Molecular determinants of receptor binding and signaling by the CX3C chemokine fractalkine.

Fractalkine/CX3CL1 is a membrane-tethered chemokine that functions as a chemoattractant and adhesion protein by interacting with the receptor CX3CR1. To understand the molecular basis for the interaction, an extensive mutagenesis study of fractalkine's chemokine domain was undertaken. The results reveal a cluster of basic residues (Lys-8, Lys-15, Lys-37, Arg-45, and Arg-48) and one aromatic (Phe-50) that are critical for binding and/or signaling. The mutant R48A could bind but not induce chemotaxis, demonstrating that Arg-48 is a signaling trigger. This result also shows that signaling residues are not confined to chemokine N termini, as generally thought. F50A showed no detectable binding, underscoring its importance to the stability of the complex. K15A displayed unique signaling characteristics, eliciting a wild-type calcium flux but minimal chemotaxis, suggesting that this mutant can activate some, but not all, pathways required for migration. Fractalkine also binds the human cytomegalovirus receptor US28, and analysis of the mutants indicates that US28 recognizes many of the same epitopes of fractalkine as CX3CR1. Comparison of the binding surfaces of fractalkine and the CC chemokine MCP-1 reveals structural details that may account for their dual recognition by US28 and their selective recognition by host receptors.

Fractalkine/CX3CL1 is a membrane-tethered chemokine that functions as a chemoattractant and adhesion protein by interacting with the receptor CX3CR1. To understand the molecular basis for the interaction, an extensive mutagenesis study of fractalkine's chemokine domain was undertaken. The results reveal a cluster of basic residues (Lys-8, Lys-15, Lys-37, Arg-45, and Arg-48) and one aromatic (Phe-50) that are critical for binding and/or signaling. The mutant R48A could bind but not induce chemotaxis, demonstrating that Arg-48 is a signaling trigger. This result also shows that signaling residues are not confined to chemokine N termini, as generally thought. F50A showed no detectable binding, underscoring its importance to the stability of the complex. K15A displayed unique signaling characteristics, eliciting a wild-type calcium flux but minimal chemotaxis, suggesting that this mutant can activate some, but not all, pathways required for migration. Fractalkine also binds the human cytomegalovirus receptor US28, and analysis of the mutants indicates that US28 recognizes many of the same epitopes of fractalkine as CX3CR1. Comparison of the binding surfaces of fractalkine and the CC chemokine MCP-1 reveals structural details that may account for their dual recognition by US28 and their selective recognition by host receptors.
Chemokines are proinflammatory proteins that coordinate the immune response by directing the migration of leukocytes (1)(2)(3). For humans alone, the chemokine superfamily comprises over 40 members that cluster into four families (CC, CXC, C, and CX3C) according to the number and spacing of conserved cysteines (4). Chemokines mediate their effects by binding and signaling through seven-transmembrane G-protein-coupled receptors (7TMRs), 1 of which 18 human receptors have been identified thus far (2,5). The patterns of ligand-receptor recognition are complex since chemokines and chemokine receptors typically interact with multiple partners. Furthermore, a given chemokine can act as an agonist, antagonist, or inverse agonist in the context of different receptors (6 -10). Despite the high level of promiscuity, interactions generally occur between receptors and ligands within a particular subfamily, an important exception being viral proteins that typically show broad spectrum recognition profiles (11)(12)(13)(14). A large number of studies have aimed to dissect the molecular determinants of chemokine receptor binding, signaling, and specificity (15)(16)(17)(18)(19)(20). However, these details remain poorly understood because structural information on the receptors is difficult to obtain, and the important features of the ligands are encrypted in remarkably conserved tertiary structures.
