MODULATION OF THE PROLINE KINK IN TRANSMEMBRANE 6 BY A ROTAMER TOGGLE SWITCH*

In many rhodopsin-like G-protein-coupled receptors, agonist binding to a cluster of aromatic residues in TM6 may promote receptor activation by altering the configuration of the TM6 Pro-kink and by the subsequent movement of the cytoplasmic end of TM6 away from TM3. We hypothesized that the highly conserved Cys, in the vicinity of the conserved Pro, modulates the configuration of the aromatic cluster and the TM6 Prokink through specific interactions in its different rotamer configurations. In the 2 adrenergic receptor, mutation of Cys to Thr, which in an -helix has a different rotamer distribution from Cys and Ser, produced a constitutively active receptor, whereas the Ser mutant was similar to wild-type receptor. Use of the biased Monte Carlo technique of Conformational Memories showed that the rotamer changes among Cys/Ser/Thr, Trp, and Phe are highly correlated, representing a rotamer “toggle switch” that may modulate the TM6 Pro-kink. Differential modulation of the accessibility of Cys and an engineered Cys in wild type and a constitutively active background provides experimental support for the association of this rotamer switch with receptor activation.

G-protein-coupled receptors (GPCRs) 1 are comprised of an extracellular N terminus, seven transmembrane ␣-helical segments (TMs) connected by intracellular and extracellular loops, and a cytoplasmic C terminus. In the ␤ 2 adrenergic receptor (␤ 2 AR) and related rhodopsin-like GPCRs that bind small molecules, the binding site is formed among their seven TMs in a water-accessible binding site crevice (1). The transmembrane domain of these receptors, therefore, transduces the binding of ligands, such as biogenic amines, to the activation of intracellular G-proteins, which in turn mediate downstream signal transduction pathways. Even in peptide and glycoprotein hor-mone receptors, in which the binding site is formed, at least in part, by extracellular loops or by large N-terminal domains, respectively, agonist binding must still be transduced via the transmembrane segments to the intracellular G-protein-interacting regions of these receptors.
Understanding the function of GPCRs at a molecular level requires an understanding of how agonist binding to the receptor is converted into receptor activation (2). Studies based on electron paramagnetic resonance spectroscopy, fluorescence spectroscopy, alterations in cysteine accessibility, and engineering of metal-binding sites have altogether pointed to a key role for conformational changes of TM3 and TM6 in receptor activation (3)(4)(5)(6)(7)(8). It has been suggested that protonation of the Asp in the highly conserved (D/E)RY motif at the cytoplasmic side of TM3 leads to release of constraining intramolecular interactions, thereby resulting in movements of TM3 and TM6 and conversion of the receptor to the active state. This hypothesis has been supported by the observation that charge-neutralizing mutations of the Asp/Glu 3.49 in TM3 lead to increased agonist-independent activation of a number of GPCRs (7, 9 -12). Experiments using cysteine cross-linking and engineered metal ion-binding sites (3,4,13) suggest that residues at the cytoplasmic ends of TM3 and TM6 face each other. We proposed that in the inactive state Arg 3.50 , in addition to interacting with Asp 3.49 , also interacts with the conserved Glu 6.30 at the cytoplasmic end of TM6, and that this interaction contributes to maintaining the receptor in the inactive state by holding together the cytoplasmic ends of TM3 and TM6. Our experimental data (14) together with the high-resolution structure of rhodopsin (15) suggest that ionic interactions between Asp⁄Glu 3.49 , Arg 3.50 , and Glu 6.30 may constitute a common switch governing the activation of many rhodopsin-like receptors. Disruption of the interaction between TM3 and TM6 produces constitutive receptor activation and the extent of constitutive activation is highly correlated with the extent of conformational rearrangement in TM6 (14).
Experimental and computational simulation studies indicate that conformational switches in transmembrane ␣-helices can be generated via Pro-containing motifs that form flexible molecular hinges (16). Although Farrens et al. (3) predicted that a rigid body motion of TM6 relative to TM3 was associated with the activation of rhodopsin, the movements crucial to activation may involve flexibility about the hinge formed by the highly conserved Pro in TM6 (Pro-288 6.50 in ␤ 2 AR) (17). In rhodopsin, and presumably in the ␤ 2 AR, TM6 exists in a highly bent configuration, with its cytoplasmic end near to the cytoplasmic end of TM3 (15).
In the ␤ 2 AR and related aminergic receptors, a cluster of highly conserved aromatic residues surrounding the Pro-kink, including Phe 6.44 , Trp 6.48 , Phe 6.51 , and Phe 6.52 in catecholamine receptors, face the binding-site crevice (reviewed in Ref. 1). Binding of agonist to a residue or residues in this "aromatic cluster" has been hypothesized to induce or stabilize an altered configuration of the side chains within this cluster that promotes receptor activation (17). Such a conformational rearrangement of the aromatic cluster may be associated with an alteration in the configuration of the TM6 Pro-kink and the subsequent movement of the cytoplasmic end of TM6 away from TM3. In the ␤ 2 AR the conserved endogenous Cys-285 6.47 , positioned i-3 from the conserved Pro-288 6.50 , becomes accessible to methanethiosulfonate ethylammonium (MTSEA) in the binding-site crevice in constitutively active mutants (CAMs) (6,7,14). Cys, Ser, and Thr can hydrogen bond (H-bond) to the backbone in an ␣-helix, and this interaction may impact significantly not only the local conformation but may also lead to long range conformational changes through bends and distortions generated in the helix backbone (18 -20). The highly conserved Cys-285 6.47 near the Pro-kink at 6.50 can H-bond to different backbone carbonyls in different rotamer positions, and its rotamer position may modulate the TM6 kink.
