Monitoring Receptor Oligomerization Using Time-resolved Fluorescence Resonance Energy Transfer and Bioluminescence Resonance Energy Transfer

Oligomerization of the human δ-opioid receptor and its regulation by ligand occupancy were explored following expression in HEK293 cells using each of co-immunoprecipitation of differentially epitope-tagged forms of the receptor, bioluminescence resonance energy transfer and time-resolved fluorescence resonance energy transfer. All of the approaches identified constitutively formed receptor oligomers, and the time-resolved fluorescence studies confirmed the presence of such homo-oligomers at the cell surface. Neither the agonist ligand [d-Ala2,d-Leu5]enkephalin nor the inverse agonist ligand ICI174864 were able to modulate the oligomerization status of this receptor. Interactions between co-expressed δ-opioid receptors and β2-adrenoreceptors were observed in co-immunoprecipitation studies. Such hetero-oligomers could also be detected using bioluminescence resonance energy transfer although the signal obtained was substantially smaller than for homo-oligomers of either receptor type. Signal corresponding to the δ-opioid receptor-β2-adrenoreceptor hetero-oligomer was increased in the presence of agonist for either receptor. However, substantial levels of this hetero-oligomer were not detected at the cell surface using time-resolved fluorescence resonance energy transfer. These studies demonstrate that, following transient transfection of HEK293 cells, constitutively formed oligomers of the human δ-opioid receptor can be detected by a variety of approaches. However, these are not regulated by ligand occupancy. They also indicate that time-resolved fluorescence resonance energy transfer represents a means to detect such oligomers at the cell surface in populations of intact cells.

Recent studies have started to provide a significant body of evidence to support a concept of constitutive homo-oligomerization of a range of G protein-coupled receptors (GPCRs) 1 (1,2).
GPCRs for which such evidence exists include the ␤ 2 -adrenoreceptor (3,4), the D 2 dopamine receptor (5), the M 3 muscarinic acetylcholine receptor (6), the V 2 vasopressin receptor (7), the ␦-opioid (8,9) and -opioid receptors (9), the histamine H 2 receptor (10), and the CCR5 receptor (11). Furthermore, recent evidence has also indicated a requirement for the constitutive hetero-oligomerization of distinct GPCRs, such as between the GABA B R1 and GABA B R2 receptors, to generate a functional receptor expressed at the cell surface (12). Two important issues, however, remain contentious. The first of these is whether ligand occupancy alters the extent of GPCR oligomerization, and the second is the likely extent of GPCR hetero-oligomerization. In a range of reports, GPCR homo-oligomerization has been reported to be increased (3,4,11), decreased (8), or unaffected (6,7) by the addition of receptor ligands. Similarly, in GPCR heterodimerization studies, interactions have been indicated to be unaffected (12), regulated (13), or almost entirely dependent upon (14) the addition of receptor agonists and hetero-oligomers have recently been reported to form between quite distinct (14), as well as between closely related (12,13), GPCR sequences.
The earliest studies on GPCR oligomerization relied on the capacity to co-immunoprecipitate co-expressed but differentially epitope-tagged forms of a GPCR (see Ref. 15 for review). Because of the hydrophobic nature of the seven trans-plasma membrane helices of GPCR family members, care must be taken, however, to exclude nonspecific interactions between GPCR pairs resulting from detergent dissolution of cellular membranes. More recent studies have employed various forms of either fluorescence resonance energy transfer (FRET) (13,14) or bioluminescence resonance energy transfer (BRET) (4) to explore GPCR homo-and hetero-oligomerization in living cells.
Herein we use each of co-immunoprecipitation, BRET, and time-resolved FRET to examine constitutive homo-oligomerization of the human ␦-opioid receptor and its possible regulation by agonist and inverse agonist ligands. We demonstrate that such constitutive oligomerization can be observed for this GPCR using each of the three approaches but that ligands do not regulate these interactions appreciably. We also demonstrate that it is possible to detect a hetero-oligomeric interaction between the human ␦-opioid receptor and the human ␤ 2 -adrenoreceptor in co-immunoprecipitation studies. However, the energy transfer approaches indicated such interactions to be less prevalent than homo-oligomerization between either of the two receptor types.