Fractalkine/CX3CL1 and its receptor CX3CR1 are currently the only known members of the CX3C family (21,22). Unlike other chemokines except CXCL16 (23), fractalkine is a membrane-anchored protein consisting of a chemokine module attached to the cell membrane via a mucin-like stalk, which serves as a presentation vehicle (24). A soluble form, generated by protease cleavage near the membrane, has been observed in the supernatants of fractalkine cell transfectants (25). This form induces chemotaxis of monocytes, T cells, and natural killer cells. No difference in binding or signaling is observed between fractalkine variants with different lengths of mucin stalk (26), and functional responses can be stimulated by the chemokine module alone, which indicates that it is the effector domain (22,25). Fractalkine is up-regulated on the surface of activated endothelial cells and can induce firm adhesion of CX3CR1-expressing cells, independent of G-protein activation (27). These observations suggest that fractalkine and CX3CR1 fulfill special roles in leukocyte trafficking at the endothelium. Fractalkine and CX3CR1 probably also have important homeostatic and developmental roles, given their high constitutive expression in a variety of non-hematopoietic tissues, including the brain (25,28). Their functions in the central nervous system are not fully defined, but there is recent evidence that fractalkine serves as an anti-apoptotic factor promoting microglial and hippocampal neuron survival (29,30). Like many other chemokine receptors, CX3CR1 can function as an human immunodeficiency virus cofactor in vitro, and fractalkine can inhibit viral entry into cells, but it is unclear whether these activities are relevant in vivo. Although it has been reported that a variant haplotype of CX3CR1 (CX3CR1-I249 M280) is associated with accelerated progression to AIDS (31), these data were recently called into question (32).
We previously solved the NMR structure of fractalkine's chemokine domain and described its interaction with an Nterminal fragment of CX3CR1 (33). In the present study, we use mutagenesis to identify residues that are important for receptor binding and signaling. Although CX3CR1 is the endogenous receptor, fractalkine also binds with subnanomolar affinity to US28, a 7TMR encoded by human cytomegalovirus (14). Accordingly, we examined binding of the mutants to US28 to determine whether the viral receptor recognizes the same "hotspots" on fractalkine as CX3CR1. These studies further define the structural and chemical elements necessary for fractalkine:CX3CR1 interactions and provide insight into the molecular basis of chemokine:receptor specificity. Since chemokines and their receptors have been implicated in a number of diseases (34 -37), the results may also aid in the development of agents that block their interaction (38).

EXPERIMENTAL PROCEDURES
Gene Construction of Human Fractalkine Variants for Expression in E. coli-With the exception of an N-terminal deletion variant of fractalkine (⌬NT), all constructs were made by polymerase chain reaction mutagenesis of the WT template (human fractalkine residues 1-76) and cloned into a T7-driven expression vector, pAED-4 (39). Sequences were confirmed by double-stranded DNA sequencing.
Protein Expression and Purification in Escherichia coli-Proteins were expressed in E. coli BL21(DE3) pLysS cells. Cells were grown at 37°C in Luria Broth containing 100 g/ml ampicillin and 34 g/ml chloramphenicol. When the cell density reached 0.5-0.6 A 600 , protein expression was induced by addition of isopropyl-D-thiogalactopyranoside to a final concentration of 0.5 mM. After 1 h, rifampicin was added to a final concentration of 20 g/ml. 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 in MOPS minimal medium containing 1 g/liter [ 15 N]ammonium sulfate (40).
All fractalkine mutants were purified from inclusion bodies. Typically, cells from a 0.5-liter growth were sonicated in 50 ml of lysis buffer containing 10 mM K 2 PO 4 , 1 mM EDTA, 100 mM NaCl, pH 7.0, and centrifuged at 10,000 rpm for 20 min. Refolding and oxidation were accomplished by solubilizing the cell pellet in a minimal amount of 6 M guanidinium chloride and diluting 50-fold into a redox buffer containing 100 mM Tris, 1 mM EDTA, 0.2 mM oxidized glutathione, 1 mM reduced glutathione, pH 8.0. After overnight stirring at 4°C, the solution was adjusted to pH 2.0, centrifuged, and filtered. Final purification was achieved by reversed-phase HPLC using a 4.6 ϫ 250-mm C 18 column. Proteins were eluted with an increasing acetonitrile gradient containing 0.1% trifluoroacetic acid, typically at 35 Ϯ 5% acetonitrile. The proteins were then lyophilized, redissolved in water at 1-5 mg/ml and stored in small aliquots at Ϫ80°C. Yields generally ranged from 0.25 to 2 mg of purified protein/0.5 liter of growth.
The ⌬NT variant was obtained by dissolving WT fractalkine chemokine domain at 1 mg/ml in 35 mM Tris, pH 8, incubating with 15 g of aminopeptidase (Peprotech, Rock Hill, NJ) per 1 mg of protein for 36 h at room temperature, and repurifying by reversed-phase HPLC. This treatment resulted in the removal of all residues N-terminal to the first cysteine.
The molecular weights and purity of all proteins were confirmed by electrospray-ion trap mass spectrometry (Bruker Esquire) to mass accuracy within 1 atomic mass unit. All protein concentrations were determined using a ⑀ 280 extinction coefficient calculated from the amino acid composition (41).
