Adenosine A2A Receptors in Striatal Glutamatergic Terminals and GABAergic Neurons Oppositely Modulate Psychostimulant Action and DARPP-32 Phosphorylation

Adenosine A2A receptors (A2AR) are located postsynaptically in striatopallidal GABAergic neurons, antagonizing dopamine D2 receptor functions, and are also located presynaptically at corticostriatal terminals, facilitating glutamate release. To address the hypothesis that these two A2AR populations differently control the action of psychostimulants, we characterized A2AR modulation of cocaine-induced effects at the level of DARPP-32 phosphorylation at Thr-34 and Thr-75, c-Fos expression, and psychomotor activity using two lines of cell-type selective A2AR knockout (KO) mice with selective A2AR deletion in GABAergic neurons (striatum-A2AR-KO mice), or with A2AR deletion in both striatal GABAergic neurons and projecting cortical glutamatergic neurons (forebrain-A2AR-KO mice). We demonstrated that striatum-A2AR KO mice lacked A2ARs exclusively in striatal GABAergic terminals whereas forebrain-A2AR KO mice lacked A2ARs in both striatal GABAergic and glutamatergic terminals leading to a blunted A2AR-mediated facilitation of synaptosomal glutamate release. The inactivation of A2ARs in GABAergic neurons reduced striatal DARPP-32 phosphorylation at Thr-34 and increased its phosphorylation at Thr-75. Conversely, the additional deletion of corticostriatal glutamatergic A2ARs produced opposite effects on DARPP-32 phosphorylation at Thr-34 and Thr-75. This distinct modulation of DARPP-32 phosphorylation was associated with opposite responses to cocaine-induced striatal c-Fos expression and psychomotor activity in striatum-A2AR KO (enhanced) and forebrain-A2AR KO mice (reduced). Thus, A2ARs in glutamatergic corticostriatal terminals and in GABAergic striatal neurons modulate the action of psychostimulants and DARPP-32 phosphorylation in opposite ways. We conclude that A2ARs in glutamatergic terminals prominently control the action of psychostimulants and define a novel mechanism by which A2ARs fine-tune striatal activity by integrating GABAergic, dopaminergic and glutamatergic signaling.


Introduction
Striatal circuits, composed mainly of GABAergic medium spiny neurons (MSN), are the principal entry point of the basal ganglia and the primary site for processing of motor, motivational and cognitive behaviors [1]. MSN are driven by cortico-thalamic excitatory glutamatergic projections and modulated by nigral dopaminergic inputs. MSN project either directly (striatonigral MSN) or indirectly (striatopallidal MSN) to output nuclei [2]. Adenosine A 2A receptors (A 2A R) are highly expressed in striatopallidal MSN where they antagonize dopamine D 2 receptor (D 2 R) function [3]. In addition, A 2A R are also located in striatal glutamatergic terminals [4] where they are involved in the modulation of glutamate release and corticostriatal synaptic transmission [5,6,7,8]. Notably, blockade of A 2A R in extra-striatal forebrain neurons attenuates behavioral responses to psychostimulants such as cocaine [9], amphetamine [10,11] or L-DOPA [12]. This led us to propose that presynaptic A 2A R in corticostriatal glutamatergic terminals exert their excitatory effects by facilitating glutamate release to counteract the inhibitory effect of postsynaptic A 2A R in GABAergic MNS [3,9]. This working model places A 2A R in a unique position, integrating GABAergic, glutamatergic and dopaminergic neurotransmission to fine-tune striatal activity.
