Calmodulin Binding to G Protein-coupling Domain of Opioid Receptors*

The ubiquitous intracellular Ca2+ sensor calmodulin (CaM) regulates numerous proteins involved in cellular signaling of G protein-coupled receptors, but most known interactions between GPCRs and CaM occur downstream of the receptor. Using a sequence-based motif search, we have identified the third intracellular loop of the opioid receptor family as a possible direct contact point for interaction with CaM, in addition to its established role in G protein activation. Peptides derived from the third intracellular loop of the μ-opioid (OP3) receptor strongly bound CaM and were able to reduce binding interactions observed between CaM and immunopurified OP3 receptor. Functionally, CaM reduced basal and agonist-stimulated35S-labeled guanosine 5′-3-O-(thio)triphosphate incorporation, a measure of G protein activation, in membranes containing recombinant OP3receptor. Changes in CaM membrane levels as a result of overexpression or antisense CaM suppression inversely affected basal and agonist-induced G protein activation. The ability of CaM to abolish high affinity binding sites of an agonist at OP3 further supports the hypothesis of a direct interaction between CaM and opioid receptors. An OP3 receptor mutant with a Lys273→ Ala substitution (K273A-OP3), an amino acid predicted to play a critical role in CaM binding based on motif structure, was found to be unaffected by changes in CaM levels but coupled more efficiently to G proteins than the wild-type receptor. Stimulation of both the OP1 (δ-opioid) and OP3 wild-type receptors, but not the K273A-OP3 mutant, induced release of CaM from the plasma membrane. These results suggest that CaM directly competes with G proteins for binding to opioid receptors and that CaM may itself serve as an independent second messenger molecule that is released upon receptor stimulation.

tory protein found in virtually every class of living organism from plants to alga, from humans to protozoa (1). Originally identified as the Ca 2ϩ -dependent factor responsible for activating 3Ј,5Ј-cyclic-nucleotide phosphodiesterase (2)(3)(4)(5), the role of CaM as a regulator of cytoplasmic enzymes has broadened to include adenylate cyclases, Ca 2ϩ /CaM-dependent kinases and phosphatases, ion channels, Ca 2ϩ -ATPases, and others (1,6,7). Mapping relationships between the multitude of enzymes regulated by CaM is essential to the understanding of biochemical cascades linked to activation of G protein-coupled receptors (GPCRs).
While previous studies have revealed broad involvement of CaM in downstream signaling pathways initiated by G proteincoupled receptors, progress is also being made toward uncovering interactions more proximal to the receptor. Recent examples include the discovery of a CaM binding site on a G protein ␤-subunit (8) and another report that describes a CaM binding site on the metabotropic glutamate subtype 5 receptor (9). The CaM binding domain found on the metabotropic glutamate subtype 5 receptor is located in the extended C-terminal region and appears unique to this receptor. Whereas core regions of the receptor involved in G protein coupling have not been previously implicated in CaM interactions, circumstantial evidence suggests such an involvement of CaM at the receptor level. The ability of certain CaM-binding peptides, like mastoparan and mellitin, to activate G␣ as well, in essence substituting for the receptor (10 -12), leads to suspicion that GPCRs may contain core regions of structural similarities to CaMbinding domains.
In this report, we identify an i3 loop region within receptors of the opioid family, in particular OP 1 and OP 3 , as a potential site for CaM interaction. We also demonstrate that CaM inversely affects G protein coupling by OP 1 and OP 3 and propose that CaM itself could function as an independent second messenger molecule. 3 Receptor i3 Loop Peptides-Peptides corresponding to sections of the i3 loop of the OP 3 receptor (see Table I) were prepared on an Applied Biosystems 433A automated peptide synthesizer (Foster City, CA) using para-methylbenzylhydramine resin and standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid phase methods. Peptides were purified on a Dynamax SD-200 high pressure liquid chromatograph (Varian Chromatography Systems, Walnut Creek, CA) using a reversephase C-18 column and eluting across a gradient of 10 -50% acetonitrile in water containing 0.1% (v/v) trifluoroacetic acid.
Site-directed Mutagenesis of the hOP 3 Receptor-The FLAG epitope was attached to N terminus of the hOP 3 receptor (18) using Pfu DNA polymerase (Stratagene, La Jolla, CA) and the oligonucleotides 5Ј-GC-TCTAGAGCTTAGGGCAACGGAGCAGTTTC-3Ј and 5Ј-CCCAAGCTT-GGGATGGACTACAAGGACGATGATGACAAGGACAGCAGCGCTGC-CCCCACGAAC-3Ј. The amplified PCR product was subcloned into the pcDNA3 (Invitrogen, Carlsbad, CA), using HindIII and XbaI. Mutation of Lys 273 to Ala in the i3 loop region of the FLAG-hOP 3 receptor was accomplished using the QuikChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA) and confirmed by DNA sequencing (Biomolecular Resource Center, University of California San Francisco).
