(454e) Yeast Surface Display Techniques Enhance Development of Chimeric Antigen Receptors for Hematologic Malignancies

Chimeric antigen receptor (CAR) engineered T cells have energized the field of cancer immunotherapy with their proven ability to treat CD19⁺ malignancies in the clinic and emerging efficacy in treating other diseases. The synthetic CAR receptor imparts T cells with the ability to recognize antigens independent of peptide presentation by major histocompatibility complexes. The CAR antigen recognition domain often requires much engineering to generate binding interactions resulting in target-specific activation. Unfortunately, developing a functional CAR with current standard methods is empirical and inefficient, requiring individual cloning of combinations of binding domains (often scFvs) and spacer regions followed by low throughput well plate and mouse experiments to verify activity. The scFvs incorporated into CARs often derive from published antibody sequences reformatted into scFvs, which frequently results in a loss of affinity and/or stability. These antibodies are regularly of murine origin, raising the risk of immune response that may hamper treatment.

This work aims to enhance the efficiency and success of CAR development by incorporating protein engineering and combinatorial library screening using yeast surface display techniques to facilitate scFv humanization and affinity tuning, and applying the resulting CARs for treatment of hematologic malignancies.

Two computational humanization methods were applied to the antigen recognition domain of a CD19-targeted CAR previously developed in the Forman lab. Yeast surface display techniques were applied to screen expression, foldedness, and affinity of the humanized variants, adding significant efficiency to the CAR development process that generally requires lentiviral transduction of cell lines to verify these characteristics. The two humanized scFvs displayed well, but exhibited markedly diminished affinity (8 ± 1 µM and 45 ± 30 µM, respectively) compared to the murine parent (4.8 ± 0.7 nM). The humanized scFvs were formatted into CARs and expressed in human T cells. Despite the decreased affinity, the resulting CARs showed comparable activity to the equivalent murine versions in degranulation, internal cytokine staining, proliferation, and long-term killing assays against CD19+ Daudi, TM-LCL, and SUP-B15 cells. Using a systemic CD19+ NALM6 model in NSG mice, the 8 µM affinity humanized variant showed comparable therapeutic efficacy to the murine parent. Additional efforts to affinity mature these humanized variants using yeast surface display will be discussed.

In a second project, a CD123-targeted CAR developed previously in the Forman lab was humanized. The resulting clone suffered a 575-fold affinity decrease (0.26 ± 0.1 nM vs 150 ± 40 nM) relative to the murine parent. Error-prone PCR was applied to the humanized variant, and the resulting sublibrary was yeast displayed as a novel CAR-like scFv-IgG4 hinge fusion and sorted for increased or diminished binding, yielding softly mutated variants with 5.5 ± 0.2 nM, 62 ± 2 nM, and 8.6 ± 2 µM affinity. In ongoing experiments, the resulting CARs will be screened for selectivity against CD123⁺ and CD123-dim cells for treatment of acute myeloid leukemia.

Overall, this work will greatly increase the efficiency of CAR development by streamlining this process using computational tools and high throughput screening. The insight that we gain from the interplay of CAR affinity, antigen density, and T cell response will be valuable in informing future CAR design.