INTRODUCTION

Desensitization is defined as loss of a biological response to a ligand despite its continuous presence. This phenomenon of desensitization is observed with almost every receptor, and some of the mechanisms involved in desensitization of G protein-coupled receptors have been well characterized using the -adrenergic receptor (AR)¹ and rhodopsin. For these receptors, the primary mechanism of desensitization appears to be receptor phosphorylation mediated by G protein-coupled receptor kinases (GRKs). Since these kinases have the unique feature of phosphorylating receptors only when they are in an active or a ligand-bound conformation, GRK-mediated phosphorylation is considered to be the main mechanism for agonist-induced desensitization. These kinases have been shown to phosphorylate a number of G protein-coupled receptors including the AR (1, 2), rhodopsin (3, 4), ₂-adrenergic receptor (₂AR) (5, 6), platelet-activating factor receptor (7), thrombin receptor (8), and C5a anaphylatoxin receptor (9).

Endogenous opioids play an important role in a variety of physiological processes through pharmacologically distinct, three major receptor subtypes that are coupled to inhibitory G proteins (10). Although the acute action of opioids can induce a number of beneficial effects, chronic use of opioids produces tolerance and dependence (11, 12), which are among the major factors limiting the clinical use of opioids. The molecular mechanisms underlying these phenomena are poorly understood, and receptor desensitization has been implicated as a possible mechanism.

It has been shown recently that opioid receptor desensitization is directly related to receptor phosphorylation. Protein kinase C (PKC) is involved in the functional uncoupling of the delta opioid receptor (DOR) from G protein in the striatum of guinea pigs (13). However, cellular depletion of PKC failed to alter DOR phosphorylation in HEK 293 cells (14), indicating a role for other kinases as well. In the case of the µ opioid receptor (MOR), it was reported that activation of PKC enhanced desensitization of the receptor, leading to a decrease in the receptor-activated potassium channel current (GIRK1) in Xenopus oocytes (15, 16). However, the PKC inhibitor staurosporine showed differential effects on the PKC-mediated response and the agonist-induced response. Staurosporine blocked PKC-activated MOR phosphorylation and eliminated PKC influence on desensitization, but failed to block desensitization induced by repeated administration of agonist (16). This result suggests participation of other kinases in agonist-mediated desensitization. Indeed, a recent study of agonist-induced phosphorylation of DOR suggests involvement of one or more GRKs in receptor desensitization (14). It has also been shown that overexpression of a functionally impaired mutant form of GRK2 blocks  receptor desensitization in COS-7 cells (17). In contrast to PKC manipulation, agonist-dependent phosphorylation of DOR in HEK 293 cells was altered by overexpression of GRK2, GRK 5, or a functionally impaired mutant form of GRK2 (14). These data suggest that GRKs are more directly involved in receptor desensitization and function together with PKC-mediated mechanisms.

In a previous study, we reported that 1-h pretreatment of MOR with the agonist DAMGO caused agonist-specific desensitization (18). The mechanisms underlying agonist-induced desensitization could include receptor phosphorylation, similar to that described for other G protein-coupled receptors. Recently, a splice variant of the MOR cDNA has been cloned from rat that differs at the 3-end (19). This variant (MOR1B) shares 100% amino acid sequence identity with MOR1 up to amino acid 386 but differs from residue 387 to the carboxyl tail (rMOR1, ³⁸⁷LENLEAETAPLP³⁹⁸; rMOR1B, ³⁸⁷KIDLF³⁹¹). Within that region of the carboxyl tail, one potential site for phosphorylation by GRKs is found in MOR that is absent in MOR1B. Interestingly, this shorter variant, rMOR1B, has been shown to be resistant to agonist-induced desensitization, unlike the longer form, rMOR1 (19). The cytoplasmic tail of the MOR contains potential phosphorylation sites for GRKs which may explain the differences between the variants and has led us to map the amino acid residues important in the functional desensitization of MOR.

