A library of 7TM receptor C-terminal tails. Interactions with the proposed post-endocytic sorting proteins ERM-binding phosphoprotein 50 (EBP50), N-ethylmaleimide-sensitive factor (NSF), sorting nexin 1 (SNX1), and G protein-coupled receptor-associated sorting protein (GASP).

Adaptor and scaffolding proteins determine the cellular targeting, the spatial, and thereby the functional association of G protein-coupled seven-transmembrane receptors with co-receptors, transducers, and downstream effectors and the adaptors determine post-signaling events such as receptor sequestration through interactions, mainly with the C-terminal intracellular tails of the receptors. A library of tails from 59 representative members of the super family of seven-transmembrane receptors was probed as glutathione S-transferase fusion proteins for interactions with four different adaptor proteins previously proposed to be involved in post-endocytotic sorting of receptors. Of the two proteins suggested to target receptors for recycling to the cell membrane, which is the route believed to be taken by a majority of receptors, ERM (ezrin-radixin-moesin)-binding phosphoprotein 50 (EBP50) bound only a single receptor tail, i.e. the beta(2)-adrenergic receptor, whereas N-ethylmaleimide-sensitive factor bound 11 of the tail-fusion proteins. Of the two proteins proposed to target receptors for lysosomal degradation, sorting nexin 1 (SNX1) bound 10 and the C-terminal domain of G protein-coupled receptor-associated sorting protein bound 23 of the 59 tail proteins. Surface plasmon resonance analysis of the binding kinetics of selected hits from the glutathione S-transferase pull-down experiments, i.e. the tails of the virally encoded receptor US28 and the delta-opioid receptor, confirmed the expected nanomolar affinities for interaction with SNX1. Truncations of the NK(1) receptor revealed that an extended binding epitope is responsible for the interaction with both SNX1 and G protein-coupled receptor-associated sorting protein as well as with N-ethylmaleimide-sensitive factor. It is concluded that the tail library provides useful information on the general importance of certain adaptor proteins, for example, in this case, ruling out EBP50 as being a broad spectrum-recycling adaptor.

Interaction of receptors with adaptor and scaffolding proteins is important for their biogenesis, their cellular sorting and targeting to the cell membrane, and their function at the membrane in complex with transducer molecules and downstream effector molecules as well as the subsequent internalization and post-endocytotic sorting of the receptors (1,2). These interactions among receptors, adaptors, and scaffolding proteins are highly regulated processes that can be controlled by phosphorylation events (3), expression of receptor activating, or inactivating variants of adaptor proteins (4), by competition among adaptor proteins, and by competition between adaptor proteins and effector molecules (5). For the large family of G protein-coupled seven-transmembrane segment receptors (7TM 1 receptors), this field is still in its infancy and only a rather sketchy picture has emerged of relative importance of specific adaptor and scaffolding proteins for the biogenesis, function, and desensitization of these receptors. Methods such as yeast two-hybrid screening, co-immunoprecipitation, and affinity chromatography using immobilized receptor fragments as bait have been used to identify potential receptor-binding proteins. The proposed functional roles of these interacting proteins are very diverse. Examples include promotion or inhibition of agonist-induced receptor internalization (6 -8), inhibition of mitogen-activated protein kinase activation (9), regulation of constitutive activity (10), retention of receptors in the endoplasmic reticulum (11,12), coupling to second messenger systems (13)(14)(15), and spatial organization of synapses (16).
Thus, a number of cases have been described where a specific adaptor protein has been biochemically and/or functionally linked to a single or several related receptors. However, to what degree such interactions are of general importance for 7TM receptors or for specific subsets of receptors or, in fact, only a single or a few receptors is in most cases still unclear. To address the question of the importance of specific adaptor scaf-folding proteins for the function of 7TM receptors in general, we chose a systematic biochemical approach by establishing a library of 7TM receptor tails fused to glutathione S-transferase (GST). 7TM receptors expose several intracellular loops for potential interaction with intracellular proteins. However, it is especially the C-terminal tail of the receptors that interacts with adaptor and scaffolding proteins (17)(18)(19)(20)(21)(22)(23)(24)(25). Although, for example, intracellular loop 3 is critically involved in the recognition process between the receptor and transducer/effector molecules such as the heterotrimeric G proteins and arrestins, these proteins also interact with parts of the C-terminal receptor tail (26 -30). Thus, the tail contains recognition sequences and epitopes for effector as well as scaffolding proteins. Besides the so-called "helix VIII" region, i.e. a relatively short, amphipathic, and helical segment located between the intracellular end of TM-VII and a frequently occurring palmitoylated Cys motif, very little information is available concerning the secondary and tertiary structures of 7TM receptor tails, which in the available x-ray structures have appeared to be rather unordered (31). Nevertheless, it is known that recognition motifs for adaptor and scaffolding proteins in the tails can be coiledcoil domains or C-terminally located PDZ recognition sequences (32,33).
