Control of conformational equilibria in the human B2 bradykinin receptor. Modeling of nonpeptidic ligand action and comparison to the rhodopsin structure.

A prototypic study of the molecular mechanisms of activation or inactivation of peptide hormone G protein-coupled receptors was carried out on the human B2 bradykinin receptor. A detailed pharmacological analysis of receptor mutants possessing either increased constitutive activity or impaired activation or ligand recognition allowed us to propose key residues participating in intramolecular interaction networks stabilizing receptor inactive or active conformations: Asn(113) and Tyr(115) (TM III), Trp(256) and Phe(259) (TM VI), Tyr(295) (TM VII) which are homologous of the rhodopsin residues Gly(120), Glu(122), Trp(265), Tyr(268), and Lys(296), respectively. An essential experimental finding was the spatial proximity between Asn(113), which is the cornerstone of inactive conformations, and Trp(256) which plays a subtle role in controlling the balance between active and inactive conformations. Molecular modeling and mutagenesis data showed that Trp(256) and Tyr(295) constitute, together with Gln(288), receptor contact points with original nonpeptidic ligands. It provided an explanation for the ligand inverse agonist behavior on the WT receptor, with underlying restricted motions of TMs III, VI, and VII, and its agonist behavior on the Ala(113) and Phe(256) constitutively activated mutants. These data on the B2 receptor emphasize that conformational equilibria are controlled in a coordinated fashion by key residues which are located at strategic positions for several G protein-coupled receptors. They are discussed in comparison with the recently determined rhodopsin crystallographic structure.

The understanding of the mechanisms of G-protein coupled receptor (GPCR) 1 activation has been improved by numerous experimental supports to the existence of multiple conformations, active or inactive (1). Mutation-induced constitutive ac-tivation phenomena have been widely exploited to dissect the intramolecular interaction networks which stabilize inactive receptor conformations (1), including random mutagenesis approaches to get deeper insight into the obvious complexity of these networks (2,3).
In a preceding work (4) we evidenced constitutive activation of the human B 2 bradykinin receptor induced by mutations of Asn 113 in TM III and Trp 256 in TM VI. The role of this tryptophan residue, which is fairly conserved in the GPCR family, was suggested by our previous modeling studies of the AT 1 receptor (5) and the constitutive activation of the Ala 113 mutant is reminiscent of previous observations for the AT 1 receptor mutated at the homologous Asn (6). The exceptionally high constitutive activation of the Ala 113 B 2 receptor demonstrated the crucial role of Asn 113 in stabilizing an inactive B 2 receptor conformation and the extreme conformational flexibility of this receptor. This flexibility was further supported by the striking changes in the pharmacological properties of peptidic or nonpeptidic ligands induced by Asn 113 or Trp 256 mutations (4). These data make the B 2 receptor an interesting model to study conformational changes associated to activation. In this respect the original nonpeptidic ligands designed by Fournier Research Laboratories, which constitute potential anti-inflammatory drugs (7,8), turned out to be interesting tools to modulate conformational equilibria. The purpose of the present work was to take advantage of constitutive activation phenomena to delineate some features of inactive receptor conformations and to propose a model of nonpeptide antagonist interaction with the B 2 receptor. These models are discussed with reference to those built for other GPCR (9,10), the recent crystallographic data for rhodopsin (11) and current knowledge about their activation processes. This study emphasizes the roles of some key residues that are located at strategic positions for several GPCR, including rhodopsin, and simultaneously control ligand recognition and conformational equilibria (Fig. 1).
Site-directed Mutagenesis and Receptor Expression-The human B 2 receptor sequence has been determined by Hess et al. (12). The WT and mutated receptors were systematically tagged through the addition of a 10-amino acid epitope from the c-myc oncogene at the N termini of receptors truncated at the Asn 3 residue. The cDNA sequences included a Kozak sequence. The various mutations were carried out as described previously (4). Receptors were transiently expressed in COS-7 cells using the electroporation transfection method (4). Pharmacological characterizations were performed on cells cultured for 2 days at 37°C in Dulbecco's modified Eagle's medium, 4.5 g/liter glucose, 10% fetal calf serum, 100 units/ml penicillin, 100 units/ml streptomycin.
Binding Assays-The binding of [ 3 H]BK or 125 I-HPP-HOE 140 to crude membranes from transfected COS-7 cells was carried out as previously described (4).
The binding of [ 3 H]BK to intact cells (24 or 48 well dishes), including evaluation of expression levels, and IP production (12-well dishes) were routinely carried out as described in Ref. 4. In experiments devoted to the effects of Zn 2ϩ ions, the media were modified to ensure their perfect solubility: [ 3 H]BK binding was carried out in 25 mM HEPES, 140 mM NaCl, 140 g/ml bacitracin, 10 M captopril, pH 7.4; IP accumulation was measured in 20 mM HEPES, 116 mM NaCl, 11 mM glucose, 2.5 mM CaCl 2 , 4.7 mM KCl, 140 g/ml bacitracin, 10 M captopril, pH 7.4. Molecular Modeling-Models were built on an "Octane" Silicon Graphics computer, using software from Molecular Simulations Inc.
(Insight, Discover, Homology, Modeler). Molecular dynamics simulations carried out in the absence of restraints of some transmembrane helices were performed using the computing facilities of CINES. The B 2 receptor "experimental model" was built using the Modeler software, using as template a first AT 1 receptor model (13) in which the position of TM VI has been refined (about 30°rotation) so as to fit pharmacological data relative to losartan (5). As compared with the AT 1 template a slight rotation of helix VII was introduced to take into account binding properties on nonpeptidic ligands to receptors mutated at some positions of these helices (Thr 287 and Gln 288 ).
A second B 2 receptor model was built using the rhodopsin crystallographic structure as template ("rhodopsin-like" model). For both models the positions of the transmembrane helix peptidic backbones were strictly ascribed to those of their respective templates with the exception of transmembrane domains displaying differences in proline residues: in the "rhodopsin-like model" the 35-48 (N-terminal part of TM I) and 285-290 (N-terminal part of TM VII) sequences were ascribed to helix structures; in the experimental model the 278 -288 sequence was forced to an helix structure.
