Adenosine A2A Receptor Signaling and Golf Assembly Show a Specific Requirement for the γ7 Subtype in the Striatum*

The adenosine A2A receptor (A2AR) is increasingly recognized as a novel therapeutic target in Parkinson disease. In striatopallidal neurons, the G-protein αolf subtype is required to couple this receptor to adenylyl cyclase activation. It is now well established that the βγ dimer also performs an active role in this signal transduction process. In principal, sixty distinct βγ dimers could arise from combinatorial association of the five known β and 12 γ subunit genes. However, key questions regarding which βγ subunit combinations exist and whether they perform specific signaling roles in the context of the organism remain to be answered. To explore these questions, we used a gene targeting approach to specifically ablate the G-protein γ7 subtype. Revealing a potentially new signaling paradigm, we show that the level of the γ7 protein controls the hierarchial assembly of a specific G-protein αolfβ2γ7 heterotrimer in the striatum. Providing a probable basis for the selectivity of receptor signaling, we further demonstrate that loss of this specific G-protein heterotrimer leads to reduced A2AR activation of adenylyl cyclase. Finally, substantiating an important role for this signaling pathway in pyschostimulant responsiveness, we show that mice lacking the G-protein γ7 subtype exhibit an attenuated behavioral response to caffeine. Collectively, these results further support the A2AR G-protein αolfβ2γ7 interface as a possible therapeutic target for Parkinson disease.

address these questions, a gene-targeting approach has been used to delete the various ␣ subunit genes in mice, leading to the identification of physiological functions for most of them (5). By contrast, little attention has focused on the ␤ and ␥ subunit genes. In particular, several features of the ␥ subunit genes suggest they may perform heterogeneous functions in vivo. Analogous to their ␣ partners, the various ␥ subtypes show substantial structural diversity and exhibit pleiotropic patterns of expression (2). Accordingly, we have undertaken a gene targeting approach to systematically ablate the individual ␥ subtypes in mice (6,7), with the ultimate goal of elucidating their biological functions.
Our recent work has demonstrated that knock-out of Gng7, encoding the ␥ 7 subtype, produces a behavioral phenotype resulting in part from a localized defect in dopamine D 1 receptor (D 1 R) 2 signaling in the brain (6). Within the brain, the striatum collects and processes information from the cerebral cortex and thalamus affecting the control of voluntary movements (8). Accounting for Ͼ90% of neurons within the striatum, the medium spiny neurons are comprised of two distinct subpopulations that are classified on the basis of their distinct circuitries (9). The striato-nigral (SN) neurons projecting to the substania nigra pars reticulata and entopeduncular nucleus constitute the direct tract, whereas the striato-pallidal (SP) neurons projecting to the lateral part of the globus pallidus comprise the indirect tract (8). Typically, a coordinated balance between these two tracts produces normal movements, whereas a preponderance of one tract over the other is implicated in producing motor abnormalities associated with basal ganglia disorders (10,11).
In the SN neurons, the D 1 R acts through the G-protein ␣ olf subunit to stimulate cAMP production (12). Based on recent analyses of Gng7 Ϫ/Ϫ mice, this action is also dependent on the ␥ 7 subtype (6). Intriguingly, in the SP neurons, the adenosine A 2A receptor (A 2A R) also couples through the G-protein ␣ olf subunit to enhance cAMP production even though this pathway produces the opposite behavioral effect (12). In the present study, we explored whether a specific G-protein ␣ olf ␤ 2 ␥ 7 subunit combination is required for this pathway in SP neurons. Using mice with targeted deletions of Gng7 or Gnal, lacking the * This work was supported, in whole or in part, by National Institutes of Health G-protein ␥ 7 (6) or ␣ olf (13) subunits, respectively, we showed that levels of the ␣ olf and ␤ 2 proteins were selectively and coordinately reduced in Gng7 Ϫ/Ϫ mice, whereas levels of ␥ 7 were largely unaffected in the Gnal Ϫ/Ϫ mice. Notably, these results indicate that assembly of the ␣ olf ␤ 2 ␥ 7 heterotrimer is an ordered process that is controlled by the amount of the ␥ 7 subtype. Moreover, loss of the G-protein ␥ 7 subunit led to defects in both D 1 R (6) and A 2A receptor activation of adenylyl cyclase without producing any gross alterations in locomotor behavior typical of Parkinson disease. Importantly, these findings contribute to a growing literature that suggests that blockade of A 2A R signaling in the striatum may be an effective strategy for treating various neurological and addictive disorders.

