Cooperative Conformational Changes in a G-protein-coupled Receptor Dimer, the Leukotriene B4 Receptor BLT1*

We have used an isolated receptor, the leukotriene B4 receptor BLT1, to analyze the mechanism of receptor activation in a G-protein-coupled receptor dimer. The isolated receptor is essentially a dimer whether the agonist is present or not, provided the detergent used stabilizes the inactive dimeric assembly. We have produced a receptor mutant where Cys97 in the third transmembrane domain has been replaced by a serine. This mutation leads to an ∼100-fold decrease in the affinity for the agonist. 5-Hydroxytryptophan has then been introduced at position 234 in the C97A mutant sixth transmembrane domain. Agonist binding to the labeled receptor is associated with variations in the fluorescence properties of 5-hydroxytryptophan due to specific agonist-induced conformational changes. The C97A mutant labeled with 5-hydroxytryptophan has then been associated with a wild-type receptor in a dimeric complex that has been subsequently purified. The purified complex activates its G-protein partner in a similar manner as the wild-type homodimer. Due to the difference in the affinity for the agonist between the wild-type and mutant protomers in this dimer, we have been able to reach a state where one of the protomers, the mutant, is in its unliganded state, whereas the other, the wild type, is loaded with the agonist. We show that agonist binding to the wild-type receptor induces specific changes in the conformation of the unliganded protomer, as evidenced by the variations in the emission of the 5-hydroxytryptophan residue in the mutant receptor. These data provide a direct demonstration for agonist-induced cooperative conformational changes in a GPCR dimer.

G-protein-coupled receptors (GPCRs) 1 are versatile biological sensors that are responsible for the majority of cellular responses to hormones and neurotransmitters as well as for the senses of sight, smell, and taste (1,2). Signal transduction is associated with a set of changes in the tertiary structure of the receptor that are recognized by the associated G-protein (3). A growing body of evidence point to the fact that GPCRs exist as homo-or heterodimers (4 -6). However, the role of dimerization in receptor function and the molecular mechanisms associated with receptor activation in the dimeric assembly are less well understood. Receptor dimerization is, in some cases, required for a correct addressing of the receptor to the membrane. This is clearly demonstrated for class C GPCRs such as the GABA B receptor (7)(8)(9)(10) and for some class A receptors (11)(12)(13). There is also evidence that receptor dimerization and activation are intricately associated. In this context, an interesting mechanism to consider is trans-activation. In this mechanism, ligand binding to one of the monomers in the dimeric assembly results in activation of the other receptor within the dimer. Trans-activation is well documented for class C GPCRs, in particular for the GABA B receptor. In this case, it is clearly demonstrated that one subunit, GB1, binds the agonist, whereas the other, GB2, activates the Gprotein (14 -16). The situation is not that clear for the other classes of receptors. In the case of the LH/CG receptor, co-expression of a mutant receptor defective in hormone binding and another mutant defective in signal generation rescued hormoneactivated cAMP production (17,18). Similarly, Carillo et al. (19) have shown, using a series of G␣ fusion proteins and specific mutant of the histamine H1 and ␣ 1b -adrenergic receptors, that co-expression of two nonfunctional but complementary fusion constructs reconstituted agonist-mediated signaling. All of these results point to the fact that dimers of these receptors could function via trans-activation. Finally, dimerization is also likely to be required for an efficient interaction with intracellular partners such as the G-protein (20) and certainly plays a role in receptor internalization (21,22).
We have produced the leukotriene B 4 receptor BLT1 as a functional protein isolated in a detergent medium (23). Binding of the agonist to the isolated receptor is associated with a series of specific conformational changes leading to the active state of the receptor (23). Isolated BLT1 therefore appears as a particularly convenient in vitro model system to analyze the relationship between dimerization and activation. By using a receptor dimer where one of the protomers displays a lower affinity for the agonist, we establish here that agonist-induced activation of one of the monomers in the dimeric assembly affects the conformation of the unliganded monomer. This clearly indicates that cooperative conformational changes occur in this GPCR dimer.
