Multi-Omic Transcription Regulatory Network Mapping Identifies Targetable Oncogenic TF-Cofactor Relationship between IRF4 and p300 in Multiple Myeloma, and Is Used to Improve TF Protein-Protein Interaction Methodology

Background: Chromatin dysregulation is a hallmark of blood cancers and is a focused area of therapeutic development. Although most chromatin regulators bind sites of transcription, their inhibition yields differential effects on gene expression and dependency that are often linked to specific transcription factor (TF) programs. This observation motivated efforts targeting chromatin regulators and TF cofactors to selectively inhibit oncogenic activity of TFs such as MYC, RUNX1, HOX/MEIS, MYB, and IRF4 for blood cancers. However, in most cases, the mechanistic basis linking oncogenic TFs to cofactors is not apparent.

We developed transcription regulatory network (TRN) mapping, which integrates orthogonal datasets such as genomic occupancy, functional genomics, inferred genetic regulation, and protein-protein interaction (PPI), to elucidate mechanisms of transcriptional selectivity between TFs and chromatin regulators. Owing to the broad availability of multi-omic data, we find that TRN maps can be generated in many blood cancer contexts. However, both the availability and quality of TF PPI data are limited by insufficient experimental methods and over-reliance on non-specific cell lines.

Previously we have shown that this method can be used to infer favored TF-cofactor relationships (e.g., IRF4 and p300 in multiple myeloma) motivating the broader use of TRN mapping for target discovery. To overcome the challenges associated with TF PPI data and improve PPI definitions within our TRN models, we developed an affinity purification-mass spectrometry (AP-MS) method for chromatin-bound PPIs and coupled it with quantitative proteomics to calculate the stoichiometry of TF-cofactor interactions. This approach was validated on the IRF4-p300 relationship and improves resolution and quantification of endogenous PPIs that enables us to identify essential TF-cofactor relationships and mechanisms.

Methods: To map TRNs, we integrated datasets of PPIs, genomic occupancy (ChIP-seq) and regulation (RNA-seq), genetic dependency (CRISPR), and druggability to produce directional networks that can be used to depict TF-cofactor associations. For internal TF PPI profiling, endogenous TF complexes were enriched from nuclear extract by specific antibodies and analyzed by quantitative proteomics. Experimental conditions were optimized to preserve endogenous protein complexes without sacrificing specificity. Validation of TF-cofactor relationships such as IRF4 and p300 were performed using chemical genetic perturbations followed by profiling of chromatin occupancy, gene expression or protein levels.

Results: We find that TRN maps reveal context specific TF-cofactor relationships that are not easily observed in a single dataset. For example, despite p300 having similar genomic colocalization with oncogenic TFs IRF4 and Ikaros, p300 demonstrates a closer association to IRF4 when we incorporate multiple data layers. In this case, we identified a preferential PPI between IRF4 and p300 and show a statistically similar co-network of interactors. We validated IRF4 and p300’s association through chemical perturbation and show that p300 perturbation phenocopies IRF4, while having limited effect on other TF programs such as Ikaros. Given these results, we propose TRN mapping as a methodology that can be applied in other contexts to demonstrate how chromatin regulators can function as context specific dependencies.