Köfalvi et al. BMC Biology (2020) 18:9 https://doi.org/10.1186/s12915-020-0739-0 RESEARCH ARTICLE Open Access Control of glutamate release by complexes of adenosine and cannabinoid receptors Attila Köfalvi1†, Estefanía Moreno2†, Arnau Cordomí3†, Ning-Sheng Cai4, Victor Fernández-Dueñas5,6, Samira G. Ferreira1, Ramón Guixà-González3, Marta Sánchez-Soto4, Hideaki Yano4, Verònica Casadó-Anguera2, Rodrigo A. Cunha1,7, Ana Maria Sebastião8,9, Francisco Ciruela5,6*, Leonardo Pardo3*, Vicent Casadó2* and Sergi Ferré4* Abstract Background: It has been hypothesized that heteromers of adenosine A2A receptors (A2AR) and cannabinoid CB1 receptors (CB1R) localized in glutamatergic nerve terminals mediate the integration of adenosine and endocannabinoid signaling involved in the modulation of striatal excitatory neurotransmission. Previous studies have demonstrated the existence of A2AR-CB1R heteromers in artificial cell systems. A dependence of A2AR signaling for the Gi protein-mediated CB1R signaling was described as one of its main biochemical characteristics. However, recent studies have questioned the localization of functionally significant A2AR-CB1R heteromers in striatal glutamatergic terminals. Results: Using a peptide-interfering approach combined with biophysical and biochemical techniques in mammalian transfected cells and computational modeling, we could establish a tetrameric quaternary structure of the A2AR-CB1R heterotetramer. This quaternary structure was different to the also tetrameric structure of heteromers of A2AR with adenosine A1 receptors or dopamine D2 receptors, with different heteromeric or homomeric interfaces. The specific quaternary structure of the A2A-CB1R, which depended on intermolecular interactions involving the long C-terminus of the A2AR, determined a significant A2AR and Gs protein-mediated constitutive activation of adenylyl cyclase. Using heteromer-interfering peptides in experiments with striatal glutamatergic terminals, we could then demonstrate the presence of functionally significant A2AR-CB1R heteromers with the same biochemical characteristics of those studied in mammalian transfected cells. First, either an A2AR agonist or an A2AR antagonist allosterically counteracted Gi- mediated CB1R agonist-induced inhibition of depolarization-induced glutamate release. Second, co-application of both an A2AR agonist and an antagonist cancelled each other effects. Finally, a CB1R agonist inhibited glutamate release dependent on a constitutive activation of A2AR by a canonical Gs-Gi antagonistic interaction at the adenylyl cyclase level. (Continued on next page) * Correspondence: Fciruela@ub.edu; Leonardo.Pardo@uab.es; vcasado@ub.edu; sferre@intra.nida.nih.gov † Attila Köfalvi, Estefanía Moreno and Arnau Cordomí contributed equally to this work. 5 Unitat de Farmacologia, Departament Patologia i Terapèutica Experimental, Facultat de Medicina, IDIBELL, Universitat de Barcelona, L’Hospitalet de Llobregat, Spain 3 Laboratori de Medicina Computacional, Unitat de Bioestadística, Facultat de Medicina, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain 2 Department of Biochemistry and Molecular Biomedicine, Faculty of Biology, and Institute of Biomedicine, University of Barcelona, 08028 Barcelona, Spain 4 Integrative Neurobiology Section, National Institute on Drug Abuse, Intramural Research Program, National Institutes of Health, Baltimore, MD 21224, USA Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Köfalvi et al. BMC Biology (2020) 18:9 Page 2 of 21 (Continued from previous page) Conclusions: We demonstrate that the well-established cannabinoid-induced inhibition of striatal glutamate release can mostly be explained by a CB1R-mediated counteraction of the A2AR-mediated constitutive activation of adenylyl cyclase in the A2AR-CB1R heteromer. Keywords: Adenosine A2A receptor, Cannabinoid CB1 receptor, GPCR heteromers, Adenylyl cyclase, Glutamate transmission, Striatum, Background activating the Gi-coupled A1R [3]. On the other hand, Adenosine and endocannabinoids, such as anandamide high concentrations of adenosine produce the opposite and 2-arachidonylglycerol (2-AG), are very ubiquitous effect, by activating the Gs-coupled A2AR, which pro- non-classical neurotransmitters that modulate the trans- motes glutamate release by activating adenylyl cyclase mission ensured by other more classical neurotransmit- (AC)-PKA signaling and allosterically counteracting A1R ters. In the striatum, the modulatory role of adenosine signaling within the A1R-A2AR heteromer [3, 6]. The and endocannabinoids converge in excitatory synapses, quaternary structure of the A1R-A2AR heteromer has where their signaling is integrated by G protein-coupled been recently proposed and shown to be heterotetra- receptors (GPCRs) localized in glutamatergic nerve termi- meric, constituted by homodimers of A1R and A2AR nals [1]. Adenosine is produced from the conversion of coupled to their cognate G proteins [6]. This is similar ATP (by ectonucleotidases) that is co-released with to the also recently described quaternary structure of the glutamate from the nerve terminals and from astrocytes A2AR-dopamine D2 receptor (D2R) heteromer [7], [1, 2]. Adenosine can then bind and activate adenosine re- which is localized postsynaptically in the striatum [8]. ceptors of the A1 or A2A subtype (A1R or A2AR, respect- It has been recently demonstrated that the A2AR-D2R ively) localized presynaptically, promoting inhibition or heteromer forms part of functional complexes that in- facilitation of glutamate release, respectively [1–3]. On the clude A2AR-D2R heterotetramers, constituted by A2AR other hand, endocannabinoids are produced “on demand” and D2R homodimers with their respective cognate Gs from endocannabinoid precursors by the action of en- and Gi proteins, and AC subtype AC5 [7]. The quater- zymes localized in the postsynaptic plasma membrane [4]. nary structure of these complexes is stabilized by specific One of the best studied functions of endocannabinoids is interactions between transmembrane domains (TMs) of “retrograde signaling” with stimulation of presynaptic can- the receptors, which determine the homomeric and het- nabinoid CB1 receptors (CB1R) and the consequent inhib- eromeric interfaces in the A2AR-D2R heterotetramer, as ition of neurotransmitter release [4]. well as between TMs of the receptors and TMs of AC5 We have previously hypothesized that adenosine and [7]. In addition, interactions between intracellular do- endocannabinoids exert a fine-tune modulation of striatal mains play a significant role in the stabilization of the glutamate release from striatal glutamatergic terminals, by complex, namely a strong electrostatic interaction be- which low adenosine plus high endocannabinoid tone tween the C-terminal domain of the A2AR (A2AR-CT) would produce the weakest, while high adenosine plus and the intracellular end of TM 5 of the D2R [9–12] low endocannabinoid tone would produce the strongest and between the N-terminal domain (NT) of AC5 and glutamate release [1]. We also hypothesized that this fine- βγ-subunits of the G proteins [13]. The predicted qua- tune modulation depends, not only on the adenosine and ternary structure of the A2AR-D2R heterotetramer in endocannabinoid tone, but on the ability of specific sub- complex with AC5 provided the frame for the canonical types of adenosine and cannabinoid receptors to form het- Gs-Gi antagonistic interaction at the AC level [7]. This eromers, mainly A1R-A2AR and A2AR-CB1R heteromers, canonical interaction implies the ability of an activated and on their unique biochemical properties (reviewed in Gi-coupled receptor to inhibit AC activation by a Gs- [1]). A1R-A2AR heteromers have been characterized both coupled receptor [14] and requires the simultaneous functionally and structurally, but some inconsistencies re- respective interaction of the Ras-GTPase domain of the main about the functional properties of A2AR-CB1R het- α-subunits of the Gs and Gi proteins with the C2 and eromers and even about their existence in striatal C1 catalytic domains of AC [15]. glutamatergic terminals (see below). Importantly, the heteromeric and homomeric interfaces The A1R-A2AR heteromer acts as a concentration- of the A1R-A2AR heterotetramer were found to be differ- dependent switch that mediates the adenosine control of ent from those of the A2AR-D2R heterotetramer [6, 7]. striatal glutamatergic transmission [3]. Since adenosine The consequent different conformation of the A1R-A2AR binds with higher affinity to A1R than to A2AR [5], low heterotetramer was associated with its inability to sustain concentrations of adenosine inhibit glutamate release by a canonical Gs-Gi antagonistic interaction at the AC level Köfalvi et al. BMC Biology (2020) 18:9 Page 3 of 21 [6]. Particularly striking was the involvement of A2AR- [12]. On the other hand, when an agonist and an antag- CT, because its deletion enabled the canonical Gs-Gi an- onist bind simultaneously to the two orthosteric sites of tagonistic interaction by the A1R-A2AR heteromer. On the A2AR homodimer, they counteract each other’s ef- the other hand, contrary to the A2AR-D2R heterotetra- fects [12]. At the biochemical level, a negative allosteric mer, A2AR-CT deletion did not disrupt A1R-A2AR het- interaction between orthosteric A2AR agonists and an- eromerization, indicating its lack of involvement on the tagonists in an A2AR homodimer could be demon- stabilization of the quaternary structure of the A1R-A2AR strated in dissociation kinetic binding experiments of a heterotetramer [6]. These results also indicated that the radiolabeled A2AR antagonist versus A2AR agonists and ability of A1R in the A1R-A2AR heteromer to mediate in- antagonists, including the non-selective antagonist caf- hibition of glutamate release in the striatal glutamatergic feine [12]. At the behavioral level, these homomeric allo- terminals was not dependent on the canonical Gs-Gi an- steric interactions could better explain the psychomotor tagonistic interaction at the AC level and, therefore, on stimulant effects of caffeine and selective A2AR antago- the inhibition of A2AR-mediated AC-PKA signaling. In- nists than the classically assumed competitive antagon- stead, it would be most probably dependent on the clas- ism between A2AR antagonists and endogenous sical inhibitory effect of βγ-subunits on presynaptic adenosine for the same orthosteric site [22, 23]. Thus, calcium channels [16, 17]. the allosteric homomeric interactions between A2AR ag- The ability of A2AR and CB1R to heteromerize was onists and antagonists predicted a counterintuitive coun- first suggested from results obtained in artificial cell sys- teraction of a high locomotor depressant dose of an tems using biophysical techniques [18]. In the same A2AR antagonist in the rat with an also locomotor de- study, signaling experiments performed in a neuroblast- pressant dose of an A2AR agonist in the rat [22]. oma cell line indicated the existence of the conventional There is therefore a significant amount of evidence G protein coupling for both receptors and a dependence that supports the existence of functional A2AR-CB1R on A2AR signaling for the expression of the Gi- heteromers in the striatal glutamatergic terminals with mediated inhibition of AC activity by CB1R agonists similar structural and biochemical characteristics to [18]. As expected from the canonical interaction at the those of the A2AR-D2R heteromers and that they con- AC level, a CB1R agonist could counteract an A2AR stitute a main population of CB1R that play an essential agonist-induced AC activation. But, A2AR blockade also role in striatal synaptic transmission and plasticity [24– counteracted the ability of a CB1R agonist to inhibit 28]. However, in a recent study, A2AR-CB1R complexes forskolin-induced AC activation [18]. It was then sug- were identified by the proximity ligation assay in the gested that activation of A2AR in the A2AR-CB1R het- mouse striatum and suggested to represent A2AR-CB1R eromer allows the effective coupling of CB1R to Gi heteromers localized postsynaptically in GABAergic proteins and, consequently, that CB1R signaling is en- striatopallidal neurons, but not