G protein β4 as a structural determinant of enhanced nucleotide exchange in the A2AAR-Gs complex

Adenosine A2A receptors (A2AAR) evoke pleiotropic intracellular signaling events via activation of the stimulatory heterotrimeric G protein, Gs. Here, we used cryoEM to solve the agonist-bound structure of A2AAR in a complex with full-length Gs α and Gβ4γ2 (A2AAR-Gs α:β4γ2). The orthosteric binding site of A2AAR-Gs α:β4γ2 was similar to other structures of agonist-bound A2AAR, with or without Gs. Unexpectedly, the solvent accessible surface area within the interior of the complex was substantially larger for the complex with Gβ4 versus the closest analog, A2AAR-miniGs α:β1γ2. Consequently, there are fewer interactions between the switch II in Gs α and the Gβ4 torus. In reconstitution experiments Gβ4γ2 displayed a ten-fold higher efficiency over Gβ1γ2 in catalyzing A2AAR dependent GTPγS binding to Gs α. We propose that the less constrained switch II in A2AAR-Gs α:β4γ2 accounts for this increased efficiency. These results suggest that Gβ4 functions as a positive allosteric enhancer versus Gβ1.


Introduction
Heterotrimeric G proteins, composed of a Ga subunit and Gbg dimer, interact with hundreds of different cell surface GPCRs, transducing extracellular stimuli such as light, odorants and hormones to intracellular effectors 1 .The Gs-coupled A2A adenosine receptor (A2AAR) has been an important therapeutic target for ameliorating the effects of inflammation 2 , cancer 3 , cardiovascular 4 and neurological disease 5 .
Upon binding agonist, the A2AAR, aided by Gbg, catalyzes the exchange of GDP for GTP on Gs a, eliciting activation.The GTP bound Gs a has a lower affinity for Gbg, resulting in the dissociation of the Gs heterotrimer from A2AAR, thereby liberating Gs a and Gbg 6 , which can regulate effectors such as adenylyl cyclase to increase intracellular cAMP.GTP bound Gs a is deactivated by an intrinsic GTPase activity that converts GTP to GDP, a process that can be accelerated by Regulators of G protein Signaling (RGS) proteins 7 .The GDP bound Gs a has a higher affinity for Gbg and can bind sites on Gb that overlap effector binding sites, which inhibits signaling by Gbg.The reassembled Gs heterotrimer then binds the plasma membrane to repeat the cycle with another GPCR.
The magnitude of A2AAR signaling can be modulated by a variety of factors.For example, stimulation of A2AAR by a partial agonist results in a modest rate of nucleotide exchange in Gs, whereas a full agonist produces a more robust level of nucleotide exchange 8 .These differences in activity have been ascribed to distinct active states of the A2AAR-Gs complex, which have the same affinity between Gs and the A2AAR, but can be distinguished by structural differences in the vicinity of transmembrane helix 6 (TM6) of the A2AAR, as measured by Fluorine NMR 8 .In addition, the composition of the Gbg dimer in a Gs heterotrimer can affect the coupling efficiency at the A2AAR, exemplified by the ten-fold increase in efficiency of Gb4g2 over Gb1g2 at catalyzing A2AAR dependent GTPgS binding to purified Gs a 9 (Fig. 1A).Furthermore, the Gs a:b4g2 heterotrimer can convert a much higher proportion of A2AAR into the high affinity state than Gs a:b1g2 10 .However, the high ligand binding affinity of A2AAR was the same, whether the Gs heterotrimer contained Gb1 or Gb4.
One condition where modulation of A2AAR signaling would be biologically important is inflammation 11 .For example, inflammatory cytokines upregulate expression of the A2AAR in human dermal microvascular endothelial cells 12 ; which increases A2AAR signaling through Gs, resulting in increased cellular cAMP, countering the effects of inflammation.However, in addition to increasing A2AAR expression, inflammatory cytokines also increase Gb4 expression 12 .
Considering the functional attributes of Gb4 with respect to the A2AAR, increasing the ratio of A2AAR in the high affinity state, and increasing the efficiency of nucleotide exchange, it is not surprising that a cell would increase expression of both A2AAR and Gb4 to maximize signaling from A2AAR in response to an inflammatory stimulus.
Understanding the mechanism by which Gbg improves receptor coupling by 10-fold is essential for complete understanding of GPCR-G protein signaling.To this end, we solved the first cryoEM structure of agonist-bound A2AAR in a complex with full-length Gs a and the novel Gb4g2 heterodimer.Comparison of the A2AAR-Gs a:b4g2 complex with the previously solved A2AAR-miniGs a:b1g2 complex 13 enabled interrogation of structural differences between Gb1g2 and Gb4g2.Surprisingly, A2AAR-Gs a:b4g2 displayed significantly greater solvent accessible surface area that we propose accounts for the ten-fold increased efficiency of Gb4g2 over Gb1g2 in catalyzing A2AAR dependent GTPgS binding to Gs a.These results suggest that Gb4 functions on the cytoplasmic surface as a positive allosteric modulator, like partial and full agonists on the extracellular surface.

