Monomeric monocyte chemoattractant protein-1 (MCP-1) binds and activates the MCP-1 receptor CCR2B.

To address the role of dimerization in the function of the monocyte chemoattractant protein-1, MCP-1, we mutated residues that comprise the core of the dimerization interface and characterized the ability of these mutants to dimerize and to bind and activate the MCP-1 receptor, CCR2b. One mutant, P8A*, does not dimerize. However, it has wild type binding affinity, stimulates chemotaxis, inhibits adenylate cyclase, and stimulates calcium influx with wild type potency and efficacy. These data suggest that MCP-1 binds and activates its receptor as a monomer. In contrast, Y13A*, another monomeric mutant, has a 100-fold weaker binding affinity, is a much less potent inhibitor of adenylate cyclase and stimulator of calcium influx, and is unable to stimulate chemotaxis. Thus Tyr13 may make important contacts with the receptor that are required for high affinity binding and signal transduction. We also explored whether a mutant, [1+9-76]MCP-1 (MCP-1 lacking residues 2-8), antagonizes wild type MCP-1 by competitive inhibition, or by a dominant negative mechanism wherein heterodimers of MCP-1 and [1+9-76]MCP-1 bind to the receptor but are signaling incompetent. Consistent with the finding that MCP-1 can bind and activate the receptor as a monomer, we demonstrate that binding of MCP-1 in the presence of [1+9-76]MCP-1 over a range of concentrations of both ligands fits well to a simple model in which monomeric [1+9-76]MCP-1 functions as a competitive inhibitor of monomeric MCP-1. These results are crucial for elucidating the molecular details of receptor binding and activation, for interpreting mutagenesis data, for understanding how antagonistic chemokine variants function, and for the design of receptor antagonists.

Chemokines are small secreted proteins that function as intercellular messengers to control migration and activation of specific subsets of leukocytes (1)(2)(3). This process is mediated by the interaction of chemokines with seven transmembrane G-protein-coupled receptors on the surface of target cells. Interest in these proteins was first stimulated by the observation of elevated levels in a number of inflammatory diseases (4,5) including rheumatoid arthritis (6,7), arteriosclerosis (8,9), and asthma (10). Although it is not clear whether excessive production is the cause or consequence of these diseases, the demonstration that neutralizing antibodies (11)(12)(13) reduced symptoms in a number of animal models generated optimism that receptor antagonists may have therapeutic benefit (14). Recently, it has been shown that certain chemokine receptors also serve as obligate coreceptors for entry of the human immunodeficiency virus into CD4ϩ cells (15)(16)(17) and that viral replication can be inhibited by the ligands of the coreceptors (18 -21). Thus a wide range of clinically important diseases are associated with chemokines and their receptors, motivating many studies to understand the molecular details of chemokine function.
Chemokines have been classified into two major families based on their pattern of cysteine residues, their chromosomal location, and their cell specificities (22). ␣-Chemokines such as IL-8 1 have a conserved CXC cysteine motif and act predominantly on neutrophils, whereas ␤-chemokines have a CC signature and attract monocytes and T-cells. The recently discovered chemokines lymphotactin (23) and fractalkine/neurotactin (24,25) are characterized by C and CX 3 C motifs, respectively, and chemoattract T-cells and NK cells. Mutagenesis studies, particularly of ␣and ␤-chemokines, have provided some insight into the structural determinants of receptor binding and the specificity of these proteins, but many details have yet to emerge.
Considerable effort has been devoted to characterizing the stochiometry of chemokine-receptor complexes because most chemokines oligomerize to an extent that depends on concentration and pH (26 -30). Above micromolar concentrations many form homodimers, whereas at nanomolar concentrations the monomeric form predominates in solution. High resolution structures of IL-8 (31), MGSA/GRO (32), MCP-1 (33,34), RAN-TES (35,36), and MIP-1␤ (37) have also revealed a striking correlation between chemokine class and mode of dimerization (38), suggesting that dimerization may play an important role in chemokine function. A key question, however, is whether dimerization is required for receptor binding and activation or whether it plays a more subtle role in protein stability, regu-lation, surface presentation and retention, formation of the chemotactic gradient, or other processes unrelated to chemotaxis.
