Physical Association between Neuropeptide FF and μ-Opioid Receptors as a Possible Molecular Basis for Anti-opioid Activity*

Neuropeptide FF (NPFF) modulates the opioid system by exerting functional anti-opioid activity on neurons, the mechanism of which is unknown. By using a model of SH-SY5Y cells, we recently postulated that anti-opioid activity likely takes place upstream from the signaling cascade, suggesting that NPFF receptors could block opioid receptors by physical interaction. In the present study, fluorescence techniques were used to monitor the physical association and the dynamic of NPFF2 and μ-opioid (MOP) receptors tagged with variants of the green fluorescent protein. Importantly, cyan fluorescent protein-tagged NPFF2 receptors retained their capacity to antagonize opioid receptors. Fluorescence resonance energy transfer (FRET) and coimmunoprecipitation studies indicate that NPFF and MOP receptors are close enough to generate a basal FRET signal. The opioid agonist Tyr-d-Ala-Gly-NMe-Phe-Gly-ol disrupts by 20-30% this FRET signal, mainly because it concomitantly induces 40% internalization of receptors. In contrast, the NPFF analog 1DMe significantly increases by 10-15% the basal FRET signal, suggesting an association between both receptors. In addition, 1DMe reduces, by half, MOP receptor internalization, indicating that, besides a functional blockade of opioid receptors, the NPFF analog also inhibits their internalization. Finally, as a first report showing the modulation of the mobility of a G-protein-coupled receptor by another one, fluorescence recovery after photobleaching analysis reveals that 1DMe modifies the lateral diffusion of MOP receptors in the cell membrane, changing them from a confined to a freely diffusing state. By promoting NPFF-MOP receptor heteromerization, 1DMe could disrupt the domain organization of MOP receptors in the membrane, resulting in a reduction of opioid response.

NPFF (FLFQPQRFamide) 2 is representative of a family of RFamide peptides shown to regulate pain, cardiovascular func-tions, appetite, thirst, and body temperature (1,2). NPFF is part of a neuropeptidergic system composed of two precursors, pro-NPFF A and pro-NPFF B , and of two specific G-protein-coupled receptors (GPCRs), NPFF 1 and NPFF 2 (3). A close relationship between NPFF and opioid systems exists in the central nervous system, because the pharmacological and physiological effects of opioids, especially in pain perception, are modulated (increased or decreased, depending on dose and site of action) by NPFF agonists (4 -7). The NPFF system is therefore considered to be an opioid-modulating system involved in homeostasis that counteracts the action of opioids and, thus, plays a role in opiate tolerance (8).
Much data indicate that, at the cellular level, stimulation of NPFF receptors inactivates opioid receptors by a molecular mechanism still not understood (for review, see Ref. 9). For example, the NPFF analog 1DMe blocks presynaptic ␦-opioid auto-receptors in the spinal cord, leading to an increase in K ϩ -evoked met-enkephalin release (10). 1DMe also inhibits the reduction induced by morphine, but not by noradrenalin, of the electrically evoked acetylcholine release in the guinea pig myenteric plexus (11). Furthermore, on isolated neurons from rat periventricular and dorsal raphe nuclei, that co-express nociceptin receptors and NPFF 1 or NPFF 2 receptors, respectively, NPFF analogs inhibit the nociceptin-induced reduction of N-type Ca 2ϩ channel conductance (12,13).
To elucidate the molecular mechanisms responsible for the anti-opioid activity of NPFF on neurons, we have recently developed a suitable cellular model by stably transfecting human NPFF 1 or NPFF 2 receptors in the neuroblastoma-derived cell line SH-SY5Y (14,15). The neuroblastoma clone SH-SY5Y, derived from a human sympathetic ganglion tumor, displays many properties of mature sympathetic neurons, in particular the synthesis and depolarization-evoked, Ca 2ϩ -dependent, secretion of noradrenalin (16). In these cells, -opioid and nociceptin receptor activation leads, as in neurons, to inhibition of adenylyl cyclase and voltage-gated N-type Ca 2ϩ channels. We have demonstrated that activation of NPFF 1 and NPFF 2 receptors in the recombinant cell lines inhibits the response to opioid agonists in two different paradigms involv-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed: Tel. ing Ca 2ϩ signaling, thus reproducing the cellular anti-opioid activity observed with isolated neurons (12,13). Therefore, this cellular model, in contrast to those generally used (Chinese hamster ovary and COS), exhibits properties and gathers all protagonists of a neuronal cell, allowing a detailed molecular exploration of the opioid and anti-opioid signaling cross-talk that could have physiological significance. Using this model, we have accumulated data suggesting that classic cross-desensitization between both transduction systems does not occur. Indeed, NPFF and opioid receptors regulate similar effectors (adenylyl cyclase, voltage-gated N-type Ca 2ϩ channel, and activated phospholipase C) and are coupled to essentially the same G-protein ␣ subunits (14). However, only the pretreatment with NPFF agonists abolishes the opioid response, the reverse being not true (15). This effect is not accompanied by a loss of opioid receptor binding affinity. Moreover, the antagonistic effect of 1DMe on the opioid-induced reduction of Ca 2ϩ transients is not prevented by the addition of protein kinase A and C inhibitors (14). Altogether, these results led us to consider that the anti-opioid activity of NPFF receptors could take place upstream from the signaling cascade, as the result of a physical interaction between NPFF and opioid receptors leading to non-functional complexes. Oligomerization of GPCR has been proposed to play a role in receptor biogenesis and recycling, pharmacological diversity, and regulation of signal transduction (17,18). Interestingly, the pharmacological and signaling properties of heteromers differ from those of monomers and sometimes lead to inactive receptors as demonstrated for angiotensin receptors AT1 in AT1/ AT2 heterodimers (19) or somatostatin sst 3 receptors in sst 2A ⅐sst 3 complexes (20).