Molecular dynamics was initially applied to the conformational analysis of LF 16-0687. The analysis of the corresponding trajectory revealed, as expected, a great flexibility of the molecule. The lowest energy conformations were compact, with an interaction between the quinoline and benzamidine moieties. However, it appeared impossible to perform any preliminary positioning of such conformations into our receptor model. As a consequence we selected extended conformations and performed their manual docking so as to fit the experimental data.
The quinoline of LF 16-0687 or LF 18-1300 was positioned so as to interact simultaneously with Tyr 295 and Trp 256 . Then the rest of the molecule was orientated toward the extracellular side of the TMs, applying appropriate torsion angles to the ligand bonds, with the exception of the dichlorophenyl-SO 2 -prolyl portion which was considered as playing an essential orientation role; therefore the geometry of this portion was kept close to that delineated by the ligand conformational analysis. No remarkable interaction could be found for the dichlorophenyl group of the ligand which was positioned between TMs II and VII inside the intrahelix space. Thr 287 and Gln 288 thus appeared as possible candidates for an interaction with the carbonyl function of the prolyl ligand moiety. As the mutation of Thr 287 to Ala or Leu, its B 1 counterpart, had no incidence on affinity, we favored the existence of an hydrogen bond interaction between Gln 288 and the carbonyl group of the prolyl moiety and we slightly rotated helix VII, by ϳ20°, so as to fulfil this interaction preference while preserving the "sandwich" interaction of the quinoline with the Trp 256 and Tyr 295 residues.

Strategy: Rationale of AT 1 and B 2 Receptor Comparison
Our previous modeling studies of the AT 1 receptor (5,13) and experimental data revealed the essential roles of the conserved Asp in TM II (Asp 74 ), Asn 111 (TM III), the fairly conserved Trp in TM VI (Trp 253 ), His 256 located one helix turn above Trp 253 and Tyr 292 (TM VII) which occupies a position homologous to that of the rhodopsin Lys 296 , the site of retinal covalent binding. An inactive receptor conformation would be stabilized by a  Sequence comparisons of TMs II, III, VI, and VII of the B 2 receptor with those of some other GPCR reveal the residues, located at privileged positions, which have been shown to be involved in ligand binding or activation properties, or both. The residues located in boxes have been experimentally shown to be important for the function of a given receptor. The roles of the mentioned B 2 receptor residues have been assessed in previous works (4,19) or constitute the matter of the present paper.
FIG. 2. Structure of nonpeptidic ligands specific of the human B 2 bradykinin receptor. The compounds LF 16-0335 and LF 16-0687 are quite similar: they possess the same substituted quinoline at one extremity and a benzamidine moiety at the other, the latter compound displaying a significantly higher affinity for the receptor. LF 18-1300 differs from LF 16-0687 by the imidazole substitution on the quinoline moiety, resulting in improved affinity. LF 18-1833 possesses the same imidazole-substituted quinoline, while being shortened at its other extremity which no longer possesses the benzamidine function. double interaction of Asn 111 (TM III) with Tyr 292 (TM VII) and Trp 253 (TM VI), homologous of the rhodopsin Trp 265 ; the Asn 111 -Trp 253 proximity was postulated upon model refinement (helix VI rotation) (5) to make it consistent with the pharmacological properties of the nonpeptidic ligand losartan (5,14). We postulated that the interaction of the Tyr 4 residue of AII with Asn 111 would destabilize the above mentioned network and induce rearrangements resulting in a Asp 74 -Tyr 292 interaction (13) and an intrahelical interaction between Trp 253 and His 256 . The experimental supports can be summarized as follows: receptor inactivation upon Y292F (15), W253F (5), and H256A mutations (5,16); strong constitutive activation upon Asn 111 mutations (6,17); detailed analysis of the induction of intermediary or activated conformations (17,18). Interestingly these residues are conserved in the B 2 receptor, with the exception of His 256 which is replaced by Phe 259 . That this latter residue is a BK interaction site (proposed for the rat receptor (19), and extended to the human receptor in the present work) and previous results demonstrating a role for Asn 113 and Trp 256 (4) reinforces the interest of a mechanistic comparison of the AT 1 and B 2 receptors which display a sequence identity similar to that of the AT 1 /AT 2 or B 1 /B 2 pairs (about 30%).

Role of Trp 256 in the Activation Process
Our previous study (4) showed that Trp 256 mutation to Phe or Gln induced marked increases in the human B 2 receptor constitutive activity, together with changes in the pharmacological properties of the peptidic compound HOE 140 and the nonpeptidic ligand LF 16-0335. We further investigated the crucial role of Trp 256 by evaluating the activation properties of the W256F, W256Q, and W256A mutant receptors, as well as the double mutants displaying the additional Asn 113 to Ala mutation. Fig. 3 shows that, at similar expression levels, the W256F and W256Q mutants were significantly constitutively activated (CAM receptors) as compared with the wild-type receptor, the activation factor varying from 4 to 10 according to experiments. Combined with the previously described exceptionally high constitutive activation displayed by the Ala 113 mutant (Fig. 3), we postulated that inactive receptor conformation(s) might be stabilized by an Asn 113 -Trp 256 interaction.
The first interesting finding was the lack of constitutive activation of the W256A mutant, with preservation of the BK activation properties (Fig. 3). This result has two implications: taking into account the proximity of Asn 113 and Trp 256 (validated in experiments described in the next paragraph) the hypothesis that Asn 113 and Trp 256 might stabilize an inactive conformation through interactions with separate unidentified partners appeared unlikely as Trp to Ala mutation would be expected to be activating in such a situation; the role of Trp 256 cannot be restricted to an interaction with Asn 113 stabilizing an inactive conformation. Trp 256 might also contribute to the stabilization or the generation of an active conformation in which the Asn 113 -Trp 256 interaction would be disrupted.
