Previously we correlated the efficacy for G protein activation with that for arrestin recruitment for a number of agonists at the μ-opioid receptor (MOPr) stably expressed in HEK293 cells. We suggested that the endomorphins (endomorphin-1 and -2) might be biased toward arrestin recruitment. In the present study, we investigated this phenomenon in more detail for endomorphin-2, using endogenous MOPr in rat brain as well as MOPr stably expressed in HEK293 cells. For MOPr in neurons in brainstem locus ceruleus slices, the peptide agonists [d-Ala2,N-Me-Phe4,Gly5-ol]-enkephalin (DAMGO) and endomorphin-2 activated inwardly rectifying K+ current in a concentration-dependent manner. Analysis of these responses with the operational model of pharmacological agonism confirmed that endomorphin-2 had a much lower operational efficacy for G protein-mediated responses than did DAMGO at native MOPr in mature neurons. However, endomorphin-2 induced faster desensitization of the K+ current than did DAMGO. In addition, in HEK293 cells stably expressing MOPr, the ability of endomorphin-2 to induce phosphorylation of Ser375 in the COOH terminus of the receptor, to induce association of arrestin with the receptor, and to induce cell surface loss of receptors was much more efficient than would be predicted from its efficacy for G protein-mediated signaling. Together, these results indicate that endomorphin-2 is an arrestin-biased agonist at MOPr and the reason for this is likely to be the ability of endomorphin-2 to induce greater phosphorylation of MOPr than would be expected from its ability to activate MOPr and to induce activation of G proteins.
Currently there is much interest in the phenomenon of biased agonism, whereby different agonists at a G protein-coupled receptor (GPCR) can induce the receptor to couple to distinct downstream signaling pathways (Kenakin, 2011; Reiter et al., 2012). The most likely explanation for biased agonism is that different agonists stabilize distinct active conformations of the GPCR (Kahsai et al., 2011), which couple differentially to downstream signaling pathways. One commonly observed form of biased agonism is that between G protein-dependent and arrestin-dependent signaling (Reiter et al., 2012), although there are likely to be many other variations, such as bias in GPCR coupling to different G protein subtypes. The importance of biased agonism is that ligands that could selectively activate certain downstream signaling pathways might be developed, which has the potential to improve the therapeutic potential to manage diseases and to avoid adverse effects.
Agonists at the μ-opioid receptor (MOPr) are extremely important drugs for the management of moderate to severe pain (Corbett et al., 2006), but use of these drugs often leads to undesirable effects, including respiratory depression, constipation, and tolerance, and there is the potential for abuse (Morgan and Christie, 2011). Therefore, there is a need to develop new analgesic drugs with fewer of the unwanted effects associated with classic opioids such as morphine (Corbett et al., 2006).
In a recent study, we investigated the ability of 22 opioid agonists to activate G proteins and to recruit arrestin-3 in HEK293 cells stably expressing MOPr (McPherson et al., 2010). We showed that there was a strong correlation between the two signaling outputs for most of those agonists; however, there seemed to be bias toward arrestin recruitment for a few agonists, particularly the endomorphins. Because arrestins are implicated in signaling as well as GPCR regulation (Shenoy and Lefkowitz, 2011), this finding might have important consequences for the ability of new ligands based on the endomorphin structure to induce a different array of responses, compared with older, morphine-like drugs.
Endomorphin-1 (Tyr-Pro-Trp-Phe-NH2) and endomorphin-2 (Tyr-Pro-Phe-Phe-NH2) are opioid peptides with high affinity and selectivity for the MOPr (Zadina et al., 1997). Although the peptides were originally identified in extracts from mammalian brain (Zadina et al., 1997), investigators' inability to identify a precursor protein for either of these peptides has raised the question of whether they function as endogenous opioids in the mammalian central nervous system (Corbett et al., 2006; Terskiy et al., 2007). Analogs of these peptides have potential as novel analgesic agents, and one such drug, Cyt-1010, is reported to be in development (http://www.cytogelpharma.com/news.html).
The purpose of this study was to determine whether endomorphin-2 is a biased agonist at MOPr. Our data indicate that endomorphin-2 is biased toward arrestin recruitment over G protein activation. Furthermore, we show that the likely source for the arrestin bias is the ability of endomorphin-2 to induce greater phosphorylation of MOPr than would be predicted from the ability of this peptide to activate G protein-coupled responses.
Materials and Methods
Morphine hydrochloride was obtained from Mcfarlan Smith (Edinburgh, UK), etorphine from Research Triangle Institute (Research Triangle Park, NC), and DAMGO from Bachem (Bubendorf, Switzerland). Noradrenaline, β-funaltrexamine (β-FNA), prazosin, and cocaine were obtained from Sigma-Aldrich (Gillingham, UK). Endomorphin-2 and chelerythrine were obtained from Tocris Bioscience (Bristol, UK).
Brain Slice Preparation.
