A Generic Approach for the Purification of Signaling Complexes That Specifically Interact with the Carboxyl-terminal Domain of G Protein-coupled Receptors*S

G protein-coupled receptors (GPCRs) constitute the largest family of membrane receptors and are major drug targets. Recent progress has shown that GPCRs are part of large protein complexes that regulate their activity. We present here a generic approach for identification of these complexes that is based on the use of receptor subdomains and that overcomes the limitations of currently used genetics and proteomics approaches. Our approach consists of a carefully balanced combination of chemically synthesized His6-tagged baits, immobilized metal affinity chromatography, one- and two-dimensional gel electrophoresis separation and mass spectrometric identification. The carboxyl-terminal tails (C-tails) of the human MT1 and MT2 melatonin receptors, two class A GPCRs, were used as models to purify protein complexes from mouse brain lysates. We identified 32 proteins that interacted with the C-tail of MT1, 14 proteins that interacted with the C-tail of MT2, and eight proteins that interacted with both C-tails. Several randomly selected proteins were validated by Western blotting, and the functional relevance of our data was further confirmed by showing the interaction between the full-length MT1 and the regulator of G protein signaling Z1 in transfected HEK 293 cells and native tissue. Taken together, we have established an integrated and generic purification strategy for the identification of high quality and functionally relevant GPCR-associated protein complexes that significantly widens the repertoire of available techniques.

G protein-coupled receptors (GPCRs) 1 constitute the largest family of membrane receptors with more than 800 members (1,2). By binding to a great variety of ligands (photons, odorants, amino acids, nucleotides, peptides, proteins, and lipids), GPCRs are key receptors of numerous physiological processes such as neurotransmission, cell metabolism, secretion, cell differentiation, and growth and are targeted by about half of the drugs prescribed for human diseases (3). It is now well established that GPCRs do not only couple to heterotrimeric G proteins but can also physically associate with other less well known intracellular proteins regulating receptor trafficking, subcellular localization, signaling, and desensitization (4,5). Intracellular proteins can interact directly or indirectly, via adaptor proteins, with intracellular receptor domains. Among these domains, the carboxyl-terminal tail (Ctail) is considered a key domain able to recruit intracellular proteins in large submembrane signaling networks (6,7). Several approaches have been described in the literature to identify proteins that interact with GPCRs. The yeast twohybrid assay has been used to screen for proteins that bind to cytosolic domains of GPCRs (8 -10). However, this system, which is very sensitive for the detection of in vivo proteinprotein interactions, has shown several limitations including the generation of many false positives and negatives, the detection of only binary interactions, and the non-physiological relevance of the identified interactions (e.g. occurring in the yeast nucleus). Moreover this approach is not designed to identify protein complexes. To overcome these limitations, proteomics approaches were developed based on peptide affinity chromatography coupled to protein identification by mass spectrometry. The first generation used the entire receptor C-tail expressed as a GST fusion protein in Escherichia coli to purify the interactome of the C-tail from brain lysates (11). Although some interacting proteins have been identified for the 5-HT 2C receptor, several limitations prevented a more general application of this approach. Indeed the amount of nonspecifically retained proteins is high because of the presence of contaminating bacterial proteins, full-length and truncated GST fusion proteins, and proteins that nonspecifically bind to the GST carrier. In addition, the complex sample has to be separated by two-dimensional (2D) electrophoresis before mass spectrometry analysis of the recruited proteins, thus minimizing the number of potential candidates because of the limitations inherent to 2D electrophoresis for hydrophobic, basic, and large proteins. The second generation of peptide affinity chromatography is based on the use of short synthetic peptides encompassing a specific interaction motif that recruit only the proteins that interact with this specific motif. The amount of nonspecific proteins is indeed largely reduced in this approach, and the pattern of specific interaction partners is much less complex. Consequently this approach has been successfully used for several GPCRs and is expected to be applied to multiple GPCR interaction motifs (12,13). However, this evolution of peptide affinity chromatography does not respond to the initial task, the identification of the interactome of the entire C-tail, without prior knowledge of specific interaction motifs.
In the present study, we present an improved peptide affinity chromatography that has the potential to become the first generic approach for the purification of GPCR C-tailassociated protein complexes. We combined the use of chemically synthesized His 6 -tagged peptides encompassing the entire receptor C-tail combined with metal affinity immobilization on a Ni-NTA matrix to recover protein complexes from mouse brain lysates. The C-tails of the MT 1 and MT 2 melatonin receptors, typical class A GPCRs, were used as model receptors. Major features of this approach are the low nonspecific binding, the high integrity of recovered complexes, and the compatibility with 1D and 2D electrophoresis. We report the identification of 40 and 22 proteins that specifically associate with the C-tails of MT 1 and MT 2 , respectively. To demonstrate the functional relevance of the identified protein complexes, we selected the interaction between MT 1 and the regulator of G protein signaling Z1 (RGSZ1) for further characterization in transfected HEK 293 cells and native tissue.

EXPERIMENTAL PROCEDURES
Peptide Affinity Chromatography-Peptides encompassing the Ctails of the MT 1 and MT 2 receptors were chemically synthesized by NeoMPS (Strasbourg, France) with a His 6 tag at the amino terminus.