in corticostriatal termi- tirely dependent on A2AR signaling [18]. Strong evi- nals. This conclusion was based on results obtained in dence for physical interactions between A2AR and CB1R mice with differential genetic blockade of CB1R in fore- in striatal glutamatergic terminals was afterwards brain GABAergic neurons versus dorsal telencephalic reported, including a robust co-localization and co- glutamatergic neurons [29]. In addition, parallel experi- immunoprecipitation [19]. In the same study, an A2AR ments in conditionally immortalized striatal neuroblasts agonist significantly decreased the potency of a CB1R suggested that A2AR-CB1R heteromers signal by Gq agonist to inhibit striatal glutamate release and pre- protein coupling, but not through Gs and Gi proteins synaptic corticostriatal glutamatergic transmission [19]. [29]. In view of these apparently controversial results, More in line with the expected dependence on A2AR acti- the goals of the present study were, first, to establish the vation within the A2AR-CB1R heteromer, some studies quaternary structure and biochemical properties of the also found evidence for counteraction of CB1R-mediated A2AR-CB1R heteromer and to compare them with those corticostriatal transmission by A2AR antagonists [20, 21]. of the A1R-A2AR and A2AR-D2R heterotetramers. Sec- Although apparently incompatible with the involve- ond, we aimed at establishing the localization, biochem- ment of a single population of A2AR, forming hetero- ical properties, and functional significance of the A2AR- mers with CB1R, similar effects of A2AR agonists and CB1R heteromers in the striatal glutamatergic terminals. antagonists have also been obtained with the A2AR-D2R heteromer and found to depend on allosteric interac- Results tions that depend on its tetrameric structure. When ei- A2AR and CB1R in the A2AR-CB1R heteromer couple to ther an A2AR agonist or an A2AR antagonist binds to their respective cognate Gs and Gi proteins in HEK-293T the orthosteric sites of the A2AR homodimer in the cells A2AR-D2R heterotetramer, they both produce an allo- The complemented donor-acceptor resonance energy steric decrease in the affinity and efficacy of D2R ligands transfer (CODA-RET) assay was first used to analyze the Köfalvi et al. BMC Biology (2020) 18:9 Page 4 of 21 preferred functional G protein coupling of A2AR and results represent a readout of a negative allosteric inter- CB1R in the A2AR-CB1R heteromer, in transiently action between two orthosteric agonists within the A2AR- transfected HEK-293T cells. In this assay, two comple- CB1R heteromer, by which an A2AR agonist decreases the mentary halves of the bioluminescent protein Renilla lu- potency of a CB1R agonist-mediated Gi protein activation. ciferase (Rluc8 variant; nRluc and cRluc) are separately If the A2AR-CB1R heteromer would exhibit a tetrameric fused to two different GPCR units (protomers) putatively structure similar to the A2AR-D2R heterotetramer, we able to oligomerize, and a fluorescent protein, the mVe- could expect the same allosteric interaction with A2AR ag- nus variant of the yellow fluorescence protein (YFP), is onists and antagonists and their counteracting effect when fused to a Gα protein subunit. Ligand-induced changes simultaneously applied (see “Background” and ref. [12]). In in CODA-RET measurements imply, first, a successful fact, both NECA and caffeine produced a concentration- complementation of Rluc and, therefore, oligomerization dependent decrease of the change in BRET ratio values in- of the corresponding protomers. Second, although duced by an EC50 concentration of CP55940 (2 μM; Fig. CODA-RET does not provide an estimate of the degree 1d, e). In addition, effective concentrations of NECA (10 of oligomerization (affinity or stoichiometry), it repre- μM; Fig. 1d) and caffeine (3 mM; Fig. 1e) became ineffective sents the reading of a specific G protein activation when co-applied (Fig. 1e). These results therefore recapitu- through the GPCR heteromer [30–32]. A2AR was fused late allosteric interactions within the A2AR-D2R heterote- to cRluc, and CB1R was fused to nRluc and co- tramer and, therefore, suggest an also tetrameric structure transfected with either Gαs, Gαi1 or Gαq fused to YFP. of the A2AR-CB1R heteromer. Concentration-response curves of ligand-induced We also analyzed the possible opposite allosteric inter- changes in BRET were then determined in the presence action by which a CB1R agonist could modify the potency of different concentrations of the non-selective adeno- of an A2AR agonist-mediated Gs protein activation (cells sine receptor agonist NECA or the selective CB1R agon- transfected with Gαs-YFP). CP55940 did not produce a sig- ist CP55940. A concentration-response curve with nificant shift in the concentration-response curve of NECA. NECA could only be obtained when the cells were trans- EC50 values of the concentration-response curves of NECA fected with Gαs-YFP (EC50 values, in mean ± S.E.M: in the absence and presence of CP55940 (10 μM) were, in 0.35 ± 0.08 μM; n = 5 with triplicates), but not with mean ± S.E.M, 0.52 ± 0.23 μM and 0.57 ± 0.13 μM, respect- Gαi1-YFP or Gαq-YFP (Additional file 1: Figure S1a; ively (paired t test: p > 0.05; n = 5 with triplicates); Emax Additional file 2: Data values 1). On the other hand, a values in the absence and presence of CP55940 (10 μM) concentration-response with CP55940 could only be ob- were, in mean ± S.E.M, 0.009 ± 0001 and 0.009 ± 0.001 tained when the cells were transfected with Gαi-YFP BRET ratio units (BRU), respectively (paired t test: p > 0.05; (EC50 values, in mean ± S.E.M: 1.05 ± 0.39 μM; n = 5 n = 5 with triplicates) (Additional file 1: Figure S2; Add- with triplicates), but not with Gαs-YFP or Gαq-YFP itional file 2: Data values 2). These results indicate the exist- (Additional file 1: Figure S1b; Additional file 2: Data ence of a unidirectional allosteric modulation between values 1). A positive control of Gq coupling by a GPCR orthosteric ligands in the A2AR-CB1R heteromer. heteromer was obtained with the previously reported serotonin 5-HT2A receptor (5-HT2AR)-D2R heteromer The A2AR-CB1R heteromer has a different quaternary [33, 34]. A concentration-response curve was obtained structure compared to the A2AR-D2R and the A1R-A2AR with serotonin, but not dopamine, when the cells were heterotetramers transfected with 5-HT2AR fused to cRluc, D2R fused to The possible heterotetrameric structure of the A2AR- nRluc and Gαq-YFP (EC50 values, in mean ± S.E.M: 0.21 CB1R heteromers was further evaluated in HEK-293T ± 0.05 μM; n = 5 with triplicates) (Additional file 1: transfected cells by a BRET experiment based on the Figure S1c; Additional file 2: Data values 1). These re- double complementation of both BRET bioluminescent sults indicate that within the A2AR-CB1R heteromer, and fluorescent proteins [12, 31], with complementary A2AR and CB1R preferentially couple to their respective halves of Rluc and YFP separately fused to different pro- cognate Gs and Gi proteins. Nevertheless, the results do tomers of A2AR and CB1R. As shown in Additional file not discard the possible coupling to Gq by A2AR and 1: Figure S3, significantly higher BRET values could be CB1R in the A2AR-CB1R heteromer in another cellular obtained with co-transfection of A2AR-cRluc, A2AR- environment, such as in the conditionally immortalized nRluc, CB1R-cYFP, and CB1R-nYFP, as compared striatal neuroblasts and in the striatum, in the GABAer- with controls, where the A2AR or CB1R constructs gic striatopallidal neurons or astrocytes [29]. were substituted by dopamine D1 receptor (D1R) con- Remarkably, NECA (10 μM) produced a shift to the right structs, D1R-cRluc and D1R-nRluc or D1R-cYFP and in the concentration-response curve of CP55940 (Fig. 1a), D1R-nYFP. with a significant increase in EC50 values (Fig. 1b) and no Bimolecular fluorescence complementation (BiFC) ex- significant difference in the Emax values (Fig. 1c). These periments with synthetic interfering peptides were then Köfalvi et al. BMC Biology (2020) 18:9 Page 5 of 21 Fig. 1 Modulation by A2AR ligands on CB1R-mediated G protein activation in the A2AR-CB1R heteromer. a–e CODA-RET experiments, where two complementary halves of Rluc (cRluc and nRluc) are respectively fused to the A2AR and CB1R and YFP is fused to the α-subunit of Gi. HEK-293T cells were transiently transfected with cDNAs of A2AR-cRluc (3.33 μg), CB1R-nRluc (1.67 μg), Gαi1-YFP (5 μg), and non-fused β1 and γ2 subunits (4.5 μg and 5 μg, respectively). a Concentration-response curves of the effect of the selective CB1R agonist CP55940 on the ligand-induced BRET changes, which are determined by changes in the interaction of the A2AR-CB1R heteromer with Gi, in the presence (blue plot) and absence (red plot) of the non-selective adenosine agonist NECA (10 μM). Data are means ± S.E.M. of triplicate BRET ratio values of a representative experiment. b, c EC50 and Emax values of 12 independent experiments performed in triplicate, expressed as means ± S.E.M.; the EC50 and Emax values were obtained by non-linear regression fitting to a sigmoidal concentration-response curve and analyzed statistically with a paired t-test (*: p < 0.05, compared with the absence of NECA). d, e Modification by NECA (1 and 10 μM), caffeine (CAFF, 1 and 3 mM) and NECA (10 μM) plus caffeine (3 mM) on the effect of an EC50 concentration of CP55940 (2 μM); values are means ± S.E.M. (n = 10–11 with triplicates) of the percentage of the effect of CPP55940 alone and analyzed statistically with repeated measures ANOVA, followed by Dunnett’s multiple comparison test (**: p < 0.01, compared with CPP55940 alone) performed to elucidate the homomeric and heteromeric heteromers, indicate that under in vivo conditions (live interfaces of the heterotetramer, which is the same strat- cell preparations) BiFC complex formation can be re- egy that was used to reveal the precise quaternary struc- versible [6, 7, 31, 36–38]. Complementary halves of YFP ture of the A1R-A2AR and A2AR-D2R heterotetramers were separately fused to A2AR (A2AR-cYFP) and CB1R [6, 7]. While BiFC complex formation under in vitro (CB1R-nYFP) or two different molecules of A2AR conditions (purified complementary fluorescent pro- (A2AR-nYFP and A2AR-cYFP) or CB1R (CB1R-nYFP teins) has been considered to be essentially irreversible and CB1R-cYFP), in the absence and presence of syn- [35], several studies, including our own on GPCR thetic peptides with the amino acid sequence of all Köfalvi et al. BMC Biology (2020) 18:9 Page 6 of 21 possible TMs of both receptors fused to the cell- cYFP and CB1R-nYFP with TM5 and TM6 of both penetrating HIV transactivator of transcription (TAT) A2AR and CB1R (Fig. 2a, b). These results point to the sequence (TMs are abbreviated TM 1, TM 2,…; TM involvement of TM 5/6 in the A2AR-CB1R heteromer peptides are abbreviated TM1, TM2,…). We could first interface. Notably, the previously studied heteromer demonstrate the selective interference of BiFC of A2AR- interface of A1R-A2AR was also TM 5/6 [6], whereas Fig. 2 Heteromeric and homomeric interfaces in the A2AR-CB1R heterotetramer. a–f BiFC experiments with synthetic peptides with the amino acid sequence of all TMs of A2AR and CB1R. HEK-293T cells were transiently transfected with a, b cDNAs of A2AR and CB1R separately fused to complementary halves of YFP (A2AR-cYFP and CB1-nYFP; 6 μg in both cases); c, e cDNAs of two different molecules of A2AR separately fused to complementary halves of YFP (A2AR-cYFP and A2AR-nYFP; 3 μg in both cases) without (c) or with (e) co-transfection with cDNA of non-fused CB1R (1 μg); d, f cDNAs of two different molecules of CB1R separately fused to complementary halves of YFP (CB1R-cYFP and CB1R-nYFP; 3 μg in both cases) without (d) or with (f) co-transfection with cDNA of non-fused A2AR (1 μg). Cells were treated for 4 h with medium (control) or 4 μM of indicated TM peptides (numbered 1–7). Values are means ± S.E.M. (n = 6 with triplicates in all the experiments) of the percentage of the fluorescence in the control group and analyzed statistically with repeated measures ANOVA, followed by Dunnett’s multiple comparison test (***: p < 0.001, compared with control) Köfalvi et al. BMC Biology (2020) 18:9 Page 7 of 21 the heteromer interface of A2AR-D2R was found to be in the crystal structure of the μ-opioid receptor [39]. TM 4/5 [7]. BiFC of A2AR-nYFP and A2AR-cYFP (in The modeling of a symmetrical TM 4/6 interface for the absence of CB1R) was only significantly reduced by homomerization is not straightforward as it is not steric- TM6 (Fig. 2c), as previously reported [7], and BiFC of ally feasible that these two helices in one protomer sim- CB1R-nYFP and CB1R-cYFP (in the absence of A2AR) ultaneously interact with the same helices in the other was only reduced by TM4 (Fig. 2d). These results indi- protomer. To fit with the experimental data, in each ho- cate that TM 6 forms the A2AR interface and TM 4 modimer, TM 4 of the internal protomer engaged in forms the CB1R interface for homodimerization, when heteromerization should interact with TM 6 of the ex- each receptor is expressed alone. Significantly, the inter- ternal protomer (an asymmetrical interface; Fig. 3a). In face for both CB1R-CB1R and A2AR-A2AR homodimers fact, crystal structures and molecular dynamics (MD) changed in the presence of the other non-fused molecu- simulations have already provided evidence for the pos- larly different receptor (Fig. 2e, f). Thus, TM4 and TM6 sibility that GPCRs show several possible symmetrical of A2AR reduced BiFC of A2AR-cYFP and A2AR-nYFP and asymmetrical homomeric TM interfaces [40, 41]. in the presence of non-fused CB1R (Fig. 2e) and TM4 What the present results indicate is that the preferred and TM6 of CB1R reduced BiFC of CB1R-cYFP and homomeric interfaces can be determined by their het- CB1R-nYFP in the presence of non-fused A2AR (Fig. eromeric partner in a GPCR heterotetramer. 