Paired Activation of Two Components within Muscarinic M 3 Receptor Dimers Is Required for Recruitment of (cid:1) -Arrestin-1 to the Plasma Membrane*

(cid:1) -Arrestins regulate the functioning of G protein-cou-pled receptors in a variety of cellular processes including receptor-mediated endocytosis and activation of signaling molecules such as ERK. A key event in these processes is the G protein-coupled receptor-mediated recruitment of (cid:1) -arrestins to the plasma membrane. However, despite extensive knowledge in this field, it is still disputable whether activation of signaling path-ways via (cid:1) -arrestin recruitment entails paired activation of receptor dimers. To address this question, we investigated the ability of different muscarinic receptor dimers to recruit (cid:1) -arrestin-1 using both co-immunopre-cipitation and fluorescence microscopy in COS-7 cells. Experimentally, we first made use of a mutated muscarinic M 3 receptor, which is deleted in most of the third intracellular loop (M 3 -short). Although still capable of activating phospholipase C, this receptor loses almost completely the ability to recruit (cid:1) -arrestin-1 following carbachol stimulation in COS-7 cells. Subsequently, M 3 - short was co-expressed with the M 3 receptor. Under these conditions, the M 3 /M 3 -short heterodimer could not recruit (cid:1) -arrestin-1 to the plasma membrane, even though the control M 3 /M 3 homodimer could. We formalin, and then permeabilized with Triton X-100. In order to recognize the 3HA-M 3 receptor, cells were exposed to a primary anti-HA antibody and then to a secondary antibody labeled with rhodamine. The red fluorescence (rhodamine) and the GFP images were acquired with a fluorescence microscope. The two images were merged with Image Photoshop software running on a MacIntosh computer. After 30 min of carbachol incubation, it was possible to see a clear co-localization of (cid:1) -arrestin-1 and 3HA-M 3 in endocytotic vesicles of cells transfected with 3HA-M 3 . The co-localization of (cid:1) -arrestin-1 and 3HA-M 3 is pre- vented by the M 3 -short receptor. fluorescence (rhodamine) and the GFP images were acquired with a fluorescence microscope. The two images were merged with Image Photoshop software running on a MacIntosh com-puter. After 15 min of carbachol stimulation, it was possible to observe a clear co-localization of (cid:1) -arrestin-1 with 3HA-M 3 in endocytotic vesicles that increased significantly at 30 min.

Over the past several years, receptor dimerization has become an established concept in the field of G protein-coupled receptors (1,2). Although the mechanism(s) by which receptors may undergo dimerization has yet to be elucidated in detail, it is becoming ever more clear that the phenomenon is playing a key role in receptor maturation, G protein coupling, and downstream signaling besides regulating such processes as internalization and desensitization. One of the most critical issues in receptor homodimerization and heterodimerization is whether signal transduction may require pair activation of receptor dimers. Many of the recent reports on G protein coupling consider separate stimulation of receptor dimers to be sufficient for activating G proteins in co-transfected cells (3)(4)(5)(6)(7). Accordingly, heterodimerization would not necessarily interfere with G protein coupling and the receptor monomer should be sufficient by itself to activate G protein. This conclusion is also supported by the observation that heterodimerization between receptors that bind to distinct G proteins quite often leaves their coupling selectivity unaltered. For instance, ␤ 2 adrenergic receptors that couple with stimulatory G proteins or ␦ and opioid receptors that couple with inhibitory G proteins both form heteromeric complexes but their heterodimerization does not significantly alter ligand binding or their coupling properties (8).
However, this interpretation does not take into account the possibility that simultaneous activation of both receptors with selective agonists presupposes a different pharmacology from that foreseeable for two receptors activated separately (6). In addition, the above conclusion is contradicted by a number of recent evidence. For instance, Baneres and Parello (9) have been able to demonstrate unambiguously with a combination of mass spectrometry and neutron scattering in solution that only one G protein trimer binds to a leukotriene B 4 (LTB 4 ) 1 receptor BLT1 dimer (2xBLT1.LTB 4 ), thus forming a stoichiometrically defined (2xBLT1.LTB 4 )G␣ i2 ␤ 1 2 pentameric assembly. They suggested that receptor dimerization may be crucial for LTB 4induced signaling. Similar conclusions have been reached in a recent paper by Chinault et al. (10) who demonstrated that yeast ␣-factor receptor oligomerization is required for G protein activation. These two latter papers point to the receptor dimer as the minimal structural configuration necessary to sustain receptor functioning. In line with these findings, we have recently demonstrated that paired activation of two receptor monomers within muscarinic M 3 dimers is required to stimulate ERK1/2 phosphorylation. In fact, a mutant muscarinic M 3 receptor (M 3 -short), which by itself cannot stimulate ERK1/2 phosphorylation, reduces by a large extent the ability of M 3 to activate ERK1/2 (11). However, the most compelling evidence supporting the requirement for receptor dimerization in ERK1/2 phosphorylation was obtained with chimeric ␣ 2 /M 3 and M 3 /␣ 2 receptors (12,13). We demonstrated that both adrenergic and muscarinic components of the ␣ 2 /M 3 and M 3 /␣ 2 chimeric receptor heterodimers must be activated for ERK1/2 to be phosphorylated (11). In the same study, we provided additional evidence that activation of ERK1/2 by M 3 receptors in COS-7 cells occurs independently of G protein activation. Because ␤-arrestins may activate mitogen-activated protein kinase by recruiting signaling molecules into a complex with agonistoccupied receptors (14), we suggested that activation of ERK1/2 by M 3 receptors may be due to the recruitment of ␤-arrestin to the site of the activated receptor.