EXPERIMENTAL PROCEDURES
Production of Mice-Disruption of Gng7, the gene encoding the G-protein ␥ 7 subunit in mice, was described previously (6). Gng7 ϩ/Ϫ mice were backcrossed to C57BL/6 mice (Jackson Laboratories, Bar Harbor, ME) for 5 generations, or separately to BALB/c mice (Jackson Laboratories) for 5 generations. Gng7 ϩ/Ϫ mice were intercrossed to produce the Gng7 Ϫ/Ϫ mice and wild-type littermates used in these experiments. In this work, we will describe these mice as "on a C57BL/6 background," or "on a BALB/c background." However, after only 5 generations of backcrosses, there is still some contribution to their genetic makeup from the original ES cells (129SvEv Brd , Lexicon Genetics, Inc., The Woodlands, TX), from the dam that was bred with the chimera (C57BL/6 albino, Lexicon), and from the Cre recombinase expressing strain that we utilized (BALB/c-TgN(CMV-Cre)#Cgn, Jackson Laboratories). Hence, it was essential to use littermates to control for the possible influence of genetic background to observed responses. On the C57BL/6 background, 16 Gng7 Ϫ/Ϫ mice (8 males and 8 females) and 16 wild-type littermates (8 males and 8 females) were studied. On the BALB/c background, 12 Gng7 Ϫ/Ϫ mice (6 males and 6 females), and 14 wild-type littermates (4 males and 10 females) were studied. Genotypes were determined by PCR analysis of tail biopsy DNA as described previously (6).
Mice with a disrupted Gnal gene, that lack the G-protein ␣ olf , were described previously (13). These were backcrossed for up to 9 generations with C57BL/6 mice to obtain homozygous (Gnal Ϫ/Ϫ ) and their control littermates (Gnal ϩ/ϩ ). For comparing homozygous mutant and wild-type mice, 6-week-old male and female mice were used for experiments.
Animal Care and Approval-Mice were segregated by sex and group housed in plastic microisolator cages in ventilated racks (Thoren Caging Systems, Inc., Hazelton, PA). Mice were given ad lib access to water and Mouse Diet 9F (Purina Mills, LLC, St. Louis, MO). Environmental factors included temperature and humidity control and a 12-hour light/dark cycle. The animal facility is maintained as virus antibody-free and parasite-free. The Geisinger Clinic Institutional Animal Care and Use Committee approved animal research protocols.
Adenylyl Radioligand Binding Assays-Radioligand binding to A 2A R in striatal membranes prepared from Gng7 Ϫ/Ϫ mice and wild-type littermates was performed using either the radiolabeled agonist [ 125 I]2-[2-(4-amino-3-iodo-phenyl)ethylamino]adenosine ( 125 I-APE) or the radiolabeled antagonist 125 I-ZM241385 as described previously (14,15). Binding was performed using striatal membranes prepared from wild type or Gng7 Ϫ/Ϫ mice in buffer containing 10 mM HEPES pH 7.4, 1 mM EDTA, 5 mM MgCl 2 , and 1 unit/ml adenosine deaminase. The agonist, 125 I-APE binds to two affinity states of the A 2A R, a high affinity state corresponding to receptor-G-protein complexes, and a low affinity state corresponding to receptors uncoupled from G-proteins (15). Because GTP␥S added to membranes uncouples receptors from G-proteins, 50 M GPT␥S was added to some membranes to measure agonist binding to largely uncoupled receptors. In the absence of added GTP␥S, 125 I-APE binds preferentially to G-protein coupled receptors. 125 I-APE also binds to uncoupled receptors, but may under measure total receptor number because a fraction of the radioligand may dissociate for the low affinity site during washing of filters. Hence, the total number of receptors (B max ) was more accurately detected as the number of specific binding sites for the high affinity antagonist, 125 I-ZM241385. Nonspecific radioligand binding was measured in the presence of 50 M N-ethylcarboxamidoadenosine (NECA).
