Chemokine CXCL1 Dimer Is a Potent Agonist for the CXCR2 Receptor*

Background: Chemokines exist reversibly as monomers and dimers, but dimer activity remains poorly defined. Results: A disulfide-linked CXCL1 dimer is highly active, and NMR studies show that dimer binds CXCR2 like the monomer. Conclusion: The potent activity of CXCL1 dimer is novel. Significance: Chemokine dimers can be highly active to completely inactive, indicating that dimerization fine-tunes chemokine-specific in vivo functions. The CXCL1/CXCR2 axis plays a crucial role in recruiting neutrophils in response to microbial infection and tissue injury, and dysfunction in this process has been implicated in various inflammatory diseases. Chemokines exist as monomers and dimers, and compelling evidence now exists that both forms regulate in vivo function. Therefore, knowledge of the receptor activities of both CXCL1 monomer and dimer is essential to describe the molecular mechanisms by which they orchestrate neutrophil function. The monomer-dimer equilibrium constant (∼20 μm) and the CXCR2 binding constant (1 nm) indicate that WT CXCL1 is active as a monomer. To characterize dimer activity, we generated a trapped dimer by introducing a disulfide across the dimer interface. This disulfide-linked CXCL1 dimer binds CXCR2 with nanomolar affinity and shows potent agonist activity in various cellular assays. We also compared the receptor binding mechanism of this dimer with that of a CXCL1 monomer, generated by deleting the C-terminal residues that stabilize the dimer interface. We observe that the binding interactions of the dimer and monomer to the CXCR2 N-terminal domain, which plays an important role in determining affinity and activity, are essentially conserved. The potent activity of the CXCL1 dimer is novel: dimers of the CC chemokines CCL2 and CCL4 are inactive, and the dimer of the CXC chemokine CXCL8 (which is closely related to CXCL1) is marginally active for CXCR1 but shows variable activity for CXCR2. We conclude that large differences in dimer activity among different chemokine-receptor pairs have evolved for fine-tuned leukocyte function.

The CXCL1/CXCR2 axis plays a crucial role in recruiting neutrophils in response to microbial infection and tissue injury, and dysfunction in this process has been implicated in various inflammatory diseases. Chemokines exist as monomers and dimers, and compelling evidence now exists that both forms regulate in vivo function. Therefore, knowledge of the receptor activities of both CXCL1 monomer and dimer is essential to describe the molecular mechanisms by which they orchestrate neutrophil function. The monomer-dimer equilibrium constant (ϳ20 M) and the CXCR2 binding constant (1 nM) indicate that WT CXCL1 is active as a monomer. To characterize dimer activity, we generated a trapped dimer by introducing a disulfide across the dimer interface. This disulfide-linked CXCL1 dimer binds CXCR2 with nanomolar affinity and shows potent agonist activity in various cellular assays. We also compared the receptor binding mechanism of this dimer with that of a CXCL1 monomer, generated by deleting the C-terminal residues that stabilize the dimer interface. We observe that the binding interactions of the dimer and monomer to the CXCR2 N-terminal domain, which plays an important role in determining affinity and activity, are essentially conserved. The potent activity of the CXCL1 dimer is novel: dimers of the CC chemokines CCL2 and CCL4 are inactive, and the dimer of the CXC chemokine CXCL8 (which is closely related to CXCL1) is marginally active for CXCR1 but shows variable activity for CXCR2. We conclude that large differences in dimer activity among different chemokine-receptor pairs have evolved for fine-tuned leukocyte function.
Since their original discovery about 25 years ago, it is now firmly established that chemokines function as crucial immune regulators and that dysregulation in their function leads to a variety of inflammatory and infectious diseases (1,2). Humans express ϳ50 chemokines, and all share the following fundamental properties: they reversibly exist as monomers and dimers, mediate function by activating G protein-coupled receptors expressed on leukocytes, and regulate leukocyte trafficking by binding to glycosaminoglycans on endothelial cells and the extracellular matrix.
