Microtubule disassembly by caspases is an important rate-limiting step of cell extrusion

Abstract

The expulsion of dying epithelial cells requires well-orchestrated remodelling steps to maintain tissue sealing. This process, named cell extrusion, has been mostly analysed through the study of actomyosin regulation. Yet, the mechanistic relationship between caspase activation and cell extrusion is still poorly understood. Using the Drosophila pupal notum, a single layer epithelium where extrusions are caspase-dependent, we showed that the initiation of cell extrusion and apical constriction are surprisingly not associated with the modulation of actomyosin concentration and dynamics. Instead, cell apical constriction is initiated by the disassembly of a medio-apical mesh of microtubules which is driven by effector caspases. Importantly, the depletion of microtubules is sufficient to bypass the requirement of caspases for cell extrusion, while microtubule stabilisation strongly impairs cell extrusion. This study shows that microtubules disassembly by caspases is a key rate-limiting step of extrusion, and outlines a more general function of microtubules in epithelial cell shape stabilisation.

DeXtrusion: Automatic recognition of epithelial cell extrusion through machine learningin vivo

“…We first aimed at detecting cell extrusion in the Drosophila pupal notum: a single-layer epithelium on the back of the developing Drosophila pupal thorax with a high rate of cell extrusion/apoptosis that follows stereotypical patterns ( Fig.1A ) (Guirao et al, 2015; Levayer et al, 2016; Marinari et al, 2012; Valon et al, 2021; Villars et al, 2022). We used a dataset generated in the laboratory covering a large number of extrusion events (6700 in the training set and 2320 in the test set see Supplementary tables 1 and 2 ).…”

“…To filter out false positives which are usually detected on smaller areas and fewer time windows ( Fig. 2C ), we also thresholded the size of the positive volume to fit the minimal volume associated with extrusion events (which last between 20 and 30 minutes (Villars et al, 2022)). The probability threshold and volume threshold were optimised to maximise the precision and recall ( Fig.…”

“…Of note, the accuracy was much lower for the test movie of the colcemid injected pupae ( Fig. 3E-E”, deep blue, movie ID18 ), which phenotype was poorly represented in the training dataset and corresponded to an extreme condition where the mode of extrusion and the architecture of the tissue are quite different (Villars et al, 2022). After all these optimisation steps, the final model (2nets) can still detect all cellular events on a movie covering the full pupal notum (1200 by 1200 pixel^2, 200 time points) in less than 40 min on a regular PC with 1 GPU.…”

“…We further challenged DeXtrusion by testing its performance on a squamous epithelial, the larval epithelial cells of the pupal abdomen, where cells have very different shapes from the pupal notum and where extrusions occur through slightly different mechanisms (Hoshika et al, 2020; Michel and Dahmann, 2020; Teng et al, 2017; Villars et al, 2022) ( Fig. 4 A,D,D’ ).…”

“…For instance, there are more than 1000 extrusions, distributed all over the epithelium in the Drosophila pupal notum in less than 12h (Guirao et al, 2015; Valon et al, 2021). So far, quantitative analyses of cell death/extrusion were performed using laborious manual detection of these events (Moreno et al, 2019; Valon et al, 2021; Villars et al, 2022). This highly time-consuming task remains one of the main bottlenecks for comparing a high number of conditions with precise quantitative readouts.…”