De Novo Design of a TROP-2-Targeting Peptide PET Tracer via AI-Guided Binding Interface Mimicry

Abstract

Trophoblast cell surface antigen 2 (TROP-2) is a transmembrane glycoprotein overexpressed across a range of epithelial malignancies, yet its clinical assessment still relies largely on invasive immunohistochemical analysis of biopsied tissues. In this study, we employed the PepMimic artificial intelligence platform to de novo design a TROP-2-targeting peptide, TR23, via binding interface mimicry, which was subsequently conjugated with DOTA and radiolabeled with ⁶⁸Ga to yield the peptide-based PET probe [⁶⁸Ga]Ga-DOTA-TR23. The tracer bound TROP-2 with nanomolar affinity (KD = 26.8 nM), showed selective cellular uptake in TROP-2-positive cells, and exhibited high radiochemical purity (>95%) with excellent stability in saline and serum. Micro-PET/CT imaging in pancreatic (BxPC-3), prostate (PC3), and thyroid (BCPAP) xenograft models revealed rapid, specific tumor accumulation with high contrast and low background uptake, while competitive blocking studies with excess unlabeled TR23 substantially suppressed tumor uptake, confirming receptor-mediated specificity. Quantitative analysis further demonstrated strong positive correlations between SUVmax and TROP-2 expression levels verified by immunohistochemistry across all three tumor types. Collectively, these findings establish [⁶⁸Ga]Ga-DOTA-TR23 as a promising peptide-based PET tracer for rapid, specific, and quantitative imaging of TROP-2-positive tumors, and highlight the PepMimic-enabled strategy as a versatile platform for peptide-based probe discovery, supporting the translational potential of this tracer as a pan-cancer diagnostic agent for precision molecular imaging.