Adenosine A2A and A3 Receptors Are Able to Interact with Each Other. A Further Piece in the Puzzle of Adenosine Receptor-Mediated Signaling

The aim of this paper was to check the possible interaction of two of the four purinergic P1 receptors, the A2A and the A3. Discovery of the A2A–A3 receptor complex was achieved by means of immunocytochemistry and of bioluminescence resonance energy transfer. The functional properties and heteromer print identification were addressed by combining binding and signaling assays. The physiological role of the novel heteromer is to provide a differential signaling depending on the pre-coupling to signal transduction components and/or on the concentration of the endogenous agonist. The main feature was that the heteromeric context led to a marked decrease of the signaling originating at A3 receptors. Interestingly from a therapeutic point of view, A2A receptor antagonists overrode the blockade, thus allowing A3 receptor-mediated signaling. The A2A–A3 receptor heteromer print was detected in primary cortical neurons. These and previous results suggest that all four adenosine receptors may interact with each other. Therefore, each adenosine receptor could form heteromers with distinct properties, expanding the signaling outputs derived from the binding of adenosine to its cognate receptors.


Introduction
The laboratories of Susan George and Lakshmi Devi were the first to demonstrate that two G-protein-coupled receptors (GPCRs), for the same neurotransmitter, may form functional heterodimers. In fact, receptor heteromers composed of different opioid receptor types were reported circa 20 years ago [1][2][3][4]. Similarly, dopamine D 1 -D 2 receptor heteromers were described with a particular feature, Adenosine receptors are divided into high affinity A 1 R and A 2A R types, and low affinity A 2B R and A 3 R types [23,24]. Members of this family interact to form heteromeric complexes, however, this possibility has not yet explored for the A 2A R and A 3 R types. To first address a possible interaction between these two receptors, we performed immunocytochemical assays to look for colocalization. We used a heterologous system consisting of human embryonic kidney HEK-293T cells, which were transfected with A 3 R fused to yellow fluorescent protein (YFP), A 2A R fused to Rluc, or both. The results in Figure 1A show a marked colocalization of the signal of the two receptors; the label was mainly in the cell surface (notice that the planes of the confocal images are taken near the surface of the slide) and did not substantially change by previous treatment with agonists. Because of the fact that this technique cannot demonstrate a direct interaction, we transfected HEK-293T cells with a constant amount of the cDNA for A 2A R-Rluc and increasing quantities of the cDNA for A 3 R-YFP, then bioluminescence resonance energy transfer (BRET) was determined. Interestingly, a saturation BRET curve indicating a specific interaction between A 2A R and A 3 R was obtained, as shown in Figure 1B. The calculated parameters were BRET max = 51 ± 4 mBU and BRET 50 = 12 ± 3. As negative control, the same assay was developed with a constant amount of A 2A R-Rluc and increasing quantities of D 4 R-YFP, obtaining a nonspecific linear signal that indicates a lack of interaction between these receptors ( Figure 1B). In summary, co-expression of the two receptors in a heterologous system leads to A 2A R-A 3 R heteromer (A 2A R/A 3 R-Het) formation. Taking into account previous knowledge, receptor heteromers are formed in the endoplasmic reticulum and traffic to the cell surface, where they complex with heteromeric G proteins to acquire further stabilization. Indeed, the final structure is significantly modified by agonist binding [16,17,19,21].
Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 3 of 14 that this technique cannot demonstrate a direct interaction, we transfected HEK-293T cells with a constant amount of the cDNA for A2AR-Rluc and increasing quantities of the cDNA for A3R-YFP, then bioluminescence resonance energy transfer (BRET) was determined. Interestingly, a saturation BRET curve indicating a specific interaction between A2AR and A3R was obtained, as shown in Figure  1B. The calculated parameters were BRETmax = 51 ± 4 mBU and BRET50 = 12 ± 3. As negative control, the same assay was developed with a constant amount of A2AR-Rluc and increasing quantities of D4R-YFP, obtaining a nonspecific linear signal that indicates a lack of interaction between these receptors ( Figure 1B). In summary, co-expression of the two receptors in a heterologous system leads to A2AR-A3R heteromer (A2AR/A3R-Het) formation. Taking into account previous knowledge, receptor heteromers are formed in the endoplasmic reticulum and traffic to the cell surface, where they complex with heteromeric G proteins to acquire further stabilization. Indeed, the final structure is significantly modified by agonist binding [16,17,19,21]. Immunocytochemical assays were performed in HEK-293T cells expressing A2AR-Rluc (0.5 µg cDNA), which was detected by a mouse anti-Rluc antibody and a secondary anti-mouse Cy3 conjugated antibody (red), and/or A3R-YFP (0.5 µg cDNA), which was detected by its own fluorescence (green

