GPR88 in A2AR Neurons Enhances Anxiety-Like Behaviors

Abstract GPR88 is an orphan G-protein-coupled receptor highly expressed in striatal dopamine D1 (receptor) R- and D2R-expressing medium spiny neurons. This receptor is involved in activity and motor responses, and we previously showed that this receptor also regulates anxiety-like behaviors. To determine whether GPR88 in D2R-expressing neurons contributes to this emotional phenotype, we generated conditional Gpr88 knock-out mice using adenosine A2AR (A2AR)-Cre-driven recombination, and compared anxiety-related responses in both total and A2AR-Gpr88 KO mice. A2AR-Gpr88 KO mice showed a selective reduction of Gpr88 mRNA in D2R-expressing, but not D1R-expressing, neurons. These mutant mice showed increased locomotor activity and decreased anxiety-like behaviors in light/dark and elevated plus maze tests. These phenotypes were superimposable on those observed in total Gpr88 KO mice, demonstrating that the previously reported anxiogenic activity of GPR88 operates at the level of A2AR-expressing neurons. Further, A2AR-Gpr88 KO mice showed no change in novelty preference and novelty-suppressed feeding, while these responses were increased and decreased, respectively, in the total Gpr88 KO mice. Also, A2AR-Gpr88 KO mice showed intact fear conditioning, while the fear responses were decreased in total Gpr88 KO. We therefore also show for the first time that GPR88 activity regulates approach behaviors and conditional fear; however, these behaviors do not seem mediated by receptors in A2AR neurons. We conclude that Gpr88 expressed in A2AR neurons enhances ethological anxiety-like behaviors without affecting conflict anxiety and fear responses.


Introduction
G-protein-coupled receptors (GPCRs) are the target for ϳ40% of marketed drugs, and are major players in biomedicine (Rask-Andersen et al., 2011). Orphan GPCRs, whose ligands remain unknown and functions have been little studied, offer great promise (Levoye et al., 2006;Rask-Andersen et al., 2011;Ghanemi, 2015). The orphan GPCR GPR88 has been implicated in a number of behaviors related to psychiatric disorders. Mice lacking Gpr88 present a complex behavioral phenotype that includes motor coordination deficits, reduced prepulse inhibition, stereotypies, and altered cue-based learning (Logue et al., 2009;Massart et al., 2009;Quintana et al., 2012;Ingallinesi et al., 2015). These behaviors can all be related to the strong enrichment of GPR88 in the striatum (Swerdlow et al., 2001;Lewis and Kim, 2009;Liljeholm and O'Doherty, 2012). In humans, the Gpr88 gene was associated with bipolar disorders and schizophrenia (Del Zompo et al., 2014). Recently, we found that Gpr88 deletion in mice also decreases anxiety-like behavior (Meirsman et al., 2016), implicating this receptor in emotional processing and in the evaluation of environmental stimuli value. Concordant with this finding, Gpr88 expression was shown regulated by antidepressant and moodstabilizer treatments in both rodent models and humans (Ogden et al., 2004;Brandish et al., 2005;Böhm et al., 2006;Conti et al., 2007).
Several lines of evidence suggest that GPR88 alters behavior by modulating striatal transmission. In the striatum (dorsal and ventral), GPR88 is most abundant in medium spiny neurons (MSNs) expressing dopamine D 1 receptors (Rs; D 1 R-MSNs coexpressing substance P) and dopamine D 2 R MSNs coexpressing adenosine A 2A R (A 2A R) and regulates the excitability of these neurons, possibly by acting on glutamatergic, GABAergic, and dopaminergic receptors activity (Logue et al., 2009;Quintana et al., 2012). Conversely, glutamatergic and dopaminergic depletion differentially alters Gpr88 expression in these distinct MSN subpopulations (Logue et al., 2009). However, the precise mechanism by which GPR88 regulates the transmission of MSNs to alter behavior remains unknown. In recent years, research on MSNs subtype function has revealed that these two neuronal populations differentially regulate not only motor behaviors but also responses to rewarding and aversive stimuli (Durieux et al., 2009;Lobo et al., 2010;Kravitz et al., 2012). For instance, it has been suggested that altered D 2 R MSN transmission may disrupt inhibitory controls and avoidance in a decision conflict task (Hikida et al., 2010(Hikida et al., , 2013. Moreover, studies in humans and rodents suggest that the D 2 R modulates reward and emotional processing (Hranilovic et al., 2008;Peciña et al., 2013;Brandão et al., 2015), while the activation of D 2 R neurons in mice induced depressive-like behavior (Francis et al., 2015).
To gain a better understanding of how GPR88 in D 2 R MSNs regulates emotional processing, we generated a conditional knockout (KO) of Gpr88 in neurons expressing A 2A R (A 2A R-Gpr88 KO) known to be selectively expressed in D 2 R MSNs (Schiffmann et al., 1991;Schiffmann and Vanderhaeghen, 1993). To evaluate the contribution of GPR88 in D 2 R MSNs, we compared behavioral responses of A 2A R-Gpr88 KO mice with those of total Gpr88 KO animals using behavioral tests measuring anxiety-like behaviors and fear responses. We show that Gpr88 expressed in these neurons is responsible for ethological anxiety-like behaviors, but does not regulate conflict anxiety and fear responses.

