(cid:1) -Arrestin Binding to CC Chemokine Receptor 5 Requires Multiple C-terminal Receptor Phosphorylation Sites and Involves a Conserved Asp-Arg-Tyr Sequence Motif*

Agonist binding to the CC chemokine receptor 5 (CCR5) induces the phosphorylation of four distinct serine residues that are located in the CCR5 C terminus. We established a series of clonal RBL-2H3 cell lines expressing CCR5 with alanine mutations of Ser 336 , Ser 337 , Ser 342 , and Ser 349 in various combinations and explored the significance of phosphorylation sites for the ability of the receptor to interact with (cid:1) -arrestins and to undergo desensitization and internalization upon ligand binding. Receptor mutants that lack any two phosphorylation sites retained their ability to recruit endogenous (cid:1) -arrestins to the cell membrane and were normally sequestered, whereas alanine mutation of any three C-terminal serine residues abolished both (cid:1) -arres-tin binding and rapid agonist-induced internalization. In contrast, RANTES (regulated on activation normal T cell expressed and secreted) stimulation of a S336A/ S349A mutant triggered a sustained calcium response and enhanced granular enzyme release. This mutational analysis implies that CCR5 internalization largely depends on a (cid:1) -arrestin-mediated mechanism Agonist-dependent intracellular calcium mobilization was measured in transfected RBL-2H3 cells as described (19). Calcium decays from 80% of the peak height to basal levels were fitted to an exponential ( a (cid:5) b (cid:1) e (cid:6) t/ (cid:4) ), where the time constant (cid:4) reflects the ability of CCR5 variants to evoke a more or less sustained calcium response (25).

The CC chemokine receptor CCR5 1 is a member of the large family of heptahelical receptors that transduce extracellular signals into the cell by activating heterotrimeric G proteins. CCR5 is expressed, among other cells, on memory T lymphocytes, macrophages, and dendritic cells (1). Physiological ligands that bind to this receptor with high affinity include the CC chemokines RANTES, MIP-1␣, MIP-1␤, and MCP-2 and a proteolytically processed variant of HCC-1. CCR5 is also the principal co-receptor for macrophage-tropic (or R5) strains of HIV-1 (2). Individuals homozygous for a nonfunctional CCR5⌬32 allele express a truncated receptor that fails to reach the cell surface and are thus highly resistant to HIV infection. Chemokines and small molecule receptor antagonists inhibit HIV-1 infection in vitro by blocking the binding of the viral envelope glycoprotein gp120 to CCR5. Another mechanism that underlies the pronounced antiviral effect of certain chemokines and their derivatives relates to their ability to effectively downmodulate CCR5 expression by inducing receptor endocytosis and to thereby decrease co-receptor availability on the cell surface (3).
According to a current concept of G protein-coupled receptor (GPCR) regulation, largely extrapolated from detailed studies with the prototypic ␤ 2 -adrenergic receptor, agonist activation of receptors leads to their phosphorylation by both second messenger-dependent protein kinases and GPCR kinases (GRKs) (4,5). This, in turn, promotes binding of members of the arrestin family. Arrestins bind to the ligandactivated/phosphorylated receptor and thereby sterically interfere with further binding of heterotrimeric G proteins. With many GPCRs, ␤-arrestins are essential for receptor internalization via clathrin-dependent mechanisms by their ability to directly couple the phosphorylated receptor to clathrin heavy chain and to recruit accessory proteins that participate in the formation of clathrin-coated pits (6). ␤-Arrestins were shown to interact with the ␤ 2 -adaptin subunit of the heterotetrameric AP-2 adaptor complex (7) and to act as co-factors in the activation of small G protein ADPribosylation factor 6 (8), two processes that fulfill essential roles in the internalization of certain GPCR. Recent evidence suggests that arrestins play a broader role as signaling scaffolds, linking them to the activation of Src family tyrosine kinases and certain mitogen-activated protein kinase modules, in addition to their role in the regulation of receptor trafficking (9). Together, these findings point to a central role for arrestins both in the termination and in the initiation of certain aspects of GPCR signaling.
