Abstract

G protein coupled receptors (GPCRs) exhibit a spectrum of functional behaviors in response to natural and synthetic ligands. Recent crystal structures provide insights into inactive states of several GPCRs. Efforts to obtain an agonist-bound active-state GPCR structure have proven difficult due to the inherent instability of this state in the absence of a G protein. We generated a camelid antibody fragment (nanobody) to the human β₂ adrenergic receptor (β₂AR) that exhibits G protein-like behavior, and obtained an agonist-bound, active-state crystal structure of the receptor-nanobody complex. Comparison with the inactive β₂AR structure reveals subtle changes in the binding pocket; however, these small changes are associated with an 11Å outward movement of the cytoplasmic end of transmembrane segment 6, and rearrangements of transmembrane segments 5 and 7 that are remarkably similar to those observed in opsin, an active form of rhodopsin. This structure provides insights into the process of agonist binding and activation.

GPCR Active States

GPCRs activated by diffusible ligands exhibit a spectrum of functional states¹. A GPCR may activate more than one G protein isoform or a G protein independent pathway such as arrestin. In the absence of a ligand, many GPCRs exhibit some basal, agonist independent activity towards one or more of these signaling pathways. Orthosteric ligands (compounds that occupy the native hormone binding pocket) are classified according to their efficacy,i.e., the effect that they have on receptor signaling through a specific pathway. Inverse agonists inhibit basal activity while agonists maximally activate the receptor. Partial agonists induce submaximal activity, even at saturating concentrations. Neutral antagonists have no effect on basal activity, but sterically block the activity of other ligands. Moreover, the efficacy profile of ligands for a given GPCR can differ for different down-stream signaling pathways. The presence of some activity in the unliganded receptor implies low energy barriers between functional states, such that thermal fluctuations significantly sample activating conformations, and ligands with distinct efficacy profiles act by stabilizing distinct subsets of conformations.

We know little about the structural basis for the functional versatility of GPCRs. Only rhodopsin has been crystallized in different conformational states^2,3,4,5. The first structures of rhodopsin covalently bound to 11-cis-retinal represent a completely inactive state with virtually no basal activity⁵. Structures of opsin, the ligand free form of rhodopsin, obtained from crystals grown at pH 5.6 likely represent active conformations^2,3. The FTIR spectrum of opsin at acidic pH resembles that of metarhodopsin II, the light activated form of rhodopsin⁶. For rhodopsin, the light-induced transition from the inactive to the active state is very efficient. Rhodopsin is activated by photoisomerization of a covalent ligand, with efficient transfer of energy from the absorbed photon to the receptor. Crystal structures of low-pH opsin reveal that the protein conformation is the same in the presence or absence of a peptide from the alpha subunit of transducin (Gt), its cognate G protein, consistent with the notion that metarhodopsin II can adopt a fully active conformation in the absence of Gt.

The crystal structures of GPCRs activated by diffusible ligands, including the human β₂AR^7,8,9,10, the avian β₁AR¹¹, and the human adenosine A2A receptor¹², represent inactive conformations bound by inverse agonists. Unlike the activation of rhodopsin by light, agonists are much less efficient at stabilizing the active state of the β₂AR, making it difficult to capture this state in a crystal structure. Fluorescence lifetime studies show that even saturating concentrations of the full agonist isoproterenol do not stabilize a single active conformation¹³. This may be due to the relatively low affinity and rapid rates of association and dissociation for β₂AR agonists. Experiments using a β₂AR labeled with a conformationally sensitive fluorescent probe show that stabilization of the active state requires both agonist and Gs, the stimulatory G protein for adenylyl cyclase¹⁴. Efforts to obtain an agonist-GPCR-G protein complex are of great importance; however, this is a particularly difficult endeavor due to the biochemical challenges in working with both GPCRs and G proteins, and the inherent instability of the complex in detergent solutions. As an alternate approach, we developed a binding protein that preferentially binds to and stabilizes an active conformation, acting as a surrogate for Gs.

