Agonist-promoted internalization of a ternary complex between calcitonin receptor-like receptor, receptor activity-modifying protein 1 (RAMP1), and beta-arrestin.

The calcitonin receptor-like receptor (CRLR) is a seven-transmembrane domain (7TM) protein that requires the receptor activity-modifying protein 1 (RAMP1) to be expressed at the cell surface as a functional calcitonin gene-related peptide (CGRP) receptor. Although dimerization between the two molecules is well established, very little is known concerning the trafficking of this heterodimer upon receptor activation. Also, the subcellular localization and biochemical state of this ubiquitously expressed protein, in the absence of CRLR, remains poorly characterized. Here we report that when expressed alone RAMP1 is retained inside the cells where it is found in the endoplasmic reticulum and the Golgi predominantly as a disulfide-linked homodimer. In contrast, when expressed with CRLR, it is targeted to the cell surface as a 1:1 heterodimer with the 7TM protein. Although heterodimer formation does not involve intermolecular disulfide bonds, RAMP-CRLR association promotes the formation of intramolecular disulfide bonds within RAMP1. CGRP binding and receptor activation lead to the phosphorylation of CRLR and the internalization of the receptor as a stable complex. The internalization was found to be both dynamin- and beta-arrestin-dependent, indicating that the formation of a ternary complex between CRLR, RAMP1, and beta-arrestin leads to clathrin-coated pit-mediated endocytosis. These results therefore indicate that although atypical by its heterodimeric composition and its targeting to the plasma membrane, the CGRP receptor shares endocytotic mechanisms that are common to most classical 7TM receptors.

the calcitonin family of regulatory peptides (1)(2)(3). Although a cDNA encoding the calcitonin receptor was cloned over 10 years ago (4), it was only recently that McLatchie et al. (5) established that co-expression of an orphan calcitonin-receptor like receptor (CRLR) with a newly discovered protein named receptor activity-modifying protein 1 (RAMP1) created a CGRP receptor. CRLR shares 55% amino acid sequence identity with the calcitonin receptor. Both belong to the family B of seventransmembrane domain (7TM) receptor. The family also includes the receptors for parathyroid hormone, parathyroid hormone-related protein, secretin, glucagon, and vasoactive intestinal polypeptide/pituitary adenylate cyclase activating polypeptide (6,7). RAMP1, for its part, is a member of a family that comprises 3 members designed RAMP1, RAMP2, and RAMP3. These small intrinsic membrane proteins share less than 30% of sequence identity and possess a large extracellular N terminus (ϳ100 amino acids), a single transmembrane domain, and a very short intracellular domain (10 amino acids) (5). One of the roles proposed for these proteins is that of chaperone/escort promoting the cell surface targeting of CRLR (5,8). Interestingly, like CRLR, RAMPs are also retained intracellularly when expressed alone suggesting that an interaction between the two proteins is essential to their common cell surface expression (5, 9 -11). However, the retention site and the fate of each protein, when expressed alone, has never been directly investigated. Given that at least one of the RAMPs is expressed in all tissues tested to date and that they are expressed independently of CRLR (1-3, 5, 12), a better characterization of the subcellular localization of RAMP could shed some light on additional roles that could be played by these proteins.
When associated with CRLR, RAMPs has been shown to determine the pharmacological property of the receptor formed. Co-expression of CRLR with RAMP1 leads to a CGRP receptor whereas RAMP 2 or 3 promote the formation of an adrenomedullin receptor (5,9). Although the contribution of RAMPs to the formation and trafficking of functional receptors has now been well established (13)(14)(15), the fate of the complex following stimulation by agonists has not been systematically investigated. For instance, many G protein-coupled receptors (GPCRs) have been shown to undergo agonist-promoted internalization through classical dynamin-dependent endocytosis. In these cases, agonist-promoted phosphorylation of the receptor by a G protein-coupled receptor kinase has been shown to promote association with ␤-arrestin leading to endocytosis through its interaction with clathrin and the adaptor protein AP-2 (16 -19). Some receptors, such as the V2-vasopressin receptor, internalize as a stable complex with ␤-arrestin while others, like the ␤ 2 -adrenergic receptor, rapidly dissociate from the receptor upon internalization (20). An alternative, and less well characterized, endocytotic pathway involving caveolae has also been proposed for some receptors (18,(21)(22)(23)(24). Although, agonist-promoted internalization of CRLR has been documented (10) the mechanisms involved in the endocytosis of this atypical receptor have not been investigated. Also very little is known concerning the fate of RAMPs following agonist stimulation. RAMP 2 and 3 were found in the lysosome following a treatment of 30 min with adrenomedullin (10), but whether the CRLR⅐RAMP complexes are endocytosed as a stable complex remains to be determined.