In the present study, different cellular, biochemical, and biophysical approaches were used to investigate the possible heteromerization between NPFF 2 and -opioid (MOP) receptors. Analysis of such an interaction in living cells has been greatly aided by the development of biological fluorescent probes derived from the native green fluorescent protein (GFP) of the jellyfish Aequora victoria, in association with biophysical and fluorescence spectroscopy techniques (21). Therefore, SH-SY5Y cells were stably transfected with NPFF and MOP receptors tagged with the cyan (CFP) and yellow (YFP) fluorescent proteins, respectively, to investigate their possible interaction by fluorescence resonance energy transfer (FRET) and fluorescence recovery after photobleaching (FRAP). FRET is a well suited technique to monitor protein-protein interactions or proximity in the 1-10 nm range in living cells and is appropriate for studying GPCR oligomerization (17,(22)(23)(24). FRAP has been extensively used to measure lateral diffusion of molecules in biological membranes (25) and has been applied to explore the confinement of GPCR in plasma membrane microdomains (26,27). In FRAP experiments, a small area of the cell membrane is bleached with a brief pulse of a high intensity laser beam, and recovery of fluorescence into the bleached area is used to measure the membrane diffusion rate of the fluorescent molecule of interest. We used FRAP at variable observation radius (vrFRAP), which allows one to identify the presence of microdomains and to characterize their size (28).
Results presented in this study provide convincing evidence for a close proximity between NPFF 2 and MOP receptors in the plasma membrane, which could be considered as heterodimerization. Interestingly, activation of NPFF receptors not only promotes receptor association, as visualized by FRET, but also modifies considerably the lateral mobility of MOP receptors in the plasma membrane by reducing the proportion of MOP receptors confined in microdomains. Upon activation, NPFF receptors could therefore move MOP receptors out of their signaling platform, preventing activation of their associated transduction systems, as well as their subsequent internalization. This hypothesis is proposed to explain the cellular antiopioid activity of NPFF agonists.

EXPERIMENTAL PROCEDURES
Materials-Oligonucleotides were synthesized by Sigma-Genosys (UK). Restriction enzymes were purchased from New England Biolabs (Ozyme, France). Mouse monoclonal anti-GFP antibody was obtained from Invitrogen, rabbit polyclonal anti-GFP antibody was from Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit polyclonal anti-MOP receptor antibody was a generous gift from Drs. Huda Akil and Stanley Watson (University of Michigan), but antibodies from Abcam (United Kingdom) were also tested. NPFF-related peptides were synthesized with an automated synthesizer (Applied Biosystems model 433A). DAMGO (Tyr-D-Ala-Gly-NMe-Phe-gly-ol) was purchased from Bachem (France), and levorphanol was from Hoffmann Laroche (France  (29). Radioiodinated peptides were stored at 4°C in the presence of 0.1% bovine serum albumin (Sigma, France).
Vector Constructions-pcDNA3 plasmids containing the enhanced cyan (mCFP) and yellow (mYFP) monomeric variants (A206K mutation) of GFP (30) were generous gifts from Dr. Roger Y. Tsien (University of California). Because the stop codon was lacking in these constructions, a short adapter containing two TAA codons was obtained by annealing sense (5Ј-CTAAGGGTAAT-3Ј) and antisense (5Ј-CTAGATTACCCT-TAGGTAC-3Ј) oligonucleotides. This adapter was then inserted at the 3Ј-end of each cDNA between KpnI and XbaI restriction sites of pcDNA3, which introduced three additional amino acids (Ala-Gly-Thr) at the C-terminal end of the proteins. cDNA were then transferred into pBluescript II SK Ϫ (Stratagene) using HindIII-XbaI sites for further cloning. We took advantage of the presence of an NcoI restriction site in the initiation codon to introduce the GFP cDNAs in-frame with the C-terminal end sequence of NPFF 2 and MOP receptors.
The human NPFF 2 receptor (NPFF 2 ) cDNA was obtained as previously reported (31), and the human MOP receptor cDNA was kindly provided by Dr. Laurent Emorine (Laboratoire de Biologie Cellulaire et Moléculaire du Contrôle de la Prolifération, Toulouse France). For the construction of NPFF 2 -CFP fusion protein, receptor cDNA was amplified by PCR with a forward primer containing a BamHI restriction site (5Ј-CGCG-GATCCACCATGAATGAGAAATGGGAC-3Ј), and a reverse primer (5Ј-TACACCATGGCAGATTCAATCTCACTGCTG-TTAGTAGTTTC-3Ј) designed to both suppress the stop codon and introduce a NcoI restriction site at the 3Ј-end of the sequence. This construction introduced three additional amino acids (Glu-Ser-Ala) at the C-terminal end of the receptor. The PCR product was ligated into PCR4Blunt-TOPO vector (Invitrogen) and subcloned into pBluesript II SK Ϫ using EcoRI. mCFP was ligated to the 3Ј-end of NPFF 2 using NcoI and XbaI sites. The resulting construct contained residues Glu-Ser-Ala between the sequences of NPFF 2 and mCFP. The construct was finally inserted into the EcoRI-XbaI sites of the mammalian expression vector pEFIN3 bearing the neomycin selection marker (31). For the construction of MOP-YFP fusion protein, receptor cDNA was amplified with a forward primer designed to replace the NcoI site present in the initiation codon by EcoRV (5Ј-TCGCGATATCATGGACAGCAGCGCTGCCC-3Ј) and a reverse primer designed to both suppress the stop codon and introduce NcoI and XbaI restriction sites at the 3Ј-end of the sequence (5Ј-TAGTTCTAGATTACACCAT-GGGACCCTGTAAG-3Ј). The PCR product was ligated to PCRT7/CT-TOPO (Invitrogen) and subcloned into pBluescript II SK Ϫ using HindIII-XbaI. mYFP was ligated to the 3Ј-end of MOP receptor using NcoI and XbaI sites. The resulting construct contained residues Leu-Gln-Gly-Pro between the sequences of MOP receptor and mYFP. The completed construct was finally inserted into the EcoRV-XbaI sites of the mammalian expression vector pEFIB3 bearing the blasticidin selection marker. All constructs were verified by sequencing (Genome Express, France).
