Regulation of Membrane Cholecystokinin-2 Receptor by Agonists Enables Classification of Partial Agonists as Biased Agonists*

Given the importance of G-protein-coupled receptors as pharmacological targets in medicine, efforts directed at understanding the molecular mechanism by which pharmacological compounds regulate their presence at the cell surface is of paramount importance. In this context, using confocal microscopy and bioluminescence resonance energy transfer, we have investigated internalization and intracellular trafficking of the cholecystokinin-2 receptor (CCK2R) in response to both natural and synthetic ligands with different pharmacological features. We found that CCK and gastrin, which are full agonists on CCK2R-induced inositol phosphate production, rapidly and abundantly stimulate internalization. Internalized CCK2R did not rapidly recycle to plasma membrane but instead was directed to late endosomes/lysosomes. CCK2R endocytosis involves clathrin-coated pits and dynamin and high affinity and prolonged binding of β-arrestin1 or -2. Partial agonists and antagonists on CCK2R-induced inositol phosphate formation and ERK1/2 phosphorylation did not stimulate CCK2R internalization or β-arrestin recruitment to the CCK2R but blocked full agonist-induced internalization and β-arrestin recruitment. The extreme C-terminal region of the CCK2R (and more precisely phosphorylatable residues Ser437-Xaa438-Thr439-Thr440-Xaa441-Ser442-Thr443) were critical for β-arrestin recruitment. However, this region and β-arrestins were dispensable for CCK2R internalization. In conclusion, this study allowed us to classify the human CCK2R as a member of class B G-protein-coupled receptors with regard to its endocytosis features and identified biased agonists of the CCK2R. These new important insights will allow us to investigate the role of internalized CCK2R·β-arrestin complexes in cancers expressing this receptor and to develop new diagnosis and therapeutic strategies targeting this receptor.

and/or expression of a constitutively active variant may contribute to human diseases (10). These findings have stimulated interest in the identification of antagonists of CCK2R. To date, a large panel of chemically distinct CCK2R antagonists have been discovered and used to assess the functions mediated by CCK2R in animals and humans (13). Furthermore, several of these compounds have reached clinical evaluation stages for indications such as anxiety and panic disorders, sleep disorders, drug dependence, pain, gastroesophagus reflux, and gastric secretion disorders (13). However, some reference molecules believed to be pure antagonists turned out to be endowed with some agonist activity in the stomach and pancreas as well as cells expressing CCK1R or CCK2R (14).
In previous works, we investigated the intrinsic activation mechanism of the CCK2R and the regulation of this activity by different ligands (15)(16)(17). This led us to understand how two structurally close nonpeptide ligands display opposite activities (partial agonist and inverse agonist) (16). Moreover, we delineated the role and mechanism of action of RGS-2 (regulator of G-protein signaling-2) in CCK2R-induced inositol phosphate production (17). However, the mechanism and consequences of CCK2R regulation at the cell surface by its natural ligands or by synthetic ligands are not yet precisely known (18,19).
In the current study, given the importance of cell surface regulation of GPCR by pharmacological agents, we investigated the mechanisms whereby the CCK2R is regulated after stimulation by its natural agonist ligand CCK or after exposure to several synthetic ligands displaying distinct pharmacological activity with respect to CCK2R-mediated production of inositol phosphates. Results from this study led us to classify the CCK2R as a member of the class B GPCRs with respect to its endocytosis features. Furthermore, this study led to identification of biased agonists that are partial agonists on CCK2R-induced inositol phosphate production and antagonists on ␤-arrestin recruitment and subsequent receptor endocytosis. The C-terminal region of the CCK2R and more precisely phosphorylatable motif Ser 437 -Xaa 438 -Thr 439 -Thr 440 -Xaa 441 -Ser 442 -Thr 443 were identified as critical determinants for ␤-arrestin recruitment. However, this motif and ␤-arrestins were dispensable for CCK2R internalization. By providing new important insights into the regulation mechanism of the CCK2R by natural agonists and pharmacological agents, this study represents a solid basis to investigate the contribution of internalized CCK2R⅐␤-arrestin complexes to CCK2R-induced intracellular signaling in cancer cells that endogenously express this receptor. It will also allow us to develop new diagnosis and therapeutic strategies based on the targeting of the CCK2R.
