The PDZ Protein Mupp1 Promotes Gi Coupling and Signaling of the Mt1 Melatonin Receptor*

Intracellular signaling events are often organized around PDZ (PSD-95/Drosophila Disc large/ZO-1 homology) domain-containing scaffolding proteins. The ubiquitously expressed multi-PDZ protein MUPP1, which is composed of 13 PDZ domains, has been shown to interact with multiple viral and cellular proteins and to play important roles in receptor targeting and trafficking. In this study, we show that MUPP1 binds to the G protein-coupled MT1 melatonin receptor and directly regulates its Gi-dependent signal transduction. Structural determinants involved in this interaction are the PDZ10 domain of MUPP1 and the valine of the canonical class III PDZ domain binding motif DSV of the MT1 carboxyl terminus. This high affinity interaction (Kd ∼ 4 nm), which is independent of MT1 activation, occurs in the ovine pars tuberalis of the pituitary expressing both proteins endogenously. Although the disruption of the MT1/MUPP1 interaction has no effect on the subcellular localization, trafficking, or degradation of MT1, it destabilizes the interaction between MT1 and Gi and abolishes Gi-mediated signaling of MT1. Our findings highlight a previously unappreciated role of PDZ proteins in promoting G protein coupling to receptors.

The concept of organized networks has emerged in the field of cellular signaling in the last few years. Assembling the different partners in close proximity optimizes the spatial and temporal organization and the specificity of the cellular response. The assembly of these multimolecular complexes occurs through the interaction of modular domains recognizing their target counterparts. PDZ domains are widely spread modules exhibiting this function. Data bank exploration with SMART (1) identifies 584 PDZ domains in 328 different proteins in the human genome.
G protein-coupled receptors (GPCRs) 5 constitute the largest family of membrane receptors, and many of the more than 750 members have been shown to interact with PDZ domain-containing proteins, either constitutively or upon agonist activation (2). Binding of PDZ proteins to GPCRs has been reported to primarily regulate subcellular localization, trafficking, and stability of receptors (3). For instance, binding of MUPP1 and syntrophins to the ␥-aminobutyric acid type B (GABA B ) receptor and the ␣ 1D -adrenergic receptor, respectively, significantly increases receptor stability (4,5). In other cases, PDZ scaffolds determine the subcellular localization of GPCRs (6) and receptor endocytosis as shown for PSD-95 and the 5-HT 2A serotonin and ␤ 1 -adrenergic receptors (7,8). PDZ proteins, such as NHERF and hScrib, are also important for the recycling of receptors to the cell surface (9 -11).
Binding of PDZ proteins to GPCRs also modulates receptor signaling by assembling proteins involved in signal transduction. NHERF family proteins are known to regulate the activity of the Na ϩ /H ϩ exchanger through association with NHERF-1 (12) and to form a ternary complex with phospholipase C␤3 and GPCRs, which enhances the signaling efficiency of the receptor-mediated activation of the phospholipase C␤/Ca 2ϩ pathway (13)(14)(15). Binding of GIPC (GAIP-interacting protein, COOH terminus) to the COOH terminus of the D3 dopamine and the ␤ 1 -adrenergic receptor (16) decreased G␣ i -mediated signaling of these receptors most likely through RGS19, which binds to GIPC (17). Further examples of PDZ scaffolds that regulate GPCR signaling include a ternary complex formation around the PDZ scaffold MAGI-3, which binds to the GPCR frizzled-4 and Ltap to regulate the JNK signaling cascade (18), as well as PDZ-domain-containing Rho guanine nucleotide exchange factors that interact with lysophosphatidic acid 1 and 2 receptors to activate RhoA (19).
To identify proteins that specifically interact with the G protein-coupled human MT 1 and MT 2 melatonin receptors, we performed a yeast two-hybrid screen using the cytoplasmic domains of these receptors as baits. The multi-PDZ domain protein MUPP1 was identified as interacting partner of the carboxyl-terminal tail (C-tail) of MT 1 but not of MT 2 . Importantly, this interaction was necessary for the stabilization of the MT 1 -G i complex and efficient G i -dependent signaling of MT 1 .