In the case of the CXC chemokine IL-8, monomeric mutants were shown to recruit and activate neutrophils in vitro as efficiently as wild type (26,30), consistent with the view that it interacts with its receptor as a monomer. For the CC chemokine, MCP-1, some data suggest that a dimer may be the receptor-bound form of the protein (29). In these studies, a heterogeneous mixture of chemically cross-linked MCP-1 was shown to be active in chemotaxis assays. In addition, a deletion mutant lacking residues 2-8 ([1ϩ9 -76]MCP-1), which acts as a receptor antagonist, was also shown to inhibit chemotaxis by wild type but not by chemically cross-linked MCP-1. Finally, [1ϩ9 -76]MCP-1 containing a C-terminal FLAG epitope tag coprecipitated iodinated MCP-1 in an immunoprecipitation assay. Based on these results, it was postulated that MCP-1 interacts as a dimer and that [1ϩ9 -76]MCP-1 acts as an dominant negative antagonist by formation of inactive heterodimers with the wild type protein. Although this conclusion is consistent with the data, the heterogeneous nature of the cross-linked MCP-1 species allows alternative interpretations. Other data including the function of IL-8 obligate monomers, geometric considerations based on the three-dimensional structure of MCP-1, and the best data and models regarding Gprotein-coupled receptor structure and function (39 -42) also seem difficult to reconcile with a heterodimer model. Because resolving this issue is necessary for a complete understanding of the molecular mechanisms of leukocyte migration and for the modeling and utilization of structural data for the design of receptor antagonists, we felt compelled to investigate further whether MCP-1 can bind and activate its receptor as a monomer.
To explore the requirement of dimerization in the interaction of MCP-1 with its receptor, we mutated residues that contribute significantly to the dimer interface and assessed the effect of these mutations on receptor binding, on activation, and on dimerization. We also characterized the aggregation state of [1ϩ9 -76]MCP-1 at concentrations up to the millimolar range to assess whether it can form a homodimer. Finally, to elucidate the mechanism by which [1ϩ9 -76]MCP-1 acts as a receptor antagonist, we carried out binding competition experiments to address whether it functions as a dominant negative or classic competitive inhibitor.

Gene Construction of Human MCP-1 Variants for Expression in
Escherichia coli-With the exception of [1ϩ9 -76]MCP-1, all mutants were made in the context of MCP-1 M64I, referred to as WT*, and expressed in E. coli. We have demonstrated that WT* behaves identically to WT in binding and activity assays (see Table I). This alteration in the primary structure improves the purity and homogeneity of the mutants by eliminating the formation of species containing methionine-sulfoxide at position 64.
The gene for WT* MCP-1 was constructed by standard gene synthesis techniques with optimal codon usage for expression in E. coli (43). Mutant constructs were made by polymerase chain reaction mutagenesis of the WT* template and cloned into a pET3 based plasmid, pAED-4 (44). All sequences were confirmed by double-stranded DNA sequencing. Plasmids were then transformed into TAP302 cells, 2 which are BL21 pLys S cells engineered with a thioredoxin reductase knockout to make the intracellular redox potential more conducive to disulfide bond formation. Using this strain, disulfide bonds appear to be formed in the cell, eliminating the need for a refolding step.
Protein Expression and Purification in E. coli-Cells were grown in Luria broth containing 100 g/ml of ampicillan and 34 g/ml chloramphenicol at 30°C in standard 2.8-liter shaker flasks. When the cell density reached 0.6 A 600 , protein expression was induced by adding isopropyl-␤-D-thiogalactopyranoside to a final concentration of 0.5 mM. After 1 h, rifampicin was added to a final concentration of 24 M. Cells were incubated with shaking for an additional 3-4 h and harvested by centrifugation. 15 N-Labeled proteins were expressed in the same manner except with MOPS (45) minimal medium containing 15 N-ammonium sulfate (Cambridge Isotope Laboratories, Andover, MA) in place of Luria broth. Post-induction times varied from 3 to 18 h.