Glial Cell Cultures-Mixed glial cell cultures expressing CX3CR1 were established as described previously using cortexes from newborn rats (43).
Transfected Cell Cultures-Full-length cDNA for human CX3CR1 (GenBank accession no. NM_001337) was cloned into the BamHI site of the mammalian expression vector pIRES-hyg (CLONTECH, Palo Alto, CA). Human embryonic kidney cells (HEK-293) constitutively expressing the EBNA-1 protein of Epstein-Barr virus (293E EBNA; Invitrogen, Carlsbad, CA) were transfected by calcium phosphate coprecipitation (MBS kit; Stratagene, La Jolla, CA) and selected with 200 g/ml G418 and 250 g/ml hygromycin (Calbiochem, San Diego, CA) for 2 weeks. The stable pool was cloned by limiting dilution into 96-well plates and stable clones maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum (fetal calf serum), 10 mM HEPES buffer, 2 mM glutamine, 2 mM sodium pyruvate, and 0.01 mg/ml gentamycin. Functional characterization of CX3CR1-expressing clones was performed by measurements of calcium mobilization (see below) and equilibrium binding of 125 I-fractalkine. Stable transfectants expressed between 35,000 and 40,000 specific binding sites/cell based on Scatchard analysis (43).
US28 transfectants were prepared as described previously (14). US28 DNA (GenBank accession no. L20501, originally cloned from the VHL/E strain of human cytomegalovirus) was inserted into a pTEJ-8 expression vector. The vector was transfected via the calcium phosphate method, into COS-7 cells that had been grown in 10% CO 2 at 37°C in DMEM 1885 supplemented with 10% fetal calf serum, 2 mM glutamine, and 0.01 mg/ml gentamycin.
Binding Assays-Whole cell binding of fractalkine to microglia and 293E-CX3CR1 cells was performed in a 96-well format, with ϳ5 million PBS-rinsed cells/well (microglia) or 1 million cells/well (HEK-293E transfectants). Cells were incubated for 1 h at 22°C in binding buffer (25 mM HEPES, 5 mM MgCl 2 , 1 mM CaCl 2 , 120 mM NaCl, 0.1% BSA, pH 7.2) containing 125 I-labeled human fractalkine chemokine domain (100 Ci/mmol, Amersham Pharmacia Biotech) and various concentrations of unlabeled chemokine mutants in a final volume of 100 l. Following incubation, reactions were terminated by filtration and washing onto GF/B filter plates (Whatman) using a Filtermate 196 Harvester (Packard Instruments, Downers Grove, IL), dried, scintillant added, and counted on a Top Count (Packard). All analyses (EC 50 and IC 50 ) were performed using Prizm software (GraphPad, San Diego, CA). For binding to US28 COS transfectants, 80,000 cells were transferred to 24-well culture plates 1 day after transfection. Two days after transfection, competition binding experiments were performed on whole cells for 3 h at 4°C using 12 pM 125 I-labeled human fractalkine (2000 Ci/mmol, Amersham Pharmacia Biotech) plus variable amounts of unlabeled fractalkine mutants suspended in binding buffer (0.5 ml, 50 mM Hepes pH 7.4, 1 mM CaCl 2 , 5 mM MgCl 2 , and 0.5% BSA). After incubation, the cells were washed four times in 4°C binding buffer supplemented with 0.5 M NaCl to reduce nonspecific binding. Determinations were made in triplicate. IC 50 values were determined by nonlinear regression using Inplot 4.0 software (GraphPad).
Chemotaxis Assays-Chemotaxis assays were performed using a 48well chamber apparatus (Neuroprobe, Cabin John, MD). Briefly, microglia cultures were detached with Versene (Life Technologies, Inc.) and resuspended in DMEM containing 1% BSA at a density of 4 ϫ 10 6 cells/ml. Various dilutions of the chemokines were prepared in DMEM ϩ 1% BSA. Aliquots of 26 l were distributed in quadruplicates in the lower wells. An 8-m pore size polycarbonate filter, coated with poly-D-lysine on the lower surface to favor microglia adhesion, separated the upper wells containing 50 l of cell suspension. The chamber was incubated for 2 h at 37°C in a moist 5% CO 2 atmosphere. After incubation, the non-migrating cells adherent to the upper surface of the filter were washed and scraped. The filter was subsequently fixed in methanol, stained with Diffquick (Dade, Aguada, Puerto Rico), and dried on a glass slide. The number of migrating cells was counted (magnification, ϫ400 or ϫ1000), according to their density. At least five high power fields were examined in each well. The results are expressed as the mean cell number Ϯ S.D.
