggest that RUNX1 cooperates with other oncogenic factors (e.g., NOTCH1, TAL1) to drive the expression of proto-oncogenes in T cell acute lymphoblastic leukemia (T-ALL) but the molecular mechanisms controlled by RUNX1 and its cooperation with other factors remain unclear. Integrative chromatin and transcriptional analysis following inhibition of RUNX1 and NOTCH1 revealed a surprisingly widespread role of RUNX1 in the establishment of global H3K27ac levels and that RUNX1 is required by NOTCH1 for cooperative transcription activation of key NOTCH1 target genes including MYC, DTX1, HES4, IL7R, and NOTCH3. Super-enhancers were preferentially sensitive to RUNX1 knockdown and RUNX1-dependent super-enhancers were disrupted following the treatment of a pan-BET inhibitor, I-BET151.


INTRODUCTION
Since the cloning of runt-related transcription factor 1 (RUNX1) over 30 years ago significant effort has been directed to understanding its functional role in hematopoietic and epithelial cancers where RUNX1 translocation, mutation, copy number gain, and over-expression have been described. [1][2][3] The RUNX family of transcription factors cf. DNA binding capacity to the heterodimeric core binding factor (CBF) transcription complex. 4 Members of the CBF complex are commonly perturbed during leukemogenesis and are frequent targets of translocation in T cell acute lymphoblastic leukemia (T-ALL), myelodysplastic syndromes (MDS), acute myeloid leukemia (AML), and B-cell acute lymphoblastic leukemia (B-ALL) [5][6][7] Missense mutations within the DNA binding domain of RUNX1 are recurrent events in T-ALL (10-15% cases) suggesting a tumor suppressor role. [8][9][10] This interpretation is complicated by the observation that RUNX1 is overexpressed in T-ALL and required for leukemic cell growth and survival in murine T-ALL models. [11][12][13] T-ALL accounts for 15% of pediatric and 25% of adult acute lymphoblastic leukemia cases. 14 T-ALL subgroups are characterized by the arrest of thymocytes at different stages of development and each subgroup can be classified by the over-expression or genetic activation of specific transcription factors. 15,16 Independent of T-ALL subgroups, more than 55% of T-ALL cases have recurrent gain-of-function mutations in NOTCH1 17,18 in the background of expressed RUNX1. 10 Following nuclear translocation, NOTCH1 is thought to co-bind enhancer elements with RUNX1 to drive expression of key NOTCH1 target genes in T-ALL. 11,19 Cooperativity between Runx and Notch appears to be conserved in Drosophila hemocytes, where Runx is required for Notch responsive enhancers to be competent to respond to Notch. 20 RUNX1 drives expression of key myeloid differentiation genes (e.g., PU.1, MPO, GM-CSF) and its recruitment to DNA induces an active chromatin state including acetylation of histone 3 lysine 27 (H3K27ac) and, in the context of gene promoters, trimethylation of histone lysine 4 (H3K4me3). 21-23 RUNX1 itself does not possess enzymatic activity but serves to recruit histone acetyltransferases (CBP/P300) and lysine methyltransferases (MLL) to deposit H3K27ac and H3K4me3 on nearby histones, respectively. 21,22 Interaction with CBP/P300 also leads to acetylation of RUNX1 that enhances its DNA binding activity. 24 RUNX1 recruitment to, and activation of, gene promoters provide a direct mechanism of transcriptional activation. However, two-thirds of the total (20,000-30,000) RUNX1 binding sites are located outside of promoters 19,25 and the contribution of RUNX1 to the chromatin landscape more broadly is not known. iScience Article Gain-of-function mutation of NOTCH1 is a defining feature of T-ALL 18 and is thought to act in part through disruption of polycomb repressive complex 2 (PRC2) mediated repression. 26 The NOTCH1 intracellular domain (ICN1) directly interacts with the H3K27me3 demethylase, JMJD3, and through its recruitment to NOTCH1 binding elements is thought to oppose PRC2 mediated transcriptional repression. 26,27 A tumor-suppressive role for PRC2 in T-ALL is reinforced by the observation of frequent loss-of-function mutations in the members of the complex. 27 But the specific role of NOTCH1 mediated H3K27me3 loss in T-ALL has yet to be established.
