AbstractObjectivesCAR‐T cells are being investigated as a novel immunotherapy for glioblastoma, but clinical success has been limited. We recently described fibroblast activation protein (FAP) as an ideal target antigen for glioblastoma immunotherapy, with expression on both tumor cells and tumor blood vessels. However, CAR‐T cells targeting FAP have never been investigated as a therapy for glioblastoma.MethodsWe generated a novel FAP targeting CAR with CD3ζ and CD28 signalling domains and tested the resulting CAR‐T cells for their lytic activity and cytokine secretion function in vitro (using real‐time impedance, flow cytometry, imaging and bead‐based cytokine assays), and in vivo (using a xenograft mimicking the natural heterogeneity of human glioblastoma).ResultsFAP‐CAR‐T cells exhibited target specificity against model cell lines and potent cytotoxicity against patient‐derived glioma neural stem cells, even when only a subpopulation expressed FAP, indicating a bystander killing mechanism. Using co‐culture assays, we confirmed FAP‐CAR‐T cells mediate bystander killing of antigen‐negative tumor cells, but only after activation by FAP‐positive target cells. This bystander killing was at least partially mediated by soluble factors and amplified by IL‐2 which activated the non‐transduced fraction of the CAR‐T product. Finally, a low dose of intravenously administered FAP‐CAR‐T cells controlled, without overt toxicity, the growth of subcutaneous tumors created using a mixture of antigen‐negative and antigen‐positive glioblastoma cells.ConclusionsOur findings advance FAP as a leading candidate for clinical CAR‐T therapy of glioblastoma and highlight under‐recognised antigen nonspecific mechanisms that may contribute meaningfully to the antitumor activity of CAR‐T cells.