Targeting Transcription Factors in Cancer

Human cancers undergo an extremely diverse range of DNA mutations and rearrangements to generate oncogenes and to inactivate tumor suppressors in the process of their malignant transformation. As a consequence, decades of research have focused on the genes and the molecular pathways as well as the mechanisms that promote the growth of specific tumor cell types. This in turn has motivated the discovery and development of precision drugs that can target tumor-specific mechanisms. Although notable successes have been achieved, particularly with the targeting of oncogenes that encode protein kinases, many known drivers of human cancers have remained hitherto “undruggable” (1). These include unmutated transcription factors (TFs) such as Myc, Myb, E2F, and nuclear factor κB or chimeric transcriptional regulators generated by chromosomal translocations and novel gene fusions such as those encoding E2A-PBX1 and E2A-HLF. The DNA-binding domains of such oncogenic TFs, which confer specificity to their genomic actions, have proven to be intractable surfaces for selective and potent small molecules that can be developed into drugs.

The Roeder laboratory, along with other groups, has been analyzing the molecular properties and functions of the E2A-PBX1 fusion protein given its import as an oncogenic driver in 5 to 7% of pediatric acute lymphoblastic leukemias (ALLs). The chimeric protein is generated by the t(1;19)(q23;p13) chromosomal translocation and contains the DNA-binding domain (homeobox) of PBX1 and the transcriptional activation domains of E2A (2). In PNAS, Li et al. (3) provide deeper molecular insight into the mechanism by which E2A-PBX1 promotes dysregulated expression of a set of target genes, including an E2F family member, that in turn drive the uncontrolled growth of the leukemic cells. The molecular insight demonstrates a biochemical and functionally important interaction of an E2A transactivation domain contained within E2A-PBX1 with a subunit (MED1) of the RNA polymerase II (Pol II) Mediator complex. The latter complex functions as a highly versatile adaptor that enables the communication of TFs bound to promoters and enhancers with the RNA Pol II complex. These findings raise the specific possibility of therapeutically targeting the interaction surface between the E2A transactivation domain and MED1 in E2A-PBX1–driven pediatric ALL. More generally, the results invoke the possibility of targeting other Mediator subunits and their interaction surfaces with the aforementioned oncogenic TFs.

E2A-PBX1 Recruitment to Target Genes

Earlier work on the E2A-PBX1 fusion protein had focused on the DNA-binding homeodomain of PBX1 as the determinant of its specificity of genome action (2). Consistent with this possibility, chromatin cross-linking and immunoprecipitation (ChIP) analysis using DNA microarrays had shown that promoter regions of E2A-PBX1–regulated genes in transformed B-lymphocyte progenitors were enriched for the PBX1 DNA-binding motif. Furthermore, the Wnt1 and Wnt16 genes were invoked as transformation-relevant targets. Prior to the study in PNAS, the Chen and Roeder laboratories had recently uncovered evidence for a novel means of recruitment of E2A-PBX1 to its target genes, namely via interaction of its PBX1 domain with DNA-bound RUNX1 (4). ChIP-sequencing (ChIP-seq) analysis of an epitope-tagged E2A-PBX1 protein showed a modest enrichment for the PBX1 motif but a substantially higher enrichment for the RUNX motif. Importantly, the PBX1 domain within E2A-PBX1 could be shown to directly interact with RUNX1 in vitro and to be required for its recruitment to a large set of target genes in vivo. Crucially, the recruitment of E2A-PBX1 to its target genes, mediated by RUNX1 bound to DNA, was not dependent on the DNA-binding properties of the homeobox domain of PBX1. Finally, many of these E2A-PBX1–regulated genes in leukemic cells could be shown to also be coregulated by RUNX1.