Generation of MEF Cells Expressing the CCK2R-Replication-defective, self-inactivating lentiviral vectors were generated in a BSL-3 facility (BiviC core vector production, IFR 150, Toulouse, France) through transient transfection of HEK 293FT cells with packaging and lentiviral vector plasmids using polyethyleneimine. Lentiviral vectors were concentrated by centrifugation using Vivaspin columns and stored at Ϫ80°C. All batches were checked to be replicative virus-free. The viral titers were determined on HT1080 cells and expressed in transduction units/ml. In addition, vector concentrations were quantified by a p24 enzyme-linked immunosorbent assay (ELISA) (Innotest, Ingen, Paris). To generate MEF cells expressing the CCK2R, cells were incubated with the lentiviral vectors (multiplicity of infection ϭ 1) for 24 h and subsequently selected with Zeocin (Invivogen, Toulouse, France) at 25 g/ml for 2 weeks. Efficiency of viral transduction and selection was estimated by measuring the percentage of cells expressing GFP by flow cytometry analysis. Expression of CCK2R was assessed by binding Alexa F 647-CCK.
Receptor Binding Assay-Cells were plated onto 10-cm culture dishes and grown in DMEM containing 10% FBS in a 5% CO 2 atmosphere at 37°C. After overnight incubation, cells were transfected with 1 g/plate (excepted when mentioned) of pcDNA5/FRT, pEGFP-N1, and pRluc-N1(h) vectors containing the cDNA for the wild-type or mutated CCK2Rs using a Fugene 6 transfection reagent (Roche Applied Science). Cells were transferred to 24-well plates 24 h after transfection. HEK 293 cells stably expressing CCK2R (Flp-In TM CCK2R-293) were directly plated onto 24-well plates. Approximately 24 h after transfer to 24-well plates, binding assays were performed using 125 I-CCK according to the protocol previously described in detail (16). Receptor density (B max ) and K d were calculated from homologous 125 I-CCK competition binding experiments using Ligand software (Kell, Cambridge, UK). K i values for competitors were calculated using the non-linear curve fitting software GraphPad Prism (GraphPad Software, San Diego, CA).
Internalization Assay by Flow Cytometry-Flp-In TM CCK2R-293 cells were plated onto poly-L-lysine (Sigma-Aldrich)-coated 6-well plates and grown in DMEM containing 10% FBS in a 5% CO 2 atmosphere at 37°C. After overnight incubation, cells were treated with CCK (0.1 M) or synthetic ligands (1 M) in DMEM containing HEPES (20 mM) at various times. Following stimulation, bound ligand was dissociated from the cells three times with DMEM/HEPES (20 mM), and the remaining receptors at the cell membrane were labeled using a saturating concentration of Alexa F 647-CCK (1 M) for 2.5 h on ice. Cells were washed twice with PBS, 0.1% BSA. Cells were detached from 6-well plates and transferred to FACS tubes. Cell-associated fluorescence was determined using a BD FACSCalibur TM flow cytometer, with Flp-In TM 293 cells that do not express CCK2R as a negative control.
Recycling Assay by Flow Cytometry-Flp-In TM CCK2R-293 cells were plated onto poly-L-lysine (Sigma-Aldrich)-coated 6-well plates and grown in DMEM containing 10% FBS in a 5% CO 2 atmosphere at 37°C. After overnight incubation, cells were treated with 0.1 M CCK in DMEM/HEPES (20 mM) for 1 h at 37°C. Following stimulation, cells were washed three times with DMEM/HEPES. The receptors at the cell surface were labeled using Alexa F 647-CCK (1 M) for 45 min at 37°C at various times (1, 2, and 3 h) after the end of the CCK stimulation. Binding of Alexa F 647-CCK was carried out in the presence of dynasore (80 M) in order to block internalization and possible fast recycling during this step of receptor quantification.
BRET Studies-HEK 293 T or Flp-In TM 293 cells were plated onto 10-cm culture dishes and grown in DMEM containing 10% FBS in a 5% CO 2 atmosphere at 37°C. After overnight incubation, cells were transfected with 0.2 g of Rluctagged CCK2R (wild type or mutants) and 10 g of either ␤-arrestin1-YFP or ␤-arrestin2 using polyethyleneimine transfection reagent (Polyplus) or Fugene 6 transfection reagent (Roche Applied Science). 24 h after transfection, cells were plated in 96-well clear bottom plates (Corning Glass) at a density of 100,000 cells/well in phenol red-free DMEM. After overnight incubation, the phenol red-free medium was removed from HEK293T cells and replaced by PBS. The BRET assay was initiated by adding 10 l of the cell-permeant substrate specific for Renilla luciferase, coelenterazine H, to the well to yield a final concentration of 5 M. The agonist/ antagonist activity of the compounds was measured by adding them to the well 5 min after the Rluc substrate. Readings started 10 min after the addition of Rluc substrate. BRET experiments were performed at room temperature in a Mithras LB940 instrument (Berthold) that allows the sequential integration of the signals detected in the 465-505 nm and 515-555 nm windows using filters with the appropriate band pass and by using MicroWin 2000 software. For titration experiments, the acceptor/donor ratio was calculated as described previously (25). For time course analysis of the interactions between CCK2R and ␤-arrestin1/2 interactions, coelenterazine H was added 10 min before the addition of PBS, CCK, or compounds. Readings were then collected at 12.5-s intervals for the next 20 min.

