Allosteric coupling from G protein to the agonist binding pocket in GPCRs

G protein-coupled receptors (GPCRs) remain the primary conduit by which cells detect environmental stimuli and communicate with each other1. Upon activation by extracellular agonists, these seven transmembrane domain (7TM)-containing receptors interact with heterotrimeric G proteins to regulate downstream second messenger and/or protein kinase cascades1. Crystallographic evidence from a prototypic GPCR, the β2-adrenergic receptor (β2AR), in complex with its cognate G protein, Gs, has provided a model for how agonist binding promotes conformational changes that propagate through the GPCR and into the nucleotide binding pocket of the G protein α-subunit to catalyze GDP release, the key step required for GTP binding and activation of G proteins2. The structure also offers hints on how G protein binding may, in turn, allosterically influence ligand binding. Here we provide functional evidence that G protein coupling to β2AR stabilizes a ‘closed’ receptor conformation characterized by restricted access to and egress from the hormone binding site. Surprisingly, the effects of G protein on the hormone binding site can be observed in the absence of a bound agonist, where G protein coupling driven by basal receptor activity impedes the association of agonists, partial agonists, antagonists and inverse agonists. The ability of bound ligands to dissociate from the receptor is also hindered, providing a structural explanation for the G protein-mediated enhancement of agonist affinity, which has been observed for many GPCR-G protein pairs. Our studies also suggest that in contrast to agonist binding alone, coupling of a G protein in the absence of an agonist stabilizes large structural changes in a GPCR. The effects of nucleotide-free G protein on ligand binding kinetics are shared by other members of the superfamily of GPCRs, suggesting that a common mechanism may underlie G protein-mediated enhancement of agonist affinity.

inverse agonists. The ability of bound ligands to dissociate from the receptor is also hindered, providing a structural explanation for the G protein-mediated enhancement of agonist affinity, which has been observed for many GPCR-G protein pairs. Our studies also suggest that in contrast to agonist binding alone, coupling of a G protein in the absence of an agonist stabilizes large structural changes in a GPCR. The effects of nucleotide-free G protein on ligand binding kinetics are shared by other members of the superfamily of GPCRs, suggesting that a common mechanism may underlie G protein-mediated enhancement of agonist affinity.