Effect of the A3R Expression on Homogeneous Ligand Binding to the A2AR
A2AR and A3R are GPCRs that, upon activation by the endogenous agonist, adenosine, initiate diverse signaling pathways. Thus, the analysis of ligand binding to receptors in a heteromeric context becomes greatly relevant. Accordingly, competition experiments were performed using a homogenous assay as described in Methods. HEK-293T cells expressing HALO-A2AR labeled with Lumi4-Tb were incubated with 20 nM of a fluorophore-conjugated selective A2AR antagonist (red Immunocytochemical assays were performed in HEK-293T cells expressing A 2A R-Rluc (0.5 µg cDNA), which was detected by a mouse anti-Rluc antibody and a secondary anti-mouse Cy3 conjugated antibody (red), and/or A 3 R-YFP (0.5 µg cDNA), which was detected by its own fluorescence (green). Colocalization is shown in yellow. Cells were previously treated with 100 nM CGS 21680, 100 nM IB-MECA, or vehicle. Cell nuclei were stained with Hoechst (blue). Scale bar: 20 µm. (B) A 2A R and A 3 R interact in transfected HEK-293T cells. BRET assays were performed in HEK-293T cells transfected with a constant amount of cDNA for A 2A R-Rluc (0.1 µg) and increasing concentrations of cDNA for A 3 R-YFP (0.1 to 1 µg) or D 4 R-YFP (0.1 to 1 µg) for negative control. Values are the mean ± S.E.M. (n = 6 in triplicates). mBU: milliBRET units.

Effect of the A 3 R Expression on Homogeneous Ligand Binding to the A 2A R
A 2A R and A 3 R are GPCRs that, upon activation by the endogenous agonist, adenosine, initiate diverse signaling pathways. Thus, the analysis of ligand binding to receptors in a heteromeric context becomes greatly relevant. Accordingly, competition experiments were performed using a homogenous assay as described in Methods. HEK-293T cells expressing HALO-A 2A R labeled with Lumi4-Tb were incubated with 20 nM of a fluorophore-conjugated selective A 2A R antagonist (red SCH 442416) and increasing concentrations (0-10 µM) of the A 2A R agonist, CGS 21680. As observed in Figure 2A, CGS 4 of 13 21680 competed the binding of red SCH 442416 in a monophasic fashion and with a K i value in the nanomolar range (35 ± 1 nM), which matches with values obtained using radiolabeled ligands. At present, there is no selective fluorophore-conjugated agonist to perform homogenous assays of binding to HALO-A 3 R expressing cells. Therefore, we measured the competition of the binding of red SCH 442416 to cells co-expressing HALO-A 2A R and A 3 R ( Figure 2B). The K i value for CGS 21680, used as competitor, was one order of magnitude higher (427 ± 4 nM), thus suggesting that, when A 2A R forms complexes with A 3 R, the structure of the orthosteric site of the A 2A R is modified; it cannot be ruled out that this apparent increase in affinity is in part owing to CGS 21680 binding to the A 3 R.
SCH 442416) and increasing concentrations (0-10 µM) of the A2AR agonist, CGS 21680. As observed in Figure 2A, CGS 21680 competed the binding of red SCH 442416 in a monophasic fashion and with a Ki value in the nanomolar range (35 ± 1 nM), which matches with values obtained using radiolabeled ligands. At present, there is no selective fluorophore-conjugated agonist to perform homogenous assays of binding to HALO-A3R expressing cells. Therefore, we measured the competition of the binding of red SCH 442416 to cells co-expressing HALO-A2AR and A3R ( Figure  2B). The Ki value for CGS 21680, used as competitor, was one order of magnitude higher (427 ± 4 nM), thus suggesting that, when A2AR forms complexes with A3R, the structure of the orthosteric site of the A2AR is modified; it cannot be ruled out that this apparent increase in affinity is in part owing to CGS 21680 binding to the A3R.

