*Unitat de Farmacologia, Facultat de Medicina, Departament de Patologia i Terapèutica

Experimental, Universitat de Barcelona, L’Hospitalet de Llobregat, Barcelona, Spain

†Molecular Recognition Section, Laboratory of Bioorganic Chemistry, National Institute of Diabetes

and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, Maryland, USA

‡Department of Neuroscience, Karolinska Institutet, Stockholm, Sweden

AbstractIn the CNS, an antagonistic interaction has been shownbetween adenosine A2A and dopamine D2 receptors (A2ARsand D2Rs) that may be relevant both in normal and patho-logical conditions (i.e., Parkinson’s disease). Thus, the molec-ular determinants mediating this receptor–receptor interactionhave recently been explored, as the fine tuning of this target(namely the A2AR/D2R oligomer) could possibly improve thetreatment of certain CNS diseases. Here, we used a fluores-cence resonance energy transfer-based approach to examinethe allosteric modulation of the D2R within the A2AR/D2Roligomer and the dependence of this receptor–receptor inter-action on two regions rich in positive charges on intracellular

loop 3 of the D2R. Interestingly, we observed a negativeallosteric effect of the D2R agonist quinpirole on A2AR ligandbinding and activation. However, these allosteric effects wereabolished upon mutation of specific arginine residues (217–222 and 267–269) on intracellular loop 3 of the D2R, thusdemonstrating a major role of these positively chargedresidues in mediating the observed receptor–receptor interac-tion. Overall, these results provide structural insights to betterunderstand the functioning of the A2AR/D2R oligomer in livingcells.Keywords: A2AR, D2R, FRET, oligomerization, allosterism,fluorescent agonist.J. Neurochem. (2012) 123, 373–384.

The existence of oligomeric complexes comprising adeno-sine and dopamine receptors (i.e., A2AR/D2R) has beenpostulated to be relevant for proper striatal function both innormal and pathological conditions (Fuxe et al. 2010). Forinstance, these A2AR/D2R oligomers have been implicated inthe control of striatal processes such as motor activitycontrol, motor learning, or some forms of associative andvisual learning (Bornstein and Daw 2011). Interestingly, theA2AR–D2R interaction occurs in the somatodendritic area ofGABAergic enkephalinergic striatal neurons (Cabello et al.2009; Ferre et al. 2007a,b; Fuxe et al. 1998, 2003). At thislevel, the coexistence of reciprocal antagonistic interactionsbetween these receptors has been further described, namelyan intramembrane interaction in which the A2AR mediatesinhibition of the D2R, thus modulating neuronal excitabilityand neurotransmitter release, and an interaction at the level ofadenylate cyclase (AC), in which the D2R inhibits A2AR-mediated protein phosphorylation and gene expression

(Shindou et al. 2002; Svenningsson et al. 2000). Thus, therationale to study the precise mechanism behind these A2AR–D2R molecular and functional interactions is well suited asthe A2AR/D2R oligomer has been postulated to be respon-sible for the neuronal signal integration coming from twodifferent neurotransmitter systems (i.e., dopaminergic andadenosinergic) (Ferre et al. 2007b).

Received June 4, 2012; revised manuscript received August 9, 2012;accepted August 10, 2012.Address correspondence and reprint requests to Francisco Ciruela,

Unitat de Farmacologia, Facultat de Medicina, Departament Patologia iTerapèutica Experimental, Universitat de Barcelona, L’Hospitalet deLlobregat, Barcelona 08907, Spain. E-mail: fciruela@ub.eduAbbreviations used: AGT, alkyltransferase; APEC, 2-[[2-[4-[2-

(2-aminoethyl)-aminocarbonyl]ethyl]phenyl]ethylamino]-5′-N-ethyl-carboxamidoadenosine; FRET, fluorescence resonance energytransfer; HEK, Human embryonic kidney; IL3, third intracellularloop 3; RIPA, radio immunoprecipitation assay .

Journal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384 373© 2012 The Authors

JOURNAL OF NEUROCHEMISTRY | 2012 | 123 | 373–384 doi: 10.1111/j.1471-4159.2012.07956.x

Importantly, in recent years the existence of the A2AR/D2Rheteromer has been revealed in living cells, while in native tissuethe results are still not conclusive, by biochemical (i.e.,co-immunoprecipitation) and biophysical (i.e., resonance energytransfer-based) approaches (Agnati et al. 2005; Cabello et al.2009; Canals et al., 2003; Genedani et al. 2005; Kamiya et al.2003). In addition, computerized modeling and techniques of pulldown and mass spectrometry have shown the heteromerizationbetween the A2AR and the D2R to depend on a Coulombicinteraction between the third intracellular loop (IL3) of the D2Rand the C-terminal tail of the A2AR (Canals et al., 2003; Ciruelaet al. 2004; Woods and Ferre 2005; Woods et al. 2005). Thus,two positively charged arginine richmotifs (215VLRRRRKRV223

and 265EVRRRNV271) located at the N-terminal part of theD2R-IL3 may interact with two distinct negatively charged motifsfrom the C-terminal tail of the A2AR. These motifs are the388HELKGVCPEPPGLDDPLAQDGAVGS412 domain, whichcontains two adjacent aspartic acid residues, or the 370SAQEpSQGNT378 domain, which contains a phosphorylatable serineresidue (S374), and thus their interaction with the D2R formselectrostatic bonds of covalent-like strength (Canals et al., 2003;Ciruela et al. 2004; Woods and Ferre 2005; Woods et al. 2005).Interestingly, we recently demonstrated that the point

mutation of serine 374 to alanine (S374A) reduced A2AR/D2R heteromerization and blocked the allosteric modulation ofthe D2R by the A2AR (Borroto-Escuela et al. 2010a,b), aresult that was further confirmed (Navarro et al. 2010). Also,when the S374A point substitution was accompanied bymutation of the two negatively charged aspartates on the A2ARC-terminal tail (D401A/D402A, see above), a synergisticreduction in the physical A2AR/D2R interaction and the loss ofantagonistic allosteric modulation over the A2AR/D2R inter-face were observed (Borroto-Escuela et al. 2010a,b). Simi-larly, it was recently shown the importance of the serineresidue on the present A2AR–D2R interaction (Azdad et al.2009). Thus, by means of competitive peptides mimicking theserine-containing epitope it was possible to preclude,performing perforated-patch-clamp recordings on brain slices,the ability of the A2AR to counteract the effects of D2Ractivation (Azdad et al. 2009). In addition, by using synthetictransmembrane (TM) a-helical peptides of the D2R, the role ofhelical interactions within the A2AR/D2R heteromeric TMinterface was also recently explored. Thus, TMs-IV and V ofthe D2R may play a critical role in the A2AR/D2R heteromericinterface as incubation with peptides corresponding to thesedomains significantly reduced the ability of the A2AR and D2Rto heteromerize. Also, the incubation with TM-IV or TM-Vpeptides blocked the allosteric modulation normally found inthe A2AR/D2R heteromer (Borroto-Escuela et al. 2010a,b).Overall, the above-mentioned results further corroborated theexistence of an electrostatic interaction between the C-terminaltail of the A2AR and IL3 of the D2R. On the other hand, whilethe A2AR-mediated allosteric effects on the D2R have beenobject of numerous studies, the reciprocal interaction is a less

studied aspect of this antagonistic receptor–receptor interac-tion. Hence, in the present work, we aimed to study the D2R-mediated allosteric modulation of the A2AR functionalitywithin the framework of A2AR/D2R oligomers, and todemonstrate that this receptor–receptor allosterism is mostlymediated by two regions rich in positive charges locatedwithin IL3 of the D2R.

