N-Glycan-mediated Quality Control in the Endoplasmic Reticulum Is Required for the Expression of Correctly Folded δ-Opioid Receptors at the Cell Surface*

A great majority of G protein-coupled receptors are modified by N-glycosylation, but the functional significance of this modification for receptor folding and intracellular transport has remained elusive. Here we studied these phenomena by mutating the two N-terminal N-glycosylation sites (Asn18 and Asn33) of the human δ-opioid receptor, and expressing the mutants from the same chromosomal integration site in stably transfected inducible HEK293 cells. Both N-glycosylation sites were used, and their abolishment decreased the steady-state level of receptors at the cell surface. However, pulse-chase labeling, cell surface biotinylation, and immunofluorescence microscopy revealed that this was not because of intracellular accumulation. Instead, the non-N-glycosylated receptors were exported from the endoplasmic reticulum with enhanced kinetics. The results also revealed differences in the significance of the individual N-glycans, as the one attached to Asn33 was found to be more important for endoplasmic reticulum retention of the receptor. The non-N-glycosylated receptors did not show gross functional impairment, but flow cytometry revealed that a fraction of them was incapable of ligand binding at the cell surface. In addition, the receptors that were devoid of N-glycans showed accelerated turnover and internalization and were targeted for lysosomal degradation. The results accentuate the importance of protein conformation-based screening before export from the endoplasmic reticulum, and demonstrate how the system is compromised when N-glycosylation is disrupted. We conclude that N-glycosylation of the δ-opioid receptor is needed to maintain the expression of fully functional and stable receptor molecules at the cell surface.

are immediately subjected to scrutiny by the ER quality control machinery. It ensures that only proteins with native conformation are transported to their final destinations. This is achieved by aiding the folding of nascent protein molecules and retaining and targeting aberrant folding products for degradation (1). The glycoprotein quality control is an essential part of general ER quality control. It relies on N-linked glycosylation, which occurs on asparagine residues at the Asn-X-(Ser/Thr) consensus sequence (where X is any amino acid except proline) of a folding substrate, as it emerges into the ER (2,3). Removal of two glucose residues from the core N-glycans allows the nascent glycoprotein to interact with ER-localized molecular chaperones calnexin and calreticulin, and subsequent release and rebinding to these chaperones is regulated by two enzymes, glucosidase II and UDP-glucose:glycoprotein glycosyltransferase, that remove and add glucose to the core glycan, respectively (3,4). The latter enzyme recognizes folding intermediates but disregards native proteins and substrates that are targeted for ER-associated degradation because of terminal misfolding. This so-called calnexin cycle retains incompletely folded proteins in the ER and with the help of interacting foldases, like protein-disulfide isomerase family member ERp57, assists in folding.
G protein-coupled receptors (GPCRs) are membrane proteins that, apart from a few exceptions, contain several consensus sites for N-glycosylation in their extracellular domains (5). They are therefore obvious targets for ER glycoprotein quality control. Nevertheless, the functional significance of their N-glycans in folding and intracellular transport has remained elusive. For example, the lack of N-glycosylation may lead to virtually complete intracellular retention, as has been reported for human vasoactive intestinal peptide 1 and rat AT 1a angiotensin II receptors (6,7). Frequently, however, the expression levels merely decrease, as happens for rat prostaglandin E 2 and human AT 1 receptors (8,9), or the non-N-glycosylated receptors are fully functional and properly localized at the plasma membrane, as is the case for canine histamine H2 and porcine m2 muscarinic acetylcholine receptors (10,11). Furthermore, these attributes are not necessarily dependent on N-glycosylation as such but on a particular N-glycan, in a specific position. Bovine rhodopsin has two N-glycans positioned in its N termi-
DNA Constructs-The h␦OR-pFT-SMMF construct encoding the h␦OR (GenBank TM accession number U10504 (56) with a cleavable influenza hemagglutinin signal peptide, N-terminal Myc tag, and C-terminal FLAG-tag has been described previously (18). The h␦OR-pcDNA5 encoding the h␦OR with a C-terminal FLAG tag was prepared by XhoI/KpnI digestion of the h␦OR-FLAG-pcDNA3 (14) and cloning into the pcDNA5 vector (Invitrogen). Preparation of the h␦OR-pcDNA5 encoding the h␦OR with a hemagglutinin signal peptide and N-terminal HA-tag will be described elsewhere. 3 The wild-type receptor constructs were modified to create the corresponding constructs encoding the N18Q, N33Q, and N18Q/N33Q mutants with QuikChange site-directed mutagenesis kit (Stratagene), using oligonucleotides 5Ј-CCGCTCTTCGCCGAAG-CCTCGGACGCC-3Ј and 5Ј-GGCGTCCGAGGCTTCGGC-GAAGAGCGG-3Ј to generate the N18Q modification, and 5Ј-GCGCTGGCGCCGAAGCGTCGGGGCC-3Ј and 5Ј-GGC-CCCGACGCTTCGGCGCCAGCGC-3Ј to generate the N33Q modification.
Metabolic Labeling-Cells were pretreated with 500 ng/ml tetracycline for 60 min (Figs. 2, 3, and 5A) or 16 h (Fig. 5B) in complete DMEM, and the reagent was maintained in the culture medium throughout depletion and labeling. Methionine and cysteine were depleted by incubating cells in methionineand cysteine-free DMEM for 60 min, and labeling was performed in a fresh medium with 100 -200 Ci/ml of [ 35 S]methi-onine/cysteine for 30 -60 min, as specified in figure legends. The labeling pulse was terminated by washing, and cells were chased in complete DMEM, supplemented with 5 mM methionine, for various periods of time as specified in figures. Inhibitors of lysosomal degradation were added to the chase medium (NH 4 Cl 10 mM, chloroquine (Sigma) 0.2 mM, and leupeptin (Alexis) 500 g/ml), whereas the N-glycosylation inhibitor tunicamycin (Alexis) was added 2.5 h before depletion, at a concentration of 5 g/ml. During depletion and labeling, the tunicamycin concentration was raised to 25 g/ml. Before harvesting, cells were incubated in phosphate-buffered saline (PBS), supplemented with 20 mM N-ethylmaleimide, for 10 min on ice.
