Mapping complex multimodal phenotypes in tissue with mosaic genetic screens

The function of organs in an animal requires the coordinated activity of thousands of genes in multiple, spatially organized cell types. Understanding the basis of this emergent tissue function will require new approaches to dissect how gene activity controls diverse cellular and tissue phenotypes in vivo. Here, we develop paired imaging and sequencing methods to construct large-scale spatial and transcriptional genotype-phenotypes maps in tissue infected in a mosaic with pooled genetic perturbations. Using imaging, we identify perturbed cells in fixed tissue sections while simultaneously measuring gene expression and subcellular morphology. Using single-cell sequencing, we then measure transcriptome-wide cellular states caused by the same genetic perturbations in dissociated fixed cells. We apply this multimodal approach to study hundreds of knockouts in a genetic mosaic mouse liver. We discover novel regulators of hepatocyte zonation, characterize a liver-specific component of the unfolded protein response, and identify distinct pathways that cause convergent hepatocyte steatosis. Our approach will enable new ways of interrogating the genetic basis of complex cellular and organismal physiology and will provide crucial training data for emerging machine learning models of cellular function. Overall design: Hepatocytes were dissociated from PFA-fixed liver tissue and enriched by FACS either by forward scatter and side scatter (WT samples) or by forward scatter, side scatter, mTurquioise expression (reflecting expression of an sgRNA) and GFP expression (reflecting expression of Cas9); or, fixed K562s were sorted for sgRNA expression.