One can reasonably postulate that Asn 113 and Trp 256 belong to a network of concerted interactions comprising one or several other residues and subtly tuned by the residue at position 256. Constitutive activation observed upon Trp mutation to Phe or Gln would then result from the perturbation of a balance between these interactions.

Biochemical Verification of the Asn 113 -Trp 256 Proximity
Asn 113 and Trp 256 were substituted, either separately or simultaneously, by histidine, and the binding of Zn 2ϩ ions to the single or double His mutants was evaluated. Results in Fig.  4A show that His mutations by themselves induce no or little perturbation of BK or HOE 140 recognition. Competition binding of Zn 2ϩ ions using [ 3 H]BK as tracer ligand revealed a striking leftwards shift of the dose-response curve for the double N113H, W256H mutant as compared with the WT receptor (Fig. 4); it indicated an increased affinity for Zn 2ϩ ions, interpreted as a spatial proximity between the two histidines. This effect was not observed for the W256H single mutant. It was much less pronounced for the N113H single mutant; the slight shift observed in this case probably resulted from the ability of Zn 2ϩ ions, which can display different modes of interactions with amino acid side chains (20 -22), to coordinate N113H to another residue. These results were confirmed by functional assays which first indicated that His replacement of Asn 113 , or Trp 256 or both induced neither constitutive activation, nor modification of BK-stimulated IP production. Zn 2ϩ ions, which had no effect on basal activities, were able to counteract BK-induced IP production in COS-7 expressing the N113H, W256H mutant ( Fig. 4B), with Zn 2ϩ inactivation characteristics similar to those observed for BK binding.

Concerted Roles of Trp 256 and Asn 113 in the Human B 2 Receptor Activation
The pivotal role of Trp 256 was confirmed by the pharmacological properties of B 2 receptors mutated at this position. HOE 140, which behaved as an inverse agonist of the WT receptor in experimental situations where the basal IP production is high enough (4), became a fairly potent agonist of the three W256F, W256Q, and W256A mutant receptors, with a markedly decreased maximal stimulation for the latter mutant (Fig. 5A). The binding of a radioiodinated derivative of HOE 140 ( 125 I-HPP-HOE 140) revealed moderate decreases in affinity for the mutant receptors (Table I). The behavior of the nonpeptidic compound LF 16-0335 was strikingly different: it only activated the W256F mutant (Fig. 5B). These properties of LF 16-0335 were reproduced for LF 16-0687, LF 18-1300, and LF 18-1833 which were agonists of the W256F mutant, and not of the W256Q and W256A mutants (not shown). The K i values relative to LF 16-0335 and LF 16-0687, determined in competition binding assays using 125 I-HPP-HOE 140 as tracer ligand, were slightly increased for the three mutants, the most significant increase being observed for the W256A receptor (Table I). Similar data were obtained when [ 3 H]BK was used as tracer ligand (not shown). These results suggest that Trp 256 might directly interact with the nonpeptidic ligand through an aromatic-aromatic interaction (which does not exclude additional interactions such as hydrogen bonding or amino-aromatic interactions) and that preservation of this interaction in the W256F CAM receptor participates to the ligand agonist behavior (see paragraph devoted to modeling).
The essential role of Trp 256 was further supported by the properties of receptors doubly mutated at positions 113 and 256 which were still constitutively activated: as previously described (4), LF 16-0335 displayed agonist properties on the N113A receptor which possesses an extremely high constitutive activity. This agonist property was only maintained for the N113A,W256F double mutant (not shown). Therefore the residue at position 256 modulates the activation properties of a receptor which is strongly destabilized by Asn 113 mutation to Ala. Taken together the data indicate that Trp 256 controls in a subtle way the balance between the stabilization of receptor FIG. 4. Histidine engineering and Zn 2؉ effects on BK binding and BKinduced IP production. Asn 113 or Trp 256 or both residues were mutated to His and the WT and mutant B 2 receptors were transiently expressed in COS-7 cells as described under "Experimental Procedures." Zn 2ϩ ion affinities for the various receptors were indirectly evaluated by their ability to inhibit [ 3 H]BK binding (1 nM) to intact cells (A) or 0.5 nM BK-induced IP production (B).

FIG. 5. Incidence of the residue at position 256 on the properties of HOE 140 (A) and LF 16-0335 (B).
The WT or Phe 256 , Gln 256 , or Ala 256 mutant B 2 receptors were transiently expressed in COS-7 cells at similar levels (6 ϫ 10 5 -10 6 sites/cell) and the ability of the peptidic and nonpeptidic ligands to modulate IP production evaluated as indicated under "Experimental Procedures." The presented typical experiment is representative of three separate experiments. In these specific experiments designed for mutant receptors and ligand comparisons, the basal WT activity was not high enough to evidence the inverse agonist activity of HOE 140 and LF 16-0335, as previously observed (4 inactive conformations and the stabilization of intermediary or active conformations and that Asn 113 and (or) Trp 256 possess other interacting partners at some steps of the activation process.

Characterization of New Mutations Modulating Activation
Properties: Roles of Tyr 115 and Tyr 295 The Pro 258 or Phe 252 mutations to Ala did not induce any constitutive activation, so that the data obtained upon muta-tion of these residues in TM VI of other GPCR (3,20) cannot be generalized. The mutation of Gln 260 (TM VI) to His (B 1 homologous residue) did not change the basal activity; its mutation to Ala induced either a weak (2-fold) or no constitutive activation according to experiments (Fig. 3).