Preparation and Solubilization of Membranes and Whole Cell Extracts-Cells were homogenized for radioligand and GTP-binding assays in buffer A (25 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 2 mM EDTA, 5 g/ml leupeptin, 5 g/ml soybean trypsin inhibitor, 10 g/ml benzamidine) and for immunoprecipitation and the heat inactivation assay in buffer B (25 mM Tris-HCl, pH 7.4, 2 mM EDTA, 2 g/ml aprotinin, 0.5 mM phenylmethylsulfonyl fluoride, 2 mM 1,10-phenanthroline, 5 g/ml leupeptin, 5 g/ml soybean trypsin inhibitor, 10 g/ml benzamidine) as described previously (18). Membranes were resuspended in buffer A for ligand-binding and GTP-binding assays and in buffer B for the heat inactivation assay, at a concentration of 1 mg/ml, and stored at Ϫ70°C. Membranes subjected to immunoprecipitation were suspended in buffer C (buffer B, containing 0.5% (w/v) n-dodecyl-␤-D-maltoside (Alexis) and 140 mM NaCl) and extracted for 60 min on a magnetic stirrer, and solubilized receptors were separated from the insoluble fraction by centrifugation at 100,000 ϫ g for 60 min. Metabolically labeled cells were sonicated with a tip sonicator three times for 15 s in buffer B, and membranes were recovered by centrifugation at 45,000 ϫ g for 30 min, washed twice, and extracted in buffer C as above. Total cellular lysates were prepared by suspending frozen cell pellets in buffer C (Figs. 1E and 5) or buffer D (buffer C, containing 0.5% (w/v) digitonin (Alexis) and 10 mM CaCl 2 but without EDTA; Fig. 3) for 30 min, and the insoluble material was removed by centrifugation at 16,000 ϫ g for 30 min. All steps were performed at 4°C. Protein concentration was determined using DC protein assay kit (Bio-Rad), with bovine serum albumin (BSA) as a standard.
Immunoprecipitation-Solubilized receptors were purified with either one-step (Figs. 1, B-D, and 2) or two-step immunoprecipitation (Figs. 1E and 5 and supplemental Fig. 1) with the anti-FLAG M2 or anti-HA antibody affinity resin, as described earlier (14,16), and eluted with 200 g/ml FLAG or HA peptide, respectively. Alternatively (Fig. 3), cellular lysates were divided into 3 aliquots, supplemented with 0.1% (w/v) BSA, and precleared with 20 l of protein G-Sepharose (GE Healthcare) for 60 min at 4°C. One aliquot was subjected to immunoprecipitation with anti-FLAG M2 antibody affinity resin, and the two others were incubated for 60 min at 4°C with anti-calnexin antibody (1:100) or rabbit preimmune serum (1:100), followed by a 2-h incubation with protein G-Sepharose. Elution was performed with 80 l of 1% (w/v) SDS in 25 mM Tris-HCl, pH 7.4, by incubating for 15 min at 22°C and for 5 min at 95°C. The eluates were diluted with buffer C, supplemented with 0.1% BSA, and immunoprecipitated with anti-FLAG M2 antibody affinity resin as described above.
Deglycosylation of Immunoprecipitated Receptors-Immunoprecipitated receptors were eluted from the anti-FLAG M2 antibody affinity resin with 1% (w/v) SDS, 50 mM sodium phosphate, pH 5.5, by incubating samples for 15 min at 22°C, and for 5 min at 95°C. The eluates were diluted 5-fold with buffer E (0.5% (w/v) n-dodecyl-␤-D-maltoside, 50 mM sodium phosphate, pH 5.5, 4 mM calcium phosphate, 0.5 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 5 g/ml soybean trypsin inhibitor, and 10 g/ml benzamidine) for O-glycosidase and neuraminidase digestions. The enzymes were added to a final concentration of 50 milliunits/ml, and samples were incubated at 30°C for 16 h before terminating the reaction by adding 40 l of SDS-sample buffer. Endo H and PNGase F digestions were performed as described (19).
Cell Surface Biotinylation-Receptor expression was induced for 16 h, and cell surface proteins were biotinylated with sulfo-NHS-biotin as described previously (14).
Radioligand Binding Assay-Binding assays were performed using 1-40 g of membrane protein in a final volume of 300 l of buffer A, containing 0.1% (w/v) BSA. For saturation binding experiments, triplicate samples were incubated with an increasing concentration of [ 3 H]diprenorphine, the final concentration ranging from 0.01 to 10 nM. For one-point binding assays, each triplicate contained 5 nM [ 3 H]diprenorphine. Nonspecific binding was determined using 10 M NTX. Samples were harvested, and data were analyzed as described earlier (18).
GTP-binding Assay-To produce cellular membranes with similar low amounts of receptor, cells were induced with 10 ng/ml (h␦OR, h␦OR(N18Q), and h␦OR(N33Q)) or 20 ng/ml (h␦OR(N18Q/N33Q)) tetracycline for 24 h. GTP-binding assays were performed using DELFIA GTP-binding kit (PerkinElmer Life Sciences), using a method modified from Frang et al. (20). Ten g of membrane protein was preincubated in a reaction buffer, containing 50 mM HEPES, 5 mM MgCl 2 , 3 M GDP, 100 mM NaCl, and 100 g/ml saponin, in a final volume of 100 l for 30 min on a plate shaker prior to addition of 10 l of 100 nM europium-labeled GTP. Receptor activation was induced with 0.01-1000 nM SNC-80, and nonspecific binding was determined by adding 5 M GTP␥S into the reaction buffer. After 30 min, the reaction was terminated by vacuum filtration (Multiscreen Vacuum Manifold, Millipore) and washed twice with ice-cold GTP-Wash Buffer. The bound europium-labeled GTP was measured with VICWallac 1420 VICTOR 2 Multilabel microplate reader, using the europium measurement protocol (excitation 350 nm, emission 615 nm, 0.4 ms delay, 0.4-ms window).