Materials and methods

Plasmid constructs

We created an A2AR sensor (A2ARCFP) to perform fluorescence

resonance energy transfer (FRET) experiments with the fluorescentagonist MRS5424, a nucleoside labeled strategically with AlexaFluor 532. To this end, a pcDNA3.1 vector (Invitrogen, Carlsbad,CA, USA) containing (from 5′ to 3′) a signal peptide that targets thereceptor to the cell membrane, the fluorescent protein CyanFluorescence Protein (CFP), and the parathyroid hormone receptortype 1 (PTHR1) was used as a template (kindly provided by Dr. J.P. Vilardaga, University of Pittsburgh, Pittsburgh, PA, USA). Thus,the cDNA encoding the A2AR (Cabello et al. 2009) was amplifiedby a polymerase chain reaction using the primers FA2ABam (5′-CGCGGATCCATGCCCATCATGGGCTCC-3′) and RA2AXho(5′-CGCCTCGAGTCAGGACACCCTGTCTCC-3′), and subclonedinto the BamHI/XhoI sites replacing the PTHR1 in the above-mentioned template vector.

The mutations at the IL3 of the D2R were introduced into the cDNAencoding the wild-type D2R cloned in pEYFP-N1 (i.e., D2R

YFP)(Clontech, Heidelberg, Germany) (Cabello et al. 2009). Thus, tworegions rich in positive charges (amino acids 217–222 and 267–269)within the IL3 were sequentially mutated to alanine by using theQuickChangeTM site-directed mutagenesis kit (Stratagene Europe,Amsterdam, The Netherlands) and following the manufacturer’sinstructions. Accordingly, the 217–222 region was first mutated fromRRRRKR to AAAAKA, and thereafter, the obtained 217–222 mutantD2R cDNAwas used as a template for the second set ofmutations (267R–269R–267A–269A). The mutagenic primers were designed usingStratagene’s web-based program: F217–222A (5′-ATCAAGATCTA-CATTGTCCTCGCCGCAGCCGCCAAGGCAGTCAACACCAAACGCAGCAGC-3′) and R217–222A (5′-GCTGCTGCGTTTGGTGTTGACTGCCTTGGCGGCTGCGGCGAGGACAATGTAGATCTT-3′) for the first set of mutations, and F267–269A (5′-TGGGAGTTTCCCAGTGAAGCGGGCGGCAGTGGTGCAGGAGGCTGC-3′) and R267–269A (5′-GGCAGCCTCCTGCACCACTGCCGCCGCGTTCACTGGGAAACTCCC-3′) for the second. The result-ing mutated D2R construct, D2Rmut

YFP, was verified by DNAsequencing with BIGDYE® terminator V3.1 cycle sequencing kit(Applied Biosystems Hispania S.A., Alcobendas, Spain).

Finally, to visualize the presence of the D2R during our real-timeFRET experiments, we tagged D2R constructs at their N-terminuswith the O6-alkylguanine-DNA alkyltransferase (AGT), which is a24-kDa protein that acts as a suicide enzyme to covalently transfermodifications from DNA bases onto it. Interestingly, once the AGT isfused to the protein of interest, it is possible to stain it by means offluorescent AGT substrates, which if non-permeable allow theidentification of cell surface proteins over periods of hours becauseof the fact that the complex formed is extremely stable (Ciruela et al.2010). To this end, a pcDNA3.1 vector containing (from 5′ to 3′) the

Journal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384© 2012 The Authors

374 V. Fernandez-Duenas et al.

signal peptide of the metabotropic glutamate 5 receptor, the AGTprotein and adenosine A1 receptor was used as a template (kindlyprovided by Dr. J. P. Pin, Université de Montpellier and Institut deGénomique Fonctionnelle, Montpellier, France). Thus, the cDNAencoding the D2R and the D2Rmut were amplified by polymerasechain reaction using the primers FD2SNAP (5′-GCCGCTCGAGGATCCACTGAATCTGTCCTGG-3′) and RD2SNAP (5′-GCCGAAGCTTTCAGCAGTGGAGGATCTTCAGGAAGGCC-3′),and subcloned into the XhoI/HindIII sites replacing the A1 receptorin the above-mentioned template vector.

Synthesis of the A2AR fluorescent agonist

APECAlexa532 (MRS5424) is a functional agonist at the adenosineA2AR, in which the fluorescent dye Alexa Fluor 532 is covalentlyattached to a functionalized A2AR agonist 2-[[2-[4-[2-(2-aminoethyl)-aminocarbonyl]ethyl]phenyl]ethylamino]-5′-N-ethyl-carboxamido-adenosine (APEC). Briefly, similar to the method previouslydescribed for a related APEC conjugate (Brand et al. 2008),MRS5424 was synthesized as follows: Alexa Fluor 532 carboxylicacid, N-succinimidyl ester (1.0 mg, 1.38 lmol) was dissolved inanhydrous DMF (200 lL). Freshly prepared sodium tetraboratelabeling buffer (0.1 M, 1 mL, pH 8.5) and APEC (1.12 mg,2.07 lmol), which was dissolved in anhydrous DMF (200 lL),were added. The reaction mixture was protected from light and afterstirring for 18 h, the mixture was diluted with H2O (600 lL) andpurification was performed by HPLC with a Luna 5 l RP-C18(2)semipreparative column (250 9 10.0 mm; Phenomenex, Torrance,CA, USA) under the following conditions: flow rate of 2 mL/min;10 mM triethylammonium acetate (TEAA)-CH3CN from 100 : 0(v/v) to 70 : 30 (v/v) in 30 min. The homogeneous product wasisolated in the triethylammonium salt form with an HPLC retentiontime of 13.5 min. Analytical purity of this conjugate was checkedusing a Hewlett-Packard 1100 HPLC equipped with a Zorbax SB-Aq 5 lm analytical column (50 9 4.6 mm; Agilent TechnologiesInc, Palo Alto, CA, USA). Mobile phase: linear gradient solventsystem: 5 mM TBAP (tetrabutylammonium dihydrogenphosphate)-CH3CN from 80 : 20 to 40 : 60 in 13 min; the flow rate was0.5 mL/min (retention time 9.08 min). Peaks were detected by UVabsorption with a diode array detector at 254, 275, and 280 nm, andthe yield of MRS5424 was 0.67 mg (31%). ESI-HRMS m/z1150.4142 [M + H]+, C55H63N11O13S2�H+: Calcd. 1150.4127).

Cell culture, transfection, and membrane preparation

Human embryonic kidney (HEK293) cells were grown in Dul-becco’s modified Eagle’s medium (Sigma-Aldrich) supplementedwith 1 mM sodium pyruvate, 2 mM L-glutamine, 100 U⁄mLstreptomycin, 100 mg⁄mL penicillin and 5% (v⁄v) fetal bovineserum at 37°C and in an atmosphere of 5% CO2. HEK293 cellsgrowing in 25-cm2

flasks or six-well plates containing 18 mmcoverslips were used for western blot analysis and fluorescenceimaging, respectively, were transiently transfected with the cDNAencoding the specified proteins using TransFectinTM Lipid Reagent(Bio-Rad Laboratories, Hercules, CA, USA). Membrane suspen-sions from transfected HEK293 cells were obtained as describedpreviously (Burgueno et al. 2003, 2004).

cAMP determinations

The functionality of both the D2RYFP and D2Rmut

YFP and theirallosteric modulation on activity of the A2AR were assessed by

means of a dual luciferase reporter assay (Promega, Stockholm,Sweden). In brief, upon co-transfection of a plasmid containing theCRE fused to firefly luciferase, it is possible to indirectly detectvariations of cAMP, as the expression of the reporter leads toincreased levels of luminescence, which are directly proportional tonewly generated cAMP. Thus, cells were co-transfected withplasmids corresponding to the following constructs as follows:1 lg firefly luciferase-encoding experimental plasmid (pGL4-CRE-luc2p; Promega), 1 lg of D2R

YFP or D2RmutYFP expression vectors,

1 lg of A2ARCFP expression vector (only in the allosteric studies),

and 50 ng Renilla luciferase-encoding internal control plasmid(phRG-B; Promega). Approximately 36 h post-transfection, afterthe cells were treated for 4 h with forskolin (Sigma-Aldrich,St. Louis, MO, USA), quinpirole (Sigma-Aldrich), CGS21680(Tocris Bioscience, Ellisville, MI, USA), and/or ZM 241385 (TocrisBioscience), they were harvested with passive lysis buffer(Promega), and the luciferase activity of cell extracts was deter-mined in a FLUOStar Optima plate-reader (BMG Labtech, Durham,NC, USA) using a 30-nm bandwidth excitation filter at 535 nm.Firefly luciferase was measured as firefly luciferase luminescenceover a 15 s reaction period. The luciferase values were normalizedagainst Renilla luciferase luminescence values.

