Clathrin-mediated endocytosis (CME) in HeLa cells (left panel; CME is visualized in green and DAPI staining in blue) and a model-predicted computational image of the same cells (right panel; the range from low to high predicted CME levels is indicated in a red to yellow gradient). Image courtesy of B. Snijder, ETH Zurich, Switzerland.

The underlying causes of cell-to-cell variation in cell populations are poorly understood. Using systems biology approaches, Lucas Pelkmans and colleagues demonstrate that the characteristics of a cell population partly determine the attributes of individual cells, including their susceptibility to viral infection, endocytic features and membrane lipid composition. These findings also suggest that accounting for cell population-dependent heterogeneity is crucial for data interpretation.

To study the cell-to-cell variability in cellular characteristics, Pelkmans and co-workers infected adherent cells with different viruses and measured levels of infectivity. They also measured cell surface levels of the sphingolipid GM1 and clathrin-mediated endocytosis (CME) in non-infected cells. To understand the underlying causes of the specific patterns they observed, the authors measured parameters of the population context and population-determined properties of single cells, such as population size, local cell density, position of the cell islet edge and cell size. Probability models for viral infection, CME activity and the amount of GM1 on the cell surface, which were based on these population-determined properties, provided good fits. So, the heterogeneity patterns of cellular characteristics could be predicted accurately, based solely on the quantitative assessment of the population context and the population-determined state of individual cells.

The authors established that variation due to population context is a major component of total variation observed between cells. The fact that much of the variation in cellular activities is determined by the adaptation of cells to their population context, Pelkmans and colleagues suggest, “reveals a fundamental problem in our current methods of studying differences in these activities between cell populations.” So, it is crucial to use methods that can distinguish between a change in cellular activity that is a consequence of an altered population context and a change that is due to a direct perturbation of the activity.

The authors' analysis further suggested a mechanistic basis for the population-determined heterogeneity of virus infection. For example, simian virus 40 (SV40) infection correlates with cells in sparsely populated areas that have higher amounts of GM1. In addition, focal adhesion kinase had previously been shown to be required for SV40 infection. By applying Bayesian network learning, Pelkmans and co-workers determined a network of causal interactions, some of which they subsequently validated. They also showed that synergy between GM1 levels and stronger focal adhesion kinase activation in sparse cells enhances the population-determined pattern of SV40 infection.

Together, these findings demonstrate the important contribution of population context to cellular activities and provide a mechanistic basis for the origin of population-dependent heterogeneity in adherent cells.

Author: Arianne Heinrichs