Structure of the adenosine A_2A receptor bound to an engineered G protein

G protein-coupled receptors (GPCRs) are essential components of the signalling network throughout the body. To understand the molecular mechanism of G protein-mediated signalling, structures of receptors are necessary in inactive conformations and in the active conformation coupled to a G protein^1,2. Here we present the structure of the adenosine A_2A receptor (A_2AR) bound to an engineered G protein³, mini-G_s, to 3.4 Å resolution. Mini-G_s binds to A_2AR through an extensive interface (1048 Ų) that is similar, but not identical, to the interface between the β₂-adrenergic receptor (β₂AR) and G_s⁴. The transition of the receptor from an agonist-bound active-intermediate state^5,6 to an active G protein-bound state is characterised by a 14 Å shift of the cytoplasmic end of transmembrane helix 6 (H6) away from the receptor core, slight changes in the positions of the cytoplasmic ends of H5 and H7 and rotamer changes of the amino acid side chains Arg^3.50, Tyr^5.58 and Tyr^7.53. There are no significant differences in the extracellular half of the receptor around the ligand binding pocket. The A_2AR–mini-G_s structure highlights both the diversity and similarity in G protein coupling to GPCRs⁷ and hints at the potential complexity of the molecular basis for G protein specificity.

Structures of A_2AR bound to either inverse agonists^8–10 or agonists^5,6,11 have elucidated the molecular determinants of subtype specificity and ligand efficacy¹². However, the mechanism of activation of the receptor to allow G protein coupling and the basis of G protein selectivity is not fully understood. Structures of A_2AR in the inactive state have been determined bound either to the antagonists ZM241385^8–10, XAC⁸, caffeine⁸ or 1,2,4-triazines¹³, and all the structures are very similar. An intramembrane Na⁺ ion that can act as an allosteric antagonist was identified in the highest resolution structure (1.8 Å)¹⁴, and a homologous Na⁺ ion has been subsequently identified in other high-resolution structures of GPCRs^15–17. Four agonist-bound structures of A_2AR have also been determined after co-crystallisation with either adenosine⁵, NECA⁵, CGS21680¹¹ or UK432097⁶. All the structures are very similar and are thought to represent an active-intermediate conformation of the receptor, but not the fully active receptor that binds a G protein⁵. Observations that support this conclusion include the presence of rotamer changes of conserved amino acid residues associated with activation of other GPCRs¹⁸, and the absence of a large-scale movement of the cytoplasmic end of transmembrane helix 6 (H6) away from the receptor core¹². The G protein-coupled state of A_2AR exhibits higher binding affinity for agonists compared to the uncoupled state¹⁹, but it is unclear whether the agonist-bound structures determined so far depict the binding pocket in a high affinity or low affinity conformation. Therefore, in order to elucidate the structure of the activated state of A_2AR, we have determined its structure bound to a high-affinity agonist and an engineered G protein.

There is a single reported structure of a GPCR bound to a heterotrimeric G protein, namely G_s-bound β₂AR⁴, which showed that virtually all the atomic contacts between the receptor and G protein were formed by the Gα subunit. To facilitate the crystallisation of any GPCR-G_s complex, we developed a minimal G protein, mini-G_s, that comprised a truncated form of the GTPase domain of Gα_s and included 8 point mutations to stabilise the protein in the absence of Gβγ and in the presence of detergents³. In addition, 3 truncations removed the switch III region, 25 amino acids from the N-terminus and the α-helical domain . Mini-G_s reproduced the increase in agonist affinity that occurred upon incubation of the receptor in the presence of the heterotrimeric G protein G_s and it also showed identical sensitivity to the presence of the allosteric antagonist Na⁺ (Fig. 1 and Extended Data Fig. 1). In addition, mini-G_s readily formed a complex with A_2AR in the presence of the agonist NECA and the complex was considerably more thermostable, particularly in short chain detergents, than A_2AR with only NECA bound (Extended Data Fig. 2). This complex was crystallised in the detergent octylthioglucoside by vapour diffusion, a data set collected from two crystals, the structure determined by molecular replacement (see Online Methods) and refined to 3.4 Å (Extended Data Table 1). Of the two A_2AR–mini-G_s complexes per crystallographic asymmetric unit, the density in complex AC was better defined and was therefore used for all subsequent analyses below (see Supplementary Discussion).

