Arousal Effect of Caffeine Depends on Adenosine A_2A Receptors in the Shell of the Nucleus Accumbens

Abstract

Caffeine, the most widely used psychoactive compound, is an adenosine receptor antagonist. It promotes wakefulness by blocking adenosine A_2A receptors (A_2ARs) in the brain, but the specific neurons on which caffeine acts to produce arousal have not been identified. Using selective gene deletion strategies based on the Cre/loxP technology in mice and focal RNA interference to silence the expression of A_2ARs in rats by local infection with adeno-associated virus carrying short-hairpin RNA, we report that the A_2ARs in the shell region of the nucleus accumbens (NAc) are responsible for the effect of caffeine on wakefulness. Caffeine-induced arousal was not affected in rats when A_2ARs were focally removed from the NAc core or other A_2AR-positive areas of the basal ganglia. Our observations suggest that caffeine promotes arousal by activating pathways that traditionally have been associated with motivational and motor responses in the brain.

Introduction

Caffeine is the most consumed psychoactive compound in the world. It is readily available through dietary products, such as coffee, tea, soft drinks, and chocolate treats, but is also added to nonprescription medications, such as pain-relievers and cold remedies. Regardless of the source, the worldwide average caffeine consumption has been estimated to be just under 80 mg/d, although the levels of intake in countries such as Sweden and Finland are in the range of 400 mg of caffeine per day (Fredholm et al., 1999).

Caffeine is widely used to promote wakefulness and to counteract fatigue. Caffeine binds with very similar affinity to adenosine A₁ (A₁Rs) and A_2A (A_2ARs) receptors, and, at doses commonly consumed by humans, adenosine actions at both receptors are antagonized. Adenosine is an inhibitory neuromodulator involved in sleep–wake regulation (Porkka-Heiskanen et al., 1997; Huang et al., 2011). Using global genetic knock-outs of A₁Rs and A_2ARs, in which the receptor is deleted from the entire animal, we demonstrated previously that the A_2AR, but not the A₁R, mediates the arousal effect of caffeine (Huang et al., 2005). However, the neurons with A_2ARs on which caffeine acts to produce wakefulness have not yet been identified.

A_2ARs are densely expressed on striatopallidal neurons in the indirect pathway of the basal ganglia (BG), in which dopamine D₂ receptors (D₂Rs) are coexpressed with the A_2ARs and contribute to the control of locomotor activity, motivation, and addiction, all activities that require wakefulness (Rosin et al., 1998; Svenningsson et al., 1999a). The striatopallidal neurons also facilitate movement by operating in parallel with dopamine D₁ receptor (D₁R)-bearing striatonigral neurons in the direct pathway of the BG. Abilities to maintain arousal are compromised under low dopamine conditions, such as Parkinson's disease (Arnulf et al., 2002; Qu et al., 2010), but the extent to which A_2ARs in the BG contribute to the regulation of wakefulness is not known and the role of A_2ARs in other brain regions is unclear.

In the present study, we used site-specific gene deletion strategies based on the Cre/loxP technology in mice and also silenced focally the expression of A_2ARs in rats by using stereotaxic microinjections of adeno-associated virus (AAV) carrying short-hairpin RNA (shRNA). We found that the A_2ARs in the shell region of the nucleus accumbens (NAc) are responsible for the effect of caffeine on wakefulness.

Materials and Methods

Genetic mouse models.

Animals were handled according to the NIH Guide for the Care and Use of Laboratory Animals and in accordance with the protocols approved by the Institutional Animal Care and Use Committee (IACUC) at the Boston University School of Medicine, the Legacy Research Institute IACUC, and the Animal Research Committee at the Osaka Bioscience Institute. All mice (weighing 24–28 g, 11–13 weeks old) and male Sprague Dawley rats (weighing 150–180 g, 6 weeks old; Shizuoka Laboratory Animal Center) used in the present study were housed at a constant temperature (24 ± 0.5°C) with a relative humidity of 60 ± 2% on an automatically controlled 12 h light/dark cycle (light on at 7:00 A.M.). Three genetic mouse lines on a C57BL/6 background were used in the present study: (1) global A_2AR knock-out mice (A_2AR KO) (Chen et al., 1999), (2) basal ganglia–A_2AR knock-out mice (BG–A_2AR KO), exclusively lacking BG A_2ARs (Shen et al., 2008), and (3) a mouse line with a loxP-site-inserted A_2AR gene that is amenable to conditional disruption by the injection of Cre recombinase-expressing AAV.

Vigilance state assessment using electroencephalogram/electromyogram/locomotor activity recordings.

