Regulation of Hippocampal Cannabinoid CB₁ Receptor Actions by Adenosine A₁ Receptors and Chronic Caffeine Administration: Implications for the Effects of Δ⁹-Tetrahydrocannabinol on Spatial Memory

Abstract

The cannabinoid CB₁ receptor-mediated modulation of γ-aminobutyric acid (GABA) release from inhibitory interneurons is important for the integrity of hippocampal-dependent spatial memory. Although adenosine A₁ receptors have a central role in fine-tuning excitatory transmission in the hippocampus, A₁ receptors localized in GABAergic cells do not directly influence GABA release. CB₁ and A₁ receptors are the main targets for the effects of two of the most heavily consumed psychoactive substances worldwide: Δ⁹-tetrahydrocannabinol (THC, a CB₁ receptor agonist) and caffeine (an adenosine receptor antagonist). We first tested the hypothesis that an A₁–CB₁ interaction influences GABA and glutamate release in the hippocampus. We found that A₁ receptor activation attenuated the CB₁-mediated inhibition of GABA and glutamate release and this interaction was manifested at the level of G-protein activation. Using in vivo and in vitro approaches, we then investigated the functional implications of the adenosine–cannabinoid interplay that may arise following chronic caffeine consumption. Chronic administration of caffeine in mice (intraperitoneally, 3 mg/kg/day, for 15 days, >12 h before trials) led to an A₁-mediated enhancement of the CB₁-dependent acute disruptive effects of THC on a short-term spatial memory task, despite inducing a reduction in cortical and hippocampal CB₁ receptor number and an attenuation of CB₁ coupling with G protein. A₁ receptor levels were increased following chronic caffeine administration. This study shows that A₁ receptors exert a negative modulatory effect on CB₁-mediated inhibition of GABA and glutamate release, and provides the first evidence of chronic caffeine-induced alterations on the cannabinoid system in the cortex and hippocampus, with functional implications in spatial memory.

INTRODUCTION

The fine tuning of neuronal activity to suit specific cognitive functions is a major task of endogenous neuromodulators, of which adenosine and the endocannabinoids are two important examples. Both modulators are released by neurons and activate G-protein-coupled receptors (GPCRs) that represent some of the most widely and densely expressed GPCRs in the brain (Dunwiddie and Masino, 2001; Herkenham et al, 1990). In the hippocampus, the predominant adenosine and cannabinoid receptors are the A₁ and CB₁ receptors, respectively. Several forms of learning, memory, and other cognitive functions require the integrity of the hippocampal circuitry, in which A₁ and CB₁ receptors were shown to have important roles owing to their presynaptic regulation of neurotransmitter release (eg, Ohno and Watanabe, 1996; Wise et al, 2009). Moreover, in the hippocampus, these receptors are the main targets for the cognitive effects of two of the most heavily consumed psychoactive substances worldwide: caffeine and Δ⁹-tetrahydrocannabinol (THC) (Barone and Roberts, 1996; Leggett, 2006).

Caffeine is present in various dietary sources, such as coffee, tea, and soft drinks, and at moderate doses is an adenosine receptor antagonist with cognitive-enhancing properties (Fredholm et al, 1999; Ribeiro and Sebastião, 2010). As customary in most coffee consumers, long-term intake of caffeine leads to the development of tolerance to some of its acute effects by mechanisms not yet fully understood, although most studies found an increased number of A₁, but not A_2A receptors, in several brain areas (Jacobson et al, 1996). Chronic caffeine intake has also been associated with increased behavioral effects of some drugs of abuse, for example, amphetamine and cocaine (Gasior et al, 2000; Justinova et al, 2009). THC is the main psychoactive constituent of the cannabis plant, which is consumed recreationally or used for medicinal purposes; it mainly activates cannabinoid CB₁ receptors in the central nervous system to produce motor- and cognitive-disrupting effects (see Pertwee, 2008).

Hippocampal CB₁ receptors are primarily found in presynaptic terminals of cholecystokinin (CCK)-expressing γ-aminobutyric acidergic (GABAergic) interneurons from the CA1 and CA3 subfields (Hájos et al, 2000; Katona et al, 1999). CCK-expressing GABAergic interneurons regulate the temporal coordination in the activity of principal cell assemblies, which is critical for the integrity of hippocampal-dependent memory (Freund and Katona, 2007). Accodingly, it was recently shown that presynaptic CB₁ receptors at GABAergic, but not glutamatergic, neurons are required for THC-induced amnesia (Puighermanal et al, 2009). In contrast, A₁ receptors mostly affect excitatory synaptic transmission (Dunwiddie and Fredholm, 1989; Sebastião et al, 1990), having no direct influence upon GABAergic transmission in mature hippocampal neurons (Jeong et al, 2003; Lambert and Teyler, 1991; Li and Henry, 2000; Yoon and Rothman, 1991) or on GABA release from isolated nerve terminals (Cunha and Ribeiro, 2000). However, A₁ receptors are present in hippocampal GABAergic interneurons (Ochiishi et al, 1999), in which they control the actions of vasoactive intestinal peptide (Cunha-Reis et al, 2008).

Both A₁ and CB₁ receptors regulate synaptic transmission through activation of G-protein α_i/o-subunits (Straiker et al, 2002), which inhibit adenylyl cyclase, block voltage-gated calcium channels (VGCCs), and activate inwardly rectifying potassium channels (Dunwiddie and Masino, 2001; Howlett, 2005). In the cerebellum, A₁ receptors modulate the motor incoordination effects induced by acute administration of THC or CB₁ receptor agonist CP55,940 (Dar, 2000; Dar and Mustafa, 2002; DeSanty and Dar, 2001). Furthermore, prolonged intracerebellar administration of a CB₁ or A₁ agonist induces cross-tolerance (DeSanty and Dar, 2001), and similar observations were obtained in two subsequent studies (Kouznetsova et al, 2002; Selley et al, 2004). A more recent study observed that CB₁-mediated inhibition of excitatory synaptic transmission in the hippocampus is modulated by endogenous adenosine, through A₁ receptor activation (Hoffman et al, 2010; but see Serpa et al, 2009). These previous findings raised the hypothesis that a functional interaction between A₁ and CB₁ receptors in the hippocampus may have cognitive and pathophysiological implications, particularly for the effects of cannabis and caffeine consumption in humans.

This study initially focused upon the possibility that an A₁–CB₁ interaction influences GABA and glutamate release. We found that A₁ receptor activation attenuated the CB₁-mediated inhibition of GABA and glutamate release, and that this interaction is manifested at the level of G-protein activation. We then evaluated the functional consequences of chronic caffeine administration on the memory deficits induced by acute THC administration. Caffeine (intraperitoneally, 3 mg/kg/day, for 15 days, >12 h before trials) increased A₁ receptor levels, and did not by itself cause measurable effects on spatial memory, but led to an A₁-mediated exacerbation of the CB₁-dependent acute effects of THC in a spatial memory task.

MATERIALS AND METHODS

Animals

Adult male 6–8 weeks old Wistar rats (Harlan Interfauna Iberica, Barcelona, Spain) and 12–16 weeks old C57Bl/6J mice (Harlan-Olac, Bicester, UK) were used. Animals were housed in a temperature- and humidity-regulated room with a 12 h dark/light cycle, and free access to food and water. Experiments were performed during the light phase. All experimentation followed the UK Animals (Scientific Procedures) Act, 1986, Portuguese and European Union law concerning animal care. C57Bl/6J mice were used in all experiments involving chronic caffeine administration for logistic advantages and because mice have been extensively used in behavioral studies where the systemic effects of cannabinoids on motor and cognitive function have been assessed (Lichtman et al, 2002).