We hypothesized that the configuration of the TM6 Pro-kink in a GPCR would be correlated with constitutive activity and that both Cys 6.47 and the aromatic cluster may impact this configuration. We tested these hypotheses in the ␤ 2 AR by assessing the binding and activation properties of Cys 6.47 mutants. The biased Monte Carlo technique of Conformational Memories was used to probe for correlations between the Cys-285 6.47 rotamer position, the configuration of the aromatic cluster in TM6, and the TM6 Pro-kink conformation. We carried out cysteine accessibility experiments to test a prediction that arose from the simulations, and the experimental results supported the existence of a critical rotamer "toggle switch" (17) associated with receptor activation.

Numbering of Residue and Residue Index
Residues are numbered according to their positions in the human ␤ 2 AR sequence. We also index residues relative to the most conserved residue in the TM in which it is located (21). By definition, the most conserved residue is assigned the position index "50," e.g. Pro-288 6.50 and therefore Leu-287 6.49 and Phe-289 6.51 . This indexing simplifies the identification of aligned residues in different GPCRs. Mutants are named as WT residue, residue number, superscript index number, and mutant residue where the residues are given in single letter codes.

Nomenclature of 1 Rotamer
Different nomenclatures have been used to describe the rotamers of side chain torsion angles. When the heavy atom at the ␥ position is at a position opposite to the backbone nitrogen, when viewed from ␤-carbon to ␣-carbon, we define the 1 rotamer as in the trans (t) position; from the same viewing angle, when the heavy atom at the ␥ position is at a position opposite to the backbone carbon, we define the 1 rotamer as in the gaucheϩ (gϩ) position; when the ␥-carbon is opposite to the ␣-hydrogen, we define the 1 rotamer as in the gaucheϪ (gϪ) position.

␤ 2 -Plasmids and Site-directed Mutagenesis
The mutations were generated by the polymerase chain reactionmediated mutagenesis using Pfu polymerase (Stratagene). The polymerase chain reaction-generated DNA fragments containing the mutations were subcloned into the appropriate plasmid. The mutations were identified initially through the presence of diagnostic restriction sites introduced via the mutated oligonucleotides and subsequently verified by DNA sequencing analysis of the polymerase chain reaction-generated segments. For stable expression in HEK 293 cells, this cDNA was subcloned into the bicistronic expression vector pCIN4-SF␤ 2 AR6His (22).

Membrane Preparation and [ 3 H]CGP-12177 Binding
To increase the expression of the constitutively active mutant, a 10 M concentration of the inverse agonist sotalol was added to the growth medium of HEK 293 cells stably expressing WT or mutant ␤ 2 AR for 1-2 days. Membranes were prepared as described previously (14). Aliquots of diluted membrane suspension (200 l) were incubated in binding buffer either with six different concentrations of the antagonist [ 3 H]CGP-12177 between 40 and 1100 pM (in saturation binding experiments) or with increasing concentrations of agonists or antagonists in the presence of 100 pM [ 3 H]CGP-12177 (in competition binding experiments). The total volume was adjusted to 1.0 ml, and the binding experiments and data analysis were performed as described previously (22,23).

cAMP Accumulation Assay
HEK 293 cells stably expressing the WT ␤ 2 AR or the mutants were plated in 96-well cell culture plates (pretreated with poly-L-lysine, 0.1 mg/ml) at a density of 40,000 -60,000 cells/well. After incubation overnight at 37°C in 5% CO 2 , the cells were 95-100% confluent. The medium was removed, and 100 l of 0-60 M N-ethoxycarbonyl-2ethoxy-1,2-dihydroquinoline (EEDQ) in Opti-MEM (Invitrogen) was added to normalize the concentration of receptor (22). The concentration of EEDQ used for a particular experiment was adjusted based on the level of cell surface expression of the particular mutant, as higher expressing receptors such as WT required higher concentrations of EEDQ for appropriate normalization. The level of cell surface expression of WT and the mutants was determined by specific binding of the hydrophilic radioligand [ 3 H]CGP-12177 at a concentration 10-fold higher than its K D (0.9 nM). cAMP accumulation assay was performed as described previously (22).

Reactions of F290 6.52 C Mutants with MTS Reagents
Whole cells from a 35-mm dish were resuspended in 400 l of buffer A. Aliquots (50 l) of cell suspension were incubated with freshly prepared MTSEA (Toronto Research Biochemicals) at the stated concentrations at room temperature for 2 min. Cell suspensions were then diluted 20-fold, and 300-l aliquots were used to assay for [ 3 H]CGP-12177 (1.2 nM) binding in triplicate as described above. The fractional inhibition was calculated as 1-[(specific binding after MTS reagent)/ (specific binding without reagent)].

