Biased signaling pathways in β₂-adrenergic receptor characterized by ¹⁹F-NMR

Abstract

Extracellular ligand binding to G protein-coupled receptors (GPCRs) modulates G-protein and β-arrestin signaling by changing the conformational states of the cytoplasmic region of the receptor. Using site-specific ¹⁹F-NMR labels in the β₂-adrenergic receptor (β₂AR) in complexes with various ligands, we observed that the cytoplasmic ends of helices VI and VII adopt two major conformational states. Changes in the NMR signals reveal that agonist binding primarily shifts the equilibrium towards the G protein specific active state of helix VI. In contrast, β-arrestin-biased ligands predominantly impact the conformational states of helix VII. The selective effects of different ligands on the conformational equilibria involving helices VI and VII provide insights into the long-range structural plasticity of β₂AR in partial and biased agonist signaling.

Signaling through G protein-coupled receptors (GPCRs) is a fundamental component of eukaryotic cellular communication. As a result, GPCRs comprise a large fraction of the “druggable proteome”, with more than 30% of available pharmaceuticals targeting GPCRs (1). In performing their functions, GPCRs such as the β₂-adrenergic receptor (β₂AR) recognize a diverse array of ligands and transmit signals through the cellular membrane to cytoplasmic G-proteins. In parallel to G-protein signaling, activated β₂AR can be phosphorylated and binds to β-arrestin, initiating desensitization, endocytosis of β₂AR and β-arrestin-dependent signaling (2). A number of β₂AR ligands, including the FDA-approved β-blocker carvedilol (3) and the agonist isoetharine (4), have been shown to impart differing degrees of signaling in G-protein and arrestin pathways, a phenomenon called “functional selectivity” or “biased signaling” (5). Understanding the mechanism of biased signaling can provide leads for designing more specific and efficient drugs.

Crystal structures have been determined for β₂AR with bound inverse agonists, antagonists, and agonists (6–10), including active-state β₂AR complexes with a G-protein mimetic nanobody (8) and with a heterotrimeric G-protein (11). The β₂AR–G-protein structure, which has also been analyzed by EM (12), reveals the location of the G-protein binding site, showing that the G-protein interacts with helices V and VI, and the intracellular loops ICL2 and ICL3, but makes no substantial contacts with helices VII and VIII (Fig. 1A,B). A comparison of inactive and active-state crystal structures of β₂AR, rhodopsin (13) and the A_2A adenosine receptor (14, 15) suggest a common pattern of structural differences between inactive and active states in GPCRs (Fig.1C), where a coupled motion of helices V and VI, and changes in helices III and VII are accompanied by distinct side chain rotamer switches on the cytoplasmic side of the protein (16, 17). Activation of β₂AR has been investigated previously via fluorescence-based biophysical experiments (18–25) and NMR (26), and conformational changes during receptor activation were observed in helix VI and the cytoplasmic surface.

Fig. 1.

Locations of ¹⁹F-NMR labels in β₂AR and activation-related changes in GPCR crystal structures. (A) Side view of β₂AR in the active G protein-bound form (PDB ID 3SN6, shown in green). The trans-membrane helices I to VII and the C-terminal helix VIII are identified. The full agonist BI-167107 in the ligand-binding site is shown as a stick diagram. Green and yellow spheres highlight the three cysteine residues used for TET labeling, i.e., Cys265^6.27 and Cys327^7.54 at the cytoplasmic ends of helices VI and VII, respectively and Cys341 at the C-terminus. The bound G-protein heterotrimer is shown as red ribbons and surfaces. (B) Cytoplasmic view of the structure in (A), with the G-protein contact sites outlined by a broken red line. (C) Plot of distance root mean square deviations (RMSD) of individual residues between crystal structures of inactive and active-states of three GPCRs. Crystal structures used (from top to bottom): β₂AR (PDB IDs 2RH1 vs. 3SN6), rhodopsin (PDB IDs 1GZM vs. 3DQB), A_2A adenosine receptor (A_2AAR; PDB IDs 3EML vs. 3QAK). The horizontal axes represent the amino acid sequences (β₂AR residues 34–341, bovine rhodopsin residues 38–320, A_2AAR residues 6–302). The vertical axis shows all-heavy-atom RMSDs per residue, while the color code defined in the upper right corner of the panel indicates corresponding C^α deviations. For each protein, selected residues are identified (see text). The locations of the helices I to VIII are indicated at the top. Periplasmic loop regions are highlighted in red, while cytoplasmic loops and helix VIII are highlighted in blue. The cytoplasmic ends of helices VI and VII, which contain Cys265^6.27 and Cys327^7.54, are “hot spots” with large conformational rearrangements between the crystal structures of inactive and active states; other transmembrane helices and intracellular helix VIII, which includes Cys341, show only small displacements.

