Metal Ion Site Engineering Indicates a Global Toggle Switch Model for Seven-transmembrane Receptor Activation*

Much evidence indicates that, during activation of seven-transmembrane (7TM) receptors, the intracellular segments of the transmembrane helices (TMs) move apart with large amplitude, rigid body movements of especially TM-VI and TM-VII. In this study, AspIII:08 (Asp113), the anchor point for monoamine binding in TM-III, was used as the starting point to engineer activating metal ion sites between the extracellular segments of theβ2-adrenergic receptor. Cu(II) and Zn(II) alone and in complex with aromatic chelators acted as potent (EC50 decreased to 0.5 μm) and efficacious agonists in sites constructed between positions III:08 (Asp or His), VI:16 (preferentially Cys), and/or VII:06 (preferentially Cys). In molecular models built over the backbone conformation of the inactive rhodopsin structure, the heavy atoms that coordinate the metal ion were located too far away from each other to form high affinity metal ion sites in both the bidentate and potential tridentate settings. This indicates that the residues involved in the main ligand-binding pocket will have to move closer to each other during receptor activation. On the basis of the distance constraints from these activating metal ion sites, we propose a global toggle switch mechanism for 7TM receptor activation in which inward movement of the extracellular segments of especially TM-VI and, to some extent, TM-VII is coupled to the well established outward movement of the intracellular segments of these helices. We suggest that the pivots for these vertical seesaw movements are the highly conserved proline bends of the involved helices.

Seven-transmembrane (7TM) 2 receptors, which couple through G proteins but also through a number of G protein-independent signaling pathways such as arrestin-mediated kinase activation, constitute the largest family of proteins in the human genome (1)(2)(3). These receptors are activated by basically all kinds of chemical messengers in the body, and they act as chemosensors for a multitude of external chemical signals. 7TM receptors are also the target for the majority of drugs, which, however, address only a small fraction of the receptor repertoire (4). Because they are involved as regulators of key physiological functions throughout the body, 7TM receptors are one of the major focus areas of the pharmaceutical industry (4,5).
7TM receptors are activated by agonists that span a chemical diversity range from large glycoprotein hormones over neuropeptides, lipid messengers, and nucleotides to small monoamines, amino acids, and even metal ions such as calcium (1)(2)(3). Notably, although there apparently is no common "lock" for all these chemically highly diverse "keys" (6), it is nevertheless expected that there is a common molecular activation mechanism for 7TM receptors as such. The conformational changes that accompany receptor activation have been characterized in rather great detail for the intracellular segments of the seven-helix bundle (7)(8)(9)(10)(11)(12)(13)(14). This has been achieved through a number of different biochemical and biophysical approaches, including systematic EPR spectroscopic analysis of all intracellular segments of the transmembrane helices (TMs) as well as the intracellular loops and tail of rhodopsin by site-directed spin labeling (8,10,14). In brief, these studies show that, during receptor activation, the TMs (especially TM-VI) move apart and thereby disclose receptor epitopes, which are believed to be recognized by the intracellular signaling molecules, G proteins, arrestin, etc. (14 -16). However, our knowledge about the structural changes that occur in the extracellular parts of 7TM receptors during activation is very limited, as systematic site-directed EPR analysis has not yet been performed in this area. Moreover, although several x-ray structures of rhodopsin are available, all of these represent only the inactive state of the protein molecule (17)(18)(19)(20).
To try to understand the activation mechanism for 7TM receptors, in this study, we have taken an alternative, biochemical approach by making the smallest and best understood ligand (a metal ion) act as an agonist. Because the coordination of metal ions in proteins is understood in great detail from a large number of especially x-ray structures, it is expected that reliable distance constraints concerning residues (and thereby the TMs) can be obtained from activating metal ion sites systematically built into the main ligand-binding pocket in the extracellular part of the receptors. Previously, we constructed an activating metal ion site between TM-III and TM-VII in the ␤ 2 -adrenergic receptor and transferred this site to the NK1 substance P receptor (21,22). However, because a high resolution structure of a 7TM receptor was not available at that time, it was not possible to use this distance constraint to propose specific models for receptor activation. Nevertheless, we suggested that the reason why only ϳ25% signaling efficacy was achieved in the III:08/ VII:06 site was due to lack of proper interactions with the important TM-VI (22). Thus, on the basis of the III:08/VII:06 site, in this study, we have made a number of constructs with metal-binding residues in these two positions as well as in different positions in TM-VI (see Fig. 1). Molecular modeling based on the distance constraints imposed by the activating metal ion sites combined with the knowledge from the x-ray structure of the inactive form of rhodopsin indicate that the extracellular parts of the helices will have to move toward each other during 7TM receptor activation to coordinate the activating metal ion. In view of the fact that there is ample evidence indicating that the intracellular parts of these helices move in the opposite direction (i.e. away from each other) during receptor activation, we consequently propose a "global toggle switch model" for the activation of 7TM receptors in which these helical movements are conjoined in vertical seesaw movements involving especially TM-VI.

Molecular Biology
The point mutations were constructed using oligonucleotide-directed mutagenesis and recombinant PCR (21). cDNAs encoding wildtype and mutant receptors were cloned into the eukaryotic expression vector pTEJ-8; all mutations were verified by restriction endonuclease mapping and DNA sequencing.

Cell Biology
Cloned receptors were transiently expressed in COS-7 cells transfected 2 days before analysis. COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM glutamine, and 10 g/ml gentamycin.