Rotamer Statistics of Cys/Ser/Thr of Membrane Proteins
The distributions of the 1 rotamer of Cys/Ser/Thr were calculated from all the high resolution (Յ3 Å) membrane protein structures currently available. The structures were downloaded from the Protein Data Bank. The list and definitions of transmembrane ␣-helical segments were updated and modified using the Membrane Protein Topology data base (24). Cys/Ser/Thr in transmembrane ␣-helical segments were only included if at least 5 residues N-terminal to the identified position were within the ␣-helix. Only regions with a B-value Յ40 (25) and structures with a resolution Յ3 Å were included in the analysis. Our rationale was that a resolution Յ3 Å can differentiate whether or not a region is ␣-helical and B-values Յ40 of the Cys/Ser/Thr side chain atoms ensure that coordinate information was reliable.

Conformational Memories
The approach of Conformational Memories, which is a two-phase biased Monte Carlo simulation, was described previously (26). We used this algorithm, which was implemented in CHARMM (27), to probe the conformational space of TM6 of human ␤ 2 AR.
Initial Structures-The template was TM6 (6.40 -6.56) of chain A of the bovine rhodopsin structure (Protein Data Bank number 1HZX). The side chains of the residues were mutated to those of the aligned positions in the human ␤ 2 AR. Four Ala residues in a standard ␣-helical conformation were added at the N and C termini. Thus, there are totally 25 residues in the system. The ␣-helix was capped by acetamide at the N terminus and N-methylamide at the C terminus. 1 of 6.47 (Cys/Ser/Thr) was manually placed in gaucheϩ (gϩ), gaucheϪ (gϪ), or trans (t). 1 of 6.48 and 6.52 were manually placed in gϩ or t, and in gϩ/gϩ, gϩ/t, and t/t combinations (see "Results"). The initial structures were energy minimized by CHARMM (27b1). During the minimization, the backbone atoms were harmonically constrained with a force constant of 10, and the 1 of the 6.47, 6.48, and 6.52 side chains were constrained by well shaped energy barriers around a specific rotamer.
Biased (Constrained) Exploring Phase-In the initial conformational memories, the backbone angles ( and ) of 6.44 -6.52 were distributed with Ϯ50°freedom from the rhodopsin structure values (10% in each of the designated 10°buckets); those of 6.40 -6.43 and 6.53-6.56 were distributed within Ϯ25°freedom (20% at each of designated 10°buckets). The dihedral angles of the alanines in the N and C termini were set to those of a standard ␣-helix ( ϭ Ϫ65° ϭ Ϫ40°). Side chain torsion angles were able to freely rotate except for the 1 of 6.47. The 1 of 6.47 (Cys/Ser/Thr) was constrained to gϩ, gϪ, or t as specified under "Results." Forty runs of Monte Carlo simulated annealing were carried out within the above defined conformational memories, with a distance dependent dielectric with a coefficient of 2. Each run of Monte Carlo simulated annealing consisted of a walk from 2070 to 310 K in 18 temperature intervals with a cooling schedule of T nϩ1 ϭ 0.9 T n , and 25,000 steps per temperature. At each step, 2 dihedral angles were randomly selected to change within the allowed conformational memories, and the trial conformation was accepted or rejected according to the standard Metropolis criteria with a Boltzmann probability function defined at the given temperature.
Biased Sampling Phase-Several hundred structures were collected for each combination of 1 of 6.47, 6.48, and 6.52 (see "Results"), based on the biased conformational memories created by the biased (constrained) exploration phase. Each structure was generated after a run of biased Monte Carlo simulated annealing. Each run in this phase consisted of a walk of 15,000 steps at 310 K. At each step, 2 dihedral angles were randomly selected to change within the output conformational memories from the biased exploring phase. Difference structures were generated by changing the odd random number seed.

Analysis of the Output Simulation Structures
The resulting simulation structures were first classified according to their 1 rotamer positions of 6.48 and 6.52 (see "Results"). For each structure in each group, the corresponding coordinates were extracted to calculate the following parameters: the and of 6.44 -6.52 and their individual averages for each group; the population distribution of 1 of 6.44, 6.47, 6.48, 6.51, and 6.52; the distances between the atoms at the ␥ position of Cys/Ser/Thr (6.47) to the carboxyl oxygen at iϪ3 (6.44) and iϪ4 (6.43) and the average for each group.
The Pro-kink can be described with bend angle, wobble angle, and face shift (28). To assess the configuration across the Pro-kink rather than only the configuration N-terminal to the Pro, we have defined the face shift differently from Visiers et al. (28). We used the angle between the projection of the average of the vectors connecting the ␣-carbons of the (iϪ3) and (iϪ4) amino acids with the origin in the y,z-plane (A Ϫ34 ) and the projection of the average of the vectors connecting the ␣-carbons of the (iϩ3) and (iϩ4) amino acids with the origin in the y,z-plane, instead of the angle between A Ϫ34 and the projection of the vector connecting the proline ␣-carbon with origin in the y,z-plane in Ref. 28. The transformation of the coordinates was similar to Visiers et al. (28). The calculation was implemented in a Perl script that automatically calls CHARMM to calculate the helix axis and extract the results from the output.
