Abstract 4009: Decoding the medulloblastoma surfaceome prioritizes the oncofetal antigen GPC2 for potent CAR-T cell therapy

Abstract Key determinants for efficacious CAR-T antigens remain near absence in vital tissues and that antigen levels in clinical specimens exceed CAR-T detection thresholds. CAR-T cells meeting these criteria have demonstrated clinical benefit in aggressive pediatric brain tumors, as seen in H3K27M-mutant diffuse midline glioma responding to GD2-directed CAR-T cell therapy. Medulloblastoma (MB) is a devastating childhood brain tumor, for which standard-of-care is fails in 30% of patients, and leaves many survivors with debilitating long-term effects. Survival rates of recurrent disease reside below <10%. Increasing cure rates and diminishing treatment-related morbidity necessitates targeted therapies, but success using CAR-T cells in MB has been limited by a dearth of suitable cell-surface targets. We profiled the CAR-T antigen landscape of MB through integrating unbiased surfaceome analysis from tumor and normal tissue transcriptomes and quantitative antigen density mapping of candidate targets on n= 24 MB tumor biopsies across all molecular subtypes. We found that MB closely resembles prenatal brain and credentialled oncofetal antigens as the predominant target pool in MB. A quantitative flow cytometry-based assay to precisely quantify antigen densities of candidate targets in clinical specimens prioritized GPC2 as the top candidate. GPC2 antigen densities ranged from ∼1700-24500 molecules/cell across MB subgroups (mean: 6374 +/-4968). Because antigen density greatly influences CAR-T potency, we previously engineered GPC2-CAR-Ts tuned towards clinical densities on neuroblastoma and demonstrated that cJUN-overexpression (OE) augments their potency by lowering antigen detection thresholds (Heitzeneder et. al, 2022). Here, we employ orthotopic xenograft models and locoregional CAR-T delivery to study the interplay between GPC2 antigen density and CAR potency and to inform a phase-1 trial design. We found that both GPC2-CAR-T constructs (+/-cJUN) achieved durable disease control against GPC2-intermediate GR4-MB (ICB1299: ∼9000 mol/cell and MBT375, a newly established PDX: ∼6500 mol/cell). In a highly aggressive, MYC-amplified, GPC2-high GR3-MB model (SU_MB002: ∼18800 mol/cell), recurrence occurred in 3/5 mice post GPC2-CAR, whereas repeated dosing or cJUN-OE significantly improved persistence and long-term anti-tumor control. In a GPC2-low, GR3-MB model (HDMB03: ∼3000 mol/cell) used to further dissect antigen detection limits, only cJUN.GPC2-CAR-T cells maintained anti-tumor responses. This work informed an ongoing phase-1 clinical trial at Stanford (NCT07087002), currently testing repeated, locoregional GPC2-CAR-T infusions and is planned to employ cJUN overexpression in patients with recurrent/refractory MB, aiming to provide a targeted treatment option for this devastating malignancy. Citation Format: Diren Usta, William Gwynne, Yujin Suk, Yiyun Chen, Molly T. Radosevich, Daria Chernova, Yushen Feng, Emon Nasajpour, Maria Trissal, Rebecca Poetschke, Alberto Delaidelli, Christopher Dunham, Louai Labanieh, Jasper van der Lugt, Stefan Nierkens, Mariella G. Filbin, Claudia Petrisch, Poul H. Sorensen, Katie Ryan, Chitra Venugopal, Elena Sotillo-Piñeiro, Crystal L. Mackall, Sheila Kumari Singh, Sabine Heitzeneder. Decoding the medulloblastoma surfaceome prioritizes the oncofetal antigen GPC2 for potent CAR-T cell therapy [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 4009.