Calcium  Results calibrated to internal standard curves according to standard protocols, and expressed as nanomolar shifts in intracellular calcium mobilization, are representative of n ϭ 3 independent trials. A complete summary of the binding, chemotaxis, and calcium flux data is included as supplementary material, available in the on-line version of this article.

RESULTS
Mutagenesis Strategy-The surface-exposed residues of fractalkine's chemokine domain were identified by computing the solvent-accessible surface area from the NMR solution structure (33) using the program ROC (44). Residues with solventaccessible surface area Ͼ 30% were individually mutated to alanine. We focused primarily on residues with basic or large hydrophobic side chains because they showed the largest effects in a similar mutagenesis study of MCP-1 (45,46). Those affected were subsequently mutated to other amino acids in order to probe the required physicochemical features of the side chain. In total, 29 mutants were generated at an average purity level of 95% based on analytical HPLC.
The N Terminus of Fractalkine Is Important for CX3CR1 Binding and Signaling, but Extension with an N-terminal Methionine Has No Effect-It is well known that chemokines are sensitive to modifications of their N termini (18,(47)(48)(49)(50). We had previously expressed the chemokine domain of fractalkine (residues 1-76) in E. coli and found that the initiating methionine was retained (33). Consequently, before embarking on a large scale mutagenesis study, it was necessary to determine if the presence of the N-terminal Met affected the interaction of fractalkine with CX3CR1. Fig. 1A shows representative radioligand competition profiles of Metϩ fractalkine (recombinant) and MetϪ fractalkine (synthetic protein from Gryphon Sciences, South San Francisco, CA) binding to CX3CR1 on rat microglia. Within experimental error, the affinity of Metϩ was identical to MetϪ on both microglia and CX3CR1 transfected HEK-293E cells (Table I, Fig. 1A), and both were equipotent in inducing chemotaxis (Fig. 1B) and calcium flux of microglia (Fig. 2). Metϩ fractalkine also inhibited calcium flux by MetϪ and vice versa (Fig. 2). Since no significant differences were observed between the two variants, all of the mutants except ⌬NT retain the initiating methionine, and the residues are numbered according to Met-1. Henceforth, we will refer to Metϩ fractalkine as WT*.
In order to assess the importance of the N terminus, a truncated variant of fractalkine (⌬NT) was prepared. The binding affinity of ⌬NT was reduced by ϳ2 orders of magnitude on both microglia and CX3CR1 transfectants (Table I, Fig. 1A). Fur-thermore, it was unable to induce chemotaxis (Fig. 1B) or calcium flux (Fig. 2) and could not desensitize microglia to WT* (Fig. 2).
Lys-8 in the N Terminus Is Important for Binding and Signaling-To identify the critical features of the N terminus, individual residues were also mutated. Two mutants, H4A and T7A, were not characterized because they did not express. H3A, V6A, and N10A produced no measurable differences from WT* in binding competition (Fig. 3) or calcium flux assays (data not shown), whereas chemotaxis was slightly reduced, particularly for N10A at subnanomolar concentrations (data not shown). Much larger effects were observed upon mutation of Lys-8. K8A showed 37-and 114-fold reductions in binding to microglia and CX3CR1 transfectants, respectively (Table I, (Table I, Fig. 3), suggesting that electrostatics contribute little to the stability of the interaction. Although not completely suppressed, chemotaxis to K8A was significantly reduced relative to WT*, particularly at subnanomolar concentrations (Fig. 4A). K8A also elicited a barely detectable calcium flux and could not desensitize the WT-like mutant, R68A (Fig.  5). As expected from the similarity in binding affinities, the chemotaxis profile of K8E mirrored that of K8A within experimental error (Fig. 4A). Important Basic Residues in the "Core Domain": Lys-15, Lys-19, Lys-37, Arg-38, and Arg-45-In the extended loop following the CX3C motif, Lys-15 was identified as an important residue. The affinity of K15A was reduced by a factor of 22 and 13 on microglia and transfectants, respectively. Although K15A elicited a WT-like calcium flux and could desensitize the receptor to further chemokine stimulation (Fig. 5), chemotaxis was significantly compromised (Fig. 4B). The glutamate variant, K15E, caused a dramatic 677-fold reduction in binding to transfectants, suggesting that Lys-15 may interact with an acidic residue in the receptor. Surprisingly, this difference was not observed for binding to microglia (Table I, Fig. 4B). Although cellular differences modulating chemokine function could explain the differential effects, it is interesting to note that the only region in the human (transfectant) receptor that contains acidic residues not present in the rat (microglia) receptor is the third extracellular loop surrounding Cys-265.