Biophysical and functional and characterization of A2AAR-Gs complexes
To investigate the role of Gb4 in A2AAR-Gs signaling, we used anti-FLAG M1 affinity resin (Supplemental Fig. 1A) to form a complex between heterotrimeric Gs a:b4g2 and A2AAR with a bound agonist UK-432,097 (Supplemental Fig. 1B).The eluted A2AAR-Gs a:b4g2 complex was monodisperse as judged by size exclusion chromatography (Supplemental Fig. 1C) and free of contaminating proteins as visualized by SDS-PAGE (Supplemental Fig. 1D).Functional activity of the A2AAR-Gs a:b4g2 complex was verified by the presence of GTPgS sensitive high affinity agonist binding, as measured by 5′-(N-Ethylcarboxamido)adenosine (NECA) displacement of bound 3 [H] ZM-241,385 (Supplemental Fig. 1E).

Structures of A2AAR and A2AAR-Gs
In the course of our work on solving the A2AAR-Gs complex structure, we solved the X-ray structure of agonist-bound A2AAR.The A2AAR construct for lipidic cubic phase crystallization contained T4 lysozyme (T4L) inserted in intracellular loop 3 (ICL3), and the structure was nearly identical to the published X-ray structure of A2AAR with the bound agonist UK-432,097 (RMSD 0.395 Å) 14 .As noted above, the Gb4g2 heterodimer was ten-fold more efficacious than the Gb1g2 heterodimer in catalyzing A2AAR dependent GTPgS binding to Gs a 9 (Fig. 1A), which was the overarching rationale for solving an A2AAR-Gs structure containing Gb4.For our cryoEM studies, expression of A2AAR in Sf9 insect cells was increased by truncating the C-tail and including an Nterminal T4L fiducial marker.Single-particle cryoEM of A2AAR-Gs a:b4g2 yielded a Coulomb potential (CP) map with a resolution ranging between 3.1 and 4.7 Å with an overall, gold-standard Fourier shell correlation (FSC) resolution of 3.5 Å (Fig. 1B).Corresponding models of the A2AAR-Gs a:b4g2 complex (Fig. 1C) recapitulate the general architecture of A2AAR in a complex with Ga (lacking the a-helical domain (so-called miniG)) and b1g2 13 .
The interior solvent accessible volume is larger in A2AAR-Gs a:Gb 4 g 2 versus A2AAR-miniGs a:Gb 1 g 2.
Structural alignment can be very useful for examining differences in subdomains between similar proteins; however, comparison of global structural changes in A2AAR, or any GPCR, under different conditions by alignment is inherently challenging.Significant differences can be averaged into many small changes throughout the receptor, which can be difficult to interpret.Upon initial inspection, the A2AAR from the complex presented here did not appear to differ from the A2AAR in the A2AAR-miniGs a:Gb1g2 13 (RMSD 1.22 Å).Not surprisingly, comparisons of the orthosteric binding site to other structures containing agonist bound A2AAR were also very similar, including UK-432,097 bound A2AAR-Gs a:Gb4g2, NECA bound A2AAR-miniGs a:Gb1g2 13 , NECA bound A2AAR-miniGs a 15 , and UK-432,097 bound A2AAR 14 .However, measurement of the solvent accessible surface area (SASA) is an orthogonal tool that can also be used to interrogate subtle structural differences between the same protein under different conditions.Using this technique, the solvent accessible volume below the orthosteric binding site and above the G protein binding site was larger in A2AAR-Gs a:b4g2 versus A2AAR-miniGs a:b1g2 (Fig. 2A, B, upper red circles), which was quantified in Fig. 3B, C. In particular, amino acid residues at the lower half of the receptor that interact with the G protein have increased SASA in A2AAR-Gs a:b4g2 versus A2AAR-miniGs a:b1g2.