To examine the contribution of RUNX1 and NOTCH1 in the establishment of a pathogenic transcriptional signature in T-ALL, we performed an integrative analysis of epigenetic and transcriptional states following RUNX1 knockdown and NOTCH1 inhibition. Analysis of these data revealed that RUNX1 cooperates with NOTCH1 to establish active chromatin states and drive the expression of key T-ALL oncogenes. RUNX1 regulates H3K27ac occupancy genome-wide and is required to maintain a majority of super-enhancers in T-ALL, including its own intronic super-enhancer that drives a feedforward transcriptional loop. Bromodomain and extraterminal (BET) proteins (BRDT, BRD2, BRD3, and BRD4) are the epigenetic reader of H3K27ac modification and are themselves recruited to super-enhancers. 28 Treatment with the pan-BET protein inhibitor I-BET151 led to a broad reduction in H3K27ac levels, disrupted RUNX1 driven pathogenic super-enhancers, and down-regulated key RUNX1 and NOTCH1 target genes and proliferation of T-ALL cell lines. Our results indicate that RUNX1 plays a surprisingly broad role in the maintenance of H3K27ac density at super-enhancers and suggests that targeted disruption of super-enhancers through I-BET151 treatment may provide therapeutic benefit in RUNX1-driven malignancies.

RUNX1 contributes to H3K27ac occupancy genome-wide
To quantitate histone modification changes following RUNX1-KD, we called enriched regions (peaks) from the ChIP-seq datasets using FindER 29 and generated a union of peaks from RUNX1-CTL and RUNX1-KD samples. Normalized read densities were used to identify the peaks that show a 2-fold change between RUNX1-CTL and RUNX1-KD. Among the six histone modifications, H3K27ac demonstrated the most significant reduction following RUNX1-KD with 50% of peaks losing H3K27ac density ( Figures S1E and 1D). In contrast to the genome-wide loss observed for H3K27ac (marking active enhancers), occupancy of H3K4me1 (primed enhancers) remained largely stable following RUNX1-KD ( Figure 1D). RUNX1-dependent H3K27ac peaks were highly enriched with RUNX1 ChIP-seq signal compared to the peaks gained  iScience Article or stable upon RUNX1-KD ( Figure 1E). To replicate the genome-wide dependence of H3K27ac levels on RUNX1, we performed H3K27ac ChIP-seq following RUNX1-KD in an independent T-ALL cell line (CUTLL1) using a third shRNA (shRUNX1-90). Despite the variations observed in knockdown efficiencies of shRUNX1 constructs, there were significant overlaps (hypergeometric p value 0) of H3K27ac peaks lost using independent shRUNX1 constructs across two T-ALL cell lines ( Figures 1D, S2A, and S2B). The shRNA-associated H3K27ac density reduction varied across cell lines to a greater degree than within a single line. However, in all treatments a reduction in H3K27ac density was observed.

RUNX1 establishes oncogenic super-enhancers
Examination of the pattern of H3K27ac loss following RUNX1-KD in KOPTK1 revealed a contraction of H3K27ac density within the majority of peaks ( Figure 2A). The combination of contraction and concomitant peak loss of H3K27ac following RUNX1-KD was associated with a 46% reduction in the total number of super-enhancers ( Figure 2B). Super-enhancers were more sensitive to RUNX1-KD compared to typical-enhancers ( Figure S3E). Upon RUNX1-KD, 70% of the super-enhancer associated H3K27ac peaks showed R2-fold reduction of read density whereas there was 47% loss at classical enhancers. Genomic Regions Enrichment of Annotations Tool (GREAT) 33 analysis of genes proximal to the depleted super-enhancers showed enrichment in lymphocyte activation and differentiation pathways (Binomial False Discovery Rate, FDR = 0) ( Figure 2C). These included a set of known and suspected proto-oncogenic drivers (e.g., FGR, RUNX1, STAT5A, MYO7B, TMEM26, and CRTAM) in T-ALL which lost H3K27ac peaks at the adjacent super-enhancers and were down-regulated (FDR % 0.05) following RUNX1-KD ( Figures 2D and 2E). FGR is a member of the Src Family Tyrosine Kinase which is highly expressed in a significant proportion of AML patients and associated with poor prognosis. 34,35 We observed an 80% overall reduction in H3K27ac peaks (25/31) within a 70 kb window encompassing the FGR super-enhancer and FGR expression was significantly (FDR = 0.0009) down-regulated following RUNX1-KD ( Figures 2E and 2F).
