The conformational change responsible for AT1 receptor activation is dependent upon two juxtaposed asparagine residues on transmembrane helices III and VII.

A model of the angiotensin AT1 receptor and site-directed mutagenesis were used to identify key residues involved in ligand binding. Receptors were stably expressed in human embryonic kidney 293 cells, and their binding properties compared. Wild type receptors exhibited low and high affinity binding sites for peptides. Substitution of Asn111, situated in the third transmembrane helix, resulted in a significant alteration in ligand binding with only high affinity binding of the peptides, angiotensin II, angiotensin III, and [p-amino-Phe6]angiotensin II and a marked loss in the binding affinity of the AT1 receptor selective non-peptide antagonist losartan. From our model it was apparent that Asn111 was in close spatial proximity to Asn295 in the seventh transmembrane helix. Substitution of Asn295, produced identical changes in the receptor's pharmacological profile. Furthermore, the Ser111AT1A and Ser295AT1A mutants did not require the association of a G-protein for high affinity agonist binding. Finally, the Ser295AT1A mutant maintained higher basal generation of inositol trisphosphate than the wild type, indicating constitutive activation. We propose that substitution of these residues causes the loss of an interaction between transmembrane helices III and VII, which allows the AT1 receptor to “relax” into its active conformation.

and [p-amino-Phe 6 ]angiotensin II (pNH 2 F 6 AII) (4), both display selectivity for the AT 2 receptor over the AT 1 receptor. Furthermore, several studies suggest that AT 1 and AT 2 receptors display subtle differences in their binding affinities for Ang II and angiotensin III (Ang III). AT 2 receptors bind Ang II and Ang III with equal affinity (5)(6)(7)(8), while AT 1 receptors exhibit a slightly higher affinity for Ang II, between 3-and 20-fold greater than Ang III (9 -11). Many of the recognized actions of Ang II appear to be mediated through AT 1 receptors, predominantly through activation of the phospholipase C signal transduction pathway.
AT 1 receptors have been cloned from a variety of species and tissues (9 -15). All these AT 1 receptors consist of a single polypeptide, 359 amino acids in length, arranged with a topography comprising seven ␣-helical transmembrane regions (TM I-VII), typical of the G-protein-coupled receptor family. They display a high degree of sequence identity at the amino acid level (over 94% identical between all mammalian species). Two closely related isoforms (AT 1A and AT 1B ) encoded by different genes have been identified in the rat and mouse (13,16) Recently AT 2 receptors have been cloned (17)(18)(19), which are also members of the G-protein-coupled receptor family. They consist of 363 amino acids and share only 34% sequence identity at the amino acid level to the AT 1 receptors.
Mutagenesis studies of the AT 1 receptor have recently identified residues important in the binding of the non-peptide antagonist, losartan, and the natural ligand Ang II. These residues appear to be mutually exclusive, since amino acid substitutions, which affect losartan binding, do not alter Ang II binding, and vice versa. The major determinants of losartan binding appear to be residues located within the transmembrane helices III-VII (20 -23). In contrast, the major determinants of Ang II binding appear to be residues present in the extracellular regions, particularly in the N-terminal extension adjacent to the first transmembrane helix and in the C-terminal part of the third extracellular loop (24,25). However, a conserved residue, Lys 199 in the fifth transmembrane helix (TM V) of the AT 1 receptor, has recently been reported to be crucial for the binding of both peptide and non-peptide ligands (26). In the present study we constructed a three-dimensional model of the AT 1A receptor using the method of Donnelly et al. (27) to guide site-directed mutagenesis aimed at identifying potential residues involved in ligand binding. The consequences of substituting either Asn 111 , in the third transmembrane helix (TM III), or Asn 295 , in the seventh transmembrane helices (TM VII), on ligand binding and receptor coupling are presented. These data suggest that a specific interaction between TM III and TM VII may be a major determinant of AT 1 receptor isomerization from an inactive to active conformation. * This work was supported by research grants from the British Heart Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Site-directed Mutagenesis-The entire coding region of the rat AT 1A receptor (a HindIII-NotI fragment of 2 kilobases) was subcloned into the polylinker sites of pBluescript KS ϩ , and single-stranded DNA was rescued from CJ236 cells using M13K07 helper phage. Mutagenesis was performed with a Mutagene Phagemid In Vitro Mutagenesis Version 2 kit and confirmed by dideoxy sequencing using Sequenase II (U. S. Biochemical Corp.). The wild type and mutated AT 1A cDNAs were subcloned into the mammalian expression vector pCEP4 using the restriction enzymes HindIII and NotI.