The output simulation structures were clustered using Macromodel/ Xcluster. Representative structures and the largest clusters were selected for illustration and these were superimposed on the backbone atoms of 6.49 to 6.60 (extracellular region).

Potential Conformations and Backbone
Interactions of Cys 6.47 -Cysteine residues in an ␣-helical context are restricted to the gaucheϩ (gϩ) and trans (t) rotamer conformations (Table I), because the gaucheϪ (gϪ) configuration induces a steric clash between the sulfur atom and the backbone carbonyl of the iϪ3 position (20). The rotamer distribution of these side chains in an ␣-helical context is quite similar in soluble and membrane proteins (Table I) and is consistent with the theoretical free energy calculated for different rotamers (29).
In the gϩ rotamer, Cys 6.47 can H-bond the backbone carbonyl of the iϪ4 residue, and this is the most common rotamer configuration for a Cys residue in an ␣-helix. In an ␣-helix, the cyclic side chain of proline prevents its H-bonding to the iϪ4 backbone carbonyl and also induces steric clash, thereby bending the helix. The resulting Pro-kink conformation also prevents H-bonding of the iϩ1 backbone amide to the iϪ3 backbone carbonyl (16,30). Cys 6.47 is iϪ3 from Pro 6.50 , and in the t rotamer, the backbone carbonyl of 6.47 can act as an H-bond acceptor for its own side chain. Thus, Cys 6.47 can adopt two different rotamers, and each rotamer can interact with the TM6 backbone differently, with the potential to induce different conformations of the adjacent Pro-kink. Unlike Cys and Ser, which can populate the t rotamer, the ␤-branched Thr does not exist in t, because of steric clash between the side chain methyl and the backbone carbonyls at iϪ3 and iϪ4. We therefore mutated Cys-285 6.47 to Thr to assess the effect of eliminating the t rotamer at this position, and to Ser as a control.
Pharmacological Characterization of Mutant Receptors-The affinity of the antagonist [ 3 H]CGP-12177 was not significantly different in the mutants (C285 6.47 T and C285 6.47 S) and WT (Table II). C285 6.47 T had a 5-fold higher affinity for the agonist isoproterenol than did WT, whereas the affinity of C285 6.47 S was reduced 2-fold (Table II). In intact cells, the maximum level of [ 3 H]CGP-12177 binding to C285 6.47 T was ϳ20% that of WT, whereas that of C285 6.47 S was ϳ50% of WT (data not shown). After normalization of receptor number with the alkylating reagent EEDQ, the mean and standard error values for the EC 50 of the agonist epinephrine in the accumulation of cAMP were 6.0 Ϯ 1.2 (n ϭ 3), 9.4 Ϯ 1.2 (n ϭ 3), and 1.6 Ϯ 0.43 nM (n ϭ 3) for WT, C285 6.47 S, and C285 6.47 T, respectively (Fig.  1). Thus, the potency of epinephrine was increased ϳ4-fold in C285 6.47 T. A higher potency for agonist activation and increased agonist affinity in competition experiments are features commonly associated with CAMs (31).
The Increased Basal cAMP Accumulation of C285 6.47 T Can Be Reduced by Inverse Agonist-We measured cAMP accumulation after attempting to equalize the number of receptors by inactivating WT and C285 6.47 S by pretreatment with EEDQ. Under these conditions, basal cAMP accumulation in C285 6.47 T was about 2-fold higher than in WT and 6-fold higher than in C285 6.47 S ( Fig. 2A). In the presence of an appropriate concentration of forskolin, which directly activates adenylate cyclase, we still observe an effect of receptor activation on cAMP accumulation because of the synergistic effects of forskolin and G␣ s on adenylate cyclase. The increase in cAMP produced by forskolin, however, raises "basal" levels to a level where our assay is more sensitive. Therefore, we also examined the effects of the mutations of 6.47 on cAMP accumulation in the presence of 10 M forskolin. In the presence of forskolin the 2-fold elevation in cAMP in C285 6.47 T relative to WT and 6-fold relative to C285 6.47 S was also observed (Fig. 2B), consistent with the basal determinations reported above. Most importantly, treatment with the inverse agonist timolol (1 M) did not significantly affect cAMP accumulation in WT or C285 6.47 S (Fig. 2). 2 In contrast, the cAMP accumulation in C285 6.47 T was significantly decreased to a level similar to that of WT and C285 6.47 S in both the basal and the forskolin background (Fig. 2).
Correlation between the 1 Rotamers of Cys 6.47 , Trp 6.48 , and Phe 6.52 -The side chain conformations of Trp and Phe residues in an ␣-helical environment are restricted to gϩ or t conformations, given that the gϪ rotamer gives rise to steric clashes with the helix backbone (20). In the simulations, when Trp 6.48 adopted the t rotamer, Phe 6.52 also adopted the t rotamer, whereas when Trp 6.48 adopted the gϩ rotamer, Phe 6.52 was found in both conformations. On the other hand, when Phe 6.52 adopted the gϩ rotamer, Trp 6.48 also adopted the gϩ rotamer, whereas when Phe 6.52 adopted the t rotamer, Trp 6.48 was found in both conformations. This pattern most likely results from a steric clash between Trp 6.48 in t and Phe 6.52 in gϩ (Fig. 3A), which is dependent upon their positioning within the Pro-kink.
