Activation of adenosine A2A receptor by lipids from docosahexaenoic acid revealed by NMR

NMR studies of A2AAR in lipid bilayers of nanodiscs revealed the mechanism underlying activation of A2AAR by DHA chains.


INTRODUCTION
G protein-coupled receptors (GPCRs) are one of the largest membrane protein families in eukaryotes. The GPCR family includes receptors of various neurotransmitters and hormones, and more than 30% of modern drugs target GPCRs. Drug binding to GPCRs leads to the induction or inhibition of signal transduction mediated by G proteins, -arrestins, and various other effectors. Under physiological conditions, GPCRs are embedded in lipid bilayers, and accumulating evidence has shown that the signaling activities of GPCRs are affected by the surrounding lipids.
Adenosine A 2A receptor (A 2A AR), a class A GPCR, controls inflammation, neurotransmission, blood flow, and immune responses (1). In the brain striatum, A 2A AR is involved in neurotransmissions that controls motor skills, and A 2A AR is a drug target for the reduction of dyskinesia in Parkinson's disease (1).
In cells, A 2A AR is embedded in membranes containing acyl chains derived from docosahexaenoic acid (DHA), a polyunsaturated fatty acid with 22 carbons and six double bonds, and arachidonic acid (ARA), a polyunsaturated fatty acid with 20 carbons and four double bonds. Dietary DHA intake increases and decreases the populations of the DHA and ARA chains, respectively (2) . In the lipid bilayers of the mammalian brain striatum, in which A 2A AR is highly expressed (1), DHA constitutes 12 to 14% of the phospholipid acyl chains, and the population of the DHA chains is remarkably decreased from 3 to 8% by feeding DHA-deficient diets (3,4). The effects of DHA on the activities of A 2A AR are unknown, although the signaling activities of rhodopsin and cannabinoid CB1 receptor, which share ~10% sequence identity with A 2A AR, are reportedly affected by the population of the DHA chains in the lipid bilayer (5,6). Clarification of the effects of DHA on the conformation and activity of A 2A AR would be helpful for understanding the signal transduction through A 2A AR under physiological conditions.
The three-dimensional structures of A 2A AR have been solved in various forms, including the inverse agonist bound, full agonist bound, and ternary complex composed of A 2A AR, a full agonist, and an engineered G protein, by x-ray crystallography and cryo-electron microscopy (7)(8)(9)(10). In the structures of the ternary complex, the cytoplasmic half of transmembrane 6 (TM6) was shifted outward and rotated clockwise, when viewed from the cytoplasmic surface, to form a cavity in the cytoplasmic surface, as compared to the structures in the inverse agonist-bound form. The C-terminal helix of the engineered G protein was inserted into the cavity (7,10). The cytoplasmic cavities were also found in the crystal structures of other GPCRs bound to G protein or arrestin (11). In the solution state, GPCRs exist in an equilibrium between inactive and active conformations, and the populations and exchange rates between these two conformations are related to their functions, as exemplified by nuclear magnetic resonance (NMR) and other spectroscopic studies (12). In the case of A 2A AR, the NMR signals originating from the isoleucine 1-methyl groups (13) and chemically introduced CF 3 probes (14) indicated that the TM region of A 2A AR exchanges between multiple conformations in the presence of the full agonist. In addition, the resonances from the glycine NH and tryptophan NH groups of A 2A AR were used to observe the allosteric modulation of W246 by the D52N mutation (15). However, little is known about the effects of the DHA chains on the conformations and the signaling activities of GPCRs. Therefore, the function-related dynamics of A 2A AR in lipid bilayers containing DHA chains must be clarified.
Reconstituted high-density lipoproteins (rHDLs), also known as nanodiscs (16), can accommodate membrane proteins within a 10-nm-diameter disc-shaped lipid bilayer. rHDLs reportedly provide lipid environments with more native-like properties, as compared with liposomes, in terms of the curvature and phase behavior (17). rHDLs and other disc-shaped lipid bilayers have been used in various in vitro GPCR studies (18)(19)(20)(21), and our NMR studies of GPCRs in rHDL lipid bilayers revealed that the populations of the active conformation of GPCRs in rHDLs correlated better with the signaling levels than those in detergent micelles (20). These methods for the reconstruction of GPCRs into the lipid bilayers of rHDLs allow examination of the effects of the composition of lipid bilayers on the activities and conformational equilibria of GPCRs.
Here, we observed the NMR signals of the methionine residues of A 2A AR in the lipid bilayers of rHDLs, containing the DHA and