Cells were transfected by using FuGENE 6 according to the manufacturer's instructions (Roche Applied Science). Selection was achieved by adding to the culture medium 400 g/ml G418 (Invitrogen) for pEFIN3 constructs or 2 g/ml blasticidin (Cayla, France) for pEFIB3 constructs, or both for double transfected cells. Individual clones were isolated by limit dilution in 96-well plates and grown for at least 6 -8 weeks, because scattering strongly slowed down growth of neuroblastoma cells. After amplification, clones were characterized for their binding and functional properties. The NPFF 2 -CFP (clone B10) cell line was obtained first and then transfected with the MOP-YFP vector to obtain the double transfected cell line NPFF 2 -CFP/MOP-YFP (clone B8). The MOP-YFP vector was also used to transfect the SH 2 -D9 clone previously characterized (14) to give the (SH 2 -D9)MOP-YFP (clone B7) cell line that expresses nontagged NPFF 2 receptors and YFP-tagged MOP receptors.
Coimmunoprecipitation Assay-Cell membranes (300 -350 g), prepared as described previously (14), were suspended and incubated for 1 h at 4°C with gentle agitation in freshly prepared solubilization buffer consisting of 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40 and protease inhibitor mixture (Complete EDTA-free, Roche Applied Science). The homogenate was then centrifuged at 20,000 ϫ g for 2 min at 4°C. The supernatant was precleared by incubation with 25 l of EZView red protein A affinity gel (Sigma) for 3 h at 4°C and centrifugation at 12,000 ϫ g for 30 s. Immunoprecipitation was then performed with 2 g of mouse monoclonal anti-GFP antibody for 1 h at 4°C, followed by overnight incubation at 4°C with protein A-agarose beads. The washing procedure was exactly as described by the manufacturer (Invitrogen). The immunoprecipitate was resuspended in Laemmli sample buffer (Sigma), boiled for 5 min, and subjected to 7.5% SDS-PAGE. Proteins in the gel were transferred to a polyvinylidene difluoride membrane for immunoblotting under standard conditions in Tris-buffered saline (20 mM Tris, pH 7.6, 137 mM NaCl) containing 0.1% Tween and 1% bovine serum albumin. Immunoreactivity was revealed with peroxidase-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories) using the ECL Plus Western blotting detection kit (Amersham Biosciences).
Quantitative FRET Imaging-Cells were seeded on glass coverslips in 6-well plates. The following day, cells were rinsed three times with KRH buffer (124 mM NaCl, 5 mM KCl, 1.25 mM MgSO 4 , 1.5 mM CaCl 2 , 1.25 mM KH 2 PO 4 , 25 mM HEPES, 8 mM glucose, 0.5 mg/ml bovine serum albumin, pH 7.4) and then incubated for 5, 10, or 15 min at room temperature with 2 ml of buffer (control) or 2 ml of buffer plus 1 M 1DMe or DAMGO. Cells were rinsed one time with KRH, one time with PBS, and were fixed with 3.8% formaldehyde (Sigma) in PBS for 20 min at room temperature. After three washes with 2 ml of PBS, slides were mounted in PBS on CoverWell imaging chambers (Sigma).
Images were captured with a Leica DM 5000B (Germany) upright fluorescence microscope equipped with a 100/1.4 oil immersion objective, a HBO mercury lamp and a cooled charge-coupled device camera (CoolSNAP ES , Photometrics, Roper Scientific France). FRET measurements were performed under the "Sensitized FRET" configuration (32-34) of the Metafluor software (Universal Imaging Corp.). Three images of the same field were successively collected at 5-s intervals through three series of filters (Omega Optical Inc.): 1) FRET (Ex: 440AF21; 455DRLP; Em: 535AF26), (2) CFP (Ex: 440AF21; 455DRLP; Em: 480AF30), and 3) YFP (Ex: 500AF25; 525DRLP; Em: 545AF35). The integration time was 800 ms, without binning. Background correction was performed by subtracting an image of an empty field, acquired with the corresponding filter set under the same conditions. In each background-subtracted image, linear region of interest surrounding each cell were selected to quantify the mean fluorescence intensity at the plasma membrane, using Metamorph software (Universal Imaging Corp.).
