Mutation of Individual Serine Residues in the C-terminal Tail of the Lutropin/Choriogonadotropin Receptor Reveal Distinct Structural Requirements for Agonist-induced Uncoupling and Agonist-induced Internalization*

We have previously mapped the agonist-induced phosphorylation of the rat lutropin/choriogonadotropin receptor (rLHR) to a locus of four serines (Ser635, Ser639, Ser649, and Ser652) located in the C-terminal tail. The removal or mutation of this locus delays the time course of agonist-induced uncoupling of the rLHR from its effector system without affecting the overall magnitude of uncoupling, and it retards the endocytosis of the agonist-receptor complex. We have now prepared and analyzed four new rLHR mutants in which each of these serines were individually mutated to alanines. The data presented show that each mutation reduces agonist-promoted rLHR phosphorylation by 20–40%. Mutation of Ser635 or Ser639 delayed the time course of agonist-induced uncoupling to about the same extent as the simultaneous mutation of all four serines. Mutation of Ser635 or Ser639 also retarded agonist-induced internalization, but the magnitude of this decrease was less than that induced by the simultaneous mutation of all four serines. Mutation of Ser649 had no effect on agonist-induced uncoupling but retarded agonist-induced internalization to the same extent as the simultaneous mutation of all four serines. Mutation of Ser652 has little or no effect on either of these two parameters. Co-transfection studies with dominant-negative arrestins and dominant-negative dynamin reveal that, despite differences in their rates of internalization, rLHR-wild-type, rLHR-S639A, and rLHR-S649A are internalized by an arrestin- and dynamin-dependent pathway. These data show that the structural requirements needed for the agonist-induced uncoupling and internalization of the rLHR are distinct.

Phosphorylation of G protein-coupled receptors (GPCRs) 1 on serine and/or threonine residues is an important event in agonist-induced desensitization. GPCR phosphorylation by second messenger-dependent kinases attenuates signaling by uncoupling the receptors from their cognate G proteins, whereas phosphorylation by the G protein-coupled receptor kinases (GRKs) facilitates the interaction of the receptors with a family of inhibitory proteins called arrestins (1,2). This phosphorylated receptor-arrestin interaction uncouples the receptors from their cognate G proteins and targets the activated receptor to clathrin-coated pits for subsequent internalization (1)(2)(3)(4). Thus, the complex formed by the phosphorylated GPCR and arrestin serves as a common intermediate for the uncoupling of the receptor from its cognate G protein and for receptor internalization.
Using human kidney 293 cells stably transfected with the rat lutropin/choriogonadotropin receptor (rLHR) cDNA, we showed that, like many other GPCRs, the rLHR becomes phosphorylated on serine residues when the cells are stimulated with an agonist (lutropin (LH) or choriogonadotropin (CG)) (5). The identity of the kinases that mediate the agonist-induced phosphorylation of the rLHR is not known; however, neither kinase A nor kinase C can fully account for the agonist-induced phosphorylation of the rLHR (5,6). The involvement of one of the GRKs in the agonist-induced phosphorylation of the rLHR is suggested by the finding that overexpression of GRK2, GRK4, or GRK6 enhances agonist-induced phosphorylation. 2 A GRKcatalyzed phosphorylation of the rLHR is also suggested by functional studies showing that co-transfection of the rLHR with GRK2 or GRK4 diminishes the hCG-induced cAMP response (7).
By analogy with what is known about other GPCRs (see above) we proposed that the agonist-induced phosphorylation of the rLHR was responsible for the agonist-induced uncoupling of this receptor from its effector system (5). Further analysis of rLHR mutants truncated at residues 653, 631, or 628 (designated rLHR-t653, rLHR-t631, or rLHR-t628) and a full-length rLHR mutant with multiple serine substitutions (designated rLHR-5S/T3 A) mapped the agonist-induced phosphorylation to a cluster of four serine residues (Ser 635 , Ser 639 , Ser 649 , and Ser 652 ) present in the C-terminal tail and established some functional consequences of phosphorylation of this cluster (6,8,9). Rat LHR-t653, a truncated form of rLHR that retains Ser 635 , Ser 639 , Ser 649 , and Ser 652 , displays little or no reduction in the agonist-induced phosphorylation, as well as a normal time course and magnitude of agonist-induced uncoupling. On the other hand, rLHR-t631 and rLHR-t628, two truncated forms of the rLHR that lack Ser 635 , Ser 639 , Ser 649 , and Ser 652 , and a full-length receptor mutant in which these four residues were simultaneously mutated to alanines (i.e. rLHR-5S/T3 A) display a 90 -100% decrease in agonist-induced phosphorylation and a delay in the rate of agonist-induced uncoupling. The magnitude of agonist-induced uncoupling observed under prolonged agonist stimulation is unaffected, however (8,9).
It has been known for many years that one of the consequences of agonist binding to the LHR is the endocytosis of the agonist-receptor complex (10). Although the endocytosis of the agonist-bound LHR has been shown to occur via coated pits (11), the rate of endocytosis of the agonist-receptor complex is rather slow (half-life of 60 -120 min, depending on the cell type; see Refs. 9,10,12,and 13), and the majority of the agonistreceptor complex is routed to the lysosomes, where both the agonist and the receptor are degraded (11,14). This pathway ultimately leads to an agonist-induced reduction in the density of cell surface receptors by routing the receptor to a degradation rather than a recycling pathway (14,15). In keeping with current views on the internalization of other GPCRs, we have shown that the activation of the rLHR is necessary for efficient endocytosis (12,13). Moreover, the importance of the phosphorylation of the four-serine locus mentioned above in the endocytosis of the receptor-bound agonist was documented by the finding that cells expressing rLHR-5S/T3 A internalize the bound agonist at a slower rate than cells expressing the wildtype rLHR (9).