RUNX1 itself drives an auto-regulatory feedforward circuit with other key hematopoietic transcription factors including GATA3, TAL1, and MYB in T-ALL. 25,36 Consistent with this we observed a 144.5 kb superenhancer within the first intron of RUNX1 that is conserved across T-ALL and AML cell lines and contains 15 binding sites for RUNX1 ( Figure 2G). RUNX1-KD reduced H3K27ac levels 2.5-fold within the RUNX1 super-enhancer and reduced H3K4me3 density at both RUNX1 promoters. Collectively, these results suggest iScience Article that RUNX1 plays an essential role in maintaining super-enhancers and the expression of proto-oncogenic targets of RUNX1 is sensitive to the loss of RUNX1-dependent super-enhancers.
RUNX1 and NOTCH1 cooperate to establish active promoter states NOTCH1-mediated JMJD3 recruitment is essential for leukemic growth by modulating H3K27me3 at the promoters of proto-oncogenes. 26 NOTCH1 peaks are frequently co-occupied by RUNX1 peaks and RUNX1 is required for the expression of key NOTCH1 target genes in T-ALL models. 11,19 Based on these observations, we hypothesized that at the co-occupied peaks, NOTCH1 demethylates H3K27me3 through JMJD3 allowing RUNX1 to acetylate the demethylated H3K27 residue via recruitment of CBP/P300. To test this directly, we first examined the relationship of 28,156 RUNX1 and 13,986 NOTCH1 peaks determined by ChIP-seq in the CUTLL1 T-ALL cell line. 19 The majority (84%) of NOTCH1 peaks intersected with RUNX1 peaks and 76% of the co-occupied peaks were marked by active promoter modifications (H3K4me3 and H3K27ac) ( Figure 3A). RUNX1-KD led to a significant reduction in H3K27ac density around RUNX1 peaks whereas NOTCH1-INB led to an increase in H3K27me3 around NOTCH1 peaks ( Figures 3B and S4D), supporting a model of NOTCH1 dependent demethylation of H3K27me3. 26,27 At RUNX1 and NOTCH1 cooccupied peaks, RUNX1-KD reduced H3K27ac enrichment by 2-fold and NOTCH1 inhibition increased the H3K27me3 enrichment by 4-fold ( Figure 3C). The co-occupied peaks intersecting with promoters had relatively higher enrichment of NOTCH1 and JMJD3 compared to the distal regions and the gain of H3K27me3 following NOTCH1-INB was specifically observed at gene promoters ( Figures 3D and S4A). Thus, our analysis supports a cooperative model for NOTCH1 and RUNX1, where NOTCH1 drives the demethylation of H3K27me3 through JMJD3 recruitment to promoters of its target genes allowing for RUNX1-dependent deposition of H3K27ac and establishment of an active chromatin state.

RUNX1 regulates the Notch-dependent MYC enhancer element (N-ME)
In T-ALL, the oncogenic activity of NOTCH1 is dependent on MYC upregulation. 37 A cluster of enhancers named the Notch-dependent MYC enhancer element (N-ME) influences the expression of MYC in T-ALL through long-range enhancer-promoter interaction. 38,39 N-ME is composed of five super-enhancers with an average length of 50 kb within a span of 706 kb ( Figure 4C). NOTCH1 and RUNX1 both co-occupy MYC promoter and distal N-ME super-enhancers that were originally proposed as NOTCH1-dependent enhancers. 38,39 In support of previous findings, we observed a modest (1.8-fold on average) reduction in H3K27ac density at the super-enhancers embedded within the N-ME following NOTCH1 inhibition in KOPTK1 cells. However, this reduction was not evenly distributed across the N-ME despite NOTCH1 binding, with three out of five super-enhancers (e.g., SE-3, -4, and -5) disrupted and the largest super-enhancer (e.g., SE-2 of 97 kb) unchanged following NOTCH1 inhibition. In contrast, RUNX1-KD led to a more pronounced (3.1-fold on average) H3K27ac loss at the N-ME super-enhancers and disrupted all super-enhancers within N-ME. Loss of H3K27ac density at the MYC promoter and disruption of N-ME super-enhancers following RUNX1-KD were associated with a significant down-regulation of MYC at both transcript and protein levels in KOPTK1 ( Figures 4C-4E). H3K27me3 density increased following NOTCH1 inhibition at the MYC promoter and was associated with a down-regulation of MYC transcription ( Figures 4C and 4D).

