Allosteric Transinhibition by Specific Antagonists in CCR2/CXCR4 Heterodimers*

Chemokine receptors are presently used as targets for candidate drugs in the frame of inflammatory diseases and human immunodeficiency virus infection. They were shown to dimerize, but the functional relevance of dimerization in terms of drug action remains poorly understood. We reported previously the existence of negative binding cooperativity between the subunits of CCR2/CCR5 heterodimers. In the present study, we extend these observations to heterodimers formed by CCR2 and CXCR4, which are more distantly related. We also show that specific antagonists of one receptor inhibit the binding of chemokines to the other receptor as a consequence of their heterodimerization, both in recombinant cell lines and primary leukocytes. This resulted in a significant functional cross-inhibition in terms of calcium mobilization and chemotaxis. These data demonstrate that chemokine receptor antagonists regulate allosterically the functional properties of receptors on which they do not bind directly, with important implications on the effects of these potential therapeutic agents.

Chemokine receptors constitute an attractive family of drug targets in the frame of inflammatory diseases, human immunodeficiency virus, and cancer (1,2). However, the complexity and the functional redundancy of the chemokine system have represented significant hurdles in this development. There is therefore a need for a better comprehension of the chemokine signaling network and how molecules (agonists and antagonists) acting on specific receptors will affect the global function of the chemokine system. One of the challenges in this area is the recent demonstration that G protein-coupled receptors, including chemokine receptors, are able to form dimers, and the functional consequences of this new concept have not been fully appreciated so far.
CCR2 is a member of the CC chemokine receptor family that plays an important role in the recruitment of monocytes to atherosclerotic lesions and in the formation of intimal hyperplasia after arterial injury (3). Two isoforms of CCR2 have been described that are generated by alternative splicing and differ in their C-terminal tail, but only the longer variant (CCR2b) is expressed at high level in leukocytes. CCR2 interacts with MCP-1 (CCL2) and other CC chemokines such as MCP-2 (CCL7), MCP-3 (CCL8), and MCP-4 (CCL13) (4). CCR2 was shown to form homodimers, as well as heterodimers with CCR5, its closest structural relative within the chemokine receptor family, resulting in a negative binding cooperativity of allosteric nature (5)(6)(7)(8). CCR2 was also reported recently to form heterodimers with CXCR4, a more distantly related chemokine receptor (9). However, no pharmacological characterization of this heterodimer was performed so far. CXCR4 binds specifically SDF-1␣, an evolutionary conserved chemokine known to play a role in a number of physiological processes, such as the homing of T cell populations to sites of inflammation and the maintenance of the cellular microenvironment in bone marrow (10). SDF-1␣ and CXCR4 have also been shown to control the migration of neuronal cell populations during brain development (11), to regulate the growth and/or survival of tumoral cells (12,13), and to play a role in the metastatic spread of various tumors (14,15). In addition, CXCR4 is the main coreceptor for the T cell line tropic human immunodeficiency virus, type 1 strains (16). CXCR4 and CCR2 share only 34% of amino acid identity over their entire length. The ligand binding specificity of CCR2 and CXCR4 has been mapped to the extracellular domains of the receptors, particularly the second extracellular loop (17). Both CXCR4 and CCR2 are expressed by memory T lymphocytes and the monocyte-macrophage lineage (18,19). As other chemokine receptors, they are coupled to the G␣ i class of heterotrimeric G proteins, inhibit adenylyl cyclase, promote intracellular calcium mobilization, stimulate the mitogen-activated protein kinase and phosphatidylinositol 3-kinase pathways, and promote chemotaxis. With the aim of understanding better the functional consequences of chemokine receptor oligomerization, we have generated cells coexpressing CCR2 and CXCR4 and characterized them in terms of binding and signaling properties. We confirmed that CCR2 and CXCR4 form constitutive heterodimers, which appear unaffected by chemokine stimulation. As previously described for the CCR2/CCR5 heterodimer, negative binding cooperativity was observed between CCR2 and CXCR4 when coexpressed, a receptor heterodimer binding only a single chemokine with high affinity. In addition, it was * This work was supported by the Actions de Recherche Concertées of the shown that specific antagonists of one receptor can inhibit the binding of chemokines to receptors on which these antagonists do not bind, resulting in a cross-inhibition of the functional response, and these observations were attributed as well to an allosteric interaction between the two binding sites of a dimer. Finally, similar observations were made on primary T cells ex vivo, demonstrating the functional relevance of CCR2/CXCR4 heterodimers in native cells.