As the time-resolved FRET assays were designed to detect only cell surface oligomers, these studies also demonstrate for the first time the presence of ␦-opioid receptor homo-oligomers at the plasma membrane. EXPERIMENTAL  Construction of Receptor Plasmids-The human ␦-opioid receptor in pcDNA4 was used as the starting point for the construction of the opioid receptor plasmids used in this study. Primers were made to introduce an ApaI site, followed by the FLAG™ epitope tag at the 5Ј end (5Ј-AA-AAAAGGGCCCGCCACCATGGACTACAAGGACGACGATGATAAGG-AACCGGCCCCCTCCGCC-3Ј) and to remove the stop codon and add an XbaI site at the 3Ј end (5Ј-TGCTCTAGAGGCGGCAGCGCC-3Ј) of the receptor. The resulting fragment was cloned into pcDNA3.1(Ϫ) (Stratagene). PCR of both Renilla reniformis luciferase and enhanced yellow fluorescent protein (eYFP) was performed to construct the fusion plasmids using primers, which introduced an XbaI and an XhoI site at the 5Ј and 3Ј ends, respectively. The primers were as follows: Renilla luciferase (forward, 5Ј-GCGTCTAGAACTTCGAAAGTTTATG-3Ј; reverse, 5Ј-TCGCTCGAGTTATTGTTCATTTT-3Ј), eYFP (forward, 5Ј-TGATCT-AGAATGGTGAGCAAGGGCGA-3Ј; reverse, 5Ј-AGACTCGAGTTACTT-GTACAGCTCGTC-3Ј). The resulting products were cloned into the plasmid containing the FLAG epitope-tagged ␦-opioid receptor to give the plasmids p␦ORluc and p␦ORYFP. All plasmids were sequenced to ensure fidelity of the PCR amplification and the maintenance of the correct open reading frame. Equivalent constructs containing an Nterminal c-Myc epitope tag sequence were generated in a similar fashion. FLAG™-and fluorescent protein-tagged forms of the human ␤ 2 -adrenoreceptor were produced as described previously (16,17). A positive control vector for BRET was constructed by linking together Renilla luciferase and eYFP as previously described by Xu et al. (18).
Cell Culture and Transfection-HEK293 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% newborn calf serum and 2 mM L-glutamine. Transient transfections were performed on cells that were at 70 -80% confluence with LipofectAMINE reagent (Life Technologies, Inc.) according to the manufacturer's instructions. Cells were harvested 48 h after transfection.
Co-immunoprecipitation-48 h after transfection, cell lysates were prepared. The cells were washed three times with 6 ml of PBS, followed by resuspension in 800 l of RIPA buffer (RIPA: 50 mM HEPES, 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 1% Triton X-100, 10 mM NaF, 5 mM EDTA, 0.1 mM NaPO 4 , 5% ethylene glycol). The cells were lysed for 1 h at 4°C on a rotating wheel before removal of the cell debris by a 10-min spin on a benchtop centrifuge at 13,000 rpm. 500 g of cell lysate protein was mixed with protein G-Sepharose (Sigma) for 1 h before removal of the protein G-Sepharose and its replacement by protein G-Sepharose with the appropriate antibody for the immunoprecipitation (1.6 g of anti-c-Myc antibody (A14, Santa Cruz), 6.8 g of anti-FLAG™ (M5 monoclonal, Sigma), or 1.6 g of anti-␤ 2 -adrenoreceptor (polyclonal, Santa Cruz). Samples were incubated overnight at 4°C on a rotating wheel. The protein G-Sepharose was then washed with RIPA buffer before resuspending the pellet in reducing SDS-PAGE sample buffer.
BRET-Cells were harvested 48 h after transfection. Media were removed from cell culture dishes, and cells were washed twice with PBS before cells were detached using TEM buffer (75 mM Tris, 1 mM EDTA, 12 mM MgCl 2 , pH 7.4). All samples were subjected to fluorescenceactivated cell sorting analysis to confirm the presence of eYFP in cells transfected with acceptor-tagged receptors. Approximately 4 ϫ 10 6 cells in 1.5 ml of TEM buffer were then added to a glass cuvette; an equal volume of TEM containing 10 M coelenterazine was then added and the contents of the cuvette mixed. The emission spectrum (400 -600 nM) was immediately acquired using a Spex fluorolog spectrofluorimeter with the excitation lamp turned off. For comparisons between experiments, emission spectra were normalized with the peak emission from Renilla luciferase in the region of 480 nm being defined as an intensity of 1.00. In some cases a BRET signal was calculated by measuring the area under the curve between 500 and 550 nm. Background was taken as the area of this region of the spectrum when examining emission from the isolated Renilla luciferase.
