Abstract 4027: Co-targeting GPC2 and B7-H3 with CAR T cells achieves durable responses in preclinical high-risk neuroblastoma and medulloblastoma models

Abstract Glypican 2 (GPC2) is highly expressed in pediatric cancers including high-risk neuroblastomas and medulloblastomas, and GPC2 CAR T cells are now in phase 1 trials (NCT05650749, NCT07087002). However, durable responses are limited by heterogeneous antigen expression, treatment-induced antigen loss, and poor CAR T persistence. In neuroblastoma, interconvertible adrenergic (ADRN) and mesenchymal (MES) states further complicate targeting, as state switching downregulates GPC2 and GD2, whereas B7-H3 remains consistently expressed. B7-H3 is also abundant across other GPC2+ cancers, supporting a dual GPC2.B7-H3 CAR strategy. To approach GPC2.B7-H3 dual CAR T cell targeting, we engineered seven GPC2/B7-H3 co-targeting bicistronic CAR constructs containing CD28, 4-1BB, or OX40 costimulatory domains and either two (dual, d1-3) or one (parallel, p1-4) CD3-ζ signaling domain. Among these, d1 and p2 were prioritized because they drove strong activation of Jurkat NFAT-GFP CAR T cells in co-cultures with isogenic NALM-6 cells expressing either or both antigens and showed superior cytokine secretion in human CAR T co-incubation assays compared with single-antigen B7-H3 (B) or GPC2 (G) CARs. In vivo, dual-targeting CAR T cells - double-transduced (B/G) or expressing d1 or p2 - achieved potent clearance of COG-N-453x neuroblastoma PDXs, with 16-week cure rates of 83.3%, 83.3%, and 100%, exceeding single CARs (B, 57.1%; G, 0%). Human CAR T cell persistence was detected in bone marrow and spleen of cured mice at 16 weeks, suggesting a role of enhanced T cell persistence in durable remissions. In the aggressive COG-N-561x model, only bicistronic CARs d1 and p2 achieved durable 14-15-week cures (d1, 100%; p2, 71.4%), surpassing double-transduced (B/G, 33.3%), co-administered (B+G, 20%), or single CARs (B, 20%; G, 0%). Upon rechallenge, only d1 and p2 cohorts rejected tumors and remained tumor-free long-term (d1, 71.4%; p2, 60%), again with persistent CAR T cells in bone marrow and spleen through the 36-week study endpoint. Finally, across in vitro medulloblastoma models, repeat tumor-stimulation assays confirmed superior activity of d1 and p2. In an aggressive orthotopic group 3 medulloblastoma 7316-10374 xenograft model where 100% of GPC2 CAR-treated mice relapsed, the p2 CAR outperformed d1 and the single B7-H3 CAR, with 16-week cure rates of 100%, 71.4%, and 60%, respectively, confirming the superiority of dual GPC2.B7-H3 targeting and a potential advantage of parallel CD3ζ-sharing designs. Altogether, GPC2.B7-H3 bicistronic CARs induced durable and curative responses across neuroblastoma and medulloblastoma models, nominating this approach as a next-generation GPC2 CAR T strategy for clinical testing. Citation Format: Patrick M. Schürch, Benjamin Draper, Vincent P. Zecchino, Hehai Pan, Brandi Nelson, Anna M. Giudice, Jessica B. Foster, Laura K. Donovan, Kristopher R. Bosse. Co-targeting GPC2 and B7-H3 with CAR T cells achieves durable responses in preclinical high-risk neuroblastoma and medulloblastoma models [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4027.