Distinct Structural Changes in a G Protein-coupled Receptor Caused by Different Classes of Agonist Ligands*

The activity of G protein-coupled receptors can be modulated by different classes of ligands, including agonists that promote receptor signaling and inverse agonists that reduce basal receptor activity. The conformational changes in receptor structure induced by different agonist ligands are not well understood at present. In this study, we employed an in situ disulfide cross-linking strategy to monitor ligand-induced conformational changes in a series of cysteine-substituted mutant M3 muscarinic acetylcholine receptors. The observed disulfide cross-linking patterns indicated that muscarinic agonists trigger a separation of the N-terminal segment of the cytoplasmic tail (helix 8) from the cytoplasmic end of transmembrane domain I. In contrast, inverse muscarinic agonists were found to increase the proximity between these two receptor regions. These findings provide a structural basis for the opposing biological effects of muscarinic agonists and inverse agonists. This study also provides the first piece of direct structural information as to how the conformations induced by these two functionally different classes of ligands differ at the molecular level. Given the high degree of structural homology found among most G protein-coupled receptors, our findings should be of broad general relevance.

The superfamily of G protein-coupled receptors (GPCRs) 2 represents the largest group of cell surface receptors found in nature (1)(2)(3)(4)(5). GPCR-dependent signaling pathways play critical roles in regulating an extraordinarily large number of important physiological functions. The major structural hallmark shared by all GPCRs is a transmembrane (TM) core formed by a bundle of seven TM helices (TM I-VII) (1-4) that are connected by three intracellular and three extracellular loops. Drugs acting on specific GPCRs are of great therapeutic relevance, representing ϳ40% of drugs in current clinical use (4).
A better understanding of how GPCRs function at the molecular level is considered essential for the development of novel classes of clinically useful drugs. The structural features of the ligand binding and G protein coupling domains have been studied extensively for a large number of GPCRs (1)(2)(3)(4)6). Likewise, the conformational changes involved in the activation of the photoreceptor rhodopsin have been mapped in considerable detail (7)(8)(9)(10)(11). In contrast, relatively little is known about the sequence of molecular events that trigger the activation of GPCRs that bind diffusible ligands such as neurotransmitters and hormones.
According to their pharmacological characteristics, GPCR ligands are currently classified into agonists, neutral antagonists, and inverse agonists (12)(13)(14). Inverse agonists are drugs that can reduce GPCR-mediated G protein activation observed in the absence of ligands (basal GPCR activity). In fact, accumulating evidence suggests that most classic GPCR antagonists need to be reclassified as inverse agonists (12)(13)(14). At present, little is known about the nature of the structural changes that inverse agonists induce in their target receptors. Biophysical studies carried out with different adrenergic receptor subtypes strongly suggest that adrenergic agonists endowed with different efficacies, including inverse agonists, induce or stabilize different receptor conformations (15)(16)(17)(18)(19)(20). However, how these various conformations differ at the molecular level remains unclear at present.
To shed light on this issue, we have used the rat M 3 muscarinic acetylcholine receptor, a prototypic class I GPCR (21,22), as a model system. To monitor ligand-induced changes in receptor structure, we employed an in situ disulfide cross-linking strategy that allows the detection of disulfide bond formation between Cys residues that are adjacent to each other in the three-dimensional structure of the receptor (23)(24)(25)(26). One major advantage of this strategy is that ligand-dependent conformational changes can be detected in receptors present in their native membrane environment, without the need for any receptor purification and reconstitution steps.
In this study, we examined whether different classes of muscarinic ligands (full versus inverse muscarinic agonists) had different effects on the relative orientation of helix 8 relative to the C terminus of TM I (Fig. 1). Helix 8 represents a cytoplasmic ␣-helical extension of TM VII to which it is connected via a short linker sequence ( Fig. 1) (27). Considerable evidence suggests that helix 8 plays an important role in productive receptor/G protein coupling (6, 28 -31). High resolution structural data of bovine rhodopsin (27), complimented by biophysical and biochemical studies (11), indicate that several residues contained within helix 8 are located close to the cytoplasmic end of TM I. We therefore hypothesized that Cys residues substituted into this segment of TM I might serve as useful reporters to detect potential ligand-induced movements of helix 8 in disulfide cross-linking studies.
