Investigating Roles for Cellular Heterogeneity in Cancer

Cell populations, even those derived from a single clone, can exhibit a high degree of phenotypic variability. However, most biological studies take measurements as averages of entire populations without consideration for the underlying distribution of cellular phenotypes. Though there is growing evidence that variability within cellular populations has some functional consequences, the significance of cell to cell heterogeneity is still poorly understood. Here, we present an analytical platform that represents heterogeneity of cell populations as mixtures of distinct cell phenotypes, or subpopulations, based on immunofluorescent images. These "subpopulation profiles" make the heterogeneity of cell populations more tractable and comparable. We go on to demonstrate that subpopulation profiles can be predictive of clonal populations' drug responses. This separation is shown to be independent of the population's cell-cycle distribution. The subpopulation profiles are then shown to be robust population readouts and used to classify diverse cell lines. We show that, in diverse panels of cell populations, the relationship between basal state heterogeneity and drug response tends to break down. We also show, however, that the subpopulation profiles of diverse cell lines can be useful for identifying independently informative biomarkers. Taken together, these results demonstrate that a subpopulation level reduction of heterogeneity can be a useful readout of cell populations with many potential applications.