Analysis of Cell Surface Receptor Expression by Flow Cytometry-The total pool of cell surface receptors was labeled with anti-c-Myc antibody (9E10, ascites fluid 1:500) or anti-HA antibody (HA-7, 1:500) and phycoerythrin (PE)-conjugated rat anti-mouse IgG 1 (BD Biosciences, 1 g/ml) as described earlier (15). Fluorescein-NTX was used to label the ligand binding competent pool of cell surface receptors (Table 2 and  Quantitation of Receptor Internalization by Flow Cytometry-For assessing constitutive internalization, cell surface receptors were labeled with anti-c-Myc antibody (9E10, 2 g/ml; Santa Cruz Biotechnology) in complete DMEM for 30 min at 37°C. Cells were then thoroughly washed and chased in the medium for up to 4 h at 37°C, washed again with cold PBS, and labeled with PE-conjugated secondary antibody (1 g/ml) in PBS/FBS on ice. Agonist-induced internalization was initiated by supplementing the culture medium with 1 M Leu-enkephalin, and incubating cells at 37°C for up to 60 min. Cells were washed with ice-cold PBS, and cell surface receptors were labeled with anti-c-Myc antibody (2 g/ml; Santa Cruz Biotechnology) and PE-conjugated secondary antibody (1 g/ml) in PBS/FBS on ice. For all internalization assays, dead cells were labeled with 7-amino-actinomycin D, and triplicate samples of 10,000 cells were analyzed as described above.
Heat Inactivation Assay-Cellular membranes (200 g of protein) were incubated in Eppendorf Thermomixer (650 rpm) at 37°C for different periods of time as indicated in supplemental Fig. 3. After terminating the incubations by placing the samples on ice, membranes were diluted to 100 -150 g/ml, and MgCl 2 concentration was adjusted to 5 mM. The one-point binding assay was performed using 100 l of the suspension and a saturating concentration of [ 3 H]diprenorphine (5 nM) as described above.
SDS-PAGE, Fluorography, and Western Blotting-Protein samples were reduced in the presence of 50 mM dithiothreitol, incubated for 2 min at 95°C for complete denaturation, and analyzed in 10% SDS-polyacrylamide gels, using reagents from Bio-Rad or Amresco. Western blotting as well as fluorography, densitometric scanning, and data analysis for the detection and analysis of radioactively labeled proteins were performed as described previously (18,21). Films were scanned with Umax PowerLook 1120 color scanner and Image Master 2D Platinum 6.0 software.
Data Analysis-Data were analyzed using GraphPad Prism 4.02 software. Statistical analyses were performed using the regression analysis or analysis of variance with post-test Tukey analysis for experiments with more than one variable. Student's t test was performed for experiments with one variable only. The limit of significance was set at p Ͻ 0.05, and the data are presented as mean Ϯ S.E.

RESULTS
h␦OR Carries Two N-Linked Glycans, Both of Which Are Dispensable for Receptor Cell Surface Expression-The h␦OR contains two putative consensus sequences for N-linked glycosylation (Asn 18 -Ala-Ser and Asn 33 -Ala-Ser) at its N terminus (Fig.  1A). To determine whether both sites are occupied, and to study the effect of N-glycosylation on h␦OR processing and trafficking, the asparagine residues were mutated to glutamines by modifying the h␦OR construct that encodes the receptor with a cleavable hemagglutinin signal sequence (22) and Myc and FLAG epitopes at the N and C termini, respectively (18) (Fig. 1A). The modified receptor constructs (h␦OR(N18Q), h␦OR(N33Q), and h␦OR(N18Q/N33Q)) were stably transfected into a HEK293 i cell line (18,19) that allowed expression of all constructs from the same chromosomal integration site.
To identify the receptor species that were expressed, the stably transfected HEK293 i cells were treated with or without 500 ng/ml tetracycline for 24 h to induce receptor expression, and receptors were immunoprecipitated from solubilized membranes with immobilized anti-FLAG M2 antibody and analyzed by SDS-PAGE and Western blotting. All four cell lines displayed unique patterns of receptor species, but only under tetracycline induction (Fig. 1B). As we have shown previously (18,23), one major species of M r 61,000 and two minor ones of M r 47,000 and M r 51,000 were immunoprecipitated from induced cells that were transfected with the wild-type receptor construct (Fig. 1B, lane 2). The smallest receptor species of M r 47,000 was sensitive to Endo H digestion (Fig. 1C, lane 2), indicating that it represents the receptor precursor. In contrast, the two larger ones were Endo H-resistant but were digested with PNGase F (Fig. 1C, lanes 2 and 3, respectively), thus corresponding to mature receptor forms carrying complex-type N-glycans. When Asn 18 was mutated to glutamine, the immunoprecipitated receptor migrated as two species of M r 44,000 and M r 52,000 (Fig. 1B, lane 4), both of which were sensitive to PNGase F (Fig. 1C, lane 6). Two PNGase F-sensitive receptor forms of M r 45,000 and M r 51,000 were also detected when Asn 33 was mutated to glutamine ( Fig. 1, B, lane 6, and C, lane 9). The smaller molecular weight mutant receptor forms represent receptor precursors, as they were digested with Endo H (Fig. 1C, lanes  5 and 8). These results clearly indicate that both N-glycosylation sites of the h␦OR are used. The differences in the electrophoretic mobility of the mature mutant receptor forms (compare lanes 4 and 6 in Fig.  1B) indicate that the two N-glycans are processed in a dissimilar manner in the Golgi and/or otherwise have a different effect on electrophoretic mobility of the receptor. In addition, because the smaller Endo H-resistant wild-type receptor form co-migrated with the Endo H-resistant h␦OR(N33Q) mutant (compare lanes 2 and 6 in Fig. 1B), the M r 51,000 wild-type receptor is likely to correspond to a species that carries only one N-glycan at Asn 18 .