To date, four members of the arrestin family have been identified. Only ␤-arrestin 1 (arrestin 2) and ␤-arrestin 2 (arrestin 3) are ubiquitously expressed throughout the body and are predominantly localized in lymphoid and neuronal tissues (10). The structural determinants that underlie the stable association of ligand-activated GPCR and arrestins have been studied in detail in the rhodopsin/visual arrestin system (11)(12)(13)(14)(15). Much less is known about structural features required for the interaction of these proteins in nonvisual receptor systems. The recent determination of the structures of visual arrestin (16) and ␤-arrestin 1 (17,18) at atomic resolution has allowed us to propose a molecular mechanism by which arrestins interact with ligand-activated GPCR. According to this model arrestin in its inactive state is composed of two N-and C-terminal domains and is maintained in this basal conformation by multiple intramolecular hydrophobic interactions among highly conserved residues. Upon binding, the receptor-attached phosphates disrupt these intramolecular forces and thereby induce a structural reorganization of the arrestin molecule. This conformational change is accompanied by the exposure of currently unknown secondary binding sites that allow arrestin to bind with high affinity to the ligand-activated receptor. So far, the structural determinants on ligand-activated receptors that allow arrestins to interact with hundreds of different GPCR are not well characterized.
Previously, we have shown that CC chemokines induce the rapid phosphorylation of CCR5 on four C-terminal serine residues, a process that is mediated by the combined action of GRKs and protein kinase C (19). In a subsequent study the ligand-induced association of ␤-arrestins with CCR5 was monitored in real time by fluorescence resonance energy transfer (20). Replacement of all four phosphorylated serine residues by alanine abrogated the ability of ligand-activated receptors to interact with ␤-arrestin, and this was accompanied by a defect in CCR5 desensitization and internalization. Here, we have further investigated the structural determinants that underlie ␤-arrestin binding to the chemokine-activated CCR5. To this end, we tested several CCR5 Ser/Ala replacement mutants that were stably expressed in RBL cells for their ability to recruit ␤-arrestins upon chemokine stimulation. An additional ␤-arrestin 1-binding site within a conserved region in the second cytoplasmic loop of CCR5 was identified by surface plasmon resonance (SPR) analysis. The functional significance of these findings was determined by investigating internalization and desensitization characteristics of the various receptor mutants.

EXPERIMENTAL PROCEDURES
Materials-Tissue culture media and cell culture supplies were from Biochrom; RBL-2H3 and HEK-293 cells were from the American Type Culture Collection; ECL Western blotting reagents were from Amersham Biosciences; nitrocellulose membranes were purchased from Schleicher & Schü ll; geniticin, detergents, and protease inhibitors were from Calbiochem; the peptides were synthesized by Jerini; anti-␤-arrestin 1 monoclonal antibody (clone 10) was from Becton Dickinson Transduction Laboratories; horseradish peroxidase-labeled secondary antibodies and phycoerythrin-labeled goat polyclonal anti-mouse F(abЈ) 2 were from Dako; 125 I-RANTES was from PerkinElmer Life Sciences; and all other reagents, unless otherwise indicated, were purchased from Sigma.
Plasmids and RBL Cell Lines-The pEF-BOS expression vectors that encode human CCR5 with various combinations of C-terminal serine to alanine substitutions have been described before (19). Rat basophilic leukemia cell 2H3 subline (RBL-2H3) cells were transfected by electroporation (19), and selection of transfected cells was accom-plished by adding 600 g/ml geniticin in the cell culture medium.
␤-Arrestin Translocation Assay-The RANTES-induced translocation of endogenous ␤-arrestins present in RBL cells from the cytosol to the membrane fraction was analyzed by subcellular fractionation. RBL-CCR5 cells (1 ϫ 10 7 /135-mm dish) were incubated in the presence of varying concentrations of RANTES for up to 10 min at 37°C. The cells were placed on ice and scraped into 3 ml of buffer A (10 mM PIPES, 100 mM KCl, 3 mM NaCl, 3.5 mM MgCl 2 , pH 7.0) containing 50 g/ml phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, 10 g/ml leupeptin, and 5 g/ml aprotinin. The cells were homogenized by Dounce homogenization (10 strokes) and sonication (four 20-s bursts at 100 W). The nuclei were removed by centrifugation at 1000 ϫ g for 20 min. The supernatant was loaded on a discontinuous gradient of 50, 35, and 20% sucrose in buffer A and centrifuged at 160,000 ϫ g and 4°C for 2 h. The supernatant (cytosol) was removed, and the 20/35% sucrose interphase (membrane) was collected, diluted in 3 ml of buffer A, and recentrifuged at 160,000 ϫ g and 4°C for 15 min. The pellet was resuspended in 40 l of detergent buffer (20). The protein content in 10-l aliquots of the cytosolic and the membrane fraction was determined by the Bio-Rad DC protein assay kit, and 10 (cytosol) or 25 g (membrane) of each protein sample was loaded onto 10% SDS-PAGE gels. The proteins were transferred to nitrocellulose membranes, and nonspecific binding sites were blocked by incubation for 1 h with 4% nonfat dry milk in TBS, 0.1% Tween 20, pH 7.4. ␤-Arrestin 1 and ␤-arrestin 2 were detected using monoclonal anti-␤-arrestin 1 antibodies (1:500) and horseradish peroxidase-labeled secondary antibodies (1:2000). The effect of RANTES stimulation on membrane-associated ␤-arrestin levels was quantitated by densitometric analysis (ImageMaster TotalLab software; Amersham Biosciences) of enhanced chemiluminescence films.
