Mutations in DNA methyltransferase 3A (DNMT3A) are prevalent in various myeloid and lymphoid malignancies. The most common DNMT3A R882 mutations inhibit methyltransferase activity of the remaining wild-type DNMT3A proteins at a heterozygous state due to their dominant-negative activity. Reports and COSMIC database analysis reveal significantly different frequencies of R882 mutations in myeloid versus T-cell malignancies, inspiring us to investigate whether downregulation of DNMT3A regulates malignancies of different lineages in a dose-dependent manner. In a competitive transplant setting, the survival of recipients with Kras G12D/+; Dnmt3a+/-bone marrow (BM) cells was significantly shortened than that of recipients with Kras G12D/+ cells. Moreover, all of the recipients with Kras G12D/+; Dnmt3a+/-cells developed a lethal T-cell acute lymphoblastic leukemia (T-ALL) without significant myeloproliferative neoplasm (MPN) phenotypes, while~ 20% of recipients with Kras G12D/+ cells developed MPN with or without T-ALL. This is in sharp contrast to the recipients with Kras G12D/+; Dnmt3a-/-cells, in which~ 60% developed a lethal myeloid malignancy (MPN or acute myeloid leukemia [AML]). Our data suggest that in the context of oncogenic Kras, loss of Dnmt3a promotes myeloid malignancies, while Dnmt3a haploinsufficiency induces T-ALL. This dose-dependent phenotype is highly consistent with the prevalence of DNMT3A R882 mutations in AML versus T-ALL in human.