In this study, a systematic analysis of RAMP1 and CRLR subcellular distribution under various conditions allowed us to determine that: 1) unlike CRLR, which is retained in the endoplasmic reticulum (ER) when expressed alone, RAMP1 is found both in the ER and the Golgi even in the absence of CRLR indicating a possible additional role for RAMP1 that could involve cycling between the two compartments, 2) following agonist stimulation of cells co-expressing CRLR and RAMP1, the complex undergoes rapid phosphorylation of the CRLR but not the RAMP moiety that is followed by dynaminand ␤-arrestin-dependent internalization.
3) The complex is endocytosed as a stable ternary complex including ␤-arrestin that remains associated with CRLR-RAMP1 once it is internalized.

Materials
Human ␣CGRP was from Bachem (Bubendorf, Switzerland). Isoproterenol and arginine vasopressin were purchased from Sigma (Saint Louis, MO). Bis(sulfosuccinimidyl)suberate (BS 3 ) and disuccinimidyl glutarate (DSG) were from Pierce (Rockford, IL). Rabbit anti-myc and anti-HA antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). Rat anti-HA antibodies (3F10 clone) were from Roche Molecular Biochemicals (Mannheim, Germany). Mouse anti-p58K antibodies were from Sigma and rabbit anti-calnexin antibodies were from Stressgen Biotechnologies Corp. (Victoria, BC, Canada). Mouse anti-myc antibodies (9E10 clone) and mouse anti-HA antibodies (12CA5 clone) were produced by our core facility as ascites fluids. Horseradish peroxidaseconjugated secondary antibodies were from Amersham Pharmacia Biotech (Little Chalfont, United Kingdom). Oregon Green-conjugated and Texas Red-conjugated secondary antibodies were purchased from Molecular Probes Inc. (Eugene, OR). Phycoerythrin-conjugated secondary antibody was from Immunotech (Westbrook, MA). Protein G-Sepharose was from Amersham Pharmacia Biotech. The renaissance chemiluminescence kit was from PerkinElmer Life Sciences. Pefabloc was from Roche Molecular Biochemicals. Airvol mounting media was purchased from Air Products and Chemicals Inc. All other reagents were analytical grade and obtained from various commercial suppliers.
Transient transfections were performed using calcium phosphate co-precipitation (26). Cells in 6-well plates or in 100-mm Petri dishes were grown to 70% confluency and transfected with the indicated plasmids. Experiments were carried out 48 h after transfection.

Immunofluorescence Microscopy and Flow Cytometry
Labeling in Permeabilized Cells-HEK293T cells stably expressing HA-CRLR or myc-RAMP1 were fixed with 3% paraformaldehyde in phosphate-buffered saline (PBS) for 15 min and permeabilized for 10 min with 0.25% Triton X-100 in blocking buffer (PBS containing 0.2% BSA). The cells were then incubated with primary antibodies in blocking buffer for 1 h. The following antibodies were used at 1:100: rabbit or mouse anti-myc, mouse anti-p58K, rat anti-HA, and rabbit anti-calnexin. The coverslips were then washed in blocking buffer and incubated in the same buffer with the appropriate secondary antibodies (Texas Red-conjugated secondary antibodies for visualization of myc-RAMP1 or HA-CRLR and Oregon Green-conjugated secondary antibodies for visualization of p58K or calnexin) at room temperature for 30 min in the dark. After extensive washing, the coverslips were mounted onto slides using Airvol mounting media. Double labeling was viewed on a Bio-Rad MRC 600 laser scanning confocal microscope using a Zeiss Plan-Apo 63 ϫ 1.40 Na oil immersion lens.
Cell-surface Labeling-HA-CRLR or empty vector were transiently transfected into HEK293T cells stably expressing myc-RAMP1. DMEM/Hepes blocking buffer. Cells were then washed in ice-cold PBS and fixed with 3% paraformaldehyde in PBS for 15 min. Cells were then prepared as described in the previous section using Texas Red-conjugated goat anti-rabbit antibodies for visualization of myc-RAMP1 or Oregon Green-conjugated anti-rat antibodies for the detection of HA-CRLR.
Internalization-HEK293T cells stably expressing HA-CRLR or myc-RAMP1 were transfected with RAMP1 and ␤-arrestin2-YFP or CRLR and ␤-arrestin2-YFP, respectively, whereas ␤-arrestin2-YFP alone was transiently transfected in cells stably expressing HA-␤ 2 -adrenergic or myc-V2 vasopressin receptors. After washes with PBS, cells were incubated in DMEM/Hepes blocking buffer for 30 min before adding rabbit anti-HA or mouse anti-myc antibodies in the same buffer for 1 h. The cells were then washed and treated with the appropriate agonist (100 nM CGRP, 100 nM arginine vasopressin, or 1 M isoproterenol) for different times at 37°C. The internalization process was stopped on ice and the cells were fixed in 3% paraformaldehyde in PBS for 15 min. The cells were then prepared as before using the appropriate Texas Redconjugated secondary antibodies for visualization of the receptors and RAMP1. The fluorescent antibodies and the ␤-arrestin2-YFP were visualized by confocal microscopy as above.