Subjects
Mice (male and female) 9 -15 weeks of age were bred in-house in a group house of three to five animals per cage. Animals where maintained on a 12 h light/dark cycle at a controlled temperature (22 Ϯ 1°C). Food and water were available ad libitum throughout all experiments except for the novelty suppressed feeding test. All experiments where approved by the local ethics committee (COMETH 2014-029). For all experiments, Gpr88 A2A-Cre mice were compared with their littermates (Gpr88 flx/flx ), and Gpr88 Ϫ/Ϫ mice were compared with Gpr88 ϩ/ϩ mice. An independent cohort of naive animals was used for each behavioral paradigm, except for the fear conditioning that was performed in the same cohort as the light/ dark test 48 h after the latter. All behavioral testing was performed and analyzed blind to genotypes.

Tissue preparation and fluorescent in situ hybridization
Mice (n ϭ 4 Gpr88 flx/flx ; 4 Gpr88 A2A-Cre ) were killed by cervical dislocation, and fresh brains were extracted and embedded in OCT (Optimal Cutting Temperature Medium, Thermo Scientific), frozen, and kept at Ϫ80°C. Frozen brains were coronally sliced into 20 m serial sections by using a cryostat (model CM3050, Leica) and placed on SuperFrost slides (Thermo Scientific). In situ hybridizations were performed using the RNAscope Multiplex Fluorescent Assay. GPR88 probes were coupled to FITC, while D 1 R and D 2 R probes were coupled with Tritc and Cy5, respectively.

Drd2-positive cells
Image acquisition was performed with the slide scanner NanoZoomer 2 HT and fluorescence module L11600-21 (Hamamatsu Photonics). To verify the specific excision of Gpr88 in D 2 R MSNs neurons, in situ hybridization images were analyzed using NDP.viewer software. For each brain, four slices were selected, as follows: two slices for the caudate-putamen (CPu; rostral, ϩ0.98 mm from bregma; caudal, Ϫ0.58 mm from bregma); one slice for the nucleus accumbens (Nac; ϩ1.34 mm relative to bregma); and one slice for the central nucleus of the amygdala (CeA; Ϫ1.22 from bregma). For each structure, regions of interest (ROIs) were determined by drawing twodimensional boxes with defined surfaces. Counting was performed on one ROI with a surface of 1 mm 2 for the Nac, 0.250 mm 2 for the CeA, and two ROIs of 0.5 mm 2 for the each CPu slice (to include both dorsomedian and dorsolateral striatum; Fig. 1). Counting was balanced between right and left hemispheres. To evaluate the expression of Gpr88 mRNA in D 1 R-and D 2 R-expressing cells, counting was performed manually using the NDP.view software counting add-up. First, cells expressing Drd1 mRNA, but not Drd2 mRNA, were marked and counted. For each Drd1-positive cell, the coexpression of Gpr88 was verified and counted separately. This process was repeated for Drd2 mRNA-positive cells. Relative Gpr88 expression is represented as a percentage of the total Drd1-or Drd2-positive cells counted [(number Drd1 or Drd2 expressing cells coexpressing Gpr88 ϫ 100)/total number of Drd1-or Drd2-expressing cells]. Given the lack of difference in Gpr88 expression between lateral and medial CPu, the relative percentage of each was pooled for graphical representation and statistical analysis.
To evaluate the activation of GPR88 in PFC, CPu, Nac, CeA, and hippocampus, structures were punched in six animals of each genotype (three males, three females), as previously described (Meirsman et al., 2016), and were pooled for membrane preparation. To perform [ 35 S]-GTP␥S assays on whole-striatum mice were killed by cervical dislocation, and both striata were rapidly manually removed, frozen in dry ice, and stored at Ϫ80°C until use. Three membrane preparations were used for each genotype, gathering tissue from three animals each (males and females). Results are expressed as mean measures from the three membrane preparations. All assays were performed on membrane preparations. Membranes were prepared by homogenizing the tissue in an ice-cold 0.25 M sucrose solution (10 vol/ml/g wet weight of tissue). Samples were then centrifuged at 2500 ϫ g for 10 min. Supernatants were collected and diluted 10 times in buffer containing 50 mM Tris-HCl, pH 7.4, 3 mM MgCl 2 , 100 mM NaCl, and 0.2 mM EGTA, following which they were centrifuged at 23,000 ϫ g for 30 min. The pellets were homogenized in 800 l of ice-cold sucrose solution (0.32 M) and kept at Ϫ80°C. For each [3 5S ]GTP␥S binding assay, 2 g of protein/well was used. Samples were incubated with and without ligands, for 1 h at 25°C in assay buffer containing 30 mM GDP and 0.1 nM [ 35 S]GTP␥S. Bound radioactivity was quantified using a liquid scintillation counter. B max and K d values were calculated. Nonspecific binding was defined as binding in the presence of 10 M GTP␥S, and binding in the absence of agonist was defined as the basal binding.