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screen-The cDNA sequences corresponding to the intracellular loops i2 (residues 123-141), i3 (residues 208 -239), and the C-tail (residues 294 -350) of MT 1 or the C-tail alone were inserted in frame in the pB6 yeast expression vector derived from the original pAS2⌬⌬ (20). A random-primed, size-selected (mean insert size 800 bp) cDNA library of differentiated human brown adipocyte PAZ6 cells (21) was constructed in the pB6 vector derived from the original pGADGH vector. Plasmids able to rescue yeast growth were amplified by PCR and sequenced at their 5Ј and 3Ј junctions on a PE3700 sequencer. The resulting sequences were used to identify the corresponding interactors in the GenBank TM data base (NCBI) using a fully automated procedure.
Plasmid Constructions and Cell Culture-The GW1-HA-MUPP1 plasmid containing the coding region of the rat MUPP1 as well as GST fusion constructs expressing PDZ1-3, PDZ4-5, PDZ6, PDZ7, or PDZ8-9 were a gift from Dr. Javier and have been described elsewhere (22,23). GST fusion constructs containing PDZ9, PDZ10, or PDZ11-12 were a gift from Dr. Mancini (24). GST constructs were expressed in Escherichia coli and purified on immobilized glutathione according to standard protocols. The FLAG-tagged MT 1 receptor has been described elsewhere (25). Mutants of the MT 1 PDZ binding motif were generated by PCR. Human CCR5 and human SSTR2 somatostatin receptor expression vectors were generously given by Drs. Marullo (Paris, France) (26) and Bousquet (Toulouse, France) (27), respectively. HEK 293 cells were grown and transfected as described (25).
Peptide Affinity Chromatography-Synthetic peptides corresponding to the C-tail of MT 1 (residues 294 -350) or MT 2 (residues 305-364) tagged with His 6 were provided by Servier (Suresnes, France). Brains harvested from C57BL6 mice were washed in phosphate-buffered saline; crushed in solubilization buffer composed of 20 mM NaH 2 PO 4 , 10 mM CHAPS, 150 mM NaCl, 2 mM Na 3 VO 4 , 10 mM NaF, protease inhibitor mixture EDTA-free (Roche Diagnostics, Meylan, France), pH 8; and incubated for 2 h at 4°C. The homogenates were centrifuged at 10,000 ϫ g (1 h, 4°C), the supernatants were collected, and 10 mg of solubilized brain proteins were incubated overnight at 4°C with 20 l of Ni 2ϩ -nitrilotriacetic acid beads (Qiagen, Courtaboeuf, France) coated with His 6 -tagged MT 1 or MT 2 COOH-terminal peptides (300 -350 g), in the presence of 20 mM imidazole to reduce nonspecific binding. Bound proteins were eluted by 2% SDS, separated by SDS-PAGE, and checked by Western blotting for the presence of MUPP1 using anti-MUPP1 antibodies.
Immunofluorescence-HEK293 cells stably expressing the FLAG-MT 1 receptor and transfected by HA-MUPP1 DNA were fixed in phosphate-buffered saline containing 4% paraformaldehyde and 0.2% Triton X-100 for 30 min. Monoclonal anti-FLAG M2 antibody and polyclonal anti-HA antibody HA.11 (BAbCo, Richmond, CA) were applied, followed by rhodamine-tagged anti-mouse and fluorescein isothiocyanatetagged anti-rabbit antibodies. Cells were examined by confocal fluorescence microscopy.
ELISA-The ELISA was adapted from Stricker et al. (30). Lysates from E. coli expressing different MUPP1 regions fused to GST were coated on ELISA plates. The amount of lysate used for coating corresponded to 50 l of a 0.1 M NaHCO 3 solution, pH 9.5, containing a 20 g/ml concentration of the fusion protein (as evaluated by Coomassie Blue staining of SDS-PAGEseparated lysate proteins). HEK293 cells transfected by the different FLAG-tagged receptor plasmids were lysed, and 50 l of lysate were added to the wells for a 3-h incubation at room temperature. ELISA was carried out with anti-FLAG M2 antibody (2 g/ml), followed by horseradish peroxidase-coupled anti-mouse antibody. Staining was performed with 2-2Ј-azinobis(3-ethylbenzthiazoline-6-sulfonic acid), and the color intensity was measured at 450 nm.
To determine the affinity of the receptor/PDZ interaction, 50 l of 0.1 M NaHCO 3 , pH 9.5, buffer containing 2 g/ml GST fusion proteins were adsorbed on ELISA plates. The amount of receptor was quantified by [ 125 I]MLT binding, and then the cells were lysed, and the solubilized receptor concentration was adjusted to 5 nM. Dilutions of the solubilized receptor were incubated on preadsorbed GST-fused PDZ domains, and the ELISA was performed as described above.