Site-directed Mutagenesis-The C97A mutation was introduced in both the wild-type receptor and the W41L,W83L,W142L,W161L mutant by PCR-mediated mutagenesis using the QuikChange multisitedirected mutagenesis kit (Stratagene) and, as a template, either the wild-type BLT1 or the BLT1W41L,W83L,W142L,W161L constructs described by Banères et al. (23). Mutations were confirmed by nucleotide sequencing.
R⅐R 0 Purification-We used here the wild-type BLT1 and BLT1W41L,W83L,C97A,W142L,W161L construct described previously (23). In both cases, a stop codon (TGA) was inserted by site-directed mutagenesis immediately after the sequence encoding the receptor to prevent expression of the His tag sequence at the receptor C terminus. We inserted an S tag encoding sequence (Novagen) at the 5Ј-end of the sequence encoding BLT1 and a hexahistidine encoding sequence at the 5Ј-end of the sequence encoding BLT1W41L,W83L,C97A,W142L,W161L. These two tags were inserted between the NdeI and BamHI sites in the pET21b-BLT1 vector (23). The BLT1 receptor was expressed and purified, under denaturing conditions, as previously described (23), except that 1.25% (w/v) sarcosyl was added to the denaturing buffer to increase the solubilization ratios. 5HW was incorporated in BLT1W41L,W83L,C97A,W142L,W161L by using the CY(DE3)pLysS Escherichia coli strain (20). The 5HW-labeled protein was purified under denaturing conditions under the same conditions as the wild-type receptor (23). The wild-type (R) and mutant (R 0 ) receptors were mixed in equimolecular amounts before refolding, and the protein was refolded as described in Banères et al. (23). Unfolded proteins were discarded as described in Banères et al. (23), and the functional receptor dimers were further purified on a 5-20% isokinetic sucrose gradient. The conformational features of the recovered protein were systematically checked by CD and shown to be similar to those previously reported (23). For R⅐R 0 purification, the refolded proteins were first immobilized on an S-protein-agarose column (Novagen) previously equilibrated in buffer B. A first protein fraction (F1; see Fig. 4) eluted during washing with buffer B. The retained proteins were then eluted with buffer B containing 1 M MgCl 2 . The protein fraction recovered under these conditions was directly loaded on an Ni 2ϩ -nitrilotriacetic acid superflow column (Qiagen). A first protein fraction (F2; see Fig. 4) was eluted during the washing step with buffer B. Elution was then carried out with buffer B containing 100 mM imidazole to recover the last protein fraction (F3; see Fig. 4). F1, F2, and F3 were extensively dialyzed in buffer B.
Ligand Binding Assays-LTB 4 binding was assayed as previously described (20,23). The titration data were analyzed using the PRISM software (Graphpad Inc.) by considering a set of usual models for describing the ligand-receptor interactions.
GTP␥S Binding Assays-[ 35 S]GTP␥S binding assays were performed as previously described (20). Data are presented as percentage of bound GTP␥S normalized to the binding obtained with the LTB 4 -saturated R⅐R complex (20).
Alexa Fluor Labeling-The receptor purified in its unfolded state was extensively dialyzed in 12.5 mM sodium borate, pH 7.3, 1.25% (w/v) sarcosyl to remove urea while maintaining the receptor soluble. This pH value was determined from a series of labeling reactions carried out at different pH (ranging from 7 to 9) to define the optimal value for labeling only the protein N terminus and not the lysyl residues. The fluorescent probes were added to the protein solution (reactive reagent/ protein molar ratio 10:1), and the reaction was carried out at room temperature for 5 h under constant stirring. The conjugate was separated from the unreacted labeling reagent on a Sephadex G-50 column (10 ϫ 300 mm) equilibrated in 12.5 mM sodium borate, pH 7.8, 1.25% (w/v) sarcosyl. The relative efficiency of the labeling reaction was determined by measuring the absorbance of the protein at 276 nm (23) and that of the dye at its absorbance maximum (346 and 495 nm for Alexa Fluor 350 and 488, respectively). A ratio of 0.7-0.8 of fluorescent probe per receptor molecule was routinely found. The effectiveness of the N-terminal labeling was verified by proteolytic digestion of the receptor and N-terminal sequencing.