2f). The results in Fig. 2 are expressed as means ± In our model of the A2AR-CB1R heterotetramer, Gi S.E.M. of the percentage of the fluorescence in the con- and Gs bind to the external protomer of the CB1R and trol group, where control values (without interfering A2AR homodimers, respectively, as we have previously peptides) were always between 12,000 and 15,000 rela- suggested for the A1R-A2AR and A2AR-D2R heterote- tive fluorescence units (RFU; value that depends on the tramers [6, 7]. The fact that the C-terminal α5-helix of gain setting in the measurement equipment). Non- the G protein binds an intracellular cavity of the recep- transformed data show no significant change in fluores- tor that is opened mainly by the outward movement of cence units upon co-transfection with a non-fused TM 6 (TM 5 also moves but to a lesser extent), adds receptor. complexity to the modeling process. Thus, the TM 4/6 These results therefore demonstrate a change in a interface was modeled with the internal protomer in the GPCR homodimeric interface imposed by an additional inactive conformation (TM 6 closed) and the external molecular interaction with another GPCR. Clearly, the protomer in the active conformation (TM 6 open). The observed TM 4/6 interface for CB1R-CB1R and A2AR- stability of these interfaces was evaluated using MD sim- A2AR homodimerization in the A2AR-CB1R heterote- ulations (see “Methods”), which yielded a converged tramer differs from the TM 4/5 interface for A1R-A1R structure for the complex with steady values for both and A2AR-A2AR homodimerization in the A1R-A2AR root mean squared deviations and distance between pro- heterotetramer [6] and from the TM 6 interface for tomers (not shown). Figure 3 shows a scheme (Fig. 3a) A2AR-A2AR and D2R-D2R homodimerization in the and two different views of the computational model (Fig. D2R-A2AR heterotetramer [7] (Fig. 3a). The almost 3b). The mechanism for receptor catalyzed nucleotide complete reduction of fluorescence to background fluor- exchange in G proteins involves a large-scale opening of escence levels induced by the TM6 peptide of A2AR in the α-helical domain (αAH) of the α-subunit, from the BiFC experiments with A2AR-nYFP and A2AR-cYFP in- Ras domain, allowing GDP to freely dissociate [42, 43]. dicate that, without the presence of A1R or CB1R, Notably, the opening of αiAH and αsAH domains in the homodimerization with a TM 6 interface is the preferred proposed models of the A2AR-CB1R and A2AR-D2R oligomerization state of the A2AR, while higher-order heteromers are pointing towards different directions A2AR oligomerization (homotrimers, heterotrimers) is (Fig. 3a [7];), while they are facing each other in the clearly unlikely. Even though concomitant homomeriza- A1R-A2AR heteromer (Fig. 3b [6];). tion of A2AR by TM 6 or TM 4/5 interfaces is theoretic- ally possible, our present and previous results indicate The C-terminal domain of the partner receptor that the latter interface is conditional upon heteromeri- determines the homomeric interfaces in the A2AR-CB1R zation with A1R. heterotetramer We next constructed, using computational tools (see A major question is the mechanism by which A2AR “Methods”), a structural model of the A2AR-CB1R het- changes the CB1R-CB1R homomeric interface and erotetramer using the information of the TM interfaces, CB1R changes the A2AR-A2AR interface (see above). A for homo- (TM 4/6) and hetero- (TM 5/6) dimerization, remarkable difference between the quaternary structure derived from the BiFC experiments in the absence and of the A2AR-CB1R heterotetramer and the A1R-A2AR presence of the TAT-fused peptides. The symmetrical and A2AR-D2R heterotetramers is the proximity of the TM 5/6 interface for heteromerization was modeled as C-terminal domain of the internal protomers to TM 5 Köfalvi et al. BMC Biology (2020) 18:9 Page 8 of 21 Fig. 3 Quaternary structure of the A2AR-CB1R heterotetramer. a Schematic representation of the A2AR-CB1R heterotetramer (left) viewed from the extracellular side (colored arrows indicate the C-terminal segments of the internal protomers able to reach the external G protein-bound protomers) and the previously reported A2AR-A1R (middle) and A2AR-D2R (right) heterotetramers. Inactive/internal and active/external protomers of A2AR are shown in light and dark green, respectively, and inactive/internal and active/external protomers of CB1R, A1R, or D2R are shown in orange and red, respectively. The α-subunit of the G protein is shown in light gray, the β-subunit in dark gray, and the γ-subunit in purple. The α- helical domain (αAH) of the α-subunit, which performs a large-scale opening from the Ras domain, is shown in yellow. b A representative molecular representation of the A2AR-CB1R heterotetramer obtained from the MD simulation viewed from the membrane (left) or from the extracellular side (right). Color codes are as in panel a. c Proposed interaction between phosphorylated residues at the C-terminus of the CB1R (Gi-bound external protomer) with arginine residues at the end of TM 5 of the A2AR (internal protomer). d Proposed interaction between phosphorylated residues at the C-terminus of the A2AR (Gs-bound external protomer) with arginine residues at the end of TM 5 of the CB1R (internal protomer) Köfalvi et al. BMC Biology (2020) 18:9 Page 9 of 21 and TM 6 of the external protomers (see arrows in Fig. the results in Fig. 4 are expressed as means ± S.E.M. of 3a). Moreover, another difference among the three Gi- the percentage of the fluorescence in the control group, coupled partners of the A2AR is the long C-terminal do- where control values (without interfering peptides) were main of CB1R (73 residues) as compared to that of the always between 12,000 and 15,000 RFU. Non-transformed A1R (34 residues) or the D2R (12 residues). We there- data show no significant change in fluorescence units fore speculated that the C-terminal domain of the part- upon co-transfection with a non-fused receptor. Thus, we ner receptor could be responsible for the change in the can conclude that the C-terminal domain of A2AR modi- homomeric interface. To test this hypothesis, we engi- fies the CB1R-CB1R homomeric interface stabilizing the neered an A2AR mutant lacking the C-terminal end asymmetric TM 4/6 interface. Since the structure of the (A2AΔCTR) and evaluated BiFC of CB1R-nYFP and heterotetramer is symmetrical, we hypothesize that the C- CB1R-cYFP, in the presence and absence of each of the terminal domain of CB1R also modifies the A2AR-A2AR peptides of CB1R, and in the presence of A2AΔCTR (Fig. homomeric interface stabilizing the asymmetric TM 4/6 4a). While TM4 and TM6 reduced fluorescence in the interface. These conclusions are in agreement with previ- presence of A2AR (Fig. 2f, see above), TM4 and TM5, ous results in which we described strong electrostatic but not TM6, reduced fluorescence in the presence of interactions between phosphorylated Thr467-Ser468 at the A2AΔCTR (Fig. 4a). As an additional control, we tested C-terminal domain of CB1R and Arg205(5.66)- whether A2AΔCTR could modify the interface of the Arg206(5.67) at the intracellular end of TM 5 of A2AR as A1R homodimer in the A1R-A2AR heterotetramer. well as between phosphorylated Ser374 at the C-terminal Clearly, the C-terminal domain of A2AR had no influence domain of A2AR and the 215(5.64)VLRRRRKRVN224 epi- in the A1R homodimer (Fig. 4b, c). The same as for Fig. 2, tope at the intracellular end of TM 5 of D2R [11]. Here, Fig. 4 A2AR C-terminus-guided homomeric interface of the CB1R homodimer in the A2AR-CB1R heterotetramer. a–c BiFC experiments with synthetic peptides with the amino acid sequence of the TMs of CB1R and A1R. HEK-293T cells were transiently transfected with the following: a cDNA of two different molecules of CB1R separately fused to complementary halves of YFP (CB1R-cYFP and CB1R-nYFP; 3 μg in both cases) with co-transfection with cDNA of non- fused A2AR mutant lacking the C-terminal end (A2AΔCTR; 1 μg); b, c cDNAs of two different molecules of A1R separately fused to complementary halves of YFP (A1R-cYFP and A1R-nYFP; 3 μg in both cases) with co-transfection with cDNA of non-fused A2AR (b; 1 μg) or non-fused A2AΔCTR (c; 1 μg). Cells were treated for 4 h with medium (control) or 4 μM of indicated TM peptides (numbered 1–7). Values are means ± S.E.M. (n = 6 with triplicates in all the experiments) of the percentage of the fluorescence in the control group and analyzed statistically with repeated measures ANOVA, followed by Dunnett’s multiple comparison test (***: p < 0.001, compared with control) Köfalvi et al. BMC Biology (2020) 18:9 Page 10 of 21 Fig. 5 (See legend on next page.) Köfalvi et al. BMC Biology (2020) 18:9 Page 11 of 21 (See figure on previous page.) Fig. 5 Canonical Gs-Gi antagonistic interaction at the AC level in the A2AR-CB1R heterotetramer. a–h Effect of TM peptides on the cAMP formation induced by forskolin (500 nM) or by the A2AR agonist CGS21680 (CGS, 500 nM) and counteractive effects of the CB1R agonist CP55940 (200 nM). HEK- 293T cells were transiently transfected with the cDNAs of CB1R and A2AR (2 μg in both cases) and treated for 4 h with 4 μM of peptides TM4, TM5, TM6, or TM7 of the A2AR or CB1R. Values are means ± S.E.M. (n = 5 with triplicates in all experiments) of the percentage of forskolin-induced cAMP formation and analyzed statistically with repeated measures ANOVA, followed by Dunnett’s multiple comparison test (*, **, and ***: p < 0.05, p < 0.01, and p < 0.001, respectively, compared with basal; #, ##, and ###: p < 0.01 and p < 0.001, respectively, compared to forskolin or CGS) we propose similar interactions for the A2AR-CB1R both A2AR and CB1R, but not TM4 or TM7 (negative heterotetramer. Although it is not possible to model the control), counteracted the canonical interaction (Fig. full-length C-terminal domain of either A2AR or CB1R, 5a–h; Additional file 2: Data values 3). Since TM 4 and we modeled phosphorylated Ser374 of A2AR and TM 5 are selectively involved in homo- and heteromeri- phosphorylated Thr467-Ser468 of CB1R, connected by an zation, respectively, while TM 6 is involved in both arbitrary C-terminal domain (dotted lines in Fig. 3c, d). homo- and heteromerization, these results demonstrate Figure 3c shows the electrostatic interactions between that A2AR-CB1R heteromerization provides the neces- phosphorylated Thr467-Ser468 of CB1R and Arg205(5.66)- sary frame for the canonical Gs-Gi antagonistic inter- Arg206(5.67) of A2AR, and Fig. 3d shows