Assessment of vigilance states was performed on adult male conditional A_2AR KO mice and their respective WT littermates (n = 4–5, per genotype and drug dose). Under anesthesia using 1.5% isoflurane in N₂O/O₂ (2:1), mice were implanted with electroencephalogram (EEG) and electromyogram (EMG) electrodes for polysomnographic recordings. To monitor EEG signals, two stainless steel EEG recording screws (Plastics One) were implanted epidurally over the frontal cortical area (1 mm anterior to bregma, 1.5 mm lateral to the midline) and over the parietal area (2 mm posterior to bregma, 3 mm lateral to midline) of the right hemisphere. EMG activity was monitored by stainless steel, Teflon-coated wires (0.2 mm in diameter; Plastics One) bilaterally placed into both trapezius muscles. Finally, the electrode assembly was anchored and fixed to the skull with Super6-Bond (Sun Medical Co.) and dental cement. After a 10 d recovery period, the mice were placed in experimental cages for a 4 d habituation/acclimatization period with connection of counterbalanced recording leads.

All mice that were subjected to EEG recordings received vehicle and drug treatment on 2 consecutive days. On day 1, mice were treated with vehicle (saline, intraperitoneally) at 9:00 A.M., and the 24 h recordings performed on day 1 were used as baseline data. On day 2, mice were treated with caffeine (intraperitoneally, in a volume of 10 ml/kg body weight), and EEG/EMG signals were recorded for 24 h. The EEG/EMG signals were amplified and filtered (EEG, 0.5–30 Hz; EMG, 20–200 Hz), then digitized at a sampling rate of 128 Hz, and recorded using SLEEPSIGN software (Kohtoh et al., 2008). In addition, locomotor activity (LMA) was recorded with an infrared photocell sensor (Biotex). The vigilance states were scored offline by 10 s epochs into three stages, including waking, rapid-eye movement (REM) sleep, and non-REM (NREM) sleep, according to standard criteria (Mizoguchi et al., 2001). As a final step, defined vigilance stages were examined visually and corrected when necessary.

Assessment of activity and inactivity.

Assessment of LMA and inactivity was performed in adult male A_2AR KO mice, BG–A_2AR KO mice, and their respective WT littermates (n = 8, per genotype and drug dose). At 9:00 A.M., all animals received an intraperitoneal injection of either vehicle or caffeine at one of the following doses: 2, 10, or 30 mg/kg. Inactivity/activity was used to assess sleep and wakefulness based on a previous report (Pack et al., 2007) after caffeine treatment. LMA was recorded in standard polypropylene cages with seven infrared photocell beams (San Diego Instrument) in 50 s bins during the two experimental days. Inactivity and activity were defined based on LMA as follows: time spent in inactivity (no beam break/50 s) was used as an approximation of sleep; each period of activity was subdivided into high (at least two beam breaks/50 s, assessing ambulation) versus low (one beam break/50 s, assessing rest activity with fine movements) LMA. These three levels of activity were used to analyze the motor stimulant effects of caffeine as opposed to the arousal effects of caffeine, which were assessed using polygraphic recordings as described above.

Generation of AAV vectors.

For the generation of the AAV–shRNA–mCherry vector plasmids, a U6–shA_2AR cassette was amplified by PCR from the psiSTRIKE–hMGFP plasmid (Promega) containing the rat A_2A receptor shRNA (target sequence 1913) (Chen et al., 2004) or a control shRNA (shCTRL) sequence (GTCAGGCTATCGCGTATCG) and was inserted into the MluI site of the pAAV–hrGFP plasmid (Stratagene). Subsequently, the hrGFP gene was replaced by the gene encoding mCherry. For the generation of the AAV–Cre plasmid, the hrGFP in the pAAV–hrGFP plasmid was replaced by the Cre recombinase coding sequence derived by PCR from the pBS185 plasmid (Sauer and Henderson, 1990). The AAVs of serotype rh10 were generated by tripartite transfection (AAV–rep2/caprh10 expression plasmid, adenovirus helper plasmid, and AAV–vector plasmid) into HEK293A cells and purified by iodixanol density step-gradient centrifugation, as described previously (Zolotukhin et al., 1999). The virus distributed in the 40% density step was concentrated and dialyzed against PBS with a centrifugal concentrator (molecular weight cutoff, 100 kDa; Sartorius) and then titered by quantitative PCR.

Stereotaxic AAV injection and placement of EEG/EMG electrodes.

Surgeries for AAV injections were conducted under pentobarbital anesthesia (50 mg/kg, i.p.). Using aseptic techniques, 6-week-old rats were injected stereotaxically into the NAc and other BG nuclei with recombinant AAV–shA_2AR or AAV–shCTRL (250 nl/injection, 6 × 10¹² particles/ml) with a glass micropipette and an air pressure injector system (Chamberlin et al., 1998). Also, 8- to 10-week-old conditional A_2AR KO mice were injected with AAV–Cre or AAV–mCherry. Table 1 summarizes coordinates used for bilateral injections into selected BG nuclei of rats or conditional A_2AR KO mice, according to the atlases of Paxinos and Watson (2001, 2007). At 3 weeks after the AAV injection, rats underwent surgery for implantation of electrodes for EEG and EMG recordings as described previously (Matsumura et al., 1994), whereas EEG/EMG electrodes in the conditional KO mice were implanted as described above. Postoperatively, animals were housed individually for 8–10 d. The caffeine treatment was performed as described above; intraperitoneal injections were made at 9:00 A.M. (mice) or 10:00 A.M. (rats). In addition, at least 1 week after the caffeine treatment, each animal received an injection of vehicle or modafinil on a 2 d schedule as described above. Modafinil (Sigma-Aldrich) was dissolved in saline containing 10% DMSO and 2% (w/v) cremophor immediately before use and administered intraperitoneally at 9:00 A.M. on the experimental day at a dose of 45 mg/kg.