Specifically, we introduced pairs of Cys residues into a modified version of the M 3 muscarinic receptor that lacked most native Cys residues and contained two factor Xa cleavage sites within the third intracellular loop (this receptor is referred to as "M3Ј(3C)-Xa" receptor throughout this study; see "Experimental Procedures" for details; Fig. 1). Previous studies have shown that the M3Ј(3C)-Xa receptor exhibits ligand binding and G protein coupling properties similar to the wild-type M 3 muscarinic receptor (35). We generated 20 double Cys mutant M 3 receptors, all of which contained one Cys substitution within the cytoplasmic end of TM I (Ala 91 -Asn 95 ) and a second one within the N-terminal segment of helix 8 (Lys 548 -Arg 551 ; Fig. 1).
Disulfide cross-linking studies using membranes prepared from transfected COS-7 cells showed that muscarinic agonists and inverse muscarinic agonists had different effects on the efficiency of disulfide bond formation in specific double Cys mutant M 3 receptors. In conjunction with a three-dimensional model of the M 3 muscarinic receptor, this study provides the first piece of direct structural information as to how the receptor conformations induced (or stabilized) by GPCR agonists and inverse GPCR agonists differ from each other at the molecular level.
Construction of Cys-substituted Mutant M 3 Muscarinic Receptors-All Cys substitutions were introduced into a pCDbased expression plasmid coding for a modified version of the rat M 3 muscarinic receptor referred to as M3Ј(3C)-Xa (35) (Fig.  1). The M3Ј(3C)-Xa receptor construct contains an N-terminal hemagglutinin epitope tag and lacks all five potential N-terminal N-glycosylation sites and most endogenous Cys residues, except for Cys 140 , Cys 220 , and Cys 532 , as described previously (35). In addition, the central portion of the i3 loop (Ala 274 -Lys 469 ) was replaced by two factor Xa cleavage sites. Cys residues were substituted into the M3Ј(3C)-Xa construct by using the QuikChange TM site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. The identity of all mutant receptors was confirmed by sequencing the entire receptor coding sequences.
Transient Expression of Cys-substituted Mutant M 3 Muscarinic Receptors in COS-7 Cells-The M3Ј(3C)-Xa receptor construct and all M3Ј(3C)-Xa-derived Cys-substituted mutant receptors were transiently expressed in COS-7 cells. Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C in a humidified 5% CO 2 incubator. About 24 h prior to transfections, ϳ1 ϫ 10 6 cells were seeded into 100-mm dishes. Cells were transfected with 4 g/dish of receptor plasmid DNA using the Lipofectamine Plus kit (Invitrogen), according to the manufacturer's instructions. To achieve higher receptor expression levels, transfected cells were incubated with 1 M atropine for the last 24 h of culture, as described previously (23)(24)(25)(26).
Preparation of Membranes from Transfected COS-7 Cells-To prepare cell membranes, COS-7 cells were harvested ϳ48 h after transfections. Initially, cells were washed twice (10 min each wash) with 10 ml of ice-cold phosphate-buffered saline, pH 7.4. This washing step was included to ensure the complete removal of atropine that was present in the culture medium during the last 24 h of culture. Subsequently, 2 ml of ice-cold buffer A (25 mM sodium phosphate and 5 mM MgCl 2 , pH 7.4) was added to each 100-mm dish, followed by a 15-min incubation at 4°C. Cells were then scraped off the plates and homogenized using a Polytron tissue homogenizer (setting 5; 20 s). After a 15-min centrifugation at 20,000 ϫ g (4°C), membrane pellets were resuspended in buffer A (1 ml/100-mm dish), rehomogenized, frozen on dry ice, and stored at Ϫ70°C until use. Protein concentrations were determined using the Micro BCA protein assay reagent kit using bovine serum albumin as a standard.
Radioligand Binding Studies-The NMS and carbachol binding properties of the receptors analyzed in this study were determined in radioligand binding assays. [ 3 H]NMS saturation and carbachol competition binding studies were carried out essentially as described previously (24). In brief, membrane homogenates prepared from transfected COS-7 cells (ϳ10 -20 g of membrane protein per tube) were incubated with the muscarinic antagonist/inverse agonist, [ 3 H]NMS, for 2 h (22°C) in 1 ml of buffer A. In saturation binding assays, six different [ 3 H]NMS concentrations ranging from 20 to 3,000 pM were employed. In competition binding assays, a fixed concentration of [ 3 H]NMS (500 pM) was present in all tubes, using 10 different concentrations of the cold competitor, carbachol. Nonspecific binding was assessed as binding remaining in the presence of 1 M atropine. Binding reactions were terminated by rapid filtration over GF/C Brandel filters, followed by three washes (ϳ4 ml per wash) with ice-cold distilled water. The amount of radioactivity that remained bound to the filters was determined by liquid scintillation spectrometry. To analyze the saturation and carbachol competition binding data, the nonlinear curve-fitting program Prism 4.0 (GraphPad) was used.
