Sustained activation of a G protein-coupled receptor via "anchored" agonist binding. Molecular localization of the salmeterol exosite within the 2-adrenergic receptor.

An inherent therapeutic limitation of many G protein-coupled receptor agonists is a short duration of action due to rapid dissociation from receptors. Salmeterol is a modified beta-adrenergic receptor (betaAR) agonist that has a long duration of action at the beta2AR (but not the beta1AR) both in vitro and in vivo and that is persistent despite extensive washout of the agonist. It has been proposed that salmeterol binds not only to the active site of the beta2AR (localized to receptor transmembrane spanning domains (TMDs) 3 and 5) but also to another site (termed the "exosite") that anchors it to the receptor and provides for repetitive active-site binding events. To identify the location of this exosite, we used site-directed mutagenesis to replace beta2AR amino acids 149-173 (within TMD4) with beta1AR sequence. The resulting constructs were then expressed in COS-7 cells for radioligand binding studies. Using this approach, when this domain was replaced with the analogous beta1AR sequence, the ability of salmeterol to persist at the receptor under washout conditions was reduced by 67%. The results from more selective mutants (S-(149-166), S-(164-173), and S-(149-158)) indicated that a limited 10-amino acid region (beta2AR residues 149-158), localized at the interface of the cytoplasm and the transmembrane domain, contains a critical determinant for exosite binding. Whereas CHW cells stably expressing wild-type beta2AR displayed persistent salmeterol-promoted cAMP accumulation despite agonist washout, substitution of beta2AR residues 149-158 with beta1AR sequence resulted in a 56% attenuation of salmeterol-promoted cAMP accumulation under identical washout conditions. A reverse chimera was also studied, which consisted of a substitution of beta2AR residues 152-156 into the beta1AR. This substitution was found to confer exosite binding to the beta1AR. None of these mutations decreased the affinity of salmeterol for the receptor at the active site as assessed in competition binding studies. Anchored binding to this motif thus represents a novel mechanism by which agonists like salmeterol can repetitively activate receptors. Conceivably, with other G protein-coupled receptors that have similar motifs, anchored ligands can be designed to provide for long durations of action by this mechanism.

The active sites of the ␤ 2 -adrenergic receptor (␤ 2 AR) 1 for binding catecholamine-like agonists lie within the transmembrane spanning domains (TMDs) of the receptor and include key interactions at Ser-204 and Ser-207 of TMD5 (which form hydrogen bonds with hydroxyl groups on the catecholamine ring) and Asp-113 of TMD3 (which interacts with the amine head group) (1)(2)(3). Salmeterol is an analogue of the saligenin ethanolamine salbutamol (albuterol), with an aralkyloxyalkyl substitution at the amine group (Fig. 1a). In addition to increased lipophilicity, salmeterol exhibits binding and functional characteristics that suggest it binds not only to the active site but also to another site providing for a long duration of action (4 -6). In tissue organ bath and cell culture systems, a single exposure to salmeterol results in persistent activation of ␤ 2 AR despite removal of the drug and extensive washing. Furthermore, although this persistent activation may be readily reversed by ␤AR antagonists, subsequent washout of the antagonist results in reassertion of the agonist activity (4,7,8). Kinetic analyses of the interaction between salmeterol and artificial lipid membranes suggest that this prolongation of action is not due solely to its increased lipophilicity; rather, the data implicate an additional membrane-bound factor(s) (such as the receptor) to explain the persistence of the ligand-receptor interaction (9). Furthermore, such a prolonged duration of action is not observed in tissue and cell preparations expressing the ␤ 1 AR, even though salmeterol does bind to this receptor (albeit with a lower affinity than to the ␤ 2 AR) (8). In addition, a recent study examining the functional activation of ␤ 2 AR expressed in L cells has also suggested that salmeterol is able to access the receptor despite concurrent blockade, apparently by forming a high affinity interaction with the receptor at a second, undefined locus (6).
