Objective:

To evaluate the effectiveness of AI-designed protein binders in enhancing CAR T cell activity against tumors compared to traditional antibody-derived binders.

Approach:

• Study Design: The study involved generating and screening 1,758 de novo binders against BCMA, CD19, and CD22, followed by assessments using protein-binding assays, CAR activation studies, and mouse models.
• AI Binder Development: AI-designed binders were created from scratch, focusing on their small size and ability to target specific epitopes.
• Identification of Issues: The researchers identified issues such as tonic signaling, inaccessible binding sites, and off-target activity, and developed rules to address these problems.
• Performance Evaluation: The performance of the binders was evaluated in mouse models, with a focus on BCMA as a target for multiple myeloma.

Key Findings:

• The AI-designed binder B5 showed antigen-dependent activation and cytokine release in primary CAR T cells.
• An optimized variant, B5.I0, controlled tumor growth more effectively than existing BCMA-targeting therapies.
• The study identified excessive positive charge as linked to unwanted tonic signaling and high alanine content as a barrier to effective binding.
• Modifications to binders targeting CD22 improved on-target activity while reducing off-target activation.

Limitations:

• The study remains preclinical with small treatment groups in mouse experiments.
• The CD19 program showed limited success due to obstruction of the targeted epitope by CD81.

Sources:

• Memorial Sloan Kettering Cancer Center