Address correspondence and reprint requests to Sergi Ferré, National Institute on Drug Abuse, Intramural Research Program, National Institutes of Health, Department of Health and Human Services, 5500 Nathan Shock Dr, Baltimore, MD 21224, USA. E-mail: email@example.com
There has been a long debate about the predominant involvement of the different adenosine receptor subtypes and the preferential role of pre- versus post-synaptic mechanisms in the psychostimulant effects of the adenosine receptor antagonist caffeine. Both striatal A1 and A2A receptors are involved in the motor-activating and probably reinforcing effects of caffeine, although they play a different role under conditions of acute or chronic caffeine administration. The present review emphasizes the key integrative role of adenosine and adenosine receptor heteromers in the computation of information at the level of the striatal spine module (SSM). This local module is mostly represented by the dendritic spine of the medium spiny neuron with its glutamatergic and dopaminergic synapses and astroglial processes that wrap the glutamatergic synapse. In the SSM, adenosine acts both pre- and post-synaptically through multiple mechanisms, which depend on heteromerization of A1 and A2A receptors among themselves and with D1 and D2 receptors, respectively. A critical aspect of the mechanisms of the psychostimulant effects of caffeine is its ability to release the pre- and post-synaptic brakes that adenosine imposes on dopaminergic neurotransmission by acting on different adenosine receptor heteromers localized in different elements of the SSM.
Caffeine is the most consumed psychoactive drug in the world, with about 90% of the population (including children) in the United States regularly consuming caffeine-containing beverages or foods (Frary et al. 2005). Caffeine produces the same behavioral effects as classical psychostimulants, such as cocaine and amphetamine, mainly motor activation, arousal, and reinforcing effects (Fredholm et al. 1999; Griffiths et al. 2003). Furthermore, in humans, caffeine produces subjective effects that are qualitatively similar to classical psychostimulants (Griffiths et al. 2003). In fact, in drug discrimination experiments in animals (which provide reliable laboratory models for studying the interoceptive effects of central acting drugs; see Solinas et al. 2006), classical psychostimulants can substitute for behaviorally relevant doses of caffeine and the other way around (Mumford and Holtzman 1991; Munzar et al. 2002; Solinas et al. 2005).
Although the motor activating and arousal effects of caffeine are very well established (Fredholm et al. 1999), there has been some resistance in the literature to accept the reinforcing effects of caffeine, in spite of the fact that its considerable worldwide consumption provides a compelling circumstantial evidence. Nevertheless, there is already enough unambiguous experimental evidence supporting that caffeine can function as a reinforcer under certain conditions both in laboratory animals and humans (Griffiths and Woodson 1988; Griffiths et al. 2003).
Drug reinforcement can be defined as the ability of a drug to maintain self-administration or choice behavior. The reinforcing efficacy of a drug refers to the relative effectiveness in maintaining behavior on which the delivery of the drug is dependent. The experiments with self-administration (oral or intravenous) of caffeine or with choice behavior of caffeinated versus non-caffeinated foods in animals and humans have given very similar results. Those results indicate that caffeine has a weaker reinforcing efficacy than classical psychostimulants (Griffiths and Woodson 1988; Griffiths et al. 2003). In fact, when considering the criteria for drug dependence and abuse established by the Diagnostic and Statistical Manual of Mental Disorders, 4th edition, caffeine users often fulfill the criteria for drug dependence, but not for drug abuse (Nehlig 1999; Griffiths et al. 2003).
Physical dependence through suppression of the aversive effects of withdrawal symptoms has also been demonstrated to play an important role in human habitual caffeine consumption (Griffiths et al. 2003; Juliano and Griffiths 2004; Tinley et al. 2004; James and Rogers 2005). Caffeine withdrawal syndrome includes headache, fatigue, decreased energy and activeness, decreased alertness, drowsiness, decreased contentedness, depressed mood, difficulty concentrating, irritability, and not clearheaded (Juliano and Griffiths 2004). Onset of symptoms already occurs 12–24 h after abstinence and, although the incidence or severity of the symptoms increases with increases in daily dose, symptoms can appear with doses as low as 100 mg/day (equivalent to the content of caffeine in a regular cup of drip coffee) (Juliano and Griffiths 2004). Experiments in humans even suggest that the performance and mood enhancing effects of caffeine can only be appreciated in caffeine-dependent individuals and, therefore, just represent withdrawal reversal (James and Rogers 2005). Some studies suggest that caffeine can only produce conditioned flavor preference in caffeine-dependent individuals (Tinley et al. 2004), supporting a key role of negative reinforcing in caffeine consumption. In summary, both positive reinforcing effects along with suppression of withdrawal symptoms after short periods of abstinence can explain caffeine dependence by humans.
Caffeine and the central dopaminergic system
According to their main mechanism of action, psychostimulants are often classified as ‘dopamine uptake blockers’ and ‘dopamine releasers.’ Dopamine uptake blockers, such as cocaine, bind to and inhibit dopamine transporter function (Torres et al. 2003; Elliott and Beveridge 2005). Dopamine releasers, such as amphetamine, also bind to dopamine transporter and they enhance neurotransmitter release by reversing dopamine transport (Elliott and Beveridge 2005; Sulzer et al. 2005). Furthermore, dopamine uptake blockers and dopamine releasers exert differential effects on vesicular monoamine transporter-2 function (Hanson et al. 2004). The common final effect of both major types of psychostimulants is a significant increase of dopamine in the extracellular space of dopamine innervated brain areas. Thus, classically, psychostimulants have been considered as indirect dopamine receptor agonists, to differentiate them from direct dopamine receptor agonists, which directly activate dopamine receptors.