The peptides that correspond to cytoplasmic domains of CCR5 and derivatives thereof (see Table II) were synthesized by standard solid phase methods. Preparative purification of the peptides was achieved by reversed phase HPLC. The peptides displayed the correct mass spectrum and were Ͼ90% pure by analytical HPLC.
SPR Analysis-The association of ␤-arrestin 1 with the CCR5 cytoplasmic loop and C-tail derived peptides was analyzed in real time by SPR using a BIAcore 3000 biosensor (BIAcore AB). The peptides were immobilized to CM5 biosensor chips via the thiol group of an N-terminal cysteine residue according to the manufacturer's instructions. The peptide corresponding to the CCR5 cytoplasmic IL-3 loop was synthesized with a biotin moiety attached to the N-terminal amino acid and immobilized to a streptavidin sensor surface (SA5 chip) according to standard procedures. Recombinant ␤-arrestin 1 was used at concentrations ranging from 50 to 600 nM in running buffer (20 mM HEPES, pH 7.3, 50 mM NaCl, 10 mM KCl, 2 mM MgCl 2 , 0.2 mM dithiothreitol). All of the measurements were recorded at a flow rate of 20 l/min. Association (2 min) was followed by dissociation (2 min), during which running buffer was perfused. The sensor surface was regenerated after each experimental cycle by a pulse injection (15 s) of 10 mM NaOH, 0.5% SDS. The kinetic parameters of the interaction between ␤-arrestin 1 and receptor peptides and the equilibrium constant were calculated using the BIAevaluation software provided by the manufacturer (more details can be found in Ref. 22).
Functional Assays-The RANTES-induced N-acetyl-␤-D-glucosaminidase release from CCR5 expressing RBL-2H3 cells was determined as described (24). The values were expressed as percentages of total enzyme present in cells after lysis with 0.1% Triton X-100, and the data were analyzed using nonlinear regression applied to a sigmoidal dose response model with the Ligand Binding module of Sigma-Plot software (SPSS).
Agonist-dependent intracellular calcium mobilization was measured in transfected RBL-2H3 cells as described (19). Calcium decays from 80% of the peak height to basal levels were fitted to an exponential (a ϩ b⅐e Ϫt/ ), where the time constant reflects the ability of CCR5 variants to evoke a more or less sustained calcium response (25).

Characterization of RBL-CCR5
Cell Lines-Previously, our laboratory analyzed a series of CCR5 serine to alanine mutants that were transiently expressed in COS-7 cells for their abilities to undergo ligand-induced phosphorylation (19). In this study, several of these receptor constructs were stably transfected into RBL-2H3 cells, and we examined whether the constructs bind ␤-arrestins and undergo receptor internalization and desensitization upon RANTES stimulation. RBL-2H3 cells are of myeloid origin and have been shown to express high endogenous levels of GRK2/3 and both ␤-arrestins (arrestin 2 and arrestin 3) (19,26,27). Therefore, these cells are well suited for the analysis of GRK/arrestin-driven mechanisms of receptor regulation. CCR5 expression levels of the various cell lines were determined by flow cytometry using a monoclonal antibody (T21/8) with specificity for a CCR5 N-terminal epitope, and signaling characteristics were evaluated by testing the abilities to release glucosaminidase after RANTES stimulation (Table I). In whole cell phosphorylation experiments with these RBL-CCR5 cell lines (data not shown), we confirmed the effect of serine to alanine substitution in various combinations on receptor phosphorylation that we had previously observed when receptor constructs were transiently expressed in COS-7 cells (19).