FRET values were corrected (FRETc) for the contribution of CFP and YFP emission in the FRET channel due to bleedthrough of fluorophores, according to Equation 1, where I FRET , I CFP , and I YFP , correspond to the mean fluorescence intensity of a given region of interest in the FRET, CFP, and YFP channels, respectively, and a and b are the ratio values I FRET /I CFP and I FRET /I YFP determined in cell lines expressing only CFP-or YFP-tagged receptors, respectively (32)(33)(34). An illustration of the fluorophore contribution in wild-type and transfected cells is given in Fig. 4. Non-transfected SH-SY5Y cells displayed a moderate and non-homogeneous autofluorescence that was only detected in the CFP and FRET channels, giving an I FRET /I CFP ratio of 2 (Fig. 4). In CFP cells, the contribution of CFP bleed-through into the FRET channel was b ϭ 0.67 Ϯ 0.02 (3). No fluorescence was detected at the acceptor excitation wavelength (Fig. 4). In YFP cells, the contribution of YFP fluorescence in the FRET channel was a ϭ 0.54 Ϯ 0.02 (3), calculated with FRET intensity values corrected for the autofluorescence contamination (twice the intensity recorded in the CFP channel). Because fluorescence intensity varied among cells, results were expressed as normalized FRET: Confocal Imaging-Cells seeded on glass coverslips were incubated for 30 min at room temperature with 5 M agonist, after a 30-min preincubation in KRH buffer or KRH buffer containing 5 M 1DMe. After fixation, the cells were mounted in Vectashield medium (AbCys, France) and observed on an upright laser scanner confocal microscope (Leica TCS SP2, Germany) with a 40/1 oil immersion objective. The excitation wavelength was 514 nm, and images were acquired without frame average or frame accumulation, with a zoom of 2.97.
vrFRAP Experiments on Living Cells-(SH 2 -D9)MOP-YFP cells were grown on glass coverslips in 6-well plates for 36 h. Each coverslip was washed with 3 ml of KRH buffer, placed on a homemade stainless steel slide, and left for 5 min at room temperature with 500 l of KRH buffer (control) or 500 l of KRH buffer containing 1 M 1DMe. Cells were then covered with another coverslip and observed for 30 -40 min at room temperature (20 -22°C).
A detailed description of the conditions for FRAP experiments with variable observation radius (vrFRAP) has been previously reported (26). The vrFRAP work station consisted of an upright Leitz Ortholux II fluorescence microscope coupled to a photomultiplier detection system and equipped with a 63/1.3 oil immersion objective combined to a set of diaphragms allowing the diameter of the observed field to vary between 1.4 and 3.45 m. The cell membrane was illuminated with an argon laser beam ( ex ϭ 488 nm), and the fluorescence was recorded over a period of 30 s, including 3 s of acquisition before bleaching. Shutter and laser settings were adjusted to prevent photobleaching during recording and to obtain a 40 -60% bleaching rate during the 20-to 40-ms bleaching time. For a given observation radius, the measurements were repeated in at least 30 cells. No variation in cell responses was observed over the 30 -40 min of observation.
Fluorescence recovery curves were analyzed by fitting experimental data to a mathematical model of diffusion (35) with a customized minimization algorithm. The apparent diffusion coefficient (D app ) and the mobile fraction (M) corresponding to the fluorescent molecules that can diffuse through the bleached area (25) were deduced from the pool of recovery curves of each examined sample. The radius (r) of the diffusion area of the mobile fraction could be estimated from the relationship between M and the observation radius (R) described by Equation 2 (28), where M p represents the permanent mobile fraction able to freely diffuse over long distances and which is independent of R. For closed areas, the real diffusion coefficient (D real ) in the domain can be calculated from Equation 3.
Analysis of Data-Non-linear regression and statistical analysis of the data were performed using Prism 4.01 (GraphPad software Inc.).

Characterization of SH-SY5Y Cells Stably Expressing Receptors Tagged with Spectral Variants of GFP-SH-SY5Y
cells transfected with CFP-tagged NPFF 2 receptors were selected for an expression level comparable (B max ϭ 367 fmol/mg, Fig. 1A) to that of endogenous opioid receptors (100 -300 fmol/mg) in SH-SY5Y cells (14). CFP-tagged NPFF 2 receptors retained high ligand affinity (Fig. 1A) and were still able to inhibit adenylyl cyclase (Fig. 1B), although with a slightly lower potency compared with non-tagged receptors (Fig. 1, B versus D). Importantly, the anti-opioid activity of NPFF 2 -tagged receptors was verified ( Fig. 2 and text below) to make sure of the validity of the model, and the selected clone was further transfected with YFPtagged MOP receptors to give the NPFF 2 -CFP/MOP-YFP cell line. In the doubly transfected clone, the affinity of [ 3 H]DAMGO (K D ϭ 0.34 nM, Fig. 1C) was similar to that measured in SH-SY5Y cells (14), and the MOP receptor density was an order of magnitude larger (B max ϭ 1.08 pmol/mg, Fig. 1C). Therefore, the total level of fluorescent receptors in the NPFF 2 -CFP/MOP-YFP cell line was moderate, which prevented generation of FRET arising from random collision due to overexpression (36). This also yielded an optimal stoichiometry of donor and acceptor (1/2) for measuring FRET (37,38).
To investigate the effect of the NPFF agonist on the internalization of MOP receptors as well as on their lateral displacement on the membrane surface, MOP-YFP cDNA was also introduced in the SH 2 -D9 clone expressing NPFF 2 receptors, previously characterized as a model for NPFF anti-opioid activity (14). The ability of YFP-tagged MOP receptors to inhibit forskolin-induced cAMP in (SH 2 -D9)MOP-YFP cells was verified (Fig. 1D). Receptors were functional, with a better efficacy (80% versus 50%) and potency (ϳ3 nM versus 0.5-1 M) than native SH-SY5Y opioid receptors (14).