Drugs

4-Amino-[2,3-³H]butyric acid ([³H]GABA), L-[G-³H]glutamic acid ([³H]glutamate), 1,3-[³H]-dipropyl-8-cyclopentylxanthine ([³H]-DPCPX), and [³H]SR141716A were obtained from GE Healthcare Life Sciences (Buckinghamshire, UK). Guanosine 5′-[γ-³⁵S]-thio)triphosphate ([³⁵S]-GTPγS) was from Perkin-Elmer NEN Radiochemicals (Boston, MA, USA). Adenosine deaminase (ADA, EC 3.5.4.4) was from Roche Diagnostics (Indianapolis, IN, USA). THC (>98% purity) was from THC Pharm (Frankfurt, Germany) or Tocris Bioscience (Bristol, UK). CdCl₂, caffeine (anhydrous base), GABA, aminooxyacetic acid (AOAA), guanosine diphosphate (GDP), and guanosine 5′-O-[γ-thio]triphosphate (GTPγS) were from Sigma (St Louis, MO, USA). 1-(4,4-Diphenyl-3-butenyl)-3-piperidinecarboxylic acid hydrochloride (SKF89976A), N⁶-cyclopentyladenosine (CPA), 8-cyclopentyl-1,3-dipropylxanthine (DPCPX), 8-[4-[(2-aminoethyl) amino]carbonylmethyloxyphenyl] xanthine (XAC), (R)-(+)-[2,3-dihydro-5-methyl-3-(4-morpholinylmethyl) pyrrolo[1,2,3-de]-1,4-benzoxazin-6-yl]-1-naphthalenyl-methanone mesylate (WIN55,212-2), (3S)-[2,3-dihydro-5-methyl-3-(4-morpholinylmethyl)pyrrolo[1,2,3-de]-1,4-benzoxazin-6-yl]-1-naphthalenyl-methanone monomethanesulfonate (WIN55,212-3), and N-(piperidin-1-yl)-5-(4-iodophenyl)-1-(2,4-dichlorophenyl)-4-methyl-1H-pyrazole-3-carboxamide (AM251) were from Tocris Bioscience. (RS)-4-amino-3-(4-chlorophenyl)butanoic acid (baclofen) was from Ascent Scientific (Bristol, UK). All other reagents were from Sigma. For in vitro experiments, non-water-soluble drugs were dissolved in dimethyl sulfoxide (DMSO), aliquoted, and stored at −20 °C. The amount of DMSO in solutions was normalized for all conditions in any given experiment, and always <0.02%.

Synaptosomal Preparation

For each experiment, hippocampal synaptosomes were prepared from two rats, or cortico-hippocampal synaptosomes from one mouse. Animals were decapitated under halothane anesthesia and synaptosomes were prepared as described previously (Assaife-Lopes et al, 2010), with modifications. Briefly, tissue was dissected in a continuously oxygenated (95% O₂, 5% CO₂) ice-cold artificial cerebrospinal fluid (aCSF) of the following composition (mM): NaCl 125, KCl 3, NaH₂PO₄ 1, NaHCO₃ 25, CaCl₂ 1.5, MgSO₄ 1.2, glucose 10, pH 7.4. Samples were homogenized in ice-cold 0.32 M sucrose solution containing 1 mM EDTA and 10 mM HEPES, pH 7.4. The homogenate was then centrifuged at 3000 g for 10 min, and the supernatant obtained was centrifuged at 14 000 g for 12 min to obtain a stratified pellet, containing synaptosomes (McMahon et al, 1992; Phelan and Gordon-Weeks, 1997).

[³H]Neurotransmitter Release Assays

For [³H]GABA release assays, the pellet was resuspended and synaptosomes were incubated for 20 min, at 37 °C, with [³H]GABA (1.5 μCi/ml, 1.85 nM), and 0.625 μM of unlabelled GABA to decrease specific activity of [³H]GABA to 2.3 μCi/nmol. Incubation and superfusion solutions consisted of oxygenated aCSF containing the GABA transaminase inhibitor AOAA. For [³H]glutamate experiments, the synaptosomal pellet was resuspended in aCSF, which did not contain AOAA, and synaptosomes were incubated for 5 min, at 37°C with 10 μCi/ml [³H]glutamate. Synaptosomes were then layered over GF/C filters (Milipore, MA, USA) on an eight-chamber superfusion (0.8 ml/min) apparatus (Raiteri et al, 1974). This constant and rapid flow rate washes out endogenously released substances, thus ensuring drug effect specificity (see Raiteri and Raiteri, 2000). After a 30-min washout period, samples were continuously collected for 36 min, in 2-min fractions. Synaptosomes were stimulated during 2 min with 15 mM K⁺ (isomolar substitution of Na⁺ with K⁺) at the 6th (S₁) and 24th (S₂) minutes of collection time. CB₁ agonists were added to the superfusion medium from the 18th minute onward, to measure effects on S₂. The CB₁ antagonist, AM251, or A₁ and GABA_B receptor ligands were added from the 15th minute of the washout period onward, that is, present during S₁ and S₂ in order to assess their ability to modify the effect of WIN55,212-2 (applied before S₂). Each condition was tested in duplicate, as commonly accepted in this paradigm (eg, Cunha and Ribeiro, 2000). Under similar conditions, the percentage of GABA and glutamate in the K⁺-evoked outflow is >90% of the total tritium in the sample (Cunha et al, 1997; Lopes et al, 2002). Fractional [³H]neurotransmitter release was expressed as the percentage of total radioactivity present in the synaptosomes at each time point (fractional release). The amount of tritium released after each pulse of K⁺ (S₁ or S₂) was calculated by integration of the peak area. Effects were calculated by normalizing the S₂/S₁ values of corresponding controls from the same batch of synaptosomes to 0% effect. For example, the effect of WIN55,212-2 (added before S₂), in the presence of CPA (during S₁ and S₂), was calculated using the S₂/S₁ of CPA alone (during S₁ and S₂) as a control, which was obtained from the same experiment and batch of synaptosomes.

Binding Assays

Rat hippocampal or mouse cortico-hippocampal membranes were prepared as described previously (eg, Cunha et al, 1999), with modifications. Tissue was homogenized in ice-cold 0.32 M sucrose solution containing 2 mM EGTA, 1 mM DTT, and 50 mM Tris, pH 7.6. The homogenate was centrifuged at 1000 g for 10 min, and the supernatant obtained was centrifuged at 14 000 g for 12 min. The pellets were resuspended in assay buffer and incubated with 4 U/ml ADA for 30 min at 37 °C, followed by centrifugation at 14 000 g for 12 min and resuspension in assay buffer. Assay buffer composition, in mM, for radioligand binding assays was: Tris 50, MgCl₂ 2, pH 7.4 and for [³⁵S]GTPγS binding assays it was: Tris 50, MgCl₂ 5, NaCl 100, EGTA 0.2, pH 7.4. Protein content was determined by the Bradford method (Bradford, 1976). For [³H]DPCPX binding assays, membranes (40 μg protein) were incubated for 1 h at room temperature in a final incubation volume of 300 μl containing 4 U/ml ADA, and using 2 μM of XAC to measure nonspecific binding. For [³H]SR141716A binding, membranes (50 μg protein) were incubated for 1 h at 30 °C in a final volume of 300 μl containing 1 mg/ml BSA and using 1 μM of AM251 to measure nonspecific binding. For [³⁵S]GTPγS binding assays, membranes (10 μg of protein) were incubated with 0.1 nM [³⁵S]GTPγS and 0.1 nM–10 μM of CB₁ agonist, in the absence or presence of 100 nM CPA or 100 μM baclofen, in assay buffer containing 30 μM GDP, in a total volume of 500 μl, for 30 min at 37 °C. At this GDP concentration, WIN55,212-2 has been shown to induce high-affinity [³⁵S]GTPγS binding (Breivogel et al, 1998). Specific binding was calculated by subtracting nonspecific binding obtained by incubation with 10 μM GTPγS. The effect of co-application of CPA or baclofen with CB₁ agonists was calculated by subtracting the increase in [³⁵S]GTPγS induced by CPA or baclofen alone. The reactions were stopped by vacuum filtration through GF/C filters, followed by washing with ice-cold buffer.

In Vivo Drug Administration

Mice were randomly assigned to various groups and habituated to the handling during 5 days before testing began. For chronic treatment with caffeine, animals received caffeine (3 mg/kg/day), or vehicle (saline: 0.9% NaCl), >12 h before trials, for at least 15 days before experimental days, and throughout the course of behavioral testing, in order to avoid withdrawal effects. Total caffeine exposure was for 22–24 days, and euthanization occurred 24 h after last injection. The half-life of caffeine for doses lower than 10 mg/kg ranges from 0.7 to 1.2 h in the rat and mouse (Fredholm et al, 1999); therefore, the estimated concentrations of caffeine present in plasma or brain during behavioral testing were negligible. For acute administration, animals received a single dose of vehicle (8% Tween-80 in saline), THC (5 mg/kg), AM251 (3 mg/kg), DPCPX (1 mg/kg), or WIN55,212-2 (1 mg/kg). THC was prepared in Tween-80 as described previously (Pertwee et al, 1992); AM251, DPCPX, and WIN55,212-2 were suspended in the vehicle and carefully sonicated. All drugs were given by intraperitoneal injection in a volume of 2 ml/kg weight. The concentration of Tween-80 used was previously shown not to affect motor activity in mice (Castro et al, 1995).