Flow Cytometry-Following anti-myc antibody incubation as described above, HEK293T cells transfected with CRLR and myc-RAMP1 or myc-CRLR and RAMP1 were washed with PBS containing 0.2% BSA and incubated with phycoerythrin-conjugated secondary antibody in PBS/BSA for 1 h at room temperature. The cells were washed again in PBS/BSA, detached using PBS containing 2 mM EDTA, and fixed in 3% paraformaldehyde. Fluorescence-activated cell sorting (FACS) analysis was performed on a FACS TM calibur Becton-Dickson flow cytometer. Cell viability was assessed by Trypan blue exclusion and 10,000 cells were sorted in each experiment.

Immunoprecipitation
Total Extracts-Cells grown in 100-mm Petri dishes were washed twice with PBS and lysed for 20 min at 4°C in a buffer containing 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 50 mM iodoacetamide, 1 mM Pefabloc, 0.1 mg/ml bacitracin, 0.1 mg/ml benzamidine, 2.5 g/ml leupeptin (lysis buffer) and centrifuged at 12,000 ϫ g for 15 min at 4°C. Immunoprecipitation was then initiated by adding mouse monoclonal anti-myc antibody (clone 9E10) or rat monoclonal anti-HA antibody (clone 3F10) to the supernatant. Following an overnight incubation at 4°C, protein G-Sepharose was added for an additional 3 h. Protein G-Sepharose⅐antibody⅐antigen complexes were collected by centrifugation and washed four times with cold lysis buffer containing 150, 250, 350 and finally 150 mM NaCl. The final pellet was resuspended in sample buffer containing 60 mM Tris-HCl, pH 6.8, 2% SDS, 4.5 M urea with or without 100 mM dithiothreitol (DTT) and prepared for Western blot analyses.
Cell Surface-Cells grown in 100-mm Petri dishes were washed three times in PBS and incubated with blocking buffer (PBS containing 0.2% BSA) for 1 h on ice. Whole cells were then incubated with anti-myc (9E10) or anti-HA (12CA5) antibodies (1:150 dilution) in blocking buffer on ice for 1 h, washed twice in blocking buffer and twice in PBS. After removing the free antibodies, the cells were lysed and prepared for immunoprecipitation as described above.

Western Blot Analysis
Protein samples were resolved by 10 or 14% SDS-polyacrylamide gel electrophoresis. They were transferred to nitrocellulose and subjected to immunoblotting using rabbit polyclonal anti-myc antibodies (1:8000 dilution), or rat monoclonal anti-HA antibodies (1:3000 dilution). Renaissance chemiluminescence kit was used for Western blot development.

Covalent Cross-linking Analysis
Cells expressing myc-CRLR, myc-RAMP1, or both were grown in 6-well plates, washed three times in cold PBS and incubated on ice for 1 h in PBS containing 1 mM BS 3 . The cross-linking reaction was stopped by incubation in PBS containing 25 mM glycine. Cells were then lysed and prepared for immunoprecipitation as described above using mouse monoclonal anti-myc antibody (9E10).

Whole Cell Phosphorylation
Following an incubation of 1 h in phosphate-free medium containing 1% FCS, cells expressing myc-RAMP1 or myc-RAMP1 and myc-CRLR were incubated for 2 h in phosphate-free medium containing [ 32 P]P i (1.7 mCi/10 6 cells) and treated with 100 nM CGRP for the indicated times. They were then solubilized in lysis buffer (see above) containing 0.2 mM sodium orthovanadate, 10 mM sodium fluoride, 10 mM sodium phosphate and centrifuged at 12,000 ϫ g for 15 min at 4°C. Immunoprecipitation was then performed as described above using monoclonal anti-myc antibody (9E10). Protein samples were resolved by 14% SDSpolyacrylamide gel electrophoresis. They were transferred to nitrocellulose and [ 32 P]P i was detected with BioMax MR Kodak films (Amersham Pharmacia Biotech).

CRLR-RAMP1 Complex Internalization
Cells stably expressing myc-RAMP1 were transfected with myc-CRLR in combination with ␤-arrestin2, ␤-arrestin2(319 -418), K44A dynamin I, or an empty vector. Following a 1 h incubation in serum-free media, cells were pretreated (or not) for 30 min with internalization inhibitor (0.5 M sucrose) and exposed to 100 nM CGRP for the indicated periods at 37°C. The cells were then quickly chilled, washed twice with ice-cold PBS, and incubated in PBS containing 1 mM BS 3 for 1 h. For other experiments, crude membrane preparations were prepared as previously described (5)  were lysed for immunoprecipitation and prepared for Western blot analyses as described above. Immunoreactivity was quantified by densitometry using chemiluminescence radiography and the results analyzed using the NIH Image 1.62 program.