Anti-opioid Activity of CFP-tagged NPFF 2 Receptor-The antagonistic activity of NPFF agonists on the opioid-induced reduction of the Ca 2ϩ transients triggered by depolarization in SH-SY5Y cells has been previously well characterized (14). Although the magnitude of the response to depolarization was unchanged in cells treated for 30 min with 0.1 M 1DMe (⌬(F/ F 0 ) ϭ 2.233 Ϯ 0.132, n ϭ 123 in treated cells compared with 2.227 Ϯ 0.167, n ϭ 110 in control cells), the response to DAMGO was reduced by 40 -50%. As shown in Fig. 2, the presence of CFP at the C-terminal end of the NPFF 2 receptor did not prevent its anti-opioid activity. DAMGO (1 M) reduced by 40 -50% the magnitude of the Ca 2ϩ transients evoked by KCl depolarization in wild-type and in NPFF 2 -or NPFF 2 -CFP-transfected SH-SY5Y cells. 1DMe (0.1 M, for 30 min at room temperature) was inactive in wild-type SH-SY5Y cells but halved the inhibitory effect of DAMGO in NPFF 2 as well as in NPFF 2 -CFP transfected cells, indicating that CFP-tagged receptors retained their ability to functionally antagonize MOP receptors.
Modulation of MOP Receptor Internalization by 1DMe-The effect of 1DMe on the internalization of MOP receptors was explored by quantitative confocal imaging in (SH 2 -D9)MOP-YFP cells.
As described in the literature, MOP receptors were internalized after 30-min stimulation by 5 M DAMGO (Fig. 3A). The fluorescence intensity measured at the cell membrane was reduced by 40% (Fig.  3B). Neither internalization of MOP receptors (Fig. 3A) nor a significant decrease of membrane fluorescence (Fig. 3B) was observed following a 30-min incubation with 5 M 1DMe, indicating that stimulation of NPFF receptors does not produce MOP receptor endocytosis and that anti-opioid activity of NPFF is probably not due to a loss of MOP receptors at the cell surface. In contrast, in cells preincubated for 30 min with 5 M 1DMe before DAMGO stimulation, internalization of MOP receptors was still observable (Fig. 3A), but the reduction of membrane fluorescence was only 20% (Fig. 3B), indicating that stimulation of NPFF receptors prevents internalization of MOP receptors. This experiment describes therefore a novel anti-opioid effect exerted by NPFF on opioid receptor endocytosis.
Visualization of the Interaction between NPFF 2 and MOP Receptors by FRET-FRET experiments were carried out to test if receptor interaction or proximity could be involved in the functional blockade of opioid receptors by NPFF. FRET was measured by fluorescence imaging using the three-filter method that allows for correction of FRET (FRETc) for the contribution of fluorophores and autofluorescence in the FRET channel (see details under "Experimental Procedures"). Fig. 4 shows images acquired through the three filter sets (CFP, YFP, and FRET) for wild-type, CFP-transfected, and YFP-transfected SH-SY5Y cells, giving an illustration of the autofluorescence and fluorophore bleed-through in the FRET channel. In preliminary time-lapse experiments on living cells, we noticed that the FRET signal changed over time. However, this was difficult to quantify because of variations in fluorescence due to cell movement, membrane shape fluctuations, and photobleaching. Therefore, FRET was measured on fixed cells, first treated for different times with agonists.   These results clearly indicate therefore that basal FRET occurs and is modulated by agonists in a timedependent fashion in cells cotransfected with GFP-tagged NPFF 2 and MOP receptors. Basal FRET significantly increased (10 -15%) upon stimulation with the NPFF agonist, whereas it decreased (22-33%) upon stimulation with the opioid agonist (Fig. 6B), probably because of the internalization of MOP receptors as described above.

FIGURE 1. Pharmacological characterization of SH-SY5Y cells expressing CFP-tagged NPFF 2 receptors (A and B), YFP-tagged MOP receptors (D) or both (C). A and C, saturation curves of three experiments performed in duplicate corresponding to the specific binding of [ 125 I]EYF on the membrane of cells stably expressing NPFF 2 -CFP receptors (A) or the specific binding of [ 3 H]DAMGO on membrane of cells co-expressing MOP-YFP and NPFF 2 -CFP receptors (C). B and D, dose-response curves of the inhibition of forskolin-induced intracellular cAMP accumulation by 1DMe (black square) or DAMGO (white square) in cells expressing either NPFF 2 -CFP receptors (B) or non-tagged NPFF 2 receptors together with YFP-tagged MOP receptors (D). Curves
Biochemical Characterization of NPFF 2 -MOP Receptor Interaction-The presence of heteromeric NPFF 2 ⅐MOP receptor complexes in the membrane of cells was further investigated by coimmunoprecipitation studies. Anti-GFP antibodies, which recognize both variants of the green fluorescent protein, were used to immunoprecipitate CFP-tagged NPFF 2 receptors in NPFF 2 -CFP cells, and the presence of endogenous MOP receptor in the immunoprecipitate was visualized using anti-MOP receptor antibodies. Fig. 7A shows that anti-GFP antibodies specifically recognize a band of 110 -120 kDa in solubilized membranes from MOP-YFP-transfected cells that corresponds to the fusion of YFP (27 kDa) with MOP receptor. The deduced size for MOP receptor (83-93 kDa) is in accordance with that reported for overex-

. Fluorescence microscopy in SH-SY5Y wild-type cells (top) or cells transfected with NPFF 2 -CFP alone (middle) or MOP-YFP alone (bottom) receptors.
