This item is under embargo and not available online per the author's request. For access information, please visit

Date of Award

Spring 5-15-2016

Author's School

Graduate School of Arts and Sciences

Author's Department

Biology & Biomedical Sciences (Immunology)

Degree Name

Doctor of Philosophy (PhD)

Degree Type



Loss of function mutations in the DNA methyltransferase DNMT3A are highly recurrent in acute myeloid leukemia (AML). DNMT3A and the highly homologous gene DNMT3B encode the two methyltransferases that are primarily responsible for mediating de novo methylation of specific DNA sequences during cellular differentiation. DNMT3A mutations are mutually exclusive of several translocations that create oncogenic fusion genes (PML-RARA, RUNX1-RUNX1T1, CBFB-MYH11, and MLL-X), suggesting that these fusions may require functional DNMT3A to initiate leukemogenesis. Using bone marrow cells from a constitutive Dnmt3a null mouse, we show that loss of Dnmt3a caused a striking loss of DNA methylation throughout the genome of bone marrow cells, and a complete loss of methylation at hundreds of specific loci, suggesting that these regions are entirely dependent on Dnmt3a for maintaining normal methylation patterns. Using both retroviral vectors and a transgenic model, we demonstrated that the methyltransferase activity of Dnmt3a but not Dnmt3b is required for aberrant self-renewal ex vivo that is driven by PML-RARA (but not RUNX1-RUNX1T1 or MLL-AF9); further, the PML-RARA competitive transplant advantage and leukemia generation both required Dnmt3a. In contrast, Dnmt3a was not required for leukemia generation caused by MLL-AF9, which is known to have a requirement for Dnmt1 activity. Together, these findings demonstrate that PML-RARAs specifically dependent on Dnmt3a to drive leukemogenesis, and may explain why DNMT3A mutations are mutually exclusive of PML-RARA in AML patients.

While the most common mutation in DNMT3A in AML patients is the missense mutation R882H, other heterozygous mutations produce frameshifts, premature stop codons, or deletions of the entire coding sequence of the gene, strongly suggesting that these mutations lead to simple haploinsufficiency for DNMT3A. To test the hypothesis that Dnmt3a haploinsuffiency may initiate AML, we performed a long-term tumor watch comparing wild-type mice (Dnmt3a+/+) to mice carrying one wild-type Dnmt3a allele and one targeted allele that contains a neomycin-resistance cassette inserted into the sequence coding for the catalytic domain of the protein, producing a true null allele (Dnmt3a+/- mice). At 6 weeks of age, Dnmt3a+/- mice have normal hematopoiesis, with no detected differences from wild-type littermates in myeloid, lymphoid, erythroid, or stem/progenitor populations in the bone marrow or spleen. However, after 1.5 years of age, 15/43 Dnmt3a+/- mice (35%) became moribund and were euthanized for pathologic evaluation, and at conclusion of the tumor watch at 2 years similar pathologic findings were observed in an additional 9 Dnmt3a+/- mice, for an overall disease penetrance of 24/43 (56%). In contrast, 0/20 WT littermate control animals developed myeloid malignancies over the same time period. Based on flow cytometric and morphologic findings, we classified 16 splenic tumors according to the Bethesda criteria: 11/16 had myeloid proliferative disease/MPD, 2/16 had myeloid leukemia with maturation, 2/16 had MPD-like myeloid leukemia, and 1 case had myeloid sarcoma. Six tumors out of 18 tested were able to successfully engraft and lead to lethal disease in sublethally irradiated wild-type recipients, providing further evidence that these tumors represent transplantable, cell-autonomous myeloid malignancies. Exome sequencing of engrafted tumors revealed mutations in the Ras/MAPK pathway, including the canonical gain-of-function mutation Kras G12C, a Ptpn11 E76K mutation, and a missense mutation in the tumor suppressor Neurofibromatosis 1 (Nf1). Importantly, 9/51 AML samples with DNMT3A mutations in the TCGA AML cohort also contained activating NRAS or KRAS mutations. Examination of the Dnmt3a locus in 4 sequenced samples revealed no evidence for mutations in or deletions of the residual wild-type Dnmt3a allele. These data strongly suggest that Ras/MAPK pathway mutations can cooperate with Dnmt3a haploinsufficiency to induce AML in C57Bl/6 mice and in humans. .


English (en)

Chair and Committee

Timothy J. Ley

Committee Members

John Edwards, Daniel Link, Eugene Oltz, Jacqueline Payton,


Permanent URL:

Available for download on Thursday, October 12, 2017

Included in

Genetics Commons