Characterization of Flp-In TM CCK2R-293
Cells-To investigate the regulation of the CCK2R at the cell surface, we established a HEK 293 cell line that permanently expressed the human CCK2R and was named Flp-In TM CCK2R-293. Homologous competition binding studies with these cells indicated that expressed CCK2R bound 125 I-CCK with a dissociation constant (K d ) of 4.8 Ϯ 1.0 nM and a maximum binding capacity of 1.98 Ϯ 0.18 pmol/10 6 cells. The CCK2R displayed a typical CCK2R-like pharmacology, as shown by the high inhibition constant of selective CCK2R ligands (K i values as follows: gastrin, 13.2 Ϯ 1.5 nM; JB93,242, 40.6 Ϯ 4.7 nM; JB93,182, 4.2 Ϯ 0.3 nM; PD135,158, 11.7 Ϯ 0.8 nM) and the low inhibition constant of the specific CCK1R antagonist SR27,897 (K i Ն 10 M) (Fig. 1, A and B). Furthermore, the CCK2R efficiently coupled with phospholipase C as shown by the dose-response curve for CCK-induced inositol phosphate production (EC 50 of 4.0 Ϯ 0.4 nM, maximal production ϳ50fold basal level after 1 h of stimulation). In agreement with our previous results (14,16), JB93,242 and PD135,158 behaved as partial agonists with respect to inositol phosphate turnover. These two compounds showed efficacies that repre-sented 22 and 42% of that of the full agonists CCK and gastrin (Fig. 1C). However, JB93,182, which acted as a partial inverse agonist on COS-7 cells overexpressing the CCK2R, did not exhibit this activity in Flp-In TM CCK2R-293 due to the lack of detectable basal constitutive activity of the CCK2R in HEK cells (14) (Fig. 1C). Finally, the effect of nonpeptide ligands on ERK1/2 phosphorylation was evaluated because the activation of the CCK2R is recognized to trigger this signaling pathway (10). As shown in Fig. 1D, CCK-stimulated phosphorylation of ERK1/2 was rapid and sustained over time. JB93,182, JB93,242, and PD135,158 also stimulated phosphorylation. Autoradiography scanning indicated that maximal ERK1/2 phosphorylation achieved with JB93,182, JB93,242, and PD135,158 represented ϳ15, 30, and 60 -70% of the maximum achieved with CCK, respectively (not shown).
Characterization of CCK2R Internalization-Endocytosis and trafficking of the CCK2R were traced using fluorescent CCK2R (CCK2R-GFP) and fluorescent CCK (Alexa F 647-CCK). Prior to the use of these tools, biological experiments were carried out in order to check their pharmacological features (supplemental Fig. 1, A and B). When transiently expressed in HEK 293 cells, CCK2R-GFP bound 125 I-CCK with a dissociation constant of 1.15 Ϯ 0.20 nM (versus 1.17 Ϯ 0.25 nM for CCK2R) and efficiently stimulated production of inositol phosphates after CCK stimulation (EC 50 of 2.00 Ϯ 0.23 nM versus 0.53 Ϯ 0.04 nM for CCK2R). Furthermore, Alexa F 647-CCK competed with 125 I-CCK to CCK2R similarly to unmodified CCK (K i ϭ 2.4 Ϯ 0.1 nM for Alexa F 647-CCK inhibited 1251-CCK binding to CCK2R with the same potency as native CCK. Confocal microscopy observation of living HEK 293 cells transiently expressing CCK2R-GFP showed that CCK2R was expressed uniformly at the cell surface as the unmodified CCK2R immediately after its labeling by Alexa F 647-CCK (Fig. 2). However, soon after the addition of CCK or Alexa F 647-CCK, the membrane fluorescence was relocated in numerous membrane clusters and progressively entered the interior of the cells in punctuate peripheral vesicles. The number of these fluorescent vesicles continuously increased to the detriment of membrane fluorescence, whereas they became more widely distributed within the cytoplasm, including in perinuclear areas, over the time. During CCK2R internalization, the ligand remained trapped together with the receptor in endocytosic vesicles for periods of time as long as 3 h (Fig.  2, merged panels). Gastrin-or CCK-mediated internalization was blocked by incubating the cells at 4°C. PD135,158, a partial agonist of the CCK2R exhibiting about 40% of maximal CCK-induced inositol phosphate production (Fig. 1C), did not stimulate internalization of CCK2R-GFP, even at supramaximal concentrations, but fully inhibited CCK-induced internalization of CCK2R-GFP. The nonpeptide ligands of the CCK2R, JB93,242 and JB93,182, characterized as a partial agonist and an antagonist/inverse agonist, respectively, were also unable to stimulate CCK2R-GFP internalization and abolished internalization of the receptor induced by CCK (Fig.  2B). The CCK1R antagonist SR27,897 neither induced CCK2R-GFP internalization nor prevented receptor internal-ization induced by CCK. Last, gastrin stimulated CCK2R-GFP internalization as did CCK.