A Tripartite Complex of RUNX1:E2A-PBX1:Mediator

Given this backdrop involving recruitment of E2A-PBX1 to a subset of RUNX1 target genes in transformed B lymphocyte progenitors, the authors were motivated to analyze the interaction of E2A-PBX1 with subunits of the Mediator and then to determine the functional import of such interactions in the proliferation of transformed pre-B cells. The experiments involve strongly complementing biochemical, molecular, genetic, and genomic approaches (3). Coimmunoprecipitation as well as protein–protein cross-linking analyses were used to demonstrate a direct interaction of an E2A activation domain region (AD2) with the MED1 subunit of the Mediator. These experiments were extended using immobilized DNA template assays to show that DNA-bound E2A could recruit the Mediator complex to a probe containing E2A binding sites, presumably via the AD2–MED1 interaction surface. In independent biochemical experiments using an immobilized DNA template the authors demonstrated the recruitment of E2A-PBX1 by DNA-bound RUNX1. What remains to be shown is that such a binary complex of the two TFs can in turn recruit the Mediator to the template via the AD2–MED1 interaction surface. Nevertheless, the biochemical experiments are consistent with the RUNX1 recruitment of E2A-PBX1 to its target genes via the PBX1 domain and then the subsequent recruitment of the Mediator by the E2A AD2 domain. The authors use ChIP-seq analyses to examine the genome-wide binding landscapes of these various TFs and the Mediator. Analyses of the E2A-PBX1 and MED1 ChIP-seq datasets delineated ∼35% of the MED1 peaks to overlap with E2A-PBX1 peaks. Conversely, ∼39% of the E2A-PBX1 peaks coincided with MED1 peaks. Importantly, a large majority of these E2A-PBX1 and MED1 cobound regions (∼92%, n =∼3,500) were also bound by RUNX1 in vivo. Thus, the combined in vitro and in vivo DNA binding/recruitment/chromatin cross-linking analyses reinforce a model invoking a tripartite molecular assembly (RUNX1:E2A-PBX1:MED) on select target genes containing RUNX1 binding sites. This assembly cannot entirely account for the observed targeting of E2A-PBX1 to RUNX1-bound regulatory sequences since only a subset of those sequences are seen to cobind E2A-PBX1. The authors provide evidence that another B-lineage TF, EBF1, could additionally stabilize such tripartite complexes, thereby further increasing specificity of E2A-PBX1 genomic transactions.

To genetically test the requirement of MED1 in E2A-PBX1–mediated growth of pre-B ALL cell lines, CAS9-mediated genomic editing was performed. This resulted in an impairment in cell-cycle progression which could be rescued by reexpression of MED1. Notably, the loss of MED1 did not result in disassembly and degradation of the Mediator complex. The E2F5 and IGLL1 genes, the latter representing a component of the signaling-competent pre-B cell receptor, appear to satisfy key functional criteria as E2A-PBX1 targets that contribute to the transformation of pre-B cells in ALL. As predicted by the tripartite assembly model these genes are also positively regulated by RUNX1 and MED1 in E2A-PBX1–transformed pre-B cell lines.

A systematic effort to map and solve the structures of all TF–Mediator interaction surfaces could generate a valuable resource for the development of a novel class of drugs targeting oncogenic TFs, which have to date been considered to be undruggable.

Targeting TF–Mediator Interaction Surfaces in Cancer Therapy

The current work further amplifies two fundamental issues pertaining to the transcriptional control of mammalian genes by TFs. First, are all mammalian TFs able to achieve their genomic target specificities via two distinct mechanisms involving either direct DNA binding or recruitment by other DNA-bound TFs? ChIP-seq analyses suggest both direct as well as indirect modes of genomic binding by many TFs. Are these differing modes of DNA targeting of capable of generating similar specificities of TF genomic transactions and temporal dynamics?

Second, what is the nature of protein interaction surfaces between mammalian TFs and Mediator subunits? Furthermore, how are such multiple and likely simultaneous interactions that undoubtedly occur in the context of combinatorial binding of TFs to promoters and enhancers integrated to control the activities of specific genes? Nevertheless, the latter consideration raises the possibility that a systematic effort to map and solve the structures of all TF–Mediator interaction surfaces could generate a valuable resource for the development of a novel class of drugs targeting oncogenic TFs, which have to date been considered to be undruggable.