The High Affhity State of the Pz-Adrenergic Receptor Requires Unique Interaction between Conserved and Non-conserved Extracellular Loop Cysteines*

A disulfide bond between two extracellular cysteines, conserved in all G-protein-coupled receptors, is believed to be critical for stabilization of the ligand-binding pocket. The &-adrenergic receptor (&-AR) contains two conserved cysteines (CYS'~' and Cys") as well as two other extracellular cysteines ( C ~ S ' ' ~ and Cys'"). The specificity of the interactions between these four cysteines has not yet been clearly established. Mutants en- coding alanines for specific extracellular cysteines in the &-AR gene were constructed and expressed in COS-1 and Chinese hamster ovary cells. Ala106, Ala1s4*1w)Je1, and Ala1"Js47'soJ91 mutants displayed low affinity for the &antagonist, 12SI-cyanopindolol and in- sensitivity to dithiothreitol (DTT). The Ala106*'91 mutant displayed an intermediate affinity and D'N' sensitivity. Mutants Ala". Ala184Js0, and displayed high affinity and DTT sensitivity, indicating that a solvent-accessible disulfide bond(s) is present in these mutant receptors as in the wild-type &AR. Additionally, thermal stability studies provided evidence that the extra- cellular disulfide bonds are essential for stabilization of the high affinity state of the receptor. These studies in- dicate that the covalent linkage between loops 1 and 2 of the &-AR extracellular domains involves of on Binding-To examine the effect of on lZ5I-CYP or were with 0.1, 1, 5, or 10 m~ 37 preincubation, (50 was incu- 37 for 60 with lZ5I-CYP at least 4-fold higher than its respective Kd value. examine the time course effect of DTT, wild-type and mutant receptors were preincubated with 1 m~ method (27). 5-50 pg of membrane protein were used in the assay. The adenylate cyclase activity of the membranes expressing each receptor was determined both basally and after stimulation with M (-)-isoproterenol. To standardize the ad- enylate cyclase activity among the different membrane preparations, adenylate cyclase was stimulated maximally with 10 p~ forskolin.

The abbreviations used are: G-proteins, guanine nucleotide binding threitol; CYF' , cyanopindolol; PAGE, polyacrylamide gel electrophoresis. The nomenclature of single residue mutants is Ala (alanine) preceding the residue number. Mutants containing multiple Cys + Ala changes serial number as designated in Table I. are designated extracellular cysteine to alanine, ECCA followed by a photoreceptor, rhodopsin. Therefore, these receptors are presumed to be governed by similar structure-function relationships (2). A large number of G-protein-coupled receptors are believed to share the basic structural motif of seven transmembrane a-helical segments that form the binding pocket for the ligand. Several other structural features are conserved in the entire family of G-protein-coupled receptors, prompting the basis for a fundamental relationship in their functioning. Among these conserved structural features, the covalent linking of the first and second extracellular loops (loops B-C and D-E) by a disulfide bond is thought to be important for the generation and stabilization of receptor structure in the entire family of Gprotein-coupled receptors.
There are 15 cysteine residues in the hamster lung pz-AR ( Fig. 1, Table I). Four of these cysteines (106,184,190, and 191) are in the putative extracellular domain; five cysteines (77,116, 125,285, and 327) are in the putative transmembrane domain; and six cysteines (265, 341, 371, 383, 395, and 411) are in the putative cytoplasmic domain, C Y S~~' undergoes palmitoylation, which is thought to be important for anchoring the COOHterminal tail to the lipid bilayer (3). The remaining 14 cysteine residues are not accessible to alkylation unless the protein is reduced under denaturing conditions, suggesting their involvement in a disulfide bond(s) (4). The compact structure thus produced is important because thiol reducing agents inactivate the pz-AR, presumably by cleaving critical disulfide bonds.
Thiol reduction of the p 2 -m can be blocked by agonists and antagonists demonstrating that the structurally critical disulfide bond(s) is masked by ligand occupancy ( 5 ) and, therefore, may be in proximity to the ligand-binding pocket. Furthermore, the reconstitution work of Pederson and Ross (6) demonstrated that the reduction of one or more disulfides functionally activates the P 2 -m and consequentially renders lability to the receptor structure. Site-directed mutational analyses showed that four cysteines critical for producing a p z -A R with high affinity ligand binding are localized to the putative extracellular domain (4,7,8). However, the specificity of their interaction has not been clearly established by systematic analysis of mutants and their thiol sensitivity or disulfide linkage. Such studies have been carried out, for example, with the m l muscarinic receptor and with bovine rhodopsin (9-12).