Typically, proteins were purified from 1.5-liter E. coli cells, harvested by centrifugation for 20 min at 6000 ϫ g. Cells were lysed in 200 ml of buffer by sonication on ice for three 5-min cycles with 5 min of cooling between cycles, using a 1.25-cm horn type sonicator at 80 -90% power. The lysis buffer contained 10 mM K 2 PO 4 , pH 7.5, 100 mM NaCl, 5 mM MgCl 2 , and 3000 units DNase I. Cell lysate was cleared by centrifugation at 13,800 ϫ g for 1 h at 4°C and then loaded on a 40-ml SP-Sepharose column equilibrated with lysis buffer. Protein was eluted with a gradient of NaCl in 10 mM K 2 PO 4 , pH 7.5; typically MCP-1 mutants elute at 0.4 Ϯ 0.1 M NaCl. Peak fractions were pooled and further purified by reversed-phase HPLC using a 4.6 ϫ 250 mm C18 column with a 5-m particle size and 300-Å pore size. Proteins were eluted using a gradient of increasing acetonitrile containing 0.1% trifluoroacetic acid; typically, proteins eluted at 34 Ϯ 5% acetonitrile. They were then lyophilized, dissolved at 1 mg/ml in 35 mM Tris, pH 8, reacted with 15 g of aminopeptidase (Peprotech, Rock Hill, NJ)/1 mg of protein for 36 h at room temperature, and repurified by reversed-phase HPLC. Aminopeptidase treatment removes only the N-terminal methionine when proline is the second residue of the mature protein. The protein was then lyophilized, redissolved in water at 1-5 mg/ml, and stored in small aliquots at Ϫ80°C. These methods typically yielded 0.5-5 mg of purified protein/1.5 liter of bacteria.
Vectors for Expression in Pichia pastoris Yeast-The N-terminal truncation mutant, [1ϩ9 -76]MCP-1, was expressed as a secreted protein in Pichia strain GS115 (Invitrogen, San Diego, CA) using expression vectors, pPIC9 (Invitrogen). Expression cassettes consist of the AOX1 promoter, a secretion sequence, [1ϩ9 -76]MCP-1 cDNA derivative, and a transcription terminator. Vector pPIC9 uses the Saccharomyces ␣-mating factor prepro-secretion sequence, and we inserted the [1ϩ9 -76]MCP-1 gene immediately after the Lys-Arg codons of the ␣-mating factor sequence. Thus the sequence of the junction between these two proteins is GVSLEKR2QVTCCY, where the downward facing arrow represents the cleavage site of the fusion protein, pSRF224. Expression cassettes were integrated into the AOX1 genes. This was done by first linearizing the expression plasmids at the edges of the 5Ј AOX1 and 3Ј AOX1 homology region using BglII and subsequently transforming into yeast. Successful integration was determined by histidine prototropy, mut Ϫ phenotype, and MCP-1 expression.
Protein Expression in P. pastoris-Pichia strains containing MCP-1 expression cassettes were grown to saturation by shaking for up to 2 days at 30°C in noninducing medium. Noninducing medium is BMGY (Invitrogen), containing buffered complex medium and 1% glycerol. Two liters of cells could be grown in this manner. Expression of MCP-1 was induced by resuspending the cells in 400 ml of medium containing 0.5% methanol and either BMMY (Invitrogen) containing buffered complex medium or synthetic medium. The cultures were shaken at 30°C for an additional 2-6 days. Buffered complex medium contains: 1% yeast extract, 2% peptone, 100 mM K 2 PO 4 , pH 6.0, 1.34% yeast nitrogen base without amino acids, and 0.4 g/ml biotin. Synthetic medium contains: 1% casamino acids, 58 (46,47). These methods typically yielded 30 mg/liter of crude secreted MCP-1 derivative and often 20 mg/liter of purified protein. Proteins were purified and characterized as described above for E. coli material, starting with the cell growth medium but without the aminopeptidase step.
Analytical Characterization-The molecular masses of all proteins were characterized by electrospray mass spectrometry and differed by no more than 1 Da from the expected value. Protein purity was analyzed using reversed-phase HPLC; the average purity was 95 Ϯ 5%. All protein concentrations were determined using an ⑀ 280 extinction coefficient calculated from the amino acid composition (48).
Binding Assay-A complete description of our binding assay can be found elsewhere. 3 Briefly, binding was measured using membranes prepared from two cell lines, THP-1 and CCR2-CHL. Each assay was composed of membranes, 50 pM 125 I-MCP, MCP buffer, protease inhibitors, and test protein. Equilibrium was achieved by incubation at 28°C for 90 min. Membrane-bound 125 I-MCP was collected by filtration through GF/B filters presoaked in polyethyleneimine and bovine serum albumin, followed by four rapid washes with approximately 0.5 ml of ice-cold buffer containing 0.5 M NaCl and 10 mM HEPES, pH 7.4. MCP buffer consists of 50 mM HEPES, pH 7.2, 1 mM CaCl 2 , 5 mM MgCl 2 , and 0.1% bovine serum albumin. Protease inhibitors include 0.1 mM phenylmethylsulfonyl fluoride, 1 M leupeptin, and 0.35 g/ml pepstatin. THP-1 cells are from a human monocyte cell line (ATCC TIP-202) that expresses both CCR1 and CCR2. CCR2-CHL cells are Chinese hamster lung cells (ATCC CRL-1657) that have been stably transformed with an expression vector, pSW104, bearing the human CCR2b receptor and a neomycin resistance marker plasmid as described previously (49). The CCR2-CHL and the THP-1-4X cells express approximately 10,000 CCR2 receptors per cell, whereas THP-1, HEK-293-CCR2b, and CHO-K1-CCR2b-cAMP-Luc-neo-22 cells express approximately 5000 CCR2 receptors/cell (see below).