EXPERIMENTAL PROCEDURES
Antibodies-Phycoerythrin-labeled anti-CXCR4 (MAB173), anti-CCR2 (FAB151P) monoclonal antibodies, and recombinant chemokines were obtained from R & D Systems. The DOC-1 anti-CCR2 antibody was kindly provided by Matthias Mack (University of Regensburg, Regensburg, Germany). TAK-779 was obtained from the National Institutes of Health AIDS Research and Reference Reagent Program, Division of AIDS, NIAID and AMD3100 from Sigma-Aldrich.
Cell Lines and Leukocyte Populations-CHO-K1 cells were cultured in Ham's F12 medium supplemented with 10% fetal bovine serum (Invitrogen), 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen). The CCR2 coding sequence was cloned between the BamHI and XbaI sites of the bicistronic expression vector pEFIB3, as described (20), and the construct was transfected by FuGENE (Roche Applied Science) into a CHO-K1 cell line expressing apoaequorin, G␣ 16 and wild-type CXCR4. Cells expressing CCR2 were selected by 10 g/ml blasticidin (Invitrogen). Human peripheral blood lymphocytes were prepared from buffy coats of healthy blood donors as described previously (8). CD4 ϩ blasts were generated by incubating the lymphocytes with anti-CD3 (1:100) and anti-CD28 (1:1000) antibodies for 3 days. The cells were maintained in a medium supplemented with human interleukin 2 (2 ng/ml; R & D Systems) for an additional 7 days.
Bioluminescence Resonance Energy Transfer Assays-The cDNAs encoding EYFP 3 and a humanized form of Renilla luciferase were fused in frame to the 3Ј end of CXCR4 and CCR2 cDNAs in the pcDNA3.1 vector. A BRET protocol adapted to cell monolayers was used (21). Briefly, human embryonic kidney cells (HEK-293T) were transfected by the calcium phosphate precipitation method, using a constant amount of plasmid DNA but various ratios of plasmids encoding the fusion protein partners (22). A control corresponding to mock transfected cells was included to subtract raw basal luminescence and fluorescence from the data. Expression of EYFP fusion proteins was estimated by measuring fluorescence at 535 nm following excitation at 485 nm, using a Mithras LB940 multilabel reader (Berthold). Expression of Rluc fusion proteins was estimated by measuring the luminescence of the cells after incubation with 2.5 M coelenterazine H (Promega). In parallel, BRET was measured as the fluorescence of the cells at 535 nm at the same time points. The BRET ratio is defined as [(emission at 510 -590)/(emission at 440 -500)] Ϫ C f where C f corresponds to (emission at 510 -590)/(emission at 440 -500) for the ϪhRluc construct expressed alone in the same experiment.
Binding Assays-Competition binding experiments were performed as previously described (7). Briefly, the membrane preparations were incubated in the assay buffer (50 mM Hepes, pH 7.4, 1 mM CaCl 2 , 5 mM MgCl 2 , 0.5% BSA) with 0.2 nM 125 I-SDF-1␣ or 0.1 nM 125 I-MCP-1 as tracers and variable concentrations of unlabeled competitors. The samples were incubated for 1 h, and bound tracer was separated by filtration through GF/B filters presoaked in 1% BSA. The filters were counted in a ␥-scintillation counter. Binding parameters were determined with the PRISM software (Graphpad Softwares) using nonlinear regression applied to single-site or two-site binding models. The software compared the sum-of-square and the degree of freedom of each regression by using the F test and selected the most appropriate equation.
Dissociation Kinetics Experiments-Ligand dissociation experiments were performed as previously described (8). The membrane preparations were first incubated in assay buffer (50 mM Hepes, pH 7.4, 1 mM CaCl 2 , 5 mM MgCl 2 , 0.5% BSA) with 0.1 nM 125 I-MCP-1 or 125 I-SDF-1␣ in a final volume of 500 l. After 1 h, the membranes were centrifuged, and the unbound radioligand was removed by aspiration. The membrane pellet was washed once with assay buffer and resuspended in 2.5 ml of assay buffer, with or without unlabeled ligands. At different time points, the aliquots were collected, the bound tracer was separated by filtration through GF/B filters presoaked for 1 h in 1% BSA, and the filters were counted for 1 min in a ␥-scintillation counter. In all experiments, total binding and total tracer remaining at the initiation of the dissociation phase represented less than 10% of the amount of tracer engaged initially. The data are presented as the ratio between bound cpm at the various dissociation time points and total bound cpm at time 0 of dissociation. The curves were fitted with PRISM software (Graphpad Softwares) using nonlinear regression and a single phase decay model.