Chemokines are broadly divided into CXC and CC classes based on whether the N-terminal conserved cysteines are contiguous or separated by an amino acid. Structures of a number of chemokines have been solved by solution NMR and x-ray crystallography, which reveal that the subfamilies dimerize using different regions of the protein. However, the monomerdimer equilibrium constants for both CXC and CC chemokines can vary by orders of magnitude, with both CXC and CC chemokines showing weak to very strong dimerization potencies (3). Dimerization is also sensitive to solution conditions such as pH and ionic strength, with some chemokines showing dramatic changes from very high order oligomers at physiological pH to folded monomers and dimers at lower pH values (4 -6). However, deciphering the functional roles of monomer and dimer is challenging, as the very process of monomerdimer equilibrium interferes in the study of one or the other form, and so studies geared toward understanding monomer and dimer function continue to be an area of active research.
The strategy of creating nondissociating dimers has shown that the disulfide-linked dimeric forms of the CC chemokines CCL2 and CCL4 are inactive, the CXC chemokine CXCL12 2 is differentially active, and CXCL8 is marginally active for CXCR1 but differentially active for CXCR2 (7)(8)(9)(10)(11)(12). However, there have been no studies investigating the functional roles of the dimeric form of CXCL1 or related neutrophil-activating chemokines, which specifically and selectively activate the chemokine receptor CXCR2. CXCL1 (also known as melanoma growth stimulatory activity or GRO␣), a member of the ELR-CXC subset of chemokines, plays a crucial role in recruiting neutrophils in response to microbial infection and tissue injury. CXCL1 dimerizes at micromolar concentrations (K d ϳ20 M). Cell-based in vitro assays have shown that WT CXCL1 activates CXCR2 at nanomolar concentrations, and a nonassociating monomer also shows WT-like activity, indicating that WT is active as a monomer (13)(14)(15)(16). The structure of the WT CXCL1 dimer is known, as by necessity structure determination studies are carried out at high (millimolar) concentrations where CXCL1 exists as a dimer (17,18). CXCL1 is up-regulated under conditions of insult; under basal conditions, its concentration is negligible (nanomolar to picomolar), and it exists essentially as a monomer. However, during active neutrophil recruitment, its concentration could reach levels high enough so that dimers exist at different locations and times, indicating that both monomers and dimers orchestrate neutrophil recruitment. Therefore, knowledge of the receptor activities of the dimer, and the molecular mechanisms by which monomers and dimers activate CXCR2, is essential to describe how CXCL1 monomers and dimers regulate in vivo neutrophil function.
To address this missing knowledge, we have now designed a disulfide-linked CXCL1 N27C dimer and have characterized its receptor function. Because the receptor N-terminal domain (N-domain) functions as a critical docking site and plays an important role in determining affinity and activity, we also characterized the binding of the trapped dimer and a designed monomer to the CXCR2 N-domain using NMR spectroscopy.
Our data indicate that the functional characteristics of the CXCL1 dimer are distinctly novel. Together, our results also emphasize that the receptor affinities, selectivities, and activities of chemokine monomers and dimers vary for different chemokine-receptor pairs, and we propose that dimerization plays chemokine-specific differential roles for fine-tuned in vivo leukocyte function.

EXPERIMENTAL PROCEDURES
Expression and Purification of CXCL1 Variants-All variants were cloned in a pET32 Xa vector and expressed and purified as His-tagged thioredoxin fusion proteins, as described previously (19,20). The CXCL1(1-67) monomer was generated by introducing a stop codon after residue 67 in the WT sequence. The trapped N27C dimer was generated using primers in which the codon of Asn-27 was replaced with Cys. PCR amplification was carried out using the QuikChange Site-directed Mutagenesis kit (Strategene).
The CXCL1 variants were expressed in the Escherichia coli BL21 (DE3) strain, and 15 N-labeled proteins were expressed in minimal medium containing 15 NH 4 Cl as the nitrogen source. Transformed E. coli BL21 (DE3) cells were grown to an A 600 of 0.8 and induced with 1 mM isopropyl ␤-D-thiogalactopyranoside for 8 h at 37°C. The purity and molecular weight of the proteins were confirmed using analytical high pressure liquid chromatography (HPLC) and matrix-assisted laser desorption/ ionization mass spectrometry, respectively.
Cell Culture and Differentiation-HL60 cells were induced to differentiate into the neutrophilic lineage as follows. Briefly, the cells were maintained in RPMI 1640 medium supplemented with L-glutamine, 10% fetal bovine serum, and 50 units/ml Pen-Strep (Invitrogen). Cells were subcultured every 2-3 days and differentiated in antibiotic-free medium containing 1.25% dimethyl sulfoxide (endotoxin-free; Sigma) and cultured for 6 -7 days.