Functional Characterization of A2AR/A3R-Hets in HEK-293T Cells
After detecting the existence of A2AR/A3R-Hets in co-transfected HEK-293T cells, we questioned the functional implication of this newly discovered protein-protein interaction. Thanks to A2AR coupling to Gs protein, agonists such as CGS 21680 lead to adenylate cyclase activation and to increased cytosolic cAMP levels. In contrast, A3R couples to Gi and its activation leads to decreased cytosolic cAMP concentrations. Taking into account these facts, confirmed in cells transfected with the plasmid for one of the two receptors, we measured cAMP levels in HEK-293T cells transfected with cDNAs for A2AR (0.3 µg) and for A3R (0.4 µg), and treated with selective agonists. While CGS 21680 stimulation induced a marked rise in cAMP concentration, IB-MECA had no significant effect over forskolin-induced increase in cAMP levels ( Figure 3A). This result indicates that A3R-Gi coupling is blocked in the heteromeric context, which can be considered a print to detect A2AR/A3R-

Functional Characterization of A 2A R/A 3 R-Hets in HEK-293T Cells
After detecting the existence of A 2A R/A 3 R-Hets in co-transfected HEK-293T cells, we questioned the functional implication of this newly discovered protein-protein interaction. Thanks to A 2A R coupling to G s protein, agonists such as CGS 21680 lead to adenylate cyclase activation and to increased cytosolic cAMP levels. In contrast, A 3 R couples to G i and its activation leads to decreased cytosolic cAMP concentrations. Taking into account these facts, confirmed in cells transfected with the plasmid for one of the two receptors, we measured cAMP levels in HEK-293T cells transfected with cDNAs for A 2A R (0.3 µg) and for A 3 R (0.4 µg), and treated with selective agonists. While CGS 21680 stimulation induced a marked rise in cAMP concentration, IB-MECA had no significant effect over forskolin-induced increase in cAMP levels ( Figure 3A). This result indicates that A 3 R-G i coupling is blocked in the heteromeric context, which can be considered a print to detect A 2A R/A 3 R-Hets in native tissues/cells. When cells were simultaneously treated with the two agonists, the effect was similar to that obtained upon A 2A R activation, thus reinforcing the hypothesis that A 2A R stimulation blocks A 3 R induced signaling. In the case of cells pretreated with the selective antagonists (SCH 442416 for the A 2A R or PSB-10 for the A 3 R), we found that SCH 442416 blocked CGS 21680-induced effects, while it potentiated the G i -mediated effect elicited by IB-MECA. These results suggest that antagonist binding to A 2A R leads to a structural reorganization in the A 2A R/A 3 R-Hets that blunts the A 2A R-mediated blockade of A 3 R-G i coupling ( Figure 3A). For its part, the A 3 R antagonist (PSB-10) had no effect on A 2A R activation; no cross-antagonism of A 3 R over A 2A R was detected. antagonist (PSB-10) had no effect on A2AR activation; no cross-antagonism of A3R over A2AR was detected. In cAMP (in the absence of forskolin), ERK1/2 phosphorylation, and β-arrestin 2 recruitment assays, one-way analysis of variance (ANOVA) followed by Bonferroni's multiple comparison post hoc test were used for significance analysis. * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001 versus vehicle treatment (basal); # p < 0.05 significance versus agonist treatment; &&&& p < 0.0001 versus forskolin treatment.