Keywords
GPCR; nucleotide-free; receptor conformation Sequencing of the human genome revealed the magnitude of the GPCR superfamily, identifying over 800 genes encoding GPCRs, making this class of receptors the third-largest gene family 3 . Despite the varying nature of the chemical stimuli, which range from photons to small-molecule odorants and hormones to larger peptides and proteins, the generation of G protein-mediated signals proceeds by a common mechanism. Following activation, the receptor engages a heterotrimeric G protein and catalyzes release of GDP from the G protein α-subunit (Gα). Intracellular GTP then binds the nucleotide-free G protein, allowing it to regulate downstream effectors (adenylyl cyclase, phospholipase C, ion channels, etc.) to elicit cellular responses 4 . We recently used x-ray crystallography 2 , hydrogen-deuterium exchange mass spectrometry 5 , and electron microscopy 6 to characterize an agonist-GPCR-G protein ternary complex in the absence of nucleotide. These studies revealed dramatic conformational changes in G that are stabilized by binding to agonist-activated receptor and provided insight into the mechanism by which GPCRs bind G proteins to promote nucleotide exchange. Here, we suggest an explanation for the allosteric communication which links the nucleotide binding site on the G protein to the hormone binding site on the receptor, with a focus on conformational changes in the extracellular face of the receptor that alter access to the hormone binding site.
GPCR-G protein interactions have historically been monitored using radioligand binding assays. Observations as early as the 1970s suggested that G protein coupling enhances agonist affinity for the receptor, and can be abolished by uncoupling the G protein from the receptor with guanine nucleotides 7 . These and other data formed the basis for the ternary complex model of agonist-receptor-G protein interactions 8,9 . In this paradigm, the active state of the receptor is stabilized by both the agonist and G protein, and enhancement of agonist affinity arises due to the positive cooperativity between agonist and G protein. However, using purified β 2 AR•Gs complexes, we observed peculiar binding characteristics of the antagonist [ 3 H]dihydroalprenolol ([ 3 H]DHAP) to β 2 AR (Fig. 1a). As illustrated, addition of GDP increases the observed binding of a saturating concentration of [ 3 H]DHAP, whereas removal of GDP using a nucleotide lyase, apyrase, decreases [ 3 H]DHAP binding. The apyrase-mediated decrease in [ 3 H]DHAP binding is reversed upon addition of excess GDP, suggesting that the decrease is indeed due to the formation of nucleotide-free β 2 AR•Gs complexes. Removal of GDP from the β 2 AR•Gs complex relies on the constitutive of Gs, Gsα. The nucleotide-free status of Gsα in these β 2 AR•Gs complexes was confirmed by rapid [ 35 S]GTPγS binding kinetics (Extended Data Figure 1) 10 Figure 2). GDP enhances [ 3 H]DHAP association in a concentration-dependent manner, with similar effects achieved by complete β 2 AR•Gs uncoupling with GTPγS. Although nucleotides do not significantly affect the affinity (K d ) of [ 3 H]DHAP, their modulatory capacity is gamma phosphate-dependent since GTPγS is ∼10fold more potent than GDP (Extended Data Figure 3). Thus, β 2 AR bound to nucleotide-free G protein adopts a conformation characterized by restricted access to the hormone binding site.
Crystallographic and pharmacological evidence suggests that the active conformation of β 2 AR is stabilized by nucleotide-free Gs or by a single-chain camelid antibody raised against agonist-bound β 2 AR (nanobody Nb80) (Fig. 2a) 2,11,12 . As illustrated in Fig. 2b (and Extended Data Figure 4a), Nb80 stabilizes a conformation of β 2 AR that restricts [ 3 H]DHAP association, similar to nucleotide-free Gs. Importantly, Nb80 also slows the association of full agonist, [ 3 H]formoterol (Fig. 2c), as well aspartial agonist, [ 3 H]CGP-12177 (Fig. 2d). These data suggest that in the nucleotide-free Gs-or Nb80-stabilized active state, the β 2 AR adopts a 'closed' conformation impairing access to the orthosteric binding site, regardless of the orthosteric ligand's cooperativity with G protein. These data agree with our previous observation that the inverse agonist ICI-118,551 blocks the formation of β 2 AR•Gs complexes, but is unable to disrupt pre-formed complexes 10 . Nb80 also impairs binding of inverse agonist [ 3 H]carvedilol to β 2 AR by modestly decreasing the observed association rate ( Fig. 2e) but significantly decreasing total binding, suggesting that Nb80 and [ 3 H]carvedilol do not simultaneously occupy β 2 AR.
Agonist-promoted G protein engagement and subsequent nucleotide loss would be expected to stabilize the active, closed receptor conformation, thus trapping the agonist in the orthosteric site and enhancing its observed affinity. Indeed, uncoupling G protein from receptor using the GTP analog GppNHp has been shown to accelerate agonist dissociation from β 2 AR 13 . Such agonist-G protein cooperativity is not predicted for neutral antagonists like alprenolol, which do not stimulate G protein coupling and thus should not stabilize the closed conformation. However, we have previously demonstrated that Gs can be 'forced' to form a complex with β 2 AR bound to antagonist alprenolol 10 , provided that free nucleotide is removed, indicating that antagonist-bound β 2 AR retains enough basal activity to engage Gs. Analysis of access to the hormone binding sites in inactive-and active-state β 2 AR structures provides a structural rationale for the slowing of agonist and antagonist association (Figs. 3, Extended Data Figure 6 and Supplementary Video 1). The binding of Gs or Nb80 to β 2 AR stabilizes a rearrangement of the cytoplasmic end of TM7 ( Fig. 4a & b) immediately above the ligand binding site and a change in the structure of the extracellular loop between TM4 and TM5 (ECL2). In comparison to the inactive β 2 AR, the structure of the β 2 AR-Gs or -Nb80 (or related Nb6B9 14 ) complex identifies two aromatic residues, Phe193 (5.25 or ECL2) and Tyr308 7.35 , that move approximately 2-2.5 Å closer to each other to form a lid-like structure over the orthosteric binding site. Lys305 7.32 also contributes to capping the orthosteric site by trading its salt bridge 15 with Asp192 ECL2 for an interaction with the backbone carbonyl of Phe193 ECL2 (Fig. 4c). These structural changes are stabilized in the active forms of β 2 AR bound to either the ultra-high affinity agonist BI-167107 or the smaller, low affinity agonist adrenaline 15 , and formation of this 'lid' would be expected to sterically obstruct both ligand association and dissociation.