Time-resolved FRET-Cells were harvested 48 h after transfection. A 2-h incubation was performed at room temperature with 500,000 cells in a total volume of 100 l containing 15 nM Eu 3ϩ -labeled anti-c-Myc antibody (Wallac) and 45 nM APC-labeled M5 anti-FLAG™ antibody (in house) and 50% newborn calf serum/PBS. After incubation the cells were washed twice with PBS and resuspended in 30 l of PBS before placing into wells of a 384-well microtiter plate for FRET analysis using a Victor 2 (Wallac) configured for time-resolved fluorescence. The Eu 3ϩlabeled anti-c-Myc antibody was excited at 320 nm and emissions monitored at 615 nm (Eu 3ϩ emission) and 665 nm (energy transfer). A 200-s reading was taken after a 50-s delay to allow for decay of short-lived endogenous fluorescence signals.
[ 3 H]Ligand Binding Studies (BRET)-As the BRET experiments cannot distinguish between receptors present at the cell surface and in intracellular membranes, membrane preparations were used to obtain total cell expression levels.
␦-Opioid Receptor Binding Assays-For binding to the ␦-opioid receptor, a single concentration (4 nM) of [ 3 H] diprenorphine in the absence and presence of 300 M naloxone was used to define total and nonspecific binding. Assays were performed at 25°C for 1 h in a buffer comprising 50 mM Tris, 5 mM EDTA, 15 mM CaCl 2 , 5 mM MgCl 2 , 5 mM KCl, 120 mM NaCl, pH 7.4. ␤ 2 -Adrenoreceptor Binding Assays-For binding to the ␤ 2 -adrenoreceptor, a single concentration (2 nM) of [ 3 H]dihydroalprenolol in the absence and presence of 10 M propranolol was used to define total and nonspecific binding. Assays were performed at 30°C for 1 h in a TEM buffer comprising 75 mM Tris, 5 mM EDTA, 12.5 mM MgCl 2 , pH 7.4.
All binding experiments were terminated by filtration through Whatman GF/C filters followed by three washes with ice-cold TEM.
For each assay equivalent unlabeled cells were counted and membranes prepared to calculate receptor levels per cell (see Table I).
[ 3 H]Ligand Binding Studies (Time-resolved FRET)-As the timeresolved FRET experiments monitor only receptors delivered to the cell surface, intact cells were used for these studies in conditions akin to those used for the FRET experiments (see above). [ 3 H]DADLE (5 nM), in the absence or presence of naloxone (300 M), was used to define total and nonspecific binding.
Intact Cell Adenylyl Cyclase Activity Measurements-Were performed essentially as described in Refs. 19 and 20. Cells were split into wells of a 24-well plate, and the cells were allowed to reattach. Cells were then incubated in medium containing [ 3 H]adenine (0.5 Ci/well) for 16 -24 h. The generation of [ 3 H]cAMP in response to treatment of the cells with various ligands and other reagents was then assessed.

RESULTS
The human ␦-opioid receptor was modified at the N terminus to include either a c-Myc or a FLAG™ epitope tag. Following transient expression of either form of the receptor in HEK293 cells, these could be immunoprecipitated with appropriate antic-Myc or anti-FLAG™ antibodies (Fig. 1a). No immunoprecipitation was observed, however, when the antibody/epitope-tagged GPCR combinations were reversed, confirming the specificity of immunoprecipitation ( Fig. 1a and data not shown).
Immunoblotting of SDS-PAGE resolved membrane fractions expressing the c-Myc-tagged ␦-opioid receptor with the antic-Myc antibody resulted in detection of a 60-kDa polypeptide (Fig. 1b). Such a polypeptide was not detected by the anti-c-Myc antibody in membranes expressing the FLAG™-tagged form of the receptor (Fig. 1b). Co-expression of the c-Myc and the FLAG™ epitope-tagged forms of the ␦-opioid receptor followed by immunoprecipitation with the anti-FLAG™ antiserum and immunoblotting with the anti-c-Myc antibody also resulted in detection of the 60-kDa c-Myc-tagged ␦-opioid receptor (Fig.  1b). Equivalent results were obtained when the protocol was reversed and immunoprecipitation of cells co-expressing the two epitope-tagged forms of the ␦-opioid receptor was performed with the anti-c-Myc antibody followed by immunoblotting with the anti-FLAG™ antibody (data not shown). However, expression of either the c-Myc or FLAG™-tagged ␦-opioid receptor alone failed to result in detection of the 60-kDa polypeptide using either of these two protocols ( Fig. 1b and data not shown). Separate expression of the c-Myc-and the FLAG™ epitope-tagged forms of the ␦-opioid receptor followed by physical mixing of cell lysates prior to immunoprecipitation with either antibody also failed to result in co-immunoprecipitation of the two forms of the receptor (data not shown). Such results confirm previous data on the ability to detect homooligomers of co-expressed but differentially tagged forms of the ␦-opioid receptor (8).