Determination of Intracellular cAMP Accumulation
COS-7 cells were seeded in 6-well culture dishes 1 day after transfection at a density of 400,000 cells/well and supplemented with 2 Ci/ml [ 3 H]adenine (Amersham Biosciences). Two days after transfection, cells were washed twice with HEPES-buffered saline (25 mM HEPES, 0.75 mM NaH 2 PO 4 , and 140 mM NaCl, pH 7.2) and incubated in HEPESbuffered saline supplemented with 1 mM 3-isobutyl-1-methylxanthine. Pindolol, free metal ion, or metal ion complex was added, and the cells were incubated for 30 min at 37°C. The assay was terminated by aspirating the buffer, followed by addition of ice-cold 5% trichloroacetic acid containing 0.1 mM unlabeled cAMP and ATP. cAMP was isolated by applying the supernatant to AG 50W-X4 resin (Bio-Rad), followed by alumina resin (catalog no. A9003, Sigma). Determinations were made in duplicate. Because COS-7 cells express endogenous catecholamine receptors, which are stimulated by isoproterenol, giving a cAMP accumulation of up to 96 fmol/10 5 cells, but not by pindolol, this compound was used as agonist for the ␤ 2 -adrenergic constructs in this study.

Data Analysis
The cAMP curves were analyzed, and EC 50 values were determined by computerized nonlinear regression using GraphPad Prism.

Molecular Modeling of Receptors
Analysis of Metal Ion Site Distances-The homology modeling of the ␤ 2 -adrenergic receptor utilized the crystal structure of bovine rhodopsin (Protein Data Bank code 1F88; solved to 2.8-Å resolution) as a structural template (17,23). Sequence alignment of the ␤ 2 -adrenergic receptor and bovine rhodopsin was performed on the basis of conserved fingerprint residues and motifs in the seven TMs (24). Two models of mutant forms of the ␤ 2 -adrenergic receptor were produced: 1) the CysVI:16/CysVII:06 double mutant with the wild-type AspIII:08 metal ion-binding residue in TM-III constituting part of the proposed tridentate metal ion-binding site and 2) the HisIII:08/CysVI:16/CysVII:06 triple mutant. The software MODELLER was configured to produce 25 three-dimensional comparative homology models of each of the two mutant forms of the ␤ 2 -adrenergic receptor (25)(26)(27). For each receptor construct, the 7TM domains of the 25 resulting models were superimposed on the 7TM domain of bovine rhodopsin, considering heavy backbone atoms only. The calculated backbone root mean square deviations were between 0.06 and 0.28 Å. The resulting models were evaluated based on manual visualization and the MODELLER fitness score. Inspection of the individual structures showed that a number of residues, especially in the extracellular part of the 7TM segments, in the main ligand-binding pocket had several possible side chain conformations. The different side chain conformations could be mapped in general to different discrete rotamer states with 1 / 2 angles within a 30°w indow from corresponding side chain rotamers as listed in the Lovel rotamer library. Considering the packing environment for the different side chains, all possible combinatorial accessible rotamer states for the residue positions composing the two introduced metal ion-binding sites (AspIII:08/CysVI:16/CysVII:06 and HisIII:08/CysVI:16/CysVII:06) were produced in SYBYL and the biopolymer module using the Lovel rotamer library (SYBYL 7.0, Tripos, Inc., Louis, MI). The heavy atom distance distributions between either of the two acidic oxygens of AspIII:08 and the sulfur of CysVI:16 and CysVII:06 in the AspIII:08/CysVI:16/Cys-VII:06 construct and the ⑀or ␦-nitrogen in the imidazole ring of HisIII:08 and the sulfurs of CysVI:16 and CysVII:06 in the HisIII:08/ CysVI:16/CysVII:06 receptor construct were calculated using customized scripts.
Searches in Relibase (available at relibase.rutgers.edu), which is a data base of protein-ligand complexes (28), for metal ion complexes in which Zn(II) simultaneously interacts with a nitrogen atom of a His residue and the sulfur atom of a Cys residue showed that, among 2862 sites, the most frequently observed interaction distance between Zn(II) and sulfur is 2.2 Å and that between Zn(II) and the imidazole ⑀-nitrogen is 1.8 -2 Å and that the most populated angles between imidazole nitrogen, Zn(II), and sulfur are 90 -120°. Thus, in a Zn(II) metal ion-binding site, the distance between the imidazole ⑀-nitrogen and the sulfur of a Cys residue is typically between 3.35 and 3.5 Å and can be up to 4.06 Å, which, however, is observed in only ϳ1% of the complexes.
