Functional reconstitution in situ of 5-hydroxytryptamine2c (5HT2c) receptors with alphaq and inverse agonism of 5HT2c receptor antagonists.

Membranes prepared after infection of Sf9 cells with recombinant baculovirus containing the rat 5HT2c receptor DNA, but not after infection with wild-type virus, expressed high affinity binding sites for 125I-lysergic acid diethylamide and [3H]mesulergine. The receptor site density reached an optimum of 50-70 pmol/mg membrane protein at 60 h postinfection. Extraction of peripheral membrane proteins from the postnuclear membrane fraction with 6 M urea depleted GTPgammaS-binding 4-fold without decreasing 5HT2c receptor binding activity. Urea-extracted Sf9 membranes expressing the 5HT2c receptor catalyzed the activation of squid retinal alphaq but not bovine retinal alphat or bovine alphao/alphai. Productive interaction of 5HT2c receptors with squid alphaq was enhanced by the addition of betagamma dimers prepared from either bovine brain or bovine rod outer segment discs. While the addition of serotonin increased 5HT2c receptor-catalyzed GTPgammaS binding to alphaq, the unoccupied receptor was also catalytically active. The 5HT2c receptor antagonists, mesulergine, mianserin, and ketanserin competitively inhibited 5HT activation of the receptor with predicted rank-order affinities; and mianserin and ketanserin markedly inhibited basal 5HT2c receptor activity. Interestingly, this "inverse agonist" efficacy did not correlate with antagonist affinity for the 5HT2c receptor. Baculoviral expression of the 5HT2c receptor and urea extraction of postnuclear Sf9 cell membranes have provided a high density of in situ, uncoupled, G-protein-linked receptor useful for reconstitution with purified G-protein subunits. This has allowed for independent manipulation of receptor and G-protein chemical concentrations and has revealed that a G-protein-linked receptor can possess a significant basal catalytic activity and that antagonist compounds can act as inverse agonists of this basal activity at the level of receptor activation of G-proteins.

The 5HT 2c 1 receptor belongs to a recently reclassified subfamily of 5-hydroxytryptamine receptors (1) and to the larger superfamily of rhodopsin-homologous, seven transmembrane domain receptors (2). Rhodopsin-homologous receptors transmit extracellular signals across biological membranes by activating specific signal transducing GTP-binding proteins (heterotrimeric G-proteins) that in turn regulate a variety of intracellular effector pathways (3)(4)(5). Agonist stimulation of the 5HT 2 subfamily of serotonin receptors initiates intracellular phosphoinositide (PI) hydrolysis and a subsequent rise in intracellular calcium (6,7). The 5HT 2c receptor is distinguished from 5HT 2a and 5HT 2b by its ligand affinities and tissue distribution. The 5HT 2c receptor is expressed in neurons of the central nervous system, with especially high levels of expression in the choroid plexus (8). The neurotransmitter, serotonin, has been extensively studied and has been implicated in mental diseases ranging from affective and mood disorders to autism and dementia (9). As we more fully understand the physiological diversity exhibited by the large family of serotonin receptors (10), we will gain further insights into the neuropharmacology of serotonin.
Members of the 5HT 2 receptor subfamily, as do other members of the extensive gene family of rhodopsin homologous receptors, signal through heterotrimeric G-proteins. Members of this receptor family, when stimulated, act to catalyze the exchange of GTP for tightly bound GDP on the G-protein ␣ subunits, which in turn activate intracellular effectors such as adenylyl cyclase or phospholipase C-␤ subtypes (PLC-␤s). The ␣ subunits are subclassified into four groups based upon sequence homology (11,12). At least four ␣ gene products have been found to stimulate phosphoinositide hydrolysis by activating PLC-␤s (13); these are designated as the ␣ q subfamily (14,15). G␤␥ subunits have also been noted to activate PLC-␤s (16,17).
Because the 5HT 2 subfamily of receptors stimulate PI hydrolysis it has been assumed that agonist stimulation of 5HT 2c receptors leads to GDP-GTP exchange on ␣ q which in turn activates a PLC-␤ isozyme(s). However, no reports have directly examined the activation of G q proteins by 5HT 2 receptors. The in vitro studies of activation of G-proteins by receptors have proven difficult. In general, difficulties relate to purification of the quantities of proteins necessary to study the molecular interactions between a single receptor and distinct G-proteins (18,19). Thus, although previous work has led to the inference that 5HT 2c activates ␣ q , the coupling of 5HT 2c and ␣ q , like that of many receptors with purified G-protein subunits, has not been studied directly.
