Distinct cortical and striatal actions of a β-arrestin–biased dopamine D2 receptor ligand reveal unique antipsychotic-like properties

Determinants of in Vitro D2R-βarr2 Functional Selectivity.

Clinically effective APDs are either antagonists or partial agonists at both D2R-G protein and D2R-βarr2 signaling pathways (39, 40). We have previously shown that the βarr2-biased D2R ligands UNC9994A (94A) and UNC9975A (75A), which are based on the scaffold of ARI, are weak selective partial agonists at D2R-βarr2 interactions and have little or no agonist activity at D2R-G protein signaling (33). However, we show in Fig. 1 that the agonist versus antagonist profiles of these ligands at D2R-βarr2 interactions in HEK293 cells can be modulated by the complement of GRK2 and βarr2. In a bioluminescence resonance energy transfer (BRET)-based assay (Fig. 1 A, C, and E), compared with DA ARI, 94A and 75A are very weak partial agonists at mediating D2R-βarr2 interactions, but increasing GRK2 levels in these cells markedly enhanced the partial agonist activity of only ARI and 94A. Consistent with pharmacological principles, when tested as antagonists (Fig. 1 B, D, and F), all three ligands fully antagonized DA-mediated D2R-βarr2 interactions, and high GRK2 levels reduced the antagonist efficacy of only ARI and 94A. The profile of 75A was not significantly changed by GRK2, suggesting that it may have slightly different properties than ARI or 94A. However, in a D2R–G protein (GloSensor) assay, ARI and 75A but not 94A behaved as antagonists (Fig. S1 and Table S1), suggesting that only 94A is a completely selective D2R-βarr2 ligand. These results are consistent with the established concept that βarr-dependent GPCR functions are not only dependent on agonist activation but also, enhanced by phosphorylation of the receptor by GRKs (11, 41) and that the expression levels of βarr2 and GRK2 can regulate GPCR signaling and the pharmacological profile of ligands (42–44).

Previous studies have suggested that the levels of GRKs and βarrs can vary significantly between tissues and brain areas (45, 46). Protein levels of βarr2 and GRK2 but not βarr1 are higher in the prefrontal cortex (PFC) compared with the striatum (STR) in mice (Fig. 2 A–D) and humans (Fig. S2 A and B) (Ps < 0.05). Because ARI and 94A gain D2R-βarr2 partial agonism on GRK2 overexpression, we hypothesized that these compounds might behave as agonists in the PFC (high βarr2/GRK2). Previous studies have shown that local injection of the D2R agonist quinpirole can exert antipsychotic-like effects, because it inhibits locomotion induced by the NMDA receptor antagonist PCP (47). PCP-induced locomotion is a pharmacologically induced behavior commonly inhibited by APDs, and the behavioral effects of PCP are thought to be mediated by cortical disinhibition (48). To assess whether PFC D2R-βarr2 agonism can elicit antipsychotic-like effects, we measured the effects of injecting quinpirole, ARI, and 94A locally into the PFC of WT mice on PCP-induced locomotion. Consistent with previous findings, local bilateral PFC injection of the unbiased full-agonist quinpirole significantly inhibited PCP-induced locomotion (Fig. 2E, QUIN) (P < 0.01); 94A mimicked this D2R agonist action (Fig. 2E, 94A) (P < 0.05) and inhibited PCP-induced locomotion, whereas the effect of ARI was weaker and did not reach significance. Coadministration of a pharmacological GRK2 inhibitor [compound 101 (cpd101)] prevented the 94A-mediated inhibition of PCP-induced locomotion (Fig. 2F, cpd101/94A). These results are consistent with our in vitro data, showing that, in a D2R-βarr2 interaction BRET assay, the ability of GRK2 to enhance the agonist effects of both DA and UNC9994A is lost in presence of the GRK2 inhibitor (cpd101) (Fig. S2C). These observations show that D2R-βarr2 agonism in the PFC is sufficient to inhibit PCP-induced locomotion, and this effect is dependent on βarr and GRK2.

Distinct Striatal Vs. Cortical Role of D2R on PCP-Mediated Behavioral Response.