Intracellular Retention of CRLR and RAMP1-Previous
studies have demonstrated that both CRLR and RAMP1 are retained inside the cell when expressed individually (5, 9 -11). However, the precise site of retention for the two proteins has not been identified. Thus, the subcellular distribution of HAtagged CRLR (HA-CRLR) and myc-tagged RAMP1 (myc-RAMP1) was assessed by immunofluorescence confocal microscopy. As shown in Fig. 1, neither myc-RAMP1 nor HA-CRLR were detected at the surface of HEK293 cells stably expressing myc-RAMP1 (A) or HA-CRLR (B) independently. This contrasts with the cell surface localization of the two proteins when the myc-RAMP1 expressing cells were transiently transfected with HA-CRLR (C and E) and when myc-RAMP1 was transfected in the cells stably expressing HA-CRLR (D and F). Panels G and H further show the co-localization of the two proteins at the plasma membrane upon co-expression. To determine the intracellular site of retention of each protein ex-pressed individually, co-localization with the endoplasmic reticulum (ER) and Golgi markers, calnexin (27) and p58K (28), was assessed. HA-CRLR was found to co-localize with calnexin but not p58K ( Fig. 2A) indicating that it is retained in the ER in the absence of co-expressed RAMP. In contrast, myc-RAMP1 was found to co-localize with both calnexin and p58K suggesting that RAMP1 can traffic from the ER to the Golgi in the absence of CRLR (Fig. 2B). Given that no myc-RAMP1 could be detected at the cell surface in the absence of co-transfected CRLR, these data show that RAMP1 alone can be transported from the ER to the Golgi. Active trafficking between the ER and the Golgi has previously been described for molecular chaperones such as ERGIC-53 and vesicular integral membrane protein 36, that are known to operate as cargo receptors for the transport of glycoproteins from the ER to the ERGIC or Golgi (29 -31), suggesting that RAMPs may have additional roles that are independent of their association with CRLR. The Golgi expression of RAMP is probably not a mislocalization resulting from overexpression because it was observed in stable clones with a wide range of expression (data not shown). Similar expression levels of a myc-tagged CRLR never led to Golgi expression (data not shown).
Taken together, these results suggest that when expressed individually the CRLR is retained in the ER while RAMP1 can be transported from the ER to the Golgi. Upon co-expression, the two proteins would thus have to interact in the ER, their association would then allow the complex to transit through the Golgi "en route" to the plasma membrane.
CRLR-RAMP1 Complex Displays a 1:1 Stoichiometry at the Cell Surface-To biochemically confirm the influence of CRLR on the cell surface expression of RAMP1 and to verify that once expressed at the cell surface, CRLR and RAMP1 remain as a stable complex, total extract and cell surface immunoprecipitation of myc-RAMP1 and HA-CRLR were carried out. In the absence of co-transfected HA-CRLR (Fig. 3A, lanes 3 and 4), myc-RAMP1 was detected in the total extract but no Myc immunoreactivity was observed at the cell surface consistent with the immunofluorescence data presented in Fig. 1. However, when HA-CRLR and myc-RAMP1 were co-expressed A, total and cell surface myc-RAMP1 immunoreactivity were assessed in cells expressing myc-RAMP1 alone or in combination with HA-CRLR. Cells were transfected with pcDNA3 alone (lanes 1 and 2), myc-RAMP1 (lanes 3 and 4), or myc-RAMP1 and HA-CRLR (lanes 5-8) and the total population (T) or cell surface (S) myc-RAMP1 were immunoprecipitated (IP) using mouse anti-myc antibodies (lanes 1-6) or mouse anti-HA antibodies (lanes 7 and 8). Immunocomplexes were then analyzed by SDSpolyacrylamide gel electrophoresis (14%) and immunoblots (IB) were probed using rabbit anti-Myc antibodies. The Western blot shown is representative of three independent experiments. B, direct interaction between RAMP1 and CRLR at the cell surface was investigated using the cell impermeable cross-linker BS 3 . Cells transfected with pcDNA3 alone (lanes 1 and 2), myc-RAMP1 (lanes 3 and 4), myc-CRLR (lanes 5 and 6), or myc-RAMP1 and myc-CRLR (lanes 7 and 8) were incubated in the presence (ϩ) or absence (Ϫ) of 1 mM BS 3 . Immunoprecipitations (IP) were performed on total cell extracts using mouse anti-Myc antibodies and the immunocomplexes were analyzed by SDS-polyacrylamide gel electrophoresis, 10% (top panel) or 14% (bottom panel). Immunoblots (IB) were probed with rabbit anti-myc antibodies. Arrow indicates the myc-CRLR⅐myc-RAMP1 complex. The Western blot shown is representative of four independent experiments. (lanes 5 and 6), RAMP1 was both in the total extract and at the cell surface. Moreover, RAMP1 was found to be tightly associated with CRLR because myc-RAMP1 could be co-immunoprecipitated with HA-CRLR both in total extract (lane 7) and upon cell surface immunoprecipitation (lane 8), indicating that the complex remained stable once the two proteins reach the plasma membrane. Immunoreactive HA-CRLR was also detected upon total extract and cell surface immunoprecipitation of myc-RAMP1 (data not shown).