EXPERIMENTAL PROCEDURES

The full-length cDNA for the rat MOR was cloned into the mammalian expression vector pRC/CMV (InVitrogen, San Diego, CA). This construct was used as a template for mutagenesis and for subsequent stable transfection of the wild type and mutant receptor cDNAs into Chinese hamster ovary (CHO) cell lines. Mutant MORs such as T394A, T/S363-383A (AT), T/S363-379A (ATT), E388-393Q (EQ) (Fig. 1) were constructed by substituting single or multiple amino acid residues with alanine(s). In case of the mutant EQ, glutamic acids between residues 388 and 393 were substituted with glutamine. Site-directed mutagenesis was performed using a polymerase chain reaction-based technique as described previously (20). Briefly, oligonucleotide primers corresponding to unique Eco47III and ApaI restriction sites (located at amino acids 304-305 and at vector, respectively) were utilized in combination with two mutagenic primers. Polymerase chain reaction fragments were digested with unique restriction enzymes and subcloned into the corresponding restriction sites of pRC/CMV encoding a MOR. All mutations were verified by dideoxy sequencing. For stable expression, cell line CHO K-1 (no. CCL61; American Type Culture Collection) was grown to 60% confluency in a 100-mm dish and subsequently transfected with wild type or mutant constructs in pRC/CMV vector using a calcium phosphate transfection kit (Life Technologies, Inc.) according to the manufacturer's recommendations. Stable transfectants were selected in 1 mg/ml geneticin (Life Technologies, Burlington, Ontario, Canada) and clones with the appropriate expression level were screened by radioligand binding assay. 30-90 clones expressing varying numbers of receptors were screened to select those with comparable expression levels.

Cells were grown until apparent confluency and then washed twice with 12 ml of ice-cold phosphate-buffered saline, harvested and centrifuged at 100 × g for 10 min. Cells were then lysed in hypotonic buffer (5 mM Tris-HCl, pH 7.8, 2 mM EDTA, containing a protease inhibitor mixture (10 mg/ml leupeptin, 5 mg/ml soybean trypsin inhibitor and 5 mg/ml benzamidine)) with a Polytron homogenizer (Brinkman Instruments, Westbury, NY) for two 30-s bursts at the 5.5 setting. The lysate was then centrifuged at 100 × g for 10 min to pellet unbroken cells and nuclei. The supernatant was collected and centrifuged at 30,000 × g for 30 min, and the resulting pellet was resuspended in buffer containing 50 mM Tris-HCl, pH 7.8, 5 mM MgCl₂, 1 mM EGTA with the protease inhibitor mixture and used immediately for radioligand binding studies. Protein content was determined according to the method of Bradford (21). For saturation experiments, cell membranes (20-30 µg of protein/tube) were incubated with increasing concentrations of [³H]naloxone (specific activity, 58.5 Ci/mmol; DuPont NEN) in a total volume of 1 ml. Specific binding was determined by calculating the difference in binding of the radiolabeled ligand in the absence and presence of the µ-selective antagonist naltrexone (10 µM). Competition experiments were performed using [³H]naloxone at approximately its K_d value and competing cold drug, DAMGO at concentrations ranging from 10¹¹ to 10⁴ M. After 2 h of incubation at room temperature, bound ligand was isolated by rapid filtration through a 48-well cell harvester (Brandel, Montreal, Quebec, Canada) onto GF/C Whatman filters. Filters were washed with 10 ml of ice-cold 50 mM Tris-HCl, pH 7.4, and placed in glass vials with 5 ml of scintillation fluid (Cytoscint; ICN, Costa, Mesa, CA) and counted for tritium. All experiments were performed in duplicate, and each experiment was repeated at least three times.