Drugs
The GPR88 agonist compound 19 (Meirsman et al., 2016) was synthesized by Prestwick Chemicals and dissolved in water.

Behavioral analysis Open field locomotion
To assess basal locomotor activity in a novel environment, mice were placed in a dimly lit (15 lux) open field arena placed over a white Plexiglas infrared-lit platform. Locomotor activity was recorded during 30 min via an automated tracking system (Videotrack, ViewPoint). Only movements with speed exceeding 6 cm/s were taken into account for this measure.

Elevated plus maze
Anxiety-like behavior was first evaluated using the ethological (also known as unconditioned) anxiety test elevated plus maze (EPM). The EPM was a plus-shaped maze elevated 52 cm from the base, with a black Plexiglas floor consisting of two open and two closed arms (37 ϫ 6 cm each) connected by a central platform (6 ϫ 6 cm). The experiments were conducted under low-intensity light (15 lux). The movement and location of the mice were analyzed by an automated tracking system (Videotrack; View Point, Lyon, France). Each mouse was placed on the central platform facing a closed arm and were observed for 5 min. Anxiety-like behavior was assessed by mea- Light/dark test Anxiety-like behavior was next evaluated using the light/ dark apparatus, which was composed of two rectangular compartments (20 ϫ 20 ϫ 14 cm) separated by a tunnel (5 ϫ 7 ϫ 10 cm; Imetronic). One compartment is constituted of a black floor and walls dimly lit (5 lux), whereas the other is constituted of a white floor and walls intensely lit (1000 lux). The apparatus is equipped with infrared beams and sensors. Mice were placed in the dark compartment, and behavior was automatically recorded for 5 min.

Novelty-suppressed feeding test
Conflict-based anxiety was measured using the noveltysuppressed feeding test. All mice were subjected to fasting 24 h before the beginning of the test, but water was provided ad libitum. Mice were isolated in a single cage 30 min before the beginning of the test. During the test, three food pellets (regular chow) were placed on a square piece of white filter paper positioned in the center of a brightly illuminated (60 lux) open field (50 ϫ 50 cm) filled with ϳ2 cm of sawdust bedding. Each mouse was placed in a corner of the open field facing the open field wall. The latency to the first bite of the food pellet was recorded. The cutoff time was defined as 15 min. After the test was over, the animal was placed in the home cage and left alone for 5 min. The food intake during this period was scored.

Social interaction test
Social interaction was assessed on an open field (50 ϫ 50 cm) dimly lit (Ͻ10 lux) using naive wild-type mice of the same age and weight as interactors. On the first day, all mice were individually placed in the open field arena and left for a 10 min period of habituation. The next day, mice were placed in the open field arena with a wild-type naive interactor, and a 10 min session was recorded. Nose and paw contacts as well as following and grooming were measured. If an interactor failed to engage in any interaction, data from the respective mice were excluded from analysis (one mouse was excluded).

Marble burying
Defensive burying was measured using the marbleburying test performed with 20 small glass marbles (15 mm) evenly spaced in a transparent single cage (21 ϫ 11 ϫ 17 cm) over 4 cm sawdust bedding. The cage was covered by a plastic lid in a room illuminated at 40 lux. The mice were left in the cage for 10 min, and the number of marbles buried more than halfway in the sawdust was counted.

Novelty preference
Novelty preference was assessed in unbiased computerized boxes that have been previously described (Imetronic; Le Merrer et al., 2012). Briefly, the apparatus was composed of two chambers separated by a central alley. Two sliding doors separated the compartments from the central alley. Chambers differed in global shape (but same total surface) and floor texture. Mice were confined to one of the chambers (the familiar chamber) for 15 min before being placed in the central corridor for 5 min. Then, both sliding doors were opened, and mice were allowed to freely explore the apparatus. Time spent in each chamber was recorded, and the novelty preference was calculated as the percentage of time spent in the unfamiliar compartment.