Nanobody-stabilized β₂AR active state

The active G protein coupled state of the β₂AR (and many other family A GPCRs) exhibits characteristic functional properties. Agonists promote Gs binding to the β₂AR and G protein binding to the receptor increases agonist affinity. We identified a camelid antibody fragment that exhibits G protein-like behavior towards the β₂AR. Tylopoda (camels, dromedaries and llamas) have developed a unique class of functional antibody molecules that are devoid of light chains¹⁵. A nanobody (Nb) is the recombinant minimal-sized intact antigen-binding domain of such a camelid heavy chain antibody and is approximately 25% the size of a conventional Fab fragment. To generate receptor specific nanobodies, a llama was immunized with purified agonist bound β₂AR reconstituted at high density into phospholipid vesicles. A library of single chain nanobody clones was generated and screened against agonist bound receptor. We identified seven clones that recognized agonist bound β₂AR. Of these, Nb80, was chosen because it exhibited G-protein-like properties upon binding to both wild type β₂AR and β₂AR-T4L, the β₂AR-T4 lysozyme fusion protein used to obtain the high-resolution inactive state crystal structure^7,9.

We compared the effect of Nb80 with Gs on β₂AR structure and agonist binding affinity. β₂AR was labeled at the cytoplasmic end of TM6 at C265 with monobromobimane and reconstituted into HDL particles. TM6 moves relative to TM3 and TM5 upon agonist activation (Fig. 1A), and we have previously shown that the environment around bimane covalently linked to C265 changes with both agonist binding and G protein coupling, resulting in a decrease in bimane intensity and a red shift in λ_max¹⁴. As shown inFig. 1B, the catecholamine agonist isoproterenol and Gs both stabilize an active-like conformation, but the effect of Gs is greater in the presence of isoproterenol, consistent with the cooperative interactions of agonist and Gs on β₂AR structure. Nb80 alone has an effect on bimane fluorescence and λ_max of unliganded β₂AR that is similar to that of Gs (Fig. 1C). This effect was not observed in β₂AR bound to the inverse agonist ICI-118,551. The effect of Nb80 was increased in the presence of 10 μM isoproterenol. These results show that Nb80 does not recognize the inactive conformation of the β₂AR, but binds efficiently to agonist occupied β₂AR and produces a change in bimane fluorescence that is indistinguishable from that observed in the presence of Gs and isoproterenol.

Figure 1D and E shows the effect of Gs and Nb80 on agonist affinity for β₂AR. β₂AR was reconstituted into HDL particles and agonist competition binding experiments were performed in the absence or presence of Nb80 and Gs. In the absence of either protein, isoproterenol has an inhibition constant (Ki) of 107 nM. In the presence of Gs two affinity states are observed, because not all of the β₂AR is coupled to Gs. In the Gs coupled state the affinity of isoproterenol increases by 100-fold (Ki = 1.07 nM ) (Fig 1D andTable S1). Similarly, in the presence of Nb80 the affinity of isoproterenol increases by 95-fold (Ki = 1.13 nM) (Fig. 1E andTable S1). In contrast, Nb80 had little effect on β₂AR binding to the inverse agonist ICI-118,551 (Supplementary Figure 1 andTable S1). These binding data suggest that Nb80 stabilizes a conformation in WT β₂AR that is very similar to that stabilized by Gs, such that the energetic coupling of agonist and Gs binding is faithfully mimicked by Nb80.

The high-resolution structure of the inactive state of the β₂AR was obtained with a β₂AR-T4L fusion protein. We previously showed that β₂AR-T4L has a higher affinity for isoproterenol than WT β₂AR⁷. Nevertheless, in the presence of Nb80 the affinity increased by 60-fold, resulting in an affinity (Ki = 0.56 nM) comparable to that of WT β₂AR bound to Nb80 (Fig 1F andTable S1). While we cannot study G protein coupling in β₂AR-T4L due to steric hindrance by T4L, the results show that T4L does not prevent binding of Nb80, and the nearly identical Ki values for agonist binding to wild-type β₂AR and β₂AR-T4L in the presence of Nb80 suggest that Nb80 stabilizes a similar conformation in these two proteins. The most likely explanation for the ability of Nb80 to bind to β₂AR-T4L while Gs does not is the difference in size of these two proteins. Nb80 is approximately 14 kDa while the Gs heterotrimer is approximately 90 kDa.