Male Wistar rats (130–170 g) were killed through cervical dislocation, and horizontal brain slices (200–250 μm thick) containing the locus ceruleus (LC) were prepared as described previously (Bailey et al., 2003). All experiments were performed in accordance with the UK Animals (Scientific Procedures) Act of 1986, the European Communities Council Directive of 1986 (86/609/EEC), and the University of Bristol ethical review documents.
Whole-Cell Patch-Clamp Recordings.
Slices were submerged in a slice chamber (0.5 ml) mounted on a microscope stage and were superfused (2.5–3 ml/min) with artificial cerebrospinal fluid composed of 126 mM NaCl, 2.5 mM KCl, 1.2 mM MgCl2, 2.4 mM CaCl2, 1.2 mM NaH2PO4, 11.1 mM d-glucose, 21.4 mM NaHCO3, and 0.1 mM ascorbic acid, saturated with 95% O2/5% CO2 at 33–34°C. For patch-clamp recordings, LC neurons were observed with Nomarski optics by using infrared light and individual cell somata were cleaned with a gentle flow of artificial cerebrospinal fluid from a pipette. Whole-cell voltage-clamp recordings (Vh = −60 mV) were made by using electrodes (3–6 MΩ) filled with a solution containing 115 mM potassium gluconate, 10 mM HEPES, 11 mM EGTA, 2 mM MgCl2, 10 mM NaCl, 2 mM MgATP, and 0.25 mM Na2GTP (pH 7.3; osmolarity, 270 mOsM). Recordings of whole-cell currents were filtered at 2 kHz by using an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA) and were analyzed off-line by using pClamp (Molecular Devices). Activation of MOPr evoked a transmembrane K+ current and, with the use of whole-cell patch-clamp recordings, a real-time index of MOPr activation could be recorded continually. The opioid-evoked current was continuously recorded at a holding potential of −60 mV. MOPrs and α2-adrenoceptors couple to the same set of K+ channels in LC neurons (North and Williams, 1985). To reduce variations between cells, the amplitudes of opioid-evoked currents were normalized to the maximal α2-adrenoceptor-mediated current in the same cell evoked by 100 μM noradrenaline (NA) applied in the presence of 1 μM prazosin and 3 μM cocaine.
All drugs were applied in the superfusing solution at known concentrations. Concentration-response curves for MOPr agonists were obtained through cumulative addition. To reduce the influence of desensitization on the slopes and maxima of the curves, each concentration of drug was added for only 2 min, by which time the response had reached a steady state. Each individual cell was exposed to a limited number of high concentrations (greater than the EC50) of the drug and to one supramaximal concentration. The maximal response for each drug obtained in this way was not different from the maximal response observed in other cells exposed to a single supramaximal concentration of that drug. For example, the amplitude of the maximum of the cumulative concentration-response curve for etorphine was 123.0 ± 11.1% (n = 3) of the response to 100 μM NA, whereas that evoked by a single, maximally effective concentration of etorphine (1 μM) was 142.4 ± 14.3% (n = 4) of the response to 100 μM NA (unpaired t test, p = 0.36).
HEK293 cells were maintained at 37°C, in 95% O2/5% CO2, in Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum, 10 U/ml penicillin, and 10 mg/ml streptomycin. In addition, the culture medium for the HEK293 cells stably expressing MOPr tagged at the NH2 terminus with HA contained 250 μg/ml G-418 (Geneticin; PAA Laboratories GmbH, Pasching, Austria).
Agonist-induced phosphorylation of MOPr at Ser375 was assessed by using an IN Cell Analyzer 1000 high-content imaging platform (GE Healthcare, Chalfont St. Giles, Buckinghamshire, UK), as described previously (Caunt et al., 2010), with minor variations. Cells cultured in 96-well plates were incubated with different concentrations of various MOPr agonists for 10 min at 37°C. Cells were then fixed, permeabilized, and immunostained with rabbit anti-phospho-Ser375 polyclonal antibody (1:1000 dilution; Cell Signaling Technology, Danvers, MA), followed by incubation with Alexa Fluor 488-conjugated goat anti-rabbit antibody (Invitrogen) and 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI) nuclear stain (600 nM). Images were acquired and then analyzed with IN Cell Investigator software (Workstation 3.5; GE Healthcare), with a dual area object analysis algorithm. Fluorescence from DAPI staining was used to define the nuclear area and thus the presence of a cell. The intensities of the Alexa 488 fluorescence in the whole-cell areas of the observed cells were averaged, and these values were used for further analysis. Alexa 488 fluorescence intensity values, which indicated phosphorylation of Ser375 at the MOPr, underwent background subtraction and then were normalized to the values obtained with phosphate-buffered saline (basal) and with 100 μM DAMGO.
