Abstract

Adenosine and endocannabinoids are very ubiquitous non-classical neurotransmitters that exert a modulatory role on the transmission of other more ‘classical’ neurotransmitters. In this review we will focus on their common role as modulators of dopamine and glutamate neurotransmission in the striatum, the main input structure of the basal ganglia. We will pay particular attention to the role of adenosine A_2A receptors and cannabinoid CB₁ receptors. Experimental results suggest that presynaptic CB₁ receptors interacting with A_2A receptors in cortico-striatal glutamatergic terminals that make synaptic contact with dynorphinergic medium-sized spiny neurons (MSNs) are involved in the motor-depressant and addictive effects of cannabinoids. On the other hand, postsynaptic CB₁ receptors interacting with A_2A and D₂ receptors in the dendritic spines of enkephalinergic MSNs and postsynaptic CB₁ receptors in the dendritic spines of dynorphinergic MSN are probably involved in the cataleptogenic effects of cannabinoids. These receptor interactions most probably depend on the existence of a variety of heteromers of A_2A, CB₁ and D₂ receptors in different elements of striatal spine modules. Drugs selective for the different striatal A_2A and CB₁ receptor heteromers could be used for the treatment of neuropsychiatric disorders and drug addiction and they could provide effective drugs with fewer side effects than currently used drugs.

This article is part of a themed issue on Cannabinoids. To view the editorial for this themed issue visithttp://dx.doi.org/10.1111/j.1476-5381.2010.00831.x

The striatal spine module

The concept of ‘local module’ facilitates the understanding of the integrated role that a neurotransmitter has in a particular brain area. Also it facilitates understanding the role of interactions between different neurotransmitters. Local module is defined as the minimal portion of one or more neurons and/or one or more glial cells that operates as an independent integrative unit (Ferréet al., 2007a). In a recent review we analysed the different types of local modules centred in the dendritic spines of striatal γ-aminobutyric acidergic (GABAergic) efferent neurons, also called medium-sized spiny neurons (MSNs), which constitute more than 95% of the striatal neuronal population (Ferréet al., 2009a). MSNs receive two main extrinsic inputs: glutamatergic afferents from cortical, limbic and thalamic areas and dopaminergic afferents from the mesencephalon [substantia nigra pars compacta and ventral tegmental area (VTA)] (Gerfen, 2004). The glutamatergic and dopaminergic terminals make synaptic contact with the head and neck of the dendritic spines of the MSN (Gerfen, 2004). GABAergic intrinsic inputs from GABAergic interneurons and from MSN collaterals also contact the neck of the dendrite or the dendritic shaft close to the base of the dendritic spine (Ferréet al., 2009a). Because the dopaminergic and GABAergic inputs preferentially target the distal and proximal portions of the MSN dendritic tree, at least two different types of local modules can be distinguished. Each ‘striatal spine module’ includes the dendritic spine of the MSN, the glutamatergic terminal wrapped by glial processes and dopaminergic and/or GABAergic nerve terminals (Ferréet al., 2009a). But there are different phenotypes of MSNs that give rise to several additional types of striatal spine modules.

Two classes of MSNs, which are homogeneously distributed in the striatum, can be differentiated by their output connectivity and their expression of dopamine and adenosine receptors and neuropeptides. In the dorsal striatum (mostly represented by the nucleus caudate-putamen), enkephalinergic MSNs connect the striatum with the globus pallidus (lateral globus pallidus) and express the peptide enkephalin and a high density of dopamine D₂ and adenosine A_2A receptors (they also express adenosine A₁ receptors), while dynorphinergic MSNs connect the striatum with the substantia nigra (pars compacta and reticulata) and the entopeduncular nucleus (medial globus pallidus) and express the peptides dynorphin and substance P and dopamine D₁ and adenosine A₁ but not A_2A receptors (Ferréet al., 1997;Gerfen, 2004;Quirozet al., 2009). These two different phenotypes of MSN are also present in the ventral striatum (mostly represented by the nucleus accumbens and the olfactory tubercle). However, although they are phenotypically equal to their dorsal counterparts, they have some differences in terms of connectivity. First, not only enkephalinergic but also dynorphinergic MSNs project to the ventral counterpart of the lateral globus pallidus, the ventral pallidum, which, in fact, has characteristics of both the lateral and medial globus pallidus in its afferent and efferent connectivity. In addition to the ventral pallidum, the medial globus pallidus and the substantia nigra-VTA, the ventral striatum sends projections to the extended amygdala, the lateral hypothalamus and the pedunculopontine tegmental nucleus. Finally, unlike the dorsal striatum, the substantia nigra pars reticulata is not a main target area for the ventral striatum, which preferentially directs its midbrain output to the substantia nigra pars compacta and the VTA (Heimeret al., 1995;Robertson and Jian, 1995;Ferré, 1997). It is also important to mention that a small percentage of MSNs have a mixed phenotype and express both D₁ and D₂ receptors (Surmeieret al., 1996).