Permanent Expression of Receptors in HEK293 Cells-HEK293 were cultured in modified Eagle's medium containing Earle's salts, supplemented with 10% (v/v) fetal calf serum, 1% nonessential amino acids, 50 g/ml gentamicin, 100 units/ml penicillin G, 100 g/ml streptomycin, and 0.25 g/ml amphotericin at 37°C in a humidified atmosphere of air/CO 2 (19:1). Cells were transfected with 12 g of pCEP4 containing either the cDNA for the wild type or mutated AT 1A receptors using Transfectam, according to the manufacturer's instructions. Stable expression of AT 1A receptors in HEK293 cells was achieved by addition of 200 g/ml hygromycin B to the medium 3 days after transfection and for all subsequent passages of the cells.
Binding Assay-Preparation of transfected HEK293 cell membranes and subsequent radioligand-receptor binding assays were undertaken as described previously (28) with the following modifications. All competitors were diluted in assay buffer which consisted of 100 mM NaCl, 20 mM Hepes (pH 7.4), 1 mM EDTA, 10 mM MgCl 2 , bacitracin (0.1% w/v), and bovine serum albumin (0.1% w/v). Nine to 11 concentrations of competitor were examined, each concentration in duplicate. Membranes were incubated with 125 I-[Sar 1 ,Ile 8 ]angiotensin II and competitors for 60 min at room temperature. In the experiment to determine the effect of GTP␥S on Ang II binding, the total reaction volume was 300 l, consisting of 200 l of membranes (10 -50 g of protein), 50 l of 125 I-angiotensin II (1.2 nM), and 50 l of GTP␥S (60 M) or assay buffer.

Measurement of Intracellular Calcium in Transfected HEK293
Cells-Cells were subcultured on glass coverslips and used at subconfluence. The coverslips were incubated with the cell permeant fluorescent probe fura 2-AM (5 M) in a Krebs-Ringer buffer (pH 7.4) consisting of 145 mM NaCl, 5 mM KCl, 1.3 mM MgCl 2 , 1.2 mM NaH 2 PO 4 , 1.3 mM CaCl 2 , 10 mM glucose, 20 mM Hepes, at 20°C for 1 h. Loaded cells were then washed free of extracellular dye, and the coverslip was mounted in an open culture chamber which was then placed on the stage of an inverted Nikon microscope and maintained at 37°C in Krebs-Ringer buffer. Dynamic video imaging was carried out as described previously (29), using Magical hardware and TARDIS software (Applied Images, Sunderland, UK) and a low light level intensified charged coupled device camera (Photonic Science, Robertsbridge, UK).

Measurement of Inositol 1,4,5-Trisphosphate (IP 3 ) Production in Transfected HEK293
Cells-Cells were subcultured on 30-mm Petri dishes and used at confluence. Each dish was washed with assay medium (inositol-free Dulbecco's modified Eagle's medium containing 20 mM HEPES, 0.25% (w/v) bovine serium albumin, 0.25% (w/v) baci-tracin, pH 7.4; three times at 2 ml each). After washing, 1 ml of medium containing either no reagents or 100 nM angiotensin II was added, and incubations were performed for 15 s. The assay was stopped by the addition of ice-cold 20% trichloroacetic acid (0.5 ml). Subsequent extraction and measurement of IP 3 production was undertaken using an inositol 1,4,5-trisphosphate 3 H radioreceptor assay kit according to the manufacturer's instructions. Functional studies for both wild type and Ser 295 AT 1A receptors were routinely performed in a single assay.