K19A and R38A showed modest 8-fold reductions in binding to CX3CR1 transfectants and either no change (K19A) or a 16-fold change (R38A) on microglia. Mutation of either residue to Glu caused, at most, another 2.5-fold loss in affinity indicating that charge is not important for the stability of the interaction. Calcium flux was similar to WT* for both alanine mutants, whereas chemotaxis at 1 nM protein was slightly

FIG. 3. Summary of the binding affinity of fractalkine mutants to microglia and CX3CR1 transfectants.
Black columns represent binding to rat microglia. Gray columns represent binding to transfectant cells. All values are normalized to those obtained for WT* binding to the corresponding cell type: fold change ϭ mutant K d /WT* K d . The affinities for R48E binding to both microglia and transfectants and F50A binding to microglia (highlighted by asterisked arrows) were too weak to quantify. The sequence of WT* and the secondary structure of the protein are shown above the figure. reduced (data not shown). R45A had a larger impact on affinity (24-and 18-fold on microglia and transfectants, respectively) and chemotaxis, although calcium flux at 10 nM was robust (data not shown). No additional effect was observed upon mutation to Glu. Finally, mutation of Lys-37 to either Ala or Glu showed a 24 -50-fold reduction in binding affinity and corresponding changes in chemotactic activity, although calcium flux was like WT* (data not shown).
The Hotspots of Fractalkine: Arg-48 and Phe-50 -Apart from the N terminus and Lys-15, mutation of Arg-48 and Phe-50 produced by far the largest effects on binding. The affinity of R48A compared with WT* was reduced by 93-and 208-fold on microglia and CX3CR1 transfectants, respectively. Furthermore, R48A was unable to elicit a calcium response, desensitize to further chemokine stimulation (Fig. 5), or induce chemotaxis (Fig. 4C). Mutation to Gln (R48Q) largely restored affinity (Table I), but, interestingly, calcium flux and chemotaxis were still significantly compromised (Table I). No detectable binding was observed for R48E, providing evidence for a probable interaction between Arg-48 and an acidic residue of the receptor. Even more dramatic results were observed for F50A, which showed a 489-fold loss in affinity for CX3CR1 transfectants and no detectable binding to microglia. Accordingly, no calcium flux or chemotaxis could be measured for this mutant (Fig. 5). Mutation to leucine (F50L) restored some of the binding. However, it was still reduced by 78-fold, and both calcium flux and chemotaxis were markedly affected (Table I, Fig. 4D).
US28 Recognizes a Similar Binding Surface on Fractalkine-Using the same panel of fractalkine mutants, we investigated binding to the human cytomegalovirus receptor US28. Overall, the extent to which the mutations affected binding was not as dramatic as for CX3CR1 (Fig. 6). Like CX3CR1, no difference in binding was observed between WT and WT*, but deletion of the N terminus resulted in a large (128-fold) reduction in affinity. Of the alanine mutants that were tested, R48A had the largest effect, producing a 26-fold loss in binding, whereas K8A and F50A showed modest 5-6 fold reductions. As was true for CX3CR1, R48E showed very weak binding, and an accurate K d could not be derived from the data. Since mutation to Gln restored much of the binding, a positive charge may be important at this position. Overall, we did not identify any residues necessary for interaction with US28 that were not involved in the interaction with CX3CR1.
The Fractalkine Mutants Are Structurally Unperturbed-To investigate the structural integrity of the fractalkine mutants, we recorded two-dimensional 1 H-15 N HSQC spectra of 15 Nlabeled WT* and three mutants that had a large impact on binding: K8A, R48A, and F50A. All mutants showed the expected number of cross-peaks. The NMR signals were uniform in size and well dispersed, indicating that the mutations did not cause misfolding of the proteins. Fig. 7 shows representative spectra of WT* and R48A to illustrate the similarity of the chemical shift patterns. The differences are localized to residues near the site of mutation: Arg-48 and Leu-49 are missing, whereas Gln-46 and His-47 are slightly shifted. Similarly, only small differences were observed for the mutants K8A and F50A when compared with WT* (data not shown). These data confirm that the observed effects on binding and signaling are a direct consequence of the mutation and not an indirect consequence of structural perturbations.