To determine whether the interaction between CRLR and RAMP1 is direct or involves an additional intermediary(ies), covalent cross-linking studies using the membrane impermeable homobifunctional cross-linker BS 3 were performed. As shown in Fig. 3B, BS 3 had no effect on the electrophoretic mobility of either myc-RAMP1 (M r 17,000) or immature myc-CRLR (M r 58,000) when expressed alone. However, when the two proteins were co-expressed (lanes 7 and 8), a component of M r ϳ85,000 was detected in addition to the fully processed form of myc-CRLR (M r 66,000) and myc-RAMP1 (M r 17,000). The change in electrophoretic mobility observed for myc-CRLR upon myc-RAMP1 co-expression, even in the absence of BS 3 , is consistent with the previously reported (5,9) Golgi maturation that results from the interaction between the two proteins. Decreased amount of free mature myc-CRLR (M r 66,000) was observed in cross-linked samples (compare lanes 8 and 7), consistent with a shift of a part of the receptor into the larger molecular weight complex of M r ϳ85,000. The presence of the M r 66,000 species in the cross-linked samples should not be taken as evidence for the existence of free CRLR at the cell surface but most likely reflects the low cross-linking efficacy of BS 3 . The size of the cross-linked complex (M r 85,000) is consistent with a direct interaction between one molecule of RAMP1 (M r 17,000) and one molecule of mature receptor (M r 66,000). High molecular forms of RAMP1 could be detected in the absence of CRLR, even in the absence of any cross-linker (lanes 3 and 4) suggesting the possibility of oligomeric structures of RAMP1 that are at least partly resistant to SDS denaturation. Such species were previously described for RAMP1 and RAMP3 (5, 12), however, their relation with the formation of CRLR⅐RAMP complexes has not been investigated.
Distinct Subcellular Localization of RAMP1 Homodimers and CRLR⅐RAMP1 Complex- Fig. 4 shows that in the absence, of CRLR, myc-RAMP1, expressed as M r 17,000 and M r 32,000 species corresponding to monomer and dimer, is retained intracellularly. In nonreducing conditions, the dimer was the major form detected whereas the monomer was predominant in the presence of DTT, indicative of disulfide bond contribution in dimer formation. Whether these lead to a covalently bound dimer or participate to intramolecular conformational requirement for dimerization cannot be determined at this point. However, the DTT-sensitive dimer does not result from spurious cross-linking following cell lysis since it was not inhibited by alkylating agents such as iodoacetamide. Co-expression of CRLR significantly increased the expression of RAMP1 and dramatically changed the dimer/monomer ratio such that the monomer became the major species. This indicates that the FIG. 5. Effect of agonist stimulation on CRLR⅐RAMP1 complex stability and phosphorylation. A, the stability of the CRLR⅐RAMP1 complex upon CGRP treatment was assessed by co-immunoprecipitation. Cells transfected with pcDNA3 alone (MOCK) or with myc-RAMP1 and HA-CRLR were incubated with 100 nM CGRP for the indicated times. Immunoprecipitations (IP) were performed on total cell extracts using mouse anti-HA antibodies. Immunocomplexes were analyzed by SDS-polyacrylamide electrophoresis gels (14%) and immunoblots (IB) were probed with rabbit anti-myc antibodies. The Western blot shown is representative of three independent experiments. B, the phosphorylation states of the receptor and RAMP1 was studied following metabolic labeling of the cells with [ 32 P]P i . Cells transfected with pcDNA3 alone (lane 1), myc-RAMP1 (lane 2), or myc-CRLR and myc-RAMP1 (lanes 3-6) were loaded with [ 32 P]P i for 2 h and incubated with 100 nM CGRP for the indicated times. Immunoprecipitation (IP) were performed on total cell extracts using mouse anti-myc antibodies and immunocomplexes were analyzed by SDS-polyacrylamide gel electrophoresis (14%) and autoradiography. The autoradiogram shown is representative of two independent experiments. Arrow, phosphorylated form of myc-CRLR. formation of CRLR-RAMP1 heterodimer is favored over that of RAMP1 homodimers, either as a result of a CRLR-promoted dissociation of preformed RAMP1 dimers or through the sequestration of the neosynthesized RAMP1 into a more stable CRLR⅐RAMP1 complex. Given the level of energy that would be required to break the disulfide bonds involved in the RAMP1 homodimer, the second possibility appears more likely. Interestingly, the formation of the CRLR⅐RAMP1 complex is associated with a change in the electrophoretic mobility of the RAMP1 monomer (M r 16,000 when co-expressed with CRLR versus M r 18,000 when expressed alone) that most likely reflects the formation of intramolecular disulfide bonds. Indeed, the M r 18,000 monomeric species was found to be insensitive to DTT whereas the M r 16,000 that is formed upon CRLR coexpression is sensitive to reducing agents slowing its migration and leading to a species that has the same apparent size as RAMP1 expressed alone (M r 18,000). Taken together, these results indicate that CRLR interaction promotes conformational changes that involve disulfide bond reorganization that are different from those involved in the RAMP1 homodimer. The physical interaction between CRLR and the DTT-sensitive M r 16,000 RAMP species was further confirmed by the coimmunoprecipitation of myc-RAMP1 with HA-tagged CRLR. Although leading to changes in disulfide organization, the in-teraction between RAMP1 and CRLR most likely do not involve intermolecular bonds (14).
