Recruitment of activated G protein-coupled receptors to pre-existing clathrin-coated pits in living cells.

The process of clathrin-mediated endocytosis tightly regulates signaling of the superfamily of seven-transmembrane G protein-coupled receptors (GPCRs). A fundamental question in the cell biology of membrane receptor endocytosis is whether activated receptors can initiate the formation of clathrin-coated pits (CPs) or whether they are simply mobilized to pre-existing CPs. Here, using various approaches, including a dynamic assay to monitor the distribution of CPs and GPCR-beta-arrestin complexes in live HeLa cells, we demonstrate for the first time that activated GPCRs do not initiate the de novo formation of CPs but instead are targeted to pre-existing CPs.

Many cell surface receptors and membrane proteins are internalized through specialized structures of the plasma membrane, called clathrin-coated pits (CPs), 1 which gradually invaginate and ultimately detach from the plasma membrane forming clathrin-coated vesicles. The endocytic machinery comprises two major structural proteins, clathrin and the adapter protein complex AP2, which plays a central role in CP assembly. A number of more recently identified cytosolic or integral-membrane proteins regulate CP formation and the fission of clathrin-coated vesicles. In the conventional view of constitutive endocytosis, readily accessible tyrosine-or di-leucine-based endocytic motifs, present in the cytoplasmic tails of membrane proteins that are to be internalized, interact with the AP2 adapter complex. In the case of ligand-induced endocytosis, these endocytic signals are cryptic and thought to be unmasked upon receptor activation by its cognate ligand (1)(2)(3).
G protein-coupled receptors (GPCRs), one of the largest families of membrane receptors (4), represent a different model of ligand-induced endocytosis through CPs due to their use of specific adapter proteins called ␤-arrestins that form a bridge between activated GPCRs and the clathrin coat (5). Endocytosis of GPCRs plays an important role in the regulation of their signaling cycle (6). Agonist-stimulated GPCRs initiate cell responses by modulating effector molecules via the activation of heterotrimeric G proteins. These receptors are then rapidly phosphorylated by specific kinases. This phosphorylation promotes the binding of cytosolic nonvisual arrestins (␤-arrestin1 and ␤-arrestin2) to GPCRs (7) resulting in their desensitization (8,9), a transient state during which receptors become refractory to any further stimulation. To recover a full signaling function, GPCRs have to be internalized (10), dephosphorylated in the endosomal compartment (11), and then recycled back to the plasma membrane (12,13). Previous studies have indicated that nonvisual arrestins play the role of adapter molecules during this process (14). The overexpression of arrestins can rescue an endocytosis-defective mutant of the ␤ 2adrenergic receptor, a member of the GPCR family and the overexpression of dominant negative forms of arrestins inhibits GPCR internalization (15). In addition, activated GPCR-arrestin complexes concentrate in punctate areas of the plasma membrane where they colocalize with clathrin (14,16) and AP2 (17,18). The fact that arrestins have been shown to directly interact with both clathrin (14) and AP2 (19) provided a potential mechanism by which GPCRs are internalized through CPs. In this model, arrestins recruit activated GPCRs into CPs by connecting the cytosolic domain of GPCRs to clathrin-coat components. Recent studies confirmed this scenario by showing that mutants of ␤-arrestin2 lacking the AP2 binding motif are unable to efficiently recruit activated GPCRs into CPs (17).