With the identification of over 100 cDNAs encoding seven transmembrane domain receptors (exclusive of olfaction) and the molecular cloning of cDNAs encoding multiple gene products for each of the G-protein subunit families, a central issue in cell regulation is how these protein families contribute to the diversity of cellular responses while conserving the specificity of each response. One level of specificity is likely encoded in the thermodynamics of distinct gene product interactions, and to examine this possibility we have in the past exploited the high chemical concentration and unique gene expression found in photoreceptor membranes (20 -22). To extend such studies to other G-protein-coupled receptors, we and others have turned to high level expression systems. The utility of the baculovirus expression system for obtaining reconstitutively active G-protein-coupled receptors has been described previously (23). However, previous investigations of recombinant G-protein-linked receptors have employed detergent solubilization of the receptor from its native membrane and subsequent reconstitution in artificial liposomes (18,19,24,25). We have avoided liposome reconstitution by exploiting the utility of urea for extraction of extrinsic membrane proteins from postnuclear membrane fractions without disrupting membrane integrity.
In this report we present the development of in situ reconstitution procedures for rat 5HT 2c receptors expressed in insect Sf9 cells and the quantitative analysis of the catalytic activation of squid photoreceptor ␣ q (26,27). Our studies reveal a basal catalytic activity of this receptor that is enhanced by agonist ligands and differentially inhibited by 5HT 2c receptor antagonists. The unique potential of these methods for testing theories of ligand-receptor interaction, with respect to G-protein activation, is discussed.

EXPERIMENTAL PROCEDURES
5HT 2c Receptor Expression in Sf9 Cells-The full-length rat 5HT 2c cDNA (28) was subcloned into the BacPAK8 transfer plasmid in the XbaI and EcoRI positions of the multiple cloning site. Sf9 cells were co-transfected with BSU361-digested BacPAK6 baculovirus DNA and the 5HT 2c -containing recombinant BacPAK8 plasmid by standard techniques (29). After amplification and plaque titer of plaque-purified recombinant baculovirus, Sf9 cells, grown in suspension culture (100 rpm in a Lab-Line Environ-Shaker) with Life Technologies, Inc. Sf-900 serum-free medium at 27°C, were infected during log phase growth (1 ϫ 10 6 cells/ml) with 5-10 plaque-forming units/cell. Cells were harvested 60 h postinfection by sedimentation at 500 rpm for 5 min in a Beckman TJ6 centrifuge and resuspended in 10 mM MOPS, pH 7.5, 1 mM EGTA, 100 M AEBSF at 4°C and incubated on ice for 30 min before cell lysis and membrane homogenization with a Dounce homogenizer (25 strokes with A pestle). Nuclei and cell debris were removed by centrifugation at 2700 rpm in the TJ6 centrifuge for 10 min, and the postnuclear fraction was collected at 40,000 rpm in a Beckman 70.1 Ti rotor for 20 min at 4°C. These membranes were either used immediately for urea extraction or resuspended in 10 mM MOPS, pH 7.5, at about 2 mg/ml protein and frozen at Ϫ80°C for storage.
Urea Treatment of Sf9 Cell Membranes-Sf9 cell membranes were extracted three times by incubation for 30 min on ice in a solution of 6 M urea dissolved in Solution A (50 mM MOPS, pH 7.5, 1 mM EDTA, 3 mM MgSO 4 ) at 0.75 mg/ml membrane protein followed by sedimentation at 40,000 rpm for 30 min in a Beckman 70.1 Ti rotor, with a final wash in Solution A without urea. The urea-washed membranes were resuspended at about 2 mg/ml protein in Solution A, and aliquots were stored at Ϫ80°C.