In this study, proteins proposed to be involved in post-endocytotic sorting of receptors were probed for interactions with the library of 7TM receptor tail-fusion proteins. The vast majority of 7TM receptors are internalized upon agonist stimulation. In the classical, arrestin-mediated pathway, the activated receptor is phosphorylated by G protein-coupled receptor kinases, which leads to recruitment of arrestin. Arrestin functions as an adaptor protein interacting with clathrin and AP2, thereby targeting the receptor to clathrin-coated pits and subsequent endocytosis. Following endocytosis, the receptors may enter one of two pathways (see Fig. 1). In the recycling pathway, which has been described for the ␤ 2 -adrenergic receptor, the -opioid receptor, and the tachykinin NK 1 receptor, the ligand dissociates in the acidic pH of the endosomal compartment and the receptor is dephosphorylated and subsequently returned to the plasma membrane. In contrast, in the lysosomal pathway used by the ␦-opioid receptor and protease-acti-vated receptor 1 (PAR1), the receptor is targeted for degradation in lysosomes. The mechanism behind this targeted sorting of receptors is poorly understood. However, a number of proteins have been proposed to govern the differential sorting event. ERM-binding phosphoprotein 50 (EBP50, also called Na ϩ /H ϩ -exchanger regulatory factor) and N-ethylmaleimidesensitive factor (NSF) have both been suggested to be responsible for the recycling of the ␤ 2 -adrenergic receptor (21,22). In contrast, sorting nexin 1 (SNX1), which originally was demonstrated to be required for the lysosomal sorting of the epidermal growth factor receptor, was recently suggested to be involved in the lysosomal sorting of PAR1 as well (24,34). Protease-activated receptors are irreversibly activated by enzymatic digestion of the N-terminal segment of the receptor, and the sorting of activated receptors to lysosomes rather than recycling is critical for terminating signaling for these receptors. Another protein called G protein-coupled receptor-associated sorting protein (GASP) was recently suggested to be involved in the preferential lysosomal sorting of the ␦-opioid receptor (23). As shown in Table I, the four proteins, EBP50, NSF, SNX1, and GASP, which have been proposed to function as adaptor proteins involved in the post-endocytotic sorting of 7TM receptors, are structurally very different and have been implicated in various other cellular functions. Here, these proteins are probed for their ability to bind to the C-terminal tails of 59 different 7TM receptors as determined by GST pull-down assays, which routinely have been used to confirm protein interactions identified by co-immunoprecipitation and yeast two-hybrid screening (8,23,(35)(36)(37). In selected cases, interactions were further studied by surface plasmon resonance (SPR) technology or the interaction was characterized in more detail through gradual deletion mutagenesis of the tail protein.

Expression and Purification of Receptor Tail GST Fusion Proteins-
The cDNA corresponding to the C-terminal tails of the 59 7TM receptors listed in Table II were cut approximately four amino acid residues after the tyrosine of the NPXXY motif in TM-VII, ensuring that the tails all include the helix 8 motif (38). For receptors that did not express the NPXXY motif (e.g. family C receptors), tails were identified by classical hydrophobicity plots (Kyte and Doolittle). To generate fusion proteins encoding the receptor C-terminal tails after GST, receptor tails were amplified by PCR, digested with BamHI and XhoI, and ligated into the pGEX-4T-1 vector. Tails containing endogenous BamHI or XhoI sites were cloned as BamHI EcoRI or EcoRI XhoI fragments. Plasmids were transformed into XL1-Blue bacteria, and the DNA sequence was verified by sequencing. Plasmids were transformed into Escherichia coli strain BL21 bacteria for protein expression. Bacterial cultures were grown in 2-liter volumes, and fusion protein production was induced by the addition of 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h at 37°C. Fusion proteins were purified on 1 ml of glutathione-Sepharose 4B beads essentially as described by the manufacturer (Amersham Biosciences). Protein was eluted from the beads using 3 ml of 10 mM glutathione (in 50 mM Tris-HCl, pH 8). The purified protein was dialyzed three times against STE buffer (10 mM Tris-HCl, pH 8, 150 mM NaCl, 1 mM EDTA) using Slide-A-Lyzer cassettes (3,500 molecular weight cut-off, Pierce). Finally, protein was frozen in 10% glycerol and 1 mM dithiothreitol in 100-l aliquots and kept at Ϫ80°C. At the time of use, protein was thawed, bound to beads, and washed three times in cold STE buffer. The amount of fusion protein on the beads was determined by comparison with bovine serum albumin standards on Coomassie Blue-stained Bis-Tris PAGE gels (10% NuPAGE Bis-Tris gel, Invitrogen).
Quality Control of Purified Proteins-The sequences of all of the receptor tails in pGEX-4T-1 were verified by sequencing, and the quality of the fusion protein was determined by polyacrylamide gel electrophoresis. Most of the proteins showed a single band of the right size. However, some proteins showed more than one band, which indicates that the protein was partly degraded. Lowering the temperature from 37 to 30°C or increasing the composition or amount of protease inhibitors did not significantly improve the quality of these proteins. In all of the cases, the amount of fusion protein was determined only from the protein band of the right size excluding degradation products. The protein quality of two tail constructs, CCR3 and AT2, was not satisfactory, and these receptors were excluded from the library.
