Activation of (cid:1) -Opioid Receptors Transfers Control of G (cid:2) Subunits to the Regulator of G-protein Signaling RGS9-2 ROLE IN RECEPTOR DESENSITIZATION*

In mouse periaqueductal gray matter (PAG) membranes, the (cid:1) -opioid receptor (MOR) coprecipitated the (cid:2) -subunits of the G i/o/z/q/11 proteins, the G (cid:3) 1/2 subunits, and the regulator of G-protein signaling RGS9-2 and its partner protein G (cid:3) 5 . RGS7 and RGS11 present in this neural structure showed no association with MOR. In vivo intracerebroventricular injection of morphine did not alter MOR immunoreactivity, but 30 min and 3 h after administration, the coprecipitation of G (cid:2) subunits with MORs was reduced by up to 50%. Furthermore, the association between G (cid:2) subunits and RGS9-2 proteins was increased. Twenty-four hours after receiving intracerebroventricular morphine, the G (cid:2) subunits left the RGS9-2 proteins and re-associated with the MORs. However, doses of the opioid able to induce tolerance promoted the stable transfer of G (cid:2) subunits to the RGS9-2 control. This was accompanied by Ser phosphorylation of RGS9-2 proteins, which increased their coprecipitation with 14-3-3 proteins. In the PAG membranes

The -opioid receptor (MOR), 1 which belongs to the family of seven-transmembrane G-protein-coupled receptors (GPCRs), is a heavily N-glycosylated protein (1, 2) that regulates G i/o/z/q/11 proteins (Ref. 3 and references therein) through a direct interaction (4,5). In the central nervous system, MORs play an important role in the antinociceptive action of opioids, but they become desensitized after repeated administration. Such tolerance is observed even after a single dose of an opioid agonist and can persist for ϳ3 days (6,7). This common characteristic of opioids that act via MORs is a serious drawback with respect to their long-term use as analgesics for the treatment of chronic pain. The phosphorylation of intracellular residues, followed by internalization, is the most widely accepted mechanism involved in GPCR desensitization. Thus, upon agonist challenge and the release of activated G␣-GTP subunits, GPCRs undergo phosphorylation by G␤-bound G-protein-coupled receptor kinases (GRKs), followed by arrestin binding (8). This has been amply documented for the -, ␦-, and -opioid receptors in in vitro cell expression systems (for example, see Ref. 9).
The desensitization and internalization of MOR are agonistdependent processes. In contrast to what has been observed for opioids such as etorphine and [D-Ala 2 ,N-Me-Phe 4 ,Gly 5 -ol]enkephalin, a number of reports describe the ability of morphine to activate the MAPK (mitogen-activated protein kinase) pathway and to desensitize the MORs without causing receptor phosphorylation and ␤-arrestin/dynamin-dependent internalization (10 -15). It is therefore possible that receptor downregulation is promoted only by agonists with high binding affinity for these receptors (16). Indeed, tightly bound agonists increase the probability that GRKs will inactivate the GPCRs since these kinases act only on agonist-occupied receptors (for example, see Ref. 17). The in vivo attained desensitization of MORs can be influenced by agents targeted to the specific RGS (regulator of G-protein signaling) proteins belonging to the R7 subfamily. The mammalian RGS proteins that act as GTPaseactivating proteins (GAP) for G␣-GTP subunits are grouped into five subfamilies according to structural and genetic similarities: Rz, R4, R7, R12, and RA (18). The members of the RGS-R7 subfamily in the central nervous system (RGS6, RGS7, RGS9-2, and RGS11) associate mostly with cell membranes. Their sequences contain the GGL (G-protein ␥-subunit-like) domain that binds to the G␤ 5 protein, but not to the other G␤ subunits (19,20). In nervous tissues, the RGS-R7 and G␤ 5 proteins are always found as dimers, indicating that this association is required for their GAP function on the corresponding G␣-GTP subunits. The RGS-R7 proteins show negative regulatory activity on the intensity of signals originating at MORs; thus, they participate in the development of tolerance to ago-nist effects (21,22). RGS9-2 is particularly important with respect to this function because its impairment provokes an increase in the potency of -opioid agonists and prevents (or at least delays) the appearance of MOR desensitization (7,21,23).
The RGS-R7 proteins display stronger affinity for the G␣ i/o/z subunits in their transition state, when the G␣ subunit initiates the spontaneous metabolism of GTP into the effectorinactive GDP form, than for the G␣-GTP form (24). Because RGS-R7 proteins can efficiently activate only the GTPase in G␣ o subunits (25), it has been proposed that, upon receptor activation, the subsequent binding of G␣ i/z subunits to RGS9-2 proteins serves to control the intensity of agonist signaling by reducing the pool of receptor-regulated G-proteins (22). In this scenario, a fraction of the morphine-activated G␣ i/z subunits is retained by the RGS9-2 proteins, leading to MOR desensitization. To determine whether this is the case, we analyzed the influence of RGS9-2 proteins on the changes that morphine induces in MOR regulation of G-proteins. The study was performed using membranes isolated from periaqueductal gray matter (PAG), a neural structure that plays a major role in mediating the effects of opioids when administered by the intracerebroventricular route (26). Morphine was found to induce a dose-dependent transfer of G␣ subunits from the MORs to the RGS9-2 proteins that correlates with the attenuation of the morphine-induced activation of G␣ subunits and its antinociceptive effects.