Fear conditioning
Context-and cue-related conditioned fear responses were evaluated using a fear-conditioning paradigm. Experiments were conducted in four dimly lit operant chambers (28 ϫ 21 ϫ 22 cm; Coulbourn Instruments), with a Plexiglas door and a metal bar floor connected to a shocker (Coulbourn Instruments). Chambers had a permanent house light and were equipped with a speaker for tone delivery. An infrared activity monitor, which was used to assess animal motion, was placed on the ceiling of each chamber. The activity/inactivity behavior was monitored continuously during a 100 ms period. Data are expressed in the duration of inactivity per second, and the total time of inactivity displayed by each subject during training and testing sessions was counted. The procedure was similar to one previously described (Goeldner et al., 2009). Briefly, on the first day, animals underwent one conditioning session; on the second day, contextual and cued fear conditioning were tested. The conditioning session was initiated with a 4 min habituation period followed by a 20-s-long tone of 20 KHz/75 dB [conditional stimulus (CS)] coupled with a 0.4 mA footshock [unconditional stimulus (US)] during the last second. Two minutes later, a similar CS-US pairing was presented, and the mice were removed from the apparatus 2 min after the footshock. On the following day, mice were exposed again to the conditioning chamber, and immobility was measured over 4 min to assess contextual fear conditioning. The same day, 5 h after context fear was measured, cued fear conditioning was assessed in modified chambers.