Receptor Affinity-The binding affinity of WT CXCL1 and the trapped dimer was measured using HL-60 cells expressing CXCR2 (HL60-CXCR2) as described previously (21). HL60-CXCR2 cells (10 7 cells/ml) suspended in phosphate-buffered saline (PBS) containing 0.1% (w/v) bovine serum albumin and 20 mM HEPES (pH 7.4) were incubated at 4°C for 4 -6 h in the presence of 0.15 nM 125 I-CXCL1 and increasing concentrations (10 Ϫ11 to 10 Ϫ6 M) of unlabeled CXCL1 WT and trapped dimer. The radioactivity in the pellet was measured using a ␥-counter, and binding affinities were calculated from at least two independent experiments, with each experiment performed in duplicate.
Intracellular Ca 2ϩ Mobilization-Changes in intracellular Ca 2ϩ were measured with a FlexStation III microplate reader (Molecular Devices, Sunnyvale, CA) using the Calcium 5 assay kit (FLIPR No-wash kit; Molecular Devices). Briefly, differentiated HL60-CXCR2 cells in Hanks' buffered saline solution were plated in a flat-bottomed black microtiter plate (2 ϫ 10 5 cells/ well). The cells were loaded with Calcium 5 dye for 1 h, and changes in fluorescence on the addition of WT CXCL1 and the trapped dimer at various doses were monitored ( ex 485 nm, em 525 nm) every 5 s for 240 -500 s at room temperature. The agonist response was determined by expressing the maximum change in fluorescence in arbitrary units over base line.
Chemotaxis-Chemotaxis was measured using a 96-well chemotaxis chamber with a 3.2-m pore diameter filter membrane (NeuroProbe). Briefly, HL60-CXCR2 cells were harvested, washed, and incubated with 5 g/ml calcein-AM (Invitrogen) for 30 min at 37°C. CXCL1 WT or trapped dimer at 10 and 100 nM and the washed cells (2 ϫ 10 5 ) were added to the lower and upper chambers, respectively, and the entire assembly was incubated at 37°C for 90 min. The number of migrated cells in the lower chamber was measured using a fluorescence plate reader. Data are expressed as the chemotactic index, calculated as ratio of fluorescence of migrated cells due to chemoattractant and due to control medium alone.
ERK Phosphorylation-Cell lysates for the detection of ERK phosphorylation were obtained as described previously (22). Differentiated HL60-CXCR2 cells (5 ϫ 10 7 cells/ml) were stimulated for 1, 5, and 10 min with 10 nM CXCL1 WT or trapped dimer and lysed with an equal volume of radioimmuneprecipitation assay buffer (Cell Signaling) supplemented with proteinase and phosphatase inhibitors. Cell lysates were sonicated for 5 s (Sonifier 250; Branson Ultrasonics, Danbury, CT) and centrifuged at 10,000 ϫ g for 15 min. The supernatants were subjected to SDS-PAGE, transferred onto a PVDF membrane at 100 V for 1 h, and blocked in 5% nonfat skim milk in TBST (10 mM Tris (pH 8.0), 150 mM NaCl, and 0.05% Tween 20) for 1 h at room temperature. Membranes were immunoblotted with primary antibody overnight and then treated with secondary antibody for 1 h. Protein bands were visualized using the ChemiDoc XRS system (Bio-Rad), and quantified using Quantity One software. The rabbit anti-phospho-ERK (Thr-202/Tyr-204) and p44/p42 total ERK antibodies were from Cell Signaling, and secondary goat anti-rabbit IgG antibody was from Santa Cruz Biotechnology.
NMR Characterization of CXCL1 Binding to CXCR2 N-domain-A synthetic CXCR2 N-domain peptide (MESDS-FEDFWKGEDlSNYSYSSTLPPFLLDAAPSEPESLEINK) was purchased from AAPPTec (Louisville, KY). The native CXCR2 N-domain sequence has a cysteine that was mutated to a serine (underlined). Two-dimensional 1 H-15 N HSQC spectra were acquired at 30°C with 2048 complex points in the direct 1 H dimension and 128 complex points in the indirect 15 N dimension. A stock solution of CXCR2 N-domain was titrated into the CXCL1 solution to final mole ratios (peptide: CXCL1 variant) of 4.9, 7.1, and 5.2 for the CXCL1(1-67) monomer, WT, and trapped N27C dimer, respectively. Apparent dissociation constants (K d ) were determined by fitting the binding-induced chemical shift changes (⌬␦) as described previously (25).