Immunoprecipitation, gel electrophoresis, and immunoblotting

For immunoprecipitation, membrane suspensions of transientlytransfected HEK293 cells were solubilized in ice-cold radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-HCl, pH7.4, 100 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate,0.2% sodium dodecyl sulfate, and 1 mM EDTA) for 30 min on ice.The solubilized preparation was then centrifuged for 30 min at13200 rpm at 4°C. The supernatant (1 mg/mL) was processed forimmunoprecipitation, each step of which was conducted withconstant rotation at 0–4°C. The supernatant was incubated overnightwith 1 lg of mouse anti-A2AR monoclonal antibody (clone 7F6-G5-A2; Millipore, Temecula, CA, USA). Then, a suspension of proteinA cross-linked to agarose beads (40 lL, Sigma-Aldrich) was added,and the mixture was incubated overnight. The beads were washedtwice with ice-cold RIPA buffer, twice with ice-cold RIPA bufferdiluted 1 : 10 in (Tris-buffered saline; 50 mM Tris-HCl pH 7.4,100 mM NaCl), and once with Tris-buffered saline and aspirated todryness with a 28-gage needle. Subsequently, 30 lL of sodiumdodecyl sulfate polyacrylamide gel electrophoresis (SDS–PAGE)sample buffer (8 M urea, 2% sodium dodecyl sulfate, 100 mMdithiothreitol, 375 mM Tris, pH 6.8) was added to each sample.Immune complexes were dissociated by heating to 37º C for 2 h andresolved by SDS–PAGE. Gel electrophoresis was performed using10% polyacrylamide gels. Proteins were transferred to PVDFmembranes using a semi-dry transfer system and immunoblottedwith either rabbit anti-A2AR polyclonal antibody (Ciruela et al.2004) or rabbit anti-D2R polyclonal antibody (Millipore), and thenhorseradish peroxidase-conjugated goat anti-rabbit IgG (1/60000;Thermo Fisher Scientific, Inc., Rockford, IL, USA). The immuno-reactive bands were developed using a chemiluminescent detectionkit (Pierce, Rockford, IL, USA) (Ciruela and McIlhinney 1997).

BRET assay

For BRET experiments, HEK293 cells, transiently transfected withconstant (1 lg) amount of plasmid encoding A2AR

Rluc and increasing

© 2012 The AuthorsJournal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384

Molecular determinants of A2AR–D2R allosterism 375

amounts (0.5–6 lg) of plasmids encoding for D2RYFP or D2RmutYFP,

were rapidly washed twice in phosphate-buffered saline, detached,and resuspended in the same buffer. Cell suspensions (20 lg protein)were distributed in duplicate into 96-well microplate black plates witha transparent bottom (Corning 3651; Corning, Stockholm, Sweden)for fluorescence measurement or white plates with a white bottom(Corning 3600) for BRET determination. For BRET1 measurement,h-coelenterazine substrate (Life Technologies Corp., Grand Island,NY, USA) was added at a final concentration of 5 lM, and readingswere performed 1 min after using the POLARstar Optima plate-reader (BMG Labtech) that allows the sequential integration of thesignals detected with two filter settings [485 nm (440–500 nm) and530 nm (510–560 nm)]. The BRET ratio was defined as previouslydescribed (Canals et al., 2003; Ciruela et al. 2004; Woods andFerre 2005; Woods et al. 2005).

SNAP labeling

Eventually, we identified cells to perform FRET experiments andalso quantified the expression of the D2R constructs by means of theSNAP-tag technology (see above). Briefly, cells transfected with theD2R

AGT and D2RmutAGT constructs were seeded into 18 mm

diameter glass coverslips or in 96-wells plates and then stainedwith 5 lM or 100 lM, respectively, of the SNAP-Surface 647ligand (New England BioLabs, Ipswich, MA, USA) in supple-mented Dulbecco’s modified Eagle’s medium during 30 min at 37°C, 5% CO2. Finally, cells were washed three times with phenol red-free Hank’s balanced salt solution containing 1 g/L glucose (HBSS:137 mM NaCl, 5.4 mM KCl, 0.3 mM Na2HPO4, 0.4 mM KH2PO4,4.2 mM NaHCO3, 1.3 mM CaCl2, 0.5 mM MgCl2, 0.6 mMMgSO4, 5.6 mM glucose, pH 7.4) and imaged at 688 nm uponexcitation at 647 nm with a Cy5 filter set (Zeiss, Oberkochen,Germany) or fluorescence measured at 680 nm upon excitation at640 nm in a POLARstar Optima plate-reader (BMG Labtech).

Microscopic FRET measurements

Binding of MRS5424 in cells expressing the A2ARCFP or PTHR1CFP

construct was determined by real-time single-cell FRET experi-ments. In brief, transiently transfected cells seeded into 18 mmdiameter glass coverslips were mounted in an Attofluor holder andplaced on an inverted Axio Observer microscope (Zeiss) equippedwith a 639 oil immersion objective and a dual-emission photometrysystem (TILL Photonics, Gräfelfing, Germany). Then, cells werecontinuously superfused with the fluorescent ligand dissolved inHBSS, applied with the aid of a focal drug application system(OCTAFLOWTM; ALA Scientific Instruments, Westbury, NY,USA). A Polychrome V (Till Photonics) was used as the lightsource and signals detected by avalanche photodiodes were digitizedusing a Digidata 1440A analog/digital converter (MolecularDevices, Sunnyvale, CA, USA). pCLAMP (Molecular Devices)and GraphPad Prism (GraphPad Software, La Jolla, CA, USA)softwares were used for data collection and analysis. FRET wasmeasured upon excitation at 430 ± 10 nm (beam splitter dichroiclong-pass 460 nm) and an illumination time set to 10 ms at 10 Hz.Then, the emission light intensities were determined at535 ± 15 nm (F535) and 480 ± 20 nm (F480) with a beam splitterdichroic long-pass of 505 nm. No corrections for spilloverbetween channels or direct Alexa532 excitation were made. Theincrease in FRET ratio (F535/F480) was fitted to the equation:

r(t) = A 9 (1 � e�t/τ), where τ is the time constant (s) and A is themagnitude of the FRET signal. When necessary for calculating τ,agonist-independent changes in FRET as a result of photobleachingwere subtracted. Cell images were taken with a Zeiss AxioCamMR3and processed with the Axiovision 4.8 software (Zeiss).

Statistics

The number of samples (n) in each experimental condition isindicated in the figure legends. When two experimental conditionswere compared, statistical analysis was performed using an unpairedt-test. Otherwise, statistical analysis was performed by one-wayanalysis of variance (ANOVA) followed by the Student–Newman–Keuls post hoc test. Statistical significance was set as p < 0.05.