After a unilateral lesion of the ascending dopaminergic pathways in rodents, the systemic administration of indirect dopamine agonists induces an asymmetric motor activation, a turning behavior ipsilateral to the lesioned side, because of the stronger stimulation of dopamine receptors in the non-denervated compared with the dopamine-denervated striatum. On the other hand, the administration of direct dopaminergic agonists induces contralateral turning, because of the development of adaptative changes that increase the sensitivity of dopamine receptors to agonists in the denervated striatum (Ungerstedt 1971; Pycock 1980). The effects of caffeine in this so-called Ungerstedt’s model have been very difficult to interpret. Thus, caffeine (and other methylxanthines) was initially found to induce contralateral turning behavior (Fuxe and Ungerstedt 1974). However, it was also found that caffeine potentiates both the ipsilateral and contralateral turning induced by indirect and direct dopamine receptor agonists (Fuxe and Ungerstedt 1974). More recently, Morelli and coworkers have demonstrated that the ability of caffeine to induce contralateral turning (when systemically administered) depends on the previous repeated administration, ‘priming,’ of direct dopamine receptor agonists (Cauli and Morelli 2005). In fact, when administered to non-primed animals, caffeine induces a weak although significant ipsilateral turning behavior (Cauli et al. 2003; Cauli and Morelli 2005).
The results from turning behavior experiments illustrate what it has been confirmed with several other animal models, that caffeine strongly interacts with the central dopaminergic systems and that it mimics and potentiates the behavioral effects of direct or indirect dopamine receptor agonists (for reviews, see Ferréet al. 1992; Garrett and Griffiths 1997; Fisone et al. 2004; Cauli and Morelli 2005). Furthermore, an important amount of experimental evidence supports a key role of dopamine in the behavioral effects of caffeine in animals and humans. Thus, dopamine depletion or blockade of dopamine receptors significantly impairs the motor and discriminative stimulus effects of caffeine (Ferréet al. 1992; Garrett and Griffiths 1997). The results obtained with Ungerstedt’s model, with the ability of eliciting either ipsilateral or contralateral turning depending on the experimental paradigm, suggest that both pre- and post-synaptic mechanisms are involved in the ability of caffeine to influence dopaminergic neurotransmission. But these mechanisms should definitively involve the blockade of endogenous adenosine. Thus, it is now well established that the multiple central and non-central effects of methylxanthines are mostly because of antagonism of endogenous adenosine (Fredholm 1980; Rall 1982; Snyder 1985). Therefore, the key to elucidate the mechanisms of action responsible for the psychostimulant effects of caffeine is to understand how adenosine modulates dopaminergic neurotransmission in the brain.
Adenosine A1 or A2A receptor antagonism?
Among the four cloned adenosine receptors (A1, A2A, A2B, and A3 receptors), A1 and A2A receptors are the ones predominantly expressed in the brain. Caffeine is a non-selective adenosine receptor antagonist, with reported similar in vitro affinities for A1, A2A, and A2B receptors and with lower affinity for A3 receptors (Fredholm et al. 2001a; Solinas et al. 2005). Physiological extracellular levels of adenosine can be sufficient to occupy and, therefore, stimulate A1 and A2A receptors. On the other hand, A2B receptors have a lower affinity for adenosine and they are only activated by high pathological extracellular levels of adenosine (Fredholm et al. 2001a). Thus, A1 and A2A receptors seem to be the preferential targets for caffeine in the brain, although their involvement in the psychostimulant effects of caffeine still remains controversial. A1 receptors are widely expressed in the brain, while A2A receptors are highly concentrated in the striatum (Fredholm et al. 2001a; Schiffmann et al. 2007). The striatal localization of both receptors can underlie the motor-activating and reinforcing effects of caffeine (see below). On the other hand, A1 receptors localized in the brainstem and basal forebrain and A2A receptors localized in the hypothalamus have been suggested to be involved in caffeine-induced arousal (see below).
It was initially suggested by Snyder et al. (1981) that A1 receptors are preferentially involved in the motor stimulant effects of caffeine, based upon a correlation between potencies of a series of methylxanthines in stimulating locomotor activity of mice and in competing at adenosine receptors labeled with a tritiated A1 receptor agonist. Later studies in mice and monkeys suggested that there was a better correlation between the locomotor depressant effects of several adenosine receptor agonists and their ability to bind to A2A receptors (Durcan and Morgan 1989; Spealman and Coffin 1986). Also, there seemed to be a better correlation between the ability of several methylxanthines to produce motor stimulation in monkeys (increase in response rate in schedule-controlled behavior) and their ability to bind A2A receptors (Spealman 1988). It was then suggested that blockade of A2A receptors was mainly responsible for the motor stimulating effects of caffeine (Spealman 1988). This was supported by more recent findings demonstrating the ability of A2A receptor antagonists and the inability of some A1 receptor antagonists to reproduce some biochemical and behavioral effects of caffeine (Svenningsson et al. 1977; El Yacoubi et al. 2000) and the inability of caffeine to induce motor activation in A2A receptor knockout mice (Ledent et al. 1997). Furthermore, some results even suggested that A1 receptor antagonism induces motor depression, which could be responsible for the motor depressant effect observed with high doses of caffeine (El Yacoubi et al. 2000). However, other studies showed strong evidence that A1 receptor antagonists can induce motor activation and potentiate the motor activating effects of A2A receptor antagonists (Jacobson et al. 1993; Popoli et al. 1998).