To validate this structural model, we tested whether a residue smaller than tyrosine could modify the capacity of Nb80 to slow ligand association. Mutation of Tyr308 7.35 to alanine, previously shown to lower agonist affinity for β 2 AR 16 , significantly diminishes the capacity of Nb80 to slow the association of [ 3 H]DHAP and even the agonist [ 3 H]formoterol (Extended Data Figure 5), as suggested by recent molecular dynamics simulations 17 . Interestingly and in contrast to [ 3 H]DHAP association, preincubation with 10 μM Nb80 also enhances the extent of [ 3 H]formoterol binding in the Y308A mutant. Eliminating barriers that impairs access to the orthosteric site (eg. Y308A) allows the agonist to at least enter the receptor where it can stabilize nanobody binding. The enhancement, therefore, is a reflection of the capacity of the agonist [ 3 H]formoterol to cooperatively stabilize Nb80 binding and vice-versa and concomitantly slow the dissociation of the bound agonist (Extended Data Figure 5d). The data also suggest that while Tyr308 7.35 significantly limits access to the orthosteric site, other residues may work in concert with Tyr308 7.35 in the active β 2 AR conformation to slow agonist dissociation.
It is noteworthy that the movement of Phe193 ECL2 and Tyr308 7.35 is not fully observed in the crystal structure of β 2 AR bound to an agonist alone 18 , nor in the inactive-state structure of β 1 AR bound to the agonist isoprenaline 19 (Extended Data Figure 6, Supplementary Video 1, and Supplementary Video 2). Binding of G protein or G protein-mimetic (nanobody) is sufficient to stabilize the closed, active conformation since their effects on ligand binding kinetics (as in Figs. 1 and 2) are agonist-independent. An agonist may enhance G protein engagement but poorly stabilizes the closed, active conformation by itself. Additionally, the data presented here suggest that formation of the closed, active conformation stabilized by the nucleotide-free G protein can occur due to basal receptor activity, in keeping with predictions of more recent models of GPCR pharmacology such as the extended and cubic ternary complex models 20,21 (see supplemental materials for extended discussion). Moreover, conformational changes stabilized by the nucleotide-free G protein influence not only agonist binding, but ligand binding in general, implying that the role of nucleotides needs to be included in an updated version of ternary complex model.
The capacity of G proteins to stabilize a closed receptor conformation explains the poorly defined GTPγS-mediated increase in radiolabeled antagonist binding observed with several GPCRs, including muscarinic, α-adrenergic, adenosine, and opioid receptors [22][23][24][25] (as in Extended Data Figure 7 & Extended Data Figure 8). We analyzed the behavior of the M2 muscarinic acetylcholine receptor (M2R) and the μ-opioid receptor (MOPr) to determine whether GTPγS-mediated uncoupling relieves a G protein-stabilized closed conformation. We focused on these receptors since structural models are available for both inactive and active conformations [26][27][28][29] , and to determine whether the mechanism we propose for β 2 AR is shared among other GPCRs. The active-state structure of the M2R, in particular, revealed similar conformational changes to the β 2 AR in that a 'lid-like' structure is formed above the orthosteric site 27 (see Supplementary Video 4 & Supplementary Video 5). Although the structural changes are not identical, the effect of G proteins (or nanobodies) on the association and dissociation of ligands at the orthosteric sites is shared among β 2 AR, M2R, and MOPr (Extended Data Figure 9 and Supplementary Video 3), suggesting that the allosteric effects of G proteins on orthosteric agonists may be manifested by conceptually common mechanisms. More discussion of the details and implications are described in the supplementary materials. Additionally, many recent studies have focused on the allosteric effect of sodium ions on Class A GPCR ligand binding and signaling 30 . Outward movement of TM6 during receptor activation collapses the sodium binding pocket in many Class A GPCRs, thus it appears that loss of bound sodium is necessary for G proteins to stabilize a closed, active receptor conformation.
The formation of the closed conformation also has particular significance to the development of allosteric modulators targeting Class A GPCRs. Most allosteric modulator binding sites have focused on the extracellular vestibule located above the orthosteric binding sites. For the muscarinic M2R for example, the potent allosteric positive modulator LY2119620 utilizes residues that form the 'lid' in the active, closed conformation as described here, as the floor of the vestibule 27 . Stabilization of this closed conformation may therefore be an important aspect on the differentiation between positive allosteric modulators, which enhance agonist binding and activation, and negative allosteric modulators, which decrease agonist binding.
In summary, we provide pharmacological and biochemical evidence suggesting that the closed, active conformation of GPCRs is stabilized by the nucleotide-free G protein, allowing G proteins to influence passage of ligands to the orthosteric binding site. The dramatic effect of G proteins on either ligand association or dissociation is consistent with, and in fact validates, structural models generated from x-ray crystallography where G protein coupling on the intracellular face of the receptor allosterically influences the structure of the extracellular face. Agonist or hormone binding enhances G protein engagement, where by formation of the active receptor conformation is accompanied by nucleotide loss from the G protein. Therefore, the capacity of G proteins to enhance agonist binding affinity is structurally and energetically linked to the agonist's capacity to promote nucleotide loss from Gα.

