Anti-CAR19 CART cells to deplete CAR19⁺ leukemic cells. (A) Model: anti-CAR19 CART (αCAR19-CART) to recognize CAR19-expressing leukemic cells and trigger cytotoxicity. (B) Two novel CAR constructs were designed by using an anti-CAR19 scFv (clone 136.20.1) (two light chain orientations, H-L and L-H), a CD8 hinge, the 4-1BB costimulatory domain, CD3 zeta-signaling domain, and the pTRPE vector backbone. (C) The H-L and L-H αCAR19 CAR are efficiently expressed on T cells as detected by flow cytometry using a goat anti-human Ig Fab. (D) In vitro αCAR19-CART (L-H and H-L) efficiently kill CAR19⁺ B-ALL (NALM6) but not wild-type (WT) NALM6. The L-H construct exhibits a higher antileukemia effect and was therefore also tested in vivo. (E) NOD/SCID γ chain–deficient (NSG) mice were engrafted with luciferase-positive relapsed patient #107 leukemia blasts; then, at day 14, mice were randomized to receive no treatment, control T cells (untransduced [UTD]), CART19, CART22, or αCAR19-CART (1 × 10⁶ cells/mouse). (F) CART19 and UTD cells are not able to control disease progression, but CART22 and αCAR19-CART L-H cells exhibit significant leukemia control. All experiments were repeated twice. (G) In an additional experiment, NSG mice were engrafted with luciferase-positive CAR19⁺ leukemia blasts obtained at relapse from patient #107; at day 14, mice were randomized to receive no treatment, UTD, αCAR19-CART H-L, or αCAR19-CART L-H. UTD cells are not able to control disease progression, but both αCAR19-CARTs, in particular L-H, show an advantage in overall survival. Survival curves were compared by using the log-rank test. Asterisks are used in each figure to represent P values (*P < .05, **P < .01, ****P < .0001).