Live imaging of SARS-CoV-2 infected airway epithelium cultures

SARS-CoV-2 infects the conducting airways, where mucociliary clearance inhibits pathogen penetrance. Mucociliary clearance is a dynamic system, and both the host and the pathogen can influence it. To better understand how SARS-CoV-2 changes MCC, we performed live imaging of infected differentiated primary human bronchial epithelium cultures over multiple days. We used a fluorescent reporter virus and fluorescent markers for tubulin and apoptotic cells to understand changes in the culture in both infected and bystander cells. In whole culture movies, we saw that SARS-CoV-2 infection foci traced the motion of mucus, suggesting that mucociliary clearance shapes the spread of the virus. We then monitored how mucociliary clearance changed during infection. We found that SARS-CoV-2 infection induced defects in ciliary motion in both infected and bystander cells, taking ~4 days for the infected cells to become numerous and old enough to impact culture-wide ciliary motion.

Methods
Live imaging of inverted ALI cultures was performed on a DeltaVision Elite microscope outfitted with a stagetop incubator. For movies of entire cultures, imaging was done in a 5x5 panel of 1024 x 1024 FOVs across 4 Z planes spanning 80 um with a 4x air lens (Olympus UPLXAPO4X), with panels acquired every 2 hours. For ciliary motion movies, 256x256 pixel FOVs were acquired with a 10x air lens at 125 frames per second. Centered on the smaller FOV of each ciliary motion movie, 1024 x 1024 reference images of all channels spanning approximately 20 microns of Z were also acquired with the same 10x lens. Channels include SPY650-Tubulin (far red), NucView 530 (orange), GFP (green), and white light. Details of each experiment are provided in the accompanying .csv.

Ciliary motion movies were denoised using cellpose 3 cytoplasm denoising model. Ciliary beat frequency was calculated using scipy.signal.periodogram on ciliary motion movies. Whole culture movies were stitched in ImageJ using the Grid/Collection stitching plugin (doi: 10.1093/bioinformatics/btp184), Z projected and concatenated across multiple days of imaging, registered using the Linear Stack Alignment with SIFT Multichannel, and cropped to remove areas outside the culture. The code used for these tasks is provided in the GitHub repository accompanying the manuscript (link in Related Works).