As Ser 111 was postulated to be involved in peptidic ligand recognition specificity (21), it was mutated to either Ala or Lys, its B 1 counterpart. These mutations induced moderate but significant losses of BK affinity. However, none of these muta-

H]BK binding was carried out on intact cells; intact cells
were not suitable for 125 I-HPP-HOE 140 binding because of passive ligand penetration, which led to overestimated values. The expression levels for a given plasmid amount (25-50 ng) was similar for the WT and all mutant receptors (1-3 ϫ 10 5 sites/cell), with the exception of N198A and P258A, which displayed 10-fold reduced expression, and N297A, which was not expressed at the plasma membrane but was intracellularly detectable. The K i values relative to the binding of LF 16 -0335 or LF 16 -0687 were evaluated on membrane preparations, in competition binding assays using 125 I-HPP-HOE 140 as tracer ligand. Each value is the mean Ϯ S.D. of at least three separate determinations. tions induced any significant modification of basal IP production activities.
The most clearcut findings refer to the role of Tyr 115 , located two residues below Asn 113 and homologous of the rhodopsin Glu 122 . Important constitutive activations were reproducibly elicited by Tyr 115 mutation: while its mutation to Phe (the residue located at the homologous position in the B 1 receptor) had no effect, its mutation to Ala reproducibly induced a 5-fold increase in the basal IP production (Fig. 3). The pharmacological properties of the nonpeptidic compound LF 16-0687 for the Y115A receptor were unchanged as it behaved as in inverse agonist, at variance with the CAM N113A and W256F receptors. Examination of the various models suggested some residues as possible partners of Tyr 115 ; a role for some of them was experimentally ruled out: Ser 162 , Asn 202 , and Phe 206 located in the vicinity of Gly 205 , homologous of rhodopsin His 211 which was shown to interact with Glu 122 (11).
The mutation of the conserved Asn 48 in TM I did not induce constitutive activation. These results differ from those reported for the ␣ 1B and TRH receptors (22,23) and does not allow to draw any conclusion about the connections between TMs I, II, and VII (23)(24)(25).
We checked for possible interactions between Trp 256 and residues of TM V which are expected to face TM VI. Phe 206 mutation to Ala neither induced constitutive activation, nor significant modification of BK activation properties, leading to the conclusion that the proposed interaction between the TM VI Trp and TM V Phe in the TRH receptor (26) cannot be extrapolated. No significant perturbation of activation properties could be detected for the D202A mutant.
Tyr 295 mutation to Phe (4) or to Ala in TM VII of the B 2 bradykinin receptor changed neither its basal IP production activity nor its BK activation properties (Fig. 3 for basal activities, data not shown for BK activation properties). These results show that Tyr 295 does not play, by itself, a role similar to that of the homologous Tyr 292 in the AT 1 receptor. Nevertheless, as this residue, homologous of the rhodopsin Lys 296 , was demonstrated to be an essential interaction site with nonpeptidic compounds, together with Gln 288 (see modeling section), we constructed some double mutants to check a possible role of Tyr 295 in association with other residues. Interestingly we found that while the Y295A and Q228A mutant displayed moderately decreased recognition of BK and HOE140 (Table I), the Q228A,Y295A double mutant no longer recognized BK, and displayed biphasic binding curves for 125 I-HPP-HOE 140 with the appearance of low affinity binding sites. Moreover the simultaneous Ala mutation of Tyr 295 , Trp 256 , and Phe 259 drastically affected the maximal BK stimulated IP production, which was not observed for the W256A,F259A double mutant despite decreased BK potency (not shown). It indicated that Tyr 295 plays some subtle role in receptor activation.
Ala mutations of other TM VII residues, Thr 287 , Ser 291 , Ser 296 , and Ser 298 , did not significantly alter the activation properties (Fig. 3), while the properties of the Asn 297 mutant could not be adequately evaluated because of its poor expression. In this respect the behavior of helix VII appeared significantly different in AT 1 and B 2 receptors, as, in addition to Tyr 292 mutation, mutations of Asn 294 and Asn 295 caused inhibition of hormone-induced AT 1 receptor activation (27).
On the Role of Phe 259 in the B 2 Receptor Activation Mechanism-Phe 259 , located about one helix turn above Trp 256 , is supposed to directly interact with BK, together with Thr 263 ; indeed we found dramatic losses in BK affinity upon Phe 259 or (and) Thr 263 mutations to Ala in the human B 2 receptor (Table  I), consistent with previous findings for the rat receptor (19). Therefore Phe 259 might be a "switch" residue playing a key role in agonist-induced destabilization of inactive conformations. Previous data on the AT 1 receptor have suggested that the transition from inactive to active conformations might involve an interaction between Trp 253 and His 256 (5), the residues homologous of the B 2 receptor Trp 256 and Phe 259 respectively. It is noteworthy that His 256 and Asn 111 were shown to interact with the Phe 8 (16,18) and Tyr 4 (6, 13) residues of angiotensin II, respectively. There exists no evidence for or against the possible interaction of Asn 113 with BK in the B 2 receptor; the expected increase of agonist affinity for CAM receptors, recently observed for the B 1 receptor mutated at its conserved Asn residue (28), was found neither for the Ala 113 B 2 receptor (4) nor for the N111A AT 1 receptor (6). It might result from compensatory effects through the loss of BK-Asn 113 interaction. Answering this question would imply an exhaustive structurefunction analysis, as performed for the AT 1 receptor using both receptor mutants and hormone analogs (17,18). An alternative explanation is that Asn 113 mutation mimicks a receptor perturbation induced by BK through its interaction with Phe 259 . In this respect it is tempting to postulate that there is some link between the roles of Phe 259 and Asn 113 ; indeed the lack of increase in BK affinity upon Asn 113 to Ala mutation might reflect some "redundancy" between the molecular events (and associated free energy changes) triggered by BK-Phe 259 interaction and Asn 113 mutation-induced facilitation of transition to active conformations. It was corroborated by the fact that the N113A,F259A (and also N113A,T263A) double mutant displayed striking increases in BK affinity or potency for IP production as compared with the F259A single mutant (Table II), indicating that the Asn 113 mutation is also able to favor some other molecular events which are mediated by the Phe 259 -BK (and Thr 263 -BK) interaction.
