Ultrastructure and Function of the Fractalkine Mucin Domain in CX3C Chemokine Domain Presentation*

Fractalkine (FKN), a CX3C chemokine/mucin hybrid molecule on endothelium, functions as an adhesion molecule to capture and induce firm adhesion of a subset of leukocytes in a selectin- and integrin-independent manner. We hypothesized that the FKN mucin domain may be important for its function in adhesion, and tested the ability of secreted alkaline phosphatase (SEAP) fusion proteins containing the entire extracellular region (FKN-SEAP), the chemokine domain (CX3C-SEAP), or the mucin domain (mucin-SEAP) to support firm adhesion under flow. CX3C-SEAP induced suboptimal firm adhesion of resting peripheral blood mononuclear cells, compared with FKN-SEAP, and mucin-SEAP induced no firm adhesion. CX3C-SEAP and FKN-SEAP bound to CX3CR1 with similar affinities. By electron microscopy, fractalkine was 29 nm in length with a long stalk (mucin domain), and a globular head (CX3C). To test the function of the mucin domain, a chimeric protein replacing the mucin domain with a rod-like segment of E-selectin was constructed. This chimeric protein gave the same adhesion of peripheral blood mononuclear cells as intact FKN, both when immobilized on glass and when expressed on the cell surface. This implies that the function of the mucin domain is to provide a stalk, extending the chemokine domain away from the endothelial cell surface to present it to flowing leukocytes.

Leukocyte trafficking of cells out of the bloodstream and into sites of inflammation requires multiple steps (1,2). In the classic pathway of leukocyte migration, the first step involves transient, selectin-mediated interactions between the rolling leukocytes and the endothelium (3,4). In the next step, integrins are activated by locally produced chemokines to induce firm adhesion of the leukocyte to endothelial cells (5)(6)(7). Leukocytes then extravasate through the vascular wall and into the tissue.
Recently, a new pathway by which leukocytes can be induced to firmly adhere to endothelium and traffic into inflamed tissues has been described. This process is mediated by the chemokine fractalkine (FKN), 1 expressed on endothelial cells, and its G-protein coupled receptor, CX 3 CR1, expressed on PBMC. Fractalkine (neurotactin in the mouse) is the first, and thus far only, member of the CX 3 C chemokine family, so named because the first two conserved cysteine residues are separated by three amino acids (8,9). Another unique characteristic of fractalkine is that it is the only chemokine that is expressed on the cell surface. By nucleotide sequence analysis, fractalkine consists of a chemokine head tethered to the cell surface by a mucin stalk, followed by a single transmembrane spanning domain and a short cytoplasmic tail (8,9). In addition, fractalkine expression on endothelium is increased by proinflammatory cytokines interleukin-1 and tumor necrosis factor-␣. The receptor for fractalkine, CX 3 CR1 (V28), is expressed on T cells, monocytes, macrophages, and natural killer cells (10,11). Fractalkine and CX 3 CR1 function as cell adhesion molecules under both static and dynamic conditions (10,12). Unlike other chemokine/G-protein coupled receptor interactions that require signal transduction and integrin activation for cell adhesion to occur, the adhesive interaction between fractalkine and CX 3 CR1 is independent of signal transduction or integrin function (12,13). Therefore, fractalkine and CX 3 CR1 provide an integrin-independent mechanism for leukocyte migration.
Compared with other chemokines, fractalkine has at least three unique features that may mediate its function as a cell adhesion molecule: 1) it is a transmembrane molecule with a cytoplasmic tail that may participate in signal transduction; 2) it has a mucin domain; 3) it is the only CX 3 C chemokine and has a three-dimensional structure that is slightly different from other chemokines (14). In this report, we sought to determine the role of the various fractalkine domains, but in particular the mucin domain, in cell adhesion.