CCK2R internalization was quantified by measuring binding of Alexa F 647-CCK to transfected cells prestimulated with CCK for different times. As shown in Fig. 3A, prestimulation with CCK resulted in a rapid decrease of the amount of fluorescent CCK bound to the cells, indicating that CCK2R progressively internalized during prestimulation with CCK. Half of the CCK2R population was not accessible to Alexa F 647-CCK after 60 min of incubation, a result in satisfactory agreement with confocal microscopy observations ( Fig. 2A). In contrast, prestimulations with PD135,158, JB93,242, or JB93,182 did not significantly alter the presence of CCK2R at the cell surface. We further examined the possibility of the population of CCK2R at the cell surface containing a pool of recycled CCK2R in addition to CCK2R that have not yet been internalized. For this purpose, Flp-In TM CCK2R-293 cells stimulated with 0.1 M CCK at 37°C for 60 min were then incubated at 37°C for various times in the absence of CCK, after which CCK2R density was measured using Alexa F 647-CCK. As shown in Fig. 3B, for times up to 3 h, the binding of Alexa F 647-CCK remained stable, which suggested that internalized CCK2R was not significantly recycled at the cell surface within this period of time. This absence of apparent CCK2R recycling within a period shorter than 3 h was confirmed in experiments aimed at quantification of internalization in the presence of the recycling inhibitor, monensin, which had no significant impact on receptor density at the plasma membrane after a challenge with CCK (Fig. 3B). However, as expected, in the presence of monensin, the size of endocytosis vesicles increased (26).
The trafficking and fate of internalized CCK2R under prolonged stimulation was investigated by co-expressing CCK2R-GFP and fluorescent Rab proteins, which are small GTPases regulating intracellular trafficking between functionally distinct membrane compartments within the cell (27). As shown in Fig. 4, following the addition of CCK, CCK2R-GFP immediately and exclusively co-localized in Rab5-illuminated vesi-   cles probably corresponding to early endosomes. Indeed, colocalization with Rab5 was seen in 40% of cells immediately after the addition of CCK; it was maximal (100%) after 1 h and decreased to 70% at 2 h. Localization of CCK2R-GFP in late endosomes or lysosomes containing Rab7 was observed at times longer than 5-6 h poststimulation (80% of cells with co-localization). Localization of CCK2R-GFP to slow recycling endosomes expressing Rab11 was obvious at times longer than 2 h (85% of cells with co-localization) and was maintained for times up to 6 h (100% of cells with co-localization). These co-localization experiments are in line with results indicating that internalized CCK2R was not rapidly recycled to the cell surface but was directed to late endosomes for possible slow recycling or degradation. CCK2R Internalization Involves the GTPase Dynamin-and Clathrin-coated Pits-Internalization of many GPCRs following agonist stimulation occurs through clathrin-coated pits and involves recruitment of ␤-arrestins, the clathrin adaptor AP-2, and the GTPase dynamin (1). We therefore tested whether CCK2R used this major route for internalization. The dependence of CCK2R internalization on dynamin was investigated by using the dynamin inhibitor, dynasore (28), which efficiently abolished internalization, as shown by cell membrane labeling by Alexa F 647-CCK (Fig. 5A). In the pres-ence of chlorpromazine, an inhibitor of clathrin-coated pit formation (29), at a concentration (100 M) that inhibits Alexa Fluor 488-transferrin endocytosis (not shown), CCK2R labeled by Alexa F 647-CCK failed to internalize (Fig. 5A). Hence, CCK2R internalization in HEK cells is dependent on the membrane GTPase dynamin and most likely occurs through clathrin-coated pits.