Design and Characterization of CXCL1 Dimer and Monomer-
A trapped CXCL1 dimer was designed by substituting a cysteine for Asn-27. The WT dimer structure shows that Asn-27 of the first ␤-strand constitutes the 2-fold symmetry axis, its side chain is solvent-exposed and not involved in dimer interactions, and its backbone amide is H-bonded to the Ile-41 carbonyl of the adjacent ␤-strand within the monomer (Fig. 1) (17, 18, 26). Gel electrophoresis, mass spectrometry, and sedimentation velocity studies show that mutating Asn-27 to Cys results in the formation of a trapped nondissociating dimer (supplemental Fig. S1). Sedimentation equilibrium experiments showed no evidence of monomer-dimer equilibrium and showed molecular weights consistent with a dimer (data not shown). NMR NOE data show characteristic sequential NOEs for the ␤-strand, intramolecular NOEs between ␤-strands, and intermolecular NOEs across the dimer interface residues, indicating that introducing the disulfide does not perturb the native structure (Fig. 2).  WT CXCL1 dimerizes at micromolar concentrations (K d ϳ20 M) and is therefore a dimer at the millimolar concentrations used in NMR studies (16). We designed a CXCL1 monomer by deleting the last six C-terminal helix residues, as the WT CXCL1 dimer structure showed that these residues are involved in packing interactions with residues of the other monomer across the dimer interface ( Fig. 1) (17, 18). Sedimentation velocity showed that the CXCL1(1-67) deletion mutant is a monomer (supplemental Fig. S1). Further, the characteristic intermolecular NOEs of dimer interface residues were absent; for instance, dimer interface NOEs from Val-28 to Ser-25 and Val-26 were not observed, indicating that the deletion mutant is a monomer (Fig. 2).
CXCR2 Receptor Function-It is well established that CXCL1 binds and triggers various receptor activities at nanomolar concentrations, where WT essentially exists as a monomer (13,14,27). The binding affinity of the CXCL1 disulfide-linked dimer and WT was measured by competitive binding to the CXCR2 receptor stably expressed on HL-60 cells (Fig. 3A). The trapped CXCL1 dimer effectively competed with the binding of 125 I-CXCL1 like the WT (K d of WT, 1.0 Ϯ 0.2 nM; dimer, 1.6 Ϯ 0.3 nM). Our measured receptor affinity of 1 nM for the WT is similar to those reported in other studies (13,14,27).
We characterized the receptor activity of the disulfidelinked dimer by measuring Ca 2ϩ release, chemotaxis, and ERK phosphorylation (Fig. 3, B-D). The calculated EC 50 values for Ca 2ϩ release of 9 Ϯ 2 nM for WT and 29 Ϯ 9 nM for the trapped dimer, and similar activities for chemotaxis and ERK phosphorylation, indicate that the CXCL1 dimer is a potent CXCR2 agonist.
Structural Characterization of CXCL1 Monomer and Dimer-The HSQC spectrum of the disulfide-linked dimer showed excellent dispersion and a single set of peaks, indicating a well folded and structured protein (Fig. 4). The backbone C ␣ H shifts are exquisitely sensitive to secondary, tertiary, and quaternary structure. The chemical shift difference plot between the trapped dimer and WT dimer reveals that the shifts are very similar with some localized differences in residues around the site of mutation, indicating that the structure of the disulfide-linked dimer is essentially similar to the WT dimer. Analysis of NOEs also indicated that introducing the disulfide does not perturb the packing interactions between the strands within the monomer and across the dimer interface (Fig. 2).
The HSQC spectrum of CXCL1(1-67) showed the characteristic shifts of a folded protein (supplemental Fig. S2). Moreover, functional data indicated that it binds the receptor with affinity similar to that of the WT (K d 1.1 Ϯ 0.1 nM) and is highly active in the Ca 2ϩ release assay (supplemental Fig. S3). Interestingly, the NMR spectrum of the monomer showed two peaks for several residues indicating the presence of two conformers. Reducing the pH minimized heterogeneity, and the spectrum at pH 4.5 indicated essentially a single major conformer (supplemental Fig. S2). Analytical ultracentrifugation studies showed that the CXCL1 deletion mutant is a monomer at all pH values, indicating that the conformational heterogeneity is not due to dimerization. The effect of temperature at pH 4.5 showed that higher temperatures (40 versus 20°C) further minimized heterogeneity (data not shown). These data show that the differences in conformational dynamics are unlikely to play a role in function; whether it plays some other in vivo role remains to be determined.