Next, β-arrestin 2 recruitment was analyzed in cells expressing A2AR/A3R-Hets and the findings were similar to those obtained in cAMP determination assays obtained in the absence of forskolin ( Figure 3C). In contrast, the results of ERK1/2 phosphorylation showed significant responses produced by either the A2AR or the A3R agonist ( Figure 3B). Interestingly, when the receptors were simultaneously exposed to the two agonists, ERK1/2 phosphorylation was milder than in individual receptor engagement. This phenomenon may be considered as negative crosstalk. On the one hand, when cells were pretreated with the selective antagonist for the A2AR (SCH 442416), a complete blockade of CGS 21680-induced MAPK activation was observed, while the antagonist was ineffective on A3R activation. On the other hand, pretreatment with the A3R antagonist (PSB-10) induced a partial cross-antagonism on A2AR-induced ERK1/2 phosphorylation. Dynamic mass redistribution (DMR) is a label-free technique widely used in drug discovery, especially in the field of GPCRs, that serves to analyze cell responses in the absence of any exogenous reagent (apart from receptor ligands). The DMR equipment detects changes upon time of the wavelength of light reflected by cells; picometer shifts in the wavelength of photons occur when a GPCR is activated on the cell surface [25]. DMR responses showed that the signal due to A3R activation was blocked when the A2AR was . cAMP accumulation (A) was determined by HTRF as described in Methods. When indicated, cells were subsequently treated for 15 min with 0.5 µM forskolin. ERK1/2 phosphorylation (B) was analyzed using an AlphaScreen ® SureFire ® kit (Perkin Elmer) and β-arrestin 2 recruitment (C) was determined by BRET. ERK1/2 phosphorylation and β-arrestin 2 recruitment data are expressed as increases in percentage over basal, whereas cAMP data are expressed as percentage with respect to values obtained with forskolin. Dynamic mass redistribution (DMR) tracings (D) represent the picometer-shifts of reflected light wavelength over time. All data are the mean ± S.E.M. of eight different experiments performed in triplicates. In cAMP (in the absence of forskolin), ERK1/2 phosphorylation, and β-arrestin 2 recruitment assays, one-way analysis of variance (ANOVA) followed by Bonferroni's multiple comparison post hoc test were used for significance analysis. ** p < 0.01, *** p < 0.001, and **** p < 0.0001 versus vehicle treatment (basal); # p < 0.05 significance versus agonist treatment; &&&& p < 0.0001 versus forskolin treatment.
Next, β-arrestin 2 recruitment was analyzed in cells expressing A 2A R/A 3 R-Hets and the findings were similar to those obtained in cAMP determination assays obtained in the absence of forskolin ( Figure 3C). In contrast, the results of ERK1/2 phosphorylation showed significant responses produced by either the A 2A R or the A 3 R agonist ( Figure 3B). Interestingly, when the receptors were simultaneously exposed to the two agonists, ERK1/2 phosphorylation was milder than in individual receptor engagement. This phenomenon may be considered as negative crosstalk. On the one hand, when cells were pretreated with the selective antagonist for the A 2A R (SCH 442416), a complete blockade of CGS 21680-induced MAPK activation was observed, while the antagonist was ineffective on A 3 R activation. On the other hand, pretreatment with the A 3 R antagonist (PSB-10) induced a partial cross-antagonism on A 2A R-induced ERK1/2 phosphorylation. Dynamic mass redistribution (DMR) is a label-free technique widely used in drug discovery, especially in the field of GPCRs, that serves to analyze cell responses in the absence of any exogenous reagent (apart from receptor ligands). The DMR equipment detects changes upon time of the wavelength of light reflected by cells; picometer shifts in the wavelength of photons occur when a GPCR is activated on the cell surface [25]. DMR responses showed that the signal due to A 3 R activation was blocked when the A 2A R was co-expressed. Furthermore, a cross-antagonism was detected, that is, the A 2A R-induced signal was blocked by pretreatment with either A 2A R or A 3 R antagonists ( Figure 3D).