Methods
Large-scale Purification of β 2 AR β 2 AR bearing an N-terminal FLAG tag and C-terminal 10×-His tag was expressed in Sf9 cells (Invitrogen) and purified as previously described 2 .

Expression and purification of G protein and Nanobodies
Gs and Go heterotrimer were expressed in HighFive™ (Invitrogen) insect cells using recombinant baculovirus and purified by chromatography on Ni-NTA, MonoQ, and Superdex 200 resin, as previously described 31 . Nanobodies were expressed in Escherichia coli and purified as previously described 11,14,27 .

Membrane Preparations
HEK293T cells (ATCC) were used for small-scale expression and purification of β 2 AR and mutants. Cells were grown in DMEM + 10% FBS to ∼70% confluency, then transfected with mYFP-β 2 AR (pCMV5, 6 μg DNA per 10-cm plate) using Lipofectamine 2000. Cells were harvested 40-48 hrs post-transfection in ice-cold lysis buffer buffer (50 mM HEPES, pH 8.0, 65 mM NaCl, 1 mM EDTA, 35 μg/ml phenylmethylsulfonyl fluoride, 32 μg/ml each tosyl-L-phenylalanine-chloromethylketone and tosyl-L-lysine-chloromethylketone, 3.2 μg/ml leupeptin, 3.2 μg/ml ovomucoid trypsin inhibitor). The cell suspension was sonicated using a Branson Sonifier and centrifuged for 20 minutes at 25,000g. The pellet was resuspended in wash buffer (50 mM HEPES, pH 8.0, 100 mM NaCl with protease inhibitors listed above) using a Dounce homogenizer, then centrifuged for 20 minutes at 25,000g. The pellet was resuspended and homogenized in minimal wash buffer and the volume was adjusted to reach a final protein concentration of 5 mg/ml as measured by the Bradford protein assay. Membranes were frozen by slowly pouring into liquid nitrogen and stored at -80 °C until use.

Enrichment of β 2 AR and β 2 AR-Y308A from HEK293T cells
Frozen membranes were thawed on ice and NaCl, MgCl 2 , and GTPγS were added to reach final concentrations of 300 mM, 1 mM, and 10 μM, respectively. Timolol was then added to a final concentration of 1 μM and the membranes were incubated for 10 minutes on ice. Receptors were solubilized for 1 hour at 4 °C in the presence of 1% dodecylmaltoside (DDM) and 0.1% cholesterol hemisuccinate (CHS). Following centrifugation for 30 minutes at 25,000g, the supernatant was applied to Ni-NTA agarose. The column was slowly washed with 20 column volumes of 20 mM HEPES, pH 8.0, 300 mM NaCl, 0.1% DDM, 0.01% CHS to remove bound timolol. Receptor was eluted in the same buffer plus 200 mM imidazole and concentrated using an Amicon 30 kDa-cutoff spin concentrator for addition to the rHDL reconstitution mixture.