Cells were fixed, and images from a field were acquired for 800 ms without binning with a ϫ100 objective in the CFP (left), YFP (middle), and FRET (right) channels. Images were corrected for background fluorescence and represented in pseudocolor. Images from wild-type cells show that CFP, but not YFP, excitation wavelength gives rise to intracellular autofluorescence of the highest intensity in the FRET channel. Images from cells transfected with CFP-tagged or YFP-tagged receptors show (i) specific labeling of the membrane at their respective excitation wavelength, (ii) no signal in the other one, except autofluorescence in the CFP channel for MOP-YFP cells, and (iii) bleed-through of fluorophores in the FRET channel. Quantification of the membrane fluorescence intensity was performed on 10 -15 images acquired in these conditions (four experiments) to correct FRET images for autofluorescence and contribution of fluorophores in FRET experiments (see details in the "Experimental Procedures").
pressed MOP receptors in SH-SY5Y cells (39) 3 and indicates that MOP receptor is highly glycosylated in SH-SY5Y. In contrast, no band could be specifically detected in extracts from NPFF 2 -CFP cells that express 10-to 20-fold less tagged receptors than MOP-YFP cells.
After immunoprecipitation with mouse monoclonal anti-GFP antibody, polyclonal anti-GFP antibodies selectively detected bands in extracts from cells expressing tagged receptors, but not in SH 2 -D9 cells (Fig. 7B, left blot). The major band in the MOP-YFP lane corresponds to that observed in solubilized extract from (SH 2 -D9)MOP-YFP cells (Fig. 7A), and the band above 72 kDa in the NPFF 2 -CFP lane corresponds to the expected size of the fusion of NPFF 2 receptor (49 kDa) and GFP (27 kDa). This indicates that monoclonal anti-GFP antibodies were able to specifically precipitate GFP-tagged receptors. Immunodetection with anti-MOP receptor antibodies in the immunoprecipitates revealed bands only in cells expressing NPFF 2 -CFP receptors but not in cells expressing wild-type NPFF 2 receptors (Fig. 7B, right). The apparent molecular mass of the band around 90 kDa corresponds to the estimated size of MOP receptor expressed in SH-SY5Y (83-93 kDa), the other bands being not always observed. This result indicates that MOP receptor is coimmunoprecipitated with NPFF 2 -CFP receptors, suggesting the existence of heteromeric complexes in membranes of doubly transfected cells. We failed to determine if these complexes were increased after treatment with 1DMe, because variation in the yield of coimmunoprecipitation from one experiment to another prevented the necessary accuracy for quantitative analysis.
Modulation of the MOP Receptor Diffusion in the Plasma Membrane by 1DMe-Because receptor interaction was strongly suggested by FRET and coimmunoprecipitation studies, we investigated whether the stimulation of NPFF receptors could modulate the lateral mobility of MOP receptors at the cell surface. FRAP experiments were conducted at different illumination areas with radii (R) ranging from 1.4 to 3.45 m on membranes of (SH 2 -D9)MOP-YFP cells that express wild-type NPFF 2 receptors in addition to tagged MOP receptors. The apparent diffusion coefficient (D app ) and the mobile fraction (M) were deduced from the fluorescence recovery curves recorded for each radius on cells treated or not treated by 1 M 1DMe (Fig. 8A).
For (SH 2 -D9)MOP-YFP cells in control conditions, the mobile fraction linearly increased with 1/R, giving a straight line that intercepted the y axis at M p ϭ 46 Ϯ 8% (Fig. 8B). According to Equation 2 ("Experimental Procedures"), this indicates that ϳ54% of MOP receptors are compartmentalized in areas with a mean radius of 1.03 Ϯ 0.26 m, as deduced from the slope. Their real diffusion coefficient in these domains, deduced from Equation 3, is estimated to be between 0.02 and 0.07 m 2 /s. The other fraction (46%) of receptors freely diffuses either over long distances between closed domains, or through domains with partial permeability. These results are in accordance with a recent report showing that GFP-tagged MOP receptors expressed in SH-SY5Y diffuse (D real around 0.03 m 2 /s) in a restricted area of 0.7 m radius, with a 55% permanent mobile fraction (40).
In contrast, when cells were treated with 1DMe, neither the mobile fraction (Fig. 8B) nor the apparent diffusion coefficient varied with the observation radius, indicating that, whatever the bleached area, ϳ80% of the receptors were freely moving on the cell surface with D app ϭ D real ϭ 0.2 m 2 /s. ϳ20% of the fluorescence was never recovered, suggesting either compartmentalization of receptors in very small domains with a radius smaller than the observation radius or a total immobilization of receptors.
To make sure that the effect of 1DMe was not due to a nonspecific rearrangement of the membrane structure, FRAP experiments were also performed in the GFP-MOP SH-SY5Y cell line (40), which does not express NPFF receptors. In these conditions, the mobile fraction of GFP-tagged MOP receptors at the 3.45-m observation radius was not significantly different (p Ͼ 0.05, unpaired Student's t test) in control (M p ϭ 66.10 Ϯ 1.72%, n ϭ 3) and 1DMe-treated cells (M p ϭ 62.60 Ϯ 2.63%, n ϭ 3), whereas it was significantly different in (SH 2 -D9)MOP-YFP cells (M p ϭ 56.23 Ϯ 4.15% in control versus 82.43 Ϯ 5.48% in 1DMe-treated cells, n ϭ 3). Therefore, these data clearly demonstrate that 1DMe modulates the lateral mobility of opioid receptors by a mechanism that requires the presence of NPFF receptors.