Enkephalinergic and dynorphinergic MSNs can also be differentiated into two phenotypically different groups of neurons, which are heterogeneously distributed in the patch (striosomes) or the matrix compartments. One main characteristic of the patch compartment is that it has a much higher expression of µ-opioid receptors than the matrix compartment (Mansouret al., 1987;Gerfen, 2004). But there are also connectivity differences: patch-MSNs receive cortical input predominantly from periallocortical areas (such as the infralimbic and prelimbic cortices) while matrix-MSNs receive input from neocortical areas (Gerfen, 2004). Furthermore patch-dynorphinergic-MSNs project predominantly to the substantia nigra pars compacta and matrix-dynorphinergic-MSNs project predominately to the substantia nigra pars reticulata (Gerfen, 2004). In summary, there are different types of striatal spine modules, based on variations in localization on the dendritic tree and in enkephalinergic versus dynorphinergic and in patch versus matrix phenotypes (Ferréet al., 2009a).

Adenosine and endocannabinoids in the striatal spine modules

The main sources of extracellular adenosine in the striatum still need to be established, but there is evidence suggesting that an important source is ATP co-released with glutamate from glia and glutamatergic terminals (Pascualet al., 2005;Schiffmannet al., 2007;Ferréet al., 2007a). ATP is then converted to adenosine by means of ectonucleotidases. In fact, most extracellular adenosine, ATP and glutamate seem to be glial in origin, but to depend on neuronal glutamate release. When glutamate and ATP receptors localized in astrocytic membranes closely apposed to glutamatergic synapses are activated, astrocytes release further glutamate and ATP (Hertz and Zielke, 2004). Therefore, an increased extracellular concentration of adenosine, coming from ATP released from nerve terminals and amplified by the adjacent glial processes, represents a signal of increased glutamatergic neurotransmission (Schiffmannet al., 2007;Ferréet al., 2007b). Recent studies also suggest that ATP could also be co-released with dopamine in the striatum (Cechova and Venton, 2008).

Most of the effects of adenosine in the central nervous system (CNS) are mediated by A₁ and A_2A receptors, which are usually coupled to G_i/o and G_s/olf proteins respectively (Fredholmet al., 2001). A_2A receptors are more concentrated in the striatum than anywhere else in the brain, while A₁ receptors are more widespread (Fredholmet al., 2001;Rosinet al., 2003;Schiffmannet al., 2007;Quirozet al., 2009). In the striatal spine module, A_2A receptors are localized predominantly postsynaptically in the dendritic spine of enkephalinergic but not dynorphinergic MSNs, co-localized with D₂ receptors (Ferréet al., 2007a,b;Quirozet al., 2009). Striatal A_2A receptors are also found presynaptically in glutamatergic terminals, where they form receptor heteromers with A₁ receptors (Ciruelaet al., 2006). Recent studies have shown that these presynaptic A_2A receptors are also segregated and that they are predominantly localized in glutamatergic terminals that make synaptic contact with dynorphinergic MSNs (Quirozet al., 2009). In fact,in vitro andin vivo experiments indicate the existence of a selective presynaptic A_2A receptor-mediated modulation of glutamatergic neurotransmission to dynorphinergic MSNs (Quirozet al., 2009). Therefore, the recently proposed relative expression of A_2A receptors in the different subtypes of striatal spine modules (Ferréet al., 2009a) already needs to be amended. With respect to the other elements of the striatal spine module, A_2A receptors show little expression in the dopaminergic and GABAergic terminals (Rosinet al., 2003;Gomeset al., 2009). Nevertheless, recent studies suggest the existence of functional A_2A receptors in the striatal collaterals of enkephalinergic MSNs, which activation facilitates GABA release (Shindouet al., 2008). Less clear is the significance of results from some studies suggesting the existence of striatal A_2A receptors localized in GABA terminals whose activation inhibits GABA release (Mori and Shindou, 2003). Although, contrary to A_2A receptors, no segregation seems to apply to A₁ receptors, segregation of other receptors allows them to form different receptor heteromers in different striatal spine modules. As mentioned before, A₁ receptors heteromerize with A_2A receptors in the glutamatergic terminals that contact dynorphinergic MSNs and also in these neurons they seem to form heteromers with D₁ receptors (Ferréet al., 1997;2007a;).

Endocannabinoids are membrane-derived signalling lipids that stimulate G protein-coupled receptors (GPCRs) that are targeted by delta9-tetrahydrocannabinol (THC), the main addictive ingredient of marijuana. Two major endocannabinoids, anandamide and 2-arachidonylglycerol (2-AG), have been discovered. Like classical neurotransmitters they are released from neurons following neuronal depolarization and Ca²⁺ influx into the cell. Unlike classical neurotransmitters, they are not stored in vesicles, but are produced ‘on demand’ from endocannabinoid precursors by the action of enzymes localized in the plasma membrane (Di marzoet al., 1998;Freundet al., 2003;Piomelli, 2003). The enzymes necessary for the biosynthesis of anandamide are the Ca²⁺-dependent N-acyltransferase and N-acylphosphatidylethanolamine-phospholipase D (NAPE-PLD). For the biosynthesis of 2-AG, the main enzymes involved are the Ca²⁺-dependent and G_q-11-coupled receptor-activated phospholipase C and diacylglycerol lipase (DGL). The action of endocannabinoids is terminated by uptake or diffusion into cells followed by intracellular metabolism. Fatty acid amide hydrolase and monoacylglycerol lipase are the two primary enzymes involved in intracellular metabolism of anandamide and 2-AG respectively (Di Marzoet al., 1998;Freundet al., 2003;Piomelli, 2003). Finally, the existence of an endocannabinoid membrane equilibrative transporter has been postulated to explain cellular uptake of both anandamide and 2-AG (Piomelli, 2003). Some studies suggest that this transporter, which has not yet been identified, plays a role in endocannabinoid release (Ronesiet al., 2004;Adermark and Lovinger, 2007). However, the existence of a transporter has been questioned and, instead, fatty acid hydrolases maybe important for uptake of anandamide and 2-AG (Kaczochaet al., 2006).