Two mechanisms of action of salmeterol have been considered to account for these observations, 1) the drug, which is highly lipophilic, interacts with the lipid bilayers of the cell membrane in the vicinity of the receptor thereby providing a local depot, or 2) it specifically binds to the ␤ 2 receptor itself at an anchoring region that allows for persistent and/or repetitive stimulation via the active site. This latter interaction, commonly referred to as "exosite binding," would be a novel mechanism for G protein-coupled receptor agonism, and the molecular domains responsible for this potential interaction were investigated. To distinguish between the two aforementioned mechanisms, we developed a recombinant cell model by which the interaction of salmeterol with the ␤ 2 AR could be measured. We then used this model system, together with site-directed mutagenesis of the receptor, to explore the molecular basis of this interaction. We demonstrate herein that salmeterol indeed binds to the ␤ 2 AR in a receptor-specific, "exosite"-type manner; furthermore, the molecular determinants of this novel binding mechanism are localized to a small domain within the transmembrane-spanning domain of the receptor, which is distinct from those regions classically associated with the receptor active site. Thus, the "anchored" binding of salmeterol to the ␤ 2 AR represents a novel ligand interaction of an agonist with a G protein-coupled receptor.

EXPERIMENTAL PROCEDURES
Constructs-Plasmids encoding the cDNAs for the wild-type ␤ 1 AR and ␤ 2 AR were used as described previously (10). Mutations of the wild-type ␤ 2 AR in which various sequences were replaced by the analogous region of the ␤ 1 AR were constructed by site-directed mutagenesis similar to that previously reported (11); briefly, the KpnI-PstI fragment of the ␤ 2 AR was subcloned into M13 phage, and oligonucleotide-directed mutagenesis was performed by the method of Kunkel (12), after which the mutated fragment was reintroduced into the wild-type ␤ 2 ARexpressing plasmid ␤ 2 pBC12BI. Mutations of the ␤ 1 AR were similarly constructed, except that the entire coding region was subcloned into M13 phage; following mutagenesis, the BstXI-AscI fragment was subcloned into the plasmid containing the otherwise wild-type sequence. All mutations were confirmed by dideoxy sequencing. The predicted protein sequences of the ␤ 2 AR wild-type and mutated receptors are depicted in Fig. 1. In each case, numbering of the mutation refers to the regions of the wild-type receptor that were replaced by analogous sequence. Specifically, the largest mutation (mutant ␤ 2 S-(149 -173)) consisted of replacement of the entire ␤ 2 AR TMD4 (residues 149 -173) with the analogous ␤ 1 AR sequence. Similarly, smaller mutations included substitution of the ␤ 2 AR TMD4 amino acids in closest proximity to the cytoplasmic face of the receptor (mutant ␤ 2 S-(149 -158)), residues of the lower two-thirds of the ␤ 2 TMD4 (mutant ␤ 2 S-(149 -166)), and residues closest to the extracellular face of the receptor (mutant ␤ 2 S-(164 -173)), with the analogous ␤ 1 AR sequences.
Transfections and Cell Culture-Plasmid DNA containing each receptor construct were transiently expressed in subconfluent COS-7 cells by either electroporation or by the DEAE/dextran technique (13). 1-10 g of plasmid DNA was used for either transfection procedure and was adjusted to yield equivalent receptor expression levels, as determined in radioligand binding studies described below. Cells were maintained in DMEM media supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin in an atmosphere of 5% CO 2 at 37°C. Cells were assayed when confluent, typically 48 h following transfection. For functional studies, plasmids were stably transfected in CHW-1102 cells by calcium phosphate precipitation as described previously (10); 30 g of plasmid DNA were cotransfected with 3 g of pSV2neo, and recombinant cells were selected in 300 g/ml G418. Cell lines resistant to G418 were maintained in DMEM with serum, antibiotics, and 80 g/ml G418, and ␤AR expression was determined by radioligand binding as described below. Clonal lines expressing equivalent amounts of receptor (ϳ600 fmol/mg protein) were selected for study.
Radioligand Binding Assays-Radioligand binding studies were performed in COS-7 membranes in a manner similar to that described previously (10,11). Briefly, confluent COS-7 cultures were scraped in 5 mM Tris (pH 7.4), 2 mM EDTA at 4°C, and crude membranes collected by centrifugation at 40,000 ϫ g for 15 min at 4°C. The resulting pellets were homogenized in a Brinkman Polytron, recentrifuged, and suspended (final concentrations) in 75 mM Tris (pH 7.4), 12.5 mM MgCl 2 , 2 mM EDTA, 100 M GTP (to inhibit high affinity ternary complex formation (14)) at a protein concentration of approximately 1 mg/ml. For saturation studies, membranes were incubated along with 5-400 pM 125 I-cyanopindolol (ICYP) in a final volume of 250 l for 2 h at 22°C. Reactions were stopped by dilution and rapid filtration over glass fiber filters. The wet filters were then counted in a ␥ counter at 75% efficiency. Nonspecific binding was defined as ICYP counts that occurred in the presence of 1 M propranolol. Competition studies were performed similarly, except that a fixed concentration of ICYP (30 pM) was used, and membranes were incubated with 0 -1 mM unlabeled competing ligand. In all assays, bound ICYP counts were typically Ͻ10% of total counts added. Protein for this and all other assays was determined by a copper bicinchoninic acid method (15).