Table 1.

Coordinates for bilateral injections of AAV–Cre or AAV–mCherry in rats and mice according to Paxinos and Watson (2001, 2007)

Area	Coordinates (mm)
	Anterior to bregma	Lateral to midline	Below dural surface
Rat
OLT	1.8	1.5	8
	1.3	1.3	8
	0.8	1.3	8
NAc	1.8	1.5	7
	1.3	1.2	7
	0.8	1	6.8
CPu	1.8	2.0	4.5
	1.3	2.5	4
	0.8	2.5	4.5
GP	−0.4	2.2	7
	−0.9	2.8	6
Mouse
NAc	1	0.75	4.2
	1.4	0.8	3.9
SI/HDB	0	1.4	4.6

Each animal received two or three bilateral injections using different sets of coordinates. A_2AR-positive areas in the BG: CPu, NAc, OLT, and GP. Arousal-related areas in the basal forebrain: SI and HDB.

Immunohistochemistry.

After all of the above procedures, animals were deeply anesthetized with an overdose of chloral hydrate (500 mg/kg, i.p.) and perfused through the left ventricle of the heart with saline, followed by neutral buffered 10% Formalin. Brains were removed and placed in 10% sucrose in PBS overnight at 4°C to reduce freezing artifacts. The brains were then frozen on dry ice and sectioned at 30 μm (mice) or 40 μm (rats) on a freezing microtome. Immunohistochemistry was performed on free-floating sections as described previously (Estabrooke et al., 2001). In brief, sections were rinsed in PBS, incubated in 3% hydrogen peroxide in PBS for 30 min at room temperature, and then sequentially at room temperature in 3% normal donkey serum and 0.25% Triton X-100 in PBS (PBT) for 1 h and primary antibody diluted in PBT with 0.02% sodium azide overnight. Primary antibodies included rabbit anti-Cre (1:10,000; EMD Biosciences), rabbit anti-mCherry (1:10,000; Clontech), mouse anti-A_2AR (1:2000; Millipore), goat anti-A_2AR (1:1000; Santa Cruz Biotechnology), chicken anti-β-galactosidase (β-gal) (1:4000; Abcam), and mouse anti-neuronal-specific nuclear protein (NeuN) (1:2000; Millipore). After incubation with the primary antisera overnight, sections were rinsed and incubated for 2 h in biotinylated anti-rabbit, anti-goat, anti-chicken, or anti-mouse secondary antiserum (Jackson ImmunoResearch) at a dilution of 1:1000. Immunoreactions for D₂R with a rabbit anti-D₂R antibody (1:500; Millipore) were conducted over two nights at 4°C and one night at room temperature. All tissue sections were then treated with avidin–biotin complex (1:1000; Vectastain ABC Elite kit; Vector Laboratories) for 1 h, and immunoreactive cells were visualized by reaction with 3,3′-diaminobenzidine and 0.1% hydrogen peroxide. Tissue sections mounted on glass slides were scanned with Aperio ScanScope, and digital photomicrographs were analyzed with Aperio ImageScope software version 10. The region of A_2AR knockdown by shA_2AR was identified by the expression of mCherry and confirmed by the absence of A_2AR immunoreactivity, whereupon the area of loss of A_2ARs in the NAc was quantified on photomicrographs of sections containing the rostral, central, and caudal NAc. Digital photomicrographs were adjusted for optimal display for the output levels of the contained color values and then imported into NIH ImageJ 1.42 software for area measurements of mCherry expression in the NAc versus total nucleus extension. Double immunofluorescence staining for β-gal and choline acetyltransferase (ChAT) was also performed. Sections were incubated overnight at room temperature in a mixture of anti-β-gal and goat anti-ChAT (1:200; Millipore) primary antibodies in PBT with donkey normal serum. On the next day, sections were incubated for 2 h in a mixture of biotinylated anti-chicken and Alexa Fluor-594-conjugated anti-goat secondary antibodies (Invitrogen) at a dilution of 1:500. After several washes, sections were incubated for 1 h in Alexa Fluor-488-conjugated streptavidin (Invitrogen) at a dilution of 1:500. Fluorescence microscopy with tissue sections mounted on glass slides was performed with a Carl Zeiss laser scanning confocal microscope.

Statistical analysis.

The data were presented as the mean ± SEM. Statistical comparisons between two groups were performed by using the unpaired Student's t test. Comparisons among combined multiple parameters (genotype, experimental conditions, and more than two groups) were performed by one-way ANOVA, followed by Bonferroni's post hoc comparisons.