Results

Role of the D2R IL3 Arg-rich domains on A2AR/D2R

oligomerization

It is well established that a Coulombic interaction betweenthe C-terminal tail of the A2AR and IL3 of the D2R isinvolved in the A2AR/D2R oligomerization (Ciruela et al.2004). Interestingly, although the role of the A2ARC-terminal tail in this phenomenon has been largely studied(Borroto-Escuela et al. 2010a,b,a,b), the impact of Arg-richdomains of the D2R IL3 on the oligomeric function has beenless explored. Thus, we aimed to investigate the role of tworegions rich in positive charges within the D2R IL3 on thenegative allosteric receptor–receptor interaction associatedwith the A2AR/D2R oligomer. To this end, and by means ofsite-directed mutagenesis, we generated a mutated D2Rconstruct (D2Rmut

YFP), where the arginine (Arg) residues inIL3 of the D2R were substituted by alanine (Ala) (Fig. 1a).First, we examined whether the Arg to Ala substitutionsaffected the functionality of the receptor, as it had to bechallenged in further experiments with the selective agonistquinpirole, and a loss of efficacy could lead to misinterpre-tation of the results. Therefore, by means of a luciferasereporter assay, we determined changes in CRE transcription,which is increased after the activation of the CREB.Specifically, we examined the ability of both D2R

YFP andD2Rmut

YFP to antagonize the forskolin-mediated activation ofAC through coupling to Gao/i, as previously described(Obadiah et al. 1999). Interestingly, both D2R constructsequally reduced the forskolin-mediated cAMP accumulationafter quinpirole (25 nM) challenge (Fig. 1b), suggesting thatthe Arg to Ala substitutions did not affect the receptorfunctionality.Next, we evaluated the ability of the D2Rmut

YFP tooligomerize with the A2AR. To this end, we examined thephysical interaction with the A2AR by means of a classicalbiochemical approach (i.e., co-immunoprecipitation experi-ments) and by a biophysical assay (i.e., BRET experiments).Thus, from extracts of HEK293 cells transiently co-trans-fected with A2AR

CFP plus D2RYFP or D2Rmut

YFP, the mouse

Journal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384© 2012 The Authors

376 V. Fernandez-Duenas et al.

anti-A2AR antibody co-immunoprecipitated a protein ofmolecular weight ~75 kDa corresponding to the D2R

YFP

(either the wild type or the mutant) (Fig. 2a, IP: anti-A2AR,upper panel, lanes 4 and 5). Importantly, this protein banddid not appear in singly transfected (A2AR

CFP, D2RYFP or

D2RmutYFP) cells immunoprecipitated with the same antibody

(Fig. 2a, IP: anti-A2AR, upper panel, lanes 1, 2, and 3).Finally, we confirmed the presence of the A2AR by means ofa rabbit polyclonal antibody (Fig. 2a, IP: anti-A2AR, lowerpanel, lanes 1, 4, and 5). Overall, these results suggested thatby mutating the Arg residues of IL3 of the D2R, itsinteraction with the A2AR was not qualitatively altered.Also, we analyzed the degree of the A2AR-D2R interaction

by means of BRET experiments. Thus, a BRET saturationcurve was constructed in HEK293 cells co-transfected with aconstant amount of an A2AR

Rluc construct, while increasingthe concentrations of plasmids containing D2R

YFP orD2Rmut

YFP. Interestingly, a positive BRET signal wasobtained for the transfer of energy between A2AR

Rluc andboth the D2R

YFP and the D2RmutYFP. The BRET signal, as

reflected in the BRET ratio, increased as a hyperbolicfunction of the concentration of the D2R

YFP fusion constructs(Fig. 2b). However, the pairing of A2AR

Rluc and D2RmutYFP

led to a substantial reduction in the maximal BRET signal(BRETmax) when compared with the A2AR

Rluc and D2RYFP

pair (45 ± 2 mBU and 81 ± 2 mBU, respectively) (Fig. 2b).Importantly, when cells singly expressing the A2AR

Rluc andD2R

YFP-expressing cells were mixed, no BRET signal wasobserved (Fig. 2c). Overall, the results obtained suggested

that while the Arg to Ala substitutions within the D2R IL3did not abolish A2AR/D2R oligomerization, the proximitybetween the two receptors was significantly reduced. Alter-natively, it could also be possible that the mutations wouldlead to a lower proportion of oligomers, thus the partialBRET signal detected would be the sum of a smaller numberof direct receptor–receptor interactions.

D2R-mediated allosteric modulation of A2AR agonist binding

Antagonistic A2AR-mediated allosteric modulation of D2Rfunction has been largely studied; however, the reversecondition, that is, D2R-mediated allosteric modulation of theA2AR, is still elusive. Consequently, we decided to inves-tigate both the possible allosteric modulation exerted by theD2R on A2AR function, when these receptors are formingpart of an oligomer, and also, the contribution of Arg-richdomains of IL3 of the D2R to this receptor–receptorallosterism. To this end, we implemented a real-timeFRET-based A2AR agonist binding procedure in cellsco-expressing both A2ARs and D2Rs. Accordingly, werecorded the FRET engaged between a novel fluorescentA2AR agonist synthesized in this study (MRS5424) and theA2AR

CFP construct (Fig. 3b). First, we checked that theagonist-receptor interaction was observed at the cell surfaceof the transfected cells, and that the binding was completelyblocked in the presence of an A2AR-specific antagonist,ZM241,385 (Fig. 3a). And thereafter, a specific FRET signalbetween MRS5424 and the A2AR

CFP was measured in areal-time mode (Fig. 3c). Thus, under these experimental

D2R

IL3

V V V

0

20

40

60

80

100

120

**

V

V

EK

V

V

V

E

N NL

RR

RR

R

RRR A A

AA

AAA

A

L

K

D2Rmut

D2RYFP

cAM

P a

ccum

ulat

ion

(% f

orsk

olin

stim

ulat

ed)

D2RmutYFP

IL3mut

Mutation

(a) (b)

Fig. 1 Characterization of the D2R construct. (a) Schematic repre-sentation of the D2Rmut construct. The mutations performed at the third

intracellular loop 3 (IL3) are indicated. The D2R structure was suitedusing the PDB 1JGJ of rhodopsin and prepared using PyMOL (PyMOLMolecular Graphics System, DeLano Scientific, San Carlos, CA,

USA). (b) Determination of cAMP accumulation by means of theluciferase reporter assay system. HEK293 cells transiently transfectedwith pGL4-CRE-luc2p/phRG-B plus D2R

YFP (red bars) or D2RmutYFP

(blue bars) were incubated either with forskolin (1 lM) plus Hank’sbalanced salt solution buffer (solid bars) or quinpirole (25 nM) (empty

bars). Light emission was normalized assigning the 100% of effect tothat obtained when incubating cells with forskolin. Data are expressedas the mean ± SEM of three independent experiments. (*) indicate

statistically significant differences (p < 0.05; Student’s t-test) whencomparing Hank’s balanced salt solution versus quinpirole treatmentin both D2R

YFP and the D2RmutYFP transfected cells.

© 2012 The AuthorsJournal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384

Molecular determinants of A2AR–D2R allosterism 377

conditions, we analyzed changes in the FRET signal exertedby a quinpirole challenge on cells expressing A2AR

CFP pluseither D2R or D2Rmut. It is important to mention here that toensure the presence of both A2AR and D2R within the sameassayed cell, we engineered a D2R

AGT construct (Fig. 4a),which allowed the subsequent visualization of the receptor atthe cell surface (Fig. 4b). Indeed, in Fig. 4b, it is possible tovisualize that the A2AR

CFP and the D2R-AGT fusionconstructs (D2R

AGT and D2RmutAGT) labeled with the SNAP

substrate are co-distributed at the plasma membrane of co-transfected cells, thus providing the ideal conditions to assessour real-time FRET-based A2AR agonist binding experi-ments.