A commonly proposed paradigm for receptor-mediated endocytosis is that the endocytic machinery is recruited to the inner face of the plasma membrane from the cytoplasm by cytoplasmic tails of membrane proteins (nucleation) (20,21). In the case of ligand-activated receptors, such as GPCRs, this would occur after stimulation, permitting CP formation at random sites on the plasma membrane through the association of individual coat components. Recent evidence, however, has challenged this model of nucleation and suggests that regulation of CP levels is independent of endocytic signal levels. In these studies transferrin receptors (22) or chimeric integral membrane reporter proteins containing cytoplasmic endocytic signals (23) were expressed to levels that saturated the internalization machinery, but no evidence to suggest a change in CP levels was found. In addition CPs have been shown to exhibit limited mobility within the membrane and form repeatedly at distinct sites or "hot spots" (24). In the model of pre-existing CPs, the receptor would perhaps be mobilized to these sites upon activation. To investigate this issue in a model of GPCR endocytosis we studied arrestin translocation and redistribution of receptor-arrestin complexes in CPs during the early stages of ligand-mediated endocytosis of the rat thyrotropin-releasing FIG. 1. Time course of TRHR internalization and its colocalization with ␤-arrestin2-GFP following stimulation. a, HeLa cells transiently cotransfected with VSV-TRHR and ␤-arrestin2-GFP constructs were subsequently treated with 40 nM [ 3 H]TRH for different times, and surface and internal [ 3 H]TRH binding was measured. Data are expressed as the percentage of surface receptors at 0 min (cells incubated on ice for the entirety of the experimental period) and represent means ϩ S.D. from three independent experiments performed in triplicate. b-g, HeLa cells transiently cotransfected with VSV-TRHR and ␤-arres-tin2-GFP constructs were fixed and processed for fluorescence microscopy using the P5D4 antibody directed against the VSV epitope and revealed by a Texas Redlabeled secondary antibody. The cells were observed under an epifluorescence microscope attached to a cooled CCD camera. The focus was on the planar plasma membrane adherent to the coverslip. b, d, and f, green fluorescence emitted by GFP in unstimulated cells and in cells stimulated with 10 M TRH for 2 and 15 min, respectively. c, e, and g, red fluorescence emitted by Texas Red corresponding to VSV-TRHR in unstimulated cells and in cells stimulated with 10 M TRH for 2 and 15 min, respectively. Insets show higher magnifications of representative areas of the plasma membrane. Bar, 15 M. In parallel control experiments, TRH had no effect on ␤-arrestin2 distribution in HeLa cells transfected with ␤-arrestin2-GFP alone (i.e. without TRHR, data not shown).
hormone receptor-1 (TRHR). In particular, we have conducted dynamic studies in living cells expressing a ␤-arrestin 2-enhanced green fluorescent protein (GFP) and a chimera between Eps15, a constitutive component of CPs (25,26), and the red fluorescent protein (RFP) Ds-Red. Our results show that agonist-activated TRHRs do not promote the formation of specific

EXPERIMENTAL PROCEDURES
Materials-Cell culture media and fetal bovine serum were from Life Technologies. TRH, neomycin, and 100/2 antibody were purchased from Sigma. FuGENE 6 reagent and monoclonal anti-GFP antibody were from Roche Molecular Biochemicals. Mouse monoclonal antibodies AP.6 (anti-AP2 (27)) and OKT9 (anti-human transferrin receptor (TfR)) were obtained from ATCC. Monoclonal antibody P5D4 (IgG1) against the YTDIEMNRLGK epitope of the G protein of vesicular stomatitis virus (VSV) has been described previously (28). Rabbit polyclonal anti-Eps15 antibody was obtained as in Ref. 29 using GST-Eps15 proteins. Texas Red and Alexa 594-conjugated goat anti-mouse, goat anti-rabbit immunoglobins, and Alexa 594-conjugated Tf were obtained from Molecular Probes. [ 3 H]TRH (specific radioactivity 98 Ci/mmol) was from PerkinElmer Life Science.
Plasmids-The construction of the VSV-TRHR plasmid has been described previously (30) and was a kind gift from Dr G. Milligan. p␤-arrestin2-EGFP was generated by excising the ␤-arrestin2 cDNA fragment from p␤-arrestin2-YFP (31) (generously provided by S. Angers) using NheI and cloning this into the NheI site of EGFPN1 (CLON-TECH). p␤-Arrestin1-EGFP was generated by excising the XhoI/SacII fragment (containing amino acids 1-414 of bovine ␤-arrestin1) from pBC.␤-arrestin (7) and cloning this into the XhoI/SacII sites of EGFPN1. pRFP-Eps15 was generated by performing PCR of the first 944 bases of Eps15, corresponding to the NH 2 -terminal 314 amino acids, introducing a BamHI site in the upper primer. The PCR product was digested with BamHI and HindIII and cloned into pGEX5.1-Eps15⌬P/C-(32) digested BamHI/HindIII to create pGEX5.1-Eps15⌬P/ C-2. The full-length Eps15⌬P/C cDNA was then excised from pGEX5.1-Eps15⌬P/C-2 using BamHI/XhoI and subsequently introduced downstream and in frame of DsRed using the BglII/SalI sites of pDsRed1C1 (CLONTECH). Constructs were verified by nucleotide sequencing (Institut Cochin de Génétique Moléculaire sequencing facility). Sequences of the primers used to make RFP-Eps15 are available on request.