Results

Deletion of A_2ARs in the basal ganglia of mice abolishes the arousal effect of caffeine

We first examined sleep–wake profiles in our previously developed BG–A_2AR KO mice based on the Cre/loxP technology (Shen et al., 2008). The existence of A_2ARs in arousal-related cell groups surrounding the striatum, such as the nucleus of the horizontal limb of the diagonal band of Broca (HDB), the substantia inominata (SI), or the ventrolateral preoptic area (VLPO), remains elusive (Svenningsson et al., 1997; Rosin et al., 1998). However, to assess Cre-dependent knock-out of A_2ARs in these adjacent areas, we crossbred the Dlx5/6–Cre transgenic mouse, which was used to create the BG–A_2AR KO mouse, with a Rosa26 reporter line (Soriano, 1999) expressing β-gal only in the presence of Cre recombinase (Rosa26–Dlx5/6–Cre). The limit of this method may, however, hinge on the barely existing expression of A_2ARs in various regions of the brain. As shown in Figure 1A, the striatum, including the olfactory tubercle (OLT), caudate–putamen (CPu), and NAc, showed robust β-gal staining. The vast majority of neurons in the striatum are GABAergic medium-sized spiny output neurons, but double immunofluorescence staining for β-gal and ChAT (Fig. 1B) revealed that the A_2AR knock-out in the striatum occurred also in the cholinergic interneurons. Outside of the striatum, sparse β-gal expression was detected in the HDB, whereas β-gal staining was absent in the SI and the VLPO (Fig. 1C). In addition, only scattered cells with β-gal immunostaining were observed in the septum, cerebral cortex, thalamic nuclei, and hippocampus (Fig. 1A or data not shown). Brain sections from Cre-negative mice of the Rosa26–Dlx5/6–Cre line did not show any β-gal staining (data not shown).

Figure 1.

Arousal effect of caffeine was abolished in BG–A_2AR KO mice. A–C, Typical sections from the Rosa26–Dlx5/6–Cre reporter mouse were stained with mouse polyclonal antibodies against β-gal to visualize Cre-expressing neurons indirectly. A, Robust expression of β-gal is seen in the striatum of the reporter mouse. B, At a single-cell level, double immunofluorescence for β-gal (green) and ChAT (magenta) on an adjacent section to A shows that cholinergic interneurons in the striatum also express β-gal. The arrows in the top and bottom of B indicate neurons with dual immunolabeling for β-gal and ChAT. C, In the classical arousal/sleep-related cell groups of the basal forebrain and anterior hypothalamus, i.e., the nucleus of the HDB, SI, or VLPO, only moderate immunoreactivity for β-gal is detected in the HDB. β-gal immunolabeling is absent in neurons of the SI and VLPO. Scale bars: A, 500 μm; B, 20 μm; C, 250 μm. D, E, The BG–A_2AR KO mice and WT littermates were treated with vehicle or caffeine (2 or 30 mg/kg, i.p.). Time course (D) and total time (E) of wakefulness during the first 3 h after caffeine injection (arrows) were assessed with EEG/EMG recordings. Data are presented as the mean ± SEM (n = 4–5). *p < 0.05, **p < 0.01 compared with vehicle treatment within corresponding genotype. ^##p < 0.01 compared with corresponding WT littermates. ^††p < 0.01 compared between caffeine doses.

We then recorded EEG and EMG for 2 consecutive days in the BG–A_2AR KO mice and their wild-type littermates (Fig. 1D,E). On day 1, the mice were treated with vehicle (intraperitoneally) at 9:00 A.M. in the early phase of the light (inactive) period, and the recordings made on that day served as the baseline data. The animals were then treated with caffeine 24 h later (either 2 or 30 mg/kg, i.p.). The vigilance states were classified offline into three stages: waking, REM sleep, and NREM sleep. Caffeine dose-dependently increased wakefulness in control WT mice 2-fold and 3.2-fold after the 2 and 30 mg/kg doses, respectively (Fig. 1D,E). This arousal effect of caffeine was almost completely eliminated in the BG–A_2AR KO mice.

Deletion of A_2ARs in the basal ganglia of mice abolishes caffeine-induced locomotor activity

We measured activity in global A_2AR KO (Chen et al., 1999) and BG–A_2AR KO mice to determine whether caffeine induces in these mice the motor pattern that is typical for an animal during caffeine-induced wakefulness. We injected intraperitoneally male A_2AR KO and BG–A_2AR KO mice with vehicle (saline) or caffeine in a physiologically relevant range of 2–30 mg/kg at 9:00 A.M. and monitored the behavior of the mice in a field of infrared photocell beams to assess inactivity versus low and high levels of activity. Caffeine dose dependently decreased time spent in inactivity in the control WT mice but not in A_2AR KO and BG–A_2AR KO mice (Fig. 2A,B). Interestingly, although the time spent in low levels of locomotor activity (fine movement or quiet wakefulness) was not altered by caffeine in the A_2AR KO and BG–A_2AR KO mice (data not shown), the waking period characterized by high LMA (ambulation) was dose dependently increased by caffeine only in the control WT mice of both A_2AR KO genotypes (global vs BG) (Fig. 2C,D). Thus, caffeine acts at A_2ARs in the BG to promote wakefulness and associated locomotion.