Other experimental findings are in favor of a concerted role of Asn 113 and Phe 259 : the behavior of B 2 receptors possessing either N113A and F259A mutations or N113A and T263A mutations deserves a comment: the properties of these double mutants revealed a discrepancy between the BK K d values determined by direct [ 3 H]BK binding (not possible in the single F259A and T263A mutants) and the K i values evaluated in competition binding assays using 125 I-HPP-HOE 140 as tracer ligand (Table II). It might result from the existence of various conformational states, displaying either high affinity for HOE 140 and low affinity for BK (preferentially binding the tracer ligand) or a high affinity for BK (preferentially binding BK at low concentrations), together with a very slow interconversion between these two states, as already described for other receptors (29). In the B 2 receptor context, the observed discrepancies confirm that the mutations of residues 113 and 259 profoundly affect the pathways of conformational changes; the additional mutation of Phe 259 (and therefore suppression of BK-Phe 259 interaction) in the CAM N113A receptor might in some way limit the reversibility of specific steps in these pathways.
A concerted role for Phe 259 and Trp 256 could also be inferred from the properties of receptors mutated at positions 256 and 259. The CAM W256F and W256Q receptors displayed no significant increases in BK affinity. Phe 259 was also mutated to His, its AT 1 counterpart. While the F259H receptor displayed properties similar to those of the F259A mutant, i.e. strongly decreased affinity for BK (Table II) and no constitutive activation, the F259H, W256F and F259H, W256Q receptors were constitutively activated, similarly to the single Phe 256 and Gln 256 mutants (not shown) and the EC 50 for BK stimulation were dramatically decreased as compared with the W256H receptor (these effects were much less pronounced for the F259H, W256A receptor, which was not constitutively activated) ( Table II).
As a consequence our hypothesis is that the balance between active and inactive conformations might be controlled by a network of concerted interactions involving Asn 113 , Trp 256 , and Phe 259 (Fig. 8) but obviously not restricted to these residues. Besides its interaction with Trp 256 , Asn 113 might interact with another residue located in another TM than TM VI. By analogy to the AT 1 receptor, Tyr 295 appeared as a plausible candidate, thus pointing to coordinated roles of TMs III, VI, and VII. As detailed in Fig. 8, all predictions about receptor interaction networks should take into account the possible existence of a statistical distribution of multiple conformational states each displaying only part of the postulated interactions.

Docking of Nonpeptidic Ligands in a B 2 Receptor Model
Nonpeptidic ligands designed by Fournier Laboratories, inverse agonists of the WT receptor, are expected to stabilize or induce inactive receptor conformations. They might induce inactive conformations characterized by a loss of Asn 113 -Trp 256 interaction. However, the fact that they become agonists of the Ala 113 CAM receptor (4) which is no longer stabilized by this interaction, together with the lack of experimental evidence for their interaction with Asn 113 , suggest that the Asn 113 -Trp 256 interaction is essentially preserved in the nonpeptide-WT receptor complex. As a consequence, nonpeptide ligands were docked into a receptor model built by homology to the AT 1 receptor (13), characterized by the Asn 113 -Trp 256 and Asn 113 -Tyr 295 proximities. Interestingly the Asn 111 -Trp 253 interaction resulted from a refinement of TM VI position in initial models (5), so as to allow the double interaction of the tetrazole moiety of the nonpeptide losartan with Lys 199 and His 256 (5) (Fig. 6). In parallel, we also performed a docking of nonpeptidic ligands into a B 2 receptor model homology-derived from the rhodopsin crystallographic structure (11). Three main candidates to nonpeptidic ligand recognition emerged from the results collected in Table I (Table  I). Tyr 295 appeared to be essential for nonpeptide affinity; its mutation to Phe and Ala induced a gradual increase in K i values for LF 16-0687 (8 -10-fold and 130 -140-fold increases, respectively). Interestingly the loss of affinity upon Tyr to Phe mutation was significantly attenuated, irrespective of the tracer ligand used, for the compound LF 18-1300 which possesses an imidazole moiety at position 4 of the quinoline (Table  III). The interpretation is that Tyr 295 interacts with the quinoline and that the loss of hydroxyl group upon its mutation to Phe is compensated by the presence of the imidazole in LF 18-1300. The similar affinity losses upon Tyr mutation for LF 18-1300 and LF 18-1833, a related compound lacking the benzamidine moiety at its extremity, precludes a Tyr interaction with the benzamidine, which significantly contributes to the affinity of LF 16-0335, LF 16-0687, and LF 18-1300. The Tyr- quinoline interaction is consistent with the essential role of quinoline emerging from the chemical design of LF 16-0687 and the dramatic loss of affinity induced by Ala mutation of Tyr 295 .
Significant losses in nonpeptide affinity were observed for the W256A mutant (Fig. 5, Table I, Table II); their relatively moderate intensities can be interpreted as resulting from compensatory effects involving the reinforcement of Tyr 295 -quinoline interaction in the mutant receptor. The decreases in nonpeptide affinity were lower for the W256F and W256Q mutants which were constitutively activated and therefore displayed modified conformational equilibria with increased representations of active conformations. The changes in the pharmacological properties of the nonpeptide for the various mutants support the hypothesis of nonpeptide-Trp 256 interaction: as LF 16-0335, LF 16-0687, LF 18-1300, and LF 18-1833 became agonists of the W256F receptor, but not of the W256Q receptor, we postulated that an aromatic-aromatic interaction between the W256F receptor and the ligand was responsible for the agonist action of this latter; as detailed in the recapitulative scheme (Fig. 8) the ligand conformation interacting with the W256F CAM receptor should be different from that interacting with the inactive conformation of the WT receptor.
As a consequence we postulated the existence of a sandwich packing comprising Tyr 295 -quinoline-Trp 256 . The drastic loss of nonpeptide affinity for the W256A,Y295A double mutant (Table III) supports this hypothesis. The alternative hypothesis of Tyr 295 and Trp 256 separate interactions with the quinoline and dichlorophenyl ligand groups, respectively, appeared impossible to be fulfilled, because of the steric hindrance of the dichlorophenyl group and the importance of its link to the SO 2 -prolyl portion in the ligand conformation. The hypothesis of Trp 256 interaction with the benzamidine moiety of LF 16-0687 and LF 18-1300 was also excluded because of the similar behavior of LF 18-1833.