The experiments presented here were designed to determine which of the four serine residues present in this locus of the rLHR become phosphorylated and to more carefully define the role of each of these residues in the agonist-induced uncoupling and internalization of the rLHR. To this end, we constructed and analyzed four new rLHR mutants in which Ser 635 , Ser 639 , Ser 649 , or Ser 652 was individually mutated to an alanine residue in the context of the full-length rLHR. These mutants were analyzed for phosphorylation, uncoupling, and internalization.

MATERIALS AND METHODS
Plasmids and Cells-The cloning of the rat luteal LH/CG receptor cDNA and the template plasmid containing the full-length coding region plus portions of the 5Ј-and 3Ј-untranslated regions of the wild-type rLHR cDNA have been described previously (16). The individual Ser to Ala mutants were constructed using polymerase chain reaction strategies to alter the nucleotides coding for these residues. The sequence of the entire region of each mutant cDNA generated by polymerase chain reaction was verified by automated DNA sequencing. The mutant and wild-type rLHR cDNAs were subcloned into the eukaryotic expression vector pcDNAI/Neo (Invitrogen) for transfection. A plasmid encoding for an HA-tagged dominant-negative mutant form of dynamin (i.e. dynamin-K44A; see Ref. 17) was obtained from Sandra Schmid and subcloned into pcDNA3.1 (Invitrogen) for transfection. The expression vectors (all in pcDNA3.1) encoding for ␤-arrestin, arrestin-3, ␤-arrestin-V53D, and ␤-arrestin (319 -418) have also been described (4).
Transient co-transfections of human embryonic kidney (293) cells were done using calcium phosphate as described by Chen and Okayama (18). Cells plated in 100-mm dishes were transfected when 70 -80% confluent using 10 g of each plasmid. After an overnight incubation, the cells were washed and incubated in growth medium for 1-2 h at 37°C. The cells were then trypsinized, distributed into 35-mm wells (5-10 ϫ 10 5 cells/well), and used 24 h later. Stably transfected cell lines were obtained following G418 selection and cloning as described elsewhere (6,13). The establishment and properties of two clonal cell lines expressing rLHR-wt (designated 293L(wt-12) and 293L(wt-17)) and a clonal cell line expressing a mutant in which Ser 635 , Thr 638 , Ser 639 , Ser 649 , and Ser 652 were simultaneously mutated to alanines (designated 293L(5S/T3 A-2) have been described (8,9,19).
Intact Cell Phosphorylation Assays-Stably transfected cells were plated in 100-mm dishes and were metabolically labeled during a 4-h incubation in phosphate-free medium containing 200 Ci/ml 32 P i . Receptor phosphorylation was ascertained after incubating the 32 P i -prelabeled cells at 37°C with buffer only, or with 1000 ng/ml oLH for 5 min. These conditions were chosen to elicit a maximal response (5,6,8). One cell line expressing rLHR-wt (either 293L(wt-12) or 293L(wt-17)) and one cell line expressing one of the rLHR mutants was used in each experiment. Following lysis of the cells, the amount of wt and mutant receptor used for immunoprecipitation was equalized based on 125 I-hCG binding assays (see below). Immunoprecipitations were performed using Bugs (a rabbit polyclonal antibody to the rLHR (20)) or a mixture of AntiL and R02 (two polyclonal antibodies directed against synthetic peptides derived from the known amino acid sequence of the rLHR (19,21)). The methodology used for immunoprecipitation and SDS gels was the same as that described earlier (5,6,8,19). Autoradiograms of the dried gels were obtained using Kodak BioMax MS film and intensifying screens. The autoradiograms were scanned using a Bio-Rad Molecular Imaging System and captured in a digital format for presentation.
Hormone Binding and Signal Transduction Assays-Equilibrium binding parameters for hCG were measured during an overnight incubation (4°C) of intact cells with a fixed concentration of 125 I-hCG and increasing concentrations of hCG as described previously (6,8). Concentration-response curves for the hCG-induced increases in cAMP accumulation were obtained by measuring total cAMP levels in cells that had been incubated with at least five different concentrations of hCG for 30 min at 37°C in the presence of a phosphodiesterase inhibitor. The different parameters that describe the concentration response curves were calculated as described elsewhere (6,8).
Measurements of agonist-induced uncoupling in stably transfected cells were performed as follows. Cells expressing rLHR-wt or the mutant receptors were divided into two groups and incubated without (group A) or with (group B) 100 ng/ml oLH or hCG for 15 min. 3 All cells were then washed with neutral and acidic buffers to remove the free and bound hormone, respectively, and each group of cells was subdivided into two groups, which were then incubated without (groups A1 and B1) or with (groups A2 and B2) 100 ng/ml hCG for 15 min at 37°C (6,8). Intracellular levels of cAMP were measured at the end of this incubation, and agonist-induced uncoupling was calculated as follows: ((B2 Ϫ B1)/(A2 Ϫ A1)) ϫ 100. This same assay could not be used to measure uncoupling in transiently transfected cells because of their reduced responsiveness. Thus, when using transiently transfected cells, uncoupling was simply assessed by measuring the total levels of cAMP accumulated by the cells during a 15-min incubation with a saturating concentration of hCG (100 ng/ml) in medium containing a phosphodiesterase inhibitor (6 -8).
Internalization Assays-The endocytosis of 125 I-hCG was measured in cells that had been briefly preincubated (i.e. 10 min at room temperature) with 40 ng/ml 125 I-hCG. At the end of this preincubation, the free hormone was removed by washing, and the cells were re-incubated in fresh, hormone-free medium at 37°C for up to 4 h. After the desired interval, the cells were placed on ice, and the medium was saved. The cells were briefly treated with an isotonic pH 3 buffer (10, 13). The radioactivity that was released by the acid treatment was considered to be surface-bound, whereas the radioactivity that remained cell-associated was considered to be internalized. The medium was precipitated with trichloroacetic acid, and the acid-insoluble and -soluble radioactivity were considered to be undegraded and degraded hormone, respectively (10,13). Because there is little or no dissociation of the receptorbound hCG during this incubation (10), the rate of disappearance of the receptor bound hCG can be used to measure the rate of internalization. The rates of internalization (ke) were thus calculated from the slopes of linear regression fits to plots of the ln of the surface-bound hCG versus time. The half-life of internalization (t1 ⁄2 ) is therefore defined as 0.693/ke.