Real-time RT-PCR-To examine levels of mRNA for ␣ olf , ␤ 2, ␥ 7 , and DARPP-32, RNA was prepared from striatum of six Gng7 Ϫ/Ϫ mice and six wild-type littermates at the N24 back-cross to C57BL/6 and real-time RT-PCR was conducted as described previously (20). Primers were as follows: Gnal (CCT  TCC TAC TTG CCT GAC CGC; TGA CGA TAG TGC TTT  TCC  Locomotor Activity-Locomotor activity was quantified in CLAMS cages (Columbus Instruments, Columbus, OH). The cages consist of clear plastic boxes (20 cm ϫ 10 cm ϫ 12.5 cm) fitted with three rows of 8 photoelectric sensors (x, y, and z directions). The mice were placed in the CLAMS cages at 11 am and remained in the cages for 4 h. During this time the mice had ad lib access to water. Every minute the numbers of individual (total) and consecutive (ambulatory) photobeam breaks for each of 3 sensor arrays, and the number of contacts with the sipper tube were recorded. For caffeine trials, mice were removed from the cages after 1 h and given an intraperitoneal injection of saline or caffeine (5 ml/kg). Mice that had been habituated to the CLAMS cages and to the injection procedure with saline were used in drug trials. The locomotor response to drug is expressed as an increment over the response to saline. Locomotor activities were studied at age 10.0 Ϯ 0.8 weeks for C57BL/6 background mice and age 8.5 Ϯ 0.3 weeks for BALB/c background mice. The response to caffeine was studied at age 23.6 Ϯ 3.0 weeks for C57BL/6 background mice and age 14.6 Ϯ 0.3 weeks for BALB/c background mice.
Statistical Analysis-Sample statistics and Student's t-tests were computed using Excel (Microsoft). Data are presented as means Ϯ S.E. of the mean. Locomotor activity was compared by repeated measures multivariate analysis of variance (MANOVA), using JMP (SAS Institute Inc., Carey, NC).

RESULTS
The A 2A R is primarily responsible for the psychostimulant action of caffeine (21,22). Because blockade of this receptor has been shown to reverse the hypolocomotor phenotype resulting from dopamine deficiency (23) and dysfunctional dopamine signaling (24), the A 2A R signaling components are increasingly recognized as valid therapeutic targets for treating Parkinson disease and for reducing the side effects of levodopa therapy (25,26). In SP neurons, the G-protein ␣ olf subunit has been shown to positively couple this receptor to stimulation of cAMP production (27,12). However, little information is available on the obligatory ␤ and ␥ components involved in this context. By applying biochemical and behavioral approaches to a novel mouse model, we found that the G-protein ␥ 7 subunit is specifically required for both A 2A R signaling and psychostimulant response to caffeine. Identifying a mechanistic basis for this requirement, we show that the ␥ 7 protein drives the preferential assembly of a G-protein ␣ olf ␤ 2 ␥ 7 heterotrimer in the striatum that is involved in a key signaling pathway controlling locomotion and reward.
Defective A 2A R Signaling in Mice Lacking the ␥ 7 Protein-Regulation of cAMP production in medium spiny neurons rep-resents a primary target of many neurotransmitters and psychoactive drugs that affect short-and long-term locomotor responses (28). Virtually all medium spiny neurons, including SN and SP neurons, express substantial levels of ␥ 7 mRNA (29,30,31). Because adenylyl cyclase signaling by the D 1 R was shown to be dependent on ␥ 7 expression in SN neurons (6), we explored whether adenylyl cyclase signaling by the A 2A R is similarly dependent on expression of ␥ 7 in SP neurons. As a biochemical assay, we used the selective A 2A R agonist, CGS-21680, to stimulate adenylyl cyclase activity in striatal membranes prepared from Gng7 Ϫ/Ϫ mice. By comparison to their wild type littermates, adenylyl cyclase activity in response to 10 M or 25 M CGS-21680 was reduced by 30 -40% in striatal membranes from Gng7 Ϫ/Ϫ mice (Fig. 1A). Furthermore, cAMP production in response to 100 M forskolin was also reduced by ϳ35% in striatal membranes from Gng7 Ϫ/Ϫ mice (Fig. 1B).