Generation of an "Active" Receptor Conformation-The helical bundle of an inactive model of the ␤ 2 -adrenergic receptor was used as a template to produce a model of the active conformation. A tridentate construct was used for convenience because it is envisioned that, during activation, movements of both TM-VI and TM-VII will occur and because the two bidentate sites converge on position III:08. The active receptor conformation was produced by Monte Carlo simulated annealing protocols (29) using distance constraints applied by the nuclear Overhauser effect (NOE) functionality of CHARMM corresponding to 1) the metal ion site, a 2-Å distance constraint between Zn(II) and the heavy atom of the three involved residues and a 3.2-4.0-Å constraint between the sulfur of the Cys residues and the nitrogens of the His residues; 2) the intracellular segment of TM-VI, a 16-Å NOE constraint for the backbone C ␣ -C ␣ distance between ArgIII:26 and LysVI:04 consistent with an outward movement of 6 Å plus a 9-Å constraint for the backbone C ␣ -C ␣ distance between TyrV:24 and LysVI:04 to ensure a close spatial relationship between the intracellular ends of TM-V and TM-VI (14); 3) movement of the intracellular segment of TM-VII, a C ␣ -C ␣ distance constraint of 19 Å between ArgIII:26 and TyrVII:20 consistent with an outward movement of 3 Å (14); and 4) the conserved hydrogen bond network between TM-I, -II, and -VII (17,30,31), a 2.8 -3.5-Å heavy atom distance constraint between the side chain nitrogen of AsnI:18 and the free backbone carbonyl of SerVII:13 and between the side chain carboxylate oxygen of AspII:10 and the nitrogen of AsnVII:16. To allow TM-VI to possibly straighten out, the / backbone torsion angles of TrpVI:13, which are responsible for the pronounced kink in this helix, were allowed to be perturbed to obtain energetically favorable conformations to stabilize the helix locally. Similarly, to allow TM-VII to possibly straighten out, the backbone / torsion angles of ValVII:11 to AsnVII:16 were allowed to be perturbed. To allow TM-VII to straighten out, distance constraints were in some simulations added to optimize the backbone hydrogen bonds toward a more regular ␣-helix. An NOE distance constraint of 2.8 -3.5 Å between the amide nitrogen and carbonyl oxygen of the i and iϪ4 positions was applied to LeuVII:18 and GlyVII:14, Asn-VII:16 and AsnVII:12, PheVII:15 and ValVII:11, GlyVII:14 and TyrVII: 10, and SerVII:13 and GlyVII:09.
An initial temperature of 1000 K was used. In the temperature interval of 1000 to 100 K, the temperature was lowered in steps of 100 K. In the interval of 100 to 0 K, the temperature was lowered in steps of 10 K. For each temperature, the system was optimized in 1000 Monte Carlo steps, i.e. a total of 20,000 Monte Carlo steps were applied in the simulation. The coordinates were saved every 200 steps. The Monte Carlo command (MOVE ADD) in CHARMM was used to construct the move sets (32). During the simulation protocol, perturbations were applied to translations, rotations, and torsional degrees of freedom for the defined move sets. For simplicity, the backbones of TM-I-V were fixed throughout the entire simulation. We used the defined CHARMM move types for rigid translations (RTRN) along a random uniformly distributed vector for the two move sets comprising all atoms in TM-VI and TM-VII, respectively. Similarly, rigid rotations (RROT) were applied to the same two move sets. The origin of the uniformly distributed rotation vector was located at the geometric center for TM-VI and TM-VII, respectively. Concerted perturbations of individual backbone torsions (CROT) were applied to perturb the central region of TM-VI and TM-VII, respectively, to locally allow modifications of the helical kinks induced by ProVI:15 and ProVII:17. Torsional perturbations were applied to all torsions of the side chains in the main ligand-binding crevice, consistent with the template retinal-binding site in rhodopsin defined as all atoms (selected by residue) within a 6-Å radius from any retinal atom. The following move criteria were applied to the different move sets. A D max of 0.1 Å was used for rigid translation. For rigid rotations, a D max of 5°was applied, whereas the D max for torsions of side chains in the binding was 180°to explore alternative rotamer states and side chain packings. The CROT main chain perturbation step was subsequently followed by 10 steps of conjugated gradient minimization, i.e. in addition to a main chain perturbation, all atoms in the binding site region were minimized before the Metropolis acceptance criteria were applied. Each Monte Carlo step consisted of 1) randomly picking a group of move instances (a subset of the protein side chain atoms, the chelator functional group, the metal ion, or the helical segments), 2) performing the "MOVE" (i.e. perturb Cartesian Coordinates of that group; each move was subsequently minimized for 10 steps of steepest decent prior to the Metropolis acceptance criterion) (29); 3) calculating the energetic contribution in the new configuration; and 4) accepting or rejecting the new position with the probability Exp(⌬V ϪMC /kT ).

Connecting the III:08/VII:06 Activating Metal Ion Site with TM-VI-
Originally, a bidentate activating metal ion site was constructed in the ␤ 2 -adrenergic receptor between position III:08, with either the naturally occurring amine-binding Asp residue or a His residue introduced at this position, and a Cys residue introduced at position VII:06 ( Fig. 1) (21). As movement of TM-VI has been strongly implicated in the activation mechanism for 7TM receptors, we used the III:08/VII:06 bidentate site as the starting point to probe the inner face of TM-VI for potential additional metal ion attachment sites. His or Cys residues were systematically introduced at positions VI:16, VI:19, and VI:20 in the AspIII:08/ CysVII:06 background in the ␤ 2 -adrenergic receptor (Fig. 1). The mutant receptors were expressed transiently in COS-7 cells, and cAMP production was measured in response to Cu(II) or Zn(II) either alone or in complex with either bipyridine or phenanthroline and compared with the response to the partial agonist pindolol, which, in the wild-type receptor, increased cAMP accumulation from a basal level of 6.8 Ϯ 0.7 to 45 Ϯ 5 fmol/10 5 cells (Table 1). 3 Introduction of a Cys or His residue at position VI:19 or VI:20 did not improve the ability of the metal ions either alone or in complex with chelators to stimulate cAMP production more than was observed in the background construct, AspIII:08/CysVII:06 (Table 1, Part A). In fact, in most of these constructs, a somewhat reduced response was observed. In the case in which a His residue was substituted for the natural Phe residue at position VI:16 in the AspIII:08/CysVII:06 background, a totally non-responsive receptor was obtained (Table 1, Part A). However, when a Cys residue was introduced at this position in the AspIII: 08/CysVII:06 background, the dose-response curve for Cu(II) in complex with bipyridine and phenanthroline with respect to stimulation of cAMP production was shifted by 5-and 10-fold, respectively, to the left compared with the background bidentate construct, reaching a potency of 0.8 M for Cu(II)-phenanthroline in the AspIII:08/CysVI:16/Cys-VII:06 construct (Fig. 2). This potency corresponds to the metal ion affinity measured in the best tridentate-inactivating metal ion sites constructed previously, i.e. between residues located in the outer segments of TM-V and TM-VI in the NK1 or -opioid receptor (33)(34)(35) or in the muscarinic or opioid receptors (36,37). The maximal signaling efficacy observed for Zn(II) or Cu(II) in complex with chelators was apparently higher than in the bidentate background construct (Table 1, Part A). For example, Cu(II)-bipyridine increased cAMP production from 6.0 Ϯ 4 to 70 Ϯ 7 fmol/10 5 cells in the AspIII:08/CysVI:16/CysVII:06 tridentate construct (Asp 113 /F289C/N312C in Table 1, Part A) compared with a maximal signaling efficacy of 24 Ϯ 4 fmol/10 5 cells in the corresponding bidentate construct, AspIII:08/CysVII:06 (Asp 113 /N312C in Table 1). It should be noted that in no case was an increase in basal activity observed in any of these or the other constructs presented below compared with the wild-type ␤ 2 -adrenergic receptor (Table 1).