CHO cells expressing wild type or mutant MORs at confluent monolayer were treated with either medium alone or 1 µM DAMGO (final concentration) and incubated for 1 h at 37 °C in 5% CO₂. Incubation was terminated by washing plates three times with 12 ml of ice-cold phosphate-buffered saline. Membranes were prepared as described above. Adenylyl cyclase assays were conducted essentially as described previously (22). The assay mix contained 20 µl of membrane suspension (15-20 µg of protein), 12 µM ATP, 0.1 mM cAMP, 53 µM GTP, 2.7 mM phosphoenolpyruvate, 0.2 unit of pyruvate kinase, 1 unit of myokinase, and 0.13 µCi of [³²P]ATP in a final volume of 50 µl. Enzyme activities were determined in triplicate assay tubes containing increasing concentrations (10¹¹ to 10³ M) of DAMGO with 10 µM forskolin. Adenylyl cyclase activity was also measured in the presence of buffer alone (basal) or forskolin alone and reactions were incubated at 27 °C for 20 min. Reactions were stopped by the addition of 1 ml of an ice-cold solution containing 0.4 mM ATP, 0.3 mM cAMP and [³H]cAMP (25,000 cpm). [³²P]cAMP and [³H]cAMP were isolated by sequential column chromatography using Dowex cation exchange and aluminum oxide columns. The amount of [³H]cAMP was used to quantify individual column recovery. Data were analyzed by nonlinear least-squares regression.

RESULTS

We previously reported that the MOR was desensitized after 1 h treatment with an agonist, DAMGO 1 µM (18). During agonist exposure, basal adenylyl cyclase activity and forskolin-stimulated adenylyl cyclase activity did not change, suggesting this desensitization was specifically agonist-induced, with no evidence of heterologous desensitization. The underlying mechanism of this agonist-induced form of desensitization is likely due to the involvement of GRKs. Since phosphorylation of receptors plays an important role in receptor desensitization, and the carboxyl tail region of MOR contains a potential phosphorylation site for GRKs (Thr³⁹⁴), Thr³⁹⁴ was substituted by alanine in the longer variant of the receptor to test the involvement of this residue in agonist-induced desensitization of MOR.

Initially, stable CHO cell lines expressing wild type or mutated receptor in which the Thr³⁹⁴ was changed to alanine (T394A) were prepared (Fig. 1). The binding of antagonist [³H]naloxone to the wild type or T394A mutant receptor was saturable and of high affinity. The mu agonist, DAMGO detected two affinity states of the wild type and mutant receptors. With each of the receptors, 70% of the total receptor population was in a high affinity state for agonist with a dissociation constant of 3 nM, and 30% of the receptor population was in a low affinity state with a dissociation constant of 330 nM. To determine the functional coupling of the wild type and mutant receptors (T394A) to adenylyl cyclase, dose-response curves for the inhibition of forskolin-stimulated cAMP accumulation with increasing concentrations of DAMGO were tested. The IC₅₀ values for adenylyl cyclase inhibition for both receptors in CHO cells were identical; 8.5 ± 2.6 nM (n = 7) for wild type receptors and 5.6 ± 1.2 nM (n = 4) for mutant (T394A) receptors (Fig. 2, A and B). However, the maximum percentage inhibition of adenylyl cyclase activities was slightly different, but statistically insignificant (p > 0.05), suggesting that mutation of the Thr³⁹⁴ did not affect the G protein-coupling property or the functional coupling of receptors to the effector, adenylyl cyclase. In wild type receptor pretreated with 1 µM DAMGO for 1 h, agonist-dependent inhibition of forskolin-stimulated cAMP accumulation was decreased to 60% of the maximum inhibition occurring in untreated cells and IC₅₀ shifted to 200 ± 13 nM (Fig. 2A). Strikingly, in the mutant receptor (T394A), agonist-induced desensitization of receptors was completely abolished, as indicated by the lack of rightward shift of the IC₅₀ and a lack of reduction in the maximum inhibition of forskolin-stimulated cAMP accumulation. The IC₅₀ for wild type receptor after 1 µM DAMGO treatment for 1 h was 200 ± 13 nM (n = 7) and IC₅₀ for the T394A mutant was 5.3 ± 0.8 nM (n = 4) (Fig. 2, A and B). Thus, the Thr³⁹⁴ at the carboxyl tail of the receptor is critical for agonist-induced desensitization occurring within 1 h.