Data Analysis-Radioligand binding results are expressed as a percentage of the control (i.e. specific binding in the absence of any competitor). The binding data were analyzed and IC 50 values determined by nonlinear regression analysis using GraphPad Prism (GraphPad Software Inc., San Diego, CA). Competition curves were simultaneously fitted to one-and two-site models, and an F test was used to determine the most appropriate model. All IC 50 values are quoted as the mean Ϯ S.E.  Table I) and the ability of Ang II to stimulate a rise in intracellular calcium ( Fig. 2A). GraphPad Prism Analysis of the Ang II, Ang III, and pNH 2 F 6 AII competition curves fitted best a two-site model. These ligands competed for both high and low affinity binding sites, with estimated IC 50 values (nM) of 0.8 Ϯ 0.1, 3.5 Ϯ 0.2, and 63 Ϯ 30 for high, and 12 Ϯ 2.8, 120 Ϯ 12, and 51810 Ϯ 4075 for low affinity binding sites, respectively. The AT 1 selective non-peptide antagonist losartan competed for a single high affinity binding site (IC 50 value of 26.1 Ϯ 1.5 nM), while the AT 2 selective ligands, CGP42112A, and PD123319 (an analogue of PD123177) competed for single low affinity binding sites (IC 50 values of 5930 Ϯ 1466 and 102650 Ϯ 6582 nM, respectively). The functional coupling of the receptor was demonstrated by measuring intracellular calcium. Addition of Ang II (100 nM) to transfected cells resulted in a rapid, transient increase in intracellular calcium, which reached maximum levels by 5-10 s and then declined toward resting levels by 60 s (Fig. 2A). Neither this response nor specific binding of the radioligand 125 I-[Sar 1 ,Ile 8 ]angiotensin II was observed in untransfected cells (data not shown).

Characterization of the Wild Type AT
Characterization of the Ser 111 and Ala 111 Mutant AT 1A Receptors-The replacement of an asparagine residue (Asn 111 ), located in the third transmembrane helix, with a serine, resulted in both subtle and large changes in the binding affinities of the AT 1 -and AT 2 -selective ligands studied here ( Fig. 1B and Table I). In contrast to the findings with the wild type receptor, all of the ligands now competed for binding to a single site. Of particular interest was the observation that the resulting IC 50 values for Ang II, Ang III, and pNH 2 F 6 AII (0.9 Ϯ 0.1, 0.6 Ϯ 0.1, and 68 Ϯ 13.6 nM, respectively) were very similar to the IC 50 values recorded for the binding of these ligands to their high affinity sites displayed by the wild type receptor. One consequence of this altered binding was that the mutated receptor now had equal affinity for Ang II and Ang III; an observation typical of an AT 2 receptor. Furthermore, the AT 2 selective peptide ligand, CGP42112A, also displayed an increased binding affinity (28 fold), in keeping with this trend toward AT 2 receptor pharmacology. However, the AT 2 selective non-peptide ligand PD123319 showed only a modest 3-fold increase in binding affinity for the mutated receptor.
In contrast to the above increases in binding affinity, the ability of the specific AT 1 non-peptide antagonist losartan to bind to the mutated receptors was severely impaired (a 148fold decrease in binding affinity). The rank order of potency was thus Ang III ϭ Ang II Ͼ pNH 2 F 6 AII Ͼ CGP42112A Ͼ losartan Ͼ PD123319. The pharmacological profile of the Ala 111 AT 1A receptor was identical to that of the Ser 111 AT 1A receptor, with only minor differences in the IC 50 values for all the ligands tested ( Fig. 1C and Table I). Addition of Ang II (100 nM) to HEK293 cells transfected with the two Asn 111 mutant AT 1A receptors resulted in a rapid, transient increase in intracellular calcium, similar to that observed in HEK293 cells transfected with the wild type receptor (Fig. 2, B and C).
Characterization of the Ser 295 AT 1A Receptor-Analysis of the molecular model of the AT 1A receptor revealed that Asn 111 located in the third transmembrane helix was in spatial proximity to Asn 295 in the seventh transmembrane helix (Fig. 3, A  and B), raising the possibility of an interaction between these two residues. Since this interaction may be disrupted by the mutation of Asn 111 to either an alanine or serine residue, its loss may be responsible for the observed changes in ligand binding. This hypothesis was tested by creating a Ser 295 AT 1A receptor. The pharmacological profile of the Ser 295 AT 1A receptor was identical to that of the Asn 111 mutants, with only minor differences in the IC 50 values for all the ligands tested ( Fig. 1D and Table I). Furthermore, addition of Ang II (100 nM) to HEK293 cells transfected with the Ser 295 AT 1A receptor resulted in a rapid, transient increase in intracellular calcium, similar to that observed in HEK293 cells transfected with the wild type receptor (Fig. 2D).