There are two subtypes of cannabinoid receptors so far identified and characterized, CB₁ and CB₂ receptors. The CB₁ subtype is the one predominantly expressed in the adult CNS and it is considered the most abundant GPCR in the brain (Herkenhamet al., 1990;1991;), although CB₂ receptors have also been found to be expressed in neurons and other brain cells (Ashtonet al., 2006;Gonget al., 2006;Bruscoet al., 2008a,b;). CB₁ receptors are usually coupled to G_i proteins (Demuth and Molleman, 2006) and are often localized presynaptically, where their stimulation usually inhibits neurotransmitter release (Di Marzoet al., 1998;Piomelli, 2003). In the striatal spine module CB₁ are localized both pre- and postsynaptically (reviewed inFerréet al., 2009a). Presynaptically, CB₁ receptors are localized in GABAergic terminals of interneurons or collaterals from MSNs, and also in glutamatergic but not in dopaminergic terminals (Pickelet al., 2004;2006;Köfalviet al., 2005;Mátyáset al., 2006;Uchigashimaet al., 2007). Postsynaptically, CB₁ receptors are localized in the somatodendritic area of MSN (Rodriguezet al., 2001;Pickelet al., 2004;2006;Köfalviet al., 2005) and both enkephalinergic and dynorphinergic MSNs express CB₁ receptors (Martínet al., 2008).

One of the best studied functions of endocannabinoids is retrograde signalling with stimulation of presynaptic CB₁ receptors and the consequent inhibition of neurotransmitter release. 2-AG, rather than anandamide, seems to be mainly responsible for endocannabinoid-mediated retrograde signalling in the striatum and, probably in most brain areas (Hashimotodaniet al., 2007). A recent study has demonstrated that in the striatum DGL is mostly localized in the plasma membrane of the dendritic spines of MSNs (Uchigashimaet al., 2007). The same study showed that, in MSNs, inhibition of DGL abolishes depolarization-induced suppression of inhibition and depolarization-induced suppression of excitation (Uchigashimaet al., 2007), which depend on activation of postsynaptic G_q-11-coupled receptors and on activation of presynaptic CB₁ receptors with inhibition of GABA and glutamate release respectively (Hashimotodaniet al., 2007). As mentioned before, Ca²⁺-dependent and G_q-11-coupled receptor-activated phospholipase C is a main enzyme in 2-AG synthesis. The metabotropic glutamate receptor mGlu₅ is a G_q-11-coupled receptor that is co-expressed with DGL in the dendritic spines of MSNs and it has been found to be particularly involved in endocannabinoid-dependent retrograde signalling (Uchigashimaet al., 2007). Studies performed in the hippocampus suggest that, unlike 2-AG, anandamide may be preferentially involved in anterograde signalling, as NAPE-PLD was found to be concentrated presynaptically, where it was associated with intracellular calcium stores in several types of excitatory axon terminals (Nyilaset al., 2008). Furthermore, fatty acid amide hydrolase is mostly localized postsynaptically (Freundet al., 2003).

Adenosine A_2A receptor–cannabinoid CB₁ receptor interactions

The very high expression of A_2A and CB₁ receptors in the striatum (Herkenhamet al., 1990;1991;Fredholmet al., 2001;Rosinet al., 2003;Schiffmannet al., 2007;Quirozet al., 2009) suggests that direct or indirect interactions between A_2A and CB₁ receptors are involved in the modulation of motor activity and goal-directed behaviours. In fact, there is experimental evidence indicating that adenosine and adenosine ligands, by acting on A_2A receptors, modulate some of the pharmacological effects of cannabinoids, particularly those dependent on striatal function (see below). If we also take into account that the density of A_2A receptors is much higher in the striatum than anywhere else in the brain (Fredholmet al., 2001;Rosinet al., 2003;Schiffmannet al., 2007;Quirozet al., 2009), we can infer that those pharmacological effects induced by cannabinoids that are strongly dependent on A_2A receptor function probably involve CB₁ receptors localized in the striatum.

Cannabinoids produce a myriad of pharmacological (behavioural and autonomic) effects (Chaperon and Thiébot, 1999). Among those effects, hypolocomotion, hypothermia, antinociception and catalepsy constitute the so-called ‘tetrad syndrome’ of cannabinoid activity in experimental animals (Chaperon and Thiébot, 1999;Monoryet al., 2007). These pharmacological effects of cannabinoids depend on CB₁ receptor activation, as they are counteracted by pharmacological blockade or genetic inactivation of CB₁ receptors (Ledentet al., 1999;Zimmeret al., 1999;Varvelet al., 2005;Monoryet al., 2007). Surprisingly, recent studies using conditional knockout mouse lines lacking the expression of CB₁ receptors in different neuronal subpopulations indicate that the tetrad syndrome does not depend on functional expression of CB₁ receptors on GABAergic interneurons (Monoryet al., 2007), as previously thought. Those studies strongly support the involvement of CB₁ receptors localized in cortico-striatal glutamatergic neurons in the hypolocomotor effects of THC, while the cataleptic effects of THC seem to depend on the functional expression of CB1 receptors in MSNs, particularly D₁ receptor-containing neurons (Monoryet al., 2007).