Dopamine-and cAMP-regulated phosphoprotein (DARPP-32) is a key signaling molecule coordinating MSN responsiveness, where its activity is regulated by its phosphorylation status on different residues, namely Thr-34 and Thr-75 [13]. The phosphorylation of striatal DARPP-32 at Thr-34 and Thr-75 is under tight control of dopamine, adenosine and glutamatergic signalling [13]. DARPP-32 phosphorylation at Thr-34 is controlled by the G s /G i -cAMP-PKA signalling pathway via D 1 receptors (D 1 R) in the direct pathway and A 2A R/D 2 R activation in the indirect pathway. DARPP-32 phosphorylation at Thr-75 in MSN is competitively inhibited by and inversely correlated with the activation of cAMP signalling and is additionally modulated by glutamate signalling via cdk5 kinase [14]. Studies with global [13] or striatal pathway-selective genetic deletion of DARPP-32 [15,16,17] confirmed that DARPP-32 activation in the direct and indirect pathways oppositely determines motor responses to psychoactive drugs. Specifically, the selective deletion of DARPP32 in the indirect pathway enhances psychomotor activity while the selective deletion of DARPP-32 in the direct pathway attenuates the psychomotor effect [15,16]. Thus, Thr-34 and Thr-75 phosphorylation of DARPP-32 integrates the glutamatergic drive with dopaminergic extrinsic modulation as well as with intrinsic striatal modulation such as through adenosine [13]. We therefore hypothesize that A 2A R in GABAergic and glutamatergic neurons modulates the action of psychostimulants through a putative opposite control of striatal DARPP-32 phosphorylation.
To test this hypothesis, we developed and characterized two cell type-selective A 2A R knockout (KO) lines with selective deletion of A 2A R either in inhibitory GABAergic striatopallidal neurons (striatum-A 2A R KO, st-A 2A R KO) or in excitatory glutamatergic cortical neurons in addition to GABAergic MSN (forebrain-A 2A R KO, fb-A 2A R KO). Their use allowed us to demonstrate that A 2A Rs in GABAergic MSN and in corticostriatal glutamatergic terminals control the action of psychostimulants in opposite manners at the levels of (i) DARPP-32 phosphorylation; (ii) cocaine-induced c-Fos expression; and (iii) cocaine-induced psychomotor activity. This suggests that A 2A R control the action of psychostimulants through the regulation of DARPP-32 phosphorylation (at Thr-34 and Thr-75) in striatopallidal neurons. Furthermore, these results define a novel function of A 2A R in glutamatergic terminals and GABAergic striatopallidal neurons to fine-tune striatal neuronal activity and the action of psychostimulants through the integration of GABAergic, glutamatergic and dopaminergic signaling pathways.  (Figure 2A and 2B). In contrast, CGS21680-mediated facilitation of 3 H-glutamate release was completely abolished in striatal synaptosomes from fb-A 2A R KO mice (n = 4, p.0.05 compared to 0%) ( Figure 2C), but was unaffected in synaptosomes from st-A 2A R KO mice (n = 4, p,0. 05 compared to 0%) ( Figure 2D). These findings support the selective preservation of presynaptic A 2A R function in glutamatergic terminals in st-A 2A R KO but not fb-A 2A R KO mice.   [18,19]. Also in agreement with previous studies [18,20], acute treatment with cocaine (25 mg/kg, i.p.) produced a marked increase of DARPP-32 phosphorylation at Thr-34 and a concomitant reduction of DARPP-32 phosphorylation at Thr-75 in WT mice (st-WT and fb-WT, Figure 3). As predicted from a direct, postsynaptic facilitatory effect of A 2A R in GABAergic neurons, cocaine-induced DARPP-32 phosphorylation at Thr-34 was significantly attenuated in st-A 2A R KO mice compared to their WT littermates (n = 6 per group, p,0.05 comparing cocaine with saline treatment) ( Figure 3D). In contrast, the acute treatment with cocaine markedly increased DARPP-32 phosphorylation at Thr-34 in fb-A 2A R KO mice compared to WT littermates (n = 4, p,0.05 comparing cocaine with saline treatment) ( Figure 3A and 3B), consistent with a reduced glutamate release and dis-inhibition of glutamate suppression of DARPP-32 phosphorylation at Thr-34 in fb-A 2A R KO mice [14]. Additional fluorescence immunohistochemistry using brain sections showed that DARPP32 phosphorylation at Thr-75 was markedly reduced 45 minutes after cocaine treatment in fb-WT and fb-A 2A R KO mice (data not shown), a finding consistent with Western blot analysis. These findings demonstrated that, following cocaine treatment, presynaptic A 2A R in glutamatergic terminals exert an opposite and predominant effect over postsynaptic A 2A R in GABAergic neurons on striatal DARPP-32 phosphorylation at Thr-34 and Thr-75.    