High affinity β₂AR agonist

To further stabilize the active state of the β₂AR, we screened over 50 commercial and proprietary β₂AR ligands. Of these, BI-167107 (Boehringer Ingelheim) had the most favorable efficacy, affinity and off-rate profile. BI-167107 is a full agonist that binds to the β₂AR with a dissociation constant K_d of 84 pM (Supplementary Figure 2A and B). As shown inSupplementary Fig 2C and D, BI-167107 induces a larger change in the fluorescence intensity and λ_max of bimane bound to C265 than does the agonist isoproterenol. Moreover, the rate of dissociation of BI-167107 was extremely slow. Displacement of BI-167107 with an excess of the neutral antagonist alprenolol required 150 hours to complete as compared with 5 sec for isoproterenol.

Crystallization of β₂AR-T4L-Nb80 complex

The β₂AR was originally crystallized bound to the inverse agonist carazolol using two different approaches. The first crystals were obtained from β₂AR bound to a Fab fragment that recognized an epitope composed of the amino and carboxyl terminal ends of the third intracellular loop connecting TMs 5 and 6⁸. In the second approach, the third intracellular loop was replaced by T4 lysozyme (β₂AR-T4L)⁷. Efforts to crystallize β₂AR-Fab complex and β₂AR-T4L bound to BI-167107 and other agonists failed to produce crystals of sufficient quality for structure determination. We therefore attempted to crystallize BI-167107 bound to β₂AR and β₂AR-T4L in complex with Nb80. While crystals of both complexes were obtained in lipid bicelles and lipidic cubic phase (LCP), high-resolution diffraction was only obtained from crystals of β₂AR-T4L-Nb80 grown in LCP. These crystals grew at pH 8.0 in 39–44% PEG400, 100 mM Tris, 4 % DMSO, and 1% 1,2,3-heptanetriol.

A merged data set at 3.5 Å was obtained from 23 crystals (Supplementary Table S2). The structure was solved by molecular replacement using the structure of the carazolol-bound β₂AR and a nanobody as search models.Supplementary Figure 3A shows the packing of the β₂AR-T4L-Nb80 complex in the crystal lattice. The receptor has interactions with lattice neighbors in several directions, and is relatively well ordered (Supplementary Figure 3A), with readily interpretable electron density for most of the polypeptide. Nb80 binds to the cytoplasmic end of the β₂AR, with the third complementarity determining region (CDR) loop projecting into the core of the receptor (Figure 2A, andSupplementary Figure 4).

Agonist-stabilized changes in the β₂AR

Figure 2B–D compares the inactive β₂AR structure (from the carazolol boundβ₂AR-T4L structure) with the agonist bound β₂AR component of the β₂AR-T4L-Nb80 complex. The largest differences are found at the cytoplasmic face of the receptor, with outward displacement of TM5 and TM6 and an inward movement of TM7 and TM3 in the β₂AR-T4L-Nb80 complex relative to the inactive structure. There are relatively small changes in the extracellular surface (Fig. 2C). The second intracellular loop (ICL2) between TM3 and TM4 adopts a two-turn alpha helix (Fig. 2D), similar to that observed in the turkey β₁AR structure¹¹. The absence of this helix in the inactive β₂AR structure may reflect crystal lattice contacts involving ICL2.

Figure 2A andSupplementary Figure 4A–C show details of interaction of Nb80 with the cytoplasmic side of the β₂AR. An eight amino acid sequence of CDR 3 penetrates into a hydrophobic pocket formed by amino acids from TM segments 3, 5, 6 and 7. A four amino acid sequence of CDR1 provides additional stabilizing interactions with cytoplasmic ends of TM segments 5 and 6. CDR3 occupies a position similar to the carboxyl terminal peptide of transducin in opsin² (Supplementary Fig 4C,D). The majority of interactions between Nb80 and the β₂AR are mediated by hydrophobic contacts.

When comparing the agonist- and inverse agonist-bound structures, the largest change is observed in TM6, with an 11.4Å movement of the helix at Glu268^6.30 (part of the ionic lock) (superscripts in this form indicate Ballesteros–Weinstein numbering for conserved GPCR residues¹⁶) (Fig. 2D). This large change is effected by a small clockwise rotation of TM6 in the turn preceding the conserved Pro288^6.50, enabled by the interrupted backbone hydrogen bonding at the proline and repacking of Phe282^6.44 (see below), which swings the helix outward.