Measurement of Receptor-mediated Phosphatidylinositol Hydrolysis-To study whether the different mutant receptors analyzed in this study retained the ability to activate G proteins, we determined carbachol-mediated increases in intracellular inositol monophosphate (IP) levels. Transiently transfected COS-7 cells grown in 6-well plates were labeled with myo-[ 3 H]inositol (3 Ci/ml) for the last 20 -24 h of culture. On the day of the assay, the labeled cells were incubated with increasing concentrations of carbachol in the presence of 10 mM LiCl for 1 h at 37°C. The IP fraction was isolated and quantitated as described previously (24). The nonlinear curve-fitting program Prism 4.0 (GraphPad) was used to derive EC 50 and E max values from carbachol concentration-response curves.
Disulfide Cross-linking, Solubilization, and Factor Xa Digestion of Cys-Substituted M 3 Muscarinic Receptors-Disulfide cross-linking studies were carried out as described in detail previously (24). In brief, receptor-containing membranes prepared from one 100-mm dish (ϳ1 mg of protein) were suspended in 1 ml of buffer A containing 25 M Cu(II)-phenanthroline, either in the presence or in the absence of different muscarinic ligands. Reactions were carried out for 10 min at room temperature (22°C) and then terminated by the addition of EDTA and N-ethylmaleimide (10 mM each), followed by a 10-min incubation on ice. Membrane proteins were then solubilized by incubating samples with 1.2% digitonin, as described by Han et al. (24). Receptor-containing membrane lysates (ϳ15 g of protein) were then treated in 30 l of factor Xa digestion buffer with factor Xa protease (final concentration, 0.1 g/l) at room temperature for 16 h. Reactions were terminated by addition of 1 l of mammalian protease inhibitor mixture (1:25 dilution; see Ref. 24). Samples were then stored at Ϫ70°C or used directly for SDS-PAGE.
Urea Treatment of Receptor-containing Membranes-To inactivate heterotrimeric G proteins, membranes prepared from transfected COS-7 cells were treated with a high concentration of urea (23,33). Receptor-containing membranes were incubated on ice for 30 min either in the presence or in the absence of 5 M urea and then used for disulfide cross-linking studies, exactly as described by Ward et al. (23).
Western Blot Analysis-SDS-PAGE was carried out as described previously (23). Factor Xa-treated membrane lysates were incubated with Laemmli loading buffer for 30 min at 37°C, either under nonreducing or under reducing conditions (in the absence or presence of 50 mM DTT, respectively). Samples were then immediately loaded onto 10 -20% Tris-glycine polyacrylamide gels and run at 125 V in the presence of 0.1% SDS. Western blotting studies were performed essentially as described (24), using a rabbit polyclonal anti-M 3 receptor antibody directed against the C-terminal 18 amino acids of the rat M 3 receptor sequence (34). M 3 receptor proteins were visualized by using enhanced chemiluminescence detection reagents and autoradiography. To quantitate the intensities of immunoreactive bands, we used scanning densitometry employing the program NIH ImageJ.

Generation of Cys-substituted Mutant M 3 Muscarinic
Receptors-All Cys substitutions were introduced into the M3Ј(3C)-Xa construct, a modified version of the M 3 muscarinic receptor lacking most native Cys residues and containing a pair of factor Xa cleavage sites within the third intracellular loop (Fig. 1). We generated 20 double Cys mutant M 3 receptors harboring one Cys substitution within the cytoplasmic end of TM I (Ala 91 -Asn 95 ) and a second one within the N-terminal segment of helix 8 (Lys 548 -Arg 551 ; Fig. 1; Table 1). All receptor constructs were transiently expressed in COS-7 cells and subsequently characterized in pharmacological and biochemical studies.