Although D2R PFC agonism inhibits PCP-induced locomotion, striatal D2R antagonism is thought to be the primary mechanism of action of most APDs (49, 50). To broadly explore this paradoxical role of D2R function behaviorally, we selectively deleted D2Rs in either the PFC by injecting an mCherry-Cre adeno-associated virus (AAV) in the previously described D2R floxed (D2^f/f) mice (51) or the STR by crossing D2^f/f mice with adenosine 2A receptor Cre (A2aCre) mice (Fig. S3), and we tested the effect on PCP-induced locomotion. We observed that, compared with controls, deletion of D2Rs in the PFC resulted in an enhanced response to acute PCP injection (Fig. 3A) (Ps < 0.01), consistent with D2R agonism playing an inhibitory role on the PCP response (47). However, deletion of D2Rs in the STR attenuated the PCP response (Fig. 3B) compared with controls (Ps < 0.01), consistent with striatal D2R antagonism inhibiting PCP-mediated responses (49, 50). These pharmacological and behavioral data suggest distinct roles for D2R signaling in the PFC versus the STR in mediating antipsychotic-like effects.

Generation and Characterization of βarr2 Floxed Mice.

To further evaluate the potential opposite role of D2R-βarr2 signaling in the PFC and STR in antipsychotic-like effects, we generated a βarr2 floxed (βarr2^f/f) mouse line for region- and cell-specific Cre-dependent deletion of βarr2 (details are in Materials and Methods and SI Materials and Methods; Fig. S4 A and B); βarr2^f/f mice were indistinguishable from C57BL/6J mice in standard behavioral tests (Fig. S4 C–E). Deleting βarr2 in all neurons (βarr2^f/f CMV-Cre) recapitulated the decrease in the AMPH locomotor response originally observed in whole-body βarr2KO mice (27) (Fig. S4 F–H). We have previously shown that the AMPH locomotor response is mediated at least in part by D2Rs in a βarr2-dependent manner (27, 29), but D2Rs are expressed in multiple regions of the brain, such as midbrain, STR, and PFC (24, 52–57). To confirm which D2R+ population of neurons is responsible for the AMPH response, we deleted βarr2 in several distinct neuronal populations; βarr2^f/f mice were crossed with either D2Cre (all D2R+ neurons) or A2aCre (postsynaptic D2R+ striatal neurons) mice as well as the D1Cre (D1R+ neurons) or ChaTCre (cholinergic interneurons) mice as controls. Using immunohistochemistry (IHC) and real-time quantitative PCR techniques, we confirmed selective deletion of βarr2 in appropriate neuronal populations of all Cre lines crossed with the βarr2^f/f mice (Fig. S5). As shown in Fig. 4, βarr2 in striatal D2R+ neurons seems to play the most prominent role in the AMPH response, because the locomotor response is reduced only in D2R+ or A2aR+ neurons lacking βarr2 (Fig. 4 E–H) but not affected by deletion of βarr2 in either D1R+ neurons alone or cholinergic interneurons (Fig. 4 C, D, and I). These data suggest that deleting βarr2 in D2R striatal neurons is sufficient to mimic antipsychotic-like activity. Importantly, the mice in which βarr2 is inactivated in select D2R+ neuronal populations provide unprecedented models to critically examine the in vivo antipsychotic-like function of our D2R-βarr2–biased compounds.

Region-Specific Responses of Antipsychotics and βarr2-Biased D2R Ligands.

AMPH- and PCP-induced hyperlocomotion are the two commonly used pharmacological models to test APD efficacy. Most APDs are D2R partial agonists or antagonists with varying potencies and efficacies (31, 32, 40, 58) and inhibit either the AMPH- or PCP-induced locomotor response in mice (59). The AMPH-induced locomotor response is dependent on striatal DA release, whereas the behavioral effects of PCP are thought to be mediated by cortical disinhibition and activation of the corticostriatal pathway (48, 50). We tested the ability of representative first, second, and third generation APDs, such as haloperidol, clozapine, and ARI, respectively, along with the βarr2-biased D2R ligands 94A and 75A in both of these pharmacological models. Optimal doses for APDs and the D2R-βarr2–biased ligands were based on previous studies (30, 33, 35).