The changes in agonist bound β₂AR-T4L-Nb80 relative to the inactive carazolol-bound β₂AR-T4L are remarkably similar to those observed between rhodopsin and opsin^2,3 (Fig. 2E). The salt bridge in the ionic lock between highly conserved Arg131^3.50 and Asp/Glu130^3.49 is broken. In opsin, Arg135^3.50 interacts with Tyr223^5.58 in TM5 and a backbone carbonyl of the transducin peptide. Arg131^3.50 of β₂AR likewise interacts with a backbone carbonyl of CDR3 of Nb80. However, Nb80 precludes an interaction between Arg131^3.50 and Tyr219^5.58, even though the tyrosine occupies a similar position in opsin and agonist bound β₂AR-T4L-Nb80. As in opsin, Tyr326^7.53 of the highly conserved NPxxY sequence moves into the space occupied by TM6 in the inactive state. In carazolol-bound β₂AR-T4L we observed a network of hydrogen bonding interactions involving highly conserved amino acids in TMs 1, 2, 6 and 7 and several water molecules⁷. While the resolution of the β₂AR-T4L-Nb80 structure is inadequate to detect waters, it is clear that the structural changes we observe would substantially alter this network.

In contrast to the relatively large changes observed in the cytoplasmic domains of β₂AR-T4L-Nb80, the changes in the agonist-binding pocket are fairly subtle.Figure 3 shows a comparison of the binding pockets of the inverse agonist and agonist bound structures. An omit map of the ligand-binding pocket is provided inSupplementary Figure 5. Many of the interactions between the agonist BI-167107 and the β₂AR are similar to those observed with the inverse agonist carazolol. The alkylamine and the β-OH of both ligands form polar interactions with Asp113^3.32 in TM3, and with Asn312^7.39 and Tyr316^7.43 in TM7. The agonist has a longer alkyl substituent on the amine, which ends with a phenyl ring that lies in a hydrophobic pocket formed by Trp109^3.28, Phe193^5.32 and Ile309^7.36.

The greatest difference between inactive and active structures in the ligand-binding site is an inward bulge of TM5 centered around Ser207^5.46, whose Cα position shifts by 2.1 Å (Fig. 4A). In addition, there are smaller inward movements of TM6 and TM7. The basal activity displayed by the β₂AR suggests that the protein structure surrounding the binding pocket is relatively dynamic in the absence of ligand, such that it samples active and inactive conformations. The presence of Pro211^5.50 in the following turn, which cannot form a hydrogen bond with the backbone at Ser207^5.46, likely lowers the barrier to the transition between the conformations observed in the presence of carazolol and BI-167107. There are extensive interactions between the carbonyl oxygen, amine and hydroxyl groups on the heterocycle of BI-167107 and Ser203^5.42 and 207^5.46 in TM5, as well as Asn293^6.55 in TM6 and Tyr308^7.35 in TM7. In contrast, there is only one polar interaction between the nitrogen in the heterocycle of carazolol and Ser203^5.42. Interactions of Ser203^5.42, Ser204^5.43 and Ser207^5.46 with catecholamine hydroxyls have been proposed based on mutagenesis studies showing that these serines are important for agonist binding and activation^17,18. While Ser204^5.43 does not interact directly with the ligand, it forms a hydrogen bond with Asn293^6.55 on TM6, which is in turn linked to Tyr308^7.35 of ECL3 (Fig. 3A). This tyrosine packs against Phe193^5.32 of ECL2, and both residues move to close off the ligand-binding site from the extracellular space.

Asn293^6.55 contributes to enantiomeric selectivity for catecholamine agonists¹⁹. The β-OH of BI-167107 does not interact with Asn293^6.55, but forms hydrogen bonds with Asp113^3.32 and Asn312^7.39, similar to what is observed for carazolol in the inactive structure. The chirality of the β-OH influences the spatial position of the aromatic ring system in β₂AR ligands, so the effect of Asn293^6.55 on β-OH enantiomeric selectivity may arise from its direct interaction with the aromatic ring system of the ligand, as well as its positioning of Ser204^5.43 and Tyr308^7.35, which also interact with this portion of the ligand. However, BI-167107 is not a catecholamine, and it is possible that the β-OH of catecholamine agonists, such as epinephrine and norepinephrine, has a direct interaction with Asn293^6.55, since mutation of Asn293^6.55 has a stronger influence on the preference for the chirality of the β-OH of catecholamine agonists, compared with non-catechol agonists and antagonists¹⁹.