Purification of G-protein Subunits-G-proteins were isolated from bovine brain, bovine retina, and squid retina. Squid photoreceptor membranes were prepared from 100 frozen, enucleated squid eyes by thawing 25 eye cups per 35 ml of Solution B (10 mM MOPS, pH 7.5, 1 mM EGTA, 3 mM MgSO 4 , 100 mM NaCl) with 100 M AEBSF at room temperature. Eyes were vortexed for 1 min to separate the photoreceptor membranes from the rest of the eye cup, and all subsequent steps were carried out on ice or at 4°C. The cups were removed and the photoreceptor membranes pooled and Dounce-homogenized (B pestle with 10 strokes). The homogenate was pelleted at 20,000 rpm for 10 min in Beckman JA20 rotor and resuspended in Solution B containing 34% sucrose to a final volume of 360 ml. The resuspended homogenate was distributed in 30-ml aliquots into Beckman SW28 centrifuge tubes, overlaid with 7 ml of Solution B, and centrifuged at 28,000 rpm for 30 min. The photoreceptor membranes were collected as a layer floating at the sucrose interface, and the harvested membranes were washed twice by dilution to 80 ml with Solution B and centrifugation at 20,000 rpm in a Beckman JA20 rotor. The final pellet was resuspended to 40 ml with Solution B and stored at Ϫ80°C.
Squid photoreceptor membranes (200 mg of membrane protein) were thawed and washed by dilution with Solution C (20 mM Tris/HCl, pH 8, 1 mM EDTA, 3 mM MgCl 2 , 1 mM DTT) with 100 M AEBSF and collection at 20,000 rpm in a Beckman JA20 rotor. The washed membranes were then extracted at 1 mg/ml protein with Solution C containing 1% cholate by stirring at 4°C for 30 min. The extracted membranes were sedimented at 50,000 rpm for 1 h in a Beckman type 35 rotor, and the supernatant was collected. The pelleted fraction, which contains G-protein-depleted rhodopsin, was washed by dilution with Solution B and centrifugation at 20,000 rpm for 20 min in a Beckman JA20 rotor, resuspended with Solution B, and stored at Ϫ80°C. The cholate-extracted supernatant was treated by sequential addition of MgCl 2 to 11 mM, AlCl 3 to 100 M, and NaF to 10 mM (AMF) and incubated at room temperature for 30 min. The AMF-treated extract was then chilled and applied to a 200-ml (2.6 ϫ 40 cm) column of DEAE-Sephacel (Pharmacia Biotech Inc.) that had been equilibrated with 3-bed volumes of Solution D (20 mM Tris, pH 8, 1 mM EDTA, 3 mM MgCl 2 , 1 mM DTT, 1% sodium cholate, 20 M AlCl 3 , and 10 mM NaF). After application of the extract, the column was washed with 1-bed volume of Solution D and then eluted with a 4-bed volume gradient from 0 to 200 mM NaCl in Solution D, collecting 100 fractions and monitoring protein concentration by absorbance at 280 nm. All chromatography was performed at 4°C. GTP-binding activity of purified ␣ q was detected by reconstitution of chromatography fractions with squid rhodopsin, as described below, and typically eluted around 80 mM NaCl. Squid rhodopsin catalyzed GTP␥S binding to a protein with an apparent molecular mass of 44 kDa, as predicted for squid transducin (␣ q ) (26). The DEAE chromatography fractions were visualized by SDS-polyacrylamide gel electrophoresis, pooled according to purity and activity, concentrated on Amicon PM30 membranes, and subjected to gel filtration on Ultrogel AcA44 (IBF) using Solution E (20 mM Tris, pH 8, 1 mM EDTA, 100 mM NaCl, 0.2% cholate). Fractions were evaluated by functional reconstitution, SDSpolyacrylamide gel electrophoresis, and amido black protein measurement (30). The typical yield for 100 retinas was 2 mg of purified ␣ q . Aliquots of pooled peak fractions were stored at Ϫ80°C. When necessary, ␣ q was concentrated on centricon 30 spin concentrators (Amicon) and desalted over G-50 resin using Solution E.
Functional Reconstitution of in Situ 5HT 2c Receptors or Squid Rhodopsin with ␣ q 2 -The receptor-catalyzed binding of GTP␥S by squid ␣ q was determined by modification of the method described for bovine rhodopsin-transducin (20,21). Reactions were carried out in 12 ϫ 75 siliconized borosilicate glass test tubes at a volume of 50 l of Solution F (50 mM MOPS, pH 7.5, 100 mM NaCl, 1 mM EDTA, 3 mM MgSO 4 , 1 mM DTT, 3 mg/ml bovine serum albumin) containing 1 M GTP␥S with trace [ 35 S]GTP␥S (about 1000 dpm/pmol). Exchange reactions were initiated by addition of Sf9 membranes or squid rhodopsin, incubated in a 30°C water bath for varying times, stopped by addition of 2 ml of Solution G (20 mM Tris/HCl, pH 8, 25 mM MgCl 2 , 100 mM NaCl) at 4°C, and immediately filtered over nitrocellulose membranes on a vacuum manifold by washing four times with 2 ml of cold Solution G. Filters were dried and radioactivity quantitated by liquid scintillation in a Wallac 1219 beta counter.