When the c-Myc-tagged ␦-opioid receptor was co-expressed along with the human ␤ 2 -adrenoreceptor, co-immunoprecipitation experiments akin to those described above, but now using combinations of the anti-c-Myc antibody and an anti-␤ 2 -adrenoreceptor antibody, were able to provide evidence for the presence of hetero-interactions between these two GPCRs (Fig.  2). Immunoprecipitation of the ␤ 2 -adrenoreceptor resulted in the presence of the c-Myc-tagged ␦-opioid receptor in the precipitated sample, which could be detected by immunoblotting following resolution of the sample by SDS-PAGE. A second polypeptide with mobility consistent with a dimer containing the c-Myc-tagged ␦-opioid receptor was also detected (Fig. 2). Neither of these bands was detected when the human ␤ 2adrenoreceptor was expressed in the absence of the c-Myctagged ␦-opioid receptor and then immunoprecipitated (Fig. 2). Equivalent results were obtained when the c-Myc-tagged ␦-opioid receptor was co-expressed with a form of the ␤ 2 -adrenoreceptor that had been C-terminally tagged with eYFP or with a form of the ␤ 2 -adrenoreceptor tagged at the N terminus with the FLAG™ epitope and at the C terminus with green fluorescent protein (GFP) (Fig. 2).
Immunoprecipitation of either of these modified forms of the ␤ 2 -adrenoreceptor resulted in co-precipitation of the c-Myctagged ␦-opioid receptor and detection of both monomeric and potential dimeric species. These rather unexpected observations led us to consider whether such co-immunoprecipitation approaches following transient transfection of cells might produce artifactual results following solubilization of GPCRs from the membrane environment.
Recently, ␤ 2 -adrenoreceptor homo-oligomerization has been observed in intact cells by monitoring the interactions between forms of this GPCR C-terminally epitope-tagged with either Renilla luciferase or eYFP (4). We thus constructed C-terminally tagged Renilla luciferase and eYFP forms of the ␦-opioid receptor and the ␤ 2 -adrenoreceptor as well as a Renilla luciferase and eYFP fusion protein akin to that described by Xu et al. (18) to act as a positive control for BRET. Transient expression of either isolated Renilla luciferase (data not shown) or ␤ 2 -adrenoreceptor-Renilla luciferase in HEK293 cells, followed by the addition of the cell permeant luciferase substrate coelenterazine, resulted in emission of light with a single peak centred at 480 nm (Fig. 3a). Expression of the Renilla luciferase-eYFP fusion construct and addition of coelenterazine resulted in both a peak at 480 nm and the appearance of a second peak centred at 527 nm (Fig. 3a). This second peak represents energy transfer from Renilla luciferase to eYFP and its subsequent emission with lower energy. Co-expression of ␤ 2adrenoreceptor-Renilla luciferase and ␤ 2 -adrenoreceptor-eYFP followed by addition of coelenterazine again produced the dual peak consistent with energy transfer between the BRET partners, which is reliant on their close physical proximity (Fig. 3a). However, co-expression of the isolated Renilla luciferase along with ␤ 2 -adrenoreceptor-eYFP did not result in energy transfer upon addition of coelenterazine (data not shown), indicating that there were not direct interactions between these two constructs. Furthermore, separate expression of ␤ 2 -adrenoreceptor-Renilla luciferase and ␤ 2 -adrenoreceptor-eYFP followed by mixing of the cells prior to addition of coelenterazine also failed to produce an energy transfer signal (Fig. 3b), demonstrating a requirement for physical proximity for energy transfer. Equivalent studies with the co-expression of ␦-opioid receptor-Renilla luciferase and ␦-opioid receptor-eYFP again produced a pattern of light emission following addition of coelenterazine consistent with energy transfer and thus the proximity of the BRET partners and their associated GPCRs (Fig. 4a). As before, no energy transfer signals were observed when the Renilla luciferase and eYFP-tagged forms of this GPCR were transiently expressed in different populations of HEK293 cells, which were then mixed prior to addition of coelenterazine (data not shown).