For the AMPH-induced locomotor response, AMPH was injected at a dose (3 mg/kg) at which there were no significant genotype differences between mice (Fig. 4 C–H). This higher dose allowed us to achieve better separation of effects when treating with APDs and biased compounds. Consistent with their antagonist activity, all tested APDs and biased D2R ligands significantly inhibited AMPH-induced locomotion in all control βarr2^f/f mice (Fig. 5, black bars) (Ps < 0.01). No genotype differences were observed in mice lacking βarr2 in D1R+ neurons for all compounds (Fig. 5 A and B) (Ps < 0.01). All three APDs and 75A also inhibited the AMPH response (Ps < 0.01) in mice lacking βarr2 in either all D2R+ (Fig. 5C) or striatal A2aR+ neurons (Fig. 5E). However, in these mice, 94A showed markedly diminished antipsychotic-like activity compared with genotype controls (Ps < 0.05), consistent with this compound being a selective βarr2-biased D2R antagonist. Compared with other APDs, the observations with 94A indicate that D2R-βarr2 antagonism is sufficient but not necessary for efficacious antipsychotic-like activity in the AMPH pharmacological model, which is consistent with previous observations (60).

For PCP-induced responses, in mice lacking βarr2 in D1R+ neurons, all drugs significantly inhibited PCP-induced locomotion compared with vehicle (VEH)-treated controls (Fig. 6A) (Ps < 0.05), and there were no genotype differences. ARI and 75A also inhibited PCP-induced locomotion in mice lacking βarr2 in D2R+ striatal neurons (Fig. 6B) (Ps < 0.05) or all D2R+ neurons (Fig. 6C) (Ps < 0.05) compared with genotype and VEH controls. In contrast to the AMPH response (Fig. 4), however, we observed a loss of 94A activity only in mice lacking βarr2 in all D2R+ neurons (Fig. 6B) (P = 0.2526) but not in striatal D2R+ neurons (Fig. 6C) (P < 0.05), suggesting a role for βarr2 outside the STR, because A2aCre is essentially STR-selective. Although the D2Cre is expressed in STR, PFC, and midbrain, a cortical role for βarr2 in 94A-mediated inhibition of PCP-induced locomotion is most likely, because evidence suggests against a role for midbrain DA neuron βarr2 (61) and the primary site of action for PCP is in the PFC. However, to further rule out the contribution of midbrain presynaptic βarr2 and confirm a role for cortical βarr2, we deleted βarr2 in PFC and STR or PFC alone by injecting mCherry-Cre AAV in the PFC of βarr2^f/f A2aCre (PFC + STR βarr2KO) or βarr2^f/f mice (PFC βarr2KO) (Fig. S6); 94A inhibited PCP-induced locomotion in PFC βarr2KO mice (Fig. 6D) (P < 0.05) but lost its antipsychotic-like activity when βarr2 was deleted in both PFC and STR (Fig. 6D), thus confirming a dual dependence on cortical and striatal βarr2. These data suggest that cortical and striatal βarr2 are necessary for the antipsychotic-like effect of 94A. Thus, our behavioral data further support our initial supposition that distinct mechanisms might regulate the antipsychotic-like effect of D2R-βarr2 signaling in the PFC (agonism) versus the STR (antagonism). We next wanted to confirm the neuronal mechanism of these distinct phenomena electrophysiogically in the PFC and STR.

Effect of 94A on Excitability of Cortical D2R+ Fast-Spiking Interneurons.