DISCUSSION
Our goal is to understand the molecular details of how chemokines bind and activate their receptors. Given that chemokines have remarkably similar folds, biochemical studies are necessary to determine what features control the specific recognition of some receptors over others. Understanding how viral receptors like US28 achieve high affinity interactions with a broad spectrum of chemokines is also an important endeavor. To this end, we identified residues of fractalkine that contribute to the interaction with the host receptor CX3CR1, and with the cytomegalovirus receptor US28. Fig. 8 (left) shows a GRASP model of fractalkine highlighting residues that, when mutated, had a large (Ͼ10-fold) impact on CX3CR1 binding and/or signaling (Lys-8, Lys-15, Lys-37, Arg-45, Arg-48, and Phe-50). The identification of Lys-8 and Arg-48 as important for the interaction with CX3CR1 is also consistent with a recent mutagenesis analysis of basic residues in fractalkine (26). Below, we discuss the results in more detail and compare them to what is known about other chemokine receptor systems, particularly MCP-1:CCR2.
The N Terminus Contributes to Binding Affinity and Signaling Specificity-Like other chemokines, the N terminus of fractalkine is disordered in the solution structure (33) and probably becomes ordered only upon receptor engagement. The importance of chemokine N termini in receptor activation is well established (15,18,51), and fractalkine is no exception, as removal of the first 8 residues abolishes function. Although the contribution of chemokine N termini to binding varies considerably, deletion of fractalkine's N terminus caused a loss in affinity of more than 2 orders of magnitude. By comparison, N-terminal deletion of MCP-1 causes only a 7-fold reduction in affinity for CCR2 (45,52), whereas short peptides derived from the N terminus of SDF-1 have sufficient affinity to bind CXCR4 and stimulate chemotaxis in the absence of the rest of the chemokine (53).
The importance of specific residues within the N termini also varies significantly among chemokines, thereby encoding func-tional specificity. For fractalkine, extension of the N terminus with a Met had no effect on binding or function whereas similar extensions of MCP-1 (18) and RANTES (47) produce receptor antagonists. Among the N-terminal residues, Lys-8 is particularly crucial, somewhat akin to the ELR motif of interleukin-8 (20,54), and specific residues identified as critical to receptor activation in RANTES (19), SDF-1 (15), and eotaxin (55). In contrast, the length rather than the amino acid composition of the N terminus is important for MCP-1 (45).
K15A Displays Unique Signaling Characteristics-The functional responses of CX3CR1 to the fractalkine variant K15A are particularly instructive from the standpoint of general issues of 7TMR signaling. In contrast to mutants that showed marked effects on both calcium and chemotaxis (⌬NT, F50A, and R48A), K15A produced a normal calcium flux and could desensitize to further stimulation, but had a greatly diminished chemotaxis response. For certain chemokines including MCP-1 (56) and fractalkine (43), receptor binding has been shown to lead to mitogen-activated protein kinase activation, a critical event controlling actin polymerization and cytoskeletal reorganization, which in turn are necessary for migration. In preliminary analysis of mitogen-activated protein kinase activation stimulation, we noted no appreciable activity induced by K15A compared with WT fractalkine (data not shown), suggesting a possible cause for the impaired chemotaxis. Differential activation of pathways was observed previously with MCP-1, where the Y13A mutant was able to inhibit adenylate cyclase and stimulate calcium flux, but not chemotaxis (46). Based on these findings, we suggested that certain molecular features of chemokines are required for activation of some pathways but not others, and fractalkine's K15A provides additional support for this hypothesis. Interestingly, the location of Lys-15 on fractalkine's surface is similar to that of Tyr-13 on MCP-1 (Fig. 8).