RANTES-induced Translocation of ␤-Arrestins in RBL-CCR5 Cells-We have shown by fluorescence resonance energy transfer technology that wild type CCR5, but not a phosphorylation-deficient CCR5 mutant with alanine substitution of all four C-terminal serine residues rapidly interacts with ␤-arrestin upon RANTES stimulation (20). We now established a simple translocation assay based on subcellular fractionation of RBL cellular lysates that measures ligand-induced recruitment of ␤-arrestins from the cytosol to the cell membrane. This assay is based on an anti-␤-arrestin 1 monoclonal antibody, which significantly cross-reacts with ␤-arrestin 2 (Fig. 1A). As shown in Fig. 1B, treatment of RBL-CCR5 cells with 10 nM RANTES, but not with phorbol 12-myristate 13-acetate, leads to a more than 3-fold increase of two membrane-associated proteins that were identified as ␤-arrestin 1 and ␤-arrestin 2, respectively, by virtue of being immunoreactive with the anti-␤-arrestin antibody and migration at the same apparent molecular masses of 54 and 48 kDa as the ␤-arrestin 1/2 cDNA products from HEK-293 cells. Because this effect was absent in untransfected RBL cells, it directly reflects the ability of ligand-activated CCR5 to recruit ␤-arrestins to the cell membrane. Translocation of ␤-arrestins was induced by as little as 1 nM and reached a maximum at 10 nM RANTES (EC 50 ϭ 2.4 nM; Fig.  1C). ␤-Arrestins rapidly associate with the membrane fraction after stimulation with saturating concentrations of RANTES; maximum levels were observed after stimulation for 3-10 min (Fig. 1D).
To determine the minimal requirements of receptor phosphorylation for CCR5/␤-arrestin association, we investigated RAN-TES-induced ␤-arrestin translocation to the cell membrane in RBL cells that express various CCR5 Ser/Ala mutants (Fig. 2). All CCR5 mutants with alanine replacements of one or two C-terminal serine residues were capable of inducing translocation of ␤-arrestin 1 or ␤-arrestin 2 in a significant (p Ͻ 0.05)  manner and to a similar degree (2.6 -4.6-fold increase of membrane-associated ␤-arrestin upon RANTES stimulation) to that of wild type CCR5. In contrast, alanine mutation of any three C-terminal serine residues essentially eliminated the ability of the receptor to recruit ␤-arrestins. Overall, no significant differences were observed in the binding of the two ␤-arrestin isoforms to the various receptor mutants. We conclude from these experiments that ␤-arrestin 1 or ␤-arrestin 2 binding to ligand-activated CCR5 requires the presence of at least two intact C-terminal receptor phosphorylation sites, but the exact position of these sites is not critical.
Receptor Internalization of CCR5 Ser⁄Ala Mutants-Next, we asked how the abilities of various RBL-CCR5 mutants to induce ␤-arrestin membrane translocation correlate with their capacities to internalize activated receptors into the intracellular membrane compartment of RBL cells. We had previously reported that 125 I-RANTES is rapidly taken up by RBL-CCR5 cells and that endocytosis is significantly impaired in RBL cells that express a phosphorylation-deficient CCR5 mutant with alanine mutation of all four serine residues (20). We now show (Fig. 3) that internalization of receptor mutants with alanine replacements of any single or two serine residues proceeds in a time-dependent manner that closely resembles wild type receptor internalization. In contrast, substitution of three C-terminal serine residues by alanine in various combinations significantly impaired the early rapid phase of ligand uptake to a similar degree as was observed with the completely phosphorylation-deficient CCR5-⌬4 mutant. Again, the exact position of the various mutated C-terminal phosphorylation sites did not appear to be critical.
Desensitization of the CCR5-mediated Calcium Response and Enzyme Release-In a previous study we had shown that phosphorylation-deficient CCR5 mutants that do not bind ␤-arrestins and internalize much slower than wild type CCR5 also show a defect in agonist-induced receptor desensitization (20). This was most clearly demonstrated by different time courses of the RANTES-induced intracellular calcium mobilization in the various mutant cell lines. In all of the RBL-CCR5 mutants that were analyzed in the present study, a robust calcium flux was observed which peaked within 10 -20 s after stimulation with a saturating concentration (40 nM) of RANTES (Fig. 4A). This confirms the intact G protein coupling abilities of these receptor mutants. However, with several mutant cell lines a significant prolongation of the rise in intracellular calcium levels was observed, consistent with the idea that a desensitization-deficient receptor induces a sustained calcium response. These differences in receptor desensitization were expressed as calcium decay time constants as described before (20,25). As shown in Fig. 4B, alanine mutation of serine 349 alone (CCR5-⌬1A) did not prolong calcium signaling, but in combination together with S336A (CCR5-⌬2C), a significantly higher time constant for calcium decay was determined that was almost as high as that observed with the fully phosphorylation-deficient receptor (CCR5-⌬4). Combination of the S336A mutation together with S337A (CCR5-⌬2A) or with S342A (CCR5-⌬2B) had a smaller effect or no effect on the ability to induce a sustained calcium response. A significant defect in receptor desensitization was again observed in all Ser/Ala triple mutants.