NMR resonances from methionine residues of A 2A AR in rHDLs
To observe the NMR resonances from A 2A AR in rHDLs with a molecular mass of ~180 kDa, composed of ~35 kDa A 2A AR, ~80 kDa lipids, and ~60 kDa MSP1E3, we used 13 C selective labeling of the methionine methyl groups, which reflect the spatial arrangement of the neighboring aromatic residues (12), along with sensitivity enhancement by deuteration. The deuteration and methionine methyl-selective 13 C labeling in the yeast expression system were accomplished by using medium containing deuterium oxide and [,,-2 H, methyl-13 C] methionine. In the 1 H- 13 S2A). For the assignment of the methionine resonances, we mutated the methionine residues. For example, M211 5.72 [superscripts indicate GPCR database (GPCRdb) numbering (25)] was assigned by introducing the M211T mutation into A 2A AR. A resonance was absent in the spectrum of the M211T mutant, revealing that this resonance is from M211 5.72 (fig. S2B). The resonances from the other methionine residues were assigned in a similar manner ( fig. S2, C to G). Assignments for the methionine residues in the presence of the full agonist were accomplished in the same manner as those in the presence of the inverse agonist ( fig. S2, H S2N).
To investigate the conformation of A 2A AR at the cytoplasmic surface, we newly introduced methionine residues as structural probes to the cytoplasmic surface by the I106M, A232M, V239M, and I292M mutations (Fig. 1B), according to the comparison with other class A GPCRs: The residues corresponding to I106 3.54 , A232 6.34 , V239 6.41 , and I292 8.47 are methionine in at least one of the class A GPCRs. We confirmed that these mutants retained the ligand binding activity  S4, E to H), one resonance was additionally observed upon the methionine residue introduction, revealing that these resonances are from M232 6.34 , M239 6.41 , M292 8.47 , and M106 3.54 . The remaining methionine methyl signals did not exhibit substantial chemical shift changes in the presence of these mutations, and thus, they did not affect the global conformation of A 2A AR. The chemical shifts of the resonances from M106 3.54 , M140 4.61 , M211 5.72 , M232 6.34 , M239 6.41 , and M292 8.47 of A 2A AR in rHDL(POPC/POPG) in the presence of the inverse agonist were markedly different from those in the presence of the full agonist ( Fig. 1, C to G), suggesting that a large conformational change occurs upon A 2A AR activation.

Conformational equilibrium of A 2A AR in rHDLs bound to the full agonist and partial agonist
To examine the resonances undergoing conformational exchanges, we also recorded the spectra at lower and higher temperatures, 288 and 308 K, respectively. In the spectra observed at 288 and 308 K, the resonance from M106 3.54 remarkably shifted from that at 298 K ( fig. S5, A and B). In addition to the resonance from M106 3.54 , the resonances from M140 4.61 and M292 8.47 exhibited the temperaturedependent shift (fig. S5, C to F). In the cases of M140 4.61 and M292 8.47 , the chemical shift changes corresponding to these temperaturedependent shifts were not induced by the type of surrounding lipids (vide infra). Thus, these temperature-dependent shifts reflect other conformational equilibria, which are not regulated by the lipids. The M211 5.72 signals exhibited almost identical chemical shifts at 288, 298, and 308 K. These results suggest that A 2A AR in the presence of the full agonist assumes multiple active substates, exchanging faster than the chemical shift difference (>100 s −1 ).
In the 1 H- 13 Fig. 1H) and the full agonist (M106 A , M140 A , M211 A , and M292 A in Fig. 1H). The M106 A signal exhibited the temperaturedependent shift ( fig. S5, G to I). In addition, the intensity ratios of the active state signals (M106 A , M140 A , M211 A , and M292 A ), relative to the inactive state signals (M106 I , M140 I , M211 I , and M292 I ), were decreased at the lower temperatures ( fig. S5J). These signal intensity changes of A 2A AR in the presence of the partial agonist suggest that there is an exchange between the inactive and active states that occurred at a slower rate (<100 s −1 ) than the chemical shift differences, and the temperature-dependent shift of the M106 A signal suggests that the active state exists in equilibria among multiple substates, with a faster exchange rate (>100 s −1 ) than the chemical shift differences.