the interaction action at the AC level. This effect was also present in between phosphorylated Ser374 of A2AR and Lys300(5.66)- the A2AR-D2R heterotetramer [7] but absent in the Arg307(5.73)-Arg311(5.77) of CB1R. We hypothesize that A1R-A2AR heterotetramer, which was devoid of canon- these electrostatic interactions are the driving force for the ical interaction [6]. This is rationalized by the fact that conformational change, from the TM 4 interface of the simultaneous opening of αiAH and αsAH is feasible in CB1R homodimer in the absence of A2AR and from the the A2AR-CB1R and A2AR-D2R heterotetramers (αAH TM 6 interface of the A2AR homodimer in the absence of opening is towards different direction) and is not feasible CB1R, to the TM 4/6 interface that orients TM 5 towards in A1R-A2AR because αiAH and αsAH face each other the C-terminal domain of the partner receptor in the (Fig. 3a) [6]. A2AR-CB1R heterotetramer. Experiments were also performed in cells trans- fected with both receptors and pre-treated with per- The A2AR-CB1R heterotetramer provides a frame for the tussis toxin (PTX) or cholera toxin (CTX) (for details, canonical Gs-Gi antagonistic interaction at the AC level see legend to Additional file 1: Figure S4), which alter The ability of A2AR-CB1R heterotetramer to modulate Gi- and Gs-mediated signaling, respectively. As ex- AC signaling was then tested in HEK-293T cells by pected, we observed blockade of CP55940-induced using heteromer-interfering peptides. As expected from cAMP decrease by PTX, and blockade of CGS21680- their respective coupling to Gi and Gs proteins, the induced cAMP increase by CTX (Additional file 1: CB1R agonist CP55940 (200 nM) decreased cAMP for- Figure S4d and S4e). In addition, PTX and CTX also mation induced by forskolin (500 nM) and the selective blocked the effect of CGS21680 and CP55940, re- A2AR agonist CGS21680 (500 nM) increased cAMP in spectively (Additional file 1: Figure S4d and S4e). cells transfected with CB1R or A2AR, respectively (Add- These results indicate that both A2AR- and CB1R- itional file 1: Figure S4a and S4b). The same effect of mediated signaling in the A2AR-CB1R heterotetramer both ligands was observed in cells co-transfected with are dependent on the functional integrity of both Gs A2AR and CB1R, where, in addition, CP55940 counter- and Gi proteins. The same phenomenon has been acted CGS21680-mediated increase in cAMP (Additional previously described for the A1R-A2AR heteromer [6, file 1: Figure S4c). This is in agreement with the ability 44], while the toxins maintained their expected select- of A2AR and CB1R to signal with their cognate G pro- ivity for either Gs or Gi in the A2AR-D2R heteromer teins and to establish the canonical Gs-Gi antagonistic [7]. This cross-communication is not fully understood interaction at the AC level in the A2AR-CB1R heterote- but it could be related to the closer proximity of the tramer. Interfering peptides (TM4 to TM6 of CB1R and α-subunits of Gs and Gi coupled to the A2AR-CB1R A2AR) were then used to demonstrate the dependence and A1R-A2AR heterotetramers. on the integrity of the A2AR-CB1R heterotetramer for the existence of this canonical Gs-Gi antagonistic inter- Presynaptic A2AR-CB1R heterotetramers mediate the action, as previously shown for the A2AR-D2R heterote- cannabinoid-induced inhibition of striatal glutamate tramer [7]. None of these peptides significantly modified release the effect of either CP55940 or CGS21680 when activat- The demonstration of the presence and functional sig- ing their respective receptor (Fig. 5a–h; Additional file 2: nificance of A2AR-CB1R heteromers in striatal gluta- Data values 3). On the other hand, TM5 and TM6 of matergic terminals was then approached by analyzing Köfalvi et al. BMC Biology (2020) 18:9 Page 12 of 21 the ability of the CB1R agonist WIN55,212-2 to inhibit would reflect functional correlations of the allosteric in- depolarization-induced glutamate release from rat stri- teractions within the A2AR-CB1R heterotetramer, they atal synaptosomes (see “Methods” and ref. [19]). We should be counteracted by a synthetic peptide that de- would first expect that, as previously reported [19, 21], stabilizes the heteromeric interface. Finally, we should A2AR agonists would reduce the effect, and more specif- also find a correlate of the Gs-Gi canonical interaction ically the potency, of WIN55,212-2. Second, we should at the AC level. Indeed, as shown in Fig. 6a–d, WIN55, be able to demonstrate that A2AR antagonists would 212-2 inhibited evoked glutamate release (EC50, 5.5 nM) produce the same effect as A2AR agonists and that when in a concentration-dependent fashion, reaching its max- co-applied they would counteract each other’s effect. imal effect (77.0 ± 3.0% of DMSO control) at 0.03 μM. Third, if these effects of A2AR agonists and antagonists The concentration-response inhibitory curve of WIN55, Fig. 6 A2AR-CB1R heteromer-mediated control of the evoked glutamate release from striatal nerve terminals. a Concentration-response curves of the inhibitory effect of WIN55,2121-2 (WIN) on the evoked glutamate release, expressed as the difference in % of the control value, alone or in the presence the A2AR agonist CGS21680 (CGS, 0.03 μM), the A2AR antagonist SCH58261 (SCH, 0.1 μM), or both. Data are means ± S.E.M. of 7–14 experiments with duplicates per treatment and concentration. b–d Individual results from the experiments shown in a with WIN at the concentrations of 0.01 μM (b), 0.03 μM (c), and 0.1 μM (d), administered alone or in the presence of CGS (0.03 μM), SCH (0.1 μM), or both (yellow dots). e Effect of the peptides TM5 and TM7 of the A2AR on the inhibitory effect of WIN (0.03 μM) on the evoked glutamate release, expressed as the difference in % of the control value, alone or in the presence of CGS (0.03 μM) or SCH (0.1 μM). f Effect of CGS (0.03 μM), SCH (0.1 and 10 μM), or SCH (10 μM) plus WIN (0.3 μM). b–f Values are means ± S.E.M. (n = 7–14 with duplicates) of the evoked glutamate release, expressed as the difference in % of the control value, and analyzed statistically with one-sample t-test (*: p < 0.05/3; see “Methods”) Köfalvi et al. BMC Biology (2020) 18:9 Page 13 of 21 212-2 was shifted to the right both by the A2AR agonist to inhibit evoked glutamate release. Indeed, at 10 μM CGS21680 (EC50, 36.8 nM) and the A2AR antagonist (but not 0.1 μM), SCH58261 significantly inhibited the SCH58261 (EC50, 40.0 nM) alone, but not when co- evoked glutamate release, which could not be further administered (EC50, 8.2 nM). In detail, CGS21680 (0.03 inhibited by a relatively high concentration of WIN55212- μM) and SCH58261 (0.1 μM) fully prevented the inhib- 2 (0.3 μM; Fig. 6f). These results therefore indicate that a ition by 0.01 and 0.03 μM, but not by 0.1 μM of WIN55, main mechanism by which CB1R localized in striatal 212-2 of glutamate release (Fig. 6a–d). On the other nerve terminals modulate glutamate release is by counter- hand, co-application of both CGS21680 (0.03 μM) and acting a constitutive A2AR-mediated facilitation of glu- SCH58261 (0.1 μM) did not significantly modify the ef- tamate release by a canonical Gi-Gs antagonistic fect of WIN55,212-2 (Fig. 6a–d). interaction dependent on A2AR-CB1R heteromerization. Striking results were obtained when evaluating the ef- The existence of a significant constitutive activity of fect of the destabilizing A2AR-CB1R heteromer peptides. A2AR in the A2AR-CB1R heterotetramer was confirmed TM5, but not TM7, of A2AR significantly blocked with experiments in HEK-293T cells transiently trans- WIN55,212-2-induced inhibition of glutamate release fected with A2AR alone or with additional transfection and no additional change was observed with the applica- with CB1R, A1R, or D2R. A2AR showed a high constitu- tion of the A2AR ligands (Fig. 6e). These results would tive activity (about 250% increase in cAMP formation in indicate that WIN55,212-2-induced inhibition of glu- cells only transfected with A2AR, as compared with cells tamate release is dependent on A2AR-CB1R heteromeri- transfected with an empty control vector; see “Methods”) zation and would confirm the initial hypothesis of the (Fig. 7; Additional file 2: Data values 4). Very dramatic permissive role of A2AR for CB1R signaling in the differential results were obtained when A2AR was co- A2AR-CB1R heteromer [18]. But the results would also transfected with either CB1R, A1R, or D2R. Thus, A2AR indicate that the permissive role of A2AR would not rely lost its constitutive activity when co-transfected with on a ligand-dependent A2AR activation. A1R or D2R, but it was not significantly modified with Our assumption about presynaptic A2AR-CB1R het- co-transfection with CB1R (Fig. 7; Additional file 2: Data eromers determining all interactions between A2AR and values 4). Therefore, these results indicate that the CB1R ligands at the level of striatal glutamatergic trans- mission bears the question of how the activation of A2AR can be needed for CB1R signaling while, at the same time, A2AR ligands significantly decrease the po- tency of CB1R ligands. A possible answer to this appar- ent conundrum is the existence of a significant constitutive activity of A2AR in the A2AR-CB1R hetero- mer, which would produce a constitutive activation of AC. A CB1R receptor ligand, including endocannabi- noids, would inhibit A2AR-mediated constitutive AC activation by the canonical Gi-Gs antagonistic inter- action in the absence of A2AR ligands. In fact, a high constitutive activity of A2AR has been previously re- ported [45–47]. This would first imply that the CB1R- mediated inhibition of glutamate release is dependent on a Gi-coupling with or without heteromerization with A2AR. Synaptosomes were incubated for 4 h at 4 °C Fig. 7 Differential A2AR constitutive activity (CA) in different A2AR heteromers. Constitutive activation of AC in cells expressing A2AR with or without PTX (2 μg/ml) before the regular release alone or with A2AR plus CB1R, A1R, or D2R. HEK-293T cells were experiment procedure. While WIN55,212-2 (0.1 μM) transiently transfected with the cDNA of A2AR alone (3 μg) or with significantly inhibited glutamate release in the toxin- the cDNA of A2AR (1.5 μg) and the cDNA of CB1R, A1R, or D2R (1.5 naïve synaptosomes (in mean ± S.E.M and expressed as μg in all cases). CA values are expressed as means ± S.E.M. (n = 5 the difference in % of the control value: 82.9% ± 3.6%, n with quintuplicates in all experiments) of the % of basal values, calculated by dividing the levels of cAMP obtained in each = 6; one-sample t-test: p < 0.05/3; see “Methods”), PTX condition by the levels of cAMP in cells transfected with an empty fully prevented this inhibition (in mean ± S.E.M and vector (pcDNA3, 3 μg). In addition, the values were normalized by expressed as the difference in % of the control value: the levels of cell surface expression of A2AR, which was determined 114.6% ± 6.8, n = 6; one-sample t-test: p > 0.05/3). Fi- by means of SNAP staining (see “Methods”). Statistical analysis was nally, one would also expect that a high concentration of performed by using one-way ANOVA, followed by Tukey’s multiple comparison test (** and ***: p < 0.01 and p < 0.001, respectively, the A2AR inverse agonist, SCH58261 [46], should lead compared with cells only transfected with A2AR) to a significant reduction of the ability of WIN55,212-2 Köfalvi et al. BMC Biology (2020) 18:9 Page 14 of 21 constitutive activity of A2AR is blunted by heteromeriza- significant A2AR-mediated constitutive activation of AC. tion with either A1R or D2R, but not by heteromerization Using the peptide-interfering approach in rat striatal with CB1R; activation of CB1R in the A2AR-CB1R hetero- synaptosomal preparations, we could then identify the tetramer can, nevertheless, blunt the A2AR-mediated con- A2AR-CB1R heteromer in striatal glutamatergic termi- stitutive activation of AC (see “Discussion”). The nals, where it provides the main target by which canna- molecular mechanism of this heteromerization-dependent binoids inhibit glutamate release. The present study control of the constitutive activation of A2AR seems diffi- presents a new conceptual view about the mechanism by cult to rationalize, but