The PFC comprises multiple neuronal cell types, and many of these neurons, in particular GABA interneurons and pyramidal glutamatergic neurons, express D2Rs (56, 62–64). GABAergic interneurons are thought to play a critical role in schizophrenia pathophysiology in humans and animal models of schizophrenia. Postmortem brain analyses of patients (65–71) and behavioral studies in rodents have implicated GABAergic parvalbumin+ (PV+) fast-spiking interneurons (FSIs) in altered excitation–inhibition imbalance and cognitive impairment in schizophrenia (72, 73). The D2R agonist quinpirole increases FSI excitability in adult rodents in a similar fashion as D1R agonists (64), suggesting that this excitatory effect is not mediated by the canonical inhibitory Gα_i/o activation but presumably, is through a G protein-independent pathway. Our data suggest that the higher levels of βarr2 and GRK2 in the cortex might support this G protein-independent agonist signaling of D2R ligands (Figs. 1 and 2). Additionally, the elevated βarr2 and GRK2 expression is present in PFC PV+ FSI (Fig. S7) and presumably, pyramidal neurons (non-PV+ cells) as well. Given the pharmacological, genetic, and behavioral evidence for the importance of FSIs in schizophrenia pathophysiology, we chose to determine the functional impact of higher cortical levels of βarr2 and GRK2 in FSIs. We performed whole-cell, current clamp slice recordings in prefrontal GAD67+ FSI from control (βarr2^+/+) and global βarr2KO mice and assessed the effects of UNC9994A. FSIs were visually identified in acute slices from Gad1-eGFP adult mice and further identified by their responses to hyperpolarizing and depolarizing current injections (Fig. 7A). As expected from previous studies with D2R agonists, like quinpirole (64), ARI and 94A increased prefrontal FSI excitability as measured by changes in FSI action potential frequency in response to minimal amounts of depolarizing current injection. The D2R partial agonist ARI (10 µM) elicited a modest increase in action potential firing (+6.2 ± 1.4 Hz after 20 min of exposure relative to a 10 min predrug baseline; n = 6) (Fig. 7B). This effect was prevented by the D2R antagonist eticlopride (10 µM) and even led to a slight decrease in FSI excitability (−2.1 ± 2.2 Hz after 20 min of exposure of 10 µM eticlopride and 10 µM ARI; n = 4) (Fig. 7B). These data suggest that ARI has modest D2R agonist-like activity on FSIs in the PFC and are consistent with the partial agonist activity observed with ARI when GRK2 and βarr2 are overexpressed in HEK293 cells (Fig. 1 B, D, and F). Interestingly, 94A (10 µM) elicited a significantly more robust increase in FSI excitability compared with ARI (+25.8 ± 3.8 Hz after 20 min of 94A; +6.1 ± 1.4 Hz after 20 min of 10 µM ARI; n = 4 and n = 6, respectively; P < 0.01) (Fig. 7 C and D). Bath application of eticlopride (10 µM) also prevented the increase in FSI excitability by 94A and again, led to a slight decrease in FSI excitability [−5.0 ± 12.6 Hz after 20 min of exposure to eticlopride and 94A (10 µM); n = 3]. Additionally, the effects of 94A were dependent on βarr2-signaling, because UNC9994A failed to enhance FSI excitability in βarr2KO mice [+1.3 ± 1.6 Hz after 20 min of 94A (10 µM)] (Fig. 7E). Finally, the presence of the membrane-permeable GRK2 inhibitor cpd101 (30 μM) prevented the increase in FSI excitability elicited by 94A (Fig. 7F) (−2.3 ± 3.2 Hz after 20 min of exposure of cpd101 and 94A; n = 5). These data further highlight that 94A requires GRK2 in addition to βarr2 to produce its D2R-dependent, agonist-like effects in cortical FSI.

Effect of 94A on Excitability of Striatal D2R+ Medium Spiny Neurons.

Based on the above behavioral and biochemical observations, because βarr2 and GRK2 levels are significantly lower in the STR than in the PFC, one might expect that the agonist activity of 94A would be reduced in striatal neurons. The D2R agonist quinpirole decreased excitability of striatal medium spiny neurons [MSNs; −6.8 ± 2.2 Hz after 20 min of quinpirole (10 µM); n = 8] (Fig. 8B, black line). These findings are consistent with previous reports showing that D2R agonists reduce the excitability of striatal MSNs (74). Unlike quinpirole, 94A produced a negligible reduction in striatal MSN excitability [−2.6 ± 1.0 Hz after 20 min of 94A (10 µM); n = 8; P = 0.0137 compared with quinpirole] (Fig. 8 B and C). Thus, 94A lacks agonist-like activity in striatal MSNs, showing that D2R-βarr2–biased ligands can have distinct electrophysiological actions in the PFC and STR.