Fluorescence Measurements-Fluorescence emission spectra were recorded at 20°C on an Eclipse spectrofluorimeter (Cary) with an excitation wavelength of 315 nm (bandwidth 2 nm) for the 5HW fluorescence measurements and 350 nm (bandwidth 2 nm) for the fluorescence resonance energy transfer (FRET) experiments. For the excitation spectra, the emission was collected at 340 nm. For the emission spectra of the agonist-loaded receptor, emission was recorded 30 min after adding the ligand. Protein concentrations in the 10 Ϫ8 to 10 Ϫ9 M range were used. Buffer B was used in all measurements. Buffer contributions were subtracted under the same experimental conditions.

RESULTS
BLT1 Dimerization-We previously showed that agonistloaded BLT1 was essentially a dimer, whereas the free receptor corresponded to a mixture of monomer and dimer (20). As we previously discussed, this could be due to agonist-induced dimerization or could translate the fact that the unliganded dimer is less stable than the agonist-loaded one so that it partly dissociates in a detergent medium. To assess these different possibilities on an experimental basis, we explored the stability of the ligand-free BLT1 dimer in the presence of different detergents. The monomer/dimer ratio was defined by chemical cross-linking, as previously described (20). The results obtained for some of these detergents are given in Fig. 1. The amount of receptor dimer in the absence of agonist is clearly dependent on the detergent used. Whereas, as previously reported (20), the agonist-free receptor corresponds to a mixture of monomer and dimer in the presence of LDAO or ␤-D-dodecyl maltoside, the unliganded receptor is essentially dimeric when reconstituted with detergents such as hexadecyl-␤-D-maltoside ( Fig. 1). BLT1 was also found to be dimeric when reconstituted in proteoliposomes (not shown). This indicates that BLT1 is likely to be essentially dimeric but that the unliganded dimer is less stable than the agonist-loaded one so that it dissociates depending on the detergent used (see "Discussion"). Hexadecyl-␤-D-maltoside has the same head group as the widely used ␤-D-dodecyl maltoside detergent but displays a longer alkyl chain (16 and 12 atoms of carbon, respectively), and this may have an effect on dimer stability. Indeed, we previously showed that detergents with long hydrophobic chains stabilized the purified receptor better than detergents with short chain (23). We used in the following work hexadecyl-␤-D-maltoside that allows a maximal stabilization of the receptor dimer, even in the absence of agonist.
BLT1 C97A Mutant-We produced a receptor mutant where Cys 97 in the third TM domain has been substituted by an alanine. Cys 97 was identified on the basis of photolabeling experiments as being part of the receptor ligand-binding pocket and directly responsible for specific BLT1-LTB 4 interactions. 2 Substituting Cys 97 for an alanine does not affect the structural features of the complex, as inferred from circular dichroism measurements (data not shown). However, as expected on the 2 J.-L. Banères, manuscript in preparation. basis of the photolabeling data, the C97A mutant is characterized by an ϳ100-fold decrease in the affinity for LTB 4 (K d ϭ 2.4 ϫ 10 Ϫ7 and 1.8 ϫ 10 Ϫ9 M for the mutant and wild-type receptors, respectively) ( Fig. 2).