For Western blotting, cells were washed three times with ice-cold phosphate-buffered saline after agonist treatment and were lysed in immunoprecipitation buffer (50 mM Tris, pH 7.5, 10 mM EDTA, 150 mM NaCl, 0.5% deoxycholate, 0.1% SDS, 1% Nonidet P-40, 50 mM NaF, 10 mM glycerol-2-phosphate, 200 μΜ sodium orthovanadate, 25 mM sodium pyrophosphate, and 1× complete mini-protease inhibitor; Roche Diagnostics, Indianapolis, IN). Cell lysates were clarified through centrifugation at 12,000 rpm at 4°C in a microcentrifuge and underwent immunoprecipitation with anti-HA antibody (HA-11; 3 μg per sample; Covance Research Products, Princeton, NJ) and protein G/A-agarose overnight at 4°C. Beads were washed three times with immunoprecipitation buffer, and proteins were eluted with the addition of SDS-sample buffer for 3 min at 95°C. Proteins were then resolved through 8% SDS-polyacrylamide gel electrophoresis and were transferred to polyvinylidene difluoride membranes. Membranes were incubated with the anti-phospho-Ser375 polyclonal antibody (1:1000). Blots were stripped and reprobed with anti-HA antibody (HA-11; 1:1000). Signal detection was performed with enhanced chemiluminescence methods.
These experiments were performed exactly as described previously (McPherson et al., 2010). HEK293 cells cotransfected with MOPr-yellow fluorescent protein (YFP), arrestin-3-cyan fluorescent protein (CFP), and G protein-coupled receptor kinase 2 (GRK2) (this increased the agonist-induced FRET signal obtained) were plated on poly-l-lysine-coated glass coverslips. Cells were mounted on a Nikon Eclipse TE2000S inverted microscope (Nikon, Melville, NY) and were observed by using an oil-immersion 63× lens, a Polychrome V monochromator for excitation, and a dual-emission photometric system (TILL Photonics, Gräfelfing, Germany). MOPr agonists were applied by using a computer-assisted superfusion system. Fluorescence was measured at 535 ± 15 nm (FYFP) and 480 ± 20 nm (FCFP) after excitation at 436 ± 10 nm. Signals detected by avalanche photodiodes were digitized by using an analog/digital converter (Digidata 1322A; Molecular Devices), and data were stored on a personal computer with Axoscope 9.2 (Molecular Devices). FRET was calculated as the FYFP/FCFP ratio. Off-line analyses of the FRET (FYFP/FCFP) data were performed to determine the kinetics and extent of the MOPr-YFP and arrestin-3-CFP interactions induced by different agonists.
Cell Surface Receptor Loss.
Receptor loss was assessed with enzyme-linked immunosorbent assays, as described previously (McPherson et al., 2010). HEK293 cells stably expressing HA-tagged MOPr were seeded into 24-well tissue culture dishes coated with 0.1 mg/ml poly-l-lysine, 24 h before experimentation. For time course experiments, cells were washed and then challenged for 0 to 30 min at 37°C with Dulbecco's modified Eagle's medium containing opioid agonist. Reactions were terminated by fixing the cells with 3.7% formaldehyde. Cells were then incubated with primary antibody (anti-HA-11, 1:1000) for 1 h at room temperature. For investigations of the agonist concentration dependence of internalization, cells were prelabeled with primary antibody for 1 h at 4°C before incubation with the agonists for 30 min at 37°C. Cells were then incubated with secondary antibody (goat anti-mouse antibody conjugated with alkaline phosphatase, 1:1000; Sigma-Aldrich), a colorimetric alkaline phosphatase substrate (Bio-Rad Laboratories, Hemel Hempstead, UK) was added, and samples were assayed at 405 nm in a microplate reader. Changes in surface receptor expression were determined by normalizing data from each treatment group to corresponding control surface receptor levels determined from cells not exposed to opioid agonist, and values were expressed as either percentage of surface receptor expression or cell surface loss (as percentage of no-drug control), depending on the primary antibody labeling method used, with the background signal from HEK293 cells being subtracted from all receptor-transfected cell values. All experiments were performed in triplicate.
All data were analyzed by using GraphPad Prism (GraphPad Software Inc., San Diego, CA). Agonist concentration-response data from LC neurons were fitted to sigmoidal curves with variable slopes, with the bottom of the curve in each case being constrained to 0, for determination of agonist EC50 and Emax values and for graphical representation of the data. Parameters from this fitting were then used as initial values for fitting of the concentration-response data to the operational model of pharmacological agonism (Black and Leff, 1983; Bailey et al., 2009; McPherson et al., 2010), E = Emτn[A]n/(Ka + [A])n + τn[A]n, where the response E is expressed in terms of the molar concentration of agonist A, the theoretical maximal effect Em (greater than that which can be functionally attained) (Black and Leff, 1983), the equilibrium dissociation constant Ka, the transducer ratio τ, and n, which is the slope of the curve (n in this equation is not the same as the Hill slope, although the values may be similar). Em and n are intrinsic properties of the receptor/cell and are independent of the agonist used, whereas τ depends on the cell, receptor function, and agonist used. For each pair of curves (i.e., concentration-response curves for DAMGO, etorphine, and endomorphin-2 in the presence and absence of 30 nM β-funaltrexamine), values for Em, Ka, and n (shared for the paired curves for each agonist) and for τ were determined.