Synthesis of 35 S-Labeled Sorting Proteins-[ 35 S]Methionine was incorporated into human EBP50, rat NSF, human SNX1, human cGASP (C-terminal 497 residues of GASP), and GASP in a coupled in vitro transcription and translation reaction according to the manufacturer's instructions (rabbit reticulocyte lysate system, Promega L5010). Proteins were expressed from T7 promoters in pCDNA3 (EBP50, NSF) and pCDNA3.1ϩ (SNX1, GASP, cGASP). All of the plasmids were linearized with XbaI to increase expression with the exception of GASP and cGASP, which contain endogenous XbaI sites.
GST Fusion Protein-binding Assay-GST fusion protein on beads (3 g on 15 l of settled beads) was incubated with 500 l of blocking buffer (20 mM Hepes, pH 7.4, 100 mM NaCl, 2 mM MgCl 2 , 0.1% Triton X-100, 1% ovalbumin) for 30 min at room temperature and was collected by centrifugation (500 ϫ g for 5 min), and 2 l of the in vitro translation reaction mixture was added along with 18 l of wash buffer (20 mM Hepes, pH 7.4, 100 mM NaCl, 2 mM MgCl 2 , 0.1% Triton X-100). This mixture was mixed for 1 h at room temperature. Beads were washed three times for 5 min in ice-cold wash buffer and collected by centrifugation, eluted in 4ϫ SDS sample buffer, and subjected to Bis-Tris PAGE (10% NuPAGE Bis-Tris gel, Invitrogen). Gels were Coomassie Bluestained and dried overnight (DryEase gel-drying system, Invitrogen). Finally, radioactive bands on the gel were visualized on a Phosphor-Imager (Amersham Biosciences) and developed in Molecular Imager software (Bio-Rad). Quantitative determination of radioactive band intensities was done with Molecular Imager software using background subtraction. Band intensities were subsequently normalized to the band intensity of the reference lane, which was loaded with 2 l of the in vitro translation reaction.

7TM Receptor Tails and Post-endocytic Adaptor Proteins
Expression and Purification of SNX1-For use in surface plasmon resonance analyses, SNX1 was expressed in a cell-free E. coli system (RTS 500 expression system, Roche Applied Science) and purified using the ProBond purification system (Invitrogen). The isoelectric point of SNX1 had to be increased from 5.08 to 5.75 by adding 4 lysines and 12 histidines to the C-terminal end of the protein to achieve efficient adsorption/preconcentration at the carboxymethyl dextran chip surface (described below). The coding sequence of SNX1 (with 4 lysines and 12 histidines added to the C-terminal) was amplified by PCR and cloned into the pET101D vector containing a C-terminal six-histidine tag and a T7 promoter using the TOPO expression kit (Invitrogen).

7TM Receptor Tails and Post-endocytic Adaptor Proteins
Runs were conducted at 25°C with HBS-EP buffer (10 mM Hepes, pH 7.4, 150 mM sodium chloride, 3 mM EDTA, 0.005% (v/v) surfactant P20) at a flow rate of 10 l/min. SNX1 was covalently attached to the carboxymethyl dextran surface using standard amine coupling.

RESULTS
The receptors, which were included in the library of 7TM receptor C-terminal tails used in this study, were selected from all three major classes of 7TM receptors, i.e. 44 Family A, rhodopsin-like receptors, 5 Family B receptors, and 10 Family C receptors (Table II). The Family A receptors comprise 13 biogenic amine receptors, 21 peptide receptors, 2 proteaseactivated receptors, 1 lipid messenger receptor, 3 chemokine receptors, and 3 virally encoded chemokine receptors plus one orphan receptor. The library was somewhat biased as the complete set of receptor subtypes for a particular ligand in some cases were included in order to be able to probe for potential subtype-specific interactions with the adaptor proteins (for example, all five muscarinic receptors and all five somatostatin receptors). The length of the tails varied considerably between the Family A and B receptors from only 17 residues (5-HT 1D , H 1 , and MC 1 ) to 106 amino acid residues (NK 3 ). Among the Family C receptors, the length varied from 32 to 354 residues (Table II).
