Optimization of Antigen-Targeting Strategies for Chimeric Antigen Receptor T-Cell Therapies

ABSTRACT

Chimeric antigen receptor (CAR) T-cells are a powerful therapeutic technology that has shown exceptional efficacy in B-cell cancers and multiple myeloma, but only limited effects in other hematological malignancies or solid cancers. A major challenge has been developing and optimizing strategies for targeting cancer-specific cell surface targets without targeting healthy tissues, especially given the heterogeneity of cell-surface protein expression in most tumors.Here, I first develop a dual-antigen targeting CAR T-cell therapy against the antigens CD70 and the active conformation of integrin β2 (aITGB2), two antigens known to be expressed in acute myeloid leukemia (AML) but on very few healthy tissues. I show that an OR-gated approach for these antigens significantly increases the proportion of AML blasts that can be targeted. I identify dual-targeting CAR-T constructs with superior anti-tumor cytotoxicity in vitro against AML cell line and patient-derived xenograft models. I further show significantly improved in vivo tumor clearance and survival for a dual-targeting CAR in murine models of AML tumor heterogeneity. Finally, I show that this dual-targeting CAR does not increase off-tumor toxicity, especially against hematopoietic stem and progenitor cells. Together, these findings demonstrate a promising translatable approach for the treatment of AML without the notable toxicities commonly seen in other leading CAR-T targets for this disease. Next, I explore the use of deep learning algorithms to develop an optimized CAR binder against CD70 based on its native receptor, CD27. I generate several CD27-based CAR variants identified by a novel computational pipeline. I then test these variants in a battery of in vitro and in vivo tests, identifying a lead candidate mutation, denoted CD27N88A, with improved tumor killing, effector cytokine expression, and T-cell persistence. Molecular dynamics simulations of the receptor-ligand interaction predict that CD27N88A forms a more dynamic and energetically favorable interaction with CD70 compared to CD27WT. I utilize biochemical assays and fluorescent microscopy to support this prediction, revealing that CD27N88A CAR-Ts display a stronger affinity and avidity to CD70, greater segregation between receptors and ligand at the CAR synapse, and improved cytotoxic granule trafficking to the synapse. I further show that CD27N88A maintains therapeutic safety, with no observed off-target binding and no significant impact on hematopoietic stem and progenitor cell populations. Finally, we show that the CD27N88A CAR is highly effective in many CD70-expressing tumors, including AML, high-risk multiple myeloma, and renal cell carcinoma, demonstrating that this best-in-class anti-CD70 CAR-T therapy has great translational potential for a number of cancer indications.