Intracellular Calcium Mobilization Assay-The functional responses were analyzed with an aequorin-based assay as described (23). The cells were incubated for 4 h in the dark in the presence of 5 M coelenterazine H (Promega Corporation) and then diluted 5-fold before use. Variable concentrations of chemokines were added to cell suspension (25,000 cells/well), and luminescence was measured for 30 s in an EG & G Berthold luminometer (PerkinElmer Life Sciences). Half-maximal effective concentrations (EC 50 ) were determined with the PRISM software (Graphpad Softwares) using nonlinear regression applied to a sigmoidal dose-response model.
Chemotaxis Assay-The migration of CD4 ϩ T cells was performed in 96-well transwell chambers (5-m pore size; Costar, Corning NY). The lower compartment of the chambers was loaded with serial dilutions of SDF-1␣ or MCP-1 in Dulbecco's modified Eagle's medium-Ham's F-12 medium supplemented with 0.1% BSA. The upper compartment was loaded with 5 ϫ 10 5 cells preincubated or not with specific antagonists, and migration was allowed for 1 h at 37°C. The cells were harvested in the lower chamber and counted by the ATPlite luminescence assay kit (PerkinElmer Life Sciences). The results are expressed as the chemotaxis index, i.e. the ratio of cells migrating in response to the chemoattractant over cells migrating toward medium.

Heterodimerization of CCR2 and CXCR4 in Recombinant
Cell Lines-In previous studies, we showed that CCR2 forms constitutive homodimers, as well as heterodimers with its close structural relative CCR5 (6,7). The functional consequence of this heterodimerization was a negative binding cooperativity between the CCR5 and CCR2 ligands, as a result of an allosteric interaction between the two binding sites of the heterodimer (8). In this work, we investigated further the ability of CCR2 to interact with more distant members of the chemokine receptor family and whether a negative binding cooperativity could apply as well for other such heterodimers. In a first step, we investigated by a BRET technique the extent of heterodimerization between CCR2 and CXCR4 in living cells. As described previously, a constitutive energy transfer was measured between CXCR4-hRluc and CXCR4-EYFP and between CCR2-hRluc and CCR2-EYFP expressed in HEK293T cells (7,9) (not shown). A constitutive energy transfer was also observed between CXCR4-hRluc and CCR2-EYFP, as well as between CCR2-hRluc and CXCR4-EYFP. The BRET 50 and BRET max parameters of energy transfer for CCR2/CXCR4 heterodimers were similar to those reported for homodimers (Table 1 and supplemental Fig. S1). Stimulation by chemokines specific to CCR2 or CXCR4 did not change the BRET 50 values but modified to some extent the BRET max . Indeed, MCP-1 increased the maximal energy transfer between CCR2-hRluc and CXCR4-EYFP but not between CXCR4-hRluc and CCR2-EYFP. These data are similar to those reported previously (9) and support the conclusions following which the stimulation by chemokines modifies the conformation of preformed CCR2/CXCR4 dimers rather than the amount of dimers.
Construction and Characterization of Cell Lines Coexpressing CXCR4 and CCR2-To study the functional consequences of the heterodimerization, CHO-K1 cell lines stably coexpressing CXCR4 and CCR2 were isolated and analyzed for the level of expression of the two receptors by FACS and saturation binding assays (data not shown). The parental CHO-K1 cell line expressing CXCR4 (B max ϭ 3.1 Ϯ 0.1 pmol/mg membrane proteins) was used as the recipient for CCR2 coexpression, and the C2X4 -40 clone expressing both receptors was selected for further analysis. The estimated K D values for CXCR4 and CCR2 measured on this clone (respectively 0.22 Ϯ 0.03 nM and 0.16 Ϯ 0.03 nM) are consistent with values described previously for cells expressing single receptors (0.20 Ϯ 0.03 nM for CXCR4 and 0.15 Ϯ 0.01 nM for CCR2) (7). In saturation binding assays, the B max was estimated to 3.2 Ϯ 0.07 pmol/mg proteins for CXCR4 and to 2.5 Ϯ 0.04 pmol/mg proteins for CCR2. FACS analysis demonstrated that the selected clones were homogeneous in terms of receptor expression and regular testing confirmed the stability of this expression over time.