Quantitation of 5HT 2c Receptor Sites-5HT 2c binding sites were quantitated by analysis of [ 3 H]mesulergine or 125 I-LSD binding to membrane preparations. Radioligand binding assays were carried out in solution F at 30°C identically as functional reconstitution except that reactions were filtered over GF-C glass fiber filters. 125 I-LSD was quantitated on a Wallac 1470 gamma counter. [ 3 H]Mesulergine was counted on the Wallac 1219 beta counter.
Analysis of Data and Curve Fitting-All results shown are representative of data obtained from three or more separate experiments. All curves presented were best fits to a simple exponential model for progress curves or a single site binding model for saturation isotherms using the program "Enzfitter." Materials-The 5HT 2c receptor cDNA was kindly provided by Beth Hoffman (LCB, NIMH). The baculovirus DNA and transfer plasmid were purchased from Clontech. Sf-900 serum-free medium was from Life Technologies, Inc. [N-6-methyl-3 H]Mesulergine was obtained from Amersham Corp. 2-[ 125 I]-iodo-lysergic acid diethylamide and [ 35 S] GTP␥S were from DuPont NEN. Squid (Loligo forbesi) eyes were pur-chased from Calamari, Inc., in Woods Hole, MA. The serine protease inhibitor, [4-(2-aminoethyl)-benzenesulfonylfluoride, HCl] (AEBSF), was purchased from Calbiochem. Cholic acid was obtained from Sigma and repurified as described (34). MOPS was from Fluka. The DEAE-Sephacel and Ultrogel AcA-44 chromatography resins came from Pharmacia and IBF, respectively. Bovine serum albumin used as standard for protein measurements was from Pierce. The hydrochloride salts of mesulergine, mianserin, and ketanserin were obtained from Research Biochemicals, Inc. GTP␥S was from Sigma. HAWP nitrocellulose filters and the vacuum manifold were from Millipore. The GF-C glass fiber filters were from Whatman.

RESULTS
After co-transfection of Sf9 cells with baculoviral DNA and the 5HT 2c -containing transfer plasmid, recombinant baculovirus was plaque-purified twice. Three of three isolated plaques expressed 125 I-LSD-binding sites. One of these was selected, amplified, and used to obtain a viral stock that was utilized in all subsequent infections (29). The time course of 5HT 2c receptor expression was assessed by 125 I-LSD labeling of receptor sites and by activity of membranes reconstituted with ␣ q at 12-h intervals after viral infection. We observed maximal expression at 60 h, by both receptor site abundance and functional activity (data not shown). Infecting with multiplicities of infection of 5 and 40 plaque-forming units/cell provided an equal abundance of 5HT 2c receptor sites at 60 h post-infection (data not shown). 5HT 2c receptor expression was diminished by 20 -40% when recombinant baculovirus-infected Sf9 cells were cultured in the presence of 1 M mianserin, consistent with findings of others (35) (data not shown). Membranes prepared from Sf9 cells infected with the wild-type baculovirus vector did not have specific binding sites for 125 I-LSD or [ 3 H]mesulergine, and they did not catalyze GTP␥S-binding to ␣ q (data not shown).
Infection of a 0.5-liter suspension culture (5 ϫ 10 8 cells) yielded approximately 20 mg of total membrane protein in the postnuclear fraction. Urea extraction removed 60 -70% of the membrane-associated protein. Five infections have yielded specific activities of 50 -70 pmol/mg total membrane protein which were increased to 150 -210 pmol/mg after urea extraction. Fig.  1 shows the results of an experiment analyzing the saturation of 5HT 2c binding sites with [ 3 H]mesulergine in urea-extracted membranes. These data are consistent with the expression of a single class of binding sites with a K d of 5 nM for mesulergine and 190 pmol binding per mg of membrane protein in this preparation.