Water Maze Experiments (Trials to Criterion Task)

We performed two separate sets of water maze experiments, in which mice were randomly assigned to four experimental groups of seven to eight subjects (total of 57 animals). The protocol is a version of the Morris water maze test that is sensitive to hippocampal-dependent short-term spatial learning (Chen et al, 2000; Daumas et al, 2007). To form a stable representation of the environment, mice were first trained to quickly find a hidden platform at a fixed platform location for 5 consecutive days. Subjects then performed several tasks, each consisting of a new platform position. Each animal was given a maximum of eight trials per day, to perform the task until reaching a performance criterion of ⩽7 s average latency on three consecutive trials. A 15 min intertrial interval was applied, during which animals were allowed to dry under a ceramic heat lamp. Once the criterion was reached, trials stopped and a new task began on the following testing day. Animals first performed 4–5 training tasks in order to learn to optimize their search strategies, and then the effects of acute drug administration were tested in separate tasks, as described in the Results section.

A black infrared-translucent Perspex tank (1 m in diameter) of water (temperature, 22±1 °C) was placed over an infrared lightbox (Tracksys, Nottingham, UK) in a room with various visible external cues. A transparent platform was ∼0.5 cm below the water surface and its position varied between several possible locations, on two concentric circles, according to the original protocol. An infrared-sensitive automated tracking system (Noldus Ethovision 7.0, Noldus Information Technology, Wageningen, the Netherlands) monitored all performances.

Statistical Analysis

Statistical significance was tested using paired Student's t-test, one- or two-way analysis of variance (ANOVA), followed by Dunnett's or Tukey's post hoc tests, as indicated. The two-way ANOVA and post hoc tests were performed using the Predictive Analytics Software 18.0 (SPSS, an IBM Company, Chicago, IL). GraphPad Prism 5.0 (GraphPad Prism Software, San Diego, CA, USA) was used for all other statistical tests and nonlinear regression curve fitting. Differences in parameters between binding curves were tested using extra sum-of-squares F test.

RESULTS

A₁ Receptor Activation Attenuates the CB₁ Receptor-Dependent Inhibition of [³H]GABA and [³H]Glutamate Release from Rat Hippocampal Nerve Terminals

In control conditions, pooling data from all experiments performed, the average basal release of [³H]GABA from rat hippocampal synaptosomes was 0.76±0.02% (n=31, average of first 6 min of collection; Figure 1a) of the total tritium retained by synaptosomes at the same time points. Depolarization of the hippocampal synaptosomes with K⁺ (15 mM) for 2 min induced a threefold increase in the [³H]GABA release with an average peak of 2.5±0.1% during S₁, and 2.3±0.1% (n=31; Figure 1a) during S₂, giving an average S₂/S₁ of 0.94±0.01. Depolarization by K⁺ mainly induced a calcium-dependent release of [³H]GABA, as blockade of VGCCs by Cd²⁺ (CdCl₂, 200 μM) inhibited its release by 70±3.0% (n=8, p<0.001, data not shown). The GABA transporters account for the remaining percentage of [³H]GABA released upon K⁺ depolarization, as blockade of GABA transporters with SKF89976a (20 μM) inhibited its release by 34±1.6% (n=8, p<0.001, data not shown).

Figure 1

Inhibition of K⁺-evoked, Ca²⁺-dependent release of [³H]GABA from rat hippocampal synaptosomes by WIN55,212-2 (WIN). (a) Fractional release of [³H]GABA evoked by two 15 mM K⁺ stimuli of 2-min duration, as indicated (S₁ and S₂); in the test assay, WIN (1 μM) was applied before S₂, as indicated by the horizontal bar. Data represent mean±standard error of mean (SEM) from 31 experiments performed in duplicate. (b) Percentage inhibition of [³H]GABA release induced by WIN (1 μM) in the absence or in the presence of the calcium channel blocker, CdCl₂ (Cd²⁺, 200 μM), or the GABA transporter inhibitor, SKF89976A (SKF, 20 μM), as indicated below each bar. Note that Cd²⁺ fully blocked the effect of WIN (n=5, ^**p<0.01, paired Student's t-test vs effect of WIN alone within the same batch of synaptosomes), whereas SKF did not alter the WIN-induced inhibition (n=5, p>0.05). Data represent mean±SEM from five experiments, performed in duplicate. (c) Concentration-dependent inhibition of K⁺-evoked release of [³H]GABA induced by WIN (0.01–10 μM) in the absence or in the presence of the CB₁ receptor antagonist, AM251 (1 μM); the effect of partial CB₁ receptor agonist, Δ⁹-tetrahydrocannabinol (THC) (1 μM), as well as of a WIN enantiomer that is inactive at the CB₁ receptor, WIN55,212-3 (WIN-3; 1–10 μM), is also shown. WIN significantly inhibited [³H]GABA release at all concentrations (p<0.01), except for the lowest concentration tested (0.01 μM, p>0.05); THC (1 μM) also significantly inhibited [³H]GABA release (p<0.01). Note that WIN-3 was devoid of the effect at 1 μM (p>0.05), but not at 10 μM (p<0.01), and that AM251 antagonized the effect of 1 μM, but not of 10 μM WIN, indicating that WIN is CB₁ receptor selective at 1 μM, but not at 10 μM. Each point represents the mean±SEM of 4–10 independent experiments performed in duplicate, except (n=2) for 0.01 and 3 μM WIN, and 1 μM WIN-3. The S₂/S₁ values from corresponding controls were taken as 0% within each experiment. P-values were obtained by a one-way analysis of variance (ANOVA) test with Dunnett post hoc, compared with control (0%). SKF, Cd²⁺, and AM251 were applied 15 min before the start of sample collection and were continuously perfused throughout the experiment, being therefore present during S₁ and S₂ (S₁+S₂); WIN, WIN-3, and THC were added before S₂ (see Materials methods for further details).

PowerPoint slide

To induce a CB₁ receptor-dependent effect on [³H]GABA release, we used the potent cannabinoid receptor agonist WIN55,212-2, which has been previously shown to inhibit evoked [³H]GABA release from hippocampal synaptosomes through a CB₁-specific mechanism, having a maximum CB₁-selective effect at 1 μM (Köfalvi et al, 2007). Application of 1 μM WIN55,212-2 6 min before S₂ caused a decrease of basal [³H]GABA outflow and inhibited evoked GABA release (Figure 1a) with an average S₂/S₁ of 0.78±0.01 (n=31) that represents an inhibition of 16.7±1.4% (n=31), when compared with control S₂/S₁ within each experiment. Blockade of VGCCs by Cd²⁺ (CdCl₂, 200 μM) completely abolished the effect of WIN55,212-2 at its maximum CB₁-specific concentration (n=5, p<0.01; Figure 1b). Conversely, blockade of GABA transporters with SKF89976a (20 μM) did not alter the effect of WIN55,212-2 (n=5; Figure 1b), which suggests that the effect of 1 μM WIN55,212-2 upon [³H]GABA release is exerted through the inhibition of Ca²⁺-dependent exocytotic release. The effect of WIN55,212-2 (0.01–10 μM) on K⁺-evoked [³H]GABA release was concentration dependent (Figure 1c). As WIN55,212-2 is known to directly block N-type VGCCs at concentrations above 1 μM (Németh et al, 2008; Shen and Thayer, 1998), we tested the specificity of its effect in our preparation using the CB₁ antagonist AM251, as well as WIN55,212-3, an enantiomer of WIN55,212-2 that does not activate the CB₁ receptor, but maintains the Ca²⁺ channel-blocking properties (Shen and Thayer, 1998). AM251 (1 μM) fully blocked the effect of 1 μM, but not of 10 μM of WIN55,212-2 (Figure 1c). Higher concentrations of AM251 were not used to avoid loss of selectivity (see Köfalvi, 2007, 2008). The enantiomer had no significant effect applied at 1 μM, but it inhibited evoked [³H]GABA release by 22.4±1.4% at 10 μM (n=5, p<0.05; Figure 1c), which indicates that the effect of 1 μM WIN55,212-2 upon [³H]GABA release is CB₁ receptor dependent. It is worth noting that the effect of 10 μM WIN55,212-2 was larger than the effect of 10 μM WIN55,212-3, which indicates that the effect of 10 μM WIN55,212-2 still encompasses a CB₁ receptor-dependent component. The partial CB₁ agonist THC (1 μM) inhibited K⁺-evoked [³H]GABA release by 8.9±0.9% (n=8, p<0.05; Figure 1c).