The involvement of A1 and A2A receptors in the motor-activating properties of caffeine has recently been revisited with more comprehensive quantitative and qualitative analysis. In a systematic study that compared the counteraction of the motor depressant effects of selective A1 and A2A receptor agonists by caffeine and selective A1 and A2A receptor antagonists in rats, caffeine showed a profile of a non-selective adenosine receptor antagonist with preferential A1 receptor antagonism (Karcz-Kubicha et al. 2003). Factor analysis of a detailed qualitative study of motor activity also depicted a very similar profile for caffeine and an A1 receptor antagonist, which was significantly different from the profile of an A2A receptor antagonist (Antoniou et al. 2005). Recent experiments in wild-type mice or in mice with genetic inactivation of A1, A2A or both receptors also indicate that A1 receptor contributes to the stimulatory but not the inhibitory motor-activating effects of caffeine (Halldner et al. 2004; Kuzmin et al. 2006). Also, recent drug discrimination experiments support a key role of A1 receptors in the psychostimulant effects of caffeine (Solinas et al. 2005). A selective A1, but not A2A, receptor antagonist produced significant caffeine-like discriminative effects in rats trained to discriminate an injection of a motor-activating dose of caffeine from saline. Furthermore, a selective A1, but not A2A, receptor agonist dose-dependently reduced caffeine’s discriminative effects (Solinas et al. 2005). In conclusion, both A1 and A2A receptors are involved in the motor-activating effects of caffeine, while A1 receptor is the adenosine receptor subtype mostly involved in its discriminative-stimulus effects.
Importantly, chronic exposure to caffeine differentially modifies its motor effects dependent on A1 and A2A receptor blockade. Thus, chronic exposure to caffeine in the drinking water results in partial tolerance to the motor effects of an additional acute administration of caffeine and total cross-tolerance to the motor effects of an A1 but not an A2A receptor antagonist (Karcz-Kubicha et al. 2003). This suggests that tolerance to the effects of A1 receptor blockade is mostly responsible for the tolerance to the motor-activating effects of caffeine and that the residual motor-activating effects of caffeine in tolerant individuals might be mostly because of A2A receptor blockade. These results agree with other experimental findings showing a lack of tolerance to the motor activating effects of an A2A receptor antagonist after its repeated administration (Halldner et al. 2000). Similarly, in rats with a unilateral lesion of the ascending dopaminergic pathways, chronic treatment with caffeine or repeated administration of an A2A receptor antagonist was not associated with tolerance to the A2A receptor antagonist-induced potentiation of turning behavior produced by dopamine receptor agonists (Popoli et al. 2000; Pinna et al. 2001).
The well-known arousal-enhancing properties of caffeine depend on its ability to antagonize the sleep-promoting effects of adenosine (Porkka-Heiskanen et al. 2000; Basheer et al. 2004;Huang et al. 2007; Ferréet al. 2007a). Although previous results suggested that A1 receptors are involved in sleep regulation by inhibiting ascending cholinergic neurons of the basal forebrain (Porkka-Heiskanen et al. 2000; Basheer et al. 2004), more recent studies, which include experiments with A2A and A1 receptor knockout mice, indicate that A2A receptors play a crucial role in the sleep-promoting effects of adenosine and the arousal-enhancing effects of caffeine (Huang et al. 2007; Ferréet al. 2007a). Those A2A receptors are localized in the ventrolateral pre-optic area of the hypothalamus and their stimulation promotes sleep by inducing GABA release in the histaminergic tuberomammillary nucleus, thereby inhibiting the histaminergic arousal system (Huang et al. 2007; Ferréet al. 2007a).
Finally, the relative involvement of A1 and A2A receptors in the reinforcing effects of caffeine still needs to be determined. But, most probably, the same basic dopamine-mediated mechanisms (see below) are involved in the motor-activating and reinforcing effects of caffeine, as it happens with classical psychostimulants (Wise 2004;Ikemoto 2007).
Adenosine in the striatal spine module
The striatal efferent neuron, also called the medium spiny neuron, constitutes more than 95% of the striatal neuronal population (Gerfen 2004). It is an inhibitory neuron that uses GABA as its main neurotransmitter. There are two subtypes of medium spiny neurons, which selectively express one of two peptides, enkephalin or dynorphin. Enkephalinergic medium spiny neurons predominantly express dopamine D2 (D2L isoform) and A2A receptors, while dynorphinergic medium spiny neurons, which also express the peptide substance P, predominantly express dopamine D1 receptors and adenosine receptors of the A1 subtype (Ferréet al. 1997; Agnati et al. 2003; Gerfen 2004) (Fig. 1). It must be pointed out that a detailed analysis of the mRNA expression of the different receptor subtypes indicated that there is a limited subset of striatal neurons (about 15% of all GABAergic efferent neurons) with a mixed phenotype of GABAergic enkephalinergic and GABAergic substance P-dynorphinergic neurons, with D1 and D2 receptors (Surmeier et al. 1996). The same degree of D1–D2 receptor co-localization has been suggested from recent experiments using bacterial artificial chromosome transgenic mice that express an enhanced green fluorescence protein under the control of the D1 or D2 receptor promoter (Lee et al. 2006).
