Crowdsourcing: Spatial clustering of low-affinity binding sites amplifies in vivo transcription factor occupancy

Abstract

To predict in vivo occupancy of a transcription factor (TF), current models consider only the immediate genomic context of a putative binding site (BS) - impact of the sites spatial chromatin context is not known. Using clusters of spatially proximal enhancers, or archipelagos, and DNase footprints to quantify TF occupancy, we report for the first time an emergent group-level effect on occupancy, whereby BS within an archipelago experience greater in vivo occupancy than comparable BS outside archipelagos, i.e. BS not in spatial proximity with other homotypic BS. This occupancy boost is tissue-specific and scales robustly with the total number of BS, or enhancers, for the TF in the archipelago. Interestingly, enhancers within an archipelago are non-uniformly impacted by the occupancy boost; specifically, archipelago enhancers that are enriched for BS corresponding to degenerate motifs exhibit the greatest occupancy boost, as well as the highest overall accessibility, evolutionary selection, and expression at neighboring gene loci. Strikingly, archipelago-wide activity scales with expression of TFs with degenerate, but not specific, motifs. We explain these results through biophysical modelling, which suggests that spatially proximal homotypic BS facilitate TF diffusion, and induce boosts in local TF concentration and occupancy. Together, we demonstrate for the first time cooperativity among genomically distal homotypic BS that is contingent upon their spatial proximity, consistent with a TF diffusion model. Through leveraging of three-dimensional chromatin structure and TF availability, weak archipelago binding sites crowdsource their occupancy as well as context specificity, with coordinated switch-like effect on overall archipelago activity.