11

3

However, the role of 7 other potential phosphorylation sites within the carboxyl tail of MOR (2 serines and 5 threonines between residues 363 and 383) remained of interest. To test whether the Thr³⁹⁴ was sufficient to mediate desensitization, a mutant was prepared where all 7 potential phosphorylation sites (T/S363-383A) were replaced by alanines except for the Thr³⁹⁴ which was left intact (AT mutant) (Fig. 1). This mutant (AT) was essentially identical to wild type receptor in expression level, agonist affinity, G protein-coupling properties, and functional coupling to adenylyl cyclase (IC₅₀ = 8.6 ± 4.1 nM, n = 8) (Table I). The ability of this mutant receptor (AT) to inhibit forskolin-stimulated cAMP accumulation after 1 h treatment with 1 µM DAMGO (Fig. 2C) was blunted, with maximal inhibition only 70% of that observed with wild type receptor; IC₅₀ for wild type receptor after 1 h of agonist treatment was 200 ± 13 nM and that for AT mutant receptor was 92 ± 28 nM (n = 8) (Fig. 2, C and D). These data indicate there may be other residues in addition to the Thr³⁹⁴ required for full desensitization (Fig. 2D). To identify the other site(s), another mutant receptor was created in which 6 potential phosphorylation sites, 2 serines and 4 threonines, between residues 363 and 379 were substituted by alanines but the last 2 threonines at residues 383 and 394 were left intact (ATT) (Fig. 1). This mutant was also indistinguishable from wild type receptor with respect to expression level, agonist binding affinity and G protein-coupling properties (Table I). The dose-response curve for this mutant receptor (ATT) to inhibit forskolin-stimulated cAMP accumulation by DAMGO (25 ± 9 nM, n = 4) was also not different from wild type receptor (p > 0.05) (Fig. 3A). Furthermore, the maximal inhibition induced by DAMGO by interaction with this mutant receptor (ATT) was identical with that for wild type receptors, 52 and 54%, respectively. One hour of pretreatment with 1 µM DAMGO of cells expressing mutant receptors (ATT) showed an identical degree of desensitization compared with wild type receptors; IC₅₀ for wild type receptor after agonist treatment was 200 ± 13 nM (n = 7) and IC₅₀ for the mutant receptor was 190 ± 12 nM (n = 4). The maximal inhibition of adenylyl cyclase activity by the mutant receptor completely overlapped with that by wild type receptors (Fig. 3B). Taken together, results from these three mutants (T394A, AT and ATT) suggest the Thr³⁹⁴ is the primary recognition site by GRKs, but another threonine at residue 383 that is closest located, is also required for complete desensitization of the MOR.

Table I. Binding parameters for the stably transfected wild type or mutant µ opioid receptors KH and KL, Ki are values of the high (H) and low (L) affinity states; RH and RL, proportion of receptor in high (H) and low (L) affinity states. Kd and Bmax values of naloxone binding were tested with increasing six concentration of [3H]naloxone. Ki and proportion of high and low affinity states were determined by competition assay, by incubation with 2 nM [3H]naloxone and increasing concentration of DAMGO. The binding data from saturation and competition experiments were fitted by nonlinear least-squares regression using the data analysis program LIGAND. Data shown are the mean ± S.E. of three independent experiments. [3H]Naloxone [2H]Naloxone/DAMGO KH KL RH RL nM % Wild type Bmax 3.8  ± 0.9 pmol/mg 2.9  ± 0.7 326  ± 25 69  ± 2 31  ± 2 Kd 0.9  ± 0.2 nM T394A Bmax 3.5  ± 0.5 pmol/mg 2.7  ± 1.4 329  ± 14 74  ± 2 26  ± 2 Kd 1.0  ± 0.3 nM T/S363-383A (AT) Bmax 4.3  ± 0.3 pmol/mg 2.2  ± 1.7 370  ± 23 70  ± 1 30  ± 1 Kd 1.0  ± 0.1 nM T/S363-379A (ATT) Bmax 4.0  ± 0.1 pmol/mg 3.2  ± 0.2 383  ± 97 71  ± 2 29  ± 2 Kd 1.0  ± 0.1 nM E388-393Q (EQ) Bmax 2.0  ± 0.1 pmol/mg 2.9  ± 1.8 431  ± 215 73  ± 2 27  ± 2 Kd 2.5  ± 0.9 nM