First, we recorded FRET in cells singly transfected withthe A2AR

CFP. Thus, upon excitation at 430 nm (CFPexcitation), the signals recorded from single HEK293 cellsexpressing A2AR

CFP challenged with the fluorescent ago-nist were analyzed at emissions of 480 nm (Fig. 3b, lowerpanel, CFP) and 535 nm (Fig. 3b, lower panel, Alexa532),and the corresponding normalized ratio F535/F480 plotted(Fig. 3b, upper panel). After addition of MRS5424(2 lM), a symmetrical decrease in CFP emission and anincrease in Alexa532 emission indicated that the changewas indeed because of an increase in FRET betweenA2AR

CFP and MRS5424 as a consequence of the ligandbinding (i.e., increase in the ratio F535/F480) (Fig. 3b).Interestingly, control experiments with cells expressingPTHRCFP showed no FRET when MRS5424 was super-fused on these cells (data not shown). On the other hand,the selection of the MRS5424 concentration (2 lM) wasbased on previous results (Fernández-Dueñas, unpublisheddata), indicating that this concentration elicited the best-defined response to be subsequently modulated with theD2R agonist.Next, we analyzed the D2R-mediated allosteric modulation

of A2AR agonist binding in cells co-expressing the A2ARCFP

and D2RAGT constructs. Interestingly, the proteins were

shown to co-localize at the cell surface (Fig. 4b), and it couldalso be assessed that the plasma membrane expression of theD2R

AGT constructs were unaffected because of the mutations(Fig. 4c). Thus, upon identification of a doubly transfectedcell (as in Fig. 4b), we proceeded to analyze the binding ofMRS5424 at the A2AR

CFP in the absence or the presence of

Fig. 2 Role of the D2R third intracellular loop 3 Arg-rich domains onA2AR/D2R oligomerization. (a) Co-immunoprecipitation experiments.

Membrane suspensions of HEK293 cells transiently expressingA2AR

CFP (lane 1), D2RYFP (lane 2), D2Rmut

YFP (lane 3), A2ARCFP plus

D2RYFP (lane 4), A2AR

CFP plus D2RmutYFP (lane 5) were solubilized

and processed for immunoprecipitation using mouse anti-A2ARmonoclonal antibody (IP). Immunoprecipitates were analyzed bysodium dodecyl sulfate–polyacrylamide gel electrophoresis and im-

munoblotted using rabbit anti-D2R polyclonal antibody and rabbitaffinity-purified anti-A2AR polyclonal antibody (IB). The presented blotsare representative of three different experiments with similar qualita-tive results. (b) BRET experiments. BRET saturation curves for the

A2AR/D2R oligomers (A2ARRluc/D2R

YFP: filled red circles and A2ARRluc/

D2RmutYFP: filled blue circles) at increasing expression levels of the

YFP-tagged vectors. Plotted on the X-axis is the fluorescence value

obtained from the YFP, normalized with the luminescence value ofA2AR

Rluc 10 min after h-coelenterazine incubation. Results areexpressed as mean ± SEM (n = 8 in triplicate). (c) Comparison of

the BRETmax obtained in (b). Values represent percentages ofmaximal saturable BRET responses (BRETmax). Mean ± SEM.(n = 8 in triplicate). ***: A2AR

Rluc/D2RmutYFP group is significantly

different compared with the A2ARRluc/D2R

YFP group (p < 0.001); +++:

control group (A2ARRluc+D2Rmut

YFP) is significantly different comparedwith the A2AR

Rluc/D2RmutYFP and A2AR

Rluc/D2RYFP groups (p < 0.001).

Crude

1 2

0

0

20

40

***

+++

60

80

100

0 1 2 3 4 5 6 7

10

20BR

ET

rat

io (

mB

U)

BR

ET

max

(m

BU

)

30

40

50

60

70

80

3 4 5 1 2 3 4 5

IP: anti-A2AR

D2RYFP

A2ARCFP

A 2AR

Rluc /D

2RYFP

A 2AR

Rluc /D

2Rm

utYFP

A 2AR

Rluc +D

2RYFP

(YFP-YFP0)/Rluc (AU)

IB: anti-D2R

IB: anti-A2AR

(a)

(b)

(c)

Journal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384© 2012 The Authors

378 V. Fernandez-Duenas et al.

the D2R agonist quinpirole (100 lM). Importantly,MRS5424 binding was unaffected by quinpirole in cellstransfected only with the A2AR

CFP (data not shown). Thus,upon D2R

AGT co-expression and quinpirole challenge, asignificantly lower FRET signal was observed, indicatingthat D2R activation diminished the binding of the A2ARagonist (Fig. 5a, left panel). Noteworthy, in this Figure, theinverse of the decline of CFP fluorescence (1/FCFP) was usedas readout of the ligand binding event, because the diminu-tion of CFP fluorescence was exclusively as a result of FRETand also to avoid any contamination of the FRET determi-nations. Furthermore, we analyzed the kinetic data by fittingthe CFP decline to a monoexponential decay curve, and a

decrease was observed in the association rate upon additionof quinpirole (0.134 ± 0.025 s�1 with quinpirole vs. 0.028 ±0.005 s�1 with buffer). This indicated that the association ofthe fluorescent ligand was significantly affected by theconformational change of the D2R upon quinpirole activa-tion. Conversely, in cells co-expressing the A2AR

CFP andD2Rmut

AGT, the binding of the fluorescent A2AR ligand wasunaffected by quinpirole superfusion (Fig. 5a, right panel).In addition, when adjusting the CFP decline to a monoex-ponential decay curve, the kinetic data revealed that theassociation rate constant was not as affected by theD2Rmut

AGT (0.046 ± 0.007 s�1) as with the wild-typeD2R

AGT. Therefore, Arg mutations diminished the ability

+MRS5424+MRS5424 +MRS5424+MRS5424

–ZM241,385–ZM241,385 +ZM241,385+ZM241,385

MRS5424MRS5424

(APECAlexa532)

A2ARCFP

FRET

0.950 20 40 60 80

1.00

1.05

0.95

1.00

1.05

1.10HN

NHN

HN

NN

N

OHO

OO

H

HNO

-

HN OO

SO3–

CO2–

NH2

CH3CH2NH

SO3H

Ratio

Alexa532

CFP

F53

5/F

480

(nor

mal

ized

)F

/F0

Time (s)

(a)

(b) (c)

Fig. 3 Determination of A2AR agonist binding by real-time fluores-cence resonance energy transfer (FRET) in single living cells. (a)Specific MRS5424 binding to the A2AR

CFP. HEK293 cells transiently

transfected with A2ARCFP were superfused with MRS5424 (2 lM)

during 5 min, washed and observed on an inverted microscope without(upper image) or with (down image) ZM241,385 (1 lM), a specific

A2AR antagonist. Scale bar: 10 lm. (b) Schematic representation ofthe FRET experiment between the fluorescent A2AR agonist MRS5424(APECAlexa532) and the A2AR

CFP construct. The A2AR and CFP

structures were prepared using PyMOL and the PDBs 3EML and

1EMA, respectively. (c) Time-resolved changes in CFP and Alexa532fluorescence emission signals in single cells transfected with A2AR

CFP.The emission intensities of CFP (F480, blue trace), Alexa 532 (F535,

yellow trace), and the ratio F535/F480 (black trace) in response toMRS5424 were recorded simultaneously from single HEK293 cellsexpressing the A2AR

CFP. Shown are the changes induced by rapid

superfusion with 2 lM MRS5424. The increase of the ratio F535/F480

was fitted by a simple monoexponential curve giving a time constant(τ) in this experiment of 12.0 ± 0.3 s. Traces are representative of

eight separate experiments.

© 2012 The AuthorsJournal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384

Molecular determinants of A2AR–D2R allosterism 379

of the D2R to exert a complete allosteric modulation ofbinding at the A2AR. Overall, these results demonstrated thatthe IL3 Arg-rich domains of D2R played a major role in theallosteric effects of the D2R within the A2AR/D2R oligomer,as the negative allosteric modulation exerted by the D2R(74.3 ± 3.1%) disappeared (94.4 ± 4.3%; p < 0.05) follow-ing mutagenesis (Fig. 5b).