Gel Mobility Shift Assays-Gel mobility shift assays involving CaM and test peptides were performed using 12% nondenaturing polyacrylamide gel electrophoresis essentially as described by Liu et al. (19). Briefly, 60 M of purified CaM (Calbiochem) was incubated for 30 min at room temperature with test peptides dissolved in 10 mM HEPES, pH 7.0, containing 1 mM CaCl 2 or 2 mM EGTA. Proteins were visualized with Coomassie Brilliant Blue staining.
Measurement of Peptide-CaM Interactions Using Fluorescence Assays-Dansylation of CaM was performed as described by Kincaid et al. (20). Fluorescence emission spectra of dansylated CaM (dansyl-CaM) were measured from 400 to 600 nm using a Fluorolog Series 2 spectrofluorometer (SPEX, Edison, NJ) and an excitation wavelength of 340 nm. Test peptides were added to 0.6 M dansyl-CaM in the absence or presence of 200 M Ca 2ϩ in 10 mM HEPES pH 7.0 buffer.
Interactions between i3 loop peptides and dansyl-CaM were also examined indirectly by measuring test peptide-induced changes in fluorescence intensity of dansyl-CaM complexed with Ca 2ϩ /CaM-KII-(290 -309) (21), in a Fluostar '97 fluorimeter, equipped with microinjectors (BMG Lab Technologies, Durham NC). Samples were divided evenly into solid white 96-well microtiter plates (135 l/well) containing 150 nM dansyl-CaM, 100 nM Ca 2ϩ /CaM-KII-(290 -309), and test peptide. Fluorescence intensity was recorded at 100-ms intervals using an excitation wavelength of 340 nM and emission of 490 nm. Ten consecutive steady-state base-line observations were averaged for each peptide using Ca 2ϩ -free buffer (10 mM HEPES, pH 7.0, 100 M EGTA), and replicates of each averaged value were combined. Concentration-response curves were plotted using a logistic curve-fitting equation as described by De Lean et al. (21). Subsequently, the effect of adding Ca 2ϩ was determined on the competition between test peptides, Ca 2ϩ /CaM-KII-(290 -309), and dansyl-CaM. Using the same starting conditions outlined above, base-line measurements were then followed by an injection of 15 l of a 0.11 mM CaCl 2 solution yielding a final free [Ca 2ϩ ] of ϳ10 M in 150 l. Changes in fluorescence intensity were monitored at 100-ms intervals for 18 s. In the absence of dansyl-CaM, no fluorescence was observed under any condition. In the presence of dansyl-CaM alone, Ca 2ϩ injection caused a reproducible increase in fluorescence that reached maximum intensity within 200 ms and then remained at that new plateau level. When a mixture of Ca 2ϩ /CaM-KII-(290 -309) and dansyl-CaM was injected with Ca 2ϩ (in the absence of test peptides), there was an increase in fluorescence intensity that peaked within 200 ms, followed by an exponential decay in fluorescence intensity, which reflects the kinetic interaction of Ca 2ϩ /CaM-KII-(290 -309) with dansyl-CaM in its altered Ca 2ϩ -bound form. The ability of peptides to inhibit the decay in fluorescence intensity was used as an indication of the test peptide's ability to disrupt formation of this new complex between dansyl-CaM and Ca 2ϩ /CaM-KII-(290 -309), providing an indirect measure of relative binding potency in the presence of Ca 2ϩ and plotted as percentage change in fluorescence intensity (F f /F i ϫ 100), where F i is the average fluorescence intensity between 200 and 700 ms after injection of Ca 2ϩ and F f is the average fluorescence intensity between 15.8 and 17.7 s after injection.
Binding of Purified FLAG-mOP 3 Receptor to CaM Affinity Gels-HEK-FLAG-mOP 3 cells or pRC/CMV empty vector-transfected HEK293 control cells were homogenized in phosphate-buffered saline (PBS; 1 mM KH 2 PO 4 , 10 mM NaHPO 4 , 137 mM NaCl, 2.7 mM KCl) and centrifuged at 30,000 ϫ g for 20 min, and the membrane pellet was solubilized for 30 min at 4°C with 1% NP 40 in 20 mM Tris-HCl, 100 mM NaCl, pH 7.5 (Buffer A), containing a protease inhibitor mixture (Roche Molecular Biochemicals). This mixture was then centrifuged at 50,000 ϫ g for 40 min at 4°C, and the supernatant was concentrated to 1 ⁄10 its volume using an Ultrafree centrifugal filter with a nominal molecular weight limit of 50 kDa. Concentrated supernatant was diluted 10-fold in buffer A to reduce the concentration of NP 40 to 0.1%, and solubilized material was incubated with 1 ⁄10 volume of anti-FLAG mAb (M2) affinity gel (Eastman Kodak Co.) for 1 h at 4°C with gentle agitation. The gel was then washed with buffer A plus 0.1% NP 40 , and FLAG-mOP 3 receptor was eluted by adding 400 M FLAG peptide and incubating at room temperature for 30 min. FLAG peptide was removed by three cycles of concentration and dilution. Immunopurified FLAG-mOP 3 receptor was then incubated with 20 l of immobilized bovine CaM affinity gel (Calbiochem) for 1 h at 4°C in buffer A plus 0.1% NP 40 and 1 mM CaCl 2 .