The medium spiny neuron receives two main afferents: glutamatergic afferents from cortical, thalamic and limbic areas and dopaminergic afferents from the mesencephalon (substantia nigra and ventral tegmental area). Both sets of afferents converge on the dendritic spine of the medium spiny neuron. Glutamatergic and dopaminergic terminals make preferential synaptic contacts with the head and the neck of the dendritic spine, respectively (Gerfen 2004) (Fig. 1). ‘Local module’ has been recently 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. 2007b). The dendritic spine, glutamatergic terminal, dopaminergic terminal, and astroglial processes that wrap the glutamatergic synapse constitute the most common striatal local module, which can be called the striatal spine module (SSM) (Ferréet al. 2007b) (Fig. 1). It must, however, be pointed out that the segregation of dopamine and adenosine receptors in the two subtypes of medium spiny neurons implies the existence of at least two subtypes of SSMs, which could be called enkephalinergic SSM and dynorphinergic SSM (without forgetting the small percentage of mixed enkephalinergic–dynorphynergic SSM; see above).
In the SSM, adenosine is a key modulator of dopaminergic and glutamatergic neurotransmission. Until recently it was believed that the main source of extracellular adenosine was a paracrine-like formation. Extracellular adenosine would come mostly from intracellular adenosine, the concentration of which depends upon the breakdown and synthesis of ATP, which is metabolized to AMP and, then, by means of 5′ nucleotidases it is converted to adenosine, which can be transported to the extracelular space by means of equilibrative transporters (Ferréet al. 2005). However, recent studies suggest that astroglia play a fundamental role in the formation of extracellular adenosine which affects synaptic transmission. Astrocytes express glutamate receptors (mostly metabotropic) and ATP receptors which, when activated, induce astrocytes to release glutamate and ATP (Newman 2003; Hertz and Zielke 2004) (Fig. 1). Astroglial-released ATP can be converted to adenosine in the extracellular space by means of ectonucleotidases (Pascual et al. 2005). Finally, there is an increasing amount of data suggesting the existence of a neurotransmitter-like formation of adenosine, a synaptic pool of adenosine. In this case, adenosine would come from ATP co-released with glutamate, which is metabolized to adenosine by means of ectonucleotidases (Ferréet al. 2005) (Fig. 1).
In the SSM, A2A receptors are localized post-synaptically in the dendritic spine of enkephalinergic medium spiny neurons, co-localized with D2 receptors and pre-synaptically in glutamatergic terminals, co-localized with A1 receptors (Ferréet al. 1997, 2005, 2007b; Rosin et al. 2003; Schiffmann et al. 2007) (Fig. 1). In addition to the glutamatergic terminals, A1 receptors are localized in a fraction of dopaminergic nerve terminals (Borycz et al. 2007) and, post-synaptically, in the dendritic spine of dynorphinergic medium spiny neurons, co-localized with D1 receptors (Ferréet al. 1997, 2005) (Fig. 1). Although most of these findings have been demonstrated in the dorsal striatum, functional studies indicate that the same distribution of adenosine receptors can be found in the ventral striatum (Ferréet al. 2007a). Antagonistic interactions between A2A and D2 receptors modulate the function of the enkephalinergic medium spiny neuron and antagonistic interactions between A1 and D1 receptors modulate the function of the dynorphinergic medium spiny neuron (Ferréet al. 1993, 1996). This gives the explanation at the neuronal level of an important number of pharmacological findings indicating a selective modulation of A1 and A2A receptor ligands on D1 and D2 receptor-mediated behavioral effects, respectively (reviewed in Ferréet al. 1992, 1997, 2001). For instance, in relation to A1 and A2A receptor blockade, selective A1 or A2A receptor antagonists are not able to induce turning behavior in rats with a unilateral lesion of the ascending dopaminergic pathways, but they selectively potentiate the contralateral turning induced by low equipotent doses of D1 and D2 receptor agonists, respectively (Popoli et al. 2000; Ferréet al. 2001). Predictably, non-selective adenosine receptor antagonists, such as caffeine and theophylline potentiate both D1 and D2 receptor-mediated behavioral responses (Ferréet al. 1991a; Jiang et al. 1993).
The molecular mechanisms responsible for the selective antagonistic A1–D1 and A2A–D2 receptor interactions have been found to involve ‘intramembrane receptor–receptor interactions,’ a common property of neurotransmitter receptor heteromers (Agnati et al. 2003; Ferréet al. 2007c). In fact, there is compelling evidence for the existence of A1–D1 and A2A–D2 receptor heteromers in artificial cell systems (Ginés et al. 2000; Hillion et al. 2002; Canals et al. 2003) and in the striatum (Ferréet al. 2007c). In the present review, I adopt the broad definition of ‘neurotransmitter’ by Snyder and Ferris (2000), i.e. a molecule, released by neurons or glia, which physiologically influences the electrochemical state of adjacent cells. Neurotransmitter receptor heteromers are functional entities with distinctive biochemical properties different from those of the individual components of the heteromer. These biochemical characteristics include changes in ligand binding characteristics and signaling (Marshall 2001; George et al. 2002; Prinster et al. 2005; Ferréet al. 2007c). A receptor unit in the heteromer can display several biochemical properties, which can be simply dependent on the presence of the other unit, i.e. just as a consequence of the heteromerization, or on co-stimulation of the two (or more) receptor units in the heteromer. In case of dependence on co-stimulation, the neurotransmitter receptor heteromer acts as a ‘processor’ of computations that modulate cell signaling (Ferréet al. 2007c).