Brains of C57/Bl6 mice were crushed in 20 ml of buffer containing 20 mM NaH 2 PO 4 , 2 mM Na 3 VO 4 , 10 mM NaF, protease inhibitor mixture EDTA-free (Roche Applied Science), pH 8.0, using an Ultra-Turrax T25 (Janke-Kunkel). CHAPS (10 mM final concentration) and NaCl (150 mM final concentration) were added, and the homogenates were incubated for 3 h at 4°C under gentle end-over-end mixing. After centrifugation (10,000 ϫ g for 1 h at 4°C), the supernatants were collected, and the protein concentration was determined by BCA assay (Pierce).
Supernatants (10 mg of solubilized brain proteins) were incubated overnight at 4°C with 20 l of Ni-NTA-agarose beads (300 -350 g of immobilized peptide) in the presence of 20 mM imidazole to reduce nonspecific binding. The beads were washed five times in washing buffer (20 mM NaH 2 PO 4 , 2 mM Na 3 VO 4 , 10 mM NaF, protease inhibitor mixture EDTA-free, 10 mM CHAPS, 150 mM NaCl, 20 mM imidazole, pH 8), and proteins retained by affinity were eluted with either 50 l of 2% SDS in PBS (95°C for 10 min) for BCA measurements, 1D electrophoresis, and immunoblotting or 200 l of rehydration buffer (8 M urea, 2 M thiourea, 4% CHAPS) for 45 min at room temperature for 2D electrophoresis.
For 2D electrophoresis, eluted proteins in rehydration buffer were first separated according to their isoelectric point along a nonlinear IPG strip (pH 3-11 nonlinear, 11 cm long) using the IPGphor apparatus (GE Healthcare). DeStreack Reagent (15 mg/ml of rehydration solution; GE Healthcare) and ampholytes (1% IPG buffer; GE Healthcare) were added to the sample, and then sample loading for the first dimension was performed by passive in-gel rehydration. Focalization was stopped when 27,000 V-h were reached. After the first dimension, the IPG strips were equilibrated for 15 min in buffer A (6 M urea, 2% SDS, 30% glycerol, 50 mM Tris/HCl, pH 8.6) containing 1% DTT and then for 15 min in buffer A containing 4.7% iodoacetamide. For the second dimension, the strips were loaded on 8 -18% SDS-polyacrylamide gels. The gels were then silver-stained according to Rabilloud and Charmont (14).
In-gel Trypsin Digestion-Spots of interest were excised from the gel, washed with 200 mM ammonium bicarbonate (ABC) and immediately destained according to Gharahdaghi et al. (15). In-gel trypsin digestion was carried out as described in a protocol based on the ZipTip Plate (Millipore) with minor modifications. After destaining, spots were rinsed three times with water and shrunk with 50 mM ABC, 50% ACN for 20 min at room temperature. Gel pieces were dried using 100% ACN for 15 min and then incubated in 50 mM ABC containing 10 mM DTT for 1 h at 56°C. The solution was then replaced by 55 mM iodoacetamide in 50 mM ABC for 30 min in the dark at room temperature. The gel pieces were washed twice with 50 mM ABC, finally shrunk with 25 mM ABC, 50% ACN for 30 min, and dried using 100% ACN for 10 min. Gel pieces were rehydrated in 20 l of 40 mM ABC, 10% ACN, pH 8.0, containing 12.5 g/ml sequencing grade modified trypsin (Promega). Proteins were digested overnight at 37°C. After digestion, the gel pieces were shrunk with 100% ACN, and peptides were extracted with 0.2% TFA. Peptides were then desalted using C 18 phase on a ZipPlate. Two elutions were performed successively to recover products from the C 18 phase, first with 50% ACN, 0.1% TFA and then using 90% ACN, 0.1% TFA. Pooled elutions were concentrated using a vacuum centrifuge (Ep-pendorf), and generated peptides were redissolved in 3 l of 1% formic acid before mass spectrometry analysis.
Protein Identification by MALDI-TOF Mass Spectrometry-Digested samples were spotted directly onto a 96 ϫ 2 MALDI plate (Applied Biosystems) using the dried droplet method (0.5 l of the sample mixed on the plate with 0.5 l of ␣-cyano-4-hydroxycinnamic acid matrix in ACN/water/TFA (50:50:0.1%). Droplets were allowed to dry at room temperature. Mass spectrometry analysis were performed on a MALDI-TOF Voyager DE-PRO mass spectrometer (Applied Biosystems) and acquired in positive ion reflector mode. Spectra were obtained in a reflectron-delayed extraction in source over a mass range of 500 -3500 daltons. Raw spectra were treated automatically according to the Mascot wizard algorithm (version 1.1.2.0, Matrix Science) using the default parameters with the exception of the maximum peak count reduced to a value corresponding to the expected number of tryptic peptides according to the experimental molecular weight of the protein. The protein database was restricted to the rodent entries of the comprehensive nonredundant protein sequence database (NCBInr database; National Center for Biotechnology Information) (from version October 2005, 118,719 entries to version October 2007, 217,820 entries). When non-significant scoring was obtained, a manual treatment was performed according to the following explanations. First raw spectra were treated using a noise filter algorithm (correlation factor, 0.7) and the default advanced base-line correction. Then monoisotopic masses were generated on a fully detected spectrum after internal calibration using autodigestion tryptic peptides or external calibration using a mixture of five external standards (PepMix 1, LaserBio Labs, Sophia Antipolis, France). Mains peaks were selected according to local background, molecular weight of the protein, and known peak contamination from trypsin and keratin using PeakErazor software (version 1.49, Lighthouse Data, Odense, Denmark). Searches were carried out using four different algorithms for protein identification: Mascot (version 2.1, Matrix Science), ProFound (version 2005.02.14, Proteometrics), MS-Fit (version 3.2.1, ProteinProspector), and Aldente (version 2006.09.11, ExPASy) softwares on the rodent entries of the NCBInr database. A combination of results from different algorithms offers the option to cross-validate and consolidate the identification through the complementary use of several packages. Allowed variable modifications were oxidation of methionine, acrylamide-modified cysteine, and carbamidomethylation of cysteine. Up to one missed tryptic cleavage was considered, and a mass accuracy in the range of 25-50 ppm was used for all tryptic mass searches. Identification was considered positive if the protein was identified with the most elevated score using at least two algorithms. For a protein with a score close to the threshold value, the identification was rejected. Identified proteins are listed in Tables I and II. If peptides matched to multiple members of a protein family, the protein with the highest number of matched peptides and experimental molecular weight and isoelectric point close to theoretical parameters was indicated in the table, and NCBI accession numbers of the other isoforms are mentioned at the bottom of the table.