When both Asn 18 and Asn 33 of the h␦OR were mutated, the double mutant migrated as a single M r 45,000 species on SDS-PAGE (Fig.  1B, lane 8). Occasionally an ill-defined species migrating slightly faster than the M r 45,000 one was also apparent. As expected, the M r 45,000 species was insensitive to both Endo H and PNGase F (Fig. 1C, lanes 11 and 12, respectively). However, its molecular weight was higher than that of the deglycosylated precursors of the wild-type receptor or single mutants (Fig. 1C, compare lane 10 to lanes 2, 5, and 8), suggesting that it contains other post-translational modifications. As the h␦OR has been shown to undergo O-glycosylation (14), we tested the possibility that the M r 45,000 receptor form might also contain O-glycans by subjecting the immunoprecipitated receptor to glycosidase digestions. As seen in Fig. 1D, the h␦OR(N18Q/N33Q) was sensitive to O-glycosidase, but only if the samples were treated simultaneously with neuraminidase. This points out that the M r 45,000 receptor species represents a receptor form that is devoid of N-glycans but contains sialylated O-glycans. O-Glycosylation of the h␦OR is known to take place in the Golgi (14), suggesting that the M r 45,000 form of the h␦OR(N18Q/N33Q) represents a receptor species that has been exported from the ER. To further confirm that the M r 45,000 non-N-glycosylated receptor is indeed able to reach the plasma membrane, cell surface proteins were labeled with membrane-impermeable sulfo-NHS-biotin. The immunoprecipitated receptors were then analyzed by Western blotting, using either horseradish peroxidase-conjugated streptavidin or anti-FLAG M2 antibody. As expected, streptavidin was only able to detect the Endo H-resistant forms of the wild-type receptor and single mutants, whereas the antibody detected all receptor forms (Fig. 1E, compare lanes 1-3 in the upper and  lower panels). Streptavidin was also able to detect the immunoprecipitated M r 45,000 h␦OR(N18Q/N33Q) species but not the ill-defined M r 42,000 one (Fig. 1E, lane 4), indicating that the latter represents an intracellular form of the double mutant. Importantly, this intracellular form of the non-N-glycosylated receptor was relatively less abundant than that of the other receptor constructs. The same appeared to apply also to the cell surface form (Fig. 1E, compare lanes 1-4 in the lower panel). The lower cell surface expression level of the non-N-glycosylated receptor was confirmed by flow cytometry, using antic-Myc antibody and PE-conjugated secondary antibody (Fig.  1F, see also Table 2). Similar results were also obtained by transient transfection of Flp-In-293 and Flp-In-CHO cells with constructs encoding the wild-type h␦OR and the N18Q/N33Q mutant with Myc, HA, or FLAG epitope tags (supplemental Fig. 1).
N-Glycans of the h␦OR Are Not Essential for Receptor Function-To assess the effect of N-glycosylation on the ligandbinding ability of the h␦OR, crude cellular membranes from induced stably transfected HEK293 i cells were subjected to saturation binding assays, using the opioid antagonist [ 3 H]diprenorphine as a radioligand. As seen in Table 1, the maximal binding capacity (B max ) measured for the wild-type receptor expressing cells was significantly higher than for the receptor mutant cell lines. The lowest expression level was measured for the non-N-glycosylated receptor, with a B max of only 36% of the wild-type receptor level. However, the affinities for the ligand were similar for all receptors and were comparable with values reported previously for the wild-type receptor (24), thus indi-cating that the h␦OR does not need N-glycans for ligand binding.
The effect of N-glycosylation on downstream signaling of the h␦OR was examined by measuring stimulation of nucleotide exchange at heterotrimeric G proteins. The ability of increasing concentrations of SNC-80, a ␦OR-specific agonist, to stimulate GTP binding to G proteins was assessed for cellular membranes prepared from stably transfected HEK293 i cells that were induced with low concentrations of tetracycline. This was necessary to obtain sufficiently low and comparable amounts of receptor expression and to avoid saturation of the reaction components (ϳ2-4 pmol/mg of membrane protein). There were no differences in the basal stimulation, and the agonist induced about 2-fold increase in GTP binding for all receptor constructs. The potency of SNC-80 was, however, decreased by 1.9-fold for the non-N-glycosylated receptor (Table 1).