We then analyzed the ability of the C97A mutant of BLT1 to be specifically activated by the agonist. We previously established that agonist binding to a receptor mutant where all of the tryptophan residues, besides Trp 234 in TM6, had been replaced by leucines (W41L,W83L,W142L,W161L mutant) is associated with specific changes in the fluorescence emission spectrum of the receptor (23). These changes are probably due to a difference in the immediate environment of Trp 234 between the inactive and active states of the receptor. We introduced here the C97A mutation in the W41L,W83L,W142L,W161L mutant of BLT1. We previously showed that all the Trp 3 Leu mutations, with the exception of W234L, resulted in no significant alteration of the K a value of the receptor for its LTB 4 agonist (23). The W41L,W83L,C97A,W142L,W161L mutant therefore displays the same affinity for LTB 4 than the single C97A mutant. We then produced this mutant with 5HW instead of tryptophan at position 234 (see "Experimental Procedures"). As previously reported (20), the advantage of labeling the receptor with 5HW is that this residue can be selectively excited between 310 and 320 nm without affecting the other Trp residues (see excitation spectra of the unlabeled and labeled receptor in Fig. 3A). The 5HW-labeled W41L,W83L,C97A,W142L,W161L mutant has been systematically used throughout this work and is named R 0 , as opposed to the wild-type receptor that will be named R. We then recorded the fluorescence spectra of R 0 upon specific excitation of the 5HW residue at 315 nm in the absence or presence of LTB 4 . As shown in Fig. 3B, agonist binding to R 0 is associated with a significant increase in the fluorescence of the 5HW residue. The emission features observed upon binding of LTB 4 to R 0 are very similar to those we reported in the case of R (20), indicating that the same kind of agonist-induced conformational changes are likely to occur for the wild-type and mutant receptors. This effect is agonist-specific, since no significant change was observed upon binding of the 5b␣ antagonist (not shown). The changes in 5HW emission features induced by agonist binding therefore provide a convenient method for monitoring changes in R 0 conformation.
Finally, we analyzed the dimerization properties of R 0 by chemical cross-linking, as described above for the wild-type receptor. No difference was observed between the mutant and wild-type receptor at the dimerization level: R 0 , when reconstituted in hexadecyl-␤-D-maltoside, is essentially in a dimeric state in the absence or presence of its agonist (not shown). R 0 therefore represents a receptor with the same properties than the wild-type receptor in terms of activation and dimerization but with a decreased affinity for its agonist.
R⅐R 0 Dimer-In order to produce a receptor dimer containing a single R 0 molecule, we selectively labeled R and R 0 with two different purification tags, namely an S tag and a His tag. Briefly stated (see "Experimental Procedures" for details), both receptors were mixed before refolding in equimolecular amounts and then refolded. Since R 0 displays the same structural features and dimerization properties as the wild-type receptor (see above), one should expect, after refolding, a mixture of both R⅐R and R 0 ⅐R 0 homodimers as well as the R⅐R 0 "heterodimer." The latter was purified through two successive steps involving the two different affinity tags on R and R 0 (Fig.  4). The protein complex obtained after the second purification step on the nickel column (F3; see Fig. 4 for the definition of the fractions) displays two classes of agonist binding sites (Fig. 5). One is of high affinity, identical to that of the wild-type receptor, and the other displays an affinity close to, although slightly higher than (see "Discussion"), that of the C97A mutant. A series of cross-linking experiments were also carried out to assess the stoichiometric features of the protein complex in the F3 fraction. As expected, a single species was observed after chemical cross-linking with an electrophoretic mobility compatible with that of a receptor dimer (see inset in Fig. 5A), indicating that the major species obtained under such conditions is the dimeric assembly. All of these data indicate that the protein complex purified under such conditions is the R⅐R 0 "heterodimer." The interesting feature with the R⅐R 0 complex is that the difference in the affinity for LTB 4 between R and R 0 makes it possible to reach a state where only one of the protomers, R, is loaded with the agonist (see titration plot in Fig. 5). We confirmed R⅐R 0 dimerization using FRET. For these measurements, R 0 and R were labeled at their N terminus with a fluorescence donor (Alexa Fluor 350) and acceptor (Alexa Fluor 488) molecule, respectively. To label only the N terminus of the proteins, the labeling reaction was carried out in vitro at a pH value where only the N-terminal amine and not the lysil ⑀-amino group is protonated and therefore able to react with the probe (see "Experimental Procedures"). We then analyzed the fluorescence transfer properties of each of the protein complexes obtained from the two successive purification steps on the S-protein-agarose and Ni 2ϩ -nitrilotriacetic acid columns (see Fig. 4). As shown in Fig. 6, a significant energy transfer signal was observed only for the protein complex in the second peak eluted from the Ni 2ϩ -nitrilotriacetic acid column (F3). This indicates that both the wild-type and the mutant receptor are associated in the dimeric complex, confirming the conclusion inferred from the ligand-binding data.