Figure 2.

Locomotion in A_2AR KO and BG–A_2AR KO mice after caffeine treatment. Rest–activity assessment was performed based on the amount of LMA in 50 s bins for 3 h after caffeine (2, 10, or 30 mg/kg) treatment. Time spent in inactivity (A, B) or high LMA (ambulation; C, D) is presented as percentage of total time (3 h). Data are presented as mean ± SEM (n = 8, per genotype and drug dose rest–activity assessment). *p < 0.05, **p < 0.01 compared with vehicle treatment within corresponding genotypes. ^#p < 0.05, ^##p < 0.01 compared with corresponding WT littermates. ^††p < 0.01 compared between caffeine doses.

Selective deletion of A_2ARs in the NAc of mice eliminates the arousal effect of caffeine

The A_2AR agonist CGS21680 (2-p-(2-carboxyethyl)phenethylamino-5′-N-ethylcarboxamidoadenosine), which induces sleep, also produces c-Fos expression in the NAc (Scammell et al., 2001). We therefore tested whether the arousal effect of caffeine depended on A_2ARs in the NAc by using a mouse strain with a loxP-modified A_2AR gene conditionally deletable by Cre recombinase. An AAV vector that contained the gene for Cre recombinase under the control of the cytomegalovirus (CMV) promoter was stereotaxically injected bilaterally into the NAc of loxP-modified A_2AR mice to generate NAc–A_2AR KO. At 3 weeks after the injection of AAV–Cre, immunohistochemistry for A_2AR confirmed the loss of A_2ARs in the NAc (Fig. 3A, right photomicrograph), whereas A_2AR expression was unchanged in the control group of loxP-modified WT mice injected with the red fluorescent protein mCherry-expressing AAV (Fig. 3A, left photomicrograph). Microinjections of AAV vectors do not induce inflammation at the injection site, and tissue injury is minimal, as after the injection of saline (Lazarus et al., 2007). A_2ARs are known to be coexpressed with D₂Rs on neurons of the NAc (Svenningsson et al., 1997; Durieux et al., 2009), and the unaltered D₂R staining confirmed the integrity of the NAc of AAV–mCherry- and AAV–Cre-injected mice except for the absence of the A_2ARs (Fig. 3A,B).

Figure 3.

Arousal effect of caffeine was abolished in NAc–A_2AR KO mice. A, B, Typical sections of conditional A_2AR KO mice after injection with mCherry-expressing AAVs (WT; left photomicrograph) and AAV carrying Cre recombinase (NAc–A_2AR KO; right photomicrograph) were stained with a goat polyclonal antibody against A_2AR (Santa Cruz Biotechnology) to visualize the presence (black circle) or loss (red dashed circle) of A_2ARs in the NAc. Immunostaining with a rabbit polyclonal antibody against D₂R (Millipore) confirms the integrity of the NAc (black and blue circles) in the WT and NAc–A_2AR KO mice (B). C, D, The NAc–A_2AR KO and WT mice were treated with caffeine (15 mg/kg, i.p.) or modafinil (45 mg/kg, i.p.). The time course (C) and total time (D) of wakefulness for the first 3 h after caffeine treatment were assessed from EEG/EMG recordings, as was modafinil-induced wakefulness for 3 h (D, right). Arrows in C indicate the time of caffeine injection. E, F, Typical examples of EEG, EMG, LMA, and hypnograms after administration of caffeine (15 mg/kg, i.p., top panels), vehicle for caffeine administration (middle panels), or modafinil (45 mg/kg, i.p., bottom panels) in a WT (E) and NAc–A_2AR KO (F) mouse. Arrows in E and F indicate the time of injection. Data are presented as the mean ± SEM (n = 4–5). *p < 0.05, **p < 0.01 compared with vehicle treatment within corresponding AAV injection. ^##p < 0.01 compared with AAV treatment. Scale bars: A, B, 500 μm.

Next, we injected both mouse groups with caffeine (15 mg/kg, i.p.) and recorded their EEG and EMG (Fig. 3C,D). Typically, the effect of caffeine on wakefulness was strongly attenuated in the NAc–A_2AR KO mice generated by the AAV–Cre injection compared with the control group injected with AAV–mCherry (Fig. 3C). The time spent in waking in the control mice was increased twofold during a 3 h period after the 15 mg/kg dose of caffeine but was indistinguishable from the vehicle injection in mice with a deletion of A_2ARs in the NAc (Fig. 3D, left). In addition, modafinils (45 mg/kg, i.p.), a wakefulness-inducing compound that primarily requires D₂R (Qu et al., 2008), induced strong arousal in the NAc–A_2AR KO and control mice, causing almost complete insomnia during a 3 h period after injection (Fig. 3D, right). Typical examples of EEG, EMG, LMA, and hypnograms are shown in Figure 3, E and F, after the administration of caffeine (15 mg/kg, i.p.; top panels) or modafinil (45 mg/kg, i.p.; bottom panels) in a WT and NAc–A_2AR KO mouse. The vehicle control is only shown for the caffeine administration (middle panels), because the vehicle response in the modafinil and caffeine experiment was similar. Caffeine increased wakefulness in the WT mouse but not in the mouse with a deletion of A_2ARs in the NAc (Fig. 3F, top panel), whereas modafinil induced long-lasting suppression of sleep in both the NAc–A_2AR KO and control mice. Because the possibility of a knock-out of A_2ARs in the HDB of the BG–A_2AR KO mice cannot be entirely excluded (Fig. 1C), we also injected AAV–Cre bilaterally into the basal forebrain (BF) region, including the HDB and SI, of loxP-modified A_2AR mice and found that caffeine (15 mg/kg, i.p.) induced wakefulness (155 ± 7 min/3 h, n = 4) at similar levels as in WT mice (153 ± 3 min/3 h) (Fig. 1D). These experiments indicate that A_2ARs in the NAc were specifically required for the caffeine-induced arousal and that dopamine D₂R functions of these neurons were not affected by the deletion of the A_2ARs in the same neurons.