Animals and Drugs.

Mice were age- and sex-matched, and for all experiments, at least eight mice were used for each genotype/treatment group (unless specified in the figures). All experiments were performed in an unblinded fashion during the light cycle and in drug-naïve animals. D1Cre, D2Cre, and A2aCre were all obtained from GENSAT and backcrossed with the C57BL/6J strain for at least 10 generations (30). The Cre lines were crossed with the βarr2^f/f mice (C57BL/6J:129SvJ mixed strain), such that littermates were βarr2^f/f (controls) and βarr2^f/f Cre (knockouts). The D2^f/f mice were previously described (51), and the ChAT-IRES-Cre mice were obtained for Jackson Laboratories (stock 018957) (89). AMPH, PCP, and haloperidol were purchased from Sigma-Aldrich and freshly prepared in saline (SAL); ARI was purchased from Sequoia Biosciences, dissolved in minimal glacial acetic acid, brought up to volume with distilled water, and dilutions were made in SAL. Clozapine was dissolved in a stock solution of Na-tartrate, and additional dilutions were made in SAL. University of North Carolina biased compound stock solutions were prepared in a minimal amount of DMSO, sonicated, and dissolved in 20% (vol/vol) hydroxylpropyl cyclodextrin (Sigma-Aldrich), and dilutions were subsequently made in SAL. All drugs were injected i.p. at a volume of 5 μL/g body weight, and the doses chosen were based on previous studies (30, 33).

Generation of βarr2 Targeting Vector and Floxed βarr2 Mice.

To generate floxed βarr2 mice, we inserted LoxP sites flanking exon 2 of the Arrb2 gene. The genomic region encompassing and flanking exon 2 of the Arrb2 gene was isolated from a bacterial artificial chromosome library (Coriell Institute) containing genomic DNA from the 129SvJ mouse strain. The homology arms were generated by PCR amplification of regions upstream and downstream of exon 2; inserted into the pEZ-Frt-Lox-DT vector generated in Klaus Rajewsky’s laboratory, Max Delbruck Center, Berlin; and obtained from Addgene (11736). Exon 2 of the Arrb2 gene was inserted into the pEZ-Frt-Lox-DT vector, such that it was now flanked by LoxP sites and next to the neomycin selection cassette that was flanked by flippase recognition target (FRT) sites for excision by flippase (FLP) recombinase (Fig. S4A). The targeting vector was then injected into C57BL/6J-129SvJ ES cells at the Duke University Transgenic Facility, selected for recombination by neomycin resistance, and confirmed by PCR. Two optimal ES cell clones were injected into blastocysts that were reimplanted into recipient female mice. Chimeric pups obtained were selected by coat color, crossed with C57BL/6J (Jackson Laboratories) mice to confirm germline transmission, and screened by PCR for appropriate insertion (Fig. S4B). The floxed βarr2 (+/f) mouse was then crossed with an FLPe recombinase mouse (Jackson Laboratories) to remove the neo^R selection cassette. Subsequently, the floxed βarr2 (+/f) mice were either propagated to generate βarr2^f/f mice for all experimental studies or crossed with a CMV-Cre mouse line (stock 006054; Jackson Laboratories) to globally delete the βarr2 allele (^−/−) for confirmation of Cre-lox recombination. These mice were then propagated, and genotypes were confirmed by Western blot for deletion of the Arrb2 gene (Fig. S4F).

Prepulse Inhibition.

The βarr2^f/f mice were habituated for 10 min to white noise (64 dB) in the prepulse inhibition (PPI) apparatus (San Diego Instruments). They were tested with seven pulse-alone trials (40-ms bursts of 120-dB white noise) followed by combinations of the prepulse–pulse (20-ms prepulse stimuli 4, 8, or 12 dB above the white noise background), pulse alone, and null trials (64 dB) and terminated with seven pulse-alone trials (35). PPI responses were calculated as %PPI = [1 − (prepulse trials/startle-only trials)] × 100.