EXPERIMENTAL PROCEDURES
Preparation of Membranes from Mouse PAG-Male albino CD-1 mice (Charles River Laboratories Españ a, S. A., Barcelona, Spain) weighing 22-25 g were killed by cervical dislocation, and the PAG was removed. About 1 mm of tissue around the aqueduct was taken from 2-mm-thick coronal sections (MP-600 micropunch, Activational Systems Inc.). The structures from 20 mice (for direct analysis) or from six mice (for immunoprecipitation studies) were washed and pooled in ice-cold 25 mM Tris-HCl (pH 7.7), 1 mM EGTA, and 0.32 M sucrose supplemented with a protease inhibitor mixture (catalog no. P8340, Sigma), a phosphatase inhibitor mixture (catalog no. P2850, Sigma), and H-89 (catalog no. B1427, Sigma). The PAG membranes were then obtained. The tissue was homogenized in a Polytron homogenizer (Model PT 10/35) for 15 s at setting 3. The homogenate was centrifuged (Sorvall RC5C centrifuge, SS-34 rotor) at 1000 ϫ g for 10 min to remove the nuclear fraction. After the pellet was discarded, the supernatant was centrifuged at 20,000 ϫ g for 20 min to obtain the crude synaptosomal pellet (P2). After two cycles of washing, followed by resuspension in buffer and centrifugation, the final pellet was diluted in Tris buffer and analyzed as described below.
Co-immunoprecipitation of Signaling Proteins-Affinity-purified IgGs raised against MORs (2, 3) and RGS9-2 proteins (catalog nos. sc-8142 and sc-8143, respectively; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were labeled with biotin (catalog no. B1022, Sigma) following the manufacturer's instructions. MORs and RGS9-2 proteins were immunoprecipitated from the solubilized P2 fraction of mouse PAG as described (22) with minor modifications. Negative controls were performed with IgGs heated for 10 min at 100°C or pre-absorbed with 0.1 mg of antigenic peptide for 1 h at room temperature. Pilot assays were performed to adjust the amount of IgGs and sample protein and to determine the incubation period required to precipitate the desired protein in a single run. Thus, in any second precipitation, only a remnant of the immunosignal would be observed. The procedure required ϳ0.8 mg of protein from sonicated (two cycles of 5 s each) PAG membranes in a volume of 400 l of ice-cold buffer containing 50 mM Tris-HCl (pH 7.7), 50 mM NaCl, 1% Nonidet P-40, 50 l of protease and phosphatase inhibitor mixtures, and H-89. Membranes were solubilized overnight at 4°C and centrifuged at 10,000 ϫ g for 10 min. The supernatant was then cleared with 20 l of streptavidin-agarose (catalog no. S1638, Sigma) pre-equilibrated for 1 h at 4°C, followed by centrifugation at 3000 ϫ g for 5 min. The solubilized material was incubated overnight at 4°C with 10 l (3 g) of affinitypurified biotinylated IgGs raised against either MORs or RGS9-2 proteins. Fifty microliters of streptavidin-agarose were added, and incubation was continued for an additional 90 min at 4°C. The samples were then centrifuged at 3000 ϫ g for 5 min, and the supernatant was removed. The agarose pellets were subjected to five cycles of washing, followed by centrifugation and resuspension in 1 ml of Nonidet P-40 buffer. At the end of this process, the agarose pellets were heated in 300 l of 40 mM Tris-HCl and 1% SDS for 10 min at 100°C to detach and denature the proteins contained in the immunocomplexes. The mixture was then cooled to room temperature, and the streptavidin-agarose was separated in centrifugal filter devices with a 0.45-m pore (Ultrafree-MC, catalog no. UFC30GV, Millipore Iberica S. A.). To prevent interference in the visualization of Western blots, the biotinylated IgGs detached during the initial heating were selectively removed by the addition of octyl thioglucoside at a final percentage of 0.65% in 400 l plus 30 l of fresh streptavidinagarose. After 2 h at 4°C, the samples were centrifuged for 5 min at 10,000 ϫ g, and the streptavidin-agarose with the attached biotinylated IgGs was discarded. The proteins in the soluble fraction were concentrated in centrifugal filter devices (10,000-Da nominal molecular mass limit; Amicon Microcon YM-10, catalog no. 42407, Millipore Iberica S. A.). The proteins were solubilized in 2ϫ Laemmli buffer with mercaptoethanol by heating at 100°C for 3 min and then left to cool to room temperature before resolving by SDS-PAGE (10 -16% total acrylamide concentration, 2.6% bisacrylamide cross-linker concentration). The above procedure supplied enough protein to load four to six gel lanes. The proteins were transferred to polyvinylidene difluoride membranes (Bio-Rad) for Western blot analysis and incubated with the antibodies in Hoefer TM Deca-Probe chambers (catalog no. PR150, Amersham Biosciences, Barcelona).