Site-specific knockdown of A_2ARs in the NAc of rats blocks caffeine-induced arousal

We next aimed to validate our findings by using a knockdown of A_2ARs and to define the extent to which the A_2AR-positive neurons in the core and shell regions of the NAc were required for the arousal effect of caffeine. Because rats are more suitable for anatomical work than mice, we used rats to dissect the contribution of A_2AR-positive neurons in the core and shell of the NAc, as well as in several other BG regions [the OLT, CPu, and globus pallidus (GP)] by stereotaxically injecting AAV vectors that contained short-hairpin interfering RNA specific for A_2AR (shA_2AR) and the reporter gene mCherry (Fig. 4A,B). The A_2AR shRNA sequence was derived from a previously validated small-interfering RNA target for A_2AR (Chen et al., 2004). At 4 weeks after the injection, A_2ARs were completely eliminated at the site of injection (Fig. 4D). A_2AR expression was not attenuated when injections were made with shCTRL with no homology to any known sequences in the rat genome (Fig. 4C). NeuN staining confirmed the absence of neuronal toxicity at the injection site in the NAc of both AAV–shCTRL- and shA_2AR-injected rats (Fig. 4E,F).

Figure 4.

Site-specific deletion of A_2ARs in rats by focal RNA interference. A, Generation of AAV vectors that contained shA_2AR and the red fluorescent protein mCherry as a reporter gene. B, AAV–shA_2AR vectors were stereotaxically injected into the A_2AR-positive core and shell of the NAc, as well as into the OLT, CPu, and GP. C, D, Typical sections from rats injected bilaterally with AAV carrying shCTRL or shA_2AR into the NAc were stained with mouse monoclonal antibody against A_2AR. Immunoreactivity for A_2AR is depleted selectively in the NAc of the AAV–shA_2AR-treated rats (D, dashed red circle), but it is unaffected in the AAV–shCTRL-treated rats (C, red circle). E, F, NeuN staining confirms the integrity of NAc neurons (green circles) in AAV–shCTRL-injected (E) and AAV-shA_2AR-injected (F) rats. Scale bars: C, E, 500 μm (also apply to D, F, respectively).

We then examined the effects of an intraperitoneal injection of caffeine (15 mg/kg) in rats that had been bilaterally injected with AAV–shCTRL or AAV–shA_2AR into the NAc (Fig. 5). EEG/EMG recordings made during a 3 h period after caffeine and vehicle injections were analyzed to obtain the total duration of wakefulness after each injection, and the difference in total amount of wakefulness after caffeine versus vehicle injection was defined as caffeine-induced wakefulness. We also measured the area of reporter expression as an indicator for the loss of A_2A receptors, using tissue sections of the rostral, central, and caudal NAc that were immunostained for the reporter. A Pearson's correlation of r = 0.8 (p < 0.01) between the reporter immunoreactivity and the caffeine-induced wakefulness (Fig. 5A) showed that the disruption of A_2ARs in the NAc was proportional to the loss of the arousal effect of caffeine. In contrast, the caffeine-induced arousal was not affected in rats when AAV–shA_2AR was bilaterally injected into other A_2AR-positive areas of the BG, including the CPu, OLT, and GP, or in AAV–shCTRL-treated or AAV-untreated rats (Fig. 5B).

Figure 5.