In contrast, this combination of rotamer conformations is tolerated without clash in a standard ␣-helix in a number of known proteins (data not shown). The clash between Trp 6.48 t and Phe 6.52 gϩ implies the following correlated motions: if Trp 6.48 t 3 Phe 6.52 t and if Phe 6.52 gϩ 3 Trp 6.48 gϩ. Thus, as proposed by Visiers et al. (17), these two aromatic residues act as a toggle switch.
In our initial conformational searches, we found that particular starting conformation/rotamer positions significantly impacted upon the final distribution of distinct conformations. For example, if we started the conformational search from Trp 6.48 t/Phe 6.52 t, Ͼ90% of the resulting structures contain Trp 6.48 t/Phe 6.52 t (data not shown). This likely resulted from the favorable interaction between the aromatic side chains, the propensity for aromatic residues to be in t in ␣-helices (32), and the need for complex coordinated changes of multiple side chains and/or backbone angles for movement of these side chains. In contrast, if the search started from Trp 6.48 gϩ/Phe 6.52 gϩ, the position of these two side chains in the resulting structures was highly correlated, with either gϩ/gϩ (59 and 79%) or t/t (34 and 9%), depending on the rotamer at 6.47 (gϩ or t, respectively; see below).
In an attempt to minimize the impact of the starting structure, we started simulations with Trp 6.48 in gϩ and Phe 6.52 in t, thereby breaking the correlation between the rotamers at these two positions. If we started from Trp 6.48 gϩ/Phe 6.52 t, a high percentage of the resulting structures leave the starting uncorrelated state and become correlated, either in t/t or gϩ/gϩ 2 Although the lack of effect of timolol on the level of basal activity in WT and C285 6.47 S was not surprising given their lack of substantial constitutive activity (see Ref. 14), it is somewhat surprising that the basal activity in the presence of timolol was higher in WT than in C285 6.47 S or C285 6.47 T. This might conceivably relate in part to the 2-fold higher expression level of WT after EEDQ treatment and/or to differences in cAMP levels in the different stable pools that were independent of ␤ 2 receptor expression.   6.47 S to that of C285 6.47 T, but the number of WT receptors was only reduced to twice that of the other mutants (data not shown). (Table III). We analyzed the relationship in the final structures between the 6.47 rotamer and the different rotamer combinations of 6.48 and 6.52 (Table III). There was a clear correlation between the rotamer position of Cys 6.47 and the resulting rotamer conformation of 6.48. Specifically, when Cys 6.47 was in gϩ, 73% of 6.48 moved from the initial gϩ to t. In contrast, when 6.47 was in t, 87% of 6.48 remained in gϩ (Table III). Similarly, with Thr 6.47 in gϩ or gϪ, 73 and 67% of 6.48, respectively, moved from the initial gϩ to t. With Ser 6.47 in gϩ or gϪ, 61 and 68% of 6.48, respectively, moved from the initial gϩ to t, whereas with Ser 6.47 in t, 70% of 6.48 remained in gϩ. Because Thr does not exist in t, it is likely, therefore, that in the C285 6.47 T mutant the Trp 6.48 is significantly overrepresented in t compared with its rotamer distribution with Cys or Ser at 6.47.
Modulation of the Pro-kink in TM6 by the 1 Rotamer of 6.47, 6.48, and 6.52-The resulting simulation structures can be clustered into two major groups based on the rotamer conformations of 6.47, 6.48, and 6.52, one with Cys 6.47 t/Trp 6.48 gϩ/ Phe 6.52 gϩ and the other with Cys 6.47 gϩ/Trp 6.48 t/Phe 6.52 t. These two clusters have significant differences in the parameters representing the global characteristics of the Pro-kink, which results primarily from the significant alteration in the backbone torsion angles of 6.47 and 6.48 (Table IV). Even a small local change in backbone resulting from the different rotamer combinations could have a large impact on the position of the cytoplasmic end of TM6.
Thr at 6.47 exists predominantly in the gϩ rotamer, and it does not exist in the t rotamer. Because the Thr 6.47 mutant shows an activated phenotype, the 6.47 gϩ conformation seems to favor and/or mimic the active state. In contrast, Ser, which can populate the t rotamer, showed a distribution of states similar to that of WT. Thus, we hypothesize that Cys 6.47 t/ Trp 6.48 gϩ/Phe 6.52 gϩ represents the inactive conformation of the receptor and that Cys 6.47 gϩ/Trp 6.48 t/Phe 6.52 t represents the active state (see "Discussion").
