How cellular and organismal complexity emerges from combinatorial expression of genes is a central question in biology. High-content phenotyping approaches such as Perturb-seq (single-cell RNA-sequencing pooled CRISPR screens) present an opportunity for exploring such genetic interactions (GIs) at scale. Here, we present an analytical framework for interpreting high-dimensional landscapes of cell states (manifolds) constructed from transcriptional phenotypes. We applied this approach to Perturb-seq profiling of strong GIs mined from a growth-based, gain-of-function GI map. Exploration of this manifold enabled ordering of regulatory pathways, principled classification of GIs (e.g., identifying suppressors), and mechanistic elucidation of synergistic interactions, including an unexpected synergy between CBL and CNN1 driving erythroid differentiation. Finally, we applied recommender system machine learning to predict interactions, facilitating exploration of vastly larger GI manifolds.

Inhibitors targeting KRASG12C, a mutant form of the guanosine triphosphatase (GTPase) KRAS, are a promising new class of oncogene-specific therapeutics for the treatment of tumors driven by the mutant protein. These inhibitors react with the mutant cysteine residue by binding covalently to the switch-II pocket (S-IIP) that is present only in the inactive guanosine diphosphate (GDP)–bound form of KRASG12C, sparing the wild-type protein. We used a genome-scale CRISPR interference (CRISPRi) functional genomics platform to systematically identify genetic interactions with a KRASG12C inhibitor in cellular models of KRASG12C mutant lung and pancreatic cancer. Our data revealed genes that were selectively essential in this oncogenic driver–limited cell state, meaning that their loss enhanced cellular susceptibility to direct KRASG12C inhibition. We termed such genes “collateral dependencies” (CDs) and identified two classes of combination therapies targeting these CDs that increased KRASG12C target engagement or blocked residual survival pathways in cells and in vivo. From our findings, we propose a framework for assessing genetic dependencies induced by oncogene inhibition.

The CFTR gene lies within an invariant topologically associated domain (TAD) demarcated by CTCF and cohesin, but shows cell-type specific control mechanisms utilizing different cis-regulatory elements (CRE) within the TAD. Within the respiratory epithelium, more than one cell type expresses CFTR and the molecular mechanisms controlling its transcription are likely divergent between them. Here, we determine how two extragenic CREs that are prominent in epithelial cells in the lung, regulate expression of the gene. We showed earlier that these CREs, located at −44 and −35 kb upstream of the promoter, have strong cell-type-selective enhancer function. They are also responsive to inflammatory mediators and to oxidative stress, consistent with a key role in CF lung disease. Here, we use CRISPR/Cas9 technology to remove these CREs from the endogenous locus in human bronchial epithelial cells. Loss of either site extinguished CFTR expression and abolished long-range interactions between these sites and the gene promoter, suggesting non-redundant enhancers. The deletions also greatly reduced promoter interactions with the 5′ TAD boundary. We show substantial recruitment of RNAPII to the −35 kb element and identify CEBPβ as a key activator of airway expression of CFTR, likely through occupancy at this CRE and the gene promoter.

Genomic sequencing of thousands of tumors has revealed many genes associated with specific types of cancer. Similarly, large scale CRISPR functional genomics efforts have mapped genes required for cancer cell proliferation or survival in hundreds of cell lines. Despite this, for specific disease subtypes, such as metastatic prostate cancer, there are likely a number of undiscovered tumor specific driver genes that may represent potential drug targets. To identify such genetic dependencies, we performed genome-scale CRISPRi screens in metastatic prostate cancer models. We then created a pipeline in which we integrated pan-cancer functional genomics data with our metastatic prostate cancer functional and clinical genomics data to identify genes that can drive aggressive prostate cancer phenotypes. Our integrative analysis of these data reveals known prostate cancer specific driver genes, such as AR and HOXB13, as well as a number of top hits that are poorly characterized. In this study we highlight the strength of an integrated clinical and functional genomics pipeline and focus on two top hit genes, KIF4A and WDR62. We demonstrate that both KIF4A and WDR62 drive aggressive prostate cancer phenotypes in vitro and in vivo in multiple models, irrespective of AR-status, and are also associated with poor patient outcome.