Non-N-glycosylated h␦ORs Mature Faster and Have a Shorter Half-life than the Wild-type Receptor-The findings that the non-N-glycosylated h␦OR existed mainly in the mature cell surface form but was expressed at a much lower level than the wild-type receptor suggest that N-linked glycosylation may have a role in receptor maturation and/or turnover. Thus, to investigate these possibilities further, metabolic pulse-chase labeling experiments were performed. Receptor expression was induced for 2 h, and methionine-depleted cells were labeled with [ 35 S]methionine/cysteine and chased for various periods of time. At each time point, cells were harvested, and solubilized membrane extracts were subjected to immunoprecipitation with anti-FLAG M2 antibody and immunoprecipitated receptors were analyzed by SDS-PAGE and fluorography. As seen in Fig. 2A, two receptor forms of M r 47,000 and M r 44,000 were detected for the wild-type receptor at the end of the pulse (lane 1). In time, these receptor forms, which were both Endo H-sensitive (data not shown), started to disappear, and two higher molecular weight forms of M r 61,000 and M r 51,000 started to accumulate ( Fig. 2A, lanes 2-6). These results confirm that the wild-type receptor exists in two glycoforms, containing either one or two N-glycans, the latter one being the predominant species. In line with this notion, only one low

binding and G protein-coupling properties of wild-type and mutant h␦ORs
Ligand-and GTP-binding assays were carried out using membranes that were prepared from stably transfected HEK293 i cells, expressing the wild-type h␦OR or the N18Q, N33Q, or N18Q/N33Q mutants. Cells were treated for 24 h with 500 ng/ml tetracycline for the ligand-binding assays or with either 10 ng/ml (h␦OR, h␦OR(N18Q), h␦OR(N33Q)) or 20 ng/ml (h␦OR(N18Q/N33Q)) tetracycline for the GTP-binding assays. Saturation binding assays were performed using 0.01-10 nM ͓ 3 H͔diprenorphine to obtain the K d and B max values. The GTP-binding assays were carried out using 0.01-1000 nM SNC-80 to obtain the EC 50 and E max values. The receptor densities in membranes used for the GTP-binding assays were measured by a one-point binding assay using 5 nM ͓ 3 H͔diprenorphine, and the values were 2.9 Ϯ 0.6, 2.7 Ϯ 0.5, 2.9 Ϯ 0.3, and 2.6 Ϯ 0.6 pmol/mg protein for the wild-type h␦OR, h␦OR(N18Q), h␦OR(N33Q), and h␦OR(N18Q/N33Q), respectively. Analysis of the data was performed using GraphPad Prism. Data represent the mean Ϯ S.E. of three independent experiments, performed in triplicate. molecular weight Endo-H-sensitive receptor species was detected for the single mutants, the h␦OR(N18Q) and h␦OR(N33Q) (Fig. 2, C and D, lane 1, respectively). In contrast, two low molecular weight receptor species were detected for the double mutant at the end of the pulse (Fig. 2B, lane 1). As the smaller one disappeared in time, it represents the precursor form of the non-N-glycosylated receptor that was barely detectable by Western blotting (see Fig. 1B, lane 8).

Receptor
Precursors of the wild-type receptor and the N18Q mutant were converted to the mature form with similar overall kinetics, reaching completion within 4 h (Fig. 2H). The N33Q mutant, on the other hand, displayed somewhat enhanced kinetics of maturation that was complete already within 2 h (Fig. 2H).
Interestingly, faster maturation kinetics appeared to characterize also the wild-type receptor that carried only one N-glycan (Fig. 2H). This is in line with the notion that the carbohydrate is attached to Asn 18 in the one N-glycan form. The differences in the maturation kinetics were not reflected in the maturation efficiency. It varied from 20 to 60% for both the wild-type receptor and the N18Q and N33Q mutants.
In contrast to what was observed for the single mutants (Fig.  2F), the precursor of the double mutant disappeared significantly faster than that of the wild-type receptor (half-time for disappearance 43 Ϯ 4 and 76 Ϯ 10 min for the h␦OR(N18Q/ N33Q) and wild-type h␦OR, respectively, p Ͻ 0.05) (Fig. 2E). Furthermore, a substantial amount of the non-N-glycosylated receptor was in the mature form already at the beginning of the chase (Fig. 2B, lane 1), and the half-time for maturation was significantly shortened (42 Ϯ 9 and 110 Ϯ 9 min for the h␦OR(N18Q/N33Q) and wild-type h␦OR, respectively, p Ͻ 0.001) (Fig. 2G). In addition, the mature non-N-glycosylated receptor started to disappear after 2 h, and less than half of the receptors were apparent after 6 h (Fig. 2G). This was in contrast to receptors that carried the N-glycan at Asn 33 . They were more stable, and their amount started to disappear only after 4 h (Fig. 2H).
To confirm that the distinct changes in the maturation of the h␦OR(N18Q/N33Q) did not result from the replacement of the two asparagines in the receptor N terminus, pulse-chase labeling experiments were conducted using wild-type receptor expressing cells that were treated with tunicamycin, which inhibits co-translational addition of core N-glycans to glycoproteins (25). As expected, the receptor precursors disappeared faster in these cells (Fig. 2E), and the maturation kinetics was enhanced (Fig. 2G).
h␦ORs Carrying Either One or Two N-Glycans Interact with Calnexin-The results of the pulse-chase labeling experiments indicate that the non-N-glycosylated h␦OR is exported from the ER faster than receptors carrying either one or two N-glycans. Because the ER lectins calnexin and calreticulin are known to be involved in the quality control and retention of glycoproteins in the ER (1), it can be hypothesized that the enhanced ER export might be due to the absence of interaction with these ER chaperones. We therefore tested the ability of the different receptor constructs to associate with calnexin by sequential co-immunoprecipitation. Stably transfected HEK293 i cells that were induced to express the different receptors were labeled with [ 35 S]methionine/cysteine for 30 min, and cellular lysates were prepared. The total pool of receptors was purified by two-step immunoprecipitation using anti-FLAG M2 antibody, and the calnexin-interacting receptor pool was recovered from the denatured anti-calnexin immunoprecipitate by second immunoprecipitation with anti-FLAG M2 antibody. As we have previously reported (18,23), the wild-type receptor precursor was readily recovered from the calnexin immunoprecipitate, whereas only a trace amount of the non-N-glycosylated receptor was recovered (Fig. 3B, lanes 1 and 4,  respectively). If the calnexin antibody was replaced with preimmune serum, no wild-type receptor precursor was detected (Fig. 3C, lane 1), confirming the specificity of the observed interaction. The precursors of the two single mutants were also recovered from the calnexin immunoprecipitate, but only about half of the receptors were actually recovered (compare lanes 2 and 3 in Fig. 3, A and B). In comparison, hardly any of the wild-type receptor precursors carrying one N-glycan were immunoprecipitated with calnexin (Fig. 3B, lane 1), although they were quite abundant in the anti-FLAG M2 antibody immunoprecipitate (Fig. 3A, lane 1).