Values that were reported for selected cations and complexes on indicated clones (21). by simple washing (Fig. 2C), as also previously found in other constructs (21). In 7TM receptors, disulfide formation occurs rather easily between Cys residues introduced into the loop regions and N-terminal extension, but with more difficulty within the helical bundle (38). Notably, Zn(II) (which is not redox-active) in complex with phenanthroline also acted as a very efficacious agonist in the AspIII:08/CysVI:16/CysVII:06 construct, increasing cAMP activation from 6.0 Ϯ 0.4 to 32 Ϯ 4 fmol/ 10 5 cell (Table 1, Part A). Activating Bidentate Metal Ion Sites between Positions III:08 and VI:16-Having identified Cys at position VI:16 as the optimal metal ion interaction partner on the inner face of TM-VI, we turned back to TM-III and position III:08, where only the natural Asp residue had been explored. In a bidentate setting, AspIII:08 appeared to be the optimal partner for CysVI:16, as both CysIII:08 and HisIII:08 in combination with CysVI:16 gave constructs with lower or no efficacy for the metal ions (Table 1, Part B). Notably and somewhat surprisingly, the AspIII: 08/CysVI:16 bidentate construct displayed a phenotype rather similar to that observed with the tridentate construct described above, i.e. AspIII:08/CysVI:16/CysVII:06. Both Cu(II) and Zn(II) in complex with either of the two chelators stimulated cAMP production with high efficacy in the bidentate site (Table 1, Parts A and B). Also the potency of Cu(II)-phenanthroline or Cu(II)-bipyridine, for example, was high in the AspIII:08/CysVI:16 bidentate construct, i.e. 1.3 and 2.1 M, respectively (Fig. 3), similar to the potencies of these two Cu(II)-chelator complexes in the AspIII:08/CysVI:16/CysVII:06 tridentate construct, 0.8 and 9 M, respectively (Fig. 2). One indication that the additional Cys-VII:06 could in fact still play a role in the tridentate construct was that the efficacies of the phenanthroline complexes with both Cu(II) and Zn(II) were somewhat higher in the tridentate setting than in the bidentate setting: 54 Ϯ 7 versus 18 Ϯ 4 fmol/10 5 cells and 32 Ϯ 4 versus 12 Ϯ 4 fmol/10 5 cells, respectively (Table 1, Parts A and B).
In conclusion, an activating metal ion site can be built between TM-III and TM-VI, with metal ion-binding residues located at positions III:08 and VI:16 as demonstrated in both the AspIII:08/CysVI:16 bidentate and AspIII:08/CysVI:16/CysVII:06 tridentate settings, displaying rather similar properties. However, with an Asp residue at position III: 08, an aromatic chelator was required as a co-ligand for the metal ion to obtain an appreciable agonistic efficacy (Table 1, Parts A and B). In fact, this was also the case in the original TM-III/TM-VII bidentate site, AspIII:08/CysVII:06 (21). Notably, the chelators had no agonistic property without a metal ion in any of the engineered metal ion site receptor constructs of the present or previous studies, indicating that it is the combined metal ion-chelator complex that acts as an agonist in the engineered metal ion sites.
Improving the Efficacy for Free Metal Ions through His Substitution at Position III:08-In the case of the original III:08/VII:06 bidentate construct, high efficacy and potency for the free metal ions were obtained by introducing a longer His residue instead of the natural Asp residue at position III:08 (21). We therefore tested whether a His residue at position III:08 could improve the metal ion activation mechanism in the tridentate setting. In contrast to the situation with an Asp residue at position III:08, both Cu(II) and Zn(II), as the free metal ions, acted as highly efficacious agonists in the HisIII:08/CysVI:16/CysVII:06 tridentate construct ( Table 1, Part C). The agonistic efficacies of the free metal ions were rather similar to those obtained in the HisIII:08/CysVII:06 bidentate construct (Table 1, Part C), whereas the potency for Cu(II) was improved apparently from ϳ5 to 1.7 M in the HisIII:08/CysVI:16/ CysVII:06 tridentate construct (Fig. 4).
