Designing Rigid Protein Fiducials to Visualize GPCR Conformational States

Abstract

G-protein coupled receptors (GPCRs) mediate precise ligand-specific signaling profiles, yet structural visualization of how ligands alter receptor conformational landscapes in the absence of signaling partners or mimetics has proven incredibly challenging. Here we show that by combining generative protein design with deep-learning based conformational ensemble prediction we can reliably design fiducial markers to facilitate cryogenic electron microscopy (cryoEM) of GPCRs at arbitrary fusion points, enabling the visualization of previously intractable states. We validate the approach with high-throughput determination of inactive state structures of four pharmaceutically relevant GPCRs, allowing for key details of receptor pharmacology to be resolved in each case. We then engineered an extracellular fiducial marker for the prototypical {beta}2-adrenergic receptor that enabled direct structural characterization of the rearrangement of key intracellular motifs in the absence of G-protein. Comparison with recent co-folding models highlights gaps in current methods for predicting ligand-induced GPCR conformational changes. These results present a generalizable framework for accessing traditionally inaccessible structural states of small, dynamic proteins.