Assessment of Prolonged Agonist Binding to the ␤ 2 AR-35-mm tissue culture dishes containing COS-7 cells expressing each of the receptors were preincubated at 37°C for the indicated times in serum-free DMEM containing the indicated amount of agonist or vehicle. The media were then removed, and the attached cells were washed continuously at room temperature with calcium-free phosphate-buffered saline (PBS) at a constant flow rate of 20 ml/min for the indicated times using a multichannel peristaltic pump. (In preliminary studies, analysis of the perfusate after such washing revealed no detectable agonist by high performance liquid chromatography.) Cells were then lysed, membranes prepared, and the apparent receptor density (i.e. the number of receptors not occupied by drug) was determined by radioligand binding as described above. Because of day-to-day variability in the transient receptor expression levels (range, 400-6000 fmol/mg protein), apparent receptor density was normalized as the percentage of ICYP binding sites observed in similarly treated membranes preincubated with vehicle alone. For wild-type ␤ 2 AR, 38.2 Ϯ 2.3% of the receptor remained occupied by salmeterol after this washout procedure, indicative of persistent binding to the wild-type receptor. Subsequent data are presented as the loss of this exosite binding activity for the various mutant receptors.
cAMP Accumulation Studies-Functional studies were performed using confluent 35-mm tissue culture dishes containing stably-transfected CHW-1102 cells expressing either wild-type or mutated ␤ 2 AR at equivalent expression levels as described above. Media from each dish were aspirated, cells washed twice with PBS, and 1 ml of serum-free DMEM containing either the indicated concentration of agonist plus vehicle or vehicle alone was added. Dishes were incubated for 10 min at 37°C, after which the media were removed and saved for measurement of cAMP. (In preliminary studies, salmeterol-promoted accumulation of cAMP in CHW cells expressing either wild-type or mutated ␤ 2 AR was linear and equivalent up to at least 30 min.) The cells were washed by perfusion with PBS for 2.5 min at room temperature in the same manner as described for the binding studies above. At the end of this perfusion, the PBS was aspirated, 1 ml of serum-free DMEM (without agonist) was added, the cells were returned to the 37°C incubator for 10 min, and the supernatant was then removed for cAMP determination. This process of washing, incubation, and cAMP determination was repeated until the total aggregate perfusion time reached 15 min. Total cAMP content in the saved DMEM fractions was determined by an acetylated radioimmunoassay method as described previously (11,16). The cAMP content of each fraction, determined as pmol/ml media, was expressed as the percentage of cAMP measured in the initial fraction obtained prior to any washing.
Data Analyses-Comparisons between means were by paired or unpaired t tests, as appropriate. p values Ͻ 0.05 were considered significant. Data shown are mean values Ϯ S.E. of n tests, except where indicated. Radioligand binding data were analyzed by nonlinear regression of raw counts, using software from GraphPad (San Diego, CA). K i values were determined from EC 50 values by the method of Cheng and Prusoff (17).
Materials-DMEM and fetal calf serum were from JRH Biosciences. G418 was from Life Technologies, Inc. ICYP was from DuPont NEN. 125 I-cAMP-tyrosine methyl ester, used in the cAMP radioimmunoassays, was from Hazelton-Washington. Oligonucleotides were from the University of Cincinnati DNA Core Facility. Restriction endonucleases were from New England BioLabs. Salmeterol and formoterol were a gift from Glaxo-Wellcome. All other reagents were from Sigma.

