To investigate the role of a NUP98::NSD1 fusion gene in vivo, we generated transgenic mice that expressed a NUP98::NSD1 fusion gene in the hematopoietic compartment under the control of Vav1 regulatory elements by microinjection of fertilized C57BL/6 embryos (Fig. 1A-B). Two of seven founders (mice A10 and I8) developed leukemia during a 20-month observation period, at 181 and 232 days respectively (Fig. 1C). CBCs from both founders showed leukocytosis, mild anemia, and circulating blasts (Supp Table S1). Flow cytometry demonstrated infiltration of bone marrow (BM), spleen, and thymus with Mac-1+Gr-1+ positive myeloblasts, and BM cytospin reveled sheets of myeloblasts. (Fig. 1D-E) Necropsy findings revealed hepatosplenomegaly, and IHC demonstrated myeloperoxidase (MPO) positive blast invasion of perivascular regions of parenchymal tissues such as liver and lung (Fig. 1F). Only one of these two leukemic mice (A10) was able to successfully breed with wild-type (WT) mates prior to death from AML. Surprisingly, F1 mice from this founder did not show a survival difference between WT and transgenic mice, and only one animal (C970) in a cohort of 30 F1 transgenic mice showed evidence of AML (Supp Fig S1).
To determine if there may be a mouse strain, or integration effect leading to lack of transmission of the leukemic phenotype, we generated a new cohort of transgenic mice, this time on an FVB/NJ background. In this cohort, only one mouse (MT1502) developed clear evidence of AML during a 20 month observation period. Similar to the NUP98::NSD1 mice generated on a C57BL/6 background, the leukemic MT1502 mouse displayed leukocytosis and anemia in the peripheral blood (Supp Table S1) and invasion of myeloblasts by flow cytometry, May-Grünwald-Giemsa (MG) staining, and IHC. (Supp Fig. S2A-C) The MT1502 founder was bred to a WT mate, but similar to findings with the C57BL/6 founder, there was no survival difference in the F1 generation between transgenic and WT F1 generation mice (Supp Fig S2D).
Given the frequent co-occurrence of NUP98::NSD1 and FLT3-ITD in human AML, we generated double transgenic mice that expressed both the NUP98::NSD1 fusion and a FLT3-ITD mutation. The double transgenic mice (n = 37) had decreased survival compared to WT (n = 20; p < 0.0001), FLT3-ITD (n = 25; p = 0.0040), or NUP98::NSD1 (n = 31; p < 0.0001). (Fig. 2A) We were able to perform detailed necropsies on 17 of the deceased NUP98::NSD1/FLT3-ITD mice. Fourteen animals (82.4%) developed AML, and three animals developed a precursor T-cell lymphoblastic leukemia/lymphoma (pre-T LBL) (Supplemental Table S2); FLT3-ITD only mice developed myeloproliferative disease, as previously shown. [12] The double transgenic AML were characterized by Mac-1+Gr-1+ blasts, whereas the pre-T LBL cases displayed T-lineage lymphoblasts. (Fig. 2B-C). BM cytospin and IHC shows invasion of myeloblasts, consistent with AML. (Fig. 2D-E) CBC results at 1 year, prior to the onset of most cases of leukemia (see Fig. 2A) showed moderate macrocytosis, thrombocytopenia, and a non-significant trend toward anemia and leukocytosis in the NUP98::NSD1/FLT3-ITD and FLT3-ITD mice (Fig. 2F). Supplemental Table S2 summarizes data from NUP98::NSD1/FLT3-ITD leukemic mice, including survival, diagnosis, relevant immunophenotype, and CBC results. Prominent, recurrent findings included severe macrocytic anemia, leukocytosis, and occasional thrombocytopenia.
We used WES to identify acquired mutations in the NUP98::NSD1/FLT3-ITD leukemias. Previously, we identified acquired mutations involving Ras (Kras, Nras, Ptpn11, Nf1 and Cbl) or tyrosine kinase (Flt3, Kit, Jak, Stat, and Sh2b3) genes in 20–72% of leukemias driven by NUP98 fusions or the related CALM-AF10 fusion.[15–18] Surprisingly, we found no recurrent acquired mutations in 13 NUP98::NSD1/FLT3-ITD mice with AML. Rare Tier1 acquired mutations involving known leukemia genes such as Notch1 and Jak1 were identified in AML and pre-T LBL respectively. (Supplemental Fig. S3A, Supplemental Table S3). Given that loss of the WT allele is a frequent event in AML patients who have a FLT3-ITD, we evaluated NUP98::NSD1/FLT3-ITD AML sample for loss of the WT Flt3 allele. Four of 13 (31%) AML samples showed loss of the WT Flt3 allele (Supplemental Fig. S3B).
We used RNA-Seq to generate gene expression profiles for NUP98::NSD1/FLT3-ITD AML, and compared them to gene expression profiles from WT unfractionated BM and WT BM enriched for hematopoietic stem and progenitor cells (Lineage negative BM) (Supplemental Table S4). Principal component analysis (PCA) demonstrated clear distinction between these three groups (Supplemental fig. S4). Unsupervised hierarchical clustering also separated the samples into anticipated groups for NUP98::NSD1/FLT3-ITD AML, WT unfractionated BM and Lineage negative BM (Fig. 3A). We used GSEA to identify global gene expression patterns that were similar to that of NUP98::NSD1/FLT3-ITD AML. We generated a set of genes that was > 2 fold differentially expressed at p < 0.05 between NUP98::NSD1/FLT3-ITD AML and Lineage negative BM, and used this set to interrogate the “cell type signature” sets available on Molecular Signatures Database (MSigDB) v7.5.1.4. Gene sets that had a normalized enrichment score (NER) > 1.5 are listed in Supplemental Table S5. The majority of these gene sets represent tissue macrophages or neutrophils; several examples are shown in Fig. 3B. Given that the GSEA analysis suggested that the NUP98::NSD1/FLT3-ITD AML were of myelomonocytic origin, we stained NUP98::NSD1/FLT3-ITD AML with additional antigens. All five AML analyzed showed a similar pattern, Mac-1+Gr-1+CD16/32+F4/80hetCD13−, consistent with myelomonocytic cells (Supplemental Fig. S5).
GSEA comparison to previously described AML signatures revealed strong similarity to “Valk_AML_Cluster_5”, which primarily consisted of patients with a monocytic or myelomonocytic (M4 or M5) AML subtype,[19] and a group of AML patients with MLL gene fusions (Fig. 4A).[20] Finally, we compared the NUP98::NSD1/FLT3-ITD AML gene signature to one of human NUP98::NSD1 AML extracted from publicly available data.[8] Again, there was strong similarity between human NUP98::NSD1 AML (most of whom also had FLT3-ITD mutations) and murine NUP98::NSD1/FLT3-ITD AML (Fig. 4B).