Receptor-induced Internalization of Selective Peptidic μ and δ Opioid Ligands

The binding and internalization of radioiodinated and fluorescent μ and δ opioid peptides in mammalian cells were quantitatively studied by biochemical techniques and directly visualized by confocal microscopy. The labeled peptides were prepared by inserting either a 125I-Bolton-Hunter group or a fluorescent probe into the C-terminal part of 5-aminopentylamide derivatives of deltorphin-I and [Lys7]dermorphin. The purified derivatives kept most of their specificity and selectivity toward δ and μ opioid receptors, respectively. Biochemical and confocal microscopy data showed that both μ and δ opioid peptides were internalized in mammalian cells transfected with the corresponding opioid receptor according to a receptor-mediated mechanism. The internalization process was time- and temperature-dependent and was completely blocked by the endocytosis inhibitor phenylarsine oxyde. Internalization of both δ and μ ligands occurred from a single large cap at one pole of the cell, indicating that polymerization of ligand-receptor complexes preceeded internalization. Finally, green and red fluorescent analogues of deltorphin-I and [Lys7]dermorphin, respectively, were found to internalize through partly distinct endocytic pathways in cells co-transfected with μ and δ receptors, suggesting that each of these receptors interacts with distinct proteins mediating intracellular sorting and trafficking.

of internalization of the tritiated enkephalin agonist [ 3 H]D-Ala, D-Leu-enkephalin in cultured neuroblastoma cells, morphological studies clearly failed to observe internalization of a fluorescent derivative of enkephalin in the same cell system (10 -12). More recently, confocal microscopic studies carried out on transfected cells have shown a rapid endocytosis of (13,14) and ␦ (13) antigenic epitope-tagged receptors following exposure to enkephalins, but not to morphine. Whether this endocytosis occurs in conjunction with that of the bound ligand, however, remains unclear. In the present study, we have reinvestigated the fate of receptor-bound opioid peptides using newly developed radioactive and fluorescent derivatives of the selective and ␦ opioid agonists dermorphin and deltorphin.
Dermorphin, isolated from the skin of the frog Phylomedusa sauvagei (15), was the first natural peptide described as having high affinity and selectivity for the opioid receptor. Another unusual property of dermorphin is the D configuration of the alanine residue in position 2, which is responsible for its strong resistance to enzymatic degradation and for its good opioid binding site recognition. The most interesting analogue of this peptide is [Lys 7 ]dermorphin (16), which possesses the highest affinity for the opioid receptor (less than 10 Ϫ10 M) together with the highest /␦ specificity (10,000-fold more specific for the than for the ␦ receptor). Two other peptides having also a D-alanine in position 2 have been purified from the skin of another frog, Phylomedusa bicolor (17). They have the same first three amino acids (Tyr-D-Ala-Phe) but differ in their C-terminal sequence (Asp-Val-Val-Gly for deltorphin-I and Glu-Val-Val-Gly for deltorphin-II). Deltorphin-I was chosen for our study because of its superior affinity and specificity for the ␦ opioid receptor.
Sequences of [Lys 7 ]dermorphin and deltorphin-I may be divided into two parts, the same first three amino acids (Tyr-D-Ala-Phe) confirming something like an opioid master key and the last four being responsible for the /␦ selectivity and affinity. The only site that could be modified without changing the properties of these peptides is the C-terminal end. We have thus incorporated radiolabeled and fluorescent groups into Cterminal extensions of deltorphin-I and dermorphin and used these specific tools to compare the cellular distribution and fate of specifically labeled and ␦ opioid receptors. Radioactive compounds were used to quantitatively assess the binding and internalization of the opioid derivatives, while fluorescent compounds were used to study their distribution in the confocal microscope, an approach that has been successfully resorted to for a variety of other neuropeptides, including fluorescent analogues of cholecystokinin (18), gastrin-releasing peptide (19), neurotensin (20), thyrotropin-releasing hormone (21,22), substance P (23,24), and somatostatin (25).