The combined evidence that M 3 receptors might activate ERK1/2 via ␤-arrestin and that ERK1/2 activation is dependent upon receptor dimerization strongly suggests that the binding of M 3 receptors to ␤-arrestin depends on both receptor components of the muscarinic M 3 dimers. To test this hypothesis, we have extensively characterized the binding of ␤-arrestin-1 to a number of muscarinic M 3 receptor dimers using co-immunoprecipitation and fluorescence microscopy in COS-7. Our findings provide strong evidence that the binding of ␤-arrestin-1 to muscarinic M 3 receptor dimers requires a paired activation of two receptor components within the same receptor dimer. Eukaryotic Expression Vectors-Construction of M 3 -short and chimeric adrenergic/muscarinic ␣ 2 /M 3 and M 3 /␣ 2 receptors has been described elsewhere (12,13). In particular, M 3 -short was obtained by deleting 196 amino acids from the third cytoplasmic (i3) loop of M 3 (the remaining i3 loop was 43 amino acids long). Chimeric adrenergic/muscarinic ␣ 2 /M 3 and M 3 /␣ 2 receptors were obtained by exchanging the last two transmembrane regions between the adrenergic ␣ 2C and the muscarinic M 3 receptor. In both chimeras, the third cytoplasmic loop was from the M 3 receptor. A schematic representation of the receptor mutants used in this study is shown in Fig. 1. For co-immunoprecipitation experiments, we used a triple HA N-terminal tagged human M 3 receptor inserted in a pcDNA3 vector from Guthrie cDNA Resource Center and a c-Myc N-terminal tagged human M 2 receptor kindly provided by Dr. Tatsuya Haga. The chimeric M 3 /␣ 2 receptor was tagged at its N terminus with the HA antigen (HA-M 3 /␣ 2 ). The C-terminal tagged FLAG-␤-arrestin-1 was kindly provided by Dr. Robert Lefkowitz.

Materials and Reagents-N-[
Cell Cultures and Transfection-COS-7 cells were incubated at ϩ37°C in a humidified atmosphere (containing 5% CO 2 ) and grown in Dulbecco's modified Eagle's medium, which was supplemented with 5% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomy-cin. Cells were seeded at a density of 7.5 ϫ 10 5 /100-mm dish, and 24 h later, they were transiently transfected with the plasmid DNA by the DEAE-dextran chloroquine method. The total amount of DNA used for each transfection was brought to 4 g by adding an appropriate amount of vector DNA.
Membrane Preparation and Binding Assay-On day 1, COS-7 cells were transfected with the plasmid(s) of interest. Three days after transfection, confluent plates of cells were lysed by replacing the medium with ice-cold hypotonic buffer (1 mM Na-HEPES, 2 mM EDTA). After 20 min, cells were scraped off the plate and centrifuged at 17,000 rpm for 20 min at ϩ4°C. The lysed cell pellet was homogenized with a Polytron homogenizer in ice-cold binding buffer (50 mM Tris-HCl, pH 7.4, 155 mM NaCl, 0.01 mg/ml bovine serum albumin). Binding was carried out at ϩ30°C in a final volume of 1 ml, and atropine 1 M was used to define nonspecific binding. The bound ligand was separated from the unbound ligand using glass-fiber filters (Whatmann, GF/B) with a Brandel cell harvester, and the filters were counted with a scintillation ␤-counter.
Phosphatidylinositol Breakdown Assay-Transfected COS-7 cells were incubated with myo-[ 3 H]inositol (3 Ci/ml) for 48 h. Immediately prior to the experiment, the cells were washed twice with phosphatebuffered saline and incubated for 15 min in Eagle's minimal essential medium containing 10 mM LiCl and 20 mM HEPES. The medium was then replaced by 0.25 ml of the same medium containing the experimental agents. After a 1-h incubation at ϩ25°C, the reaction was arrested by the addition of 0.75 ml of 7.5% (w/v) ice-cold trichloroacetic acid followed by a 30-min incubation on ice. The trichloroacetic acid was extracted with water-saturated diethyl ether (3 ϫ 4 ml), and the levels of inositol monophosphate were determined by anion exchange chromatography.
Adenyl Cyclase Assay-COS-7 cells were transfected with the plasmid(s) containing the receptor of interest plus the adenyl cyclase V. 24-h after transfection, the cells were trypsinized and recultured in 24-well plates, and after an additional 24 h, they were assayed for adenyl cyclase activity. The cells in the 24-well plates were incubated for 2 h with 0.25 ml/well fresh growth medium containing 5 Ci/ml [ 3 H]adenine, and this medium was replaced with 0.5 ml/well DMEM containing 20 mM HEPES, pH 7.4, 0.1 mg bovine serum albumin, and the phosphodiesterase inhibitors 1-methyl-3-isobutylxanthine (0.5 mM) and RO-20-1724 (0.5 mM). Adenyl cyclase activity was stimulated by the addition of 1 M forskolin in the presence or absence of carbachol. After 10 min of incubation at ϩ30°C, the medium was removed and the reaction was terminated by the addition of perchloric acid containing 0.1 mM unlabeled cAMP followed by neutralization with KOH. The amount of [ 3 H]cAMP formed was determined by a two-step column separation procedure.
