Regulation of cell identity by 3D super-enhancers

Gene regulation and genome structure are intimately linked, but accurate predictions of the elements that form topological regulatory assemblies in mammalian genomes remain elusive. To address this challenge, we developed BOUQUET, a graph theory-derived algorithm that uses multiple facets of genome topology to discern and interrogate communities of cis-regulatory elements (CREs). BOUQUET’s unbiased approach reveals otherwise undetectable CRE communities, including those that cross presumed insulating boundaries and comprise different categories of CRE. Critically, some communities, which we term “3D super-enhancers (3D-SEs)”, accrue exceptional amounts of regulatory apparatus. 3D-SEs are associated with cell identity-defining genes across mammalian tissues. Microscopy analyses show co-localization and co-expression of genes from the same 3D-SE community within transcriptional co-factor puncta, resembling microcompartments. Gene pairs within 3D-SE communities are co-expressed, and CREs within 3D-SEs are co-accessible. Overall, our model more accurately represents the architectural complexity of CRE interactions, the coordination between chromatin conformation and gene regulation, and the connection between super-enhancers and cell identity.