We have prepared 15 &-AR mutants by replacing the putative extracellular domain cysteine residues, in various combinations with alanine (Table I and  bordered approximately by the horizontal lines. The peptide segments connecting the helices (e.g. D and E) are designated as a loop region (D-E). The sequence, Glu-Thr-Ser-Gln-Val-Ala-Pro-Ala, at the carboxyl terminus, is the octapeptide epitope derived from the bovine opsin carboxyl terminus. This octapeptide sequence can be used for detection of the expressed protein using a mouse monoclonal antibody 1D4 (14). Of the 15 cysteines highlighted in the protein only Cys341 is palmitoylated. The cysteines numbered are the residues examined in this study by mutagenesis. The disulfide bond connectivity shown is based on the findings of the present study. consistent with a promiscuous interaction between CYS'~' and C y~'~~/ C y s '~~ in the production of the high affinity state of the p,-AR. The conserved Cyslg4 might be important for discrimination between C Y S '~~ and C Y S '~~. A disulfide interaction between Cyslg4 and Cyslg0 could be responsible for providing a unique partner to the conserved CYS'~', and, in addition, affording stability to the receptor. Thus, the role of C Y S '~~ in the folding and stabilization of the p,-AR is fundamentally different from the role suggested for the analogous residue in bovine rhodopsin or in m l muscarinic receptor (9-12).

MATERIALS AND METHODS
Cloning of the Hamster Lung &AR cDNA a n d Synthetic (17,18). Three days after transfection, membranes were prepared by Parr Bomb disruption (19) in Hanks' balanced salt solution containing protease inhibitors. The membrane pellet was resuspended in 50 m Hepes (pH 7.2), 12.5 mM MgCl,, and 1.5 m EGTA (HME buffer) containing 10% glycerol and stored at -70 "C. Protein concentration was determined by the Bradford method (20).

Zmmunoblotting of Receptor Protein Expressed in
COS-1 Cells-Membranes (50 pg of protein) were mixed with SDS-polyacrylamide gel electrophoresis (PAGE) buffer, allowed to remain at room temperature for 3 h, and then were analyzed by SDS-PAGE. The proteins in the gel were transferred to Immobilon-P membrane (Millipore) for 2.5 h a t 450 mA. The Immobilon-P membrane was blocked with 4% bovine serum albumin and then incubated with 2.5 pg/ml of 1D4 monoclonal antibody. After washing, the membrane was incubated with 1 pCi/ml Iz5I-antimouse Ig (Amersham Corp.). The membrane was washed, dried, and exposed to x-ray film (Kodak) a t -70 "C. The relative intensities of the bands were estimated by PhosphorImager analysis.
Photoaflinity Labeling OfEpressed Receptor Protein in COS-I Cell-Membranes (30 pg of protein) and 5 nM of 1251-CYP-diazirine (Amersham Corp.) were incubated for 3 h at 25 "C in the dark. Nonspecific labeling was determined in the presence of M propranolol (Sigma). Photolysis was performed by irradiation a t 366 nm for 20 min a t 0 "C (21,22). After photolysis, ice-cold 1 mM glutathione (Sigma) was added and the membranes were centrifuged for 10 min at 30,000 x g at 4 "C.
The membrane pellets were resuspended in 6 pl of 10% deoxycholic acid (Sigma j and added to SDS-PAGE sample buffer. Covalent incorporation of radioactivity was determined from the autoradiography. Deglycosylation of Wild-type p,-AR-A membrane (50 pg of protein) expressing the wild-type receptor was incubated in the presence or absence of 82.1 unitdml N-glycosidase F (Boehringer Mannheim) at room temperature overnight.
Equilibrium Binding Studies-Membranes expressing wild-type or mutant receptors were incubated with 0.01-10 n M lZ5I-CYP (Amersham Corp.) in HME buffer at 37 "C for 60 min. Nonspecific binding to the membranes was determined from '251-CYP binding in the presence of M propranolol. The binding reaction was stopped by filtering under vacuum (Brandel type M-24R) on FP-200 GFK filters (Whatman). Filter-bound I25I-CYP was quantitated in a y-counter (Packard). Equilibrium binding kinetics were determined using the computer program Ligand (23).