Mapping Interactions of Receptor Tail with Sorting Protein by GST Pull-down Assay- Fig. 2 shows autoradiographic images and the corresponding quantitative data obtained in the GST pull-down experiments with a representative selection of GST tail-fusion proteins and the four proposed 35 S-labeled sorting proteins. The volume intensity of the radioactive bands was determined, and the relative amount of the bound radioactive protein was quantified by dividing the volume intensity of each band by that of the probe band, which was defined as the relative binding. Using relative binding instead of absolute band intensity values allowed for a comparison between experiments performed on different days and analyzed in different gels. Unspecific binding of the radioactive probe (sorting protein) to the GST beads was estimated by running the GST protein alone without tail-fusion in parallel in all of the assays, The relative binding of receptor tails to sorting proteins in GST pull-down experiments are shown as the mean and mean Ϯ S.E. Band intensities were normalized to the band intensity of a reference lane so that 100% binding corresponded to the retention of all of the added radioactive sorting protein (see Fig. 2). The cut-off values between binders and non-binders for NSF (5.5%), SNX1 (4.1%), and cGASP (6.3%) are marked by dotted lines. Only the ␤ 2 receptor bound to EBP50, and therefore, no cut-off line is marked for EBP50. which typically gave a relative apparent binding of 0 -0.5% and always below 1%. Similar experiments were performed for the full GST tail protein library. Positive interactions were repeated at least twice in independent experiments, and the relative binding was averaged.
Adaptor Proteins Suggested to be Involved in Receptor Recycling-EBP50 has been proposed to be involved in the recycling of the ␤ 2 receptor through a phosphorylation-sensitive binding to a C-terminal PDZ recognition sequence (39). Accordingly, it was found that the GST fusion of the C-terminal tail of the ␤ 2 receptor strongly bound the 35 S-labeled EBP50 protein (Fig. 3,  panel A). However, as shown in Fig. 3A, surprisingly, the ␤ 2 receptor tail was the only tail protein among the 59 fusion proteins tested, which bound EBP50 convincingly. A small signal was observed also for the mGlu 1b tail-fusion and a few other tail proteins; however, in view of the strong signal from the ␤ 2 receptor tail and the fact that the mGlu 1b tail, for example, does not have a PDZ recognition sequence, only the ␤ 2 receptor was considered a true hit (Fig. 3, panel A).
NSF is another protein that has been suggested to be required for the recycling of the ␤ 2 receptor (22). Also, in this case, the ␤ 2 receptor tail protein served as a convenient positive control binding strongly to the 35 S-labeled NSF protein (Fig. 3,  panel B). However, in contrast to EBP50, NSF bound to a number of the other tail-fusion proteins, albeit not as strongly as the ␤ 2 receptor tail. For an additional number of tail-fusion proteins, a weak signal was observed. Statistical analysis revealed that the data could be considered to be composed of two Gaussian distributions of what could be coined "non-binders" and "binders," respectively, and that a cut-off value in relative binding, which would ensure Ͻ1% contamination of the binders with non-binders, would be 5.5% for NSF (Fig. 3B, dotted line; see Supplemental Data). According to this finding, the following tail proteins, besides the ␤ 2 receptor, bound to NSF: the muscarinic M 1 , M 3 , M 4 , and M 5 , receptors; the tachykinin NK 1 and NK 2 receptors; the somatostatin SST 1 receptor; the DOP receptor, and as the virally encoded chemokine receptors US27 and US28 from human cytomegalovirus. However, a couple of tail proteins fall just under the relatively stringent cut-off value ensuring less than 1% false-positives, i.e. PAR1 and mGlu 1a (see Fig. 3B). These receptors should probably also be considered true NSF binders (see Supplemental Fig. S1 and "Discussion").
Adaptor Proteins Suggested to Be Involved in Lysosomal Receptor Targeting-SNX1 has been proposed to be responsible for the targeting of PAR1 to lysosomes (24). Pull-down analyses between 35 S-labeled SNX1 and the library of 7TM tail-fusion proteins showed a number of strong binders (Fig. 3C). Surprisingly, however, the GST fusion protein of the C-terminal tail of the human PAR1 bound only weakly to SNX1 in comparison to the virally encoded US28 receptor tail, for example. Statistical analysis of the relative binding data gave a cut-off value of 4.1% ensuring Ͻ1% false positive binders (see Supplemental Data). According to this cut-off value, the following tail proteins were considered to bind to SNX1: the muscarinic M 1 , M 4 , and M 5 receptors; the tachykinin NK 1 , NK 2 , and NK 3 receptors; the oxytocin receptor; the DOP receptor; US28; and the GLP 1 receptor (Fig. 3C). Four receptors can be considered to be borderline binders that are probably positive: the muscarinic M 3 receptor; the CXCR2 chemokine receptor; and the mGlu 1a and mGlu 5a metabotropic glutamate receptors ( Fig. 3C and Supplemental Fig. S1).
GASP is a large protein of 1394 amino acids that has recently been implicated in selective lysosomal sorting of the DOP receptor as opposed to the MOP receptor (23). Originally, yeast two-hybrid screening with the DOP receptor tail gave four positive clones, all of which contained sequences exclusively from the C-terminal 497 amino acid fragment of GASP (called cGASP).