(Fig. 3D-F).Note also that alignment of receptors emphasizes the downward shift of Gs in A2AAR-Gs a:b4g2 versus A2AAR-miniGs a:b1g2 (Fig. 3D), which increases the solvent accessible volume (Fig. 2A).In addition, the cavity space is increased between Gs a and Gb4 versus miniGs a and Gb1 (Fig. 2A, B, lower red circles).
The increased SASA in A2AAR-Gs a:b 4 g 2 elicits a proline switch in TM7.
Many of the receptor residues that displayed greater SASA in A2AAR-Gs a:b4g2 were localized in TM6 and TM7 (Fig. 3B, C).Closer examination revealed that TM6 of A2AAR of A2AAR-Gs a:b4g2 aligns very well with TM6 of A2AAR-miniGs a:b1g2 (Fig. 4A).However, this alignment emphasizes differences in the TM7 backbones.P285 7.50 (superscript refers to Ballesteros-Weinstein numbering system 16 ) of the NPXXY motif resides at the lower third of TM7, which is an important conformational switch in GPCR activation 17 .The proline in the NPXXY motif induces a kink in TM7 (Fig. 4A) 18 .Importantly, P285 7.50 in A2AAR of A2AAR-Gs a:b4g2 is in the endo conformation, whereas the analogous proline in A2AAR-miniGs a:b1g2 is in the exo conformation (Fig. 4B).This switch is triggered by the increased solvent accessible volume in A2AAR-Gs a:b4g2, beginning at the proline 285 kink (Fig. 4A).A consequence of the increased solvent accessible volume is a loss of the hydrogen bond between S234 6.36 in TM6 and R291 7.56 in TM7 in A2AAR-Gs a:b4g2 (Fig. 4C), whereas this hydrogen bond is retained in A2AAR-miniGs a:b1g2 (Fig. 4D).This interpretation is supported by the cryoEM maps (Fig. 4E, F).
Expansion of the Gs a-Gb 4 interface elicits an extension in the N-terminal helix (HN) of Gs a, which functions as an "actuator" to enlarge the cavity of the complex.
To highlight a prominent left lateral shift of Gs a HN in A2AAR-Gs a:b4g2, (gold) vs A2AAR-miniGs a:b1g2 (grey), we aligned the complexes using the Gs a Ras domains (Fig. 5A).This lateral shift of HN is associated with an increase in the length of the Gs a subunit in A2AAR-Gs a:b4g2 along the HN-H4 axis by > 3 Å (as measured between a carbons of residues Q35 G.HN.52 and Y339 G.H4.08 ) than the analogous measurement (Q35-Y329) in miniGs a in A2AAR-miniGs a:b1g2 (Fig. 5A, red dashed line).SASA analysis of the Gs a-Gb contact surface revealed a larger area (3x to 50x) in A2AAR-Gs a:b4g2 than the analogous contact surface in A2AAR-miniGs a:b1g2 (Fig. 5C, D and E).With the Ras domains aligned, we note that HN in Gs a:b4g2 is shifted left with respect to ICL2 (Fig. 5A, red circle), compared with HN in miniGs a:b1g2.This shift occurs between Q31 and H41 at a contact site with ICL2 (Fig. 5B).Given the tight contacts between Gs a HN and the Gb torus (Fig. 5A and C, red ovals), the left lateral extension of HN in Gs a:b4g2 (gold arrow) results in lateral translation of the b4 torus compared with the b1 torus (Fig. 5A), thereby increasing the SASA in Gs a:b4g2.In this way we propose that HN functions as an "actuator" to manifest this expansion.These conformational changes are depicted as cartoons in Fig. 6A.

