Outcomes for patients with glioblastoma (GBM) remain poor despite aggressive multimodal therapy. Immunotherapy with genetically modified T cells expressing chimeric antigen receptors (CARs) targeting interleukin (IL)-13Rα2, epidermal growth factor receptor variant III (EGFRvIII), or human epidermal growth factor receptor 2 (HER2) has shown promise for the treatment of gliomas in preclinical models and in a clinical study (IL-13Rα2). However, targeting IL-13Rα2 and EGFRvIII is associated with the development of antigen loss variants, and there are safety concerns with targeting HER2. Erythropoietin-producing hepatocellular carcinoma A2 (EphA2) has emerged as an attractive target for the immunotherapy of GBM as it is overexpressed in glioma and promotes its malignant phenotype. To generate EphA2-specific T cells, we constructed an EphA2-specific CAR with a CD28-ζ endodomain. EphA2-specific T cells recognized EphA2-positive glioma cells as judged by interferon-γ (IFN-γ) and IL-2 production and tumor cell killing. In addition, EphA2-specific T cells had potent activity against human glioma-initiating cells preventing neurosphere formation and destroying intact neurospheres in coculture assays. Adoptive transfer of EphA2-specific T cells resulted in the regression of glioma xenografts in severe combined immunodeficiency (SCID) mice and a significant survival advantage in comparison to untreated mice and mice treated with nontransduced T cells. Thus, EphA2-specific T-cell immunotherapy may be a promising approach for the treatment of EphA2-positive GBM.