Translating Ribosome Affinity Purification.

Translating ribosome affinity purification (TRAP) was performed as described previously (90, 91), with minor modifications. Briefly, WT or βarr2^f/f mice harboring a Cre-inducible TRAP transgene (ribosomal L10a-GFP) were crossed with βarr2 wt/wt A2aCre⁺ or βarr2^f/f A2aCre⁺ mice for cell type-specific expression of the TRAP transgene in D2R-MSNs. These mice were euthanized, striata were rapidly dissected, and TRAP was performed. The mRNAs derived from WT or βarr2^−/− mice were converted into cDNA using Superscript VILO Mastermix (ThermoFisher Scientific) and then, subjected to Taqman qPCR. Two Primetime Standard qPCR Assays were designed (IDT) to specifically detect exon 2 of βarr2 (ENSMUST00000108568). Probe set 1: forward primer 5′-CCCTAACTGCAAGCTCACC-3′, reverse primer 5′-CTTTCCGGTCCTTCAAGTAGTC-3′, and probe 5′-56-FAM/AGCGCGACT/ZEN/TTGTAGATC ACCTGG/3IABkFQ/-3′ and probe set 2: forward primer 5′-GCGCACCATGGGAGAAA-3′, reverse primer 5′-AAGTACACGGTGAGCTTGC-3′, and probe 5′-/56-FAM/AGGGTCTTC/ZEN/AAGAAGTCGAGCCCTA/3IABkFQ-3′. Taqman gene expression assays (ThermoFisher) were ordered for Drd2 (MM00438545.m1), PENK (MM01212875.m1), Drd1 (MM01353211.m1), Tac1 (MM01166996.m1), and GAPDH (Mm99999915.g1). Real-time PCR was performed with TaqMan Fast Advanced Master Mix (4444557; ThermoFisher) on a StepOnePlus Real-Time PCR System (ThermoFisher). Relative mRNA abundance was determined using the ∆∆CT (threshold cycle) method (92).

cAMP Inhibition Assay.

To measure D2R Gi-mediated cAMP inhibition, HEK293T cells were cotransfected in a 1:1:2 ratio with human D2_Long receptor, GRK2, or pcDNA3.1 and a split luciferase-based cAMP biosensor (GloSensor; Promega), respectively. After at least 24 h, transfected cells were plated in polylysine-coated 384-well white clear-bottom cell culture plates with DMEM and 1% dialyzed FBS at a density of 15,000–20,000 cells per 40 µL per well and incubated overnight. Next day, drug dilutions were prepared in fresh assay buffer (20 mM Hepes, 1× HBSS, 0.3% BSA, 0.03% ascorbic acid, pH 7.4) at 5× and 6× concentrations for APDs and DA, respectively. APD concentrations corresponding to dose–response curves were added in 5-µL volume to each well containing 20 µL per well assay buffer (20 mM Hepes, 1× HBSS, pH 7.4). Plates were allowed to incubate for exactly 30 min in the dark at room temperature, and immediately afterward, 5 µL DA (10 nM final concentration) was added and allowed to incubate for an additional 15 min. To stimulate endogenous cAMP production via β-adrenergic receptor-mediated Gαs activation, 10 µL 4× isoproterenol (200 nM final concentration) diluted in assay buffer supplemented with GloSensor assay reagent was added per well, and luminescence intensity was quantified 15 min later using a Wallac TriLux Microbeta (Perkin-Elmer) Luminescence Counter.

βarr BRET Assay.