we hypothesize that the postulated which cannabinoids presynaptically control striatal exci- interaction between TM 5 of the Gs-bound A2AR proto- tatory neurotransmission. Instead of the classically mer and the C-terminal domain of the Gi-unbound CB1R assumed direct inhibition of glutamate release by CB1R- protomer (Fig. 3c), may help to sustain the constitutive ac- mediated signaling, we demonstrate that cannabinoids tivity of A2AR in the A2AR-CB1R heterotetramer, in con- inhibit glutamate release indirectly, by counteracting the trast to the short C-termini of the A1R or D2R in the stimulation of glutamate release induced by A2AR- A1R-A2AR and A2AR-D2R heterotetramers (Fig. 3a). In mediated signaling. This mechanism is completely fact, TM 5 and TM 6 are the helices that move upon re- dependent on the heteromerization with A2AR, which ceptor activation. provides the frame for the canonical Gi-Gs antagonistic interaction at the AC level. In addition, this mechanism Discussion operates in the absence of an adenosine-mediated activa- In the present study, using a peptide-interfering approach tion of the A2AR, by the ability of CB1R activation to combined with biophysical and biochemical techniques in cancel A2AR-mediated constitutive activation of AC mammalian transfected cells and computational modeling, within the A2AR-CB1R heteromer. We could also dem- we could establish the tetrameric quaternary structure of onstrate that the constitutive activity is also present the A2AR-CB1R heterotetramer, having the same hetero- when the A2AR is not forming heteromers and that it is meric but different homomeric interfaces than those of lost upon heteromerization with A1R or D2R. the A1R-A2AR heterotetramer. The long A2AR-CT We recently provided the proof of concept of a ligand- played a significant role, determining the homomeric independent ability of one of the protomers in a GPCR interface of the CB1R homodimer. The A2AR-CB1R het- heteromer to modify the constitutive activity of the other erotetramer showed the same allosteric interactions and molecularly different protomer. That was the ability of the canonical Gs-Gi antagonistic interaction at the AC level as dopamine D4 receptor (D4R) to modify the constitutive the A2AR-D2R heterotetramer. The utilization of TAT- activity in the D2R-D4R heteromer [32], also localized in fused TM peptides in mammalian transfected cells and striatal glutamatergic terminals, where it mediates a pre- striatal primary neuronal cultures allowed demonstrating synaptic dopaminergic inhibitory control of glutamate re- that the canonical interaction at the AC level between lease [50–52]. This effect was even dependent on the A2AR and D2R ligands required the right quaternary polymorphic variant of the D4R, demonstrating for the structure of the A2AR-D2R heterotetramer in the A2AR- first time a functional difference between the two most D2R-AC5 complex. Application of TM peptides that spe- common D4R polymorphic variants, D4.4 and D4.7 [52]. cifically interfere with the heteromeric interface or with The D4.7R-mediated increase in the constitutive activity the receptor-AC5 interface specifically counteracted the of the D2R could explain the recently reported gain of ability of a D2R agonist to inhibit AC5 activation by an function of the D2R in the control of striatal glutamate A2AR agonist, but not with the ability of D2R ligands to release when forming heteromers with the ADHD- counteract forskolin-induced AC activation [7]. More gen- associated polymorphic variant D4.7R [52]. The constitu- erally, these results indicate that to inhibit a Gs-coupled tive activity of the D2R in the D2R-D4R heteromers and receptor-mediated AC activation, the ligands need to sim- the constitutive activity of the A2AR in the A2AR-CB1R ultaneously interact with heterotetramers of the corre- seem therefore to determine the basal degree of sensitivity sponding Gs-coupled and Gi-coupled receptors. This has of striatal glutamatergic terminals to release glutamate been so far shown for the A2AR-D2R [7], the A1R-D1R upon depolarization. [48], the D1R-D3R [49], and now the A2AR-CB1R hetero- The two A2AR heteromers, A2AR-CB1R and A1R- tetramers, where destabilization of their heteromeric A2AR, allow an adenosine-mediated fine-tuned modula- interface leads to the disruption of the canonical Gs-Gi tion of presynaptic striatal glutamatergic transmission, antagonistic interaction at the AC level without disruption which is moderated by the extracellular levels of endo- of the Gi-coupled receptor-mediated inhibition of cannabinoids [2] (Additional file 1: Figure S5). First, the forskolin-induced AC activation. A2AR loses its constitutive activity when it forms het- Different from the A2AR-D2R and A1R-A2AR hetero- eromers with A1R and adenosine binds with more affin- tetramers, the A2AR-CB1R heterotetramer showed a ity to A1R than to A2AR. As mentioned in the Köfalvi et al. BMC Biology (2020) 18:9 Page 15 of 21 “Background”, low extracellular concentrations of ad- three different GPCRs [57]. Experiments are in progress enosine should predominantly activate A1R within the to determine if the striatal Gq protein coupling of post- A1R-A2AR heteromer, leading to a predominant inhib- synaptic A2AR and CB1R depends on their ability to ition of glutamate release (Additional file 1: Figure S5a). separately or simultaneously form heteromers with This would represent a mechanism of silencing corticos- mGlu5R. triatal “noise”. Retrograde endocannabinoid release fol- The important role of the A2AR-CB1R heteromer in lowing postsynaptic metabotropic glutamate receptor the inhibitory control of corticostriatal glutamate release activation [25] would further guarantee the low noise, by makes it a potential target for neuropsychiatric disorders counteracting the A2AR-mediated constitutive activation putatively associated with increased corticostriatal trans- of AC by the A2AR-CB1R heteromer (Additional file 1: mission, which includes obsessive-compulsive disorder Figure S5a). Upon strong corticostriatal activity, the in- [58], schizophrenia [59], and substance use disorders crease in the presynaptic release and degradation of [60]. In addition, changes in the stoichiometry of the ATP would create locally high levels of adenosine, which presynaptic A1R-A2AR and A2A-CB1R receptor hetero- in turn would activate A2AR to shut down A1R signal- mers could have important implications in the patho- ing in the A1R-A2AR heteromer [3, 6] and to counteract genesis of neuropsychiatric disorders. For instance, in CB1R signaling in the A2AR-CB1R heteromer, switching Restless Legs Syndrome (RLS), there is recent preclinical the effect of adenosine on glutamate release from inhib- evidence for the existence of downregulation of striatal ition to facilitation (Additional file 1: Figure S5b). This A1R [61], which promotes an increased sensitivity of high-pass filter is expected to increase the reliability of corticostriatal terminals to release glutamate [51, 62]. information passage and should be maximized under This should be associated with a relative functional up- conditions of low extracellular concentrations of endo- regulation of presynaptic A2AR due to an increase in cannabinoids [2]. the proportion of A2AR not forming heteromers or The present study seems to be in contradiction with forming heteromers with CB1R, both endowed with con- our recent previous study, where A2AR-CB1R hetero- stitutive activity. As expected, the increased sensitivity of mers were described to be preferentially coupled to Gq corticostriatal terminals observed in the experimental and postsynaptically localized in the striatum, in animal could be counteracted by blocking the constitu- GABAergic striatopallidal neurons [53], while the tive signaling of A2AR, with A2AR inverse agonists or present study shows their presynaptic localization in glu- CB1R agonists, or by activating A1R, by moderately in- tamatergic terminals and their preferential coupling to creasing extracellular levels of adenosine with an inhibi- Gs and Gi proteins. Although CB1R couples preferen- tor of the adenosine transporter. There is in fact tially to Gi/o proteins, its ability to couple to Gq has preliminary clinical evidence for the successful applica- already been reported and shown to be ligand- tion of these strategies in RLS patients [63–65]. dependent [53]. Thus, WIN55,212-2 was found to be Finally, more studies need to be done to evaluate the more and less potent than CP55940 at eliciting Gq- possible existence of A2AR-CB1R heteromers, and dependent intracellular calcium increase and at inhibit- therefore, the same type of adenosine-cannabinoid- ing forskolin-induced cAMP production, respectively mediated modulation of glutamate release, in other areas [53, 54]. Nevertheless, despite its relative functional se- of the CNS. It is very plausible that A2AR-CB1R hetero- lectivity for Gq protein-mediated signaling, in the mers modulate neurotransmitter release in many other present study, WIN55,212-2 was shown to inhibit stri- brain areas, such as the hippocampus. Those heteromers atal glutamate release by a PTX-sensitive mechanism. could mediate the reported ability of A2AR antagonists Therefore, the results of our previous and present stud- to prevent cannabinoid-induced memory and LTP im- ies indicate that the preference for a specific G protein pairments [66, 67]. coupling of the A2AR-CB1R heteromer depends mostly on the cellular environment. They also suggest that Gq Conclusions coupling depends on an additional component that is We demonstrate that the ability of cannabinoids to con- present in immortalized striatal neuroblasts and trol excitatory transmission in the striatum depends on GABAergic striatopallidal neurons (previous study) and adenosine transmission, which is mediated by hetero- absent in HEK-293T cells and glutamatergic terminals mers of specific subtypes of adenosine and cannabinoid (present study). One very plausible candidate is the receptors, A2AR and CB1R. The molecular analysis of canonical Gq-coupled glutamate mGlu5 metabotropic the A2AR-CB1R heteromers supports our hypothesis receptor, which is abundantly expressed postsynapyti- about GPCR heteromers being often constituted by het- cally in the striatum [25]. Thus, mGlu5R was previously eromers of homodimers [68]. It also supports that these reported to form functional heteromers with A2AR [55, GPCR heterotetramers often include Gs- and Gi- 56] and suggested to form complexes that include the coupled homodimers, providing the frame for the Köfalvi et al. BMC Biology (2020) 18:9 Page 16 of 21 canonical Gs-Gi antagonistic interaction at the adenylyl TAT fused to a TM GPCR peptide can be inserted ef- cyclase level, forming part of GPCR signaling complexes fectively into the plasma membrane as a result of both that include GPCR heteromers and their common inter- the penetration capacity of the TAT peptide and the acting effectors [7, 69]. The comparison of the A2AR- hydrophobic property of the TM peptide [72]. To obtain CB1 receptor with other heterotetramers of A2AR added the right orientation of the membrane-inserted peptide, new information about the properties of GPCRs. It re- HIV TAT peptide was fused to the C-terminus of vealed that different heteromeric partners of A2AR de- peptides with the amino acid sequence of TM 1, TM 3, termine profound differences in their pharmacological TM 5, and TM 7 of CB1R, A2AR, or A1R (TM1, TM3, properties. This implies the need to consider the imme- TM5, and TM7 peptides, respectively) or to the N- diate context of a GPCR to truly understand its pharma- terminus of TM 2, TM 4, and TM 6 of CB1R, A2AR, or cological properties and, therefore, its role in drug A1R (TM2, TM4, and TM6 peptides, respectively). All development. peptides were synthesized by Genemed Synthesis, Inc. Their sequences were as follows: Methods Expression vectors and fusion proteins VYITVELAIAVLAILGNVLVCWAVWYGRKKRRQRRR Sequences encoding amino acids (aa) residues 1-229 and for TM1 of A2AR, 230-311 of Rluc (Rluc8 variant) and amino acid residues YGRKKRRQRRRYFVVSLAAADIAVGVLAIPFAITI for 1-155 and 156-238 of YFP (mVenus variant) were sub- TM2 of A2AR, cloned in pcDNA3.1 