Because forskolin-stimulated adenylyl cyclase activity was reduced, we could not assess whether the impaired response to the A 2A R agonist was due to a defect in G-protein coupling and/or adenylyl cyclase activation. Therefore, as a second biochemical assay, we used high affinity agonist binding to directly measure the actual interaction between the A 2A R and the G-protein. Initial saturation binding experiments were performed with the A 2A R agonist, 125 I-APE, on pooled samples of striatal membranes from either Gng7 Ϫ/Ϫ mice or their wildtype littermates (12 mice in each group). In the wild-type sample, the addition of GTP␥S dramatically reduced agonist binding by greater than 75%, indicating a significant portion of the A 2A R was associated with G-protein ( Fig. 2A). In contrast, in the knock-out sample, addition of GTP␥S produced little reduction in A 2A R agonist binding (Fig. 2B), suggesting most A 2A R was no longer coupled to G-protein. Finally, there was no significant difference in the binding of the A 2A R antagonist, 125 I-ZM241385, between the samples (Fig. 2C), indicating the total number of A 2A R was comparable between the two genotypes. Taken together, these results indicate a striking reduction in the fraction of the A 2A R that was coupled to G-protein in striatal membranes from Gng7 Ϫ/Ϫ mice. Subsequent binding studies were performed on striatal membranes from individual mice representing each genotype. To calculate the fraction of the A 2A R pool that was coupled to G-protein, we determined the ratio of specific GTP␥S-sensitive 125 I-APE agonist binding sites relative to 125 I-ZM421385 antagonist binding sites in striatal membranes from both genotypes. By comparison to their wild-type littermates, the fraction of the A 2A R pool that was coupled to G-protein was markedly reduced in striatal membranes from Gng7 Ϫ/Ϫ mice (p Ͻ 0.001) (Fig. 2D). Collectively, these results confirm that A 2A R signaling shows a specific requirement for G-protein ␥ 7 expression and that its loss is associated with an impaired ability of this receptor to couple to G-protein.
Impaired Assembly of G olf Heterotrimer in Mice Lacking the ␥ 7 Protein-One mechanism that could account for the observed defects in both G-protein coupling and adenylyl cyclase activation is a coordinate reduction in the cellular amount of the G-protein ␣ s or ␣ olf subunit. These two structurally related isoforms are both able to stimulate cAMP production (12) and are both expressed in the striatum (32). Previously, we showed that loss of the ␥ 7 subunit coordinately reduced levels of the ␣ olf protein in the striatum (6). To confirm and extend this finding, we performed immunoblot analysis on micropunch samples from dorsal striatum (caudate) and ventral striatum (nucleus accumbens) of Gng7 Ϫ/Ϫ mice on two different genetic backgrounds (i.e. C57BL/6 and BALB/c). By comparison to their wild-type littermates, ␣ olf protein levels were strikingly reduced by Ͼ85% in both dorsal and ventral striatal membranes from knock-out animals on a C57BL/6 genetic background (Fig. 3A). The effect was remarkably specific in that ␣ s protein levels (i.e. 45-and 52-kDA forms) were not affected in the dorsal striatum and were reduced by only 20% in the ventral striatum of knock-out mice (Fig. 3B). Attesting to the generality of this finding, immunoblot analysis of the corresponding regions of knock-out mice on a BALB/c genetic background yielded similar results (supplemental Fig. S1). Taken together, these results demonstrate that the cellular level of the ␣ olf but not the ␣ s subunit is dependent on expression of the ␥ 7 subunit.
Next, we investigated how loss of the ␥ 7 protein impacts the levels of particular ␤ protein(s) in the striatum. Because targeting of the ␤␥ dimer to the plasma membrane is dependent upon post-translational lipid modifications of the ␥ subunit (33,34,35), we reasoned that loss of the ␥ 7 protein could affect the level of a specific ␤ subtype in striatal membranes from Gng7 Ϫ/Ϫ mice. By comparison to their wild-type littermates, ␤ 2 protein levels were selectively reduced by 31% with no significant changes in the amounts of ␤ 1 and ␤ 4 proteins (Fig. 3C). Taken together, these findings show both coordinate and selective suppression of the ␣ olf , ␤ 2 , and ␥ 7 subunits at the protein level.
To assess whether the ␣ olf subunit plays a reciprocal role in this process, we used the Gnal Ϫ/Ϫ mouse model (13) to determine whether loss of the ␣ olf protein causes a corresponding suppression of the ␥ 7 protein. By comparison to their wild type littermates, ␥ 7 protein levels were not significantly different in striatal membranes from Gnal Ϫ/Ϫ mice (Fig. 4), indicating that the expression of the ␥ 7 but not the ␣ olf protein drives the assembly of a specific G-protein heterotrimer in the striatum.