The A_2AR–mini-G_s complex contained density for the agonist NECA bound to A_2AR and density for a molecule of GDP bound to mini-G_s (Extended Data Fig. 3). The presence of GDP in the mini-G_s structure is a reflection of the properties of the engineered G protein, which, after complex formation, is insensitive to GTPγS-mediated dissociation³. Mini-Gs is in a conformation virtually identical to that observed in the β₂AR–Gs structure (see below) and therefore represents an active state of the G protein, consistent with its ability to couple to A_2AR and induce high-affinity agonist binding (Fig. 1). The interface between A_2AR and mini-G_s is formed between 20 amino acid residues from the receptor and 17 residues in mini-G_s (Fig. 2, Extended Data Figs. 4-6, Supplementary Table 1), comprising a total buried surface area of 1048 Ų on the receptor. In mini-G_s, contacts are made predominantly by the α5 helix involving 14 amino acid residues that pack against residues in H3, CL2, H5, H6, H7 and H8 of A_2AR. Additional interactions include residues in S1, S3, the S2-S3 loop and α5 that form a hydrophobic pocket in which the side chain Leu110 in CL2 of A_2AR is sequestered (Extended Data Fig. 7). Amino acid residues in A_2AR and mini-G_s form complementary surfaces that pack together predominantly via van der Waals interactions (~90% of contacts) with 6 polar interactions across the interface. Helix α5 protrudes into the cleft within the cytoplasmic face of A_2AR created through the outward bending of the cytoplasmic end of H6. The apex of the α5 helix, Tyr391^H5.23 (superscript refers to the CGN system for G proteins⁷) makes extensive van der Waals interactions with Arg102^3.50 (superscript refers to the Ballesteros-Weinstein numbering system for GPCRs²⁰) that forms the whole upper surface of the cleft (Fig. 3).

Superposition of the receptors in the A_2AR–mini-G_s complex and the β₂AR–G_s complex⁴ shows that the receptors have very similar architectures (rmsd 1.7 Å over 1239 atoms). The intracellular faces of the receptors align very well, including the large outward shift of the cytoplasmic end of H6 on activation. However, mini-G_s does not superimpose exactly on the Gα subunit of the heterotrimeric G protein bound to β₂AR (Fig. 3). There is a difference in orientation of ~15°, although the difference is smaller (~10°) for the α5 helix. This is probably a consequence of the different amino acid residues in A_2AR compared to β₂AR (Fig. 3 and Extended Data Fig. 5), which results in a slightly different packing of the G proteins to the receptors, although we cannot discount the possible influence of lattice contacts. Alignment of mini-G_s with Gα_s bound to β₂AR shows that they are essentially identical (rmsd 0.92 Å over 1158 atoms), with the most significant difference being an 8° tilt between the respective α5 helices, resulting in a 3.7 Å displacement of the Cα of Tyr391 in mini-G_s away from the core of the G protein (Extended Data Fig. 8). Overall, there are 14 contacting residues in common between β₂AR and A_2AR G_s complexes, with an additional 6 contacting residues present only in A_2AR and another 10 present only in β₂AR (Supplementary Table 1). Many of the contacts between residues in the α5 helix of the G protein and the receptors are conserved, although the exact orientation and atomic contacts may differ (Fig. 3; Extended Data Fig 6, Supplementary Table 1). Similarly, there is a highly conserved interaction between a hydrophobic residue in the centre of CL2, Leu110 in A_2AR and Phe139 in β₂AR, and residues His41^S1.2, Val217^S3.1 and Asp215^s2s3.1 in Gαs (Extended Data Fig. 7). The most significant difference between the A_2AR–mini-G_s interface compared to the β₂AR–G_s interface occurs as a result of the different amino acid sequences at the H7-H8 boundary. In A_2AR, H7 terminates with Arg291^7.56 and forms the sequence R^7.56IREFR (amino acid residues in italics do not contact mini-G_s), compared to the sequence S^7.56PDFRI in the equivalent position of β₂AR, where none of the residues make contacts with Gα_s. Another region of the receptors that differs in the presence/absence of contacts to their respective G proteins is at the end of H5, due to the extension of H5 in β₂AR by an additional turn compared to A_2AR (Fig. 3; Extended Data Fig 6, Supplementary Table 1). From these examples it is clear that although the majority of amino acid residues at the interface between the receptor and G protein are identical, the specific atoms involved in the contacts differ either in terms of the amino acid side chains involved, their relative dispositions at the interface and/or the nature of the interaction.