To measure D2R-mediated βarr2 recruitment, HEK293T cells were cotransfected in a 1:1:15 ratio with mouse D2_Long receptor fused to C-terminal renilla luciferase (RLuc8), GRK2, or pcDNA3.1 and a Venus-tagged N-terminal βarr2 (a gift from Jonathan Javitch, Columbia University, New York), respectively. After at least 24 h, transfected cells were plated in polylysine-coated 96-well white clear-bottom cell culture plates with DMEM and 1% dialyzed FBS at a density of 250,000–500,000 cells in 200 µL per well and incubated overnight. Buffers used for the BRET assay and to dilute drugs were exactly the same as for the cAMP inhibition assay. The next day, media were decanted, cells were washed twice with assay buffer, and APDs were added at concentrations corresponding to the dose–response curves and allowed to incubate for exactly 30 min. Afterward, EC₈₀ of DA was added to each well and allowed to incubate for exactly 5 min. The RLuc substrate, coelenterazine h (5 µM final concentration; Promega), was added per well, and exactly 5 min later, both luminescence at 485 nm and fluorescent enhanced yellow fluorescent protein (eYFP) emission at 530 nm were measured for 1 s per well using a Mithras LB940 Multimode Multiplate Reader (Berthold Technologies). The ratio of eYFP/RLuc was calculated per well, and the net BRET ratio was calculated by subtracting the eYFP/RLuc per well from the eYFP/RLuc ratio without Venus-Arrestin present. The net BRET ratio was plotted as a function of drug concentration using Graphpad Prism 5 (Graphpad Software Inc.). The GRK2/3 inhibitor cpd101 (93) was purchased from Hello Bio.

Western Blot.

Western blot analyses were performed on human postmortem or drug-naïve mice brain tissue as described previously (30). Briefly, STR and frontal cortex regions were rapidly dissected from mouse brains on ice, or frozen human postmortem tissue (six patient samples from PFC area 10 and caudate) was obtained from Craig Stockmeier, Postmortem Brain Core, University of Mississippi Medical Center, Jackson, MS. Tissue samples were immediately sonicated and boiled in preheated 1% SDS solution. Loading buffer was added to the samples, which were loaded onto 10% (wt/vol) SDS gels (Nupage; Invitrogen), transferred to nitrocellulose membranes, and exposed to antibodies for GRK2/3 (1:500; 05–465; EMD Millipore), βarr1/2 (A2CT clone; gift from Robert J. Lefkowitz, Duke University, Durham, NC), and GAPDH (1:5,000; MAB374; EMD Millipore). Incubation with primary antibody was followed by exposure to IR secondary antibody (LICOR), and the blots were developed using an LICOR Odyssey Detection System and quantified using ImageJ (NIH).

Human Brain Tissue.

Human prefrontal cerebral cortex (area 10) and caudate nucleus were obtained at autopsy from psychiatrically normal individuals (Table S2), rapidly frozen, and stored at −80 °C. The procedures used for brain tissue collection, toxicology, and psychological autopsy have been published elsewhere (94, 95). Toxicology screens revealed nothing remarkable.

Animals and Drugs.

All mouse studies were conducted in accordance with the NIH guidelines for animal care and use and through animal protocols approved by the Duke University Animal Care and Use Committee and Pfizer’s Institutional Animal Care and Use Committee. All mice were housed in a 12-h light–dark cycle at a maximum of five per cage, provided with food and water ad libitum, and tested at 10–20 wk of age. Details of mouse lines used, generation of βarr2^f/f animals, and drugs and chemicals used are in SI Materials and Methods.

Locomotor Activity.

Activity was measured in an Accuscan Activity Monitor (Accuscan Instruments) and performed as described (30, 88). Briefly, mice were allowed to habituate to the open field for 30 min, injected with various drugs, and returned to the open field. Locomotor activity was measured in 5-min intervals, and data were analyzed for the distance traveled in 5-min increments over 120 min. All APDs and tool compounds were administered i.p. and injected 10 min before AMPH or PCP injections.

IHC.

Fifty-micrometer-thick vibratome cut sections of formalin-fixed mouse brains were processed for IHC analyses as described previously (30). To assess neuron-specific deletion of βarr2, antibodies to βarr2 (generated in rabbit; at 1:300; gift from Jeff Benovic, Thomas Jefferson University, Philadelphia), Cre recombinase (mouse antibody MAB3120; at 1:500; EMD Millipore), and D2 neuron marker enkephalin (ENK; rabbit antibody; AB5026; at 1:500; EMD Millipore) were used. Because both antibodies to βarr2 and ENK were from rabbit, they were used in combination with antibodies to Cre only. Additionally, because βarr2 levels in the STR are much lower, a TSA Amplification System (Perkin-Elmer) was used to enhance detection of βarr2. Antibodies to PV (1:500; PVG-213; Swant Inc.) and GRK2/3 (1:500; 05–465; EMD Millipore) were used to label PV+ interneurons and assess the levels of GRK2, respectively, in these cells. All representative images are from IHC analyses from at least two to three mice for each group.