Analysis of Mutant M 3 Muscarinic Receptors in Radioligand
Binding and Functional Studies-To increase mutant M 3 receptor expression levels, transfected COS-7 cells were incubated with atropine (1 M) for the last 24 h of culture (23)(24)(25)(26). We first characterized the ligand binding properties of the 20 double Cys mutant M 3 receptors in radioligand binding studies, using [ 3 H]NMS as a radioligand. NMS, like atropine and essentially all other classic muscarinic antagonists, has recently been reclassified as an inverse muscarinic agonist (36,37). [ 3 H]NMS saturation binding studies with membranes prepared from transfected COS-7 cells showed that all double Cys mutant M 3 receptors were expressed at levels Ն1 pmol/mg membrane protein (Table 1). Moreover, all receptors were able to bind  Table 1). To determine the affinity of the different mutant M 3 receptors for the muscarinic agonist carbachol, we carried out [ 3 H]NMS/carbachol inhibition binding experiments. These studies showed that all double Cys mutant M 3 receptors were able to bind carbachol with affinities that were similar to or even higher than that observed with the M3Ј(3C)-Xa receptor from which all mutant receptors were derived (range of carbachol K i values, ϳ1-14 M; Table 1).
To examine whether the analyzed mutant receptors retained the ability to couple to G proteins, we studied their ability to mediate carbachol-induced increases in IP production. As shown in Table 1, carbachol stimulation led to a significant increase in IP accumulation in all mutant receptors studied. With the exception of the N95C/R551C construct, which showed a very low carbachol potency (EC 50 ϭ 12,100 nM), all other double Cys receptors mediated carbachol-induced IP production with high agonist potency (EC 50 Յ 423 nM; Table 1). Several of the analyzed mutant receptors showed a significant reduction in maximum IP responses (E max ; F92C/K548C, F92C/T549C, F92C/R551C, K93C/K548C, and K93C/T549C; Table 1), consistent with the demonstrated role of helix 8 in modulating the efficiency of receptor/G protein interactions (6, 28 -31).
The A91C/T549C and F92C/F550C Receptors Show Disulfide Cross-linking That Is Inhibited by Carbachol Treatment-To reveal ligand-induced conformational changes in the M 3 receptor structure, we examined whether the different double Cys mutant receptors underwent ligand-dependent changes in disulfide cross-linking. To promote the formation of disulfide bonds, receptor-containing membrane preparations were incu-bated in the presence of a low concentration (25 M) of the oxidizing agent Cu(II)-phenanthroline. Incubations were carried out either in the presence or in the absence of different muscarinic ligands. Following complete digestion with factor Xa, receptor proteins were visualized via Western blotting (nonreducing conditions), using an antibody directed against the C terminus of the M 3 receptor ( Fig. 1; see "Experimental Procedures" for details). Under these conditions, the appearance of a full-length receptor band (ϳ38 kDa) is indicative of successful disulfide cross-linking (23)(24)(25)(26).
An initial screening of the 20 double Cys mutant receptors showed that two of the mutant receptors, A91C/T549C and F92C/F550C, gave a pronounced cross-linking signal even in the absence of ligand ( Fig. 2A). Strikingly, incubation with the full muscarinic agonist, carbachol (1 mM), led to a significant reduction in the intensity of this signal in both mutant receptors ( Fig. 2A). Six additional mutant receptors also displayed some degree of disulfide cross-linking (V94C/T549C, V94C/ F550C, V94C/R551C, N95C/T549C, N95C/F550C, and N95C/ R551C; Fig. 2A). However, the intensity of these cross-linking signals (ϳ38-kDa bands) remained unaffected by carbachol (1 mM) treatment ( Fig. 2A). A three-dimensional model of the M 3 muscarinic receptor (24) suggests that the two Cys residues contained in each of these mutant receptors are relatively far apart (Ͼ12 Å) and do not face each other in the inactive state of the receptor (also see Fig. 9), suggesting that the relatively faint cross-linking signals observed with these mutant receptors may be due to the presence of partially unfolded (or misfolded) receptor subpopulations. Fig. 3 shows that all 20 double Cys mutant receptors could be detected readily in Western blotting studies, consistent with the results of the radioligand binding experiments (Table 1). In this case, samples were treated in exactly the same fashion as described above ( Fig. 2A), except that the factor Xa digestion step was omitted. Taken together, these data suggest that the absence of disulfide cross-linking signals in Fig. 2A is not caused by the poor expression of specific mutant receptor proteins.