Preparation of Fluorescent Peptides
Peptide precursors were reacted with the N-hydroxysuccinimide esters of Bodipy 503/512 or Bodipy 576/589 (Molecular Probes). NHS-Bodipy (2 mol in 400 l of dimethyl sulfoxide) was incubated with DLT-I 5APA or [K7]DRM 5APA (2 mol) in a final volume of 1 ml of boric acid (50 mM), sodium phosphate (50 mM) buffer, pH 8.5, for 3 h at 4°C. The different derivatives were purified by reverse phase HPLC on a C18 Ultrosphere ODS column (10 x 250 mm, Beckmann) eluted in 0.1% trifluoroacetic acid with a linear gradient of acetonitrile from 20 to 60% during 60 min. Fluorescent peaks were tested for their ability to displace the specific binding of ⑀-BH * [K7]DRM 5APA and -BH * DLT-I 5APA to the and ␦ opioid receptors, respectively (27). Seven fluorescent peaks were found to have a high binding activity and were further characterized by Edman degradation.

Binding of Fluorescent Peptides to COS-7 Cell Membranes
Binding properties of fluorescent derivatives were determined by displacement of -BH * DLT-I 5APA and ⑀-BH * [K7]DRM 5APA-specific binding to membranes of cells transfected with pcDNAI-DOR and pcDNAI-MOR, respectively. Forty-eight hours after transfection, cells were rinsed with PBS Ϫ , scraped in Tris/EDTA (5 mM/5 mM, pH 7.5) and centrifuged at 100,000 ϫ g. Radioactive ligands (0.2 nM) were incubated with transfected cell membranes (5-20 g of membrane proteins) and increasing concentrations of fluorescent peptides in 250 l of 0.2% bovine serum albumin, 50 mM Tris-Cl, pH 7.5, during 30 min at 25°C as described previously (27). Incubations were stopped by filtration on GF/C filter presoaked in 0.3% polyethyleneimine-Cl, pH 7.5. Filters were rinsed three times with 3 ml of binding buffer and counted in a ␥ counter. IC 50 values were determined by measuring the concentration of fluorescent peptide that displaced 50% of the bound radioactive ligand.

Internalization Studies
Cell Preparation-Forty-eight hours after transfection, cells were rinsed in PBS, detached with a PBS solution containing 0.05% trypsin and 0.53 mM EDTA, and equilibrated with 10% fetal calf serum in Dulbecco's modified Eagle's medium during 2 h at 37°C. They were then centrifuged at 1000 rpm during 5 min and equilibrated for 15 min at 37°C in binding buffer (Earle-HEPES buffer, pH 7.4, supplemented with 0.09% glucose and 0.2% of bovine serum albumin) at a final concentration of 50 -200,000 cell/ml.
Radioactive Binding Experiments-Cells transfected with pcDNAI-DOR and pcDNAI-MOR were incubated for 0 -60 min at 37°C with -BH * DLT-I 5APA and ⑀-BH * [K7]DRM 5APA (0.1-10 nM, 2000 Ci/mmol), respectively, in the presence or absence of 10 M of the endocytosis inhibitor, phenylarsine oxide. The incubation was terminated by adding 3 ml of hypertonic acid buffer (Earle-HEPES, pH 4, acetic acid, 0.4 M NaCl) or of control buffer (Earle-HEPES at neutral pH) for 2 min, after which the cells were filtered on GF/C filters presoaked in binding buffer and rinsed three times with 3 ml of binding buffer. Cell-bound ligand was determined by counting the radioactivity retained on filters with a ␥ counter. Nonspecific binding was determined by carrying the incubation in the presence of 10 M naloxone. Temperature sensitivity was verified by incubating additional cells for 60 min at 0°C with 0.2 nM radioactive ligand. Stability of the ligands was determined by reverse phase HPLC analysis of a fraction of ligand recovered at the end of the incubation. Fractions collected after HPLC were counted and compared with initial solutions.