Immunoprecipitation and Western Blotting-For immunoprecipitation, confluent cells were treated with the selected compound 3 days after transfection. Following incubation, the reaction was stopped by washing the cells twice with ice-cold sterile 0.9% NaCl. Cells were then scraped with 300 l of buffer solution A (0.2% digitonin, 1 mM EDTA in phosphatebuffered saline) containing a mixture of protease inhibitors (1.5 M pepstatin, 4 M leupeptin, 0.01 M aprotonin, 500 M phenylmethylsulfonyl fluoride). Lysed cells were incubated on ice for 20 min and then centrifuged at 6000 rpm for 10 min at ϩ4°C, and the supernatant was recovered for immunoprecipitation. The amount of proteins was assayed with the protein assay kit of Bio-Rad, and aliquots of 60 g of proteins (if not otherwise specified) were used for immunoprecipitation.
To eliminate proteins nonspecifically bound to magnetic beads, the cell extract was preexposed to 25 l of protein G magnetic beads (New England Biolab) for 1 h at ϩ4°C. A magnetic field was applied to the side of the tube, and the supernatant was recovered and transferred to a clean 1.5-ml microcentrifuge tube. For immunoprecipitation, 5 g of a monoclonal anti-HA (Roche Applied Science) or an anti-c-Myc antibody (Sigma) was added to the tube and the suspension was incubated at ϩ4°C for 1 h. At the end of this incubation period, 25 l of protein G magnetic beads were added and incubated at ϩ4°C for an additional hour. A magnetic field was applied to pull beads to the side of the tube, and the supernatant was carefully removed. The magnetic beads were washed with 500 l of buffer solution B (0.1% Niaproof, 0.2 M NaCl, 1 mM EDTA, 50 mM Tris, pH 7.5) including the protease inhibitor mixture. Bead pellets were recovered by applying a magnetic field, and they were resuspended in 30 l of Laemmli buffer and incubated at ϩ70°C for 5 min.
A magnetic field was applied to the sample, and the supernatant was recovered and loaded on 10% SDS-polyacrylamide gel. Proteins were run at 90 V for 2-3 h and transferred to the nitrocellulose membrane for 1 h at 100 V. Co-immunoprecipitated proteins were determined by immunoblotting with specific antibodies such as polyclonal anti-␤-ar-FIG. 1. Schematic representation of wild type muscarinic M 3 receptor and the derived receptor mutants. The M 3 -short has been obtained by cutting out 196 amino acids from the i3 loop of the M 3 receptor (black strip). The ␣ 2 /M 3 and M 3 /␣ 2 chimeras were obtained by exchanging the last two transmembrane domains between the muscarinic M 3 and the adrenergic ␣ 2C receptor (green strip). In both chimeras, the i3 loop was from the M 3 receptor.
restin-1 (Santa Cruz Biotechnology), polyclonal anti-i3-loop M 3 (Santa Cruz Biotechnology), monoclonal anti-FLAG (Sigma), and monoclonal anti-c-Myc antibody. Data from separate experiments were digitized on a flat bed scanner and analyzed with a Kodak Scientific Imaging Systems software.
Immunocytochemistry-COS-7 cells were transfected using the DEAE-dextran chloroquine method on 100-mm plates, and the day after transfection, they were split and seeded on coverslips previously placed on 6-well plates at a confluence of 2 ϫ 10 5 cells/well. Three days after transfection, they were treated with 100 mM carbachol and then processed for immunocytochemistry. Cells were fixed for 10 min in 3.7% paraformaldehyde, permeabilized for 5 min with 0.2% Triton X-100, and then incubated in 1% BSA for 1 h to reduce background staining. They were then exposed to a rabbit polyclonal anti-HA antibody (Santa Cruz Biotechnology) for 1 h at a working dilution of 1/100 followed by another incubation with rhodamine-labeled donkey anti-rabbit secondary antibody (Jackson ImmunoResearch) at a working dilution of 1/100. Coverslips were washed with phosphate-buffered saline and mounted on slides using Prolong Antifade kit (Molecular Probe). The coverslips were analyzed with a Zeiss Axioplan fluorescence microscope using ϫ63ϫ/1.25 oil Zeiss objective.

RESULTS
To characterize the expression of endogenous ␤-arrestin-1 in COS-7 cells, protein extracts from mock-transfected COS-7 cells and from cells transfected with a C-terminal tagged FLAG-␤-arrestin-1 were immunoblotted with an antibody directed against the C terminus of ␤-arrestin-1. In mock-transfected COS-7 cells, a band corresponding to the molecular mass of ␤-arrestin-1 (55 kDa) was detected on the blot ( Fig. 2A). In COS-7 cells transfected with FLAG-␤-arrestin-1, the same antibody recognized a broad band corresponding to the FLAG-␤arrestin-1. In COS-7 cells transfected with FLAG-␤-arrestin-1, the anti-FLAG antibody recognized the same band as the anti-␤-arrestin-1 antibody, whereas it did not detect any signal in mock-transfected COS-7 cells (Fig. 2B).