Immunoblots-Expression of the transfected arrestins and dynamin was ascertained by immunoblots using the ECL system of detection. The different arrestin constructs were detected using a polyclonal antibody (KEE) raised against a C-terminal 16 residue peptide (22) or a mouse monoclonal antibody (F4C1) directed against an epitope common to all known arrestins (23). The HA-tagged dynamin was detected using the 12CA5 monoclonal antibody (Boehringer Mannheim).
Hormones and Supplies-Purified hCG (CR-127) and oLH (AFP-5551B) were obtained from the National Hormone and Pituitary Agency of the NIDDK, National Institutes of Health. 125 I-hCG was prepared as described previously (24), to give a specific radioactivity of 25,000 -30,000 cpm/ng. [ 32 P]Orthophosphate was obtained from NEN Life Science Products. Phosphate-free DMEM was purchased from ICN Biomedicals (Irvine, CA). Nonidet P-40, protease inhibitors, N,NЈ,NЉtriacetylchitotriose, protein A-agarose, and bovine serum albumin were from Sigma. Okadaic acid and cypermethrin were purchased from Alexis Biochemicals (Woburn, MA). Wheat germ agglutinin agarose was from Vector Laboratories. Cell culture supplies and reagents were obtained from Corning (Corning, NY) and Life Technologies, Inc., respectively. All other materials were obtained from commonly used suppliers.

Preparation and Functional Properties of rLHR Mutants-
Phosphoamino acid analysis and phosphorylation experiments utilizing three different C-terminal truncations of the rLHR and a full-length mutant with multiple Ser to Ala mutations have identified Ser 635 , Ser 639 , Ser 649 , and Ser 652 as the major locus of rLHR phosphorylation in transfected cells (6,8,9). For the experiments presented here, we prepared four new fulllength rLHR mutants in which each of these four serines was individually mutated to alanine. Each mutant cDNA (designated rLHR-S635A, rLHR-S639A, rLHR-S649A, and rLHR-S652A) was transfected into human kidney 293 cells, and clonal lines stably expressing each of the mutants were obtained. All mutant receptors bound hCG with an affinity comparable to that detected in cells expressing rLHR-wt (i.e. about 200 pM, see Table I). Of the several clonal lines obtained with each mutant, we chose those with the highest cell surface receptor density for further study (Table I).
The data presented in Table I also show a comparison of the concentration response curves for hCG-induced cAMP accumulation in the clonal lines expressing the different mutants and in two clonal lines, 293L(wt-12) and 293L(wt-17), stably transfected with the rLHR-wt. We chose 293L(wt-12) and 293L(wt-17) as control cell lines because their receptor density (100,000-200,000 receptors/cell) is comparable to that of the cell lines expressing the mutant receptors, and at these levels of receptor expression, the cAMP response is basically independent of receptor density (6,8,19). In an attempt to correct for an inherent variability in the basal and hCG-stimulated levels of cAMP, we also measured the levels of cAMP in the different cell lines stimulated with cholera toxin and calculated a response ratio by dividing the maximal hCG response by the maximal cholera toxin response. As shown in Table I, this ratio shows that all mutants respond to hCG as well as or better than cells expressing rLHR-wt. It is noted that the concentration of hCG required to elicit the half-maximal cAMP response is 1.5-2.0-fold higher in the cells expressing the mutants than in the cells expressing rLHR-wt. This finding is consistent with previous data obtained using a mutant rLHR in which all four serine residues in question were simultaneously mutated to alanine (9) Agonist-induced rLHR Phosphorylation-The effect of individual Ser to Ala mutations on rLHR phosphorylation was determined by immunoprecipitation of the rLHR from the different stable cell lines that had been metabolically labeled with 32 P i . Each phosphorylation experiment consisted of four samples. Two samples were derived from 32 P i -labeled 293L(wt-12) or 293L(wt-17) cells incubated with or without agonist, and two samples were derived from 32 P i -labeled 293L(S635A-5), 293L(S639A-5), 293L(S649A-14), or 293L(S652A-5) cells, also incubated with or without agonist.
In earlier phosphorylation experiments, we used hCG as the agonist (5,6,8,9). In more recent experiments, we switched to using oLH as the agonist because the actions of LH and CG are indistinguishable under these conditions, but the rate of dissociation of the bound oLH is much faster than that of the bound hCG (25,26). Thus, when using oLH as the agonist, most of the bound oLH dissociates from the rLHR prior to the immunoprecipitation, whereas a substantial amount of the hCG remains receptor-bound. This turns out to be an important consideration in immunoprecipitation experiments because some of the receptor antibodies that we have started to use (see under "Materials and Methods") do not recognize the agonist-bound receptor. 4 As such, when using hCG as the agonist, it is possible to misinterpret a decrease in the 32 P signal as a decrease in phosphorylation, whereas in reality, the decrease is simply due to the inefficient immunoprecipitation of the hCG-receptor complex as compared with the free receptor.