MTSEA Reaction Rate of F290 6.52 C in WT and CAM Background-Cys-285 6.47 in ␤ 2 AR is responsible for the inhibitory effect of MTSEA on ligand binding in several ␤ 2 AR CAMs (6,7,14). We inferred that this increased accessibility of Cys 6.47 to MTSEA resulted from a rotation and/or tilting of TM6 associated with receptor activation. Although a substantial backbone movement of TM6 around the Pro-kink might contribute to such a conformational rearrangement, it is also possible that a local change in the relative orientations of Cys 6.47 and Trp 6.48 may contribute to the difference in the accessibility of Cys 6.47 in the inactive and active states. In the simulations we noted an interaction between the side chain of Cys 6.47 in t and that of Trp 6.48 in gϩ (Fig. 3B). In the putative inactive state, this interaction may impact on the dielectric environment of Cys 6.47 , making it more hydrophobic and likely decreasing the ionization and therefore the reactivity of the thiol. Furthermore, the bulky Trp 6.48 side chain may also sterically block MTSEA from accessing Cys 6.47 . In the putative active state, when Trp 6.48 is in the t position, the Trp side chain would be much farther from Cys 6.47 , and therefore Cys 6.47 would be more reactive with MTSEA because of the removal of the factors described above (Fig. 3B).

TABLE III
The correlation of 1 of Cys 6.47 , Trp 6.48 , and Phe 6.52 The ratios of different combination of rotamers of resulting structures from the conformational search are shown according to the ending rotamer of 6.48 and 6.52. The starting combination of 1 6.48/6.52 was gϩ/t (see text). The 1 of 6.47 was restrained at gϩ, gϪ, or t, as indicated.  . 3. Three-dimensional molecular representations of the rotamer toggle switch. A, a three-dimensional molecular representation of the effect of local rotamer conformation on the Prokink configuration. The representative structures of Cys 6.47 t/Trp 6.48 gϩ/Phe 6.52 gϩ (green) and Cys 6.47 gϩ/Trp 6.48 t/Phe 6.52 t (red) were taken from the Conformational Memories simulations. 200 -300 structures from each cluster were superimposed on the backbone of 6.49 -6.60, and their backbone ribbons are shown. Side chains of 6.47, 6.48, 6.50, and 6.52 are shown in stick representation. B, left panel shows the interaction the side chain of Cys 6.47 in the t rotamer with Trp 6.48 in gϩ, representing the putative inactive state, which lowers the reactivity of Cys 6.47 with MTSEA (see text). Right panel shows the side chain of Cys 6.47 in the gϩ rotamer and Trp 6.48 in t, representing the putative active state, with increased accessibility and reactivity of Cys 6.47 . C, left panel shows the lack of interaction between Trp 6.48 in gϩ with a substituted Cys at 6.52 in the putative inactive state. Right panel shows the interaction between Trp 6.48 in t and a substituted Cys at 6.52 in the putative active state, thereby decreasing the accessibility and reactivity of Cys 6.52 (see text). The Cys 6.47 sulfur is shown in yellow.
In considering ways to test this hypothesis, we noted that a Cys residue at position 6.52 would be able to have a similar stabilizing interaction with Trp 6.48 in the t conformation (Fig.  3C). Thus, the F290 6.52 C mutant would have an activation-dependent MTSEA accessibility phenotype opposite to that of WT Cys 6.47 . That is, in the active state, Trp 6.48 in t would be expected to interact with Cys 6.52 , thereby decreasing its reactivity with MTSEA. Thus, we predicted that F290 6.52 C would react with MTSEA more rapidly in the WT background than in a CAM background. To test this prediction, we made the mutants F290 6.52 C in WT and F290 6.52 C in a well characterized CAM background that contains several mutations at the cytoplasmic end of TM6 (31). The binding affinity of these mutants for [ 3 H]CGP12177 was significantly lower than that of WT receptor (5.8 Ϯ 0.19 nM, n ϭ 4, and 1.4 Ϯ 0.35 nM, n ϭ 3, for F290 6.52 C in WT and F290 6.52 C in CAM, respectively). This is consistent with the known role of Phe 6.52 in ligand binding (reviewed in Ref. 1). Consistent with the prediction stated above, we found that the rate of reaction of MTSEA with F290 6.52 C in WT (261 Ϯ 54 M Ϫ1 s Ϫ1 , n ϭ 5) was 5-fold faster than the rate of reaction with F290 6.52 C in CAM (48 Ϯ 6 M Ϫ1 s Ϫ1 , n ϭ 4) (Fig. 4). DISCUSSION Interactions at the cytoplasmic ends of TM3 and TM6 constrain the relative mobility of these segments in the inactive state of the ␤ 2 AR and related rhodopsin-like receptors (14). Upon release of this "ionic lock" the mobility of TM6 and its position relative to TM3 may be modulated by the configuration of its Pro-kink. If, as we hypothesized, the rotamer position of 6.47 modulates the TM6 Pro-kink, then because the ␤-branched Thr does not populate the t rotamer, C285 6.47 T should favor conformations of the receptor driven by the gϩ rotamer, and destabilize those conformations driven by the t rotamer. Because Ser can exist in the t rotamer with a roughly similar frequency as Cys, C285 6.47 S would be expected to resemble WT receptor in its activation properties.
We found that C285 6.47 T had multiple properties of a CAM: it had a higher basal level of cAMP accumulation, it had a higher basal level of cAMP in the presence of forskolin, it had a lower EC 50 for agonist stimulation of adenylyl cyclase, basal and forskolin-stimulated basal activity was reduced substantially by the inverse agonist timolol, and it had higher affinity for agonist in competition binding studies. Therefore, C285 6.47 T has the classical characteristics of a CAM, in which a greater fraction of receptor is present in the active state (31,33). In contrast, C285 6.47 S had properties similar to that of WT.
