(814) Decoding the Hallmark Features of Modular CAR Domains and Their Impact on CAR T Cell Tumoricidal Activity

Introduction/Rationale: Engineering T cells to express a Chimeric Antigen Receptor (CAR) directs potent tumor cytotoxicity, and this cellular immunotherapy has yielded robust success in hematological malignancies. However, in solid tumor models, including neuroblastoma, regression after CAR administration is only transient. The linked domains of CARs determine their activation threshold and can be optimized to overcome efficacy barriers, including exhaustion, persistence, and activation threshold. Yet, as to how each modular domain and its variants influence T cell phenotype and function is poorly understood.

Methods: We generated a library of 12 constructs that combinatorially link 3 distinct hinge regions (CD8, CD28, or IgG), 2 transmembrane domains (CD8 or CD28), and 2 co-stimulatory regions (41BB or CD28), all frequently used in FDA-approved CARs. We used a scFv targeting the oncofetal antigen GPC2, an antigen with moderate to low expression in neuroblastoma. Expressing these unique CARs in human T cells, we will evaluate differences in T cell cytokine production, killing capacity, phenotype, and persistence using preclinical models of neuroblastoma. These assays will generate a high-dimensional dataset, and we will employ machine learning to identify an in vitro biosignature of CAR T cells that predicts in vivo CAR T cell killing, as well as a favorable phenotype for persistence and effector qualities.

Results: Preliminary results identified a broad range of CAR T cell cytotoxicity against neuroblastoma lines in vitro, with the transmembrane domain driving the greatest variation in killing, as well as significant interaction effects between domains.

Conclusion: Overall, we aim to create a model to establish the role of each domain and how they interact to propagate an activation signal, recruit a unique signalosome, and dictate the phenotype and function of CAR T cells.