Plasmids-The cDNA encoding HA-RGSZ1 was purchased from the University of Missouri Rolla cDNA Resource Center (Rolla, MO), and pGEX-4T-1 vector was from Amersham Biosciences. The cDNA encoding Myc-RGS10 was a kind gift from Dr. P. J. Casey (Durham, NC). The FLAG-MT 1 protein construct has been described elsewhere (16). The FLAG-MT 2 construct was obtained by a PCR-based approach using a Myc-MT 2 cDNA (16) and the Phusion High-Fidelity DNA Polymerase (Finnzymes). Restriction sites for EcoRI and XbaI were introduced immediately adjacent to the initiation and termination codons of MT 2 by PCR. After digestion by the respective restriction enzymes, the resulting insert was ligated into a pcDNA3.1 vector encoding FLAG. All constructs were confirmed by sequencing.
Immunoprecipitation-HEK 293 cells grown in 100-mm culture dishes were transiently transfected by 4 g of either FLAG-MT 1 or FLAG-MT 2 cDNAs together with 4 g of HA-RGSZ1 or Myc-RGS10 cDNAs. 48 post-transfection, cells were stimulated, or not, by 10 Ϫ7 M melatonin for 15 min at 37°C, washed three times in PBS, and lysed in 1.5 ml of cold lysis buffer (75 mM Tris, 2 mM EDTA, 12 mM MgCl 2 , 10 mM CHAPS, pH 7.4). After sonication, receptors were solubilized for 3 h at 4°C under gentle end-over-end mixing, and lysates were centrifuged at 12,000 ϫ g for 1 h at 4°C. Receptor immunoprecipitation was performed on 1 ml of supernatant using 4 g of polyclonal anti-FLAG antibodies preadsorbed on protein G-Sepharose beads (Sigma). Immunoprecipitated proteins were eluted with 50 l of SDS-PAGE sample buffer (62.5 mM Tris/HCl, pH 6.8, 2% SDS, 10% glycerol, 100 mM DTT, 0.5% bromphenol blue) for 10 min at 95°C and subjected to SDS-PAGE and immunoblotting.
Immunoblotting-Proteins resolved by 1D electrophoresis were transferred onto nitrocellulose membrane (Whatman GmbH). After incubation in PBS supplemented with 5% skimmed milk and 0.05% Tween 20 for 1 h at room temperature, membranes were incubated in primary antibody solution in PBS supplemented with 5% skimmed milk and 0.05% Tween 20 for an additional 1 h at room temperature. After washes in PBS supplemented with 0.05% Tween 20, immunoreactivity was revealed using a horseradish peroxidase-conjugated secondary antibody (1:10,000) and the ECL chemiluminescent reagent (Perbio).
Immunofluorescence Microscopy-HEK 293 cells transiently transfected by 2 g of FLAG-MT 1 together with 1 g of RGSZ1-YFP cDNA were grown on sterile coverslips. 48 post-transfection, cells were stimulated, or not, by 10 Ϫ7 M melatonin for 15 min at 37°C, washed three times in PBS, and fixed with 4% paraformaldehyde in PBS for 15 min. After blocking with 3% BSA in PBS for 30 min, cells were incubated with monoclonal anti-FLAG M2 (1:500) antibodies in PBS containing 0.3% BSA for 1 h at room temperature. Coverslips were then washed three times with PBS and incubated with Cy3-conjugated goat anti-mouse antibodies (1:800) in PBS containing 0.3% BSA for 40 min at room temperature. Coverslips were mounted and analyzed by confocal laser microscopy (Leica TCS SP2 AOBS).