Figure 5D); this finding is consistent with our results using two color, sequential immunohistochemistry of c-Fos and dynorphin ( Figure 4) and also agrees with previous reports that cocaine induces c-Fos expression predominantly in the D 1 Rcontaining striatonigral neurons (e.g. [17]). In fb-WT animals, we also observed a cocaine-induced c-Fos expression in the D 2 Rcontaining indirect pathway, likely attributed to a postsynaptic (striatopallidal) A 2A R effect since cocaine-induced c-Fos expression was reduced in fb-A 2A R KO mice ( Figure 5C). Thus fb-A 2A R KO mice displayed a reduced cocaine-induced c-Fos expression in the direct pathway as well as the indirect pathway, although the majority of cocaine-induced modifications of c-Fos expression in fb-A 2A R KO mice were attributed to the direct pathway (see Figure 5). This finding suggests that the elimination of presynaptic glutamatergic A 2A R mainly affects the direct pathway to control psychomotor activity and c-Fos expression. This indicates that forebrain A 2A R exert their control of cocaine action predominantly through the regulation of glutamate release, which challenges previous views attributing those actions to the control of the responsiveness of striatal GABAergic neurons. The most intriguing aspect of A 2A R function in glutamatergic terminals is their ability to over-ride the effect of A 2A R in striatopallidal neurons, which have a nearly 20-fold higher A 2A R density [3]. This preferential engagement of A 2A R in glutamatergic terminals is heralded by the observations that psychostimulants [23,24,25,26] as well as NMDA receptor activation [27,28] can enhance the local striatal extracellular levels of adenosine, preferentially near glutamatergic but not GABAergic terminals [25]. Thus, the pattern of generation of adenosine by psychostim-  [19,29]. Furthermore, the increase of DARPP-32 phosphorylation at Thr-34 in fb-A 2A R KO mice is best explained by the selective changes of DARPP-32 in the indirect pathway since the attenuation of cocaine-psychomotor activity is strongly correlated with enhanced DARPP-32 in the indirect pathway (not the direct pathway) as clearly demonstrated by the elegant work using cell-type specific DARPP-32 KO [15,16]. Our findings are also in line with the concept that the striatopallidal pathway exerts a general inhibitory effect on behavior such as instrumental learning [30], psychostimulant activity [9], and aversive behavior [31], as revealed by selective destruction of the indirect pathway using targeted toxin expression [32] and by optogenetic silencing [33,34].

Selective preservation of
However, the A 2A R control of c-Fos expression in the striatum seems to result mainly from the c-Fos response in the direct pathway since we now demonstrated that cocaine-induced c-Fos expression was detected mainly in dynorphin-positive neurons. This effect could either result from recurrent collateral connections between striatopallidal and striatonigral MSN [35] or from an enhanced D 2 R-mediated release of endocannabinoids, which would decrease glutamate release from corticostriatal terminals projecting to both the indirect as well as the direct pathway [36]. This also explains the ability of A 2A R to control D 1 R-mediated responses such as rotational behavior [37,38], c-Fos expression in striatopallidal neurons [39] and DARPP-32 phosphorylation [29,40]. In addition, the c-Fos expression may also be a secondary functional consequence of the enhanced psychomotor activity by selective deletion  (Figure 1) and the consequent abolishment of A 2A R-facilitated glutamate release from striatal nerve terminals (Figure 2), the different regulation of DARPP-32 phosphorylation by A 2A R in fb-A 2A R KO mice likely results either from the impact of presynaptic A 2A R on glutamate release alone or from the combined effect of presynaptic A 2A R and postsynaptic A 2A R actions, an issue that will require the use of selective deletions of A 2A R in presynaptic glutamatergic corticostriatal terminals to be resolved. In fact, we are concluding that the differences between the phenotypes of fb-A 2A R KO and st-A 2A R KO mice are mostly due to the effects of presynaptic A 2A R in glutamatergic corticostriatal terminals since the most evident differentiating factor in fb-A 2A R KO mice is the deletion of presynaptic A 2A R and the abolishment of A 2A R-mediated facilitation of glutamate release. Since increased DARPP-32 phosphorylation at Thr-34 in the direct pathway is expected to produce enhanced cocaine psychomotor activity [15,17], the increased DARPP-32 phosphorylation at Thr-34, together with the attenuation of cocaine-induced psychomotor activity in fb-A 2A R KO mice strongly suggests that glutamate release by A 2A R in corticostriatal terminals preferentially affects DARPP-32 phosphorylation in the indirect pathway. Conversely, fb-A 2A R KO mice display an altered c-Fos expression in the direct and indirect pathways with the direct pathway being prominent one. Overall, the molecular and behavioral responses found in fb-A 2A R KO mice suggest a selective modification of DARPP-32 phosphorylation in the indirect pathway and a prominent modification of cocaine-induced c-Fos expression in the direct pathway in tight correlation with cocaine-induced psychomotor activity. This is in line with the findings from cell-type specific deletion of DARPP-32, which showed that cocaine-induced psychomotor activity was attenuated by selective inactivation of DARPP-32 in the direct pathway [15]. While these results suggest that A 2A R activity in glutamatergic terminals and GABAergic neurons may influence the action of psychostimulants by controlling DARPP-32 phosphorylation selectively in the indirect pathway, with the c-Fos response being secondary to the psychomotor effect, additional experiments are clearly warranted to clarify the cellular substrate linking the presynaptic A 2A R control of glutamate release and its impact on psychomotor activity.