Statistical analysis
All data are expressed as the mean group value Ϯ SEM and were analyzed using Student's test or two-way ANOVA, whenever it was appropriate. When relevant, data were submitted to Sidak's or Tukey's multiplecomparison post hoc analysis. The criterion for statistical significance was p Ͻ 0.05. All statistics were performed using Prism 6 (GraphPad Software).
Converging evidence supports the inhibitory function of striatal D 2 R MSNs in motor output systems. Optogenetic bilateral excitation of these neurons was shown to decrease the initiation of locomotor activity , while ablation or disruption of these neurons increased motor activity (Durieux et al., 2009(Durieux et al., , 2012Bateup et al., 2010). Several reports (Logue et al., 2009;Quintana et al., 2012), including our own work (Meirsman et al., 2016), indicate that Gpr88 gene deletion leads to general hyperactive behavior. In the present report, we show that the deletion of Gpr88 from A 2A R neurons is sufficient to increase locomotor activity. Considering the predominant expression of A 2A Rs in D 2 R MSNs, this observation suggests that GPR88 normally promotes the demonstrated inhibitory function of D 2 R MSNs on locomotor activity. The mechanisms underlying the positive modulatory role of GPR88 on D 2 R MSNs remain to be clarified. Importantly, both total and A 2A R-Gpr88 KO mice show a similar hyperactivity when tested under the same experimental conditions, strongly suggesting that the presence of GPR88 in D 2 R MSN fully accounts for this phenotype.
In a previous report, we showed that Gpr88 Ϫ/Ϫ mice present decreased anxiety-like behaviors (Meirsman et al., 2016). Also, recent reports suggest that D 2 Rs and D 2 R MSNs regulate emotional processing and goaldirected behavior (Hranilovic et al., 2008; Kravitz et al., 2012;Peciña et al., 2013;Brandão et al., 2015;Francis et al., 2015). In the light/dark and elevated plus maze tests, both Gpr88 A2A-Cre and Gpr88 Ϫ/Ϫ mice displayed similar decreased anxiety-like behaviors with increased exploration of the light compartment/open arm of the apparatus. This strongly suggests that GPR88 in D 2 R neurons, but not in D 1 R neurons, regulate anxiety-like behaviors; however, we cannot exclude that GPR88 also regulates emotional behavior in A2AR-expressing neurons at extrastriatal sites (Wei et al., 2014). Note that this anxiety phenotype cannot be explained by their overall hyperactive behavior since the total distance traveled or the number of entries did not differ from that of control animals.
In these ethological anxiety tests, the tendency to avoid threatening stimuli (bright light/exposed arms) is confronted with the inner drive toward exploration, and this conflict is thought to inhibit exploration (Crawley, 2000;Sousa et al., 2006;Bailey and Crawley, 2009;Aupperle and Paulus, 2010). As such, the low-anxiety phenotype of mice lacking GPR88 could result from increased drive toward novelty exploration, decreased avoidance of a threatening environment, or both factors. We therefore evaluated avoidance behavior in the marble-burying test that measures ethological defensive burying (Borsini et al., 2002;De Boer and Koolhaas, 2003). Both Gpr88 Ϫ/Ϫ and Gpr88 A2A-Cre mice buried fewer marbles than control littermates, showing decreased defensive burying that is consistent with reduced threat avoidance in these mice. To tackle approach behavior, we assessed novelty preference in both KO lines. Total but not A 2A R-Gpr88 KO mice showed enhanced preference for the novel compartment when presented with a choice for novel or familiar environment. Similarly, in the novelty-suppressed feeding test, total but not A 2A R-Gpr88 deletion decreased the latency to start eating. In this conflict test, both approach and avoidance components are enhanced by starving and neophagia, respectively. The absence of phenotype of Gpr88 A2A-Cre mice in these two tests could therefore be Figure 6. Total but not A2AR-Gpr88 gene deletion impairs fear conditioning. To assess whether Gpr88 deletion affects fear responses, we tested mice in a fear-conditioning test. A, D, During the conditioning session, mutant and control animals displayed similar levels of immobility before and after tone-shock pairing when compared with control mice. E, Twenty-four hours later, Gpr88 Ϫ/Ϫ mice displayed significantly lower context fear than Gpr88 ϩ/ϩ mice. F, The percentage of immobility of Gpr88 Ϫ/Ϫ was also decreased when tested for cued fear memory. Deletion of Gpr88 in A 2A R-expressing neurons did not affect context (B) or cued (C) fear memories. n ϭ 11, Gpr88 A2A-Cre ; n ϭ 11, Gpr88 flx/flx (A-C); n ϭ 10, Gpr88 Ϫ/Ϫ ; n ϭ 10, Gpr88 ϩ/ϩ (D-F). Solid asterisks: ‫ء‬p Ͻ 0.05 (Student's t test); text asterisks: ‫ء‬p Ͻ 0.05, ‫‪p‬ءء‬ Ͻ 0.01 (post hoc Sidak's multiple comparisons test). explained by unaltered motivation toward new environment or food reinforcement.
Recent reports suggest that D 1 R MSNs encode predictive reward and mediate approach behavior, while D 2 R MSNs mediate aversive and defensive behavior (Hranilovic et al., 2008;Durieux et al., 2009;Hikida et al., 2010Hikida et al., , 2013Kravitz and Kreitzer, 2012;Kravitz et al., 2012Kravitz and Bonci, 2013;Calabresi et al., 2014). For instance, D 2 R MSN neurotransmission blocking in the nucleus accumbens was found to disrupt aversive but not reward learning (Hikida et al., 2010). Moreover, the same authors showed that the impaired aversive behavior was dependent on D 2 R activation (Hikida et al., 2013). Interestingly, in the present report, we show decreased threat avoidance in mice lacking GPR88 in A 2A R neurons. Together with the increased locomotion observed in A 2A R-Gpr88 KO mice, we may therefore hypothesize that the lack of GPR88 in D 2 R MSNs disrupted the activity of this pathway. Further studies using electrophysiological approaches would help to confirm this hypothesis. Also, results from Gpr88 Ϫ/Ϫ mice showing both decreased avoidance behavior and increased novelty and food approach suggest that GPR88 in D 1 R MSNs normally regulates approach behavior.
Finally, we tested the fear responses of total and A 2A R-Gpr88 KO mice for the first time. Total deletion of Gpr88 impaired both context and cued fear responses. Reduced fear responses in total KO mice is in agreement with altered cue-based learning previously reported in Gpr88 KO animals. The lack of phenotype in Gpr88 A2A-Cre mice may be due to the D 1 R neuron-mediated mechanisms contributing to these behaviors. Alternatively, and because central amygdala functioning is essential in the acquisition and expression of fear conditioning (Wilensky et al., 2006), the partial Gpr88 deletion at the level of the amygdala may be insufficient to alter fear responses. Further studies using viral approaches will define the precise role of GPR88 function in amygdala-mediated fear responses.
In sum, our analysis of Gpr88 A2A-Cre mice shows that GPR88 in A 2A R MSNs regulates locomotor and anxiety behaviors. These results represent a first step toward understanding the circuit mechanisms underlying GPR88 function in the brain. Future studies will evaluate the role of GPR88 in D 1 R MSNs, and how this receptor regulates the D 1 R/D 2 R MSN balance. Finally, further demonstration of GPR88 implication in anxiety-related behaviors and threat evaluation definitely posit GPR88 blockade as a new target for the treatment of anxiety-related disorders (Aupperle and Paulus, 2010).