Trp^6.48 is highly conserved in Family A GPCRs, and it has been proposed that its rotameric state plays a role in GPCR activation (rotamer toggle switch)²⁰. We observe no change in the side chain rotamer of Trp286^6.48 in TM6 (Fig. 4A), which lies near the base of the ligand-binding pocket, although its position shifts slightly in concert with rearrangements of nearby residues Ile121^3.40 and Phe282^6.44. While there is spectroscopic evidence for changes in the environment of Trp^6.48 upon activation of rhodopsin²¹, a rotamer change is not observed in the crystal structures of rhodopsin and low-pH opsin. Moreover, recent mutagenesis experiments on the serotonin 5HT4 receptor demonstrate that Trp^6.48 is not required for activation of this receptor by serotonin²². These observations suggest that while changes in hydrophobic packing alter the conformation of the receptor in this region, changes in the Trp^6.48 rotamer do not occur as part of the activation mechanism.

It is interesting to speculate how the small changes around the agonist-binding pocket are coupled to much larger structural changes in the cytoplasmic regions of TMs 5, 6 and 7 that facilitate binding of Nb80 and Gs. A potential conformational link is shown inFigure 4. Agonist interactions with Ser 203^5.42 and 207^5.46 stabilize a receptor conformation that includes a 2.1 Å inward movement of TM5 at position 207^5.46 and 1.4 Å inward movement of the conserved Pro211^5.50 relative to the inactive, carazolol-bound structure. In the inactive state, the relative positions of TM5, TM3, TM6 and TM7 are stabilized by interactions between Pro211^5.50, Ile121^3.40, Phe282^6.44 and Asn318^7.45. The position of Pro211^5.50 observed in the agonist structure is incompatible with this network of interactions, and Ile121^3.40 and Phe282^6.44 are repositioned, with a rotation of TM6 around Phe282^6.44 leading to an outward movement of the cytoplasmic end of TM6.

Although some of the structural changes observed in the cytoplasmic ends of transmembrane domains of the β₂AR-T4L-Nb80 complex arise from specific interactions with Nb80, the fact that Nb80 and Gs induce or stabilize similar structural changes in the β₂AR, as determined by fluorescence spectroscopy and by agonist binding affinity, suggests that Nb80 and Gs recognize similar agonist stabilized conformations. The observation that the transmembrane domains of rhodopsin and the β₂AR undergo similar structural changes upon activation provides further support that the agonist-bound β₂AR-T4L-Nb80 represents an active conformation and is consistent with a conserved mechanism of G protein activation.

However, the mechanism by which agonists induce or stabilize these conformational changes likely differs for different ligands and for different GPCRs. The conformational equilibria of rhodopsin and β₂AR differ, as shown by the fact that rhodopsin appears to adopt a fully active conformation in the absence of a G protein²³ whereas β₂AR cannot¹⁴. Thus, the energetics of activation and conformational sampling can differ among different GPCRs, which likely gives rise to the variety of ligand efficacies displayed by these receptors. An agonist need only disrupt one key intramolecular interaction needed to stabilize the inactive state, as constitutive receptor activity can result from single mutations of amino acids from different regions of GPCRs²⁴. Thus, disruption of these stabilizing interactions either by agonists or mutations lowers the energy barrier separating inactive and active states and increases the probability that a receptor can interact with a G protein.

METHODS SUMMARY

METHODS (insupplemental information)

Supplementary Material

Acknowledgments

We acknowledge support from National Institutes of Health Grants NS028471 and GM083118, and GM56169 (W.I.W.), a grant from Boerhinger Ingelheim (B.K.K.), the Mathers Foundation (B.K.K and W.I.W), the Lundbeck Foundation (Junior Group Leader Fellowship, S.G.F.R.), the Fund for Scientific Research of Flanders (FWO-Vlaanderen) and the Institute for the encouragement of Scientific Research and Innovation of Brussels (ISRIB) (E.P. and J.S.).

Footnotes

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.