Activity-by-Contact model of enhancer specificity from thousands of CRISPR perturbations [RNA-Seq mouse]

Mammalian genomes harbor millions of noncoding elements called enhancers that quantitatively regulate gene expression, but the mechanisms that determine which enhancers regulate which genes are not well understood. Here we describe a high-throughput approach, based on CRISPR interference, RNA FISH, and flow cytometry (CRISPRi-FlowFISH), to perturb enhancers in the genome and apply it to test >3,000 potential regulatory enhancer-gene connections across multiple genomic loci. Seemingly complex patterns of enhancer-gene connections in this dataset can largely be explained by a simple quantitative Activity-by-Contact (ABC) model, in which the effect of a candidate enhancer on a target gene is predicted by the intrinsic activity of the enhancer multiplied by the contact frequency between the enhancer and promoter. This ABC score can identify regulatory enhancer-gene connections and predict their quantitative effects from DNase-Seq, H3K27ac ChIP-seq, and Hi-C measurements. The ABC model can explain even complex regulatory phenomena including the ability of an enhancer to “skip” over one gene to regulate a more distant one. Our approach provides a foundation for dissecting mechanisms of enhancer-gene regulation and predicting enhancer-gene connections across cell types. We developed a genetic approach based on classical cis-trans tests to distinguish between (i) primary effects on the expression of nearby genes resulting from direct local functions of an enhancer and (ii) secondary effects on nearby genes resulting from indirect downstream consequences of one of the cis targets influencing another gene in the region. We generated clonal cell lines carrying heterozygous genetic modifications at enhancer loci and compared the expression of nearby genes within 1 megabase on the cis and trans alleles (i.e., on the modified and unmodified homologous chromosomes) in the same cells. Changes in neighboring gene expression that involve only the cis allele likely result from local functions, while changes that involve both the cis and trans alleles likely result as downstream effects of non-local functions. We performed genetic modifications in 129/Castaneus F1 hybrid mouse embryonic stem cells (mESCs) that contain ~1 polymorphic site every 140 basepairs (bp), enabling us to distinguish the two alleles using RNA sequencing, and we checked for consistency between knockouts on each genetic background to control for potential haplotype-specific effects. We initially characterized allele-specific gene expression using RNA sequencing combined with hybrid selection.