After the initial screen of the 20 double Cys mutant receptors ( Fig. 2A), the A91C/T549C and F92C/F550C receptors, which showed carbachol-sensitive disulfide cross-linking, were studied in greater detail. By using scanning densitometry of immunoreactive bands, we estimated that about 20 -30% of all A91C/  Table 2 for a quantitative analysis of carbachol-mediated inhibition of disulfide bond formation in the A91C/T549C and F92C/F550C receptors). The full-length receptor bands were no longer detectable when Western blotting studies were carried out under reducing conditions (in the presence of 50 mM DTT; data not shown; note that actual Western blots involving the use of carbachol and the A91C/T549C and F92C/F550C mutant receptors run under both nonreducing and reducing conditions are included in Fig. 5 for control purposes). These observations suggest that the full-length A91C/T549C and F92C/F550C receptor bands that could be observed after factor Xa treatment under nonreducing conditions were because of the formation of disulfide bridges rather than incomplete digestion by factor Xa.
The A91C/T549C and F92C/F550C Receptors Do Not Form Intermolecular Cross-links-Like other GPCRs (38), the M 3 muscarinic receptor has been shown to form dimers or oligomers (39,40). To exclude the possibility that the disulfide bonds formed by the A91C/T549C and F92C/F550C receptors were because of inter-rather than intra-molecular interactions, we carried out an additional set of cross-linking studies. We coexpressed M3Ј(3C)-Xa-based mutant receptors containing the A91C and F92C single point mutations with M3Ј(3C)-Xa-derived constructs containing the T549C and F550C single point mutations, respectively. In this case, no specific disulfide cross-linking signal was observed, either in the absence or in the presence of carbachol (1 mM; Fig. 4, A and  B, top rows). The A91C/T549C and F92C/F550C double Cys receptors gave cross-linking signals (full-length receptor bands) that were significantly reduced in intensity after treatment of receptor-expressing membranes with carbachol (1 mM ;   Fig. 4, A and B, top rows), in agreement with the results shown in Fig. 2. All Cys mutant receptors used for these studies could be detected readily in Western blotting studies when the factor Xa incubation step was omitted (Fig. 4, A and B, lower rows). Taken together, these observations support the concept that the Cys residues contained in the A91C/T549C and F92C/F550C constructs formed intra-molecular rather than inter-molecular disulfide bonds.
Muscarinic Agonists and Inverse Muscarinic Agonists Exert Opposite Effects on Disulfide Bond Formation in the A91C/T549C and F92C/F550C Receptors-We next examined the effects of several other muscarinic ligands on the disulfide cross-linking signals (ϳ38-kDa bands) observed with the A91C/T549C and F92C/F550C receptors. Like carbachol, the full muscarinic agonist, oxotremorine-M (1 mM), inhibited the formation of disulfide cross-links in both receptors (Fig. 5). Strikingly, treatment of receptor-containing membrane preparations with atropine and NMS, two inverse muscarinic agonists, led to opposite effects on disulfide bond formation in the A91C/T549C and F92C/F550C receptors (Fig. 5). Traditionally, atropine and NMS have been considered classic muscarinic antagonists. However, recent studies have shown that these agents can suppress signaling by constitutively active mutant muscarinic receptors (36,37), indicating that these ligands need to be reclassified as inverse muscarinic agonists. As shown in Fig. 5, atropine or NMS treatment (100 nM each) led to significantly enhanced disulfide cross-linking in the A91C/T549C and F92C/F550C receptors. When Western blotting studies were carried out under reducing conditions (in the presence of 50 mM DTT), the full-length receptor bands observed under nonreducing conditions were no longer detectable (Fig. 5), confirming the involvement of disulfide bridges. Additional studies indicated that NMS facilitated the formation of disulfide cross-links in these two receptors in a concentration-

muscarinic receptors analyzed in this study
The indicated double Cys mutant M 3 muscarinic receptors were transiently expressed in COS-7 cells. All mutant receptors were derived from the M3Ј(3C)-Xa construct (see "Experimental Procedures"). Radioligand binding studies and functional assays (carbachol-mediated IP production) were carried out as detailed under "Experimental Procedures." Data are given as means Ϯ S.E. from two to five independent experiments, each performed in duplicate.  Table 2 for a quantitative analysis of NMS-mediated stimulation of disulfide bond formation in the A91C/T549C and F92C/F550C receptors).