RESULTS
In order to develop a useful model of salmeterol/␤ 2 AR binding, initial experiments were carried out using the ␤ 2 AR and the closely related ␤ 1 AR subtype expressed in the same parental cell. Human ␤ 2 AR (18) and ␤ 1 AR (19) were recombinantly expressed in COS-7 cells and exposed to 100 M isoproterenol (a short acting agonist), 10 M formoterol (a moderately longacting agonist thought to act via a membrane depot effect), or 1 M salmeterol. After 10 min, the cells were washed extensively, and the number of receptors occupied by drug was assessed by means of radioligand binding. For ␤ 2 AR, Ͻ2% of the receptors remained occupied by isoproterenol or formoterol after such washout, but 38 Ϯ 2.3% (i.e. 1696 Ϯ 151 of 4464 Ϯ 242 fmol/mg protein, n ϭ 11) remained occupied by salmeterol. For the ␤ 1 AR, Ͻ2% of the receptors remained occupied after treatment with any of the three agonists, even at concentra-tions up to 10 times their respective K i (as determined in radioligand binding studies; see Table I and discussion below) and subsequent washout. Given that these two receptors were expressed in an identical membrane environment, these results suggest that salmeterol's persistence at ␤ 2 AR results from its binding specifically to a domain within the ␤ 2 AR receptor itself, rather than simply within the membrane bilayers.
To identify a specific domain within the ␤ 2 AR that is required for persistent binding, we constructed chimeric ␤ 2 AR having ␤ 1 AR sequences substituted into TMD4 (Fig. 1b). This segment was chosen for mutagenesis because it displays the greatest difference in primary amino acid sequence between ␤ 1 AR and ␤ 2 AR TMDs (ϳ55% identity), and because earlier mutagenesis studies by our group (20) and others (21) have suggested that this domain may be important in agonist side chain interactions with ␤ 2 AR. Of note, the TMD3 and TMD5 segments that have the key determinants of active site binding are virtually identical (19) between the two receptors. Radioligand binding data from competition studies for these receptors are shown in Fig. 2 and are summarized in Table I. As indicated in the table, substitution of the entire ␤ 1 AR TMD4 into the ␤ 2 AR (mutant ␤ 2 S-(149 -173)) had no significant effect on the affinity of the radioligand ICYP for the receptor (dissociation constant K D ϭ 41.7 Ϯ 2.3 versus 38.0 Ϯ 5.6 pM for wild type, n ϭ 3). In addition, in radioligand binding competition studies this mutant displayed no significant differences in af-finity for isoproterenol (inhibition constant K i ϭ 67.0 Ϯ 4.0 versus 105 Ϯ 17.0 nM) or epinephrine (327 Ϯ 52.4 versus 477 Ϯ 69.9 nM) compared with the wild-type ␤ 2 AR. Importantly, the affinity of the ␤ 2 S-(149 -173) chimera (as well as each of the other ␤ 2 AR chimeras studied) for salmeterol was no different than that observed for the wild-type ␤ 2 AR (K i ϭ 0.63 Ϯ 0.09 versus 1.29 Ϯ 0.51 nM, respectively; n ϭ 6 -8, p ϭ 0.30). No major alterations in affinities were detected with any of the other receptors, and the indicated agonists, although small, statistically significant differences were noted in some cases (Table I). It should be noted that although one report (21) using different chimeras suggested that ␤ 2 AR versus ␤ 1 AR agonist specificity is predominantly dictated by TMD4, this was not observed in the current study (Fig. 2). Thus, neither the ␤ 2 AR rank order of agonist affinities (isoproterenol Ͼ epinephrine Ͼ Ͼ norepinephrine) was changed by these mutations nor was the affinity of the ␤ 2 AR-specific agonist albuterol altered. Taken together, these results suggest that the overall integrity of the active binding site for agonists remained intact for each mutant receptor.