Mechanism for Coordinate Suppression of the G-protein ␣ olf Subunit in Mice Lacking the ␥ 7 Subunit-One mechanism that could account for suppression of the ␣ olf protein is a reduced level of the corresponding mRNA transcript. To test this possibility, we performed real time RT-PCR analysis on striatal tissue from Gng7 Ϫ/Ϫ mice and their wild-type littermates on the C57BL/6 background. Despite the loss of ␣ olf protein (Fig. 4), the level of ␣ olf mRNA was not significantly reduced in the striatum of Gng7 Ϫ/Ϫ mice (Fig. 5A). Moreover, the level of ␤ 2 mRNA was not significantly different in wild type and Gng7 Ϫ/Ϫ striatum. Finally, using a well validated marker of medium spiny neurons, the level of DARPP-32 (Ppp1r1b) mRNA was not significantly reduced in the striatum of Gng7 Ϫ/Ϫ mice (Fig. 5A), indicating that there was no overt loss of medium spiny neurons from the brains of knockout mice. These findings point to a post-transcriptional mechanism responsible for the coordinate suppression of the ␣ olf and ␤ 2 proteins in mice lacking the ␥ 7 protein.
Because plasma membrane binding of the ␣ subunit is facilitated by ␤␥ association (36, 37), we examined whether the decreased amount of ␣ olf protein in the striatal membrane fractions was associated with increased accumulation of this protein in the corresponding cytosolic fractions. Despite loss from the membrane fraction (Fig. 5B), no ␣ olf protein was detectable in the cytosolic fraction of Gng7 Ϫ/Ϫ mice (Fig. 5C).
Basis for ␥ 7 Selectivity-To determine why ␥ 7 has a special place in the assembly of the G olf heterotrimer in the native context, we performed immunoblot analysis to provide a quantitative accounting of the G-protein subunits that are expressed in the striatum of wild-type and Gng7 Ϫ/Ϫ mice ( Table 1). The wild-type striatum contains roughly equimolar levels of ␣ olf and ␥ 7 proteins (supplemental Fig. S2A and Table 1). Therefore, in the Gng7 Ϫ/Ϫ striatum, the 85% reduction in ␣ olf could be almost completely accounted for by loss of ␥ 7 . Likewise, the wild-type striatum contains a Ͼ2-fold molar excess of ␤ 2 relative to ␥ 7 (supplemental Fig. S2A and Table 1). Therefore, in the knock-out striatum, the 30% reduction in ␤ 2 could also be attributed to loss of ␥ 7 . While our accounting is necessarily limited by the availability of G-protein subunit antibodies and standards, these results strongly suggest the existence of a specific G-protein ␣ olf ␤ 2 ␥ 7 complex that normally functions downstream of the A 2A R in SP neurons. To explore the mech-  anistic role of the ␥ 7 protein in this process, we examined the relative levels of the various ␥ subtypes in the striatum. Compared with the ␥ 7 protein, there is a 4-fold molar excess of the ␥ 2 protein and an equimolar amount of the ␥ 3 protein in the striatum (7). Moreover, in the knock-out striatum, there is a further increase in the level of the ␥ 3 protein (supplemental Fig. S2B). Hence, the combined levels of the ␥ 2 and ␥ 3 proteins appear to be sufficient to compensate for loss of the ␥ 7 protein, suggesting other ␥ subtypes are not functionally interchangeable.