Intramembrane receptor–receptor interactions are changes in binding characteristics that are dependent on co-activation of the receptor units of the receptor heteromer (Ferréet al. 2007c). They imply an intermolecular crosstalk between both receptor units in the heteromer at the membrane level, without intervention of signaling pathways (Agnati et al. 2003; Ferréet al. 2007c). In the A2A–D2 receptor heteromer, the stimulation of the A2A receptor decreases the binding of dopamine to the D2 receptor (Ferréet al. 1991b). This intramembrane interaction seems to control neuronal excitability and, consequently, neuronal firing and neurotransmitter (GABA) release by the enkephalinergic medium spiny neurons (Ferréet al. 1993; Stromberg et al. 2000). This is related to the ability of D2 receptors to suppress Ca2+ currents through L-type voltage-dependent calcium channels by a cAMP–protein kinase A (PKA)-independent and Gq/11–phospholipase C-dependent signaling pathway (Hernandez-Lopez et al. 2000; Schiffmann et al. 2007) (Fig. 2a). Thus, stimulation of striatal A2A receptor does not produce a significant effect on its own, but it strongly counteracts the depressant effects of D2 receptor stimulation on neuronal firing and neurotransmitter release (Ferréet al. 1993; Stromberg et al. 2000; Schiffmann et al. 2007).
In addition to the intramembrane interaction, a strong antagonistic interaction between A2A and D2 receptors has been found at the second messenger level, by which stimulation of D2 receptors counteracts the activation of adenylyl-cyclase induced by stimulation of A2A receptors (Kull et al. 1999; Hillion et al. 2002). Stimulation of A2A receptor can potentially stimulate adenylyl-cyclase, with consequent activation of cAMP–PKA signaling pathway and induction of the expression of different genes, such as c-fos and preproenkephalin by activating the constitutive transcription factor cAMP-response element binding protein and the mitogen-activated protein kinase pathway (Kull et al. 1999; Ferréet al. 2005, 2007b; Schiffmann et al. 2007) (Fig. 2a). A2A receptor-mediated activation of PKA can also induce phosphorylation of dopamine and cAMP-regulated phosphoprotein 32 kDa (DARPP-32) at threonine 34 (Thr34) and a phosphatase 2-mediated dephosphorylation of DARPP-32 at Thr75 (Fisone et al. 2004; Schiffmann et al. 2007). Furthermore, A2A receptor-mediated activation of PKA can phosphorylate α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) glutamate receptors (Schiffmann et al. 2007), which plays a crucial role in the initial plastic changes of glutamatergic synapses, which includes synaptic recruitment of AMPA receptors (Song and Huganir 2002) (Fig. 2a).
From experiments with gene inactivation of Gαolf or DARPP-32, which show a significant decrease in caffeine-induced motor activation, it has been suggested that the psychostimulant effects of caffeine depend mostly on the blockade of a basal adenosinic tone able to activate A2A receptor–Gs-olf–cAMP–PKA signaling (Hervéet al. 2001; Lindskog et al. 2002). However, under basal conditions, stimulation of A2A receptors can poorly activate cAMP–PKA signaling and gene expression, because of a strong tonic inhibitory effect of endogenous dopamine and D2 receptor stimulation on adenylyl-cyclase (reviewed in Ferréet al. 2005, 2007b; Schiffmann et al. 2007). For instance, the systemic administration of A2A receptor antagonists do not significantly modify the basal PKA-dependent pshosphorylation of DARPP-32 or AMPA receptors, but it significantly counteracts DARPP-32 and AMPA receptor phosphorylation induced by D2 receptor antagonists (Svenningsson et al. 2000; Hakansson et al. 2006). Furthermore, genetic inactivation of AC5, the main adenylyl-cyclase isoform coupled to adenosine and dopamine receptors in the striatum, practically eliminates the inhibitory effects of D2 and stimulatory effects of A2A receptor stimulation on cAMP accumulation, but does not modify the motor depressant effects of A2A agonists and motor activating effects of caffeine (Lee et al. 2002).
We have therefore evidence for the coexistence of two reciprocal and apparently incompatible interactions between A2A and D2 receptors in the GABAergic enkephalinergic neurons, which could be involved in the psychostimulant effects of caffeine. Thus, either co-stimulation of A2A and D2 receptors results in blockade of the D2 receptor-Gq/11-phospholipase C signaling pathway, by means of the intramembrane A2A–D2 interaction, or it results in a blockade of the A2A receptor–Gs-olf–cAMP–PKA signaling, by means of the A2A–D2 interaction at the adenylyl-cyclase level. It is possible that when the D2 receptor is not forming heteromers, but homomers, it couples preferentially to Gi, while when forming heteromers with the A2A receptor the D2 receptor couples preferentially to Gq/11. This would be a similar situation to that of the recently described D1–D2 receptor heteromer (Rashid et al. 2007). However, in the D1–D2 receptor heteromer, both receptors couple and signal through Gq/11 (Rashid et al. 2007), while in the A2A–D2 receptor heteromer the main function of the A2A receptor seems to be the control of D2 receptor signaling through Gq/11. In this case, A2A receptor does not couple to Gs/olf, or else there should be a clear activation of the cAMP–PKA signaling under basal conditions. Another possibility for A2A receptors to activate Gs-olf–cAMP–PKA signaling is to couple to another receptor, such as the metabotropic mGlu5 glutamate receptor. There is evidence indicating that A2A receptors heteromerize with mGlu5 receptors and that their co-activation synergistically potentiates A2A receptor function (Ferréet al. 2002), including A2A receptor-mediated PKA activation with phosphorylation of DARPP-32 (Nishi et al. 2003). Thus, in vivo experiments have demonstrated that co-activation of A2A and mGlu5 receptor agonists induce an increase in the striatal expression of c-fos. It is therefore possible that in the mGlu5–A2A–D2 receptor heteromer, D2 receptor predominantly uses Gi–cAMP–PKA signaling.