Intracellular Signaling and Chemotaxis-The methods for measurement of intracellular signaling and chemotaxis have been described in a separate paper. 4 Briefly, adenylate cyclase inhibition was measured by direct cAMP measurement using stable HEK293-CCR2b-transfected cells that were labeled with [H 3 ]adenine, stimulated with forskolin, and inhibited by exposure to the appropriate mutant for 15 min. cAMP was isolated with ion exchange chromatography and quantified by radioactivity determination. Adenylate cyclase inhibition was also measured using the firefly-luciferase reporter gene attached 3Ј to a cAMP response enhancer, CREB; these elements were contained within the cell clone, CHO-K1-CCR2b-cAMP-Luc-neo-22. The assay measured adenylate cyclase inhibition after 6 h of exposure to the appropriate mutant in the presence of forskolin. Luciferase expression was quantified using the Luclite substrate (Packard Inst, Downers Grove, IL). Cytosolic calcium influx was measured in THP-1-4X cells loaded with the fluorescent dye Fura-2-AM (50). Quantitation of fluorescence intensity was done by integrating the signal for 82 s following the addition of mutant protein. Conversion of signal intensity to calcium molarity was done by standard methods (51). 4 Chemotaxis was measured over 1 h using THP-1-5X cells in a 96-well Boyden chamber apparatus. Cell migration through the polycarbonate filter was quantified by fluorescent staining using propidium iodide in 0.1% Triton X-100. These assays typically gave a stimulated to unstimulated migration of 4 -10-fold (average 6-fold) using 3 nM MCP-1, a maximally effective concentration for this protein.
Statistical Analysis of Binding and Signaling-All 441 determinations of the K d for the WT protein were tested for statistical normality; the data are log-normally distributed as would be expected from the ratio definition of K d . Similarly, the WT tests of CCR2-CHL binding, inhibition of cAMP synthesis, calcium flux, and chemotaxis were found to be log-normally distributed. Thus all of the statistical tests on the K d values and IC 50 values were performed as pK d values and pIC 50 values. Significance is noted in Table I as explained in the footnotes to the  table. Sedimentation Equilibrium Ultracentrifugation-Protein samples were dialyzed into 10 mM K 2 PO 4 , 250 mM NaCl at pH 7.5. Centrifugation was performed in a Beckman XL-A centrifuge at 25°C using six-chamber cells with an optical pathlength of 1.2 cm. Equilibrium sedimentation was performed with starting concentrations of 5.9, 11.8, and 58.9 M as determined by composition (48). After 20 h at 25,000 rpm the cells were scanned across their radius at 280 nm. Two hours later a second scan was taken and compared with the first to confirm equilibrium. The speed was then increased to 30,000 or 35,000 rpm, and after overnight equilibration the cells were scanned as before. In some cases additional data were measured at 45,000 rpm in the same way. For the very low concentrations required to measure an accurate association constant for WT*, absorbances were also measured at 220 nm after equilibration from starting concentrations ranging from 0.1 to 1 M.
Analysis was performed using the Microcal Origin software provided by Beckman for the XL-A. In the absence of significant nonideal behavior, all runs were first fit to the ideal, noninteracting model.
where A r is the absorbance at radius x, A o is the absorbance at reference radius x o , H is a constant that accounts for partial specific volume of the solute, solvent density, rotor speed, and temperature, M is the gram molecular weight of the solute, and E is a base-line offset. Data were fit to the equation by nonlinear least squares. The quality of the fit was characterized by 2 , the sum of the squares of the residuals, and examination of the residuals for systematic deviation. Data from multiple samples at different concentration ranges and centrifuge speeds were then fitted simultaneously using the software provided by Beckman. By this method, it is possible to obtain an association constant. Those constants are reported only when they fall within the concentration range observed in the experiment.