Consistent with previous observations (Cunha and Ribeiro, 2000), adenosine A₁ receptor activation with the selective agonist CPA (100 nM) before S₂ did not affect K⁺-evoked [³H]GABA release (2.7±3% of control S₂/S₁, n=3, p>0.05, data not shown). To evaluate the influence of A₁ receptors on the CB₁-mediated inhibition of GABA release, we tested the effect of WIN55,212-2 (applied before S₂) in the presence of CPA (100 nM, applied throughout the experiment, S₁+S₂). Under these conditions, the effect of WIN55,212-2 at 1 μM (n=10) and at 10 μM (n=7) was significantly attenuated (p<0.01 and <0.05, respectively; Figure 2a). The average effect of 1 μM WIN55,212-2 was 18±3% (n=10) and CPA attenuated this effect to 12±2% (p<0.01, n=10), which represents a 33±13% decrease in the average effect of WIN55,212-2 alone. To test if the attenuation of the effect of 10 μM WIN55,212-2 caused by CPA occurred through a Ca²⁺ channel-dependent mechanism and not involving CB₁ receptors, we performed the experiments using WIN55,212-3. Application of WIN55, 212-3 (10 μM, S₂) by itself inhibited [³H]GABA release by 22±2%, whereas in the presence of CPA (100 nM), its effect was unaltered (22±2%, n=3; Figure 2a). This indicates that the CPA-induced attenuation of the effect of WIN55, 212-2 was exerted at the CB₁ receptor-dependent component, but not upon the Ca²⁺channel-dependent mechanisms, affected by high micromolar concentrations of WIN55,212-2. The blockade of A₁ receptors with the antagonist DPCPX (50 nM) did not, on its own, alter the effects of 1 and 10 μM WIN55,212-2, but it fully prevented the CPA-induced attenuation (p>0.05, n=7, paired Student's t-test; Figure 2b).

Figure 2

A₁ receptor activation significantly attenuates the CB₁-mediated inhibition of K⁺-evoked [³H]GABA release from rat hippocampal synaptosomes. (a) Effects of WIN55,212-2 (WIN, 1 and 10 μM) and of its CB₁ receptor-inactive enantiomer, WIN55,212-3 (WIN, 3, 10 μM), in the absence or in the presence of the selective A₁ receptor agonist, CPA (100 nM), as indicated below each column. Note that CPA significantly attenuated the effect of 1 and 10 μM WIN (^**p<0.01, ^*p<0.05, respectively, compared with the effect of WIN in the absence of CPA, in the same experiments), whereas the effect of WIN55,212-3 (WIN-3) (10 μM) was unaffected by CPA (p>0.05). (b) WIN (1 and 10 μM) was tested in the absence or in the presence of selective A₁ receptor antagonist, DPCPX (50 nM), alone and in combination with CPA (100 nM). (c) WIN (1 μM) was tested in the absence or in the presence of the selective GABA_B receptor agonist, baclofen (10 μM), as indicated below the column. Note that in the presence of DPCPX, CPA did not attenuate the inhibitory effect of WIN (p>0.05), and that baclofen did not modify the effect of WIN (1 μM) (p>0.05). Bars represent the mean±standard error of mean (SEM) of 3–10 individual experiments performed in duplicate. The S₂/S₁ values from controls were taken as 0% within each experiment. P-values were obtained by paired Student's t-test, compared with corresponding controls within the same batch of synaptosomes. WIN and WIN-3 were added before S₂, whereas the other drugs were applied 15 min (CPA or baclofen) or 30 min (DPCPX) before the start of sample collection, being therefore present during S₁ and S₂ (see Materials and methods for further details).

PowerPoint slide

The signaling pathways of CB₁, A₁, and GABA_B are known to converge when co-expressed in cerebellar neurons (Selley et al, 2004). Furthermore, both CB₁ and GABA_B receptors are present in inhibitory interneurons (Katona et al, 1999; Sloviter et al, 1999), couple to the same Gα_i/o-subunits (Straiker et al, 2002), and exhibit reciprocal inhibition (Cinar et al, 2008) in hippocampal neurons. We therefore evaluated whether the A₁ receptor-dependent attenuation of the effect of WIN55,212-2 was mimicked by the activation of GABA_B receptors. As shown in Figure 2c, the effect of WIN55,212-2 (1 μM) was unchanged (p>0.05, n=5, paired Student's t-test) by the presence of the GABA_B receptor agonist baclofen (10 μM). Altogether, these results indicate that the cannabinoid CB₁ receptors in GABAergic nerve terminals are under the modulatory influence of adenosine A₁ receptors, but not GABA_B receptors.

Despite the predominant influence of CB₁ receptors in hippocampal circuitry being exerted through the inhibition of GABA release, CB₁ receptors at glutamatergic presynaptic nerve terminals (Katona et al, 2006; Kawamura et al, 2006) also inhibit the K⁺-evoked release of glutamate (eg, Cannizzaro et al, 2006; D’Amico et al, 2004; Köfalvi et al, 2007). Importantly, the regulation of excitatory synaptic transmission is the most relevant role of the hippocampal A₁ receptors (Dunwiddie and Fredholm, 1989; Sebastião et al, 1990). To investigate whether A₁ receptors also regulate the CB₁-dependent effects upon glutamate release, the influence of A₁ receptor activation on the CB₁-mediated inhibition of K⁺-evoked [³H]glutamate release was also tested. The absence of a tonic activation of A₁ receptors by endogenous adenosine was first assessed by using A₁ receptor blocker DPCPX. As Figure 3a shows, WIN55,212-2 (1 μM) inhibited the release of [³H]glutamate by 16±2% and blockade of A₁ receptors with DPCPX (50 nM) did not modify the effect of WIN55,212-2 (15±1%, p>0.05, n=4). This indicates that endogenous adenosine was effectively washed out by the continuous vertical flow of superfusion medium (see Materials and methods).

Figure 3

A₁ receptors modulate the CB₁-mediated inhibition of [³H]glutamate. (a) Blockade of A₁ receptors by DPCPX (50 nM) did not modify (p>0.05, n=4) the effect of WIN55,212-2 (WIN, 1 μM). (b) The inhibition induced by WIN (1 μM) was significantly (^**p<0.01, n=5) attenuated by CPA (100 nM), but the effect of CPA alone (applied before S₂) was not modified (p>0.05, n=4) when applied in the presence of WIN (1 μM). Bars represent the mean±standard error of mean of 4–5 individual experiments performed in duplicate. The S₂/S₁ values from controls were taken as 0% within each experiment. P-values were obtained by paired Student's t-test, compared with corresponding controls within the same batch of synaptosomes.

PowerPoint slide

As shown in Figure 3b, the inhibitory effect of 1 μM WIN55,212-2 upon glutamate release (19±2%) was significantly attenuated to 7±2% (n=5, p<0.01; Figure 3b) in the presence of CPA. We then evaluated whether the A₁ receptor-mediated inhibition of glutamate release is also under the modulatory control of CB₁ receptors, by comparing the effect of CPA (before S₂) in the absence and in the presence of 1 μM WIN55,212-2. CPA (100 nM) inhibited [³H]glutamate release by 12±4%, and this effect was not modified in the presence of 1 μM WIN55,212-2 (13±2%, n=4, p>0.05; Figure 3b). These findings further indicate that A₁ receptors negatively modulate the CB₁-mediated effects in the hippocampus and support recent evidence that CB₁-mediated inhibition of excitatory synaptic transmission in the hippocampus is modulated by A₁ receptor activation (Hoffman et al, 2010).