However, in the HisIII:08/CysVI:16/CysVII:06 construct, Cu(II) in complex with phenanthroline stimulated cAMP accumulation with the highest potency yet observed in an engineered metal ion site of a 7TM receptor, i.e. 0.49 M, which represents an almost 10-fold increase in potency from 4.6 M observed in the corresponding HisIII:08/Cys-VII:06 bidentate site (Fig. 4). This supports the notion that all three potential metal ion-chelating residues could potentially participate in a tridentate metal ion site in the HisIII:08/CysVI:16/CysVII:06 construct. This was further substantiated by the dose-response curve for Cu(II) in complex with bipyridine, which was clearly biphasic, with a very high potency component (i.e. 0.54 M), corresponding to that observed in complex with phenanthroline, as well as a lower potency component of 64 M (Fig. 4). In the HisIII:08/CysVII:06 bidentate construct, Cu(II)bipyridine stimulated cAMP accumulation in a simple monophasic manner, with a potency of 10 M (Fig. 4).
Thus, introduction of His at position III:08 imposed a truly high submicromolar potency for both Cu(II) and Zn(II) in complex with the aromatic chelators in the III:08/VI:16/VII:06 tridentate construct and ensured high efficacy for the free metal ions, which they did not display with the shorter Asp residue at position III:08, in analogy with the III: 08/VII:06 bidentate construct (21). The efficacies for the various ligands were rather similar in the tridentate and bidentate constellations; however, the activation profiles suggested that all three potential metal ionbinding residues participated in the binding and activation process of the HisIII:08/CysVI:16/CysVII:06 construct (Fig. 4).

Molecular Modeling of the Metal Ion Sites in the Inactive Rhodopsinlike Receptor
Conformation-Homology models of the mutant forms of the ␤ 2 -adrenergic receptor with metal ion-binding residues introduced at positions III:08, VI:16, and/or VII:06 were built over the x-ray structure of the inactive state of rhodopsin, preserving the overall backbone  structure. Special attention was paid to the side chain conformations of residues in the main ligand-binding pocket. Considering the packing environment, all combinatorial accessible rotamer states for the side chains of the metal ion-binding residues were produced, and the distances between the heavy atoms involved in metal ion binding were calculated (Fig. 5). With Asp at position III:08, the distribution of distances between the oxygen of AspIII:08 and the sulfur of CysVII:06 was 4.0 -9.9 Å (mean ϭ 7.2 Å), that between the sulfur of CysVII:06 and the sulfur of CysVI:16 was 4.2-7.0 Å (mean ϭ 5.5 Å), and that between the oxygen of AspIII:08 and the sulfur of CysVI:16 was 6.3-10.3 Å (mean ϭ 7.8 Å) (Figs. 5 and 6). With His at position III:08, these distances were somewhat smaller in general (Figs. 5 and 6). However, the distance between, for example, the ⑀-nitrogen of a His residue and the sulfur of a Cys residue, coordinating, for example, a zinc ion in a tetrahedral conformation, is, according to Relibase (see "Experimental Procedures"), generally between 3.35 and 3.5 Å, which is shorter than even the shortest possible theoretical distance between two of the metal ion-coordinating heavy atoms in the mutant receptors.

. Activation of cAMP production by metal ions in complex with chelators in ␤ 2 -adrenergic receptors engineered with potential bidentate metal ion sites between TM-III and TM-VI and between TM-III and TM-VII. A, dose-response experiments with Cu(II)-bipyridine in the N312C mutant
We conclude that, in receptor models built over the inactive state of rhodopsin, the distances between metal ion-binding residues at positions III:08 and VI:16 and between positions III:08 and VII:06, i.e. the original activating bidentate metal ion site (21), and between positions VI:16 and VII:06 all are too long to form a metal ion site whether in a bidentate or tridentate setting. Thus, during activation, the extracellular ends of TM-III, TM-VI, and TM-VII will have to move closer toward each other to bring the metal ion-binding residues (introduced at positions III:08, VI:16, and VII:06) into configurations in which the heavy atoms of these residues can in fact function as ligands for the activating metal ions in either a III:08/VI:16 or III:08/VII:06 bidentate setting or a potential tridentate setting.

Molecular Modeling of a Putative Active Receptor Conformation Based on the Distance Constraints from the Activating Metal Ion
Sites-In the inactive model of the ␤ 2 -adrenergic receptor, the most favorable possible metal ion site was generated through a search of side chain conformations (see "Experimental Procedures"), and a Zn(II) metal ion was manually docked to initially satisfy the experimental 2-Å distance constraint to the ⑀-nitrogen of HisIII:08. In this setting, the distances from the metal ion to the sulfur atoms of CysVI:16 and CysVII:06 were 5.45 and 4.56 Å, respectively (Fig. 7A). Thus, to form an optimal metal ion-binding site, these distances would have to be diminished by at least ϳ1.7 and 1 Å, respectively.
Putative active receptor conformations were generated through Monte Carlo simulated annealing protocols, applying NOE distance constraints corresponding to the metal ion sites between HisIII:08, CysVI:16, and/or CysVII:06 located in the extracellular segments used in this study plus distance constraints corresponding to the EPR analysis of the movements of the intracellular segments of TM-VI and TM-VII (14). With the application of only these two sets of distance constraints, TM-VI performed a vertical seesaw movement, with an inward movement of the extracellular part and an outward movement of the intracellular part, while it was partly straightened out. At the same time, the intracellular part of TM-VII moved out, but the extracellular part of TM-VII did not change position significantly. During the simulation, the NOE partition of the total energy function converted to energies close to 0 kcal/mol, i.e. all NOE distance constraints were essentially satisfied. Similarly, the total energy of the system decreased as the simulation progressed and converged to a stable level at the end of the simulation. Analysis of the trajectory showed that only minor conformational changes were obtained at the end of the simulation.