We also obtained indirect but compelling evidence for a main role of striatal CB₁ receptors in the motor-depressant effects of cannabinoids (Carribaet al., 2007). In our experiments, local infusion of the CB₁ receptor agonist WIN 55212-2 produced a decrease in exploratory locomotor activity in rats, which was completely counteracted by the previous systemic administration of two different CB₁ receptor antagonists, SR141716A and AM251. Furthermore, systemic administration of the selective A_2A receptor antagonist MSX-3 completely counteracted the motor-depressant effects of WIN 55212-2 (Carribaet al., 2007). Previous studies had also shown that pharmacological blockade or genetic inactivation of A_2A receptors reduced the cataleptic effects induced by systemic administration of the CB₁ receptor agonist CP55,940 in mice (Anderssonet al., 2005) and that A_2A receptor knockout mice show a decrease in the rewarding and aversive effects of low and high doses of THC respectively (Soriaet al., 2006). Furthermore, we found that low doses of the A_2A receptor antagonist MSX-3 reduce the reinforcing effects of THC and anandamide under a fixed-ratio schedule of intravenous drug injection in squirrel monkeys (Justinovaet al., 2008). These studies suggest that not only the motor but also the addictive effects of THC are mediated by CB₁ receptors localized in the striatum and that A_2A receptor stimulation is involved in those effects.

At this point it is important to mention that there has been, and there is still, debate about the main localization of CB₁ receptors involved in the addictive effects of THC (Gardner, 2005;Lupica and Riegel, 2005). A common molecular mechanism contributing to the development of drug addiction that is shared by various drugs of abuse (including cannabinoids) is their ability to increase levels of extracellular dopamine in ventral striatum, especially in the shell of the nucleus accumbens (Koob, 1992;Robbins and Everitt, 1996;Di Chiara, 2002). The systemic administration of THC increases the firing rate of the VTA neurons and induces dopamine release in the nucleus accumbens (Chenet al., 1990;Frenchet al., 1997;Tandaet al., 1997;Solinaset al., 2007). Although the involvement of CB₁ receptors from the VTA has been suggested,in vivo administration of THC directly into the VTA does not induce dopamine release in the nucleus accumbens, while direct infusion of THC into the nucleus accumbens does (Schlicker and Kathmann, 2001;Gardner, 2005;Lupica and Riegel, 2005). It has then been suggested that presynaptic CB₁ receptors that control striatal glutamate release (Robbeet al., 2001) are main targets for the dopamine-releasing effects of cannabinoids. Their stimulation would decrease the excitability of the striatal GABAergic dynorphinergic neurons that project to the mesencephalon and tonically inhibit dopaminergic cells in the VTA, therefore stimulating dopamine release in the nucleus accumbens (Schlicker and Kathmann, 2001). Relevant to the present review and adding more evidence for the striatal localization of CB₁ receptors involved in the addictive effects of THC, we found that in rats a behaviourally active dose of the A_2A receptor antagonist MSX-3 significantly counteracted THC-induced, but not cocaine-induced, increases in extracellular dopamine levels in the shell of the nucleus accumbens (Justinovaet al., 2008).

Going back to the striatal spine modules, and without discarding the possibility of indirect interactions between receptors localized in different neurons, there are two main localizations where close interactions between A_2A and CB₁ receptors can take place: the dendritic spine of enkephalinergic MSNs and the glutamatergic terminals that make synaptic contact with dynorphinergic MSNs. From the experiments with conditional knockout mice and dopamine release mentioned above, it seems most probable that interactions between A_2A and CB₁ receptors localized in glutamatergic terminals that contact dynorphinergic MSNs are primarily involved in the hypolocomotor and rewarding effects of THC. However, from results obtained with biochemical and electrophysiological experiments, it has been suggested that postsynaptic mechanisms are also involved in striatal A_2A receptor-dependent CB₁ receptor function (Anderssonet al., 2005;Tebanoet al., 2009).Anderssonet al. (2005) found that systemic administration of the CB₁ receptor agonist CP55,940 in mice produces catalepsy and striatal protein kinase A (PKA)-dependent phosphorylation at threonine 34 of 32 kDa dopamine- and adenosine 3′,5′-monophosphate-regulated phosphoprotein (DARPP-32). These effects were counteracted by a CB₁ receptor antagonist, by pharmacological blockade of A_2A receptors or by genetic inactivation of A_2A or D₂ receptors. Furthermore, CB₁ receptor agonist-induced catalepsy was significantly reduced in DARPP-32 knockout mice (Anderssonet al., 2005). These results suggest that CB₁ receptors in enkephalinergic MSNs utilize G_s/olf protein-dependent signalling, which is probably related to simultaneous molecular interactions between these CB₁ receptors with A_2A and D₂ receptors (see below) and may be involved in the cataleptogenic effects of cannabinoids. In fact, in co-transfected cells or primary striatal neurons in culture, co-stimulation, but not individual receptor stimulation of CB₁ and D₂ receptors results in a G_s/olf protein-dependent (pertussis toxin-insensitive) adenylyl cyclase activation (Glass and Felder, 1997;Kearnet al., 2005). Another study suggested that simply co-expressing CB₁ and D₂ receptors is sufficient to induce stimulation of adenylyl cyclase in response to CB₁ receptor activation (Jarrahianet al., 2004). Postsynaptic CB₁ receptors are also localized on dynorphinergic MSNs, where they probably have inhibitory effects, which are also probably involved in the cataleptogenic effects of cannabinoids (Monoryet al., 2007). Those postsynaptic CB₁ receptors localized in dynorphinergic MSNs are probably coupled to G_i/o proteins (Martínet al., 2008). As also pointed out byAnderssonet al. (2005), their biochemical results could also involve a presynaptic mechanism, by a CB₁-mediated inhibition of striatal glutamate release in dynorphinergic MSNs (as glutamate receptor stimulation results in dephosphorylation of DARPP-32 at threonine 34;Svenningssonet al., 2004).

