Image-based cell subpopulation identification through automated cell tracking, principal component analysis, and partitioning around medoids clustering.

In vitro cell culture model systems often employ monocultures, despite the fact that cells generally exist in a diverse, heterogeneous microenvironment in vivo. In response, heterogeneous cultures are increasingly being used to study how cell phenotypes interact. However, the ability to accurately identify and characterize distinct phenotypic subpopulations within heterogeneous systems remains a major challenge. Here, we present the use of a computational, image analysis-based approach-comprising automated contour-based cell tracking for feature identification, principal component analysis for feature reduction, and partitioning around medoids for subpopulation characterization-to non-destructively and non-invasively identify functionally distinct cell phenotypic subpopulations from live-cell microscopy image data. Using a heterogeneous model system of endothelial and smooth muscle cells, we demonstrate that this approach can be applied to both mono and co-culture nuclear morphometric and motility data to discern cell phenotypic subpopulations. Morphometric clustering identified minimal difference in mono- versus co-culture, while motility clustering revealed that a portion of endothelial cells and smooth muscle cells adopt increased motility rates in co-culture that are not observed in monoculture. We anticipate that this approach using non-destructive and non-invasive imaging can be applied broadly to heterogeneous cell culture model systems to advance understanding of how heterogeneity alters cell phenotype. This work presents a computational, image-analysis-based approach-comprising automated contour-based cell tracking for feature identification, principle component analysis for feature reduction, and partitioning around medoids for subpopulation characterization-to non-destructively and non-invasively identify functionally distinct cell phenotypic subpopulations from live-cell microscopy image data.