Fluorescent Binding Experiments-Cells transfected with pcDNAI-DOR and pcDNAI-MOR were preincubated as described above and incubated for 15-90 min at 37°C in binding buffer containing 10 nM -Bodipy 503/512 DLT-1 5APA or -Bodipy 503/512 [K7]DRM 5APA in the presence (nonspecific binding) or absence (total binding) of 10 M naloxone. In some experiments, the incubation was carried out in the presence of phenylarsine oxide, to prevent ligand endocytosis. At the end of the incubation, cells were washed with either hypertonic acid or isotonic neutral buffers as above, centrifuged at 2000 rpm during 1 min, deposited in 10 l of Earle-HEPES on glass microscope slides, air-dried, and examined by confocal microscopy.
Concomitant labeling of ␦ and opioid receptors was carried out on COS cells co-transfected with the pcDNAI-DOR and pcDNAI-MOR plasmids. Co-transfected cells were incubated at 37°C for 90 min with a mixture of 10 nM of -Bodipy 503/512 DLT-I 5APA (green) and -Bodipy 576/89 [K7] DRM 5APA (red). At the end of the incubation, cells were centrifuged at 2000 rpm, deposited on glass slides, and air-dried for confocal microscopic examination.

Confocal Microscopy
Labeled COS cells were examined under a Leica confocal laser scanning microscope configured with a Leica Diaplan inverted microscope equipped with an argon/krypton laser with an output power of 2-50 mV (Leica, St. Laurent, Canada). Images of cells were acquired as single midcellular optical sections and averaged over 32 scans/frame. For double labeling experiments, ␦ and ligand images were acquired in the green and red channels, respectively.

RESULTS
Biochemical Studies-The binding and internalization of the selective opioid agonists deltorphin-I and dermophin were first assessed quantitatively in COS-7 cells transfected with cDNA encoding ␦ and opioid receptors, using 125 I-labeled Bolton-Hunter derivatives of deltorphin-I (-BH * DLT-I 5APA) and dermorphin (⑀-BH * [K7]DRM5APA), respectively. These derivatives have been documented to selectively bind to ␦ and opioid receptors with respective K d values of 0.7 and 0.14 nM (27). Binding and internalization kinetics were established for each of these compounds by incubating whole cells at 37°C. As can be seen in Fig. 1, in which total specific binding (open symbols) corresponds to the sum of acid wash-resistant specific binding (corresponding to internalized ligand; closed symbols) and of acid-washable specific binding (corresponding to membrane-bound ligand; not shown), total binding kinetics were very rapid (t1 ⁄2 of about 1 min), whereas internalization kinetics were about 10 times slower (t1 ⁄2 ϭ 13 and 10 min for ␦ and receptors, respectively). The main difference between the two opioid systems was the maximal proportion of internalizable ligand, which represented approximately 55 and 25% of the total specific binding in cells transfected with ␦ and receptors, respectively. As shown in Fig. 2, this proportion was not very sensitive to ligand concentration as increasing the concentration of each radioligand by a factor of 100 (from 0.1 to 10 nM) enhanced maximal internalization by a factor of less than 2. Thus, independent of ligand concentration, the internalization process was approximately twice as efficient for the ␦ as for the ligand.
Internalization of both ␦ and ligands was almost totally inhibited by blocking endocytosis with phenylarsine oxide or by lowering the temperature of incubation, as reflected by the reduction in size of the acid wash-resistant fraction (Fig. 3). Somewhat surprisingly, the addition of phenylarsine oxide also markedly reduced the total binding of 125 I-labeled deltorphin to cells transfected with the ␦ receptor (Fig. 3A), but not that of 125 I-labeled dermorphin to cells transfected with the receptor (Fig. 3B). In both systems, lowering the temperature of incubation to 0°C resulted in large losses of total specific binding (35 and 55% of the binding observed at 37°C for ␦ and systems, respectively). Finally, neither binding nor internalization of either 125 I-deltorphin or 125 I-dermorphin were observed in COS-7 cells transfected with the noncorresponding opioid receptor.