In summary, A2A–D2 receptor interactions play a key role in the motor-activating effects of caffeine. In fact, these interactions have been given a lot of attention in the literature, more recently with the application of A2A receptor antagonists as an adjuvant therapy for l-DOPA in Parkinson’s disease (for recent review, see Muller and Ferré 2007). But, as mentioned before, A1–D1 receptor interactions are also of significant functional and pharmacological importance. Thus, A1 receptor antagonists selectively potentiate the motor activating effects of D1 receptor-mediated motor activation in different animal models (Popoli et al. 1996, 2000). Similarly to what happens with the A2A–D2 receptor heteromers, A1 and D1 receptors antagonistically interact both at intramembrane level and at the adenylyl-cyclase level (Fig. 2b). However, in this case, the interactions are not reciprocal, and stimulation of A1 receptors inhibits both the binding of dopamine to the D1 receptor (Ferréet al. 1994, 1998) and the D1 receptor-mediated activation of cAMP–PKA signaling pathway and the expression of genes, such as c-fos and preprodynorphin (Ferréet al. 1998, 1999) (Fig. 2b).
Pre-synaptic mechanisms: glutamate and dopamine release
It was first postulated that the post-synaptic A1 and A2A receptors, which form heteromers with D1 and D2 receptors, respectively, were mostly responsible for the interactions between adenosine and dopamine involved in the motor and reinforcing effects of caffeine (Ferréet al. 1992, 1997). However, previous studies had repeatedly shown that A1 and A2A receptors exert opposite modulatory roles on extracellular levels of glutamate and dopamine in the striatum, with activation of A1 receptors inhibiting and activation of A2A receptors stimulating glutamate and dopamine release (Wood et al. 1989; Lupica et al. 1990; Zetterstrom and Fillenz 1990; Ballarin et al. 1995; Popoli et al. 1995; Okada et al. 1996, 1997; Golembiowska and Zylewska 1997; Corsi et al. 2000). More recently, systemic administration of caffeine was found to produce a significant increase in the extracellular concentrations of dopamine and glutamate in the ventral striatum, particularly in the most medial part, the medial shell of the nucleus accumbens (Solinas et al. 2002; Quarta et al. 2004a; Borycz et al. 2007). The same effect was observed with the systemic administration of an A1 receptor antagonist, while an A2A receptor antagonist produced the opposite effect, a small but significant decrease in the extracellular concentration of dopamine and glutamate in the nucleus accumbens (Quarta et al. 2004a). Furthermore, chronic administration of caffeine in the drinking water completely counteracted the effects of caffeine or the A1 receptor antagonist on dopamine and glutamate, while the effect of A2A receptor antagonist was not modified (Quarta et al. 2004a). Thus, these biochemical changes correlate with previous studies on motor activity (Karcz-Kubicha et al. 2003), suggesting an involvement of pre-synaptic mechanisms in the psychostimulant effects of caffeine. This interpretation is strongly supported by results of in vivo microdialysis experiments where caffeine is administered through the dialysis probe (reverse dialysis). Local perfusion of caffeine or an A1 receptor antagonist, but not an A2A receptor antagonist, significantly increased the extracellular concentration of dopamine in the dorsal striatum (Okada et al. 1996, 1997). The same increasing effects on dopamine and also glutamate were obtained when caffeine or an A1 receptor antagonist, but not an A2A receptor antagonist, were perfused in the shell of the nucleus accumbens (Quarta et al. 2004b). Furthermore, caffeine-induced dopamine release in the shell of the nucleus accumbens was found to be dependent on glutamate release and stimulation of NMDA receptors most probably localized in dopaminergic terminals (Quarta et al. 2004b).