In order to probe for subtle conformational changes involved in receptor–ligand binding (27), ¹⁹F-NMR spectroscopy was employed. This highly sensitive method was used to provide information based on the observation of line shapes and chemical shifts of strategically located ¹⁹F-labels in the cytoplasmic region of the receptor (Fig. 1A, B) (28). Similar approaches have previously been successfully applied to studies of a number of membrane proteins (29–31). Three native cysteine residues (Cys265^6.27, Cys327^7.54, Cys341) (32) of β₂AR (33) in the cytoplasmic region of the receptor are accessible for covalent labeling with 2,2,2 trifluoroethanethiol (TET) (31), and their NMR signals were unambiguously assigned through the use of site-specific mutagenesis of these residues and all other native cysteine residues that might be affected by TET (Fig. 2A) (34). To probe for structural changes, we primarily used singly-labeled receptors, which have reduced complexity of the ¹⁹F-NMR spectra when compared with wild type ^TETβ₂AR (Fig. 2A). Cys265^6.27 and Cys327^7.54 are located at the cytoplasmic ends of helices VI and VII, respectively, both of which are known to undergo large conformational changes during receptor activation (Fig. 1C). In contrast, Cys341 is located in the non-transmembrane helix VIII at the C-terminus and β₂AR(C265A, C327S, ^TETC341) was observed to be unresponsive to binding of different ligands (Fig. 2B). This C-terminal probe therefore served as an internal negative control relative to Cys265^6.27 and Cys327^7.54 in studies of the effects from the binding of different pharmacological ligands.

Fig. 2.

1D ¹⁹F-NMR spectra of different TET-labeled β₂AR constructs under variable solution conditions. (A) ¹⁹F-NMR resonance assignments for carazolol-bound ^TETβ₂AR at 298 K. Spectra of the following constructs were recorded: wt ^TETβ₂AR; β₂AR (^TETC265, C327S, C341A), β₂AR (C265A, ^TETC327, C341A) and β₂AR (C265A, C327S, ^TETC341). The vertical lines connect peaks in the ¹⁹F-NMR spectrum of ^TETβ₂AR with the corresponding peaks in the spectra of the single-residue TET-labeled mutants. At the top, the peak assignments are indicated by the one-letter amino acid code and the residue number. (B) β₂AR (C265A, C327S, ^TETC341) in complex with an inverse agonist (carazolol), a biased agonist (isoetharine), and a full agonist (isoproterenol) at 280K, 298K and 310K. (C), (D) and (E) β₂AR (^TETC265, C327S, C341A) and β₂AR (C265A, ^TETC327, C341A) free and bound to nine different ligands at 280K, 298K and 310K. In (B) to (E), the temperature and the ligands are indicated at the top and on the right, respectively. For all experiments, the following parameter settings were used to collect and process the spectra: data size 1024 complex points, acquisition time 51 ms, 24576 scans per increment. The data were multiplied with an exponential function with a line-broadening factor of 30 Hz, and zero-filled to 2048 points prior to Fourier transformation.

The ¹⁹F-NMR spectra ofβ₂AR(^TETC265, C327S, C341A) and β₂AR(C265A, ^TETC327, C341A) (Fig. 2C–E) show prominent changes upon addition of saturating concentrations of a collection of ligands (Table S1) which represent a full spectrum of ligand efficacy (35). In general, the signals at 280K of both β₂AR(^TETC265, C327S, C341A) and β₂AR(C265A, ^TETC341A, C327) consist of two components, suggesting that the β₂AR structure contains two independent equilibria between locally different conformations manifested at these two amino acid positions. While ligand-free and inverse agonist-bound ^TETβ₂ARs have almost identical spectra, in which the populations of C327^I and C265^I are much more prominent than those of C327^A and C265^A, spectra of ^TETβ₂AR bound to the full agonist isoproterenol exhibit approximately equal populations at both residues (Fig. 3A). Shifts of this equilibrium are observed with different agonists bound to ^TETβ₂AR, suggesting that the peaks A and I are likely to represent the active and inactive state of both helices VI and VII of β₂AR, respectively. The temperature dependence of the ¹⁹F-signals with regard to line width and peak intensity (Figs. 2C–E, S4) suggests that the active state of both helices VI and VII of β₂AR exhibits a larger degree of conformational plasticity than the inactive state, i.e., the active state samples a wider range of conformers with slightly different chemical shifts, either in a static ensemble or in a manifold of conformers that also includes conformational exchange.

