NUP98-HOXA9-Initiates Molecular and Biologic Features of Disease Progression in a Humanized Model of CML

Acute myeloid leukemias (AMLs) are heterogenous diseases often resulting from the acquisition of multiple genetic alterations that deregulate hematopoietic precursor proliferation and block normal differentiation. Chronic myeloid leukemia offers a unique opportunity to identify molecular mechanisms that interfere with normal differentiation in the context of a highly proliferative hematopoietic stem cell clone that produces massive number of functional differentiated myeloid cells due to the presence of a BCR-ABL1 fusion gene. Since the NUP98-HOXA9 (NA9) fusion gene has been identified in some CML blast crisis patients or de novo AML, we asked whether a humanized model of CML progression to AML would result from its lentiviral-mediated introduction into primary CML CD34+ cells from 3 chronic phase CML patients in which >97%, >96%, and 44% of the CD34+ longterm culture-initiating cells were Ph+/BCR-ABL1+.

In vitro experiments showed the NA9-transduced cells produced a hugely increased number of granulo-monocytic progenitors in long-term cultures (up to 1000-fold, p=0.03; t-ratio test) and enhanced the serial replating activity of directly clonogenic cells as compared to matched samples of cells transduced with a control vector. In vivo experiments showed that 90% of sublethally irradiated NOD-SCID IL2Rgc-null mice expressing human IL3, GM-CSF and SCF constitutively that were transplanted with these same NA9-transduced cells developed evidence of a progressed myeloproliferative neoplasia. This included tissue infiltrates of eosinophils, basophils and mastocytes, exclusive myeloid differentiation and signs of an imminently fatal leukemia between 8 and 27 weeks post-transplant in 50% of cases, although an excess blast population was not seen in these.

RNA-sequencing of cells analyzed just 48 hours after transduction revealed a signature of 53 genes that were more highly expressed in CD34+ CML cells carrying the NA9 fusion gene in comparison to control CML cells. This signature included many genes expressed by hematopoietic stem cells (HSCs), such as 5' HOXA genes, 3' HOXB genes, PBX3, MEIS1,ARID5B, AHR, REL, BMP6, GDF10, SFRP5, PPBP,PLA2G4A suggesting that NA9 induces the expression of a partial HSC program in later CML progenitor types that make up the bulk of the CD34+ CML population. Consistent with this hypothesis, the NA9 signature separated the HSCs, CMPs and GMPs of chronic CML patients as well as those of normal subjects (GSE47927; p=1x10^-6, p=2x10^-4 respectively). Most genes in the NA9 signature were also found to be over-expressed in CD34+ cells of patients in the accelerated phase and in the blast phase of CML, including ARID5B, AHR, STARD9, TOX, or FOXP1 (p=9.10^-8; ANOVA) (GSE4170). The NA9 signature was also significantly enriched in transcripts of genes that are over-expressed in the blasts of AML patients carrying NPM1 mutations or MLL fusions and was predictive of overall survival in the AML cohort of the Cancer Genome Atlas (p=0.02; log-rank test, n=200).

This stem cell signature was also associated with an increase in the number of H3K27ac marks (on average 7805 ± 960 peaks for NA9 vs 5888 ± 2739 peaks for control) at 48 hours post-transduction. De novo H3K27ac peaks in NA9+ CD34+ cells were located in proximal and distal enhancers of 993 genes (GREAT parameters TSS ± 2kb from TSS and 100kb max extension). These were significantly enriched genes in the GM-CSF signaling pathway (MySigDB, binomial p-value=2.19x10^-8) and that are upregulated in granulocytes and monocytes following LPS exposure. H3K27ac peaks also matched with GATA1, RELA, MEF2A and IKZF1 transcription factor binding sites previously mapped by ChIPSeq experiments (adjusted p-value <0.05). Finally, we identified super-enhancers in 12 genes among which were PBX3, ANGPT1, MBNL1 and PRKACB.

Overexpression of the NA9 fusion gene in chronic phase CD34+ CML cells thus appears to reprogram the expression of HSC genes as well as those associated with GM-CSF pathway activation and inflammatory responses via H3K27 acetylation of associated loci leading to a picture of advanced accelerated phase/disease progression but not the complete differentiation arrest seen in terminal blast crisis or frank AML. These findings highlight the multiplicity of biologically important molecular alterations that can result from a single epigenetic perturbation but, nevertheless, are insufficient to create an overt AML phenotype.