Binding Properties of Cells Expressing CXCR4 and CCR2-We next examined the ability of the ligands of each receptor to compete for 125 I-SDF-1␣ and 125 I-MCP-1 binding to membranes of cells expressing CXCR4 and/or CCR2. As expected, SDF-1␣ inhibited with high affinity the binding of 125 I-SDF-1␣ on membranes containing CXCR4 only, whereas MCP-1 had no effect (Fig. 1A). Similarly, MCP-1 but not SDF-1␣ competed efficiently the binding of 125 I-MCP-1 on membranes containing exclusively CCR2 (Fig. 1B). A drastic modification of the competition patterns was observed when CXCR4 and CCR2 were coexpressed. In the 125 I-SDF-1␣ binding assay, SDF-1␣ behaved similarly, but MCP-1 now competed partially (IC 50 ϭ 0.02 Ϯ 0.01 nM) (Fig. 1C). Conversely, in the 125 I-MCP-1 binding assay, competition by MCP-1 was unaffected, whereas a significant competition by SDF-1␣ was observed (IC 50 ϭ 0.78 Ϯ 0.09 nM) (Fig. 1D). The IC 50 values calculated for heter-  ologous competition were similar to those obtained for homologous competition in cells expressing single receptors ( Table  2). Heterologous competitions were, however, incomplete, representing ϳ35% of the specific binding. These results suggest that about one-third of each receptor takes part with heterodimers and that within these heterodimers, ligands specific for one receptor are able to compete for the binding of the tracer on the other. We next tested the ability of the CXCR4 antagonist AMD3100 and the CCR2 inverse agonist TAK-779 to promote similar heterologous competition. As expected, AMD3100 and TAK-779 did not compete for respectively 125 I-MCP-1 and 125 I-SDF-1␣ binding when the receptors were expressed alone (Fig. 2, A and B). However, AMD3100 was able to compete for 125 I-MCP-1 binding as efficiently as SDF-1␣ on membranes prepared from cells coexpressing CXCR4 and CCR2 (Fig. 2D). Similarly, TAK-779 efficiently competed the binding of 125 I-SDF-1␣ in the same conditions (Fig. 2C). Yet again, the IC 50 values were similar to those obtained for cells expressing single receptors (Table 2), and the extent of the competition reached about 35% of the specific binding. The fact that small molecule antagonists can behave similarly to chemokines in these cross-competition experiments indicates that receptors do not need to reach their active state to inhibit the binding capacity of their heterodimerization partner. These results also suggest that the observed phenomenon is not linked to the steric occupancy of both receptors by a single ligand. We investigated further the potential contribution of steric hindrance at the outer surface of the receptor heterodimers by testing anti-CXCR4 and anti-CCR2 monoclonal antibodies in binding assays. The anti-CXCR4 MAB173 prevented 125 I-SDF-1 binding on membranes containing CXCR4 or the two receptors but did not affect 125 I-MCP-1 binding. Similarly, the anti-CCR2 DOC-1 prevented 125 I-MCP-1 binding but not 125 I-SDF-1 binding (supplemental Fig. S2).