The resin was pelleted and washed three times with buffer A plus 0.1% NP 40 and 1 mM CaCl 2 for 10 min at 4°C. Samples were then eluted by incubating with 20 l of Laemmli sample buffer at 65°C for 5 min. FLAG-mOP 3 receptor was separated on 8% SDS-PAGE and detected by Western blot using rabbit anti-OP 3 receptor C-terminal antiserum (Incstar, Stillwater, MN), followed by an AP color development kit (Bio-Rad).
Binding of Biotinylated CaM to FLAG-mOP 3 Receptor-HEK-FLAG-mOP 3 cells or pRC/CMV vector-transfected HEK cells were solubilized in PBS, pH 7.4, containing 1% NP 40 and a protease inhibitor mixture. The reaction mixture was centrifuged at 50,000 ϫ g for 40 min at 4°C. The supernatant was then gently agitated for 9 h at 4°C with 3 g of anti-FLAG mAb (M2) (Sigma) and 75 l of a 20% slurry of washed protein G-Sepharose beads (Amersham Pharmacia Biotech); total volume was 0.7 ml. Beads were pelleted by centrifugation, washed three times with PBS plus 0.1% NP 40 , and gently mixed with 2.2 g of biotinylated CaM (Calbiochem) in 300 l of PBS and 0.1% NP 40 for 2 h at 4°C, either in the presence of 1 mM CaCl 2 or 2 mM EGTA. Beads were then washed with PBS containing 0.1% NP 40 (with CaCl 2 or EGTA), dissolved in 20 l of Laemmli sample buffer, and analyzed on 10% SDS-polyacrylamide gel electrophoresis. Biotinylated CaM was detected using an AP-conjugated avidin and an AP color development kit.
Expression of Sense and Antisense CaM Vectors in HEK-rOP 3 Cells-cDNA encoding chicken type II CaM (from Dr. Anthony R. Means of Duke University (22)) was subcloned into pcDNA3, Zeo(ϩ)/pcDNA3 (sense CaM), and Zeo(Ϫ)/pcDNA3 (antisense CaM) using BamHI and XbaI. (Amino acid sequences of chicken and mammalian CaMs are identical.) Transfection of HEK-rOP 3 cells was accomplished by using Superfect reagent (Qiagen, Santa Clarita, CA). The sense CaM-pcDNA3 construct was used for transient expression, and cells were harvested after 36 h. Stable cell lines were established with CaM-Zeo(ϩ)/pcDNA3 (sense) or CaM-Zeo(Ϫ)/pcDNA3 (antisense) plasmids with 400 g/ml Zeocin TM (Invitrogen, Carlsbad, CA) for initial selection and then lowering the concentration of Zeocin TM to 100 g/ml once established. Expression levels of rOP 3 receptor, determined with [ 3 H]diprenorphine as described earlier, were not measurably altered by changes in CaM expression.
Measurement of CaM Levels in Plasma Membranes-Plasma membranes were prepared as described by Nehmad et al. (23). CaM was then extracted from the plasma membrane fractions, containing approximately 70% of total [ 3 H]naloxone binding sites, and CaM levels were determined by measuring phosphodiesterase activity using [ 3 H]cAMP as described previously (24). Standard curves were generated with known quantities of purified CaM (Sigma), with correlation coefficients ranging between 0.95 and 0.98. CaM recovery over the assay procedure was 65-68%. Results were confirmed by Western blot analysis.
[ 35 S]GTP␥S Binding Studies-Membranes for [ 35 S]GTP␥S binding assays were prepared from cells washed with PBS and then homogenized in 40 mM Tris-HCl buffer, pH 8.0, in the absence (control membranes) or presence of 2 mM EGTA (EGTA-washed membranes), incubated on ice for 15 min, and centrifuged at 30,000 ϫ g for 20 min at 4°C. Thus, washing in pH 8/EGTA buffer serves to reduce the CaM content of the membranes (23). Pellets were then resuspended briefly in 10 mM HEPES buffer, pH 7.4, centrifuged, and used for the [ 35 S]GTP␥S binding assay as described by Burford et al. (25).
[ 3 H]Naloxone Binding Experiments-Membranes for binding studies, prepared as described above for [ 35 S]GTP␥S binding assays, were suspended in 10 mM HEPES buffer containing 10 mM MgCl 2 and 100 mM NaCl, pH 7.4. Morphine (final concentrations ranging from 0 to 10 M) and 1 nM [ 3 H]naloxone (specific activity 69 Ci/mmol; Amersham Pharmacia Biotech) were added to each membrane sample (containing ϳ20 g of protein/sample), incubated for 2 h at room temperature, filtered, and washed at 4°C, and filter-bound [ 3 H]naloxone was determined by scintillation counting (13). Where indicated, membranes were preincubated for 30 min at room temperature with 60 M CaM, and then diluted 4 times before [ 3 H]naloxone binding was determined. Binding curves were analyzed using one-or two-site models by GraphPad Prism (San Diego, CA).
cAMP Measurements-cAMP measurements were performed by radioimmunoassay as described elsewhere (13). Cell suspensions were incubated with 10 M forskolin for 10 min at 37°C in the presence of various concentrations of morphine (0 -1 M).

