Disease relapse remains the major clinical challenge in treating T-cell acute lymphoblastic leukemia (T-ALL), particularly those with PTEN loss. We hypothesized that leukemia-initiating cells (LIC) are responsible for T-ALL development and treatment relapse. In this study, we used a genetically engineered mouse model of Pten(-/-) T-ALL with defined blast and LIC-enriched cell populations to demonstrate that LICs are responsible for therapeutic resistance. Unlike acute and chronic myelogenous leukemia, LICs in T-ALL were actively cycling, were distinct biologically, and responded differently to targeted therapies in comparison with their differentiated blast cell progeny. Notably, we found that T-ALL LICs could be eliminated by cotargeting the deregulated pathways driven by PI3K and Myc, which are altered commonly in human T-ALL and are associated with LIC formation. Our findings define critical events that may be targeted to eliminate LICs in T-ALL as a new strategy to treat the most aggressive relapsed forms of this disease.