DISCUSSION
Heteromerization is considered to be a general characteristic of GPCR functioning and is now being targeted in new drugscreening strategies (41). Depending on the type of receptors, 3 L. Moulé dous, personal communication.

FIGURE 5. Detection of FRET between NPFF 2 -CFP and MOP-YFP receptors.
Example of images, acquired as in Fig. 4, from NPFF 2 -CFP/MOP-YFP cells incubated for 10 min at room temperature in KRH buffer (Control) or in KRH buffer with 1 M NPFF agonist (1DMe) or 1 M opioid agonist (DAMGO), before fixation. FRETc corresponds to images corrected for background, autofluorescence, and contribution of CFP and YFP bleed-through, as described under "Experimental Procedures," and displayed using quantitative pseudocolor scale. FRET is detected at the membrane of control cells and is increased after incubation with the NPFF agonist. In contrast, no FRET is detectable at the membrane of cells treated with the opioid agonist. dimerization is essential for functional activity (GABA B receptors (42)) or provides new pharmacological entities with enhanced, or reduced, efficiency (17). Heteromeric complexes may either pre-exist at the cell surface or are formed or disrupted, upon agonist stimulation (36). With regard to the opioid field, MOP receptors have been found associated with ␦-opioid, nociceptin, somatostatin (sst 2A ), substance P (NK1), chemokine (CCR5), adrenergic (␣ 2 ), and cannabinoid (CB 1 ) receptors by using biochemical and/or bioluminescence techniques (43)(44)(45)(46)(47)(48)(49)(50)(51). In the particular case of MOP/␦-opioid assembly, allosteric modulation by agonists improves the binding and functional activity of the heteromer, revealing important consequences in terms of analgesia, because low doses of ␦ agonists or antagonists increase binding and signaling of -selective agonists (44). Our present work provides biochemical and biophysical evidence for a physical interaction between MOP and NPFF 2 receptors as a molecular basis for the anti-opioid activity of NPFF agonists in neurons.
Using FRET, we demonstrate that the NPFF analog 1DMe promotes NPFF 2 ⅐MOP receptor association, whereas the opioid agonist DAMGO disrupts it. The constitutive association suggested by basal FRET is strengthened by immunoprecipitation experiments showing that both receptors are present in interacting complexes. 1DMe gradually increases FRET from 10 to 15 min, which corresponds to the minimal duration of preincubation necessary to observe anti-opioid activity in response to NPFF agonist, suggesting that both phenomena are correlated in time (12)(13)(14)(15). Similar kinetics (maximal effect from 5 to 15 min) for FRET or BRET signals upon agonist stimulation have been reported for other receptors (36), indicating that induction of dimeric complexes by agonist is a dynamic process. On the other hand, the reduction of FRET by DAMGO is also time-dependent and could be related to the concomitant opioid-induced receptor endocytosis that removes the acceptor partner in the FRET couple. This has been reported for endothelin ET A / ET B heterodimers (52). A return to the basal FRET level after 15 min could be due to the recycling of receptor in the time course of the internalization process. Therefore, NPFF 2 ⅐MOP receptor heteromeric complexes exist at the basal level and are differently modulated by NPFF and opioid agonists. This correlates with the fact that NPFF agonists impair the opioid response, whereas opioid agonists are ineffective on NPFF receptor activation in SH-SY5Y cells (14,15). An inhibition of opioid response by pretreatment or simultaneous addition of the heterologous agonist is also described in cells cotransfected with MOP receptors and ␣ 2 adrenergic (50), nociceptin (45), or cannabinoid (CB 1 ) (51) receptors, but, in contrast to the NPFF-MOP interaction, there is not yet evidence that it occurs in isolated neurons. Moreover, unlike the MOP⅐NPFF 2 receptor association, ␣ 2 adrenergic and CB 1 cannabinoid receptors are reciprocally inhibited by opioid agonists, indicating cross-desensitization in these cases and suggesting a different mechanism for NPFF receptors.
1DMe also modulates the trafficking of MOP receptors by reducing (40%) their internalization, likely as a consequence of receptor heteromerization (18). Much data show that receptors in heteromers are co-internalized upon stimulation with one of the two agonists, and this is generally associated with crossphosphorylation and cross-desensitization, as for MOP⅐NK1, MOP⅐sst 2A , MOP⅐CCR5 receptor heteromers (46 -48). But it is also described that the receptor showing the slowest endocytosis prevails in the heteromer and acts as a dominant-negative for internalization. This has been demonstrated for the endothelin ET B receptor whose sequestration mediated by the nonselective endothelin-1 agonist is delayed when complexed to ET A receptor (52). Similarly, the non-internalizing ␤ 3 -adrenergic receptor blocks the recruitment of ␤-arrestin by the ␤ 2  receptor and the subsequent endocytosis of ␤ 2 /␤ 3 heterodimer (53). Such a mechanism could occur in the case of the inhibition of MOP receptor internalization by 1DMe. Although we have no quantitative data on the trafficking of NPFF 2 receptors, essentially because the CFP signal is too weak for an accurate quantification under our conditions of confocal microscopy, a clear pattern of NPFF 2 receptor endocytosis was not observed after 30-min incubation with NPFF or opioid agonists, suggesting that NPFF receptors do not undergo internalization upon stimulation and could therefore act as a dominant-negative for MOP receptor endocytosis. Whether the blockade of opioid receptor endocytosis is the direct consequence of the physical interaction between each protomer or results from a reduced opioid receptor activity remains to be addressed.