Transfections and Immunofluorescence-HeLa cells were grown in DMEM supplemented with 10% fetal calf serum, 2 mM L-glutamine, penicillin, and streptomycin. Subconfluent HeLa cells were used for transient expression of the different constructs. Transfections were performed using the FuGENE 6 Transfection System according to the manufacturer's instructions. For immunofluorescence studies, HeLa cells were seeded on coverslips in six-well plates, transfected with 2 g of VSV-TRHR and 0.2 g ␤-arrestin2-GFP the following day, and used for immunofluorescence 1 day post-transfection. Cells were fixed and processed for fluorescence microscopy as described previously (32). Endocytosis of Alexa594-conjugated Tf was performed on HeLa cells grown on coverslips 1 day after transfection. The cells were first incubated for 20 min at 37°C in DMEM 20 mM Hepes, pH 7.2, to eliminate receptorbound Tf and then incubated in DMEM 20 mM Hepes, pH 7.2, 1 mg/ml bovine serum albumin containing 100 nM Alexa594-conjugated Tf. After incubation at 37°C for 15 min, the cells were rapidly cooled to 4°C, washed twice in cold PBS, and then fixed as described above. Samples were examined either under an epifluorescence microscope (Zeiss) attached to a cooled CCD camera (Spot 2, Diagnostic Instruments) or under a confocal microscope (LSM 510, Zeiss). HeLa cells stably expressing ␤-arrestin2-GFP were generated by standard procedures using G418 (600 g/ml) selection.
Subcellular Fractionation, Gel Electrophoresis, and Immunoblotting-Subconfluent HeLa cells stably expressing ␤-arrestin2-GFP in 10-cm cell culture dishes were transiently transfected with 10 g VSV-TRHR using FuGENE 6. 48 h after transfection the cells were incubated with TRH at 37°C for 2 or 5 min as indicated in the figure legend. Following the appropriate treatments, cells were rinsed twice in ice-cold PBS and detached using 10 mM EDTA in PBS. After pelleting at 1000 ϫ g, cells were resuspended in PBS containing Complete TM protease inhibitors mixture (Roche Molecular Biochemicals). Homogenization was performed by six passes through a ball-bearing homogenizer (cell cracker EMBL) with a clearance of 10 M. Crude membranes were prepared by centrifugation at 45,000 ϫ g in a Sorvall SS-34 rotor for 20 min. Insoluble fractions were collected and resolved by electrophoresis on 7% SDS-polyacrylamide gels and transferred onto Immobilon (Millipore) membranes. Western blot analysis was performed using monoclonal antibodies specific for either AP2 (100/2, 1 in 2000 dilution) or GFP (1 in 1000 dilution). Horseradish peroxidase-conjugated secondary antibodies were visualized using a chemiluminescent detection system (Pierce). Quantification of bands was performed using NIH Image 1.62 software.
Confocal Laser Microscopic Imaging of ␤-Arrestin2-GFP and RFP-Eps15 in Living Cells-For real-time imaging purposes, HeLa cells were seeded on coverslips in six well plates, transfected the following day with 2 g of VSV-TRHR, 0.2 g of ␤-arr2-GFP and 0.2 g RFP-Eps15, and used for imaging 2 days post-transfection. Cells on coverslips were mounted on the imaging chamber and maintained in Hepesbuffered medium at 37°C. The cells were observed with a laser scanning confocal microscope (Bio-Rad MRC 1024) using a 40ϫ oil immersion objective and electronic zoom of 2. Images were acquired sequentially every 5 s using simultaneous excitation at 488 and 568 nm. 10 M TRH was applied directly over the selected cells following acquisition of the second image. The images were analyzed using LaserSharp 3.2 software.