DHA chains enhanced G protein activation by A 2A AR
To investigate the effects of the DHA chains on the conformation and signaling activities of A 2A AR, A 2A AR was reconstituted into rHDLs composed of lipids with the DHA or ARA chains. Here, rHDLs containing the DHA and ARA chains are referred to as rHDL(DHA) and rHDL(ARA), respectively. Lipids containing 1-stearoyl-2docosahexaenoyl-phosphocholine (SDPC), 1-stearoyl-2docosahexaenoyl-phosphoglycerol (SDPG), and 1-stearoyl-2arachidonoyl-phosphocholine (SAPC), in which one of the acyl chains is derived from DHA or ARA, were used for the preparation of A 2A AR in rHDL(DHA) and A 2A AR in rHDL(ARA) ( fig. S6A). Considering the fact that more than 40% of the phospholipid acyl chains in cellular membranes are composed of those derived from DHA or ARA (26), we prepared A 2A AR in rHDL(DHA) and A 2A AR in rHDL(ARA) using lipids with SDPC:SDPG:POPC:POPG = 9:3:3:1 and SAPC:POPG = 3:1, respectively, in which 40% of the total acyl chains are derived from DHA or ARA ( Fig. 2A and fig. S6A). The populations of the anionic lipids (SDPG, SAPG, and POPG) of A 2A AR in rHDL(DHA) and A 2A AR in rHDL(ARA) were the same as those of A 2A AR in rHDL(POPC/POPG) ( Fig. 2A and fig. S6A).
We used NMR to examine the lipid composition of A 2A AR in rHDLs. In the 1 H-1D spectrum of A 2A AR in rHDL(DHA), an ensemble of resonances from the methyl groups of the DHA chains and another ensemble of resonances from the methyl groups of the other acyl chains were observed with 4:6 signal intensity ratio, suggesting that DHA constitutes 40% of the acyl chains in A 2A AR in rHDL(DHA) ( fig. S6B). In the 1 H-1D spectrum of A 2A AR in rHDL(ARA), a resonance from the methyl groups of the polar head of SAPC, an ensemble of resonances from methylene groups connecting the C═C double bonds in the ARA chain, the methylene groups farther away from the C═C double bonds, and the methyl groups of the acyl chains were observed ( fig. S6C). The signal intensities of these resonances indicate that ARA constitutes ~35% of the fatty acids in A 2A AR in rHDL(ARA). Therefore, the populations of DHA and ARA in the acyl chains are almost identical to those of the lipids used for the preparation of the rHDLs. In the 31

Effect of lipids on the conformational equilibrium of A 2A AR in rHDLs
To investigate the effects of the lipids on the conformational equilibrium of A 2A AR in rHDLs, the 1 H-13 C HMQC spectra of A 2A AR/ I106M/I292M, labeled with [[,,,-2 H, methyl-13 C] Met, u-2 H] and embedded in rHDL(DHA) and rHDL(ARA), were observed in the presence of the partial agonist. In these spectra, M106 3.54 , M140 4.61 , M211 5.72 , and M292 8.47 exhibited two resonances, and their chemical shifts were similar to those of the inactive (M106 I , M140 I , M211 I , and M292 I ) and active (M106 A , M140 A , M211 A , and M292 A ) states in rHDL(POPC/POPG) (Fig. 3, A to C). The intensity ratios of the active state signals, relative to the inactive ones, of A 2A AR in rHDL(DHA) were less than 1.1-fold higher than those of rHDL(ARA), suggesting that the populations of the inactive and active states are not affected by the type of surrounding lipids, which cannot explain the 1.7-fold higher GDP/GTP reaction rate of A 2A AR in rHDL(DHA) than A 2A AR in rHDL(ARA) (Fig. 2B). In contrast, the M106 A signal of A 2A AR in rHDL(ARA) exhibited a remarkably different chemical shift from that of A 2A AR in rHDL(DHA) (