Production and Purification of HA-RGSZ1-To express HA-RGSZ1 as a GST fusion protein, restriction sites for BamHI and XhoI were introduced immediately adjacent to the initiation and termination codons by PCR of the plasmid encoding HA-RGSZ1 using the Phu-sion High-Fidelity DNA Polymerase. After digestion by BamHI and XhoI, the resulting insert was ligated into pGEX-4T-1 vector. DNA sequences were confirmed by sequencing. The resulting plasmid was transformed into BL21(DE3) cells (Invitrogen), and cells were grown in LB medium containing 50 g/ml ampicillin. Expression of GST-HA-RGSZ1 protein was induced by adding 1 mM isopropylthiogalactoside (Sigma) to midlog cultures. Cultures were harvested after 7 h at 37°C. Cells were centrifuged at 6000 ϫ g for 15 min, and the pellet was stored at Ϫ20°C. For purification of HA-RGSZ1, the pellet was resuspended in cold PBS containing protease inhibitor mixture EDTAfree, crushed with an Ultra-Turrax T25, and sonicated. After adding 1% Triton X-100 (Sigma), the cell suspension was centrifuged at 10,000 ϫ g for 30 min, and the supernatant was collected and applied to glutathione-agarose beads (Sigma) that had been equilibrated with 1% Triton X-100 in PBS. After extensive washes, HA-RGSZ1 proteins were eluted by 0.04 unit/l thrombin (Amersham Biosciences) in 200 l of PBS (2 h at 22°C). Eluates were applied to new, freshly prepared glutathione-agarose beads to eliminate residual contaminant GST and bacterial proteins. The purity of HA-RGSZ1 was assessed by SDS-PAGE after silver staining. Radioligand Binding-Ovine pituitary pars tuberalis was collected, and crude membranes were prepared essentially as described previously (17). Radioligand binding was performed with pars tuberalis mem-branes (1.5-2 mg of protein) in 1 ml of TEM buffer (75 mM Tris, 12 mM MgCl 2 , 2 mM EDTA, protease inhibitor mixture EDTA-free, pH 7.4) containing 400 pM 2-[ 125 I]melatonin (PerkinElmer Life Sciences) for 90 min at 37°C. After membrane solubilization with 1% digitonin (overnight at 4°C) and centrifugation (90 min at 18,000 ϫ g at 4°C), RGSZ1 was immunoprecipitated from solubilized proteins using 10 l of polyclonal anti-RGSZ1 antibodies preadsorbed on protein G-Sepharose beads. Beads were washed, and radioactivity was measured after rapid filtration through GF/F glass fiber filters (Whatman).

Optimization of the Peptide Affinity Chromatography-
Metal affinity chromatography on an Ni-NTA matrix was chosen to purify proteins interacting with the amino-terminally His 6 -tagged C-tails of MT 1 and MT 2 (61 and 59 amino acids long, respectively) in whole brain lysates. Binding of His 6tagged proteins is maintained under low concentrations of imidazole, which is known to minimize nonspecific binding to Ni-NTA-agarose beads by competing for nickel ion binding with dispersed histidine residues in proteins. To determine the optimal imidazole concentration, beads were incubated with different amounts of brain lysates in the presence of different imidazole concentrations. After several washes, retained proteins were recovered and quantified (Fig. 1A). In the absence of imidazole, the nonspecific binding was high and increased concurrently to the amount of brain proteins added. At 20 mM, the nonspecific binding was dramatically decreased irrespective of the amount of lysate used. Further increase in the Optimization of the peptide affinity chromatography conditions. A, determination of the optimal concentration of imidazole to minimize nonspecific binding on the beads. Different amounts of brain proteins were added to non-coated beads (20 l) and incubated overnight at 4°C in the presence or absence of 10 or 20 mM imidazole. After washes, proteins were eluted from the beads, and the nonspecific binding was quantified. B, determination of the optimal amount of brain protein lysates. Non-coated beads (20 l) were incubated with 2, 4, 6, 8, or 10 mg of protein lysates in the presence of 20 mM imidazole overnight at 4°C. After washes, proteins were eluted from the beads, and the nonspecific binding was quantified. C, determination of the binding kinetics of the His 6 (6xHis)-tagged peptides to the beads. 500 g of His 6 -tagged peptides were dissolved in a binding buffer containing 20 mM NaH 2 PO 4 , 6 M urea, pH 8, at 1 mg/ml. Absorbance of the peptide solutions was measured at 280 nm after 15, 30, 60, and 90 min of incubation with beads. D, determination of the amount of brain proteins specifically recruited by the C-tails of MT 1 and MT 2 . 10 mg of brain protein lysate were incubated with 20 l of beads coated with His 6 -tagged baits in the presence of 20 mM imidazole overnight at 4°C. After washes, the amount of retained proteins was quantified. C, non-coated beads. Results are expressed as mean Ϯ S.E. (n ϭ 5 for non-coated beads, n ϭ 3 for MT 1 , and n ϭ 6 for MT 2 ). amount of brain lysate (up to 10 mg of protein) did not change the level of nonspecific binding (10 -12 g) (Fig. 1B). We thus incubated Ni-NTA beads with 10 mg of brain lysate in the presence of 20 mM imidazole in all further experiments.
We then determined the binding kinetics of the His 6 -tagged peptides to the Ni-NTA beads in phosphate buffer at pH 8, the optimal pH for binding of His 6 -tagged peptides (according to the manufacturer) and for the preservation of protein complexes (18). Binding of peptides was rapid and reached a plateau after 30 min. After 90 min, 295 Ϯ 21 g (n ϭ 4) and 351 Ϯ 16 g (n ϭ 4) of His 6 MT 1 and MT 2 C-tails, respectively, were immobilized on Ni-NTA beads (Fig. 1C). Using these optimized experimental settings, the C-tails of MT 1 and MT 2 typically recruited 80.7 Ϯ 4.3 g (n ϭ 3) and 26.4 Ϯ 2.7 g (n ϭ 6) of proteins, respectively. Under our conditions, nonspecific binding in the absence of peptide was 11.9 Ϯ 1.4 g (n ϭ 6).