In a previous study we showed that a fully phosphorylationand desensitization-deficient CCR5 mutant releases significantly more glucosaminidase upon maximal RANTES stimulation than wild type receptor-expressing cells (20). This finding was confirmed in the present work and extended to other receptor mutants with only partial phosphorylation defects (Table I). Overall, alanine mutation of C-terminal serine residues in various combinations affected the capacity to induce a sustained calcium response or to release a higher fraction of the total cellular pool of granular enzymes from RBL cells in a parallel manner. This indicates that both cellular parameters reflect the abilities of receptors to undergo desensitization.
Binding of ␤-Arrestin 1 to Intracellular Domains of CCR5-To characterize ␤-arrestin 1 binding to intracellular domains of CCR5 in quantitative terms and also to identify additional ␤-arrestin 1-binding sites on the receptor, we measured the affinity of ␤-arrestin 1 binding to synthetic receptor peptides by SPR spectroscopy. To this end various synthetic peptides that correspond to the three intracellular loops and the C-terminal tail of CCR5 (Table II) were immobilized on the biosensor chip and probed for the binding of recombinant ␤-arrestin 1 (Fig. 5) The two different C-terminal receptor peptides represent the membrane-proximal (CT-1) and the C-terminal (CT-2) parts of this receptor domain, which are separated by a membrane anchor/palmitoylation site. The sensorgram shown in Fig. 6 indicates that ␤-arrestin 1 binds to CT-2 with high specificity because control surfaces with an immobilized unrelated peptide or without any peptide did not bind any ␤-arrestin 1. For the binding of ␤-arrestin 1 to the CCR5 C-terminal tail peptide, we calculated an equilibrium rate constant of 5.5 M (Table II). Unexpectedly, binding of ␤-arrestin 1 to a phosphorylated version of the same C-terminal tail peptide exhibited a similar affinity of 3.6 M. Thus, in vitro phosphorylation of the CCR5 C-terminal domain has only a minor stimulatory effect on the binding of ␤-arrestin 1 to this region.
We next determined whether ␤-arrestin 1 binds to other cytoplasmic loop regions of CCR5 (Table II). Synthetic peptides corresponding to the cytoplasmic loops IL-1 to IL-3 as well as the C-terminal sequence preceding the palmitoylation site were incubated with ␤-arrestin 1 to record binding. Although CCR5 IL-1 as well as IL-3 did not bind to ␤-arrestin 1 at all, we were able to detect arrestin binding to IL-2 at a rate constant similar to that observed for CT-2 (6 M). In addition, we also detected arrestin binding to CT-1, although with 50% reduced affinity (12 M). Because the binding of ␤-arrestin 1 to IL-2 was unexpected, we tried to further delineate amino acid residues that mediate arrestin binding. For these experiments a shorter IL-2 peptide was used as a positive control as well as several IL-2 mutant peptides. As summarized in Table II, we could show that ␤-arrestin 1 binding to IL-2 depends on an intact Nterminal DRY-motif. Single point mutations within the DRY region resulted in a 2-6-fold reduced binding. Additionally, the mutation of all three residues to alanine totally abolished ␤-arrestin binding. In conclusion, our in vitro binding studies demonstrate specific binding of ␤-arrestin 1 to the IL-2 loop and to the CT-2 region of CCR5 as well as low affinity binding of arrestin to CT-1. Interestingly, binding to IL-2 involves a conserved sequence motif that is located at the junction of the third transmembrane domain and the second cytoplasmic loop of most GPCR.

DISCUSSION
␤-Arrestins are central regulators of GPCR signaling and have been implicated in the desensitization and internalization of chemokine receptor CCR5. The aim of this study was to identify phosphorylation-dependent as well as phosphorylation-independent determinants on the receptor that enable ␤-arrestins to bind with high affinity to ligand-activated CCR5. Another aspect of our work relates to the question of whether receptor desensitization and endocytosis are both dependent on the same arrestin-mediated mechanism or whether these processes are regulated differentially.
Previously, we used phosphoamino acid analysis and alanine scanning mutagenesis to identify four distinct C-terminal serine residues as the exclusive phosphorylation sites on this receptor (19). Elimination of all serine phosphorylation sites by alanine mutation resulted in the inability of ␤-arrestin to interact with CCR5 upon ligand binding, and this also signifi-

FIG. 3. Time-dependent internalization of 125 I-RANTES by wild type (WT) or mutant CCR5 in RBL-2H3 cells.
Radioligand was allowed to bind to RBL-2H3 cells that expressed wild type CCR5 or various mutant receptors for 90 min at 4°C. After incubation of the cells at 37°C for the times indicated, surface-bound labeled RANTES was removed by washing with chilled low pH buffer. Internalization was calculated as the amount of acid resistant radioactivity divided by total specific cell-associated radioactivity. The results are presented as the means Ϯ S.E. of at least two independent experiments performed in triplicate.