Conformations of A 2A AR in rHDL(DHA) and A 2A AR in rHDL(ARA) investigated by solvent PRE experiments
To examine the structural characteristics of A 2A ARs in rHDL(DHA) and rHDL(ARA), the solvent accessibilities of the methionine residues were examined by solvent paramagnetic relaxation enhancement (PRE) experiments. In these experiments, Gd-diethylenetriamine pentaacetic acid-bismethylamide (Gd-DTPA-BMA), a highly water-soluble paramagnetic probe that lacks a site for protein binding (27), was added to the sample containing A 2A AR in rHDL (Fig. 5A). The PRE arising from the magnetic dipolar interactions between the nuclear spin and the unpaired electron spin of the diffusing paramagnetic probe enhances the relaxation of the nuclear spins in a distance-dependent manner, leading to the intensity reduction of the NMR signals of the solvent-exposed residues (Fig. 5A). In the 1 H-13 C HMQC spectrum of A 2A AR in rHDL(DHA) in the presence of the full agonist and Gd-DTPA-BMA, the M232 6.34 signal intensity was remarkably lower than that in the absence of Gd-DTPA-BMA (Fig. 5B). In the case of A 2A AR in rHDL(DHA), the M232 6.34 signal intensity was more affected by Gd-DTPA-BMA than that of A 2A AR in rHDL(ARA) (Fig. 5C), suggesting that M232 6.34 is more accessible to the solvent in rHDL(DHA) than in rHDL(ARA). In contrast, the On the basis of the comparison between the A 2A AR crystal structures in the forms bound with an inverse agonist and in the ternary complex with a full agonist and a G protein mimetic, the following structural mechanism for the A 2A AR activation was proposed (fig. S8) (7-9). The full agonist directly induces the movements of H250 6.52 , T88 3.36 , W246 6.48 , S277 7.42 , and H278 7.43 , which, in turn, induce the repositioning of V186 5.47 , P189 5.50 , I92 3.40 , F242 6.44 , and N280 7.45 , along with the movements of the intracellular sides of TM3, TM5, TM6, and TM7 ( fig. S8). The movement of TM6 requires the reorganization of the side chains of Y197 5.58 and Y288 7.53 . Y288 7.53 constitutes a highly conserved NPxxY motif with N284 7.49 and P285 7.50 , and the side-chain rotamer of Y288 7.53 determines the activation state of the NPxxY motif ( fig. S8) (28). The reorganization of the Y288 7.53 side chain allows R102 3.50 to adopt a fully extended conformation, packing against the side chain of Y391 in the 5 helix of the engineered G protein ( fig. S8). Therefore, the side-chain rotamer of Y288 7  of the inverse agonist [2.23 parts per million (ppm)] exhibited a larger downfield ring current shift than that in the presence of the full agonist (1.9 ppm) ( Fig. 1G and fig. S4, C and G). In the crystal structures of A 2A AR in the inverse agonist-bound form, the ring current from Y288 7.53 should induce a downfield shift in the 1 H methyl atoms of M292 8.47 , whereas Y288 7.53 is further away from M292 8.47 in the crystal structure of the ternary complex form with a full agonist and a G protein mimetic (fig. S8). These structural characteristics are in good agreement with the abovedescribed conformations indicated by the 1 H chemical shifts, suggesting that the inactive and active states correspond well to the states containing the inactive and active NPxxY motifs: NPxxYoff and NPxxYon, respectively.
The relative intensity of the M211 A signal exhibited a strong correlation with the G protein activation rates observed for A 2A AR in the presence of the inverse agonist, partial agonist, and full agonist (Fig. 6A), suggesting that the population of the NPxxYon conformations in the equilibrium is closely related to the levels of the G protein signaling activities. Various class A GPCRs also exist in equilibria between the inactive and active states, and there are good correlations between the population of the active state and the ligand efficacies (12). We found that the DHA chains enhanced the G protein activation  by A 2A AR and that the acyl chains derived from DHA did not affect the populations of the NPxxYoff and NPxxYon states in the presence of the partial agonist (Fig. 3, A to C). In contrast, the DHA chain induced a distribution change among the multiple NPxxYon substates (Figs. 3, D and E, and 4). The chemical shifts of M106 3.54 and M232 6.34 correlated well with the G protein activation rates observed for A 2A AR in rHDL(POPC/POPG), A 2A AR in rHDL(ARA), and A 2A AR in rHDL(DHA) (Fig. 6, B and C), suggesting that the NPxxYon state includes the substates that are more preferable for the G protein activation than the other NPxxYon substates. Here, the former and latter NPxxYon substates are referred to as the NPxxYon2 and NPxxYon1 substates, respectively.
Chemical shift differences among A 2A AR-rHDL(DHA), A 2A AR-rHDL(ARA), and A 2A AR-rHDL(POPC/POPG) were observed for M106 3.54 and M232 6.34 , which exist at the cytoplasmic ends of TM3 and TM6 of A 2A AR (Figs. 3 and 4), suggesting that the structures of these regions are affected by the type of surrounding lipids. In addition, the solvent PRE experiments revealed that the DHA chains shift the equilibrium of A 2A AR toward the conformations with the higher solvent accessibility of M232 6.34 (Fig. 5). The crystal structures of A 2A AR and other GPCRs indicated that the solvent accessibilities of the GPCR residues corresponding to M232 6 S7E). The crystal structures of A 2A AR (7,9) revealed that the G protein mimetic induces large TM6 rotation by >25°. These results, along with the solvent PRE experiments, suggest that TM6 would undergo a large clockwise rotation, when viewed from the cytoplasmic surface, in the conformations in the NPxxYon2 substates ( fig. S9, E and F, and see Supplementary Text). The TM6 rotation of A 2A AR would facilitate the association between A 2A AR and the G protein, which could be the rate-limiting step in the activation of the G protein (29), because the TM6 rotation is required for the van der Waals interactions between the residues in TM6 of a GPCR and the C-terminal helix of the G protein, formed in the crystal structures of GPCRs bound to both a full agonist and a trimeric G protein or its mimetics (7,11). Therefore, the NPxxYon2 conformation, which is highly populated in A 2A AR-rHDL(DHA), would be preferable for the G protein activation.
On the basis of this structural interpretation of the resonances from M106 3.54 and M232 6.34 , we propose the following signal regulation mechanism (Fig. 6D). A 2A AR bound to the inverse agonist primarily adopts the inactive NPxxYoff conformation. In the presence of the full agonist, A 2A AR exists in equilibria among the multiple active NPxxYon conformations, including the NPxxYon1 and NPxxYon2 conformations. The NPxxYon2 conformation adopts the large clockwise TM6 rotation and is more preferable for the G protein activation, as compared to the NPxxYon1 conformation. In the presence of the partial agonist, A 2A AR exists in equilibrium between the aforementioned NPxxYoff and NPxxYon conformations. The DHA chains shift the equilibria among the NPxxYon confor-mations toward the NPxxYon2 conformation, which is more active than the NPxxYon1 conformation, leading to the enhanced G protein activation by A 2A AR.