Discussion
Comparison of the cryoEM structures of A2AAR-Gs a:b4g2 and A2AAR-miniGs a:b1g2 revealed dramatic expansion of the solvent accessible volume in the complex containing Gb4 versus Gb1.However, we note several caveats in this comparison.The receptor in A2AAR-Gs a:b4g2 was deglycosylated by treatment with PNGaseF, whereas the receptor A2AAR-miniGs contained a N154A mutation to prevent glycosylation.In addition, A2AAR-miniGs a:b1g2 is so named because it lacks the a-helical domain and Switch III of Gs a and contains several mutations that are distinguished from full-length, wild-type Gs a in our complex (Supplemental Fig. 2).
Furthermore, A2AAR-Gs a:b4g2 is prenylated, whereas A2AAR-miniGs a:b1g2 contains a mutated Gg2 isoform to prevent prenylation.Lastly, A2AAR-miniGs a:b1g2 differs from A2AAR-Gs a:b4g2 by binding NECA instead of UK-432,097, which has higher receptor affinity.Notwithstanding these differences, profound functional differences have been observed in reconstitution 9 and expression systems 10 that contained identical A2AAR, agonist and G protein subunits, with the Gb1 or Gb4 isoform being the only variable (Fig. 1A).Consequently, structural differences would be expected to arise from A2AAR-Gs complexes differing only by the Gb isoform.Scrutiny of the structures reveals that Gb4 elicits a dramatic increase in the solvent accessible volume within the complex (Fig. 2).Significant structural correlates are an extension of the Gs a helix HN and a switch at proline 285 in receptor helix TM7 (Fig. 4A, B).The mechanism for the increased distance between the Gb4 torus and the Gs a Ras domain is not clear; however, it appears to be enabled by an elastic region of the Gs a HN that contacts ICL2 (Fig. 5B).In the case of A2AAR-Gs a:b4g2, it is likely that the structural relationship between Gb4 and Gs a expands this elastic region of HN, employing HN as an "actuator switch" to enlarge the interior cavity of A2AAR via interactions with ICL2 in a process that is accommodated by the proline 285 switch (Fig. 4A, B).
The increased separation between Gs a and Gb in A2AAR-Gs a:Gb4g2 versus A2AAR-miniGs a:Gb1g2 (Fig. 5 and depicted as cartoons in Fig. 6A), suggests a mechanism for the tenfold increased coupling efficiency of Gb4g2 over Gb1g2 in catalyzing A2AAR dependent GTPgS binding to Gs a 9 (Fig. 1A).In particular, the conformational change required of switch II to bind GTP would be less constrained by Gb4 (Fig. 6B).In contrast, tighter interactions between miniGs a and Gb1 (Fig. 6C) would provide a higher energy barrier for switch II to undergo the GTP dependent conformational change.Fig. 6D illustrates H3 and Switch II of the GDP bound Gs a:b1g2 heterotrimer 19 , highlighting the proximity of Switch II with Gb1.In contrast, Fig. 6E shows the same orientation of the GSP activated Gs a subunit 20 , with Switch II adopting a different conformation.Note that R231 G.H2.04 of Switch II binds to Gb1 in the GDP bound form of Gs a (Fig. 6D), but swings around to form a salt bridge with E268 G.H3.04 of H3 in the activated GSP bound state (Fig. 6E); this ionic bond has been referred to as a "hasp," and was determined to be necessary to stabilize the GTP bound form of Gs a 21 .Thus, the nature of the Gs a:Gb interface is likely related to the kinetics of nucleotide binding, with a more open interface allowing more rapid binding of GTP.Whether the effect of Gb4 on the active states of GPCRs extends beyond A2AAR has yet to be determined.Intriguingly, reconstitution experiments have shown that the Go a:b4g2 heterotrimer displayed an increased level of M2 Receptor (M2R) dependent nucleotide exchange compared to the Go a:b1g2 heterotrimer, suggesting an enhanced ability of Gb4 to activate M2R 22 .
We propose that the enhancement of signaling efficiency by Gb4 versus Gb1 is a result of the increased solvent accessible volume in the complex with Gb4 versus Gb1.This observation that occurs on the cytoplasmic side of the A2AAR complex is remarkably parallel to the activity of full and partial agonists in the orthostatic site on the extracellular side of A2AAR.One possible explanation for the enhanced signaling efficiency of full vs partial agonists was proposed in a study that observed different conformations and levels of signaling efficiency of A2AAR, depending on the specific agonist 8 .Modulation of active states may be similarly occurring on the cytoplasmic side of A2AAR.That is, the effect of Gb4 is analogous to a full agonist, inducing an active state with higher signaling activity than Gb1, which would be analogous to a partial agonist.