Receptor Internalization-For constitutive internalization, suspended HEK-FLAG-MT 1 cells were incubated with anti-FLAG M2 antibody (1 h, 4°C). Aliquots were then incubated at 37°C for variable times. Cells were then transferred to ice, incubated with fluorescein isothiocyanate-coupled secondary antibody, and fixed with paraformaldehyde. The fluorescence was measured by fluorescence-activated cell sorting.
For agonist-stimulated internalization, cells in 6-well plates were incubated in culture medium containing 0.1 M MLT for variable times (0.5-3 h) and then transferred to ice. Cells were suspended and incubated with anti-FLAG M2 antibody, followed by fluorescein isothiocyanate-coupled secondary antibody to label surface receptor. Fluorescence was measured by fluorescence-activated cell sorting after fixation with paraformaldehyde.
Receptor Degradation-HEK-FLAG-MT 1 cells were plated in polylysine-coated 24-well plates. The next day, protein synthesis was inhibited by treatment for 45 min with 100 M cycloheximide. Incubation was continued for 4 h in the presence or absence of agonist (0.1 M MLT). Cells were transferred to ice, fixed, and permeabilized by treatment with cold ethanol at Ϫ20°C for 10 min. ELISAs were performed as described above.
cAMP Assay-Cyclic AMP levels were determined by HTRF using the Cisbio "cAMP femto2" kit according to the manufacturer's instructions. Samples were analyzed with a PheraStar apparatus (BGM Labtech, Offenburg, Germany).
Bioluminescence Resonance Energy Transfer (BRET) Assay, Luminescence, and Fluorescence Measurements-BRET experiments, luminescence, and fluorescence measurements were performed as described on adherent cells (25).
siRNA Treatment-siRNAs corresponding to region 955-973 of the human MUPP1 cDNA were synthesized (Eurogentech, Seraing, Belgium) and transfected with Lipofectamine 2000 (Roche Applied Science) according to the supplier's instructions. Negative control siRNA Alexa Fluor 488 was from Qiagen (catalog number 1022563).
Statistical Analysis-Results were analyzed by PRISM (GraphPad Software Inc., San Diego, CA). Data are expressed as mean Ϯ S.E. Student's t test was applied for statistical analysis.

RESULTS
The C-tail of MT 1 Specifically Interacts with MUPP1-To identify interacting partners of the human melatonin MT 1 and MT 2 receptors, we conducted yeast two-hybrid screens using the C-tails or a fusion of the intracytoplasmic loops i2 and i3 and the C-tail of these receptors as baits (Fig. 1A). The screens were performed against a cDNA library of differentiated human brown adipocytes, a cellular context known to express functional melatonin receptors (31). Among the positive clones that interacted with the two baits containing the C-tail of MT 1 , we identified 14 sequences corresponding to the multi-PDZ protein MUPP1 (Fig. 1B), which is composed of 13 PDZ domains (32). Sequence alignment of these clones unraveled a common region corresponding to nucleotides 4548 -5189 (amino acids 1516 -1729) in the coding sequence of the human MUPP1 encompassing part of PDZ9 and PDZ11 and the entire PDZ10 domain. The identification of a PDZ domain-containing protein as a specific interacting partner of the C-tail of MT 1 is consistent with the presence of a canonical class III PDZ domain-binding motif DSV-COO Ϫ at the COOH terminus. To further test the specificity of our yeast two-hybrid screen, we used the C-tail of the ␤ 2 -adrenergic receptor as bait. Despite the presence of a functional PDZ binding motif at the C-tail of this receptor, we were unable to recover MUPP1, indicating the high specificity of the MT 1 /MUPP1 interaction (data not shown).