Phe-50 and Arg-48 Are Essential Residues That Are Spatially Removed from the N Terminus-Many mutagenesis studies of chemokines have targeted the N terminus and the extended loop (the "N-loop") following the second cysteine because of their importance for binding and particularly signaling (15,19,46,55,57,58). Consequently, it is interesting that the two most critical residues of fractalkine, Arg-48 and Phe-50, lie outside of these regions. Like the N terminus and Lys-8, Arg-48 is essential for signaling because mutation to Ala completely elimi-nated calcium flux and chemotaxis, but not binding. Mutation of Arg-48 to Glu eliminated binding whereas mutation to Gln restored much of the affinity, presumably due to the size and hydrogen bonding capacity of Gln compared with Ala. Since calcium and chemotaxis are still markedly affected by R48Q, it appears that a basic charge is required for receptor activation. The significant difference between R48Q and R48E also suggest that Arg-48 interacts with an acidic region of the receptor (see below). For the mutant F50A, a 489-fold reduction in affinity was observed on transfectants and no binding could be detected on microglia. The binding and signaling of F50L was much improved over the alanine mutant; however, it was still significantly less active than the WT protein, suggesting that an aromatic side chain at this position is important. These data provide further evidence that residues critical for signaling are not confined to the N termini of chemokines but can be dispersed over the receptor-binding surface (46,55,59). Thus, there may be significant variability in the mechanisms involved in activation of different chemokine receptor systems.
Similarities in the Recognition Surfaces That May Account for Broad Spectrum Recognition by US28 -US28 not only interacts with fractalkine, but also binds CC chemokines such as MCP-1 with subnanomolar affinity (14). Our previous mutagenesis studies identified residues important for binding of MCP-1 to CCR2, its cognate receptor (45,46). Preliminary data on a subset of those mutants suggest that, as with fractalkine, US28 uses the same binding epitopes on MCP-1 as the host receptor (data not shown). Therefore, a comparison of the binding surfaces of fractalkine and MCP-1 may shed light on what enables the viral receptor to engage both chemokines.
At low resolution, the receptor binding surfaces of fractalkine and MCP-1 have a lot in common, as they consist of a largely basic surface with one aromatic residue (Fig. 8). In contrast, the CXC chemokine interleukin-8, which does not bind US28, has a much more hydrophobic surface (45,60). The distribution of basic residues on the tertiary structures of fractalkine and MCP-1 are also similar (Fig. 8). Furthermore, both chemokines have residues that are highly sensitive to charge swap mutations (Arg-48 in fractalkine, Arg-24 in MCP-1), suggesting that they may interact with acidic regions of their receptors. Many studies have demonstrated that the receptor N termini are important for binding chemokines (61)(62)(63)(64)(65). We showed previously that a DYDY sequence in the N terminus of CCR2 was critical for binding MCP-1 (45), and thus may be involved in contacting MCP-1's R24. Others have shown that this motif is tyrosine-sulfated in vivo, and the post-translational modification is important for ligand binding (66). Interestingly, CX3CR1 and US28 also contain potential tyrosine sulfation sites (EYDD and DYDD, respectively) in their N termini (67) that could be involved in interactions with basic epitopes on the chemokines. We plan to pursue these questions in future studies.
Differences in the Recognition Surfaces That Contribute to Specificity-Despite the similarities in the binding epitopes, closer inspection of the structures reveals sufficient differences to account for the class-specific recognition of fractalkine by CX3CR1 and MCP-1 by CCR2. Previously, we noted that subtle structural and chemical features create different topological landscapes on the two chemokines (33). Other salient features emerge from the mutagenesis data. Lys-8 and another critical residue, Lys-15, flank fractalkine's characteristic CX3C motif (residues 9 -13). In contrast, the corresponding region of MCP-1 contains an aromatic residue, Tyr-13. Tyr-13 was the single most important residue for binding and activation of CCR2, as mutation to Ala completely eliminated chemotaxis. It is also one of the most important features distinguishing MCP-1 and fractalkine, as it produces a large hydrophobic bump on MCP-1's surface that is not present in fractalkine. Like MCP-1, fractalkine's single most important residue is aromatic, Phe-50. Since it is located in a different region of fractalkine's surface relative to Tyr-13 in MCP-1, it is probably also a critical specificity determinant.
In summary, we have identified several residues of fractalkine that are important for binding and signaling of its host receptor CX3CR1 and for binding the viral receptor US28. Our results show that both receptors recognize similar binding epitopes on the surface of the ligand. Comparison of the recognition surfaces of fractalkine and MCP-1 suggests plausible reasons for the broad spectrum recognition of US28, as well as selective binding to their cognate receptors. We demonstrated that key signaling residues are not confined to the N termini and N-loop of chemokines. We also observed differential stimulation of signaling pathways by specific mutants, which has broad implications for 7TMR function.