Thus, in the molecular model employed and with the applied constraints, the metal ion site was sufficient to move the extracellular segment of TM-VI (but not TM-VII) inward. To investigate whether an inward tilting of the extracellular segment of TM-VII in the direction of TM-II/TM-III would be energetically feasible in this system, extra distance constraints was applied between TM-VII and TM-II and within the backbone of TM-VII (see "Experimental Procedures"). As shown in Fig. 7B, under these conditions, both TM-VI and TM-VII performed vertical seesaw movements, and notably, the total energy of the system was slightly in favor compared with the simulation in which the NOE between TM-VII and TM-II was not applied.

DISCUSSION
The molecular pharmacological data of this as well as previous studies (21,22) demonstrate that metal ions either alone or in complex with chelators can activate the ␤ 2 -adrenergic or NK1 receptor when metal ion-binding residues are placed at position III:08 in combination with positions VI:16 and/or VII:06. Based on the coordination geometry for metal ions, this indicates that the oxygen, nitrogen, or sulfur atoms of the metal ionbinding residues involved will have to be within a distance of ϳ3.5 Å from each other in the active conformation of the receptor. Notably, in receptor models built over the backbone structure of the inactive conformation of rhodopsin and probing all possible rotamer conformations of the side chains, the distances between these metal ion-binding heavy atoms are longer than that (Fig. 6). Thus, the data from this study indicate that the molecular activation mechanism for 7TM receptors must involve a movement of the extracellular ends of TM-III, TM-VI, and TM-VII inward toward each other. Due to the special structural constraint of TM-III being located in the middle of the TM-I-V "domain" of the receptor, which, according to NMR studies, is supposed to form a rather rigid core structure (39), it is suggested that the extracellular segments of TM-VI and TM-VII move inward toward TM-III during receptor activation (Fig. 7).
Bidentate Versus Tridentate Metal Ion Sites-It should be emphasized that the two individual activating bidentate metal ion sites FIGURE 5. Normalized distribution for distances measured between the metal ion-binding heavy atoms in homology models of the AspIII:08/CysVI:16/CysVII:06 and HisIII:08/CysVI:16/CysVII:06 mutants of the ␤ 2 -adrenergic receptor built over the x-ray structure of the inactive state of rhodopsin. All likely combinatorial accessible rotamer states for the side chains of the metal ion-binding residues were produced. The normalized distribution of distances between the metal ion-binding heavy atoms in AspIII:08 and CysVII:06 (A), AspIII:08 and CysVI:16 (B), and CysVI:16 and CysVII:06 (C) in the AspIII:08/CysVI:16/CysVII:06 site are shown as well the corresponding distances between HisIII:08 and CysVII:06 (D), HisIII:08 and CysVI:16 (E), and CysVI:16 and CysVII:06 (F) in the HisIII:08/CysVI:16/CysVII:06 site. The vertical dotted line in each of the panels indicates the distance between, for example, an oxygen and a sulfur coordinating Zn(II) in a tetrahedral conformation (ϳ3.5 Å). In all cases, the measured possible distances are longer than the optimal distance for metal ion binding when the receptor model is built over the inactive rhodopsin structure with a rigid backbone. FIGURE 6. Distances between the metal ionbinding heavy atoms of AspIII:08 (A) or HisIII:08 (B), CysVI:16, and CysVII:06 in homology models of the mutated ␤ 2 -adrenergic receptor built over the inactive conformation of rhodopsin. The range of calculated distances between the metal ion-chelating heavy atoms in the possible different rotamer conformations of the metal ion-chelating residues are indicated in red in angstroms (see "Experimental Procedures" and Fig. 5 for details). The distance between an oxygen and a sulfur coordinating Zn(II) in a tetrahedral conformation is ϳ3.5 Å.
between positions III:08 and VI:16, as identified in this study, and between positions III:08 and VII:06, as identified previously (21,22), are both important in establishing individual distance constraints concerning the active receptor conformation. Because no high resolution receptor structure was originally available, the structural significance of the distance constraint could not be appreciated, however, for the III:08/ VII:06 bidentate site at the time when it was identified (21,22).