Mechanism underlying the population shift of the equilibria of A 2A AR by the DHA chains
Accumulating evidence suggests that the conformations and activities of membrane proteins can be modulated by their interactions with lipids, locations in membrane domains, and the physical properties of the lipid bilayer (e.g., lipid order, membrane fluidity, curvature stress, and hydrophobic mismatch) (30). The location in the domains can be excluded from the mechanism underlying the effects of DHA observed in this study, because the purified A 2A AR in rHDLs is isolated from other proteins. Whereas the average area per lipid, bilayer thickness, lateral diffusion, and lateral pressure profiles of the lipid bilayer with the ARA chains are reportedly similar to those with the DHA chains (31), remarkable differences were successfully observed for the G protein activation rates of A 2A AR in rHDL(DHA) and A 2A AR in rHDL(ARA) in this study. We also confirmed that the G protein activation by A 2A AR in rHDLs was not affected by the addition of trifluoroethanol, which reportedly perturbs the lateral pressure profiles (table S1) (32). Thus, the changes in the physical properties of the bilayer by the DHA chains would have a limited effect on the function-related dynamics of A 2A AR. In the recent 15-s all-atom molecular dynamics simulation of A 2A AR, the local lipid environment around A 2A AR was significantly enriched with lipids containing unsaturated chains (33). Thus, it is tempting to speculate that the interaction between A 2A AR and the DHA chains facilitates the clockwise rotation of TM6.

Relevance of stimulation of A 2A AR by DHA
Although A 2A AR in cells is embedded in membranes containing the DHA chains, information about the effects of DHA on the signaling activity of A 2A AR was unavailable. In this study, the G protein activation rates of A 2A AR in rHDL(DHA), A 2A AR in rHDL(ARA), and A 2A AR in rHDL(POPC/POPG) demonstrated that the DHA chains enhanced the G protein activation by A 2A AR (Fig. 2B). In the lipid bilayers of the mammalian brain striatum, in which A 2A AR is highly expressed (1), DHA constitutes 12 to 14% of the phospholipid acyl chains, and the population of the DHA chains is remarkably decreased to 3 to 8% by feeding DHA-deficient diets (3,4). In our experiment, the initial [ 35 S]-GTPS binding rate of A 2A AR in rHDL(DHA), in which DHA constitutes 15% of the acyl chains, was 1.29 ± 0.03 times higher than that of A 2A AR in rHDL(POPC/POPG) (n = 6). The effect of the DHA chains is higher than the level observed for the partial agonist (~10%; Fig. 1A), suggesting that the DHA chains in cell membranes play an indispensable role in the activation of A 2A AR.

Effects of various lipids on the conformation and activities of GPCRs
Cholesterol, anionic lipids, and other lipids, as well as the DHA chains, reportedly influence the GPCR activities (21). The compositions of biological membranes are affected by various pathological conditions, including aging, metabolic disease, Alzheimer's disease, and cancer. Thus, knowledge about the mechanisms underlying the influences of lipids on the modulation of GPCR activities is helpful for understanding the GPCR functions under normal and pathological conditions. The effects of lipids on the signaling activities and the conformations of GPCRs may be different among various GPCRs and lipids, because the residues in the TM region directed toward the lipids are not conserved among class A GPCRs. For example, a previous ultraviolet-visible absorption spectroscopic study of rhodopsin indicated that the DHA chains facilitate the transition from metarhodopsin I to metarhodopsin II, which reportedly adopt conformations containing the inactive and active NPxxY motifs, respectively (34), whereas the populations of the NPxxYoff and NPxxYon conformations of A 2A AR were not affected by the DHA chains (Fig. 3, A to C). The methyl resonances of deuterated GPCRs embedded in rHDLs with various lipid compositions are widely applicable for analyses of the effects of lipids on the function-related dynamics of GPCRs. Solid-state NMR has been used for studying lipid membrane protein interactions and conformations of membrane proteins embedded in lipid bilayer environments (35). Thus, solid-state NMR, which allows flexibility with respect to the lipid preparations, NMR detection schemes, and accessible temperature range, will be valuable for studying GPCRs in lipid bilayers.
Our studies using A 2A AR in rHDL lipid bilayers revealed that the DHA chains enhanced the G protein activation by A 2A AR. To our knowledge, this is the first report demonstrating the activation of A 2A AR by DHA. In addition, our NMR studies of A 2A AR in rHDL lipid bilayers revealed that A 2A AR bound to the full agonist exists in equilibria among multiple NPxxYon conformations with faster exchange rates than the chemical shift difference, and DHA shifts the equilibria toward the conformations with a large clockwise rotation of TM6, which is preferable for the G protein activation. These findings provide insight into the physiological functions of A 2A AR.