Fluorescence-activated cell sorting analysis of cells to detect ␦-opioid receptor-eYFP expression demonstrated an average of 35% of the cell population was successfully transfected. 3 Hlabeled ligand binding studies using the ␦-opioid receptor antagonist [ 3 H]diprenorphine as radioligand indicated that an average of 200,000 Ϯ 10,000 copies of the receptor binding sites were present in each successfully transfected cell following co-expression of the ␦-opioid receptor-Renilla luciferase and ␦-opioid receptor-eYFP constructs ( Table I). Addition of either the agonist DADLE or the inverse agonist ICI174864 (20,21) (both up to 10 M) failed to produce a statistically significant alteration in the ␦-opioid receptor energy transfer signal (Fig.  4, a and b), indicating that these ligands were not altering the extent of ␦-opioid receptor oligomerization or the relative proximity of the two forms of this GPCR.
Co-expression of ␤ 2 -adrenoreceptor-Renilla luciferase and the ␦-opioid receptor-eYFP construct resulted in a small energy transfer signal upon addition of coelenterazine (Fig. 4, b and c). Combinations of [ 3 H]diprenorphine binding and direct monitoring of the fluorescence of eYFP clearly indicated that the ␦-opioid receptor-eYFP construct had been expressed success-fully in these experiments and at similar levels as in the receptor homo-oligomerization studies (Table I and data not  shown). Equivalent ligand binding studies used the ␤ 2 -adrenoreceptor antagonist [ 3 H]dihydroalprenolol to identify ␤ 2 -adrenoreceptor-Renilla luciferase. In cells that produced a positive BRET signal, an estimated 76,000 Ϯ 14,000 copies of this ␤ 2 -adrenoreceptor construct were expressed (Table I). Such results indicate that these two GPCRs can hetero-oligomerize in intact cells but that the extent of these interactions is substantially lower than for homo-oligomerization of either the ␦-opioid receptor or the ␤ 2 -adrenoreceptor. Addition of either the opioid receptor agonist DADLE or the ␤ 2 -adrenoreceptor agonist isoprenaline (both at 10 M) resulted in statistically significant (p Ͻ 0.05) increases in the energy transfer signal consistent with agonist-induced formation of a ␤ 2 -adrenoreceptor-␦-opioid receptor hetero-oligomer (Fig. 4b), but these signals remained small compared with those indicative of ␦-opioid receptor homo-oligomer formation (Fig. 4, b and c).
Neither the co-immunoprecipitation nor BRET studies can provide direct information on the cellular location of the detected receptor oligomers. To address whether ␦-opioid homooligomers were present at the cell surface and if at least this fraction of the GPCR homo-oligomer population would be regulated by agonist or inverse agonist treatment, we expressed combinations of N-terminally c-Myc-and Flag™-tagged forms of the ␦-opioid receptor in HEK293 cells. We then added combinations of a europium 3ϩ -labeled anti-c-Myc antibody as energy donor and an allophycocyanin (APC)-labeled anti-Flag™ antibody as acceptor to intact cells and used this pairing in time-resolved FRET studies monitored by light emission at 665 nm from APC. No increase in signal above that observed in untransfected HEK293 cells was detected upon individual expression of either the c-Myc or Flag™ forms of the ␦-opioid receptor. However, co-expression of these two forms of the ␦-opioid receptor resulted in strong time-resolved FRET (Fig.  5a), which provided a substantially greater signal to noise ratio than obtained in the BRET studies. Such a signal was only observed with addition of both the europium 3ϩ -labeled and APC-labeled antibodies. The absence of any of the four elements resulted in no energy transfer signal being detected (data not shown). Again, no time-resolved FRET signal was obtained if cells separately expressing either the c-Myc or Flag™ forms of the ␦-opioid receptor were simply mixed prior to addition of the antibodies (Fig. 5a).