The ability of Sf9-expressed 5HT 2c receptor to couple functionally to squid photoreceptor ␣ q was examined in the experiment presented in Table I. Urea extraction reduced the GTP␥S-binding activity of the Sf9 membranes by 75%. The functional activity of the receptor to catalyze GTP␥S binding by added ␣ q was retained following urea treatment, and the ratio of this catalyzed binding to the endogenous binding of GTP␥S by the membranes was increased from 3-to 10-fold. Reconstitution of membranes with ␤␥ did not increase GTP␥S binding to endogenous sites but enhanced the binding when co-reconstituted with ␣ q , and neither ␣ q nor ␤␥ nor their combination bound GTP␥S significantly in the absence of the 5HT 2c receptor (data not shown). This is evidence for the presence of an uncoupled receptor and is also consistent with a catalytic role for ␤␥ in 5HT 2c receptor-catalyzed GTP␥S-binding by ␣ q similar to that observed for reconstitution of bovine rhodopsin with ␣ t (21).
5HT activated the 5HT 2c receptor to increase catalysis of GTP␥S binding by ␣ q , but mianserin did not. Upon examining nonextracted membranes containing the 5HT 2c receptor, we noted that the addition of 5HT did not increase GTP␥S binding by the membranes. Even the addition of saturating ␤␥ did not reveal a 5HT-stimulated GTP␥S binding to endogenous Sf9 proteins while significantly enhancing that to added squid retinal ␣ q . Although others have noted an ␣ q endogenous to Sf9 cells (36), the rat 5HT 2c receptor does not appear to couple effectively to this protein. The 5HT 2c receptor in urea-extracted membranes also failed to catalyze GTP␥S binding by bovine ␣ t or bovine brain ␣ o /␣ i fractions (data not shown). We also note that in both the native and urea-extracted membranes, we observed a significant enhancement of GTP␥S binding to ␣ q in the absence of added serotonin and mianserin inhibited this.
The experiment in Table I utilized bovine retinal ␤␥ in order to eliminate detergent additions necessary when reconstituting the hydrophobic ␤␥ structures (21). Retinal ␤␥ samples, as seen by comparing the activity without and with ␣ q addition, contain no detectable ␣ q contamination. However, we have been unable to saturate the ␤␥ enhancement of 5HT 2c activation of ␣ q using the bovine retinal protein (data not shown). Therefore we examined the ␤␥ dependence of this reconstitution using bovine cortical ␤␥ preparations. In preliminary experiments we discovered that such preparations uniformly were contaminated with GTP␥S binding activity which was catalyzed by the 5HT 2c receptor but not by bovine rhodopsin. Neither did these preparations contain detectable pertussis toxin substrate (data not shown). We were able to remove the GTP␥S binding activity from the bovine brain ␤␥ preparations by additional chromatography over fast protein liquid chromatography phenyl-Sepharose, but these experiments were not uniformly successful. As with the retinal ␤␥ fraction, we have failed to saturate the 5HT 2c receptor with such brain ␤␥ fractions, but we can estimate that the K 1 ⁄2 is about 600 nM for these preparations. Parallel analyses of the saturation of bovine rhodopsin-␣ t activation by these same ␤␥ samples indicate K 1 ⁄2 values of 20 -50 nM (data not shown).
Our kinetic analyses of the ligand regulation of the rat 5HT 2c receptor included the 5HT saturation of GTP␥S exchange at differing concentrations of the receptor. Fig. 2 presents data from one of these experiments using 4 and 12 nM receptor. Consistent with a catalytic function for the receptor, the rate of GTP␥S binding to ␣ q was increased at the higher receptor concentration. The concentration of 5HT that gave half-maximal activation of the 5HT 2c receptor was approximately 100 nM at both 4 and 12 nM concentrations of the 5HT 2c receptor. Also seen in this experiment, as in Table I, is a significant 5HT 2c membrane-catalyzed exchange in the absence of 5HT. This activity also increases proportionally with the membrane concentration.
The significant "basal" activity of the recombinant 5HT 2c receptor prompted us to examine antagonist pharmacology with our reconstituted system. The effects of three antagonist ligands (mianserin, ketanserin, and mesulergine) upon 5HT 2c receptor-␣ q interactions, in the presence and absence of 5HT, are shown in Fig. 3, A and B. The results shown in Fig. 3A display the expected rank order of antagonism of 5HT activation of the 5HT 2c receptor based upon published affinities of these antagonists for the 5HT 2c receptor (8). Data presented in Fig. 3B show that ketanserin and mianserin, but not mesulergine, inhibited basal activity of 5HT 2c receptor-catalyzed nucleotide exchange by ␣ q . The IC 50 of these antagonist ligands as inhibitors of the basal activity of the 5HT 2c receptor does not correlate with their affinities for the receptor, i.e. ketanserin, with a lower affinity for the 5HT 2c receptor than mianserin, was similarly potent as an inhibitor. Furthermore, mesulergine, the most potent antagonist of 5HT activation of the  c The concentration of ␣ q was 60 nM based on completion of ␣ q GDP/GTP exchange at 1 h when reconstituted with ␤␥ and 5HT (see Fig. 4). d The concentration of ␤␥ is 100 nM based upon amido black staining. ␤␥ was purified from bovine rod outer segment discs (see "Experimental Procedures").