Functional Validation of Peptide Columns-We used the differential presence of a PSD-95/Disc-large/ZO-1 (PDZ) domain binding motif in the C-tails of MT 1 and MT 2 to evaluate the efficacy and specificity of our purification approach. Whereas the C-tail of MT 1 contains a functional class III PDZ recognition motif, MT 2 does not. PDZ domain-containing proteins are prototypical scaffolding proteins that are involved in the assembly of large molecular complexes. Based on literature reports (19,20) and our own results 2 obtained from a yeast two-hybrid screen, we probed eluates of the peptide columns for the presence of three PDZ domain-containing proteins, the multi-PDZ domain-containing protein MUPP1, nNOS, and PSD-95. The use of specific antibodies revealed the presence of MUPP1, nNOS, and PSD-95 in eluates of the MT 1 C-tail column but not in that of control or MT 2 C-tail columns (Fig. 2). These results confirm the high specificity of our approach. In addition, we identified three PDZ domain-containing proteins as new interacting proteins for MT 1 . We furthermore probed the column eluates for the presence of GRKs, which are known as specific GPCR-associated proteins (21). Indeed GRK2/3-specific antibodies confirmed the presence of these kinases in eluates of MT 1 and MT 2 C-tail columns (Fig. 2). We finally probed eluates of the peptide columns for the presence of the ␣ subunit of the heterotrimeric G i proteins known to be associated with the MT 1 and MT 2 receptors (22)(23)(24). As shown in Fig. 2, the MT 1 C-tail and, to a lesser extent, the MT 2 C-tail retained G i3 ␣ proteins.
Systematic Identification of MT 1 -and MT 2 -associated Proteins-We then performed large scale experiments to recover protein quantities that are sufficient for mass spectrometric analysis. Because of the inherent limitations of the different electrophoretic approaches, peptide column eluates were separated in three different ways: 2D electrophoresis (150 -180 g of recruited proteins) and 1D electrophoresis using either 10 or 5-9% polyacrylamide gradient gels. Fig. 3 illustrates a typical gel for each condition. Numerous 2D spots and 1D bands were reproducibly observed on silver-stained gels for MT 1 and MT 2 conditions. In agreement with Fig. 1A, the apparent number of proteins recruited by the C-tail of MT 1 was greater than for MT 2 . 1D and 2D control gels showed that non-coated beads recruited very few nonspecific proteins. Silver-stained 2D spots and 1D bands were systematically excised and digested with trypsin, and the resulting peptides were analyzed by MALDI-TOF and identified automatically with the Mascot wizard algorithm or manually with Mascot, ProFound, MS-Fit, and Aldente softwares in the NCBInr database. Proteins unambiguously identified by mass spectrometry for the C-tails of MT 1 and MT 2 and for the non-coated beads are listed in Tables I and II and supplemental Table 1, respectively. We identified 58 proteins that were retained by the non-coated beads. These proteins, having affinity for Ni-NTA-agarose beads, were considered as nonspecific binders and systematically removed when found in the lists of MT 1and MT 2 -interacting proteins. 40 proteins that bind to the C-tail of MT 1 and 22 proteins that bind to the C-tail of MT 2 were identified. Among the 40 proteins identified for the MT 1 C-tail, 30 were obtained from 2D electrophoresis, and 10 were obtained from 1D electrophoresis. Among the 22 proteins identified for the MT 2 C-tail, 11 were obtained from 2D electrophoresis, and 11 were obtained from 1D electrophoresis. For each C-tail, different ␣ and ␤ isoforms of tubulin were 2 J. Guillaume and R. Jockers, unpublished data. identified from 1D electrophoresis (10 and 5-9% gradient gels), but the corresponding spots, although easily noticeable from 2D gels, were not excised for identification. If ␣ and ␤ isoforms of tubulin are not taken into account, separation of samples by 1D electrophoresis allowed the identification of eight and nine additional high molecular mass proteins (including seven proteins of molecular mass Ն150 kDa) for MT 1 and MT 2 , respectively. This significant increase in the number of identified interacting proteins (20% of the MT 1 and 41% of the MT 2 interactomes) clearly illustrates that separation of column eluates by 1D electrophoresis is complementary to 2D electrophoresis. Identified proteins showed localization to different subcellular compartments (plasma membrane, cytosol, and cytoskeleton) and could be divided into five distinct groups: "membrane proteins," "signaling proteins," "cytoskeleton proteins," "chaperones and stress response proteins," and a last group, classified as "others," containing the remaining proteins for which the function is not well described.
Validation of MT 1 -and MT 2 -associated Proteins by Immunoblotting Screen-Using specific antibodies, we next performed an immunoblotting screen to estimate the overall reliability of our mass spectrometry data and to quantify the relative amount of proteins that are recruited by both C-tails. Seven different antibodies were tested and confirmed the presence of the interacting protein identified by MALDI-TOF, demonstrating the high quality of our mass spectrometric data (Fig. 4). These experiments validated common interaction partners for both receptors, such as the tubulin ␣ and ␤ isoforms, the catalytic subunit of the protein phosphatase PP2A, and the PKC 2, and showed higher recruitment with the C-tail of MT 1 compared with MT 2 . The immunoblotting screen also confirmed MT 1 -specific interacting partners, such as 14-3-3 ␤; the ubiquitin thiolesterase otubain 1 (OTUB1), which is known to express different alternatively spliced forms (25); and an RGSZ1 splice variant migrating at 36 kDa as reported previously in bovine brain (26). In addition, validation of 14-3-3 ␤, otubain 1, and the catalytic subunit of PP2A demonstrates the reliability of MALDI-TOF identifications from 1D gels. Using other antibodies, we also demonstrated the presence of the regulatory subunit A␣/␤ associated with the catalytic subunit of PP2A and another isoform of 14-3-3 (14-3-3 ) in the MT 1 C-tail eluates (supplemental Fig. 1). Both C-tails also recruited the ␥ isoform of tubulin.