Detection of Signaling Proteins in Mouse PAG Membranes by Electrophoresis and Immunoblotting-The proteins from P2 membranes were resolved by SDS-PAGE on 8 ϫ 11 ϫ 1.5-cm gel slabs (10 -20%). For immunodetection, 40 -60 g of PAG protein/lane were typically used. The separated proteins were transferred to 0.2-m polyvinylidene difluoride membranes and probed with the following, previously characterized antibodies: affinity-purified IgGs to MOR (2, 3) diluted 1:1000; anti-G q/11 (QL, catalog no. NEI809, PerkinElmer Life Sciences), anti-G␣ o (GC/2, catalog no. NEI804, PerkinElmer Life Sciences), and anti-G␣ i3 (catalog no. 371729-Q, Calbiochem) antibodies (27) diluted 1:1000; anti-G␣ z , anti-G␣ i2 , and anti-G␣ i1 antibodies (28,29)  The antibodies were diluted in Tris-buffered saline and 0.05% Tween 20 and incubated with the transfer membranes at 6°C for 24 h. Primary antibodies used were horseradish peroxidase-conjugated goat anti-rabbit IgG (HϩL; catalog no. 170-6515, Bio-Rad), horseradish peroxidase-conjugated donkey anti-goat IgG (catalog no. sc-2020, Santa Cruz Biotechnology, Inc.), horseradish peroxidase-conjugated goat antimouse IgM (HϩL; catalog no. 401225, Calbiochem), and horseradish peroxidase-conjugated goat anti-mouse IgG (HϩL; catalog no. 170-6515, Bio-Rad). Secondary antibodies were diluted 1:5000 in Trisbuffered saline and 0.05% Tween 20, incubated for 3 h, and revealed with the ECL Plus Western blotting detection system (catalog no. RPN2132, Amersham Biosciences). Chemiluminescence was visualized with a Peltier cooled CCD camera (Ϫ35°C, high signal-to-noise ratio, dynamic range up to 3.400) using a ChemiImager IS-5500 system (Alpha Innotech Corp., San Leandro, California). This system provides a real-time (30 frames/s) readout that can be analyzed by densitometry (Quantity One, Bio-Rad). Immunoprecipitation assays were performed on solubilized proteins typically obtained by pooling the PAG structures of six mice that had received identical morphine treatment and that were killed at the same post-opioid interval. These assays were repeated at least three times with PAG samples obtained from independent groups of mice. Equal loading was verified; and if necessary, the signal was adjusted using that obtained when probing the directly immunoprecipitated protein (MOR or RGS9-2). This was performed in re-blotting assays. However, if signal interference was observed, parallel blots generated with the same immunoprecipitated samples were used to make the necessary adjustment. Densitometries are expressed as the means Ϯ S.E. of the integrated volume (average optical density of the pixels within the object area/mm 2 ). All calculations were made using the Quantity One software. The statistical analysis of the results included analysis of variance (ANOVA), followed by the Student-Newman-Keuls test (SigmaStat, SPSS Science Software GmbH, Erkrath, Germany). Significance was set at p Ͻ 0.05. GTPase Assays-The PAG membranes were analyzed for GTPase activity by measuring [␥-32 P]GTP (6000 Ci/mmol; catalog no. NEG004Z; PerkinElmer Life Sciences) hydrolysis at 37°C for 20 min as described previously (29). The assay medium contained (final concentrations) 0.4 nM [␥-32 P]GTP (4 ϫ 10 5 cpm), 0.1 mM EDTA, 2 mM dithiothreitol, 100 mM NaCl, 5 mM MgCl 2 , 1 mM ATP, 10 mM creatine phosphate, 2.5 units of creatine phosphokinase, 0.25 mM App(NH)p, 1 mM ouabain, 10 mM Tris-HCl (pH 7.4); and 3 g of protein in a final reaction volume of 100 l. Unlabeled GTP was added at 0.3 or 1 M.
A preliminary incubation in 10 mM Tris-HCl and 0.5 mM EGTA (pH 7.4) for 2 h at 4°C equilibrated the opioid agonist with MORs in the PAG membranes. The reaction was started by the addition of 20 l of the membranes plus morphine to 80 l of the prewarmed reaction mixture. This was allowed to continue for 20 min at 37°C. The reaction was terminated by the addition of 900 l of an ice-cold suspension of 5% (w/v) activated charcoal (Norit A) in 20 mM H 3 PO 4 (pH 2.3). After 15 min of centrifugation at 11,000 ϫ g, 500 l of the supernatant were removed, and 32 P i that was released by the GTP hydrolysis was determined by liquid scintillation counting (Ecolume, ICN) in a Beckman LS-6500 scintillation counter. Blanks (without membranes) accounted for 1-2% of the total radioactivity added. Data were corrected for the background and are expressed as percentage increase over the basal GTPase activity. High affinity GTPase activity was calculated by subtracting the radioactivity ( 32 P i ) liberated from [␥-32 P]GTP in the presence of 100 M GTP (low affinity GTPase activity) from that measured at lower concentrations, typically 0.3 or 1 M [␥-32 P]GTP. The experiments were repeated three times and performed in triplicate.
Animals, Intracerebroventricular Injection, and Evaluation of Antinociception-Male albino CD-1 mice weighing 22-25 g were housed at 22°C under a 12-h light/dark cycle (lights on from 8 a.m. to 8 p.m.). Food and water were provided ad libitum. Animals were lightly anesthetized with ether, and morphine sulfate (Merck, Darmstadt, Germany) in a volume of 4 l was injected into the lateral ventricle. The response of the animals to nociceptive stimuli was determined by the warm water (52°C) tail-flick test. Antinociception is expressed as a percentage of the maximum possible effect: maximum possible effect ϭ 100 ϫ (test latency Ϫ base-line latency)/(cutoff time (10 s) Ϫ base-line latency).
Production of Acute Morphine Tolerance-Tolerance to morphine can arise after the administration of a single dose (7,24). To promote this effect, the animals lightly anesthetized with ether received a single intracerebroventricular injection of either 3 or 10 nmol of morphine (priming dose). Controls were administered saline in the same manner. To determine whether the opioid produced the sought-after tolerance, a test dose of morphine identical to the priming dose was intracerebroventricularly injected 24 h later into all morphine-primed and control mice. Analgesia was then determined by the tail-flick test at the postmorphine interval corresponding to the peak effect of the opioid, i.e. at 30 min. The development of acute tolerance was determined by reduction in antinociceptive potency. Each treatment was performed on a different group of 10 or 15 mice. The data collected were examined by ANOVA, followed by the Student-Newman-Keuls test (SigmaStat). Significance was set at p Ͻ 0.05. For immunodetection analysis, the mice were injected with saline or the morphine priming dose and killed by decapitation at intervals of 30 min, 3 h, and 24 h to obtain the PAG synaptosomal membranes.
