Chemoresistance is a common cause of treatment failure in patients with acute myeloid leukemia (AML). We generated a diphtheria toxin (DT) fusion protein composed of the catalytic and translocation domains of DT (DT388) fused to interleukin-3 (IL-3). IL-3 receptors (IL-3R) are overexpressed on blasts from many AML patients. DT388IL-3 showed cytotoxicity to leukemic blasts in vitro and in vivo and minimal damage to normal tissues in nonhuman primate models. However, only a fraction of patient leukemic samples were sensitive to the agent. To enhance the potency and specificity of the DT388IL-3 molecule, we constructed variants with altered residues in the IL-3 moiety. Two of these variants, DT388IL-3[K116W] and DT388IL-3[Delta125-133], were produced and partially purified from Escherichia coli with excellent yields. They showed enhanced binding to the human IL-3R and greater cytotoxicity to human leukemia cell lines relative to wild-type DT388IL-3. Interestingly, the results support a previously hypothesized model for interaction of the C-terminal residues of IL-3 with a hydrophobic patch on the alpha-subunit of IL-3R. Rational modification of the targeting domain based on structural analysis can produce a fusion toxin with increased ability to kill tumor cells. One or both of these variant fusion proteins merit further development for therapy of chemotherapy refractory AML.