Internalization of TRHR as Assessed by Radioligand Binding-Internalization of TRHR was performed essentially as described in Ref. 33. Briefly, subconfluent HeLa cells in 10-cm cell culture dishes were transiently transfected with 10 g of VSV-TRHR and 1 g of ␤-arres-tin2-GFP using FuGENE 6. 24 h after transfection cells were trypsinized and seeded into 12-well plates (one 10-cm dish was split into 18 wells). The day after, DMEM was removed from the cells and replaced with 0.5 ml Hepes-buffered DMEM (HD medium: serum-free DMEM, 20 mM Hepes, pH 7.2) per well, and cells were incubated at 37°C. 40 min later, plates were placed on ice and cooled for 10 min.

RESULTS AND DISCUSSION
To investigate the issue of whether ligand-activated GPCRs can initiate the formation of new CPs or whether they are recruited to pre-existing CPs, we utilized a system that would result in a high density of internalized GPCR following receptor activation. To this end, we chose to study HeLa cells overexpressing a vesicular stomatitis virus epitope-tagged form of TRHR (VSV-TRHR). Previous studies have demonstrated that the TRHR is endocytosed by a ␤-arrestin-(33, 34) and clathrindependent (35,36) mechanism. This GPCR was chosen, as in preliminary experiments using real-time confocal microscopy it was found to be one of the best GPCR translocators of ␤-arrestins to the plasma membrane among several GPCRs tested including the ␤ 2 -adrenergic receptor, the receptor for the chemoattractant formyl-methionyl-leucyl-phenylalanine, and the chemokine receptors CXCR4 and CCR5 (data not shown). In addition, the TRHR also belongs to the recently described class of GPCRs, which forms stable complexes with both ␤-arrestins1 and -2 and traffics together with ␤-arrestins in endocytic vesicles (37), allowing ␤-arrestins to act as markers of GPCRs during the entire endocytic process. VSV-TRHR was cotransfected in combination with a chimeric protein, consisting of the rat ␤-arrestin2 cDNA fused to the NH 2 -terminal end of GFP (␤-arrestin2-GFP). In this system the endocytosis rate of the TRHR (t1 ⁄2 of ϳ10 min, Fig. 1a) is similar to that obtained in rat pituitary GH 4 C 1 cells expressing endogenous TRHR (38). The cellular distribution of TRHR and ␤-arrestin2 was analyzed by immunofluorescence: VSV-TRHR was detected using the P5D4 monoclonal antibody specific to the VSV-tag, and ␤-arrestin2-GFP was directly visualized by the green fluorescence emitted by GFP. This approach therefore allowed us to observe a TRHdependent membrane recruitment of ␤-arrestin2 as described previously in other cell systems.
Trafficking of TRHR in HeLa Cells-Previous studies have reported that ␤-arrestin-GFP chimeras maintain their function in receptor endocytosis (39) and may be used as markers of GPCR endocytosis (37,40,41). To determine whether ␤-arrestin and TRHR traffic together as a complex after receptor stimulation in our model, their relative cellular distributions were studied by immunofluorescence microscopy. The images shown focus on the plasma membrane of cells that are adherent to the coverslip as described previously (32). Under basal conditions, ␤-arrestin2 displayed a characteristic diffuse cytosolic distribution (39) (Fig. 1b), whereas the TRHR was evenly distributed at the cell surface (Fig. 1c). Following TRHR stimulation by TRH for 2 min, ␤-arrestin2 (Fig. 1d) translocated to the plasma membrane as a punctate staining. The TRHR also displayed a punctate staining at the plasma membrane (Fig. 1e) that completely colocalized with ␤-arrestin2. Following longer term stimulation (15 min), ␤-arrestin2 (Fig. 1f) and TRHR (Fig. 1g) were found colocalized in internal endosomal structures, demonstrating their continued presence in a shared endocytic pathway. ␤-Arres-tin2 could therefore be used as a marker of TRHR in this FIG. 4. ␤-Arrestin2-GFP, but not AP2, is mobilized to the plasma membrane following TRHR activation. a, quantification of AP2 dots over a defined magnified area in cells before and 2 min after stimulation was performed by manual counting. b-e, HeLa cells stably expressing ␤-arrestin2-GFP were transiently transfected with VSV-TRHR. Unstimulated cells and cells 2 and 5 min after stimulation were separated into membrane and cytoplasmic fractions, and equal amounts of membrane fraction were subjected to SDS-PAGE and immunoblotted with antibodies to AP2 (b) or GFP (d). Immunoreactive bands were visualized using a chemiluminescent detection system. Data in c and e represent the mean ϩ S.E. from three individual experiments performed on two independent clones. model to follow its cell trafficking upon activation, as reported previously by Groarke et al. (34), in HEK-293 cells. This approach was preferred, in this study, instead of the direct labeling of TRHR with GFP, because it avoids any potential artifact that could arise from the fusion of GFP to the COOH-terminal tail of the receptor.
