High-Resolution Imaging of Single-Cell Behaviors in 3D Bacterial Biofilms using Lattice-Light Sheet Microscopy and Deep Learning-Based Image Processing
Ji Zhang,Yibo Wang,Mingxing Zhang,A. Achimovich,Jie Wang,S. Acton,Andreas Gahlmann
DOI: https://doi.org/10.1017/S1431927621010540
IF: 4.0991
2021-07-30
Microscopy and Microanalysis
Abstract:Myxococcus xanthus is a highly social bacterium and a model system for coordinated multicellular behaviors, including swarming motility and developmental cell differentiation. Under nutrient rich conditions, M. xanthus cells migrate over surfaces as dense 2-dimensional swarms to prey on nearby bacterial colonies. Under nutrient limited conditions, M. xanthus populations activate a developmental program that induces cells to aggregate into 3-dimensional mounds. Some of these mounds develop further into fruiting bodies with heights and diameters of up to 100 micrometers. Both swarming motility and fruiting body development involve pronounced changes in cell morphology. Swarming, rod-shaped M. xanthus cells can bend along their long axis by well over 90 degrees ( Figure 1a ). During fruiting body development, rod-shaped cells either differentiate into spherical myxospores or into peripheral rods (a persister-like cell-state), or they burst by undergoing developmental cell lysis ( Figure 1b ). While these morphological transitions have been observed for isolated cells, or in larger populations at single, fixed time-points, it has not been possible to follow the migration behaviors and developmental fates of hundreds and thousands of cells in a population over time. As such, it remains unclear how different biochemical signals and/or mechanical cues help coordinate cell migration and cell differentiation in 3D space and time. Conventional fluorescence imaging modalities suffer from phototoxicity and photobleaching limitations that prevent high-resolution, single-cell tracking of motility or developmental progression over hours and days in large 3D bacterial communities, called biofilms. To address this challenge, we adapted minimally-invasive lattice light-sheet microscopy (LLSM) for 3D bacterial biofilm imaging within optically accessible flow altered images that are more amenable to conventional mathematical image processing, such as watershed segmentation. Initial results show similar, and in some cases superior, segmentation performance compared to the initial BCM3D workflow. Importantly, the new BCM3D pipelines do not require image deconvolution as a pre-processing step and are able to recognize a larger variety of bacterial cell shapes, including straight rods, spherical cells, as well as regular and irregularly bent cells.