As shown in Fig. 3D, the binding profile for cGASP was clearly the broadest among the tested sorting proteins. Approximately one-third, i.e. 23 of the 59 tail-fusion proteins, bound 35 S-labeled cGASP with a specific binding above the cut-off value of 6.3% (see Supplemental Data). The positive receptors were as follows: the ␤ 1 -and ␤ 2 -adrenergic receptors but not the three 5-HT receptors tested; the muscarinic M 1 , M 3 , M 4 , and M 5 receptors; the three tachykinin receptors; the oxytocin but not the V 2 receptor; the AT 1 receptor; the motilin but not the homologous ghrelin receptor; the DOP but not the MOP receptor (which are, respectively, the positive and negative controls) (23); PAR1 and PAR2; CXCR2; ORF74 and US28; the GLP 1 and the VIP 1 but not the secretin and glucagon receptors; and, finally, the mGlu 1a , mGlu 5a , and mGlu 8 receptors. The M 2 and the mGlu 1b receptors could be considered as borderline binders of cGASP.
The binding of 35 S-labeled cGASP and similarly labeled fulllength GASP was compared in ten receptor tails (Fig. 4). Although the overall picture was rather similar, cGASP gave a somewhat higher binding signal with most of the receptor tails. Interestingly, for the oxytocin and GLP 1 receptor tails, this difference was large and the binding of full-length GASP was borderline or would not be considered to be significant. Table III lists all of the positive hits among members of the 7TM receptor tail library with the four different adaptor proteins proposed to be involved in receptor sorting. None of the tail proteins were positive for all four adaptor proteins. How-ever, a group of tail proteins were positive for NSF, SNX1, and cGASP, i.e. the M 1 , M 4 , and M 5 receptors; the NK 1 and NK 2 receptors; the DOP receptor; and the virally encoded US28. It should be noted that these tail proteins, with respect to their size, cover the full range of the library, because they include some of the shortest (the M 4 tail being only 23 residues long) as well as some of the longest tails (the NK 1 and NK 2 being 97 and 88 residues long, respectively) (Table II).
SPR Affinity Measurements-This study constitutes a "horizontal" analysis of a large library of 7TM receptor tails with a selection of adaptor proteins as determined in pull-down experiments. However, to confirm a few of the positive hits and quantify the binding affinities, US28 and DOP binding to SNX1 were analyzed by SPR analysis. Preliminary experiments showed that preconcentration of SNX1 at the chip surface was not efficient, which was most probably caused by its low isoelectric point. Therefore, the isoelectric point of SNX1 was increased from 5.08 to 5.75 by adding 4 lysines and 12 histidines to the C-terminal end of the protein. In Fig. 5 are shown the association and dissociation kinetics for the US28 and DOP fusion proteins to SNX1 immobilized on the chip. The KOP fusion protein did not bind to SNX1 and is shown as a negative control. When the data were fitted to a 1:1 kinetic-binding model (A ϩ B 7 AB) with the BIAevaluation 3.0 software, it was found that the tail of the virally encoded receptor US28 bound to SNX1 with a K D value of 49 nM (chi 2 ϭ 12). The tail of the DOP receptor bound with an even higher affinity, i.e. a K D value of 20 nM (chi 2 ϭ 11).
Mutational Analysis of the NK 1 Tail Sequence Required for Binding to NSF, SNX1, and cGASP-The structural basis for the strong binding of one of the tails, i.e. the one from the NK 1 receptor, which binds to three of the four adaptor proteins, was characterized through systematic deletion mutagenesis. Five truncated versions of the NK 1 receptor tail were analyzed in GST pull-down experiments: NK 1 ⌬4; NK 1 ⌬21; NK 1 ⌬42; NK 1 ⌬63; and NK 1 ⌬84, where the number following the ⌬ refers to the number of amino acids deleted from the C-terminal end of the full tail sequence, i.e. residues 311-407 in the NK 1 receptor (Fig. 6A). As shown in Fig. 6B, a fairly similar picture was obtained for all three adaptor proteins. Deletion of the last four residues, which would have eliminated the tail binding if it had been recognized through a PDZ-domain type of recognition, had no or very little effect on the binding of the NK 1 tail to NSF, SNX1, and cGASP (Fig. 6B). Deletion of the last 21 residues clearly diminished the binding to all three adaptor proteins, and subsequent further deletions gradually diminished the binding. It should be noted that the shortest version of the NK 1 tail, NK 1 ⌬84, which basically only consists of the helix 8 motif, still showed binding albeit weak binding to all three adaptor proteins (Fig. 6). These data indicate that the binding epitope for NSF, SNX1, and cGASP to the NK 1 tail is large and covers major parts of the tail structure.