FIG. 2. 5HT-activation of 5HT 2c receptor and enhancement of GDP/GTP exchange by ␣ q . Exchange reactions were carried out with 60 nM ␣ q and 400 nM brain ␤␥ at 4 nM (E) and 12 nM (å) of 5HT 2c receptor sites with the indicated concentrations of 5HT. Assays were for 5 min and were carried out as described under "Experimental Procedures." At the highest receptor and 5HT concentrations less than 35% of the ␣ q was consumed in the reaction, thus initial rates of reaction were approximated throughout.

FIG. 3. Inhibition by antagonist ligands of 5HT-activated and basally active 5HT 2c receptor-catalyzed GDP/GTP exchange by ␣ q .
A, the antagonism of 5HT activation of the receptor is observed for mianserin (f), ketanserin (E), and mesulergine (q). Sf9 membranes (30 nM of 5HT 2c receptors) were incubated with 125 nM ␣ q , 240 nM brain ␤␥, 1 M 5HT, and the indicated concentrations of antagonist. Assays were performed as described under "Experimental Procedures," and the incubation time was 12 min. B presents inhibition of the basal catalytic activity (no 5HT added) of the 5HT 2c receptor (20 nM) with 125 nM ␣ q and 320 nM brain ␤␥ and the indicated concentrations of ketanserin (E), mianserin (f), or mesulergine (q) and GTP␥S exchange assays conducted as described under "Experimental Procedures." For this experiment the incubation time was 90 min. 5HT 2c receptor, showed no inhibition of 5HT 2c basal activity when the receptor concentration was 20 nM. This strongly suggests that antagonists for the 5HT 2c receptor have differential efficacies as inverse agonists.
To address the molecular bases for the inhibition of "basal" activity of the 5HT 2c receptor, we performed a progress analysis of the activation of ␣ q in the absence of ligand and in the presence of saturating 5HT, mianserin, mesulergine, or ketanserin. These data are presented in Fig. 4. The progress curves were well-fit as simple exponentials. These differed significantly in rate; 50% exchange of the ␣ q was catalyzed in 5 min by the 5HT-activated receptor, in 40 min by the basally active receptor, and in 2 h by the mesulergine-inhibited receptor. In the presence of mianserin or ketanserin, 5HT 2c receptors exhibited detectable but much slower catalytic activities, turning over 25% of the ␣ q in 2 h. These data are consistent with general models of ligand-receptor regulation of the single ratelimiting step of GDP dissociation from ␣ q . In the presence of saturating concentrations of each of the three antagonists, the rates are appreciably lower than for the unoccupied receptor. The rate decrements are quite distinct for each of the antagonists, suggesting varying efficacies of inverse agonism among the three.
Based upon observations made from data shown in Fig. 3B, we first hypothesized that receptors saturated with mesulergine would have the same catalytic activity as the basal receptor, i.e. that mesulergine was a pure antagonist and thus devoid of inverse agonist properties. However, observations from the time course experiment (Fig. 4) contradicted this. It should be noted that in order to enhance the catalytic rates for the progress analyses in the experiment of Fig. 4, we utilized 30 nM receptor while only 20 nM was used for the experiments of Fig. 3. This suggested that mesulergine may be a low efficacy inhibitor of 5HT 2c basal activity (a "partial inverse agonist"), requiring higher concentrations of receptor in order to observe its inverse agonist properties.