CCK2R Internalization Involves ␤-Arrestins-In order to verify whether nonvisual ␤-arrestins are involved in CCK2R internalization, we first expressed ␤-arrestin1-GFP or ␤-ar-restin2-GFP in Flp-In TM CCK2R-293 cells. In the absence of stimulation, arrestin2-GFP was localized uniformly in the cytoplasm of the cells (Fig. 5B). ␤-Arrestin1-GFP was also located in the cytoplasm and slightly in the nucleus (supplementary Fig. 2). The addition of CCK caused rapid illumination (at 5 min) of the cell periphery by the green proteins supporting the translocation of cytosolic ␤-arrestin2-GFP or ␤-arrestin1-GFP to the plasma membrane. This recruitment of ␤-arrestins to the plasma membrane was quantified by measuring the decrease of cytosolic fluorescence. As shown in Fig. 5C, kinetics of ␤-arrestin2-GFP or ␤-arrestin1-GFP recruitment to the plasma membrane were similar and so rapid that half of the tagged arrestins initially seen in the cytoplasm were recruited to the plasma membrane
␤-Arrestins Are Directly Recruited by CCK2R upon Stimulation by Full Agonists-Because ␤-arrestins were involved in CCK-induced CCK2R internalization and were recruited to the plasma membrane prior to CCK2R internalization, we determined whether recruitment of ␤-arrestins was due to a direct interaction with the CCK2R and occurred only under full agonist stimulation. For this purpose, a BRET assay was performed using CCK2R-Rluc as the BRET donor and ␤-ar-restin2-YFP or ␤-arrestin1-YFP as the BRET acceptors coexpressed in HEK 293 cells. Control binding experiments and inositol phosphate assays showed that CCK2R-Rluc displayed pharmacological features identical to those of the untagged CCK2R because the dissociation constant (K d ) of CCK binding was 0.90 Ϯ 0.21 nM, and the EC 50 value for production of inositol phosphates was 1.3 Ϯ 0.31 nM (supplemental Fig. 1, C  and D). Additionally, stimulation of CCK2R-Rluc by CCK caused recruitment of ␤-arrestin2-YFP or ␤-arrestin1-YFP to plasma membrane and underwent rapid internalization, as depicted by fluorescent tracing using Alexa F 647-CCK (data not shown). Together, these results allowed us to use CCK2R-Rluc to monitor ␤-arrestin recruitment to the CCK2R by BRET. The BRET titration curve between CCK2R-Rluc and ␤-ar-restin2-YFP or ␤-arrestin1-YFP (Fig. 6A) indicated that in both unstimulated and CCK-stimulated HEK 293 cells, the BRET donor (CCK2R-Rluc) was progressively associated and finally saturated with BRET acceptor (␤-arrestin2-YFP or ␤-arrestin1-YFP) in the presence of increasing quantities of the latter. In the presence of 0.1 M CCK, the BRET signal reached 8 -10-fold that measured in the absence of CCK. The kinetics of BRET showed that half of maximal recruitment of ␤-arrestins to the CCK2R occurred 30 s after the addition of CCK or gastrin, and this recruitment remained relatively stable for at least 20 min, which is in line with confocal microscopy observations and quantifications. In contrast, stimulation by PD135,158, JB93,242, or JB93,182 did not cause any significant change in BRET signal (Fig. 6B). Recruitment of ␤-arrestins to the CCK2R induced by CCK or gastrin was dose-dependent, with EC 50 of 26.1 Ϯ 3.1 and 10.5 Ϯ 3.3 nM, respectively (Fig. 6C). These values were slightly higher than the values of the dissociation constants of the peptides determined in binding experiments (4.8 and 13.2 nM). Finally, ligands that did not stimulate ␤-arrestin recruitment to the CCK2R dose-dependently inhibited CCK-induced BRET signal (for PD135,158, IC 50 ϭ 95.9 Ϯ 8.0 nM) (Fig. 6D) (data not shown).
The C-terminal Region of CCK2R Is Critical for ␤-Arrestin Recruitment but Not for CCK2R Internalization-In many GPCRs, key residues involved in the recruitment of ␤-arrestins and receptor internalization are phophorylatable amino acids located in the C-terminal region of the receptor (30). We therefore constructed a large series of CCK2Rs in order to identify those residues. Control experiments were conducted to verify that the potency and efficacy of truncated and mutated CCK2R on the G-protein dependent signaling pathway (i.e. inositol phosphate production) were retained. Results depicted in supplemental Table 1 indicate that all of the mutants responded to CCK with potencies similar to that of the wild-type CCK2R, although several truncated and mutated receptors exhibited enhanced maximal responses relative to the wild-type receptor.
Because the above results indicate that a region downstream of residue 409 in the CCK2R is essential for ␤-arrestin recruitment, we aimed at identifying more precisely the residues or motifs involved. Results from BRET experiments with CCK2R-Rluc mutated on serine and/or threonine in the Cterminal tail indicated that double mutations of Thr 427 /Ser 429 or Ser 432 /Ser 434 and single mutation of residues Thr 440 or Thr 443 did not significantly affect recruitment of ␤-arrestin2-YFP to the receptor (Fig. 7E). In contrast, single mutation of Ser 437 , Thr 439 , or Ser 442 as well as double mutation of Ser 442 / Thr 443 significantly decreased BRET between CCK2R-Rluc and ␤-arrestin2-YFP. Simultaneous mutation of Ser 437 -Thr 339 -Thr 440 or Ser 437 -Thr 339 -Thr 440 -Ser 442 -Thr 443 caused equal maximal decrease in BRET signal (Fig. 7, C and D). This finding highlights the crucial importance of the sequence Ser 437 -Xaa 438 -Thr 439 -Thr 440 -Xaa 441 -Ser 442 -Thr 443 for recruitment of ␤-arrestin2 to the CCK2R.
