P06.04 Integrated de novo design and high-throughput screening of T cell receptor mimetic minibinders for multipronged intracellular cancer antigen targeting

Background Intracellular antigens have remained beyond the reach of most immunotherapies due to limited accessibility. To overcome this limitation, we applied artificial intelligence (AI)–guided protein design to generate peptide–major histocompatibility complex (pMHC) class I–specific minibinders that mimic T cell receptor (TCR) recognition. Previous AI-designed pMHC binders, including our own, have primarily focused on high specificity. However, natural TCRs are inherently cross-reactive, enabling recognition of multiple tumor-associated antigens (TAAs) through a single TCR. Notably, multipronged T cells were shown to enhance tumor recognition across cancer types. 1 We therefore investigated whether AI-driven design can intentionally engineer multipronged pMHC minibinders. As a proof of principle, we targeted three TAA–derived (Melan A, BST2, IMP2) peptides that are known to be cross-reactively recognized in the context of HLA-A02:01.1 Material and methods Candidate minibinders were generated using RFdiffusion 2, BindCraft3, and BoltzGen4, followed by iterative refinement of RFdiffusion and BindCraft designs using a custom refinement algorithm, yielding a total of 12,000 designs. All candidates were cloned as an oligonucleotide pool and screened in parallel using the transposon display platform.5 Enriched binders were identified by next-generation sequencing.Results Screening yielded target-specific success rates of 0.4–6.1%. Among 375 enriched minibinders, 20 demonstrated multipronged binding to at least two pMHC targets. Complementary in silico analyses were performed to assess which metrics best predict multipronged design success. Ongoing validation aims at testing soluble binders and chimeric antigen receptor T cell and bi-specific T cell engager formats, while our comprehensive in silico database supports future metric refinement.Conclusion This work introduces a paradigm shift in binder design, in which multipronged cross-reactivity is intentionally engineered as a therapeutic advantage. By targeting multiple TAAs, such minibinders may mitigate antigen escape, reduce resistance, and enable pan-cancer applicability through conserved peptide motifs, while enhancing sensitivity in low-antigen-density tumors. We establish a proof of concept for the intentional design of multipronged pMHC binders and outline both the potential and current limitations of de novo TCR-mimetic design approaches.References Dolton G, Rius C, Wall A, et al. Targeting of multiple tumor-associated antigens by individual T cell receptors during successful cancer immunotherapy. Cell. 2023;186(16):3333–3349.e27.Watson JL, Juergens D, Bennett NR, et al. De novo design of protein structure and function with RFdiffusion. Nature 2023;620(7976):1089–1100.Pacesa M, Nickel L, Schellhaas C, et al. One-shot design of functional protein binders with BindCraft. Nature. 2025;646(8084):483–492.Stark H, Faltings F, Choi M, et al. BoltzGen: toward universal binder design. bioRxiv. 2025:2025.11.20.689494.Hoch A, Fandrey CI, Coventry B, et al. Transposon-display of AI-designed binders enables manipulation of the proteome in human cells. bioRxiv.2025:2025.05.11.653342.