In contrast to the RAMP1 derived from the CRLR-RAMP1 heterodimer, which is easily detectable at the cell surface, the RAMP1 homodimer could be seen only in total extracts indicating that it is retained intracellularly. Taken with the observation that when expressed alone, RAMP1 but not CRLR is located in the Golgi, these data may indicate that the RAMP1 homodimer has a different role. Although no data is available to suggest what this role might be, RAMP1, -2, and -3 share structural features with chaperone proteins such as ERGIC-53 that have been shown to form disulfide-linked homodimers and homohexamers in the ER (31). This may indicate a role for RAMP1 in the protein synthesis quality control system. CRLR⅐RAMP1 Complex Stability and Phosphorylation upon CGRP Treatment-To determine if binding of the receptor to its natural ligand, CGRP, can modulate the CRLR-RAMP1 interaction, the effect of short time CGRP exposure (100 nM for 0 -15 min) was assessed. As shown in Fig. 5A, CGRP had no effect on the amount of myc-RAMP1 that could be co-immunoprecipitated with HA-CRLR, suggesting a stable interaction between RAMP1 and CRLR even after CGRP binding. However, CGRP treatment led to the phosphorylation of the receptor (Fig. 5B). Indeed, immunoprecipitation of both myc-CRLR and myc-FIG. 7. Agonist-promoted internalization of CRLR⅐RAMP1 complex. The CGRP-promoted internalization of the CRLR⅐RAMP1 complex was investigated using differential cross-linking of the total and cell surface complexes. Cells stably expressing myc-RAMP1 were transfected with myc-CRLR and treated for the indicated times with 100 nM CGRP. myc-CRLR⅐myc-RAMP1 complexes remaining at the cell surface (following (B) or not (A) a pretreatment with 0.5 M sucrose) were stabilized using 1 mM BS 3 (A and B) while the total complexes were stabilized using 1 mM DSG (D). Immunoprecipitations were performed on total cell extracts using mouse anti-myc antibodies and immunocomplexes were analyzed by SDS-polyacrylamide gel electrophoresis (10%). Immunoblots were probed with rabbit anti-myc antibodies. The total amount of myc-CRLR remaining after exposure to CGRP for the indicated times was assessed by Western blot (E). Quantitative analyses of the immunoreactivity (C and F) were carried out by densitometric scanning and the value expressed as % of the value obtained in the absence of CGRP (time 0 ϭ 100%). These data represent the mean Ϯ S.E.M. of four independent experiments. 1, immature myc-CRLR (M r ϳ58,000). 2, mature myc-CRLR (M r ϳ66,000). 3, myc-CRLR/myc-RAMP1 complex (M r ϳ85,000). RAMP1 from cells metabolically labeled with [ 32 P]P i revealed an agonist-dependent incorporation of radioactive phosphate into CRLR but not RAMP1 (Fig. 5B). Agonist-promoted phosphorylation is a well known phenomenon among 7TM receptors and is associated with the rapid desensitization process that follows receptor activation. The fact that maximal CRLR phosphorylation was attained within 5 min following the beginning of the stimulation is consistent with such a role of phosphorylation for this receptor. Rapid desensitization of the CGRPmediated adenylyl cyclase activation has indeed been reported in SK-N-MC cells (32). Phosphorylation has also been shown to facilitate the internalization that occurs upon receptor activation (19). Given that CRLR but not RAMP1 appears to be phosphorylated, the internalization profile of both proteins was investigated.
Simultaneous Internalization of CRLR and RAMP1 within a Stable Complex following CGRP Stimulation-The internalization kinetics of CRLR and RAMP1 were assessed by FACS analysis in cells expressing myc-CRLR and RAMP1 (Fig. 6, top) or CRLR and myc-RAMP1 (Fig. 6, bottom), respectively. Agonist treatment induced a rapid and time dependent decrease in cell surface myc-CRLR that led to a loss of 60% of the signal after 60 min of treatment. Myc-RAMP1 followed an identical internalization kinetic upon CGRP exposure (Fig. 6, bottom) suggesting that both proteins were internalized together.