Discovery of A 2A R/A 3 R-Hets in Primary Cultures of Cortical Neurons
We moved to a more physiological environment to check whether the A 2A R/A 3 R-Hets may be expressed in a natural source. It is known that the two adenosine receptors are expressed in different areas of the central nervous system. Owing to the implication of the adenosine receptor in neuromodulation, we addressed the possible expression of A 2A R/A 3 R-Hets in cortical neurons by detecting the heteromer print.
Primary cultures of cortical neurons were prepared and cAMP determination and ERK1/2 phosphorylation assays were performed. The results in Figure 4A show that the selective A 3 R antagonist (PSB-10) did not counteract the effect of the selective A 2A R agonist (CGS 21680); it is one of the features detected in HEK-293T cells. The release of the brake on A 3 R-mediated signaling by selective A 2A R antagonists ( Figure 4B) and the cross antagonism in the link to the MAPK signaling pathway ( Figure 4C) were also detected. In summary, these data constitute strong evidence of A 2A R/A 3 R-Hets expression in primary cultures of cortical neurons. co-expressed. Furthermore, a cross-antagonism was detected, that is, the A2AR-induced signal was blocked by pretreatment with either A2AR or A3R antagonists ( Figure 3D).

Discovery of A2AR/A3R-Hets in Primary Cultures of Cortical Neurons
We moved to a more physiological environment to check whether the A2AR/A3R-Hets may be expressed in a natural source. It is known that the two adenosine receptors are expressed in different areas of the central nervous system. Owing to the implication of the adenosine receptor in neuromodulation, we addressed the possible expression of A2AR/A3R-Hets in cortical neurons by detecting the heteromer print.
Primary cultures of cortical neurons were prepared and cAMP determination and ERK1/2 phosphorylation assays were performed. The results in Figure 4A show that the selective A3R antagonist (PSB-10) did not counteract the effect of the selective A2AR agonist (CGS 21680); it is one of the features detected in HEK-293T cells. The release of the brake on A3R-mediated signaling by selective A2AR antagonists ( Figure 4B) and the cross antagonism in the link to the MAPK signaling pathway ( Figure 4C) were also detected. In summary, these data constitute strong evidence of A2AR/A3R-Hets expression in primary cultures of cortical neurons.

Discussion
This paper discovers a new complex formed by two different adenosine receptors that may be expressed in a heterologous system, but also in primary cultures of cortical neurons. The heteromer print is quite unique as antagonists of the A2AR enhance A3R-mediated signaling. One-way ANOVA followed by Bonferroni's multiple comparison post hoc test were used for significance analysis. * p < 0.05, ** p < 0.01, and **** p < 0.0001 versus vehicle treatment (basal); # p < 0.05, ### p < 0.001 versus agonist treatment. &&&& p < 0.0001 versus forskolin treatment.