The experiment shown in Fig. 5 was designed to test this hypothesis. We measured the catalytic activity of the receptor in the presence or absence of saturating antagonists while varying the concentration of receptor. In the presence of 50 M mesulergine, 5HT 2c receptor concentrations below 30 nM appeared to catalyze ␣ q GDP/GTP exchange at the same rate as the basally active receptor. However at higher concentrations of receptor, an inhibition by mesulergine was observed. Also note that ketanserin displayed slightly greater inhibition of basal activity than mianserin at saturating concentrations of ligand, consistent with its being a somewhat more efficacious inverse agonist. DISCUSSION Cells responding to a diversity of environmental stimuli utilize a repertoire of cell surface receptors and intracellular signaling proteins. Presumably, a cell employs multiple representatives from the various identified families of signal transduction proteins in responding specifically to a complex variety of stimuli. Dissection of the contributions of different receptors, G-proteins, and effector proteins to the regulation of cellular responsiveness has proven to be a formidable task (12). In vitro reconstitution of the various interactions between components of signal transduction cascades is one experimental approach to this question; however, the co-expression of several members of each family of proteins within any cell complicates purification of any single component as it complicates conclusive assignment of function based upon experiments in whole cells. Additionally, because the receptors and G-proteins are the first components of an amplified cascade, they are expressed in low abundance by most cells. Thus, the technical limitations of purifying sufficient quantities of receptors and G-proteins are significant hurdles to overcome for in vitro studies of receptor-G-protein interaction. In this report we have examined a new approach to in situ reconstitution of G-proteincoupled receptors that should address many of the current obstacles for investigating receptor-G-protein coupling.
First, we have developed a method of reconstituting in situ recombinant receptors, which has advantages over previous methods. This preparation of the receptor does not involve The concentration of 5HT 2c receptor was varied from 1 to 150 nM as indicated, and ␣ q and brain ␤␥ were held constant at 175 and 240 nM, respectively. Incubation time was 45 min, and the exchange activity was assessed as described under "Experimental Procedures." At each receptor concentration, data were obtained from equivalent amounts of membranes in the absence of reconstituted G-protein subunits (not shown). These values were used as references to remove nonspecific trapping of GTP␥-35 S by membranes from the analysis. All data points shown have had the respective reference value (ranging from 675 cpm (1 nM) to 11,650 cpm (150 nM) subtracted. Sixty percent of the ␣ q sites present were bound to GTP␥-35 S in the condition with 150 nM of the basally active 5HT 2c receptor. detergent solubilization; thus, the native structure of the receptor is relatively stabilized as it is manipulated, and exposure to chromatographic resins is avoided. Maintaining the native membrane environment into which Sf9 cells incorporate the receptor eliminates the need for phospholipids, which must be used at high concentrations to reconstitute solubilized receptors in artificial liposomes (18,19,24,25). This simplifies the reconstitution and avoids issues regarding specific chemical compositions in liposomes that may be required for proper receptor function. Additionally, we feel kinetic measurements of interactions between in situ receptors and G-protein subunits more closely reflect what occurs physiologically, due to the absence of excess phospholipid and residual detergent that may alter important hydrophobic protein-protein interactions between receptors and G-proteins.
The use of recombinant receptors has obvious advantages for characterizing important regions of molecular interaction with other proteins. Others have utilized prokaryotic expression systems (37). The baculovirus expression system offers the advantages of much higher receptor expression and post-translational protein modification. Urea extraction of nonintegral membrane proteins (60 -70% of total membrane protein) reduced nonspecific GTP binding by membranes 4-fold and increased receptor site density by 3-fold. This uncoupling of the receptor allows for much greater signal to noise when reconstituting with exogenous G-protein subunits. Thus, urea-extraction of membrane preparations may be useful in searching for additional protein factors responsible for modulating receptor coupling.
Using this new system for analyzing receptor-G-protein interaction, we have established that ␣ q purified from squid photoreceptors (26,27) is useful for reconstitution in a mammalian PLC-linked signal transduction pathway. This gene product is expressed in high abundance in squid retina; it can be readily isolated in high yields, and it has been sequenced and cloned. The functional coupling of the 5HT 2c receptor with squid photoreceptor ␣ q provides reciprocal evidence that each is a component of a PLC-PI-mediated signal transduction pathway. The 5HT 2c receptor specifically interacts with ␣ q , as it does not catalyze GDP-GTP exchange on bovine ␣ t or fractions of bovine ␣ 0 /␣ i . Sf9 cells express an endogenous ␣ q that is recognized by antisera to other proteins in the family and that can activate PLC-␤s by reconstitution (36). Interestingly, the rat 5HT 2c receptor did not appear to couple to this protein. The recent expression and purification of four members of the ␣ q family for reconstitution of their ability to activate PLC-␤s (15,36) should provide sources of related but different G-proteins that, in addition to being useful for comparison, can be genetically manipulated to examine important domains of proteinprotein interaction. Squid retinal G q shares 78% sequence identity with mouse ␣ q and ␣ 11 and 74% identity with mouse ␣ 14 . Our data with the squid retinal ␣ q suggest that if the mammalian proteins can be expressed with appropriate posttranslational modification, our in situ reconstitution procedures should allow for the determination of differences among the G␣ q family members. This method also should be applicable to other calcium mobilizing receptors 3 as well as those coupling through other G-protein sub-families.