To confirm the interaction between the C-tail of MT 1 and MUPP1 in a different experimental setting, we incubated MT 1 or MT 2 C-tails, which were chemically synthesized with a His 6 tag and immobilized on beads, with whole brain lysates from mice. We then tested for the presence of MUPP1 among the retained proteins with anti-MUPP1 antibodies (Fig. 1C). MUPP1 was bound to the C-tail of MT 1 but not to that of MT 2 , confirming the specificity of the interaction. Affinity and Molecular Determinants of the MT 1 /MUPP1 Interaction-To delineate the molecular determinants of the MT 1 /MUPP1 interaction in vitro, we used receptor mutants of the PDZ binding motif and GST fusion proteins of the different PDZ domains of MUPP1 ( Fig. 2A) to perform GST pull-down experiments (see supplemental Fig. S1 for GST constructs). Only GST-PDZ10 was able to specifically precipitate the NH 2terminally FLAG-tagged MT 1 , confirming that this domain is involved in the interaction. Mutation of the very last valine residue into alanine (MT 1 -DSA) completely abolished the interaction, as did the deletion of the entire C-tail (MT 1 -⌬cter). Mutation of the aspartate residue at position Ϫ2 (MT 1 -ASV) reduced the amount of precipitated receptor.
To confirm these interactions in a more quantitative assay, we used an ELISA set-up where similar quantities of the GST-PDZ fusion proteins were immobilized (Fig. 2B) and incubated with equivalent quantities of solubilized receptors, as confirmed by Western blot (not shown). Only PDZ10 was able to interact with MT 1 . Although the MT 1 -ASV mutant was par-tially retained, MT 1 -⌬cter and MT 1 -DSA completely failed to interact with PDZ10. We then determined the dissociation constant (K d ) of the MT 1 /PDZ10 interaction with the ELISA (Fig. 2C). Whereas binding to immobilized PDZ10 was saturable and of high affinity (K d ϭ 3.8 Ϯ 0.7 nM), only background binding was detected for PDZ9-coated wells.
Overall, our in vitro data show that MUPP1 interacts with MT 1 with high affinity and that this interaction involves the PDZ10 domain of MUPP1 and the COOH-terminal DSV motif of MT 1 . The interaction mainly depends on the COOH-terminal valine residue, although other amino acids, such as the aspartate residue at position Ϫ2, also appear to be involved.
Interaction between MT 1 and MUPP1 in Mammalian Cells-To investigate the interaction of MUPP1 with MT 1 in mammalian cells, we expressed HA-tagged MUPP1 in HEK-FLAG-MT 1 cells (100 fmol/mg of protein) (25). As shown by immunofluorescence staining with anti-HA antibodies, a small amount of MUPP1 was present throughout the cytoplasm, but the majority was located at the plasma membrane, where it colocalized with MT 1 (Fig. 3, A-C). The interaction of both proteins was addressed in intact cells by co-immunoprecipitation experiments in HEK-FLAG-MT 1 cells. As shown in Fig. 3D, anti-FLAG antibodies coprecipitated endogenous MUPP1.
To demonstrate the presence of the protein complex in native tissue, we performed immunoprecipitation studies with ovine pituitary PT tissue samples known to express significant amounts of endogenous MT 1 receptors (33) and MUPP1 (data not shown). Receptors were labeled with the specific [ 125 I]MLT radioligand, and protein complexes were solubilized and immunoprecipitated with anti-MUPP1 antibodies. As shown in Fig. 3E, significant amounts of radiolabeled MT 1 were precipitated in the presence of anti-MUPP1 as compared with irrelevant control antibodies, demonstrating the existence of MT 1 -MUPP1 complexes in native tissue.
To further confirm the specificity of the interaction, we transiently expressed similar levels of FLAG-MT 1 , FLAG-MT 1 -ASV, FLAG-MT 1 -DSA, or FLAG-MT 1 -⌬cter with HA-MUPP1 (Fig. 3F). In agreement with our in vitro interaction data, HA-MUPP1 was readily precipitated from cells expressing FLAG-MT 1 , to a lesser extent from cells expressing FLAG-MT 1 -ASV, and not at all from cells expressing FLAG-MT 1 -DSA, FLAG-MT 1 -⌬cter, or HA-MUPP1 alone (Fig. 3G). The amount of co-precipitated MUPP1 did not change upon stimulation (up to 10 min) with 10 nM melatonin (MLT), the natural hormone of MT 1 (Fig. 3H). Taken together, these results show that MUPP1 and MT 1 interact constitutively in HEK 293 cells and in the PT.