EXPERIMENTAL PROCEDURES
Fusion Proteins-FKN-SEAP, CX3C-SEAP, and mucin-SEAP fusion proteins were produced as described previously. (10) A plasmid encoding CX3C-ESCR hybrid molecules was generated by polymerase chain reaction and subsequent ligation of the DNA fragments into pDREF-Hyg-SEAP. DNA encoding the CX 3 C domain was synthesized using primers 5Ј-CGCGTCGACTCAGCCATGGCTCCGATA-3Ј and 5Ј-CGCA-GATCTTAGGGCAGCAGCCTGGCG-3Ј, and digested with SalI and BglII. DNA encoding the E selectin consensus repeat (ESCR) domain was generated by polymerase chain reaction amplification of E-selectin cDNA in pMT2 (15) using the primers 5Ј-CGCAGATCTATTGTGAAC-TGTACAGCC-3Ј and 5Ј-CGCTCTAGATTCACAGGTAGGTAGCAG-3Ј. These fragments were digested with BglII and XbaI and then ligated together into the pDREF-Hyg SEAP expression vector digested with SalI and XbaI. The nucleotide sequence of both strands of the new construct (pCX3C-ESCR-SEAP) was determined to verify its identity. A plasmid encoding a CX3C-ESCR-GFP fusion protein in which the mucin domain of full-length fractalkine was replaced by ESCR was constructed by amplifying the transmembrane and cytoplasmic tail of fractalkine by polymerase chain reaction using primers 5Ј-CGCTCTA-GACAGGCGGTGGGGCTGCTG-3Ј and 5Ј-CTGAGGATCCCCACGGG-CACCAGGAC-3Ј. This fragment and the CX3C-ESCR fragment described above (prepared by digesting pCX3C-ESCR-SEAP with SalI and XbaI) were cloned into the pEGFP-N2 expression vector (CLON-TECH, Palo Alto, Ca) digested with XbaI and BamHI. The nucleotide sequence of the resulting plasmid (pCX3C-ESCR-GFP) was determined for verification. SEAP fusion proteins were produced in mammalian 293/EBNA-1 cells and soluble fractalkine was generated in a baculovirus system as described previously (10).
Parallel Plate Flow Chamber Adhesion Assay-Experiments to determine the interaction between CX 3 CR1 and various forms of fractalkine were carried out as described previously (12). Briefly, SEAP fusion proteins were immobilized on coverslips using 10 g/ml anti-SEAP antibody 8B6 (Sigma). Either K562, CX 3 CR1 transfected K562 cells (K562-CX3CR1), or purified human peripheral blood mononuclear cells were resuspended in flow buffer (1ϫ phosphate-buffered saline containing 0.75 mM CaCl 2 , 0.75 mM MgCl 2 , and 0.5% bovine serum albumin) at a concentration of 1 ϫ 10 6 cells/ml. The cells were allowed to interact with the substrate within the flow chamber at a shear stress of 0.25 dynes/cm 2 for 5 min. The shear stresses were adjusted using a Harvard Model 44 syringe pump (Harvard Apparatus, South Natick, MA). Following subsequent washes at shear stresses of 2, 5, and 10 dynes/cm 2 for 2 min each, the number of firmly attached cells to the coverslip was counted. In studies using transfected ECV304 cells as the target for binding, the adherent cells were plated onto the glass coverslips within a 6-well cluster dish the day before the experiment. All results were recorded and analyzed as described previously (12).
Gradient Sedimentation and Rotary Shadowing-Samples of fractalkine-SEAP or of fractalkine alone were layered on 15-40% glycerol gradients in 0.2 M ammonium bicarbonate and centrifuged in a SW50Ti rotor (Beckman) at 30,000 rpm for 16 h at 20°C. Sedimentation standards (catalase, 11.3 S; bovine serum albumin, 4.6 S; and ovalbumin, 3.5 S) were centrifuged in a separate gradient. Gradients were fractionated into 12 fractions, and the protein peaks were identified by SDS-polyacrylamide gel electrophoresis. Sedimentation coefficients were estimated relative to the standards. Peak fractions were supplemented to 40% glycerol, sprayed on mica, and rotary shadowed as described (16). Specimens were examined in a Phillips 301 transmission electron microscope and photographed at ϫ50,000 magnification. All length measurements were corrected for the presumed 1-nm shell of metal around the protein.
Flow Cytometry-The level of cell surface expression was determined by labeling the transfected cells with antibodies against either the chemokine domain of fractalkine (4F2) or the mucin domain (1D6) as described previously (12). The secondary reagent used was phycoerythrin-conjugated goat anti-mouse IgG (Sigma). Analysis was performed on a Coulter Epics XL flow cytometer (Beckman-Coulter, Fullerton, CA) and data analyzed using CellQuest software (Becton-Dickinson, Mountain View, CA).