A₁ Receptor Activation Attenuates CB₁ Receptor-Induced Stimulation of G Proteins in Rat Hippocampal Membranes

To test whether the adenosine–cannabinoid interaction occurs at the level of G-protein activation, we measured agonist-stimulated [³⁵S]GTPγS binding in hippocampal membranes with the full agonist of CB₁ receptors, WIN55, 212-2, or with the partial agonist THC, in the absence and in the presence of A₁ receptor agonist CPA. The basal [³⁵S]GTPγS binding in the absence of WIN55,212-2 (Figure 4a) or THC (Figure 4b) is represented as 100% in the ordinates, which corresponds to (fmol/mg protein): Figure 4a—149±16 in control (○, n=7), 311±50 in the presence of CPA (▪, n=4), and 331±74 in the presence of baclofen (♦, n=4); and Figure 4b—98±4 in control (○, n=4), 255±25 in the presence of CPA (▪, n=4), and 161±15 in the presence of baclofen (♦, n=4). When applied alone, WIN55,212-2 (0.1 nM–10 μM) concentration dependently stimulated [³⁵S]GTPγS binding, with an EC₅₀ ≈255 nM and E_max=251±7% (n=7; Figure 4a and Table 1). CPA (100 nM) by itself induced a 150±12% net increase from basal [³⁵S]GTPγS binding (n=4). Co-application of WIN55,212-2 (0.1 nM–10 μM) with 100 nM CPA (Figure 4a) significantly decreased the E_max of WIN55,212-2 to 204±10% (p<0.001, n=4; Table 1), but not the EC₅₀ (≈177 nM; Table 1). This indicates a functional interaction between colocalized CB₁ and A₁ receptors in hippocampal membranes, which impacts on the ability of CB₁ receptors to activate Gα_i/o proteins. Similarly, the co-application of THC (0.1 nM–1 μM) with 100 nM CPA (Figure 4b) significantly decreased the E_max of THC from 163±8% (when applied alone) to 129±13% (n=4, p<0.05; Table 1), but not the EC₅₀.

Figure 4

Influence of A₁ or GABA_B receptor activation on CB₁-induced stimulation of G proteins, as assayed by WIN55,212-2 or Δ⁹-tetrahydrocannabinol (THC)-induced [³⁵S]GTPγS binding. Rat hippocampal membranes (10 μg protein) were incubated for 30 min at 37 °C with 30 μM GDP, 0.1 nM [³⁵S]GTPγS, and varying concentrations of (a) WIN55,212-2 (0.1 nM–10 μM) or (b) THC (0.1 nM–1 μM), alone (○) or in combination with 100 nM CPA (▪), or 100 μM baclofen (♦). E_max and log EC₅₀ values are shown in Table 1. Data represent mean percentage of basal stimulation±SEM of n=7 (○) and n=4 (▪, ♦), performed in duplicate. Non-visible error bars are within symbols.

PowerPoint slide

Table 1 E_max and Log EC₅₀ Values of Agonist-Stimulated [³⁵S]GTPγS Binding in Rat Hippocampal Membranes

To examine if other Gα_i/o-coupled receptors are also capable of interfering with the G-protein coupling of CB₁ receptors, we tested whether combined activation of CB₁ and GABA_B receptors in the hippocampus would also affect the efficacy of WIN55,212-2 in [³⁵S]GTPγS binding. To activate GABA_B receptors, we used 100 μM baclofen, which by itself induced a 118±11% net increase from basal [³⁵S]GTPγS binding (n=4), which was not significantly different from the effect of 100 nM CPA. As shown in Figure 4a and b and Table 1, 100 μM baclofen did not affect the WIN55,212-2-induced (E_max=257±12%, n=4) or THC-induced (E_max=157±9%, n=4) stimulation of [³⁵S]GTPγS binding. The reduced efficacy in stimulation of [³⁵S]GTPγS binding by CB₁ with A₁, but not GABA_B, suggests that the adenosine A₁ receptors play a specific role in modulating CB₁ signaling in hippocampal presynaptic terminals.

Chronic Caffeine Administration Increases Acute THC-Induced Spatial Memory Deficits in Mice

The evidence that A₁ receptor activation attenuates CB₁ receptor signaling raised the hypothesis that this A₁–CB₁ interplay has a functional impact upon hippocampal-dependent memory. Chronic caffeine consumption is known to induce an increase in adenosine A₁, but not A_2A receptors (reviewed by Jacobson et al, 1996). Acute systemic THC administration induces CB₁-dependent deficits in working memory (Wise et al, 2009). Therefore, we used a hippocampal-dependent, short-term spatial memory testing protocol (Chen et al, 2000; see Materials and methods) to study the effects of chronic caffeine administration on the memory deficits induced by an acute systemic THC injection in mice. Two separate sets of experiments were performed, in which caffeine (3 mg/kg/day), or vehicle, was administered >12 h before trials, and for at least 15 days before the first test with THC.

During the training phase, all subjects learned to perform efficiently at all test parameters, in both sets of experiments (see Figures 5a–c and 6a–c). The number of trials, the total latency, and total pathlength needed to reach the criterion decreased progressively from task to task reaching a plateau in the last training task. A repeated measures two-way ANOVA on these parameters revealed a significant overall learning effect. For ‘pathlength’ (representative parameter of memory performance), significance values were: F(4, 100)=10.2, p<0.0001 (Figure 5c); and F(3, 72)=13.5, p<0.0001 (Figure 6c). The average swim speed (control parameter for motor activity) was constant throughout the training tasks (Figure 5d: F(4, 100)=0.4, p=0.8; Figure 6d: F(3, 72)=1, p=0.4), with no significant differences between groups (Figure 5d: F(3, 25)=1.8, p=0.2; Figure 6d: F(3, 24)=0.2, p=0.9). In both sets of experiments, there were no differences between groups and no ‘group × task’ interaction in any parameter during the training period (p>0.05 for all parameters). Thus, chronic caffeine administration by itself did not affect memory performance or motor activity.

Figure 5

Influence of chronic caffeine administration upon the Δ⁹-tetrahydrocannabinol (THC)-induced short-term spatial memory deficits in mice. Caffeine was given daily (3 mg/kg, >12 h before trials), for 15 days before testing the effect of THC (see Methods). (a–d) Mice familiarized with the escape strategies during the first 5 tasks (training) and all groups showed improved performance in (a) the number of trials to reach criterion, (b) the escape latency, (c) swim pathlength, whereas (d) average swim speed remained constant. THC (5 mg/kg), or vehicle, was then tested in the absence (task 6) and in the presence (task 7) of AM251 (3 mg/kg). Subjects rested for 1 day off drug after task 6, to allow full metabolization of THC. For clarity of comparison between groups, symbols were nudged at tasks 6 and 7. Note that chronic caffeine exacerbated the spatial memory deficits induced by acute THC, and this effect of THC was fully prevented by previous administration of AM251. All data represent mean±standard error of mean (SEM) of n=7–8. ^*p<0.05, two-way analysis of variance, followed by Tukey's post hoc test (see text for more details).

PowerPoint slide

Figure 6

Influence of chronic caffeine administration, and involvement of the adenosine A₁ receptors, upon the acute effects of Δ⁹-tetrahydrocannabinol (THC) and WIN55,212-2 on mice tested in a short-term spatial memory task. Caffeine was given daily (3 mg/kg, >12 h before trials), for 15 days before testing the effect of THC (see Methods). (a–d) Mice familiarized with the escape strategies during the first 4 tasks (training) and all groups showed improved performance in (a) the number of trials to reach criterion, (b) the escape latency, and (c) swim pathlength, whereas (d) average swim speed remained constant. THC (5 mg/kg), or vehicle, was tested in the absence (task 5) and in the presence (task 7) of DPCPX (1 mg/kg). A control (CTR) test in the absence of acute drugs was performed at task 6 to measure whether performance levels returned to baseline values. Subjects rested for one day off drug after each test task, to allow full metabolization of THC. The effect of WIN55,212-2 (WIN, 1 mg/kg), or vehicle, was tested at task 8. For clarity of comparison between groups, symbols were nudged in tasks 5–8. Note that chronic caffeine exacerbated the spatial memory deficits induced by acute THC and WIN. The effect of THC was prevented by the previous administration of DPCPX. All data represent mean±standard error of mean (SEM) of n=7. *p<0.05, ^**p<0.01, two-way analysis of variance, followed by Tukey's post hoc tests (see text for more details).