Competition Binding Sudies-COS-1 cell membranes expressing the wild-type p2-AR or the mutants, Ala-184, ECCA4, and ECCA5 (Table 11) were incubated at room temperature for 2 h with 60 PM 1251-CYP and various concentrations of the agonists:     Effect of DTT on '2sI-CYP Binding-To examine the dose-response effect of DTT (Aldrich) on lZ5I-CYP binding, wild-type or mutant receptors were preincubated with 0, 0.1, 0.5, 1, 5, or 10 m~ DTT in HME a t 37 "C for 30 min. After preincubation, each receptor (50 PM) was incubated at 37 "C for 60 min with lZ5I-CYP at a concentration at least 4-fold higher than its respective Kd value. To examine the time course effect of DTT, wild-type and mutant receptors were preincubated with 1 m~ DTT at 37 "C for different times (0-90 min). After preincubation, '251-CYF' was added to each receptor (50 PM) at a concentration 10-fold higher than its respective Kd value, and the final concentration of DTT was adjusted to that used for the preincubation. The samples were then incubated at 37 "C for 60 min. As a control, the same amount of receptor was preincubated for the same time, and then incubated with lZ5I-CYP for 60 min in the absence of DTT. To examine the effect of DTT on binding affinity, membranes ( 1 0 6 0 pg of protein) expressing wild-type or mutant receptors were preincubated with or without 1 m~ DTT at 37 "C for 30 min, and then with at least a 4-fold higher concentration of 1z51-CYP a t 37 "C for 60 min. Kd values were determined from the equilibrium binding data using Ligand (23).
Temperature Dependence of '2sI-CYP Binding-Membranes containing the wild-type or mutant receptors were incubated at 37 "C or at 42 "C for 0 3 h in HME buffer (pH 7.4). After incubation, each receptor (50 PM) was incubated with lZ5I-CYP, a t a concentration at least 10-fold higher than its respective Kd value, at 37 "C for 60 min. Incubations were stopped by filtering on a GF/C glass filter.
Zkansfection of Chinese Hamster Ovary Cells-CHO cells (American Type Culture Collection) were co-transfected with either the wild-type &-AR or independently with five different mutant genes, Ala-106, Ala-184, ECCA4, ECCA5, or ECCA6, together with selectable vector pSV-His (25) by the Polybrene method (26). After about 2 weeks of selection, histidinol-resistant colonies (to 2.4 m~) were expanded. Membranes were prepared by Parr Bomb disruption and used for equilibrium binding studies and adenylate cyclase assays.
Adenylate Cyclase Assay-Adenylate cyclase activity was measured using a modification of Salomon's method (27). 5-50 pg of membrane protein were used in the assay. The adenylate cyclase activity of the membranes expressing each receptor was determined both basally and after stimulation with M (-)-isoproterenol. To standardize the adenylate cyclase activity among the different membrane preparations, adenylate cyclase was stimulated maximally with 10 p~ forskolin.

Expression of Wild-type P2-AR Gene in COS-1 Cells
Expression of the native &-AR cDNA, the synthetic &-AR gene, or its mutants in COS-1 cells was examined by immunoblotting, by photoaffinity labeling, and by antagonist-binding studies ( Table 11, Figs. 2 and 3). Immunoblotting revealed heterogeneous expression with predominant molecular weight forms of -40-55 kDa ( Fig. 2 A ) . A -65-70-kDa band corresponding to about 20-25% of the expressed wild-type &-AR was also observed. The heterogeneity of the expressed p,-ARspecies resulted from differences in Asn-linked glycosylation, since treatment of the membranes with N-glycosidase F produced a single band of approximately 45 kDa, which corresponds to the calculated molecular mass of the unglycosylated pz-AR polypeptide (Fig. 2B). A 24-28-kDa band was also detected by immunoblot analysis with the 1D4 antibody. The mobility of this band was not altered by N-glycosidase F treatment. Thus, this species probably represents a proteolytic COOH-terminal fragment of the pz-AR.
Of all the bands, only the 65-70-kDa band could be photolabeled with the antagonist, 1251-CYF"diazirine ( Fig. 2C). Photolabeling of this band was specific because it could be blocked by M (-)-propranolol (Fig. 2C). A 65-70-kDa photolabeled species was also observed in membranes prepared from COS cells transfected with a native &-AR cDNA that lacks the 1D4 epitope tag. The abundance of this species was similar to that observed with the wild-type synthetic gene. In both membrane preparations, the efficiency of photolabeling was proportional to the B,,, values estimated from antagonist equilibrium binding studies (discussed later). Furthermore, it correlated with the intensity of the 65-70-kDa band as determined by immunoblotting. Since the 40-55-kDa receptor forms, as determined by immunoblotting, were expressed at higher levels than the 67-70-kDa forms, these species should have been detectable by photoaffinity labeling as well. This suggests that nascent, folding intermediates are present in the cell at a steady state due to a high level of expression. Moxham  (-)-norepinephrine. Thus, the observed heterogeneity of pz-ARs is presumably due to the production of the mature protein, as well as nascent species, which do not bind ligand, are partially glycosylated, and have not yet folded completely, rather than due to the end products of proteolysis within the cells.