All nonpeptidic ligands displayed ϳ10-fold decreased affinities for the Q288A mutant, as compared with the WT receptor (Table III). Here again the behavior of the "shortened" compound LF 18-1833 could not be distinguished from that of compounds possessing the benzamidine. As expected, the simultaneous mutation of Tyr 295 and Gln 288 elicited drastic losses in nonpeptide affinity (K i value difficult to evaluate because of multiple sites for the tracer ligand 125 I-HPP-HOE 140).
These experimental results constituted the support for the docking of nonpeptide ligands into a model of receptor inactive conformation, derived from previous AT 1 and B 2 receptor studies (4 -6, 15). The docking strategy (see details under "Experimental Procedures") positioned the benzamidine moiety of LF 16-0687 at the extracellular space boundary of the transmembrane helix bundle. Further experimentation would be required to detect receptor residues responsible for its interaction with benzamidine which is not an absolutely required pharmacophore but nevertheless improves ligand affinity. The possible candidates to an interaction with the benzamidine appeared to be Asp 266 , Thr 267 , Asp 279 , Glu 280 , and Asp 284 ; this latter residue was excluded on the basis of D284A mutant properties.
The model of the minimized LF 16-0687-receptor complex, termed as experimental model, is represented in Fig. 7A. This minimization did not significantly change the position of TM VII which was kept free of constraints together with the adjacent loops. An opened question is the participation of Asn 113 to ligand interaction, through hydrogen bonding to the oxygen atom directly linked to the quinoline. The lack of incidence of Asn 113 to Ala mutation on LF 16-0687 affinity is not necessarily inconsistent with this possibility as the mutation drastically shifts conformational equilibria to active states, conferring agonist properties to the nonpeptidic ligand. As above indicated, optimization of the geometry of the microdomain constituted by Asn 113 , Trp 256 , and Tyr 295 and the ligand quinoline, which might not be unique (and its relationship with the Asn 113 -Trp 256 interaction) is not easily feasible through use of semiempirical quantum mechanics calculations, taking into account the difficulty to assess the relevance of expectedly weak differences in energy levels of conformations corresponding to different possibilities of packing the involved chemical structures.
Preliminary positioning of LF 18-1300 into B 2 receptor models made no interaction of additional receptor residues with its imidazole moiety, suggesting that this latter probably induces a reinforcement of the quinoline interaction with its partners Tyr 295 and (or) Trp 256 . Here again the success of theoretical calculations is not guaranteed. As previously mentioned it is noteworthy that a model of losartan-AT 1 receptor interaction (Fig. 6), consistent with abundant experimental structure-function and activation data, was based on similar hypotheses of inactive receptor conformation (5,13), including the essential stabilizing Asn 111 -Trp 253 interaction.
A B 2 receptor model, homology-derived from the crystallographic data of rhodopsin (termed as rhodopsin-like model), was also built (Fig. 7B). The flexibility of the ligand LF 16-0687 made possible a docking into this model which satisfied the same interactions as in the preceding model, including a sandwich position of the ligand quinoline with Tyr 295 and Trp 256 which is postulated to be essential for the stabilization of inactive receptor conformations. In this model, the ligand extremity bearing the benzamidine moiety adopts a more bent shape, so as to keep conserved possible constraints of extracellular loops which, in the rhodopsin structure, function as a lid preventing retinal from going out of the binding pocket. In the absence of any data about residues interacting with the benzamidine, it is difficult to support any hypothesis about the position of this ligand extremity as well as the role of the B 2 receptor extracellular loops in the delimitation of the binding site. The comparison of the experimental and rhodopsin-like models was focused on residues which display interesting activation and ligand recognition properties, more particularly on Asn 113 , Trp 256 , and Tyr 295 which are essential for the relative positioning of TMs III, VI, and VII and therefore the positions of other residues located on them (Tyr 115 , Phe 259 , Gln 260 , and Gln 288 ). It is noteworthy that the proximity between Asn 113 and Trp 256 , which constitutes an essential hypothesis of the present work, is more consistent with the rhodopsin structure than with previous rhodopsin models (9,10). Indeed the distance between the ␣ carbon atoms of Gly 120 and Trp 265 , the rhodopsin residues homologs of the B 2 Asn 113 and Trp 256 is 11.5 Å in the crystallographic structure while it is strikingly higher in the model by Herzyk et al. (10) (15.1 Å); the corresponding distance in our experimental AT 1 and B 2 receptor models is 9.5 Å and the spatial arrangement of helices III and VI allows hydrogen bond or amino-aromatic interactions between the side chains, in a B 2 receptor model built by homology to the rhodopsin structure, the distance between the Asn 113 and Trp 256 side chains is not compatible with the geometrical possibility of an hydrogen bond interaction, because of a 40°difference in the orientation of TM III, but the addition of a water molecule between the two side chains would account for their functional interaction. The ability of Zn 2ϩ ions to coordinate histidine residues simultaneously introduced at these positions is an essential verification of the proximity between Asn 113 and Trp 256 . A careful analysis of models possessing the histidine mutations at these positions revealed that Zn 2ϩ ion coordination is theoretically possible in our experimental B 2 receptor model as well as in the rhodopsin-like model (minimal dis-tances between the ␦-N histidine atoms 5.3 and 5.3 Å, respectively, corresponding distances between the ⑀-N histidine atoms 2.0 and 5.4 Å, respectively). Further comparisons revealed that the Asn 111 -Tyr 292 or Asn 113 -Tyr 295 proximity in the experimental AT 1 or B 2 receptor models is consistent with the Herzyk's rhodopsin models (10) while it does not exactly match the rhodopsin-like model (difference in the distances between side chains less than 2 Å in terms of hydrogen bonding possibility). At the present stage of these studies it is difficult to privilege one of the two proposed models or to exclude that an intermediary model would be more realistic. Fig. 8 represents tentative explanations for a wide set of pharmacological data relative to ligand recognition and activation or inactivation properties of the human B 2 bradykinin receptor, with special attention to the key or switch