For graphical representation of Ser375 phosphorylation, concentration-response data were fitted to sigmoidal curves with variable slopes, with the bottom of the curve in each case being constrained to 0. Parameters from this fitting were then used as initial values for fitting of the concentration-response data to the operational model of pharmacological agonism (see above). For operational model fitting, constrained parameters were Em shared and <101, n shared and <2.0, and Ka values determined previously (McPherson et al., 2010), which were as follows: DAMGO, 228 nM; etorphine, 3.5 nM; endomorphin-2, 283 nM; morphine, 250 nM.
Ser375 phosphorylation data were also analyzed for efficacy values with the method described previously (Ehlert, 1985), e = (Emax,agonist/Emax,full agonist) × [(Ka,agonist/EC50,agonist) + 1] × 0.5, where e is the efficacy of the test agonist, Emax,agonist and Emax,full agonist are the relative maximal response values of the test agonist and an agonist yielding a full response, respectively, and Ka and EC50 are the equilibrium dissociation constant and EC50, respectively, for the test agonist. The efficacy values for agonist-induced cell surface loss were also determined with this method.
For occupancy-response relationships, data from McPherson et al., 2010 were reanalyzed to calculate fractional receptor occupancy at each concentration of agonist used in [35S]GTPγS or arrestin-3 recruitment assays. The fractional receptor occupancy was calculated by using the expression p = [D]/(Ka + [D]), where p is the fractional receptor occupancy (between 0 and 1), [D] is the agonist concentration, and Ka is the equilibrium dissociation constant, the value of which was determined previously in membranes of these cells (McPherson et al., 2010).
For desensitization in LC neurons, the desensitization-phase data (up to 10 min) from individual experiments were combined and fitted to a one-phase, exponential-decay model to obtain values for t1/2 and the rate constant k for decay, as well as maximal desensitization (plateau level). For analysis of FRET data, the t1/2 of the MOPr-YFP/arrestin-3-CFP interaction was obtained by fitting the data from the time points at which the agonist had been applied to a one-phase, exponential-association model. The extent of the agonist-induced MOPr-YFP and arrestin-3-CFP interaction was calculated as the peak of the interaction and was normalized to the interaction induced by the subsequent application of 10 μM DAMGO.
For ligand bias calculations, the method described by Rajagopal et al. (2011) was used with previously generated data (McPherson et al., 2010). For each agonist with either G protein activation or arrestin-3 recruitment, the “effective signaling” (σlig) was calculated, where σlig = log(τlig/τref); τlig is the operational eficacy of a ligand for a particular signaling pathway, and τref is the operational efficacy for the reference agonist (assumed to be unbiased) for that pathway. In this case, the reference ligand was Leu-enkephalin (McPherson et al., 2010). The bias factor (βlig) for a particular ligand was then calculated as follows: βlig = (σligpath1 − σligpath2)/√2. Statistical differences were determined, where appropriate, with Student's t test, one-sample t test, or F test for comparison of different models (the simpler model was selected unless the extra sum-of-squares F test had a p value of <0.05), by using GraphPad Prism.
Relative Operational Efficacy of Agonists at Native MOPr in Mature Neurons.
Previously we reported that, in HEK293 cells stably expressing MOPr, endomorphin-1 and -2 may be biased toward arrestin recruitment over G protein coupling (McPherson et al., 2010). To investigate whether endomorphins show bias at endogenous MOPrs in mature neurons, we needed a reliable measure of MOPr agonist efficacy for a G protein-mediated response in LC neurons. Therefore, we compared the ability of several MOPr agonists to activate G protein-coupled, inwardly rectifying, K+ channel current (GIRK) in individual rat LC neurons. We constructed concentration-response curves for endomorphin-2, DAMGO, and etorphine, before and after exposure of the brain slices to the irreversible MOPr antagonist β-FNA (30 nM, 30 min). In the absence of β-FNA, all three agonists produced the same maximal responses, which indicates that, for this response in this tissue, endomorphin-2 is a full agonist (Fig. 1). Pretreatment with β-FNA shifted the agonist concentration-response curve for each of these agonists to the right and depressed the maximal response (Fig. 1). The calculated mean ± S.E.M. values for the sigmoidal curve fitting are listed in Supplemental Table 1.
The curves for each agonist in the absence and presence of β-FNA were then fitted to the equation describing the operational model of pharmacological agonism (Black and Leff, 1983) (see Materials and Methods), and values of operational efficacy (τ) were obtained. Relative efficacy values are given in Table 1, and calculated mean ± S.E.M. values are listed in Supplemental Table 2. In a previous study, we obtained a τ value of 1.6 for morphine in LC neurons (Bailey et al., 2009). We and others reported previously that the maximal response for morphine is lower than that for the other agonists (Osborne and Williams, 1995; Bailey et al., 2009), which confirms that it is a partial agonist and must have lower efficacy than the other three agonists for this response. The rank order of τ values (Table 1) obtained from LC neurons (DAMGO > etorphine ≫ endomorphin-2 > morphine) was similar to that obtained previously with these agonists for GTPγS binding in HEK293 cells (McPherson et al., 2010) (Table 1), which indicates that the relative operational efficacy of G protein-dependent signaling for these agonists is independent of the tissue in which the MOPr is expressed. These data demonstrated that, for MOPr-G protein coupling in both a heterologous cell line and mature neurons, endomorphin-2 had much lower efficacy than DAMGO.