Discussion
This paper discovers a new complex formed by two different adenosine receptors that may be expressed in a heterologous system, but also in primary cultures of cortical neurons. The heteromer print is quite unique as antagonists of the A 2A R enhance A 3 R-mediated signaling.
Especially unexpected was the discovery of heteromers formed by two receptors of the same neurotransmitter/hormone that are in complex with different cognate (heterotrimeric) G proteins. One possibility is a shift in the G-protein-coupling. In fact, the dopamine D 1 -D 2 heteroreceptor does not couple to G s nor to G i , but to G q [5][6][7][8]. However, in the case of the A 1 R-A 2A R heteromer, there is no shift in G-protein coupling, but the overall structure allows increasing or decreasing cAMP levels depending on the concentration of the endogenous agonist. Adenosine preferentially activates the A 1 R, thus engaging G i proteins. Nevertheless, when adenosine level increases and the A 2A R is activated within the heteromer, G i -coupling is blunted and G s engagement occurs. The molecular determinants that make this possible are detailed elsewhere [16,17], but it is worth mentioning that the C-terminal domain of the A 2A R plays a fundamental role.
The A 2B R is the most enigmatic adenosine receptor; it has a very reduced affinity for the nucleoside and it is also the receptor for netrin, which belongs to a family of proteins involved in axon guidance [26]. On the one hand, comparison of the structural arrangement of A 2A R and A 2B R has led to the discovery that the second extracellular loop determines low (A 2B R) or high (A 2A R) affinity for adenosine [27]. On the other hand, as earlier commented, the expression of the A 2B R reduces the A 2A R-mediated functions by means of the formation of A 2A -A 2B heteroreceptor complexes. Thus, it seems that the A 2B R, per se or negatively acting on the A 2A R, contributes to reducing the actions derived from extracellular adenosine accumulation.
The properties of the heteromer described here are different from those previously defined for adenosine receptor heterocomplexes. A simple scheme summarizing the operation of A 2A R/A 3 R-Hets is provided in Figure 5. One of the features of A 2A R/A 3 R-Hets is common to several heteromers, namely cross-antagonism [9,28,29]. However, the most important characteristic is that the A 3 R functionality is negligible within the A 2A R/A 3 R-Het. Importantly, selective A 2A R antagonists release the brake on A 3 R activation. This finding adds a new piece in the puzzle of both purinergic and GPCR-heteromer-mediated signaling. This specific feature may be of interest in drug discovery in a time when adenosine receptors are gaining momentum after the approval (in Japan and the USA) of istradefylline, a selective A 2A R antagonist, as adjuvant therapy in Parkinson's disease [30][31][32][33][34]. Especially unexpected was the discovery of heteromers formed by two receptors of the same neurotransmitter/hormone that are in complex with different cognate (heterotrimeric) G proteins. One possibility is a shift in the G-protein-coupling. In fact, the dopamine D1-D2 heteroreceptor does not couple to Gs nor to Gi, but to Gq [5][6][7][8]. However, in the case of the A1R-A2AR heteromer, there is no shift in G-protein coupling, but the overall structure allows increasing or decreasing cAMP levels depending on the concentration of the endogenous agonist. Adenosine preferentially activates the A1R, thus engaging Gi proteins. Nevertheless, when adenosine level increases and the A2AR is activated within the heteromer, Gi-coupling is blunted and Gs engagement occurs. The molecular determinants that make this possible are detailed elsewhere [16,17], but it is worth mentioning that the C-terminal domain of the A2AR plays a fundamental role.
The A2BR is the most enigmatic adenosine receptor; it has a very reduced affinity for the nucleoside and it is also the receptor for netrin, which belongs to a family of proteins involved in axon guidance [26]. On the one hand, comparison of the structural arrangement of A2AR and A2BR has led to the discovery that the second extracellular loop determines low (A2BR) or high (A2AR) affinity for adenosine [27]. On the other hand, as earlier commented, the expression of the A2BR reduces the A2AR-mediated functions by means of the formation of A2A-A2B heteroreceptor complexes. Thus, it seems that the A2BR, per se or negatively acting on the A2AR, contributes to reducing the actions derived from extracellular adenosine accumulation.
The properties of the heteromer described here are different from those previously defined for adenosine receptor heterocomplexes. A simple scheme summarizing the operation of A2AR/A3R-Hets is provided in Figure 5. One of the features of A2AR/A3R-Hets is common to several heteromers, namely cross-antagonism [9,28,29]. However, the most important characteristic is that the A3R functionality is negligible within the A2AR/A3R-Het. Importantly, selective A2AR antagonists release the brake on A3R activation. This finding adds a new piece in the puzzle of both purinergic and GPCR-heteromer-mediated signaling. This specific feature may be of interest in drug discovery in a time when adenosine receptors are gaining momentum after the approval (in Japan and the USA) of istradefylline, a selective A2AR antagonist, as adjuvant therapy in Parkinson's disease [30][31][32][33][34].

Neuronal Primary Cultures
By the current legislation, obtaining protocol approval is not needed if animals are sacrificed to obtain a specific tissue. CD-1 strain mice handling, sacrifice, and further experiments were conducted according to the guidelines set in Directive 2010/63/EU of the European Parliament and the Council of the European Union that is enforced in Spain by National and Regional organisms; the 3R rule (replace, refine, reduce) for animal experimentation was also taken into account. Primary cultures of cortical neurons were obtained from 19-day embryos. Cells were isolated as described in [35] and plated at a confluence of 40,000 cells/0.32 cm 2 . Cells were maintained for 14 days in Neurobasal medium supplemented with 2 mM L-glutamine, 100 U/mL penicillin/streptomycin, and 2% (v/v) B27 supplement (Gibco, Paisley, Scotland, UK), in six-well plates for functional assays.