Effect of PDZ10 on MT 1 Internalization and Degradation-Interaction of GPCRs with PDZ proteins has been shown to stabilize receptors by inhibiting either their constitutive or agonist-promoted internalization (8,34) or by interfering with receptor degradation (4,5). To study the role of MUPP1 in MT 1 internalization and degradation, we used the isolated PDZ10 domain of MUPP1 as a dominant negative to disrupt the MUPP1/MT 1 interaction. As shown in Fig. 4A, the amount of MUPP1 associated with the receptor is, as expected, strongly decreased in the presence of PDZ10. We first determined the rate of constitutive internalization of MT 1   presence of PDZ10. Fig. 4B shows that the rate of internalization is not affected in the presence of PDZ10; in addition, the MT 1 -DSA mutant, unable to interact with MUPP1, presents equivalent constitutive internalization characteristics. Similarly, the binding of MUPP1 to MT 1 appears not to alter ligandinduced receptor internalization by 100 nM MLT, since equivalent internalization kinetics and maximal internalization of ϳ60% within 3 h were observed in the absence and presence of PDZ10 (Fig. 4C)    FLAG-MT 1 cells with a saturating MLT concentration (10 nM) resulted in the expected decrease of forskolin-stimulated cAMP levels of about 60% (Fig. 5A). Expression of the dominant negative PDZ10 in these cells strongly attenuated the inhibitory effect of MT 1 on cAMP production, indicating that binding of MUPP1 to MT 1 is necessary for efficient coupling of the receptor to the cAMP pathway. To verify that the displaced protein corresponds to MUPP1, we treated cells with MUPP1-specific siRNA molecules. MUPP1 expression was decreased by nearly 80% compared with the control siRNA (Fig. 5B), and the MLTinduced decrease in cAMP production was abolished to the same extent as in cells expressing the dominant negative PDZ10 domain (Fig. 5A). Importantly, control siRNA molecules did not interfere with the inhibitory effect of MLT on forskolin-stimulated cAMP levels. This strongly suggests that binding of MUPP1 to the MT 1 C-tail is essential for its efficient coupling to the AC pathway. To further support this idea, we tested the MT 1 -DSA mutant, which is unable to interact with MUPP1. Consistently, 10 nM MLT stimulation of HEK 293 cells stably expressing MT 1 -DSA receptors at similar expression levels as HEK-FLAG-MT 1 cells did not inhibit forskolin-stimulated cAMP levels (Fig. 5, A and C). To further assess the specificity of the MUPP1 interaction with MT 1 on the AC pathway, we measured the effect of MUPP1 knockdown on two other G i -coupled receptor, the human SSTR2 somatostatin receptor and the CCR5 chemokine receptor. Indeed, MUPP1 siRNA treatment of cells transiently transfected with each of these receptors had no effect on the cAMP inhibition elicited by 10 nM somatostatin or 100 nM RANTES, respectively (Fig. 5D). In addition, the siRNA treatment had no effect on the cAMP response observed for the ␤ 2 -adrenergic receptor, a G s -coupled receptor, stimulated by 1 M isoproterenol (Fig. 5D). Overall, our results show that binding of MUPP1 to the COOH-terminal DSV motif of MT 1 plays a central role in MT 1 signaling through the AC pathway.
G i Coupling and High Affinity Agonist Binding to MT 1 Depends on the Presence of MUPP1-We hypothesized that the inability of FLAG-MT 1 to couple to the AC pathway in the absence of MUPP1 may be due to a defect in G i coupling to the receptor. In agreement with previous reports (29), solubilization of MT 1 under mild conditions preserved the interaction with G␣ i proteins (Fig. 6A). In contrast, coexpression of PDZ10 strongly decreased the quantity of coprecipitated G i . Consistently, in cells stably expressing the MUPP1 binding-deficient FLAG-MT 1 -DSA mutant, G␣ i proteins were undetectable under these conditions. Decreased G protein coupling was specific for G␣ i , since G␣ q was readily precipitated under all conditions. These results show that the inability of MT 1 to signal through the AC pathway in the absence of MUPP1 is most likely due to its reduced G i coupling capacity.