PowerPoint slide

For the first set of experiments, the effect of THC (5 mg/kg), or vehicle, given at task 6 (30 min before first trial), as well as the modification of this effect by AM251 (3 mg/kg), given at task 7 (15 min before THC, or vehicle), are displayed in Figure 5a–d. After completion of task 6, each subject rested for 1 day to allow for the metabolic clearance of THC.

There was a significant effect of THC on ‘trials to criterion’ (F(1, 25)=4.24, p=0.05), ‘latency’ (F(1, 25)=6.98, p=0.01), and ‘pathlength’ (F(1, 25)=7.18, p=0.01), but no significant effect on ‘average speed’ (F(1, 25)<0.001, p=0.98). There were no effects of chronic caffeine treatment on all parameters, but a marginally significant ‘chronic caffeine × THC’ interaction on ‘pathlength’ (F(1,25)=3.11, p=0.09). Although acute THC injection (vs vehicle) did not induce significant effects in the control (vehicle-treated) group at any parameter, the effect of THC was exacerbated in the chronic caffeine group, on ‘latency’ (p=0.03, Tukey's post hoc; Figure 5b) and ‘pathlength’ (p=0.02, Tukey's post hoc; Figure 5c).

When mice received AM251 pretreatment, there were no significant effects of any treatment group on all parameters, indicating that the effects of THC were dependent on the activation of CB₁ receptors. These results show that chronic caffeine exacerbates the CB₁-dependent actions of THC in a short-term spatial memory task.

For the second set of experiments, the effect of THC (5 mg/kg), or vehicle, given at task 5 (30 min before first trial), the modification of this effect by DPCPX (1 mg/kg), given at task 7 (15 min before THC, or vehicle), as well as the effect of WIN55,212-2 (1 mg/kg), or vehicle, given at task 8 (30 min before first trial), are displayed in Figure 6a–d. After completion of each test task, subjects rested for one day off-drug to allow for metabolic clearance of THC. Task 6 was a control test in which no acute drug was given, to measure whether performance levels returned to baseline values 48 h after acute THC administration.

Consistent with the first set of experiments, there was a significant effect of THC on ‘trials to criterion’ (F(1, 24)=6.34, p=0.02), ‘latency’ (F(1, 24)=8.16, p=0.01), and ‘pathlength’ (F(1, 24)=9.19, p=0.01), but no significant effect on ‘average speed’ (F(1, 24)=2.09, p=0.16). On ‘pathlength’ there was also a significant effect of chronic caffeine treatment (F(1, 24)=4.82, p=0.04), and a marginally significant ‘chronic caffeine × THC’ interaction (F(1, 24)=3.29, p=0.08). Although acute THC injection (vs vehicle) did not induce significant effects in the control (vehicle-treated) group at any parameter, the effect of THC was exacerbated in the chronic caffeine group, on ‘latency’ (p=0.02, Tukey's post hoc; Figure 6b) and ‘pathlength’ (p=0.01, Tukey's post hoc; Figure 6c). There were also significant differences between THC on the chronic caffeine group vs THC in the vehicle group on ‘pathlength’ (p=0.04), and between THC on the chronic caffeine group vs the vehicle control group on ‘trials to criterion’ (p=0.05), ‘latency’ (p=0.01), and ‘pathlength’ (p=0.006).

When mice were tested 48 h after the last THC injection, there were no significant effects of any treatment group on all parameters, indicating that the effects of THC were not prevailing after this period. Importantly, when mice received a pretreatment of DPCPX, there were also no significant effects of any treatment group on all parameters, indicating that the effects of THC were reversed by the blockade of A₁ receptors.

Finally, there was a significant effect of WIN55,212-2 on ‘trials to criterion’ (F(1, 24)=7.64, p=0.01), ‘latency’ (F(1, 24)=7.50, p=0.01), and ‘pathlength’ (F(1, 24)=7.74, p=0.01), but no significant effect on ‘average speed’ (F(1, 24)=4.03, p=0.06). The acute injection of WIN55, 212-2 (vs vehicle) did not induce significant effects in the control (vehicle-treated) group at any parameter, but its effects were exacerbated in the chronic caffeine group, on ‘latency’ (p=0.03, Tukey's post hoc; Figure 6b) and ‘pathlength’ (p=0.03, Tukey's post hoc; Figure 6c). There were also significant differences between WIN55,212-2 on the chronic caffeine group vs the vehicle control group on ‘latency’ (p=0.03) and ‘pathlength’ (p=0.03). The lower effect of WIN55,212-2, compared with that of THC, could be owing to pharmacokinetic differences, as the penetration of WIN55,212-2 in the brain following intraperitoneal injection is much lower than that of THC (Petitet et al, 1999). Higher doses of WIN55,212-2 were not used to avoid nonspecific effects (Varvel and Lichtman, 2002). These findings show a significant chronic caffeine-induced, and A₁ receptor-mediated, exacerbation of the CB₁-dependent effects on short-term spatial memory.

Chronic Caffeine and A₁ Receptor Number

To quantify the influence of chronic caffeine administration upon A₁ receptor number and affinity in cortico-hippocampal membranes, [³H]DPCPX (0.1–10 nM) binding assays were performed (Figure 7). In vehicle-treated mice, the total number of specific binding sites obtained by nonlinear regression analysis (B_max) was 848±44 fmol/mg of protein, whereas the equilibrium dissociation constant (K_D) was 1.20±0.21 nM. In the chronic caffeine group, the B_max was increased to 980±50 fmol/mg of protein (p<0.05, n=6, vs vehicle group), but K_D (1.31±0.23 nM) was not significantly (p>0.05) affected. Thus, animals under chronic caffeine had ∼16% higher density of A₁ receptor without changes in affinity.

Figure 7

Saturation analysis of specific [³H]DPCPX binding (0.1–10 nM) to cortico-hippocampal membranes (40 μg protein) from chronic caffeine- (3 mg/kg/day, for 22 days, ♦) and vehicle- (○) treated mice. Inset: B_max obtained from nonlinear regression analysis. Nonspecific binding was determined at all [³H]DPCPX concentrations by the addition of 2 μM XAC. All points represent mean±standard error of mean (SEM) of n=6, and each saturation experiment was performed in duplicate. ^*p<0.05, vs control, calculated using the extra sum-of-squares F test.

PowerPoint slide

Chronic Caffeine and CB₁ Receptor Signaling in Mouse Cortico-Hippocampal Tissue

We first tested the consistency of our in vitro results between rats and mice by testing the effect of CPA (100 nM) on the WIN55,212-2-mediated inhibition of K⁺-evoked [³H]GABA release from cortico-hippocampal synaptosomes prepared from untreated mice (Figure 8). Consistent to previous observations in rats, the effect of 1 μM WIN55,212-2 alone was 19±1%, and it was significantly attenuated to 9±1% in the presence of CPA (p<0.01, n=4, paired Student's t-test; Figure 8). We then analyzed the influence of chronic caffeine administration upon the CB₁ receptor-mediated inhibition of K⁺-evoked [³H]GABA release (Figure 8). In control (vehicle-treated) mice, 1 μM WIN55, 212-2 inhibited [³H]GABA release by 17±1%, whereas in the chronic caffeine group, the effect of WIN55,212-2 was significantly reduced to 11±2% (p<0.05, n=4, paired Student's t-test; Figure 8).

Figure 8

Influence of the adenosine A₁ receptor agonist, CPA (100 nM), and of chronic caffeine administration on the CB₁-mediated inhibition of K⁺-evoked [³H]GABA release from mouse cortico-hippocampal synaptosomes. WIN55,212-2 (WIN, 1 μM) was tested in the absence and in the presence of CPA on synaptosomes prepared from untreated mice, as well as on synaptosomes prepared from chronic caffeine- (3 mg/kg/day, for 22 days) or vehicle-treated mice, as indicated below each column (see legend to Figure 1 for details). Note that the effect of WIN was significantly attenuated by CPA as well as by chronic caffeine consumption. Bars represent mean±SEM of four experiments, performed in duplicate. ^**p<0.01, compared with the effect of WIN alone; ^*p<0.05, compared with the effect of WIN in vehicle-treated mice (paired Student's t-test).

