Conserved binding mode of human β₂ adrenergic receptor inverse agonists and antagonist revealed by X-ray crystallography

G protein-coupled receptors (GPCRs) are the largest protein family involved in signal transduction across membranes¹. The β₂ adrenergic receptor (β₂AR) is one of the best characterized members of the GPCR family, for which pharmacologically distinct high-affinity ligands have been described as (i) agonists (compounds activating signaling), (ii) antagonists (blocking agonist signaling) or (iii) inverse agonists (blocking both agonist and basal signaling). The human β₂AR structure has previously been determined in complex with two partial inverse agonists, carazolol (^Carβ₂AR-t4l)² and timolol (^Timβ₂AR-t4l)³ and turkey β₁ adrenergic receptor has been determined in complex with the antagonist cyanopindolol⁴. A number of studies have used these structures for in-silico ligand docking and discovery of new scaffolds of β₂AR ligands^5–8. Currently, a challenge for rational drug design and docking studies is to ascertain to what degree the conformation of the ligand binding site changes upon interaction with different compounds. To assess the extent of such ligand-induced conformational differences and reveal further details of ligand binding, we determined the X-ray crystal structure of β₂AR in complex with two of the most potent inverse agonists and the well known antagonist alprenolol.

Using a previously described engineered β₂AR construct³, the co-crystal structures of β₂AR-t4l in complex with 1 (^ICIβ₂AR-t4l), 2 (²β₂AR-t4l) and 3 (^Alpβ₂AR-t4l) were determined at 2.8Å, 2.8Å and 3.1Å, respectively (Figure 1; see Supporting Information for experimental details). All three structures show the same overall fold observed for the previous ^Carβ₂AR-t4l and ^Timβ₂AR-t4l structures with a rmsd of ~0.3Å (over β₂AR Cα atoms only) between all five reported β₂AR-t4l-ligand structures (Figure 2). Ligand mass spectrometry identification, receptor thermostability analysis, and the crystal structures reported here are consistent with the presence of compound 1, 2 and 3 bound in each of the β₂AR-t4l complexes. The electron density shows the compounds bound to the same orthosteric binding site as carazolol and timolol, with minor differences in side chain orientations that reflect specific ligand-receptor interactions (Figure 2).

Figure 1.

Structural comparison of the ligand binding sites in the (a) ^ICIβ₂AR-t4l, (b) ²β₂AR-t4l and (c) ^Alpβ₂AR-t4l crystal structures. The ligands 1 (ICI 118,551), 2 and 3 (alprenolol) are colored in darker shades of orange, green and blue, respectively, while residues around the binding site are colored in lighter shades and labeled. Hydrogen bonds are depicted as black dotted lines. Chemical structures of compounds are shown in boxes.

Figure 2.

Conserved overall fold of the ^ICIβ₂AR-t4l, ²β₂AR-t4l and ^Alpβ₂AR-t4l structures compared to ^Timβ₂AR-t4l and ^Carβ₂AR-t4l. (a) Superimposition of all β₂AR-t4l crystal structures determined to date (t4l omitted); ^ICIβ₂AR-t4l (yellow), ²β₂AR-t4l (green), ^Alpβ₂AR-t4l (blue), ^Timβ₂AR-t4l (magenta) and ^Carβ₂AR-t4l (grey) (b) Close view of the ligand binding site showing the conserved binding of the hydroxy-amine motif and the differences in the aromatic system moieties. Compounds are shown as sticks and surrounding residue side chains are shown as lines. Superscripts indicate the Ballesteros-Weinstein numbering convention.

The binding pocket of β₂AR can be described as a narrow cleft surrounded by mostly hydrophobic residues, with few polar residues located at the `front' (Asp113<sup>3.32</sup>, Tyr316<sup>7.43</sup> and Asn312<sup>7.39</sup>) and `back' (Ser203^5.42, S207^5.46 and Asn293^6.55) of the binding site (Figures 1 and 2). Compounds 2, 3, carazolol and timolol contain an aliphatic oxypropanolamine moiety (compound 1 has a structurally similar oxybutanolamine), referred to as the ligand tail, and chemically and structurally diverse aromatic systems defined as the ligand head groups.

The amine and hydroxyl groups in the tails of 1, 2 and 3 establish a conserved hydrogen bond network with the receptor polar triad Asp113^3.32, Tyr316^7.43 and Asn312^7.39 in the `front' of the pocket that closely resembles the ligand interactions observed in the <sup>Tim</sup>β<sub>2</sub>AR-t4l and <sup>Car</sup>β<sub>2</sub>AR-t4l structures (Figures 1 and 2). The aromatic head groups of the ligands, however, are mostly anchored between the side chains of Val114<sup>3.33</sup> and Phe290<sup>6.52</sup> in the `back' of the binding site, where each compound establishes distinct interactions with β₂AR (Figure 1).