Previous studies have shown that high affinity agonist binding to MT 1 depends on the coupling of the receptor to G i proteins (36). Accordingly, decreased high affinity agonist binding would be expected when MUPP1 is displaced from MT 1 . We therefore incubated cell membranes prepared from HEK-FLAG-MT 1 cells in the presence or the absence of purified PDZ9 or PDZ10 and determined agonist binding using the radiolabeled MLT receptor agonist [ 125 I]MLT (Fig. 6B). In the absence of added PDZ domain, MT 1 bound [ 125 I]MLT with the expected high affinity (K D ϭ 136 Ϯ 8 pM). Similar results were obtained in the presence of PDZ9 (K D ϭ 160 Ϯ 15 pM). In contrast, in the presence of PDZ10, a significantly lower affinity was observed (K D ϭ 378 Ϯ 79 pM) (p Ͻ 0.05; MT 1 alone versus MT 1 ϩ PDZ10). The number of binding sites was not affected in any of the conditions (B max ϭ 100 Ϯ 9, 94 Ϯ 9, and 118 Ϯ 13% for MT 1 , MT 1 ϩ PDZ9, and MT 1 ϩ PDZ10, respectively). The lower affinity for [ 125 I]MLT in the presence of PDZ10 is consistent with the lower affinity of the MT 1 -DSA mutant (K D ϭ 321 Ϯ 50 pM), which is devoid of MUPP1 binding. These results indicate that binding of MUPP1 to the MT 1 C-tail participates in high affinity agonist binding, most likely by stabilizing G i binding to the agonist-activated receptor. To show the formation of a trimeric complex between MT 1 , G i , and MUPP1, we used the BRET assay, which allows detection of real time protein interactions in a cellular context (54). The previously described G␣ i1 -91-Rluc fusion protein (energy donor) (37) was coexpressed with a fragment of MUPP1 comprising PDZ domains 9 -13 fused at its amino-terminal tail to the energy acceptor YFP (YFP-PDZ9 -13).
Expression of both proteins in HEK 293 cells revealed a nonspecific interaction between G i and MUPP1 (weak signal, linear and nonsaturating behavior of BRET donor saturation curve) (Fig. 6C). In contrast, when YFP-PDZ9 -13 and G␣ i1 -91-Rluc were expressed in HEK-FLAG-MT 1 cells, G i interacted with high affinity with MUPP1 as shown by the hyperbolic and saturable behavior of the BRET donor saturation curve. Stimulation of cells with MLT did not alter BRET signals (not shown). The high affinity interaction between G i and YFP-PDZ9 -13 was specific, since no interaction was observed for YFP-PDZ1-4 and YFP-PDZ5-8 constructs in HEK-FLAG-MT 1 cells. To evaluate the functional importance of the PDZ9 -13 subdomain on the AC pathway, we co-transfected FLAGtagged MT 1 and PDZ9 -13 in HEK293 cells devoid of MUPP1 (MUPP1 siRNA-treated). Although the MLT-promoted inhibition of Fsk-stimulated cAMP production was abolished in MUPP1 siRNA-treated cells as expected, simultaneous expression of PDZ9 -13 did not allow us to reestablish the MT 1 response (Fig. 6D). Indeed, PDZ9 -13 by itself inhibits the MT 1 response, as was observed for the single PDZ10 fragment. Collectively, our results indicate that the PDZ9 -13 fragment of MUPP1 is necessary for high affinity binding of G␣ i to MT 1 but not sufficient to reconstitute a functional system to modulate the AC pathway.
Disruption of MUPP1/MT 1 Interaction Affects MLT-stimulated ERK Activation-Many GPCRs activate the mitogenactivated protein kinase pathway, although through several different mechanisms (38). A common mechanism of mitogenactivated protein kinase activation involves G ␤ ␥ subunits that are released from G i proteins upon receptor activation. To determine whether the mitogen-activated protein kinase activation by MT 1 also involves G i proteins, we treated HEK-FLAG-MT 1 cells with pertussis toxin (PTX; 10 ng/ml), which is known to inactivate G i proteins. As shown in Fig. 7A, 10 nM MLT-promoted ERK phosphorylation was abolished in PTXtreated cells, indicating that ERK activation by MT 1 is indeed G i -dependent. We studied the kinetics of ERK phosphorylation in the presence and absence of PDZ10, to determine the effect of MUPP1 on ERK activation. In both cases, the amplitude and time course of ERK activation was similar with maximal effects at 5 min of MLT stimulation (Fig. 7, B and C). Similar B max and EC 50 values were obtained for the MT 1 wild-type and the MUPP1 binding-deficient MT 1 -DSA mutant (Fig. 7, C and D). These results suggest that G i -dependent ERK signaling of MT 1 is not altered in the absence of MUPP1 binding. Such an observation is in apparent contradiction to our data on the cAMP pathway. To account for the different effect of MUPP1 on both pathways, we then hypothesized that the AC pathway can be more sensitive to alterations of the G i coupling to MT 1 than the mitogen-activated protein kinase pathway. We therefore decreased the amount of functional G i proteins by incubating cells with increasing doses of PTX and determined the degree of MLT-promoted ERK phosphorylation of wild-type MT 1 or the MT 1 -DSA mutant (Fig. 7E). Maximal doses of PTX inhibited ERK activation for both receptors to a similar extent. This shows that the ERK activation of the MT 1 -DSA mutant also depends on G i protein activation and excludes the possibility that ERK activation becomes G q -dependent in the absence of functional G i coupling. At submaximal PTX concentrations, ERK activation by the MT 1 -DSA mutant and the wild-type receptor was clearly different (IC 50 ϭ 0.10 and 0.02 ng/ml for wild-type MT 1 and MT 1 -DSA, respectively). The left shift of the dose-response curve of ERK phosphorylation of the MT 1 -DSA mutant indicates that ERK activation becomes indeed more sensitive to the amount of active G i proteins in the absence of MUPP1 binding to MT 1 . Taken together, both G idependent signaling pathways, the AC and the ERK pathways, are affected in the absence of MUPP1 binding to MT 1 .