To characterize further the negative binding cooperativity between CXCR4 and CCR2, the rate of radioligand dissociation from heterodimers, in the presence or absence of chemokines or antagonists, was assayed in "infinite" tracer dilution conditions (8). We showed that, on membranes prepared from cells expressing CCR2 only, dissociation of prebound 125 I-MCP-1 was slow in the absence of unlabeled ligands (t1 ⁄ 2 Ͼ 200 min; Fig. 3 E and Table 3), but complete dissociation was observed after 24 h (not shown). The dissociation rate of 125 I-MCP-1 was increased when CCR2 and CXCR4 were coexpressed (t1 ⁄ 2 ϭ 27.6 min; Fig. 3F). Conversely, the presence of CCR2 increased significantly the dissociation rate of prebound 125 I-SDF-1␣ from CXCR4 (t1 ⁄ 2 ϭ 35.6 min versus t1 ⁄ 2 Ͼ 200 min; Fig. 3, A and B). As expected, we showed that the specific CXCR4 ligands SDF-1␣ and AM3100 had no effect on the dissociation of bound 125 I-MCP-1 from membranes containing CCR2 only. However, both molecules accelerated the dissociation of 125 I-SDF-1␣ from CXCR4, but also of 125 I-MCP-1 when CXCR4 and CCR2 were coexpressed (t1 ⁄ 2 ϭ 14.7 and 23.0 min, respectively; Fig. 3, F  and H). Similarly, MCP-1 and TAK779 promoted a rapid dissociation of 125 I-MCP-1 from CCR2 and of 125 I-SDF1␣ when CXCR4 and CCR2 were coexpressed (t1 ⁄ 2 ϭ 15.4 and 13.8 min, respectively; Fig. 3, B and D). These data suggest that the stability of the interaction of CXCR4 and CCR2 with their respective chemokine ligands is lower in heterodimers than in homodimers. In addition, the binding of a chemokine agonist or a small molecule antagonist to one of the partners in a heterodimer modifies further the conformation in the other partner, resulting in a faster dissociation rate of the tracers.
Functional Properties of Cells Coexpressing CXCR4 and CCR2-We first compared the functional response of cells coexpressing CXCR4 and CCR2 to cell lines expressing either CCR2 or CXCR4. The concentration-action curve of calcium mobilization following stimulation by MCP-1 was similar for cells coexpressing both receptors or expressing CCR2 only.  Similarly, the functional response to SDF-1␣ was identical in CXCR4-expressing cells whether CCR2 was coexpressed or not (supplemental Fig. S3). We also showed that the costimulation by MCP-1 and SDF-1␣ at equimolar concentrations resulted in cells coexpressing the two receptors, in a functional response essentially similar to that of MCP-1, which is the ligand displaying the highest potency on its receptor. These data indicate the absence of cooperative signaling between CCR2 and CXCR4 in this calcium mobilization assay (supplemental Table S1). We next investigated the effect of specific antagonists in cells coexpressing both receptors. In line with the binding data, AMD3100 antagonized CXCR4, but also CCR2 signaling, whereas TAK-779 inhibited both CCR2 and CXCR4 signaling, demonstrating the ability of antagonists to inhibit the signaling of receptors on which they do not bind directly (Fig. 4).
Functional CCR2-CXCR4 Heterodimerization in Native Cells-Finally, we investigated whether negative binding cooperativity could be demonstrated in cells expressing CCR2 and CXCR4 endogenously. Human CD4 ϩ T lymphocytes were isolated and activated with anti-CD3 antibodies (OKT3) and interleukin 2. Specific 125 I-MCP-1 and 125 I-SDF-1␣ binding could be   1 ratio, f). The functional response of CHO-K1 cell lines expressing CCR2 (C) or both receptors (D) was measured using the aequorin-based functional assay and stimulated with MCP-1 alone (F), MCP-1 ϩ AMD-3100 (1:1 ratio, Ⅺ), or MCP-1 ϩ TAK-779 (1:1 ratio, f). Luminescence was recorded for 30 s. The results were normalized for base-line activity (0%) and the maximal response obtained (100%). The displayed data are representative of three independent experiments. All of the points were run in triplicate (error bars, S.E.).

TABLE 3 Dissociation rates from CHO-K1 cells expressing CCR2 and/or CXCR4
The binding parameters were measured on CHO-K1 cells expressing CCR2 and/or CXCR4. The rate of dissociation (half-life) of prebound tracer were obtained from kinetics experiments as displayed in Fig. 3. The values represent the means Ϯ S.E. of at least three independent experiments. detected on these cells, as demonstrated by the complete competition obtained with blocking antibodies for each receptor (Fig. 5). Expression of both receptors in the same cells was also confirmed by FACS (data not shown). On these cells coexpressing CCR2 and CXCR4, 125 I-MCP-1 binding was inhibited by unlabeled MCP-1, MCP-3, and TAK-779, but also partially by SDF-1␣ and AMD-3100, demonstrating cross-competition by CXCR4-specific ligands (Fig. 5A). Conversely, a partial inhibition of 125 I-SDF-1␣ binding was observed in the presence of MCP-1 and TAK-779 (Fig. 5B). Interleukin 8 (CCL8), a chemokine acting on CXCR1 and CXCR2, which are not present on T cells, did not affect the binding of the various tracers. To test for functional consequences of this cross-competition, we performed chemotaxis experiments on these lymphoblasts. Migration toward SDF1-␣ or a combination of SDF1-␣ and MCP-1 at equimolar concentrations was similar (data not shown), suggesting the absence of cooperative signaling in this setting. Cell migration toward MCP-1 was inhibited by TAK-779 but also partially by the CXCR4-specific antagonist AMD3100. Simi-larly, migration toward SDF-1␣ was fully blocked by AMD3100 and partially by TAK-779 (Fig. 5, C and D). Altogether, these data indicate that CCR2-CXCR4 heterodimers do exist in native cells and that negative binding cooperativity occurs between the two binding pockets of these receptor heterodimers. Importantly, these data also demonstrate that chemokine receptor heterodimerization results in a cross-inhibitory effect of small molecule antagonists on the functional responses of receptors on which these antagonists do not bind.