There has been some debate about the ability of caffeine to modulate dopamine release in the nucleus accumbens. Another experimental group claimed that the increase in dialysate dopamine by caffeine we observed in the shell of the nucleus accumbens is due to contamination by dopamine arising from the adjacent medial prefrontal cortex (De Luca et al. 2007). However, this is not a valid interpretation if we consider the difference in basal levels of extracellular dopamine between striatal compartments, which, in agreement with both research groups, is four to five times higher in the shell of the nucleus accumbens compared with the adjacent cortex (Borycz et al. 2007; De Luca et al. 2007). If anything, increases in the extracellular dopamine level in the shell of the nucleus accumbens could cause dopamine diffusion with contamination of samples in the adjacent cortex. A more detailed mapping analysis showed that systemic administration of caffeine or local perfusion of an A1 receptor antagonist could only produce a significant elevation of extracellular dopamine in the dorsal but not in the ventral part of the most medial aspect of the shell of the nucleus accumbens (Borycz et al. 2007). In fact, our ‘dorsal shell of the nucleus accumbens,’ which contains much higher basal levels of dopamine than the adjacent cortex, very much overlaps with the striatal compartment described by De Luca et al. (2007) as the ‘anterior and medial border of the shell of the nucleus accumbens,’ where they also found a significant caffeine-induced increase in extracellular dopamine. On the other hand, our ‘ventral shell of the nucleus accumbens’ very much overlaps with the striatal compartment simply described by De Luca et al. (2007) as ‘nucleus accumbens shell,’ where both studies show that caffeine does not modify the extracellular levels of dopamine. Therefore, the results of both experimental groups are basically the same and point to differential effects of caffeine in different striatal subcompartments. In fact, analyzing the effects of the intrastriatal perfusion of an A1 receptor antagonist in several other striatal compartments showed striking differences compared with the shell of the nucleus accumbens. Thus, A1 receptor blockade significantly increased the extracellular concentration of dopamine, but not glutamate, in the core of the nucleus accumbens and in the caudate–putamen and the effect was more pronounced in the most medial compartments (Borycz et al. 2007). In summary, a subregional difference in the A1 receptor-mediated control of glutamate and dopamine release exists in the striatum, most probably related to subregional differences in the level of tonic activation by endogenous adenosine (Borycz et al. 2007).
Many experimental findings indicate that dopamine release in the medial striatal compartments is involved in invigoration of approach and in some aspects of incentive learning (for recent review, see Ikemoto 2007). In relation to psychostimulants, dopamine release in the very medial striatal compartments seems to be involved in both their motor-activating and reinforcing effects (Ikemoto 2007). Therefore, the pre- and post-synaptic dopaminergic mechanisms mentioned above (striatal dopamine release and adenosine–dopamine receptor–receptor interactions) taking place in the medial striatal compartments are most probably involved in the motor and reinforcing effects of caffeine. In relation to its dopamine-releasing effects, at least two factors could explain the weaker reinforcing properties of caffeine when compared with other psychostimulants: its specific subregional effects (Borycz et al. 2007) and the development of tolerance (Quarta et al. 2004a). Also, even when considering both pre- and post-synaptic mechanisms, at behaviorally significant doses, caffeine does not reproduce the same pattern of striatal biochemical changes induced by classical psychostimulants and other drugs of abuse (Bennett and Semba 1998; Nehlig and Boyet 2000; Valjent et al. 2004). It is also important to mention that non-dopaminergic mechanisms, related to the ability of adenosine to directly or indirectly modulate the release of different neurotransmitters, such as acetylcholine, serotonin, and histamine (see below), can also be involved in the motor and rewarding effects of caffeine.
The existence of an A1 receptor-mediated glutamate-independent modulation of dopamine release suggested the presence of functional A1 receptors in striatal dopaminergic terminals. Recent studies, using a combination of immunological and pharmacological techniques in striatal nerve terminal preparations have demonstrated that A1 receptors are present in a significant fraction of dopaminergic terminals in the rat striatum, and that activation of these receptors directly inhibits dopamine release (Borycz et al. 2007). On the other hand the existence of an A1 and A2A receptor-mediated regulation of glutamate release suggested the existence of both receptors in striatal glutamatergic terminals. This was recently demonstrated with electron-microscopy and immunocytochemical experiments, which showed that the majority of glutamatergic nerve terminals contain both A1 and A2A receptors (Ciruela et al. 2006). In functional studies with striatal nerve terminal preparations, stimulation of the A1 receptor was found to decrease while stimulation of A2A receptors potentiated potassium-induced glutamate release (Ciruela et al. 2006). Importantly, when both A1 and A2A receptors were stimulated, the response was not a counteractive effect, but the same than results from A2A receptor stimulation, i.e. a potentiation of glutamate release (Ciruela et al. 2006). Furthermore, in the same kind of preparations, low concentrations of adenosine inhibited while high concentrations stimulated glutamate release (Ciruela et al. 2006). This would agree with a higher affinity for adenosine of the A1 compared with the A2A receptor (Fredholm et al. 2001b) and would provide a mechanism for a fine-tuned modulation of glutamate release by adenosine, with low concentrations inhibiting and high concentrations stimulating glutamate release.
The ability of A1 receptors to inhibit glutamate or dopamine release most probably involves a βγ-dependent inhibition of N- and P/Q-type voltage-dependent calcium channels, which is the most commonly reported mechanism for inhibition of neurotransmitter release by Gi-coupled receptors (Jarvis and Zamponi 2001) (Fig. 2c). On the other hand, the stimulatory effect of A2A receptor on striatal glutamate release is probably related to their ability to activate cAMP–PKA signaling, as this mechanism has been shown for A2A receptor-induced acetylcholine in the striatum, GABA release in the globus pallidus and serotonin release in the hippocampus (Gubitz et al. 1996; Shindou et al. 2002; Okada et al. 2001) (Fig. 2c). This effect is related to the ability of PKA to phosphorylate different elements of the machinery that is involved in vesicular fusion (Leenders and Sheng 2005). The mechanism by which A2A receptor stimulation shuts down the effects of A1 receptor stimulation is related to their ability to heteromerize and to exert intermolecular interactions of the same kind as that previously described for the A2A–D2 receptor heteromer (see above). Thus, A1–A2A receptor heteromers have been identified in transfected cells an in rat striatum (Ciruela et al. 2006). A biochemical fingerprint of the A1–A2A receptor heteromer is an intramembrane receptor–receptor interaction, by which stimulation of A2A receptors decreases the affinity of A1 receptor for agonist binding (Ciruela et al. 2006). Furthermore, the affinity of the A2A receptor for caffeine was found to be significantly lower in the A2A–A1 receptor heteromer compared with the A2A–D2 receptor heteromer or the non-heteromerized A2A receptor (Ciruela et al. 2006). The A1–A2A receptor heteromer provides a ‘concentration-dependent switch’ mechanism by which low and high concentrations of synaptic adenosine produce the opposite effects on glutamate release. The A1–A2A receptor heteromer also provides a target for caffeine that explains at least the A1 receptor-mediated glutamate-dependent modulation of dopamine release.