Validation of the Interaction between the MT 1 Receptor and RGSZ1 in Living Cells-RGS proteins are signaling modulators of heterotrimeric G␣ proteins that increase their rate of GTP hydrolysis. Recent observations indicate that RGS proteins regulate G protein-dependent signaling not only in a G protein-specific manner but also in a receptor-specific manner (27)(28)(29)(30). The molecular basis for receptor specificity is still poorly understood. However, the recruitment of RGS proteins into receptor-associated complexes, as suggested by our  data, represents one possible explanation. To confirm the physiological association of RGSZ1 with the full-length MT 1 receptor, we co-expressed either FLAG-MT 1 or FLAG-MT 2 constructs together with HA-tagged RGSZ1 in HEK 293 cells. Co-immunoprecipitation experiments performed with anti-FLAG antibodies revealed that only MT 1 co-immunoprecipitated RGSZ1 in both resting and melatonin-activated cells (Fig. 5A). The specificity of the interaction was further validated using RGS10 as a negative control. As shown in Fig.  5B, immunoprecipitation of FLAG-MT 1 or FLAG-MT 2 did not co-immunoprecipitate RGS10. As an additional control, we further performed immunofluorescence experiments in HEK 293 cells co-expressing FLAG-MT 1 together with RGSZ1-YFP (Fig. 5C). Consistent with our biochemical experiments, a distinct co-localization at the plasma membrane was observed in these cells irrespective of the activation state of the receptor.
To evaluate the functional relevance of RGSZ1 coupling to the MT 1 receptor, we determined the effect of increasing concentrations of purified RGSZ1 on [ 35 S]GTP␥S binding in response to 1 M melatonin using CHO cell membranes stably expressing high levels of MT 1 (Fig. 5D). As expected, the agonist melatonin stimulated [ 35 S]GTP␥S binding (to 221 Ϯ 27%), indicating efficient MT 1 protein coupling in the absence of RGSZ1. When purified RGSZ1 was added, an increase in the [ 35 S]GTP␥S binding was observed in response to the agonist as reported previously for RGS4 (31). This assay is designed to promote [ 35 S]GTP␥S binding in the presence of RGS proteins as GTP hydrolysis becomes rate-limiting due to saturation of GDP-GTP exchange at high concentrations of receptor (31). This results in local depletion of inactive heterotrimeric G-GDP that is reversed by RGS GTPase-activating protein activity. Indeed purified RGSZ1 accelerated melatonin-promoted [ 35 S]GTP␥S bind-
To confirm that the presence of RGSZ1 in MT 1 -associated protein complexes is physiologically relevant, we performed immunoprecipitation studies with ovine pituitary pars tuberalis tissue samples known to express significant amounts of endogenous MT 1 receptors (32, 33) and RGSZ1 (data not shown). In the absence of high affinity anti-MT 1 antibodies, receptors were labeled with the specific 2-[ 125 I]iodomelatonin radioligand, and protein complexes were solubilized and immunoprecipitated with anti-RGSZ1 antibodies. As shown in Fig. 5E, significant amounts of radiolabeled MT 1 were precipitated in the presence of anti-RGSZ1 antibodies, demonstrating the existence of MT 1 -RGSZ1 complexes in native tissue.

DISCUSSION
In this study, we present an improved proteomics approach coupling peptide affinity chromatography with 1D and 2D electrophoresis separation and mass spectrometry for the identification of proteins interacting with the C-tail of GPCRs. To date, this is the first study proposing the use of an adapted and optimized IMAC for the identification of protein complexes. Using this approach, we identified 40 proteins that bind to the C-tail of MT 1 and 22 proteins that bind to the C-tail of MT 2 .
Interactions between GPCRs and associated proteins are known to involve the intracellular loops, the transmembrane, and the C-tail of GPCRs. The most abundant and the best studied are proteins interacting with the C-tail of GPCRs. This C-tail, recently nicknamed the "magic" tail by Bockaert et al. (7), serves as a scaffold for the formation of a signalosome containing diverse effectors and clustering proteins. The nature of the interaction proteins that bind to the C-tail of GPCRs can determine not only its targeting to a specific cellular compartment but also its association with other signaling or structural proteins and the fine tuning of its signal transduction such as desensitization and resensitization (34). Therefore, the identification of the protein complexes associated with GPCRs constitutes an important step toward the development of new drugs. Indeed these compounds could be used to disrupt or strengthen specific interactions between GPCRs and their associated proteins. Different proteomics approaches coupling peptide affinity chromatography and mass spectrometry have been developed to identify proteins interacting with the C-tail of GPCRs and have proven their efficacy in identifying protein complexes associated with the C-tail of GPCRs (7,35). However, these approaches, which used either short peptides encompassing a specific interaction motif such as the PDZ ligand of the 5-HT 2A and 5-HT 2C receptors (12,13) or the entire C-tail expressed as a GST fusion protein such as for the 5-HT 2C receptor (11), have shown limitations. They were either too restrictive to some particular interaction proteins (PDZ domain-containing proteins) or used GST fusion proteins as "bait" leading to increased background. IMAC has several advantages. Indeed the Ni-NTA interaction with the His 6 tag is highly specific and strong enough to resist detergents, induces low nonspecific binding due to the addition of low concentrations of imidazole (Fig. 1A), and uses pH conditions optimal for the preservation of protein complexes and to minimize nonspecific interactions (18). Moreover the His 6 tag is much smaller (955 Da) than other commonly used tags, such as the GST tag (26 kDa), contributing to the reduction of nonspecific interactions between mice brain proteins and the tag. The use of chemically synthesized bait of high purity also considerably reduced nonspecific binding in contrast to GST-tagged bait produced in bacteria for which the 2D protein pattern is often complex due to the presence of bacterial proteins, which bind to the GST tag and/or fusion peptide during bait production in bacteria, and the major brain proteins. After optimizing parameters such as the concentration of imidazole, the amount of brain protein lysates, and the incubation time for bait coupling on the beads, we were able to considerably reduce nonspecific binding to 0.1% (10 g from 10 mg of total brain proteins). Using these experimental settings, ϳ80 and 25 g of brain proteins (from 10 mg) were specifically recruited by the C-tails of MT 1 and MT 2 , respectively.