Rate and Extent of Agonist-Induced Desensitization of MOPr-Activated K+ Current in LC Neurons.
We next assessed the ability of the MOPr agonists to induce acute desensitization of the GIRK current in LC neurons (Fig. 2). A receptor-saturating concentration of each agonist was applied for 10 min and the GIRK current was recorded. Representative traces for each agonist are shown in Fig. 2. It can be seen that DAMGO, etorphine, and endomorphin-2 induced extensive rapid desensitization of the GIRK current, whereas morphine induced less desensitization over the same period. The desensitization phase for each agonist was fitted to a one-phase, exponential-decay model to determine the rate of desensitization and the maximal amount of desensitization. The data are shown in Fig. 2 and Table 2. The fastest rate of desensitization was observed with endomorphin-2 (F test, p < 0.0001), with the order from faster to slower decay rate constant k values being as follows: endomorphin-2 (0.440 min−1) > DAMGO (0.233 min−1) > morphine (0.205 min−1) > etorphine (0.137 min−1). The ability of a number of other MOPr agonists to induce acute desensitization in LC cells was examined, but none induced faster desensitization than endomorphin-2 (Supplemental Fig. 1). These data indicate that, in rat brain neurons, the ability of endomorphin-2 to induce rapid desensitization is much greater than would be predicted from its efficacy for GIRK activation.
Agonist-Induced Phosphorylation of MOPr at Ser375.
MOPr agonists induce phosphorylation at Ser375 in the COOH terminus of the receptor, and an antiphosphoreceptor antibody was developed to identify this phosphorylation event (Schulz et al., 2004). To facilitate measurement of agonist-induced MOPr phosphorylation at Ser375, automated imaging of permeabilized HEK293 cells was undertaken (Fig. 3A). Initial experiments indicated that this Ser375 phosphorylation was mediated in part by GRK2, was independent of protein kinase C activation, and partly depended on coupling to G proteins (Supplemental Fig. 2). Concentration-response curves for phosphorylation of Ser375 in response to 10 min of agonist stimulation were constructed. DAMGO, endomorphin-2, and etorphine seemed to be full agonists in this assay, whereas morphine was a partial agonist (Fig. 3B; see Supplemental Table 3 for parameters for fitting of data to sigmoidal curves). Etorphine was the most potent agonist, followed by DAMGO and endomorphin-2 and then morphine. These agonist effects were confirmed with Western blotting of MOPr immunoprecipitated from HEK293 cells (Fig. 3C). Concentration-response data obtained from the automated imaging were then subjected to operational analysis with the binding constants for MOPr in HEK293 cells determined in our previous study (McPherson et al., 2010). Relative efficacy values are given in Table 1, and calculated mean ± S.E.M. values are listed in Supplemental Table 4. They indicate that, unlike the results from G protein coupling, the operational efficacy of endomorphin-2 for phosphorylation of Ser375 in MOPr was similar to that of etorphine and almost as high as that of DAMGO. The operational efficacy for morphine remained low in this assay. Because the operational analysis of these data produced τ values with large standard errors (Supplemental Table 4), the method described by Ehlert (1985) was used as an alternative approach to determine relative efficacy in this assay; this method uses a combination of EC50, maximal response, and binding constant values (for details, see Materials and Methods) to obtain a measure of efficacy (e). This produced a similar rank order of relative efficacy (Table 1), compared with the operational analysis. Whichever method was used to determine relative efficacy from these data, the efficacy of endomorphin-2 for Ser375 phosphorylation was close to that of DAMGO, although the latter had much higher efficacy for G protein activation.
Agonist-Induced Interaction of MOPr with Arrestin-3.
We used FRET assays to assess the MOPr/arrestin-3 interaction in intact cells after agonist application. HEK293 cells were transiently transfected with MOPr-YFP and arrestin-3-CFP; 48 h later, FRET was monitored for 3 to 5 min after the addition of DAMGO (10 μM), etorphine (10 μM), endomorphin-2 (30 μM), or morphine (30 μM). With the exception of etorphine, which did not wash out, in each case DAMGO (10 μM) was added after the first drug had been washed out, to obtain a relative measure of maximal response (Fig. 4A). Analysis of agonist-induced FRET showed that the t1/2 of the MOPr/arrestin-3 association-induced FRET was relatively low for DAMGO and endomorphin-2 and much higher for morphine (Fig. 4B). Furthermore, DAMGO and endomorphin-2 produced rapid increases in the FRET ratio that were of similar maximal amplitudes, whereas morphine produced a smaller increase in the FRET ratio, compared with the other agonists (Fig. 4C).
Agonist-Induced Cell Surface Loss of MOPr.