To investigate the nature of receptor-arrestin interaction, co-immunoprecipitation experiments were performed in COS-7 cells co-expressing ␤-arrestin-1 and the 3ϫHA N-terminal tagged M 3 receptor (3HA-M 3 ). The 3HA-M 3 was immunoprecipitated with an anti-HA antibody, and the amount of ␤-arrestin-1 pulled down was normalized to the amount of the immunoprecipitated receptor. The same amount of immunoprecipitated 3HA-M 3 (and in some experiments M 3 ) obtained with an antibody directed against the third cytoplasmic loop could be reproducibly recovered in different experiments. In cells transfected with 3HA-M 3 or M 3 (results not shown), this antibody recognizes two bands that most likely correspond to the monomeric and dimeric forms of the muscarinic receptor (Fig. 2C).
The amount of ␤-arrestin-1 co-immunoprecipitated with the 3HA-M 3 receptor in the absence of agonist (basal) was usually very low or undetectable. In contrast, stimulation of 3HA-M 3 receptors with 100 M carbachol induced a time-and concentration-dependent increase in the co-immunoprecipitated ␤arrestin-1 complex, which then reached its maximum by ϳ20 min (Fig. 3). A dose-response curve of carbachol-induced ␤arrestin-1 co-immunoprecipitation was obtained by incubating cells transfected with 3HA-M 3 receptors for 10 min at the different concentrations of the agonist. The EC 50 was between 10 and 100 M, and no further increase in ␤-arrestin-1 coimmunoprecipitation was observed up to 1 mM (Fig. 4).
It has been shown that the phosphorylated i3 loop of M 3 receptors is important for binding ␤-arrestin-1 (15). Because deletion of the i3 loop is expected to impair carbachol-induced stimulation of ␤-arrestin-1 binding, we tested the ability of a mutated muscarinic receptor in which 196 amino acids of the i3 loop were deleted (M 3 -short) to co-immunoprecipitate with ␤-arrestin-1. For this purpose, we used the N-terminal tagged HA-M 3 -short. As predicted, stimulation of this receptor with 100 M carbachol did not increase the amount of co-immunoprecipitated ␤-arrestin-1 compared with the basal value ( Fig.  5). Nonetheless, the basal amount of ␤-arrestin-1 that co-immunoprecipitated with the HA-M 3 -short receptor was significantly higher than that co-immunoprecipitated with the wild type receptor. This suggests that, in addition to the i3 loop, other segments of the M 3 receptor can also be involved in ␤-arrestin-1 binding.
We have previously shown that the muscarinic M 3 -short receptor interacts with the wild type M 3 receptor to form heterodimers (11). We calculated that up to 76.6% M 3 receptor expressed could be co-immunoprecipitated with M 3 -short in COS-7 cells co-transfected with 2 g of M 3 -short and 1 g of M 3 . In the same paper, we showed that M 3 -short impairs the ability of M 3 to stimulate ERK1/2 phosphorylation in the heterodimer. Here, we tested whether the M 3 -short could interfere with the ability of 3HA-M 3 to co-immunoprecipitate with ␤-arrestin-1. As shown in Fig. 5, co-expression of M 3 -short almost entirely prevents carbachol-dependent co-immunoprecipitation of ␤-arrestin-1 with the M 3 receptor. This result was not due to potential problems linked to co-transfection, because overexpression of wild type M 3 did not impair the amount of ␤-arrestin-1 co-immunoprecipitated with the 3HA-M 3 (Fig. 5).
These findings strongly suggest that M 3 -short interferes with the ability of 3HA-M 3 to recruit ␤-arrestin-1. To test this hypothesis, we verified whether ␤-arrestin-1 and 3HA-M 3 could co-localize following muscarinic receptor activation in living cells. For this purpose, ␤-arrestin-1 C-terminal tagged with GFP (␤-arrestin-1-GFP) and the 3HA-M 3 receptor were co-expressed in COS-7 cells. As shown in Fig. 6, the ␤-arrestin-1-GFP fluorescence was uniformly distributed in the cytoplasm (except when in the nucleus which was intensely stained). After 30 min of stimulation with carbachol (100 M), several endocytotic vesicles could be discerned at cell sites where the receptor red staining and the green fluorescence of ␤-arrestin-1-GFP overlapped (Fig. 6, Merge), indicating that muscarinic 3HA-M 3 receptors and ␤-arrestin-1-GFP were indeed co-localized in these cells. Interestingly, no co-localization of ␤-arrestin-1-GFP and muscarinic 3HA-M 3 receptor could be revealed in endocytotic vesicles of cells co-expressing 3HA-M 3 , M 3 -short, and ␤-arrestin-1-GFP after 30 min of carbachol stimulation (Fig. 6). These findings strongly suggest that deletion of the i3 loop in one of the two monomers constituting the M 3 dimer may suffice to impair recruitment of ␤-arrestin-1 to the plasma membrane.
To further test whether recruitment of ␤-arrestin-1 does indeed require two fully functional receptors within a dimer, ␤-arrestin-1 was allowed to interact with a dimer formed by muscarinic M 2 and M 3 receptors. Using a pharmacological approach, we had previously suggested that M 2 and M 3 receptors can form heterodimers (16). In a recent work, Hornigold et al. (17) have provided evidence suggesting that both second messengers and extracellular signal-regulated kinases may be regulated via a cross-talk between M 2 and M 3 muscarinic acetylcholine receptors. To determine whether M 2 and M 3 receptors can interact physically, 3HA-M 3 receptors were co-expressed with Myc-M 2 receptors. Under these conditions (Fig. 7) Contrary to our data, Zeng and Wess (18) reported no coimmunoprecipitation between the muscarinic M 2 and M 3 receptors. We do not have any explanation for this discrepancy.