Bothin vitro andin vivo electrophysiological experiments have shown that CB₁ receptor agonists produce a pronounced decrease in striatal synaptic transmission, mostly by decreasing excitatory neurotransmission and it is currently believed that this is predominantly related to inhibition of glutamate release (Robbeet al., 2001;Pistiset al., 2002;Kreitzer and Malenka, 2007).Tebanoet al. (2009) recently showed that a CB₁ receptor-mediated decrease in striatal synaptic transmission (measuring extracellular field potentials in cortico-striatal slices) can also be partially, but significantly, reduced by pharmacological blockade or genetic inactivation of A_2A receptors. The authors conclude that their results are more consistent with a postsynaptic mechanism, although a presynaptic component could not be ruled out (Tebanoet al., 2009). According to Tebanoet al. the postsynaptic mechanism in this case would depend on an interaction between A_2A and CB₁ receptors in the enkephalinergic MSN. This interpretation would not fit with the above-postulated predominant G_s/olf-coupling-dependent stimulatory role of postsynaptic CB₁ receptors in enkephalinergic MSNs.

In summary, the experimental results mentioned so far suggest that presynaptic CB₁ receptors interacting with A_2A receptors in cortico-striatal glutamatergic terminals that make synaptic contact with dynorphinergic MSNs are involved in the motor-depressant and addictive effects of cannabinoids. On the other hand, postsynaptic CB₁ receptors interacting with A_2A and D₂ receptors in the dendritic spines of the enkephalinergic MSNs (and postsynaptic CB₁ receptors in the dendritic spines of dynorphinergic MSN, according to the conditional knockout mice experiments) are probably involved in the cataleptogenic effects of cannabinoids. In both cases, basal levels of A_2A receptor activation would be necessary for CB₁ receptor function. These, so far hypothetical, mechanisms need to be reconciled with other more established functions of A_2A receptors in striatal spine modules. For instance, how is it possible that A_2A receptor stimulation in striatal glutamatergic terminals induces glutamate release (Ciruelaet al., 2006;Quirozet al., 2009) and yet A_2A receptor stimulation is necessary for a CB₁ receptor-mediated inhibition of glutamate release?

The analysis of the differential effects of A_2A receptor ligands in different animal models can already give some clues about these apparently contradictory findings. First it is important to remember that according to the most accepted model of basal ganglia function, stimulation of enkephalinergic or dynorphinergic MSNs leads to motor depression or motor activation, respectively, and that the final motor activation depends on the balanced activation of both striatal efferent neurons (Gerfen, 2004;DeLong and Wichmann, 2007). Results from different research groups have repeatedly shown that A_2A receptor antagonists produce motor activation and A_2A receptor agonists produce motor depression (seeKarcz-kubichaet al., 2003, and references therein) and that A_2A receptor agonists produce an increase in the striatal basal levels of glutamate, while A_2A receptor antagonists do not significantly modify those basal levels (Quartaet al., 2004a,b;), although they block striatal glutamate release induced by cortical stimulation (Quirozet al., 2009). These pharmacological findings suggest that there is a tonic basal stimulation of postsynaptic striatal A_2A receptors, and their blockade produces motor activation. These postsynaptic A_2A receptors are probably only partially occupied by endogenous adenosine, explaining the ability of A_2A receptor agonists to produce motor depression. On the other hand, there is little tonic activation of striatal presynaptic A_2A receptors that produce glutamate release when stimulated, indicating that these receptors are only important during increases in input from cortical glutamatergic afferents, which increases striatal glutamate release and adenosine production (see above andFerréet al., 2007a,b;Schiffmannet al., 2007). However, presynaptic A_2A receptors that facilitate CB₁-mediated inhibition of glutamate release must be tonically activated, which would explain the blockade by A_2A receptor antagonists of the putative effects of cannabinoids on dynorphinergic MSNs described above. Thus, there should be different pools of presynaptic and postsynaptic A_2A receptors with different affinities for adenosine and possibly with different G protein-coupling or signalling (see above). This is where receptor heteromers come into play.