All together, these data indicate that 1) amino acids crucial for ␤-arrestin recruitment to the CCK2R and which participate in arrestin-dependent internalization correspond to Ser 437 -Xaa 438 -Thr 439 -Thr 440 -Xaa 441 -Ser 442 -Thr 443 located in the C-terminal region of the CCK2R, and 2) truncated and poly(A)-mutated CCK2R and the wild-type CCK2R in cells lacking ␤-arrestins can efficiently use an alternative mechanism of ␤-arrestin recruitment for their internalization process; this alternative mechanism most likely involves clathrincoated pits and amino acids outside of the C-terminal region of the CCK2R.

DISCUSSION
In our effort to gain insight into the functioning and regulation of the CCK2R, which is a GPCR of crucial importance in the central nervous system and in peripheral organs (10,12), we undertook the current study. We demonstrated that the CCK2R internalizes rapidly and abundantly in response to stimulation by the two equipotent natural agonists of the receptor, CCK and gastrin. The absence of rapid recycling and colocalization of internalized receptors with fluorescent Rab7 and Rab11 indicates targeting of internalized CCK2R to late endosomes or lysosomes for slow recycling or degradation. Although fluorescent CCK or gastrin remained trapped in endocytosic vesicles over times as long as 6 h, we do not know if the peptidic ligands were still intact and bound to internalized CCK2R at these times or if they were degraded by proteases. Other studies have shown that substance P, calcitonin gene-related peptide, and somatostatin are degraded in acidified early endosomes following internalization with their cognate receptors (32). Interestingly, it was suggested that retention of ligand integrity or ligand degradation may impact signaling triggered by internalized GPCR (33).
Internalization of GPCRs can occur through two main membrane structures, caveolae and clathrin-coated pits. Here, we show that chlorpromazine, a chemical inhibitor of clathrin-coated pit formation, strongly affected receptor internalization, supporting recruitment of membrane CCK2R to clathrin-coated pits for their subsequent endocytosis. Furthermore, CCK2R internalization was blocked by dynasore, an inhibitor of the mechanochemical enzyme dynamin, which thus appears to play a key role in the fission of CCK2R-containing endosomes (28). Several lines of evidence show the involvement of ␤-arrestin1 and -2 in the internalization of the CCK2R. GFP-tagged arrestins were translocated from cytosol to the plasma membrane after CCK or gastrin stimulation; GFP-tagged arrestins co-localized with the CCK2R during endocytosis; overexpression of both ␤-arrestin1 and -2 dominant negative efficiently abolished CCK2R internalization; and BRET experiments confirmed agonist-stimulated recruitment of ␤-arrestin1 and -2 to the CCK2R.
Previous study of agonist-mediated translocation of ␤-arrestins to GPCRs identified two classes of GPCRs (34). Members of class A, which includes ␤ 2 -adrenergic receptor, -opioid receptor, and endothelin type 1 receptor, bind ␤-arrestin2 with a higher affinity than ␤-arrestin1, whereas members of class B, comprising angiotensin II type 1A receptor, vasopres- FIGURE 7. Evidence that the C-terminal region of the CCK2R is critical for ␤-arrestin recruitment to CCK2R and that this interaction is dispensable for CCK2R internalization. A, sequence of the C-terminal region of the CCK2R and positions where CCK2R was truncated (⌬409-CCK2R, ⌬429-CCK2R, ⌬439-CCK2R) or mutated (all Ser and Thr residues were substituted by Ala (poly(A)-CCK2R). B, Flp-In TM ␤-arrestin2-GFP-293 cells transfected with CCK2R constructs or Flp-In TM 293 cells co-transfected with ␤-arrestin1-GFP were stimulated with 0.1 M Alexa 647 F-CCK. Images were captured at different times (shown for 30 min), showing that although ␤-arrestin recruitment was almost abolished by C-terminal truncation or mutation of the CCK2R, endocytosis remained abundant and dependent on clathrin-coated pits because it was blocked by chlorpromazine. C, wild-type MEF cells and ␤-arrestin1/2 Ϫ/Ϫ MEF cells expressing the CCK2R were incubated with Alexa F 647-CCK for the times indicated. Images show intensive internalization of the labeled CCK2R in the two cell lines. Immunoblots show the absence of ␤-arrestin1 or -2 in ␤-arrestin1/2 Ϫ/Ϫ MEF cells. D and E, effects of threonine/serine single or multiple mutations on CCK-stimulated ␤-arrestin recruitment to the CCK2R (D, dose/response curves; E, BRET recruitment induced by 0.1 M CCK). BRET experiments were performed at least in triplicate according to the protocol described in the legend to Fig. 6. Images are representative of at least three separate experiments. *, p Ͻ 0.05; **, p Ͻ 0.01 compared with the wild-type CCK2R. Error bars, S.E. sin 2 receptor, and substance P receptor, bind both ␤-arres-tin1 and -2 with similar high affinities (34). Subsequently, the affinity with which GPCRs bind arrestins was related to the degree of receptor phosphorylation by GPCR kinases, the intracellular trafficking, and cellular signal triggered by internalized GPCRs (35). In the current study, confocal microscopy analysis of translocation of fluorescent ␤-arrestin1 or -2 to the plasma membrane indicated that recruitment of both ␤-arres-tin1 and -2 to the CCK2R occurs with rigorously identical kinetics. Moreover, both ␤-arrestin1 and -2 remained stably associated with endocytosic vesicles throughout the time. This finding was confirmed, at the molecular level, by BRET measurements that account for proximity (Յ10 nm) or direct interaction of two partner proteins. In these BRET experiments, binding affinities of the two ␤-arrestins for the CCK2R as judged by BRET 50 were similar, and BRET signal remained almost stable for times longer than 20 min, confirming that the presence of ␤-arrestins at the periphery of endocytosic vesicles was the result of a stable and direct interaction with the CCK2R. These converging results lead us to classify the CCK2R as belonging to class B of the GPCRs.