Translational Diffusion Coefficient Measurements by NMR-Diffusion coefficient measurements were performed using the water-sLED experiment (52). Sample concentrations were 2 mM in 10 mM K 2 PO 4 , 150 mM NaCl, pH 7.5. The experiments were performed on a Bruker DMX-600 spectrometer at 35°C. Gradient strengths were calibrated using a GdCl 2 doped H 2 0 sample. Data were collected using gradient strengths from 10 to 70% increasing by 5% each time. The attenuation of the NMR signal, A, is related to the translation diffusion coefficient, D, by the following expression.
where ␥ is the gyromagnetic ratio of the observed nucleus, G is the gradient strength, ␦ is the duration of the gradient pulse, and ⌬ is the delay between gradient pulses.
where M b and M f are the bound and free concentrations of MCP-1, R T is the receptor concentration, K d is the dissociation constant for WT binding, K i is the dissociation constant for binding of [1ϩ9 -76]MCP-1, and I is the concentration of [1ϩ9 -76]MCP-1. It was assumed that the concentration of I is much greater than the concentration of receptor; this is a reasonable assumption because even at the lowest concentration of [1ϩ9 -76]MCP-1 used in these experiments, it exceeds the concentration of receptor by a factor of 10.

Y13A* Affects Receptor
Binding and Activation-To investigate the role of dimerization on binding and activity we designed mutants that potentially disrupt or weaken the ability of MCP-1 to self-associate. Based on the structure (33,34) and calculations of the difference in solvent accessibility between monomer and dimer, we identified several candidates for mutation including Tyr 13 , Thr 10 , Val 9 , and Pro 8 (Fig. 1). These residues form a short stretch of ␤-sheet at the interface between the two subunits. Tyr 13 and Tyr 13 Ј are oriented toward each other and in the crystal structure (34) are hydrogen bonded through an intervening water molecule. They also make hydrophobic contacts that stabilize the dimer by packing onto Val 9 of the opposing subunit. Thr 10 and Thr 10 Ј are packed against each other on the concave face of the dimer and interact through their hydroxyl groups. In addition to these side chain interactions, there are also four hydrogen bonds stabilizing the ␤-sheet.
We mutated each of these residues individually to make the following constructs: Y13A*, V9A*, P8A*, V9E*, T10E*, and V9AϩT10A* (the asterisks indicate that the mutations are in the WT* background; WT* is M64I). The binding affinity of these proteins and [1ϩ9 -76]MCP-1 for the receptor, CCR2b (55), was tested in a THP-1 cell line and in stably transfected CCR2-CHL cells (Table I). These two cell lines were used to assess the effect of cell background on binding affinity and ensure that measurement of THP-1 affinities was not complicated by the presence of CCR1 receptors. The measured affinity of MIP-1␣ in the THP-1 assay is Ͼ100 nM, whereas MCP-1, MCP-2 and MCP-3 compete with K d values of 35, 320, and 50 pM, respectively. This indicates that we only measure binding to CCR2 in the THP-1 assay. The binding affinities of wild type and mutant proteins were similar for THP-1 and CCR2-CHL cells, also indicating that cell background does not influence the results. Fig. 2 shows representative binding isotherms for WT MCP-1 and several of the mutants on THP-1 cells, and the results are summarized in Table I. In agreement with previous reports (56), the binding affinity of [1ϩ9 -76]MCP-1 is reduced by approximately 7-fold relative to WT. However, the impact of the N-terminal deletion was not as significant as the single Y13A* mutation, which showed an increase in IC 50 of 2 orders of magnitude relative to WT and WT*. P8A* and V9A* had WT binding affinity, whereas T10E* and V9E* showed slight reductions in affinity by factors of 6 and 22, respectively. The double mutant V9AϩT10A* showed only a slight reduction in affinity of 2.3-fold.