Following stimulation, cells were lysed, and equal amounts 4 Immunoprecipitation experiments (not shown) using 293L(wt-12) or 293L(wt-17) cells metabolically labeled with [ 35 S]methionine showed that Bugs, a polyclonal antibody to the rLHR used in previous phosphorylation experiments (5,6,8,9), can immunoprecipitate equivalent amounts of receptor from cells that had been preincubated with buffer only, hCG, or oLH. Similar experiments done with some of the newer antibodies (antiL and RO2; see under "Materials and Methods") showed that these antibodies immunoprecipitate equivalent amounts of receptor from [ 35 S]methionine-prelabeled cells that had been preincubated with buffer only or oLH. The amount of receptor immunoprecipitated from cells preincubated with hCG was somewhat lower, however. a Equilibrium binding parameters were determined using intact cells during an overnight incubation at 4°C as described under "Materials and Methods." Each number represents the average Ϯ S.E. of 4 -6 independent experiments.
b Total (i.e. cells ϩ medium) levels of cAMP were measured in cells incubated in the presence of a phosphodiesterase inhibitor and increasing concentrations of hCG for 30 min or in cells incubated with a single concentration (500 ng/ml) of cholera toxin for 2 h, as described under "Materials and Methods." The response ratio shown in the far right was calculated by dividing the average maximal response obtained for hCG over the average maximal response obtained for cholera toxin. Each number represents the average (ϮS.E.) of three independent experiments for the cell lines expressing the mutant receptors or nine independent experiments for the cell lines expressing rLHR-wt. The results obtained with 293L(wt-12) and 293L(wt-17) were combined because they seem to be indistinguishable. c ND, not determined. These parameters were not remeasured in the current series of experiments, but previous data from this laboratory have shown that these two cell lines bind hCG with a K d of 200 -300 pM and display a B max of 100,000 -200,000 molecules/cell (8,9,19). of the wt and mutant receptors were immunoprecipitated, resolved on SDS gels, and visualized by autoradiography as shown in Fig. 1. The results of several experiments, such as those shown in Fig. 1, were quantitated by densitometry, and these are summarized in Table II. The results obtained with rLHR-wt are in agreement with previous data showing that this receptor is phosphorylated in unstimulated cells and that agonist-stimulation elicits ϳ2-fold increase in the 32 P signal (5,6,8,9). The results summarized in Table II also show that the individual mutation of Ser 635 , Ser 639 , Ser 649 , or Ser 652 to Ala diminishes agonist-induced phosphorylation to 57-78% of control. Because the removal or simultaneous mutation of all these four serines reduces phosphorylation to barely detectable levels (6,8,9), the simplest interpretation of the data shown in Table  II is that all four serines are phosphorylated when cells are stimulated with agonist.
If all four serines are phosphorylated to the same extent, each individual mutation should reduce phosphorylation to 75% of the rLHR-wt. However, a reduction of this magnitude (i.e. 78% of rLHR-wt) was found only with the rLHR-S635A mutant. The other three mutations reduced phosphorylation to 57-65% of the rLHR-wt, suggesting that phosphorylation is hierarchical or that not all residues are phosphorylated to the same extent. Additional studies to distinguish between these possibilities were not performed because the relatively low increase in agonist-promoted phosphorylation detected in rLHR-wt (ϳ2-fold; see Table II) makes the quantitation of partial changes in rLHR-phosphorylation rather difficult.
Agonist-induced Uncoupling-The time course of agonistinduced uncoupling of the LH/CG-sensitive adenylyl cyclase in 293 cells expressing rLHR-wt suggests the existence of two phases: a fast phase, which occurs within 15 min of agonist addition and leads to a 40 -60% reduction in agonist-stimulated cAMP synthesis, and a slower phase, which leads to a further 20 -40% reduction in the agonist-stimulated cAMP synthesis ( Fig. 2 and Refs. 8 and 9). Because the removal or mutation of Ser 635 , Ser 639 , Ser 649 , and Ser 652 affects the time course, rather than the magnitude of agonist-induced uncoupling (cf. Fig. 2 and Refs. 8 and 9), the effect of the individual Ser to Ala mutations on this process were initially tested by rechallenging cells with agonist after a short (i.e. 15-min) preincubation with or without agonist. Experiments were also  12) or 293L(wt-17) cells was prelabeled with 32 P i for 3 h. The cells were then further incubated with buffer only or with 1 g/ml oLH for 5 min as indicated. Following lysis of the cells, the amount of wt and mutant receptor used for immunoprecipitation was equalized based on 125 I-hCG binding assays, and the rLHR was immunoprecipitated as described under "Materials and Methods." The immunoprecipitates were resolved on SDS gels, and autoradiograms of the dried gels were obtained using Kodak BioMax MS film and intensifying screens. The autoradiograms were scanned using a Bio-Rad Molecular Imaging System and captured in a digital format for presentation. The results of a representative experiment are shown. Only the relevant portions of the autoradiograms are shown because previous experiments have demonstrated that the antibodies used are specific for the LHR and that the addition of preimmune serum to lysates of transfected cells or immune serum to lysates of untransfected cells do not result in the precipitation of any radiolabeled bands (5,19).
a Individual phosphorylation experiments were done with one of the cell lines stably transfected with a mutant receptor and either 293L(wt-12) or 293L(wt-17) cells as shown in Fig. 1. The autoradiograms were scanned and quantitated using a BioRad Molecular Imaging System, and the magnitude of the signal measured in agonist-treated cells was expressed as fold over basal. Basal levels of receptor phosphorylation varied by at most 30% in all cell lines, except for cells expressing rLHR-S639A, in which basal phosphorylation was enhanced 2-3-fold when compared to cells expressing rLHR-wt (cf. Fig. 1 c Not determined during the course of these experiments. However, previous results from this laboratory have shown that the basal and agonist-induced phosphorylation of rLHR-5S/T 3 A are barely detectable (9). done using either 293L(wt-12) or 293L(wt-17) cells as controls, because the agonist-induced uncoupling of cells expressing rLHR-wt is independent of receptor density (8,9).