As described above, wild-type ␤ 2 AR display persistent occupation of receptors by salmeterol after extensive washing of the cells. The mutant receptors were thus studied in a similar fashion, with the extent of persistent binding of salmeterol to wild-type ␤ 2 AR (termed exosite binding) serving as the control. The results of these studies are shown in Fig. 3. Substitution of   FIG. 1. Structure of salmeterol and the ␤ 2 AR mutant receptors. a, chemical structure of the salmeterol molecule, which consists of an aralkyloxyalkyl substitution of the saligenin ethanolamine salbutamol. b, schematic representations of the mutated ␤ 2 AR constructs, consisting of the human ␤ 2 AR with substitutions of human ␤ 1 AR sequence in the fourth TMD. Only TMDs three to five are shown for clarity. Solid circles represent ␤ 1 AR sequence substituted into the ␤ 2 AR by site-directed mutagenesis, as described under "Experimental Procedures." The numbering system refers to the amino acid residues of the ␤ 2 AR, relative to the initiator methionine, which were substituted with ␤ 1 AR residues. For mutant ␤ 2 S-(149 -158), the homologies between the ␤ 2 AR and ␤ 1 AR are shown, with dissimilar residues bracketed. the entire ␤ 1 AR TMD4 sequence into ␤ 2 AR (mutant ␤ 2 S-(149 -173)) resulted in a 67% decrease in exosite binding. In contrast, substitution of residues within TMD4 closest to the extracellular region of the receptor (mutant ␤ 2 S-(164 -173), Fig. 1b) had no effect. Consistent with this, replacement of partial TMD4 sequence in mutant ␤ 2 S-(149 -166) resulted in a loss of exosite binding similar to that found when the entire TMD4 chimera was studied. Finally, the receptor that had a small 10-amino acid substitution at the intracellular face of TMD4, ␤ 2 S-(149 -158), was also found to display a loss (66%) in exosite binding equivalent to that of the full TMD4 chimera, thus isolating a critical domain required for such binding to this small region. As shown in Fig. 1, within these 10 amino acids there are 6 that are dissimilar between the ␤ 2 AR and ␤ 1 AR. Since one of these (Lys-149) is just outside TMD4, it can be reasonably assumed that a minimal critical domain for exosite binding at the ␤ 2 AR is the sequence Val-Ile-Ile-Leu-Met at residues 152-156. As indicated above, the wild-type ␤ 2 AR displayed no exosite binding for isoproterenol or formoterol. None of the mutations imposed had any effect on this parameter (data not shown).
In order to demonstrate the functional consequences of exosite binding, stable CHW cell lines expressing either wild-type ␤ 2 AR or ␤ 2 S-(149 -158) were established, and the ability of isoproterenol or salmeterol to activate each receptor under washout conditions was examined. Receptor expression was equivalent for the two lines studied (635 Ϯ 61 fmol/mg protein for ␤ 2 AR versus 664 Ϯ 48 fmol/mg protein for ␤ 2 S-(149 -158), n ϭ 5). Fig. 4 presents the cAMP content determined in the supernatants of CHW cells expressing each receptor before washout (perfusion time ϭ 0) and following each subsequent 2.5-min perfusion washout (for simplicity, these perfusion times were summed and expressed as an aggregated time for each data point). As shown in the figure, salmeterol-promoted cAMP accumulation in CHW cells expressing wild-type ␤ 2 AR remained at or near initial values despite repeated washings, indicating persistent activation of the receptor by retained agonist. In contrast to the results obtained with the wild-type ␤ 2 AR, but consistent with the radioligand binding studies described above, salmeterol-promoted cAMP levels in cells expressing ␤ 2 S-(149 -158) demonstrated a gradual decline. Thus, at 10 min aggregate perfusion (washout) time, cAMP levels for the wild-type ␤ 2 AR was still 91.9 Ϯ 2.7% of the control (prewashout) value versus 40.5 Ϯ 7.5% for the ␤ 2 S-(149 -158) receptor, a reduction of 56% (n ϭ 5, p ϭ 0.0013). Persistent receptor activation was not observed following exposure of cells (149 -173). COS-7 cells were transiently transfected with either wild-type ␤ 2 AR (solid symbols) or ␤ 2 S-(149 -173) (open symbols) constructs, membranes prepared, and competition binding assays performed in the presence of GTP and the indicated concentrations of the agonists isoproterenol (Ⅺ, f), epinephrine (Ç, å), and norepinephrine (E, q), as described under "Experimental Procedures." In each case, the data were fit to a single-site binding model using software from GraphPad and inhibition constants K i determined from EC 50 data by the method of Cheng and Prusoff (17). For each of the agonists studied, little or no difference in agonist affinity was observed between the wild-type and mutated receptor; in addition, no effect on the typical ␤ 2 AR rank-order profile (isoproterenol Ͼ epinephrine Ͼ Ͼ norepinephrine) was observed. Shown are mean data Ϯ S.E. for four or five experiments; see also Table I. FIG. 3. Exosite binding of salmeterol to wild-type and mutant ␤ 2 AR. COS-7 cells expressing each of the receptors were preincubated at 37°C for 10 min with either 1 M salmeterol or vehicle, the media removed, and the cells washed continuously at room temperature with PBS at a constant flow rate of 20 ml/min for 30 min. Cells were then lysed, membranes prepared, and the apparent receptor density (i.e. the number of ␤ 2 AR not occupied by drug) was determined as described under "Experimental Procedures." With ␤ 2 AR, 38.2 Ϯ 2.3% of the receptor remained occupied by salmeterol after this washout procedure, indicative of persistent binding to the wild-type receptor. Data are presented as the percent loss of this amount of exosite binding for the various mutant receptors. Treatment with isoproterenol and formoterol revealed no persistent binding in the wild-type ␤ 2 AR or the mutant receptors (data not shown).