roles of residues involved in the control of conformational equilibria or the binding of nonpeptidic ligands, or both. DISCUSSION The human B 2 bradykinin receptor possesses an extreme conformational flexibility, and thus constitutes a model of choice to investigate the control of conformational equilibria by ligands. Consistently with modeling studies of the AT 1 receptor which originated these studies we hypothesized that an interaction between Asn 113 and Trp 256 could play a major role on the stabilization of an inactive B 2 receptor conformation (4). A first FIG. 7. Models of the human B 2 receptor interaction with the nonpeptidic compound LF 16-0687. The model represented in A was homology-derived from the AT 1 model as described under "Experimental Procedures," with a slight modification of TM VII position. The represented inactive receptor conformation is stabilized by a Asn 113 -Trp 256 interaction, which received experimental support. The nonpeptidic compound binding to the receptor involves interaction of the ligand quinoline moiety with Tyr 295 and Trp 256 and interaction of the carbonyl function of the prolyl moiety with Gln 288 . Interestingly receptor residues involved in nonpeptide binding are located at strategic positions, important for the control of conformational equilibria in the B 2 receptor itself or other GPCR, including rhodopsin and the AT 1 receptor. The model represented in B was homology-derived from the rhodopsin structure. The nonpeptidic ligand docking can be achieved using the same ligand-receptor contact points as in A; the bent form of the ligand extremity bearing the benzamidine moiety results from the arbitrary transposition to the B 2 receptor of the lid role of some rhodopsin extracellular loops (11). Top, side view; bottom, view from the extracellular side. essential finding of the work was the biochemical evidence for the proximity between these two residues, provided by the ability of Zn 2ϩ ions to coordinate histidine residues simultaneously introduced at these positions. It is noteworthy that this proximity is in better agreement with the recently determined rhodopsin structure (11) than previously published rhodopsin models (9,10,31).

Summarized Picture of Interaction Networks Controlling B 2 Receptor Conformational Equilibria and Their Modulation by Mutations or Ligand Action
Extensive investigations were devoted to the function of Trp 256 , fairly conserved in many GPCR. A detailed pharmacological study of the various mutants, including nonpeptide ligand recognition and activation properties, led to the conclusion that the residue at position 256 exerts a subtle control of the balance between inactive and active receptor conformations. The importance of this tryptophan in GPCR function is underlined by initial photolabeling experiments demonstrating that the homologous Trp 265 in rhodopsin lies in the retinal binding pocket (32); it was confirmed by crystallographic data (11) and a more detailed study of rhodopsin conformational changes associated to retinal photoisomerization suggested that Trp 265 keeps 11-cis-retinal as an inverse agonist (33). The same tryptophan was shown to participate in the stabilization of a inactive TRH receptor conformation (26), through interaction with TM V, a finding that could not be extrapolated to the B 2 receptor.
The search for partners interacting with Asn 113 or Trp 256 led us to address the possible role of Phe 259 , which is a BK interaction site and is located one helix turn above Trp 256 in TM VI and Tyr 295 , which is adequately located for a possible interaction with Asn 113 and might play a role in the transition to activated states, in association with other key residues. The inactivation of the AT 1 receptor (15) and the constitutive activation of the ␦-opioid receptor (34) observed when the homologous Tyr residues were mutated probably reflect some conserved features with rhodopsin which covalently binds retinal by its Lys 296 located at the same position.
The search for additional residues which might participate in the stabilization of inactive B 2 receptor conformations led us to evidence strong constitutive activation upon Ala mutation of Tyr 115 . Tyr 115 , located about half an helix turn below Asn 113 in TM III, is homologous of Glu 122 , which is important for rhodopsin function (35) and interacts with His 211 in the inactive crystal form (11,36).
We constructed a B 2 receptor model inherited of previous building (13) and refinements (5) of an AT 1 receptor model, and based on the Asn 113 proximity to Trp 256 and Tyr 295 . Interestingly these proximities can be considered as in fairly good agreement with rhodopsin crystallographic structure (11). At this stage the question is raised to which extent it is justified to deduce the various GPCR models from the structure of rhodopsin. A first consideration is the possible participation of water molecules in hydrogen bond networks which is liable to modify distances between residues which are supposed to interact on the basis of experimental results. The bacteriorhodopsin and rhodopsin examples demonstrate the importance of water molecules and their redistribution associated to retinal photoisomerization (37,38). Another consideration is that the inactive rhodopsin conformation is covalently linked to retinal which behaves as an inverse agonist in the absence of light and should contribute to induce a specific conformation which cannot be transposed to other liganded or unliganded GPCR. Indeed the role of either all-trans-retinal in bacteriorhodopsin or 11-cis-retinal in rhodopsin in exerting constraints on transmembrane helices has been well documented and retinal isomerization has been shown to induce significant changes in the helix bundle organization (33, 39 -45). An elegant recent study demonstrates that appropriate rhodopsin illumination drastically changes the site of its covalent labeling by photoactivatable retinal derivatives and that the generation of some photointermediates is accompanied by helix motions (33).
The postulated interaction network is consistent with the importance of TMs III, VI, and VII motions for GPCR activation (1,46); it would provide a plausible explanation to the destabilizing action of BK through direct interaction with Phe 259 . It is noteworthy that similar considerations can be put forward for AII interaction with Asn 111 and His 256 in the AT 1 receptor (5,16). It is necessary to emphasize that conformational states simultaneously displaying all the possible interactions inside these networks might have no physical reality but the "inactive conformation" might cover a set of conformational states each displaying only part of these interactions. The consequence of mutations or ligand binding would be a change in the statistical distribution of the conformational states or the induction of new states.