DISCUSSION
Adaptor and scaffolding proteins are highly important for the function of membrane receptors, for example, in determining their targeting to specific locations in the cell membrane (16,40), in determining the spatial and thereby functional association of receptors with various co-receptors (41), transducer proteins, and downstream effector molecules (13,14,16,42), and in determining post-signaling events such as receptor sequestration and post-endocytotic sorting (21)(22)(23)(24). The interactions of 7TM receptors with adaptor and scaffolding proteins are to a large degree governed by epitopes located in their C-terminal intracellular tails (17)(18)(19)(20)(21)(22)(23)(24)(25). We have established a library of C-terminal tails from a series of selected 7TM receptors covering families A, B, and C and representing their various subfamilies. In this study, this receptor tail library was screened for binding to four proteins, which have been proposed to be involved in the sorting of 7TM receptors, either for recycling to the cell membrane or for lysosomal degradation. Large differences were found in the patterns of interaction of the proposed sorting proteins with the GST tail-fusions from EBP50, which basically only interacts with a single receptor tail, i.e. the ␤ 2 receptor, to GASP, which binds roughly onethird of the 59 tail-fusion proteins tested. For many of the receptors, very little is known as yet regarding their postendocytotic fate, but for a number of receptors, such information is available and can be correlated to the binding pattern observed in the present study of their C-terminal tail with proteins proposed to be involved in these events (Table III). For example, the binding of the ␤ 2 receptor to EBP50 and NSF is consistent with the recycling phenotype of this receptor, whereas its binding to the proposed lysosomal targeting pro-tein, GASP, at first does not fit well into the picture. However, mutations of the recognition sequence for PDZ domain binding found in the C-terminal end of the ␤ 2 receptor or phosphorylation of this recognition sequence, which disrupts EBP50 (21) and NSF (22) binding, resulted in lysosomal targeting and degradation of the receptor. Thus, it could be hypothesized that some receptors such as the ␤ 2 receptor are "dual fate" receptors that are able to interact with several proteins and that, after endocytosis, they are sorted either for recycling or to the lysosomes depending on the phosphorylation state and on which sorting proteins are available for interaction with the receptor.
Binders versus Non-binders?-Pull-down experiments using GST fusion proteins is an established in vitro method for assessing protein-protein interactions. However, it is generally performed as part of a study where the interaction of interest is being illuminated through a series of complementary biochemical and cell biological approaches. In such "vertical" studies, where only one or a few receptors are being studied, the issue of "binding versus non-binding" is usually determined solely by comparison with a negative control, GST alone. Furthermore, the pull-down experiments are usually made either as the initial method through which the interaction partner is identified as a positive hit or it is applied to confirm an interaction already determined through some other means, i.e. focus is normally directed toward positive results. In this case, we are performing a horizontal study in which we evaluate among a large number of receptor tail-fusion proteins, which are binding to certain adaptor/sorting proteins and which are not. Although this study predicts a number of protein-protein interactions, studies performed with more physiological systems wherein intact receptors are used in co-immunoprecipitation studies, for example, may yield different results. For all four proposed sorting proteins, several unambiguous positive hits (for EBP50, only one hit) clearly stood out among the multitude of fusion proteins, many of which showed little or no binding (Fig. 3A). Nevertheless, if the binding to GST alone was used as the strict negative control, most of the tail proteins would in fact be considered to be positive hits because the weak bands, although TABLE III Significant interactions between sorting proteins and 7TM receptor tails Receptor tails with relative binding values higher than the estimated cut-off values of 5.5% (NSF), 4.1% (SNX1), and 6.3% (cGASP). Only h␤ 2 bound to EBP50, and therefore, no cut-off value was determined for EBP50. Not known rmGlu 5a Not known rmGlu 8 Not known

7TM Receptor Tails and Post-endocytic Adaptor Proteins
weak but slightly stronger than the GST band, were observed for most of the GST fusion proteins. However, the tail-binding results for each adaptor protein were in fact distributed in two Gaussian populations and we have chosen to use a cut-off value in relative binding, which would ensure Ͻ1% contamination of the binders with non-binders (see Supplemental Data). In doing so, we took into account that a certain degree of variable nonspecific protein-protein interaction probably is responsible for the low degree of binding observed with some tail-fusions. It is possible that we hereby exclude certain tail proteins, which in a cellular context may in fact form a physiologically important albeit weak interaction with a particular adaptor protein, from being considered as true binders. It will be interesting to test to what degree post-translational modifications such as the phosphorylation of particular residues in the tail will turn such weak binders into true hits with strong binding. Accordingly, it will be interesting to determine the effect of in vitro phosphorylation on the binding properties of the tail library. However, it is important to note that the effect of phosphorylation often does not result in an "all or none binding phenomenon." For example, the effect of phosphorylation of receptors on the affinity for arrestin is only ϳ5-10-fold (43,44). Although this study as described above is primarily a horizontal analysis of the receptor tail library, a few vertical experiments were included to substantiate the results. Thus, for two of the tail-fusion proteins, the virally encoded receptor US28 and the DOP receptor, the strong interaction with one of the adaptor proteins, SNX1, which had been identified through the pull-down experiments, was further analyzed by SPR to quantify the affinity of the protein-protein interaction. The observed nanomolar affinities correspond to affinities previously reported for PDZ interactions (45,46). This type of analysis will be valuable, for example, in the analysis of potential effects of post-translational modifications on the interaction of the tail library with adaptor proteins.