To determine whether the punctate areas where ␤-arrestin2 and TRHR colocalize at early times of receptor stimulation correspond to CPs, we revealed the localization of AP2 complexes using the anti-AP-2, AP6 antibody. The cellular distribution of ␤-arrestin2 relative to AP2 under basal conditions and after TRHR stimulation was thus monitored. In untreated cells, AP2 displayed a punctate staining (Fig. 2b) characteristic of plasma membrane CPs (18,23,32,42), and ␤-arrestin2 demonstrated a diffuse cytoplasmic staining (Fig. 2a). There was no colocalization of the two different proteins under these conditions. Two minutes after TRHR activation, however, ␤-ar-restin2 (Fig. 2c) was found to be completely colocalized with AP2 (Fig. 2d). Similar results were also observed with other specific CP markers, Eps15 (Fig. 2, e and f) and clathrin (data not shown). In an attempt to quantify this colocalization, the number of ␤-arrestin2-GFP and AP2 immunofluorescent dots in a defined area in the same cells 2 min after TRHR stimulation were counted. We counted more than 250 AP2 dots from four different preparations and found no significant difference in the number of ␤-arrestin2 dots. In addition, each ␤-arrestin dot was found to correspond to an AP2 dot. These results indicate that following TRHR stimulation, ␤-arrestin2 and TRHR are mobilized to CPs and that all CPs contained ␤-ar-restin2 in agreement with studies investigating ␤ 2 -adrenergic receptor redistribution (16). In contrast, at later time points (15 min; Fig. 2, g and h) there was a loss of ␤-arrestin2-GFP colocalization with AP2 due to movement of the TRHR/␤-arres-tin2 complex to the endosomal compartment (see Fig. 3

, lower panel).
To determine whether the areas that ␤-arrestin2 was mobilized to upon TRHR stimulation were areas of the plasma membrane undergoing active endocytosis, we compared the distribution of cell surface transferrin receptors (TfR), which undergo constitutive endocytosis, and ␤-arrestin2. In control cells, the TfR displayed a punctate staining due to its concentration in CPs (Fig. 3b), and ␤-arrestin2 demonstrated a diffuse cytosolic staining (Fig. 3a). Following activation of TRHR, ␤-ar-restin2 was found to colocalize with most TfR clusters, indicating that the TRHR-␤-arrestin2 complex mobilizes to the same areas of the membrane where TfR are undergoing constitutive endocytosis. This is in agreement with studies investigating the GPCR protease-activated receptor-1 in HeLa cells, which indicate a greater than 90% colocalization of protease-activated receptor-1 with TfR following short term stimulation of protease-activated receptor-1 (43). The localization of ␤-arrestin following longer TRHR stimulation was also investigated using Alexa-594-labeled transferrin to stain the early/recycling endosomal compartment. At these later times ␤-arrestin was colocalized with transferrin in perinuclear regions of the cell (Fig.  3, lower panel, 15 min TRH, yellow color) in agreement with results obtained in HEK-293 cells (33).