TABLE I Equilibrium radioligand binding data for COS-7 cell membranes expressing the indicated ␤AR constructs.
Inhibition constants K i and dissociation constants K D were determined in ICYP radioligand binding assays in the presence of GTP as described under "Experimental Procedures." Data shown are mean Ϯ S.E. for three to eight experiments, each performed in duplicate. In general, effects on agonist affinities were small, relative to wild type ␤ 2 AR. No differences in salmeterol affinity were observed for any of the ␤ 2 AR chimeras studied. ␤ 1 AR data are presented for comparative purposes only. ISO, isoproterenol; EPI, epinephrine; NOREPI, norepinephrine; ND, not determined. To examine further the role of the small ␤ 2 AR domain to function as a salmeterol exosite binding domain, we substituted the Val-Ile-Ile-Leu-Met ␤ 2 AR sequence (residues 152-156) into TMD4 of the wild-type ␤ 1 AR. The ability of salmeterol to persist at this reverse chimera, termed ␤ 1 S-(177-181), despite washout conditions was then assessed in a similar manner as described for the ␤ 2 AR constructs above. As shown in Fig. 5, under the reduced washout stringency conditions described in the figure legend, wild-type ␤ 1 AR expressed in COS-7 cells still did not exhibit any retained binding of salmeterol (Ϫ0.6 Ϯ 1.8% change in apparent receptor density versus untreated control, n ϭ 5). In contrast, COS-7 cells expressing ␤ 1 S-(177-181) demonstrated a 13.6 Ϯ 2.2% (n ϭ 5) loss of apparent receptor density following treatment with salmeterol, indicating retained receptor occupancy by drug. This difference, which represents a gain in exosite binding, was highly significant (p ϭ 0.001). These results were not due to an increase in salmeterol affinity for the mutant versus wild-type ␤ 1 AR; in fact, the affinity of salmeterol for the ␤ 1 S-(177-181) chimera was actually reduced slightly relative to the wild-type ␤ 1 AR (K i ϭ 1130 Ϯ 77 (n ϭ 5) versus 826 Ϯ 85 (n ϭ 8) nM, respectively; p ϭ 0.03).

DISCUSSION
As introduced earlier, salmeterol is a modified ␤ 2 AR agonist that exhibits markedly prolonged duration of activity (Ͼ15 h) in multiple physiologic preparations. Although salmeterol is highly lipophilic (log octanol:H 2 O partition coefficient ϭ 3.88 (4)), the activity of salmeterol in such preparations is not adequately explained by increased lipophilicity alone. For example, although activation of ␤ 2 AR by salmeterol is blocked by typical ␤AR antagonists (indicating that the drug acts via the receptor), this activity recurs rapidly following washout of the antagonist. Kinetic studies have demonstrated that the drug dissociates from model lipid membranes much faster (t1 ⁄2 Ϸ25 min) than is observed in vivo, suggesting that other components of the native cell membrane architecture, including perhaps the receptor itself, are involved in prolonging the effective duration of action of the drug (9).
Based on studies such as these, a unique model of salmeterol binding to the ␤ 2 AR has been proposed. According to this model, salmeterol binds not only to the receptor active site but in addition to a second locus within the receptor. This second locus, termed the exosite, allows the drug to persist at the receptor even in the presence of concurrent antagonist occupancy at the active site. By anchoring the agonist to the receptor, the concentration of agonist available for active site binding might be maintained, rather than decline, due to degradation, redistribution, and/or other elimination pathways. Since the equilibrium between activated (agonist-bound) and inactivated receptor depends, in part, on the concentration of agonist available to the receptor, the net result of anchored ligand binding would be an increase in the probability of receptor activation and thus a prolongation of effective agonist duration of action.