We, therefore, conclude that high level expression of &-AR is achieved in COS-1 cells by pMT3 vector, and that the level and The pattern of protein expression of all of the mutants was similar to that of the wild type ( Fig. 2A). The 40-55-kDa bands were expressed at the same level for all of the mutants, suggesting that there are no defects in the transcription/ translation of the mutant genes. A 65-70-kDa form of the receptor was observed with all of the mutants, although the amount of these species was variable. For example, with the mutants Ala-106, ECCA7, and ECCAll (Table I), the 65-70-kDa band was less abundant than that observed for the wildtype receptor. In the mutant ECCA6, the expression of the 65-70-kDa band was almost the same as that of the wild type. With all of the mutants, only the 65-70-kDa receptor could be detected by 12"I-CYF"diazirine photocross-linking (Fig. 2C). However, the extent of labeling differed significantly among the mutants. For example, the photocross-linked band intensities with the mutants Ala-184, ECCA4, and ECCA5 were similar to that of the wild type. It was considerably reduced with the mutant ECCA6, although by immunoblotting the protein expressed by this mutant gene was the same as that of wild type. As indicated below, antagonist binding with some of the mutants was not directly related to the level of the 65-70-kDa receptor. Therefore, it appears that the mutant receptors are appropriately glycosylated in COS-1 cells, but unlike the wildtype receptor, are not homogeneously folded.
Antagonist Binding to Mutants-The mutants could be classified into four groups based on their affinity for the antagonist radioligand 12sI-CYP (Table 11) (Table 11). Additionally, these in their affinity for 1251-CYP (Kd >700 PM, see Table 11) and a mutants displayed a single class of high affinity antagonist reduced ability to form a 65-70-kDa receptor (Fig. 2 A ) . The binding sites (Fig. 3). The agonist-affinity profile of these mu-presence of complex, multiple affinity binding sites for the antants was also similar to that of the native &-AR (Table 11).
Thus, these findings indicate that the presence of a cysteine at 184 is not essential for producing the wild-type-like group I phenotype. However, the B,,, of the Ala-184 mutant is reduced compared to that of the wild-type Pz-AR. Presumably, this suggests a significant role for Cysis4, either in efficient folding or in stabilization of the folded state of the Pz-AR (see "Discussion"). Mutants that retain Cys"' and Cysiso or Cysisi, in various combinations, belong to this group, suggesting an important role for these residues in the production of wild-typelike receptors.
Group I1 mutants showed a 5-7-fold decrease in their affinity for i251-CYP (Kd = 200-300 PM, see Table 11). These included Ala-190, Ala-191, and ECCA3 (Table 11). The 65-70-kDa receptor form for these mutants is produced at the same abundance as for the group I mutants (Fig. 2 A ) . All the mutants that belong to this class retain CysiS4, suggesting a role for this residue in the production of the group I1 phenotype. Among the seven mutants that retain C Y S '~~, i t appeared that the presence of CyslSo (as in the mutant ECCA3), or Cysin6, in combination with either C~S '~~ or CYS"~ (as in mutants Ala-191 and Ala-190, respectively) produce this phenotype. Complex multiple antagonist-affinity states of the receptors (Kd = 140 PM and & = 350-700 PM) were observed for the Ala-190 and Ala-191 mutants (Fig. 3). Thus, the presence of a free cysteine at position 184 (e.g. in mutants Ala190 and Ala191) might interfere with uniform folding and, thus, distinguish these mutants from group I mutants. However, the mutant ECCA3, which has C Y S '~~ and Cysig0, produced a single affinity class of receptors. This mutant presented some interesting additional phenotypic characteristics which are described below.
Group I11 mutants which lack Cyslo6, and/or all of the cys-tagonist is apparent in this group of mutants (Fig. 3).
The group IV phenotype is represented by only one mutant, ECCA6, which retains the conserved Cyslo6 and Cysis4 residues but lacks the vicinal Cysig0 and Cysigi (ECCA6). Expression of the 65-70-kDa mutant receptor is enhanced (as also indicated by its B,,,) but the affinity of the receptor for lZ5I-CYP is greatly decreased (Kd > 1500) (Table 11, Fig. 2 A ) . Also, antagonist-binding studies demonstrated the presence of multiple affinity forms in this mutant.

Effect of DTT on Antagonist Affinity
Wild-type-Incubation of the wild-type P z -A R with 1-10 m M DTT leads to the loss of high affinity antagonist binding (Fig.  4). This is thought to be due to reduction of disulfide bonds. Preincubation of the wild-type receptor with 1 m M DTT led to a decrease in the affinity of the receptor for 1251-CYP (Table 111, Fig. 4). However, incubation with 5-10 m M DTT results in a substantial decrease in affinity combined with loss of binding sites. The DTT inactivation follows a biphasic time course (Fig.  4). The initial decrease in binding represents the fast reduction of solvent accessible disulfide bond(s), by DTT. In contrast, the slower decrease in binding presumably represents the ensemble reduction of other accessible disulfide bond(s). We assumed that the initial phase involved the reduction of the extracellular disulfide bonds. Therefore, we chose 1 m M DTT to investigate the disulfide interaction of extracellular cysteines in mutants. Inactivation was complete in 30 min at 37 "C (Fig.  4). The receptor could be protected from inactivation by preincubation with 1251-CYP. Inactivation of Pz-adrenoreceptors by treatment with D'IT or P-mercaptoethanol also leads to loss of antagonist binding with similar kinetics (28).