DISCUSSION
We show here that the PDZ protein MUPP1 regulates GPCR signaling by stabilizing the coupling of G proteins to their cognate receptors. This substantially extends the role of MUPP1 in GPCR function, which was previously limited to the subcellular targeting and trafficking of receptors. The PDZ10 domain of MUPP1 is the only one out of its 13 PDZ domains that specifically associates with the C-tail of the MT 1 but not with that of the MT 2 and the ␤ 2 -adrenergic receptor. The valine of the canonical class III PDZ domain binding motif DSV of the C-tail of MT 1 is crucial for the high affinity interaction. Using several approaches to hamper the MUPP1/MT 1 association (overexpression of PDZ10 as a MUPP1 competitor, extinction of MUPP1 expression by siRNA, or expression of the MUPP1 binding-deficient MT 1 -DSA mutant), we were able to destabilize the interaction of G i proteins with MT 1 and to interfere with G i -dependent signaling pathways. Indeed, MLT-stimulated AC inhibition and ERK activation were differentially affected when the binding of MUPP1 to MT 1 was impaired.
PDZ scaffolds have been shown to modulate GPCR signaling in different ways. Binding of GPCRs to PDZ scaffolds has been reported to modulate the amount of receptor at the cell surface and consequently the amplitude of the functional response by altering the receptor's trafficking and stability (2). Other examples highlight the importance of the scaffolding properties of PDZ domain proteins in GPCR signaling. For instance, simultaneous binding of the PDZ scaffold GIPC to the G i -coupled D 2 dopamine receptor and to RGS19 favors the GDP/GTP exchange of G i by RGS19 (17). Moreover, binding of NHERF-1 to GPCRs and phospholipase C␤3 enhances the signaling efficiency of the phospholipase C␤/Ca 2ϩ pathway (13)(14)(15). Modulation of MT 1 signaling by MUPP1 probably involves a previously unappreciated regulatory mechanism. Disruption of the interaction of MUPP1 with the C-tail of MT 1 decreased signaling of MT 1 through the AC and the ERK pathway. Importantly, the interaction between G i proteins and MT 1 was also destabilized under these conditions, suggesting that MUPP1 regulates MT 1 signal transduction by stabilizing the receptor-G i protein complex. This indicates physical proximity between MUPP1 and G i proteins, as supported by our BRET experiments. However, more complex mechanisms cannot be excluded, since PDZ9 -13, which is able to restore high affinity binding of G i to MT 1 , is insufficient to reconstitute functional coupling to AC. Some PDZ proteins are indeed able to physically interact with heterotrimeric G proteins, as recently reported for PSD95, SAP97, and Veli2, which interact with G␥ 13 (39). Furthermore, ␣-syntrophin has been shown to bind to G ␤ ␥ through its PDZ domain (40). Although G␤ subunits typically code for the "LWL" or "IWN" sequence at their C-tails, several G␥ subunits (e.g. G␥ 4 , G␥ 12 , and G␥ 13 ) have the COOH-terminal "TIL" sequence, which corresponds to the class I ((S/T)XL) PDZ domain recognition sequence. Alternatively, MUPP1 and G i may be physically linked through a third protein. According to our BRET experiments, the domain that promotes G i binding to MT 1 appears to localize between PDZ9 and PDZ13, since this part of MUPP1 is sufficient to stabilize the ternary complex between MUPP1, G i , and MT 1 .