In agreement with previous results (8,9), the data presented in Table II show that the response of agonist-pretreated cells expressing the rLHR-wt is reduced to 43% of the response of cells that were not pretreated with agonist. Mutation of Ser 649 or Ser 652 had no effect on this parameter, because the responses of agonist-pretreated cells expressing either of these mutants were reduced to 37 and 35%, respectively, of the response detected in cells not pretreated with agonist. The responses of agonist-pretreated cells expressing rLHR-S635A or rLHR-S639A were reduced to 60 and 69%, respectively, of the response detected in cells not pretreated with agonist. In parallel experiments, and in agreement with previous results (9), the response of agonist-pretreated cells expressing rLHR-5S/T3 A (a mutant in which Ser 635 , Ser 639 , Ser 649 , and Ser 652 are simultaneously mutated to Ala) was reduced to 74% of the response detected in cells not pretreated with agonist. We have also reported that under the same experimental conditions, the subsequent agonist-induced response of cells expressing rLHR-t631 or rLHR-t628 (two truncation mutants in which Ser 635 , Ser 639 , Ser 649 , and Ser 652 were removed) is reduced to 69 and 66%, respectively, of the response detected when they are preincubated without agonist (8). Thus, the magnitude of the agonist-induced uncoupling in cells expressing receptors in which Ser 635 , Ser 639 , Ser 649 , and Ser 652 were removed or simultaneously mutated is similar to that detected in cells expressing receptors in which only Ser 635 or Ser 639 was mutated.
Based on these results, we conclude that both Ser 635 and Ser 639 are needed for agonist-induced uncoupling. This conclusion is further supported by the data presented in Fig. 2, which show a more complete time course of agonist-induced uncoupling in clonal lines expressing rLHR-wt, a mutant in which Ser 635 , Ser 639 , Ser 649 , and Ser 652 were simultaneously mutated to alanines (rLHR-5S/T3 A) or one of the single substitution mutants (rLHR-S639A). These results clearly show that the early time course of uncoupling is retarded to the same extent in cells expressing rLHR-5S/T3 A or rLHR-S639A and that the magnitude of uncoupling detected at later time points (i.e. 60 min or later) is basically the same in cells expressing rLHR-wt or either of these two mutants.
Agonist-induced Internalization-The effect of individual serine mutations of the rLHR on the endocytosis of hCG were measured using a protocol that allows us to follow one round of endocytosis of the bound hormone (10,13). The half-life of internalization of hCG in cells expressing the rLHR-wt is slow (1-2 h, depending on the cell type; see Refs. 9, 10, 12, and 13) when compared with the half-life of internalization of many other ligand-receptor complexes (27). In agreement with previous data, the results summarized in Fig. 3 and Table III show that 293 cells expressing rLHR-wt (either 293L(wt-12) or 293L(wt-17)) internalize the bound hCG with a half-life of 139 min. These data also show that mutation of Ser 652 had little or no effect on internalization, mutation of Ser 635 or Ser 639 increased the half-life of internalization 1.4 -1.6-fold, and the mutation of Ser 649 increased the half-life of internalization 2.8-fold. In fact, the half-life of internalization of hCG in cells expressing rLHR-S649A is similar to that measured in parallel experiments using cells expressing rLHR-5S/T3 A, a mutant in which all four serines were simultaneously mutated to alanines (Table III).

FIG. 2. Time course of agonist-induced uncoupling in clonal lines of 293 cells expressing the wild-type or mutant receptors.
One of two cell lines stably expressing rLHR-wt (either 293L(wt-12) or 293L(wt-17)) (f), 293L(S639A-5) (E), or 293L(5S/T3 A-2) (¹) cells were preincubated with agonist for the indicated times at 37°C. The free and bound agonist were then removed (see under "Materials and Methods"), and the cells were divided into two groups and incubated without or with 100 ng/ml hCG for 15 min at 37°C. At the end of this incubation, the medium was aspirated, and the cells were used to determine the intracellular cAMP content. The amount of cAMP accumulated in the group of cells incubated without agonist was then subtracted from that present in the cells incubated with agonist, and the results were expressed as a percentage of the cAMP response determined under identical conditions but using cells that were preincubated without agonist (see under "Materials and Methods" and the legend to Table II). Each data point represents the average Ϯ S.E. of four independent experiments. The results obtained using 293L(wt-12) and 293L(wt-17) cells were combined because they were indistinguishable.

FIG. 3. Endocytosis of the receptor-bound 125 I-hCG in clonal lines of 293 cells expressing the wild-type and mutant receptors.
Cell lines stably transfected with rLHR-wt or mutants were preincubated with 40 ng/ml 125 I-hCG for 10 min at room temperature. f, 293L(wt-12) or 293L(wt-17); Ⅺ, 293L(S635A-5); E, 293L(S639A-5); ‚, 293L(S649A-14); ϫ, 293L(S652A-5). At this point (t ϭ 0 in the figure), the free hormone was removed by washing, and the cells were incubated in fresh, hormone-free medium at 37°C to allow internalization and degradation of the surface-bound hormone. At the times indicated, the cells were placed on ice and used to determine the amount of surfacebound, internalized, and degraded hormone as described under "Materials and Methods." The internalized and degraded radioactivity were manually combined to simplify the presentation of the data into two, rather than three, compartments. The surface-bound (top panel) and the internalized and degraded radioactivity (bottom panel) are expressed as a percentage of the total hormone present (i.e. surface bound, internalized, degraded, and dissociated). Each data point represents the average Ϯ S.E. of five independent experiments. The results shown for rLHR-wt were obtained using 293L(wt-12) and 293L(wt-17) cells and were combined because they were indistinguishable.
Based on these results we conclude that Ser 635 , Ser 639 , and Ser 649 are needed for internalization and that the role of Ser 649 is particularly important.
Roles of Arrestin and Dynamin on Agonist-induced Internalization and Uncoupling-In order to learn more about the structural requirements for the agonist-induced internalization of the rLHR, we examined the effects of nonvisual arrestins, dynamin, and mutants thereof on the internalization of rLHR-wt, rLHR-S639A (one of the two mutants that slows down internalization and uncoupling; cf. Figs. 2 and 3, and Tables II and III), and rLHR-S649A (a mutant that blocks internalization but does not affect uncoupling; cf. Fig. 3 and Tables II and III). As already mentioned above, nonvisual arrestins are clathrin-binding proteins that act as adapters, linking GPCRs to clathrin-coated pits (3). Two mutant forms of ␤-arrestin, ␤-arrestin-V53D and ␤-arrestin(319 -418), act as dominant negative mutants of arrestin-mediated GPCR internalization because they have reduced binding affinities for the phosphorylated GPCRs (4). Dynamin is a GTPase that participates in the fission of endocytic vesicles from the plasma membrane (28). Dynamin mutants that are deficient in GTP binding (such as dynamin-K44A) block coated pit-mediated internalization of several receptors (17,29,30).