DISCUSSION
During the past few years, the number of reports demonstrating GPCR dimerization has increased tremendously, and it is presently well accepted that most GPCRs are able to form homodimers (24). Chemokine receptors make no exception to this rule, and dimerization was reported so far for four chemokine receptors: CCR2, CCR5, CXCR2, and CXCR4 (8). Coimmunoprecipitation, BRET, and fluorescence resonance energy transfer experiments have shown that CCR2 and CCR5 form both homo-and heterodimers constitutively (7). Although ligand-promoted dimerization was suggested at some point (25), it was later demonstrated that stimulation by chemokines did not affect the dimerization status of either homo-or heterooligomers (26). In a previous study, we have demonstrated that heterodimerization between CCR2 and CCR5 resulted in a strong negative binding cooperativity of allosteric nature (7,8).
To extend the concept of allosteric interactions between dimer units, we showed by BRET that CCR2 and CXCR4 were able to form constitutive homo-and heterodimers and investigated the influence of this process on the pharmacological properties of both receptors. Following coexpression of CCR2 and CXCR4, we observed negative cooperativity interactions between the two receptor-binding pockets, very similar to what we reported previously for the CCR5/CCR2 heterodimer. The CXCR4-specific ligands inhibited MCP-1 binding when CXCR4 and CCR2 were coexpressed. Conversely, CCR2-specific ligands competed for SDF-1␣ binding only in cells coexpressing both receptors. The extent of this cross-competition (35%) is compatible with the expected proportion of receptors involved in the formation of heterodimers, considering similar expression levels for the two receptors, and no major differences in their relative tendency to form homo-or heterodimers. Dissociation kinetics after extensive tracer dilution showed that chemokines specific for one receptor can promote the dissociation of the tracer bound on the other receptor, demonstrating unambiguously the allosteric nature of the negative binding cooperativity. Interestingly, the coexpression of CCR2 and CXCR4 in CHO-K1 cells increased the spontaneous dissociation rate of both SDF-1␣ and MCP-1. Such changes in basal dissociation rates were not detected when CCR5 was coexpressed with CCR2 (8) and reflect probably the modification of the conformation of the receptors, according to the partners with which they interact. Additional studies will be required to determine whether such changes in kinetics properties can be extended to other chemokine receptor heterodimers.
