We use AI to tease apart cells chemo and mechano responsiveness to its microenvironment and show based on the analysis of single cell phenotypic data of geometric measures such as area, circularity or aspect ratio, mechanical measures such as cell stiffness, and transport measures such as persistence in motion and motility can be categorized into distinct subclasses.
Alterations of the extracellular matrix (ECM), including both mechanical (such as stiffening of the ECM) and chemical (such as variation of adhesion proteins and deposition of hyaluronic acid (HA)) changes, in malignant tissues have been shown to mediate tumor progression. To survey how cells from different tissue types respond to various changes in ECM mechanics and composition, we measured physical characteristics (adherent area, shape, cell stiffness, and cell speed) of 25 cancer and 5 non-tumorigenic cell lines on 7 different substrate conditions. Our results indicate substantial heterogeneity in how cell mechanics changes within and across tissue types in response to mechanosensitive and chemosensitive changes in ECM. The analysis also underscores the role of HA in ECM with some cell lines showing changes in cell mechanics in response to presence of HA in soft substrate that are similar to those observed on stiff substrate. This pan-cancer investigation also highlights the importance of tissue-type and cell line specificity for inferences made based on comparison between physical properties of cancer and normal cells. Lastly, using unsupervised machine learning, we identify phenotypic classes that characterize the physical plasticity, i.e. the distribution of physical feature values attainable, of a particular cell type in response to different ECM-based conditions.