Cys ~ + Ala Mutants-The mutants fall into two different     Tables I1 and  111). These results suggest that reduction of putative disulfide bond(s) in other domains was not taking place under our experimental conditions. Furthermore, they indicate that these mutants do not contain a DTT-accessible disulfide bond(s). The mutants lacking all four extracellular cysteines (ECCAll), mutants lacking loop D-E cysteines (ECCA'I), and the mutant Ala-106 fall into this category. Some of the mutants retaining a single pair of extracellular cysteines also showed resistance to 1 mM DTT. For example, DTT treatment did not influence antagonist binding of mutants retaining Cyslg0 and Cyslgl (ECCAl), CyslS4 and Cyslgl (ECCM), or Cys106 and CyslS4 (ECCAG) ( Table 111). These mutant receptors, like mutants ECCA7 and ECCA11, however, were inactivated at elevated temperatures in the absence of DTT (data not shown). Therefore, it seems unlikely that cysteine pairs, CYS'~O and CYS'~', Cyslo6 and CyslS4, or Cyslg4 and Cyslgl are linked by a disul- fide bond. A small fraction of moderate to high affinity receptors observed could arise in all of these mutants, presumably, by the packing of helices and loops by non-covalent interactions. Therefore, the high affinity state is not abundant in mutants and is unstable at elevated temperatures (see below). Five mutants, Ala-184, Ala-191, ECCA3, ECCA4, and ECCA5, were similar to the wild-type receptor in their sensitivity to DTT (Table 111). In addition, all of these mutants, with the exception of Ala-191, produced a single antagonist-affinity class of receptors. The mutants Ala-184, ECCA4, and ECCA5 also showed Kd values similar to that of wild-type (Tables I1   and 111), suggesting a relationship between the high affinity state of the receptor and sensitivity to DTT. When these mutant receptors were incubated with 1 m~ D'M', a rapid loss of high affinity for antagonist binding was observed (Fig. 5 A ) . The inactivation process caused small changes in the number of binding sites, indicating that irreversible denaturation of receptor was not taking place. The mutants Ala-184, ECCA4, and ECCA5 did not show a rapid inactivation by D'M' at 0-5 min. The wild-type and ECCA3 exhibited this property (Fig. 5 A ) . Therefore, this phase of inactivation of the wild-type might involve the reduction of the putative C y~'~-C y s '~~ bond. In support of this observation, the mutant ECCA3 was fully reduced by treatment with D'M' at 0 min. This mutant was not inactivated any further by Dm. The slope of inactivation of all other mutants differed from that observed with the wild-type receptor, probably reflecting a change in the environment of the reduced disulfide bond in each mutant. The mutants Ala-184, ECCA4, and ECCA5 are likely to contain a disulfide bond involving Cys'06, which plays a critical role in the formation of the high affinity antagonist binding state in the wild-type receptor. All of these mutants uniquely lack Cysls4, which does not appear to form a disulfide bond with Cyslo6. Furthermore, it is likely that the mutants ECCA4 and ECCA5 contain a disulfide bond involving CyslOG and Cyslg0, and Cys106 and Cys"', respectively, and, therefore, these mutants produced high affinity receptors.
The mutant ECCA3 is sensitive to reduction by D'M' and, therefore, presumably contains a disulfide bond between CyslW and Cyslg0. This mutant produced a single-affinity class of receptors with moderate affinity for lZ5I CY". Thus the disulfide bond most likely represents a genuine interaction required for efficient production of high affinity receptors, Therefore, a second DTT-sensitive disulfide bond might also exist between C y P 4 and Cyslg0 in the wild-type receptor.

Temperature Dependence of Antagonist Affinity
D'M' inactivation of the mutants did not distinguish the stability effect of two putative disulfide bonds, presumably due to similar role they play in receptor structure-function. Thermal stability of the wild-type or mutant receptors was compared to distinguish the role of the two putative disulfide bonds (Fig.  5B ).