Fig. 3.

Relative populations of active (A, red) and inactive (I, blue) states of β₂AR derived from the 1D ¹⁹F-NMR spectra at 280K. (A) Peak volumes for the individual components in the 1D ¹⁹F-NMR signals of β₂AR (^TETC265, C327S, C341A) and β₂AR(C265A, ^TETC327, C341A) obtained by a non-linear least-squares fit to a double-Lorentzian function. The experimental data and the double-Lorentzian are indicated by thin and thick black lines, respectively. The fit used the chemical shift positions of peaks A and I indicated by the red and blue vertical lines. (B) Plot of the relative peak volumes for C265^A versus the relative peak volumes for C327^A. The relative peak volumes are the ratios of the volume of peak A and the sum of the volumes of peaks A and I. Agonists (tulobuterol, clenbuterol, norepinephrine (NE), isoproterenol, formoterol) are shown as black circles highlighted by a yellow background, biased ligands (carvedilol, isoetharine) as red triangles highlighted by a green background, a neutral antagonist (alprenolol) as a black square, an inverse agonist (carazolol) as a black diamond, and the apo-protein as an open square.

The ¹⁹F-NMR data were analyzed with a double-Lorentzian function to obtain quantitative information on the conformational equilibria (Fig. 3A). This revealed that positions and widths of the NMR peaks remain practically unchanged for all ligands studied, and confirmed that the major effects of agonist binding are population shifts from inactive state peaks (I) to those of active state (A). A plot of relative peak volumes for Cys265^A versus those for Cys327^A, which represent the populations of the active state in helix VI and helix VII, respectively, reveals varying degrees of agonist-mediated shifts towards the active state manifested by Cys265^6.27 and Cys327^7.54 (Fig. 3B). For most agonists the shifts along the two axes follow the same overall trend as their reported pharmacological efficacy on G-protein activation (36). For instance, the partial agonists tulobuterol, clenbuterol and norepinephrine have a less pronounced effect on Cys265^6.27 than the full agonist isoproterenol, while the full agonist formoterol elicits a greater shift in conformational equilibrium on Cys265^6.27. These results suggest that the degree to which agonists shift the equilibrium towards the active state of helix VI results in differing G-protein signaling capacity. The antagonist/weak partial agonist alprenolol can induce a small shift on Cys265^6.27, but has a minimal effect on Cys327^7.54, in agreement with previous findings that it is a weak partial agonist (25, 37). The inverse agonist carazolol, on the other hand, induces no apparent shifts as compared to the apo form. In contrast, the two β-arrestin biased ligands, isoetharine and carvedilol, cause large shifts of the equilibria in Cys327^7.54 towards the active state. Interestingly, carvedilol, like other antagonists and inverse agonists, has little influence on the equilibrium of Cys265^6.27, while isoetharine, like other agonists, also produces a prominent shift of Cys265^6.27 towards the active state.

The large activation changes in Cys327^7.54 caused by carvedilol and isoetharine suggest that helix VII primarily impacts β-arrestin signaling pathways. This is consistent with the result of a previous study that β₂AR-phosphorylation by GPCR kinases, which is a prerequisite for β-arrestin binding, primarily targets a region directly adjacent to helix VII on helix VIII (38). Furthermore, recent crystallographic (11) and EM studies (12) of aβ₂AR–G_αβγ heterotrimer complex show that the G_α subunit makes major contact with helices V and VI of β₂AR, resulting in a 14 Å outward movement of the cytoplasmic ends of these helices, whereas there is no contact between G_αβγ and helix VII (Fig. 1A,B), suggesting that the conformational changes in helix VII are not directly involved in G-protein binding and signaling.

To further characterize the environment of Cys327^7.54 in the active state, we evaluated the effect of the water-soluble paramagnetic relaxation agent gadopentetic dimeglumine (magnevist®) (39) on individual resonances of β₂AR(C265A, ^TETC327, C341A) bound to isoproterenol. A comparison of the line widths of peaks A and I in the 1D ¹⁹F-NMR spectra at various magnevist concentrations revealed that peak I is more susceptible to the line broadening by magnevist than peak A (Fig. S1A), indicating that Cys327^7.54 in the inactive state is more readily solvent-accessible than in the active state. This interpretation is consistent with crystal structure data, where Cys327^7.54 was found to be solvent-exposed in the carazolol-bound structure (6), but buried in the trans-membrane helix bundle of the receptor in its complexes with a nanobody (8) or the G-protein (11) (Fig. S1B).