CaM-binding Motif
Search-Binding of CaM to CaM-binding proteins does not involve a universal, conserved sequence motif (26); therefore, it remains difficult to define CaM binding domains in distinct protein families using sequence alignment data. To identify possible CaM binding domains in GPCRs, we used a series of sequence motifs derived from distinct protein families with known CaM binding domains for a PatScan (27) pattern search of the SwissProt protein data base. Scanning for possible binding motifs produced matches within a number of known CaM-binding proteins and a limited number of additional proteins, including GPCR sequences located in the i3 loop. Among these, the opioid receptors emerged with several motif searches showing similarities to myosin light chain kinases, Ca 2ϩ /CaM kinases, and phosphodiesterases. Shown in Fig. 1 are representative alignments, with key motif residues underlined, suggesting the possibility that the i3 loop might interact not only with G proteins (10,19,28,29) but also with CaM.
Binding of OP 3 Receptor-derived Peptides to CaM-The proposed site for interaction with CaM in the i3 loop of opioid receptor was first examined by observing mobility shifts of CaM using nondenaturing gel electrophoresis (8). CaM was incubated in varying molar ratios with selected peptides derived from the sequence of the OP 3 receptor i3 loop (Table I; Fig. 2A). Three peptides containing the C-terminal portion of the i3 loop predicted to interact with CaM, namely i3-1, i3-2, and i3-3, caused a notable shift in the electrophoretic mobility of CaM, whereas the N-terminal peptide i3-4 did not ( Fig. 2A). The small gel shift caused by the low mass C-terminal i3-1 peptide is better visualized in Fig. 2B. These results indicate that the C-terminal portion of the OP 3 receptor i3 loop is a suitable candidate for interaction with CaM. Mobility shifts were not observed in the absence of Ca 2ϩ (not shown), suggesting that peptide-CaM interactions are strengthened in the presence of Ca 2ϩ .
In contrast to mobility shifts observed with Ca 2ϩ /CaM kinase II-(290 -309) or -(281-309), which were complete at molar ratios of 1:1 (Table I), shifts in CaM mobility caused by i3 loop peptides were incomplete even at substantially higher ratios. The i3-2 peptide, which lacks the N-terminal portion of the i3 loop, for example, produced a visible gel shift only at a 100-fold excess over CaM ( Fig. 2A). The inability of i3 loop peptides to cause a complete shift in CaM under these conditions is similar to results reported for peptides derived from the CaM binding region of the metabotropic glutamate subtype 5 receptor (9).
Since motif analysis indicated that Lys 273 of the OP 3 receptor may be important for binding interaction with CaM, two modified i3-1 peptides (one in which the N-terminal Lys is removed (i3-5) and another in which Lys is replaced with Ala (i3-6)) were also synthesized and tested (Fig. 2B, Table I). Both of these modified peptides failed to induce mobility shifts of CaM, demonstrating that modifications to this critical residue can affect the ability of i3 loop peptides to interact with CaM.
Shifts in the mobility of Ca 2ϩ /CaM kinase II peptides reflect the strong interaction that occurs with CaM, consistent with previous reports (30). In contrast, several peptides representing mostly the autophosphorylation substrate domain of Ca 2ϩ / CaM kinases failed to shift CaM. Moreover, autocamtide-2 and mastoparan, two peptides with sequence motifs and predicted secondary structures resembling those of CaM binding domains also failed to shift the CaM band (Table I), although mastoparan is known to bind to CaM with high affinity when determined by fluorescence shifts of dansylated CaM (11). This suggests that unlike the i3 loop peptides, binding stability of mastoparan to CaM appears insufficient to induce a detectable shift in this gel assay.
Interaction of i3 Loop Peptides with Dansyl-CaM-CaM binding of i3 loop peptides was also examined by measuring changes in the dansyl-CaM fluorescence spectrum (20). The i3-3 peptide, which represents the amino acid sequence of the entire OP 3 receptor i3 loop, caused the largest change of dansyl-CaM fluorescence among the i3 loop peptides tested, and changes in fluorescence intensity occurred both in the presence and absence of Ca 2ϩ (Fig. 3A). This suggests that i3-3 is capable of inducing a substantial conformational change in CaM (12) and that its interaction is at least partially Ca 2ϩ -independent. Each of the other i3 loop peptides was also able to induce detectable, although smaller, changes in dansyl-CaM fluorescence at higher concentrations, although some of these peptides failed to induce a CaM gel shift, and no response was observed with the control FLAG peptide.