Analysis of phosphorylation of a synthetic peptide based on ₂AR or ₂AR sequence showed that negatively charged residues are required for a peptide to serve as a substrate for GRK2 (23). Support for the idea that the Thr³⁹⁴ serves as the primary recognition site for GRKs, compared with the other potential sites in the carboxyl tail, resulted from the stretch of acidic amino acids in front of it, a mutant receptor was created in which glutamic acid residues at 388, 391, and 393 were replaced by glutamines and Thr³⁹⁴ was left intact (EQ). This mutant (EQ) was expressed at low densities and naloxone binding affinity was marginally higher (Table I). However, K_i values for DAMGO at high and low receptor affinity states and their proportions were comparable to those of wild type receptors, indicating no changes in G-protein coupling ability. As shown in Fig. 3C, agonist-induced desensitization was completely abolished in the EQ mutant. Thus, the acidic amino acid residues preceding the Thr³⁹⁴ are absolutely necessary for Thr³⁹⁴ to play the critical role in desensitization, possibly by facilitating phosphorylation of Thr³⁹⁴ by a GRK.

DISCUSSION

Our results demonstrate that the Thr³⁹⁴ and preceding acidic amino acids at the carboxyl tail of the MOR are the key residues required for agonist-induced receptor desensitization. Agonist-induced receptor desensitization has been proposed to be due to phosphorylation of the agonist-bound form of receptors by GRKs (reviewed in Inglese et al. (24) and Lefkowitz (25)). Furthermore, the recognition sites for these kinases have been identified as repeated serines and threonines at the carboxyl tail by phosphopeptide sequencing of high pressure liquid chromatography-purified peptides derived from proteolysis of phosphorylated receptors (1, 26). Since there are usually several repeated serine/threonine residues in the carboxyl tails of G protein-coupled receptors, it is of interest to determine whether all of these residues are required to be phosphorylated to attenuate prolonged agonist-induced stimulation.

Numerous studies have attempted to answer this question by using truncation mutations of the carboxyl tail of various lengths, to interfere with both receptor phosphorylation and agonist-induced desensitization (1-8). These studies demonstrated that phosphorylation of particular residues are necessary for desensitization. These results raise the question of how GRKs distinguish certain serine or threonine residues from others. Analysis of the phosphorylation of several synthetic peptides based on ₂AR sequence identified two peptides that were phosphorylated by GRK2. Interestingly, the feature common to both substrate peptides was the presence of acidic amino acids, mostly glutamic acids, in close proximity to serine and threonine residues (1, 6, 23, 27-31). However, peptides with glutamic acid residues on the carboxyl-terminal side of the serine or threonine were not substrates for GRK2 (23). A peptide with glutamic acid residues on both sides of the serine residue was also a poor substrate. Furthermore, a peptide that was a good GRK2 substrate became poorly phosphorylated when acidic amino acids at the amino-terminal side of the phosphorylation sites were changed to their neutral counterparts (6, 23). This strongly suggests that amino-terminal acidic amino acids relative to the phosphate acceptor group is a requirement for GRK2 recognition. In contrast to GRK2, rhodopsin kinase (GRK1) recognizes a serine or threonine residue located on the carboxyl-terminal side of basic amino acids (1, 23). This seems to indicate that even though there are several GRKs in this family of kinases, they each have different substrate specificity together with different tissue distributions (24, 25).

Sequential phosphorylation of different residues has been demonstrated in the case of rhodopsin, in which light-dependent phosphorylation occurs first at the serine in position 338, and subsequently at serine 343 and then threonine 336 (4, 32-34). In the case of the human C5a anaphylatoxin receptor, phosphoamino acid analysis of C5a-dependent phosphorylation showed the maximal stoichiometry to be 6 moles of phosphate/mole of receptor. However, mutation of 2 particular serines at positions 332 and 338 completely abolished agonist-dependent phosphorylation, further suggesting multiple phosphorylations occurring in a sequential manner (9). However, in the case of ₂AR, there are 4 consecutive serines in the third intracellular loop and substitution of each serine decreased overall phosphorylation by 25% when compared with that of wild type ₂AR (6). Since both C5a anaphylatoxin receptor and ₂AR are substrates for the same kinase, GRK2, it is unlikely that these two different patterns of phosphorylation are due to different substrate and kinase interactions. It remains to be determined if the nonsequential nature of ₂AR phosphorylation reflected the close proximity of each serine residue or its third loop location. However, in our study, substitution of the Thr³⁹⁴ completely abolished agonist-induced desensitization, whereas for achieving the full agonist-induced desensitization, both Thr³⁸³ and Thr³⁹⁴ appear necessary in the carboxyl terminus. This would be best explained by a sequential phosphorylation model, in which Thr³⁸³ can be phosphorylated only after Thr³⁹⁴ is already phosphorylated. Since it has been reported that GRK2 is the most predominant of the GRK isoforms expressed in CHO cells, as evidenced by Western blotting experiments with antibodies specific for GRK2 and GRK3 (35), we postulate that these two threonine residues are the likely targets for GRK2 action, leading to receptor phosphorylation and desensitization. It remains, however, to be shown that phosphorylation does occur at these threonine residues specifically.