Statistics and general methods
No statistical methods were used to predetermine the sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiment and outcome assessment.

Expression and purification of A 2A AR
The complementary DNA fragment encoding human A 2A AR (1 to 316), with a different secretion signal (-factor) and a C-terminal 10× His-tag, was amplified by overlap extension polymerase chain reaction and cloned into the pPICZA vector (Thermo Fisher Scientific) via the Eco RI-Sal I sites. Mutations were introduced with a QuikChange site-directed mutagenesis kit (Stratagene). The M1A/M4T mutation was introduced in all constructs, unless otherwise stated. The pPICZA vector containing the A 2A AR gene was linearized with Pme I and purified with a QIAquick nucleotide extraction kit (Qiagen). The linearized vector was transformed into P. pastoris X-33 cells by electroporation.
All of the following procedures were either performed on ice or at 4°C. The cells were resuspended in buffer A [50 mM tris-HCl (pH 7.4), 200 mM NaCl, 5 mM EDTA, 0.2 mM 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride, 20 M leupeptin hemisulfate (Peptide Institute Inc.), 28 M E-64, 0.6 M aprotinin, and 30 M pepstatin A (Peptide Institute Inc.)] and disrupted by agitation at 300 rpm at 4°C using 40 ml zirconia/silica beads (BioSpec Products). Glycerol was added to the cell lysate, at a final concentration of 10% (v/v). The cell lysate was centrifuged at 10,500g for 10 min, and the resulting supernatant was centrifuged at 140,000g for 60 min. The membrane pellet was suspended in 5 ml of buffer B [50 mM tris-HCl (pH 8.0), 200 mM NaCl, and 10% (v/v) glycerol], supplemented with 10 mM theophylline, and was stored at −80°C. The membranes were solubilized in 2 to 3 ml of buffer B, supplemented with 10% DDM and 10 mM theophylline, for 4 hours, and were then centrifuged at 150,000g for 30 min. The supernatant was batch incubated overnight with 1 ml of TALON metal affinity resin (Clontech). The resin was first washed with 15 ml of buffer B, supplemented with 0.05% DDM, 10 mM theophylline, and 5 mM imidazole, and subsequently washed with 15 ml of buffer B, supplemented with 0.05% DDM and 5 mM imidazole. The protein was eluted with 15 ml of buffer B, supplemented with 0.05% DDM and 250 mM imidazole. The eluate was batch incubated for 5 hours with 1 ml of XAC-agarose gel. The resin was washed with 5 ml of buffer B, supplemented with 0.05% DDM, and the protein was eluted with 16 ml of buffer B, supplemented with 0.05% DDM and 30 mM theophylline. The eluted A 2A AR was concentrated to ~1 mg/ml, using a centrifugal filter device (Amicon Ultra-15, 30 kDa molecular weight cutoff, Millipore), and was stored at −80°C.

Preparation of A 2A AR in rHDLs
MSP1E3 was expressed and purified as previously described (20), with minor modifications, as follows. The double-stranded DNA encoding the N-terminal 7× His-tag, tobacco etch virus (TEV) protease recognition site, and MSP1E3 gene (19) was transferred into the pET43a vector (Novagen), with a modification of the nucleotides around the Psh AI site from 5′-GACAAGAGTCCGGGAGC-3′ to 5′-GACAAGTGTCCGGGCTTCTCCTCAACGATATCT-GAGC-3′, using T4 DNA polymerase and T4 DNA ligase. The gene encoding the NUS tag was removed by Nde I digestion and subsequent ligation, and a gene encoding the C-terminal 1D4 tag was inserted with the QuikChange site-directed mutagenesis kit. The plasmid for MSP1 expression was constructed in a similar manner, except for the absence of the C-terminal 1D4 tag. The E. coli BL21 (DE3) cells were transformed with the plasmid and cultured in Terrific Broth medium at 37°C. Protein expression was induced by the addition of 1 mM isopropyl 1-thio--d-galactopyranoside, when the optical density at 600 nm (OD 600 ) reached 2.5. After 3 hours of induction, the cells were harvested and sonicated in buffer containing 20 mM tris-HCl (pH 8.0), 300 mM NaCl, 0.1 mM 4-(2-aminoethyl) benzenesulfonamide, and protease inhibitor cocktail (Nacalai Tesque). The cell lysate was centrifuged at 20,000g for 30 min, and the precipitate was solubilized in buffer containing 1% (v/v) Triton X-100. After centrifugation at 100,000g for 1 hour, the supernatant was batch incubated for 2 hours with 10 ml of HIS-Select resin (Sigma-Aldrich). The resin was washed with 500 ml of buffer containing 1% (v/v) Triton X-100, followed by 500 ml of buffer containing 10 mM sodium cholate and 100 ml of buffer containing 10 mM imidazole. The protein was eluted by 50 ml of buffer containing 300 mM imidazole and was dialyzed against buffer containing 50 mM NaPi (pH 8. After a 1-hour incubation at 50°C, MicroScint (PerkinElmer) was added to the filter, and the radioactivity was detected with a liquid scintillation counter.

