Fast and long-term super-resolution STED microscopy of nanostructural cellular dynamics using a neural network

Super-resolution fluorescence microscopy is revolutionizing our understanding of cellular architecture. A unique area of application is to visualize the dynamics of nano-architectures in living cells. However, high laser intensities induce fluorophore photobleaching and sample phototoxicity, limiting the number of fluorescence images obtainable from a living cell. Here, we address these challenges by using ultra-low irradiation intensities and a neural network for image restoration through denoising, enabling extensive imaging of single living cells. The endoplasmic reticulum (ER) was chosen as the target structure due to its dynamic nature over short (seconds) and long (hours) timescales. The reduced irradiation intensity combined with denoising permitted the continuous observation of ER dynamics in living cells for several hours with a temporal resolution of seconds. This allowed for quantitative analysis of ER structural features over short (seconds) and long (hours) timescales within the same cell, and enabled fast 3D live-cell STED microscopy. Overall, the combination of ultra-low irradiation with image restoration enables comprehensive analysis of organelle dynamics over extended periods in living cells.