FIG. 4. Time course of RANTES-induced calcium mobilization in RBL-2H3 cells expressing wild type (WT) or mutant CCR5.
Stably transfected RBL-2H3 cells were loaded with Fluo-3AM, and intracellular calcium levels were recorded by spectrofluorimetry. A, calcium flux in three different cell lines in response to a saturating concentration of RANTES (40 nM). Maximal calcium levels were adjusted to 100% to better illustrate different abilities of the various cell lines to induce more or less sustained calcium responses. B, time constants () for calcium decay were calculated as described in the text. The data represent the means Ϯ S.E. from at least three independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.001. cantly inhibited receptor desensitization and internalization (20). In the present study we expressed CCR5 mutants with serine to alanine substitutions in various combinations in RBL cells and determined their abilities to recruit cytosolic ␤-arrestins to the cell membrane after RANTES stimulation. Ligand activation of wild type CCR5 expressed in RBL cells induced a 3.8-fold increase in the amount of ␤-arrestin 1 associated with the cell membrane and a similar 3.6-fold increase in the amount of membrane-associated ␤-arrestin 2. A previous report showed that in RBL-2H3 cells ␤-arrestin 2, but not ␤-arrestin 1, was selectively recruited to clathrin-coated pits upon stimulation of different GPCR (27). Because our assay measures membrane translocation rather than clathrin recruitment, these divergent results probably reflect the ϳ6-fold higher clathrin binding affinity of ␤-arrestin 2 compared with ␤-arrestin 1 (28). In contrast to other assays for the detection of agonist-induced association of arrestins with GPCR in intact cells that use co-transfection, cross-linking, and co-immunoprecipitation of the two interacting proteins (29,30), the method described here does not directly measure arrestin binding to a receptor. Nonetheless, we believe that this method represents a valid approach because ␤-arrestins associated with the cell membrane in an agonist-and time-dependent manner, which correlated well with GRK-mediated CCR5 phosphorylation, as reported before (19), and also required intact C-terminal receptor phosphorylation sites. Furthermore, this method avoids the inherent drawback of other procedures that involve arrestin overexpression and that in some cases were found to result in nonspecific binding of arrestin to nonactivated receptors (30).
The minimal receptor phosphorylation levels required for high affinity ␤-arrestin binding have not been systematically investigated in nonvisual GPCR. With the ␤ 2 -adrenergic (11) and m2 muscarinic cholinergic receptors (31), a stoichiometry of phosphorylation of 2 mol of phosphate/mol of receptor was found to be both necessary and sufficient for arrestin binding, and increasing phosphorylation levels to even higher molar ratios did not further enhance association of arrestin with the receptor. In contrast, arrestin affinity to light-activated rhodopsin increased almost linearly with each added phosphate up to a stoichiometry of 4 mol phosphate/mol receptor (32). However, these results are difficult to interpret because of the stochastic nature of receptor phosphorylation and the heterogeneity of phosphorylation sites in the different receptor preparations. By using site-directed mutagenesis to specifically eliminate distinct phosphorylation sites, we avoided this potential complication. In our study with phosphorylation-deficient receptor mutants, we found that arrestin binds to CCR5 in a nongraded manner that requires a minimum of two intact receptor phosphorylation sites. The exact position of these sites within a cluster of four C-terminal serine residues did not appear to be critical. A similar flexibility in the positions of C-terminal phosphorylation sites that are involved in high affinity ␤-arrestin binding was reported in the formyl peptide receptor system (33).
These findings are compatible with the sequential multisite model of ␤-arrestin association with ligand-activated/phosphorylated GPCR, which is based on the structural analysis of arrestins by x-ray crystallography (16 -18, 34) and mutagenesis studies (11,35). This model proposes that high affinity binding of arrestin to a receptor involves two separate activation-recognition and phosphorylation-recognition sites that reside within the N-terminal domain of arrestin as well as a secondary receptor recognition site that is located within the C-terminal domain of the molecule. Interaction between the phosphorylated receptor tail and a phosphate sensor region that forms part of the conserved polar core of arrestin releases the C-terminal tail, thereby allowing arrestin to undergo a transition into its high-affinity receptor-binding state. In transgenic mice that express various rhodopsin mutants with a decreased number of rhodopsin phosphorylation sites, a minimal number of three C-terminal rhodopsin phosphorylation sites appeared to be necessary for rapid and reproducible receptor deactivation in vivo (36). In contrast, in our study, which directly investigates ␤-arrestin binding to CCR5 rather than a Cysteine residues and biotin-␤Ala-␤Ala groups that were used for the conjugation to carrier proteins are underlined, and phosphoserine (pS) and alanine substitutions are highlighted in bold type. Single-letter amino acid codes are used. b ␤-Arrestin 1 binding to synthetic CCR5 peptides was determined as outlined in the legend to Fig. 6. From the derived sensorgrams the kinetic rate constants for association (k a ) and dissociation (k d ) were determined using BIAevaluation software, version 2.1. The derived equilibrium constants (K D ) were then calculated using the equation k d /k a . ND, not detectable. receptor deactivation, we find that receptor mutants with no more than two intact phosphorylation sites interact with ␤-arrestins as efficiently as wild type receptors.