TRHR-␤-Arrestin2 Complex Is Recruited to Pre-existing CPs-Significant cytosolic pools of clathrin and AP2 exist in cells that are often equivalent to or exceed membrane-bound forms (23, 44 -46). In the nucleation model of GPCR endocytosis, agonist activation of receptor would promote CP formation, and this would be expected to result in an increase of total CPs. This would result in a recruitment of AP2 from the cytosolic to the plasma membrane pool giving an increase in AP2 levels at the membrane. We used three different approaches to investi-gate if the TRHR is targeted to pre-existing CPs or initiates CP assembly upon activation. First, the number of CPs (as revealed by AP2 dots) was counted in a defined area of the plasma membrane in cells expressing TRHR and ␤-arrestin2-GFP in control (unstimulated) cells and after TRHR stimulation. No difference was found in the number of CPs following stimulation, indicating a lack of novel CP formation (Fig. 4a). For the second approach HeLa cells stably expressing ␤-arres-tin2-GFP were used. These cells were transiently transfected with VSV-TRHR. Crude membrane fractions of unstimulated and stimulated cells (2 and 5 min) were prepared, followed by SDS-PAGE resolution and immunoblotting with an antibody against the ␣ chain of AP2. We could detect no difference in the levels of membrane or soluble AP2 after stimulation using this approach (Fig. 4, b and c, and data not shown), whereas in the same samples a clear increase in membrane ␤-arrestin2-GFP was observed (Fig. 4, d and e) with a corresponding decrease in the soluble fraction (not shown). These data are not consistent with the idea that ␤-arrestin drives the formation of de novo CPs upon TRHR activation but do agree with the model in which the receptor-␤-arrestin complex is mobilized to pre-existing CPs. Finally, to confirm that ␤-arrestin2 is recruited to pre-existing CPs using a dynamic in vivo assay, we utilized a triple transfection model in which ␤-arrestin2-GFP and VSV-TRHR were co-expressed in combination with RFP-Eps15. Eps15 is a constitutive component of CPs (25,26) but not clathrin-coated vesicles (47) and could therefore be used as a marker of "pre-existing" CPs, i.e. CPs present before TRHR stimulation. RFP-Eps15 consists of the Eps15 cDNA lacking amino acids 763-896 (E⌬P/C), which is targeted to CPs and does not interfere with CP assembly (32) fused to the COOH terminus of the newly identified RFP from Discosoma sp. As expected, RFP-Eps15 displayed a bright membrane punctate staining (Fig. 5) that colocalized with both AP2 and clathrin (not shown). This enabled the concomitant monitoring of ␤-ar-restin2-GFP and CPs by real-time confocal immunofluorescence microscopy in living cells following ligand activation of TRHR. Using this approach we predicted that pre-existing CPs shown by RFP-Eps15 that were present before TRHR stimulation would also be stained with ␤-arrestin2-GFP after TRH stimulation if ␤-arrestin is recruited to pre-existing CPs. However, if the nucleation of CPs was driven by the TRHR-␤-arrestin complex we expected that novel RFP-Eps15 areas containing ␤-arrestin2-GFP would appear following TRHR stimulation. An acquisition of the cells was taken every 5 s following receptor stimulation. After as little as 10 -20 s ␤-ar-restin2-GFP could begin to be seen to colocalize with RFP-Eps15-stained areas of the membrane (Fig. 5, upper panel, 10 -20 s) that were present before stimulation and after 25-30 s (Fig. 5, upper panel, 25-30 s) was completely colocalized with these regions. We observed no novel RFP-Eps15-stained areas of the membrane following receptor stimulation. These results therefore clearly indicate that ␤-arrestin2 is translocated to pre-existing CPs upon TRHR stimulation. Similar results were obtained using a ␤-arrestin1-GFP chimera (Fig. 5, lower panel). Taken together the data provide compelling evidence that the TRHR-␤-arrestin complex does not initiate the novel formation of CPs, but that it is mobilized to pre-existing CPs. This model agrees with studies, which indicate that CPs do not form at sites of FcepsilonRI receptor activation (42). Furthermore, recent studies using a GFP-labeled form of clathrin light chain indicate that CP formation occurs repeatedly at defined sites or hot spots on the plasma membrane of cultured cells while excluding other areas (24). The same phenomenon exists in synapses where the endocytic machinery for synaptic vesicle recycling is found localized at membrane-associated hot-spots surrounding areas of exocytosis (48). The fact that GPCRs do not promote the nucleation of "de novo" CPs may explain why their endocytosis is saturable in some cell types even in the presence of overexpressed ␤-arrestin (49), the limiting factor being the amount of constitutive CPs where GPCRs may be recruited through ␤-arrestin. In conclusion, therefore, the cur-rent study provides novel insights into the regulation of TRHR endocytosis, which directly demonstrates for the first time that a GPCR-␤-arrestin complex is targeted to pre-existing CPs. Since the majority of GPCRs studied so far undergo ligand-and ␤-arrestin-dependent endocytosis, this model is of potential interest for other GPCRs.