To estimate the relative potency of the i3 loop peptides, changes in fluorescence intensity were measured when Ca 2ϩ / CaM kinase II-(290 -309) is complexed with dansyl-CaM, an interaction that has an approximate binding K d of 50 nM (30), in the presence of test peptides. In the presence of Ca 2ϩ , i3 loop peptides altered changes in fluorescence intensity that were dose-dependent and of different magnitude. The entire i3 loop peptide, i3-3, was most potent and efficacious with a relative IC 50 of 42 (S.D. ϭ 3) nM followed in rank order by i3-1, i3-2, and then (of similar low efficacy) i3-4, -5, and -6 ( Fig. 3B; curves for i3-1, -3, -4, 5, and -6 shown). A similar rank order of relative potency was obtained in the absence of Ca 2ϩ (Fig. 3, C and D; curves for i3-1 and i3-6, i3-3, and i3-4 shown). Thus, substituting or deleting Lys (Lys 273 of the OP 3 receptor) in the i3 loop reduced the interaction of the C-terminal peptides with CaM. The relative affinity of i3-3 appeared to be lower in the absence of Ca 2ϩ (IC 50 ϳ300 nM), indicating that Ca 2ϩ enhances CaM binding. The binding curve of the i3-3 peptide appeared to be biphasic (Fig. 3D), suggesting that more than one binding interaction may be involved.
CaM Binding to Solubilized OP 3 Receptor-FLAG-mOP 3 receptor, immunopurified using FLAG monoclonal antibody and analyzed by Western blotting with OP 3 receptor antiserum, was visible as a main band at ϳ65 kDa, a band absent in mock-transfected cell extracts (Fig. 4, A, B, and C). In the presence of Ca 2ϩ , FLAG-mOP 3 receptor was retained on CaM affinity gels but not control gels (Fig. 4A). (Binding in the absence of Ca 2ϩ with EGTA or BAPTA added could not be tested because of excessive nonspecific background binding to control gels.) C-terminal i3 loop peptides, i3-1, -2, and -3, reduced receptor binding observed to CaM affinity gels, whereas the N-terminal peptide, i3-4, did not (Fig. 4, A and B). A reduction in binding was also observed with Ca 2ϩ /CaM-II-(290 -309) but not with the control FLAG peptide (Fig. 4A). The Lys 273 truncated peptide, i3-5, and the K273A peptide, i3-6, were less effective at inhibiting OP 3 receptor binding to CaM gels (Fig. 4C). These results indicate, therefore, that OP 3 receptor binding to CaM is mediated predominantly by the Cterminal portion of the i3 loop, and this parallels the results observed in gel shift assays.
Binding of CaM to OP 3 receptor was also tested using solubilized FLAG-mOP 3 receptor, which was immunoabsorbed to Sepharose G beads with an anti-FLAG monoclonal antibody. Neither CaM nor the G i␣3 subunit of heterotrimeric G proteins (which is preferentially activated by OP 3 receptor (25)) was  Fig. 2

)
The numbering of the OP 3 -derived peptides is according to the human OP 3 sequence (single letter code). The shown sequences are identical between human and rodent OP 3 . Substitutions of the wild-type sequence are indicated by single letter amino acid code with a right superscript number indicating the substituted position.  2. Binding of OP 3 receptor i3 loop-derived peptides to CaM using a nondenaturing gel shift assay. Calmodulin was incubated with peptides derived from the OP 3 receptor i3 loop in molar ratios of 1:1, 1:10, and 1:100 (CaM/peptide as indicated) in the presence of Ca 2ϩ and separated by nondenaturing polyacrylamide gel electrophoresis including Ca 2ϩ . The addition of EGTA to chelate Ca 2ϩ suppressed the gel shift (not shown). For peptide designations see Table I. detectable by Western blot analysis when extracts of HEK-FLAG-mOP 3 cells were applied (data not shown). This suggests that if these proteins were associated with OP 3 receptor in the intact cells, they had dissociated during the solubilization process. On the other hand, externally added biotinylated CaM did bind to immobilized FLAG-mOP 3 receptor in the presence of Ca 2ϩ (Fig. 4D). In contrast, CaM was not recovered when using extracts from vector-transfected cells treated in the same fashion. Moreover, in the absence of Ca 2ϩ , CaM was not retained by FLAG-rOP 3 receptor, indicating that the binding is Ca 2ϩ -sensitive (Fig. 4D).
CaM Blocks G Protein Activation by i3 Loop Peptide i3-3-Peptides derived from both the N and C terminus of the i3 loop of various GPCRs have been shown to activate G proteins directly (29). Therefore, the ability of peptides derived from the i3 loop of OP 3 receptor to stimulate [ 35 S]GTP␥S binding to membranes of HEK-rOP 3 cells was examined. Each of the peptides tested, i3-1, -2, -3, and -4 (30 M each), caused a significant 10 -20% increase in [ 35 S]GTP␥S binding to membranes. The full-length i3 loop peptide, i3-3, produced the largest effect of 22% (SD ϭ 2%, n ϭ 6, p Ͻ 0.001, Student's t test, unpaired). The addition of an equimolar CaM concentration completely blocked G protein activation by i3-3, reducing [ 35 S]GTP␥S incorporation to the same level observed with CaM alone (ϳ93% of control). This result indicates that CaM binding to i3-3 prevents G protein activation.