Fusion Proteins and Expression Vectors
The human cDNAs for the A 2A R and A 3 R cloned in pcDNA3.1 were amplified without their stop codons using sense and antisense primers harboring either unique BamHI and XhoI sites for A 3 R and Hind III and BamHI sites for A 2A R. The fragments were subcloned to be in frame with the sequence coding for an enhanced YFP (pEYFP-N1; Clontech, Heidelberg, Germany) or a Rluc protein (pRluc-N1, PerkinElmer, Wellesley, MA, USA). Final cDNAs encoding for A 2A R-YFP, A 2A R-Rluc and A 3 R-YFP, and fusion proteins having the receptor at the N-terminal end.
After transfection with the corresponding plasmids, the health and viability of transfected cells were proved using the appropriate negative controls. In addition, expression of receptors was tested by either fluorescent confocal microscopy ( Figure 1A) or Rluc expression, and receptor function was tested by performing ERK1/2 activation assays. Using Trypan Blue solution (T8154, Sigma-Aldrich, St. Louis, MO, USA), the percentage of non-viable cells when collected for experiment performance was <15.

Homogeneous Competition Binding Assays in Living Cells
HEK-293T cells expressing HALO-tagged A 2A R in the presence or in the absence of A 3 R were seeded in six-well plates. After 48 h, medium was removed and cells were treated with 100 nM HALOTag-Lumi4-Tb, previously diluted in 3 mL TLB 1X, for 1 h at 37 • C under 5% CO 2 atmosphere in a cell incubator. Cells were then washed four times with 2 mL TLB 1X to remove the excess of HALOTag-Lumi4-Tb, detached with enzyme-free cell dissociation buffer, centrifuged 5 min at 1500 rpm, and collected in 1 mL TLB 1X. Cells at densities in the 2500-3000 cells/well range were plated in white opaque 384-well plates (10 µL). Then, 5 µL of 20 nM fluorophore-conjugated A 2A R antagonist were added in the presence of vehicle or of CGS 21680 hydrochloride at increasing concentrations (0-10 µM range, 5 µL total volume). Plates were then placed at room temperature for 2 h before signal detection. Detailed description of the HTRF assay is found in [37]. Signal was detected using an PheraSTAR microplate reader (PerkinElmer, Waltham, MA, USA) equipped with a Fluorescence Resonance Energy Transfer (FRET) optic module allowing donor excitation at 337 nm and signal collection at both 665 and 620 nm. A frequency of 10 flashes/well was selected for the xenon flash lamp excitation. The signal was collected at both 665 and 620 nm using the following time-resolved settings: delay: 150 µs and integration time: 500 µs. HTRF ratios were obtained by dividing the acceptor (665 nm) by the donor (620 nm) signals and multiplying by 10,000. The 10,000-multiplying factor is used solely for the purpose of easier data handling.

cAMP Determination
For cAMP studies, HEK-293T transfected cells and cortical neurons were prepared. Signaling experiments were performed as previously described [16,17,19,38]. Briefly, 2 h before initiating the experiment, cell-culture medium was replaced by serum-free DMEM medium. Then, cells were detached, resuspended in growing medium containing 50 µM zardaverine, and placed in 384-well microplates (2500 cells/well). Cells were pretreated (15 min) with antagonists (SCH 442416 for A 2A R and/or PSB-10 for A 3 R) and stimulated with agonists (CGS 21680 for A 2A R and/or IB-MECA for A 3 R) (15 min) before adding 0.5 µM forskolin or vehicle. Readings were performed after 1 h incubation at 25 • C. HTRF energy transfer measures were performed using the Lance Ultra cAMP kit (PerkinElmer, Waltham, MA, USA). Fluorescence at 665 nm was analyzed in a PHERAstar Flagship microplate reader equipped with an HTRF optical module (BMG Lab Technologies, Offenburg, Germany).