The importance of the C-terminal region of the CCK2R (and, more precisely, of phophorylatable Ser/Thr within this region) for ␤-arrestin recruitment and receptor internalization was further evaluated by analyzing the behavior of Cterminally truncated and mutated CCK2R. We found that determinants for ␤-arrestin binding to the CCK2R reside in motif Ser 437 -Tyr 438 -Thr 439 -Thr 440 -Ile 441 -Ser 442 -Thr 443 , which is frequently found in GPCRs (30,35). It is not known if three or more residues of this motif are phosphorylated upon CCK stimulation because BRET results agree with the view that three phosphates is the threshold number for ␤-arrestin binding to GPCRs (35).
Strikingly, we observed that in cells co-expressing poly(A)-CCK2R and YFP-or GFP-tagged ␤-arrestins, a weak translocation of ␤-arrestins from the cytosol to the plasma membrane was still observed under CCK stimulation, and 35% of maximal signal BRET could be measured between ␤-arrestin2 and the CCK2R mutant having five Ser/Thr residues substituted by alanine. Internalization of the Alexa Fluor 647-labeled CCK2R was not abolished by C-terminal tail deletion or by exchanging all serine and threonine residues for an alanine, even in conditions of normal endogenous expression of ␤-arrestins in HEK cells. However, with these CCK2R variants, no colocalization of fluorescent ␤-arrestins with internalized receptors could be observed. To explain this result, one may suggest that modified CCK2Rs within the C-terminal region form relatively unstable complexes with arrestins that would dissociate at or near the plasma membrane, as observed for GPCRs of group A (34). Alternatively, internalization of the CCK2R could switch from a ␤-arrestin-dependent to a ␤-arrestin-independent mechanism when crucial determinants for binding of this adaptor (␤-arrestin) are lacking on the receptor. The observation of abundant internalization of the CCK2R in MEF-␤-arrestin1/2 Ϫ/Ϫ and its inhibition by chlorpromazine strongly supports this second switch hypothesis.
Precedent exists showing that deletion of the distal carboxyl-terminal region of the A2b adenosine receptor switches internalization from an arrestin-and clathrin-dependent to an arrestin-and clathrin-independent pathway (36). Furthermore, the requirement of ␤-arrestin for receptor internalization is very dependent on the type of GPCR. For instance, whereas the ␤2-adrergic receptor does not internalize in MEF-␤-arrestin1/2 Ϫ/Ϫ cells, the protease-activated receptor-1, the formyl peptide receptor, and the urotensin-II receptor internalize normally in these cells, and nearly 20% of angiotensin AT1A receptor still internalizes (37)(38)(39)(40). Basal constitutive internalization of the ghrelin receptor is arrestinindependent, whereas agonist-stimulated internalization is arrestin-dependent (41).
GPCRs that internalize through clathrin-coated pits independently of ␤-arrestins can use the heterotetrameric complex AP-2 that connects cargo molecules directly to clathrin. AP-2 is able to recruit molecules through 2 or ␣/2 subunit interactions with tyrosine-based (YXX⌽) or dileucine ((D/ E)XXXL(L/I)) motifs, which are most often located in the Cterminal region of GPCRs. Such mechanisms were reported, for instance, for the P2X4 purinergic receptor, PAR1 protease-activated receptor, and CXCR2 chemokine receptor (42)(43)(44). In the case of CCK2R, an analysis of amino acid sequence highlights the presence of two tyrosine-based sorting signals that might be responsible for binding of the CCK2R with the 2-subunit of the AP-2 complex. These are located in the third intracellular loop at positions 246 -249 (YLGL) and 294 -297 (YVQL). Future studies will examine if these motifs or other not yet identified mechanisms are responsible for ␤-arrestin-independent internalization of the CCK2R and if the ␤-arrestin-independent pathway of internalization is employed in parallel with the ␤-arrestin-dependent one in the physiological context.