Alterations in Locomotor Activity in Mice Lacking ␥ 7 -To assess the functional consequences of impaired A 2A R signaling and G olf assembly, we examined the behavioral responses of Gng7 Ϫ/Ϫ mice. Because the A 2A R is primarily responsible for the psychostimulant actions of caffeine (21,22), we reasoned that Gng7 Ϫ/Ϫ mice on the C57BL/6 background might exhibit an attenuated response to the locomotor enhancing effects of this drug. As a quantitative measure of locomotor activity, Gng7 Ϫ/Ϫ mice and their wild-type littermates were placed in CLAMS cages equipped with photobeams to record their movements. After allowing sufficient time for acclimation to their surroundings, the locomotor responses for both groups of mice were calculated as the difference between the response to caffeine versus saline injection. Although showing similar responses to low dose caffeine (Fig. 6A), Gng7 Ϫ/Ϫ mice showed a significantly attenuated response to high dose caffeine that was particularly apparent at later time points (Fig. 6B; MANOVA analysis: F 1,30 ϭ 7.4, p ϭ 0.01). Because locomotor behavior is influenced by genetic background (38), these studies were repeated on Gng7 Ϫ/Ϫ mice on the BALB/c background to assess the generality of this finding. Even more strikingly, Gng7 Ϫ/Ϫ mice on the BALB/c background exhibited significantly reduced responses to both low and high dose caffeine (Fig. 6, C and D; MANOVA analyses: for the low dose, F 1,24 ϭ 16.4, p ϭ 0.0005 and for the high dose, F 1,24 ϭ 8.6, p ϭ 0.007). Taken together, these results show a clear association between defective A 2A R signaling and impaired locomotor response to caffeine in Gng7 Ϫ/Ϫ mice.
In addition to psychostimulant response, we also measured the basal locomotor activity that reflects the coordinated output from the SN and SP tracts (10,11) in these animals. Previously, we showed that the basal locomotor activity of Gng7 Ϫ/Ϫ mice was not impaired despite reduced striatal D 1 R signaling (6). Because increased D 1 R signaling is generally associated with a higher level of locomotor activity (39), this finding was somewhat surprising. To exclude the possibility that the background strain was obscuring any defect due to loss of ␥ 7 , we compared the basal locomotor activities of Gng7 Ϫ/Ϫ mice and their wild-type littermates on two different genetic backgrounds. On the C57BL/6 background, both wild-type and knock-out mice displayed comparably high levels of locomotor activity when introduced into CLAMS cages (Fig. 7A) and similar abilities to acclimate to the novel environment (Fig. 7B). That is to say, the initially high levels of locomotor activity observed in both groups decreased similarly over the 3 h following initial exposure to the CLAMS cages (Fig. 7A) and over the course of 4 days upon repeated exposure to the CLAMS cages (Fig. 7B). On the BALB/c background, Gng7 Ϫ/Ϫ mice exhibited  a significantly higher locomotor activity than their wild-type littermates when introduced into CLAMS cages ( Fig. 7C; MANOVA analysis: F 1,38 ϭ 11.3, p ϭ 0.002) but showed no differences in their abilities to acclimate to the novel environment (Fig. 7, C and D). Collectively, these findings support the possibility that combinatorial disruption of both D 1 R and A 2A R signaling pathways secondary to loss of the ␥ 7 protein produces overtly normal motor activity.

DISCUSSION
Despite their molecular cloning more than a decade ago (2), the functional significance for the large diversity of G-protein ␥ subunits is still not known. From analysis of Gng7 Ϫ/Ϫ mice, we now show that loss of the G-protein ␥ 7 subtype produces both biochemical and behavioral consequences. In the process of studying these effects, we also identify a fundamental role for the ␥ 7 subtype in driving the preferential assembly of a G-protein ␣ olf ␤ 2 ␥ 7 heterotrimer that is required for A 2A R signaling and its locomotor inhibitory effect in the striatum. These results support a growing body of data pointing to the effectiveness of A 2A R blockade to better normalize motor activity in Parkinson patients (40,41).
G olf Assembly Is Specifically Regulated by the ␥ 7 Subtype-The functions of G-protein ␣␤␥ heterotrimers are dependent on their proper assembly and trafficking to the plasma membrane (37,42). However, very little is known regarding which ␣␤␥ heterotrimers exist in the intact cell setting and how they are actually assembled. Previously, our analysis of Gng7 Ϫ/Ϫ mice provided the first in vivo demonstration that loss of the ␥ 7 protein disrupts the assembly of the G olf but not the G s heterotrimer in the striatum (6). Now, we extend this finding by showing through reciprocal analysis of Gnal Ϫ/Ϫ mice that loss of the ␣ olf protein does not substantially impact the level of the ␥ 7 protein. Taken together, these results point  to a hierarchical order of G olf formation that begins with the ␥ 7 subunit. In hindsight, the ␥ 7 subtype has several characteristics consistent with a primary role in this process. Spanning Ͼ66 kb in size, the mouse Gng7 gene produces multiple mRNA transcripts that encode the same protein, suggesting complex regulation of its expression. Moreover, the mouse ␥ 7 protein encompasses only 69 amino acids that assume an ␣-helical structure in solution (43) that approximates that seen in the crystal structure (44). This suggests that unlike its ␣ (45) and ␤ (46,47) partners, chaperone-type proteins may not be required for proper folding of the ␥ 7 protein. Finally, indicating that a "␥ subunit first" hierarchy may be generally applicable to other G-protein ␣␤␥ heterotrimers, loss of the ␥ t1 subtype has recently been shown to disrupt assembly of a specific G-protein ␣ t1 ␤ 1 ␥ t1 heterotrimer in the retina (48).
