ISMB Commentary: Using deep learning and single cell tracking to understand competitive interactions in cell populations

Cell competition is a quality-control mechanism through which tissues eliminate unfit cells. Cell competition can result from short-range biochemical inductions or long-range mechanical cues. However, little is known about how cell-scale interactions give rise to population shifts in tissues, due to the lack of experimental and computational tools to efficiently characterize interactions at the single-cell level. Here, we address these challenges by combining long-term automated microscopy with deep-learning image analysis to decipher how single-cell behavior determines tissue makeup during competition. Using our high-throughput analysis pipeline, we show that competitive interactions between MDCK wild-type cells and cells depleted of the polarity protein scribble are governed by differential sensitivity to local density and the cell type of each cell’s neighbors. We find that local density has a dramatic effect on the rate of division and apoptosis under competitive conditions. Strikingly, our analysis reveals that proliferation of the winner cells is up-regulated in neighborhoods mostly populated by loser cells. These data suggest that tissue-scale population shifts are strongly affected by cellular-scale tissue organization. We present a quantitative mathematical model that demonstrates the effect of neighbor cell–type dependence of apoptosis and division in determining the fitness of competing cell lines.

Dr Alan Lowe

Ref: Local cellular neighbourhood controls proliferation in cell competition
Bove, A., Gradeci, D., Fujita, Y., Banerjee, S., Charras, G., Lowe, A.R.
Mol. Biol. Cell (2017) 28: 3215-3228

This video by the Alan Lowe lab demonstrates the data acquired and software developed as part of the work.