Design of the A2AAR and Gs constructs
The wild-type A2AAR construct for cryoEM contained a hemagluttinin (HA) signal sequence, a FLAG epitope, a TEV cleavage site and a T4 lysozyme sequence, which was attached to the amino terminus of A2AAR lacking the first four residues.The C-terminus was truncated at position 316, and the final construct was designated HA-Flag-TEV-T4L-A2AAR-316.
The difference for the construct used for X-ray crystallography was that T4 lysozyme was inserted in the third intracellular loop rather than at the amino terminus 14 .The Gs heterotrimer contained human Gs a, Gβ4 containing a V226E mutation in order to bind a stabilizing nanobody (Nb35) which was used for stabilizing the b2-adrenergic receptor Gs complex 23 , and an amino-terminally 6His-FLAG tagged human Gγ2 24 .The plasmids containing Nb35 was kindly provided by S.G.
Rassmussen (Stanford University).A Gg2 clone from bovine brain 25 (which translates to the same protein sequence as the human Gg2) was ligated into the pDoubleTrouble vector (pDT) 26 in order to add a 6His-FLAG tag to the N-terminus of Gγ2 (Gg2FH) 24 .The Gg2FH coding sequence was subcloned into the transfer vector pVL1393, and recombinant baculoviruses were isolated as described 24 .

