3D-printed niche-matched GelMA-dECM bioinks rewire lipid metabolic reprogramming and chemosensitization in pancreatic ductal adenocarcinoma.

Abstract

Primary pancreatic ductal adenocarcinoma (PDAC) and hepatic metastases grow in anatomically distinct niches defined by organ-specific extracellular matrix (ECM) cues related to matrix composition and biomechanics such as substrate stiffness. Metabolic remodeling is a fundamental mechanism by which PDAC cells adapt to environmental stress. However, whether ECM actively imposes site-dependent metabolic liabilities remains unclear. To this end, we developed a digital light processing-printable GelMA-dECM platform incorporating decellularized pancreas- or liver-derived ECM under controlled three-dimensional culture conditions. Single-cell analysis and immunohistochemical staining of human tissues revealed a site-associated metabolic divergence, with primary pancreatic lesions enriched in cholesterol/mevalonate associated markers and liver metastases displaying stronger fatty acid synthesis signatures. Consistently, pancreas-dECM promoted HMGCS1/HMGCR-associated cholesterol and mevalonate programs in PDAC cells, whereas liver-dECM induced FASN/ACC1-associated fatty acid synthesis and lipid droplet accumulation. Increased matrix stiffness enhanced YAP activation and amplified lipid metabolic output without overriding the dECM-defined metabolic direction, indicating a mechanical amplifier role. Functionally, these matrix-encoded metabolic states predicted therapeutic vulnerability. Cholesterol/mevalonate pathway inhibition preferentially sensitized pancreas-dECM constructs and pancreatic-site tumors to gemcitabine, whereas fatty acid synthesis inhibition was more effective in liver-dECM constructs and liver-site tumors. These findings identify that organ-specific ECM defines the direction of lipid metabolic programming, whereas matrix stiffness amplifies the magnitude of this program. Together, these matrix-encoded metabolic states create site-matched therapeutic vulnerabilities in PDAC. Our study provides a potentially generalizable platform for supporting scalable microenvironment-guided modeling of site-specific tumor metabolism and therapeutic screening.