Abstract
The lack of inadequate preclinical models remains a limitation for cancer drug development and is a primary contributor to anti-cancer drug failures in clinical trials. Heterotypic multicellular spheroids are three-dimensional (3D) spherical structures generated by self-assembly from aggregates of two or more cell types. Compared to traditional monolayer cell culture models, the organization of cells into a 3D tissue-like structure favors relevant physiological conditions with chemical and physical gradients as well as cell-cell and cell-extracellular matrix (ECM) interactions that recapitulate many of the hallmarks of cancer in situ. Epidermal growth factor receptor (EGFR) mutations are prevalent in non-small cell lung cancer (NSCLC), yet various mechanisms of acquired resistance, including epithelial-to-mesenchymal transition (EMT), limit the clinical benefit of EGFR tyrosine kinase inhibitors (EGFRi). Improved preclinical models that incorporate the complexity induced by epithelial-to-mesenchymal plasticity (EMP) are urgently needed to advance new therapeutics for clinical NSCLC management. This study was designed to provide a thorough characterization of multicellular spheroids of isogenic cancer cells of various phenotypes and demonstrate proof-of-principle for the applicability of the presented spheroid model to evaluate the impact of cancer cell phenotype in drug screening experiments through high-dimensional and spatially resolved imaging mass cytometry (IMC) analyses. First, we developed and characterized 3D homotypic and heterotypic spheroid models comprising EGFRi-sensitive or EGFRi-resistant NSCLC cells. We observed that the degree of EMT correlated with the spheroid generation efficiency in monocultures. In-depth characterization of the multicellular heterotypic spheroids using immunohistochemistry and high-dimensional single-cell analyses by IMC revealed intrinsic differences between epithelial and mesenchymal-like cancer cells with respect to self-sorting, spatiotemporal organization, and stromal cell interactions when co-cultured with fibroblasts. While the carcinoma cells harboring an epithelial phenotype self-organized into a barrier sheet surrounding the fibroblasts, mesenchymal-like carcinoma cells localized to the central hypoxic and collagen-rich areas of the compact heterotypic spheroids. Further, deep-learning-based single-cell segmentation of IMC images and application of dimensionality reduction algorithms allowed a detailed visualization and multiparametric analysis of marker expression across the different cell subsets. We observed a high level of heterogeneity in the expression of EMT markers in both the carcinoma cell populations and the fibroblasts. Our study supports further application of these models in pre-clinical drug testing combined with complementary high-dimensional single-cell analyses, which in turn can advance our understanding of the impact of cancer-stroma interactions and epithelial phenotypic plasticity on innate and acquired therapy resistance in NSCLC.