The Growth Factor Independence 1 variant form GFI136N Predisposes to Acute Myeloid Leukemia by Inducing Epigenetic Changes in Oncogenes Such As Hoxa9

Abstract 223

A coding variant form of GFI1 (GFI136N) increases the risk to develop AML by 60% and is present in about 10–15 % of all Caucasian AML patients. To determine the underlying molecular mechanism and potentially develop new therapeutic approaches, we generated “knockin” mouse strains wherein the endogenous murine Gfi1 gene was replaced either by the human GFI1 variant (GFI136N, the form predisposing to AML) or by the more common form of GFI1 (GFI136S). In most hematopoietic compartments no difference was observable between GFI136N and GFI136S expressing mice; however, there was a 3–5 fold increase in the number of granulocytic monocytic progenitors (GMPs) and common myeloid progenitors (CMPs) in Gfi136N expressing (either homozygous or heterozygous) mice compared to wild-type or Gfi136S expressing mice(p≤0.01). Interestingly, both human and murine AML leukemic cells are thought to originate from GMPs and CMPs. To assess functional differences, we seeded GMPs from GFI136N or GFI136S knockin mice on methylcelluose or transplanted them into into syngenic animals. We found that GFI136N expressing GMPs proliferate faster and have an increased self-renewal capacity both in-vitro and in-vivo compared to GMPs carrying Gfi136S alleles.

A gene expression array analysis showed that GFI136N GMPs have a stem cell-like gene signature with elevated levels of Hoxa9 expression and a deregulation of a number of oncogenes involved in the development of human AML such as Trib2, Tet2 or Idh2. It is of particular interest that Hoxa9, a known GFI1 target gene, was up-regulated 3–4 fold in GFI136N GMPs compared to in GFI136S GMPs (p≤0.01). It is known that high levels of Hoxa9 accelerate AML development in mice and are associated with a poor prognosis in AML patients. GFI1 is a transcriptional repressor and exerts its function by recruiting different histone modifying enzymes, in particular LSD1, which de-methylates histone 3 (H3) at lysine 4 (K4), or histone deacetylases (HDACs), which remove acetyl groups from H3K9 residues and G9a, which initiates dimethylation of H3K9. Both H3K4 methylation and H3K9 acetylation correlate with actived gene expression, whereas H3K9dimethyl correlates with repession. Chromatin-immuno-precipitation (ChIP) of Gfi1-bound chromatin from Lin⁻Sca1⁻c-Kit⁺ cells, which contains the GMP population, showed that GFI136N binds to a lesser degree to the Hoxa9 locus than GFI136S. This diminished binding of Gfi136N correlated with an increased H3K4 dimethylation and H3K9 acetylation as well as diminished H3K9 dimethylation across the Hoxa9 locus in GFI136N cells. It is likely that these epigenetic changes lead to the increased Hoxa9 expression observed in GFI136N GMPs. A more exhaustive ChIP-Seq analysis with antibodies recognizing H3K4dimethyl in Lin⁻Sca1⁻c-Kit⁺ cells from Gfi136N or Gfi136S mice showed significant epigenetic alterations throughout the Hoxa9 locus genome and at other GFI1 target genes. It is conceivable that these epigenetic alterations explain, at least in part, the changed gene expression signatures in GFI136N GMPs.

To investigate the role of GFI136N in myeloid leukemogenesis, we induced the expression of a mutated form of KRAS (K12D) in both GFI136N and GFI136S mice. All mice developed a deadly myelo-proliferative disorder, but animals carrying the GFI136N allele succumbed to the disease within a significantly shorter latency period (17 against 31 days, p≤0.01) than GFI136S mice. We also transduced GFI136N and GFI136S GMPs with retroviral vectors directing the expression of either the AML1-Eto9a or the MLL-AF9 onco-fusion proteins typically found in human AML. We observed that GFI136N GMPs expressing MLL-AF9 or AML1-Eto9a generated 5–10 fold more colonies (p≤0.01) on methylcellulose and exhibited a higher replating efficiency than the respective GFI136S GMPs. Finally, AML blast cells from GFI136N heterozygous patients expressed higher levels of HOXA9 compared to AML blasts from GFI136S homozygous patients, suggesting that our mouse model reflects the disease predisposition in human patients. Our knockin mice are, to our knowledge, the first animal model for a human genetic variation that predisposes to leukemia. Based on the findings with this model, we propose that the human GFI136N variant predisposes to AML by inducing epigenetic changes affecting the expression of important regulators with oncogenic potential such as Hoxa9.