The goal of the current study was to identify the molecular basis of such an exosite. We hypothesized that if such a domain existed, alteration of these amino acids would attenuate the ability of salmeterol to persist at the receptor. In order to demonstrate the specificity of such a domain, we chose to substitute candidate regions of the ␤ 2 AR with corresponding regions of the ␤ 1 AR. In doing so, we made use of the observa-tion that although salmeterol binds to the ␤ 1 AR, salmeterol does not exhibit exosite binding to this receptor despite the fact that the ␤ 1 AR and ␤ 2 AR display a moderate degree of homology within the putative ligand binding domains. We selected the fourth transmembrane spanning domain (TMD4) for study for several reasons. First, the greatest degree of dissimilarity among the TMDs of these receptors is observed in TMD4; if exosite binding is indeed unique to the ␤ 2 AR, one might expect it to occur in a region mostly dissimilar to the ␤ 1 AR. Second, using chimeric ␤ 1 AR/␤ 2 AR constructs Frielle et al. (21) demonstrated that while multiple domains appeared to be involved in determining the subtype specificity of agonist binding, the ma- FIG. 4. Functional consequences of ␤ 1 AR sequence substitution within the ␤ 2 AR exosite. Wild-type ␤ 2 AR and ␤ 2 S-(149 -159) constructs were stably expressed in CHW-1102 cells, and whole-cell cAMP accumulation following a single 10-min exposure to salmeterol (0.1 nM) was measured under washout conditions as described under "Experimental Procedures." Results are normalized to the pre-washout value to illustrate the persistence of activation. In contrast, the response to washout of isoproterenol demonstrated a rapid decline with wild-type ␤ 2 AR (Ⅺ) as well as the ␤ 2 S-(149 -158) mutant (f) as shown. Initial cAMP levels were not different between the wild-type ␤ 2 AR and the mutant receptor when normalized to forskolin (1 M) responses (40.7 Ϯ 2.1 pmol/ml ϭ 21.2 Ϯ 1.1% of forskolin response versus 14.3 Ϯ 1.7 pmol/ml ϭ 26.8 Ϯ 6.0% of forskolin response, respectively; n ϭ 6, p ϭ 0.38). The persistence of activation by salmeterol at the ␤ 2 S-(149 -158) receptor, relative to the wild-type ␤ 2 AR, is attenuated at all washout intervals studied. Receptor expression, as determined by radioligand binding, was equivalent between the two cell lines (635 Ϯ 61 fmol/mg protein for ␤ 2 AR versus 664 Ϯ 48 fmol/mg protein for ␤ 2 S-(149 -158), n ϭ 5).
FIG. 5. Exosite binding of salmeterol to wild-type and mutant ␤ 1 AR. Exosite binding was measured in COS-7 cells expressing either wild-type ␤ 1 AR or ␤ 1 S-(177-181) similar to that described in the legend for Fig. 3, except that the concentration of salmeterol used in the preincubation step was increased to 10 M, the time of preincubation increased to 30 min, and the time of washout perfusion decreased to 15 min. Even under these reduced stringency conditions, no detectable retained salmeterol binding was evident for cells expressing the wildtype ␤ 1 AR (Ϫ0.6 Ϯ 1.8% change in apparent receptor density versus untreated control cells, n ϭ 5). In contrast, the chimeric ␤ 1 S-(177-181) receptor displayed 13.6 Ϯ 2.2% loss of apparent receptor density, indicating a gain in exosite binding (n ϭ 5, p ϭ 0.001 versus wild-type ␤ 1 AR). Total receptor density did not differ for the wild-type and mutant receptors (3736 Ϯ 311 versus 4269 Ϯ 798 fmol/mg protein, n ϭ 5, p ϭ 0.55). jority of such specificity was observed when TMD4 was altered. Again, since exosite binding is unique to the ␤ 2 AR, it would be reasonable to assume that a candidate exosite domain might likewise be involved in subtype-selective agonist specificity. Third, we have previously demonstrated, using a naturally occurring variant (polymorphism) of the human ␤ 2 AR within TMD4, that this region of the receptor may significantly modulate drug binding via the interaction of the ligand side chain with this domain of the receptor (20). Finally, as discussed below, initial molecular modeling (22) suggested that an interaction of salmeterol with TMD4 was reasonable, based on the molecule's predicted preferred conformation relative to the active binding sites in TMD3 and TMD5 and the length of the side chain moiety.