Biochemical Characterization of Fluorescent Analogues-In order to visualize at the cellular level the interaction of ␦ and opioid ligands with their respective receptors, we have synthesized two different fluorescent derivatives of deltorphin-I and [Lys 7 ]dermorphin. These two heptapeptides are the most potent and selective ␦ and agonists currently known. The synthesis was carried out in two steps. First, an aminopentyl group was grafted on the C-terminal carboxyl function of dermorphin and deltorphin-I. Second, peptide precursors DLT-I 5APA and [K7]DRM 5APA were reacted with the N-hydroxysuccinimide esters of green (Bodipy 503/512) or red (Bodipy 576/589) fluorescent probes. The four different mixtures resulting from the reaction of the two peptide precursors with the two fluorescent reagents were fractionated by reverse phase HPLC. Elution profiles illustrated in Fig. 4 show that it was always possible to separate the unmodified peptide (N) from its fluorescent derivatives. Peaks numbered 1-6 in Fig. 4 were selected for further characterization on the basis of their ability to compete for binding to the ␦ (peaks 1 and 2) or to the (peaks 3-6) opioid receptor. Amino acid analysis and UV-visible spectra of each fraction (not shown) indicated that fluorescent peptides 1-6 contained a single fluorescent group per mol of peptide. The position of the fluorophore into each peptide sequence was determined by Edman degradation (Table I). The sequences of peaks 1 and 2 were identical to that of deltorphin-I, indicating that the green (peak 1) and red (peak 2) Bodipy fluorophores had been incorporated on the amine function of the 5-aminopentylamide group in DLT-I 5APA. In the same way, peaks 4 and 6 were identified as the red and green -substituted analogues of [K7]DRM 5APA, because their sequences were indistinguishable from that of [K7]DRM. By contrast, the seventh cycle of sequencing of fractions 3 and 5 did not give a PTH-Lysine, indicating that the ⑀-amino group of Lys 7 has been modified. Therefore, fractions 3 and 5 were identified as the green and red ⑀-labeled analogues of [K7]DRM 5APA.
The binding properties of the six fluorescent peptides were then evaluated by measuring their ability to displace the specific binding of 125 I-labeled analogues of DLT-I 5APA and [K7]DRM 5APA to the ␦ and opioid receptors transiently expressed in COS cells. The two fluorescent analogues of DLT-I 5APA interacted with the ␦ opioid receptor with high affinity (K 0.5 ϭ 2 nM) and retained much of their selectivity since their affinity for the receptor was lower by at least two orders of magnitude (Table II). The four fluorescent derivatives of [K7]DRM 5APA bound to the opioid receptor with affinities in the nanomolar range. However, their selectivity was not as good as that of ␦ ligands, since the ␦/ ratio of IC 50 values varied only from 4 to 21 (Table II). The fluorescent analogues 4 (-Bodipy 503/512[K7]DRM 5APA) and 6 (-Bodipy 576/ 589[K7]DRM 5APA) were selected for the confocal studies described below because of their higher selectivity for the receptor as compared with derivatives 3 and 5.
Confocal Microscopic Studies-Confocal microscopic examination of COS-7 cells transfected with a cDNA encoding the ␦ opioid receptor and incubated for 15-90 min with 10 nM -Bodipy 503/512 DLT-15APA revealed selective fluorescent labeling of approximately 30% of the cells, in keeping with the documented transfection yield of this cell line (32) (Fig. 5A). This labeling was specific in that it was no longer detected when the incubation was carried out in the presence of 10 M naloxone (Fig. 5D) or with cells transfected with an empty plasmid (Fig. 5C). It was selective for ␦ opioid receptor-expressing cells, since cells transfected with cDNA encoding the site were totally devoid of fluorescent labeling (Fig. 5B).