Insect cell expression and purification of A2AAR, the Gs heterotrimer and Nb35
A2AAR and the Gs heterotrimer were expressed separately in Sf9 (S. frugiperda) insect cells using the Bac-to-Bac Baculovirus Expression System (Invitrogen).For A2AAR, Sf9 cells were grown in ESF921 media (Expression Systems) at 27 ºC, diluted to a density of 2.0 x 10 6 cells/mL and infected for 48 hrs with a high-titre baculovirus stock at a multiplicity of infection (MOI) of 3, harvested by centrifugation and stored at -80 ºC.Viral titers were performed by a flow cytometric method 27 on a Guava easycyte 8HT after staining the cells with anti-gp64-PE antibody (Expression Systems.)For the Gs heterotrimer, Sf9 cells were diluted to a density of 2.5 x 10 6 cells/mL and then infected with three separate baculoviruses (Gs a, Gβ4 and Gγ2FH) as described 9 .
Membranes were prepared from receptor-infected insect cells based on the method by Jaakola et al. 28 .All steps were performed at 4 o C unless otherwise stated.Cells were resuspended in a hypotonic Buffer A (10 mM HEPES, pH 7.5, 10 mM MgCl2, 20 mM KCl and EDTA-free protease inhibitor cocktail (Roche)) and lysed by Dounce homogenization (~20 strokes/wash); membranes were recovered by centrifugation at 100,000 x g and washed again with Buffer A. Two additional washes were performed with buffer A containing 1M NaCl (Buffer B).After the final wash, membranes were weighed, resuspended in 2.5 volumes of Buffer A containing 40% (v/v) glycerol (Buffer C), and flash frozen in liquid nitrogen.
For purification of A2AAR, frozen membranes were thawed in a 10-fold excess of 20 mM HEPES, pH 7.5, 10% glycerol, 4 mM CaCl2, 100 µM adenosine and EDTA-free protease inhibitor cocktail (Roche) (Buffer D) containing 300 mM NaCl and extracted with 0.5% n-dodecyl-b-Dmaltoside/cholesteryl hemisuccinate (DDM/CHS) for 4 hrs with gentle rotation.Receptor extracts were clarified by high-speed centrifugation at 100,000 x g and then added to a slurry of anti-FLAG M1 agarose affinity gel (Sigma-Aldrich, catalog number A4596) for overnight binding with gentle rocking at 4 0 C The slurry was collected in a column at 1 g, and the beads with bound A2AAR were washed with 30 column volumes of Buffer D containing 300 mM NaCl and 0.1% DDM/CHS, followed by a wash with 30 column volumes of Buffer D containing 500 mM NaCl and 0.05% DDM/CHS.The final wash was with 20 column volumes of Buffer D containing 150 mM NaCl and 0.025% DDM/CHS.The FLAG-M1 affinity resin containing bound A2AAR was then combined with purified Gs to assemble the A2AAR-Gs complex (described below).
Sf9 cell pellets containing Gs a, b4 and an N-terminally 6HIS-FLAG tagged g2 were lysed in a five-fold excess of buffer containing Buffer E (20 mM Tris, pH 8.0, 10 µM GDP and EDTAfree protease inhibitor cocktail).All steps were performed at 4 o C unless otherwise stated.Unbroken cells and debris were removed by centrifugation at 2,500 x g for 10 min, and membranes were collected from the supernatant by centrifugation at 100,000 x g.Membranes containing Gs were suspended and extracted with Buffer E containing 100 mM NaCl and 1% cholate 9 by gentle rotation for 1 hr.The cholate extract was clarified by centrifugation at 100,000 x g, supplemented with 20 mM imidazole, and loaded onto a Ni-NTA superflow (Qiagen) column equilibrated with Buffer E. Contaminants were removed by washing with 8 column volumes of Buffer F (20 mM HEPES, pH 7.5, 500 mM NaCl, 0.5% Genapol C-100 (Sigma-Aldrich), containing 1 mM MgCl2, 10 µM GDP, 20 mM imidazole and EDTA-free protease inhibitor cocktail.Detergent was exchanged by washing with 1.5 column volumes of Buffer G (20 mM HEPES, pH 7.5, 100 mM NaCl, 0.1% DDM/CHS, 1 mM MgCl2, 10 µM GDP, and EDTA-free protease inhibitor cocktail), and the heterotrimeric Gs was eluted with Buffer G containing 200 mM imidazole.Elution fractions containing protein were identified by SDS-PAGE and pooled and concentrated using an Amicon Ultra-15 50,000 MWCO (Millipore).Concentrated protein was diluted with Buffer H (20 mM HEPES, pH 7.5, 20 mM NaCl, 0.025% DDM/CHS, 10 mM MgCl2, 1 mM EDTA, 100 µM TCEP and EDTA-free protease inhibitor cocktail) to lower the salt concentration, and the solution was loaded onto a 1.3 mL UnoQ Q1 column (BioRad).Protein was eluted with a linear gradient of 0% Buffer H to 40% Buffer I (Buffer H with 1M NaCl instead of 20 mM NaCl) over 20 mL.Collection wells contained a 10x solution of GDP to yield a final concentration of 10 µM.Fractions containing Gs were pooled, concentrated using an Amicon Ultra-4 50,000 MWCO (Millipore), aliquoted, flash frozen in liquid nitrogen and stored at -80 o C. The 6His tagged nanobody Nb35 was expressed in E.coli and purified by Immobilized Metal Affinity Chromatography as described 29 .