Compared to carazolol, timolol and compound 2, the dihydro-indene head group of the inverse agonist compound 1⁹ is smaller and does not contain any polar groups that could accept or donate hydrogen bonds. The ^ICIβ₂AR-t4l structure shows the C4 carbon in the tail of compound 1 in the vicinity of Phe193^5.32, and the cyclopentene ring of the dihydro-indene in close proximity to the Phe289^6.51 and Phe193^5.32 sidechains in the `back' of the binding site (Figures 1 and 2). Furthermore, 1 has an additional methyl group on the aromatic system, and a comparison between all β₂AR-t4l-ligand structures shows that this compound requires some rearrangements in Ser203^5.42 (~1.2Å compared to other β₂AR-t4l structures) and a slight local shift ~0.4 Å in transmembrane helix 5 (TM V) (Figure 2).

The structure of ²β₂AR-t4l provides further structural insights into the binding mode of the strong inverse agonist compound 2⁶. The geometry adopted by compound 2 in the active site of the ²β₂AR-t4l structure overlaps well with that of carazolol in the ^Carβ₂AR-t4l structure, and we also observe a hydrogen bond between the side chain of Ser203^5.42 (TM V) and the benzofuran oxygen O3 of compound 2 (Figures 1 and 2). In addition, the ethyl-carboxylate moiety extends towards Asn293^6.55 and allows for an additional hydrogen bond interaction between the oxygen O2 and the amine group of Asn293^6.55 side chain in TM VI (Figure 2). A comparison between the available crystal structures of β₂AR-ligand complexes reveals that compound 2 is the only ligand that connects TM V and VI through hydrogen bond networks. Other than a few minor differences, the compound 2 pose in the ²β₂AR-t4l structure is similar to that predicted by Kolb et al.⁶ with an rmsd of ~0.9Å.

Unlike compound 1, 2, timolol and carazolol, which contain at least one cyclic system other than the aromatic ring, the allylbenzene head group of the antagonist compound 3¹⁰ is smaller and contains only a short prop-1-ene attached to the benzene group. Although the ^Alpβ₂AR-t4l structure has been determined at 3.1 Å resolution and therefore decreased confidence in the ligand placement (see Supplemental Material), there is sufficient electron density detail to orient the prop-1-ene chain of 3 in the same location as the cyclic system present on the other 4 compounds (Figure 2).

Although we observe a conserved binding mode for the β-hydroxy-amine motif on the ligand tails, a common feature among the `classical' scaffold of β₂AR ligands with inverse agonist, antagonist or full/partial agonist activities⁵, all β₂AR-t4l-ligand crystal structures show distinct interactions between the head groups of the ligands and the receptor (Figures 1 and 2). While the aromatic moieties of all compounds are anchored by strong hydrophobic interactions in the binding cleft, specific hydrogen bonds are also established by substituent moieties in compound 2, timolol and carazolol.

Recently performed large scale docking and virtual screening studies^6,11 suggest that the ^Carβ₂AR-t4l structure is highly efficient in screening for a wide range of antagonists and inverse agonists, though certain changes in the binding pocket may still be required for optimal binding of high affinity agonists. Since almost identical conformations were found for the ligand binding site in all five β₂AR-t4l structures, we set out to investigate whether a single complex structure could be suitable for docking a range of antagonists and inverse agonists^6,11. To test this hypothesis, we performed cross-docking experiments where each of the five ligands was docked into each β₂AR-t4l structure. The results (see Supporting Information) show excellent accuracy of docking pose predictions (rmsd<1Å) and high binding scores (ICM Score<−30 kJ/mol) for the docked compounds. The exception is compound 1, which cross-docks poorly into all other crystal structures, mostly because of its exocyclic methyl group, which cannot be optimally accommodated within the slightly smaller pockets of the other structures. Overall, these results support the applicability of different β₂AR-ligand structures for docking and virtual screening of antagonists and inverse agonists. Substantially better binding scores for self-docking (except for compound 2), however, suggest that additional ligand-receptor structures can further improve the performance of in-silico docking and can be particularly valuable for rational drug design at lead optimization stages.

Minimal structural differences between the three complexes reported here indicate that the ligands studied exert only minor local impact on the structure of the receptor. The most conserved region is the `front' part of the orthosteric binding pocket of the receptor, and therefore it is unlikely associated with distinct pharmacological properties of antagonists and inverse agonists. Instead, differences in specific interactions between the ligand and receptor TMs III, V and VI that take place through the aromatic ring system appear to define the pharmacologic effects. Note that agonists, characterized by a distinctly shorter “tail” and multiple polar substituents in the aromatic system are likely to introduce other changes in the β₂AR binding pocket associated with activation of the receptor, although the degree of these changes are yet to be structurally observed.

The result that β₂AR bound to pharmacologically distinct ligands (antagonists and inverse agonists) have virtually identical backbone conformations in the crystal structures suggests that the conformational changes capable of modifying signaling properties are very small, beyond the resolution of the obtained data. Alternatively, and more likely, the major effect of the antagonists/inverse agonists versus agonists in β₂AR is not on modifying a specific conformation but on changing the receptor dynamics. The answer to this intriguing problem should likely arrive from a combination of crystallography with techniques sensitive to dynamics, such as NMR¹², EPR¹³, and HDX¹⁴.