To directly determine if the two proteins were internalized as a complex, the fate of the CRLR-RAMP1 heterodimer was examined following covalent cross-linking. The use of the hy-drophilic cell impermeable cross-linker BS 3 allowed the exclusive stabilization of those complexes remaining at the cell surface during the course of CGRP treatment, whereas the hydrophobic cell permeable DSG has access to the total population of CRLR-RAMP1 heterodimers. To visualize the cross-linked complexes, myc-CRLR and myc-RAMP1 immunoreactivity was detected following immunoprecipitation with the anti-Myc antibody. As shown in Fig. 7A, immunoreactive species corresponding to the immature myc-CRLR (band 1: M r 58,000), the mature myc-CRLR (band 2: M r 66,000), and the myc-CRLR⅐myc-RAMP1 complex (band 3: M r 85,000) were detected. Although not shown, a band of M r 17,000 corresponding to myc-RAMP1 was also observed. Upon agonist treatment, a rapid decrease of cell surface complexes, stabilized by BS 3 (band 3), was observed such that 50% of the complex disappeared from the cell surface in less than 10 min (Fig. 7, panels A and C). In contrast, CGRP promoted very little change in the total contingent of CRLR⅐RAMP1 complexes stabilized by DSG (Fig. 7 panels D and F). This indicates that the decrease in BS 3 -stabilized CRLR-RAMP1 heterodimer results from its internalization and not its dissociation. This is consistent with the notion that the interaction between the two proteins remains stable upon agonist exposure (see Fig. 5A). This is also in agreement with the intracellular co-localization of CRLR and RAMP2 or RAMP3 observed by Kuwasako et al. (10) following adrenomedullin treatment. The agonist promoted internalization was significantly blocked by 0.5 M sucrose (Fig. 7, panels B and C), consistent with classical endocytosis through clathrin-coated vesicles and in agreement with the observation that CRLR co-localizes with transferrin upon treatment of cells with CGRP (10).
The marginal loss of total DSG-stabilized CRLR⅐RAMP complex, observed during CGRP treatment, most likely results from the degradation of the receptor after sequestration. Indeed, a similar reduction in CRLR immunoreactivity was observed in the absence of any cross-linker (compare Fig. 7, D and E). Another interesting observation that is worth mentioning is that the electrophoretic migration of the mature myc-CRLR and myc-CRLR⅐RAMP1 complexes (but not the intracellularly retained immature myc-CRLR) is retarded following agonist exposure. This band shift is most likely due to the agonistpromoted phosphorylation of the receptor that we reported in Fig. 5B.
CRLR-RAMP1 Internalization Is ␤-Arrestin and Dynamin Dependent-To determine the mechanisms involved in the internalization of the CRLR-RAMP1 heterodimer, the potential roles of ␤-arrestin and dynamin were investigated by cotransfecting their respective dominant negative mutants ␤-arrestin2(319 -418) (33) and dynamin K44A (34). The effects of expressing the carboxyl tail of ␤-arrestin2 (␤-arrestin2(319 -418)) and the GTPase-deficient dynamin K44A were assessed on the internalization of the complex using both FACS analysis and cross-linking experiments. In cells expressing both myc-CRLR and RAMP1 (Fig. 8A) or CRLR and myc-RAMP1 (Fig.  8B), FACS analyses revealed that the CGRP-promoted loss in cell surface expression of both myc-CRLR and myc-RAMP1 were sensitive to dynamin K44A and ␤-arrestin2(319 -418). The internalization of the myc-CRLR⅐myc-RAMP1 complex was also found to be blocked by the two dominant negative mutants using the cross-linking approach described above, see Fig. 8C. Taken together these results indicate the involvement of the classical clathrin-coated pit endocytosis pathway. Overexpression of the wild-type ␤-arrestin2 had only marginal effects on the extent of the complex internalization observed by FACS and cross-linking indicating that the level of ␤-arrestin expression in HEK cells is already sufficient to promote maximal endocytosis. This is consistent with previous studies ad- dressing the endocytosis of the ␤ 2 -adrenergic receptor in the same cell type (33,35). The inhibition of endocytosis by the ␤-arrestin2 dominant mutant could be interpreted as an indication of a role of this regulatory protein in CRLR-RAMP internalization as has been proposed for several GPCRs (33, 36 -38). However, because ␤-arrestin2(319 -418) inhibits receptor internalization by interacting with clathrin and not by preventing receptor-␤-arrestin association (33), it is conceivable that the dominant negative effect could result from a general interference with the clathrin-mediated endocytosis machinery and thus is not definite proof of ␤-arrestin involvement. The role of ␤-arrestin and specifically its recruitment by the receptor complex, was assessed by fluorescence microscopy. For this purpose, a ␤-arrestin2-YFP fusion construct was co-expressed with either HA-CRLR and RAMP1 or CRLR and myc-RAMP1. Cell surface CRLR or RAMP1 were prelabeled with anti-HA and anti-Myc antibodies, respectively, prior to agonist exposure. The subcellular distribution of the proteins was then revealed by immunofluorescence and direct excitation of the YFP. As shown in Fig. 9A, under basal conditions myc-RAMP1 and HA-CRLR immunoreactivity was detected at the cell surface while ␤-arrestin2-YFP was found in the cytosol as expected. As previously described, a similar distribution was observed when cells expressing the V2-vasopressin or the ␤ 2adrenergic receptors were co-expressed with ␤-arrestin2-YFP (Fig. 9A). Following a 30-min CGRP treatment, the ␤-arres-tin2-YFP was redistributed into endocytic vesicles where it co-localized with myc-RAMP1 and HA-CRLR (Fig. 9B). This distribution pattern is very similar to the one observed following V2-vasopressin receptor stimulation but is drastically dif-ferent from what is observed for the stimulated ␤ 2 -adrenergic receptor (Fig. 9B).