Functional consequences of the D2R allosteric

modulationIn view of the previous results, one would expect that D2R-mediated allosteric modulation of the A2AR would lead to

functional consequences on A2AR signaling. Therefore, weevaluated the effects of the D2R

AGT or the D2RmutAGT

challenge on A2AR-mediated AC stimulation by measuringcAMP accumulation in co-transfected cells. It is importantto mention here that in cells singly transfected withA2AR

CFP the selective A2AR agonist CGS21680 (50 nM)led to an increase of cAMP accumulation, and while theA2AR antagonist ZM241385 (1 lM) totally blocked thiseffect, the D2R agonist quinpirole (25 nM) did not produceany effect on A2AR-mediated cAMP accumulation (data notshown). Interestingly, in doubly A2AR

CFP-D2RAGT trans-

fected cells quinpirole partially and significantly (~60%;p < 0.01) diminished the effects of CGS21680 on cAMPaccumulation (Fig. 6a, red bars, CGS+Quin). On the otherhand, when cells expressing A2AR

CFP and D2RmutAGT were

challenged with quinpirole, a significant (~80%; p < 0.01)reduction of A2AR agonist-mediated cAMP accumulationwas observed (Fig. 6a, red bars, CGS+Quin). Curiously,when inhibition of D2R

AGT- and D2RmutAGT-mediated

cAMP accumulation was compared, it was shown to besignificantly (p < 0.05) different (Fig. 6a). Apparently,these contradictory results might be readily explained ifwe consider that the establishment of an A2AR/D2Rfunctional oligomer (i.e., displaying receptor–receptor allo-sterism) would allow direct A2AR–D2R cross talk, providinga balanced outcome of Gas- and Gai -mediated AC modu-lation (Fig. 6b). On the contrary, when no direct functionalA2AR-D2R cross talk would exist, a shift to more Gai -coupled D2Rs would happen, thus leading to an increasedAC inhibition outcome (Fig. 6b).Overall, the results of this study show that the Arg-rich

domains of the D2R IL3 played a key functional role inthe A2AR/D2R oligomer. Thus, when mutating these

A2ARCFP D2RAGT

A2ARCFP + D2RAGT

A2ARCFP + D2RmutAGT

+ D 2R

mut

AGT

Transfection

0

5000Flu

ores

cenc

e (A

U)

10 000

15 000

20 000

25 000

CFP SNAP Merge

CFP SNAP Merge

A 2ARCFP

+ D 2R

AGT

(a)

(b)

(c)

Fig. 4 Determination of A2AR agonist binding by real-time fluores-cence resonance energy transfer in single living cells. (a) Schematic

representation of the A2ARCFP and D2R

AGT constructs. The sche-matic alkyltransferase (AGT) (O6-alkylguanine-DNA alkyltransferase)diagram was prepared using PyMOL and PDB 1EH6. (b) Cell

surface localization of the A2ARCFP and D2R

AGT constructs in livingcells. HEK293 cells were transiently transfected with A2AR

CFP andeither D2R

AGT or D2RmutAGT, stained with the SNAP substrate and

visualized at 480 nm (CFP) and 688 nm (SNAP) upon excitation at430 nm and 647 nm, respectively, using an inverted microscope.Superposition of images (merge) revealed A2AR

CFP and D2RAGT co-

distribution at the cell surface. Scale bar: 10 lm. (c) Cell surface

quantification of the A2ARCFP and D2R

AGT constructs in living cells.HEK293 cells were transiently transfected with A2AR

CFP and eitherD2R

AGT or D2RmutAGT, stained with the SNAP substrate and plated

in wells, where fluorescence at 680 nm was measured uponexcitation at 640 nm using a POLARstar Optima plate-reader.Results are expressed as mean ± SEM (n = 3 in triplicate), after

the subtraction of the fluorescence measured in mock cells (AU,arbitrary units).

Journal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384© 2012 The Authors

380 V. Fernandez-Duenas et al.

residues it could be observed that although (i) there wereno changes on membrane density of D2R and (ii) no effecton D2R signaling (iii) it was affected but not abolished

A2AR-D2R heteromerization, and (iv) a selective abolitionof the allosteric interactions between A2AR and D2Roccurred.

Transfection: A2ARCFP

1/F

CF

P

0.975

0 20 40 60 0 20 40 60

0

20% F

RE

T s

igna

l (A

)

40

60

80

100

120

*

Time (s) Time (s)

1.000

1.025

1.050

1.075 MRS5424 MRS5424

+ D2RmutAGT + D2RAGT

+Quin +Quin

+ D 2R

mut

AGT

A 2ARCFP

+ D 2R

AGT

(a) (b)

Fig 5 Real-time fluorescence resonance energy transfer (FRET)determinations of the D2R-mediated modulation of A2AR agonistbinding. (a) Time-resolved changes in FRET signals (see Fig. 3) incells transfected with A2AR

CFP plus D2RAGT or A2AR

CFP plus

D2RmutAGT in the absence (black trace) or presence (red and blue

traces) of quinpirole (100 lM) were recorded. The FRET increase(1/F480) was fitted by a simple monoexponential curve and the

magnitude of the FRET signal (A, see Materials and methods)

calculated for each experimental condition. Traces are representativeof five separate experiments with similar qualitative and quantitativeresults. (b) The variation of the A value for each experimental conditionwas plotted. Data represent the average ± SEM values of five

independent experiments. Asterisk indicates data significantly differentfrom the control condition (i.e., in the absence of quinpirole): *p < 0.05by ANOVA with Student–Newman–Keuls multiple comparison post hoc

test.

0

CGS

CGS+

ZMCG

S+Q

uin

10

20*

30

40

50

cAM

P a

ccum

ulat

ion

(% f

orsk

olin

stim

ulat

ed) A2AR/D2R

ATP

AC AC

Gαs α α ααγ γ γ γβ β β βGαsGαi/o Gαi/o

cAMP ATP cAMP

A2AR/D2Rmut

(a) (b)

Fig. 6 Functional consequences of the D2R allosteric modulation. (a)Allosteric modulation of the A2AR

CFP activity. HEK293 cells transientlytransfected with pGL4-CRE-luc2p/phRG-B and A2AR

CFP plus D2RAGT

(red bars) or D2RmutAGT (blue bars) were incubated with forskolin

(1 lM), CGS21680 (50 nM), CGS + ZM241385 (50 nM + 1 lM,respectively), or CGS21680 + quinpirole (50 nM + 25 nM, respec-tively) and the cAMP accumulation determined by means of the

luciferase reporter assay system (see Materials and methods). Lightemission was normalized assigning the 100% of effect to that obtainedwhen incubating cells with forskolin. Data are expressed as the

mean ± SEM of three independent experiments. (*) indicates statis-tically significant differences (p < 0.05; Student’s t-test) when com-paring cells transfected with the D2R

AGT or the D2RmutAGT. (b)

Schematic representation of the oligomeric antagonistic A2AR–D2Rinteraction. In the left part, the native A2AR/D2R oligomer is shown, inwhich a balance between Gas- and Gai/o-coupling would exist, and thebidirectional negative allosteric receptor–receptor interaction may

outline the final adenylyl cyclase (AC) signaling efficacy. Indeed, atthe G-protein level the expected antagonistic interaction can beobserved as the A2AR protomer via Gas activates AC, while the D2R

protomer would partially lose its Gai/o-coupled AC inhibition. In the rightpart, the mutated A2AR/D2R oligomer is shown. This oligomer lacks thenegative allosteric receptor–receptor interaction and non-A2AR–D2R

trans-inhibition phenomena are established. Thus, in the absence ofthe powerful antagonistic A2AR–D2R interaction at the level of D2Rrecognition and Gai/o-coupling, an enhanced AC inhibition is observed.