Effect of GTP␥S on Angiotensin II Binding-The effects of GTP␥S on 125 I-angiotensin II binding were examined in membranes prepared from HEK293 cells expressing wild type, Ser 111 AT 1A , and Ser 295 AT 1A receptors (Fig. 4). GTP␥S reduced Ang II binding to the wild type receptor to 13.8 Ϯ 0.4% of that in the absence of GTP␥S (n ϭ 3), whereas it had no effect on Ang II binding to Ser 111 AT 1A (98.6 Ϯ 2.1%, n ϭ 3) or Ser 295 AT 1A (102.7 Ϯ 3%, n ϭ 3) receptors. This suggests that,  in contrast to the wild type receptor, coupling to G-proteins is not required for these mutated receptors to maintain their high affinity binding. Inositol 1,4,5-Trisphosphate Production in Transfected HEK293 Cells-IP 3 production was examined in HEK293 cells expressing either wild type AT 1A or Ser 295 AT 1A receptors (Fig.  5). Even though the level of expression of the Ser 295 AT 1A receptor (1385 Ϯ 123 fmol/mg total cell protein, n ϭ 3) was less than that of the wild type AT 1A (1990 Ϯ 178 fmol/mg total cell protein, n ϭ 3), the cells expressing the Ser 295 AT 1A receptor exhibited higher levels of agonist-independent (basal) production of IP 3 (2.2 Ϯ 0.4-fold higher than wild type receptor basal value, n ϭ 6; Fig. 5, left). Furthermore, although 100 nM Ang II stimulated both wild type AT 1A and Ser 295 AT 1A receptors, the total IP 3 production by cells expressing the Ser 295 AT 1A receptor was lower (64 Ϯ 7% of the wild type AT 1A , n ϭ 6) than that produced by cells expressing the wild type AT 1A receptor (Fig.  5, right). DISCUSSION Using our model of the rat angiotensin AT 1A receptor, we selected to mutate the asparagine residue at position 111, located in TM III. Substitution with either a serine or alanine residue resulted in a marked alteration in the binding of several peptides and the AT 1 selective non-peptide ligand losartan. Importantly, only high affinity binding was observed following Asn 111 substitutions, in contrast to the wild type receptor, which displayed both high and low affinity binding for Ang II, Ang III, and pNH 2 F 6 AII. Furthermore, the binding affinity of the AT 1 receptor-selective non-peptide antagonist losartan decreased (148 -248-fold). Identical results were obtained when we substituted a serine residue for an asparagine residue located in TM VII (Asn 295 ), identified by our three-dimensional model of the rat AT 1A receptor as being in spatial proximity to Asn 111 . All these mutations retained Ang II mobilization of intracellular calcium, indicating that these substitutions did not perturb coupling of these receptors to the phospholipase C signal transduction pathway.
The binding of agonists, but not antagonists, to G-proteincoupled receptors in membrane preparations is characterized by high and low affinity binding states. The high affinity state is believed to be the active conformation of the receptor, consisting of a ternary complex involving the agonist, the receptor, and the G-protein, while the low affinity state is believed to be an inactive binary complex of agonist and receptor (30). In the presence of guanine nucleotides, which cause dissociation of G-proteins from their receptors, the high affinity state appears to be converted into the low affinity state (31).
Ang II receptors are typical of G-protein-coupled receptors, existing in high and low affinity agonist binding states in various tissues, the higher being converted to the lower in the presence of guanine nucleotides (32)(33)(34)(35). We have also observed that wild type recombinant AT 1A receptors display both high and low affinity agonist binding states for Ang II. Furthermore, using a low concentration of 125 I-angiotensin II, which would preferentially bind to the high affinity state, we demonstrated that this high affinity state was lost when binding was undertaken in the presence of the nonhydrolyzable guanine nucleotide GTP␥S. This suggested that the high affinity state was also dependent on formation of a ternary complex involving Ang II, the AT 1A receptor and a G-protein.
Ang III and pNH 2 F 6 AII also bound to high and low affinity states of the wild type AT 1A receptor, suggesting both are agonists. While Ang III is accepted as an agonist, there are no direct reports of the agonist actions of pNH 2 F 6 AII. We included pNH 2 F 6 AII in our studies since it had been reported that it preferentially bound to AT 2 receptors and might therefore be useful as a subtype selective ligand (4). However, in keeping with our observations of its agonist properties, these authors also reported that the potency of binding of pNH 2 F 6 AII to AT 1 receptors was diminished in the presence of GTP␥S but did not attempt to estimate the high and low affinity agonist binding state values (4).
