Artificial intelligence-assisted design of MAGE-A4 ×CD16 T-cell receptor-like natural killer cell engagers.

The cancer-testis antigen MAGE-A4 is an attractive immunotherapy target due to its high expression in various malignancies and restricted expression in normal tissues. Peptides derived from intracellular MAGE-A4 are presented on the cell surface by HLA-A*02:01, enabling targeting by T-cell receptor (TCR)-like molecules. We identified a TCR-like antibody recognizing the HLA-A *02:01-presented MAGE-A4 GVY230-239 peptide complex through yeast-display screening, computational structural prioritization, and experimental validation. Diverging from T-cell engagers, we engineered this binder into a natural killer cell engager (NKCE) by fusing it with a CD16-targeting variable domain of heavy-chain-only antibody (VHH) in an IgG format. This MAGE-A4-directed NKCE demonstrated potent and specific cytotoxicity against MAGE-A4+/HLA-A*02:01+ tumor cell lines. To distinguish conventional Fc-mediated signaling from VHH-mediated CD16a engagement, we introduced the Fc-silencing N297A mutation. The mutation markedly reduced reporter activation by the parental IgG format, whereas the VHH-containing NKCE retained CD16a reporter activity. Reporter activation was measurable in the absence of target cells and increased further following addition of A375 cells. In cytotoxicity assays, the NKCE mediated target- and CD16a-dependent killing in the tested in vitro models. These findings provide in vitro proof of concept for a MAGE-A4-directed NKCE and support further preclinical evaluation, including cytokine profiling, safety assessment, and in vivo efficacy studies.