Part of the Non-N-glycosylated h␦ORs at the Cell Surface Are Incapable of Ligand Binding-As was shown by the cell surface biotinylation assay and flow cytometry and suggested by the pulse-chase labeling experiments, the h␦ORs are transported to the cell surface whether the receptor molecules are N-glycosylated or not. However, it can be hypothesized that the enhanced maturation kinetics of the non-N-glycosylated receptors, and their significantly decreased ability to exploit the ER quality control machinery, may result in defective folding. Therefore, we wanted to assess the cell surface receptor population in more detail by flow cytometry, using both antibodies and fluorescein-conjugated opioid antagonist NTX. The ability of the conjugated antagonist to bind to cell surface receptors was first validated by control experiments, assessing the ligand binding conditions and specificity, and ascertaining the membrane nonpermeable nature of the ligand (supplemental Fig. 2).
For the flow cytometry assays, the induced stably transfected HEK293 i cells were labeled with fluorescein-NTX (0.1 M) or anti-c-Myc antibody, followed by PE-conjugated secondary antibody. In agreement with the saturation ligand-binding assays, less cell surface fluorescence was detected for cells expressing the mutant receptors, whether the labeling was carried out with the antibody or the conjugated ligand (Table 2), the difference reaching a significant level for the h␦OR(N33Q) and h␦OR(N18Q/N33Q). Importantly, the non-N-glycosylated receptor appeared to bind less ligand than expected, as the ratio of the ligand binding pool and the total pool of receptors for the h␦OR(N18Q/N33Q) was only 77% that for the wild-type receptor. This suggests that a substantial fraction of the non-N-glycosylated receptor population at the plasma membrane is in a conformation that is incompatible with ligand binding.
Non-N-glycosylated h␦ORs Are More Prone to Internalization with or without Agonist-mediated Activation-As the turnover of mature non-N-glycosylated receptors was increased (see Fig.  2G), it is likely that they are internalized and targeted for degradation more extensively than the wild-type receptors. To investigate this further, receptor internalization was followed by flow cytometry. To study constitutive internalization, the stably transfected HEK293 i cells expressing the wild-type h␦OR and the N18Q/N33Q mutant were induced, plasma membrane receptors were labeled with anti-c-Myc antibody, and the remaining antibody-bound receptors at the cell surface were FIGURE 3. Both N-linked glycans of the h␦OR are required for the efficient calnexin interaction. HEK293 i cells expressing the wild-type h␦OR or the N18Q, N33Q, or N18Q/N33Q mutants (lanes 1-4, respectively) were induced to express the receptor for 2 h, pulse-labeled with 200 Ci/ml [ 35 S]methionine/cysteine for 30 min, and harvested. Cellular lysates were divided into equal aliquots and subjected to immunoprecipitation with anti-FLAG M2 (A) or anti-calnexin (B) antibodies or preimmune serum (C), and receptors were recovered from the denatured eluates by second immunoprecipitation with anti-FLAG M2 antibody. The final eluates were analyzed by SDS-PAGE and fluorography. The receptor precursor forms are indicated with open symbols. E and छ, wild-type h␦OR carrying two or one N-glycans, respectively; ‚, h␦OR(N18Q); ƒ, h␦OR(N33Q); Ⅺ, h␦OR(N18Q/N33Q). IP, immunoprecipitation.

TABLE 2 Cell surface expression of wild-type and mutant h␦ORs
Stably transfected HEK293 i cells expressing the wild-type h␦OR or the N18Q, N33Q, or N18Q/N33Q mutants were induced for 24 h. Cells were labeled with anti-c-Myc antibody (9E10), followed by PE-conjugated secondary antibody, or with 0.1 M fluorescein-NTX, and the fluorescence of live cells was measured with flow cytometry. The GeoMean values were normalized to those obtained for the wild-type h␦OR that were set to 100%. The data were obtained from three independent experiments, performed in duplicate, and analyzed with CellQuest Pro and GraphPad Prism.
labeled with PE-conjugated secondary antibody after a chase. As seen in Fig. 4A, internalization occurred for both receptors and, as expected, was more pronounced for the non-N-glycosylated receptor, affecting significantly the level of cell surface receptors remaining after 4 h (48 Ϯ 9 and 14 Ϯ 1% for wild-type h␦OR and h␦OR(N18Q/N33Q), respectively; p Ͻ 0.05). In addition, the mutant receptors internalized faster (half-time for internalization was about 55 and 140 min for the mutant and wild-type receptors, respectively). A similar difference between the two receptors was also detected following agonist-mediated activation (Fig. 4B), when cells were treated with an opioid agonist Leu-enkephalin (1 M), and the remaining cell surface receptors were labeled after 10, 30, or 60 min with anti-c-Myc antibody and PE-conjugated secondary antibody. Within 60 min, 52 Ϯ 3 and 84 Ϯ 1% (p Ͻ 0.05) of the wild-type and non-N-glycosylated receptors, respectively, were internalized (Fig. 4B).
As the agonist-mediated internalization concerns receptors that are able to bind ligand, we next tested the possibility that the ligand binding competent mutant receptors might be structurally more unstable than the corresponding wild-type ones. For this purpose, a heat inactivation assay was applied that has been widely used to assess the structural instability of GPCRs (18,26). Cellular membranes from HEK293 i cells that were induced to express the wild-type h␦OR or h␦OR(N18Q/N33Q) were incubated at 37°C for increasing periods of time, and the remaining ligand-binding ability was tested using [ 3 H]diprenorphine. The binding ability of both receptors decreased in time, and the inactivation for the non-N-glycosylated receptor appeared to be faster at early time points (supplemental Fig. 3). However, the differences were minor, and the remaining ligand-binding ability was similar at the end of the 60-min incubation.