Purification of A2AAR-Gs complexes
Formation of A2AAR-Gs complex utilized the washed A2AAR bound to FLAG-M1 affinity resin (described above) and purified Gs at a ratio of 1:1.5 A2AAR:Gs a, based on Simply Blue staining of A2AAR and Gs a gel bands.Excess GDP was removed from purified Gs by applying protein to Dowex 1x4 ion exchange resin (Sigma-Aldrich, catalog number 428612) and eluting with Buffer J (25 mM HEPES, pH 7.5, 150 mM NaCl, 10% glycerol, 0.025% DDM/CHS, 1 mM MgCl2, 4 mM CaCl2, 100 µM adenosine and EDTA free protease inhibitor cocktail).Fractions containing Gs were pooled and combined with the A2AAR bound to FLAG-M1 affinity resin.Apyrase (7.5 units; NEB catalog number MO398L) and PNGaseF (500 units; NEB catalog number PO7O4S) were added to the reaction mixture, which was allowed to incubate at room temperature for 1 hr with gentle rotation.Nanobodies (~2-fold excess) were also added during this incubation.
The FLAG-M1 affinity resin containing A2AAR-Gs complex and nanobodies was then washed with 10 mL of Buffer J in a column at 1 g to remove contaminants, which included excess Gbg, due to the location of the affinity tag on Gg.A2AAR-Gs complex was eluted with Buffer J containing 10 mM EDTA; collection tubes contained MgCl2 to give a final concentration of 15 mM MgCl2.Eluted A2AAR-Gs complex was pooled, and 10 µM UK-432,097 (Axon Medchem) was added to displace bound adenosine.Protein was concentrated using a Vivaspin 6 100,000 MWCO (GE Healthcare), and success of complex formation was initially judged by SEC using a Superdex 200 10/300 GL column (GE Healthcare Life Sciences) equilibrated with Buffer K (25 mM HEPES, pH 7.5, 100 mM NaCl, 0.025% DDM/CHS, 10 mM MgCl2, 1 mM EDTA, 100 µM TCEP and 1 µM UK-432,097.Peak fractions of the SEC purified A2AAR-Gs complex were pooled and crosslinked with 0.1% glutaraldehyde (SigmaAldrich) for 30 min at 4 o C, quenched with 0.1 M Tris, pH 8.0, and concentrated to ~400 µl.The amphipol A8-35 was then added to the A2AAR-Gs complex (100 µl of 10 mg/mL stock prepared using Buffer K without DDM/CHS) and incubated overnight at 4 o C with gentle rotation.Detergent was removed by a final SEC purification using Buffer K without DDM/CHS.Peak fractions of the A2AAR-Gs complex were pooled and concentrated in a 0.5 mL Vivaspin 100,000 MWCO concentrator prior to application onto cryoEM grids.

CryoEM grid preparation
CryoEM grids were prepared in the University of Virginia Molecular Electron Microscopy Core (MEMC) using a Vitrobot Mark IV with the blotting chamber at 4°C and 95 % relative humidity.
Whatman #1 filter paper was used to blot grids using a blot force of 7 and blot time of 14 s.UltrAuFoil 1.2/1.3,300 mesh grids (Quantifoil) were glow discharged in a Pelco EasiGlow for 60 s at 20 mA.A 2.1 µL aliquot of A2AAR-Gs was applied to the gold-film side of each grid and vitrified in liquid ethane cooled by liquid nitrogen.

CryoEM data collection
CryoEM data were collected in the University of Virginia Molecular Electron Microscopy Core (MEMC) on a ThermoFisher Scientific Titan Krios equipped with a Gatan BioQuantum-K3 energy filter and detector (Supplemental Table 1).

