High-throughput screening of 3D cell culture models necessitates imaging workflows that can support fast acquisition and quantification of underlying structures. Imaging of dynamic processes in organoids and/or spheroids, such as mitotic divisions, has to overcome limitations, such as slow acquisition, low Z resolution, excessive light scattering and phototoxicity. In their Tools and Resources article, Roland Eils, Christian Conrad and colleagues (Eismann et al., 2020) describe a workflow that combines a light-sheet microscope (dual-view inverted selective plane illumination microscopy; diSPIM) with an image processing pipeline using convolutional neuronal networks (CNN) that is suitable for automated high-content screening of 3D cultured spheroids. To put this workflow to the test, the authors knocked down 28 mitotic genes, after which cells were suspended in Matrigel and the resulting spheroids imaged in two steps; a pre-screen to register the locations of the spheroids in the imaging plate, followed by a high-content screen of each spheroid. The dual-view of the system (acquiring images from two different angles) and the fast wide-field acquisition of SPIM allowed for high spatiotemporal screening of the spheroids. Following a subsequent imaging processing step using CNN to reliably classify the cell cycle stage of each cell, this approach enabled the detection of subtle mitotic 3D phenotypes, highlighting its promise for high-throughput screening of 3D cell culture models.