Mechanism of CXCL1 Dimer and Monomer Binding-Receptor activation by chemokines involves two interactions:
between the ligand N-loop and receptor N-terminal domain residues (defined as Site I) and between the ligand N terminus and receptor extracellular loop residues (defined as Site II) (28 -30). All neutrophil-activating chemokines share a highly conserved ELR motif, and mutational studies have shown that these residues are critical for receptor activation at Site II. The N-loop residues are less conserved, conformationally flexible, play a major role in determining receptor affinity, and are intimately coupled to receptor activation at Site II (13,29).
The structure of the CXCR2 receptor is not known, but structures of CXCR1 and CXCR4 have shown that the receptor N-domain is unstructured (31, 32). Sequence analysis and structure-function studies for various chemokines have shown that binding studies with isolated receptor N-domain can capture the structural characteristics of Site I interactions in the context of intact receptor (12, 20, 25, 33-40).
We have now characterized the Site I binding interactions of the CXCL1 WT, designed monomer, and disulfide-linked dimer by titrating unlabeled CXCR2 N-domain into 15 N-labeled CXCL1 variants. Essentially the same N-loop residues were perturbed in the monomer and in the disulfide-linked and WT dimers, indicating that the same structural interactions mediate the binding of the monomer and dimer (Fig. 5). In particular, Gly-17 and Lys-21 show the largest chemical shift perturbations in both the monomer and dimer. The WT and trapped dimer bind the isolated CXCR2 N-domain with ϳ5-fold lower affinity compared with the CXCL1 monomer (Fig. 6). The affinity of the trapped dimer to the intact receptor was at most ϳ2-fold lower compared with the monomer (Fig. 3A), indicating that similar Site I and Site II interactions mediate monomer and dimer binding interactions.
NMR studies facilitate the description of how individual amino acids in CXCL1 mediate the binding process. We observe that essentially the same residues mediate receptor binding in the monomer and dimer. Binding induced significant chemical shift changes for the N-loop, second ␤-strand, 40s turn, and third ␤-strand, and the C-terminal helical residues (Fig. 5). Significantly perturbed N-loop residues include Thr-14, Leu-15, Gln-16, Gly-17, Ile-18, and Lys-21, and to a lesser extent Leu-12, Gln-13, His-19, and Asn-22. The structure shows that these residues span the entire length of the N-loop (Fig. 7), that Leu-15 and Gly-17 are packed against the protein core, and that the side chains of Ile-18 and Asn-22 are partially buried and in close proximity to each other. As the N-loop is conformationally dynamic, it is possible that the side chains of Leu-15, Gly-17, and Ile-18 become exposed and engage in binding or indirectly influence the binding of the solvent-exposed residues. Consistent with our observations, mutational studies  have shown that Leu-12, His-19, and Lys-21 are critical for receptor binding and activation (14,41).
The binding data also show significant perturbation of residues in the second and third ␤-strands and the C-terminal helix (Fig. 5). In the case of third ␤-strand residues that are perturbed, Arg-48 and Lys-49 are solvent-exposed (accessible surface area ϳ0.7), and Ala-50, Leu-52, and Asn-53 are largely buried and involved in packing interactions against the N-loop residues. Similarly, the significantly perturbed C-terminal residues Ile-62 and Ile-63 are also involved in packing interactions (Fig. 7).

DISCUSSION
NMR and x-ray structure determination of a large number of CXC and CC chemokines have shown that they share a common structural fold at the monomer level, and biophysical measurements have shown that chemokine dimerization constants can vary by orders of magnitude from very weak (K d ϳmM) to very strong (K d ϳnM) (3)(4)(5)42). The monomer structure consists of three antiparallel ␤-strands followed by a C-terminal ␣-helix. However, the dimer structures are distinctly different between CC and CXC chemokines. CXC chemokines dimerize using the first ␤-strand and ␣-helix forming a globular structure, and the dimer interface is located away from the receptor binding N-terminal and N-loop regions (Fig.  1). CC chemokines dimerize using their N-loop residues and form an extended structure, and so their dimerization and receptor binding domains overlap.

