Abstract Introduction: Epithelial to mesenchymal transition (EMT), where cells detach from basement membrane (BM) and invade interstitial matrix (IM), plays a significant role in tumor metastasis and drug resistance. Mechanobiological mechanisms of such tumor-stroma interactions are poorly understood largely due to inherent limitations of current preclinical tumor models (e.g. 2D cell culture and spheroids). The objective of this work was to apply a novel in-vitro 3D tumor-stromal model to define how extracellular matrix (ECM) composition and biophysical properties modulate EMT. Methods: This work involved pancreatic cancer cells representing different EMT phenotypes (BxPc-3: epithelial; Panc-1: intermediate; MiaPaCa-2: mesenchymal). Matrigel was used to represent BM, while tunable type I collagen Oligomers represented IM. Cells were encapsulated within 3D matrices composed of Matrigel, Oligomer of varied fibril density (matrix stiffness), or Oligomer and Matrigel combined at varied ratios. Cell phenotype was characterized based on morphology and EMT protein immunostaining (i.e., E-cadherin and vimentin). Dose response curves to gemcitabine (standard of care) were used to define relative IC50. ECM microstructure was visualized using confocal reflection microscopy, and matrix stiffness (G') was determined using rheometric oscillatory shear testing. Results: Oligomer enhanced the mesenchymal phenotype with decreasing stiffness (vimentin expression; spindle-shaped cells), while Matrigel downregulated mesenchymal properties and encouraged an epithelial phenotype (E-cadherin expression; rounded, grouped cells). While maintaining vimentin expression in all matrices, Panc-1 and MiaPaCa-2 grew as tight clusters within Matrigel and the highest stiffness Oligomer, where they showed the greatest gemcitabine resistance. BxPc-3 were most sensitive in intermediate stiffness Oligomer where they appeared most proliferative. They also showed relatively high drug resistance in Matrigel and matched-stiffness Oligomer despite each matrix supporting opposite phenotypes. Cells were also cultured within matrices of varied Oligomer:Matrigel ratio to define how BM and IM engagement guides EMT. In matrices with high Oligomer content (>50%), MiaPaCa-2 generated a notably more invasive, mesenchymal phenotype, while BxPc-3 only transitioned to a mesenchymal phenotype in pure Oligomer. Since stiffness and fibril density were similar in the 75:25 and 100:0 ratios, the presence of BM components appears to mediate the phenotypic transition between these matrices. Collectively, these results emphasize the importance of ECM composition and biophysical properties in guiding EMT. Findings suggest that epithelial tumor cells (BxPc-3) may require engagement with a low stiffness IM to undergo EMT and reduce drug sensitivity. For already transitioned cells (Panc-1 and MiaPaCa-2), decreasing stiffness enhances their mesenchymal nature but increases drug sensitivity. Together these observations align with the idea that EMT increases drug resistance, yet challenge the view that increasing matrix stiffness mediates this transition. Also, engagement with both matrix types promotes invasiveness of mesenchymal cells while exposure to BM suppresses EMT of epithelial cells. Finally, Matrigel promotes a drug resistant, epithelial phenotype which highlights its limitations for studying EMT mechanisms. Conclusion: This work represents the first use of type I collagen oligomers for mechanistic study of EMT mechanobiology and provides new insight into how IM fibril microstructure and mechanical properties guide EMT. These results challenge correlations between stromal properties and EMT established using conventional preclinical models and help identify critical parameters for engineering pathophysiologically relevant tumor-stromal models. Citation Format: TJ Puls, Xiaohong Tan, Catherine Whittington, Sherry Voytik-Harbin. Novel 3D tumor-stromal model highlights the importance of ECM composition and biophysical properties in pancreatic cancer EMT and drug resistance. [abstract]. In: Proceedings of the AACR Special Conference on Engineering and Physical Sciences in Oncology; 2016 Jun 25-28; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2017;77(2 Suppl):Abstract nr A57.