Adenosine A_2A–cannabinoid CB₁ receptor heteromers

Receptor heteromer has been recently defined as a macromolecular complex composed of at least two (functional) receptor units or protomers with biochemical properties that are demonstrably different from those of its individual components (Ferréet al., 2009b). In a recent study, using a functional complementation assay, D₂ receptor homodimers with a single G protein were found to be a minimal signalling unit, which is maximally activated by agonist binding to one of the protomers, whereas additional agonist or inverse agonist binding to the second protomer blunts or enhances signalling respectively (Hanet al., 2009). This allosteric modulation of signalling results from a direct interaction of the receptor homodimer (which should probably be called homodimeric D₂ receptor; seeFerréet al., 2009b, for nomenclature) with the single interacting G protein, rather than from a downstream effect (Hanet al., 2009). A similar situation, two receptor units and one G protein, is most probably found in a receptor heterodimer, but in this case two different ligands interact within the heteromer, and several putative allosteric interactions like the one described for D₂ receptor homomers have also been described, for instance, in the A_2A–D₂ and A₁–A_2A receptor heteromers (Ciruelaet al., 2006;Ferréet al., 2008;2009b;). In these examples, stimulation of A_2A receptors decreases the affinity of D₂ and A₁ receptors for their respective agonists (Ferréet al., 1991;Ciruelaet al., 2006). It is currently believed that an allosteric interaction in a receptor heteromer (as well as in a receptor homomer) involves an intermolecular interaction by which binding of a ligand to one of the protomers induces structural changes on the second protomer that result in functional or pharmacological changes on this second protomer (Milligan and Smith, 2007;Ferréet al., 2007b;2009b;).

In addition to allosteric modulations, GPCR oligomerizationper se can produce changes in ligand recognition and G protein-coupling. With regard to changes in ligand recognition, for instance, the A_2A receptor was found to have 13 times higher affinity for caffeine when co-transfected with D₂ than when co-transfected with A₁ receptors (Ciruelaet al., 2006). With regard to changes in G protein-coupling, recent models indicate that only one G protein can bind to two receptor units (van Rijn and Whistler, 2009;Hanet al., 2009). This means that a receptor heteromer will at least have to ‘decide’ to which G protein it should bind if its protomers are usually coupled to different G proteins. In some cases, the receptor heteromer can ‘choose’ a completely new G protein. For instance, dopamine receptors are classified as D₁-like, with D₁ and D₅ receptor subtypes, usually coupled to G_s/olf proteins, or D₂-like, with D₂, D₃ and D₄ receptor subtypes, usually coupled to G_i/o proteins. However, the D₁–D₃ receptor heteromer couples to G_s/olf (Fiorentiniet al., 2008) and the D₁–D₂ and D₂–D₅ receptor heteromers couple preferentially to G_q proteins (Rashidet al., 2007;Soet al., 2009). Although it still needs to be demonstrated, the A₁–A_2A heteromer probably couples to G_i/o and the A_2A–D₂ receptor heteromer couples to G_q (Ferréet al., 2008) (Figure 1).

The functional interactions between adenosine A_2A, cannabinoid CB₁ and dopamine D₂ receptors mentioned above provide the framework for the possible molecular interactions between A_2A and CB₁ receptors and for the existence of A_2A–CB₁ and A_2A–CB₁–D₂ receptor heteromers. A_2A–CB₁ receptor heteromerization has been demonstrated by means of bioluminescence resonance energy transfer (BRET) techniques in cells co-transfected with A_2A receptors fused toRenilla luciferase (Rluc) and CB₁ receptors fused to a yellow fluorescence protein (YFP) (Carribaet al., 2007). It is important to point out that just the existence of BRET does not demonstrate the existence of actual physical contact between the fused proteins, but that there exists a very close proximity, which could depend on oligomerization. Nevertheless, when using BRET, physical contact can be shown by using several different experimental approaches. One possibility includes BRET saturation experiments, where a constant amount of the donor fusion protein (usually a receptor fused to RLuc) is co-expressed with increasing amounts of the acceptor fusion protein (usually a receptor fused to YFP). If there is oligomerization, saturation is reached when all receptor-Rluc molecules are specifically associated with their receptor-YFP counterparts. By contrast, if the BRET signal results from random collision promoted by high receptor density, a quasi-linear curve is obtained. A clear saturation BRET curve was obtained for the A_2A−Rluc–CB₁–YFP pair when constant amounts of the cDNA for the Rluc construct were co-transfected with increasing amounts of the plasmid cDNA for the YFP construct (Carribaet al., 2007). The existence of striatal A_2A–CB₁ receptor heteromers has also been inferred from confocal immunohistochemical analysis of rat brain sections and co-immunoprecipitation experiments from rat striatal membranes (Carribaet al., 2007). A_2A–CB₁ receptor co-localization was found predominantly in fibrilar structures, compatible with both dendritic processes and nerve terminals (Carribaet al., 2007).