Additional insight from analysis of the ¹⁹F-NMR results can be obtained by examining the chemical structures of the agonists and their interactions with the individual helices in the β₂AR binding pocket (Fig. 4). The majority of known high-affinity adrenergic ligands have a common structural motif consisting of an aromatic “head group” and an ethanolamine “tail group”. The head group (e.g., catechol) of agonists is directly connected to the ethanolamine tail, while antagonists and inverse agonists, both of which act as competitive inhibitors to agonists and either have no impact or a negative impact on the receptor basal activity, have an additional two-atom linker between the head and tail groups. Crystal structures of the β₂AR complexes with various ligands (6–11), as well as biochemical (40–42) and modeling studies (43, 44) have established that the aromatic head groups interact with helices V and VI, while the ethanolamine tails are anchored by ionic and polar interactions at helices III and VII (Fig. 4A and Fig. S2). For agonists, polar interactions of their head groups with helix V side chains thereby mediate an inward shift of Pro211^5.50, and subsequent conformational changes involving Ile121^3.40 and Phe282^6.44 (8) result in the activation move of the cytoplasmic tip of helix VI.

Fig. 4.

Features of ligand binding in the β₂AR structure and a conceptual model of signaling pathways to G-proteins and arrestins in β₂AR activation. (A) Side view of the structure of active-state β₂AR in the complex with the agonist BI-167107 (PDB ID 3SN6), with helices V/VI and III/VII color-coded orange and blue, respectively, to indicate that they interact with the correspondingly colored fragments of the ligands in (B). (B) Chemical structures of the ligands used in the current ¹⁹F-NMR studies. Orange highlights the head groups, green the common ethanolamine moieties and blue the substituents to the amino group of the ethanolamine tail. Ligand names are shown on the right, with published pharmacological efficacy indicated in parentheses.

The changes observed by ¹⁹F-NMR (Fig. 3) provide additional insight into correlations between the chemical structure of the ligands and their interactions with β₂AR. Binding of full agonists (such as isoproterenol and formoterol) imparts a large shift in the conformational equilibrium of both helices VI (Cys265^6.27) and VII (Cys327^7.54). Binding of partial agonists, which either completely lack (as in tulobuterol) or have much weaker (as in clenbuterol) polar features on their aromatic heads, has a smaller influence on the conformational equilibrium of helix VI than full agonists, but has almost as much impact on helix VII. In addition, the β-arrestin-biased ligands carvedilol and isoetharine, which differ from their corresponding “unbiased” prototypes (inverse agonist carazolol and full agonist isoproterenol, respectively) only in the tail moieties, strongly impact the helix VII equilibrium. A key observation is that the 1D ¹⁹F-NMR spectrum of β₂AR(C265A, C327S, ^TETC341) bound to isoetharine, which is one of the most potent biased agonists, is similar to those of β₂AR(C265A, C327S, ^TETC341) bound to carazolol and isoproterenol (Fig. 2B). This shows that the C-terminus of the non-transmembrane helix VIII, where Cys341 is located, is not directly involved in the β-arrestin signaling pathway. In the absence of structural data on a complex of β₂AR with β-arrestin, it is tempting to further hypothesize from this data that the position of Cys341 on helix VIII is located outside of the interface with bound arrestin.

These observations on structure–activity relationships for β-arrestin biased β₂AR ligands, and unbiased full and partial agonists, provide insights into β₂AR signaling pathways that suggest the existence of parallel pathways for G-protein and β-arrestin signaling (Fig. S3). Although these pathways appear interconnected on several levels and may involve other activation-related rearrangements in the cytoplasmic and periplasmic loops (45), the present ¹⁹F-NMR results point to a significant level of decoupling of the two pathways. The implicated links between ligand structure and function may serve as a platform for future rational design of specific ligands.

Given the above observations on the correlation between ligand structure and receptor response, a connection can now be made between the chemical structures of pharmacological ligands and their signaling pathways acting either through G-proteins or β-arrestins. The general principle that links the chemical space of the ligands to the specific changes in conformational plasticity at the signaling cytoplasmic receptor surface may be preserved in other GPCRs. The use of ¹⁹F-NMR may thus prove to be a widely applicable tool for structure–activity relationship studies of GPCR ligands.