The results of the current study clearly support the notion that salmeterol exhibits a prolonged duration of activation at the ␤ 2 AR via binding of the drug to a localized receptor domain, as opposed to an interaction that is independent of the receptor structure. First, exosite binding is dependent on the presence of the ␤ 2 AR and is not evident when identical cells expressing the structurally related ␤ 1 AR are expressed instead. Second, exosite binding, as defined by resistance to washout, can be largely (although not completely) eliminated by the substitution of a small region within a transmembrane spanning domain of the receptor with the analogous ␤ 1 AR sequence. This effect on exosite binding appears to be distinct from active site binding, in that there were no significant effects on agonist affinity (Figs. 2 and 3, and Table I). Indeed, it should be emphasized that this substantial reduction in salmeterol retention during washout cannot be attributed to a decrease in affinity of salmeterol for the receptor (as defined in radioligand binding competition experiments), since these affinities, if anything, tended toward being greater for some mutant receptors. Third, the persistence of functional activation of the ␤ 2 AR by salmeterol is substantially attenuated when the exosite is altered (Fig. 4). It should be noted that persistent stimulation of ␤ 2 AR by salmeterol in the system that we utilized is a complex process that involves not only activation of the receptor but potential desensitization of the receptor as well (6). However, it is unlikely that the differences in cAMP accumulation observed between the wild-type and mutant ␤ 2 AR in the present study are due to differences in desensitization patterns, since no differences were observed for isoproterenol-promoted cAMP accumulation under these conditions. Finally, introduction of the small domain into the wild-type ␤ 1 AR (mutant ␤ 1 S-(177-181)) conferred retention of salmeterol binding under washout conditions, which was not observed for the wild-type ␤ 1 AR receptor (Fig. 5). This gain-of-function mutation strongly supports the notion of a discrete sequence within the ␤ 2 AR as necessary for exosite binding. Several possibilities exist as to the observed difference in exosite binding between the wildtype ␤ 2 AR and the chimeric ␤ 1 S-(177-181) receptors. For example, other regions of the ␤ 2 AR not present in the ␤ 1 AR, and not explored here, may also be contributing to exosite-type binding interactions with salmeterol. Or it is possible that the lower affinity of salmeterol at the ␤ 1 AR active site may be prohibitive of "full" (i.e. ␤ 2 AR-like phenotype) exosite binding. Fig. 6 is a model representation of the interaction of salmeterol with the ␤ 2 AR based upon the current studies. Shown are the interactions at Ser-204, Ser-207, and Asp-113 and the exosite interaction of the salmeterol side chain to the identified TMD4 segment. As predicted by the physical constraints of such an arrangement, the salmeterol molecule adopts a bend in the side chain in order to interact with this region. Such a bend might occur at the oxygen moiety that is a point of flexion in space filling models (22). This is consistent with recent studies using salmeterol analogues in which the oxygen was positioned at different locations along the alkyl side chain (4). In compounds where the oxygen was placed two or eight carbons from the nitrogen, the duration of action was reduced to Ͻ30 min as compared with Ͼ12 h observed with salmeterol, where the oxygen is at position six. Since the lipophilicity of these com- pounds is not changed by the altered position of the oxygen (4), this further supports the concept that the mechanism of salmeterol's persistent action is not simply one of a nonspecific lipophilic integration of the drug within the membrane but a specific interaction with a distinct exosite domain of the ␤ 2 AR. This interaction may nonetheless be related to the lipophilic nature of the salmeterol side chain since the residues identified as the anchor in the current study comprise the most hydrophobic region of TMD4.
In summary, the present work confirms the molecular basis for binding of salmeterol to a region of the ␤ 2 AR distinct from the active binding site. Elimination of this exosite domain results in loss of the ability of salmeterol to persist at the receptor, without affecting the overall affinity for the active site as determined by radioligand binding analyses. Taken together, this provides a molecular mechanism to explain, in part, salmeterol's prolonged physiologic duration of bronchodilation, as well as the ability of salmeterol to persevere at the receptor despite reversible active site blockade by antagonists. This type of binding may have significant implications for other G protein-coupled receptors. As has been demonstrated in clinical trials, inhaled salmeterol provides for a long duration of bronchodilation in the treatment of asthma, resulting in substantially improved control of the disease over that provided by traditional short-acting agents (23). To date, the mechanism of salmeterol's long duration of action had not been delineated and indeed the existence of an exosite-type mechanism intensely debated (24). With the results of the current work, which identifies a specific domain within the ␤ 2 AR that provides for exosite binding, the possibility exists that ligands for other G protein-coupled receptors could be designed based on this mechanism to provide for repetitive activation, thereby enhancing the therapeutic utility of drugs acting via these receptors.