Despite the substantially greater capacity to observe ␦-opioid receptor homo-oligomers in intact cells using time-resolved FRET compared with BRET, co-expression of the c-Myc-tagged ␦-opioid receptor and the FLAG™-␤ 2 -adrenoreceptor-GFP did not result in the production of a time-resolved FRET signal upon addition of the appropriate antibodies (Fig. 5a). These observations indicate that, at least at the cell surface, levels of a potential ␦-opioid receptor-␤ 2 -adrenoreceptor hetero-oligomer were not detectable. Despite the clear evidence for the presence of cell surface constitutive ␦-opioid receptor homooligomers, addition of neither DADLE nor ICI174864 (both at 100 nM) altered the time-resolved FRET signal (Fig. 5b). It was clearly possible that ligands would not be able to bind to the GPCRs in the presence of the antibodies required for timeresolved FRET. However, specific binding to the ␦-opioid receptor of [ 3 H]DADLE was unaffected by the presence of the antibodies (Fig. 5c). Furthermore, the presence of the epitope tag antibodies did not reduce the capacity of DADLE to mediate inhibition of cAMP production monitored in intact cell adenylyl cyclase activity assays (Table II).

DISCUSSION
It is now clear that many GPCRs have the capacity to oligomerize (1)(2)15). However, key issues that remain unresolved or contentious are how widespread this phenomenon is, whether such oligomers persist at the plasma membrane or reflect only a chaperonin-like strategy to deliver the GPCRs to the cell surface, and whether oligomerization is regulated by the binding of agonist ligands. Early studies utilized co-immunoprecipitation strategies to infer oligomerization of GPCRs and reports such as with the ␤ 2 -adrenoreceptor (3) indicated the extent of oligomerization to be increased in the presence of agonist ligands. As this has also been observed for a number of chemokine receptors (11,22,23), this increased the possibility of it being a widespread regulatory feature. However, related studies with the ␦-opioid receptor indicated that agonist ligands reduced the extent of oligomerization of this GPCR (8,9), whereas, for the -opioid receptor (9) and the M 3 muscarinic receptor (6), no effects of ligands were observed. Many of these above studies were unable to address the cellular location of the GPCR oligomers. However, the demonstration that appropriate membrane delivery and production of a functional GABA B receptor requires the co-expression of two distinct GPCR gene products suggested a chaperonin role for GPCR dimerization (Refs. 12, 24, and 25; see Ref. 26 for review). Recently, a considerable range of such GPCR hetero-oligomers have been detected (2) and, at least in certain cases such as FIG. 4. BRET-based detection of constitutive ␦-opioid receptor homo-oligomerization but little evidence for a constitutive ␦-opioid receptor-␤ 2 -adrenoreceptor hetero-oligomer. a, wavelength sweeps of light emission were conducted following addition of coelenterazine to HEK293 cells transiently expressing isolated Renilla luciferase (light blue) or ␦-opioid receptor-Renilla luciferase and ␦-opioid receptor-eYFP (purple, red, green). The agonist DADLE (red) or the inverse agonist ICI174864 (green) (both at 10 M) were present in some experiments. Data are from a representative set of experiments. b, The BRET signal was measured as described under "Experimental Procedures" for HEK293 cells expressing the ␦-opioid receptor-Renilla luciferase and ␦-opioid receptor-eYFP (columns 1-3) or ␤ 2 -adrenoreceptor-Renilla luciferase and ␦-opioid receptor-eYFP (columns 4 -6). Vehicle (columns 1 and 4), DADLE (columns 2 and 5), ICI174864 (column 3), or isoprenaline (column 6) (all at 10 M) were also present. Data represent means Ϯ S.E. from three independent experiments. Addition of either DADLE or isoprenaline to cells co-expressing ␤ 2 -adrenoreceptor-Renilla luciferase and ␦-opioid receptor-eYFP resulted in a significant (p Ͻ 0.05) increase in BRET signal. Ligands produced no significant effect when added to cells co-expressing ␦-opioid receptor-Renilla luciferase and ␦-opioid receptor-eYFP. c, wavelength sweeps of light emission were conducted following addition of coelenterazine to HEK293 cells transiently expressing the Renilla luciferase-eYFP fusion protein (purple), the ␦-opioid receptor-Renilla luciferase and ␦-opioid receptor-eYFP (red), ␤ 2 -adrenoreceptor-Renilla luciferase and ␦-opioid