TCF3-Hlf Hijacks ERG to Activate Leukemia-Critical Super-Enhancers in t(17;19) Leukemia

The translocation t(17;19), which results in a fusion of the transcriptional activation domain of the B cell developmental transcription factor TCF3 to the DNA-binding domains of HLF, defines a rare subtype of acute lymphoblastic leukemia (ALL) associated with very poor outcome. The mechanisms by which TCF3-HLF rewires gene expression in leukemia are still unclear.

To evaluate the function of TCF3-HLF in leukemia, we deleted the fusion protein by targeting HLF using CRISPR, which resulted in the dropout of targeted ALL cells ex vivo in cell cultures and in vivo using patient derived xenografts (PDX). Serial transcriptome analysis identified c-MYC (MYC) among the top genes that are downregulated early after TCF3-HLF knockout (KO). Several of previously inferred TCF3-HLF target genes including BCL2 and LMO2 were retrieved in this signature. Gene sets that are positively regulated by MYC and reactivation of gene sets that are repressed by MYC were highly enriched by gene set enrichment analysis (GSEA) after TCF3-HLF KO. Importantly, CRISPR knockout of MYC was lethal for TCF3-HLF leukemia. These results reveal MYC as an important target of TCF3-HLF.

To gain further insight into the TCF3-HLF regulome, we performed chromatin immunoprecipitation and sequencing (ChIP-seq), using engineered TCF3-HLF ALL cells (HAL01) in which we inserted a FLAG tag at the c-terminal end of the endogenous TCF3-HLF by CRISPR. ChIP-seq identified 484 high confidence binding peaks of TCF3-HLF, with a large fraction occupied introns (44.75%) or distal intergenic regions (44.21%). ChIP-seq of epigenetic marks revealed strong enrichment of active enhancers marks (H3K4me1 and H3K27ac), absence of poised enhancer marks (H3K27me3) and very low detection of promoter marks (H3K4me3) flanking the regions bound by TCF3-HLF. Importantly, 82 of the TCF3-HLF bound regions are associated with high density and long range of active chromatin mark (H3K27ac), which are annotated as super-enhancers. These regions are close to genes that are implicated in leukemia and hematological malignancies (eg. BCL2 and LMO2) and include a distal MYC enhancer that is active in acute myeloid leukemia (AML) and hematopoietic stem cells. Using chromosome conformation capture coupled with qPCR (3C-qPCR), we confirmed that this MYC enhancer is in proximity of MYC transcription start site through an enhancer-promoter loop, which is disrupted when TCF-HLF is knocked out. Furthermore, CRISPR targeting of the TCF3-HLF binding motif in this region abolished MYC expression and strongly reduced ALL cell proliferation. These results identify MYC enhancer as one of the important targets of TCF3-HLF.

De novo motif discovery identified high occurrence (399/484) of HLF motif (5'-TTACGTAA-3') enriched in the center of TCF3-HLF binding regions. To evaluate candidate TCF3-HLF binding regions functionally, we cloned 1 KB genomic DNA sequences centered around TCF3-HLF binding sites in an enhancer reporter vector. Mutagenesis of TCF3-HLF binding sites and TCF3-HLF knockout disrupted enhancer function of several candidate enhancers including the MYC enhancer mentioned above. TCF3-HLF KO lead to a marked reduction of H3K27ac abundance at positions flanking TCF3-HLF binding sites, providing functional evidence for a pioneering role of the TCF3-HLF complex in the activation of enhancer regions. Furthermore, we discovered a conserved motif grammar with ETS motifs (5'-GGA(A/T)-3') being closely associated with HLF motifs, in many instances at -1bp upstream of the HLF motif. Deletion of these ETS binding motifs or CRISPR knockout of ERG, but not of other ETS family transcription factors, compromised enhancer activity and leukemia survival. By affinity purification we confirmed that ERG is associated with the TCF3-HLF complex.

Collectively, these results brand TCF3-HLF as an oncogenic driver that acts primarily through aberrant activation of enhancer gene regulatory regions. TCF3-HLF likely cooperates with ERG to recruit transcriptional complexes to mediate enhancer activity. Indeed, we retrieved EP300 and BRD4 in the TCF3-HLF complex and show that TCF3-HLF ALL PDX were remarkably sensitive to BRD4 inhibition. Our models will enable further exploration of the TCF3-HLF complex by proteomics and systematic evaluation of candidate therapeutic agents to interfere with the fundamental oncogenic process.