© 2012 The AuthorsJournal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384

Molecular determinants of A2AR–D2R allosterism 381

Discussion

The molecular interactions between the A2AR and the D2Rwhen engaged in an oligomeric complex have been postu-lated to play a major role in the development and also thetreatment of neurodegenerative disorders, such as Parkin-son’s disease (for review see (Fuxe et al. 2010)). As a result,the elucidation of the mechanisms mediating such receptor–receptor interactions has become a main goal during recentyears. Consequently, as commented earlier, it was previouslyshown that a Coulombic interaction established between theIL3 of the D2R and the C-terminal tail of the A2AR plays apivotal role in the phenomenon of A2AR/D2R oligomeriza-tion (Canals et al., 2003; Ciruela et al. 2004; Woods andFerre 2005; Woods et al. 2005). Thus, in the present work,we aimed to examine the allosteric modulation that the D2Rexerts on the A2AR within the A2AR/D2R oligomer. Inaddition, we also aimed to demonstrate that this receptor–receptor interaction is mostly mediated by the two highlypositively charged regions located at the IL3 of the D2R, notonly at the level of the binding of A2AR agonists but alsowith respect to the functionality of the latter receptor.Hence, we performed site-directed mutagenesis to generate

a D2R mutant in which the positively charged Arg residueswere replaced by uncharged alanine residues. Interestingly,the mutation of the Arg residues did not affect the function-ality of the receptor, as it was able to inhibit forskolin-inducedcAMP accumulation at the same level as the wild-type D2R.Accordingly, it could be assumed that the results obtainedwhen examining the allosteric effects of the D2R agonist werenot because of an aberrant signaling pathway as a result ofthe mutations. Interestingly, by means of co-immunoprecip-itation experiments, we found that the Arg-rich domains ofIL3 of the D2R seemed non-essential for A2AR and D2Roligomerization as D2Rmut still co-immunoprecipitated withthe A2AR. However, the Arg?Ala substitutions within theD2R IL3 probably reduced the molecular proximity betweenthese two receptors as indicated in our BRET assays. Theseresults are consistent with those previously obtained con-cerning the role of negatively charged residues on the A2ARC-terminal tail, which in fact may interact electrostaticallywith IL3 of the D2R in A2AR/D2R oligomerization (Borroto-Escuela et al. 2010a,b). Although these Coulombic interac-tions disappeared, the complex was still formed, as theprotein–-protein interaction also depends on other receptorregions, for instance the TMs-IV and V of the D2R (Borroto-Escuela et al. 2010a,b). Needless to say, it could also bepossible for the receptor–receptor interaction to be whollyaffected and that we were observing a lower proportion ofoligomers formed. However, when assessing the expressionlevels of theD2R

AGT constructs, we did not find differences inthe presence of the wild type and the mutant receptor at thecell surface. Nevertheless, it could be concluded that themutations affected in some manner the present receptor–

receptor interaction, thus the possible effects on the presum-ably antagonistic allosteric interaction were subsequentlyexamined.Interestingly, regarding the allosteric modulation of the

D2R on A2AR binding, the generation of the mutant variantof the D2R permitted to demonstrate the relevance of thepositively charged residues located at the IL3 in suchreceptor–receptor interaction. Thus, upon co-transfectingboth the D2Rmut

AGT and the A2ARCFP, quinpirole treatment

did not affect FRET between the fluorescent A2AR agonistand A2AR

CFP, as compared with the wild-type D2R. Previousstudies have revealed that an intramembrane interactionoccurs between the A2AR and the D2R; thus, it was shownthat the A2AR inhibits the D2R-mediated neuronal excitabil-ity and neurotransmitter release and that the D2R negativelymodulates A2AR ligand binding to inhibit A2AR-mediatedprotein phosphorylation and gene expression (Canals et al.,2003; Ciruela et al. 2004; Woods and Ferre 2005; Woodset al. 2005). However, to our knowledge, this is the first timethe negative allosteric modulation of the D2R on A2ARagonist binding has been visualized in a real-time mode.Thus, we were able to determine that D2R activation partiallyinhibited and also slowed the binding association of thefluorescent A2AR agonist. Furthermore, the results obtainedclearly indicated that this important interaction was abolishedby mutating the IL3 of the D2R. In fact, we also evaluated theallosteric modulation of the D2R on A2AR agonist-mediatedcAMP accumulation, and it was shown that the D2R agonistquinpirole, by activating the D2Rmut, was able to completelycounteract A2AR agonist-mediated activation. Interestingly, ithas been previously shown that the antagonistic allostericA2AR/D2R interaction favors a rapid onset of b-arrestin-2D2R signaling through Gai/o-uncoupling (Canals et al., 2003;Ciruela et al. 2004; Trincavelli et al. 2012; Woods and Ferre2005; Woods et al. 2005). Thus, when the receptor–receptorallosteric interaction disappears, as in the absence of Arg-richdomains of the D2R IL3, the D2R would be able to mostlysignal through Gai/o-protein to inhibit cAMP accumulation.Consequently, it could be considered that in the absence ofA2AR–D2R allosteric modulation, a major outcome of Gai/o-signaling depending cascades occurs (Fig. 6b). Interestingly,this condition might be exacerbated by the fact that the levelsof Gai/o-protein may be higher than the Gas (Borroto-Escuelaet al. 2011), thus contributing to the enhanced inhibition ofAC activity. In this regard, it is currently thought that uponagonist stimulation and the subsequent receptor conforma-tional change, this information is transmitted into the IL3 (Liet al. 1996) Thus, it would seem likely that changing someresidues (Arg?Ala) of this domain would disrupt its normalfunction. Consequently, the lack of an allosteric modulationof the D2Rmut could also be explained, not only because areduced direct receptor–receptor interaction within the A2AR/D2R oligomer but also by means of other mechanismslocated at the intracellular signal transmission pathways

Journal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384© 2012 The Authors

382 V. Fernandez-Duenas et al.

(Kull et al. 1999; Zurn et al. 2009). However, our resultspoint to a scenario in which the mutation of the Arg residuesat the IL3 reduced the physical interaction between the D2Rand the A2AR, and therefore partially precluded the D2R-mediated allosteric modulation of the A2AR.In conclusion, the findings of this study provide further

evidence of the existence of antagonistic A2AR-D2R interac-tions. In particular, the Arg residues (217–222 and 267–269)on IL3 of the D2R have been identified to play a major keyrole in such receptor–receptor allosteric interaction. Thus, itwould be feasible to target this region in order to allostericallymodulate the functionality of the A2AR/D2R oligomer.

Acknowledgements

This work was supported by grants SAF2011-24779 and Consol-ider-Ingenio CSD2008-00005 from Ministerio de Cienciae Innovación and ICREA Academia-2010 from the CatalanInstitution for Research and Advanced Studies to FC. Also, VF-D, MG-S, and FC belong to the “Neuropharmacology and Pain”accredited research group (Generalitat de Catalunya, 2009 SGR232). Support to KAJ and TSK from the NIDDK IntramuralResearch Program of the National Institutes of Health, Bethesda,MD, USA is acknowledged. We also thank Esther Castaño, EvaJulià, and Benjamín Torrejón, from the Scientific and TechnicalServices (SCT)-Bellvitge Campus of the University of Barcelonafor the technical assistance.

Conflict of interest

The authors declare no conflict of interest.

References

Agnati L. F., Fuxe K., Torvinen M., Genedani S., Franco R., Watson S.,Nussdorfer G. G., Leo G. and Guidolin D. (2005) New methods toevaluate colocalization of fluorophores in immunocytochemicalpreparations as exemplified by a study on A2A and D2 receptorsinChinese hamster ovary cells. J.Histochem.Cytochem. 53, 941–953.

Azdad K., Gall D., Woods A. S., Ledent C., Ferre S. and Schiffmann S.N. (2009) Dopamine D2 and adenosine A2A receptors regulateNMDA-mediated excitation in accumbens neurons through A2A–D2 receptor heteromerization. Neuropsychopharmacology 34,972–986.

Bornstein A. M. and Daw N. D. (2011) Multiplicity of control in thebasal ganglia: computational roles of striatal subregions. Curr.Opin. Neurobiol. 21, 374–380.