It is interesting to note that, when transfected alone, the M 3 receptor migrate both as a dimer and a monomer in polyacrylamide gels (Figs. 2C and 7B), whereas when co-immunoprecipitated with M 2 , it can only be detected in a monomeric form and not in a M 3 /M 2 heterodimeric form (Fig. 7B). The observation that receptor dimers may persist in denaturating gels suggests the possibility that the two-receptor monomers might be interacting through disulfide bridges. As a matter of fact, Zeng and Wess (18) have previously shown that M 3 /M 3 homodimers are stabilized by disulfide bonds and that modification of Cys 140 and/or Cys 220 in the receptor greatly reduces the amount of M 3 detectable in the high molecular weight dimeric form. A likely explanation to account for the apparent discrepancy with our results may be that M 2 and M 3 do not interact through disulfide bonds such that the heterodimer may break apart upon running in denaturing gels.
Interestingly, the time course of carbachol-induced ␤-arrestin-1 co-immunoprecipitation is clearly left-shifted in cells coexpressing the 3HA-M 3 and M 2 receptors compared with that observable in cells expressing the 3HA-M 3 alone (Fig. 3). In fact, 2 min of carbachol stimulation suffice to cause significant immunoprecipitation of ␤-arrestin-1 and to plateau by ϳ5 min. The dose-response curve of carbachol is also shifted to the left with an EC 50 below 1 M (Fig. 4). Under our experimental conditions, Myc-M 2 receptors could not co-immunoprecipitate with ␤-arrestin-1 up to 20 min of incubation and 1 mM concentration of carbachol (Fig. 8A). When Myc-M 2 and M 3 receptors were co-expressed, a significant co-immunoprecipitation of ␤-arrestin-1 with Myc-M 2 could be observed. This observation does not rule out the possibility that M 2 might bind ␤-arrestin-1, as it appears to be the case when exogenous ␤-arrestin is co-expressed with the receptor (19). However, it suggests that ␤-arrestin-1 might bind to the M 2 receptor much more weakly than to the M 2 /M 3 heterodimer. This interpretation is corroborated by the observation that significant co-immunoprecipitation can be produced when COS-7 cells co-transfected with FLAG-␤-arrestin-1 and Myc-M 2 are stimulated with 100 M carbachol (Fig. 8B).
Altogether these findings suggest that M 2 /M 3 heterodimerization enhances the ability of each receptor monomer to recruit ␤-arrestin-1. This conclusion is further corroborated by the observations that carbachol stimulation in COS-7 cells co-expressing 3HA-M 3 and M 2 receptors causes ␤-arrestin-GFP to become redistributed within the cell. As is shown in Fig. 9, recruitment of ␤-arrestin-1-GFP in these cells is more rapid than in cells transfected solely with 3HA-M 3 . Following 15 min of carbachol stimulation and even more so at a 30-min interval, a large number of endocytotic vesicles appear labeled by both red and green fluorescence, indicating that ␤-arrestin-1 and 3HA-M 3 receptors co-localize along the plasma membrane ( Fig.  9, Merge).
To provide further support to the view that M 2 receptor may act by enhancing ␤-arrestin-1 binding to the heterodimeric M 2 /3HA-M 3 receptor, co-immunoprecipitation was performed in the presence of the selective M 2 antagonist tripitramine (20). By N-[ 3 H]methylscopolamine displacement binding analysis, we estimated that, in the presence of 4 nM tripitramine, ϳ96% M 2 receptors and only 9% 3HA-M 3 receptors are saturated by the antagonist (Fig. 10A). Furthermore, functional experiments indicated that, in COS-7 cells co-transfected with M 2 and 3HA-M 3 , 4 nM tripitramine prevented the carbachol-stimulated M 2 inhibition of adenylyl cyclase, whereas it did not prevent the M 3 -induced phosphatidylinositol hydrolysis (Fig.  10B). Thus, cells co-expressing the 3HA-M 3 and M 2 were stimulated with carbachol for 10 min in the presence and absence of 4 nM tripitramine. As it can be seen in Fig. 10C, tripitramine largely reduced the co-immunoprecipitation of ␤-arrestin-1 in cells co-expressing 3HA-M 3 and M 2 , whereas it did not affect In previous work, we demonstrated that the co-expression of chimeric adrenergic/muscarinic ␣ 2 /M 3 and M 3 /␣ 2 receptors could rescue the binding of muscarinic and adrenergic ligands (12,13). In both chimeric receptors, the i3 loop was from the muscaric M 3 receptor. This explains why the activation of both the cholinergic or the adrenergic halves of the reconstituted receptor stimulates phosphatidylinositol hydrolysis (11). The chimeric ␣ 2 /M 3 and M 3 /␣ 2 receptors are a good tool for testing whether the paired stimulation of a dimer is required to recruit ␤-arrestin-1 to the plasma membrane, because the only functional form is a dimer. For this purpose, the N-terminal tagged HA-M 3 /␣ 2 receptor (11) was co-expressed in COS-7 cells with the other ␣ 2 /M 3 chimera. Because through binding analysis the estimated amount of the two chimeras that fold correctly and form heterodimers is very low (30 -70 fmol mg proteins, ϳ5% of the wild type 3HA-M 3 receptor), to magnify the amount of translocated ␤-arrestin-1, the receptors were stimulated with the agonists for 30 min. Furthermore, for co-immunoprecipitation, we used five times the amount of proteins that we used with the other receptors. As shown in Fig. 11, ␤-arrestin-1 co-immunoprecipitation was clearly detected only when the adrenergic and the muscarinic receptor components of the heterodimer were stimulated simultaneously with carbachol and clonidine. No co-immunoprecipitation was observed when cells were stimulated with either carbachol or clonidine alone. These results further support the hypothesis that both components of the dimer need to be activated to recruit ␤-arrestin-1 to the plasma membrane. DISCUSSION In this study, we provide strong evidence that both components of muscarinic receptor dimers are required to fully recruit ␤-arrestin-1 and that this might have important functional implications on activation of the ERK1/2 pathway.