At this time, we can only hypothesize as to the mechanism by which the ␥ 7 protein sets the level of the G olf heterotrimer. Because the corresponding mRNAs were not altered, the simplest explanation for coordinately reduced ␣ olf and ␤ 2 protein levels is a post-transcriptional requirement for the ␥ 7 in the stabilization and/or trafficking of G olf to the plasma membrane. Because loss of ␥ 7 protein was not associated with accumulation of unassembled ␣ olf protein in the cytosol (Fig. 5), we speculate that formation of this G olf results from stabilization of the ␣ olf and ␤ 2 subunits by the ␥ 7 subunit, and that in the absence of this component, the unassembled ␣ olf and ␤ 2 subunits undergo active degradation.
Likewise, at this stage, we can only speculate as to the basis for the unique requirement for ␥ 7 subtype in this process. Refuting the long-standing dogma that most ␥ subunits are functionally interchangeable (reviewed in Ref. 2), the ␥ 7 subtype must possess unique features that cannot be replaced by the other ␥ subtypes. To begin to identify such features, we examined the possibility that the reportedly high abundance of the ␥ 7 subtype in the striatum (49,50) could provide an explanation. In fact, the ␥ 2 and ␥ 3 proteins are present together in 4-fold molar excess to the ␥ 7 subtype (supplemental Fig. S2 and Table 1). Therefore, abundance does not seem to be the answer. Next, we considered the possibility that the ␥ subtypes might be sequestered between different neuronal types making up the striatum. Two recent studies (30,31) comparing the translational mRNA profiles of SN and SP neurons reveal that both cell types express multiple ␥ forms (supplemental Fig. S3). Therefore, cell type specific expression does not explain the failure of other ␥ subtypes to substitute for the role of ␥ 7 in assembly process. Finally, we are contemplating the possibility that the ␥ 7 subtype could be localized within a particular subcellular compartment in neurons, similar to that shown for the ␥ 5 subtype in focal adhesions (51). Such compartmentation could result from a structural feature that is unique to either the ␥ 7 mRNA or protein. In this regard, it is notable that the ␥ 7 mRNA contains a very long 3Ј-UTR, a region that has been implicated in translational regulation and subcellular targeting of dendritically targeted proteins (52).
A 2A R Signaling Is Dependent on a Specific G-protein ␣ olf ␤ 2 ␥ 7 Heterotrimer-The A 2A R activates adenylyl cyclase activity in SP neurons (27,12,53,54). Consistent with a requirement for the ␣ olf ␤ 2 ␥ 7 heterotrimer in this pathway, both A 2A R-G-pro-tein coupling and adenylyl cyclase activation are markedly reduced in knock-out membranes (Figs. 2 and 1). In looking for a functional connection, it is tempting to compare the Ͼ75% reduction in A 2A R-G-protein coupling with the 85% reduction in ␣ olf protein ( Fig. 3 and supplemental Fig. S1). Because the A 2A R number (Fig. 2, as determined by antagonist binding) is several orders of magnitude lower than the total G olf content ( Table 1, as determined by quantitative immunoblotting), the random collision coupling model (55) would predict that the receptor represents the rate-limiting step for adenylyl cyclase activation. However, such a model is incompatible with several published reports. Notably, only about 20% of the A 2A R is actually coupled to G-protein in platelets (56) and striatum (14,57). Furthermore, in these tissues, the G-protein appears to control the rate of adenylyl cyclase activation (58). Thus, despite the apparent excess of G olf , it appears that the A 2A R interacts with only a limited pool of G olf and that the latter controls the cellular response.