PowerPoint slide

As Figure 9 and Table 2 show, in cortico-hippocampal membranes of vehicle-treated mice, WIN55,212-2 stimulated [³⁵S]GTPγS binding (% of basal) with an EC₅₀ ≈989 nM and E_max=321±11% (n=5; Figure 9a), whereas THC had an EC₅₀ ≈41 nM and E_max=167±11% (n=4; Figure 9c). Chronic caffeine administration did not affect the EC₅₀ of WIN55,212-2 or THC, but it significantly reduced the E_max of WIN55,212-2 to 269±8%, and of THC to 135±8% (p<0.05, n=4–5, extra sum-of-squares F test; Figure 9a and c). As observed in rats (Figure 4a), the co-application of 100 nM CPA in control mice significantly decreased the E_max of WIN55,212-2 (254±13%, p<0.05, n=5, extra sum-of-squares F test; Figure 9b and Table 2), but not the EC₅₀. This reduction in WIN55,212-2-stimulated [³⁵S]GTPγS binding, caused by CPA in membranes from control animals, was of similar magnitude as the decrease observed in the chronic caffeine-treated group in the absence of CPA (Table 2). In the chronic caffeine group, CPA did not induce a further decrease in the efficacy of WIN55,212-2 to stimulate G-protein activation (E_max=252±8%; Figure 9a and b and Table 2), which may suggest that chronic caffeine treatment and A₁ receptor activation do not have additive effects upon the modification of CB₁ receptor signaling.

Figure 9

Influence of chronic caffeine administration on CB₁-induced stimulation of G proteins, as assayed by WIN55,212-2- or Δ⁹-tetrahydrocannabinol (THC)-induced [³⁵S]GTPγS binding. Cortico-hippocampal membranes (10 μg protein) from chronic caffeine- (♦,▾) and vehicle- (○, ▵) treated mice were incubated for 30 min at 37 °C with 30 μM GDP, 0.1 nM [³⁵S]GTPγS, and varying concentrations of WIN (0.1 nM–10 μM) in the absence (a) or in the presence of 100 nM CPA (b), or varying concentrations of THC (0.1 nM–1 μM) (c). E_max and log EC₅₀ values are shown in Table 2. Data represent mean percentage of basal stimulation±standard error of mean (SEM) of n=5 (a, b) and n=4 (c), performed in duplicate. Non-visible error bars are within symbols.

PowerPoint slide

Table 2 E_max and Log EC₅₀ Values of WIN55,212-2 and THC-stimulated [³⁵S]GTPγS Binding in Mouse Cortico-hippocampal Membranes

The basal [³⁵S]GTPγS binding in the absence of WIN55,212-2 (Figure 9a and b) or THC (Figure 9c) is represented as 100% in the ordinates, which corresponds to (fmol/mg protein): Figure 9a and b—143±10 (○, n=5), 154±15 (♦, n=5), 324±13 (▵, n=5), and 327±25 (▾, n=5); Figure 9c—53±9 (○, n=4) and 67±10 (♦, n=4). CPA (100 nM), by itself, enhanced [³⁵S]GTPγS binding by 122±9% (over twofold net increase from basal binding) in membranes prepared from vehicle-treated subjects, and by 111±7% in the chronic caffeine group (p>0.05, n=5, Student's t-test, data not shown); hence, the ability of A₁ receptors to activate G proteins is unaltered in chronic caffeine-treated mice. Accordingly, there were no statistically significant differences in the bottom of the nonlinear regression binding curves between chronic caffeine- and vehicle-treated animals, in the presence of CPA (p>0.05, n=5, extra sum-of-squares F test; Figure 9b).

Chronic Caffeine and CB₁ Receptor Number

The effects of chronic caffeine administration upon CB₁ signaling in vitro suggested that CB₁ receptor number and/or affinity are decreased in these mice. We then directly analyzed the effect of chronic caffeine administration in CB₁ receptor number and affinity by performing [³H]SR141716A (0.1–4 nM) saturation binding assays in tissue collected from mice used in the behavioral experiments. The nonlinear regression analysis of [³H]SR141716A binding to cortical membranes of vehicle-treated mice showed a B_max=1425±123 fmol/mg of protein and a K_D=1.4±0.3 nM (n=10; Figure 10a). In hippocampal membranes, the B_max of [³H]SR141716A was 1322±97 fmol/mg of protein and the K_D=1.0±0.2 nM (n=5; Figure 10b). In the chronic caffeine group, the B_max of [³H]SR141716A binding was lower (p<0.05, compared with vehicle group), and this reduction was observed both in cortical membranes (B_max=1151±65 fmol/mg of protein, n=10) and hippocampal membranes (B_max=1089±57 fmol/mg of protein, n=5). There were no significant differences in affinity, as in the chronic caffeine group the K_D values for [³H]SR141716A binding were 1.0±0.2 nM in the cortical and 0.9±0.1 nM in the hippocampal membranes.

Figure 10

Saturation analysis of specific [³H]SR141716A binding (0.1–4 nM) to (a) cortical and (b) hippocampal membranes (50 μg protein) from chronic caffeine- (3 mg/kg/day, for 22 days) and vehicle-treated mice. Insets: B_max values, obtained from nonlinear regression analysis. Nonspecific binding was determined at all [³H]SR141716A concentrations by the addition of 1 μM AM251. All points represent mean±standard error of mean (SEM) of 5–10 experiments, each performed in duplicate. ^*p<0.05, vs control, calculated using the extra sum-of-squares F test.

PowerPoint slide

DISCUSSION

This study shows that adenosine A₁ receptors located in GABAergic and glutamatergic nerve terminals of the hippocampus exert a negative modulatory effect on the cannabinoid CB₁ receptor-mediated inhibition of GABA and glutamate release. CB₁-mediated G-protein activation is also impaired by A₁ receptor activation. In addition, chronic administration of caffeine leads to an A₁ receptor-mediated enhancement of the CB₁-dependent effects of THC upon short-term spatial memory, despite a reduction in CB₁ receptor number and signaling. This provides first evidence for chronic caffeine-induced alterations in cannabinoid actions in the cortex and hippocampus.

The CB₁–A₁ receptor cross-talk might occur at the G-protein level, as A₁ receptor activation with CPA reduced the efficacy of CB₁ receptor agonists to stimulate [³⁵S]GTPγS binding in the hippocampus. This is in accordance with a previous observation that simultaneous application of CB₁ and A₁ agonists produces less than additive stimulation of [³⁵S]GTPγS binding in cerebellar membranes (Selley et al, 2004). Similarly to the A₁ receptors, GABA_B receptors couple to Gα_i/o proteins and are expressed in the same interneuron populations as CB₁ receptors (Neu et al, 2007; Sloviter et al, 1999). However, CB₁ receptor-mediated signaling, assessed either as inhibition of [³H]GABA release or stimulation of [³⁵S]GTPγS binding, was unaffected by GABA_B receptor activation, which indicates that the modulation of CB₁ receptor signaling by A₁ receptors is not shared by all Gα_i/o-coupled receptors.

The CB₁ receptor agonist WIN55,212-2 inhibited calcium-dependent [³H]GABA release with a maximum specific effect at 1 μM, in agreement with previous studies (Katona et al, 2000; Köfalvi et al, 2007). There are clear differences in the magnitude of the reported effects of WIN55,212-2 in studies using different methodologies. For example, we and others (Köfalvi et al, 2007) observed that 1 μM WIN55,212-2 induces 15–20% inhibition of K⁺-evoked [³H]GABA release from rat hippocampal synaptosomes, whereas several reports show that the same concentration of WIN55,212-2, by activating presynaptic CB₁ receptors, inhibits GABAergic inhibitory postsynaptic currents (IPSCs) in rat hippocampal slices by ∼50% (eg, Hájos et al, 2000; Hoffman and Lupica, 2000; Wilson and Nicoll, 2001). These differences are likely owing to a combination of factors. The main reason possibly lies in the fact that [³H]GABA release assays provide a quantitative measurement of the amount of GABA released from the whole population of GABAergic nerve terminals at the hippocampus, whereas patch-clamp techniques provide a quantification of endogenous GABA release by measuring the post-synaptic responses of a single hippocampal pyramidal neuron. In addition, differences are likely owing to the type of stimulus used (electrical vs high K⁺), the time and length of WIN55,212-2 application, and to an amplifying effect of multiple afferents upon IPSC measurements. The effect of 1 μM WIN55,212-2 on the release of [³H]GABA from rat hippocampal slices (Katona et al, 1999) is also larger than in synaptosomes. Again, a longer exposure time (6 vs 18 min) to WIN55,212-2 and/or the amplification by intrinsic circuits in the slices is a likely explanation for these differences.