The main finding in support of this evidence is that a receptor dimer formed by the wild type M 3 receptor and its M 3 -short In panel A, confluent monolayers of COS-7 cells were stimulated for different times and with different concentration of carbachol. Incubation was terminated by adding icecold NaCl (0.9%). Cells were extracted in digitonin, and aliquots of 60-g proteins were incubated with an anti-Myc antibody and processed for immunoprecipitation (IP) with protein G magnetic beads (see "Experimental Procedures"). The amount of ␤-arrestin-1 co-immunoprecipitated was detected with an antibody directed against the Cterminal end of ␤-arrestin-1. In any of the conditions, it was possible to detect ␤-arrestin-1 co-immunoprecipitation. As a control, COS-7 cells were also co-transfected with Myc-M 2 and M 3 . In this case, a clear co-immunoprecipitation of ␤-arrestin-1 could be detected on the blot after stimulation with 100 M carbachol for 5 min. In panel B, COS-7 cells were transfected with Myc-M 2 and FLAG-␤-arrestin-1 and stimulated with 100 M carbachol. Cells were processed as described above, and the amount of ␤-arrestin-1 co-immunoprecipitated was detected with an anti-FLAG antibody. As it can be seen in the third line, FLAG-␤-arrestin-1 could be detected on the blot after stimulation with carbachol. The blots are representative of five (A) and three (B) experiments. IB, immunoblot; Carb., carbachol.

FIG. 9. Carbachol stimulation of ␤-arrestin-1-GFP translocation in COS-7 cells transfected with 3HA-M 3 and M 2 . COS-7 cells
were co-transfected with 1 g of 3HA-M 3 , 2 g of M 2 , and 1 g of ␤-arrestin-1-GFP. Cells were stimulated with 100 M carbachol for the times indicated, fixed in formalin, and then permeabilized in Triton X-100. In order to recognize the 3HA-M 3 receptor, cells were exposed to a primary anti-HA antibody and then to a secondary antibody labeled with rhodamine. The red fluorescence (rhodamine) and the GFP images were acquired with a fluorescence microscope. The two images were merged with Image Photoshop software running on a MacIntosh computer. After 15 min of carbachol stimulation, it was possible to observe a clear co-localization of ␤-arrestin-1 with 3HA-M 3 in endocytotic vesicles that increased significantly at 30 min. mutant, which lacks the ability of recruiting ␤-arrestin, was impaired in its ability to interact with ␤-arrestin-1. Since we and others (11,21) have previously shown that the M 3 -short was unable to activate the ERK1/2 pathway, this finding strongly supports the hypothesis that ERK1/2 activation requires the recruitment of ␤-arrestin-1 by the paired activation of the two-receptor components of the muscarinic M 3 dimers.
The interaction of ␤-arrestin-1 with the wild type M 3 receptor and its M 3 -short mutant was explored both by co-immunoprecipitation and fluorescence microscopy experiments in COS-7 cells co-expressing the different proteins. The muscarinic M 3 receptor was able to co-immunoprecipitate with ␤-arrestin-1 after agonist stimulation in a time-and concentrationdependent manner. In co-immunoprecipitation experiments, the interaction with ␤-arrestin-1 could be already observed on the blots after 5 min of stimulation with carbachol. In fluorescence microscopy, the co-localization of the M 3 and ␤-arrestin-1 was clearly observed after 30 min of carbachol stimulation. Thus, the recruitment of ␤-arrestin-1 observed in co-immunoprecipitation experiments displays slightly faster kinetics than in fluorescence microscopy. This might be due to the fact that, under the conditions of fluorescent microscopy used in our experiments, a distinct signal can only be seen when the receptor and ␤-arrestin-1 are accumulated in the defined part of the cells (clathrin-coated pits). It is likely that at earlier times the translocation of ␤-arrestin-1 to the plasma membrane could not be appreciated.