Preparation of the G protein and its mimetic
Human G s was prepared as previously described (18), with slight modifications, as follows. Double-stranded DNA encoding the N-terminal hexahistidine tag, thrombin recognition site, and human G s gene was transferred into the pET15b vector (Novagen). The E. coli BL21 (DE3) cells were transformed with the plasmid and cultured in LB medium at 37°C. Protein expression was induced by the addition of 1 mM isopropyl 1-thio--d-galactopyranoside, when the OD 600 reached 0.6 to 0.7. The cells were cultivated at 18°C for 6 hours after induction. After centrifugation at 9000g for 10 min, the cells were resuspended in buffer B [20 mM tris-HCl (pH 8.0), 5 mM 2-mercaptoethanol, 0.1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM MgCl 2 , and 0.1 mM GDP] and stored at −80°C. The cells were lysed by sonication and centrifuged at 150,000g for 30 min. NaCl was added to the resulting supernatant at a final concentration of 100 mM, and the solution was batch incubated for 2 hours with 5 ml of cOmplete His-tag purification resin (Roche), equilibrated with buffer B. The resin was washed with 50 ml of buffer B containing 500 mM NaCl, followed by 100 ml of buffer B containing 100 mM NaCl and 20 mM imidazole. The protein was eluted by 100 ml of buffer B containing 100 mM NaCl and 150 mM imidazole. The eluate was applied to 1 ml of HiTrap Q HP (GE Healthcare), equilibrated with buffer B. The protein was eluted by a linear concentration gradient of NaCl, from 0 to 500 mM, and the eluate was concentrated with a centrifugal filter device (Amicon Ultra-15, 30 kDa molecular weight cutoff). The buffer was exchanged to 20 mM tris-HCl (pH 8.0), 2 mM MgCl 2 , 0.5 mM EDTA, 20 M GDP, 0.1 mM PMSF, and 10% (v/v) glycerol, and the sample was stored at −80°C. The heterotrimeric G protein consisting of human G s , bovine G 1 , and human G 2 was prepared, as previously described (19).
The G protein mimetic was prepared as previously, with modifications as follows. The DNA fragment encoding the engineered G s gene (mini-Gs) was cloned into the above-described plasmid for the expression of MSP1, which contains the N-terminal 7× His-tag and the TEV protease recognition site, with an In-Fusion HD cloning kit (Clontech). The E. coli BL21 (DE3) codon plus RP cells were transformed with the plasmid and cultured in LB medium at 37°C. Protein expression was induced by the addition of 50 M isopropyl 1-thio--d-galactopyranoside, when the OD 600 reached 0.6 to 0.7. The cells were cultivated at 25°C for 20 hours after induction. After centrifugation at 9000g for 15 min, the cells were resuspended in a buffer containing 40 mM Hepes (pH 7.5), 100 mM NaCl, 10% glycerol, 10 mM imidazole, 5 mM MgCl 2 , 50 M GDP, 1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 1.5 M pepstatin A, 20 M leupeptin, and 100 M dithiothreitol and stored at −80°C. The cells were lysed by sonication and centrifuged at 38,000g for 45 min. The supernatant was added to the 10 ml of HIS-Select resin, and the flow-through was collected and added to the column. The column was washed with 200 ml of buffer G [20 mM Hepes (pH 7.5), 10% glycerol, 1 mM MgCl 2 , and 50 M GDP] containing 40 mM imidazole and 500 mM NaCl, followed by 100 ml of buffer G containing 40 mM imidazole and 100 mM NaCl. The protein was eluted by 30 ml of a buffer G containing 500 mM imidazole and 100 mM NaCl. TEV protease was added to a final ratio of 1:20 w/w (TEV:mini-Gs) and incubated for 18 hours at room temperature. The sample was dialysed overnight against 2 liters of buffer G containing 100 mM NaCl and added to the 8 ml of HIS-Select resin. The flow-through was collected and passed through the column five times and was concentrated with a centrifugal filter device (Amicon Ultra-15, 10 kDa molecular weight cutoff). The protein was further purified by size exclusion chromatography on a HiLoad 26/60 Superdex prep grade column and eluted with buffer containing 10 mM Hepes (pH 7.5), 100 mM NaCl, 10% glycerol, 1 mM MgCl 2 , 1 M GDP, and 0.1 mM tris(2-carboxyethyl)phosphine. After a 30-min incubation at 25°C, the reaction mixture was applied to a NAP-5 size exclusion column. The fractions containing the G protein were mixed with Optiphase Supermix, and the radioactivity was detected with the liquid scintillation counter.