Do All Three Metal Ion-binding Residues Participate in the "Tridentate" Constructs?-Rather similar efficacies were obtained in the bidentate and corresponding tridentate settings, i.e. the efficacy of AspIII:08/CysVI:16 was similar to that of AspIII:08/CysVI:16/CysVII:06, and the efficacy of HisIII:08/CysVII:06 was similar to that of HisIII:08/CysVI:16/CysVII:06 (Table 1). However, in both of the tridentate constructs, there were clear indications of improved potencies of the metal ion ligands, suggesting that the "third" residue is in fact involved in the binding of the metal ion and/or metal ion-chelator complex (Figs. 2 and 4). Nevertheless, the biphasic activation curve for Cu(II)-bipyridine in the HisIII:08/CysVI:16/CysVII:06 construct could indicate that, although a very high potency is achieved in this construct, as also indicated by the Cu(II)-phenanthroline curves, the binding mode could be more complex than predicted. Notably, the two individual distance constraints based on the two bidentate metal ion sites converge on the same position, III:08. Consequently, movement of both FIGURE 7. Top view of the suggested vertical seesaw movements of TM-VI and TM-VII in the toggle switch activation model for 7TM activation. A, molecular model of the helical bundle of the engineered metal ion site ␤ 2 -adrenergic receptor built over the x-ray structure of the inactive conformation of rhodopsin as seen from the extracellular side. For simplicity, only the metal ion-binding HisIII:08, CysVI:16, and CysVII:06 residues as well as the important rotamer switch residue (TrpVI:13) are shown in rod models, whereas the rest of the helical bundle is shown only in a blue ribbon model. In this "inactive" model, the distances between the ⑀-nitrogen of HisIII:08 and the sulfurs of CysVI:16 and CysVII:06 are 5.46 and 4.56 Å, respectively. B, molecular model of a proposed active receptor conformation generated through Monte Carlo simulated annealing using NOE distance constraints corresponding to a metal ion site between HisIII:08, CysVI:16, and CysVII:06 as well as distance constraints between the intracellular segments of TM-VI and TM-VII generated by EPR analysis (see "Experimental Procedures" for details) (14). For simplicity, the backbones of TM-I-V (indicated by blue ribbons) was not allowed to move. In this active model, in which the backbones of TM-VI and TM-VII are shown by green ribbons, the distances between the ⑀-nitrogen of HisIII:08 and the sulfurs of CysVI:16 and Cys-VII:06 are 3.47 and 3.50 Å, respectively. The yellow arrows indicate the inward movement of the extracellular segments of TM-VI and TM-VII, and the black dashed arrows indicate the outward movement of the extracellular segments of these helices. Note that TrpVI:13 (partially hidden below the extracellular segment of TM-VI) has changed its rotamer state and is no longer located between TM-III and TM-VI, but is rotated over toward TM-V. residues VI:16 and VII:06 inward toward residue III:08 (to satisfy these two constraints) would not only bring these two residues in TM-VI and TM-VII, respectively, closer to TM-III, but would also bring them closer to each other and thereby increase the possibility of forming a true tridentate metal ion site with a residue at position III:08.
Binding of Free Metal Ions Versus Metal Ion-Chelator Complexes-In this as well as previous metal ion site engineering studies (21,22,40), we used both free metal ions and metal ions in complex with small aromatic chelators as ligands. The rationale for including the chelators is that the complexes will bind to the metal ion, functioning as an "anchor point" for the aromatic chelator, which, depending on the location and surrounding of the site as well as the coordination geometry of the metal ion, can potentially establish second-site interactions with, for example, aromatic residues in the vicinity of the anchoring metal ion site. Thus, the chelator will improve, in certain cases, the potency and/or the efficacy of the metal ion through the establishment of such second-site interactions with the receptor. In this case, TrpIII:04, TrpVI:13, and PheVI:17 are examples of neighboring residues that could potentially be involved in interactions with phenanthroline and bipyridine. However, it is important for the distance constraint argument that the metal ion is the crucial anchoring part of the complex and that the chelators do not by themselves have any agonistic effect in the engineered metal ion site receptor. Second-site interactions for aromatic chelators can, in certain robust receptors, be mapped through additional mutagenesis. 4 However, in this study, no attempt was made to identify the positions of the second-site interactions for the chelators mainly because two or three mutations had already been introduced into the receptor just to generate the metal ion site as such.
Global Toggle Switch Model for Helical Movements during 7TM Receptor Activation-A series of biophysical and biochemical studies, including the systematic spin labeling studies of Hubbell and co-workers (8,14), indicate that, during activation, the intracellular poles of the TMs undergo major conformation changes, which are generally interpreted to represent relatively large amplitude movements (in the case of TM-VI, perhaps on the order of 6 Å at the level of the inner surface of the membrane) outward or away from the center of the receptor (13,41,42). It is envisioned that epitopes such as the TM-III DRY motif and the TM-VII/helix motif thereby become accessible for interaction with downstream signaling molecules (14 -16).
To accommodate both the outward movement of TM-VI and TM-VII at the intracellular ends as well as the opposite, i.e. inward, movement of the same helices at the extracellular ends, we propose a global toggle switch activation model for 7TM receptors. The basic concept is that receptor activation occurs through closing of the main ligand-binding crevice, i.e. around a small molecule agonist such as the metal ion or metal ion-chelator complex in this present study, mediated through an inward movement of the extracellular segments of especially TM-VI and TM-VII. This inward movement is then coupled to a corresponding opening of the helical bundle at the cytoplasmic face of the receptor. Thus, it is envisioned that especially TM-VI and also, to some degree, TM-VII perform a vertical seesaw movement around a "fulcrum" or pivot located close to the middle of the membrane, i.e. corresponding to the location of the conserved prolines (Fig. 7). This seesaw movement could very well be associated with rotation and partial straightening of especially TM-VI.
Conserved Proline Bends as Pivots for Helical Seesaw Movements-Proline is a "helix breaker" that cannot participate in the hydrogen bond backbone network due to the lack of a free ␣-amino group. and it thereby induces a weak point in the helix. Because proline residues are nevertheless highly conserved at certain positions in the middle of the TMs of 7TM receptors, it has been suggested that they may serve key roles as functional switches (2,43). In the inactive form of rhodopsin, a considerable kink is found around the backbone of TrpVI:13 as induced by ProVI:15 (Fig. 7A). Molecular dynamics simulations have shown that this bend is energetically very unfavorable (42). Several models for 7TM receptor activation have focused on various types of rigid body movement of the intracellular segment of TM-VI, often involving straightening of this proline kink (8,13,14,42). We propose that such a movement of the intracellular segment of TM-VI (and TM-VII) is functionally coupled to an opposite movement of the extracellular segment of TM-VI, i.e. in the main ligand-binding pocket.