Ligand-dependent Changes in Disulfide Cross-linking Patterns Remain Unaffected by Inactivation of Heterotrimeric G
Proteins-To exclude the possibility that the ligand-induced inhibitory or stimulatory effects on disulfide bond formation observed with the A91C/T549C and F92C/F550C mutant receptors were affected by precoupling of the receptors to heterotrimeric G proteins, we carried out additional cross-linking experiments. Specifically, we incubated receptor-containing membranes with a high concentration (5 M) of the chao-tropic agent urea. Studies with the M3Ј(3C)-Xa receptor construct and other GPCRs have shown that this treatment leads to the almost complete inactivation or removal of heterotrimeric G proteins, without affecting the function of the uncoupled receptors (23,33,41). After incubation of membranes expressing the A91C/T549C and F92C/F550C receptors with urea (5 M, 30 min on ice), samples were treated with Cu(II)-phenanthroline, solubilized, digested with factor Xa, and subjected to Western blotting analysis (nonreducing conditions). We found that urea treatment had no significant effect on the ability of the agonist, carbachol (1 mM), to inhibit disulfide cross-linking in the A91C/T549C and  (25 M) in the absence or the presence of the muscarinic agonist, CCh (1 mM), digested with factor Xa, and subjected to Western blotting analysis (nonreducing conditions), using the anti-M 3 antibody. All bands shown correspond to the ϳ38-kDa full-length receptor species, which is indicative of successful disulfide cross-linking (23)(24)(25)(26). Note that two of the investigated receptors, A91C/T549C and F92C/F550C, gave pronounced cross-linking signals, the intensity of which was significantly reduced in the presence of CCh. B, CCh reduces disulfide cross-link formation in the A91C/T549C and F92C/F550C receptors in a concentration-dependent fashion. Except for the use of different CCh concentrations, all other experimental conditions were the same as in A. In the experiment shown here, the cross-linking signal displayed by the F92C/F550C receptor was reduced by 36% at the highest CCh concentration used (10 mM), as compared with the control sample (no CCh treatment). The bands shown correspond to the ϳ38-kDa full-length receptor species that can be observed after successful disulfide cross-linking (23)(24)(25)(26). All Western blots shown are representative of three independent experiments. In each individual experiment, identical amounts of proteins were loaded in each lane (ϳ5 g/lane).

TABLE 2 Quantification of the effects of carbachol and NMS on disulfide bond formation in the A91C/T549C and F92C/F550C mutant receptors
Membranes were prepared from COS-7 cell expressing the A91C/T549C and F92C/F550C mutant receptors and processed for disulfide cross-linking and Western blotting studies (nonreducing conditions), as described under "Experimental Procedures." Disulfide cross-linking studies were carried out either in the absence of ligands or in the presence of increasing concentrations of the full muscarinic agonist, carbachol, or the inverse muscarinic agonist, NMS. The intensities of immunoreactive bands corresponding to cross-linked receptors were determined by scanning densitometry (NIH ImageJ). Data were analyzed by using the nonlinear curve-fitting program Prism 4.0 (GraphPad). In each individual experiment, the extent of disulfide cross-linking observed in the absence of ligands was set equal to 100%. Data are given as means Ϯ S.E. of three independent experiments.

Receptor
Carbachol NMS F92C/F550C receptors (Fig. 7). Similarly, the ability of the inverse agonist, NMS (100 nM), to promote disulfide cross-linking in these two receptors remained unaffected after incubation with urea (Fig. 7). These findings suggest that the observed ligand-specific effects on disulfide cross-linking were caused by conformational changes intrinsic to the A91C/T549C and F92C/F550C receptor proteins, rather than by disruption of precoupled receptor-G protein complexes.

Atropine and NMS Act as Inverse Agonists in a Functional
Assay-To examine whether atropine and NMS acted as inverse agonists at the A91C/T549C and F92C/F550C receptors, we carried out a series of functional studies (IP assays). Incubation of cells expressing either of the two receptors (A91C/T549C or F92C/F550C) with atropine or NMS (10 M each) led to a significant reduction in basal IP accumulation (Fig. 8), consistent with the concept that atropine and NMS act as inverse muscarinic agonists at these receptors. Moreover, atropine and NMS also suppressed basal IP accumulation in cells transfected with the M3Ј(3C)-Xa receptor from which the two double Cys mutant receptors were derived (Fig. 8). How-ever, the A91C/T549C and F92C/F550C mutant receptors exhibited greater basal activity than the M3Ј(3C)-Xa receptor (Fig. 8), suggesting that the introduced Cys substitutions , and subjected to Western blotting analysis (nonreducing conditions), using the anti-M 3 antibody. Another set of samples was processed in the same fashion, except that the factor Xa digestion step was omitted (lower rows), to ensure that all receptors were detectable via Western blotting under the chosen experimental conditions. All bands shown correspond to the ϳ38-kDa fulllength receptor species. In the Xa-digested samples (upper rows), these bands are indicative of successful disulfide cross-linking (23)(24)(25)(26). All Western blots shown are representative of three independent experiments. Identical amounts of proteins were loaded in each lane (ϳ5 g/lane).