Radioligand Association Experiments using rHDL Particles
All assays were performed in Tris-buffered saline (TBS; 25 mM Tris-HCl, pH 7.4, 136 mM NaCl, 2.7 mM KCl) with a final concentration of 0.05% w/v bovine serum albumin. Reaction components were mixed and pre-incubated at room temperature (see below) before the addition of radioligand to initiate the association time course. Aliquots were withdrawn at the indicated times and filtered over Whatman GF/B filters pre-soaked in 0.3% w/v polyethyleneimine. Filters were washed with ice-cold TBS, dried, and subjected to liquid scintillation counting on a TopCount™ NXT (Perkin-Elmer, MA). Bound ligand never exceeded 10% of the total ligand added.

Binding experiments with [ 3 H]DHAP and Gs•β 2 AR nucleotide-free complexes -For association experiments, gel-filtered samples of apyrase-treated Gs•β 2 AR-rHDL particles were incubated with 5 nM [ 3 H]DHAP to bind any receptor that was not complexed
with Gs. The experiment was started by adding varying amounts of either GDP or GTPγS. For "equilibrium" binding experiments, samples were incubated with all the indicated components at room temperature for 90 minutes before filtration. Non-specific binding was determined in the presence of 10 μM (+/-)-propranolol.  [ 3 H]Carvedilol Association to β 2 AR-Due to high amounts of non-specific [ 3 H]carvedilol (American Radiolabeled Chemicals) binding both to bovine serum albumin (BSA) and to the glass fiber filters typically used for separation, β 2 AR-rHDL was diluted into empty rHDL particles rather than into a 5× BSA solution (0.25% w/v BSA in TBS buffer) prior to addition to the assay mix. Using empty rHDL in place of BSA was critical for maintaining sample recovery from the assay plate while improving the signal-to-noise ratio of the assay. The receptor was incubated with the indicated concentrations of Nb80 for 15 minutes at room temperature, then for 30 minutes at 4°C. [ 3 H]carvedilol was added to reach 1 nM final concentration. Aliquots were withdrawn at the indicated time points and bound ligand was isolated using gel filtration on Sephadex G75 resin. Non-specific binding was determined in the presence of 10 μM (+/-)-propranolol.

Extended Data
Extended Data Figure 1
Extended Data Figure 6. The closed conformation stabilized by agonist and G protein (or mimic) Illustrated are the crystal structures of agonist-vs. inverse agonist-bound of the β 2 AR (cyan) and β 1 AR (yellow), where only β 2 AR is bound to G protein. Similarly, the μ-opioid receptor (MOPr, orange) adopts a closed conformation upon binding G protein surrogate, Nb39. As in Figure 5 (of the main text), nucleotide-free G protein-stabilized family A GPCRs experience alterations in the extracellular face of the receptor, thus affecting orthosteric binding site. In a monoamine receptor like the β 2 AR, G protein binding and GDP loss accompanies the stabilization of a closed, active conformation of the receptor, as in a). b) For family members such as MOPr or NTS-R1, where the peptide hormones/agonists are considerably larger, the influence of the G protein-mediated changes in the extracellular domain structure result in similar effects on orthosteric ligand dissociation. Rather than closing over the orthosteric site as with monoamine receptors as in a) the extracellular face may contain structures and residues that 'pinch' the larger ligands.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.   In the β 2 AR active state (cyan), the cytoplasmic end of TM6 moves away from the receptor core by ∼14 Å relative to its position in the inactive-state structure, allowing for an inward movement of TM7. b) Rotation of TM7 allows Tyr326 7.53 (of the highly conserved NPxxY motif) to fill the space vacated by the conserved aliphatic residue Ile278 6.40 . c) The rotation of TM7 repositions Tyr308 7.35 and Lys305 7.32 . This conformational change allows Lys305 7.32 to coordinate the backbone carbonyl of Phe193 ECL2 , stabilizing its movement toward Tyr308 7.35 to form a "lid" over the hormone binding site.

Figure 5. Basis for G protein-dependent high-affinity agonist binding
Agonist binding promotes the receptor-G protein interaction and GDP release from Gα. In this nucleotide-free state, the C-terminal helix of Gα remains embedded in the receptor core, stabilizing the conformational changes at both the intracellular and extracellular faces of the receptor. At the extracellular side, the orthosteric binding site closes around the bound agonist, sterically opposing agonist dissociation and thereby enhancing the observed affinity. Constitutive (basal) receptor activity may also activate the G protein, releasing GDP and thereby stabilizing the 'closed' conformation of the receptor in the absence of an agonist. See also Extended Data Figure 10.