There was a striking similarity between the estimated high affinity agonist binding state values for Ang II, Ang III, and pNH 2 F 6 AII displayed by the wild type AT 1A receptor and their corresponding single site binding affinities in the Asn 111 and Asn 295 mutant AT 1A receptors. This suggests that a consequence of these substitutions is to allow all of the receptors expressed to adopt the high affinity agonist binding state conformation. However, the effects of GTP␥S on Ang II binding indicate that, whereas the wild type receptor requires the association of a G-protein to form the ternary complex to adopt the conformation of a high affinity agonist binding state, the Ser 111 AT 1A and Ser 295 AT 1A mutants do not (Fig. 4). While, in many instances, the lack of a GTP␥S effect on agonist binding merely indicates that a mutant receptor is no longer coupled, this was clearly not the case here. All three mutants produced transient increases in intracellular calcium similar to the wild type receptor. Thus substitutions at Asn 111 and Asn 295 produce a G-protein-independent, high affinity agonist binding state conformation of the AT 1A receptor which is still capable of coupling to a G-protein. Such features are characteristic of constitutively active receptors.
In a subsequent experiment (Fig. 5) we compared IP 3 production in cells expressing either wild type AT 1A or Ser 295 AT 1A receptors. In agreement with the intracellular calcium data we confirmed that the Ser 295 AT 1A receptor was capable of coupling to a G-protein. However, we also noted two discrepancies between the cells expressing the wild type AT 1A and Ser 295 AT 1A receptors. First, although the level of expression of the Ser 295 AT 1A receptor was less than that of the wild type AT 1A receptor, cells expressing the Ser 295 AT 1A receptor exhibited higher levels of agonist-independent production of IP 3 (approximately 2-fold higher basal), suggesting that this receptor is constitutively active. Second, Ang II stimulated IP 3 production by cells expressing the Ser 295 AT 1A receptor was lower than that produced by cells expressing the wild type receptor. Since the Ser 295 AT 1A receptor appears constitutively active, the underlying mechanism may be constitutive desensitization, as previously reported for a constitutively active ␤ 2 -adrenergic receptor (36). Similar findings are anticipated for the Asn 111 mutants, but so far have not been undertaken.
Constitutive activation of a G-protein coupled receptor was first described in the ␣ 1B -adrenergic receptor (37), with substitution of amino acids in the C terminus of the third intracellular loop resulting in increases in both agonist binding affinity and basal inositol phosphate production (suggesting receptor activation in the absence of an agonist). Mutations that cause constitutive activation appear to occur at widely distributed sites in the sequences of G-protein-coupled receptors (38). However, relevant to our work is the observation that substitution of Glu 113 in TM III or Lys 296 in TM VII cause constitutive activation of rhodopsin (39). According to our model, equivalent residues to these in the AT 1A receptor would be positioned two helical turns and one helical turn above Asn 111 and Asn 295 , respectively. Of particular interest is that substitution of Cys 128 in the ␣ 1B -adrenergic receptor leads to constitutive activation (40). This residue is equivalent in position to Asn 111 in the AT 1A receptor and both receptors are coupled to phospholipase C.
Previous approaches to identifying residues important in the binding of ligands to AT 1 receptors have, in general, been undertaken comparing the effects of substituting residues on the binding of non-peptide antagonist(s) and at most only one or two peptide ligands. Although such studies have established that peptides bind differently to non-peptide Ang II receptor antagonists (20 -24), the limited range of ligands used may have resulted in inappropriate conclusions. For example, Schambye et al. (21) reported that substitution of Asn 295 with a serine residue substantially reduced non-peptide antagonist binding without affecting the binding of Ang II or [Sar 1 ,Leu 8 ]angiotensin II. They concluded that this residue was directly or indirectly interacting with the non-peptide antagonists, but was not involved in the binding of peptides. Recently, similar results and conclusions were reported independently by two groups (22,23) for Asn 111 . However, although the present study confirms some of the above observations, our alternative approach suggests a very different conclusion regarding the function of these two residues.