Although not yet identified, the cellular mechanisms responsible for limiting their interaction could include post-translational modification and/or subcellular localization. In HEK293 cells, A 2A R coupling to G s and activation of adenylyl cyclase has been shown to occur within cholesterol-rich microdomains (59). In future studies, it will be interesting to investigate the functional significance for the poor coupling between the A 2A R and G olf in the striatum and to elucidate the underlying mechanism. Finally, in an analogous fashion, the D 1 R activates adenylyl cyclase activity in SN neurons (12). Because this receptor utilizes the same G-protein ␣ olf ␤ 2 ␥ 7 heterotrimer (6), it will be interesting to explore whether poor coupling of the D 1 R is also observed in the striatum and whether the G olf represents the rate-limiting step in adenylyl cyclase activation. Supporting the latter possibility, our previous analyses of mutant Drda1a and Gnal mice have revealed that the G-protein rather than the receptor controls psychostimulant responses (60).
Taken in conjunction with our earlier studies, the G-protein ␥ 7 subtype seems to play a special role in adenylyl cyclase stimulation in various cellular contexts. In HEK293 cells, the ␥ 7 subtype is responsible for driving the assembly of a particular G s heterotrimer required for ␤ adrenergic and D 1 R signaling but not for PGE 2 and D 5 R signaling (61,62); and in the physiological relevant context of the brain, it is responsible for driving the assembly of a specific G olf heterotrimer required for D 1 R signaling in SN neurons (6) and for A 2A R signaling in SP neurons ( Figs. 1 and 2). In addition to its role in assembly, it remains to be determined whether the ␥ 7 subunit performs additional roles in the signal transduction process. In this regard, the ␤␥ dimer has been suggested to contribute to recognition of the upstream receptor (63,15), as well as downstream regulation of effectors such as the striatal-enriched adenylyl cyclase type 5 isoform (64,65). Addressing whether the ␥ 7 subunit affects these processes will require the use of complementary approaches that bypass the requirement for the ␥ 7 subtype in the assembly of G olf .
Altered Caffeine Responsiveness Is Associated with Impaired A 2A R Signaling-Previous studies of the locomotor stimulating effect of caffeine have demonstrated the primary involvement of the A 2A R (66,21). Supporting its function acting down-stream of this receptor, Gng7 Ϫ/Ϫ mice lacking the ␥ 7 subtype showed an attenuated response to caffeine that was particularly apparent on the BALB/c background (Fig. 7). Moreover, consistent with their functioning as components of the same heterotrimeric G-protein, Gnal Ϫ/Ϫ mice lacking the ␣ olf subtype also exhibited a reduced response to caffeine (16). Finally, in agreement with a role for G olf in adenylyl cyclase stimulation and protein kinase activation, Ppp1r1b Ϫ/Ϫ mice lacking DARPP-32, which is phosphorylated by protein kinase A (67), also showed an impaired response to caffeine (68). Taken together, these results establish the importance of the A 2A R-G␣ olf ␤ 2 ␥ 7 -AC5-PKA-DARPP-32 pathway in mediating the psychostimulant properties of caffeine.
Clinical Significance-Despite defects in both D 1 R and A 2A R signaling, deficiency of the G-protein ␥ 7 subtype has little impact on basal locomotor behavior (Fig. 7). This finding is similar to reported findings for Drd1a Ϫ/Ϫ and Ador2a Ϫ/Ϫ mice, i.e. slightly increased or decreased locomotor activity (69,21). In contrast, Drd2 Ϫ/Ϫ mice display a markedly reduced locomotor activity (70 -72). In this regard, the ability of A 2A R signaling to modulate D 2 R signaling could prove beneficial for the treatment of Parkinson disease. For instance, several groups have observed attenuated stimulation of locomotor activity in response to caffeine in Drd2 Ϫ/Ϫ mice (73), and a partial reversal of acute D 2 R antagonist-induced catalepsy in Ador2a Ϫ/Ϫ mice (66). Such results form the basis for clinical trials of adenosine antagonists in the treatment of Parkinson disease (26). In these trials, Istradefylline has been shown to reduce symptoms in patients with Parkinson disease on levodopa therapy (40,41). Our studies further support targeting of the A 2A R signaling pathway as an augmentative strategy for treating Parkinson disease and provide a better mechanistic understanding of the apparent effectiveness of this treatment.