RUNX1 driven super-enhancers are sensitive to I-BET151 treatment
Genes associated with super-enhancers are sensitive to perturbations of key components of the transcriptional machinery and chromatin modifiers including BET proteins occupying super-enhancers. 28, 40 The ability of oncogenic RUNX1 to establish super-enhancers raised the possibility of specifically targeting the RUNX1 driven super-enhancers using BET inhibitors. To test this directly, we used a pan-BET inhibitor (I-BET151) that has been shown to inhibit the expression of super-enhancer-associated genes in leukemia 41,42 and to deplete H3K27ac at the promoters of proinflammatory cytokines induced by b-glucan in monocytes. 43 In KOPTK1 cells, I-BET151 treatment (iBET-TRT) arrested cellular growth and proliferation in a dose-dependent manner ( Figures 5A and 5B) consistent with previous studies. 41,42,44 To perform H3K27ac ChIP-seq and RNA-seq, we used 0.5 mM of I-BET151 at which 50% of the KOPTK1 cells showed growth inhibition at 72 h of treatment. We used synthetic DNA-barcoded nucleosomes bearing distinct post-translational modifications as spike-in controls to measure antibody specificity and normalize histone modification density (HMD) at H3K27ac peaks. Analysis of the resulting ChIP-seq datasets revealed a broad reduction of H3K27ac enrichment following I-BET151 treatment where 13151 (40%) peaks reduced HMD R 2-fold (Figure 5C). H3K27ac peaks that were lost following I-BET151 treatment showed higher enrichment of BRD4 compared to the gained or stable H3K27ac peaks, suggesting those depleted H3K27ac peaks have a higher dependency on BRD4 and thus would be expected to be more sensitive to I-BET151 treatment ( Figure 5D). I-BET151 treatment displaces the BET proteins BRD2, BRD3, and BRD4 from chromatin. 44 BRD4 partners with P300 to promote histone acetylation 45 and P300 also facilitates BRD4 recruitment. 46 Thus, we hypothesize that in KOPTK1 cells I-BET151 mediated displacement of BET proteins disrupted the interaction with CBP/P300 and reduced H3K27ac HMD. A significant portion (61% (8,044/13,151); Fisher's exact test; p value 0) of depleted H3K27ac peaks following I-BET151 overlapped with RUNX1-dependent peaks suggesting RUNX1-dependent H3K27ac peaks are preferentially affected by I-BET151 treatment ( Figure 5E). The overlapping co-dependent H3K27ac peaks (n = 8044) show relatively higher enrichment of BRD4 and H3K27ac than the peaks lost uniquely by RUNX1-KD or I-BET151 treatment, and are co-occupied with RUNX1, BRD4, and P300 ( Figures 5F, S6A and S6B). This suggests that the cooperative activities of RUNX1, BRD4, and P300 are facilitating the increased acetylation of H3K27 residues making them sensitive to either RUNX1-KD or I-BET151 treatment. In support of the co-regulation of H3K27ac by RUNX1 and BET proteins, transcriptional programs altered between RUNX1-KD and I-BET151 treatment showed a significant overlap of 48 genes (Fisher's exact test; p value = 2e-16) including STAT5A, IL4, IRF8, MYO7B, ETV5, and FGR. (Figure 5G).