We used SPR analysis to further investigate the significance of C-terminal receptor phosphorylation for high affinity arrestin binding to the receptor. This technique has the advantage compared with cellular assays with overexpressed intact proteins that it allows determination of kinetic rate constants and binding affinities of arrestin to certain intracellular domains of the receptor in quantitative terms and, furthermore, allows precise control of the phosphorylation level of synthetic receptor peptides. Unexpectedly, we found that ␤-arrestin 1 bound to C-terminal CCR5 peptides in both their fully phosphorylated and nonphosphorylated forms with approximately the same micromolar affinities. In the rhodopsin system it had been shown that a synthetic heptaphosphopeptide that comprises the fully phosphorylated C terminus of bovine rhodopsin induces a conformational change in arrestin that also enables this protein to bind to light-activated unphosphorylated rho-dopsin (37). In the same study it was shown that elimination of the C terminus of rhodopsin had no effect on binding of phosphopeptide-activated arrestin to this receptor. Together, these results indicate that phosphorylation of the C terminus of a receptor does not directly enhance arrestin binding affinity to this particular region but rather acts as a molecular switch that induces a conformational change and enables secondary high affinity binding of ␤-arrestin to other parts of the receptor. The hypothesis that other receptor domains significantly contribute to ␤-arrestin binding is also supported by the observation that ␤-arrestin binding affinity to the CCR5 C terminus is in the same micromolar range as had been reported in similar peptide binding studies with other receptors (38,39), yet these values are still much lower (i.e. in the nanomolar range), and therefore the actual affinity is much higher when binding affinity is measured with whole receptors (11).
To our knowledge, this study is the first to systematically explore potential ␤-arrestin-binding sites on all cytoplasmic domains of a nonvisual GPCR. Our results indicate that two separate regions of CCR5, the C terminus and the second intracellular loop, are capable of binding ␤-arrestin 1. By screening a series of synthetic peptide mutants that correspond to the second cytoplasmic loop for their abilities to interact with ␤-arrestin 1, we show that a conserved triplet of amino acids (Asp-Arg-Tyr), located at the junction of transmembrane helix 3 and the second intracellular loop, represents an arrestinbinding site. The high degree of conservation of this so-called DRY motif within the GPCR superfamily implies that it is critically involved in G protein signaling pathways. Indeed, several studies with various GPCR have addressed the effects of mutations in the DRY motif on receptor-induced G protein activation. Molecular modeling studies with the ␣ 1b -adrenergic receptor predict that upon ligand-mediated receptor activation the Arg of the DRY sequence shifts out of a conserved polar pocket that lies between transmembrane domains I, III, VI, and VII (40,41). Thus, it becomes exposed to the exterior of the receptor molecule and creates a conformation that favors G protein coupling. Another critical event in the transition from the inactive to active state of a receptor appears to be the protonation of the acidic aspartate residue upon G protein binding (40,42). In the case of CCR5, such as with many other GPCR, mutation of the DRY motif resulted in a nonfunctional receptor that no longer couples to the G␣ subunit (43). In contrast, few studies have addressed the significance of the conserved DRY motif for binding of other proteins that also recognize ligand-induced conformational changes of GPCR, such as the arrestins. In the case of the formyl peptide receptor replacement of arginine by glycine within this site prevents not only G protein coupling but also arrestin binding (44). Interestingly, this receptor mutant was normally phosphorylated upon ligand binding, indicating that conformational changes within the DRY region are even more important determinants for agonist-induced arrestin binding to the formyl peptide receptor than C-terminal receptor phosphorylation. Similar mutations of arginine residues in the DRY motif of the vasopressin receptor (45) or in the equivalent ERY motif of rhodopsin (46) generate conformations that allow arrestin to interact with receptors in a constitutive, ligand-independent manner. Mutation of the DRS sequence in the second intracellular loop of the gonadotropin-releasing hormone receptor to DRY significantly increases the relative internalization of this receptor (47). Together, these results suggest that a sequence motif that is conserved throughout the GPCR family and is known to be involved in G protein activation also fulfills a principal role in arrestin binding. Other cytoplasmic domains of