vector to obtain complementary LFIACFVLVLTQSSIFSLLAIAIYGRKKRRQRRR for TM3 Rluc and YFP hemi-truncated proteins (nRluc, cRluc, of A2AR, nYFP, and cYFP). The cDNAs for human CB1R, A2AR, YGRKKRRQRRRAKGIIAICWVLSFAIGLTPMLGW for A1R, D1R, D2R (short isoform), and 5-HT2AR cloned TM4 of A2AR, into pcDNA3.1 were amplified without their stop codons MNYMVYFNFFACVLVPLLLMLGVYLYGRKKRRQRR R using sense and antisense primers harboring EcoRI and for TM5 of A2AR, BamHI sites to clone A2AR and D1R or EcoRI and KpnI YGRKKRRQRRRLAIIVGLFALCWLPLHIINCFTFF for T to clone A1R and CB1R. The amplified fragments were M6 of A2AR, subcloned to be in-frame with restriction sites of LWLMYLAIVLSHTNSVVNPFIYAYYGRKKRRQRRR for pcDNA3.1-nYFP, pcDNA3.1-cYFP, pcDNA3.1-nRluc, or TM7 of A2AR. pcDNA3.1-Rluc vectors to provide plasmids that express LAIAVLSLTLGTFTVLENLLVLCVILYGRKKRRQRRR for the receptor fused to nYFP, cYFP, nRluc, or cRluc on TM1 of CB1R, the C-terminal end of the receptor. The following hu- YGRKKRRQRRRFIGSLAVADLLGSVIFVYSFI for TM2 of man G protein constructs were used: Gαi1-YFP (with CB1R, FIGSLAVADLLGSVIFVYSFIYGRKKRRQRRR for YFP, mVenus variant, inserted at position 91), Gαs-YFP TM3 of CB1R, (with YFP, mVenus variant, inserted at position 154 of YGRKKRRQRRRAVVAFCLMWTIAIVIAVLPLLGW for Gαs, short isoform), and Gαq-YFP (with YFP, mVenus TM4 of CB1R, variant, inserted at position 97 of Gαs), untagged Gβ1, TYLMFWIGVTSVLLLFIVYAYMYILWYGRKKRRQRRR and untagged Gγ2. All the constructs were confirmed by for TM5 of CB1R, sequencing analysis. Several constructs were shared by YGRKKRRQRRRLVLILVVLIICWGPLLAIMVY for TM6 C. Gales at INSERM (Toulouse, France; Gαi1 construct) of CB1R, and N. Lambert (Georgia Regents University, Augusta, VFAFCSMLCLLNSTVNPIIYALYGRKKRRQRRR for T GA; Gαs). The expression of constructs was tested by M7 of CB1R confocal microscopy and the receptor-fusion protein RRRQRRKKRGYAAVAIAGCWILSFVVGLTPMFGW for functionality by ERK1/2 phosphorylation. An A2AR mu- TM4 of A1R, tant with a deletion of aa 321 to aa 412 on the C- MEYMVYFNFFVWVLPPLLLMVLIYLYGRKKRRQRRR terminal domain of A2AR (A2AΔCTR) was generated as for TM5 of A1R, previously described [70]. RRRQRRKKRGYLALILFLFALSWLPLHILNCITLF for TM6 of A1R, HIV TAT-fused TM peptides ILTYIAIFLTHGNSAMNPIVYAFRIYGRKKRRQRRR for Peptides with the amino acid sequence of TMs of the TM7 of A1R. CB1R, A2AR, and A1R were used as oligomer- destabilizing agents, as previously demonstrated [6, 7, Cell cultures and transient transfection 12, 31, 71]. To allow intracellular delivery, a peptide can Human embryonic kidney 293T (HEK-293T) cells ob- be fused to the cell-penetrating HIV transactivator of tained from ATCC were grown in Dulbecco’s modified transcription (TAT) peptide (YGRKKRRQRRR). HIV Eagle’s medium (DMEM) (Gibco, Gaithersburg, MD) Köfalvi et al. BMC Biology (2020) 18:9 Page 17 of 21 supplemented with 2 mM L-glutamine, 100 μg/ml so- only the BRET donor. For BRET measurements, cells (20 dium pyruvate, 100 U/ml penicillin/streptomycin, MEM μg protein) were distributed in 96-well microplates (Corn- non-essential amino acid solution (1/100), and 5% (v/v) ing 3600, White plates; Sigma-Aldrich) and BRET signal heat inactivated fetal bovine serum (FBS) (all supple- was collected 1 min after addition of 5 μM coelenterazine ments were from Invitrogen, Paisley, Scotland, UK). H (Molecular Probes, Eugene, OR) using a Mithras LB Cells were maintained at 37 °C in an atmosphere of 5% 940 microplate reader (Berthold Technologies, Bad Wild- CO2. For transient transfection, HEK-293T cells growing bad, Germany), which allows integration of the signals de- in six-well dishes were transfected with the correspond- tected in the short-wavelength filter at 485 nm (440–500 ing fusion protein cDNA by the PEI (PolyEthylenImine, nm) and the long-wavelength filter at 530 nm (510–590 Sigma-Aldrich, Saint Louis, MO) method. Cells were in- nm). To quantify receptor-Rluc reconstitution, lumines- cubated (4 h) with the corresponding cDNA together cence readings were also performed after 10 min of adding with PEI (5.47 mM in nitrogen residues) and 150 mM 5 μM of coelenterazine H. Both the fluorescence and lu- NaCl in a serum-starved medium. After 4 h, the medium minescence of each sample were measured before each was changed to a fresh complete culture medium. Forty- experiment to confirm similar donor expression (∼150,000 eight hours after transfection, cells were washed twice in luminescent units). Net BRET is defined as [(long-wave- quick succession in Hank’s balanced salt solution (HBSS; length emission)/(short-wavelength emission)]-Cf where containing, in mM: 137 NaCl, 5 KCl, 0.34 Na2HPO4, Cf corresponds to [(long-wavelength emission)/(short- 0.44 KH2PO4, 1.26 CaCl2, 0.4 MgSO4, 0.5 MgCl2, 10 wavelength emission)] for the Rluc construct expressed HEPES, pH 7.4), supplemented with 0.1% glucose (w/v), alone in the same experiment. BRET is expressed as milli detached, and resuspended in the same buffer. To con- BRET units (mBU; net BRET × 1000). trol the cell number, sample protein concentration was determined using a Bradford assay kit (Bio-Rad, Munich, Complemented donor-acceptor resonance energy transfer Germany) using bovine serum albumin dilutions as (CODA-RET) standards. HEK-293T cells transfected with A2AR fused to cRluc and CB1R fused to nRluc, or 5-HT2AR fused to cRluc Bimolecular fluorescence complementation (BiFC) and D2R (short isoform) fused to nRluc, and co- HEK-293T cells transfected with the receptor fused to n- transfected with either Gαs, Gαi1, or Gαq fused to YFP YFP and the receptor fused to the cYFP were treated with were harvested, washed, and resuspended in phosphate- vehicle or the indicated TAT-fused TM peptides (4 μM) buffered saline (PBS). About 200,000 cells/well were dis- for 4 h at 37 °C. To quantify protein-reconstituted YFP tributed in 96-well plates, and 5 μM coelenterazine H Venus expression, cells (20 μg protein) were distributed in was added to each well. One minute after addition of 96-well microplates (black plates with a transparent bot- coelenterazine, different concentrations of the non- tom; Porvair, King’s Lynn, UK), and emission fluorescence selective adenosine receptor agonist NECA, the selective at 530 nm was read in a Fluo Star Optima Fluorimeter CB1R agonist CP55940, dopamine, or serotonin were (BMG Labtechnologies, Offenburg, Germany) equipped added to each well. In some experiments, NECA or caf- with a high-energy xenon flash lamp, using a 10-nm- feine were added 12 min before the addition of bandwidth excitation filter at 400 nm reading. Protein CP55940. Fluorescence of the acceptor was quantified fluorescence was determined as fluorescence of the sam- (excitation at 500 nm and emission at 540 nm for 1-s re- ple minus fluorescence of non-transfected cells. Cells ex- cording) in the Mithras LB940 reader to confirm the pressing CB1R, A2AR, A1R, or D1R fused to nYFP or to constant expression level across experiments. In parallel, cYFP, as well as cells expressing nYFP or cYFP or both BRET signal from the same batch of cells was deter- complementary proteins (not fused to receptors) showed mined as the ratio of the light emitted by YFP (mVenus similar fluorescence levels to non-transfected cells. variant; 510–540 nm) over Rluc (485 nm). Results were calculated for the BRET change (BRET ratio for the cor- Bioluminescence resonance energy transfer (BRET) with responding drug minus BRET ratio in the absence of the donor and acceptor complementation drug) 10 min after addition of the agonists. Data manip- HEK-293T cells were transfected with the corresponding ulations and statistical analyses for all CODA-RET, receptors fused to nYFP, cYFP, nRluc, and cRluc (see BRET, and complementation experiments were per- figure legends). To quantify receptor-YFP expression, formed with GraphPad Prism 7.0 software (La Jolla, CA). cells (20 μg protein) were distributed in 96-well micro- plates (black plates with a transparent bottom) and cAMP accumulation fluorescence at 530 nm was read as described above. Re- cAMP accumulation was measured using the LANCE ceptor fluorescence was determined as fluorescence of Ultra cAMP kit (PerkinElmer, Waltham, MA, USA) as the sample minus the fluorescence of cells expressing previously described [73]. In brief, HEK-293T cells were Köfalvi et al. BMC Biology (2020) 18:9 Page 18 of 21 seeded in white 384-wells plates in Dulbecco’s modified inactive A2AR and CB1R were modeled as observed on Eagle’s medium (DMEM) (Sigma-Aldrich) containing the structure of the μ-opioid receptor to facilitate the zardaverine (up to 50 μM; Calbiochem, San Diego, CA, formation of the highly packed TM 5/6 interface. The USA) and, when using A2AR ligands, adenosine deami- A2AR and CB1R homodimers were modeled via the TM nase (ADA, 0.5 U/ml; Roche Diagnostics, GmbH, Mann- 4/6 interface, using the structure of CB2R with PDB id heim, Germany). Cells were incubated for 1 h at 37 °C 5ZTY [85, 86]. The large outward movement of TM 6 and for 4 h at 37 °C when using TAT-fused TM pep- for G protein coupling is not compatible with the TM 4/ tides. Subsequently, selective ligands and/or forskolin 6 interface observed in this structure due to steric were added and cells were incubated for 30 min. Finally, clashes. To avoid these clashes, the active protomer was Eu-cAMP tracer and ULight™-anti-cAMP reagents were manually rotated relative to the inactive protomer. The prepared and added to the sample following the manu- resulting tetrameric complex was refined using a 800-ns facturer’s instructions. The 384-wells plate was incu- molecular dynamics simulation (as described in detail bated 1 h at 22 °C in the dark and was then read on a elsewhere [6]). CLARIOstar or PHERAstar microplate reader (BMG Labtech, Durham, NC, USA). Measurements at 620 nm [14C]glutamate release assay from striatal glutamatergic and 665 nm were used to detect the TR-FRET signal terminals and the concomitant cAMP levels were calculated fol- Male Wistar rats (180–240 g, 8-10 weeks old) were pur- lowing the manufacturer’s instructions. Data were fitted chased from Charles-River (Barcelona, Spain), and by non-linear regression using GraphPad Prism 6.01 housed with 12-h light on/off cycles under controlled (San Diego, CA, USA). When comparing the effect of temperature (23 ± 2 °C) and ad libitum access to food different co-expression experiments, cells were labeled and water. Before brain extraction, the rats were deeply by means of SNAP staining as previously described [47] anesthetized with halothane (5%, 1 l/min flow rate; no and cAMP values were normalized according to the reaction to tail pinch or handling, while still breathing). levels of cell surface expression of A2AR in each condi- Experiments were carried out as previously described tion (A2AR, A2AR + A1R, A2AR + CB1R, A2AR + [19]. Briefly, the two striata were rapidly dissected in ice- D2R). In brief, adherent cells expressing the A2ARSNAP cold Krebs’ solution (in mM: NaCl 132, KCl 3, KH2PO4 were washed and incubated with DMEM containing 100 1.2, MgSO4 1.2, CaCl2 2.5, NaHCO3 25, glucose 5.5, nM of SNAP-surface 488 substrate (New England Bio- HEPES 10; pH 7.4) and moved into 2 ml of ice-cold su- Labs, Ipswich, MA) for 1 h at 37 °C. Cells were then crose solution (0.32 M, containing 15 mM HEPES; pH washed three times with phenol red-free Hank’s bal- 7.4) for homogenization with a Teflon homogenizer anced salt solution containing 1 g/l glucose (HBSS: 137 (Thomas Scientific, NJ, USA). The heavy debris particles mM NaCl, 5.4 mM KCl, 0.3 mM Na2HPO4, 0.4 mM in the homogenate were decanted at 1000g for 10 min at KH2PO4, 4.2 mM NaHCO3, 1.3 mM CaCl2, 0.5 mM 4 °C, and the supernatant was saved. The first pellet (P1) MgCl2, 0.6 mM MgSO4, 5.6 mM glucose, pH 7.4) and was resuspended once again and centrifuged as above. fluorescence read on the CLARIOstar microplate reader Subsequently, the supernatants from the two centrifuga- (BMG Labtech). tions were pelleted at 20,000g for 30 min, to obtain the P2 crude synaptosomal fraction. The two P2 pellets (syn- Computational model of the A2AR-CB1R heterotetramer aptosomes) from each rat were subsequently combined The structural model of the A2AR-CB1R heterotetramer and stored on ice until use. For the