Under chronic caffeine exposure, different factors might contribute to the predominant tolerance of the effects of A1 receptor blockade. Up-regulation of A1 receptors has been repeatedly demonstrated (Jacobson et al. 1996; Karcz-Kubicha et al. 2003), but its significance as a mechanism involved in caffeine tolerance has been seriously questioned (Holtzman et al. 1991). Modifications in the function of A1–A2A receptor heteromers might play a key role in the development of caffeine tolerance. In fact, as mentioned before, chronic caffeine exposure counteracts both motor activation and dopamine release in the nucleus accumbens induced by caffeine or an A1 receptor antagonist (but not and A2A receptor antagonist) (Karcz-Kubicha et al. 2003; Quarta et al. 2004b). Thus, radioligand binding experiments have shown that chronic treatment with caffeine causes an increase in the potency of A2A receptor agonist-mediated inhibition of A1 receptor agonist binding and a significant reduction in the affinity of the striatal A2A receptor for caffeine (Ciruela et al. 2006). An additional factor that might play a significant role in caffeine tolerance is the significant increase in plasma and extracellular concentrations of adenosine with chronic caffeine exposure (Conlay et al. 1997). The higher adenosine levels and low affinity of the A2A receptor for caffeine could allow endogenous adenosine to stimulate A2A receptors even in the presence of caffeine, which would not reach enough concentration to compete with adenosine for binding A2A receptors. Under these conditions, A1 receptor signaling in the A1–A2A receptor heteromer would be expected to be chronically turned down, because of its continuous blockade by caffeine and because of the strong A2A receptor-mediated inhibition of A1 receptor agonist binding. On the other hand, an additional administration of caffeine would be expected to produce a blockade of the residual A2A receptor signaling.
Conclusions and future perspectives
There has been a long debate about the predominant involvement of the different adenosine receptor subtypes and the preferential role of pre- versus post-synaptic mechanisms in the psychostimulant effects of the adenosine receptor antagonist caffeine. The present review emphasizes the key integrative role of adenosine and adenosine receptor heteromers in the computation of information at the level of the SSM. In the SSM, adenosine acts pre- and post-synaptically through multiple mechanisms, which depend on heteromerization of A1 and A2A receptors among themselves and with D1 and D2 receptors, respectively. Caffeine produces its motor and reinforcing effects by releasing the pre- and post-synaptic brakes that adenosine imposes on dopaminergic neurotransmission in the SSM. By releasing the pre-synaptic brake, caffeine induces glutamate-dependent and glutamate-independent release of dopamine. These pre-synaptic effects of caffeine are potentiated by the release of the post-synaptic brake imposed by antagonistic adenosine–dopamine receptor interactions.
This review has focused on the role of adenosine in the SSM, which is a main target for the motor and, most probably, reinforcing effects of caffeine. We still need to study in detail the local modules localized in the basal forebrain and hypothalamus that are targets for the arousal effects of caffeine (Porkka-Heiskanen et al. 2000; Basheer et al. 2004; Huang et al. 2007; Ferréet al. 2007a). Furthermore, more research still needs to be performed about the role of adenosine in the SSM. We have not taken into account other neurotransmitters, such as acetylcholine, serotonin, histamine, and endocannabinoids. Several studies suggest that A1 and A2A receptors localized in cholinergic nerve terminals, which form very few synaptic specializations and release acetylcholine in a volume transmission mode (reviewed in Ferréet al. 1997), modulate striatal acetylcholine release (Gubitz et al. 1996; Preston et al. 2000). Serotonin also plays an important functional role in the striatum and, in the hippocampus, A1 and A2A receptors have been found to modulate serotonin release (Okada et al. 2001). Maybe there are also A1–A2A receptor heteromers controlling both acetylcholine and serotonin release. The ability of adenosine to modulate the ascending histaminergic arousal system by acting on hypothalamic A2A receptors was already mentioned. It is however still possible that adenosine receptors localized in histaminergic terminals also modulate striatal histamine release. About endocannabinoids, we have recently reported that cannabinoid CB1 and adenosine A2A receptors form receptor heteromers in co-transfected cells and that they co-immunoprecipitate from striatal membrane preparations (Carriba et al. 2007). Functional studies suggest that, in the A2A–CB1 receptor heteromer, CB1 receptor signaling is completely dependent on A2A receptor activation (Carriba et al. 2007). Finally, we still need to define the mechanisms behind the physical dependence of caffeine, which might depend on the increase in plasma and extracellular adenosine associated with chronic caffeine intake (Conlay et al. 1997).
This study was supported by the intramural funds of the National Institute on Drug Abuse.