ERK Phosphorylation Assays
To determine ERK1/2 phosphorylation, HEK-293T transfected cells and cortical neurons were seeded at a density of 40,000 cells/well in transparent Deltalab 96-well microplates, and kept at the cell incubator. Two hours prior to the experiment, the medium was substituted by serum-free DMEM medium. Then, cells were pre-treated for 10 min at 25 • C with antagonists (SCH 442416 for A 2A R and/or PSB-10 for A 3 R) in serum-free DMEM medium and stimulated for an additional 7 min with the selective agonists (CGS 21680 for A 2A R and/or IB-MECA for A 3 R). Then, cells were washed twice with cold PBS before the addition of lysis buffer (15 min treatment). Then, 10 µL of each supernatant was placed in white ProxiPlate 384-well microplates and ERK1/2 phosphorylation was determined using AlphaScreen ® SureFire ® kit (Perkin Elmer, Waltham, MA, USA), following the instructions of the supplier and using an EnSpire ® Multimode Plate Reader (PerkinElmer, Waltham, MA, USA).

Dynamic Mass Redistribution (DMR) Assays
Cell mass redistribution induced upon receptor activation was detected by illuminating the underside of a biosensor with polychromatic light and measuring the changes in the wavelength of the reflected monochromatic light. The magnitude of wavelength shifts (in picometers) is directly proportional to the amount of DMR. Transfected HEK-293T cells were seeded in 384-well sensor microplates to obtain 70-80% confluent monolayers constituted by approximately 10,000 cells per well. Prior to the assay, cells were washed twice with assay-buffer (HBSS with 20 mM HEPES, pH 7.15) and incubated for 2 h with assay-buffer containing 0.1% dimethyl sulfoxide (DMSO) (24 • C, 30 µL/well). Hereafter, the sensor plate was scanned and a baseline optical signature was recorded for 10 min before adding 10 µL of antagonist solution (SCH 442416 for A 2A R and/or PSB-10 for A 3 R) for 30 min, followed by the addition of 10 µL of agonist solution (CGS 21680 for A 2A R and/or IB-MECA for A 3 R). The label-free signature was determined using an EnSpire ® Multimode Plate Reader (PerkinElmer, Waltham, MA, USA). DMR responses were monitored for at least 5000 s. The results were analyzed using EnSpire ® Workstation Software v 4.10 (PerkinElmer, Waltham, MA, USA).

β-arrestin 2 Recruitment
β-arrestin 2 recruitment was determined as previously described [17,36]. Briefly, BRET experiments were performed in HEK-293T cells 48 h after transfection with the cDNA corresponding to A 2A R-YFP or A 3 R-YFP (0.5 µg cDNA each) and 1 µg cDNA corresponding to β-arrestin 2-Rluc. Cells (20 µg protein) were distributed in 96-well microplates (Corning 3600, white plates with white bottom) and were incubated with antagonists (SCH 442416 for A 2A R and/or PSB-10 for A 3 R) for 15 min and stimulated with agonists (CGS 21680 for A 2A R and/or IB-MECA for A 3 R) for 10 min prior to the addition of 5 µM coelenterazine H (Molecular Probes, Eugene, OR, USA). One minute after coelenterazine H addition, BRET between β-arrestin 2-Rluc and receptor-YFP was determined and quantified. The readings were collected using a Mithras LB 940 (Berthold Technologies, Bad Wildbad, Germany) that allows the integration of the signals detected in the short-wavelength filter at 485 nm and the long-wavelength filter at 530 nm. To quantify protein-Rluc expression, luminescence readings were also performed 10 min after adding 5 µM coelenterazine H.

Data Handling and Statistical Analysis
Data from homogeneous binding assays were analyzed using Prism 6 (GraphPad Software, Inc., San Diego, CA, USA). K i values were determined according to the Cheng and Prusoff equation with K D = 20 nM for A 2A R red antagonist [39]. Signal-to-background (S/B ratio) calculations were performed by dividing the mean of the maximum value (µ max ) by that of the minimum value (µ min ) obtained from the sigmoid fits.
The data are mainly shown as the mean ± S.E.M. Statistical analysis was performed with SPSS 18.0 software. The test of Kolmogorov-Smirnov with the correction of Lilliefors was used to evaluate normal distribution and the test of Levene was used to evaluate the homogeneity of variance. Significance was analyzed by one-way ANOVA, followed by Bonferroni's multiple comparison post hoc test. Significant differences were considered when p < 0.05.