In contrast, carbachol did not induce co-immunoprecipitation of ␤-arrestin-1 with the M 3 -short receptor mutant lacking most of its i3 loop. This is consistent with the fact that the phosphorylated i3 loop of the receptor plays an important role in this interaction. Surprisingly, the amount of ␤-arrestin-1 co-immunoprecipitated with the receptor under basal conditions was significantly increased. This suggests that, even though the deletion of the i3 loop abolished the agonist dependence of ␤-arrestin-1 binding to M 3 , other parts of the receptor can interact with ␤-arrestin-1. This would be consistent with the finding that mutations in the highly conserved DRY region of the N-formyl peptide receptor (22) and of the ␣ 1b adrenergic receptor (23) can impair ␤-arrestin-1 binding. It is also possible receptors. COS-7 cells were co-transfected with 2 g of M 2 and 1 g of 3HA-M 3 plasmid DNA and stimulated with 100 M carbachol in the presence or absence of tripitramine. For inositol phosphate breakdown assay, cells were prelabeled for 48 h with 3 Ci/ml [ 3 H]inositol and, immediately before the assay, incubated for 15 min in Eagle's minimal essential medium containing 10 mM LiCl and 20 mM HEPES. The medium was then replaced by the same medium containing carbachol, and the cells were then incubated for 1 h at ϩ25°C. For the adenyl cyclase assay, cells were prelabeled for 2 h with 5 Ci/ml [ 3 H]adenine and then adenyl cyclase activity was stimulated for 10 min at ϩ30°C by the addition of 1 M forskolin in the presence of carbachol. In both assays, tripitramine was added 15 min before and during the assays. The graph shows the means Ϯ S.E. of two experiments. In panel C, confluent monolayers of COS-7 cells were stimulated with 100 M carbachol in the presence or absence of tripitramine for 10 min and incubation was terminated by adding ice-cold NaCl (0.9%). Cells were extracted in digitonin, and aliquots of 60-g proteins were incubated with an anti-HA antibody and processed for immunoprecipitation (IP) with protein G magnetic beads (see "Experimental Procedures"). The amount of ␤-arrestin-1 co-immunoprecipitated was detected with an antibody directed against the C-terminal end of ␤-arrestin-1 and was normalized for the amount of 3HA-M 3  that the deletion of the bulky i3 loop in the M 3 -short allows the binding of ␤-arrestin-1 to other parts of the receptor. The functional implications of this interaction remain to be investigated.
Because the receptor dimer formed by the wild type M 3 receptor and its M 3 -short mutant, which lacks the ability of recruiting ␤-arrestin, was impaired in its ability to interact with ␤-arrestin-1, this finding strongly suggests that the interaction of ␤-arrestin-1 with M 3 dimers requires the activation of both monomeric components within the dimers.
These findings can have different mechanistic interpretations. One possibility is that M 3 -short modifies the structure of wild type M 3 in such a way that ␤-arrestin-1 binding is precluded. Nevertheless, we consider this possibility unlikely since we have shown previously (13) that co-expression of the M 3short does not alter M 3 binding and function. This finding suggests that the overall structure of the M 3 -short receptor, the mechanism of activation, and probably its dimerization interface are not altered. Two alternative hypotheses are the most likely to interpret our data. One possibility is that a ␤-arrestin monomer requires two full-length i3 loops to stably bind to the receptor, thus preferentially interacting with receptor dimers. If this is the case, heterodimeric M 3 /M 3 -short receptors will provide only one i3 loop, thus resulting in lower ␤-arrestin binding. This hypothesis would be consistent with the results from a recent study on the organization of rhodopsin in native membranes (24). Arrestin, the cognate ␤-arrestin in the visual system, has a bipartite structure of two structurally homologous seven-stranded ␤-sandwiches forming two putative rhodopsin-binding groves that are separated by 3.8 nm (25,26). The positive charge arrangement of the surface of the rhodopsin dimer matches the negative charges on arrestin. Thus, Liang et al. (24) speculate that one arrestin monomer is likely to bind one rhodopsin dimer.
The second hypothesis is that a ␤-arrestin-1 dimer binds to the receptor dimer. This could occur in two ways. (i) A preformed ␤-arrestin-1 dimer binds all at once to the receptor dimer, or (ii) two ␤-arrestin-1 molecules bind sequentially to the receptor dimer with the second one stabilizing the complex. In both cases, heterodimeric M 3 /M 3 -short receptors would have only one arm able to bind the ␤-arrestin-1 dimer and the receptor/␤-arrestin-1 complex would be much weaker.
The idea of dimeric ␤-arrestin-1 as a functional unit comes from different observation. Monomeric and dimeric forms of visual arrestin are at equilibrium under physiological conditions (27). Dimerization has been demonstrated also for ␤-arrestin. ␤-Arrestin-(1-382), a C-tail truncation mutant of bovine ␤-arrestin, exists as a mixture of monomeric and dimeric species (28). It was shown that ␤-arrestin forms a tail-to-tail dimer with two C-tail domains facing each other (29 -32). Taking together these data, Han et al. (28) propose a mechanistic model of ␤-arrestin-receptor interaction in which the initial binding of the first ␤-arrestin to the receptor is followed by the displacement of its terminal C-tail and dimerization with another molecule of ␤-arrestin. They speculate that ␤-arrestin dimerization may help ␤-arrestin-receptor complexes fit better with the internalization machinery of the coated pits. Furthermore, they left open the possibility that dimerization of ␤-arrestin could play a role as scaffold for mitogen-activated protein kinase given that complexes containing ␤-arrestin and mitogen-activated protein kinase are large in size (33,34).