In another vertical analysis, the structural basis for the interaction of one of the apparently more promiscuous tailfusion proteins, i.e. the NK 1 receptor, with the three adaptor proteins, NSF, SNX1, and cGASP, was dissected through a series of systematic deletion mutants. It should be noted that the NK 1 tail does not bind totally promiscuously, because it does not bind to EBP50 (Fig. 3A) or certain other adaptor proteins (data not shown). Although certain differences could be pointed out for the three adaptor proteins, the structural analysis of the NK 1 tail gave a rather similar picture for all three proteins as a gradual loss of binding was observed in parallel with the gradual truncation of the tail from its Cterminal end (Fig. 6). Thus, it appears that the interaction of NSF, SNX1, and cGASP to the NK 1 receptor depends on an extended epitope covering most of the tail structure. This is particularly interesting because the binding of SNX1, for example, was also observed with the very short tail from the M 4 receptor (shown in Supplemental Fig. S2), which is only 23 residues long. This corresponds to the two most truncated NK 1 tail constructs (NK1⌬63 and NK1⌬84), both of which had lost most of their binding. Further mutational analysis is required to identify the epitopes and residues that determine the interactions of especially the short tail proteins. Some of these are so short that the helix 8 region may very well be involved in adaptor protein binding.
Adaptor Proteins Suggested to Be Involved in 7TM Receptor Recycling-Both EBP50 and NSF have been proposed to be responsible for the recycling of 7TM receptors as demonstrated in both cases initially for the ␤ 2 receptor.
In accordance with previously published results, we found that the tail of the ␤ 2 receptor bound strongly to EBP50. This binding has been demonstrated to be caused by the interaction of the PDZ domain in EBP50 with a type 1 PDZ recognition sequence SLL located at the far C-terminal end of the receptor tail (21,39). Although 13 of the tails in the library end in a type 1 PDZ recognition sequence (Table II), none of these tails bound EBP50. It has been suggested that EBP50 could be involved in the recycling of the KOP receptor, which does not contain a PDZ recognition sequence, as demonstrated by co-immunoprecipitation experiments (47). However, neither the KOP receptor tail nor any of the other tails in the library bound EBP50 (Fig. 3A). Thus, we concluded that, although EBP50 may be involved in the recycling of the ␤ 2 receptor and possibly a few other receptors, this protein is not an important protein for the recycling of 7TM receptors, in general.
NSF is a hexameric ATPase involved in vesicular transport and fusion throughout the exocytotic and endocytotic pathways (48,49). Recently, Cong et al. (22) demonstrated that the ␤ 2 receptor binds directly to NSF through an epitope involving the three last residues of the receptor tail, although NSF does not hold any PDZ domains, and that this interaction was required for receptor internalization and subsequent recycling (22). In this study, we found that NSF bound 11 of the 59 tail-fusions. As shown in Table III, most of these receptors are known to recycle to the membrane after endocytosis. The binding of NSF to the NK 1 receptor, which internalizes rapidly and is recycled and resensitized within 30 min after agonist stimulation (50,51), was not primarily dependent on the far C-terminal segment of the tail, in contrast to the ␤ 2 receptor (Fig. 6). NSF was also found to bind to the DOP and US28 receptor tails, i.e. receptors that are both known to accumulate in the lysosomal compartment upon endocytosis, although rapid recycling has also been demonstrated to occur for US28 in the same cells (52). It is possible that the binding of receptor tails to NSF could be improved upon phosphorylation, although this clearly is not required for the binding of the ␤ 2 receptor and the 11 other positive receptors identified in this study.
Adaptor Proteins Suggested to Be Involved in Lysosomal Targeting of 7TM Receptors-Both SNX1 and GASP have been proposed to be responsible for lysosomal targeting of 7TM receptors as demonstrated in biochemical and cell biological stud-ies using PAR1, DOP, and MOP receptors as the main model systems (23,24). SNX1 is a member of a relatively large family of sorting nexins, which are cellular trafficking proteins, all having a phospholipid-binding domain and a strong predisposition to form protein-protein complexes mainly through coil-coil formation (Table I) (53). SNX1, which is ubiquitously expressed, is found together with the homologous SNX2 in endosomes and was originally identified as being involved in the endocytotic processing of the epidermal growth factor receptor (34,54). Recently, Trejo and colleagues showed that the sorting of activated PAR1 from endosomes to lysosomes is regulated by SNX1 (24). SNX1 co-localizes with internalized PAR1 on early endosomes, and SNX1 is found associated with activated PAR1 in cellular lysates. Moreover, SNX1 deletion mutants cause significant inhibition of agonist-induced PAR1 degradation. However, in the initial report, a direct interaction between SNX1 and PAR1 was in fact not demonstrated. Subsequent studies have shown that depletion of SNX1 by small interfering RNA knockdown also causes significant inhibition of agonist-induced PAR1 degradation; however, experiments using the yeast two-hybrid system failed to detect a direct interaction between the PAR1 C-tail and SNX1. 2 This study also did not detect a direct interaction between the PAR1 C-tail and SNX1 using GST pull-down assays. Together, these findings suggest that SNX1 is critically involved in targeting the PAR1 receptor to the lysosomal sorting pathway for degradation, perhaps through an indirect interaction with the receptor or other important lysosomal sorting machinery. This means that PAR1 should not be taken as a positive control for direct SNX1 binding. In this study, we found that SNX1 bound to 10 of the 59 tails tested. Previously, we have reported that SNX1 binds to the dopamine D 5 but not any of the other four dopamine receptors (55). Interestingly, standard software for prediction 2 J. Trejo, personal communication. of coiled-coil domains, which could identify the three coiled-coil domains in SNX1, did not indicate that such a motif occurred frequently either among the 7TM receptor tails, in general, or among the positive hits for SNX1. Among the receptor tails that bound to SNX1, the oxytocin receptor, the DOP receptor, and the virally encoded US28 receptor are all known to be targeted to the lysosomal pathway, which would fit with the expected role of SNX1 (24,34). However, the M 1 and M 4 receptors, the three tachykinin receptors, and the GLP 1 receptor, which all bind SNX1, are known to be efficiently recycled to the cell membrane (Table III).