The ability of DAMGO, etorphine, endomorphin-2, and morphine to induce loss of cell surface MOPr was assessed with enzyme-linked immunosorbent assays. Both the agonist concentration and time dependence of cell surface loss were assessed (Fig. 5). The agonist concentration-response curves indicated that etorphine was the most potent agonist, followed by DAMGO and then morphine and endomorphin-2. Receptor-saturating concentrations of DAMGO, etorphine, and endomorphin-2, applied for 30 min, each induced >25% loss of cell surface MOPr (Fig. 5B); a saturating concentration of morphine induced less extensive loss of MOPr (<15%). The parameters for fitting the concentration-response data to sigmoidal curves are shown in Supplemental Table 5. It was not possible to fit the data in Fig. 5A with the operational model, even with a range of constraints; therefore, the relative efficacies were calculated with the method described by Ehlert (1985). In this analysis (Table 1), the rank order of relative efficacy (e) values was DAMGO > etorphine = endomorphin-2 > morphine. Together these results show that, although endomorphin-2 had an operational efficacy value for G protein coupling/K+ current activation much lower than that of DAMGO, it was able to induce cell surface loss of MOPr almost the same as that produced by DAMGO. Furthermore, although endomorphin-2 and morphine had comparable low values of operational efficacy for [35S]GTPγS binding in HEK293 cell studies, endomorphin-2 induced more extensive cell surface loss of MOPr than did morphine.
Comparison of Agonist Efficacy Values for Multiple Signaling Outputs.
For comparison of relative efficacy values from multiple signaling outputs, efficacy values for each response analyzed in this study are presented as a bar graph in Fig. 6, with the efficacy of DAMGO set as 1 in each case. The results show that the relative efficacy values of the agonists for GIRK activation and [35S]GTPγS binding (i.e., G protein responses) closely mirrored each other, with endomorphin-2 having low efficacy relative to DAMGO. In contrast, the efficacy of endomorphin-2 relative to DAMGO was much higher for phosphorylation of Ser375, arrestin-3 recruitment, and cell surface loss. For Ser375 phosphorylation and arrestin-3 recruitment, the efficacy of endomorphin-2 was essentially the same as that of DAMGO.
Agonist Occupancy-Response Relationships.
For further examination of the agonist-induced responses to the four MOPr agonists, occupancy-response relationships were constructed for MOPr agonists by using previously generated data (McPherson et al., 2010) for [35S]GTPγS binding and arrestin-3 recruitment assays for receptors stably expressed in HEK293 cells (Fig. 7). Transformation of the data in this way enables examination of the agonist occupancy-response relationship without the complication of binding affinity. These results show that the efficacy order for coupling to G protein was DAMGO > etorphine > endomorphin-2 = morphine (Fig. 7A) (comparing agonist responses at occupancies of 0.25, for example), whereas that for arrestin-3 recruitment was DAMGO = etorphine = endomorphin-2 ≫ morphine (Fig. 7B). It is also clear that the relationship between fractional MOPr occupancy and arrestin-3 recruitment was essentially linear, which suggests a lack of amplification in this response. Of interest, analysis of Ser375 phosphorylation and internalization data suggested that the relationships between fractional receptor occupancy and these outputs were also linear (Supplemental Fig. 3). The occupancy-response relationship in Fig. 7B might suggest that it is morphine rather than endomorphin-2 that is unusual, because the arrestin response-fractional receptor occupancy relationship for morphine lies apart from those of the other three agonists. To investigate this, we used previously published data to construct occupancy-response relationships for other agonists that had similar operational efficacies in the GTPγS assay (McPherson et al., 2010), including morphine and endomorphin-2, as well as oxycodone, 6-monoacetylmorphine, and morphine-6-glucuronide. When the occupancy-response relationships for arrestin-3 recruitment were plotted for this group of agonists, it could be clearly seen that endomorphin-2 had much higher efficacy than any of the other agonists within this group (Supplemental Fig. 4). This strongly supports the conclusion that endomorphin-2 is an arrestin-biased ligand.
Calculation of Ligand Bias.
A method to quantify ligand bias has been described (Rajagopal et al., 2011). With this method, operational efficacy (τ) values are calculated for a series of agonists for two signaling outputs and are compared with the τ values for an agonist that is unbiased between the two signaling pathways being assessed (see Materials and Methods). On the basis of previous data, we concluded that Leu-enkephalin is an unbiased agonist (McPherson et al., 2010). By using this approach, we calculated bias for the MOPr ligands that yielded significant responses in the two assays, which enabled us to calculate bias for 16 of the 22 ligands investigated in our previous study. Calculation of ligand bias indicated that endomorphin-2 was significantly biased toward arrestin-3 recruitment (Fig. 8). Although no other ligands displayed statistically significant bias, it is worth noting that endomorphin-1, etorphine, and alfentanil displayed trends toward arrestin bias, whereas DAMGO displayed a trend toward G protein bias.
