A Computational Pipeline for Quantifying Kinetochore Morphological Changes in Live Cells.

To segregate chromosomes kinetochores must resist yet deform under spindle forces. Measuring changes in kinetochore morphology can provide insight into kinetochore structure and function. This remains challenging in live cells because kinetochores are diffraction-limited with irregular, changing shapes. Here, we present a computational pipeline for quantifying kinetochore morphology in live cells, using mammalian cells with fluorescently tagged kinetochore proteins. First, the pipeline tracks, pairs and rotates kinetochores to align with their load-bearing axis. Second, it segments kinetochore signal from background, removing frames with overlapping neighboring kinetochore signals. Third, it provides metrics to define complex, non-Gaussian shape changes: (i) a non-parametric size metric that is more robust than the commonly used full-width-at-half-maximum (FWHM); (ii) analysis to classify common morphological patterns such as asymmetry, low intensity "tails" and multimodality; (iii) a 2D protein1-to-protein2 kinetochore vector as a reporter of structural rearrangements, if two kinetochore proteins were imaged. Finally, we validate the method using simulations, convolving ground-truth objects with the measured point spread function. Although kinetochore shape diversity makes assigning kinetochore size challenging, we show that our metrics better capture kinetochore size and shape changes than FWHM. Together, this pipeline provides a framework for analyzing complex kinetochore shape changes, with potential applications to other small and dynamic cellular structures.