Pioneering work quantified internalization of the rat CCK2R in NIH 3T3 cells and showed the importance of Cterminal tail and serine/threonine residues (18). Although this study did not investigate the role played by arrestins, the results also suggested a contribution to internalization of regions of the CCK2R outside of the C-terminal tail. This view agrees with the general molecular basis whereby ␤-arrestins bind to agonist-activated receptors through multiple sites of interaction on both the receptor and ␤-arrestins, which were termed "activation recognition" and "phosphorylation recognition" sites (35,45,46).
Results from the current study reinforce the importance of the C-terminal tail of the CCK2R for its downstream signaling and regulation. We previously documented interaction of the C-terminal region of the CCK2R with RGS-2, SHP-2 phosphatase, or phospholipase-C ␥1 (17,47,48). The motif involved in functional binding of SHP2 phosphatase and phospholipase-C ␥1 to the CCK2R is an ITIM motif containing Tyr 438 , which, once phosphorylated, recruits signaling proteins containing an Src homology 2 domain(s). We suggested that the binding of CCK2R to the Src homology 2 domain of phospholipase-C ␥1 in the pancreas of mice overexpressing the CCK2R may participate in the development of preneoplastic lesions and pancreatic cancer (49). Concerning interaction of RGS-2, we demonstrated that it involves Ser 434 and Thr 439 , requires phosphorylation of these residues, and serves for RGS2 recognition by the CCK2R (17).
The current study also provides new important pharmacological data on the CCK2R because it allowed identification of biased ligands of this receptor. Indeed, we show that, unlike CCK and gastrin, which are full agonists with respect to inositol phosphate production, partial agonist PD135,158 or JB93,242 on the G-protein-dependent pathway as well as the antagonist JB93,182 did not promote receptor internalization and were unable to stimulate recruitment of ␤-arrestins. However, these three nonpeptide ligands acted as antagonists on CCK-induced recruitment of ␤-arrestins to the CCK2R and its internalization. All three ligands, PD135,158, JB93,242, and, more surprisingly, JB93,182, stimulated ERK1/2 phosphorylation. Delineation of the precise mechanism linking binding of these ligands to the CCK2R with ERK activation will require further investigations. However, based on BRET and confocal microscopy data showing an inability of these ligands to stimulate ␤-arrestin recruitment and CCK2R internalization, it is unlikely that these two events are involved. Most likely, ERK activation by PD135,158 and JB93,242 would be a consequence of phospholipase-C activation (10). The mechanism leading to ERK1/2 activation by JB93,182 remains unexplained because this compound did not stimulate phospholipase-C or ␤-arrestin recruitment. Hence, PD135,158 and JB93,242 meet the criteria of biased ligands of the CCK2R because they are capable of stimulating a typical G-dependent signaling pathway but are unable to stimulate ␤-arrestin recruitment and receptor internalization. JB93182 is another biased ligand capable of stimulating ERK1/2 phosphorylation but unable to stimulate phospholipase C activation, ␤-arrestin recruitment, and receptor internalization. These findings may have a significant impact on the pharmacology of the CCK2R and cognate strategies to target it. In the recent period, biased ligands for GPCRs, such as, for example, adrenergic, dopamine, and angiotensin II receptors, have been identified and shown to present pharmacological and therapeutic interest (25,31,50,51). CCK2R joins this group of GPCRs. To explain why PD135,158 and JB93,242 stimulate production of inositol phosphates but could not induce ␤-arrestin recruitment to the CCK2R, we suggest that conformations of the CCK2R stabilized by CCK or by the two partial nonpeptide agonists are distinct and promote phospholipase-C with different efficacies, and only the state stabilized by the full agonist CCK or gastrin is able to promote ␤-arrestin recruitment. This view agrees with the general concept, now well supported by experimental data, that GPCRs exist in multiple distinct conformational states that can be stabilized by selective ligands and present distinct signaling profiles (8).
Our work represents a solid basis to evaluate the contribution of internalized CCK2R to CCK2R-induced intracellular signaling. The finding that CCK2R expressed in HEK cells remains bound to ␤-arrestins during intracellular trafficking suggests that part of the CCK2R-induced signaling and effects may be dependent on ␤-arrestins. Biased ligands identified in the current study should be able to block ␤-arrestin-dependent signaling and effects downstream of the CCK2R. This is of high importance because the CCK2R is expressed in a large variety of cancers and, according to a previous study (19), undergoes internalization in gastric, pancreatic, and colonic cell lines that endogenously express this receptor.