Down-regulation of RGS9-2 Expression-The 16-base end-capped phosphorothioate (indicated by *) antisense oligodeoxynucleotide (ODN) 5Ј-C*T*CGAATCAGTTCG*C*T-3Ј directed to nucleotides 1044 -1059 of the murine RGS9 mRNA expressed in the central nervous system (GenBank TM /EBI accession number AF125046) (21) was used to down-regulate RGS9-2 expression. As a control, an antisense ODN containing 5 mismatched bases (shown in boldface) was used: ODN-RGS9-2M, 5Ј-C*T*GCAATGAGTTGC*T*C-3Ј. These ODNs were injected into the lateral ventricles of animals lightly anesthetized with ether. Each series of injections was performed on a distinct group of mice according to the following 5-day schedule: 1 nmol on days 1 and 2, 2 nmol on days 3 and 4, and 3 nmol on day 5. The mice were killed on day 6, and PAG was obtained.

MOR-selective Linkage of RGS Proteins-In synaptosomal
preparations of mouse PAG, the antibodies against MORs precipitated glycosylated proteins of 55-65, 70 -80, and 100 -110 kDa. These proteins were recognized by antibodies directed against different epitopes of the receptor proteins. N-Glycosidase F increased the electrophoretic mobility of all these bands to the size of the protein predicted from the MOR amino acid sequence (2). Thus, the different sized moieties corresponded to degrees of glycosylation of MOR. No immunoprecipitation was seen when the anti-MOR IgGs were pre-absorbed with the antigenic peptides or when heat-inactivated IgGs were used (Fig. 1A).
The MORs coprecipitated various signaling elements: G␣ i/o/ z/q/11 subunits, G␤ 1/2 subunits, and RGS9-2 and its associated G␤ 5 protein (Fig. 1B). Notably, other members of the RGS-R7 subfamily present in PAG synaptosomal fractions (RGS7 and RGS11) showed no association with this receptor (Fig. 1C). Although, in the mouse, the 77-kDa RGS9-2 protein is found in large amounts in striatal membranes, it is also present in other neural structures such as the cerebral cortex and PAG (7,23,30). The antibody directed against the RGS9-2 N-terminal sequence gave more intense signals than that directed against the C-terminal domain (Fig. 1C). The 675-residue RGS9-2 protein bears Ser 667 , Thr 657 , and Thr 658 in its C terminus, groups with high phosphorylation potential (NetPhos Version 2.0 Prediction Server, Center for Biological Sequence Analysis). Thus, region-specific differences in the post-translational regulation of this protein could account for the differences in RGS9-2 immunosignals obtained with the anti-C-terminal domain antibody in the cerebral cortex and PAG (Fig. 1C).

Effect of in Vivo Intracerebroventricular Injection of Morphine on the Association of MOR with G␣ Subunits in PAG-
The effect of the acute administration of morphine on the association of G signaling proteins with MORs was analyzed. Two different doses of morphine (3 and 10 nmol) were studied in the production of acute tolerance. These morphine priming doses were given to distinct groups of mice by the intracerebroventricular route; and at the time the effect peaked (after 30 min), they produced analgesic effects of ϳ50 and 80% of the maximum possible effect for this test (cutoff time of 10 s). At the 120-min interval, only the 10-nmol morphine dose produced a remnant 20% of the maximum possible effect (Fig. 2, left  panel); at 24 h, no analgesic activity was detected (data not shown). To determine whether the morphine priming doses produced tolerance, the effect of the same doses given 24 h later (test doses) was examined. Only the 10-nmol priming dose greatly reduced the response of the test dose (Fig. 2, right  panel). The 3-nmol morphine priming dose induced no tolerance to a test dose of 10 nmol (data not shown). This phenomenon, known as acute tolerance, appears within hours of agonist administration and lasts for 2-3 days. The threshold dose needed to produce this long-lasting tolerance to morphine is about three to four times greater than that required to produce detectable analgesia in this test (6,21,31).
The administration of the desensitizing dose of 10 nmol of morphine promoted no noticeable changes in the pattern of precipitated MORs when studied in PAG membranes from mice killed 24 h later (Fig. 3A), and the opioid did not alter the levels of the G␣ subunits present in the PAG membranes (data not shown). However, the number of G␣ subunits that coprecipitated with the MORs was reduced at the 30-min, 3-h, and 24-h post-opioid intervals. These reductions ranged from 40 to 60% for G␣ i/o/z subunits and somewhat less for G␣ q/11 subunits (ϳ20%) (Fig. 3B). 3 nmol of morphine also reduced the association of MORs with the G␣ subunits, but to a lesser extent. This effect was observed 30 min and 3 h after injecting the opioid, but not at 24 h (Fig. 3C). This dose of the opioid was devoid of desensitizing capacity 24 h after its intracerebroventricular injection (Fig. 2).  A, upper panel, lanes 1-3, immunoprecipitation was performed with affinity-purified IgGs directed against the N-terminal epitope. Blots were probed with IgGs directed against the second external (2nd-Ext) loop sequence on MOR. Lanes 4-6, the second external loop epitope was used to precipitate the immunocomplexes. The probe was directed against the Nterminal sequence. Specificity of precipitation was determined by boiling MOR-directed IgGs at 100°C for 10 min (lanes 1 and 6) or preincubating the IgGs with 0.1 mg of the corresponding antigenic peptide for 1 h at room temperature (lanes 2 and 5) before starting the immunoprecipitation procedure. Lower panel, the proteins coprecipitated with the MORs were assayed with antibodies to G␣/␤ subunits. B, immunoprecipitates were also probed with anti-RGS7, anti-RGS9-2, anti-RGS11, and anti-G␤ 5 antibodies. No G␣␤ or RGS proteins were coprecipitated with pre-absorbed or heat-inactivated anti-MOR IgGs. C, control PAG membranes were SDS-solubilized, PAGE-resolved, Western-blotted, and probed with antibodies to the RGS and G␤ 5 proteins. Inset, synaptosomal membranes from the cerebral cortex, PAG, and striatum were probed with antibodies directed against the Nterminal (N) and C-terminal (C) sequences of RGS9-2 proteins. Morphine Induces Ser Phosphorylation