CXCL1 Binding to CXCR2
APRIL 26, 2013 • VOLUME 288 • NUMBER 17 phil function by activating two receptors, CXCR1 and CXCR2 (27). These chemokines are characterized by a highly conserved N-terminal ELR motif and share the property of reversible monomer-dimer equilibrium. CXCL8 alone is active for both CXCR1 and CXCR2 and the others only for CXCR2. CXCL8 monomer is highly active for both CXCR1 and CXCR2. However, CXCL8 dimer is marginally active for CXCR1 and differentially active for CXCR2, with the dimer being as active in some functional assays and less active in others (10). Nothing is known regarding the role of dimerization for CXCR2-specific chemokines.
In this study, we show that the CXCL1 dimer is a potent agonist and activates CXCR2 with potencies comparable with that of the WT monomer. Sequence comparison and structure-function studies have shown that the chemokine N-loop residues determine receptor selectivity. Comparison of the CXCL1 and CXCL8 structures and dynamic characteristics suggests that differences in N-loop conformational dynamics between monomers and dimers and between CXCL1 and CXCL8 must be responsible for the observed functional differences. We have shown previously that dimer interface interactions disfavor the binding of the CXCL8 dimer to the CXCR1 receptor N-domain. On the other hand, our current NMR studies indicate that the dimer interface interactions play no role in the binding of the CXCL1 dimer, providing a structural basis for the similar CXCR2 receptor binding affinities of the monomer and the dimer.
Our NMR studies also indicate that similar direct and indirect interactions mediate the binding of both the CXCL1 dimer and monomer to the receptor N-domain and that these interactions mediate receptor activation. We had previously proposed a flycasting mechanism for the binding of CXCL8 binding to the CXCR1 N-domain on the basis that both the N-loop and the N-domain residues are conformationally dynamic (25, 43). We observe similar interactions for the CXCL1/CXCR2 axis and propose that the fly-casting mechanism best describes the binding of CXCL1 monomers and dimers to the CXCR2 N-domain and that coupling between Site I and Site II mediates receptor activation.
It has been shown previously that disulfide-trapped CCL2 and CCL4 dimers are inactive (7,8). The inactivity of CC chemokine dimers is not surprising as dimerization occludes the receptorbinding N-terminal and N-loop residues. On the other hand, a disulfide-trapped CXCL12 dimer elicits robust ERK phosphorylation but weak Ca 2ϩ release and fails to elicit chemotaxis (11,12). CXCL12, like CXCL8, is a CXC chemokine, and therefore dimerization does not limit receptor accessibility. We propose that dimerization perturbs the distribution of the conformational substates, which in turn differentially affects the activation of various downstream signaling pathways.
Designed monomers of CXCL10, CCL5, and CCL4 were shown to be inactive in mouse models (44,45). However, a recent in vivo study has shown that a designed CCL2 monomer has WT-like activity in contrast to a previous study that had reported that CCL2 monomer was inactive (45,46). The differences have been attributed to differences in experimental design including the dosage, time point, and the organ studied (46). On the other hand, both the CXCL8 monomer and dimer are active in the mouse lung and peritoneum models (47,48). It has been proposed in the case of CXCL8 that the monomer-dimer equilibrium and receptor and cell surface glycosaminoglycan interactions together regulate in vivo recruitment and that differential glycosaminoglycan interactions of the monomer and dimer play crucial roles in defining gradient formation which in turn regulate the kinetics, flux, and duration of the neutrophil recruitment into the target tissue. Animal model studies using the full-length trapped monomer similar to those used in the case of CXCL8, and the trapped dimer, are necessary to understand better how CXCL1 monomer-dimer equilibrium regulates neutrophil trafficking.
In summary, our current study on the disulfide-linked CXCL1 dimer and previous studies for various chemokines collectively indicate that the role of dimerization varies from chemokine to chemokine and that differential interactions of the dimer for their receptors and glycosaminoglycans, as well as the monomer-dimer equilibrium, dictate in vivo function. We propose that chemokines have evolved to exploit the property of reversibly existing as monomers and dimers as a means to regulate a wide variety of physiological functions.