IMAGENE: Single-cell association of live cell imaging and gene expression profiles of non-adherent cells through photoactivatable adhesives

Abstract

Live cell imaging is uniquely placed to study cell behavior as it preserves spatial context and enables non-destructive observations over time. Integrating live cell imaging and molecular phenotypes with single-cell resolution is key to uncovering the relationship between the behavioral and morphological signatures of cells, and their molecular states. Non-adherent cells - as are most immune cells - however, present unique challenges in linking live cell imaging and fixed cell assays with single-cell resolution due to the difficulty of identifying individual cells across experimental modalities. To overcome this issue, we developed IMAGENE, an experimental and computational pipeline that leverages previously reported photoactivatable biocompatible adhesive material (PA-BAM) coatings for on-the-fly cell immobilization. We demonstrate the IMAGENE experimental and computational pipeline by generating a dataset of label-free time-lapse videos of primary human naive CD8+ T cells following 24 hours of polyclonal stimulation. Individual cells, including highly motile cells, can be matched to expression profiles of genes of interest obtained through KrakenFISH, a modified version of the previously reported autoFISH setup for automated, single-molecule fluorescence in situ hybridization (smFISH) experiments that supports sample parallelization. We use this data to train explainable machine learning models that predict expression levels of individual genes, with variable performance, from hand-crafted dynamic and spatial features obtained from live cell imaging.