Western Blot.

Western blot analyses were performed on postmortem human or drug-naïve mice brain tissue as described previously (30). For human brain tissue, all procedures were carried out in compliance with an approved protocol from the University of Mississippi Medical Center Institutional Review Board. Written informed consent was obtained from legally defined next of kin for tissue collection and informant-based retrospective diagnostic interviews (Table S2). Human or mouse tissue lysates were loaded onto SDS gels followed by transfer onto nitrocellulose membranes. Membrane blots were incubated with primary antibody followed by IR secondary antibody (LICOR), and the blots were developed using an LICOR Odyssey Detection System. Details are in SI Materials and Methods.

Stereotaxic Surgeries and Virus.

Deletion of βarr2 or D2Rs in the PFC was achieved by a viral approach; βarr2^f/f or D2^f/f mice were stereotaxically injected bilaterally with 0.5 μL AAV serotype 2/8 (UNC Viral Vector Core). βarr2^f/f (PFC βarr2KO) or A2aCre βarr2^f/f (PFC + STR βarr2KO) mice were injected with mCherry-Cre AAV, whereas D2^f/f mice were injected with either GFP or mCherry-Cre AAV at coordinates +2.5 mm anteroposterior (AP), ±0.3 mm mediolateral (ML), and −1.8 mm dorsoventral (DV) from bregma to target the prelimbic/infralimbic region of the PFC. Mice were allowed to recover for 3 wk to allow for viral expression of GFP or mCherry-Cre before behavioral testing.

Cannulation and Local PFC Drug Injections.

For local PFC injection of drugs, we inserted bilateral guide cannulas (Plastics One) into the PFC of WT mice. The bilateral guide cannulas were inserted at +2.0 mm AP with 1.0-mm spacing (±0.5 mm ML) and −2.0 mm DV and fixed to the skull with dental cement. Mice were allowed to recover for 2 wk, and then, drugs were injected using an automated injection system. Drugs were dissolved in VEH [10% (vol/vol) DMSO and 20% (vol/vol) hydroxypropyl cyclodextrin], and either 1 μg (Quinpirole and UNC9994A) or 0.5 μg (cpd101–GRK2/3 inhibitor; Hello Bio, Bristol, UK) per 0.5 μL were injected per side at a rate of 0.4 μL/min. After local injection, mice were placed in the activity monitors for 10 min before systemic (i.p.) injection with PCP (6 mg/kg), and then, locomotor activity was recorded.

cAMP Inhibition Assay.

To measure D2R Gαi-mediated cAMP inhibition, a split luciferase-based cAMP biosensor (GloSensor; Promega) in HEK293T cells was used. The assay was performed in a 384-well plate using a Wallac TriLux Microbeta (Perkin-Elmer) Luminescence Counter. Details are in SI Materials and Methods.

βarr BRET Assay.

To measure D2R-mediated βarr2 recruitment, a mouse D2_Long receptor fused to C-terminal renilla luciferase and a Venus-tagged N-terminal βarr2 (a gift from Jonathan Javitch, Columbia University, New York) expressed in HEK293T cells were used in a BRET assay. The assay was performed in a 96-well plate using a Mithras LB940 Multimode Plate Reader (Berthold Technologies). Details are in SI Materials and Methods.

Electrophysiology.

Statistical Analyses.

Data were analyzed by standard one- and two-way ANOVA or two- and three-way repeated-measures ANOVA (RMANOVA) tests for comparison between genotypes, treatments, or doses. Individual genotypes, treatments, or doses were compared using a post hoc Bonferroni’s test. Data were analyzed for normality with equal variance, and only parametric tests were used. Data points were excluded based on previously established criterion and set to ±2 SDs from the group mean. Data are presented as mean ± SEM.