Previous studies on the endomorphins identified these peptides as selective MOPr agonists with relatively high affinity for the receptor (Zadina et al., 1997). The endomorphins were reported to be partial agonists at MOPr in some cases, such as in GTPγS binding assays with spinal cord and thalamus (Hosohata et al., 1998; Narita et al., 1998; Sim et al., 1998; Xie et al., 2008) and in inhibition of neuronal Ca2+ currents (Connor et al., 1999), whereas the endomorphins behaved as full agonists in other cases, such as inhibition of contraction of mouse vas deferens (Al-Khrasani et al., 2001; Rónai et al., 2006); such differences are likely to depend in part on the receptor reserve in each tissue. In the present study with LC neurons, we found endomorphin-2 to be a full agonist for activation of GIRK current. However, the receptor reserve for the endomorphin-2 response in LC neurons was less than those for DAMGO and etorphine, because limited receptor depletion with a low concentration (30 nM) of β-FNA led to a greater reduction in the maximal response to endomorphin-2 (22% reduction) than in those to DAMGO (7% reduction) and etorphine (15% reduction), as well as a smaller shift in agonist EC50 (2.6-fold for endomorphin-2, 5.6-fold for etorphine, and 11.5-fold for DAMGO). To quantify the efficacy of endomorphin-2 relative to DAMGO and other MOPr agonists, we used the MOPr inactivation method that we (Bailey et al., 2009) and others (Rónai et al., 2006; Madia et al., 2009) used previously to determine agonist efficacy. In this analysis, we determined that the operational efficacy of endomorphin-2 for activation of GIRK was only 7% of that of DAMGO, with etorphine having an intermediate efficacy. These relative values reflect those obtained for GTPγS binding in membranes of HEK293 cells stably expressing MOPr, where the efficacy value for endomorphin-2 was 17% of that for DAMGO, with etorphine again exhibiting an intermediate value (McPherson et al., 2010). Because the MOPr-mediated activation of GIRK in LC neurons is a G protein-mediated response, it is perhaps not surprising that the relative efficacies in the two assays are similar.
Previous studies (Yu et al., 1997; Virk and Williams, 2008) suggested that the ability of saturating concentrations of MOPr agonists to induce acute (0–10-min) desensitization is agonist efficacy-dependent. On this basis, endomorphin-2 would be predicted to induce relatively little desensitization of MOPr-mediated GIRK activation during this relatively short period of agonist exposure, because of its lower efficacy for this effect. However, endomorphin-2-induced desensitization was as extensive and occurred more rapidly than that induced by DAMGO. The desensitization induced by morphine and etorphine was slower than that induced by DAMGO and was in line with efficacy values determined in this study for LC neurons and for [35S]GTPγS binding assays with HEK293 cells (McPherson et al., 2010). The acute desensitization induced by endomorphin-2 may be partly attributable to the ability of this ligand to induce efficient arrestin recruitment to MOPr, leading to extensive uncoupling of receptor and G protein. This cannot be the full explanation, however, because the efficacy of endomorphin-2 for arrestin-3 recruitment in HEK293 cells was the same as those of DAMGO and etorphine (McPherson et al., 2010) but acute desensitization was faster with endomorphin-2. One possibility is that the low efficacy and consequent small receptor reserve of endomorphin-2 for GIRK activation, coupled with relatively high efficacy for arrestin recruitment, may make it particularly sensitive to the onset of acute desensitization. In contrast, DAMGO is an agonist with high efficacy for GIRK activation and consequently a large receptor reserve, which would display slower desensitization because, although it induces efficient arrestin recruitment to MOPr, the high efficacy for GIRK activation means that a much greater loss of functional receptors, compared with endomorphin-2, would be required to yield significant desensitization of the GIRK response.
Differences in arrestin-2 versus arrestin-3 recruitment have not been explored to date, and it is possible that the profile of arrestin-2 versus arrestin-3 recruitment to the endomorphin-2-occupied MOPr is different from that induced by other agonists, leading to agonist- and arrestin-dependent signaling and regulation that are distinct from those of other agonists. Such a scenario was described recently for arrestin isoform interactions with MOPr in response to DAMGO and morphine (Groer et al., 2011). The precise role of arrestins in MOPr regulation in neurons remains to be fully elucidated. Although MOPr desensitization by high-efficacy agonists such as DAMGO is arrestin-dependent in cell lines (Chu et al., 2008), the acute desensitization induced by Met-enkephalin was unaffected in LC neurons from arrestin-3-knockout mice (Dang et al., 2011; Quillinan et al., 2011). These studies suggest that arrestins play a complex role in MOPr function in neurons, with recovery from acute desensitization being much faster in neurons from arrestin-3-knockout mice (Dang et al., 2011; Quillinan et al., 2011). Given that MOPr dephosphorylation can occur at the cell surface (Arttamangkul et al., 2006; Doll et al., 2011), arrestin interaction and internalization might actually reduce the rate of MOPr resensitization, which might be manifested as an enhanced rate of desensitization, as observed in our LC neuron study.