of RGS9-2 Proteins, and RGS9-2 Proteins Regulate the Association of G␣ Subunits with MORs-RGS3 and RGS7 interact with 14-3-3 proteins via a conserved phosphoserine residue in the RGS domain (32). This putative 14-3-3-binding motif, K(K/S)DpSY(P/A), is present in all RGS-R7 subfamily members and has a protein kinase C consensus sequence, (K/R)XXS (33). Residues in this motif and in the flanking regions establish critical contacts with the G␣-GDP⅐P (hydrolytic transition state of GTP) subunits (Fig.  4A). The RGS domain of RGS9 proteins also shows a consensus sequence for protein kinase A phosphorylation that includes Ser 427 and Ser 428 (34). The possible relevance of Ser phosphorylation in RGS9-2 function was therefore analyzed. In mouse PAG, RGS9-2 coprecipitates a series of signaling molecules, including the G␣ i2 , G␣ z , and G␣ o subunits; the G␤ 5 subunit that associates with RGS9-2 proteins; and the phosphoproteinbinding 14-3-3 proteins (22). Although the electrophoretic mobilities of these proteins are clearly different from that of RGS9-2, the possibility exists that unidentified phosphoproteins might coprecipitate with RGS9-2 and interfere with the phosphoserine analysis. We therefore sought to detach any accompanying proteins from the RGS9-2 proteins prior to performing immunoprecipitation. The PAG synaptosomal membranes were thus heated in 40 mM Tris-HCl and 1% SDS for 10 min at 100°C under reducing conditions. This mixture was then cooled to room temperature; the SDS concentration was reduced by adding octyl thioglucoside to a final percentage of 0.65%. Under these conditions, there is only a low probability that the denatured proteins with re-associate. The RGS9-2 immunoprecipitation was then performed with an antiserum directed against the N terminus. Of the four anti-Ser(P) antibodies studied, only clone 1C8 gave immunosignals on RGS9-2 proteins. This clone recognizes Ser(P) in sequences acted upon by protein kinases A and C. In PAG from control mice, this protein showed a certain level of Ser phosphorylation; after receiving the dose of morphine that produced acute tolerance (10 nmol), the phosphorylation was notably increased at 30 min (200%), 3 h (100%), and 24 h (50%) (Fig. 4B).
After the solubilization of PAG membranes under nonreducing conditions, the G␣ i2 , G␣ z , and 14-3-3 proteins coprecipitated with RGS9-2 proteins (Fig. 4C). The experimental conditions we used were aimed at maintaining all existing associations between RGS9-2 proteins and G␣ subunits. Therefore, solubilization of PAG membranes and immunoprecipitation were conducted in the absence of guanine nucleotides or agents that could force the G␣ subunits to copy the transition state. This situation is not directly comparable with that used to detect the binding of RGS domains or complete sequences to different classes of G␣ subunits in protein mixtures or when coupled to a gel matrix. Under these circumstances, the addition of GDP and AlF 4 is critical for the G␣ subunits to switch into the hydrolytic transition state and hence to promote their binding to the RGS domains (for example, see Refs. 35 and 36). In our assays, we pulled down physiological protein complexes formed before the animals were killed. In these complexes, it is probable that the endogenous guanine nucleotides determined whether the G␣ subunits associated with either MORs or RGS9-2 proteins. However, if RGS domains can associate only with the GTP-or GDP⅐P-bound forms of G␣, and were G␣-GTP to be metabolized into GDP during the immunoprecipitation procedure at 4°C, no association of RGS domains with G␣ would be observed. Since, at the end of the immunoprecipitation, we obtained RGS9-2⅐G␣ complexes, it is probable that a durable association was favored by the post-translational modification of these proteins or an interaction with third partner proteins.
Twenty-four hours after intracerebroventricular injection of 10 nmol of morphine, a tendency to increase the association between G␣ z subunits and RGS9-2 proteins was observed. The 14-3-3 proteins and G␣ i2 subunits showed statistically significant timedependent increases in their coprecipitation with RGS9-2 proteins. At the 30-min post-opioid interval, this was augmented by 140 and 100%, respectively; at 3 and 24 h, it diminished to ϳ60% in both cases (Fig. 4C). This pattern was paralleled by RGS9-2 Ser(P) signals. These were also attenuated at the later postmorphine intervals (Fig. 4B). The lower dose of 3 nmol of morphine produced a slight RGS9-2 phosphorylation plus a moderate association of G␣ and 14-3-3 proteins with these RGS9-2 proteins. This was observed at the 30-min and 3-h post-opioid intervals; but 24 h later, no changes were detectable (data not shown).

FIG. 3. Morphine reduces coprecipitation of G␣ subunits with
MORs. Groups of 6 -10 mice received a single intracerebroventricular (icv) injection of saline (Control) or 3 or 10 nmol of morphine and were killed at the time points indicated. The pooled PAG synaptosomal membranes for each post-opioid interval were solubilized with 1% Nonidet P-40 and incubated overnight at 4°C with affinity-purified biotinylated IgGs raised against the second external loop of the MORs. Immunocomplexes were precipitated with streptavidin-agarose; resolved by SDS-PAGE; and visualized in Western blots probed with antibodies directed against the N-terminal epitope of MOR (A) and anti-G␣ i1 , anti-G␣ i2 , anti-G␣ o , anti-G␣ z , and anti-G␣ q/11 antibodies (B and C). The experiment was repeated at least three times using solubilized PAG membranes from different groups of mice. Equal loading was verified; if necessary, the data were adjusted using the signals obtained by probing anti-MOR IgGs in parallel blots generated with the same immunoprecipitated samples. Densitometries are expressed as the means Ϯ S.E. of the integrated volume (average optical density of the pixels within the object area/mm 2 ). All calculations were made using Quantity One software. *, significantly different from the immunosignals of the control group corresponding to PAG membranes from mice that received saline instead of the opioid (ANOVA/Student-Newman-Keuls test, p Ͻ 0.05).
