Multiscale light-sheet organoid imaging framework

We present an imaging framework capable of turning long term light-sheet imaging of organoids into digital organoids. The framework takes advantage of deep learning techniques to faithfully segment single organoids, their lumen, cells and nuclei in 3D and over long periods of time. In parallel, large lineage trees for each organoid are predicted and corrected to iteratively improve the tracking and segmentation performances over time. To visualize the extracted information, we developed a web-based "Digital Organoid Viewer" that allows a unique understanding of the multivariate and multiscale data by linking 2D lineage trees with the corresponding 3D segmentation meshes. We also backtracked single cells of interest after fixation obtaining detailed information about their history within the entire organoid context. Furthermore, we show nuclei merging events that arise from cytokinesis failure and that these polyploid never reside in the intestinal crypt, hinting at a tissue scale control and feedback on cellular fidelity. Molecularly, these cytokinesis failures depend on a regenerative state of the organoids and are regulated by Lats1 and RXR and we propose a model of tissue integrity by multi-scale check points. This discovery sheds light on the robustness of a regenerative YAP cellular state, questioning the role of polyploidy in intestinal regeneration.