The experiments with the M 3 -short receptor mutant indicate that two fully active receptors are necessary for ␤-arrestin-1 binding but do not suggest whether the two receptors monomers must be activated simultaneously or not. A clear answer to this question comes from the experiments using muscarinic/adrener-gic chimeric ␣ 2 /M 3 and M 3 /␣ 2 receptors. These receptors expressed alone are not functional and do not bind either muscarinic or adrenergic ligands, whereas the only functional form is the ␣ 2 /M 3 -M 3 /␣ 2 heterodimer (12). We demonstrated that the recruitment of ␤-arrestin-1 by this chimeric ␣ 2 /M 3 -M 3 /␣ 2 receptor heterodimer requires the paired activation of both the muscarinic and adrenergic components. In fact, no ␤-arrestin-1 co-immunoprecipitation was observed when cells co-expressing the chimeric receptors were stimulated with carbachol or clonidine alone. The activation of only one arm of the dimer was not sufficient for ␤-arrestin-1 binding, probably because both the i3 loops must be phosphorylated. These results are in strong agreement with previous data showing that ERK1/2 phosphorylation was induced only by stimulation of the chimeric heterodimer with both carbachol and clonidine (11).
Another line of evidence supporting the correlation between ␤-arrestin-1 binding to receptor dimers and ERK1/2 activation comes from the results on muscarinic M 2 /M 3 receptor heterodimers. Using a pharmacological approach, we had previously suggested that the muscarinic M 2 and M 3 receptors can form heterodimers. In agreement with this hypothesis, in this work, we have demonstrated that the M 3 and M 2 receptors can co-immunoprecipitate when they are expressed together.
Surprisingly, endogenous ␤-arrestin-1 could not co-immunoprecipitate with the M 2 receptor alone but we confirmed (19) other author findings that the M 2 receptor can bind exogenously co-transfected ␤-arrestin-1. Two alternative explana- FIG. 11. Paired stimulation with carbachol and clonidine of chimeric muscarinic/adrenergic ␣ 2 /M 3 and HA-M 3 /␣ 2 receptors induced ␤-arrestin-1 co-immunoprecipitation. COS-7 cells were co-transfected with the chimeric ␣ 2 /M 3 and HA-M 3 /␣ 2 receptors (2 g of DNA each). Confluent monolayers of COS-7 cells were stimulated with carbachol (100 M), clonidine (100 M), or a mixture of the two for 30 min. Incubation was terminated by adding ice-cold NaCl (0.9%). Cells were extracted in digitonin, and aliquots of 300-g proteins were incubated with an anti-HA antibody and processed for immunoprecipitation (IP) with protein G magnetic beads (see "Experimental Procedures"). The amount of ␤-arrestin-1 co-immunoprecipitated was detected with an antibody directed against the C-terminal end of ␤-arrestin-1. Only in cells treated with both agonists (carbachol and clonidine), it was possible to see ␤-arrestin-1 co-immunoprecipitation. The blot is representative of four experiments. Bar graph shows the means Ϯ S.E. of all of the experiments. The amount of ␤-arrestin-1 co-immunoprecipitated was normalized for the amount of the ␣ 2 /M 3 and HA-M 3 /␣ 2 receptors immunoprecipitated that were detected with the anti-i3 loop M 3 antibody. *, significantly different from clonidine and carbachol alone (paired Student's two-tailed t test, p Ͻ 0.05). IB, immunoblot. tions can be foreseen to interpret these discrepancies. (i) The relative affinity between ␤-arrestin-1 and the M 2 receptor could be weak, but the increase in the expression of ␤-arrestin-1 by exogenous co-transfection could forced this interaction. (ii) It has been shown that, in COS-7 cells, the M 2 receptor is sequestered in caveolae in a ␤-arrestin/clathrin-independent pathway (35), and it is possible that an increase in the amount of ␤-arrestin-1 could recruit part of the receptor in this pathway.
When we co-transfected M 2 and M 3 receptors together, recruitment of ␤-arrestin-1 by the co-transfected receptors was much higher compared with cells transfected with M 3 alone, suggesting that M 2 /M 3 heterodimers bind ␤-arrestin-1 better than M 3 /M 3 homodimers.
The binding of ␤-arrestin-1 probably correlates with ERK1/2 activation since Hornigold et al. (17) have shown that, when the muscarinic M 2 and M 3 receptors are co-expressed in Chinese hamster ovary cells, these receptors work synergistically to activate the ERK pathway. We also found that a selective antagonist of the M 2 receptor markedly reduced the carbacholinduced co-immunoprecipitation of ␤-arrestin-1 by the muscarinic M 2 /M 3 heterodimer. This result indicates that, as has been observed with the homodimer M 3 /M 3 , the stimulation of both receptor components in the heterodimer is an essential requirement for ␤-arrestin-1 binding.
In conclusion, our findings demonstrate that the recruitment of ␤-arrestin-1 by homodimeric M 3 /M 3 and heterodimeric M 2 /M 3 muscarinic receptors as well as heterodimeric muscarinic/adrenergic chimeric receptors requires the paired activation of the single receptor components within the dimer. We propose that this phenomenon is mechanistically correlated to the activation of the ERK1/2 pathway. This is, to our knowledge, the first clear evidence that ␤-arrestin-1 binding requires dimerization of G protein-coupled receptors and provides some mechanistic hypothesis regarding receptor-␤-arrestin interaction that should be investigated in future studies. A crucial issue to address will be whether ␤-arrestin binds to G proteincoupled receptor dimers in its monomeric or dimeric form. Unraveling the molecular mechanisms of receptor-␤-arrestin interaction might have important implications for the further understanding the endocytic and signaling processes of G protein-coupled receptors.