The effects of arrestin-3, ␤-arrestin, 5 ␤-arrestin-V53D, ␤-arrestin(319 -418), or dynamin-K44A on the internalization of hCG mediated by rLHR-wt, rLHR-S639A, or rLHR-S649A were examined using transient co-transfection assays. The expression of each of these constructs has been documented before (4,31), and it was verified in the present experiments using Western blots, as described under "Materials and Methods" (data not shown). In agreement with the results obtained with the stably transfected cell lines (cf. Fig. 3), the results presented in Fig. 4 show that the internalization of hCG is greater in cells transiently transfected with rLHR-wt (42% of the bound hormone was internalized and/or degraded during a 2-h internalization assay) than in cells transfected with rLHR-S639A (35% of the bound hormone was internalized and/or degraded) and much higher than in cells transfected with rLHR-S649A, in which only 25% of the bound hormone was internalized and or degraded. Co-transfection with arrestin-3 increased the internalization of hCG mediated by all three receptors by about 2-fold, but the internalization of hCG detected in cells co-transfected with rLHR-S649A and arrestin-3 (56%) was lower than that detected in cells co-transfected with rLHR-wt and arrestin-3 (75%) or rLHR-S639A and arrestin-3 (73%).
An involvement of nonvisual arrestin in the endocytosis of hCG is best documented by the finding that the two dominantnegative mutants of ␤-arrestin inhibit hCG internalization (Fig. 4). In agreement with previous data obtained with the ␤ 2 -adrenergic receptor (4), we found that ␤-arrestin(319 -418) is somewhat more effective than ␤-arrestin-V53D in inhibiting the internalization of hCG. The effects of these two dominantnegative arrestins seem to be similar for rLHR-wt, rLHR-S639A, and rLHR-S649A, however. The inhibitory effects of dynamin-K44A on the endocytosis of hCG (40 -60% reduction) were somewhat more pronounced than those of the dominantnegative arrestins but were similar for rLHR-wt, rLHR-S639A and rLHR-S649A. Lastly, treatment of 293 cells expressing rLHR-wt with hypertonic sucrose, another manipulation that disrupts the assembly of clathrin-coated pits (32) also inhibited the endocytosis of hCG by about 50% (data not shown).
We also attempted to use transient co-transfection assays to test for the putative involvement of arrestin on the agonistinduced uncoupling of the rLHR. Unfortunately, the methods used to measure uncoupling in stably transfected cell lines (see legend to Table II) could not be used in transiently transfected cells because of their relatively weak cAMP response. As such, we were forced to measure uncoupling simply by measuring the magnitude of the cAMP response elicited by hCG in 293 cells transiently transfected with the rLHR-wt alone or together with different arrestin constructs. This paradigm has been used before by Premont et al. (7) to show that co-transfection of 293 cells with the rLHR-wt and GRK2 or GRK4 reduces the hCG-induced cAMP response of cells transfected with rLHR-wt only. The results of these experiments are presented in Table  IV; they show that under these experimental conditions, arrestin-3 or ␤-arrestin(319 -418) has a slight stimulatory effect on the hCG-induced cAMP response. 6 Thus, under the same experimental conditions where we can readily measure opposite effects of arrestin and dominant-negative arrestins on the agonist-induced internalization of the rLHR (cf. Fig. 4), we cannot measure opposite effects on uncoupling.
Taken together, these results show that the agonist-induced internalization of the rLHR is arrestin-and dynamin-dependent. They also suggest that the agonist-induced uncoupling of the rLHR is arrestin-independent. DISCUSSION Together with previous data from this laboratory (5,6,8,9), the experiments presented here are consistent with a model in which four serines present in the C-terminal tail of the rLHR (Ser 635 , Ser 639 , Ser 649 , and Ser 652 ) are phosphorylated in response to agonist stimulation. This accommodates the findings that the removal or mutation of Ser 635 , Ser 639 , Ser 649 , and Ser 652 decreases basal and agonist-induced phosphorylation by at least 90% (6,8,9), whereas the individual serine mutations reduce agonist-induced phosphorylation only partially (Table  II). However, the reduction in the magnitude of phosphorylation detected with each mutation suggests that all four serines are not phosphorylated to the same extent or that there is a Rates of internalization of hCG in clonal cell lines expressing the wild-type and mutant receptors Cell lines stably transfected with rLHR-wt or mutants thereof were preincubated with 40 ng/ml 125 I-hCG for 10 min at room temperature. At this point, the free hormone was removed by washing, and the cells were incubated in fresh, hormone-free medium at 37°C to allow internalization and degradation of the surface-bound hormone (cf. Fig. 3). The rates of internalization were then calculated from the rate of disappearance of the surface-bound hormone as described under "Materials and Methods." Each value represents the average Ϯ S.E. of five independent experiments. The results shown for rLHR-wt were obtained using 293L(wt-12) and 293L(wt-17) cells and were combined because they were indistinguishable. hierarchy of phosphorylation sites. Additional studies to distinguish between these possibilities (such as the simultaneous mutation of groups of two or three serines at a time) were not performed because the low magnitude of the phosphorylation signal ( Fig. 1 and Table II) makes the quantitation of partial changes in rLHR-phosphorylation rather difficult. Moreover, because the data presented here clearly show that the functional impact of the individual mutation of some of these serines is similar (or identical) to that observed when all four serines are mutated simultaneously (cf. Fig. 2 and Tables II  and III), additional mutagenesis studies would not contribute much new information to our understanding of the functional significance of these four serines.