receptor-eYFP (green), or Renilla luciferase alone (black). Data are from representative traces. H]diprenorphine (␦-opioid receptor), which are sufficient to occupy more than 90% of the receptors. Parallel FACS analysis of the cell populations demonstrated transfection efficiency to vary between 33% and 40%. These values were then combined to calculate the number of receptors per transfected cell. HEK293 cells do not endogenously express the ␦-opioid receptor and although many clones of HEK293 cells endogenously express a very low level of the ␤ 2 -adrenoceptor this was not detectable in these studies. Data represent means Ϯ S.E. from three independent experiments. ND, not detected.   with a ␦/-opioid receptor hetero-oligomer, these have been reported to have pharmacological characteristics distinct from (presumably) homo-oligomers of these receptors (9). Moreover, in a number of reports on hetero-oligomer formation, their effective detection has required the presence of agonist ligands (13,14). Considerable efforts have recently been given to the development of assays able to monitor the presence and regulation of GPCR oligomers in intact cells. At least partially this reflects concerns of the possibility of potential artifacts being produced in studies that rely entirely on the co-immunoprecipitation of highly hydrophobic proteins. Energy transfer approaches have been the systems of choice. Recently, FRET between forms of the ␣-factor receptor of the yeast Saccharomyces cerevisiae, C-terminally tagged with energy transfer competent cyan and yellow fluorescent proteins, has been used to demonstrate constitutive oligomerization of this receptor and that this is unaffected by the presence of ligand (27). A variation of this procedure, termed BRET, has been used to re-examine ligand regulation of oligomerization of the ␤ 2 -adrenoreceptor. Here Anger et al. (4) observed both basal oligomerization of this receptor and an increase in this signal upon addition of the agonist isoproterenol. Although these results were consistent with agonist-induced oligomerization, the authors were careful to note that it could also represent only a re-orientation of the GPCR constructs. A third variation, FRET with photobleaching, has been applied to study homo-and hetero-oligomerization of somatostatin receptor subtypes (13) and between the somatostatin SSTR5 receptor and the D 2 dopamine receptor (14). Some, but not all, somatostatin receptor subtypes could form hetero-oligomers on addition of agonist (13) and the interaction between the SSTR5 receptor and the D 2 dopamine receptor could be achieved on addition of agonists for either receptor (14).
Herein, we have used combinations of GPCR co-immunoprecipitation studies, BRET, and a further variant of FRET, which takes advantage of the long-lived fluorescence characteristics of certain lanthanide chelates to allow time-resolved fluorescence to be employed to re-explore constitutive oligomerization of the ␦-opioid receptor and the possible effects of agonist and inverse agonist ligands. This approach has recently been applied to the analysis of the polypeptide makeup of the GABA A Cl Ϫ ion channel (28). All three approaches provided evidence of constitutive oligomerization of the ␦-opioid receptor. However, unlike the studies of Devi and colleagues (8,9), we were unable to observe consistent regulation of ␦-opioid receptor homo-oligomers by either a synthetic opioid peptide agonist or the classical inverse agonist at this receptor. It thus appears that, at least for this receptor in intact cells, oligomerization status is not related to the R and R* equilibria, which determine receptor activation state (29). Devi and co-workers (8) have also proposed that ␦-opioid receptor monomerization might be required for agonist-induced endocytosis. However, at least in the case of the S. cerevisiae ␣-factor receptor, recent elegant studies have indicated that it is internalized as a dimer (30). Both co-immunoprecipitation and the BRET studies are unable to provide information on the cellular location of the GPCR oligomers that produce the signals, but the Eu 3ϩ -and APClabeled antibody pairs used for the time-resolved FRET studies only have access to GPCRs successfully delivered to the plasma membrane and these studies confirmed the presence of preformed homo-oligomers at the cell surface and their lack of regulation by ligands (Fig. 5).
