Deep learning–enhanced single-shot triorganelle STED-FLIM imaging of lipid dynamics in living cells

Lipid homeostasis is orchestrated by rapid exchange and remodeling across the endoplasmic reticulum (ER), lipid droplets (LDs), and mitochondria. However, live-cell visualization of this triorganelle network remains limited by subdiffraction structures, multiplexed labeling burden, and the ambiguity of intensity-only readouts. Here, we introduce a single-shot stimulated emission depletion-fluorescence lifetime imaging (STED-FLIM) workflow that combines Nile Red analogs with deep learning-based demultiplexing to generate compartment-resolved maps of lipid-organelle organization and dynamics. By combining the STED-resolved nanoscale ultrastructure with lifetime-encoded microenvironmental contrast, our approach separates ER, LDs, and mitochondria from a single acquisition and enables automated tricompartment quantification using a lightweight VGG16-UNet segmentation model. This platform captures coordinated remodeling across the ER-LD-mitochondria axis during lipid stress, including ferroptosis- and apoptosis-associated transitions, while simultaneously reporting nanoscale organization and microenvironmental shifts. Together, this strategy provides a practical route to high-spatiotemporal-resolution, lifetime-encoded multiorganelle lipid imaging in living cells.