The functional data presented here clearly show that the attenuation of the agonist-induced uncoupling that results as a consequence of the removal (8) or mutation of Ser 635 , Ser 639 , Ser 649 , and Ser 652 (9) can be fully reproduced by the mutation of either Ser 635 or Ser 639 (Table II and Fig. 2), whereas the slower rate of agonist-induced internalization detected in the multiple substitution mutant (9) can be fully reproduced by mutation of Ser 649 (Table III and Fig. 3). The mutation of Ser 652 had little or no effect on agonist-induced uncoupling or internalization. Taken together, these data show that the structural requirements for agonist-induced uncoupling and agonist-induced internalization of the rLHR are different. Some overlap between these structural motifs is suggested by the finding that mutation of Ser 635 or Ser 639 , which induced maximal attenuation of uncoupling, also induced partial attenuation of internalization (Figs. 2 and 3 and Tables II and III).
One of the most interesting recent advances in our understanding of the biology of GPCRs is the realization that the complex formed by the phosphorylated ␤ 2 -adrenergic receptor (␤ 2 AR) and arrestin plays a pleiotropic role in the regulation of ␤ 2 AR function (3,4,33,34). Thus, it is now generally accepted that the GRK-catalyzed phosphorylation of the ␤ 2 AR increases the affinity of the ␤ 2 AR for ␤-arrestin and that the formation of the phosphorylated ␤ 2 AR-arrestin complex sterically hinders the ␤ 2 AR-G protein association. Because ␤-arrestin can also bind clathrin with high affinity, the ␤ 2 AR-arrestin complex is targeted for internalization through clathrin-coated pits. Some variations on this theme have also begun to emerge, as illustrated by the finding that agonist stimulation of the AT 1A angiotensin (29) and the m2 muscarinic receptors (31) leads to receptor phosphorylation, but the phosphorylated receptors are internalized by a pathway that does not require the participation of arrestin or clathrin-coated pits. However, overexpression of nonvisual arrestins forces the agonist-stimulated AT 1A angiotensin and m2 muscarinic receptors to be internalized via clathrin-coated pits by a pathway that requires nonvisual arrestins (29,31).
The data presented here for the rLHR-wt expressed in 293 cells shows that the agonist-induced internalization of the rLHR, like that of the ␤ 2 AR expressed in 293 cells, requires the participation of arrestin and clathrin-coated pits (Fig. 4). Thus, the overexpression of two dominant-negative forms of ␤-arrestin and a dominant-negative form of dynamin, as well a treat- After an overnight transfection, the cells were washed, trypsinized, distributed into six-well plates, and used for internalization assays 24 h later (see under "Materials and Methods" for details). A single round of internalization assays was performed by first incubating the cells with 40 ng/ml [ 125 I]hCG for 10 min at room temperature. After washing (to remove the free hormone), the cells were incubated for 2 h at 37°C to allow for processing of the surface-bound hormone, and the amounts of surface-bound, internalized, and degraded radioactivity were measured as described under "Materials and Methods" and in the legend to Fig. 2. The internalized and degraded radioactivity levels were added manually and were expressed as a percentage of the total hormone present (i.e. surface bound, internalized, degraded, and dissociated). Each column represents the average Ϯ S.E. of three independent experiments. The numbers within each column represent percentages relative to the control cells (i.e. those transfected with the receptor plus empty vector plasmids).

TABLE IV Effects of arrestin and dominant-negative arrestin on the hCGinduced cAMP response of cells transiently expressing rLHR-wt
293 cells plated in 100-mm dishes were transfected with 10 g of an expression vector encoding for rLHR-wt together with 10 g of empty vector, arrestin-3, or ␤-arrestin(319 -418), as indicated. After an overnight transfection, the cells were washed, trypsinized, distributed into six-well plates, and used 24 h later (see under "Materials and Methods" for details). At this point, the cells were incubated (15 min at 37°C) in medium containing 1 mM isobutylmethylxanthine and the indicated additions. Total cAMP was measured as described under "Materials and Methods." Each number represents the mean Ϯ S.E. of three independent transfections. ment with hypertonic sucrose inhibit the agonist-induced internalization of rLHR-wt. In agreement with data obtained with the ␤ 2 AR expressed in 293 cells (4), we found that ␤-arrestin-V53D is less effective than ␤-arrestin(319 -418) in inhibiting the internalization of the rLHR. However, the magnitude of the inhibition of rLHR-wt internalization induced by overexpression of the two dominant-negative ␤-arrestins and the dominant negative dynamin in 293 cells is generally lower than those reported for the ␤ 2 AR in the same cell line (4,29,34,35). A more notable difference is with the effect of arrestin-3, as the overexpression of this protein in 293 cells has little or no effect on the internalization of the ␤ 2 AR (4, 29, 34, 35), but it enhances the internalization of the rLHR-wt about 2-fold (Fig. 4).
When considered together, these data suggest that the rLHR-wt has a low affinity for arrestins and that the slow rate of internalization of the rLHR-wt (t1 ⁄2 ϳ 140 min; see Table III) compared with that of the ␤ 2 -AR (t1 ⁄2 Ͻ 30 min; see Ref. 29) may be a reflection, at least in part, of the weak rLHR-arrestin interaction.
The effects of the two dominant-negative ␤-arrestins and dominant-negative dynamin on rLHR-S639A or rLHR-S649A were virtually indistinguishable from those detected with rLHR-wt (Fig. 4) and support the idea that the agonist-induced internalization of these two mutants also occurs by an arrestinand coated pit-dependent pathway. The experiments presented here are the first to address the involvement of nonvisual arrestins in this phenomenon. The involvement of clathrincoated pits in the internalization of the LHR-wt has been previously documented using ultrastructural approaches, however (11).