Similarly, approximately 30% of COS-7 cells transfected with a cDNA encoding the opioid receptor and incubated with the -selective ligand -Bodipy 503/512 [K7]DRM 5APA exhibited intense fluorescent labeling (Fig. 6A). Here again, this labeling was specific in that it was displaced by naloxone (Fig.  6D) and absent from cells transfected with an empty plasmid (Fig. 6C). It was also selective for receptor-transfected cells as it was no longer apparent in cells transfected with cDNA encoding the ␦ opioid receptor (Fig. 6B).
At all time intervals examined, the bulk of DLT-I 5APA fluorescent labeling was intracellular, as attested by its intracytoplasmic distribution in single optical sections passing through the core of the cells (Figs. 5A and 7A) and by its resistance to hypertonic acid wash (Fig. 7C). Internalization of the fluorescent ␦ agonist was totally prevented by addition of the endocytosis inhibitor, phenylarsine oxide, in which case the bound fluorescent molecules remained clustered at the periphery of the cells (Fig. 7B). This excentric labeling pattern corresponded to cell surface labeling as it completely disappeared after hypertonic acid wash (Fig. 7D).
Similarly, a sizeable fraction of -Bodipy 503/512 labeling of receptor-expressing cells was found to be acid wash-resistant, i.e. intracellular, at all times examined (Figs. 8, A and C). In keeping with our biochemical results, this fraction was smaller overall than in the case of ␦ labeling. As with the ␦ ligand, incubation in the presence of phenylarsine oxide prevented internalization of the opioid agonist and resulted in an acidwashable (Fig. 8D), cell surface clustering of the bound fluorescence (Fig. 8B).
The intracellular distribution of -Bodipy 503/512 DLT-I 5APA and -Bodipy 503/512 [K7]DRM 5APA, specifically bound to ␦ opioid-and opioid-transfected cells, respectively, varied markedly as a function of time. After 15 min of incubation, both fluorescent markers were clustered at one pole of the cell, onto and/or immediately beneath the plasma membrane (Fig. 9, A and B). At 30 min, they were detected in the form of small fluorescent particles excentrically clustered in the cytoplasm of the cells (Fig. 9, C and D). By 60 min, ␦ and opioid labeling remained highly punctate, but entirely filled the cytoplasm of the cells, sparing the nucleus (Fig. 9, E and F). However, by that time, the dermorphin-labeled fluorescent particles were on average larger and less numerous than the deltorphinlabeled ones and stood out less clearly against a greater intracytoplasmic background labeling.
In order to compare the distributional pattern of ␦ and   Fig. 10, A and B, and at higher magnification in Fig.  10, AЈ and BЈ, images acquired through distinct red and green excitation/emission channels showed labeling patterns comparable with those obtained in singly transfected cells confirming the efficiency of the co-transfection procedure. Superimposition of the two images (Figs. 10, C and CЈ) showed only partial overlap of the red and green fluorescent clusters, indicating that the and ␦ fluorescent ligands were sequestered partly in the same and partly in distinct compartments. This partial overlap could not be attributed to a bleed-through of one of the fluorophores into the other channel since Bodipy 576/589 and Bodipy 503/512 derivatives gave no signal in the green and red channels, respectively. Interestingly, the bulk of distinct green and red particles were small and predominated at the periphery of the cell (Fig. 10C). By contrast, double-labeled endosome-like particles (in yellow, Fig. 10C) were larger and mainly concentrated within the cytoplasmic core.

