Mdb-20. improving the persistence of car t cells against group 3 medulloblastomas via epigenetic regulation of exhaustion

Abstract Medulloblastomas (MB) are the most common malignant pediatric brain tumor with molecular subclass Group 3 (G3) requiring the most intense treatment regimen including surgery, intensive radiation followed by chemotherapy. Despite an overall 40% survival rate, survivors’ quality of life is often impaired due to whole-brain radiation leading to neurocognitive defects. Chimeric antigen receptor (CAR) T cell therapy has emerged clinically as a safe, tumor-targeted immunotherapy with the most success observed against hematological malignancies. Translating CAR T cell therapy to brain tumors has proven challenging due to lack of targetable antigens on the surface of tumor cells and lack of persistence of CAR T cells. We have previously identified that patient samples and PDOX-derived cell lines of G3 MBs consistently over-express B7-H3 (CD267) compared to normal tissue. Herein, we investigated B7-H3 targeted CAR T cell products against G3 MB in vitro and in vivo. Initial results show that B7-H3 CAR T cells are highly effective in clearing human MB tumors in NSG mice, however, 4 out 5 mice experienced late relapses that showed reduced, but still present, B7-H3 antigen expression. We hypothesize that the lack of CAR T cell persistence is ultimately responsible for tumor regrowth. To overcome this limitation, we have identified epigenetic regulators DNMT3A and TET2 that control T cell exhaustion. Using CRISPR/Cas9, we genetically knocked out DNMT3A and TET2 in CAR T cells and investigated the impact on persistence and anti-tumor activity. Our data supports the hypothesis that knock out of DNMT3A and TET2 improves antigen-dependent proliferation and persistence of B7-H3 CAR T cells against G3 MB tumor lines in vitro. DNMT3A KO shows a more sustained effect over TET2. Currently, we are optimizing in vivo experiments to investigate prolonged tumor clearance in both immunocompromised and fully immunocompetent murine models.