Neurobiological and therapeutic implications
Based on the opposite phenotypes of cocaine-induced molecular and behavioral changes in st-A 2A R KO and fb-A 2A R KO mice, and their association with glutamatergic, GABAergic and dopaminergic systems at presynaptic and postsynaptic sites, we propose a new model for A 2A R function in the control of striatal circuits: A 2A R in glutamatergic terminals and GABAergic neurons provide a ''fine-tuning'' mechanism, whereby they integrate and regulate dopaminergic and glutamatergic signaling in the striatum. The integrated function of A 2A R is accomplished through the opposing actions of A 2A R in GABAergic striatal neurons (through A 2A R-D 2 R antagonistic interactions) and in glutamatergic corticostriatal terminals (by modulating glutamate release). The novelty of this model is that the ''fine-tuning'' provided by A 2A R may serve to prevent over-or under-stimulation of striatal neurons, and illustrates an essential aspect of the integrated function of the adenosine neuromodulation system [41]. Since decreased glutamatergic neurotransmission and increased dopaminergic activity contribute to the pathophysiology of schizophrenia and related psychiatric disorders, the ability of A 2A R to integrate dopaminergic and glutamatergic systems indicates that adenosine acting at A 2A R may modulate both positive (by preventing hyper-dopaminergic activity) and negative (by preventing hypo-glutamatergic activity) symptoms of schizophrenia [42]. Thus, the selective manipulation of presynaptic A 2A R in glutamatergic terminals [43] may have a therapeutic value to manage a variety of neuropsychiatric behaviors such as anxiety, depression, psychosis and schizophrenia [44].  [45,46,47], to generate st-A 2A R KO mice [Dlx5/6-Cre(+)A 2A R flox+/+ ] mice [9]. Genotyping was conducted by 3 primer PCR analysis of tail DNA [10]. Fb-A 2A R KO and st-A 2A R KO mice were characterized for their selective Adora2a deletion in the forebrain (i.e., cortex, hippocampus, and striatum) [10,48] or exclusively in striatal [9] neurons, as shown in our previous studies. The selectivity in these two lines was further validated by Creexpression by X-gal staining of LacZ in a Rosa26 reporter transgenic line, PCR analysis of Cre-mediated Adora2a deletion, A 2A R immunohistochemistry and 3 H-ZM241385 radioligand binding of A 2A R density [9,10,48,49]. Our early studies showed that the behaviors of two WT genotypes [Cre(-)A 2A R flox+/+ or Cre(+)A 2A R flox2/2 ] were not distinguishable (data not shown) and so we used either WT type or in some cases two WT types were pooled in to one group referred to as simply st-WT or fb-WT, accordingly.

Drug treatments and psychomotor activity assessments
Before drug treatment, all mice were habituated in the testing environment and mice were injected with a single dose of cocaine (25 mg/kg, i.p.; Sigma, St. Louis, MO, USA). Horizontal locomotor activity was monitored for 180 min after drug administration and analyzed as described previously [9].