RUNX1 and BRD4 peaks more frequently co-occur within super-enhancers than the regions outside of super-enhancers ( Figure S6C) and super-enhancers showed reduced H3K27ac density following RUNX1-KD or I-BET151 treatment ( Figure S6D). Following I-BET151 treatment, super-enhancers associated proto-oncogenes (e.g., STAT5A, IL4, ETV6, BCL11A, CD79A, CDK6, ERG, and FGR) known to be involved in leukemogenesis and disease progression, were down-regulated ( Figure 5H). 35,47-53 A significant number (18/78; Fisher's exact test; p value = 5e-13) of RUNX1 and NOTCH1 co-regulated genes were also down-regulated following I-BET151 treatment including super-enhancer associated genes (e.g., MYC, MYO7B, IL4, and FGR) (Table S3). MYC is a known target of bromodomain inhibitors but its sensitivity to such inhibitors varies widely. 42,54 At N-ME, only SE-5 was depleted following I-BET151 treatment in KOPTK1 and there was a modest reduction (31% reduction, FDR = 0.38) in MYC mRNA levels whereas RUNX1-KD depleted all SEs and MYC was significantly down-regulated (40% reduction; FDR = 8e-06) ( Figure 5I). STAT5A, MYO7B, and IL4 are associated with super-enhancers and showed marked depletion of H3K27ac densities following RUNX1-KD or I-BET151 treatment and were significantly down-regulated (FDR % 0.05) ( Figures 5J-5L). Notably, MYC, STAT5A, MYO7B, and IL4 super-enhancers were co-occupied with RUNX1, P300 and BRD4 peaks. Collectively, the results suggest that RUNX1 and BET co-dependent RUNX1 drives the expression of CDC25A and promotes entry into S-phase of cell cycle RUNX1 stimulates cell cycle entry and progression in hematopoietic cells 55,56 and its knockdown is associated with impaired cell growth, increased apoptosis, and G1-phase accumulation in T-ALL cell lines. 11,12,25 We confirmed the reported block of S-phase with a 44%-26% drop in the proportion of cells in S-phase following RUNX1-KD in KOPTK1 cells ( Figure 6A). RUNX1 itself is not a part of the cell cycle pathway but its association with a blocked G1-S transition suggests that RUNX1 might regulate key cell cycle genes. To identify putative RUNX1 target genes involved in cell cycle regulation, we used gene set enrichment analysis (GSEA) with the REACTOME cell cycle gene set ( Figures 6B and 6C). The highest-ranked gene in GSEA was CDC25A whose over-expression is associated with a shortened G1-phase and faster S-phase entry in the cell cycle. 57,58 CDC25A is highly expressed in T-ALL ( Figure 6D) and its expression was significantly (FDR % 0.05) down-regulated (>2-fold) upon RUNX1-KD and with a concomitant loss of H3K4me3 (2.17-fold) and H3K27ac (1.65-fold) promoter density ( Figures 6E and 6F). NOTCH1 also binds to the CDC25A promoter but NOTCH1 inhibition did not alter nearby chromatin states nor transcription significantly. Chemical inhibition of CDC25A by NS95397 in KOPTK1 and HL60 cell lines led to reduced cell viability (t-test p-value < 0.001 at 72 h in triplicates) and phenocopied a RUNX1-KD mediated block of cell growth (Figure 6G). Thus, our data suggest that induction of CDC25A by RUNX1-mediated chromatin remodeling contributes to a disruption of the G1-S checkpoint to promote entry into the S-phase in T-ALL cells.

DISCUSSION
While RUNX1 was initially described as a tumor suppressor in T-ALL, 8 recent studies in human and mouse T-ALL models suggest it plays a dual oncogenic role. 11,12 A tumor-suppressor role for RUNX1 was proposed based on observed loss of function mutations to RUNX1 in T-ALL. 59 RUNX1 mutations are highly prevalent in the early T-cell precursor (ETP) T-ALL and are found in 20% of T-ALL cases. 8 In contrast to this model, we report here that wild-type RUNX1 binds to DNA elements controlling proto-oncogene expression where it facilitates H3K27ac deposition and gene activation. A RUNX1-dependent increase of H3K27ac establishes oncogenic super-enhancers that drive expression of proto-oncogenes known to be associated with T-ALL pathogenesis including Notch signaling and cell cycle pathways. Taken together, our data support a parallel role for RUNX1 as an oncogene in T-ALL.