DISCUSSION
In the present study, the binding and internalization of and ␦ opioid peptides in mammalian cells were quantitatively studied by means of biochemical techniques and directly visualized by confocal microscopy. The radiolabeled and fluorescent analogues developed for this purpose were synthesized by inserting either an 125 I-labeled Bolton-Hunter group or a fluores-   A and B, total binding; C and D, residual binding after hypertonic acid wash. Confocal optical sections acquired through the nuclear plane. In A and C, the label is almost exclusively intracellular and concentrated within small endosome-like organelles (arrows). In B, the label is confined to the plasmalemma (arrowheads), as confirmed by its disappearance following hypertonic acid wash (D). Scale bar, 10 m. cent probe into the C-terminal part of 5-aminopentylamide derivatives of deltorphin-I and dermorphin. Elongation of both peptide sequences by a C-terminal extension bearing a primary amine proved essential for introducing reporter groups into both dermorphin-I and deltorphin without a dramatic loss in binding properties. Indeed, all other labeling approaches tried by us, including direct iodination of Tyr 1 or acylation of the N-terminal amine function with Bolton-Hunter or fluorescent groups, led to dermorphin and deltorphin-I derivatives of low binding activity toward and ␦ opioid receptors. Although both radiolabeled (27) and fluorescent (Table II) analogues synthesized in the present work had a somewhat lower affinity than their native counterpart, they retained sufficient biological activity and specificity for biochemical analysis and confocal imaging of and ␦ opioid receptors in transfected cells. The reduced affinity observed after radioactive or fluorescent tagging probably resulted from the steric hindrance of the markers, since insertion of a Bolton-Hunter group into theor ⑀-amine functions of deltorphin-I and dermorphin resulted in smaller losses of affinity than the incorporation of the more bulky Bodipy fluorophores. Replacement of the red or green Bodipy probes by larger fluorophores like fluorescein resulted in further decreases in affinity of the corresponding deltorphin-I and dermorphin analogues for their specific receptors (results not shown). Biochemical and confocal microscopic data showed that both and ␦ opioid ligands were internalized in mammalian cells according to a time-and temperature-dependent process. This internalization was clearly receptor-mediated, since it was no longer observed in nontransfected cells or in cells transfected with a receptor not specifically recognized by the ligand. Furthermore, it was completely prevented by incubation with the non selective opioid antagonist naloxone. The kinetics of internalization of and ␦ opioid ligands were considerably slower than those of ligand binding (t1 ⁄2 of 10 and 13 versus 1 min for and ␦ ligands, respectively), but within the same range as reported for bradikinin receptor (33) and vasopressin V2 receptor (34) complexes (t1 ⁄2 ϭ 9 and 13 min, respectively). Slightly faster receptor-mediated internalization kinetics have been described for peptides bound to vasopressin V1 (35,36), angiotensin II 1a and 1b (37), substance P (23), gastrin-releasing peptide (38), and neurotensin (39,40) receptors, with t1 ⁄2 values ranging between 3 and 5 min. The similarity between the internalization kinetics of these different neuropeptide-receptor complexes suggests that they may be controlled by a common rate-limiting step.
Despite the fact that formation of receptor-ligand complexes was clearly critical for the initiation of ligand internalization, the percentage of internalized ligand molecules was in no way proportional to the degree of receptor occupancy. Indeed, cells transfected with the ␦ receptor and incubated with concentrations of 125 I-labeled deltorphin-I ranging from 0.025 to 10 nM internalized between 40 and 70% of the bound radiolabeled peptide, whereas receptor occupancy varied from 3.5 to 94%. The disproportion was even greater in the case of the opioid ligand, which proportionally internalized about 2-fold less than its ␦ counterpart. Variable internalization capacities have been reported for other peptidergic systems. Ratios of internalized to bound ligand ranging between 50 and 80% have been reported for substance P (23), vasopressin (34,35), angiotensin II (37,41), or neurotensin (39, 40) and of 100% for gastrin-releasing peptide (38) and bradykinin (33). The low proportion (20 -35%) of 125 I-labeled dermorphin internalized by COS cells trans- In A and C, the labeling is mainly intracellular and confined to intracytoplasmic organelles (arrows). Note the absence of labeling over the nucleus (N). In B, the ligand forms a discontinuous ring at the periphery of the cell (arrowheads). This ring corresponds to cell surface labeling as attested by its disappearance after acid wash (D). Scale bar, 10 m. fected with the opioid receptor observed in the present work is rather unusual. The only other example that we are aware of is that of the sst1 somatostatin receptor expressed in COS cells, which internalizes only 20 -25% of specifically bound 125 I-labeled somatostatin 14 (25). The mechanism that maintains an equilibrium between internalized and noninternalized peptidereceptor complexes is unknown.