The impact of these mutations on binding was paralleled by their effect on receptor activation as measured by inhibition of forskolin-stimulated cAMP synthesis (Fig. 3a) and luciferase expression (Fig. 3b), stimulation of cytosolic calcium influx (Fig. 3c), and chemotaxis (Fig. 4). In these assays, Y13A* again showed the largest reduction in activity relative to WT followed by T10E* and V9E*. These results are consistent with previous studies in which mutation of Tyr 13 to isoleucine and mutation of Thr 10 to arginine (57) were shown to cause substantial reductions in binding affinity and induction of chemotaxis. Y13A* also displays an interesting pattern of agonist activities; it inhibits cAMP synthesis and stimulates cytosolic calcium influx with EC 50 values, which are reduced compared with wild type but to a similar degree as its reduction in binding affinity. However, Y13A* was not able to drive chemotaxis even at 100 M. P8A* displayed WT if not slight superagonist efficacy in all three assays as illustrated in the chemotaxis profile of Fig. 4a. We also observed that [1ϩ9 -76]MCP-1, which is generally described as an antagonist, had substantial agonist activity as assessed by its ability to inhibit cAMP synthesis (Fig. 3). How-ever, [1ϩ9 -76]MCP-1 did not induce chemotaxis (Fig. 4a).
These results suggest that some of the N-terminal residues and particularly Tyr 13 in the dimerization interface are important for binding and activation. However, from these data alone it is not possible to conclude whether the effects are due to the fact that these residues directly contact the receptor or whether dimerization is required and the mutations destabilize the dimer structure. To explore these possibilities, we characterized the ability of each of the mutants to form homodimers.
Characterization of Aggregation States; Y13A*, [1ϩ9 -76]MCP-1, and P8A* are Monomeric-To evaluate the impact of the above mutations on dimerization, equilibrium sedimentation measurements were carried out on each variant. Representative sedimentation profiles for WT*, [1ϩ9 -76]MCP-1, and P8A* are illustrated in Fig. 5. Also shown are theoretical curves for a monomer versus dimer. As previously reported, WT* forms a homodimer (27,33) with a K d of approximately 0.5 M. The two mutants V9E* and T10E* also formed dimers but were slightly less stable than WT*, with K d values of 8 and 12 M, respectively, whereas the dimerization of V9A* was not significantly altered. As expected from the structure, dimerization of the Y13A* mutant was severely destabilized, and the protein remained monomeric up to 100 M, the maximum concentration of the experiment (data not shown). In light of the fact that V9A*, V9E*, and T10E* maintained the ability to dimerize, the most surprising result was that P8A* was monomeric (Fig. 5c). From inspection of the structure (Fig. 1), we anticipated that of all the mutated residues, Pro 8 would contribute least to the stability of the dimerization interface.
To confirm these results and evaluate aggregation states at even higher concentrations, we also carried out translational diffusion measurements by NMR. In these experiments, attenuation of the signal intensity as a function of gradient strength increases with the diffusion rate, D t (Fig. 6). For the dimeric WT protein, the measured D t was 0.931 ϫ 10 Ϫ6 cm 2 /s. For P8A*, Y13A*, and [1ϩ9 -76]MCP-1 the values range from 1.44 ϫ 10 Ϫ6 to 1.60 ϫ 10 Ϫ6 cm 2 /s (see the legend to Fig. 5). By comparison with the value of 1.62 ϫ 10 Ϫ6 cm 2 /s for ubiquitin, a model protein of approximately the same size as an MCP-1 monomer, we conclude that these three mutants are monomeric at concentrations in excess of 1 mM. Thus dimerization in these mutants is destabilized by at least 3 orders of magnitude relative to WT*.
NMR of MCP-1 Variants Indicate They Are Properly Folded-To rule out the possibility that the mutations cause major structural perturbations, we recorded two-dimensional 1 H-15 N HSQC spectra on 15 N-labeled WT*, Y13A*, P8A*, and [1ϩ9 -76]MCP-1 and one-dimensional 1 H spectra on unlabeled proteins. In all cases, the spectra were well dispersed, indicating that the mutations do not cause misfolding of the protein. In the one-dimensional 1 H spectra, all variants had two isolated upfield shifted resonances at chemical shifts similar to those observed in WT, which were previously assigned to ␥-methyl protons of Val 60 and Val 41 (33) (data not shown). These residues cluster together with Phe 43 in the core of the protein, and preservation of the shifted methyl protons in the mutants suggests that their core structures are similar to WT. The HSQC spectra (Fig. 7) are also all well dispersed and contain the expected number of cross-peaks, but there are significant chemical shift changes between WT* compared with Y13A*, P8A*, and [1ϩ9 -76]MCP-1. This is not surprising because the dimerization interface in MCP-1 is extensive, and loss of dimer contacts would be expected to affect residues from the N ter- FIG. 1. Molscript diagrams (71) of an MCP-1 monomer (a) and  dimer (b). The ␣-carbons of residues mutated in this study (Pro 8 , Val 9 , Thr 10 , and Tyr 13 ) are highlighted. minus to Asn 14 , residues in the 30s loop and residues in the 50s region. We assigned the 1 H N , 15 N, 13 C ␣ , and 13 C ␤ resonances of P8A* 6 and confirmed that only residues in the dimer interface were significantly perturbed. Because the HSQC of P8A* is similar to that of Y13A* and [1ϩ9 -76]MCP-1, many of the assignments could be transferred to these other proteins. In Fig. 7, representative cross-peaks that are remote from the dimer interface and have similar chemical shifts in all four proteins are highlighted in boxes. Representative cross-peaks that shift because of loss of dimerization contacts are indicated by arrows. Thus we can confidently conclude that the changes between these monomeric variants and WT* are due to loss of dimer contacts and not to misfolding.