Notably, at the 24-h post-opioid interval, the 10-nmol (but not the 3-nmol) dose promoted significant reductions in the number of G␣ subunits that remained associated with the MORs (Fig. 3B). Therefore, the loss of G-proteins that coprecipitated with MORs brought about by high doses of morphine correlated with increases in their association with RGS9-2 proteins and also with MOR desensitization.
The influence of RGS9-2 proteins on the regulation of MOR signaling was then studied by reducing the expression of these RGS proteins in PAG. The efficacy and selectivity of the ODN treatment used have been previously assessed using antibodies directed against RGS9-2 (7). In the present work, after mouse PAG proteins were resolved by SDS-PAGE, the same antibodies detected reductions of some 50% of the 77-kDa RGS9-2 protein caused by the ODN treatment (Fig. 2). In mice with reduced PAG levels of RGS9-2 proteins, the antinociception promoted by 10 nmol of morphine increases, and notably, tolerance to the acute dose of morphine does not develop (7,(21)(22)(23). In the present work, we observed that RGS9-2 knockdown prevented MORs from becoming desensitized by 10 nmol of intracerebroventricular morphine (Fig. 2). This correlated with the MORs maintaining their association with the G-proteins in the PAG membranes from these morphine-treated RGS9-2 knockdown mice (Fig. 4D).
Influence of RGS9-2 on MOR-stimulated GTP␥S Binding and GTPase Activity in Mouse PAG Membranes-The above results suggest that RGS9-2 is involved in the processes that, upon agonist challenge, bring about MOR desensitization. We explored the extent and efficiency of MOR coupling to G-proteins in PAG membranes from both control and morphinetolerant mice with reduced levels of RGS9-2 proteins. Notably, in PAG membranes from RGS9-2 knockdown mice, the baseline level of [ 35  Western blots were probed with antibodies to Ser(P) (B); G␣ i2 , G␣ z , and 14-3-3 proteins (C); and G␣ i2 , G␣ z , and G␣ o (D). See the legend to Fig. 3 for details. Loading was verified with anti-RGS9-2 (B and C) and anti-MOR (D) antibodies. Heat inactivation of anti-RGS9-2 and anti-MOR antibodies produced no immunoprecipitation of these proteins or of any others potentially associated with them. *, significantly different from the immunosignals observed for the control group that received saline instead of 10 nmol of morphine (ANOVA/Student-Newman-Keuls test, p Ͻ 0.05). enous activators of [ 35 S]GTP␥S binding. In PAG membranes obtained 24 h after control mice were intracerebroventricularly injected with 10 nmol of morphine, the opioid produced a lesser increase in G␣-related [ 35 S]GTP␥S binding (Figs. 5 and 6) and also low K m GTPase activity (Fig. 7). This was observed under experimental conditions favoring the agonist activation of G i/o proteins (0.3 M GTP) or G z/q/11 proteins (1 M GTP) (29). These results agree with the reduction of MOR-associated G-proteins observed 24 h after using this dose of morphine (Fig. 3B). The knockdown of RGS9-2 proteins increased the ability of morphine to stimulate [ 35 S]GTP␥S binding to PAG membranes from both control mice and mice acutely tolerant to morphine (Figs. 5 and 6) and also stimulated GTPase activity in PAG membranes from tolerant mice to the levels attained in control animals (Fig. 7). These results agree with the idea that in vivo RGS9-2 knockdown prevents morphine from reducing the association of G-proteins with MORs in PAG.

DISCUSSION
MORs are transmembrane proteins of ϳ40 kDa that, after N-glycosylation, generate apparent sizes of 55-65, 75-85, and 100 -110 kDa. Since the immunoprecipitated MORs were processed under reducing conditions, the diversity of molecular species observed for the chromatographically resolved MORs probably reflects different degrees of glycosylation of a single protein. In fact, the deglycosylated MOR moiety is of a size consistent with that predicted from its amino acid sequence (2). In the PAG membranes, MOR coprecipitated a series of signaling elements, including G␣ and G␤ subunits and RGS9-2 proteins. RGS7 and RGS11 present in PAG membranes showed no association with this receptor. Alternative splicing of RGS9 yields the RGS9-1 and RGS9-2 isoforms. RGS9-1 is found in the retina, whereas RGS9-2 is present in the central nervous sys-tem, with the highest levels in the striatum (7,23,37). RGS9-2 forms heterodimers with the short form of G␤ 5 proteins and binds activated G␣ i/o/z subunits, although it has only a moderate GAP activity on G␣ o -GTP subunits (Refs. 21 and 22 and references therein).
Although there are reports describing agonist-selective MOR down-regulation in neurons, this has usually been achieved with doses of agonists greater than those that induce opioid desensitization. In the central nervous system, strong tolerance to morphine and related opioids such as heroin develops as a consequence of a reduced affinity of MORs for opioid agonists, but no loss of MORs is observed (this work and Refs. 38 -43). In fact, agonist-induced desensitization of -opioid receptors is achieved by phosphorylation, but without their internalization, and resensitization requires replacement by newly synthesized receptors rather than dephosphorylation (44). MOR substitution has also been proposed after morphine induces MOR desensitization in mouse brain (45). In recent years, elements of the MOR transduction system involved in the molecular mechanisms responsible for desensitization independent of receptor down-regulation have come to light. These include G-proteins (45), regulators of G␤␥ dimers such as the phosducin-like proteins (31), and the RGS proteins (this work and Refs. 7 and 21-23).