G-protein Activation by the R⅐R 0 Dimer-We then checked whether the mutation in the R⅐R 0 complex affected the coupling to the G-protein. For this, we analyzed the ability of the heterodimer to catalyze GDT/GTP exchange at the level of G␣ i2 , as previously reported for the R⅐R homodimer (20). As shown in Fig. 7, the nonhydrolyzable GTP␥S analog does not significantly bind to G␣ i2 ␤ 1 ␥ 2 when the R⅐R 0 heterodimer is devoid of its agonist. In contrast, saturating the ligand-binding sites of R⅐R 0 with LTB 4 leads to the binding of GTP␥S to an amount comparable with that obtained with the wild-type homodimer (Fig. 7). As expected, no GTP␥S binding occurred in the presence of the 5b␣ antagonist. Moreover, as previously reported in the case of the R⅐R homodimer, this is a specific effect, since nearly no G-protein activation was observed when G␣ s was used instead of G␣ i (not shown). As stated above, the GTP␥S binding profiles are similar for the R⅐R and R⅐R 0 complexes, indicating that the R⅐R 0 heterodimer is able to activate the G-protein in the same way as the R⅐R homodimer does.
Activation of R 0 in the R⅐R 0 Dimer-We then analyzed the activation of R 0 in the R⅐R 0 dimer at different LTB 4 concentrations. As described above, in this "heterodimer," only the R 0 protomer includes a 5HW residue. Under such conditions, it is possible to selectively follow the conformational changes of R 0 in the R⅐R 0 complex by measuring the fluorescence variations upon specifically exciting the 5HW residue at 315 nm, with no contribution of the tryptophan residues in R. We therefore carried out a parallel titration experiment by measuring both LTB 4 binding to R⅐R 0 and the changes in the emission properties of 5HW in R 0 . As clearly shown in Fig. 8, filling the high affinity binding sites in R with LTB 4 is associated with a significant increase in the emission of 5HW in R 0 . Fitting this fluorescence variation to a single-site hyperbolic function showed EC 50 closely related to the affinity value determined for the high affinity sites in R from direct LTB 4 binding measurement. This strongly indicates that the effects observed on the emission properties of 5HW in R 0 are the direct consequence of the binding of LTB 4 to R. It is to be noted that in this state, the R protomer is, as expected, in the active conformation (intrinsic fluorescence data with an R⅐R 0 complex where only R contains 5HW; not shown). A subsequent change in the emission properties of 5HW, of lesser amplitude, was observed after loading the low affinity sites in R 0 with LTB 4 , indicating that a subsequent conformational adaptation occurs after binding of the agonist to R 0 in the R⅐R 0 dimer. In contrast to what is observed FIG. 4. Purification of the R⅐R 0 "heterodimer." Schematic representation of the method used to produce R⅐R 0 . The S tag at the R N terminus and the His tag at the R 0 N terminus are represented by the closed square and the closed circle, respectively. The first purification step on the S-protein-agarose yields the R 0 ⅐R 0 complex in the excluded peak (labeled here F1) and the R⅐R homodimer and the R⅐R 0 "heterodimer" in the protein fractions retained on the column; the second purification step on the Ni 2ϩ -nitrilotriacetic acid column allows the separation between the R⅐R dimer found in the excluded fractions (F2) and the R⅐R 0 complex retained on the column (F3). with LTB 4 , no change in the emission properties of 5HW was observed upon binding of an antagonist molecule (5b␣) to the R⅐R 0 dimer (not shown), indicating that the effects observed above were specific to the agonist. All of these observations clearly show that specific changes in the conformation of the R 0 protomer in the R⅐R 0 complex occur upon activation of R. DISCUSSION We have used here a purified receptor, the leukotriene B 4 receptor BLT1, to analyze the activation mechanism in a GPCR dimer. Since the isolated receptor is likely to recapitulate what occurs in a lipid bilayer with regard to ligand binding and receptor activation (20,23), it appears as a convenient in vitro model system to analyze the relationship between dimerization and activation. By using a receptor dimer where one of the protomers displays a lower affinity for the agonist, we show that agonist-induced activation of one of the monomers in the dimeric assembly induces specific changes in the conformation of the unliganded monomer.