The wild-type receptor was stable for 3 h at 37 "C, whereas mutants Ala-184, ECCA3, ECCA4, and ECCA5 showed about a 3 0 4 0 % loss in total binding. Thus, differences between the stabilities of different mutants could not be readily discerned at 37 "C. At 42 "C, however, thermal-stability differences between the mutants were apparent. The thermal stability of the wildtype receptor at 42 "C was similar to that observed at 37 "C. The mutants Ala-184 and ECCA4, which lack the putative C y~'~~-C y s '~~ disulfide bond, showed similar kinetics of inactivation (Fig 5B). The mutant ECCA3 that retains the putative C y~'~~-C y s '~~ disulfide bond and ECCA5 that retains the putative C y~l~~-C y s~~l disulfide bond were relatively rapidly inactivated. Therefore, it appears that the contribution of the C y~~~~-C y s l~l bond to receptor stability is greater than that of the or bond. However, with the wild-type receptor, a concerted effect of both of these bonds on the receptor stability is evident.

Modeling the Structure of Extracellular Loop D-E
The presence of a disulfide bond between 2 of the 3 cysteine residues in the D-E loop raises the question of which pair is sterically favored. The choice of partners in the formation of the two putative disulfide bonds will depend on the tertiary structure of the D-E loop. To address this issue, we carried out modeling studies to deduce the potential, interacting cysteine pairs. A pool of 27,000 disulfide-bonded protein sequences in the Swiss protein data base was searched using the computer program GeneworksTM by providing a sequence motif C-X(>4,<6)-CC. The pattern match picked 84 sequences. Structural information was available in the protein data base for six of the proteins in this pool. There is no sequence homology in the segments separating the cysteines in any of these six proteins. However, there is a remarkable similarity in the backbone structures of these segments. Thus, the structure might represent a motif that is frequently used in several proteins unrelated to the pz-AR. These segments are all present in the solvent exposed regions of their respective proteins. We reasoned that the D-E loop of the &-AR might assume a similar structure, and, therefore, performed comparative modeling studies with this segment of the &AR. The packing of residues depicts a p-strand conformation. The side chains of cysteines 184 and 190 face inward. We estimated the range of distances between their sulfur atoms through rotation of the C, bonds.
The minimal distance estimated was 1.87 A, and the maximal distance was 8.32 A, implying a potential C y~~~-C y s '~~ disultide interaction bond based on their proximity (29). Similar estimates for the interaction between Cysls4 and C Y S '~~ gave a range of distances of 6.66-11.85 A and for the inpraction between Cys1' ' and Cyslgl a distance of 5.81-9.25 A. This analysis strongly suggests that the interaction is favored in this model. The structure obtained for the D-E loop suggests that within this region a disulfide bond interaction between Cysls4 and CYS'~O is more likely than between Cysls4 and Cyslgl (Fig. 6). Additionally, a n interaction between Cyslg0 and Cysigl (8) seems unlikely.

Antagonist Binding and Agonist-induced Function in CHO Cells
Clonal CHO cell lines expressing some of the mutants (Table  IV) were established, as described earlier, for the purpose of examining antagonist affinity profiles and receptor-coupled adenylate cyclase activation. The expression level of the wild-type P,-AR was 0.74 pmoVmg and of the mutants between 0.11 t o 2.27 pmoVmg. In CHO membranes, the wild-type receptor and the mutants, Ala-184, ECCA4, or ECCA5 exhibited a single affinity state for antagonists. Two mutants, Ala-106 and ECCA6, exhibited a mixture of high and low affinity receptors as in COS-1 membranes. The heterogeneity of receptor-affinity states, therefore, seems to be an intrinsic property of the mutants rather than arising from the high level expression observed in COS-1 cells, and binding of 1251-CYP by the wild type and by each mutant receptor was similar to that observed for these receptors when expressed in COS-1 cells. The mutants Ala-184, ECCA4, and ECCA5, which showed a similar affinity for antagonists, were functionally active ( Table IV). The mutant Ala-106 also retained the ability to stimulate adenylate cyclase, but the mutant ECCA6 did not show agonist-induced stimulation of adenylate cyclase. The level of expression of each of the proteins could not be estimated independently by immunoblotting, and, therefore, the fold stimulation of cyclase by each of the mutants could not be used to interpret defects in coupling ability. These studies are currently underway.