In heteromeric complexes, the modulation of one receptor response does not necessarily require activation of the other. By using AT 2 receptors with impaired binding or functional activity, Abdalla et al. (19) have demonstrated that the reduction of the AT 1 receptor-mediated signaling in the AT 2 /AT 1 receptor dimer is only produced by the presence of AT 2 receptor. Conversely, the presence of adrenergic ␣ 2 receptor is sufficient to improve the efficacy of MOP receptors in the ␣ 2 ⅐MOP heteromer (50). The ␤ 3 -adrenergic receptor in complex with ␤ 2 blocks ␤ 2 internalization without being activated (53). In the case of NPFF receptors, binding of agonist, not necessarily with activation of the receptor, could be sufficient to produce anti-opioid activity, because (i) in neurons isolated from rat dorsal raphe and periventricular nuclei, NPFF agonists inhibit the effect of nociceptin on the voltage-gated Ca 2ϩ channel without producing measurable activity by themselves (12,13) and (ii) antiopioid activity of NPFF 1 receptors in SH-SY5Y cells is observed at concentrations of agonist 10-fold lower than that necessary to produce direct modulation of Ca 2ϩ signaling (15).
A modification of the diffusion properties of MOP receptors could be proposed as another consequence of the physical interaction with NPFF receptors. Lateral diffusion of proteins in the membrane has been studied extensively, especially in the context of the organization of the neuronal synapse, where it appears to play a key role in the regulation of synapse function during resting conditions or in plasticity (54,55). Receptors can diffuse over long distances or within constrained domains and can be transiently trapped at specific loci by factors such as scaffolding proteins, obstacles, lipid organization, attractive potentials, etc. that restrict their movement. This has led to the notion that membranes are compartmentalized in microdomains, the nature of which is still debated. Along this line, some but not all GPCR are thought to be present together with signaling molecules in structures enriched in saturated lipids and cholesterol called rafts or detergent-resistant membrane. In this regard, analysis of the lateral mobility of MOP receptors by FRAP shows that 54% of receptors at the surface of SH-SY5Y cells diffuse in restricted domains of ϳ1-m radius, with a diffusion coefficient in the range (0.01-0.1 m 2 /s) described for GPCR or other proteins in membranes (55). This result is consistent with previous studies of single-particle tracking (56) or FRAP experiments (40) in cells expressing MOP receptors. Upon application of the NPFF agonist 1DMe, nearly all MOP receptors become freely moving with a 10-fold increased diffusion coefficient, suggesting that activation of NPFF 2 receptors removes MOP receptors from their domain anchoring. This effect is in total opposition with that of DAMGO, which increases MOP receptor compartmentalization in SH-SY5Y cells (40) 4 and with that of agonists in general. Stimulation by agonist slows down the diffusion of chemokine CXCR1 (57) and luteinizing hormone receptors (58), even leading to immobilization in the case of neurokinin NK2 receptors (26), and/or accentuates receptor confinement by increasing the immobile fraction (57,58) or by reducing the size of the diffusion domain (26). In contrast, the effect of 1DMe rather looks like the action of an antagonist, which produces no effect in the case of serotonin receptors (59) or increases the mobile fraction in the case of N-formyl peptide receptors (60), suggesting therefore that the functional blockade and the mobility change of opioid receptors by NPFF could be linked.
To our knowledge, the study presented here provides the first report showing the influence of a GPCR on the mobility of another one, although this type of interaction has been recently described for the dopaminergic D1 receptor and the N-methyl-D-aspartic acid receptor channel (27). In this study, by measuring FRAP in organotypic culture from rat striatum transfected with fluorescent D1 receptors, the authors have clearly demonstrated that D1 receptors, which laterally diffuse in den- dritic membranes, are recruited in spines by a diffusion trap mechanism upon 30-s exposure to N-methyl-D-aspartic acid. This does not require activation of the channel but only allosteric modulation by the ligand. By analogy, we could postulate that agonist occupation of NPFF receptors reduces the opioid response by moving opioid receptors out of their signaling platform, or at least, by decreasing their affinity for effectors or scaffolding partners. In this sense, a close association of opioid receptors with transduction effectors is supported by the fact that nociceptin and MOP receptors have been found localized and active within rafts together with G-proteins (39,61) and that voltage-dependent inhibition of N-type Ca 2ϩ channels by opioids is mediated by a membrane-delimited pathway involving a direct interaction of the ␤␥ subunits of the heterotrimeric G protein on the ␣1 subunit of the channel (62,63). Therefore, by increasing the mobility of MOP receptors, NPFF could weaken their association with G-protein and channel subunits, resulting in a reduced inhibition of the Ca 2ϩ channel by the opioid agonist, i.e. an anti-opioid effect.
In conclusion, the present study shows that a physical interaction between NPFF 2 and MOP receptors could explain the anti-opioid activity of NPFF 2 receptors. By promoting a heteromeric association with MOP receptors, NPFF agonists change the lateral diffusion of MOP receptors from a constrained to a free state, likely moving them away from their signaling partners. As a consequence, the response to opioids is reduced. The modulation of the delivery and trafficking of ␦-opioid receptors at the cell surface is assumed to be a means for regulating MOP receptor function, hence opioid analgesia and tolerance (64). Likewise, the molecular mechanism described here for NPFF, comparable to the diffusion-trap system of the synapse, could represent another way to modulate the opioid response.