DISCUSSION
During the past few years, we have applied an in situ disulfide cross-linking strategy to gain insight into agonist-induced conformational changes in the M 3 muscarinic receptor, a prototypical class I GPCR. In a recent study (25), we demonstrated that the binding of full muscarinic agonists leads to a structural change adjacent to the acetylcholine-binding site that increases the proximity of the extracellular segments of TM III and VII. Most likely, this agonist-induced structural change represents one of the early conformational events leading to the more pronounced structural changes predicted to occur on the intracellular receptor surface (for recent reviews, see Refs. 11 and 42), ultimately triggering productive receptor/G protein coupling. Additional disulfide cross-linking studies with Cys-substituted mutant M 3 muscarinic receptors suggested that these changes involve a rotational movement of the cytoplasmic end of TM VI (23,26), consistent with previous biophysical and biochemical studies carried out with the photoreceptor, rhodopsin (7,11), and the ␤ 2 -adrenergic receptor (43)(44)(45). We also obtained data indicating that agonist activation of the M 3 muscarinic receptor leads to an increase in the proximity of the cytoplasmic ends of TM I and TM VII (24). Collectively, these structural changes are thought to enable heterotrimeric G proteins to contact previously inaccessible M 3 receptor surfaces or residues, thus leading to productive receptor/G protein coupling.
In this study, we demonstrated that muscarinic agonists inhibited disulfide cross-linking in the A91C/T549C and F92C/ F550C double Cys mutant M 3 receptors. In fact, this is the first time that we observed agonist-mediated inhibition of disulfide bond formation in Cys-substituted mutant M 3 receptors. The A91C/T549C and F92C/F550C receptors retained the ability to bind muscarinic ligands with high affinity and to couple to G proteins with high efficiency, indicating that the different Cys substitutions did not interfere with proper receptor folding. Fig. 9 shows a three-dimensional model of the cytoplasmic surface of the M 3 muscarinic receptor, established via homology modeling using the high resolution x-ray structure of bovine rhodopsin as a template (24,27). Highlighted in yellow in Fig. 9 are the key positions involved in ligand-modulated disulfide cross-linking, Ala 91 and Phe 92 at the cytoplasmic end of TM I, and Thr 549 and Phe 550 at the N terminus of helix 8,   . Predicted location of M 3 muscarinic receptor residues targeted in this study. A three-dimensional model of the inactive state of the rat M 3 muscarinic receptor was built via homology modeling using the high resolution x-ray structure of bovine rhodopsin as a template (24,27). A cytoplasmic view of a selected region of the intracellular receptor surface is shown. The C-␣ atoms and side chains of the nine residues that were subjected to Cys substitution mutagenesis are highlighted. Our cross-linking data suggest that muscarinic agonists increase the distance between residues 91 and 549 and between residues 92 and 550 (these four residues are highlighted in yellow). In contrast, inverse muscarinic agonists are predicted to reduce the distance between these residues. Amino acid position numbers according to the Ballesteros/Weinstein GPCR nomenclature (49)  respectively. Fig. 9 indicates that residues 91/92 lie adjacent to residues 549/550 in the inactive state of the M 3 receptor (estimated distance between C␣ atoms, ϳ9 -11 Å). Our data therefore strongly support a model in which full muscarinic agonists trigger a separation of the N-terminal segment of helix 8 from the cytoplasmic end of TM I, thus preventing the formation of disulfide cross-links between Cys residues introduced at positions 91/549 and 92/550.
Our findings are in agreement with the outcome of a sitedirected spin labeling study using Cys-substituted purified mutant rhodopsin proteins (46). In this study, Yang et al. (46) demonstrated that illumination increased the distance between spin labels attached to a Cys residue introduced at the cytoplasmic end of TM I (H65C) and a naturally occurring Cys residue (Cys 316 ) present in the middle of helix 8. It is therefore likely that the observed separation between helix 8 and the cytoplasmic end of TM I is a feature associated with the activation of other class I GPCRs.