Most studies aiming to identify protein complexes associated to the C-tail of GPCRs have used 2D electrophoresis for protein separation. This technique minimizes the number of potential candidates as it is not adapted for selection of hydrophobic, basic, and large proteins. To enlarge the number of interacting proteins and obtain a more detailed interactome, we performed 2D electrophoresis and 1D electrophoresis using 10 and 5-9% gradient gels. The utilization of 1D electrophoresis was possible because of the small size of the His 6 -tagged baits (7.16 kDa for the MT 1 and 6.80 kDa for the MT 2 C-tails) in contrast to the larger GST baits whose size makes the use of 1D electrophoresis difficult notably for the identification of proteins with molecular weights similar to that of the bait. Among the 40 proteins identified for the MT 1 C-tail, 30 were obtained from 2D electrophoresis, and 10 were obtained from 1D electrophoresis. Among the 22 proteins identified for the MT 2 C-tail, 11 were obtained from 2D electrophoresis, and 11 were obtained from 1D electrophoresis. These results demonstrated that the combination of 1D and 2D electrophoresis greatly increased the number of proteins identified and that the use of only 2D electrophoresis led to an underestimation of multiprotein complexes associated to the C-tail of GPCRs. Identification of proteins by MALDI-TOF mass spectrometry, which only provides peptide mass fingerprints but no sequence information, is known to require prior high quality purification. However, our study has shown that several proteins separated by 1D electrophoresis could be unambiguously identified by MALDI-TOF because of the sensitivity of the apparatus and the stringency of our experimental conditions that considerably reduce the nonspecific binding and consequently the complexity of the protein electrophoretic pattern for MS identification.
Several previously described GPCR-associated proteins were identified. In this study, we showed that the C-tail of MT 1 recruited 14-3-3 ␤ and destrin, proteins shown to be associated with mGluR5 (36), and dynamin 1, a protein known to interact with 5-HT 2C (11) and mGluR5 (36). Some proteins found in MT 2 C-tail eluates were also already described as GPCR-binding proteins, such as the clathrin heavy chain and Na ϩ /K ϩ -ATPase ␣3 subunit for mGluR5 (36) and microtubuleassociated protein 2 (MAP2) for mGluR5 (36) and 5-HT4a (13). Ten common members of MT 1 and MT 2 C-tail-associated complexes were identified, including well known GPCR-associated proteins such as G i ␣ proteins and GRK2/3 and other proteins such as dual specificity phosphatase 3 (DSP3), a protein able to dephosphorylate protein substrates containing both phosphotyrosine and phosphoserine or phosphothreonine residues, with specificity for the mitogen-activated protein kinases (MAPKs) Erk2 and c-Jun NH 2 -terminal kinase (JNK) (37). The serine/threonine protein kinase casein kinase II (CK2), recently shown to be involved in phosphorylation of the M3 muscarinic receptor (38), was also found as a common interaction member of MT 1 and MT 2 C-tails. Interestingly a role for CK2 consensus sites within GPCR C-tails has been proposed in targeting GPCRs to the ␤-arrestin-dependent pathway (39). PKC 2 and tubulins ␣ and ␤ were also recruited by both MT 1 and MT 2 C-tails and confirmed by immunoblot-ting. PKC 2 is specifically expressed in the mouse brain (40) and is a member of the atypical PKC subfamily. Tubulins ␣ and ␤ constitute the basic building block of microtubules, the major components of the cytoskeleton involved in many essential processes (41). Tubulins ␣ and ␤ have been shown to interact with the C-tails of mGluR7a and mGluR1a, respectively (42)(43)(44), and we recently demonstrated a modulation of melatonin receptors and G protein function by microtubules (45).