Effects of CaM on Opioid Receptor-G Protein
Coupling-The effect of CaM on G protein activation was examined in membranes prepared from HEK-mOP 1 (␦-opioid) and HEK-rOP 3 (-opioid) cells. Washing in Ca 2ϩ -free pH 8/EGTA buffer reduced CaM levels in plasma membranes (expressed as CaM activity/mg of protein) to 58% (S.D. ϭ 14%, n ϭ 9, p Ͻ 0.001, t test, unpaired) of its original level. [ 35 S]GTP␥S binding activity of washed membranes, on the other hand, was significantly elevated over unwashed controls (Fig. 5). The increase in [ 35 S]GTP␥S binding activity (of washed versus unwashed membranes) was observed in both absence (basal activity) and presence of agonists (OP 1 and OP 3 membranes stimulated with 1 M DPDPE and morphine, respectively) (Fig. 5). Washing  Fig. 6. Both basal and stimulated increases in [ 35 S]GTP␥S binding activity were abolished by pertussis toxin pretreatment in opioid receptor-containing membranes, and no change in activity was observed in mocktransfected (pRC/CMV) membranes under any of the conditions described above (Fig. 5). This suggests that CaM may interfere with or reduce G protein coupling in membranes containing OP 1 and OP 3 receptors. Furthermore, elevated [ 35 S]GTP␥S binding activity is observed in membranes containing either opioid receptor when compared with mock-transfected control under agonist-free conditions (Fig. 5). This enhanced activity was abolished by pertussis toxin, indicating that it reflects an increase in basal activity caused by the presence of opioid receptors.
The effect of adding CaM to washed membranes was also considered and is illustrated in Fig. 6A. The addition of CaM to washed HEK-rOP 3 membranes caused a lowering of [ 35 S]GTP␥S binding activity to levels observed before washing, across the entire concentration range of morphine, without significant changes in EC 50 values (0.84 -1.94 nM). These results reinforce the contention that CaM is responsible for the changes observed and inversely affects G protein coupling. Guanine nucleotides (GDP) were omitted in the experiment illustrated in Fig. 6A to facilitate [ 35 S]GTP␥S binding and possible competition between CaM and G proteins. In the presence of 1 M GDP, pH 8/EGTA washing similarly enhanced [ 35 S]GTP␥S binding activity over the entire concentration range of morphine used, but the addition of CaM was unable to fully reverse this enhanced activity (data not shown). This suggests that a rather stable OP 3 receptor-G protein interaction may have formed in the membranes (31).

Effect of Alteration of CaM Levels in Intact Cells on Opioid
Receptor G protein Coupling-CaM levels were also altered in intact cells by expressing plasmids encoding sense and antisense CaM cDNA. Membranes prepared from HEK-rOP 3 cells transiently transfected with sense CaM cDNA, expressing CaM at levels of 186% (CaM activity/mg protein; S.D. ϭ 24%, n ϭ 9, p Ͻ 0.001, t test, unpaired) over mock-transfected cells, had a significantly reduced level of basal and agonist-stimulated [ 35 S]GTP␥S binding activity (illustrated in Fig. 6B) This was reversed upon washing cell membrane with pH 8/EGTA buffer (Fig. 6B), which lowered CaM levels by 46% (S.D. ϭ 8%, n ϭ 9, p Ͻ 0.001, t test, unpaired). EC 50 values for morphine did not change significantly under any condition (3.5-4.3 nM). (Higher EC 50 values observed in Fig. 6B compared with Fig. 6A are due to the presence of GDP in the assay incubations.) CaM transfection not only lowered the entire dose-response curve but also blunted the maximal morphine effect on [ 35 S]GTP␥S binding activity (E max from 120 Ϯ 12 to 94 Ϯ 9% over basal, means Ϯ S.D., p Ͻ 0.05, t test, unpaired); this effect was also reversed upon washing (E max 126 Ϯ 7%).
In contrast to cell lines with sense CaM vectors, stable expression of antisense CaM reduced CaM levels to 73% (S.D. ϭ 13%, n ϭ 9, p Ͻ 0.001, t test, unpaired) of the control, and these cells were found to have significantly elevated basal and morphine-activated [ 35 S]GTP␥S binding activity (Fig. 6C). Importantly, washing of the antisense CaM cell membranes failed to affect the morphine concentration-response curve of [ 35 S]GTP␥S binding activity (Fig. 6C). Moreover, the maximal effect of morphine was enhanced by antisense CaM expression from 110 Ϯ 11 to 147 Ϯ 21% over basal (means Ϯ S.D., n ϭ 9, p Ͻ 0.05, t test, unpaired), and the EC 50 value was significantly reduced from 5.5 Ϯ 1.4 to 1.7 Ϯ 0.2 nM (means Ϯ S.D., n ϭ 9, p Ͻ 0.05, t test, unpaired). The activity of mock-transfected cell membranes was equivalent to untransfected controls (data not shown), ruling out a nonspecific effect of plasmid transfection.
G Protein Coupling of the Mutant K273A-FLAG-hOP 3 Receptor-Since the motif search and experimental results with i3 loop-derived peptides indicated that Lys 273 of OP 3 receptor could play a role in the proposed interaction with CaM, a K273A-OP 3 receptor mutant was constructed, and its ability to stimulate [ 35 S]GTP␥S binding activity was compared with wild-type FLAG-hOP 3 receptor. The FLAG-K273A-hOP 3 receptor was still retained on a CaM affinity gel, as determined by immunoblotting, but retention was relatively small and not easily quantifiable (data not shown). Thus, CaM interaction with this mutated receptor does not appear to be completely abrogated, in agreement with the i3-6 peptide's partial ability to block FLAG-mOP 3 binding to CaM (Fig. 4C).