We also attempted to explore the specificity of GPCR oligomerization using each of the three approaches. In co-immunoprecipitation studies, apparent interactions could be observed between the co-expressed ␦-opioid receptor and the ␤ 2 -adrenoreceptor (Fig. 2). However, in such co-immunoprecipitation studies, samples can be exposed to film using enhanced chemiluminescence for the period required to obtain a signal. By contrast, we were unable to observe any significant interactions between these two receptors in the time-resolved FRET-based assays (Fig. 5). We were able, however, to observe a BRET signal consistent with interactions between co-expressed ␦-opioid receptors and ␤ 2 -adrenoreceptors (Fig. 4, b  and c). This signal was small when compared with those obtained when monitoring homo-oligomerization of either of these two receptors. However, in contrast to the situation with ␦-opioid receptor homo-oligomers, signal consistent with hetero-oligomerzation between the ␦-opioid receptor and the ␤ 2 -adrenoreceptor was increased in the presence of agonists at either of the receptors. However, even this signal was still small compared with that observed for either ␦-opioid receptor or ␤ 2 -adrenoreceptor homooligomers. This measured increase in either ␦-opioid receptor-␤ 2adrenoreceptor interaction or orientation relative to each other in response to agonist ligands is intriguing. It is possible that transient expression of the same or closely related receptors, e.g. opioid receptor subtypes, may result in constitutive interaction based primarily on effects of mass action and a significant level of mutual affinity, whereas less closely related receptors may require ligand binding to promote interactions. However, it must be stressed that the signal to noise ratio in the BRET assay is poor (as also noted by Angers et al. (Ref. 4)), at least in part because Renilla luciferase and eYFP are not optimal BRET partners and this limits the current sensitivity of the approach. Further studies that take advantage of novel and rapidly improving energy transfer techniques will be required to validate and unravel the basis for these observations. A general issue in all studies of this nature, particularly when performed using transient transfection, is whether artifacts may be produced due to high level expression of the potential interacting partners. In the current energy transfer studies, we have maintained levels of expression of the receptor constructs as low as possible (Table I) with the proviso that signal had to be sufficient to monitor the interactions.
Evidence from construction of the functional GABA B receptor dimer indicates that the GABA B R1 displays little capacity to move to the cell surface without co-expression of the GAB-A B R2 (12). In this example dimer formation takes place in the endoplasmic reticulum with the GABA B R2 acting both to mask an ER retention sequence in the GABA B R1 (31) and to allow plasma membrane delivery of the functional receptor, even though the GABA B R2 appears not be be directly involved in recognition of the agonist GABA. These requirements for interactions between the GABA B R1 and GABA B R2 at the ER to allow delivery of the functional receptor suggest that the GABA B R2 functions as a chaperonin for the GABA B R1. This may be a specialized and extremely well studied example of a common process in which GPCRs dimerize in the ER to act as mutual chaperonins (32). Evidence in favor of such a model is provided by the example in which co-expression of the D 3 dopamine receptor with a naturally occurring splice variant named D 3 nf, which lacks transmembrane regions VI and VII and functions akin to a dominant negative mutant, blocks delivery of the wild type receptor to the plasma membrane (33).
In many regards time-resolved FRET provided the most useful approach employed herein and certainly provided excellent signal to noise ratios. As noted above, a key feature of this approach is that it is only able to monitor the proximity of GPCRs which have matured and had reached the cell surface. In transient expression studies, this is often not achieved by a significant fraction of the expressed protein. The maturation process of the ␦-opioid receptor is appreciated to cause problems in effective cell surface delivery for this GPCR (34) and such intracellularly retained receptors cannot be resolved from those at the plasma membrane using BRET or the co-immunoprecipitation approaches. Second, the time-resolved nature of the fluorescence assays allows fluorescence derived from excitation of other cell components to decay prior to monitoring the signal. This substantially increases the signal to noise ratio obtained in the assay. However, despite these advantages, we were unable to detect ␦-opioid receptor-␤ 2 -adrenoreceptor interactions in this mode or to monitor regulation of ␦-opioid receptor homo-oligomerization by ligands.
One potential caveat of the time-resolved FRET approach is that the antibodies used to identify the epitope-tagged receptors are bivalent and thus might be anticipated to cluster receptors. Indeed, a monoclonal anti-␤ 2 -adrenoreceptor antibody with agonist-like properties has been described in which Fab fragments behave as antagonists (35). This is at least consistent with the idea that clustering of receptors might be required for signal transduction. However, although this is an interesting issue, it is unlikely to be of importance to the results of this study. The bivalency of the anti-FLAG antibody can only cause potential dimerization of the FLAG-tagged version of a receptor and the anti-c-Myc antibody likewise only potential dimerization of a c-Myc-tagged version of the receptor. However, to obtain a time-resolved FRET signal requires interaction between a FLAG-tagged opioid receptor and a c-Myc-tagged one to produce a pairing that can generate an energy transfer signal. If antibody-induced clustering were sufficient to provide sufficient proximity of the differentially tagged receptors, we would have anticipated that signal would also be produced for the ␦-opioid receptor and ␤ 2 -adrenoreceptor pair used in the time-resolved FRET format. This was not observed (Fig. 5a).
These studies demonstrate delivery of preformed ␦-opioid receptor oligomers to the surface of HEK293 cells following transient transfection and indicate that time-resolved FRET currently represents the most sensitive means to detect such oligomers at the cell surface in populations of intact cells.