Ezh2 and Runx1 Mutations Targeted to Early Lymphoid Progenitors Collaborate to Promote Early Thymic Progenitor Leukemia

Understanding the specific cell populations responsible for propagation of leukemia is an important step for development of effective targeted therapies. Recently, the lymphoid-primed multipotent progenitor (LMPP) has been proposed to be a key propagating population in acute myeloid leukemia (AML; PMID 21251617). We have also shown that LMPPs share many functional and gene expression properties with early thymic progenitors (ETPs; PMID 22344248). This finding is of particular interest as ETP leukemias have recently been described: a distinct and poor prognostic disease entity with a transcriptional profile reminiscent of murine ETPs, showing co-expression of hematopoietic stem cell (HSC) and myeloid markers (PMID 19147408). Together, this raises the question whether ETPs can act as a leukemia-initiating/propagating cell population; however, relevant disease models to test this hypothesis are currently lacking.

Analysis of the genetic landscape of ETP leukemias has revealed frequent coexistence of inactivating mutations of EZH2 and RUNX1 (PMID 22237106). We therefore generated mice with deletions of Ezh2 and Runx1 specifically targeted to early lymphoid progenitors using Rag1Cre (Ezh2^fl/flRunx1^fl/flRag1Cre⁺; DKO mice). As anticipated, HSCs lacked significant recombination in DKO mice whereas close to 100% of purified ETPs (Lin^- CD4^- CD8^- CD44⁺ CD25^- Kit⁺ Flt3⁺) showed deletion of Ezh2 and Runx1. Strikingly, despite a 16-fold reduction in thymus cellularity caused by a block in thymocyte maturation at the DN2-DN3 transition, absolute numbers of ETPs within the thymus of DKO mice were markedly expanded (12-fold; p<0.0001). In contrast, Ezh2 or Runx1 deletion alone had no impact on numbers of ETPs. RNA-sequencing of the expanded ETPs in DKO mice revealed upregulation of HSC- and myeloid-associated transcriptional programs, reminiscent of ETP leukaemia e.g. Pbx1 (log₂FC=3.0; p<0.0001) and Csf3r (log₂FC=1.9; p=0.0038). Single-cell gene expression analysis confirmed co-expression of HSC and myeloid programs with lymphoid genes within individual DKO ETPs. Further, some key regulators of T-cell maturation which are aberrantly expressed in ETP leukemia were also disrupted in DKO ETPs e.g. Tcf7 (log₂FC=-9.5; p<0.0001). Gene expression associated with aberrant Ras signalling was also present. However, despite a continued expansion of the ETP population with age, we did not observe leukemia in DKO mice with over 1 year of follow-up.

Since ETP leukemias frequently feature activating mutations in genes regulating RAS signaling, we hypothesised that the expanded "pre-leukemic" ETPs in DKO mice would be primed for leukemic transformation by signalling pathway mutation. We therefore crossed DKO mice with a Flt3^ITD/+ knock-in mouse line, as internal tandem duplications (ITD) of FLT3 are frequent in ETP leukemias. Ezh2^fl/flRunx1^fl/flRag1Cre⁺Flt3^ITD/+ (DKOITD) mice showed dramatically reduced survival (median 9.3 weeks) resulting from an aggressive, fully penetrant acute leukemia showing a predominantly myeloid phenotype (e.g. Mac1) but with co-expression of some lymphoid antigens (e.g. intracellular CD3). Crucially, this leukaemia could be propagated in wild-type recipients upon transplantation of the expanded ETPs. DKOITD ETPs were transcriptionally very similar to DKO ETPs, retaining expression of lymphoid alongside HSC- and myeloid-associated genes.

Finally, in a lympho-myeloid cell line model (EML cells) we demonstrated that Ezh2 inactivation-induced loss of H3K27me3 is associated with a corresponding increase in H3K27Ac, a transcriptional activating signal that recruits bromodomain proteins. As such, we reasoned that our ETP leukemia model might be sensitive to bromodomain inhibitors such as JQ1. Indeed, we observed high sensitivity of expanded DKOITD ETPs to JQ1, raising the possibility of a new therapeutic approach for ETP leukemias.

This novel mouse model of ETP-propagated leukemia, driven by clinically relevant mutations, provides intriguing evidence that leukemias with a predominant myeloid phenotype, but co-expressing lymphoid genes, may initiate within a bona fide early lymphoid progenitor population. Since the functional characteristics of the cell of origin of a leukaemia may direct its progression and response to therapy, these findings could have important implications for future stratification and treatment of both AML and ETP leukemias.