Internalization of both and ␦ opioid ligands in transfected COS-7 cells was totally prevented in the presence of the endocytosis inhibitor, phenylarsine oxide (PAO), indicating that the internalization process is endocytic in nature. Accordingly, internalized ligand molecules were seen by confocal microscopy to be concentrated within small, endosome-like organelles. Current models of receptor-mediated internalization call for internalization of receptor-ligand complexes into clathrincoated pits, followed by their mobilization into early and then late endosomes (42). These endosomes eventually fuse into multivesicular bodies and lysosomes, while dissociated receptors are either recycled back to the membrane or degraded (42). The progressive shift in size and intracellular mobilization of fluorescent organelles observed in the present study are consistent with this pattern. Although the present approach did not allow a direct visualization of receptors themselves, the similarity of the punctate labeling seen here after internalization of fluorescent dermorphin with that observed by immunohistochemistry in cells expressing an epitope-tagged opioid receptor (14) and exposed to the selective agonist Tyr-D-Ala-Gly-N-Me-Phe-Gly-ol strongly suggests that the internalization involves ligand-receptor complexes. A similar mechanism has been invoked to account for the internalization of tritiated D-Ala 2 , D-[Leu 6 ]enkephalin in neuroblastoma cells (5,8) and may play a role in the agonist-induced down-regulation of ␦ opioid receptors documented in several cell lines (6,7,43). Our results are at odds, however, with the reported lack of internalization of a rhodamine-tagged Met-enkephalin derivative in cultured neuroblastoma cells (10 -12). This discrepancy may be due to differences in receptor behavior dependent on the ligand utilized. Mu opioid receptors were indeed shown to internalize following exposure to Tyr-D-Ala-Gly-N-Me-Phe-Gly-ol, but not to morphine (14).
In addition to blocking receptor-mediated internalization, phenylarsine oxide markedly decreased total specific binding of the 125 I-labeled deltorphin-I analogue to cells transfected with the ␦ receptor without affecting that of the dermorphin analogue to sites. This decrease cannot be imputed to direct effects of the drug on the plasma membrane, as it was no longer observed when the binding experiments were performed on COS-7 cell membranes, as opposed to whole cells (not shown). It is therefore likely that the loss of specific deltorphin binding observed in whole cells is the result of cellular traffic blockade by PAO, implying that and ␦ opioid receptors are differentially distributed within the cells at steady state. Specifically, our results suggest that receptors are predominantly localized on the cell membrane, since they are freely accessible to radioligand even in the presence of PAO, whereas a significant fraction (65%) of ␦ receptors are located inside the cell, within vesicular structures that can no longer be recruited to the membrane in the presence of PAO. The ability of PAO to inhibit not only endocytosis but also exocytosis has been documented in various reports dealing with trafficking properties of glucose transporters (44) and transferrin receptors (45) or with secretory mechanisms of the RBL-2H3 mast cell line (46). Furthermore, such differential distribution of and ␦ receptors in COS cells would be consistent with the results of light and electron microscopic localization studies in the central nervous system, which have shown opioid receptors to be predominantly associated with the plasmalemma (14,47,48) and ␦ receptors to be almost equally distributed between the plasma membrane and intracellular stores (48 -52). Comparable differences in the subcellular distribution of two homologous G-protein-coupled receptors have been observed previously within the family of adrenergic receptors. Thus, the ␤ 2 -and ␣ 2 -C10-adrenergic receptors expressed in transfected fibroblasts are mainly localized in the plasma membrane at steady state, whereas the ␣ 2 -C4-adrenergic receptor is found both in the plasma membrane and in a population of intracellular vesicles (53).
