Trajectory-based analysis of fibroblast cytoskeletal and morphological features enables automated strategies for drug screening, aging determination or tumor stratification

Many biological processes display a tightly-coordinated progression from an initial to a final state. Therefore, cells involved in such processes can't be considered to exist in only two discrete phenotypes, but to be rather found within a continuum of transitional states. Here we combine fluorescence image quantification of the cytoskeleton, a high-throughput single-cell multiplex approach for cell description, with our new proposed algorithm to build trajectories, termed Phenotrax. We either construct idealized cellular trajectories using two cell populations that reflect the initial and final states of the biological phenomena studied, or quantitatively position additional cells with respect to the idealized trajectory as a means to describe them with only two simplified descriptors. In particular, we apply this strategy to two different datasets involving the activation trajectories of fibroblasts. For lung fibroblasts, we apply the approach on a screen to identify drugs that inhibit cancer-associated fibroblast activation. For breast fibroblasts, we confirm the effect of donor age on the phenotypical changes of fibroblasts. Furthermore, we show that when this dominant age-related effect is accounted for, we can also identify cancer-associated changes in fibroblasts that are linked to the tumor's hormonal profile.