The CB₁ receptor-mediated modulation of GABA release from hippocampal CCK-positive interneurons, which express large quantities of CB₁ receptors, is a critical mechanism for spatial and episodic memory, as these interneurons regulate the temporal coordination of principal cell assemblies (Hájos et al, 2000; Robbe and Buzsáki, 2009; Robbe et al, 2006). However, the CCK-expressing interneuron populations mostly receive input from glutamatergic neurons (see Freund and Buzsáki, 1996; Freund and Katona, 2007), which also express CB₁ (Katona et al, 2006; Kawamura et al, 2006) and A₁ (Ochiishi et al, 1999) receptors. We found that A₁ receptor activation also attenuates CB₁ receptor-mediated inhibition of glutamate release from hippocampal synaptosomes. It was recently reported that endogenous adenosine, by activating A₁ receptors, regulates CB₁-mediated inhibition of glutamatergic synaptic transmission (Hoffman et al, 2010; but see Serpa et al, 2009). Thus, A₁ and CB₁ receptors also interact at glutamatergic neurons, which indicates that the inhibitory effect of A₁ receptor activation upon the CB₁-dependent stimulation of [³⁵S]GTPγS binding might be derived from an A₁–CB₁ receptor interaction at both GABAergic and glutamatergic neurons. Interestingly, WIN55,212-2 did not attenuate the inhibitory action of CPA upon glutamate release, suggesting that the modulatory action of A₁ receptors upon CB₁ receptors is not reciprocal.

The relevance of the GABAergic circuitry for the CB₁ receptor-mediated influences upon memory function became firmly established after the demonstration that intraperitoneal THC administration disrupts hippocampal-dependent memory through the activation of CB₁ receptors (Wise et al, 2009) in GABAergic, but not glutamatergic, neurons (Puighermanal et al, 2009). We now show that chronic administration of a moderate dose of caffeine leads to increased levels of A₁ receptors in the cortico-hippocampal membranes, and to an A₁ receptor-mediated increase of the disruptive effects of acute THC in a hippocampal-dependent short-term spatial memory task. This finding points toward a significant functional relevance of the cross-talk between A₁ and CB₁ receptors in the hippocampus. Interestingly, the motor impairments induced by THC are enhanced by acute activation of A₁ receptors (Dar, 2000). In contrast, acute administration of caffeine antagonizes THC-induced changes in cortico-hippocampal EEG wave recordings (Consroe et al, 1976). Several studies show that chronic exposure to adenosine receptor antagonists causes similar actions to acute agonist exposure (see Jacobson et al, 1996; Von Lubitz et al, 1993), whereas acute administration of caffeine is expected to have opposite effects to acute agonist exposure. The timing of caffeine administration and the presence of caffeine in the blood during testing must also be taken into account when comparing data from different studies. The behavioral tests now reported were performed in the absence of relevant plasma concentrations of caffeine (>12 h after caffeine injection), which was given 2 h after the last behavioral trial, to prevent effects on memory consolidation (Angelucci et al, 2002). It is therefore not surprising that acute caffeine administration prevents THC-induced effects (Consroe et al, 1976), whereas chronic caffeine exposure exacerbates the memory disruption induced by CB₁ receptor agonists (present work). In a recent study, chronic administration of a high dose of caffeine (210 mg/kg/day) in rats was shown to potentiate CB₁-dependent effects at striatal GABAergic, but not glutamatergic, synapses (Rossi et al, 2009). However, it is difficult to draw a comparison with this study, given the differences in the experimental approach, namely the dose of caffeine used, which is not adenosine receptor-selective and is more than about 70 times higher than the equivalent daily human intake. Exposure to high doses of caffeine (∼100 mg/kg/day) leads to altered brain levels of several receptors (Shi et al, 1993, 1994), inhibit phosphodiesterases, and may even block GABA_A receptors, among others (see Daly and Fredholm, 1998).

The conclusion that A₁ receptors are involved in the chronic caffeine-induced exacerbation of the effects of THC presently reported is reinforced by the finding that A₁ receptor blockade with DPCPX fully prevented the effects of THC in the chronic caffeine group. DPCPX was administered at a dose that occupies A₁ receptors (Baumgold et al, 1992; Hooper et al, 1996) while not affecting motor activity (present work; Von Lubitz et al, 1993). In addition, DPCPX by itself had no effects in the absence of THC, which suggests that A₁ receptors do not directly influence short-term spatial memory. Furthermore, acute application of DPCPX did not influence the action of THC in vehicle-treated animals, which further supports previous evidence (see above) that chronic and acute blockade of A₁ receptors have different functional consequences. The effects of DPCPX also exclude the involvement of A_2A receptors, which are known to modulate the actions of CB₁ receptors in the striatum (Carriba et al, 2007; Tebano et al, 2009). A₁ and A_2A receptors have similar affinities for caffeine (Fredholm et al, 1994), yet the expression of A_2A receptors in the hippocampus and cortex is much lower than that of A₁ receptors (reviewed by Ribeiro et al, 2002). Furthermore, chronic caffeine exposure does not alter the expression of A_2A receptors (Jacobson et al, 1996).

The increase in A₁ receptor expression caused by moderate doses of chronically administered caffeine results from prevention of tonic adenosine-mediated receptor downregulation (see Fredholm et al, 1999). The dose of caffeine we have administered to mice is equivalent to the estimated US average human daily caffeine consumption (Barone and Roberts, 1996) and, in addition to the expected increase in A₁ receptor levels, it also caused a decrease of cortical and hippocampal CB₁ receptors. Accordingly, in chronic caffeine-treated mice there was a reduction in the CB₁ receptor-mediated inhibition of GABA release and stimulation of G-protein activation. As tonic activation of A₁ receptors was prevented through elimination of endogenous adenosine by vertical superfusion in the [³H]GABA release assays, and by ADA in the [³⁵S]GTPγS binding assays, it is unlikely that chronic caffeine-induced A₁ receptor upregulation could be responsible for reduction of CB₁-dependent actions in the in vitro assays. Most probably, chronic caffeine intake, by inducing an imbalance in adenosinergic signaling, disturbs the A₁–CB₁ cross-talk, which reflects in CB₁ receptor downregulation. Independently of the exact mechanisms involved, it is clear that CB₁ receptors are affected after chronic caffeine exposure.

Given that chronic caffeine decreases CB₁ and increases A₁ receptor levels, and that activation of A₁ receptors inhibited the CB₁-mediated actions in the in vitro assays, it was somewhat surprising that the memory impairment caused by the CB₁ receptor agonists was exacerbated by chronic caffeine intake. It is therefore evident that changes observed in vitro do not necessarily reflect, in a linear way, the effects upon the integrated hippocampal circuitry in vivo. An imbalance in GABAergic transmission resulting from the chronic caffeine-induced alterations of A₁ and CB₁ levels may have occurred, leading to some adaptive changes in the pyramidal cells and/or in the parvalbumin-expressing (PV) GABAergic neurons, which do not express CB₁ receptors (Katona et al, 1999). Interestingly, the blockade of GABAergic transmission was shown to reverse the cognitive effects of acute THC in vivo (Varvel et al, 2005). A critical imbalance in the temporal coordination of pyramidal cell firing could have become evident when THC was administered, if there was an enhanced sensitivity to the fast spiking activity of PV cells; hence, leading to increased inhibition of pyramidal cell firing.

In summary, this work highlights two relevant factors influencing cannabinoid CB₁ signaling in the hippocampus: the activity of A₁ receptors and the chronic consumption of caffeine. This A₁–CB₁ receptor interaction therefore points toward the possibility that the pathophysiological or therapeutically relevant actions operated by CB₁ receptors can be significantly affected by interference with A₁ receptor activity, as is the case of chronic caffeine intake.