The association of MUPP1 with MT 1 may participate in the high stability of the MT 1 -G i protein complex. G i has been shown to be precoupled to MT 1 in its inactive form and to remain stably associated upon agonist stimulation despite the presence of high GTP concentrations in intact cells (29,36). Destabilization of the MT 1 -G i complex in the absence of MUPP1 has different effects on the signaling capacities of MT 1 . Whereas MUPP1 is necessary for efficient coupling of MT 1 to the AC pathway, the PTX-sensitive ERK activation is only moderately affected (only when the amount of G i proteins becomes limiting). This highlights the potential regulatory role of MUPP1 in the modulation of MT 1 signaling.
MT 1 is not the only GPCR that binds to MUPP1. The serotonin 5-HT 2C receptor was the first GPCR that has been shown to interact with MUPP1 (41). Although possibly regulated by the phosphorylation state of the receptor, the functional role of 5-HT 2C binding to PDZ10 of MUPP1 still remains poorly defined (42). Whereas MT 1 and 5-HT 2C bind to the same PDZ domain (PDZ10), their G protein coupling profiles are different. MT 1 couples preferentially to G i , and 5-HT 2C couples pref- erentially to G q . This difference may explain why the disruption of MUPP1 binding to 5-HT 2C does not abolish receptor signaling as for MT 1 . This interpretation is consistent with our observation that only coupling to G i and not G q is affected in the absence of MUPP1 binding to MT 1 . The metabotropic ␥-aminobutyric acid B (GABA B ) receptor 2 has also been reported to interact with MUPP1 (5). In contrast to MT 1 and 5-HT 2C , metabotropic GABA B receptor 2 binds to the PDZ13 domain of MUPP1. Accordingly, binding to this PDZ domain has different consequences for receptor function, namely the stabilization of the metabotropic GABA B receptor 2 and the enhancement of the duration of receptor signaling. Although other PDZ proteins have also been shown to stabilize their respective binding partners, the underlying mechanism remains unknown. The C-tails of further GPCRs have been reported to bind to MUPP1 in in vitro assays, such as 5-HT 2A , 5-HT 2B (41), and the mouse SSTR2b somatostatin receptor splice variant (43). The in vivo relevance and the functional consequences of these interactions remain to be determined.
The constitutive nature of the MT 1 /MUPP1 interaction raises the question of its regulation. The PDZ10 domain of MUPP1 is the target of multiple proteins (see Table 1). This implies that the selection of PDZ10 binding partners is highly competitive and will depend on their relative affinities and expression levels. Similarly, other PDZ proteins can bind to MT 1 6 and are consequently expected to compete with MUPP1 for MT 1 binding. This may be functionally relevant, since different PDZ proteins can differentially modulate receptor function (44). Furthermore, the MT 1 /MUPP1 interaction may be regulated by the expression level of MUPP1 itself. Despite the widespread expression of MUPP1 in the brain (45) and at multiple peripheral sites, its expression may vary depending on the cellular context and different pathological conditions. For instance, human keratinocyte infection by papillomavirus HPV18 E6 has been shown to induce epithelial hyperplasia and massive down-regulation of MUPP1 by targeting MUPP1 to the proteasome (23,46,47). During adenovirus infection by Ad9E4, responsible for estrogen-dependent mammary tumors in rat, the viral protein ORF1 has been shown to promote cytoplasmic sequestration of MUPP1 (23,48). In both cases, MUPP1 is removed far from its potential membrane partners, thus impeding any interaction with them. Regulation of the MT 1 /MUPP1 interaction is likely to occur under these circumstances, since functional MT 1 expression has been shown in keratinocytes (49) and mammary tumors (50). Recently, MUPP1 has been shown to be robustly up-regulated by hypertonicity and to be important in the osmotic stress response in tight junctions of kidney cells (51). Finally, altered MUPP1 expression levels have been shown in mice with high predisposition to alcohol and barbiturate physical dependence and withdrawal (52). Taken together, MUPP1 expression levels appear to be highly regulated in several physiological and pathological situations.
In conclusion, our study extends the previously panoply of known functions of MUPP1 on GPCR physiology. MUPP1 regulates G protein-dependent GPCR signaling by directly stabilizing the receptor-G protein complex, which may explain the previously reported high stability of the MT 1 -G i complex. Future studies will concentrate on the still largely unexplored functions of the other PDZ domains of MUPP1 on GPCR function.