Although the characterization of GPCR activated states is not presently available, with the exception of punctual information (47,48), the Asn 113 -Trp 256 and Trp 256 -Phe 259 spatial proximities in the B 2 receptor are consistent with specific local rearrangements which can be predicted to take place during activation. Based on the previous hypothesis of Trp 253 -His 256 interaction in the AT 1 receptor activation process (5), and the properties of B 2 receptors mutated at position 256, we postulated that transition to activated states of the B 2 receptor involves an increase in the representation of conformational states possessing a Trp 256 -Phe 259 interaction (Fig. 8). This hypothesis of intrahelical interaction is in agreement with previous modeling studies of biogenic amine receptors pointing out the role of TM VI aromatic residues (30,49) and data relative to the ␣ 1␤ receptor activation process. 2 Concerning possible mechanistic connections between TM VI and TM VII the question is raised whether the different behaviors of TM VII residues in the AT 1 and B 2 receptors, including Tyr 292 and Tyr 295 , are related to differences in the role of their TM VI Trp residue, i.e. inhibition of hormone-induced activation for the AT 1 (5) and constitutive activation for the B 2 receptors. Relationships between the roles of the two helices would be consistent with the presence of hormone interaction sites near the extracellular ends of TM VI and VII of the AT 1 (50,51) and B 2 (52,53) receptors and the finding that relative motions of TMs VI and VII participate in the activation of muscarinic receptors (54). The construction of chimeric B 1 /B 2 receptors (55) and point mutations in the B 1 receptor TMVII (56) emphasize the role of the 3rd extracellular loop and TM VII, respectively, in the specificity of ligand recognition.
Although it is impossible to deduce from literature data general rules about spatial restrains in inactive GPCR conformations, many data indicate that residues of TMs III and VI, some of which are located at homologous or neighboring positions, have been identified as switch residues, i.e. direct interaction sites for agonists whereby these latter initiate destabilization of inactive conformations. For instance, constitutive activation of the platelet-activating factor receptor (57), the ␣ 1B or ␤ 2 -adrenergic receptors (58,59), or the B 1 bradykinin receptor (28) could be induced by the mutation of an asparagine or a cysteine located at the homologous position of the TM III Asn of the AT 1 and B 2 receptors. The role of this Asn is consistent with membrane helix studies which emphasize that hydrogen bond interhelical interactions can be driven by Asn residues (60 -62). Asn 111 and His 256 were postulated to constitute AII interaction points (16) and are probably involved in the generation of the AT 1 receptor intermediary conformational states (17); that Phe 259 is a BK-binding residue in the B 2 receptor is consistent with the postulated role of this residue inside the intramolecular interaction networks. Interestingly residues located at the same position in TM VI of other receptors control the activation process and are involved in ligand binding: Phe 310 in the ␣ 1B receptor (63), Tyr 381 in the M 1 muscarinic receptor (64), and Tyr 282 in the TRH receptor (26). A systematic analysis of randomly mutated receptors (2,3,65) emphasize that other residues located in TMs III and VI are involved: for instance, Phe 451 and Asn 459 in the M 5 muscarinic receptor, homologous of the Phe 252 and Gln 260 B 2 residues.
A second aspect of the present work was to establish the molecular basis of nonpeptide ligand recognition and inactivation of the B 2 bradykinin receptor. Structure-function analysis allowed precise essential nonpeptide-receptor contacts: the nonpeptide quinoline moiety interacts simultaneously with Tyr 295 and Trp 256 . Gln 288 is also involved in ligand binding. On this basis we could propose a docking of the nonpeptidic ligand LF 16-0687 either in a receptor model homology deduced from a previous AT 1 model (experimental model) or in a model derived from the rhodopsin structure (rhodopsin-like structure). It is noteworthy that B 2 receptor nonpeptide antagonists studied in the present work interact with residues of TM VI and VII, possibly restricting movements of these helices required for activation, together with helix III movements. The nonpeptide interaction with these residues, with a major "locking" role of the sandwich interaction of their quinoline moiety with Trp 256 and Tyr 295 , might restrain the essential Trp 256 in a position favorable to its interaction with Asn 113 which constitutes the cornerstone of inactive conformations. This stabilizing effect would provide an explanation for the inverse agonist behavior of this ligand and its agonist behavior on the CAM N113A and W256F receptors which no longer possesses the Asn-Trp interaction; in this latter situation the high conformational flexibility of the nonpeptidic derivatives would allow them to fit with or induce CAM receptor active conformations.
Interestingly the B 2 receptor residues involved in nonpeptide 2 S. Cotecchia, unpublished results.
recognition are somehow involved in receptor activation even if the precise role of some of them, located at key positions in other GPCR, requires further elucidation. It indicates that ligands synthesized through semiempirical strategies have been implicitly selected according to their ability to interact with receptor residues which control conformational equilibria. Similar conclusions can be drawn for the AT 1 receptor, inasmuch Asn 111 and His 256 , which are switch residues for AII action, are involved in losartan and other nonpeptide ligand binding (together with Asn 295 in TM VII) (5,14,66,67). The NK 2 receptor amino acid Tyr 266 , homologous to the AT 1 receptor His 256 , was shown to participate in peptide agonist and nonpeptide antagonist binding (68). The question is raised whether such considerations should be taken into account for future attempts of rational design of nonpeptidic GPCR ligands. Experimental evaluation of the conformational flexibility of GPCR (1) and the ability of ligands to induce specific conformational changes will probably constitute a major difficulty in future structure predictions from the rhodopsin crystallographic data (11). Conflicting results were reported about the extent of structural changes associated to rhodopsin activation (33,69). A recent reassessment of catecholamine action on the ␤ 2 -adrenergic receptor (70) demonstrates the usefulness of thermodynamic studies to dissect ligand action in terms of receptor binding and modulation of conformational equilibria. Taken together our data emphasize the role of B 2 receptor key residues which control conformational equilibria and constitute ligand interaction points and are located at strategic positions for the function of other GPCR, including rhodopsin.