Non-N-glycosylated h␦ORs Are Targeted for Lysosomal Degradation-Lysosomal targeting of internalized h␦ORs was then investigated. First, metabolic pulse-chase labeling experiments were performed using the h␦OR(N18Q/N33Q) expressing HEK293 i cells that were treated with NH 4 Cl during the chase. This compound is known to increase pH in intracellular vesicles, thus inhibiting lysosomal degradation (27). As seen in Fig. 5A, more receptors were detectable at the end of the 4-and 6-h chase when cells were incubated in the presence of NH 4 Cl (compare lanes 5 and 6 in the upper and lower panels). The non-N-glycosylated receptors were also significantly stabilized with two other lysosomal degradation inhibitors, chloroquine or leupeptin. Importantly, none of these inhibitors had any effect on the wild-type receptor (Fig. 5B).
Subcellular localization of the wild-type and non-N-glycosylated h␦ORs was then investigated by confocal microscopy. Cells were induced for 7 h, permeabilized, and labeled with anti-FLAG M2 and anti-calreticulin antibodies. The latter was used as an ER marker. In agreement with our previous results (23), the wild-type h␦OR was found to reside mainly in the ER, whereas the non-N-glycosylated receptor was almost exclusively localized to the cell surface (Fig. 6, A-F). These results are in agreement with the pulse-chase labeling data that showed faster ER export for the non-N-glycosylated receptor (see Fig.  2). Trafficking of the cell surface mutant receptors to lysosomes was then studied by adding anti-c-Myc antibody and Lyso-TrackerRed into the culture medium 4 h before the cells were fixed, permeabilized, and incubated with Alexa-conjugated secondary antibody. Lysosomal degradation was inhibited by adding leupeptin into the medium. In addition to the cell surface, the receptors were now localized strongly in vesicular like structures (Fig. 6, M-O), co-localizing partially with the lysosomal marker (see inset in Fig. 6O). If anti-c-Myc antibody that recognizes the N-terminal extracellular domain of the receptor was replaced with anti-FLAG M2 antibody, which is targeted against the intracellular C terminus, no specific receptor labeling was detected (Fig. 6, J-L), confirming the specificity of the anti-c-Myc antibody staining. However, if anti-FLAG M2 antibody was added after permeabilization (Fig. 6, G-I), it was able to label the receptors in intracellular vesicles. As most of these receptors were labeled with both anti-FLAG M2 and anti-c-Myc antibodies, it appears that during the 4-h incubation a majority of the cell surface receptors was internalized, a finding that was consistent with the results of the constitutive internalization assay (see Fig. 4A).

DISCUSSION
All GPCRs, with a couple of exceptions, contain consensus sequences for N-glycosylation in their extracellular domains. A large number of these sites are co-translationally occupied by core N-glycans that are processed in the ER and Golgi as the receptor molecules are transported to the plasma membrane. An increasing amount of evidence suggests that the N-glycans mediate the interaction of nascent receptor molecules with ER glycoprotein quality control components (28 -34), preventing misfolded mutant receptors from exiting the ER (28,31,32,35).
However, the functional role of N-glycans in the scrutiny of newly synthesized wild-type GPCRs has remained obscure. In this study we directly demonstrate the critical role of N-glycans for the h␦OR, in both receptor folding and quality control.
Several lines of indirect and direct evidence support the important function of the N-glycan-mediated quality control of h␦ORs. (i) Mutation of the two consensus sites for N-linked glycosylation at the N-terminal extracellular domain resulted in a receptor that was incapable of interacting with calnexin, disappeared faster from the ER than the wild-type receptor, and was processed to the mature form with enhanced kinetics. Maturation was also facilitated if N-glycosylation of the wildtype receptor was prevented by tunicamycin. (ii) Some of the non-N-glycosylated receptors that reached the cell surface were incapable of ligand binding. (iii) In addition, the non-Nglycosylated receptors showed enhanced turnover and internalization and were targeted for degradation in lysosomes. Consequently, despite the enhanced kinetics of conversion of precursors to the mature form, and facilitated delivery of mature receptors to the cell surface, the steady-state level of h␦ORs that were devoid of N-glycans was lower compared with receptors carrying either one or two N-glycans. This was observed whether the receptor expression level was measured by saturation ligand binding, cell surface biotinylation, or flow cytometry.
Reduced expression is a common consequence following the mutation of GPCR N-glycosylation sites, and this has often been shown to accompany restricted transport of the mutated receptors to the cell surface and their intracellular accumulation (7-9, 36 -39). Nevertheless, for the h␦OR this was not the case, and the absence of N-glycans actually enhanced the delivery of receptors to the cell surface. The same phenomenon has been reported for temperature-sensitive yeast ␣-factor receptor mutants, as their trafficking defect was found to be rescued following the mutation of its two N-glycosylated asparagine residues (40). Likewise, a recent study on non-N-glycosylated ORs (41) has suggested that enhanced ER export may also characterize this receptor, although the mutant receptors were also found to accumulate intracellularly. The reason for the contradictory observations may be related to differences in the folding kinetics of the corresponding proteins, and/or the magnitude of their dependence on N-glycan interacting ER chaperones and foldases for the correct folding and processing. The h␦OR maturation is a very slow process, and it is therefore understandable that the retention of receptors in the ER is critical for the efficient expression of correctly folded receptors at the cell surface.