assays with pertussis consists of a heteromer of homodimers. The A2AR ho- toxin (PTX), synaptosomes from each animal were re- modimer is formed by an internal inactive protomer suspended in 2 ml of Krebs-HEPES solution, pregassed (interacting with CB1R), modeled with PDB id 5NM4 with 95% O2 and 5% CO2, containing PTX (2 μg/ml) [74, 75], and an external active protomer bound to Gs, and left tightly sealed at 4 °C for 4 h, then pelleted again modeled with PDB id 5G53 of A2AR in complex with a for [3H]glutamate loading. Hereafter, all assay solutions mini-Gs α-subunit [76, 77]. Gs was modeled based on contained the glutamate decarboxylase inhibitor, ami- the mini-Gs α-subunit and the βγ-subunits of the nooxyacetic acid (100 μM), to prevent the transfer of A2AR-Gs structure with PDB id 6GDG [78, 79]. The [3H] labels to releasable molecules other than glutamate. CB1R dimer is formed by an internal inactive protomer Before the release experiments, synaptosomes from each (interacting with A2AR), modeled with PDB id 5U09 animals were resuspended in 0.5 ml pregassed Krebs- [80, 81], and an external active protomer bound to Gi, HEPES solution and incubated in the presence of modeled with PDB id 6N4B of the CB1R-Gi complex [3H]glutamate (specific activity, 60 Ci/mmol; final con- [82, 83]. The A2AR-CB1R heteromer was modeled via centration, 200 nM; American Radiolabeled Chemicals the TM 5/6 interface, using the structure of the μ-opioid Inc, Saint Louis, MO) for 15 min. When necessary, incu- receptor with PDB id 4DKL [39, 84]. TMs 5 and 6 of bations with TM5 (20 μM) or with TM7 (20 μM) of Köfalvi et al. BMC Biology (2020) 18:9 Page 19 of 21 A2AR occurred also during this period. Subsequently, a Funding 8-microvolume chamber superfusion setup was filled Work supported with the intramural funds of the National Institute on Drug Abuse, “Ministerio de Ciencia, Innovación y Universidades-Agencia Estatal de with preloaded synaptosomes (chambers/rat) which were Investigación/FEDER” (SAF2015-74627-JIN, SAF2016-77830-R, SAF2017-87349- trapped by layers of Whatman GF/B filters (Sigma-Al- R) and ISCIII/FEDER (PIE14/00034), the Catalan government (2017 SGR 1604), drich). Synaptosomes were superfused continuously at a “Fundació la Marató de TV3” (20152031), FWO (SBO-140028), “Fundação para a Ciência e a Tecnologia” (PTDC/DTP-FTO/3346/2014 and PTDC/MED-NEU/ rate of 0.8 ml/min with pre-gassed Krebs-HEPES solu- 31274/2017), FEDER (QREN) through “Programa Operacional Factores de tion (37 °C) until the end of the experiments. Upon ter- Competitividade” (COMPETE 2020, POCI-01-0145-FEDER-007440, CENTRO-01- mination of a 10-min washout and after collecting three 0145-FEDER-000012-N2323P30, and UID/NEU/04539/2019), and through “Pro- grama Mais Centro” (CENTRO-01-0246-FEDER-000010 and CENTRO-07-ST24- 2-min samples as baseline, neurotransmitter release was FEDER-002006). stimulated twice with 30 mM KCl (S1, S2) with 10-min interval (Fig. 6a). The CB1R agonist WIN55,212-2 Availability of data and materials All data generated or analyzed during this study are included in this published (Tocris Bioscience, Bristol, UK), A2AR ligands (Tocris article and its supplementary information files (Additional files 1 and 2). The Bioscience), or their solvent, DMSO (0.1% v/v), were crystal structures 5NM4, 5G53, 6GDG, 5UO9, 6N4B, 4DKL, and 5ZTY, used to added to the superfusion medium 4 min before S2 and build the computational models, are available from Protein Data Bank (PDB; http://www.rcsb.org; specific PDB entries included in the reference list). continued to be present until the end of the experiment (that is, there was no washout study). Ethics approval Treatments were performed in duplicate (i.e., 1 pair of All studies with animal preparations were conducted in accordance with the control versus 3 pairs of different treatments per animal), principles and procedures outlined as “3Rs” in the guidelines of EU (86/609/ EEC), FELASA, and the National Centre for the 3Rs (Kilkenny et al., Brit. J. and the intra-chamber S2/S1 ratios served to evaluate the Pharmacol.; 2010, 160: 1577-1579), and were approved by the Animal Care effect of drug treatments and expressed as % of DMSO Committee of the Center for Neuroscience and Cell Biology of the University control (Fig. 6a). All data are represented as means ± of Coimbra, Portugal. All efforts were made to minimize the number of ani- mals used and to minimize their stress and discomfort. S.E.M. of “n ≥ 5” observations (rats) in duplicates. As control S2/S1 ratios do not have biological significance, Consent for publication they were taken as 100% in each experiment, and treat- Not applicable ment S2/S1 ratios were normalized to them. Normalized Competing interests data were then tested for normal distribution by the The authors declare that they have no competing interests. Kolmogorov-Smirnov test, and subsequently analyzed Author details with one-sample t-test against the hypothetical value of 1 CNC-Center for Neuroscience and Cell Biology, University of Coimbra, 100 (%), and p < 0.05/3 (a correction factor for three 3004-504 Coimbra, Portugal. 2Department of Biochemistry and Molecular treatments using one control) was accepted for signifi- Biomedicine, Faculty of Biology, and Institute of Biomedicine, University of Barcelona, 08028 Barcelona, Spain. 3Laboratori de Medicina Computacional, cant difference. Tests were performed using the Graph- Unitat de Bioestadística, Facultat de Medicina, Universitat Autònoma de Pad Prism 5.0 software. Barcelona, 08193 Bellaterra, Spain. 4Integrative Neurobiology Section, National Institute on Drug Abuse, Intramural Research Program, National Institutes of Health, Baltimore, MD 21224, USA. 5Unitat de Farmacologia, Supplementary information Departament Patologia i Terapèutica Experimental, Facultat de Medicina, Supplementary information accompanies this paper at https://doi.org/10. IDIBELL, Universitat de Barcelona, L’Hospitalet de Llobregat, Spain. 6Institut de 1186/s12915-020-0739-0. Neurociències, Universitat de Barcelona, Barcelona, Spain. 7Faculty of Medicine, University of Coimbra, Coimbra, Portugal. 8Instituto de Additional file 1: Figure S1. G protein coupling of A2AR and CB1R in Farmacologia e Neurociências, Faculdade de Medicina, Universidade de the A2AR-CB1R heteromer. Figure S2. Lack of modulation by the CB1R Lisboa, Lisbon, Portugal. 9Instituto de Medicina Molecular, Faculdade de agonist CP55940 on A2AR-mediated Gs protein activation in the A2AR- Medicina, Universidade de Lisboa, Lisbon, Portugal. CB1R heteromer. Figure S3. Tetrameric structure of A2AR-CB1R hetero- mer. Figure S4. Gs-dependent A2AR-mediated modulation and Gi- Received: 10 October 2019 Accepted: 13 January 2020 dependent CB1R-mediated modulation of AC signaling in HEK-293T cells. Figure S5. Control of striatal glutamate release by A2AR-CB1R and A1R- A2AR heteromers. References 1. Ferré S, Lluís C, Justinova Z, Quiroz C, Orru M, Navarro G, et al. Adenosine- Additional file 2. Data values 1. EC50 values from Additional file 1: cannabinoid receptor interactions. Implications for striatal function. Br J Figure S1. Data values 2. EC50 and Emax values from Additional file 1: Pharmacol. 2010;160:443–53. Figure S2. Data values 3. Values of cAMP formation from Fig. 5. Data 2. Cunha RA. How does adenosine control neuronal dysfunction and values 4. Values of constitutive activation (cAMP formation) from Fig. 7. neurodegeneration? J Neurochem. 2016;139(6):1019–55. 3. Ciruela F, Casadó V, Rodrigues RJ, Luján R, Burgueño J, Canals M, et al. Presynaptic control of striatal glutamatergic neurotransmission by Acknowledgements adenosine A1-A2A receptor heteromers. J Neurosci. 2006;26:2080–7. Not applicable 4. Piomelli D. The molecular logic of endocannabinoid signalling. Nat Rev Neurosci. 2003;4:873–84. Authors’ contributions 5. Fredholm BB, Irenius E, Kull B, Schulte G. Comparison of the potency of AK, EM, AC, N-SC, VF-D, SGF, MS-S, and VC-A. conducted the experiments. AK, adenosine as an agonist at human adenosine receptors expressed in EM, AC, N-SC, VF-D, SGF, MS-S, FC, LP, VC, and SF analyzed the results. AK, Chinese hamster ovary cells. Biochem Pharmacol. 2001;1:443–8. EM, AC, HY, FC, LP, VC, and SF designed the experiments. All authors contrib- 6. Navarro G, Cordomí A, Brugarolas M, Moreno E, Aguinaga D, Pérez-Benito L, uted to the writing and read and approved the final manuscript. et al. Cross-communication between G(i) and G(s) in a G-protein-coupled Köfalvi et al. BMC Biology (2020) 18:9 Page 20 of 21 receptor heterotetramer guided by a receptor C-terminal domain. BMC Biol. CB(1) Receptor Heteromers in the Dorsal Striatum. 2018;16:24. Neuropsychopharmacology. 2018;43:964–77. 7. Navarro G, Cordomí A, Casadó-Anguera V, Moreno E, Cai NS, Cortés A, et al. 30. Urizar E, Yano H, Kolster R, Galés C, Lambert N, Javitch JA. CODA-RET reveals Evidence for functional pre-coupled complexes of receptor heteromers and functional selectivity as a result of GPCR heteromerization. Nat Chem Biol. adenylyl cyclase. Nat Commun. 2018;9:1242. 2011;7:624–30. 8. Ferré S, Fredholm BB, Morelli M, Popoli P, Fuxe K. Adenosine-dopamine 31. Guitart X, Navarro G, Moreno E, Yano H, Cai NS, Sánchez-Soto M, et al. receptor-receptor interactions as an integrative mechanism in the basal Functional selectivity of allosteric interactions within G protein-coupled ganglia. Trends Neurosci. 1997;20:482–7. receptor oligomers: the dopamine D1-D3 receptor heterotetramer. Mol 9. Ciruela F, Burgueño J, Casadó V, Canals M, Marcellino D, Goldberg SR, et al. Pharmacol. 2014;86:417–29. Combining mass spectrometry and pull-down techniques for the study of 32. Sánchez-Soto M, Yano H, Cai NS, Casadó-Anguera V, Moreno E, Casadó V, receptor heteromerization. Direct epitope-epitope electrostatic interactions et al. Revisiting the functional role of dopamine D(4) receptor gene between adenosine A2A and dopamine D2 receptors. Anal Chem. 2004;76: polymorphisms: heteromerization-dependent gain of function of the D(4.7) 5354–63. receptor variant. Mol Neurobiol. 2019;56:4778–85. 10. Borroto-Escuela DO, Romero-Fernandez W, Tarakanov AO, Gómez-Soler M, 33. Borroto-Escuela DO, Romero-Fernandez W, Tarakanov AO, Marcellino D, Corrales F, Marcellino D, et al. Characterization of the A2AR-D2R interface: Ciruela F, Agnati LF, et al. Dopamine D2 and 5-hydroxytryptamine 5-HT(2A) focus on the role of the C-terminal tail and the transmembrane helices. receptors assemble into functionally interacting heteromers. Biochem Biochem Biophys Res Commun. 2010;402:801–7. Biophys Res Commun. 2010;401:605–10. 11. Navarro G, Ferré S, Cordomi A, Moreno E, Mallol J, Casadó V, et al. 34. Albizu L, Holloway T, González-Maeso J, Sealfon SC. Functional crosstalk and Interactions between intracellular domains as key determinants of the heteromerization of serotonin 5-HT2A and dopamine D2 receptors. quaternary structure and function of receptor heteromers. J Biol Chem. Neuropharmacology. 2011;61:770–7. 2010;285:27346–59. 35. Rose RH, Briddon SJ, Holliday ND. Bimolecular fluorescence 12. Bonaventura J, Navarro G, Casadó-Anguera V, Azdad K, Rea W, Moreno E, complementation: lighting up seven transmembrane domain receptor et al. Allosteric interactions between agonists and antagonists within the signalling networks. Br J Pharmacol. 2010;159:738–50. adenosine A2A receptor-dopamine D2 receptor heterotetramer. Proc Natl 36. Schmidt C, Peng B, Li Z, Sclabas GM, Fujioka S, Niu J, et al. Mechanisms of Acad Sci USA. 2015;112:E3609–18. proinflammatory cytokine-induced biphasic NF-kappaB activation. Mol Cell. 13. Sadana R, Dascal N, Dessauer CW. N terminus of type 5 adenylyl cyclase 2003;12:1287–300. scaffolds Gs heterotrimer. Mol Pharmacol. 2009;76:1256–64. 37. Guo Y, Rebecchi M, Scarlata S. Phospholipase Cbeta2 binds to and inhibits 14. Gilman AG. G proteins: transducers of receptor-generated signals. Annu Rev phospholipase Cdelta1. J Biol Chem. 2005;280:1438–47. Biochem. 1987;56:615–49. 38. Anderie I, Schmid A. In vivo visualization of actin dynamics and actin 15. Dessauer CW, Tesmer JJ, Sprang SR, Gilman AG. Identification of a Gialpha interactions by BiFC. Cell Biol Int. 2007;31:1131–5. binding site on type V adenylyl cyclase. J Biol Chem. 1998;273:25831–9. 