Genetic screens are critical for the systematic identification of genes underlying cellular phenotypes. While pooling gene perturbations can greatly increase screening throughput, this approach was not previously compatible with the high-content imaging of complex and dynamic cellular phenotypes. We set out to develop optical pooled screening, a method to link pooled perturbations with their associated visual phenotypic outcomes in mammalian cells. In this approach, libraries of genetic perturbations are demultiplexed following image-based phenotyping using targeted in situ sequencing. By applying improved sample preparation, in situ sequencing by synthesis, and microscopy protocols, we have established this approach at a genome-wide scale and report here the results of our first genome-wide optical pooled screen.

Intracellular responses to viral infection are mediated by a key innate immune signaling pathway detecting cytosolic dsRNA that results in phosphorylation and nuclear translocation of IRF3 upon detection of cytosolic dsRNA. In a screen of 80,000 sgRNA across >10 million HeLa cells infected with Sendai virus, we identified direct regulators of IRF3 translocation which modulate responses to RNA stimuli. In this project and others entailing high-content imaging of cell lines and primary cells, we apply machine learning to identify biological phenotypes associated with genomic perturbations in large pooled experiments.

Moving forward, we continue to improve optical pooled screening throughput to ease adoption of this approach by the research community and support large screening projects in a range of biological models. We also aim to integrate additional types of phenotypic readouts to further enrich cellular characterization and provide useful flexibility in screen design.