Compared with the wild-type FLAG-hOP 3 receptor, basal and morphine-activated [ 35 S]GTP␥S binding activity for the FLAG-K273A-hOP 3 receptor were elevated regardless of washing with pH 8/EGTA buffer (see Fig. 6D). These results indicate that substitution of Lys 273 with Ala induces elevated levels of basal G protein stimulation and that the receptor is no longer responsive to changes in CaM content.
CaM 7A). The displacement curve is fitted better by a two-site, rather than a one-site, receptor binding model, indicating a distribution of 34 Ϯ 2% high affinity and 64 Ϯ 8% (means Ϯ S.D.) low affinity sites.
If the high affinity site were to represent the G proteincoupled state, one would expect displacement of G proteins by CaM to abolish the high affinity site. Indeed, incubation of the membranes with CaM (final concentration, 15 M) in the presence of 50 M CaCl 2 largely eliminated the high affinity site (Fig. 7A), and the resultant data fit a one-site model (Fig. 7A). The binding curve in the presence of CaM is nearly identical to that obtained in the presence of GTP␥S, which uncouples the OP 3 receptor from its G proteins (Fig. 7A). Chelation of Ca 2ϩ by EGTA prevented the effect of CaM on the high affinity binding site, another indication that CaM-OP 3 interactions are Ca 2ϩ -sensitive.
We also measured [ 3 H]naloxone-morphine displacement curves in membrane obtained from HEK-rOP 3 cell that overexpressed CaM (HEK-rOP 3 /CaM, stably transfected, in which membrane CaM activity was 134 Ϯ 16% of that of mock-transfected cells). Overexpressing CaM eliminated the high affinity binding site (Fig. 7B). Moreover, the addition of GTP␥S had no further effect on the morphine displacement curve (Fig. 7B). Thus, elevated CaM levels abolished the high affinity agonist binding site.
To address the question of whether CaM effects on OP 3 receptor arise from direct binding to the receptor or indirectly to another protein attached to the receptor such as G␣, we determined ligand binding of the mutant receptor K273A-FLAG-hOP 3 , which effectively couples to G proteins but is functionally insensitive to CaM. Shown in Fig. 7C, the high affinity site of the K273A-FLAG-hOP 3 receptor remained unaffected by CaM addition, although the addition of GTP␥S eliminated this high affinity site. This supports the view that the changes in high affinity morphine binding to OP 3 receptors were mediated by direct CaM-OP 3 receptor interactions.
Morphine-induced CaM Release from Plasma Membranes-To determine whether OP 3 receptor activation can cause dissociation of CaM from the plasma membrane, we measured changes of CaM content in plasma membranes prepared from HEK-rOP 3 that had been exposed to 1 M morphine (Fig. 8A). There was a significant reduction of membrane CaM content within 1 min, leveling off at 15 min. A similar effect was observed with HEK-FLAG-hOP 3 and with HEK-mOP 1 cells stimulated with 1 M DPDPE (Fig. 8B). Moreover, treatment with pertussis toxin (100 ng/ml for 18 h) failed to affect the morphine-induced loss of CaM from plasma membranes in HEK-rOP 3 cells, whereas morphine had no effect on CaM levels in K273A-FLAG-hOP 3 cells (Fig. 8B). These results suggest that G proteins do not play a role in CaM translocation and, further, that CaM depletion from the membrane might be a direct result of CaM dissociation from the receptor.
Effect of CaM Levels on OP 3 Receptor Regulation of cAMP in Intact Cells-The ability of morphine to lower forskolin-stimulated cAMP levels was measured in HEK-rOP 3 and in sense and antisense CaM-expressing HEK-rOP 3 cells. Changing the CaM content failed to affect the EC 50 value of morphine significantly (range between 4 and 8 nM). However, in HEK-rOP 3 / as-CaM cells with reduced CaM content, maximum inhibition of cAMP accumulation by morphine increased from 73 Ϯ 3% (normal CaM level) to 97 Ϯ 4% (means Ϯ SD, n ϭ 6, p Ͻ 0.05, t test, unpaired). Similarly, morphine was slightly but significantly more efficacious in HEK-K273A-FLAG-hOP 3 than in HEK-FLAG-hOP 3 cells (maximum inhibition of cAMP accumulation was 82 Ϯ 3 versus 68 Ϯ 2%; means Ϯ S.D., n ϭ 6, p Ͻ 0.01, t test, unpaired), again without a significant change in the EC 50 (2 nM in each case). These results parallel those observed with [ 35 S]GTP␥S binding to cell membranes. DISCUSSION This study provides evidence in support of the hypothesis that CaM interacts with the i3 loop of OP 3 . We have demonstrated that solubilized OP 3 interacts with CaM in a Ca 2ϩsensitive manner. Moreover, peptides derived from the i3 loop of OP 3 bound to CaM and suppressed interactions between CaM and solubilized OP 3 . As expected from these results, CaM content of plasma membranes inversely affected G protein coupling by OP 3 and suppressed high affinity binding sites of morphine to OP 3 , thought to represent the G protein-coupled state of the receptor. However, these results taken alone did not rule out the possibility that CaM could modulate receptor-G protein coupling indirectly, for example, by binding to the ␤ subunit of G proteins (8). However, the solubilized and immunopurified OP 3 appeared devoid of associated G protein, since we were unable to detect any G i␣3 (a major G ␣ -coupling protein of OP 3 ) in the immunopurified rOP 3 preparation; therefore, G proteins are unlikely to mediate the observed CaM binding to purified OP 3 . To resolve the question whether a direct interaction occurs, we screened a number of OP 3 mutants with single residue substitutions in the i3 loop for loss of CaM interaction, but with preserved G protein coupling. Importantly, the OP 3 mutant K273A (selected on the basis of the role of this Lys residue in motif structure and i3-loop peptide binding to CaM) displayed enhanced G protein coupling, but it was insensitive to CaM. This result strongly supports a direct interaction between OP 3 and CaM.