Arousal effect of caffeine was abolished in rats with site-specific deletion of A_2ARs in the shell of the NAc. A, The loss of the arousal response to caffeine (15 mg/kg, i.p.) in rats that had received AAV–shA_2AR injections (black circles) into the NAc correlated closely with the knockdown of A_2ARs in the NAc, whereas there was very little variation in caffeine-induced wakefulness in control animals injected with shCTRL (white triangles). B, Caffeine-induced wakefulness for 3 h in AAV-untreated rats and in rats that received injection of AAV–shCTRL into the NAc or injection of AAV–shA_2AR into the NAc shell and core, OLT, CPu, and GP. Caffeine was given intraperitoneally at 15 mg/kg. C, D, Typical sections from two rats injected bilaterally with AAV carrying shA_2AR into the shell or core of the NAc that were stained with mouse monoclonal antibody against A_2AR showed dominant depletion of A_2ARs either in the shell (D, right photomicrograph) or the core (C, left photomicrograph) of the NAc. Adjacent sections to C and D were stained with rabbit polyclonal antibody against mCherry (Clontech) to confirm the extent to which neural cells in the NAc were transfected with AAV–shA_2AR (C, D, right photomicrographs). The red circles in the photomicrographs in D outline the entire NAc, including both the core and shell region, whereas the green circles in the photomicrographs in C delineate the core of the NAc. The polysomnographic recordings in C and D show typical examples of EEG, EMG, LMA, and hypnograms after administration of caffeine at a dose of 15 mg/kg (top polysomnographic panel) or vehicle (bottom polysomnographic panel) in two rats with AAV–shA_2AR infections of the shell (D) or the core (C) of the NAc. Data are presented as mean ± SEM (n = 5–6 per AAV-treatment). **p < 0.01 compared with AAV-untreated or AAV–shCTRL-injected rats, assessed by one-way ANOVA. Scale bars: C, D, 300 μm.

All injections into the NAc were aimed at the border area between the core and shell of the NAc (Fig. 5D), but occasionally injections fell more laterally. In several such cases, only neurons of the NAc core but not those of the shell portion were infected with AAV–shA_2AR (Fig. 5C), and those rats showed a normal response to caffeine (Fig. 5B, green column). Figure 5, C and D, shows typical examples of EEG, EMG, LMA, and hypnograms after the administration of caffeine at a dose of 15 mg/kg (top polysomnographic panels) or vehicle (bottom polysomnographic panels) in two rats with AAV–shA_2AR infections of the shell (Fig. 5D) or the core (Fig. 5C) of the NAc. The A_2AR depletion in the NAc shell attenuated the effect of caffeine on wakefulness (Fig. 5D, top polysomnographic panel), whereas the rat with a loss of A_2ARs only in the core portion of the NAc showed a normal response to caffeine (Fig. 5C, top polysomnographic panel). These results indicate that A_2AR-positive neurons in the shell of the NAc were crucial for caffeine to induce wakefulness (Fig. 5B, red column).

Discussion

Our results, using a combination of different gene ablation strategies based on the Cre/loxP technology and RNA interference in two different species, clearly demonstrate that the expression of A_2ARs by neurons in the shell of the NAc is essential for the caffeine-induced arousal. For caffeine to be effective as an A_2AR antagonist, excitatory A_2ARs on NAc shell neurons must be tonically activated by endogenous adenosine. Such tonic activation likely occurs because A_2ARs are abundantly expressed in the NAc shell and, even under the most basal conditions, a finite level of adenosine is detected in the extracellular space (Svenningsson et al., 1999a,b). Thus, adenosine activates A_2ARs on medium spiny projection neurons in the NAc shell and contributes to restrain the arousal system. As a consequence, caffeine clearly overrides the “adenosine brake” and promotes wakefulness. Therefore, based on a similarity between mouse and man, the area of the human brain in which caffeine acts to counteract fatigue, the shell of the NAc, is just about the astonishingly small size of a pea.

The depletion of A_2ARs in the NAc shell diminished the caffeine-induced wakefulness but did not change the amount of wakefulness after the vehicle injection (Fig. 3C,D), indicating that the inhibition of A_2ARs in the NAc shell is crucial for caffeine-induced, but not the basal, wakefulness. Therefore, the adenosine A_2AR system in the NAc shell is considered to function as an accessory nucleus for regulation of the main sleep center in the VLPO, which is activated by prostaglandin D₂ (Scammell et al., 1998) and adenosine acting via A_2AR (Scammell et al., 2001). The neural network accounting for the arousal effect of caffeine on A_2AR-expressing neurons in the NAc shell is summarized in Figure 6, in which blockade of the massive GABAergic output of NAc shell neurons activates classical arousal centers, such as the lateral hypothalamus (LHA), the tuberomammillary hypothalamic nucleus (TMN), and the locus ceruleus (LC), via direct or indirect projections from the NAc shell. Those arousal centers are reciprocally regulated by the primary sleep-promoting neurons in the VLPO via GABAergic inhibitory projections (Saper et al., 2005, 2010).

Figure 6.

A proposed wake-regulatory role of the A_2AR-expressing neurons in the shell of the NAc that accounts for the wake-promoting effect of caffeine. Endogenous somnogens, such as adenosine and prostaglandin D₂ (PGD₂), promote sleep by activating sleep-promoting neurons of the VLPO (Saper et al., 2005, 2010; Urade, 2011), which, in a putative flip–flop arrangement, inhibit the arousal-promoting regions, including the LHA, TMN, and LC in the brainstem. Adenosine acting at A_2ARs on medium spiny neurons in the shell of the NAc is hypothesized to exert inhibitory effects on the arousal systems via indirect (GABAergic and glutamatergic) pathways. Caffeine blocks the A_2ARs in the NAc shell, thereby removing the restraint on the arousal systems to promote wakefulness.