The side chain of TrpVI:13, which is part of the CWLP fingerprint motif of TM-VI, is in the inactive structure of rhodopsin positioned "vertically" at the interface between TM-III and TM-VI (Fig. 7A). In this position, the side chain of TrpVI:13 hinders the suggested inward movement of the outer segment of TM-VI. On the basis of computational chemistry analysis, Javitch and co-workers (41) hypothesized that a concerted change in the rotamer state of TrpVI:13 together with CysVI:12 would be coupled to straightening of the proline kink in TM-VI. In fact, there is experimental structural evidence from both cryoelectron microscopy and NMR studies supporting the notion that TrpVI:13 does change both position and interaction partners during receptor activation (20,39). We propose that rotation of the side chain of TrpVI:13 away from the interface between TM-III and TM-VI into the "empty" pocket between TM-III and TM-V, where the ␤-ionone ring of retinal is located in rhodopsin, will release a steric constraint on TM-VI by allowing the inward movement of the outer segment of TM-VI, which could be associated also with straightening of the helix (Fig. 7B).
Molecular Modeling and Analysis of a Proposed Active Receptor Conformation-The Monte Carlo simulations using NOE distance constraints from the metal ion sites between TM-III, -VI, and -VII for the extracellular segments as well as distance constraints from the EPR analysis of Hubbell and co-workers (8,14) for the intracellular segments of these helices demonstrated that the proposed vertical seesaw movements with partial straightening of the proline bends of especially TM-VI could occur in an energetically acceptable fashion (Fig. 7B). However, at present, these simulations should be taken only as indications of the possible conformational changes that occur during receptor activation. The available distance constraints are far too few and too imprecise to allow for firm conclusions concerning the precise degree and direction of the movements. Moreover, for simplicity, no movements were allowed during the simulations of the backbones of TM-I-V. In fact, we suspect that the whole seven-helix bundle will change conformation to some degree during activation and that, for example, significant conformational changes will occur also in TM-V. Nevertheless, we propose that the toggle switch activation mechanism as depicted in Fig. 7B, with the indicated conformational changes of especially TM-VI, probably represents the major common conformational switch mechanism for 7TM receptor activation.
Agonist Stabilization of the Active Conformation of the 7TM Receptor Global Toggle Switch-The ␤ 2 -adrenergic receptor used in this study represents one of the best characterized systems in relation to ligand/receptor interactions (44). Recently, several studies have supported the notion that the agonist-binding site in this receptor is not a solid mold into which the ligand fits like a key in a lock. Instead, agonist binding is a sequential process associated with a large degree of induced fit. For example, based on a comprehensive combined analysis of a series of epinephrine analogs and receptor mutants, Costa and co-workers (45) concluded that the free energy couplings in the system have to reflect global conformational changes that alter the mutual distances between several receptor domains and the entire three-dimensional configuration of the ligand-binding subsites. This concept is supported by recent studies from Kobilka and co-workers (46,47) and Javitch and co-workers (48). Interestingly, the schematic models used to illustrate this concept clearly show an inward movement of the TMs in the main ligand-binding pocket around the agonist, in agreement with the proposed toggle switch model (46).
Wess and co-workers (49) recently showed that disulfide bridge formation between a Cys residue introduced at position III:11 and a natural Cys residue located at position VII:09 in the muscarinic M3 receptor is promoted by agonists, but not by antagonists. The interpretation was that TM-III and TM-VII at this level of the receptor will have to move closer to each other during activation. This is again in agreement with our toggle switch model based on metal ion site engineering, including a site between positions III:08 and VII:06 located one helical turn "above" the residues used for disulfide bridge formation in the M3 receptor (21,49).
It can always be argued where and how diffusible ligands bind in a receptor. However, using covalently tethered ligands, Wells and coworkers (50, 51) provided strong evidence for a mode of action of small molecule agonists in the C5a receptor. They introduced a Cys residue at position VI:20, one helical turn above position VI:16 used for metal ion site engineering in this study, as an anchor point for covalent attachment of various small organic compounds through disulfide binding. Notably, they demonstrated that a space-generating substitution of the natural IleIII:08 (Ile 116 ) residue with Ala on the opposing face of TM-III improves the affinity for some compounds and that the smaller Ala residue changes compounds from functioning as antagonists to instead being agonists (50). Interestingly, introduction of a large Trp residue at position III:08 decreases the affinity as well as the agonism of the small organic compounds (50). From a toggle switch activation mechanism point of view, introduction of the smaller side chain at position III:08, at the interface between TM-III and TM-VI, will allow the covalently modified or "ligand-decorated" TM-VI to move inward to "dock" on TM-III in an active conformation, resulting in agonism. In contrast, introduction of a large Trp residue at position III:08 will cause steric hindrance for this inward movement of TM-VI and thereby will lead to antagonism instead of agonism for the same compounds (50).
Constitutive Toggle Switch Activation in 7TM Receptors-The very high (ϳ50%) constitutive activity of the ghrelin receptor is controlled by a cluster of aromatic residues located exactly where the activating metal ion sites were built in this study (52,53). Notably, the constitutive activity can systematically be tuned up and down depending on the size and aromaticity of the side chain at position VI:16 in the presence of a large hydrophobic residue at positions VII:06 and VII:09 (53). Thus, the large aromatic side chains appear to function as "tethered ligands," which, in analogy with the disulfide-tethered ligands of Buck and Wells (50), can hold the receptor in the active conformation and, in this case, lead to the proposed constitutive toggle switch activation.