Native Stem Cell Transcriptional Circuits in Normal and Malignant Early-Stage T Cells

The discovery of early T-cell precursor acute lymphoblastic leukemia (ETP-ALL) as a new subtype of T-ALL raised hopes for improved outcomes in this cancer. This subtype was first identified based on expression of stem cell genes that are characteristic of mouse ETP cells, the most primitive multipotent cells in the thymus. While the diverse mutational landscapes across ETP-ALL tumors have been extensively studied, establishing a shared framework that activates these stem cell genes remains elusive. Previously, we showed that the PIAS-like coactivator ZMIZ1 is a direct NOTCH1 cofactor in committed T-cell precursors and conventional T-ALL. In contrast, Zmiz1 drives normal ETP proliferation independently of Notch and is overexpressed in ETP-ALL. These observations led us to consider the possibility that ZMIZ1 induces a transcriptional network in ETP-ALL that is distinct from networks described in conventional T-ALL. To test this possibility, we knocked down Zmiz1 in normal and malignant ETPs. Zmiz1 deletion in normal ETPs impaired cell proliferation, reduced myeloid potential, and skewed differentiation to the T-cell fate. Conversely, overexpression of ZMIZ1 had the opposite effect, promoting ETP cell proliferation and inhibiting T-cell differentiation. ZMIZ1 knockdown in ETP-ALL cell lines and patient-derived xenografts reduced cell proliferation by 4-100-fold. These data show that Zmiz1 is essential for maintenance of undifferentiated normal and malignant ETP cells.

To determine mechanism, we performed RNA-Seq of ETP-ALL cells after ZMIZ1 knockdown. Among the top-25 regulated genes were BCL2, MYCN, MYB, and MEF2C, which are known ETP-ALL drivers. Out of deference to the ETP literature, we refer to these genes as “Phase I oncogenes”. Zmiz1 deletion or overexpression in normal ETPs showed repression or induction of these genes respectively. Pathway analysis showed that MYC was the top enriched gene set. To investigate the role of MYCN in ETP-ALL, we overexpressed MYCN or deleted Mycn in normal ETPs using conditional knockout mice. Like ZMIZ1 overexpression, MYCN overexpression induced robust ETP proliferation, skewed differentiation to the myeloid fate, and rescued Zmiz1-deficient ETP cells. Like Zmiz1 deletion, Mycn deletion impaired ETP proliferation while promoting differentiation. Next, we integrated our ZMIZ1 ChIP-Seq and ATAC-Seq datasets with 3D genome maps followed by validation with CRISPR interference and reporter assays. These studies identified essential ZMIZ1-bound developmental regulatory elements that induce expression of Phase I oncogenes. These elements were highly enriched for co-occupancy by MYB (~75%) and MEF2C (~30%), suggesting feedforward circuits. Consistently, RNA-Seq showed that MYB regulates ~70% of ZMIZ1 target genes, of which 95% were strikingly in cooperative direction including all Phase I oncogenes. Motif analysis showed that ETS motifs were the most significantly enriched at ZMIZ1-bound sites. Accordingly, ETS motif mutation impaired ZMIZ1 occupancy. Next, we sought to understand what triggers the ZMIZ1-centered circuit. MEF2C is a major initiator of ETP-ALL transformation downstream of multiple chromosomal alterations. Consistently, MEF2C occupied the ZMIZ1 enhancer. Further, MEF2C knockdown or inactivation with MARK inhibitors (which dephosphorylate MEF2C) impaired ZMIZ1 expression and cell proliferation. These data identify native ZMIZ1-dependent feedforward circuits involving teams of transcription factors that are hijacked to drive ETP oncogenesis.

There is no established framework that mechanistically explains how stem cell gene expression is activated by disparate mutational landscapes across ETP-ALL patients. Here we contribute by identifying a ZMIZ1-centered network of enhancers and transcription factors acting in feedforward circuits that are distinct from those in conventional T-ALL and are amplified from roots in normal ETPs. This network helps mechanistically explain why malignant ETPs show sensitivity to BCL2 inhibitors like venetoclax and rarely show genetic alterations of oncogenes like MYCN or MYB. ZMIZ1 induction of MYCN and BCL2 in combination might also explain the aggressive clinical behavior of ETP-ALL. Importantly, Zmiz1 is dispensable for postnatal health. Thus, the ZMIZ1-centered ETP-ALL network might guide development of safe therapies against this aggressive cancer.