The study of A_2A–CB₁ receptor interactions in co-transfected cells has shown that CB₁ receptor signalling through G_i/o receptor-coupling (inhibition of forskolin-induced cAMP accumulation) depends on selective co-activation of A_2A receptors (Carribaet al., 2007). Thus, G_i/o receptor-coupling could constitute biochemical characteristics of A_2A–CB₁ receptor heteromers localized in striatal glutamatergic terminals that contact dynorphinergic MSNs and may be a main target for the motor-depressant and addictive effects of cannabinoids (Figure 1). These heteromers must coexist with A₁–A_2A heteromers and A_2A receptor homomers (Figure 1). It is our current working hypothesis that these receptor homomers and heteromers have different affinities for adenosine. Under basal conditions, adenosine seems to bind preferentially to the A₁ receptor in the A₁–A_2A heteromers (as previously suggested inCiruelaet al., 2006) and to the A_2A receptor in the A_2A–CB₁ receptor heteromer. This is associated with a relatively low degree of glutamate release, which can be further inhibited by endocannabinoid release. However, under conditions of strong input, the increased adenosine formation might also activate A_2A receptors in the A₁–A_2A receptor heteromer (which inhibits its signalling due to the antagonistic allosteric interaction in the heteromer) and in the A_2A receptor homomers, which stimulates glutamate release (Ciruelaet al., 2006). This can provide a very fine tuning for the modulation of glutamate release to the dynorphinergic MSN that depends on the levels of adenosine and endocannabinoids, with low adenosine and high endocannabinoid tone producing the weakest and high adenosine and low endocannabinoid tone producing the strongest glutamate release (Figure 2).

A_2A–D₂ receptor heteromerization has been demonstrated in mammalian transfected cells with co-immunoprecipitation, BRET and fluorescence resonance energy transfer (FRET) techniques (reviewed inFerréet al., 2008). By using computerized modelling, pull-down techniques and mass spectrometric analysis, it was shown that an electrostatic interaction between the arginine (Arg)-rich epitope located in the amino-terminal portion of the third intracellular loop of the D₂ receptor and a single phosphate group from a casein kinase phosphorylable serine localized in the distal portion of the carboxy-terminus of the A_2A receptor is involved in A_2A–D₂ receptor heteromerization (Canalset al., 2003;Ciruelaet al., 2004). The Arg–phosphate electrostatic interaction between epitopes located in intracellular domains is obviously not the only interaction responsible for A_2A–D₂ receptor heteromerization. Thus, a significant but not complete reduction of BRET is observed when transfected cells express mutated D₂ receptors that lack the key amino acids involved in the Arg–phosphate interaction (Ciruelaet al., 2004), indicating that other receptor domains are also involved. Most probably, intramembrane domains play an important role in A_2A–D₂ receptor heteromerization, as has been demonstrated for other GPCR homomers and heteromers (Guoet al., 2008;González-Maesoet al., 2008;Hanet al., 2009). Nevertheless, the significant modification of BRET with mutated receptors indicates that the Arg–phosphate interaction is necessary to provide the final quaternary structure of the heteromer, which in fact determines its function. Patch-clamp experiments in identified enkephalinergic MSNs demonstrated that disruption of the Arg–phosphate interaction in A_2A–D₂ receptor heteromers (by intracellular addition of small peptides with the same sequence than the receptor epitopes involved in the Arg–phosphate interaction) completely eliminates the ability of the A_2A receptor to antagonistically modulate the D₂ receptor-mediated inhibition of neuronal excitability (Azdadet al., 2009).

The above-mentioned antagonistic interaction between A_2A and D₂ receptors is probably related to the existence of an allosteric modulation in the A_2A–D₂ receptor heteromer. An antagonistic A_2A–D₂ receptor interaction has been demonstrated in many different membrane preparations from different transfected mammalian cells and from rat and human striatal tissues and implies the ability of A_2A receptor stimulation to change the binding characteristics (decrease the affinity) of the D₂ receptor for agonists (Ferréet al., 1991;2008;). The antagonistic interaction in the A_2A–D₂ receptor heteromer provides a mechanism for the well-known ability of A_2A receptor agonists and antagonists to selectively counteract or potentiate, respectively, the motor-activating effects of dopamine D₂ receptor agonists. This interaction also seems to be fundamental for the motor-activating effects of A_2A receptor antagonists (and also non-selective adenosine receptor antagonists such as caffeine) and provides a rational for the use of A_2A receptor antagonists in Parkinson's disease (Ferréet al., 1997;2001;Muller and Ferré, 2007).

In addition to the allosteric modulation in the A_2A–D₂ receptor heteromer, A_2A and D₂ receptors can also interact at the second messenger level when not forming heteromers, and this has been repeatedly demonstrated both in cell culture and in the brain (reviewed inFerréet al., 2008). In this case, however, it is the stimulation of D₂ receptors that counteracts the effects of A_2A receptor stimulation. A_2A receptors, through their coupling to G_s/olf proteins, can potentially stimulate adenyl-cyclase, with phosphorylation of several PKA substrates, such as DARPP-32, cAMP-responsive elements binding (CREB) and AMPA receptors and the consequent increase in the expression of different genes, such asc-fos orpreproenkephalin in the enkephalinergic MSN (Ferréet al., 2007a,b;2008;Schiffmannet al., 2007). D₂ receptors, on the other hand, can potentially couple to G_i/o proteins and counteract the ability of A_2A receptor stimulation to signal through the cAMP/PKA cascade (Ferréet al., 2007a,b;2008;Schiffmannet al., 2007). Both types of antagonistic A_2A–D₂ receptor interactions (in the heteromer and between homomers) coexist in enkephalinergic MSNs (Figure 1). Co-stimulation of A_2A and D₂ receptors implies a simultaneous A_2A receptor-mediated inhibition of the D₂ receptor-mediated modulation of neuronal excitability and a D₂ receptor-mediated inhibition of the A_2A receptor-mediated modulation of gene expression, which provides a clear example of a functional dissociation between neuronal excitability and gene expression. This apparently incompatible coexistence of reciprocal antagonistic A_2A–D₂ receptor interactions can be explained by the presence in the same cell of A_2A and D₂ receptors homomers and heteromers.

