Core and Flanking bHLH-PAS:DNA interactions mediate specificity and drive obesity
The basic-Helix-Loop-Helix Per-Arnt-Sim (PAS) homology domain (bHLH-PAS) transcription factor (TF) family comprises critical biological sensors of physiological (hypoxia, tryptophan metabolites, neuronal activity, and appetite) and environmental (diet derived metabolites and environmental pollutants) stimuli to regulate genes involved in signal adaptation and homeostasis1. bHLH TFs bind DNA as homo or heterodimers via E-box (CANNTG) response elements, however the DNA binding specificity of the PAS domain-containing bHLH subfamily remains unresolved1. We systematically analysed cognate DNA binding hierarchies of prototypical bHLH-PAS family members (ARNT, ARNT2, HIF1, HIF2, AhR, NPAS4, SIM1) and demonstrate distinct core (NNCGTG) specificities for different heterodimer classes. The results also show that bHLH-PAS TFs bind over a large footprint 12-15bp and recognise preferential DNA sequences flanking the core. For example, specificity beyond otherwise identical core binding by SIM1 and the HIFs is mediated through N-terminal HIF-DNA interactions. We also reveal an intimate relationship between DNA shape and both core and flanking TF binding allowing motif sequence flexibility and underpinning TF binding specificity. Furthermore, DNA-shape affinity relationships revealed that novel downstream PAS-A-loop DNA interactions are associated with AT-rich sequences that lead to high-affinity binding, and that loss of this function underpins a monogenic cause of human hyperphagic obesity in a recapitulated SIM1.R171H knock-in mouse model. Importantly, models of protein-DNA binding accurately predict in vivo occupancy, while response element methylation blocks DNA binding and predicts cell type specific chromatin occupancy. These data provide a definitive and accurate map of bHLH-PAS TF specificity and target selectivity through novel flanking protein-DNA interactions that are crucial for in vivo biological function.