Recurrent Amplification of Chromosome 7 in Murine AML with Dnmt3aR878H and Npm1cA Mutations Identifies Gab2 As a Possible Therapeutic Target in AML

NPM1 and DNMT3A mutations co-occur much more frequently than expected by chance, with ~15% of AML samples containing mutations in both genes. This synergy is recapitulated in mouse models, where mice with either Dnmt3a^R878H or Npm1^cA mutations develop AML with a longer latency and lower penetrance than mice with both mutations (AML in 8-15 months with nearly 100% penetrance). To better understand the mechanisms underlying this synergy, we characterized pre-leukemic bone marrow, and 11 independent spontaneous AMLs that have arisen in these doubly-mutant mice. These spontaneous AMLs rapidly cause fatal disease in secondary recipients; whole genome sequencing revealed that each tumor has one or more human AML-like cooperating mutations, including canonical mutations in Idh2 (R172G), Ptpn11 (E69K, A72V, A465T and S506L), Kit (D818Y), Cbl (Y369H, Q365K, exon 9 deletion), Nf1 (frameshift), and an amplification of WT Flt3 associated with >20 fold overexpression. Remarkably, 10 of the 11 AMLs contained an amplification of the entirety of murine chromosome 7 as the sole structural variant. We analyzed panel sequencing data from mouse AMLs arising in this model (PMID: 30692594) and exome sequencing data from a different Npm1^cA driven tumor model (PMID: 28835438); these AMLs also developed complete or partial amplification of chromosome 7 in most cases. Through integration of these datasets, we identified a minimally amplified 8.9 Mbp region on chromosome 7 containing 209 genes, including two candidate genes (Gab2, Pak1) that are also overexpressed in both murine and human AMLs with DNMT3A and NPM1 mutations. Retroviral constructs expressing each cDNA (with an IRES-GFP tag) were transduced into lineage depleted bone marrow cells from preleukemic Dnmt3a^R878H, Npm1^cA, or Dnmt3a^R878H x Npm1^cA mice, and transplanted into recipients. Overexpression of GAB2 (but not “Empty Vector” or PAK1) induced significant expansion of hematopoietic cells from the Dnmt3a^R878H x Npm1^cA mice (and to a lesser extent, from Npm1^cA mice), but not Dnmt3a^R878H mice, suggesting that the mutations in Npm1 and Dnmt3a synergize to account for this phenotype. Overexpression of GAB2 in Dnmt3a^R878H x Npm1^cA bone marrow led to development of AML in 12/12 engrafted mice, with a median latency of 4 months (range 2.3-6.3); 10 mice engrafted with an “Empty Vector” (expressing GFP only) developed AML with a median latency of 7 months (range 5.8-9.9, p = 0.00048). Taken together, these data suggest that AML-associated mutations in DNMT3A and NPM1 allow for the selective expansion of cells overexpressing GAB2 as a key step towards transformation. To investigate whether Gab2 is required for maintenance of fully transformed AML, we utilized CRISPR-Cas9 to inactivate Gab2 (or Rosa26, as a negative control) in two independent, fully transformed murine AMLs from this model, and transplanted these cells into recipients. Sequencing from the leukemic mice showed significant selection against frameshift mutations in Gab2 (using two independent guides) but not Rosa26, suggesting that inactivation of Gab2 slows the growth of fully transformed murine AMLs. CRISPR-Cas9 mediated knockout of GAB2 in primary human AML cells (and K562 cells) showed selection against GAB2 frameshift mutations in vitro. Additional evidence suggests that GAB2 inhibition may represent a novel therapeutic target, including: 1) human AMLs from multiple independent datasets all display significantly increased GAB2 mRNA and protein, compared to normal hematopoietic cells; 2) GAB2 serves as a signaling hub, physically interacting with GRB2 and SHP2/PTPN11 to transmit signals from receptor tyrosine kinases (including FLT3) to downstream pathways; 3) GAB2 is essential for signaling from BCR-ABL, FLT3, and PTPN11 in myeloid malignancies; 4) the human GAB2 gene is located in a region of human chromosome 11q that is amplified in 1-2% of myeloid malignancies, and in some breast and ovarian cancers; and 5) mouse knockouts of GAB2 are viable, with very mild phenotypes (including normal resting blood counts; PMID: 11449275). Taken together, these observations nominate GAB2 as a therapeutic target for AML. Additional studies are underway to define the essentiality of GAB2 in primary human AML cells in patient-derived xenografts and in human CD34+ cells, and how leukemia-causing mutations in DNMT3A and NPM1 shape the fitness landscape to select for GAB2 overexpression.