Conformational Memories revealed that the rotamer conformations at positions 6.47, 6.48, and 6.52 around Pro 6.50 are correlated. A change of 6.48 from gϩ (its conformation in the rhodopsin inactive state structure), to t, must be accompanied by a corresponding change of 6.52 to t to avoid steric clash (Fig.  3A). On the other hand, a change of Phe 6.52 from t to gϩ must be accompanied by a corresponding change of 6.48 to gϩ to avoid the same steric clash. Thus in the ␤ 2 AR and related rhodopsin-like receptors these two residues may act in concert as part of a rotamer toggle switch that transduces agonist binding into an altered conformation of TM6 (17). The aromatic stacking interactions between Trp 6.48 and Phe 6.52 may stabilize the gϩ/gϩ and t/t rotamer configurations. These favorable interactions between 6.48 and 6.52 are facilitated by the presence of the Pro-kink, which brings these two residues closer to each other, altering the relationship between these two side chains from what would be much less favorable interactions in a standard ␣-helix (data not shown).
We also found that the rotamer conformation of 6.47 was significantly correlated with the rotamer of 6.48, making 6.47 a critical part of the rotamer toggle switch. Specifically, 6.47 in gϩ was associated with 6.48 in t, and 6.47 in t was associated with 6.48 in gϩ (Table III). We propose that the impact of mutation of 6.47 to Thr on activation results from both altered H-bonding as well as modulation by the 6.47 side chain of the rotamer toggle switch. Thus, we propose that Cys 6.47 gϩ/ Trp 6.48 t/Phe 6.52 t represents the active state and that Cys 6.47 t/ Trp 6.48 gϩ/Phe 6.52 gϩ represents the inactive state. With Trp 6.48 in t in the active state, Phe 6.52 must also be in t to avoid steric clash. In contrast, as discussed above, with Trp 6.48 in gϩ in the inactive state, although there is an increased propensity for 6.52 to be in gϩ, 6.52 may also exist in t, suggesting the possibility of a diversity of inactive states, with the rotamer positions of 6.47 and 6.48 being more critical to its phenotype.
This hypothesis is consistent with a number of experimental results in rhodopsin. Movement of the side chain of Trp 6.48 from gϩ to t upon activation is consistent with the inference from UV absorption spectrometry in rhodopsin that the indole side chain of Trp-265 6.48 changes orientation during the inactive to active conformational transition (34) as well as with data suggesting that a single Trp tilts toward the membrane during activation (35). Although 6.52 is an alanine in rhodopsin, in the inactive rhodopsin structure the ␤-ionone ring and carbon chain of retinal interact with 6.48 in gϩ in a configuration very similar to that of the interaction of 6.48 in gϩ with Phe 6.52 in gϩ in the ␤ 2 AR (Fig. 5). Photoisomerization of retinal from cis to trans moves the ␤-ionone ring toward TM4 (36) and thus away from Trp 6.48 , whose enhanced freedom and subsequent rearrangement is involved in the activation mechanism of rhodopsin. The  . 4. Rate of reaction of F 6.52 C in the WT background (open circle) and F 6.52 C in a CAM background (filled circle) with MTSEA. Specific binding was assayed as described under "Experimental Procedures" after a 2-min incubation with the indicated concentration of MTSEA. The mean Ϯ S.E. are shown (n ϭ 4 for WT-F290 6.52 C; n ϭ 5 for CAM-C285 6.47 S/F290 6.52 C). exquisite lack of constitutive activity of rhodopsin may result from the inability of 6.48 to move to the t rotamer when constrained by bound 11-cis-retinal. For GPCRs, such as the ␤ 2 AR, which are activated by diffusible ligands, a ligand that stabilizes Trp 6.48 in the gϩ conformation would, therefore, behave as an inverse agonist. One such mechanism of inverse agonism may involve forcing or stabilizing Phe 6.52 into the gϩ rotamer conformation, which would in turn force Trp 6.48 into the gϩ conformation, and thereby promote the inactive state. Conversely, an agonist that stabilizes Phe 6.52 in the t rotamer conformation would free Trp 6.48 to adopt the t conformation, and thereby promote the active state.
FTIR difference spectroscopy has been used to study the role of cysteine residues in the photoactivation of rhodopsin (37). At least one cysteine sulfhydryl was inferred to be structurally active during rhodopsin photoactivation and to form a stronger hydrogen bond in the activated state. The transmembrane cysteines proposed as candidates are Cys-167 4.56 , Cys-222 5.57 , and Cys-264 6.47 . Movement of Cys 6.47 into the gϩ rotamer and the associated formation of a strong H-bond with the i-4 carbonyl could be responsible for the spectroscopic observations. In the high-resolution structure of rhodopsin, which represents the inactive state of the receptor (38), the 1 of 6.48 is in gϩ and the 1 of Cys-264 6.47 is in a nonstandard rotamer position between gϩ and t. The modification of Cys-264 6.47 by mercury may have affected its rotamer position. It is also possible that the Cys 6.47 side chain may be dynamic, with its coordinates representing the average of its gϩ and t rotameric conformations.