NMR experiments
The 1 H-13 C HMQC and one-dimensional 1 H spectra of A 2A AR in rHDLs and in DDM micelles were recorded with a Bruker Avance 800 spectrometer equipped with a cryogenic probe. The one-dimensional 31 P spectra of A 2A AR in rHDLs were recorded with a Bruker Avance 400 spectrometer equipped with a broadband/ 19 F observe probe. 1 H-13 C HMQC spectra were recorded for 15 to 40 M [[,,-2 H, methyl-13 C] Met, u-2 H] A 2A AR or its mutants, in rHDLs and in DDM micelles in buffer D. In the NMR experiments with A 2A AR in DDM micelles, 0.19% DDM and 0.038% cholesteryl hemisuccinate were added to the samples. In the NMR experiments with A 2A AR in the UK432097-, NECA-, regadenoson-, ZM241385-, and LUF5834bound states, UK432097, NECA, regadenoson, ZM241385, and LUF5834 were added to the samples at final concentrations of 25, 100, 100, 100, and 25 M, respectively. In the solvent PRE experiments, Gd-DTPA-BMA was added to the samples at a final concentration of 5 mM. In the NMR experiments of the ternary complex, the engineered G protein was added to the sample with 15 M A 2A AR/A232M in the NECA-bound state at a final concentration of 30 M. The spectral widths were set to 12,800 Hz and 6400 to 7400 Hz for the 1 H and 13 C dimensions, respectively, and the interscan delays were set to 1 s. In total, 512 × 128 complex points were recorded, and 256 scans/free induction decay gave rise to an acquisition time of 20 to 23 hours for each spectrum. The spectra were referenced to 3-(trimethysilyl)-1-propanesulfonic acid sodium salt, in both the 1 H and 13 C dimensions. One-dimensional 1 H NMR spectra were recorded with a repetition period of 1.0 s, with 8096 complex data points and a spectral width of 12,800 Hz. One-dimensional 31 P NMR spectra were recorded with a repetition period of 1.0 s, with 2048 complex data points and a spectral width of 3240 Hz. The assignments of the resonances from the DHA and ARA chains were derived from the previous report (38). All of the spectra were processed and analyzed by the Topspin 2.1, 3.1, or 3.5 software (Bruker).

In situ chemical shift error analysis
The calculations were performed by in-house developed programs, written in the Python 2.7 programming language, supplemented with the extension modules NumPy 1.15, SciPy 1.1, and Cython 0.29. Synthetic two-dimensional time domain data, composed of 100 synthetic signals with relaxation rates and signal-to-noise ratios similar to those of the observed resonances, were generated using in-house developed programs. The data were processed in the same manner as the observed data, and the SDs of the chemical shifts were calculated.

GPCR structure database analysis
In total, 182 coordinate files of GPCR structures were obtained from Protein Data Bank (PDB), and the structures without residues corresponding to "5.63" or "6.27" in the GPCRdb numbering scheme (25) were excluded from the analysis, because the solvent accessibilities were overestimated for these structures. For the remaining 118 structures, the lipid bilayer was added by CHARMM-GUI Membrane Builder (39).
In each structure, hypothetical atoms with a 3.5-Å radius, corresponding to Gd-DTPA-DMA, were evenly placed at grid points spaced 1 Å surrounding GPCR, and the hypothetical atoms in which at least one of the GPCR and/or lipid atoms are within the sum of the radii were removed. The distances between the closest hypothetical atom and the C atom of the residues corresponding to M106 3.54 and M232 6.34 of A 2A AR were used for the calculation of the solvent PRE, based on the Hwang-Freed model, as reported previously (27). The full crowding effect was not implemented, considering the low concentration of A 2A AR in rHDLs, whereas the correlated motion of Gd-DTPA-BMA and A 2A AR in rHDLs were taken into account by setting the diffusion coefficient of Gd-DTPA-BMA equal to that of A 2A AR in rHDL (approach B) (27). The hydrodynamic radius of A 2A AR in rHDLs was set to 6 nm.

SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/ content/full/6/12/eaay8544/DC1 Supplementary Text Fig. S1. Characterization of A2AAR prepared in rHDLs. Fig. S2. Assignments of the resonances from the methionine residues of A2AAR. Fig. S3. Introduction of methionine residues in the cytoplasmic ends of the A2AAR TM regions. Fig. S4. Resonances from methionine residues at the cytoplasmic ends of the A2AAR TM region. Fig. S5. NMR spectra of A2AAR in rHDL(POPC/POPG) at various temperatures. Fig. S6. Lipids used for the reconstitution of A2AAR into rHDLs. Fig. S7. Signaling activity and conformation of A2AAR in rHDL. Fig. S8. Conformational changes in the TM region of A2AAR upon activation. Fig. S9. Conformation of TM6. Table S1. [ 35 S]-GTPS binding to complexes of purified G protein and A2AAR in rHDL(POPC/ POPG) with 0, 0.5, 1, and 2 mol% trifluoroethanol (TFE), in the presence of the full agonist.
View/request a protocol for this paper from Bio-protocol.