To summarize the data thus far, P8A* is well folded and has WT activity but does not dimerize. This leads to the conclusion that MCP-1 can interact and activate CCR2b as a monomer. The fact that the Hill coefficient for the binding of WT MCP-1 is 0.97 Ϯ 0.17 (average of 441 measurements) corroborates this conclusion. Consequently we suggest that the Y13A* mutation causes a loss of function because Tyr 13 interacts directly with the receptor and not because the mutation impairs dimerization or folding.
[1ϩ9 -76]MCP-1 Is a Competitive Inhibitor-If MCP-1 interacts with CCR2b as a monomer, this suggests that receptor antagonism by [1ϩ9 -76]MCP-1 probably does not occur by a dominant negative mechanism involving heterodimer formation. To explore this, we carried out equilibrium binding displacement measurements of MCP-1 in the presence of various amounts of the inhibitor. The data are shown in Fig. 8 as double-reciprocal (Benesi-Hildebrand) plots. Using a simple model of competitive inhibition, the data (including binding at additional concentrations not included in the figure) were simultaneously fit to estimate the dissociation constants for binding of WT MCP-1 (K d ), [1ϩ9 -76]MCP-1 (K i ), and the receptor concentration (R T ). The K i , K d , and R T calculated from the fit are 439, 12.3, and 9.3 pM, respectively, in good agreement with the independently measured values reported in Table I. This supports the conclusion that antagonism by [1ϩ9 -76]MCP-1 occurs through a competitive inhibition mechanism in which monomeric WT MCP-1 is displaced from the receptor by monomeric [1ϩ9 -76]MCP-1.
Demonstrating direct binding of [1ϩ9 -76]MCP-1 to the receptor would provide additional evidence in favor of competitive inhibition. However, we have determined that for the concentrations of receptor achievable in our assay, the antagonist affinity and the level of nonspecific binding, specific antagonist binding is not measurable.
Disulfide Cross-linked WT*(C77) and [1ϩ9 -76](C77) Bind and Activate CCR2b Similar to the Monomeric Counterparts-We made specific disulfide cross-linked MCP-1 dimers by adding a cysteine residue to the C terminus of WT* and [1ϩ9 -76]MCP-1. These proteins, WT*(C77) and [1ϩ9 -76](C77), were expressed and purified as monomers and then dimerized by air oxidation. The homogeneity and purity of the dimers was confirmed by mass spectrometry and reversed-phase HPLC. The WT*(Cys77) dimer interacts with CCR2b with near wild type affinity on THP-1 and CCR2-CHL    (Table I). These data suggest that cross-linking in this manner does not alter the ability of these proteins to bind and activate the receptor. DISCUSSION We have presented several pieces of evidence that support a model in which MCP-1 binds and activates CCR2b as a monomer. The most conclusive data are based on P8A*, which is fully functional in binding and activity assays yet does not dimerize even at very high concentrations. We have also shown that [1ϩ9 -76]MCP-1 does not efficiently homodimerize, suggesting that it also probably binds to the receptor as a monomer rather than an oligomer. Although it is possible that dimer formation may be more favorable on the receptor than in solution, the receptor binding data of Fig. 8 fits well to a model in which MCP-1 binds as a monomer and is competitively inhibited by monomeric [1ϩ9 -76]MCP-1. Therefore the highly conserved dimerization motif observed in most CC chemokines does not appear to be an integral part of receptor binding and activation. In fact, the increased potency of P8A compared with WT MCP-1 in the activity assays may reflect a negative influence of dimer formation on function; this is best rationalized by the fact that residues involved in stabilizing the dimer (e.g. Tyr 13 ) are also involved in interactions with the receptor. 3,4 Even so, dimerization is probably not irrelevant to the biological function of MCP-1. It seems unlikely that ␣ and ␤ chemokines would have evolved such strictly conserved modes of dimerization for no reason. Although we were surprised by the magnitude of the effect on dimerization of our proline to alanine mutation, the frequent occurrence of prolines at the boundaries of dimerization interfaces has been documented (58). It has been suggested that they impose constraints on the conformation and dynamics of the polypeptide chain in a way that favors oligomerization, often through a mechanism described as "arm exchange" (58). Importantly, the equivalent residue of Pro 8 of MCP-1 is conserved in most well characterized CC chemokines, again suggesting that dimerization is there by design and plays a functional role. An exception to this is MCP-3, which possesses a serine at the corresponding position and is monomeric (59). The viral CC chemokines, MC148R1 and MC148R2 (60), are also missing this proline, which may point to a function of oligomerization that is required for certain host chemokines but not for these viral chemokines.