We determined which residues to mutate using our threedimensional model and chose to explore a wide series of peptides and non-peptide antagonists to test predictions from the model. These included peptides that display differential binding to AT 1 and AT 2 receptors. Our findings indicate a pivotal role for the asparagines at positions 111 and 295 in the isomerization step described in the allosteric ternary complex model (41). In this model it is proposed that the receptor undergoes an isomerization from an inactive (R form) to an active (R* form) state which couples to G-proteins (R*G). This isomerization involves conformational changes which may occur spontaneously, or be induced by agonists or appropriate mutations which abrogate the normal 'constraining' function of the receptor, allowing it to "relax" into the active conformation. The identical effects of mutating Asn 111 in TM III and Asn 295 in TM VII on the binding of a series of agonists and antagonists, suggests that these two Asn residues may have a complementary role in the receptor isomerization. Due to the potential spatial proximity of these two residues identified by our three dimensional model (Fig. 3, A and B), this role may be through direct hydrogen bonding. Substitution of a serine residue would be expected to reduce and alanine to eliminate, hydrogen bonding associated with the asparagine residues. Thus one of the major determinants of the isomerization of the AT 1A receptor from the R to R* form may be the breaking of hydrogen bonding between Asn 111 in TM III and Asn 295 in TM VII.
Inverse agonists are ligands which are postulated to posses a higher affinity for the inactive R form of the receptor when compared to the active R* and R*G forms and are capable of shifting the equilibrium from R*G by stabilizing the R conformation (for review, see Milligan et al. (42)). In the present study, the marked reduction in the binding affinity of losartan observed following substitution of either Asn 111 or Asn 295 , raises the possibility that it is not an antagonist, but an inverse agonist.
A further interesting but more speculative implication of the present study is that the interaction between Asn 111 and Asn 295 may be fundamental in determining the apparent peptide binding selectivity of AT 1 and AT 2 receptors. The pharmacological profile of the mutant receptors, particularly of the peptides, was closer to that of an AT 2 receptor (5-8), than an AT 1 . Only the change in the affinity of CGP42112A is not entirely consistent with the typical AT 2 binding of peptides. This discrepancy may be due to the fact that CGP42112A also contains non-peptide modifications and thus is quite unlike Ang II, Ang III, or pNH 2 F 6 AII and therefore would be less likely to share similar "peptide" binding epitopes. Furthermore, in addition to displaying similar peptide binding affinities, the binding of Ang II to AT 2 receptors is also unaffected by the presence of guanine nucleotides (4,6,35,43) as observed here for the mutant receptors. Finally, the cloning of the AT 2 receptor has now revealed that it also has an asparagine residue in TM III at an equivalent position to Asn 111 of the AT 1 receptor. However, in TM VII of the AT 2 receptor, a serine residue corresponds to the asparagine at the equivalent 295 position in the AT 1 receptor (Fig. 6). This therefore raises the possibility that the loss (AT 2 receptor) or gain (AT 1 receptor) of this interaction between TM III and TM VII is a major determinant of peptide selectivity between Ang II receptor subtypes. It also raises the possibility that in vivo AT 2 receptors may be constitutively active. FIG. 4. Effect of GTP␥S on 125 I-angiotensin II binding. Membranes were prepared from HEK293 cells stably expressing wild type AT 1A , Ser 111 AT 1A , and Ser 295 AT 1A receptors, and the binding of 0.2 nM 125 I-angiotensin II to the membranes was measured in the presence and absence of 10 M GTP␥S as described under "Experimental Procedures." The data are expressed as the percentage of the specific binding measured in the absence of GTP␥S and are shown as means Ϯ S.E. from three independent experiments, each performed in triplicate. In summary, we have established that substitution of either Asn 111 in TM III or Asn 295 in TM VII leads to an identical conformational change in the receptor. The conformation adopted appears to be the active (R* form) of the receptor. The spatial proximity of these two residues suggests that the mechanism responsible for this conformational change may be the breaking of a hydrogen bond interaction between these two residues. The proximity of these two helices has been postulated by us and others in three dimensional models based upon the projection map of bovine rhodopsin (27,44,45). Earlier models based upon the bacteriorhodopsin structure tend to result in TM III and TM VII being more distant (46). To our knowledge, this is the first evidence of a direct interaction between TM III and TM VII in a G-protein-coupled receptor.