Based on the proximity of the side chain of Cys 6.47 in t and Trp 6.48 in gϩ in the ␤ 2 AR simulations, we propose the existence of an interaction between S␥ of Cys 6.47 and the N⑀ of Trp 6.48 (Fig. 3B). This configuration is consistent with the favored orientation of interaction between Cys and Trp in the Residue Contact Atlas (39). The sulfur-aromatic interaction has been characterized as an intermediate interaction between a purely van der Waals interaction and a H-bond (40). In the inactive state of rhodopsin, the side chain of Trp-265 6.48 forms a H-bond with another residue (34). In the bovine rhodopsin structure (15), which represents the inactive state of receptor, Cys-264 6.47 appears to be the only residue that could form a H-bond-like interaction with the indole side chain of Trp-265 6.48 . Receptor activation alters tertiary interactions and weakens this H-bond (34). We propose that the interaction between Cys 6.47 t and Trp 6.48 gϩ in the inactive state accounts for the H-bond feature of Trp-265 6.48 that is lost during the activation process.
In our simulations of the ␤ 2 AR, Cys 6.47 in t is positioned at an appropriate distance to H-bond to its own backbone car-bonyl, iϪ3 from Pro 6.50 . H-bonding of the Cys 6.47 side chain to its own carbonyl may contribute to the altered angles at 6.47 and 6.48, thereby altering the Pro-kink conformation (Table IV, Fig. 3A) and repositioning the cytoplasmic end of TM6 away from TM3 upon activation. Therefore, Cys 6.47 , through the interaction of its S␥ with the indole N⑀ of Trp 6.48 and the H-bond of its side chain to its own carbonyl, may link the conformation of the aromatic cluster with the conformation of the Pro-kink, thereby playing a critical role in the mechanism of receptor activation.
This local network of interactions may underlie the observed correlations between the rotamer conformations of Cys 6.47 and Trp 6.48 . The conformational changes associated with receptor activation must involve a change in interhelical interactions. Interestingly, in the high-resolution structure of bovine rhodopsin, there is a water molecule in the cavity surrounded by Cys 6.47 , Tyr 6.51 , Pro 6.50 , Pro 7.38 , Phe 7.41 , and Ala 7.42 (38). Okada et al. (38) have hypothesized that electrostatic interactions between TM6 and TM7 through this bound water may stabilize the inactive state, and it is further possible that alterations of these interactions may facilitate the movements associated with activation.
The activation process of rhodopsin has also been found to produce bulk dielectric changes surrounding Trp (34). The proximity of the bulky aromatic side chain of Trp might substantially decrease the accessibility of Cys 6.47 in the inactive state, both by decreasing the local dielectric and decreasing the ionization of the sulfhydryl as well as through direct steric block. This is consistent with the observation from our laboratory that Cys 6.47 is unreactive with MTSEA in the inactive state (6). In the proposed active state, Cys 6.47 and Trp 6.48 are substantially further apart, consistent with the observed reactivity of Cys 6.47 with MTSEA in the CAM background (6,7,14). Based on our simulations and our clustering of the simulated structures into what we inferred to be inactive and active configurations, we hypothesized that in the active state, with 6.48 in the t rotamer, a cysteine substituted for 6.52 might be shielded from reaction with MTSEA, much like Cys 6.47 is in the inactive state. We carried out this experiment and found that F290 6.52 C was ϳ5-fold less reactive in the CAM background than in the WT background (Figs. 3C and 4), consistent with our predictions. This provides experimental support for the proposed inactive and active configurations and for the existence of a rotamer toggle switch. Thus, with Trp 6.48 in the inactive state (gϩ), Cys 6.47 is less accessible and with 6.48 in the active state (t) Cys 6.52 is less accessible. In addition, these data are significant because they demonstrate for the first time that in a CAM particular positions are less reactive with MTSEA, thereby arguing against a simple model in which the CAM is simply a less stable, more dynamic molecule.
It is important to note that our simulations are performed with an isolated TM6 out of the context of the helical bundle and lipid. It is clear that studying a helix in isolation ignores known constraining interactions with other TMs, and is clearly inadequate to predict the global configuration of TM6. For this reason, we think it wise not to overinterpret the exact Pro-kink parameters generated by the simulations, and rather to simply note the differences between the two sets of rotamer combinations. Nonetheless, based on the consistency of the results with our experimental findings and the literature, it appears that the methodology can identify local conformations and correlations based on steric clashes and/or local intrahelical interactions, and mutations like those presented here can be designed to test the validity of these hypotheses. In summary, both our experimental and computational results are consistent with the hypothesis that coordinated rotamer changes of 6.47, 6.48, and 6.52 appear to be associated with receptor activation. This rotamer toggle switch may impact the ␣-helix backbone and may modulate the movement of TM6 about the Pro-kink during receptor activation. Because in rhodopsin-like receptors Cys 6.47 , Trp 6.48 , and Pro 6.50 are highly conserved and 6.52 is conserved as a bulky residue (except in rhodopsin as discussed above), it is likely that the rotamer toggle switch and its modulation of the TM6 Pro-kink are generally involved in GPCR activation.