The apparent delay in ER export of glycosylated h␦ORs is most probably caused by lectin-mediated chaperone interactions. The ability of h␦ORs to interact with calnexin was confirmed by sequential co-immunoprecipitation, and all three N-glycosylated receptor forms, but not the non-N-glycosylated one, were found to co-purify with calnexin. Previously, calnexin has been shown to be involved in ER retention of mutant GPCRs, like the human V2 vasopressin receptor (28), and more recently it has also been found to interact with a number of wild-type GPCRs (e.g. see Refs. 30 -34). In this study, the efficiency of co-purification with calnexin was decreased for h␦OR FIGURE 5. Non-N-glycosylated h␦ORs are degraded in lysosomes. A, HEK293 i cells expressing the h␦OR(N18Q/N33Q) were induced to express the receptor for 2 h, pulse-labeled with 100 Ci/ml [ 35 S]methionine/cysteine for 40 min, and chased for increasing periods of time, in the absence or presence of NH 4 Cl (10 mM). Alternatively (B), HEK293 i cells expressing the wildtype h␦OR or the h␦OR(N18Q/N33Q) mutant were induced for 17 h, pulselabeled for 40 min, and chased for 6 h, in the absence or presence of NH 4 Cl (10 mM), chloroquine (10 mM), or leupeptin (500 g/ml). Receptors were purified from total cellular lysates by two-step immunoprecipitation with anti-FLAG M2 antibody, and analyzed by SDS-PAGE and fluorography. Intensities of the labeled mature receptor species were obtained by densitometric scanning of fluorograms, and the values were normalized to the receptor labeling in cells not treated with lysosomal inhibitors. The values given are means Ϯ S.E. of four to six independent experiments. ***, p Ͻ 0.001; *, p Ͻ 0.05; ns, not significant; open bars, wild-type h␦OR; black bars, h␦OR(N18Q/N33Q). The data were analyzed using the unpaired t test with GraphPad Prism.
species that contained only one N-glycan, and this was particularly evident for the wild-type receptor. This may reflect a genuine difference in the binding affinity of calnexin for the different receptor forms, as it has been demonstrated earlier that the efficient binding of calnexin to glycoproteins takes place only if at least two core N-glycans are attached to the nascent protein (42,43).
It has been reported previously that N-glycans that are attached to newly synthesized glycoproteins are not necessarily equally important for folding and ER retention (12, 36, 37, 44 -46). This appears to apply also to the h␦OR, as the N18Q mutant, unlike the N33Q one, resembled the double mutant in that its maturation kinetics was enhanced compared with the wild-type receptor. This implies that the second N-glycosylation site at Asn 33 is more critical for ER retention of the h␦OR. This N-glycosylation site seems to be conserved in the opioid receptor family (J. Tuusa, Blast, ClustalW multiple sequence alignments). Surprisingly, despite the obvious importance of this N-glycosylation site, it was used inefficiently, and some of the wildtype h␦ORs harbored carbohydrate only at Asn 18 . Similarly, the N18Q mutant existed partially in a nonglycosylated form at the cell surface. The reason for this phenomenon is not known at present but may be related to local differences in the amino acid sequence close to the consensus site (47)(48)(49) or to its closer proximity to the membrane (50). For example, studies with the tissue-type plasminogen activator and hemagglutinin-neuraminidase glycoprotein of the Newcastle disease virus indicate that events affecting the local folding of a polypeptide chain (e.g. disulfide bond formation) may limit the accessibility of consensus site asparagines to oligosaccharyltransferase (51,52). Differences in the local environment of the glycosylated asparagines of the h␦OR may also explain the divergent Golgi processing of the two N-glycans, which was suggested by the observation that mature forms of the N33Q and N18Q mutants displayed a clear difference in electrophoretic mobility, a difference that was not explained by O-glycosylation. Interestingly, a similar difference in mobility was also seen for the OR, when either one of its two glycosylated asparagines was mutated to glutamine (41).
A fraction of the non-N-glycosylated h␦ORs that reached the cell surface appeared to exist in a conformation that was not compatible with ligand binding. This was demonstrated by the fact that fluorescein-NTX was able to recognize a smaller population of cell surface receptors than c-Myc antibody, whereas no such difference was observed for receptors carrying at least one N-glycan. In previous studies, only a very few other glycoproteins have been shown to be able to exit the ER in an incorrect conformation. One of these is human tripeptidyl-peptidase I, a lysosomal enzyme that was found to be secreted with non-native disulfide bonds and only with limited activity after mutation of one of its five N-glycosylation sites (46).
The mature non-N-glycosylated h␦ORs were also found to have a faster turnover than the wild-type receptors. Furthermore, they were considerably more prone to internalization and were targeted for lysosomal degradation, as was demonstrated by flow cytometry, metabolic pulse-chase labeling, and confocal microscopy. Importantly, the agonist-mediated inter-nalization of the non-N-glycosylated receptors was also enhanced compared with wild-type receptors, suggesting that not only misfolded but also folding-competent mutant receptors were prone to efficient disposal from the cell surface. This is most likely explained by the absence of N-glycan-mediated stabilization of cell surface receptors, as has been reported for numerous other proteins (53), including the non-N-glycosylated OR that also internalizes more efficiently after agonistmediated activation (41). However, no significant differences in the structural stability were observed for the wild-type and non-N-glycosylated h␦ORs in the heat inactivation assay, possibly reflecting the inability of the in vitro assay to recapitulate in vivo conditions and events that take place in living cells. In any case, it is tempting to speculate that a distinct functional quality control machinery exists at the cell surface that is able to target unstable and/or misfolded receptors for disposal, a mechanism that has also been proposed to dispose other polytopic membrane proteins, like misfolded cystic fibrosis transmembrane conductance regulator and yeast plasma membrane ATPase Pma1 (54,55). The structural determinants that target the non-N-glycosylated h␦ORs for degradation remain to be determined in future studies. It is very likely that the quality control machinery relies on conformational attributes rather than merely on the absence of N-glycans.
In conclusion, this study expands our knowledge on GPCR biosynthesis and highlights the importance of N-glycan-mediated ER quality control in ensuring that newly synthesized GPCRs fold correctly before they are inserted into the plasma membrane. In addition, it accentuates the additional structural stability that the N-glycans provide at the cell surface. It can be envisioned that ER glycoprotein quality control is especially vital for receptors that require extended periods of time to find their correct conformation in the ER, emphasizing the need for more thorough investigation on the functional significance of N-glycosylation for GPCR biosynthesis and trafficking.