The data presented in this work indicate that detergentsolubilized BLT1 is essentially a dimer independently of the presence of the ligand, provided we use a detergent that fully stabilizes the dimeric assembly. In agreement with this conclusion, the detergent-isolated receptor reconstituted in a lipid environment is dimeric whatever the conditions are (not shown). This is consistent with the view that, at least for class A GPCRs with small ligands that bind within the seven transmembrane helix bundle, ligand binding does not significantly alter dimerization (25)(26)(27). Our data suggest that the dimer composed of the receptor in its inactive state is less stable than that in the active state, since it dissociates in vitro depending on the detergent used. This could indicate that the active and inactive dimers involve different protein/protein contacts, probably as a consequence of the conformational changes associated with receptor activation. However, in the absence of accurate structural data on the BLT1-BLT1 assembly, it is difficult to further discuss our observations on a firm molecular basis.
To monitor BLT1 activation, we have used the variations in the emission properties of Trp 234 in TM6. These changes are specifically associated with the agonist-induced changes in receptor conformation (23). Agonist binding to the purified receptor is accompanied by a significant increase in the fluorescence emission intensity of Trp 234 . As previously discussed (23), this increase could be due to a decreased polarity of the immediate surrounding of Trp 234 in the active conformation of the receptor. Another explanation would be that LTB 4 binding increases the population of Trp 234 conformational substates with higher lifetimes, and this could also account for the increase in fluorescence intensity (28). Whatever the exact explanation is, the agonist-induced change in the fluorescence properties of Trp 234 provides us with a probe to specifically follow the changes in the conformation of the BLT1 receptor. To specifically follow the activation of R 0 with no contribution of the Trp residues in R, we selectively labeled R 0 with 5HW at position 234. As stated under "Results," the advantage of this strategy is that 5HW can be selectively excited between 310 and 320 nm in the presence of several Trp residues. The fluorescence of the single 5HW in R 0 can therefore be selectively monitored in the presence of the five Trp residues of R. Under these conditions, 5HW appears as a site-specific probe for conformational modifications in R 0 .
We then produced a "heterodimer" containing a wild-type protomer (R) and a protomer (R 0 ) whose affinity for LTB 4 is decreased as the result of the mutation of Cys 97 . This residue has been identified on the basis of photolabeling experiments as being part of the receptor ligand-binding pocket and directly responsible for specific BLT1-LTB 4 interactions. 2 This certainly explains the decrease in the affinity of the receptor for its agonist observed for the C97A mutant. The existence of the R⅐R 0 complex was confirmed by FRET. Indeed, a significant fluorescence transfer was observed when R 0 and R were labeled at their N terminus with a fluorescence donor and acceptor, respectively. It is to be noted that a difference in FRET efficiency was observed between the unliganded and agonistloaded complexes (not shown), although no change was observed in the dimer amount, as assessed by chemical crosslinking. This indicates that, as previously discussed (4), the changes in the fluorescence transfer properties observed upon agonist binding are likely to be associated with a different relative arrangement of the protomers in the inactive and active complexes. Such differences in the topological features of the dimer could be related to the difference in the stability between the unliganded and the agonist-loaded BLT1 dimer.