RUNX1 and NOTCH1 cooperation has been proposed through the analysis of co-dependent transcriptional changes at selected target genes. 11,19 Through the integrative analysis of chromatin and transcriptional states, we now propose a synergistic RUNX1 and NOTCH1 model that acts in a stepwise fashion to move chromatin from a repressed to an active state and have cataloged a comprehensive list of shared target genes. The synergistic relationship of NOTCH1 and RUNX1 extends to super-enhancers in T-ALL which were previously reported to be regulated by NOTCH1. 19 Here we show that RUNX1 is required for the maintenance of super-enhancers including N-ME and thus the previously reported dependence on NOTCH1 may lie in the requirement for an initial demethylation of H3K27me3 through JMJD3 recruitment. More generally, RUNX1 and NOTCH1 co-regulated genes were highly sensitive to either RUNX1-KD or NOTCH1 inhibition, suggesting that NOTCH1 and RUNX1 both contribute to forming a transcriptional regulatory unit that can be disrupted by the inhibition of either partner. In support of this, RUNX1 has been described as part of the ICN1 interactome 60 setting the stage for iScience Article future work to identify the basis and extent of physical interactions between ICN1 and RUNX1 interactomes at coregulated loci. Combining previous observations 26,27 along with those revealed in this study, we suggest that one mechanism for RUNX1 and NOTCH1 co-regulation of their target loci involves a stepwise activation of H3K27me3 marked regions with NOTCH1 mediating the removal of the repressive trimethyl modification followed by RUNX1 mediated acquisition of H3K27ac. However, this model does not refute the hypothesis that RUNX1 might be required to acetylate the enhancers first to recruit RBPJ/NOTCH1 complex at some loci which has been previously shown in Drosophila. 20 Previous studies in T-ALL implicate multiple RUNX family members, particularly RUNX1 and RUNX3, on the basis that they are frequently co-expressed and show functional redundancies. 11 However, RUNX1 is expressed across all T-ALL subtypes and forms a core regulatory circuitry with TAL1, GATA3, and MYB, driving the growth and survival of T-ALL cells. 11,12,25,36 Our data on leukemia and epithelial cancer cell lines showed RUNX1 regulates global H3K27ac levels and thus has the potential to co-bind with other oncogenic factors at H3K27ac marked enhancers. In support of this, a recent study suggests RUNX1 plays a role in regulating global H3K27ac levels through cooperation with key transcription factors in limbal stem cells. 61 It will be of interest to determine if the cooperative role of RUNX1 with oncogenic (co)factors in T-ALL cells extends to other cancers through synergistic partnerships with additional transcription factors.
Deregulation of enhancer states has emerged as a critical step in the activation and maintenance of aberrant transcriptional programs in T-ALL, 36,38,62,63 which provides rationale for targeting chromatin regulators associated with enhancers as a viable anti-cancer strategy, such as pharmacological inhibition of BET proteins. BET bromodomain inhibitors displace BET proteins from chromatin 28,41,44 and represent promising therapeutic agents in hematopoietic and solid cancers. 64 The findings presented here suggest that BET inhibition led to a reduction in H3K27ac levels preferentially at the BRD4 enriched regions and disrupted RUNX1 driven super-enhancers, suggesting a therapeutic opportunity to target RUNX1 driven enhancers in T-ALL. Taken together, our study highlights the central role of RUNX1 in establishing aberrant enhancers and provides the mechanistic basis for targeting aberrant enhancers using BET inhibitors in RUNX1-driven malignancies.

Limitations of the study
In this study, we demonstrated that a reduction in RUNX1 levels correlates with a genome-wide reduction of H3K27ac levels. While we propose a mechanistic model that directly links RUNX1 to this loss, future studies will be required to directly test this and other potential models. Despite an overall reduction in H3K27ac occupancy following RUNX1-KD a limited number of H3K27ac regions including those at super-enhancers demonstrated an increase in H3K27ac density. Our proposed model does not explain the mechanism of gain of H3K27ac following RUNX1-KD but we speculate these gains represent secondary effects of RUNX1-KD.
Our study provides evidence of a synergistic relationship between RUNX1 and NOTCH1 to establish active promoter and enhancer states. This model of synergy is based on a correlation between histone modification states and transcriptomic changes following RUNX1 and NOTCH1 perturbation in KOPTK1 and its generalizability remains to be explored. Furthermore, our analysis does not provide direct evidence that upon NOTCH1 or RUNX1 perturbation the interaction between RUNX1 and the ICN1 complex is disrupted, a prediction of our model. Finally, while we provide evidence that I-BET151 treatment is associated with a broad reduction of H3K27ac levels in KOPTK1, additional BET inhibitors and RUNX1-driven cell lines should be tested to confirm whether BET inhibition alone is sufficient to reduce RUNX1-mediated H3K27ac levels.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
Authors declare no conflict of interests.