RESULTS
To determine which domains of fractalkine may be necessary for leukocyte capture and firm adhesion, we constructed fusion proteins of the chemokine (CX 3 C) and mucin domains separately linked to SEAP. Based on our previous studies in which immobilized fusion proteins containing the entire extracellular region of fractalkine (FKN-SEAP) functioned as well in leukocyte capture and firm adhesion as cells expressing fractalkine, we reasoned that the extracellular domains (CX 3 C and mucin) contained all of the structures necessary for fractalkine-mediated cell adhesion (12). Both PBMC and K562 cells expressing CX 3 CR1 (K562-CX3CR1) on the cell surface were able to firmly adhere to immobilized fractalkine-SEAP but not to control SEAP proteins alone (Fig. 1A). CX3C-SEAP induced capture and firm adhesion of resting PBMC, but at a reduced number compared with FKN-SEAP (17 Ϯ 7 versus 105 Ϯ 13 cells/mm 2 , p Ͻ 0.005). In contrast, no interactions or firmly adherent cells were observed in association with immobilized mucin-SEAP. These results suggested that the chemokine domain provided the specificity of interaction with CX 3 CR1 and that the mucin domain acted to increase the affinity or avidity of CX 3 CR1 interaction.
The mucin domain could have acted either as a co-receptor with the CX 3 C domain or as an extender to present the CX 3 C domain to CX 3 CR1-expressing cells. To determine if the mucin domain may itself act as a cell adhesion molecule, mucin-SEAP fusion proteins were tested for their ability to bind to K562-CX3CR1 cells by flow cytometry and L1.2-CX3CR1 cells by receptor binding assays. While FKN-SEAP fusion proteins bound well to K562-CX3CR1 cells (12), mucin-SEAP did not detectably bind (not shown). Furthermore, while mucin-SEAP did not bind detectably to L1.2-CX3CR1 cells, both FKN-SEAP and CX3C-SEAP bound with similar high affinities (K d ϭ 113 and 136 pM, respectively) (Fig. 1B).
To determine if the mucin domain could act as a stalk or presentation molecule, we determined the ultrastructure of the fractalkine extracellular domain. Information on the structures of FKN-SEAP and soluble FKN was obtained by glycerol gradient sedimentation. Fig. 2 shows SDS-polyacrylamide gel electrophoresis of the two proteins after fractionating the gradi-ents. Each ran as a discrete peak, indicating a homogeneous sample. FKN-SEAP showed a broad band at ϳ150 kDa on SDS-polyacrylamide gel electrophoresis. The predicted mass of the protein backbone of FKN-SEAP is 100.6 kDa, indicating that FKN-SEAP has approximately 50 kDa of O-linked carbohydates. FKN-SEAP sedimented between catalase and bovine serum albumin, at an estimated 6.3 S. f/f min , where f is the observed frictional coefficient and f min is the frictional coefficient for an unhydrated sphere that could contain the protein mass, would be 1.1 if the protein were a monomer, and 1.7 if it were a dimer. If the protein were a monomer, the f/f min would imply a very compact, globular structure, while the value for a dimer would imply an elongated structure. It is known that SEAP forms a dimer (17), so the sedimentation data are consistent with an elongated dimer. Electron microscopy of FKN-SEAP (Fig. 3) showed a large central globular particle with two thin strands projecting from them. These thin strands are identified as the mucin stalk, and had a length of 26 nm. The mucin stalk was capped by a small globular domain about 3 nm in diameter, which we identify as the cytokine domain. The entire length of the FKN molecule is 29 nm.
The structure of the mucin domain as a long stalk suggested that it functioned to extend the chemokine domain for optimal interaction with cell surface CX 3 CR1. To confirm that the mucin domain acted as an extender and not as a co-receptor, we replaced the mucin domain in FKN-SEAP with the six short consensus repeat segments of E-selectin (Fig. 4A). ESCR is expected to have the same length (approximately 26 nm) as the mucin domain (19) and does not have any intrinsic binding capacity (15). Soluble CX3C-ESCR bound to L1.2-CX3CR1 cells with the same characteristics as CX3C-SEAP and FKN-SEAP (K d ϭ 130 pM) (Fig. 4B). Immobilized CX3C-ESCR functioned as well as FKN-SEAP in the capture and firm adhesion of resting PBMC and K562-CX3CR1 under physiologic flow conditions (Figs. 4C). These data indicated that extension was the important feature of the mucin domain.