Although transient cotransfection assays have been used by others (7) to show that GRK overexpression reduces the hCGinduced cAMP response mediated by rLHR-wt, the same paradigm failed to reveal opposite effects of arrestin-3 or a dominant-negative ␤-arrestin on the hCG-stimulated cAMP response. The overexpression of both arrestin-3 and a dominant-negative ␤-arrestin had a slight stimulatory effect on hCG-stimulated cAMP accumulation (Table IV). Although these results suggest the existence of an arrestin-independent pathway for the agonist-induced uncoupling of the rLHR, they may need to be interpreted with caution because transient co-transfection assays of 293 cells have also failed to detect an effect of ␤-arrestin or ␤-arrestin-V53D on the agonist-induced cAMP response mediated by the ␤ 2 AR (34). Likewise, transient overexpression of ␤-arrestin on JEG-3 cells has no effect on the agonist-induced desensitization of the m2 muscarinic receptor unless one of the GRKs is also co-transfected (36). In fact, ␤-arrestin overexpression has been shown to enhance the agonist-induced desensitization of the ␤ 2 AR only in stably transfected Chinese hamster ovary cells expressing large amounts of ␤ 2 AR (37). On the other hand, the possibility of an agonistpromoted but arrestin-independent pathway for rLHR-uncoupling cannot be completely dismissed, as there are studies that demonstrate that the GRK-catalyzed phosphorylation of rhodopsin can partially inhibit G protein coupling in the absence of arrestin (38 -40), and there is at least one study documenting that the GRK-catalyzed phosphorylation of the ␤ 2 AR can also directly inhibit G protein coupling in the absence of arrestin (41).
When taken together, the effects of arrestin overexpression on the agonist-induced uncoupling (Table IV) and internalization (Fig. 4) of the rLHR-wt, as well as the effects of individual Ser mutations on agonist-induced uncoupling ( Fig. 2 and Table  II) and internalization of the rLHR (Fig. 2 and Table III) are consistent with the following two models.
Model 1 assumes that Ser 635 , Ser 639 , and Ser 649 are involved in the interaction of the rLHR with arrestin, but the role of Ser 649 in this interaction is more important than that of Ser 635 and Ser 639 . This model also assumes that arrestin is involved in the agonist-induced internalization of the rLHR but not in the agonist-induced uncoupling of the rLHR. Model 1 accommodates all the findings presented here. Thus, the mutation of Ser 635 , Ser 639 , or Ser 649 slows down internalization, but the effect of S649A mutation is more pronounced than those of the S635A or S639A mutation ( Fig. 3 and Table III). Furthermore, arrestin-3 overexpression does not enhance the internalization of rLHR-S649A to the same level as that detected with rLHR-wt or rLHR-S639A (Fig. 4) because the S649A mutation reduces the binding affinity for arrestin. Because Ser 635 , Ser 639 , and Ser 649 are all involved in the receptor-arrestin interaction, Model 1 is also consistent with the finding that the dominant-negative ␤-arrestins inhibit the internalization of rLHR-wt, rLHR-S639A, and rLHR-S649A (Fig. 4). The lack of effect of the S649A mutation and the effect of the S635A and S639A mutations on agonist-induced uncoupling (Table II) are accommodated by the assumption that uncoupling is arrestinindependent. This assumption is, in turn, supported by our inability to demonstrate an effect of arrestin-3 or dominantnegative ␤-arrestins on the agonist-induced cAMP accumulation (Table IV) under conditions similar to those used to detect effects on agonist-induced internalization (Fig. 4). Model 2 assumes that Ser 635 and Ser 639 are involved in the interaction of the rLHR with arrestin, and Ser 649 is involved in the interaction of the rLHR with another protein (protein X). This model also assumes that the agonist-induced uncoupling of the rLHR is mediated by arrestin, and the agonist-induced internalization of the rLHR is mediated by arrestin and protein X. The attenuation of agonist-induced uncoupling detected in rLHR-S635A and rLHR-S639A and the lack of effect of rLHR-S649A (Table II) can thus be explained by a reduced interaction of rLHR-S635A and rLHR-S639A with arrestin and a normal interaction of rLHR-S649A with arrestin. Conversely, the strong inhibition of internalization detected in rLHR-S649A and the partial effect of rLHR-S635A and rLHR-S639A ( Fig. 3 and Table III) can be explained by a reduced interaction with protein X (rLHR-S649A) and a reduced interaction with arrestin (rLHR-S635A and rLHR-S639A). In this model, arrestin-3 overexpression does not enhance the internalization of rLHR-S649A to the same level as that detected with rLHR-wt or rLHR-S639A (Fig. 4), because the S649A mutation is predicted to reduce the binding of rLHR to protein X, and the interaction of the rLHR with both protein X and with arrestin is needed for internalization. The inhibitory effect of dominant-negative ␤-arrestins on the internalization of rLHR-wt, rLHR-S639A, and rLHR-S649A (Fig. 4) can also be accommodated by the assumption that the rLHR-arrestin interaction is partially responsible for internalization. What Model 2 cannot explain, however, is the lack of effect of arrestin-3 and a dominantnegative ␤-arrestin on agonist-induced cAMP accumulation (Table IV). If this paradigm measures uncoupling (see above), then Model 2 predicts an inhibitory effect of arrestin-3 and a stimulatory effect of the dominant-negative ␤-arrestin on hCGinduced cAMP accumulation.
Experiments are now being planned to directly measure the interaction of the rLHR (and mutants thereof) with arrestin and to search for other proteins that may interact with the C-terminal tail of the rLHR and affect its internalization. We will not be able to differentiate between the two models described above until one or both of these strategies are successful. Regardless of which of the two models proposed above is correct, it is clear that the agonist-induced uncoupling and internalization of the rLHR are affected by mutations of differ-ent serine residues. This finding mirrors the results presented in a recent publication on the m2 muscarinic receptor, in which it was concluded that the agonist-induced uncoupling and internalization of this GPCR are mediated by the distinct Ser/Thr clusters present in the third intracellular loop (42). Although the study on the m2 receptor did not define the importance of individual Ser/Thr residues, it clearly showed that the phosphorylation of only one of the two clusters promotes uncoupling, whereas the phosphorylation of both clusters is needed to promote agonist-induced internalization. Thus, the data from these two studies, conducted with different GPCRs, lead to basically the same conclusion, that the agonist-induced uncoupling and internalization are mediated by distinct Ser/Thr residues.