Abstract
Matrix stiffness is a defining feature of pancreatic ductal adenocarcinoma (PDAC) and drives malignant progression through mechanisms that remain poorly understood. Using an ensemble machine learning approach, we integrated multiomics data from 886 patients with mechano-biology models to develop a nine-gene matrix stiffness-associated gene signature (MSAGS). MSAGS demonstrated superior prognostic accuracy in independent cohorts, outperforming 87 existing signatures. We identified SLC20A1 within MSAGS as a novel mechano-immunological checkpoint where matrix stiffness activates an SLC20A1-STAT3 positive feedback loop, driving dual immunosuppression via TGF-β1-mediated CD8+ T-cell exclusion and PD-L1-induced T-cell dysfunction. Importantly, targeting SLC20A1 synergized with anti-PD-L1/TGF-beta bispecific antibody (BiTP, Y101D), enhancing tumor suppression and extending survival in orthotopic PDAC models by increasing the infiltration and function of cytotoxic CD8+ T cells. This work establishes MSAGS as a clinically translatable prognostic tool and positions SLC20A1 targeting as a transformative strategy to overcome PDAC immunotherapy resistance, thereby repositioning matrix stiffness as a druggable target.