Comparison of the active-intermediate state of UK432097-bound A_2AR⁶ with the structure of A_2AR bound to mini-G_s identified major re-arrangements in the cytoplasmic half of the receptor core to accommodate G protein binding (Fig. 4) and will be described in terms of the re-arrangements required to transition from the active-intermediate state to the activated G protein-bound conformation. Firstly, the cytoplasmic end of H6 moves away from the receptor core by 14 Å as measured between the Cα atoms of Thr224^6.26 in the two different conformations. This movement is achieved through H6 bending outwards with little discernible rotation around the helix axis. The extent of H6 movement is dictated by van der Waals interactions between Lys227^6.29, Ala231^6.33 and Leu235^6.37 in A_2AR and Leu393^H5.25 and the carboxy terminus of mini-G_s. The movement of H6 requires significant changes in the packing of the cytoplasmic end of H6 with helices H5 and H7. In particular, the side chains of highly conserved Tyr197^5.58 and Tyr288^7.53 both adopt new rotamers to fill the space previously occupied by the side chains of Leu235^6.37 (whose Cα moves by 3.7 Å) and Ile238^6.40 (Cα moves by 2.2 Å) respectively. The shift in Tyr288^7.53 allows Arg102^3.50 of the conserved DRY motif to adopt a fully extended conformation, packing against the side chain of Tyr391^H5.23 in the α5 helix of mini-G_s. It is striking that the structural change from the inactive conformation to the active-intermediate state⁵ is characterised by the concerted rotation of H5, H6 and H7, whereas the conformation change from the active-intermediate state to the active conformation upon mini-G_s binding is characterised by the bending of H6 with little further rotation. In contrast to the considerable re-arrangements of the cytoplasmic half of the receptor, there are no significant changes in the extracellular half of the receptor (Fig. 4; Extended Data Fig. 9). Thus the disposition of the ligand binding pocket described in the active-intermediate state most likely describes the high-affinity state of NECA-bound to A_2AR.

A_2AR appears to have a very different energy landscape to the β-adrenergic receptors (βARs). Both A_2AR and β₂AR exist in an ensemble of conformations whether bound to antagonists, agonists or to no ligand at all, and the presence of agonists increases the probability of formation of a fully active state^21,22. This active state is then stabilised by binding of a G protein. Structures of βARs bound to agonists are all in a conformation very similar to the inactive state^4,23–25, whereas structures of A_2AR bound to agonists are in an active-intermediate state^5,6,11 very similar to the active state. Whether there is an active-intermediate state for βARs equivalent to A_2AR is unknown, but recently, it has been proposed based on extensive EPR data that β₂AR also exists in two distinct states in the active conformation²¹. The work here shows that the active-intermediate and fully active states are distinct conformations in the intracellular half of the receptor. Given the highly conserved nature of the mechanism of GPCR activation, it is likely that the active-intermediate of A_2AR may represent a common intermediate for many Class A GPCRs, although it may exist only transiently depending on the energy landscape of the receptor.