DISCUSSION
Alteration in the properties of hormone receptors by thiolreducing agents has been known for some time (30). P Z -A R alteration by DTT was first suggested by the observation that DTT modifies beating of guinea pig atria (31). In all native systems examined, DTT alters the affinity of the receptors for their ligands without affecting the total number of receptors, due to the reduction of one or more disulfide bonds present in P-receptors (32). Protection of the receptor from thiol reduction by P-adrenergic agonists and antagonists implied that the bonds reduced by DTT are shielded in the ligand-occupied receptor (5). The reconstitution work of Pederson and Ross (6) and the in situ experiments of Moxham et al. (28,32) clearly demonstrated the presence of disulfide bridges, reduction of which causes functional activation of the receptor. influence the ligand-binding properties of the receptor. This conclusion seems to be borne out in other studies as well (4,8). The four cysteines in the extracellular domain have been examined in several studies (4,7,8,33). Substitution of CysloG or Cysls4 of the hamster lung P,-AR by valine resulted in altered agonist-binding properties, with no apparent alteration of lZ6I-CYF' binding. The possibility that these two residues might be involved in a disulfide bond was suggested by the finding that substitution of either cysteine caused a loss of affinity for ligands (7). By analogy to disulfide interaction of the conserved cysteines in other members of this receptor superfamily, the conclusion that these cysteines, CysloG and Cysla4, are similarly involved in P-AR integrity seemed obvious (7). However, direct evidence for an interaction between CysloG and CyslS4 in a disulfide bond is lacking so far. With the human P,-AR, three mutants involving C Y S '~~ and Cysigl demonstrated a loss of affinity for both antagonists and agonists. Involvement of Cyslg0 and Cyslgl in ligand binding was proposed (8). Dohlman and colleagues (4) created eight different cysteine-substitution mutants. The properties of receptors with mutations of the cysteines in the transmembrane domain were unaffected. The replacement of cysteines in the putative extracellular domain resulted in profound changes in the normal ligand binding affinities. They concluded that extracellular disulfide-bonded cysteines are critically important for forming or stabilizing the ligand binding site and are likely to be sensitive to DTT treatment (4).
Although the critical nature of the cysteines in extracellular domain is clearly established in the studies mentioned above, several questions remained. If residues CysioG and CYS'*~ are connected by a disulfide bond, then their contribution to receptor structure and function is expected to be similar. This expectation seems clearly satisfied in the mutagenesis studies of bovine opsin and the m l muscarinic receptor. The cysteine residues thus identified in these receptors were shown to be disulfide-bonded (9-12). However, the effects of CysloG mutation in the P,-AR are very different from those of Cysls4 mutation. In two independent studies, CYS"~ substitution showed near wild-type affinity toward the antagonist, 1251-CYP, but showed a greater decrease in affinity toward agonists.
The CysloG mutation had a more severe effect on both properties (4,7). Antagonist binding is dependent on the folding of the receptor, while agonist binding affinity is also modulated by cognate G-protein coupling (or precoupling). It seems possible that these properties are achieved by independent disulfide bond formation in the wild-type receptor. The differential effects of CyslB4 or CyslOG mutations on these properties then might suggest that these cysteines are not linked by a disulfide bond.
Rather, DTT inactivation of the P,-AR might actually reflect reduction of two different disulfide bonds. The work presented in this paper attempts to address these questions. Our studies began with the observation that a single mutant Ala-184 displays high affinity antagonist binding (Tables I and  11). This property differed from that of the Ala-106 mutant, the partner proposed to be involved in a critical disulfide bond. The Ala-106 mutant produced low affinity receptors, and was expressed at reduced levels. A quadruple mutant, ECCA11, and a triple mutant, ECCA7, also expressed low affinity receptors like the Ala-106 mutant, suggesting that removal of either Cys106, or removal of all three cysteines in the loop D-E, results in the low affinity receptor phenotype. Therefore, it is likely that C Y S '~~ is not essential for the formation of a high affinity p2-AR. In addition, the two double mutants (ECCA4 and ECCAS), both lacking CyslS4, and either CYS'~O or C Y S '~~, produce high affinity receptors. A double mutant lacking both CYS'~O and Cysigi produces low affinity receptors, as observed in earlier studies (8). Furthermore, inactivation of high affinity antagonist binding by DTT suggests that these residues are involved in disulfide bond formation. It seems possible, therefore, that the two cysteines, CYS'~' or Cysigi, could potentially interact with Cysio6. A promiscuous interaction between Cys'06, and either Cysigo or C Y S '~* , appeared to produce the high affinity state of the p2-AR in mutants lacking CysiS4.
Which one of these is the partner for Cysio6 in the native receptor?
