Investigating genetic modifications to enhance L1CAM-CAR T cell migration in solid tumors in a 3D bioprinted neuroblastoma model

Introduction: Effective CAR T cell infiltration into solid tumors remains a major barrier to therapy success. Despite their clinical potential, few studies have evaluated phenotypes of CAR T cells successfully invading the tumor mass following infusion. Phenotypic information would enrich our understanding of the mechanisms governing CAR T cell migration into solid tumors. Here we implemented an in vitro strategy to identify genes driving L1CAM-CAR T cell migration into a 3D tumor mass. Methods: L1CAM-CAR T cells were separated into 2 groups by their capability to infiltrate (or not) a 3D bioprinted neuroblastoma model. Single-cell and bulk RNA sequencing was performed, and infiltrating CAR T cells were compared to noninfiltrating cells to seek genetic drivers of CAR T cell migration. CRISPR/Cas9 technology was used to generate modified L1CAM-CAR T cells. Results: Tumor-infiltrating L1CAM-CAR T cells expressed lower levels of the selectin P ligand (SELPLG) glycoprotein and higher levels of the T cell-specific adaptor protein, SH2D2A. Functional characterization of L1CAM-CAR T cells genetically modified to enforce these characteristics demonstrated that neither trait negatively impacted L1CAM-CAR T cell cytotoxicity, activation and cytokine release upon coculture with neuroblastoma target cells. Transgenic SH2D2A expression did not improve CAR T cell migration in an endothelial transmembrane assay. SELPLG knockout benefited CAR T cell in vitro trans-endothelial migration, but did not enhance anti-tumor efficacy in an immunodeficient mouse model. Discussion: Our findings reveal a key limitation of murine xenograft models, which are widely used as the gold standard for preclinical CAR T cell testing. The lack of conservation between the human and murine SELPLG proteins likely accounts for the discrepancy between enhanced in vitro migration of SELPLG-deficient L1CAM-CAR T cells and their lack of improved efficacy in the mouse model. This underscores the need for more predictive human-relevant models to better preclinically evaluate CAR T cell function.