T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive malignancy largely caused by aberrant activation of the TAL1/SCL, LMO1/2, and NOTCH1 oncogenes. Approximately 30% of T-ALL patients relapse, and evidence is emerging that relapse may result from a failure to eliminate leukemia-initiating cells (LICs). Thymic expression of the Tal1 and Lmo2 oncogenes in mice results in rapid development of T-ALL; and similar to T-ALL patients, more than half the leukemic mice develop spontaneous mutations in Notch1. Using this mouse model, we demonstrate that mouse T-ALLs are immunophenotypically and functionally heterogeneous with approximately 1 of 10,000 leukemic cells capable of initiating disease on transplantation. Our preleukemic studies reveal expansion of Notch-active double-negative thymic progenitors, and we find the leukemic DN3 population enriched in disease potential. To examine the role of Notch1 in LIC function, we measured LIC activity in leukemic mice treated with vehicle or with a gamma-secretase inhibitor. In 4 of 5 leukemias examined, Notch inhibition significantly reduced or eliminated LICs and extended survival. Remarkably, in 2 mice, gamma-secretase inhibitor treatment reduced LIC frequency below the limits of detection of this assay, and all transplanted mice failed to develop disease. These data support the continued development of Notch1 therapeutics as antileukemia agents.