In additional studies, we investigated the ability of the agonists DAMGO, etorphine, endomorphin-2, and morphine to induce phosophorylation of Ser375 in the COOH-terminal tail of MOPr, to recruit arrestin-3 to MOPr in intact cells through FRET, and to induce cell surface loss of MOPr. We reasoned that the ability of endomorphin-2 to recruit arrestin-3 effectively might be attributable to more-rapid or more-extensive MOPr phosphorylation at Ser375. Although we (McPherson et al., 2010) and others (Yu et al., 1997) previously compared the ability of receptor-saturating concentrations of agonist to promote MOPr phosphorylation, in the current study we were able to construct full concentration-response curves for Ser375 phosphorylation, which allowed estimates of relative agonist efficacy to be made. This revealed that the efficacy of endomorphin-2 to induce Ser375 phosphorylation in MOPr was similar to that of DAMGO and much higher than that of morphine. This suggests that it is actually the ability of endomorphin-2 to promote phosphorylation of MOPr efficiently that underlies its ability to recruit arrestin-3 to MOPr and also to induce MOPr internalization to a greater extent than morphine. Recent studies indicated that MOPr is phosphorylated on multiple residues in the COOH terminus (Doll et al., 2011; Lau et al., 2011) (S. Oldfield, A. Butcher, A. Tobin, G. Henderson, and E. Kelly, unpublished observations) and the pattern of phosphorylation is agonist-dependent. Ser375 is undoubtedly a key residue in terms of arrestin recruitment (Schulz et al., 2004) but other residues may be important for arrestin recruitment (Lau et al., 2011), and it will be of particular interest to determine the ability of the endomorphins to induce phosphorylation of individual residues in the COOH terminus of MOPr. This does not explain, however, why endomorphin-2 induces more-efficient MOPr phosphorylation than morphine. Given that biased agonism suggests that different agonists can stabilize distinct active conformations of a GPCR (Kahsai et al., 2011), it is possible that endomorphin-2 stabilizes a conformation of the MOPr that couples relatively poorly to G proteins but is readily phosphorylated by kinases such as GRK2 or possibly kinases distinct from those that phosphorylate the DAMGO- or morphine-occupied MOPr.
The distinction between endomorphin-induced MOPr signaling and trafficking can be deduced from careful analysis of previous studies, in which [35S]GTPγS binding assays revealed the endomorphins to have low efficacy values close to those of morphine (Hosohata et al., 1998; Narita et al., 1998; Sim et al., 1998; Xie et al., 2008) but the endomorphins were able to induce efficient trafficking of MOPr under conditions where morphine was ineffective (Burford et al., 1998; McConalogue et al., 1999; Trafton et al., 2000). The advantages of quantifying efficacy with full concentration-response curves and of examining occupancy-response relationships are that such differences can be observed more clearly and can be quantified. Previous studies often used saturating concentrations of agonist to compare efficacies, which is of limited use for comparison of the efficacies of full agonists. In addition, comparison of potencies and maximal responses (intrinsic activity) sometimes does not reveal biased agonism (Molinari et al., 2010). More-sophisticated approaches to quantifying ligand bias have been proposed (Rajagopal et al., 2011), and the application of one such approach to our data from HEK293 cells (Fig. 8) identified the arrestin bias of endomorphin-2 as statistically significant.
In conclusion, endomorphin-2 is an arrestin-biased MOPr agonist, and this may be explained by the drug's ability to promote efficient phosphorylation of the receptor. Future studies will be focused on determining whether the pattern of MOPr phosphorylation induced by endomorphin-2 is different from that of other agonists, such as DAMGO and morphine. Endomorphin-related ligands are reported to have favorable profiles in terms of analgesia and tolerance/dependence and seem to produce less respiratory depression than other agonists (J. E. Zadina, unpublished data). It remains to be seen whether such profiles are attributable to the behavior of the endomorphins as arrestin-biased ligands at MOPr.
Participated in research design: Mundell, McArdle, Rosethorne, Charlton, Krasel, Bailey, Henderson, and Kelly.
Conducted experiments: Rivero, Llorente, McPherson, and Cooke.
Contributed new reagents or analytic tools: McArdle and Krasel.
Performed data analysis: Rivero, Llorente, McPherson, Cooke, Mundell, McArdle, and Kelly.
Wrote or contributed to the writing of the manuscript: Mundell, McArdle, Rosethorne, Charlton, Krasel, Bailey, Henderson, and Kelly.
This work was supported by the National Institutes of Health National Institute on Drug Abuse [Grant DA020836]; the Medical Research Council UK [Grant G0600943]; the Biotechnology and Biochemical Sciences Research Council [Grant BB/D012902/1]; and by a research fellowship from the Basque Government [Grant BFI08.131] (to G.R.).
Article, publication date, and citation information can be found at http://molpharm.aspetjournals.org.
- G protein-coupled receptor
- μ-opioid receptor
- locus ceruleus
- G protein-coupled, inwardly rectifying, K+ channel current
- yellow fluorescent protein
- cyan fluorescent protein
- G protein-coupled receptor kinase 2
- 4′,6-diamidino-2-phenylindole dihydrochloride
- guanosine 5′-O-(γ-thiotriphosphate)
- fluorescence resonance energy transfer.
- Received March 15, 2012.
- Accepted May 2, 2012.
- Copyright © 2012 The American Society for Pharmacology and Experimental Therapeutics