INCLUSION AND DIVERSITY
We support inclusive, diverse, and equitable conduct of research.  iScience Article I-BET151 treatment and SNAP-ChIP spike-ins KOPTK1 cells were seeded at 100,000 per well in 6 well-plate. DMSO (Sigma, D2650-100 ML) and I-BET151 (Selleckchem, S2780) were dosed by direct addition to the culture media at 0.1%. I-BET151 was used at a final concentration of 0.5 mM and cells were treated for 72 h. Cell count and viability were determined by Trypan Blue exclusion assay at 72 h of treatment and pellets were collected and snap-frozen for further analysis. For SNAP-ChIP, barcoded nucleosomes were obtained from EpiCypher. SNAP-ChIP was performed based on the native ChIP protocol describe by Lorzadeh et al. 82 by adding 1 mL spike-in after digestion. We normalized the H3K27ac libraries following I-BET151 treatment using DNA-barcoded nucleosomes as spike-in control following the protocol described by Grzybowski et al. 83 To measure histone modification density (HMD) score at H3K27ac peaks, we used the pipeline developed by Grzybowski et al. 83

RNA-seq
RNA was isolated from cells with TRIzol reagent followed by purification over PureLink RNA mini kit columns (Invitrogen). RNA-seq was performed using a poly-A library construction protocol with strand-specific cDNA synthesis and 8 cycles of PCR.

RNA-seq data analysis
RNA-seq paired-end reads were aligned to an extended reference transcriptome that consists of the reference genome and annotated exon-exon junctions using BWA (v0.5.7) aln. 70 We used JAGuaR (v2.0.3) pipeline to generate the custom reference transcriptome (built from NCBI GRCh37-lite reference and Ensembl v75 annotations) and reposition reads that spanned exon-exon junctions. 73 An in-house pipeline was used to generate RNA-seq quality control matrices and profiles to assess the quality of RNA-seq as described. 74 To quantify the exon and gene expression, we calculated Reads Per Kilobase Million (RPKM) matrix as described. 84 For RPKM normalization, we used the total number of reads aligned into coding exons normalized by total exon length. We excluded the reads from mitochondrial genome, ribosomal genes, and the reads falling into the top 0.5% expressed exons which were considered as a source of potential outliers. The gene RPKM was calculated by taking the average RPKM values of all exons of a given gene. Pairwise comparisons between control and treatment were performed to identify differentially expressed genes using a custom DEfine MATLAB tool (FDR % 0.05, Minimum number of aligned reads = 10). 74 The RNA-seq data for CD4 naive, CD8 naive, CD4 memory and fetal thymus were obtained from Roadmap Epigenomics Project 85 and processed uniformly using our pipeline. Additional RNA-seq datasets [85][86][87] used in this study are listed in Table S4.

Gene set enrichment analysis (GSEA)
To find the enrichment of RUNX1-regulated genes to the REACTOME cell cycle genes, we used the RPKM values of protein-coding genes for seven RUNX1 knockdown experiments in KOPTK1, HPBALL, and RPMI cell lines.

QUANTIFICATION AND STATISTICAL ANALYSIS
Numerical analyses and statistical tests were performed on R (v3.6.0). Statistical significance was defined by p values or adjusted p values.
The comparisons of the mean RUNX1 ChIP-seq read densities at gained, lost, and stable H3K27ac regions following RUNX1-KD were performed by a Kolmogorov-Smirnov test ( Figure 1E).
Three independent experiments were performed to calculate p values using unpaired two-tailed t-test where * indicates p value % 0.05 and ** for p value % 0.01 ( Figure 1G).
The comparison of H3K27ac peak width between RUNX1-CTL and RUNX1-KD samples was performed by an unpaired two-tailed t-test where *** indicates p value is < 0.001 (Figure 2A).
The comparison of H3K27me3 and H3K27ac fold-change at the promoters of the 78 co-regulated genes following NOTCH1-INB or RUNX1-KD was performed by an unpaired two-tailed t-test to calculate the p value ( Figure 3F).
Three independent experiments were performed to calculate p values at 72 h using unpaired two-tailed t-test between NT and NS95397 treated samples at 0.5, 1, 2.5 and 5 mM concentrations. *** indicates p value is < 0.001 ( Figure 6G).
For significance, one-way analysis of variance (ANOVA) was used with multiple comparison test; NS-not significant in ( Figure S2E).
The comparison of number of live cells in control (RUNX1-WT) and RUNX1-KO samples at 72 h in HL60 and DU145 cell lines was estimated using an unpaired two-tailed t-test. *** is the t-test p value % 0.05 in three independent experiments ( Figures S3A and S3B).
The comparison of normalized H3K27ac density within super-enhancers following RUNX1-KD or I-BET151 treatment in KOPTK1 cells was performed using an unpaired two-tailed t-test ( Figure S6D).