Since there are many repeated serines and threonines in the carboxyl tail of the MOR, it is important to know whether all of these residues actually need to be phosphorylated for receptor desensitization. Furthermore, another unknown factor is whether the sites identified in the cell culture study represent the sites of phosphorylation in endogenous tissues under physiological conditions. Although the stoichiometry of maximum receptor phosphorylation using biochemical experiments has been calculated for many receptors, the actual stoichiometry in vivo in living tissues has not been well studied yet. For instance, cell culture studies have suggested that multiphosphorylation of rhodopsin occurs at as many as 9 residues, but an in vivo study using retinas collected from mice demonstrated that only a single phosphate group was incorporated into the mouse rhodopsin molecule in a light-dependent manner, at serine 338 after light flashes and at serine 334 after continuous illumination (4, 36), both of which were found to be major sites for in vitro phosphorylation. Our data also shows that among the 8 serines or threonines in the carboxyl tail that we have examined, only two threonine residues at 383 and 394 are necessary for agonist-induced desensitization.

Another interesting feature of GRK-mediated receptor phosphorylation is that the time course of this event seems to be regulated by the primary recognition residue, usually serine in the carboxyl tail, because the most prominent and rapid desensitization seems to occur in parallel with phosphorylation of the primary recognition site. In the case of the serotonin type 2 receptor (5-HT2R) and the thrombin receptor, each has a similar carboxyl tail structure but a very different time course of functional desensitization, even though they interact with the same effector, phospholipase C. Thrombin receptor desensitization occurs rapidly, but 5-HT2R desensitization is slower. However, a chimeric receptor having the core of the 5-HT2R with the carboxyl-tail of thrombin receptor undergoes rapid desensitization, suggesting that the structure of the carboxyl-tail may determine the time course of desensitization (8). Given that GRKs have a preference for serine residues over threonine (23), and since the thrombin receptor has only serine residues clustered in an acidic environment in the carboxyl tail, it might be interesting to speculate that the slower time course of desensitization of the 5-HT2R might be due to its primary phosphorylation site being threonine. This speculation is very tempting to apply to our data, since the time course of desensitization of the MOR is slower than GRK-mediated desensitization of other receptors, and the primary recognition site for MOR desensitization as we have shown, appears to rely on two threonine residues in the carboxyl tail.

In summary, we have established that both Thr³⁹⁴ and a cluster of glutamic acids at residues 388, 391, and 393 preceding Thr³⁹⁴ in the carboxyl-tail of the MOR are absolute requirements for agonist-induced MOR desensitization. However, Thr³⁸³, the closest threonine to Thr³⁹⁴, plays an additional role in this desensitization if Thr³⁹⁴ is left unchanged. Since Thr³⁹⁴ with a stretch of acidic amino acid residues at its amino terminus side serves as an ideal sequence for GRK2 action, we propose the following model. Thr³⁹⁴ is the primary site that is targeted upon agonist binding to the MOR, with recognition of this putative phosphorylation site being dependent on the acidic residues preceding it. This phosphorylation event by itself can lead to submaximal (70%) receptor desensitization. Subsequent sequential phosphorylation of Thr³⁸³, after Thr³⁹⁴ phosphorylation, would effect the complete desensitization of the MOR in response to agonist.