Helix 8 is connected to the C terminus of TM VII via a short linker sequence (27) (Fig. 1). Studies with the M 3 muscarinic receptor (24) and bovine rhodopsin (11,47,48) have demonstrated that GPCR activation is associated with conformational changes within the cytoplasmic end of TM VII. We recently presented disulfide cross-linking data indicating that agonistinduced M 3 receptor activation leads to a conformational change that increases the proximity between the cytoplasmic ends of TM I and VII (24). In addition, the observed disulfide cross-linking pattern suggested that M 3 receptor activation is associated with a rotational movement of the cytoplasmic end of TM VII (24). A likely scenario therefore is that the agonistinduced conformational changes in TM VII are propagated to helix 8 via the short linker sequence connecting these two receptor regions (Fig. 1).
The concept that TM VII and helix 8 are structurally and functionally interconnected is also supported by biophysical and biochemical studies carried out with bovine rhodopsin. The crystal structure of bovine rhodopsin (27) indicates that Tyr 306 in TM VII (corresponds to Tyr 543 in the M 3 receptor) forms a hydrophobic interaction with Phe 313 in helix 8 (corresponds to Phe 550 in the M 3 receptor). These two aromatic residues are highly conserved among class I GPCRs. A mutant version of rhodopsin in which these two sites were linked via a disulfide bond was unable to activate G proteins, suggesting that rhodopsin activation requires a separation between these two residues (48).
Studies with different GPCRs have shown that helix 8 contains several residues that play a critical role in the efficiency of receptor/G protein interactions (6, 28 -31). It is therefore likely that the agonist-induced reorientation of helix 8 enables specific helix 8 residues to productively interact with heterotrimeric G proteins. It is also possible that the agonist-induced separation between the N-terminal segment of helix 8 and the cytoplasmic end of TM I removes structural restraints, allowing G proteins to productively interact with other functionally critical receptor domains.
Currently, GPCR ligands are subdivided into three major classes, agonists, neutral antagonists, and inverse agonists (12)(13)(14). Understanding the structural basis underlying this func-tional diversity of ligands is considered essential for the development of novel classes of therapeutically useful drugs (12)(13)(14). Although agonist-induced structural changes have been mapped in considerable detail in at least some class I GPCRs (see Refs. 11 and 42 for recent reviews), little is known about the structural details of the receptor conformations induced or stabilized by inverse agonists.
Interestingly, Gether et al. (15) showed that agonist treatment of purified ␤ 2 -adrenergic receptors carrying a fluorescence tag led to a decrease in fluorescence, whereas inverse agonists induced a small increase in base-line fluorescence. Moreover, using a fluorescence-based approach, Vilardaga et al. (17) recently demonstrated that inverse adrenergic agonists induce structural changes in the ␣ 2A -adrenergic receptor that differ in character and kinetics from those caused by agonist ligands. Although these studies provided important novel mechanistic insights, they did not reveal any structural details regarding the molecular nature of the underlying conformational differences. In contrast, we here provide the first piece of detailed structural information indicating how the conformational changes induced by muscarinic agonists and inverse muscarinic agonists differ from each other at the molecular level.
In this study, we used atropine and NMS as inverse muscarinic agonists. These agents, historically considered prototypic muscarinic antagonists, have been reclassified as inverse muscarinic agonists, because they are able to reduce receptor/G protein coupling that can be observed under certain experimental conditions in the absence of agonist ligands (36,37) (Fig.  8). In previous studies (23,24,26), atropine treatment did not lead to significant changes in the efficiency of disulfide bond formation in any of the investigated double Cys mutant M 3 muscarinic receptors. However, in this study, we demonstrated that atropine and NMS enhanced disulfide bond formation in the A91C/T549C and F92C/F550C mutant receptors. Interestingly, these two receptors showed reduced disulfide cross-linking in the presence of muscarinic agonists (see above). These findings strongly suggest that inverse muscarinic agonists, in contrast to muscarinic agonists, decrease the distance between the cytoplasmic end of TM I and the N-terminal portion of helix 8 (Fig. 9).
In summary, we demonstrated that muscarinic agonists and inverse muscarinic agonists induce distinct conformational changes in the M 3 receptor protein. Our study provides the first piece of direct structural information as to how these conformations differ from each other at the molecular level. Class I GPCRs are known to share a considerable degree of structural homology, which is particularly high among receptors activated by biogenic amine ligands, including the muscarinic and adrenergic receptors. It is therefore likely that the findings reported here are also applicable to other class I GPCRs.