We also analyzed whether the R⅐R 0 heterodimer was able to activate the G-protein in the same way as the R⅐R dimer. We previously showed that, upon saturation of the BLT1 homodimer ligand-binding sites with LTB 4 , GDP-GTP exchange occurs at the level of G␣ i2 , as assessed by GTP␥S binding (20). No difference was observed in this work between the R⅐R and R⅐R 0 dimers in the ability to activate the G-protein, indicating that the dimeric complex with one mutated subunit at the level of the ligand-binding site is still fully able to interact, in a specific manner, with its G-protein partner. These data clearly indicate that the heterodimeric species produced here displays the same characteristics as the wild-type homodimer in terms of coupling to the G-protein partner. It is to be noted that this observation opens the way to the analysis of the relationship between the different steps in the conformational adaptation of the receptor, as described here, and the interaction with the G-protein (work in progress).
The R⅐R 0 "heterodimer" obtained under such conditions displays two classes of agonist binding sites corresponding to the ligand-binding sites in R (high affinity site) and R 0 (low affinity site). The difference in the relative affinities for the agonist of the two protomers within the same dimeric complex allowed a sequential filling of the agonist-binding sites of R⅐R 0 . It is to be noted that a 2-3-fold increase in the affinity of R 0 for LTB 4 in the R⅐R 0 dimer was observed in comparison with what is measured for the R 0 ⅐R 0 homodimer. This could be due to the changes in R 0 conformation upon filling the ligand-binding sites in R (see below). Indeed, a careful analysis of the LTB 4 binding profiles in Fig. 2 clearly indicates that some positive cooperat- ivity occurs in the binding of LTB 4 to the stabilized receptor dimer. These cooperative effects, as well as their molecular significance, were not further investigated in this work and will not be discussed here. This will require a more detailed ligand binding analysis with the isolated dimeric receptor in the absence and presence of G-proteins (work in progress).
The data presented here show that agonist-induced activation of one of the protomers in the receptor dimer induces specific changes in the conformation of the unliganded protomer, as assessed by the changes in the fluorescence properties of 5HW in the TM6 segment of the receptor. An alternative explanation would be to consider a domain-swapped model for the BLT1 dimer. In this model, the TM3 domain with the C97A mutation and the TM6-labeled domain would be located in two different protomers as a result of domain swapping. However, we have built a series of "heterodimeric" species with rescuing mutations within TM3 and TM6. No evidence compatible with a swapped geometry of the BLT1 dimer has been obtained with these species. 3 Moreover, the same results were obtained when both the affinity-decreasing mutation and the 5HW residue were introduced in the same TM domain (not shown). All of these data indicate that the effects observed here certainly are not the consequence of domain swapping.
The question that arises is whether the conformation of R 0 after filling the ligand binding sites in R with LTB 4 is that of the ligand-loaded active state of the receptor. The changes in the fluorescence properties of 5HW observed upon binding of the agonist to R in the R⅐R 0 dimer are of the same nature as those observed upon direct binding of the agonist to the R 0 protomer, indicating that the conformation of R 0 in the halfloaded dimer could be close to the active one. However, it must be emphasized that the fluorescence changes are associated with very local changes in the environment of the Trp 234 residue, so that different conformations could lead to closely related fluorescence spectra. Moreover, a subsequent change, although of minor amplitude, was observed in the 5HW fluorescence features after the filling of the R 0 ligand-binding sites in R⅐R 0 . This clearly indicates that a subsequent conformational adaptation step, associated with agonist binding to the receptor, is required to reach the fully active conformation. The fact that a given protomer does not reach its fully activated state upon binding of the agonist to the other protomer in the dimeric assembly could explain recent observations with the vasopressin receptors where the identity of the activated protomer within the heterodimer determines the fate of the internalized receptors (29).