In addition to the A_2A–CB₁ and A_2A–D₂ receptor heteromers, CB₁–D₂ and A_2A–CB₁–D₂ receptor heteromerization has been shown by biophysical techniques (BRET, FRET, sequential BRET-FRET, BRET-bimolecular complementation) in transfected cells (Carribaet al., 2008;Marcellinoet al., 2008;Navarroet al., 2008) (Figure 1). As mentioned before, when CB₁ and D₂ receptors co-expressed in the same cells are co-stimulated they couple to G_s/olf, which results in stimulation of adenylyl cyclase (Glass and Felder, 1997;Kearnet al., 2005). It is most probable that G_s/olf-coupling is a biochemical property of CB₁–D₂ receptor heteromers (Kearnet al., 2005). One question that needs to be resolved is if CB₁–D₂ receptor heteromers couple constitutively to G_s/olf (Jarrahianet al., 2004) or if they only switch upon co-agonist treatment (Kearnet al., 2005). That G protein-coupling might depend on which protomers are stimulated in the receptor heteromer constitutes an attractive possibility. If it were a common property of receptor heteromers, it might also be involved in the presynaptic control of glutamate release mentioned above, with the A₁–A_2A receptor coupling to G_i or G_s proteins upon stimulation of A₁ stimulation or A₁–A_2A receptor co-stimulation respectively. Also a similar co-agonist treatment-dependent G protein switch could take place in the A_2A–CB₁ receptor heteromer. We also need to determine whether the A_2A–CB₁–D₂ receptor heteromers exist in enkephalinergic MSNs and, if they do, we need to determine their properties and function.

General conclusions

The adenosine–cannabinoid receptor interactions reviewed here provide a clear example of the importance of looking for the functional significance of receptor heteromers in the CNS. Local modules (such as the striatal spine module) provide the best framework for studying receptor heteromer function, which can lead to a better integrative view of the role of multiple neurotransmitters in different brain areas and circuits (such as the striatum and the basal ganglia). It is becoming evident that receptor heteromers can provide an explanation for the previously unsuspected existence of multiple functions of a single GPCR subtype, not only in different elements of a local module, but even in the same cell. These multiple functions depend on direct molecular interactions with other receptors, which results in significant changes in their ligand-binding and G protein-coupling properties. The selective localization and function, as well as the unique ligand-binding properties of receptor heteromers provide the obvious possibility of using receptor heteromers as targets for drug development. Drugs selective for different A_2A and CB₁ receptor heteromers could be used for the treatment of neuropsychiatric disorders and drug addiction, and they could provide effective drugs with fewer side effects than currently used drugs. For instance, the CB₁ receptor antagonist rimonabant was marketed as an anti-obesity drug and was also proposed as a potential anti-smoking treatment. However, it targets peripheral as well as CNS receptors, thus promoting side effects, such as anxiety and depression (Christensenet al., 2007;Mitchell and Morris, 2007). This led to FDA disapproval in USA and its removal from the market in Europe. We already mentioned that the A_2A–D₂ receptor heteromer can be a target for Parkinson's disease, while drugs selective for presynaptic A_2A receptor heteromers (A₁–A_2A and A_2A–CB₁) could be of use in diyskinetic disorders (Quirozet al., 2009) and in THC addiction (see above).

Acknowledgments

Work supported by the intramural funds of the National Institute on Drug Abuse, NIH, DHHS and by Grants from Spanish ‘Ministerio de Ciencia y Tecnología’ (SAF2008-00146 and SAF2008-03229-E/ for ERA-NET Coordination of Research Activities) and grant 060110 from ‘Fundació La Marató de TV3’.

Glossary

Abbreviations:

2-AG

2-arachidonylglycerol

BRET

bioluminescence resonance energy transfer

CNS

central nervous system

CREB

cAMP-responsive elements binding

DARPP-32

32 kDa dopamine- and adenosine 3′,5′-monophosphate-regulated phosphoprotein

DGL

diacylglycerol lipase

FRET

fluorescence resonance energy transfer

GABA

γ-aminobutyric acid

GPCR

G protein-coupled receptor

MSN

medium-sized spiny neuron

NAPE-PLD

N-acyltransferase and N-acylphosphatidylethanolamine-phospholipase D

PKA

protein kinase A

Rluc

Renilla luciferase

THC

delta9-tetrahydrocannabinol

VTA

ventral tegmental area

YFP

yellow fluorescent protein

Conflict of interest

The authors state no conflict of interest.

References