The inability of caffeine (15 mg/kg) to induce any arousal effect in mice with A_2AR gene deletions in the NAc shell (Figs. 3, 5) indicates that the blockade of A_2ARs in the rest of the brain is not sufficient at this dose to promote arousal. However, this observation does not rule out the possibility that A_2ARs in other brain regions may also contribute to caffeine-induced wakefulness. Selective reinsertion of A_2ARs in the NAc shell of A_2AR KO mice would be necessary to show that A_2ARs in the NAc shell are sufficient to produce caffeine arousal. In fact, caffeine at 30 mg/kg decreased inactivity and increased LMA compared with the vehicle treatment for the BG–A_2AR KO mice but not for the global A_2AR KO mice (Fig. 2), suggesting that caffeine at high concentrations has psychomotor effects at A_2ARs outside the BG. This minor wake-promoting effect in BG–A_2AR KO mice with 30 mg/kg caffeine (Fig. 1D,E) may be attributable to the blockade of A_2ARs in the leptomeninges near the VLPO, because these receptors are responsible for activation of sleep-promoting neurons in the VLPO (Scammell et al., 2001).

Adenosine clearly acts as an endogenous somnogen that regulates the homeostatic sleep drive (Urade and Hayaishi, 2010). There are multiple pathways through which sleep and wakefulness can be regulated. Because A_2ARs on medium spiny neurons of the NAc are colocalized with D₂Rs, which are essential in the maintenance of wakefulness (Qu et al., 2008, 2010), caffeine acting on neurons in the shell of the NAc may modulate neural substrates through which dopamine produces arousal. Adenosine acting via A₁R has also been shown to induce sleep by inhibiting the cholinergic region of the BF (Basheer et al., 2004). For example, the unilateral infusion of the BF with an A₁R-selective antagonist increased waking and decreased sleep (Strecker et al., 2000). Single-unit recording of BF neurons in conjunction with in vivo microdialysis of an A₁R-selective agonist decreased, and an A₁R antagonist increased, the discharge activity of the neurons in the BF (Alam et al., 1999). In addition, the activation of A₁Rs expressed in the TMN inhibits the histaminergic system and promotes NREM sleep (Oishi et al., 2008). Although caffeine is an antagonist for both A₁R and A_2AR, it increased wakefulness in A₁R KO and WT mice but not in A_2AR KO mice (Huang et al., 2005), and, therefore, A₁Rs are clearly not required for the effect of caffeine on wakefulness.

Instead of acting at the classical sleep–wake-regulatory neurons, such as the cholinergic BF neurons and the sleep-promoting preoptic neurons, caffeine appears to induce arousal by activating, at least initially, many neuronal pathways that have traditionally been associated with locomotion and motivational behaviors. The NAc shell has long been thought to activate, mainly through indirect pathways via the ventral pallidum and substantia innominata, midbrain–pontine areas that are involved in exploratory locomotion (Mogenson et al., 1983; Groenewegen and Trimble, 2007). In addition, reciprocal connections between the NAc shell and the ventral tegmental area (Zahm and Heimer, 1993), a site of dopamine neurons involved in motivation, reward, and motor control, promote arousal driven by motivation (Sesack and Grace, 2010).

The NAc shell is also well positioned to recruit the cortex, in particular the medial prefrontal cortex (mPFC), into sleep-regulatory circuits. The mPFC is a key executive interface between cognition and emotion but is also uniquely sensitive to sleep and sleep need (Muzur et al., 2002) and might promote sleep (Koenigs et al., 2010). In addition to its modulatory re-entrant projections to the NAc shell, the mPFC could provide a top-down modulation through its direct descending projections to sleep–wake-regulatory systems in the hypothalamus (e.g., the TMN containing histamine and the LHA containing orexins) and the brainstem, including the LC containing noradrenaline (Hurley et al., 1991; Saper et al., 2005). Many of these cortical and subcortical areas also directly or indirectly receive NAc shell outputs (Zahm and Heimer, 1993; Yoshida et al., 2006; Sano and Yokoi, 2007; Sesack and Grace, 2010) and produce strong c-Fos expression, a marker for neuronal activation, in response to systemic caffeine (Deurveilher et al., 2006). The critical role of the NAc shell and its A_2ARs in caffeine-induced arousal suggests that this unique transition area between the striatum and the stress and anxiety systems within the extended amygdala may play a regulatory role for sleep mechanisms.

Different from amphetamine, A_2AR antagonists (including caffeine) enhance motor and arousal activities but have minimal addictive potential (Fredholm et al., 1999). This difference is probably attributable to the unique cellular localization of the A_2AR in the D₂R-bearing neurons of the indirect pathway but not in the D₁R-bearing striatonigral neurons, in which the psychostimulants amphetamine and cocaine predominantly act and which constitute the major therapeutic target site implicated in drug addiction and dependence. Caffeine may, however, influence the intake or actions of dependence-producing drugs, because the blockade of A_2AR can synergize with agents that activate D₁R pathways (Le Moine et al., 1997).