Taken together, our results suggest that CaM and G proteins compete for a shared binding domain in the i3 loop of OP 3 . As proposed for G protein binding domains, both the N terminus and the C terminus of i3 may play a role in CaM binding, although the C terminus is the dominant factor. While the binding mechanism between CaM and the i3 loop remains to be resolved, it is possible that a distortion of the i3 loop upon receptor activation could lower the affinity to CaM and cause CaM dissociation. CaM binding to OP 3 was enhanced by Ca 2ϩ , but studies with i3 loop peptides indicate that Ca 2ϩ is not essential. At resting conditions with low intracellular Ca 2ϩ levels, OP 3 may thus be occupied either by CaM or by G proteins. OP 3 activation could cause dissociation of CaM so that the unoccupied receptor can subsequently rebind G proteins. Therefore, elevated CaM levels failed to abolish agonistinduced G protein coupling.
This study further establishes a significant level of basal G protein coupling in cell membranes for both OP 3 and OP 1 , as suggested previously (31)(32)(33). Basal G protein signaling was highly sensitive to CaM content; thus, overexpression of CaM obliterated, while down-regulating CaM enhanced, basal G protein signaling. This is consistent with the proposed model that CaM binding to OP 3 suppresses G protein coupling, and that agonist activation is needed to dislodge CaM and initiate G protein signaling. This appears to provide a novel mechanism of regulating basal G protein signaling. Accordingly, the CaM-insensitive mutant receptor K273A-hOP 3 displayed increased basal activity over that of wild-type hOP 3 .
The results in this study further raise the possibility that CaM, like G proteins, may serve directly as a receptor-mediated second messenger molecule, translocating signals upon release from receptor. Subcellular redistribution of CaM upon morphine administration had already been reported to occur in intact animals and in tissue culture (23,34). Similarly, enkephalins and opiate antagonists were shown to control a parallel CaM-opioid receptor redistribution in neuroblastoma ϫ glioma NG108 -15 cells (expressing OP 1 ), suggesting a tight regulatory, and possibly structural, relationship between opioid receptors and CaM (34). Since no CaM dissociation from plasma membranes was observed with K273A-OP 3 in our study, the loss of CaM from the plasma membrane appeared to be a direct result of CaM-OP 3 dissociation upon activation of the receptor rather than a secondary effect of OP 3 -G protein signaling. This finding is further supported by the observation that uncoupling OP 3 from G proteins with pertussis toxin failed to affect morphine-induced CaM loss from the plasma membrane.
CaM dissociation from the plasma membrane can account for previous findings that opioid receptors activate Ca 2ϩ /CaM-dependent cAMP phosphodiesterases in a G protein-independent fashion (35,36). Other candidate signaling pathways for CaM include adenylyl cyclases, NO synthases, CaM kinases, and ion channels. In some tissues, including HEK-OP 3 cells, morphine activation of OP 3 causes a sustained increase in intracellular Ca 2ϩ levels (37). Together with CaM mobilization, this could lead to enhanced OP 3 phosphorylation by Ca 2ϩ /CaM kinase-II (38) as a possible regulatory pathway.
Motif searches of CaM binding domains revealed a number of GPCRs as possible candidates for interacting with CaM. Among these, the opioid receptors all share nearly identical i3 loop sequences (Fig. 1). Thus, G protein coupling of OP 1 was similarly sensitive to CaM content in the plasma membrane, and it is plausible that CaM could bind to yet other GPCRs, such as the adrenergic (Fig. 1) and muscarinic receptors, that also contain a putative CaM motif. Similar to the opioid recep-tors, muscarinic receptors regulate phosphodiesterases by a Ca 2ϩ -dependent mechanism distinct from G protein-coupled regulation of adenylyl cyclases (39), possibly involving CaM.
In conclusion, we present evidence in support of the hypothesis that CaM binds to the i3 loop of opioid receptors and that CaM may serve as a second messenger molecule upon receptor activation.