GASP was recently identified as a cytoplasmic protein that selectively interacts with the DOP versus the MOP receptor. Mutagenesis experiments and overexpression of a dominant negative version of GASP, cGASP, suggest that GASP is responsible for lysosomal targeting of the DOP receptor (23). In this study, we found that cGASP was the most broad-spectrum adaptor protein tested, which bound approximately one-third of the 7TM receptor tails of the library. Among the positive hits for cGASP were the tails from six receptors, which are known to be targeted to lysosomes: the OT receptor; the AT 1 receptor; the DOP receptor; PAR1 and PAR2; and the virally encoded US28 receptor ( Fig. 3D and Table III). In the original report on GASP, it was demonstrated by pull-down experiments that GASP also bound the dopamine D 4 and the ␣ 2B -adrenergic receptor tails, both of which are also known to be degraded following endocytosis. All of these results fit very well with the suggested role for GASP as being an adaptor protein responsible for lysosomal targeting. However, cGASP clearly also binds a large number of tails from receptors, which are efficiently recycled to the cell membrane after endocytosis and therefore are not targeted to lysosomes. Among these are the ␤ 2 receptor, the M 1 , M 3 , and M 4 receptors, and the three tachykinin receptors ( Fig. 3D and Table III). This finding suggests that GASP binding to the C-terminal tail is not sufficient for lysosomal targeting of 7TM receptors. On the other hand, it has previously been argued that the binding of GASP to the ␤ 2 receptor, for example, demonstrates that the sorting and targeting process probably is governed by competing interactions of several adaptor proteins (23). Thus, a mutant form of the ␤ 2 receptor in which the far C-terminal epitope required for the recycling process had been changed was efficiently sorted to lysosomal degradation (21,22). Importantly, this lysosomal targeting of the mutant ␤ 2 receptor could be impaired through overexpression of a dominant negative form of GASP (23). Consequently, it is concluded that GASP may function as a generally important adaptor protein involved in 7TM receptor sorting for the lysosomal pathway. However, it remains to be resolved why GASP then binds certain receptors but not others, which normally are recycled efficiently. It should also be noted that monoubiquitinylation of the receptor tails may be another mechanism that controls lysosomal targeting of receptors (56,57).
It is interesting to note that, despite the relatively large number of tail proteins studied, it has not been possible to identify consensus recognition sequences for the adaptor proteins probed. However, it should be noted that it has also not been a simple bioinformatic task to find G protein interaction sites among the close to 400 known rhodopsin-like receptors. Nevertheless, in a recent analysis of our tail library using adaptor proteins with PDZ domains, a much more clear and comprehensive picture emerged.
The 7TM Receptor Tail Library-We believe that this study with multiple clear positive and multiple clear negative results demonstrates that the library of C-terminal tails from representative members of the superfamily of 7TM receptors constitutes a powerful resource for in vitro probing of receptor-adap-tor protein interactions across the family, i.e. horizontal studies. It should be noted that there are clear limitations to this approach. There may be motifs in addition to the C termini that are essential for high affinity and specific binding of sorting proteins to 7TM receptors, which means that a lack of an interaction does not rule out the possibility that the proteins may interact in a cellular context. More generally, in vitro approaches may not reflect fully the interactions occurring in a cellular context. Moreover, studies on 7TM receptors expressed in heterologous cell systems, which include the majority of published 7TM receptor-trafficking studies, may not reflect what is really happening in more native systems or in vivo. Thus, the real physiological implications of the interactions identified in this study await further studies and the results obtained with the tail library should be the basis for identification and initiation of interesting vertical studies of particular receptors and adaptor proteins using a broad variety of biochemical as well as cell biological techniques.