Borroto-Escuela D. O., Marcellino D., Narvaez M., Flajolet M., HeintzN., Agnati L., Ciruela F. and Fuxe K. (2010a) A serine pointmutation in the adenosine A2AR C-terminal tail reduces receptorheteromerization and allosteric modulation of the dopamine D2R.Biochem. Biophys. Res. Commun. 394, 222–227.

Borroto-Escuela D. O., Romero-Fernandez W., Tarakanov A. O., et al.(2010b) Characterization of the A2AR-D2R interface: focus on therole of the C-terminal tail and the transmembrane helices. Biochem.Biophys. Res. Commun. 402, 801–807.

Borroto-Escuela D. O., Romero-Fernandez W., Tarakanov A. O., CiruelaF., Agnati L. F. and Fuxe K. (2011) On the existence of a possible

A2A–D2-beta-Arrestin2 complex: A2A agonist modulation ofD2 agonist-induced beta-arrestin2 recruitment. J. Mol. Biol. 406,687–699.

Brand F., Klutz A. M., Jacobson K. A., Fredholm B. B. and Schulte G.(2008) Adenosine A(2A) receptor dynamics studied with thenovel fluorescent agonist Alexa488-APEC. Eur. J. Pharmacol.590, 36–42.

Burgueno J., Enrich C., Canela E. I., Mallol J., Lluis C., Franco R. andCiruela F. (2003) Metabotropic glutamate type 1alpha receptorlocalizes in low-density caveolin-rich plasma membrane fractions.J. Neurochem. 86, 785–791.

Burgueno J., Canela E. I., Mallol J., Lluis C., Franco R. and Ciruela F.(2004) Mutual regulation between metabotropic glutamate type1alpha receptor and caveolin proteins: from traffick to constitutiveactivity. Exp. Cell Res. 300, 23–34.

Cabello N., Gandia J., Bertarelli D. C., Watanabe M., Lluis C., FrancoR., Ferre S., Lujan R. and Ciruela F. (2009) Metabotropicglutamate type 5, dopamine D(2) and adenosine A(2a) receptorsform higher-order oligomers in living cells. J. Neurochem. 109,1497–1507.

Canals M., Marcellino D., Fanelli F., et al. (2003) Adenosine A2A-dopamine D2 receptor-receptor heteromerization: qualitative andquantitative assessment by fluorescence and bioluminescenceenergy transfer. J. Biol. Chem. 278, 46741–46749.

Ciruela F. and McIlhinney R. A. (1997) Differential internalisation ofmGluR1 splice variants in response to agonist and phorbol esters inpermanently transfected BHK cells. FEBS Lett. 418, 83–86.

Ciruela F., Burgueno J., Casado V., et al. (2004) Combining massspectrometry and pull-down techniques for the study of receptorheteromerization. Direct epitope-epitope electrostatic interactionsbetween adenosine A2A and dopamine D2 receptors. Anal. Chem.76, 5354–5363.

Ciruela F., Vilardaga J. P. and Fernandez-Duenas V. (2010) Lighting upmultiprotein complexes: lessons from GPCR oligomerization.Trends Biotechnol. 28, 407–415.

Ferre S., Agnati L. F., Ciruela F., Lluis C., Woods A. S., Fuxe K. andFranco R. (2007a) Neurotransmitter receptor heteromers and theirintegrative role in ‘local modules’: the striatal spine module. BrainRes. Rev. 55, 55–67.

Ferre S., Ciruela F., Quiroz C., et al. (2007b) Adenosine receptorheteromers and their integrative role in striatal function.ScientificWorldJournal 7, 74–85.

Fuxe K., Ferre S., Zoli M. and Agnati L. F. (1998) Integrated events incentral dopamine transmission as analyzed at multiple levels.Evidence for intramembrane adenosine A2A/dopamine D2 andadenosine A1/dopamine D1 receptor interactions in the basalganglia. Brain Res. Brain Res. Rev. 26, 258–273.

Fuxe K., Agnati L. F., Jacobsen K., et al. (2003) Receptorheteromerization in adenosine A2A receptor signaling: relevancefor striatal function and Parkinson’s disease.Neurology 61, S19–23.

Fuxe K., Marcellino D., Borroto-Escuela D. O., Guescini M., Fernandez-Duenas V., Tanganelli S., Rivera A., Ciruela F. and Agnati L. F.(2010) Adenosine-dopamine interactions in the pathophysiologyand treatment of CNS disorders. CNS Neurosci. Ther. 16, e18–42.

Genedani S., Guidolin D., Leo G., Filaferro M., Torvinen M., Woods A.S., Fuxe K., Ferre S. and Agnati L. F. (2005) Computer-assistedimage analysis of caveolin-1 involvement in the internalizationprocess of adenosine A2A-dopamine D2 receptor heterodimers.J. Mol. Neurosci. 26, 177–184.

Kamiya T., Saitoh O., Yoshioka K. and Nakata H. (2003)Oligomerization of adenosine A2A and dopamine D2 receptorsin living cells. Biochem. Biophys. Res. Commun. 306, 544–549.

Kull B., Ferre S., Arslan G., Svenningsson P., Fuxe K., Owman C. andFredholm B. B. (1999) Reciprocal interactions between adenosine

© 2012 The AuthorsJournal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384

Molecular determinants of A2AR–D2R allosterism 383

A2A and dopamine D2 receptors in Chinese hamster ovary cellsco-transfected with the two receptors. Biochem. Pharmacol. 58,1035–1045.

Li Y., Mende U., Lewis C. and Neer E. J. (1996) Maintenance of cellularlevels of G-proteins: different efficiencies of alpha s and alpha osynthesis in GH3 cells. Biochem. J. 318(Pt 3), 1071–1077.

Navarro G., Ferre S., Cordomi A., et al. (2010) Interactions betweenintracellular domains as key determinants of the quaternarystructure and function of receptor heteromers. J. Biol. Chem.285, 27346–27359.

Obadiah J., Avidor-Reiss T., Fishburn C. S., Carmon S., BayewitchM., Vogel Z., Fuchs S. and Levavi-Sivan B. (1999) Adenylylcyclase interaction with the D2 dopamine receptor family;differential coupling to Gi, Gz, and Gs. Cell. Mol. Neurobiol.19, 653–664.

Shindou T., Nonaka H., Richardson P. J., Mori A., Kase H. and IchimuraM. (2002) Presynaptic adenosine A2A receptors enhanceGABAergic synaptic transmission via a cyclic AMP dependentmechanism in the rat globus pallidus. Br. J. Pharmacol. 136,296–302.

Svenningsson P., Lindskog M., Ledent C., Parmentier M., Greengard P.,Fredholm B. B. and Fisone G. (2000) Regulation of thephosphorylation of the dopamine- and cAMP-regulatedphosphoprotein of 32 kDa in vivo by dopamine D1, dopamineD2, and adenosine A2A receptors. Proc. Natl Acad. Sci. USA 97,1856–1860.

Trincavelli M. L., Daniele S., Orlandini E., et al. (2012) A new D(2)dopamine receptor agonist allosterically modulates A(2A)adenosine receptor signalling by interacting with the A(2A)/D(2)receptor heteromer. Cell. Signal. 24, 951–960.

Woods A. S. and Ferre S. (2005) Amazing stability of the arginine-phosphate electrostatic interaction. J. Proteome Res. 4, 1397–1402.

Woods A. S., Ciruela F., Fuxe K., Agnati L. F., Lluis C., Franco R. andFerre S. (2005) Role of electrostatic interaction in receptor-receptorheteromerization. J. Mol. Neurosci. 26, 125–132.

Zurn A., Zabel U., Vilardaga J. P., Schindelin H., Lohse M. J. andHoffmann C. (2009) Fluorescence resonance energy transferanalysis of alpha 2a-adrenergic receptor activation reveals distinctagonist-specific conformational changes. Mol. Pharmacol. 75,534–541.

Journal of Neurochemistry © 2012 International Society for Neurochemistry, J. Neurochem. (2012) 123, 373--384© 2012 The Authors

384 V. Fernandez-Duenas et al.