Reconstructing hidden GPCR conformations and ligand binding modes through integrative DEER modeling.

Abstract

Binding of biased ligands to a G protein-coupled receptor (GPCR) stabilizes specific receptor conformations, leading to distinct physiological responses. Recently, intra-transducer bias has emerged as an additional mechanism that fine-tunes receptor function by stabilizing specific architectures of the receptor-transducer complex. This mechanism was demonstrated for the angiotensin receptor (AT1R) and its β-arrestin-biased ligand TRV026, which selectively promotes β-arrestin-1 engagement via the receptor C terminus while preventing interaction with the receptor core. Here, we present a computational protocol integrating double electron-electron resonance (DEER) data with deep learning-based structure prediction and molecular dynamics simulations to reconstruct all-atom models of TRV026-bound AT1R. Using a new clustering method, we identify key ligand-receptor interactions underlying high-affinity binding, localized receptor flexibility, and ligand bias. Our results provide a framework for validating neural-network predictions and physics-based simulations to capture rare signaling states of membrane proteins and identify target conformations for designing more functionally selective, efficacious therapeutics.