39. Manglik A, Kruse AC, Kobilka TS, Thian FS, Mathiesen JM, Sunahara RK, et al. 16. Herlitze S, Garcia DE, Mackie K, Hille B, Scheuer T, Catterall WA. Modulation of Crystal structure of the μ-opioid receptor bound to a morphinan antagonist. Ca2+ channels by G-protein beta gamma subunits. Nature. 1996;380:258–62. Nature. 2012;485:321–6. 17. Wu LG, Saggau P. Presynaptic inhibition of elicited neurotransmitter release. 40. Stenkamp RE. Identifying G protein-coupled receptor dimers from crystal Trends Neurosci. 1997;20:204–12. packings. Acta Crystallogr D Struct Biol. 2018;74:655–70. 18. Carriba P, Ortiz O, Patkar K, Justinova Z, Stroik J, Themann A, et al. Striatal 41. Provasi D, Boz MB, Johnston JM, Filizola M. Preferred supramolecular adenosine A2A and cannabinoid CB1 receptors form functional heteromeric organization and dimer interfaces of opioid receptors from simulated self- complexes that mediate the motor effects of cannabinoids. association. PLoS Comput Biol. 2015;11:e1004148. Neuropsychopharmacology. 2007;32:2249–59. 42. Galés C, Van Durm JJ, Schaak S, Pontier S, Percherancier Y, Audet M, et al. 19. Ferreira SG, Gonçalves FQ, Marques JM, Tomé ÂR, Rodrigues RJ, Nunes- Probing the activation-promoted structural rearrangements in preassembled Correia I, et al. Presynaptic adenosine A2A receptors dampen cannabinoid receptor-G protein complexes. Nat Struct Mol Biol. 2006;13:778–6. CB1 receptor-mediated inhibition of corticostriatal glutamatergic 43. Dror RO, Mildorf TJ, Hilger D, Manglik A, Borhani DW, Arlow DH, et al. transmission. Br J Pharmacol. 2015;172:1074–86. Structural basis for nucleotide exchange in heterotrimeric G proteins. 20. Tebano MT, Martire A, Chiodi V, Pepponi R, Ferrante A, Domenici MR, et al. Science. 2015;348:1361–15. Adenosine A2A receptors enable the synaptic effects of cannabinoid CB1 44. Cristóvão-Ferreira S, Navarro G, Brugarolas M, Pérez-Capote K, Vaz SH, receptors in the rodent striatum. J Neurochem. 2009;110:1921–30. Fattorini G, et al. A1R-A2AR heteromers coupled to Gs and G i/0 proteins 21. Chiodi V, Ferrante A, Ferraro L, Potenza RL, Armida M, Beggiato S, et al. modulate GABA transport into astrocytes. Purinergic Signal. 2013;9:433–49. Striatal adenosine-cannabinoid receptor interactions in rats over-expressing 45. Klinger M, Kuhn M, Just H, Stefan E, Palmer T, Freissmuth M, et al. Removal adenosine A2A receptors. J Neurochem. 2016;136:907–17. of the carboxy terminus of the A2A-adenosine receptor blunts constitutive 22. Fredholm BB, Bättig K, Holmén J, Nehlig A, Zvartau EE. Actions of caffeine in activity: differential effect on cAMP accumulation and MAP kinase the brain with special reference to factors that contribute to its widespread stimulation. Naunyn Schmiedebergs Arch Pharmacol. 2002;366:287–98. use. Pharmacol Rev. 1999;51:83–133. 46. Bennett KA, Tehan B, Lebon G, Tate CG, Weir M, Marshall FH, et al. 23. Ferré S. Mechanisms of the psychostimulant effects of caffeine: implications Pharmacology and structure of isolated conformations of the adenosine for substance use disorders. Psychopharmacology. 2016;233:1963–79. A2A receptor define ligand efficacy. Mol Pharmacol. 2013;83:949–58. 24. Köfalvi A, et al. Involvement of cannabinoid receptors in the regulation 47. Fernández-Dueñas V, Gómez-Soler M, López-Cano M, Taura JJ, Ledent C, of neurotransmitter release in the rodent striatum: a combined Watanabe M, et al. Uncovering caffeine's adenosine A2A receptor inverse immunochemical and pharmacological analysis. J Neurosci. 2005;25: agonism in experimental parkinsonism. ACS Chem Biol. 2014;9:2496–501. 2874–84. 48. Rivera-Oliver M, Moreno E, Álvarez-Bagnarol Y, Ayala-Santiago C, Cruz-Reyes 25. Uchigashima M, Narushima M, Fukaya M, Katona I, Kano M, Watanabe M. N, Molina-Castro GC, et al. Adenosine A(1)-Dopamine D(1) Receptor Subcellular arrangement of molecules for 2-arachidonoyl-glycerol-mediated Heteromers Control the Excitability of the Spinal Motoneuron. Mol retrograde signaling and its physiological contribution to synaptic Neurobiol. 2019;56:797–811. modulation in the striatum. J Neurosci. 2007;7:3663–76. 49. Guitart X, Moreno E, Rea W, Sánchez-Soto M, Cai NS, Quiroz C, et al. Biased 26. Gerdeman G, Lovinger DM. CB1 cannabinoid receptor inhibits synaptic G Protein-Independent Signaling of Dopamine D(1)-D(3) Receptor release of glutamate in rat dorsolateral striatum. J Neurophysiol. 2001;85: Heteromers in the Nucleus Accumbens. Mol Neurobiol. 2019;56:6756–69. 468–71. 50. González S, Rangel-Barajas C, Peper M, Lorenzo R, Moreno E, Ciruela F, et al. 27. Kreitzer AC, Malenka RC. Endocannabinoid-mediated rescue of striatal LTD Dopamine D4 receptor, but not the ADHD-associated D4.7 variant, forms and motor deficits in Parkinson's disease models. Nature. 2007;445:643–7. functional heteromers with the dopamine D2S receptor in the brain. Mol 28. Gerfen CR, Surmeier DJ. Modulation of striatal projection systems by Psychiatry. 2012;17:650–62. dopamine. Annu Rev Neurosci. 2011;34:4414–66. 51. Yepes G, Guitart X, Rea W, Newman AH, Allen RP, Earley CJ, et al. Targeting 29. Moreno E, Chiarlone A, Medrano M, Puigdellívol M, Bibic L, Howell LA, et al. hypersensitive corticostriatal terminals in restless legs syndrome. Ann Singular Location and Signaling Profile of Adenosine A(2A)-Cannabinoid Neurol. 2017;82:951–60. Köfalvi et al. BMC Biology (2020) 18:9 Page 21 of 21 52. Bonaventura J, Quiroz C, Cai NS, Rubinstein M, Tanda G, Ferré S. Key role of 74. Weinert T, Cheng R, James D, Gashi D, Nogly P, Jaeger K, et al. A2A the dopamine D(4) receptor in the modulation of corticostriatal Adenosine receptor room-temperature structure determined by serial glutamatergic neurotransmission. Sci Adv. 2017;3:e1601631. femtosecond crystallography. Protein Data Bank. https://doi.org/10.2210/ 53. Lauckner JE, Hille B, Mackie K. The cannabinoid agonist WIN55,212-2 pdb5NM4/pdb. increases intracellular calcium via CB1 receptor coupling to Gq/11 G 75. Weinert T, Olieric N, Cheng R, Brünle S, James D, Ozerov D, et al. Serial proteins. Proc Natl Acad Sci USA. 2005;102:19144–9. millisecond crystallography for routine room-temperature structure 54. Felder CC, Veluz JS, Williams HL, Briley EM, Matsuda LA. Cannabinoid determination at synchrotrons. Nat Commun. 2017;8:542. agonists stimulate both receptor- and non-receptor-mediated signal 76. Carpenter B, Nehme R, Warne T, Leslie AGW, Tate CG. Structure of the transduction pathways in cells transfected with and expressing cannabinoid adenosine A2A receptor bound to an engineered G protein. Protein Data receptor clones. Mol Pharmacol. 1992;42:838–45. Bank. https://doi.org/10.2210/pdb5g53/pdb. 55. Ferré S, Karcz-Kubicha M, Hope BT, Popoli P, Burgueño J, Gutiérrez MA, et al. 77. Carpenter B, Nehmé R, Warne T, Leslie AG, Tate CG. Structure of the Synergistic interaction between adenosine A2A and glutamate mGlu5 adenosine A(2A) receptor bound to an engineered G protein. Nature. 2016; receptors: implications for striatal neuronal function. Proc Natl Acad Sci USA. 536:104–7. 2002;99:11940–5. 78. Garcia-Nafria J, Lee Y. Cryo-EM structure of the adenosine A2A receptor 56. Nishi A, Liu F, Matsuyama S, Hamada M, Higashi H, Nairn AC, et al. bound to a miniGs heterotrimer. Protein Data Bank. https://doi.org/10.2210/ Metabotropic mGlu5 receptors regulate adenosine A2A receptor signaling. pdb6gdg/pdb. Proc Natl Acad Sci USA. 2003;100:1322–7. 79. García-Nafría J, Lee Y, Bai X, Carpenter B, Tate CG. Cryo-EM structure of the 57. Navarro G, Carriba P, Gandía J, Ciruela F, Casadó V, Cortés A, et al. Detection adenosine A(2A) receptor coupled to an engineered heterotrimeric G of heteromers formed by cannabinoid CB1, dopamine D2, and adenosine protein. Elife. 2018;7:e35946. A2A G-protein-coupled receptors by combining bimolecular fluorescence 80. Shao ZH, Yin J, Rosenbaum D. High-resolution crystal structure of the complementation and bioluminescence energy transfer. human CB1 cannabinoid receptor. Protein Data Bank. https://doi.org/10. ScientificWorldJournal. 2008;8:1088–97. 2210/pdb5u09/pdb. 58. Graybiel AM, Rauch SL. Toward a neurobiology of obsessive-compulsive 81. Shao Z, Yin J, Chapman K, Grzemska M, Clark L, Wang J, et al. High- disorder. Neuron. 2000;28:343–7. resolution crystal structure of the human CB1 cannabinoid receptor. Nature. 59. Moghaddam B, Javitt D. From revolution to evolution: the glutamate 2016;540:602–6. hypothesis of schizophrenia and its implication for treatment. 82. Krishna Kumar K, Shalev-Benami M, Hu H, Weis WI, Kobilka BK, Skiniotis G. Neuropsychopharmacology. 2012;37:4–15. Cannabinoid Receptor 1-G Protein Complex. Protein Data Bank. https://doi. 60. Kravitz AV, Tomasi D, LeBlanc KH, Baler R, Volkow ND, Bonci A, et al. Cortico- org/10.2210/pdb6n4b/pdb. striatal circuits: Novel therapeutic targets for substance use disorders. Brain 83. Krishna Kumar K, Shalev-Benami M, Robertson MJ, Hu H, Banister SD, Res. 2015;1628:186–98. Hollingsworth SA, et al. Structure of a Signaling Cannabinoid Receptor 1-G 61. Quiroz C, Gulyani S, Ruiqian W, Bonaventura J, Cutler R, Pearson V, et al. Protein Complex. Cell. 2019;176:448–58. Adenosine receptors as markers of brain iron deficiency: implications for 84. Manglik A, Kruse AC, Kobilka TS, Thian FS, Mathiesen JM, Sunahara RK, et al. Restless Legs Syndrome. Neuropharmacology. 2016;111:160–8. Crystal structure of the mu-opioid receptor bound to a morphinan 62. Ferré S, Quiroz C, Rea W, Guitart X, García-Borreguero D. Adenosine antagonis. Protein Data Bank. https://doi.org/10.2210/pdb4dkl/pdb. mechanisms and hypersensitive corticostriatal terminals in restless legs 85. Li XT, Hua T, Wu LJ, Liu ZJ. Crystal structure of human G protein coupled syndrome. Rationales for the use of inhibitors of adenosine transport. Adv receptor. Protein Data Bank. https://doi.org/10.2210/pdb5zty/pdb. Pharmacol. 2019;84:3–19. 86. Li X, Hua T, Vemuri K, Ho JH, Wu Y, Wu L, et al. Crystal structure of the 63. Decerce J, Smith LF, Gonzalez W, Sussman NM. Effectiveness and tolerability human cannabinoid receptor CB2. Cell. 2019;176:459–67. of istradefylline for the treatment of restless legs syndrome: an exploratory study in five female patients. Curr Ther Res Clin Exp. 2007;8:349–59. Publisher’s Note 64. Megelin T, Ghorayeb I. Cannabis for restless legs syndrome: a report of six Springer Nature remains neutral with regard to jurisdictional claims in patients. Sleep Med. 2017;36:182–3. published maps and institutional affiliations. 65. García-Borreguero D, Guitart X, Garcia Malo C, Cano-Pumarega I, Granizo JJ, Ferré S. Treatment of restless legs syndrome/Willis-Ekbom disease with the non-selective ENT1/ENT2 inhibitor dipyridamole: testing the adenosine hypothesis. Sleep Med. 2018;45:94–7. 66. Mouro FM, Batalha VL, Ferreira DG, Coelho JE, Baqi Y, Müller CE, et al. Chronic and acute adenosine A2A receptor blockade prevents long-term episodic memory disruption caused by acute cannabinoid CB1 receptor activation. Neuropharmacology. 2017;117:316–27. 67. Mouro FM, Köfalvi A, André LA, Baqi Y, Müller CE, Ribeiro JA, et al. Memory deficits induced by chronic cannabinoid exposure are prevented by adenosine A2AR receptor antagonism. Neuropharmacology. 2019;155:10–21. 68. Ferré S, Casadó V, Devi LA, Filizola M, Jockers R, Lohse MJ, et al. G protein- coupled receptor oligomerization revisited: functional and pharmacological perspectives. Pharmacol Rev. 2014;66:413–34. 69. Ferré S. The GPCR heterotetramer: challenging classical pharmacology. Trends Pharmacol Sci. 2015;36:145–52. 70. Burgueño J, Blake DJ, Benson MA, Tinsley CL, Esapa CT, Canela EI, et al. The adenosine A2A receptor interacts with the actin-binding protein alpha- actinin. J Biol Chem. 2003;278:37545–52. 71. He SQ, Zhang ZN, Guan JS, Liu HR, Zhao B, Wang HB, et al. Facilitation of μ- opioid receptor activity by preventing δ-opioid receptor-mediated codegradation. Neuron. 2011;69:120–31. 72. Schwarze SR, Ho A, Vocero-Akbani A, Dowdy SF. In vivo protein transduction: delivery of a biologically active protein into the mouse. Science. 1999;285:1569–72. 73. Taura J, Nolen EG, Cabré G, Hernando J, Squarcialupi L, López-Cano M, et al. Remote control of movement disorders using a photoactive adenosine A(2A) receptor antagonist. J Control Release. 2018;283:135–42.