585 Identifying cancer stem cells in patient specimens through high-content imaging - biological insights and prognostic potential

Abstract

Cancer stem cells undergo epithelial-to-mesenchymal transition (EMT) to acquire a migratory phenotype, break away from the primary tumour body and invade into the stroma. They then undergo the reverse process of mesenchymal-to-epithelial transition (MET) at a distant site to form a metastatic tumour. Our group has identified a subpopulation of oral cancer stem cells that possess a hybrid EMT/MET phenotype characterised by vimentin (VIM), Epcam (ESA) and CD24 expression. These hybrid cancer stem cells are associated with aggressive invasion, chemotherapeutic resistance and metastatic potential in in vitro assays. The aim of this study is to identify these hybrid stem cells in patient Formalin-Fixed Paraffin-Embedded (FFPE) specimens and determine their prognostic potential. We examined the expression of VIM/ESA/CD24 using immunofluorescence triple staining (IF) in 24 oral cancer FFPE specimens (12 with good prognosis and 12 with poor prognosis, based on pathological factors indicating likelihood of recurrence and metastasis). High-content imaging revealed the presence of VIM/ESA/CD24 co-expressing cells in the stroma adjacent to the tumour body. There was an enrichment of these hybrid stem cells in poor prognosis tumours compared to good prognosis tumours. We also used a machine learning approach to explore whether VIM/ESA/CD24 levels could predict prognosis. Training an artificial neural network using grey level intensities corresponding to VIM/ESA/CD24 levels, it was possible to predict prognosis with 92% accuracy in an independent validation cohort. In this study of FFPE tumour specimens, we have identified a subpopulation of oral cancer stem cells that are invading into the stroma, have a hybrid EMT/MET phenotype, and are associated with an increased likelihood of tumour recurrence and metastasis.