The difference in the ␤-arrestin redistribution patterns observed among GPCRs has been attributed to different affinities of the regulatory protein for distinct receptors. The co-localization of the receptor and ␤-arrestin in endocytotic vesicles such as it is the case for the V2-vasopressin receptor was interpreted as indicative of a stable high affinity interaction between the receptor and ␤-arrestin. The lack of internalization of ␤-arrestin upon ␤ 2 -adrenergic stimulation would reflect a rapid dissociation of ␤-arrestin from the receptor during internalization (39,40). The results obtained with the CRLR⅐RAMP1 complex confirm the involvement of ␤-arrestin in CGRP receptor internalization and indicates that it belongs to the class of receptor that forms a high affinity interaction with this regulatory protein (39,40). The originality of the finding resides in the fact that the CGRP receptor is already a stable complex of CRLR and RAMP1. This indicates that a ternary complex, involving CRLR-RAMP and ␤-arrestin, serves as a substrate for the endocytotic machinery. Recent studies have suggested that clusters of serines and/or threonines in the carboxyl tail of the receptor are responsible for the high-affinity interaction with ␤-arrestin (39,40). No such sequence is present in the short cytoplasmic tail of RAMP1 whereas a TXSTXS (TVSTIS) sequence is found in the tail of the CRLR, suggesting that it is through an interaction with the receptor that ␤-arrestin promotes the internalization of the complex. When considering other receptors belonging to the family B of GPCRs both ␤-arrestin-dependent and -independent internalization processes have been documented. For the gonadotropin-releasing hor- FIG. 9. Intracellular distribution of internalized CRLR⅐RAMP1 complexes and ␤-arrestin. HEK293T cells stably expressing the HA-tagged CRLR transiently transfected with RAMP1, myc-RAMP1 transiently transfected with CRLR, the myc-tagged V2 receptor (myc-V2R) or the HA-tagged ␤ 2 -adrenergic receptor (HA-b2AR) were transfected transiently with ␤-arrestin2-YFP (arrestin-YFP). These cells were labeled with rabbit anti-HA or mouse anti-myc antibodies under nonpermeabilized conditions prior to a 0 (A) or 30 (B) min incubation at 37°C with 100 nM CGRP, 100 nM arginine vasopressin or 1 M isoproterenol. The cells were then fixed, permeabilized, and incubated with Texas Red-conjugated goat anti-rabbit or anti-mouse to visualized RAMP1 or the receptors (Red). ␤-Arrestin2-YFP (arrestin-YFP) was visualized in green. mone, the vasoactive intestinal polypeptide 1 and the secretin receptors, no role of ␤-arrestin in the internalization was found (36,(41)(42)(43) whereas a dominant negative mutant of ␤-arrestin blocked the internalization of the thyrotropin-releasing hormone and the parathyroid hormone receptors (44 -46).
In conclusion, the results presented here reinforce the idea that CRLR and RAMP1 form a stable complex that serves as a CGRP receptor. The complex appears to be formed in the ER, is delivered to the cell surface and then internalized after activation by agonist. Although atypical by its heterodimeric composition that is required for targeting to the plasma membrane, the CGRP receptor shares endocytotic mechanisms that are common to most classical 7TM receptor. Indeed, involvement of ␤-arrestin in the internalization process suggests that phosphorylation by G protein-coupled receptor kinases, as has been shown for many receptors (19), is implicated in CRLR-RAMP1 responsiveness regulation. Consistent with this notion, phosphorylation of the CGRP receptor by G protein-coupled receptor kinase 6 was recently documented (47). The possibility that regulation might occur through the dissociation of RAMP1 from CRLR (and possibly through a competitive interaction with other RAMPs) now seems remote as the 7TM and 1TM proteins, CRLR and RAMP1, appear to behave as a unitary complex. If an equilibrium between RAMPs and CRLR does occur, it most likely takes place at the ER or Golgi level. In that respect, it is interesting that RAMP1 expressed alone was found both in the ER and Golgi while CRLR in the absence of RAMP1 was restricted to the ER. The change in the nature of the disulfide bridges found within RAMP1 on forming a complex with CRLR suggests a complex series of events upon association with CRLR. This may provide insights into the roles of RAMPs as chaperones and on the mechanism of chaperones in general. RAMPs, however, remain unique in constituting a fundamental component of at least two GPCRs, CRLR and calcitonin receptor. This may suggest how evolution enabled receptor selectivity through combinatorial mechanisms that generate maximum diversity from a relative small gene pool.