The recombinant baculovirus expression strategy has also been utilized to obtain functional ␤␥ dimers (22,38). Our studies thus far suggest that the 5HT 2c -␣ q paradigm may be useful for examination of the contribution of recombinant ␤␥ dimers to receptor-␣ subunit interaction. There is an increasing appreciation of the diverse roles that ␤␥ dimers play as determinants of selectivity in cell signaling (39). We observed that a mixture of ␤␥s purified from bovine cortex greatly facilitated the rate of 5HT 2c receptor-catalyzed GTP binding by ␣ q . Interestingly, this preparation of ␤␥ had an apparently low affinity for the 5HT 2c receptor, as we were unable to achieve saturation with the preparations of ␤␥ used in these studies. This may indicate the absence of an appropriate high affinity ␤␥ dimer in the retinal or brain preparations. Alternatively, 5HT 2c may have low intrinsic affinity for ␤␥. Further studies employing recombinant ␤␥ dimers and ␣ q gene products may yield answers to such questions.
The ability to manipulate independently all components of the reconstitution has allowed characterization of the molecular interactions between the 5HT 2c receptor and squid photoreceptor ␣ q and shown modulation of the catalytic state of the receptor by 5HT and three different antagonists. The constitutive catalytic activity of the 5HT 2c receptor made possible the study of inverse agonist properties of 5HT 2c receptor antagonists. Inhibition of basal activity of receptors is increasingly appreciated (40). It has been observed for the ␦-opioid (41), the muscarinic acetylcholine (42), the ␤-adrenergic (43), and the 5HT 2c (35) receptors.
Prior determination of the inverse agonist properties of 5HT 2c antagonists utilized studies of PI hydrolysis following stable transfection of the 5HT 2c DNA into the NIH 3T3 fibroblast cell line (35). Our studies expand upon these findings by examining the first biochemical step in the pathway. Even in the absence of the cascade of amplification provided in the measurement of inositol phosphates in intact cells, we find significant basal activity of the receptor and inverse agonist properties of 5HT 2c receptor antagonists.
Furthermore, as opposed to the intact cell approaches, our in vitro reconstitution allowed rigorous exclusion of 5HT contamination as the basis for the constitutive activity of the 5HT 2c receptor. Our Sf9 cells were cultured serum-free, and the membrane fractions we have employed were extensively washed, including with chaotropic concentrations of urea that extracted two-thirds of the total protein in the fraction. It seems unlikely that the rapidly dissociating ligand 5HT would persist bound to the receptor after such treatment. Moreover, mesulergine did not inhibit the basal activity at low receptor concentrations, whereas mianserin and ketanserin did. If the basal activity were due to a dissociable agonist ligand, all antagonists should appear as inhibitors. The possibility that mesulergine is a weak partial agonist, exactly compensating for the displacement of contaminating 5HT, would not explain the appearance of inverse agonist activity of mesulergine at higher receptor concentrations. Indeed, the alteration of apparent intrinsic activities of partial agonists or partial inverse agonists by variation in receptor abundance is a strong prediction of currently held receptor theory (44 -46). Nearly 40 years ago, it was proposed that ligand affinity and efficacy were independent parameters of receptor activation (47). For agonists with high efficacy, maximal responses were observed when the fractional occupancy of receptors was less than one. This phenomenon of "spare receptors" accounts for what we have observed with 5HT 2c receptor inverse agonists. It is the inverse of what one would see with a high efficacy agonist; rather than observing a maximal response until one goes below a threshold of receptor number, we see no inhibition of basal receptor activity until we supersede a receptor concentration threshold. One may imagine that the readout of this threshold (i.e. GDP-GTP exchange on ␣ q ) may vary with combinations and concentrations of ␣ q and ␤␥. Our ability systematically to vary each of the molecular components in vitro should allow for further detailed molecular analyses of receptor function at the level of G-protein activation and thus for more direct tests of existing receptor theories.