Can MCP-1 bind as a dimer? Although this has been demonstrated by chemical cross-linking studies (29), it is not clear that the cross-linked dimer is the same type of dimer as occurs naturally (Fig. 1b)  We have also demonstrated that mutation of Tyr 13 to alanine impairs binding and receptor activation to the extent that there is no measurable chemotaxis, suggesting that this residue makes important contacts with the receptor. Because Tyr 13 probably cannot simultaneously stabilize the dimer and interact with the receptor, we speculate that a true dimer would bind more weakly than a monomer, would be impaired in its ability to induce intracellular signaling, and would also be unable to induce chemotaxis.
If dimerization is not required for interaction with the receptor, then an important question is, why does it occur? One suggestion is that it serves a regulatory role to inhibit chemokine action at high chemokine concentrations, thus explaining the bell shape profiles observed in chemotaxis versus concentration experiments. However, all known chemotactic agents in both mammalian and nonmammalian cells elicit chemotaxis profiles that are bell shaped (61). Furthermore, in the elegant work of Lauffenburger and Zigmond and their co-workers (62,63), it was demonstrated that bell shaped profiles are a result of the receptor number, equilibrium, and concentration gradient present during the chemotaxis experiment (62). Thus dimerization does not account for the inhibitory phase of chemotaxis curves. This is clearly demonstrated by the similarity of the profiles for P8A* and WT MCP-1 (Fig. 4a) despite the completely different dimer forming potential of these two proteins.
A more likely possibility is that oligomerization may be important for retention and presentation of chemokines by surface glycosaminoglycans (64 -66). Mechanistically, this might occur through an avidity effect whereby oligomerization brings multiple chemokine heparin-binding domains together, increasing the affinity for glycosaminoglycans and facilitating the formation of surface concentration gradients necessary for haptotaxis. Consistent with this view, glycosaminoglycan-induced oligomerization does not occur for a monomeric form of IL-8 (66). Moreover, an analog of IL-8 lacking the heparinbinding domain was shown to have decreased in vivo proemigratory activity, which correlated with its compromised ability to bind heparin sulfate and endothelial cell surfaces (67). In contrast, a different mechanism is used for presentation of the CX 3 C chemokine, fractalkine. This protein does not oligomer-ize 7 but instead has a membrane anchored mucin-like stalk that tethers the chemokine domain to the cell surface. Together these domains appear to directly promote cell-cell adhesion as well as standard chemotactic responses (68).
Other roles for oligomerization are certainly possible and may be relevant to additional processes triggered by chemokine binding such as lysozomal enzyme release and generation of toxic products from the respiratory burst (69,70). Alternatively, oligomerization may be involved in as yet unidentified functions unrelated to chemotaxis. Either way, the mutants generated in this work should provide important reagents for pursuing these questions.
In conclusion we have demonstrated that a monomeric form of MCP-1 is sufficient for receptor binding and activation. However, we support the involvement of oligomerization in some function of the protein, most likely in surface display and marking of surfaces as eligible for diapedesis. We have also shown that [1ϩ9 -76]MCP-1 functions as an antagonist via competitive inhibition of the WT protein. This suggests that virally encoded receptor antagonists that have natural N-terminal deletions, such as MC148R (60), may function in a similar manner. We believe these results provide new molecular details and insight into the mechanism of leukocyte chemotaxis.