Among the proteins involved in signal transduction that were specifically recruited by the C-tail of MT 1 , we identified two proteins involved in the deubiquitination pathway and members of the ubiquitin-specific processing proteases, otubain 1 (confirmed by Western blot) and ubiquitin-specific peptidase 5, also known as isopeptidase T. Subunit 4 of COP9, a multiprotein complex of the ubiquitin-proteasome pathway, was also identified. The ␤ isoform of the catalytic subunit of protein phosphatase 2A (PPP2cb) was also found. PP2A is a highly conserved serine/threonine phosphatase that plays pivotal roles in diverse cellular functions (46). A direct interaction between this subunit and the C-tail of the ionotropic N-methyl-D-aspartate receptor has already been demonstrated by yeast two-hybrid and immunoprecipitation experiments (47). In addition, PP2A has been shown to coimmunoprecipitate with mGluR5 and regulates mGluR5-dependent mitogen-activated protein kinase/extracellular signal-regulated kinase kinase (MEK)/ERK phosphorylation in neurons (48). Finally three PDZ domain-containing proteins were also identified as specific MT 1 -associated proteins by immunoblot: MUPP1, nNOS, and PSD-95. We failed to identify these three proteins by MS probably due to their low abundance within the gels. MUPP1 has been characterized previously as a direct interaction partner of MT 1 that promotes G i protein coupling of this receptor (49). Neuronal NO synthase has been previously proposed as a potential interacting candidate of the MT 1 receptor (20). However, nothing is known about the functional consequences of this particular interaction. PSD-95 is a prototypical scaffolding protein highly enriched in the PSD that belongs to the membrane-associated guanylate kinase family (51). PSD-95 interacts with a host of proteins of diverse functions, most notably N-methyl-D-aspartate receptors, and regulates the functional integration of these receptors into signaling complexes in PSD (51). Interaction between PSD-95 and GPCRs has already been described as with the somatostatin receptors SSTR4 and SSTR1 (52) and dopamine D1 receptors (53). The coupling of PSD-95 with the C-tail of D1 receptors has been shown to regulate the surface expression and the intracellular trafficking of the receptors and to inhibit D1 receptor-mediated signaling in mammalian cells (53). G i3 ␣ proteins were also detected by immunoblot but again failed to be identified by MS. In contrast, G i ␣ was easily detectable by MS using the tandem affinity purification (TAP) approach (24) suggesting that the C-tail of MT 1 and MT 2 is involved in G i ␣ binding but that further molecular determinants provided by the entire receptor are needed for efficient G i ␣ recruitment.
By our approach, we were also able to identify membrane proteins, which are often undetectable when using 2D electrophoresis. The majority of these membrane proteins were identified following 1D electrophoresis. Surprisingly, however, some of them were identified by 2D electrophoresis, such as the vomeronasal type-1 receptor A12. Vomeronasal receptors are GPCRs that bind pheromones and are responsible for various behavioral and neuroendocrine responses between individuals (54). Heterodimerization between this vomeronasal receptor and the melatonin MT 1 receptor remains to be demonstrated. The number of identified proteins interacting with the MT 2 C-tail was lesser than that of MT 1 . However, several interesting proteins were identified, including catenin ␦-1, also known as p120-catenin. This protein is a regulator of cadherin stability and an important modulator of Rho GTPase activities (55). ␦-Catenin was first identified through its interaction with presenilin-1, the molecule most frequently mutated in familial Alzheimer disease (56), and shown to interact with mGluR1 receptors by co-immunoprecipitation experiments using rat brain homogenates and dissociate from mGluR1 upon activation by L-glutamate (57).
Approaches based on isolated GPCR subdomains, such as the C-tail, have to be compared with the recently described TAP approach that uses the entire receptor as bait to pull down GPCR-associated protein complexes (24). A careful analysis of the advantages and drawbacks shows that both approaches are rather complementary than mutually exclusive. Whereas the entire receptor allows the purification of protein complexes formed in intact cells, the C-tail approach is more easily accessible for comparing the interactomes of different tissues and under different experimental conditions (different pathophysiological models and pharmacological treatments). A complete list of advantages and drawbacks of both approaches is presented in supplemental Table 2. In contrast to the TAP approach, which is, by itself, a co-immunoprecipitation performed from living cells, interactions between the bait and the proteins identified by approaches using GPCR subdomains need to be confirmed in intact cells with the full-length receptor ideally expressed endogenously. In this study, we validated the interaction between the MT 1 receptor and RGSZ1, a brain-specific protein that belongs to the RGS family (26,58,59). RGS proteins have been described as signaling modulators of heterotrimeric G␣ proteins that increase their rate of GTP hydrolysis and terminate signaling (60 -62) and more recently as proteins able to directly interact with GPCRs (27,(63)(64)(65). We succeeded in confirming the interaction between the MT 1 receptors and RGSZ1 in cells from the ovine pars tuberalis, which expresses both proteins endogenously, and in transfected HEK 293 cells. The interaction was functional, constitutive, occurred at the plasma membrane, and was only detected with MT 1 but not with MT 2 . The constitutive nature of the interaction suggests precoupling of RGSZ1 to MT 1 at the basal state and molecular rearrangement within this pre-existing complex upon agonist stimulation. This model is in good agreement with results described for other RGS proteins (27,50,63).
In conclusion, we present here a generic strategy that represents a major methodological advance for the identification of components of GPCR C-tail-associated complexes and overcomes the limitations of currently used genetics and proteomics approaches. Moreover this approach appears to be complementary to recently developed proteomics approaches based on the use of the entire GPCR as bait. Features of this new strategy are the use of chemically synthesized His 6 -tagged baits and IMAC that resulted in low nonspecific binding, pH conditions that preserved the integrity of associated protein complexes, and the association of 1D and 2D gel electrophoresis for protein separation that increases the number of identified interaction partners. These 62 candidates will now provide a basis for future cellular studies. Our approach represents a new opportunity for the study of protein complexes associated with GPCRs that can be used to compare GPCR interactomes from different tissues and can be extended to the identification of GPCRassociated complexes under pathological conditions or after in vivo pharmacological treatments. Several interactions strongly depend on reversible post-translational modifications such as phosphorylation. In vitro phosphorylation of the serine, threonine, and tyrosine residues of GPCR subdomains or the replacement of these residues by phosphomimetic amino acids are two further extensions of our approach to specifically retain those complexes whose interaction depends on receptor phosphorylation.