Evolution of Vesicle Shape Parameters Based on A Deep-Learning Based Image Analysis Framework Over Extended Time Periods

Abstract

A precise analysis of shape instabilities in cell-sized membrane compartments requires thorough quantification of the shape dynamics of individual components over extended periods. Herein, we have used phase-contrast microscopy to image giant unilamellar membrane vesicles (GUVs), which are widely used as model systems for cellular structures. GUVs were prepared in the fluid lamellar phase of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine with and without cholesterol dissolved within a potato starch solution. The morphological changes and budding dynamics of individual GUVs with different chemical compositions were quantified using a deep-learning based image analysis framework over extended time periods, by using phase-contrast micrographs as input data. This automated framework integrates contour-based detection of the mother vesicle with precise segmentation of daughter compartments, thereby enabling explicit identification and tracking of mother and daughter vesicles at high temporal resolution. These findings further facilitate the measurement of the temporal evolution of vesicle morphometrics including area, and local membrane curvature during budding, that can be used as an input to develop computational models aimed at elucidating curvature-driven membrane remodeling in complex soft materials.