3. Glutamate release from striatal synaptosomes 3 H-glutamate release experiments were performed as previously described after purification of striatal nerve terminals using a sucrose/Percoll fractionation method [22]. Briefly, nerve terminals were equilibrated at 37uC for 10 min, loaded with 3 H-glutamate (0.2 mM, specific activity of 45 Ci/mmol, Amersham, Piscataway, NJ, USA) for 5 min at 37uC, washed, layered over Whatman GF/ C filters and superfused with oxygenated Krebs solution for 20 min before starting collection of the superfusate. Synaptosomes were stimulated with 20 mM K + at 3 min (S 1 ) and 9 min (S 2 ) after starting sample collection, triggering a release of tritium that was mostly 3 H-glutamate, released in a Ca 2+ -dependent manner [22]. The A 2A R agonist CGS21680 (Tocris, Bristol, UK), tested at a concentration that is supra-maximal but selective to activate A 2A R [22], was added 2 min before S 2 onwards and its effect was quantified by modification of the S 2 /S 1 ratio compared to control chambers. Normalized facilitation by CGS21680 of the K + -evoked 3 H-glutamate release was tested by the one-sample t-test against the hypothetical value of 0% compared to paired control experiments carried out in the same batch of nerve terminals in the absence of added drugs. P # 0.05 was considered to represent a significant difference.

Immunocytochemical detection of A 2A R in glutamatergic and GABAergic nerve terminals
Striatal nerve terminals were purified through a discontinuous Percoll gradient and platted over poly-L-lysine-coated cover-slips for immunocytochemical analysis, using antibodies that were previously validated [22,50]. Permeabilized nerve terminals were incubated for 1 hour with rabbit anti-A 2A R (1:500, Upstate Biotechnology, Lake Placid, NY, USA), and guinea pig antivesicular GABA transporters (vGAT, 1:1,000, Calbiochem, San Diego, CA, USA) or guinea pig anti-vesicular glutamate type 1 transporters (vGluT1, 1:1000, Chemicon, Temecula, CA, USA) antibodies followed by a 1 hour incubation with different AlexaFluor-labeled secondary antibodies (1:2,000, Molecular Probes, Leiden, The Netherlands), which did not yield any signal in the absence of the corresponding primary antibodies. After washing and mounting onto slides with Prolong Gold Antifading (Invitrogen, Eugene, OR, USA), preparations were visualized in a Zeiss fluorescence microscope and analyzed with MetaFluor 5.0. Each coverslip was analyzed by counting three different fields and in each field a total amount of 150 individualized elements excluding elements based on their insufficient or excessive pixel intensity and excessive size, as previously described [22,50]. Note that this approach can only globally distinguish glutamatergic from GABAergic terminals, but the anti-vGluT1 and anti-vGAT antibodies used cannot distinguish between the different types of glutamatergic terminals (projecting to the direct or indirect pathways) or GABAergic terminals (direct projections or collaterals).
6. Immunohistochemistry of c-Fos expression and double labeling of c-Fos with dynorphin or enkephalin Sequential antibody detection of c-Fos and dynorphin. Free-floating brain coronal sections (30 mm) were double stained immunohistochemically with anti-c-Fos and anti-dynorphin polyclonal antibodies using standard avidin-biotin procedures follow-ing a sequential antibody detection protocol as described previously [52,53]. For this procedure, the first antibody, i.e., a goat anti-dynorphin polyclonal antibody (1:200, sc-46313, Santa Cruz, CA, USA) was detected first, using immunoperoxidase staining enhanced with 0.08% nickel ammonium sulfate, which yields a dark grayish color. After completion of the first staining, the same sections were incubated with an avidin/biotin blocking solution in order to block free avidin/biotine sites from the first biotinylated goat anti-rabbit IgG antibody. Then, sections were processed for immunolabeling with the second primary antibody, i.e., a rabbit anti-c-Fos polyclonal antibody (1:5,000, PC-38, Calbiochem) following standard protocols using DAB, yielding a bright brown color. This method has been repeatedly shown to lack cross-labeling [52,53]. Moreover, the nuclear localization of c-Fos staining, as opposed to the cytoplasm/neuropil staining of dynorphin, makes it easy to differentiate the two types of staining.

Statistical analysis
Statistical comparisons between st-A 2A R KO vs st-WT or fb-A 2A R KO vs fb-WT were analyzed (independently for their different genetic backgrounds) using a paired or unpaired Student's t test, according to the experimental design. To determine the effect of genotype, drug treatment and their interaction, we applied a two-way ANOVA for repeated measurements followed by Bonferroni post hoc comparison.