CCR5 were found to be less important. In particular, we found no evidence FIG. 6. ␤-Arrestin binding to synthetic CCR5 peptides. The binding of ␤-arrestin 1 was recorded in real time using a biosensor. The indicated CCR5 peptides were immobilized on a CM5 sensor surface of a BIAcore 3000 biosensor followed by injection of 500 nM ␤-arrestin 1 (association phase). After 2 min of association, the surface was washed for 2 min with buffer (dissociation phase), and finally the experimental cycle was terminated by a pulse injection of NaOH/SDS (regeneration). It is shown in the upper sensorgram that phosphorylation of the CCR5 C-terminal cytoplasmic tail peptide only slightly stimulates arrestin binding (see Table II for equilibrium rate constants). Binding of arrestin to an unrelated control peptide or a blank surface was not detectable, revealing the specificity of the ␤-arrestin 1 interaction with the CCR5 CT-2 peptide. As shown in the lower sensorgram, the two peptides IL-2 and CT-1 were capable of binding to ␤-arrestin 1, whereas IL-1 and IL-3 did not interact with ␤-arrestin 1 at all.
for ␤-arrestin binding to synthetic peptides corresponding to the third intracellular loop that, in other receptors and by various different techniques, were shown to directly interact with arrestins (38, 48 -50). The failure of arrestin binding to this part of CCR5 may be due to the fact that the third intracellular loop of all chemokine receptors is unusually short.
The two classical functions of ␤-arrestins in GPCR regulation are receptor desensitization and internalization. Alanine mutation of any single or two serine residues did not affect CCR5 internalization, whereas the simultaneous replacement of three serine residues impaired rapid CCR5 internalization to the same degree as was observed with the fully phosphorylation-deficient receptor. These results indicate a high degree of correlation between the abilities of the various CCR5 mutants to bind ␤-arrestins and to undergo internalization upon receptor activation and thus imply a crucial role for ␤-arrestins in ligand-induced CCR5 endocytosis. As we noted before (20) and confirmed in this study, ␤-arrestins have a pronounced effect on rapid CCR5 internalization within the first 3-10 min but appear to be dispensable for receptor endocytosis at later time points. This suggests that other pathways for CCR5 endocytosis exist that are independent of ␤-arrestins. Such an alternative mechanism may involve caveolae (51).
Although the four C-terminal serine phosphorylation sites are involved in both CCR5 internalization and desensitization, different biochemical mechanisms seem to underlie these two regulatory processes. This is clearly shown with the CCR5-⌬2C mutant, which normally recruits ␤-arrestins to the cell membrane and is internalized upon RANTES stimulation, yet the agonist-induced calcium flux in these cells is significantly sustained, i.e. desensitization is impaired. This indicates that other phosphorylation-dependent mechanisms exist besides ␤-arrestin binding to the receptor that interfere with G protein coupling. In the ␤ 2 -adrenergic receptor system, receptor phosphorylation itself causes only minor impairment of the capacity of the receptor to activate heterotrimeric G proteins, and full desensitization is only achieved with the subsequent binding of ␤-arrestins to the phosphorylated receptor (52). However, different mechanisms appear to operate in the regulation of other GPCR. With increasing levels of rhodopsin phosphorylation, binding affinities for arrestin and G protein are oppositely affected in an incremental manner (32). In the case of the formyl peptide receptor, fully phosphorylated receptors also do not efficiently interact with G proteins independent of arrestin binding (53), but here both G protein and arrestin binding to partially phosphorylated receptors were reported to proceed in an all-or-nothing fashion (33). The results from our own study, which uses the receptor-mediated calcium response and granular enzyme release in whole cells as parameters for receptor desensitization, cannot be directly compared with these in vitro studies. Nonetheless, our findings agree with the concept that receptor phosphorylation alone can impair G protein coupling, independent of ␤-arrestin binding to the receptor.
In summary, we have shown in the present study that agonist-induced binding of ␤-arrestins to CCR5 and receptor internalization are related mechanisms that both require the presence of two C-terminal serine phosphorylation sites, yet the exact position of these sites appears not to be critical. In contrast, receptor desensitization is mediated by a different mechanism that involves phosphorylation of distinct C-terminal sites. Furthermore, we identify a sequence motif within the second intracellular loop of CCR5 that is conserved throughout the GPCR family as an additional ␤-arrestin-binding site.