Another aspect that has also to be considered is the kinetic one. It is important to emphasize that the experiments reported here have all been carried out under thermodynamic equilibrium conditions. Therefore, we cannot ascertain if the kinetic features of receptor activation through the binding of a single ligand to the dimer are similar to those resulting from the filling of both agonist-binding sites. This aspect of receptor activation will require a time-resolved analysis of receptor activation in the dimer (work in progress).
On these bases, the agonist half-loaded dimer could represent an intermediate conformational state between the inactive and fully activated dimeric assembly, with one of the protomers fully activated and the other one in a conformation close to the active one. If this is the case, such a cooperative conformational adaptation in the receptor dimer could facilitate the formation of a fully active GPCR dimer and/or lead to a higher diversity in the coupling to different intracellular partners that could be dependent on different conformational states of the receptor dimeric assembly.
In the absence of accurate data on the structural arrangement of the receptor dimer, it is difficult to discuss the possible origins of the cooperative structural changes observed here on a firm molecular basis. So far, trans-activation has been essentially shown to occur in class C receptors (30,31). These receptors are characterized by large extracellular domains that bind the agonist (1,31). Trans-activation has also been suggested to occur for a class A receptor, the LH/CG receptor (17,18). However, the LH/CG receptor is also characterized by two domains, an extracellular N-terminal exodomain that binds the hormone (32) and a membrane-associated C-terminal endodomain (33). In both cases, the underlying mechanisms for activation are likely to be different from what is observed here with BLT1. For the GABA B receptor, it is likely that the change in the respective orientation of the extracellular domains that occurs upon agonist binding to one of the subunits changes the interacting mode of the transmembrane domains, thus stabilizing the active conformation of the other subunit (31). LH/CG receptor trans-activation has been hypothesized to occur through a mechanism where a liganded exodomain trans-activates the endodomain of another unliganded receptor (17). The 3 J.-L. Banères, unpublished data.  4 to the R⅐R 0 complex (closed squares) and the percentage change in 5HW fluorescence (open squares) in the R⅐R 0 complex was monitored as a function of LTB 4 concentration. The binding ratio is the ratio of bound LTB 4 per receptor dimer. Note that only R 0 in R⅐R 0 is labeled with 5HW. The LTB 4 binding and fluorescence measurements were carried out as described under "Experimental Procedures." situation is less clear in the case of class A receptors where the ligand binds directly to the transmembrane helix bundle. As stated above, all of the data presented here are consistent with the idea that in such a receptor, BLT1, activation of one of the protomers induces some changes in the conformation of the other one in the dimer. A possibility would be that the changes in the orientation of the TM domains induced in one protomer by the binding of the agonist lead to changes in the protein/ protein interface in the dimer that would induce a change in the orientation of the TM6 domain of the other protomer. However, the exact molecular explanation for such a mechanism will require more structural information on the topological features of the BLT1-BLT1 dimer.
Finally, whether both protomers need to be in the active state to bind their ligand to activate the G-protein remains an open question. A receptor dimer with a single functional subunit can lead to in vivo signaling (19), indicating that activation of the receptor dimer could occur even when a single protomer is in the agonist-loaded state. As shown in this work, the conformational features of the unliganded protomer in the halfloaded receptor do not appear to be strictly identical to those in the fully active state. However, if one considers a complex composed of a single G-protein per receptor dimer (20), it could be possible that a single activated receptor selectively interacts with its G␣-protein partner, whereas the other protomer would only ensure the stability of the complex, by interacting with either a different region of G␣ or with the ␤␥ complex, as recently proposed (34,35). If this is the case, activation of the G␣-interacting receptor protomer could be sufficient for triggering a specific ligand-induced response. However, all of these models will require more molecular data on the receptor-Gprotein complex to be obtained (work in progress with the isolated receptor and purified G-proteins).