CryoEM image processing and reconstruction
All processing was performed in cryoSPARC 3.0.0(DOI: 10.1038/nmeth.4169).Patch motion correction in cryoSPARC Live was performed during data collection to assess grid and sample quality.Following data collection, processing was done in the standard cryoSPARC suite.
CTFFIND4 was used for CTF estimation (https://doi.org/10.1016/j.jsb.2015.08.008).3,873   micrographs with defocus values between -2.5 and -0.8 µm and estimated maximum resolution better than 3.7 Å were used for further processing.Blob picking with 100 micrographs resulted in 54,657 particles that were extracted with a box size of 256 pixels (358.4Å). 2D classification with these particles was used to generate five templates that were used for template-based particle picking from all 3,873 micrographs.Template particle picking followed by particle inspection with NCC=0.4 and local power between 121 and 155 resulted in 2,600,922 particles.Particle extraction from 1000 micrographs with a box size of 256 pixels followed by 2D classification and selection resulted in 355,217 particles that were used to generate three ab initio reconstructions.Particles were extracted from the remaining micrographs and ARG particles were purified by 40 iterations of 2D classification using 100 classes and a circular mask of 200 Å.2D class selection resulted in 1,143,600 particles that were subjected to heterogeneous refinement using the three ab initio reconstructions.The best 3D class contained 545,805 particles that were used for non-uniform refinement that resulted in a map with a cryoSPARC GSFSC resolution of 3.4 Å (FSC=0.143).
These particles were subjected to heterogeneous refinement with 6 classes and one class containing 123,274 particles.These particles were used for another round of non-uniform refinement that resulted in a map with a cryoSPARC GSFSC resolution of 3.5 Å (FSC=0.143)but that was qualitatively better in some regions.This map was used for local resolution estimation followed by local map filtering.The non-uniform refinement and the locally filtered Coulomb potential maps were used for model building and refinement.

Model building and refinement
We constructed a model of the A2AAR-Gs complex, using a combination of X-ray crystallography and homology modeling, at a time when the only experimentally determined GPCR-G protein structure was the X-ray structure of β2AR-Gs (PDB:3SN6) 23 .The agonist-bound A2AAR X-ray structure had also been solved (PDB:3QAK) 14 , but we also solved the X-ray structure of agonistbound A2AAR in the course of our work on solving the A2AAR-Gs complex structure by X-ray crystallography.Briefly, we expressed, purified, and crystallized agonist-bound (UK-4320097) A2AAR-T4L-ΔC as previously described 14 with minor modifications.Purified A2AAR-T4L-ΔC-UK-432,097 in DDM/CHS was reconstituted into a 90%/10% monoolein/cholesterol mixture and was used with the CHARMM36 force field to equilibrate the system as an NPT ensemble followed by minimization and unrestrained all-atom simulation (40 ns) as an NVT ensemble with T=310 K.
Nb35 was manually added to the MD-minimized A2AAR-Gs model based on the position of the nanobody in the β2AR-Gs structure.The A2AAR-Gs-Nb35 model was docked into the A2AAR-Gs cryoEM map using Phenix Dock In Map.The map also revealed densities consistent with cholesteryl-hemisuccinate (CHS) molecules.CHS molecules were extracted from PDB:7D76 (DOI: 10.1038/s41586-020-03083-w) and manually docked into the Coulomb potential map, five at the extracellular ends of TM1-4, three at the cytoplasmic ends of these helices and two at the extracellular end of TM6 and ECL3.The A2AAR-Gs-Nb35 model was refined using alternating cycles of Phenix Real Space Refine and manual real space refinement in COOT.

Fig. 1
Figures and Tables

Fig. 3
Fig.3The increase in SASA at the cytosolic side of the A2AAR in A2AAR-Gs a:b 4 g 2 is associated

Fig. 4 |
Fig. 4 | The proline 285 switch in A2AAR TM7 is associated with enlargement of the Gs a

Fig. 5
Fig.5The N-terminal helix (HN) of Gs a serves as an "actuator" linking ICL2 and the Gb

Fig. 6
Fig. 6 Movement of the Gs a switch II is less constrained in A2AAR-Gs a:b 4 g 2 versus A2AAR-

Fig. 1 |
Fig. 1 | CryoEM structure of the A 2A AR-Gs complex containing the Gb 4 isoform.(A) Published data from McIntire et al. 9 demonstrating ten-fold higher efficiency of Gb 4 g 2 over Gb 1 g 2 in catalyzing A 2A AR dependent GTPgS binding to Gs a. (B) Two views of the A 2A AR-Gs complex CryoEM map colored and filtered by local resolution.(C) Ribbon representation of CryoEM map of the A 2A AR-Gs complex: A 2A ARreddish purple, Gs a-yellow, Gb 4 -blue, Gg 2 -bluish green, Nb35-magenta.

Figure 6 G protein b 4
Figure 6 These Coulomb potential maps were also used to