While the above studies indicate that the mucin domain could be functionally replaced by a stalk of similar length (ESCR), they were performed with purified fusion proteins immobilized on glass coverslips. To determine whether ESCR could functionally replace the fractalkine mucin domain in the context of the cell surface, two chimeric molecules were constructed: a transmembrane fractalkine-green fluorescent protein (FKN-GFP) molecule; and a CX3C-ESCR-GFP fusion protein in which the mucin domain of FKN-GFP was replaced with the ESCR domain (CX3C-ESCR-GFP). Stable ECV-304 transfectants expressing similar surface levels of native fractalkine, FKN-GFP and CX3C-ESCR-GFP were isolated by fluorescence-activated cell sorting using monoclonal antibodies recognizing the CX3C (4F2) domain of fractalkine. As expected, mCX3C-ESCR-GFP was recognized by monoclonal antibody 4F2, but not 1D6, an antibody against the mucin domain of fractalkine (Fig. 5). Expression of a GFP tag on the cytoplasmic tail of fractalkine did not affect its function in cell adhesion (Fig. 6). ECV304 cells expressing mCX3C-ESCR-GFP induced capture and firm adhesion of CX3CR1-expressing K562 cells as efficiently as ECV304-FKN cells (Fig. 6).

DISCUSSION
In this paper we sought to determine the role of the fractalkine mucin domain in the capture and firm adhesion of CX 3 CR1-expressing leukocytes. From the results presented here, we conclude the following. 1) The FKN mucin domain enhances the ability of CX 3 C to function in cell adhesion. The mucin domain exists as an approximately 26 nm long stalk between the globular CX 3 C domain and the cell surface. 5) A primary function of the mucin domain in cell adhesion is to extend the CX 3 C chemokine domain away from the cell surface to present it to flowing leukocytes.
Mucins, including many selectin ligands like PSGL-1, are highly O-glycosylated and exist as long, thin stalk-like structures (20). For most selectin ligands, the O-linked carbohydrates are essential components for binding (21,22). FKN has 26 potential O-linked glycosylation sites, and is heavily glycosylated (8,9). Consistent with other mucins, the FKN mucin domain has a long stalk-like ultrastructure, but we could not detect any binding function of the mucin domain. In this respect, the FKN mucin domain shares similarity with the short consensus repeats of L-, E-, and P-selectin that function to extend the ligand-binding lectin ϩ epidermal growth factor domains while having no intrinsic binding activity (15,19,23).
Cell adhesion is typically mediated by protein-protein interaction rather than by proteins interacting directly with the lipid bilayer, and the interaction typically takes place at an extended distance from the cell surface. For example, integrins have two long stalks that project the ligand-binding domain 20 nm from the membrane. Cadherins project 15 nm, keeping the opposing membranes 30 nm apart. The selectins range from 15 nm (L-) to 25 nm (E-) to 38 nm (P-selectin), and their ligands on the adhering cell are also elongated (1,19,20,24). The 26-nm elongated stalk of fractalkine is thus consistent with this generalization that adhesion receptors project adhesion domains far from the membrane. What is unique about the fractalkine adhesion complex is that the receptor is a 7-transmembrane receptor that projects very little from the membrane. The largest possibility for projection is the N terminus, which could project 8.4 nm if the 24 amino acids were fully extended (0.35 nm per amino acid). It is unusual for molecules that are involved in leukocyte capture to be located so close to the cell surface.
The roles of the FKN pathway of leukocyte migration in normal and diseased states are now being elucidated. We have shown that FKN can mediate the trafficking of monocytes, CD8ϩ T cells and natural killer cells (12). FKN may participate in autoimmune kidney inflammation (25,26). CX 3 CR1 can act as a co-receptor for some strains of human imunodeficiency virus (11,27) and cytomegalovirus (28), indicating a potential role in viral pathogenesis. Fractalkine is expressed highly in neurons of the brain and interacts with CX 3 CR1 expressed in the microglia (29,30), suggesting a means by which these neural cell populations may communicate. Exogenously added fractalkine can cause the migration of primary rat microglial cells in vitro (31). A role for FKN and CX 3 CR1 during nerve regeneration has been postulated (29). Thus, molecules that inhibit FKN-CX 3 CR1 interactions may have therapeutic value in a variety of diseases.
In summary, we have studied the role of the chemokine and mucin domains of fractalkine in FKN-mediated firm adhesion. The chemokine domain is absolutely necessary for interactions with CX 3 CR1, but is suboptimal for firm adhesion under physiological shear stresses. The mucin domain is a 26-nm long stalk that functions to extend the chemokine domain away from the cell surface for optimal interaction of CX 3 C with its receptor. The FKN mucin domain is functionally similar to the stalk regions of the selectin family of molecules that also function in leukocyte capture and firm adhesion under physiologic flow conditions. FKN and CX 3 CR1 may also share other functional similarities with the selectins and their ligands.