Context Dependent Role of Polycomb Repressive Complex 2 in Acute Leukemia

Polycomb Repressive Complex 2 (PRC2) is a multi-protein complex with important roles in development and cancer. Both hyper- and hypoactivity of PRC2 are associated with blood-related malignancies. Activating mutations of the PRC2 methyltransferase EZH2 have been found in human B-lineage lymphomas. Inactivating mutations of PRC2 components EZH2, EED and SUZ12 have been described in early T-cell Precursor ALL (ETP-ALL) and inactivating PRC2-alterations are found in Myelodysplastic and Myeloproliferative Syndromes. The mechanisms underlying this paradox are incompletely understood. We here investigate the context dependent role of PRC2 in murine models. We initially studied PRC2 in normal hematopoiesis: Chip-seq analysis of the PRC2-mediated H3K27me3 chromatin mark demonstrates that many genes highly expressed in immature hematopoiesis gain H3K27me3 in the developmental transition from more immature Lin-Sca1+Kit+ (LSK) cells to lineage committed Granulocyte Macrophage Progenitors (GMPs). Transcription of these genes is enriched in EZH2^ko GMP compared to EZH2^ffGMP by Gene Set Enrichment Analysis (GSEA). These data suggest that PRC2 is important for the silencing of immature gene expression programs in the developmental transition from LSK to GMP.

We next analyzed the role of PRC2 in two murine models of acute leukemia: MLL-AF9 driven leukemia, and a model of early T-cell precursor T-ALL (ETP-ALL). In MLL-AF9 leukemia, we previously found that inactivation of Eed completely abrogate leukemogenesis in vitro and in vivo. We now report that genetic inactivation of the tumor suppressor Cdkn2a (a canonical PRC2 target) partially rescued MLL-AF9 mediated leukemia in vitro and in vivo. However, Cdkn2a^koEED^ko MLL-AF9 leukemia remained compromised. In vitro growth was reduced to approximately 10% of Eed^ff controls. While control MLL-AF9 leukemia developed in vivo in 100% of the recipients, Cdkn2a^koEED^koMLL-AF9 leukemia developed with significantly prolonged latency and incomplete penetrance (25%). RNAseq analysis revealed that high level expression of genes with established roles in MLL-AF9 leukemia such as HoxA9, Cdk6 and Jmjd1c unexpectedly depends on Eed. These data are in keeping with the absence of alterations in PRC2-components in human MLL-rearranged leukemia.

In contrast, PRC2 core components (EZH2/EED/SUZ12) are deleted or mutated in > 40% of ETP-ALL. ETP-ALL also often has direct or indirect activation of the RAS-pathway, and carries frequent deletions of the CDKN2A locus. To model the effects of EED and EZH2-inactivation in ETP-ALL, we established Cdkn2a^koEed^ff vs Cdkn2a^koEed^ko, and Cdkn2a^koEzh2^ff vs Cdkn2a^koEZH2^koleukemias by transduction with NRASQ61K followed by expansion on OP9DL1 stroma cells to activate T-lineage differentiation via Notch-signaling. Cdkn2a^ko NRASQ61K leukemia showed an immunophenotype similar to human ETP-ALL (positive for c-Kit, CD5 and myeloid markers and mostly negative for CD4/8). Inactivation of Eed or Ezh2 in this model led to a shortening of latency (p=0.03 for Eed, p=0.0001 for Ezh2). RNAseq revealed enrichment of genes associated with murine DN1 thymocytes and with human ETP-ALL in Eed^ko vs Eed^ff Cdkn2a^ko NRASQ61K leukemia. These genesets showed even more pronounced enrichment in Ezh2^kocompared to Ezh2^ff Cdkn2a^ko NRASQ61K leukemia. Genes highly expressed in early hematopoiesis were enriched in Eed^ko and Ezh2^ko cells in both, the MLL-AF9 and NRASQ61K leukemia models. However, there was an opposing effect on HoxA9 gene expression, with PRC2 inactivation leading to decreased HoxA9 expression in MLL-AF9, and increased HoxA9 expression in Cdkn2a^ko NRASQ61K leukemia. Decreased HoxA9 has been shown to impair MLL-AF9 leukemia growth. To test the functional significance of elevated HoxA9-levels in the Eed^ko and EZH2^koNRASQ61K leukemias, we co-expressed HoxA9 and NRASQ61K in the presence of intact Eed and Ezh2 loci. Preliminary data suggest that HoxA9 accelerates leukemia development in this setting. Alterations in chromatin modifiers, including PRC2, are frequent in leukemia and lymphoma. Our data demonstrate that manipulation of PRC2 can have opposite effects on leukemia phenotype and expression of key PRC2-repressed genes such as HoxA9 in the context of different tumors. We are currently characterizing the mechanisms leading to divergent outcomes of PRC2 manipulation in MLL-AF9 leukemia compared to NRASQ61K ETP-like leukemia.