A striking feature of both and ␦ labeling patterns was their clustering into a single large cap at one pole of the cell. Although comparable clustering patterns have been observed after ligand binding to insulin (54) and muscarinic cholinergic (55) receptors, most hormone-or neuropeptide-receptor (18,19,23,39) complexes have a tendency to aggregate into small, distinct patches distributed all over the cell membrane. In fact, earlier fluorescence studies have reported rhodamine-tagged enkephalins to form multiple small clusters at the surface of neuroblastoma cells (10). It may be that the extent of receptor clustering is controlled by the relative rates of ligand binding and internalization and that it varies from on type of cell to the other. When the rate of internalization is equal to (or faster than) the binding step, as observed for example in the case of neurotensin-receptor complexes (39), internalization begins as soon as small aggregates are formed, preventing the formation of large caps. By contrast, if the internalization proceeds at a much slower pace than ligand binding, as is the case for opioid receptors, polymerization of the complexes is allowed to proceed for longer times leading to the formation of large macromolecular aggregates. In any event, our kinetic results clearly show that internalization of both ␦ and ligands occurs only from a single polar cap, corroborating the idea that polymerization of the ligand-receptor complexes is a necessary step for internalization.
A major result of the present work is the demonstration that and ␦ receptors co-expressed in the same cells internalize through partly distinct endocytic pathways. Thus, during the early phase of internalization, and ␦ receptor-ligand complexes appeared to be mainly localized within mutually exclusive endosomal populations, as there was little overlap between the small, peripheral red and green fluorescent particles that likely correspond to early endosomes. Co-localization of the two fluorophores occurred only in larger organelles that were observed deeper in the core of the cells and probably correspond to late endosomes or lysosomes (56). To our knowledge, the present results provide the first direct evidence that different receptor subtypes co-expressed in the same cell may be sorted via different endocytic vesicles. A selective internalization mechanism has also been described for M␣ 2 -10H-and ␤ 2 -adrenergic receptors coexpressed in K293 cells (53). However, in that case, only the ␤ 2 receptor was internalized in response to cell stimulation by norepinephrine whereas the M␣ 2 -10H receptor remained in the plasma membrane. Interestingly, internalization vesicles of the ␤ 2 receptor were found to be distinct from those containing the M␣ 2 -10H receptor, a third homologous receptor co-expressed in the same cells, but which is already present in intracellular vesicles before any agonist stimulation (53). The subtype-specific receptor sorting documented in the present study suggests that individual receptors interact selectively with cellular proteins mediating intracellular sorting and trafficking. The selective sorting of internalized receptors could contribute to homologous desensitization and reactivation processes (57).
In conclusion, we have shown that and ␦ opioid peptides are internalized into transfected COS-7 cells as a consequence of their selective interaction with and ␦ opioid receptors, respectively. For each ligand, the internalization is likely to include the following steps: (i) ligand binding, (ii) partial and selective aggregation of homologous ligand-receptor complexes, (iii) interaction of aggregates with cellular proteins that mediate intracellular trafficking, and (iv) formation of vesicles and internalization. In this scheme, the signal of selectivity that ultimately leads to sorting of ligand-receptor complexes into specific vesicles would be given by the binding of the agonist to its specific receptor. The resulting conformational change would trigger the homologous aggregation of agonist-receptor complexes. A recent study on the mechanical origin of receptormediated endocytosis shows that membrane complex clustering can provide the stimulus for a local membrane motion toward the cytosol (58). However, it cannot be excluded that the sorting signal is given by selective interactions between the agonist-receptor complex and cellular proteins involved in the sequestration process. In addition to interacting with proteins involved in the formation of clathrin-coated (59) or non-clathrin-coated (60) vesicles, G-protein-coupled receptors have been shown to bind to proteins of the cytoskeleton (61, 62). It will be important to identify each one of these proteins and to determine whether their interaction with various receptors can be modulated by agonist-induced conformational changes of these receptors.