The key to answer this question comes from the DTT sensitivity and thermal stability experiments on the mutants (Table  111, Figs. 4 and 5). From the data presented in Table 111, some of the potential disulfide bond interactions can be ruled out, based on their temperature sensitivity and insensitivity to DTT. For example, a disulfide interaction between Cysig0 and Cysigi in the mutant'ECCAl, between CysiS4 and Cysigl in the mutant ECCA2, or between Cysio6 and CysiS4 in the mutant ECCA6 is not evident. However, these mutants are sensitive to temperature inactivation, like the mutants lacking all extracellular cysteines. Therefore, resistance to DTT suggests that the remaining cysteines bear a free sulfhydryl, rather than being involved in solvent-inaccessible disulfide bonds. It is possible that the secondary structure of the polypeptide and side chain packing precludes an interaction between the pairs of cysteines mentioned above. Complementing this observation, the mutant ECCA3, retaining CyslS4 and C~S'~O, and the mutant ECCA5, retaining Cyslo6 and C Y S '~~, were found to produce a single-affinity class of receptors, and both mutants were sensitive to DTT reduction. Extending this to wild-type receptor might indicate that two disulfide bonds linking Cysio6 and C Y S '~~, and Cys184 and CYS'~O exist in the native receptor. Only in one mutant (ECCA5) a high affinity antagonist binding state seems to be generated by a non-native disulfide interaction between Cys'06 and Cysigo. The absence of CyslS4 and Cysigi might position C~S~'~ at an interacting distance with Cysio6. Based on our modeling studies, it appears that a Cys1S4-Cysigo disulfide bond may be induced by the proximity of their sulf-hydryl groups. Since localized folding in the loop region could drive this interaction, native folding of the p2-AR might involve a multi-step process (Fig. 7).
A interaction leads to a linking of loops 1 and 2 through disulfide bond formation between Cysio6 and Cysigi. The mutants suggest that the high affinity antagonist binding property correlates with the formation of a C y~'~~-C y s~~' disulfide bond. The role of a second disulfide bond is unclear at this stage, although involvement in agonist affinity seems most likely based on previous observations on Cys184 mutants (4,7). Our studies suggest that a disulfide linkage does exist between C Y S '~~ and Cysig0. Therefore, it is not essential for the high affinity antagonist binding but might be important to achieve high B,,, values.
Two questions still remain. First, the basis for the conservation of C Y S '~~ is not clear at this stage. As discussed earlier, a cysteine residue analogous to Cysia4 is conserved in the D-E loop of the family of the G-protein-coupled receptors, which is believed to be disulfide-bonded to a similarly conserved cysteine, analogous to Cysio6, at the boundary of loop B-C and helix C (34). This linkage has been confirmed for bovine opsin and the m l muscarinic receptor (9-12). The critical role played by this pair of cysteines in receptor folding was shown by mutagenesis for a number of other receptors in addition (35). Based on their studies Khorana (36) proposed a scheme for the putative steps involved in the folding of bovine opsin; the assembly of the receptor begins by transfer of the hydrophobic segments into the endoplasmic reticulum membrane where high mannose glycosylation then takes place. The insertion of helices into the bilayer does not ensure their alignment. Rather, the correct alignment of helices is achieved by a cooperative interaction of the extracellular loops stabilized by disulfide bond formation. A unique disulfide bond between the conserved cysteines is obligatorily required for bovine rhodopsin (11, 12). The number of disulfide bonds may differ in other receptors (35,37). The folding of the p 2 -m may follow the same general scheme (see Fig. 7). For example, Kobilka (38) observed a lag time in the acquisition of specific ligand-binding properties of the p2-AR after its incorporation into the microsomal membrane. This lag might be due to a disulfide exchange reaction involving the extracellular cysteines. We speculate that the kinetically preferred route of folding may involve the initial formation of a C y~'~~-C y s '~~ disulfide bond (Fig. 7). The basic seven transmembrane a-helical motif generated at this step does not contain the ligand-binding pocket. In subsequent steps, a disulfide-exchange reaction resulting in the formation of disulfide bonds between Cys'06 and Cysigi, and CysiS4 and Cysigo, might then lead to the final structure. Heterogeneity in glycosylation might reflect such folding steps. Intracellular transport and posttranslational modification are independent processes thought to be coupled to the folding of proteins into complex structural motifs. In the folding of soluble proteins, it is known that disulfide bond formation is directed by other, Second, although epinephrine is also a natural ligand for al-adrenergic receptors, they do not share the structural features that characterize the @, -AR extracellular domain. For example, a vicinal cysteine pair in loop D-E is found only in @-adrenergic receptors. The concentration of DTT that inactivates the P2-AR does not inactivate the al-adrenergic receptor, although this latter receptor clearly contains a solvent-inaccessible disulfide bond (42). Furthermore, functional activation by thiol compounds is a unique property of the p-AR. @-Adrenergic agonists conserve redox properties, while @ -A R antagonists do not share this property. Based on this observation, involvement of a redox mechanism in the activation of the receptor has been suggested (43). Direct agonist-mediated disulfide bond reduction may, therefore, be involved in receptor activation, but this has not been shown. Therefore, it is possible that disulfide exchange is a property uniquely acquired in the divergence of the folding of @-adrenergic receptors from other adrenoreceptors and G-protein-coupled receptors.