The in vivo administration of a single dose of 3 nmol of morphine caused reductions in MOR-associated G i/o/z proteins that reverted to control levels upon the extinction of the effects of the agonist. The recovery of the G-protein pool associated with MORs was not observed when a dose of 10 nmol of morphine (which promotes long-term MOR desensitization) was used instead. This phenomenon probably increases the population of uncoupled MORs and hence of those displaying low affinity for agonist binding. Because morphine mostly activates G i2 and G z proteins to produce MOR-mediated supraspinal antinociception (for example, see Refs. 3 and 46), RGS9-2 could favor the onset of desensitization at MORs by binding and sequestering agonist-activated G␣ subunits (7,21,22). The possibility of this transference of G␣ subunits upon morphine challenge has now been substantiated. The reductions in MORregulated G␣ subunits corresponded with increases in those bound to RGS9-2 proteins, in particular those of the G␣ i2 class. Accordingly, the acutely tolerant MORs also showed a reduced response to morphine, including poor stimulation of GTP␥S binding and/or low K m GTPase activity, both of which are related to the capacity of the agonist to activate G␣ subunits.
Knockdown or knockout of the RGS9-2 proteins provokes an increase in the potency of -opioid agonists and strongly reduces the extent of MOR desensitization (this work and Refs. 7, 21, and 23). This suggests that the knockdown of RGS9-2 proteins makes a larger number of G-proteins available to MOR regulation. Thus, in PAG membranes from these RGS9-2 knockdown control mice, morphine-stimulated GTP␥S binding increased. In the presence of 0.3 M GTP, a tendency to increase the morphine-activated GTPase activity was observed as well. Morphine doses that produced MOR tolerance in naïve mice failed to promote this phenomenon in RGS9-2 knockdown mice. Twenty-four hours after morphine injection, the in vitro assays performed in PAG membranes from these mice (G␣ coprecipitation with MORs, morphine-induced GTP␥S binding, and stimulation of GTPase activity) indicated that the MORs maintained a control-like association with the G-proteins.
Together, these observations indicate a role for RGS9-2 proteins in morphine-induced desensitization of MOR. The transfer of MOR-regulated G␣ subunits to the RGS9-2 control is insufficient for long-term desensitization; stabilization of the RGS9-2⅐G␣ complexes is required. Studies performed on the physiology and structure of the retinal short C-tailed RGS9-1 protein may help in understanding the processes that lead RGS9-2 to sequester receptor-activated G␣ subunits. In the retina, cGMP phosphodiesterase activity is inhibited by its ␥-subunit (P␥) (47). This negative regulation is removed by the binding of the activated GTP-bound ␣-subunit of transducin to P␥ (48,49). The N-terminal sequence of P␥ establishes contacts with determined residues in G␣ t , although this binding is not affected by the nucleotide bound to G␣ t . This is followed by interaction with a P␥ C-terminal region that recognizes the activated state of G␣ t -GTP (50,51). Notably, RGS9-1 recognizes only G␣ t/o subunits when they are bound to P␥ subunits. The G␤ 5 long splice variant that binds the GGL domain on RGS9-1 determines this selectivity (36). The RGS9-1⅐G␤5L complex establishes an initial contact with both the P␥ and G␣ t subunits through residues in the RGS9-1 RGS domain that are not directly involved in its GAP activity (51,52). P␥ and RGS9-1 bind to distinct non-overlapping G␣ t residues located on the GTPase pocket (53) and stabilize the G␣ t -GDP⅐P hydrolytic intermediate or transition state before it is bound by the GAP region of the RGS domain (51). Thus, to perform its GAP activity on activated G␣ t subunits, RGS9-1 first requires the interaction of G␣ t -GTP with P␥. It next requires cooperative binding with the G␣ t -GTP⅐P␥ complex to induce and stabilize the G␣ t -GDP⅐P transition state. Finally, the binding of the RGS GAP section to produce G␣ t -GDP must take place. This leads to the dissociation of the signaling elements G␣ t -GDP, RGS9-1⅐G␤5L, and P␥, re-establishing negative control over phosphodiesterase function. This sequence of events warrants that G␣ remains active until it reaches and regulates the effector target. Later, since certain residues in the RGS domain can bind to the effector (52), the RGS domain binds to the G␣⅐effector complex, inactivating the G␣ subunit.
