Polymerase-Associated Factor Complex (PAFc) Protects Notch-Induced Leukemia Cells from DNA Damage and Mitochondrial Stress

About 60% of T-cell acute lymphoblastic leukemia (T-ALL) patient samples have activating NOTCH1 mutations, making NOTCH1 the most prevalent oncogene in this cancer. In early clinical trials, pan-Notch inhibitors caused excessive toxicity because Notch has essential functions in tissue homeostasis. To circumvent toxicity, we investigate an alternative strategy of targeting context-specific transcriptional cofactors of Notch. Cell division cycle 73 (Cdc73) is a scaffold component of the RNA polymerase-associated factor complex (PAF1c) that links histone modifier enzymes with transcriptional machinery at genes and enhancers. Previous studies reported that flies with mutant Paf1C display notched wings and that NOTCH1 interacts with CDC73. Here, we confirmed the NOTCH1-CDC73 interaction using co-IP assays in human T-ALL cells. ChIP-Seq/RNA-Seq analysis showed that CDC73 signals were associated with the most highly transcribed genes and active enhancers. But CDC73 co-occupied only a subset of Notch-bound elements, which was highly enriched for ETS1 occupancy. Consistently, ETS1 knockdown generally impaired Notch and CDC73 binding to chromatin. Thus, we extend previous studies by showing that chromatin context restricts the NOTCH1-CDC73 interaction.

To investigate the role of Cdc73 for Notch-induced T-cell development and leukemogenesis, we deleted Cdc73 specifically in thymocytes using conditional knockout mice. Cdc73-deficient mice showed impaired expression of a subset of Notch target genes and a block in T-cell differentiation that closely phenocopies the block in Notch-deficient mice. Induced Cdc73 deletion in leukemic mice reduced blast counts by 10-700-fold and prolonged survival. CDC73 knockdown impaired human T-ALL cell line and patient-derived xenograft growth by 2-20-fold. For mechanistic studies, we inducibly deleted Cdc73 in murine Notch-induced tumors and performed Bru-Seq (measures nascent mRNA), BruUV-Seq (measures eRNA), and ChIP-Seq. Enrichment analysis showed that Cdc73, Ets1, and Notch induced a shared set of important pathways, such as oxidative-phosphorylation. Consistently, Cdc73 deletion reduced mitochondrial membrane potential, reactive oxidative species, and oxygen consumption. Another shared pathway was DNA repair. Consistently, Cdc73 deletion induced gH2AX, apoptosis, and chromosomal damage in metaphase spreads. To test whether DNA repair might be a synthetic lethal pathway, we treated T-ALL cells with berzosertib, an ATR inhibitor. T-ALL cells were highly sensitive to this inhibitor at nanomolar GI50 (3-112nM). Cdc73 deletion reduced expression of several T-ALL drivers like Il7r, Rasgrp1, and Ets1. These effects were associated with loss of H3K27ac and eRNA signals at linked enhancers co-occupied by Notch, ETS1, and CDC73. Cdc73 deletion reduced H2K120ub1 (a marker for transcriptional elongation) at oxidative-phosphorylation and DNA repair genes with promoters co-occupied by Notch, ETS1, and CDC73. Thus, Cdc73 appears to promote eRNA synthesis to activate oncogenes while promoting mRNA synthesis to induce a gene expression program that mitigates the metabolic and replicative stress of supraphysiological Notch signaling.

It was previously unclear why Notch target gene expression is particularly reliant on Paf1C-associated factors and histone modifications. While we did confirm observations that CDC73 interacts with NOTCH1, we also revealed dependence on chromatin context. Accordingly, an "ETS code" enucleates NOTCH1 and CDC73, thus facilitating their interaction. Further, we previously showed that leukemia-associated NOTCH1 alleles are weak transactivators, which seems discrepant with general hyperexpression of genes in cancer. Our data suggests that oncogenic Notch might overcome its inherent weakness as a transactivator through local intersection at chromatin with ETS factors and powerful RNA synthesis machinery. We provide mechanistic support for testing inhibitors of DNA repair, oxidative-phosphorylation, and transcriptional machinery as T-ALL therapy. To circumvent the toxicity of pan-Notch inhibition, we highlight strategies that disable pathways that intersect with Notch at chromatin to target Notch signals without directly targeting the Notch complex.

No relevant conflicts of interest to declare.