Quantitative ECM Profiling Reveals a Recapitulation Gap in a Subcutaneous KPC-Derived Syngeneic PDAC Model

Background/Objectives: Pancreatic ductal adenocarcinoma (PDAC) develops within a dense desmoplastic stroma that shapes drug transport, mechanotransduction, and therapeutic response. Building on the recapitulation gap framework, we examined whether a commonly used subcutaneous KPC-derived syngeneic model reproduces human PDAC extracellular matrix (ECM) architecture. Methods: Dual-stain histochemistry and computational image analysis compared primary human PDAC tumor cores (n = 14 patients) with independently generated syngeneic tumors (n = 3), quantifying collagen and glycosaminoglycan (GAG) signals across compositional, spatial, and textural domains. Inference used one value per patient or animal, and multivariate separation was tested by exact permutation. Local organization was investigated by analyzing 65 simulated cores (sim-cores) from the mouse specimens. Results: Mouse tumors showed collagen-dominated architectural divergence, with lower collagen area, approximately seven-fold greater peripheral edge share, and a 2.8-fold lower collagen-to-GAG signal ratio. Human and mouse specimens separated in observed multivariate ECM space (standardized centroid distance 5.18; exact permutation p = 1/680); the separation persisted in format-matched sim-cores (p = 1/680). Analysis at the two scales identified peripheral collagen organization alongside lower spatial persistence and band fraction. Conclusions: Quantitative ECM profiling revealed a recapitulation gap in this model configuration. These findings support feature-specific, fit-for-purpose evaluation when tumor matrix architecture is relevant to the question being studied.