The 18 C-terminal residues of RGS9-1 are replaced in the central nervous system RGS9-2 by a 209-amino acid prolinerich domain (37). The P␥ subunit that increases the affinity between RGS9-1 and activated G␣ t/o (36) is structurally similar FIG. 7. Effect of RGS9-2 down-regulation on morphine-induced stimulation of low K m GTPase activity in mouse PAG synaptosomal membranes. The mice received either active ODN-RGS9-2 or the mismatched ODN (Control) for 5 days. On day 6, each group was divided into two, and the members of each subgroup received intracerebroventricular injections of either 10 nmol of morphine or saline. All were killed 24 h later, and PAG membranes were obtained. The opioid agonist was preincubated for 2 h at 4°C with the PAG membranes. Hydrolysis of [␥-32 P]GTP (0.3 and 1 M) was measured after incubation of the PAG membranes for 20 min at 37°C. Nonspecific low affinity GTP-hydrolyzing activity was assessed in the presence of 100 M unlabeled GTP and was subtracted from the total activity to define high affinity GTPase activity. Stimulation is expressed as a percentage increase above base-line high affinity GTPase activity. The base-line high affinity GTPase activity in the absence of agonist in the incubation medium with 0.3 M GTP was as follows: mismatched ODN and saline, 98 Ϯ 6 pmol/mg of protein/min; mismatched ODN and acute dose of morphine, 126 Ϯ 8 pmol/mg of protein/min; ODN-RGS9-2 and saline, 136 Ϯ 9 pmol/mg of protein/min; and ODN-RGS9-2 and acute dose of morphine, 123 Ϯ 7 pmol/mg of protein/min. The GTPase activity in the absence of agonist in the incubation medium with 1 M GTP was as follows: mismatched ODN and saline, 245 Ϯ 13 pmol/mg of protein/ min; mismatched ODN and acute dose of morphine, 223 Ϯ 12 pmol/mg of protein/min; ODN-RGS9-2 and saline, 259 Ϯ 14 pmol/mg of protein/ min; and ODN-RGS9-2 and acute dose of morphine, 263 Ϯ 15 pmol/mg of protein/min. Values are means Ϯ S.E. of three to five independent determinations. *, significantly different from the group receiving no acute morphine in vivo (ANOVA/Newman-Keuls test, p Ͻ 0.05).
to the C-terminal domain of the long splice variant RGS9-2.
Then, RGS9-2 shows the capacity to bind receptor-activated G␣ proteins without the cooperation of their target effector (54). Given that which has been described for RGS9-1 and P␥, the binding of non-GAP regions of the RGS9-2 RGS domain to residues in G␣-GTP ought to induce and stabilize the hydrolytic transition state of G␣-GDP⅐P, and RGS GAP-related residues should produce the effector-inactive G␣-GDP form.
The GAP function of RGS proteins is reduced and even blocked by mechanisms that involve phosphorylation of specific residues inside or near the RGS box and binding to 14-3-3 proteins. RGS9-1 and RGS7 are acted upon by protein kinase C at Ser 400 (Ser 434 in RGS7) and by protein kinase A at Ser 425 and Ser 428 (Ser 459 and Ser 462 in RGS7) (32,33). RGS9-1/ RGS9-2 Ser 400 is located in the G␣-interacting portion of the RGS domain and is inside a 14-3-3-binding motif (32). Asp 399 and Arg 403 , which are critical for the RGS9 GAP mechanism, flank this Ser 400 (51). Protein kinase C-mediated phosphorylation of Ser 400 and the subsequent binding of 14-3-3 impede RGS GAP function (32). Therefore, when morphine produces in vivo signals of certain magnitude, RGS9-2 GAP activity would be switched off by the action of protein kinase C (and probably of protein kinase A as well) and by the subsequent binding of 14-3-3 to Ser 400 . The G␣ subunits that exhibit long-term association with RGS9-2 proteins could be those bound while the GAP activity is interrupted. It is also possible that GAPblocked RGS9-2 proteins retain their ability to bind activated G␣ subunits through domains not directly involved in GAP function, e.g. C-terminal and RGS non-GAP residues. This mechanism could account for morphine-induced desensitization of MOR since there is a parallel between reduced association of G␣ subunits with MORs, increases in RGS9-2 Ser phosphorylation, and coprecipitation of 14-3-3 proteins and G␣ subunits with RGS9-2 proteins (this work and Ref. 22). Reduced tolerance to morphine also develops in mice intracerebroventricularly treated with protein kinase A and C inhibitors (55,56).
The action of GPCR agonists increases the degradation rates of G␣ subunits (57,58). It is possible that the retention of G␣ subunits by RGS9-2 proteins precedes their degradation via the proteasome pathway. This seems to apply to the RGS-Rz subfamily, which binds to the dileucine-rich region of GIPN (GAIP-interacting protein N terminus), a putative ubiquitinprotein isopeptide ligase that links these RGS proteins with G␣ degradation (59). Since functional opioid tolerance produced by single doses lasts for several days, new synthesis of G␣ subunits would help to restore the control responses (6,31). There is increasing evidence that RGS proteins exhibit regulatory properties that cannot be attributed to their GAP activity. RGS proteins can inhibit signaling pathways by different mechanisms and produce effector antagonism. In this respect, the direct binding of RGS2, RGS3, RGS4, and GAIP to phospholipase C␤1 interferes with the regulatory action of activated G␣ q subunits (60, 61). It has been described that the direct association of RGS-R4 proteins with the intracellular domain of muscarinic acetylcholine receptor subtypes helps to deactivate or sequester G␣-GTP subunits before they regulate their corresponding effectors, leading to receptor desensitization (62). The RGS proteins that delay or inhibit G␣-GTP hydrolysis also reduce receptor-mediated activation of downstream effectors. Indeed, the RGS domain of GRK2 is known to bind to and sequester activated G␣ q subunits, blocking the regulation of phospholipase C␤ (63). Similarly, RGS-Rz proteins bind receptor-activated G␣ q subunits, but exhibit no GAP activity on them, and this interaction reduces G␣ q -mediated calcium mobilization (64). Thus, the sequestering of activated G␣ subunits by RGS9-2 proteins may best explain MOR desensitization promoted by single doses of morphine (this study and Ref. 22).
In summary, the number of MOR-associated G i/o/z proteins was reduced during the time course of morphine effects. This MOR/G-protein association recovered after the cessation of opioid effects. However, it was not recovered when morphine produced MOR desensitization independent of down-regulation. By retaining mostly MOR-activated G␣ i2 subunits, the RGS9-2 proteins play a key role in the onset of this phenomenon. Thus, effects of certain magnitude induce the transfer of MOR-associated G␣ subunits to RGS9-2 proteins and also bring about the phosphoserine-dependent binding of 14-3-3 that stabilizes their interaction. These processes could exclude or precede the loss of the agonist-activated MORs.