AI-assisted development of a TCR-like T cell engager targeting MAGE-A4 against melanoma

MAGE-A4 is a promising immunotherapeutic target due to its frequent overexpression in cancers and limited expression in normal tissues. Intracellularly processed MAGE-A4 peptides, frequently presented by HLA-A*02:01 on the cell surface, can be recognised by T-cell receptor-mimetic agents. Utilising artificial intelligence-driven in silico screening and experimental validation, we generated a T-cell receptor-like antibody that specifically binds to the HLA-A*02:01-restricted GVY230-239 peptide complex. We subsequently engineered a T cell engager (TCE) in an IgG(L)-scFv format by fusing it with an anti-CD3 scFv. This MAGE-A4-targeting TCE demonstrated potent and specific cytotoxicity against multiple HLA-A*02:01 + /MAGE-A4 + tumour cell lines. To eliminate potential interference from Fc-mediated ADCC on T-cell-mediated killing, an N297A mutation was introduced into the Fc region, and the loss of ADCC activity was confirmed using a CD16 reporter cell line. Furthermore, in an HLA-A*02:01 + /MAGE-A4 + cell-derived xenograft (CDX) mouse model, the TCE exhibited significant antitumor efficacy in vivo. Compared to most currently available TCR-like antibodies, which primarily target haematological malignancies, this TCE showed superior tumour-killing activity against melanoma in vitro and in vivo. This work is supported by FDCT/009/2023/RIC and FDCT/0150/2025/AFJ.