Our study also reveals unexpected properties of chemokine receptor antagonists. AMD3100 is a bicyclam molecule described as a specific CXCR4 antagonist that inhibits the bind- FIGURE 5. Competition binding and chemotaxis assays on CD4 ؉ lymphoblasts. Competition binding assays were performed on CD4 ϩ lymphoblasts by using 125 I-MCP-1 (A) or 125 I-SDF-1␣ (B) as tracer and chemokines (300 nM), antagonists (300 nM), and monoclonal antibodies (10 g/ml) as competitors. C and D, CD4 ϩ lymphoblasts were either left untreated (f), or treated with 300 nM of TAK779 (Ⅺ) or AMD3100 (F) for 30 min, and the chemotactic response of cells to MCP-1 (C) or SDF-1␣ (D) was determined. The data were normalized for nonspecific binding (0%) and specific binding in the absence of competitor (100%). Statistical significance as compared with 100% values was tested by two-way analysis of variance test followed by Tukey's test. ***, p Ͻ 0.001. The displayed data are the means of three independent experiments performed with lymphoblasts prepared from three different donors. All of the data points were performed in triplicate (error bars, S.E.). OCTOBER 12, 2007 • VOLUME 282 • NUMBER 41 ing of SDF-1␣ but has no effect on other chemokine receptors, including CCR2 and CCR5 (27). We showed that the heterodimerization of CXCR4 with CCR2 allowed AMD3100 to antagonize the binding of CCR2-specific ligands. Conversely, TAK779, an antagonist of CCR2, CCR5, and CXCR3 (28), inhibited strongly the binding of SDF-1␣ in cells coexpressing CXCR4. These results challenge previous data claiming that AMD3100 and TAK-779 antagonize only receptors to which they bind (27,29,30). However, most of the data supporting these studies were obtained from recombinant cells overexpressing a single receptor, thereby ignoring potential effect of receptor partners. The cross-competition by antagonists demonstrate that negative binding cooperativity does not require receptor activation (7,21,32). Importantly, the consequences of heterodimerization on binding and functional properties of CCR2 and CXCR4 were also observed on CD4 ϩ lymphoblasts expressing the two receptors naturally, demonstrating that our observations are not the artifactual consequence of receptor overexpression. The extent of cross-inhibition in native cells is not as strong as compared with the situation in CHO-K1 cells, but lymphoblasts express other GPCRs, and the proportion of CCR2/CXCR4 heterodimers is therefore not known. AMD3100 also inhibited partially the binding and signaling of MCP-1 on purified monocytes (not shown), in contradiction with a previous study reporting that AMD3100 had no effect on CCR2 signaling on monocytes (29). This is the first report demonstrating the existence of functional CCR2/CXCR4 heterodimers in native cells. Because CCR2 and CXCR4 belong respectively to each of the two main classes of chemokine receptors, CCR2/CXCR4 heterodimers may be representive of the situation prevailing when two (or more) chemokine receptors are coexpressed in a leukocyte population.

Allosteric Transinhibition in Heterodimers
It should be noted that both on recombinant systems and native cells, the inhibition of functional responses is stronger that the binding cross-competition. In CHO-K1 cells, whereas heterologous competition reached 35% of specific binding, the functional responses were almost totally blocked by heterologous antagonists. In native cells, functional inhibition was not as complete in both directions but proportionally more important than binding cross-competition. This observation suggests that the functional interaction between different receptors may go beyond binding inhibition across heterodimers. It may involve functional cross-talk in large arrays of dimers, as proposed for the organization of rhodopsin in photoreceptor discs (34,35). Such arrays might also involve receptors belonging to other classes. CXCR4 was proposed to interact with the T cell receptor (TCR-CD3), allowing SDF-1␣-promoted CXCR4 signaling through a pre-existing TCR-ZAP70 complex and leading to more robust Ras and ERK (extracellular signal-regulated kinase) activation (36). Interestingly, TCR ligation was also reported to inhibit the CXCR4-and CXCR3-mediated signaling and chemotaxis in T cells, indicating a close and complex relationship between T cell activation and chemokine signaling (31,37).
The heterodimerization of chemokine receptors has important implications in the field of drug development and the validation of specific receptors as drug targets. Antagonists characterized as selective for a chemokine receptor may indeed inhibit as well the functional response of other receptors coexpressed in leukocyte populations. This cross-inhibition can therefore lead to effects in vivo more diverse than predicted on the basis of receptor selectivity determined in vitro. This can imply either an increased therapeutic benefit, as a result of the partial blockade of other receptors contributing to an inflammatory process or to the development of unexpected and detrimental side effects. These findings also imply that mouse invalidated for chemokine receptors may not represent ideal models for the pharmacological blockade of these receptors. Indeed, besides other established limitations (compensatory mechanisms during development, chronic versus acute inactivation, interspecies differences), the blockade of a receptor involved in heterodimers may have significantly different consequences than the absence of the receptor and therefore of all heterodimers that may form in cells. In addition, similar situations likely prevail in other receptor classes regulating other physiological functions. It was indeed reported recently that antagonists of the cannabinoid CB1 and the orexin-1 receptors regulate the cellular localization and the function of both receptors (33), although such demonstration was not extended to primary cells. Future evaluation of the therapeutic benefit of acting on chemokine receptors or other GPCR classes will therefore have to consider the existence of heterodimers and the allosteric interactions that characterize them. We thus believe that these observations will have broad implications in the field of immunology and more generally in all aspects of biology involving GPCRs and their use as therapeutic targets.