Eotaxin Selectively Binds Heparin

An important feature of chemokines is their ability to bind to the glycosaminoglycan (GAG) side chains of proteoglycans, predominately heparin and heparan sulfate. To date, all chemokines tested bind to immobilized heparin in vitro, as well as cell surface heparan sulfate in vitro and in vivo. These interactions play an important role in modulating the action of chemokines by facilitating the formation of stable chemokine gradients within the vascular endothelium and directing leukocyte migration, by protecting chemokines from proteolysis, by inducing chemokine oligomerization, and by facilitating transcytosis. Despite the importance of eotaxin in eosinophil differentiation and recruitment being well established, little is known about the interaction between eotaxin and GAGs and the functional consequences of such an interaction. Here we report that eotaxin binds selectively to immobilized heparin with high affinity (Kd = 1.23 × 10-8 m), but not to heparan sulfate or a range of other GAGs. The interaction of eotaxin with heparin does not promote eotaxin oligomerization but protects eotaxin from proteolysis directly by plasmin and indirectly by cathepsin G and elastase. In vivo, co-administration of eotaxin and heparin is able to significantly enhance eotaxin-mediated eosinophil recruitment in a mouse air-pouch model. Furthermore, when heparin is co-administered with eotaxin at a concentration that does not normally result in eosinophil infiltration, eosinophil recruitment occurs. In contrast, heparin does not enhance eotaxin-mediated eosinophil chemotaxis in vitro, suggesting protease protection or haptotactic gradient formation as the mechanism by which heparin enhances eotaxin action in vivo. These results suggest a role for mast cell-derived heparin in the recruitment of eosinophils, reinforcing Th2 polarization of inflammatory responses.

glycans, predominately heparin and heparan sulfate. To date, all chemokines tested bind to immobilized heparin in vitro, as well as cell surface heparan sulfate in vitro and in vivo (1). In vitro, chemokine-heparin or heparan sulfate binding may be competitively inhibited by the GAGs chondroitin sulfate and dermatan sulfate (1,2), suggesting that some chemokines may also bind these GAGs. Different chemokines bind heparin or heparan sulfate with varying affinity (1)(2)(3), and four different models, or docking modes, have been proposed for the interaction of chemokines with heparin and heparan sulfate (4).
Heparin and heparan sulfate play an important role in modulating the function of chemokines (5) and have been proposed to control chemokine action in three ways. First, it is well established that heparan sulfate immobilizes chemokines on the luminal surface of endothelial cells (6,7), which is essential for the formation of stable chemokine gradients within the vascular endothelium (7)(8)(9)(10)(11) and directing leukocyte migration (12). Second, the interaction of chemokines with heparin and heparan sulfate can protect them from proteolysis (13,14) and induce chemokine oligomerization. Together, these processes can increase the local chemokine concentration and therefore enhance chemokine receptor signaling (3,15). For example, heparin and heparan sulfate protect stromal derived factor-1 (CXCL12) from degradation by the serine protease dipeptidylpeptidase IV (16) and IL-8 from elastase released by neutrophils migrating toward IL-8 (14). Importantly, dipeptidylpeptidase IV is a cell-surface-expressed protease that co-distributes with CXCR4, the stromal derived factor-1 receptor, at the surface of CXCR4-expressing cells. Additionally, it has been suggested that oligomerization of chemokines through interactions with GAGs is essential for maximal chemokine activity (3,17). RANTES (CCL5), MCP-1 (CCL2), IL-8, and MIP-1␣ (CCL3) all bind GAGs, and this appears to increase their binding affinity for their respective receptors (3), whereas mutations in the GAG binding sites of the chemokines RANTES, MIP-1␤ (CCL4), and MCP-1 abolish their ability to induce leukocyte recruitment in vivo (18). Furthermore, a non-GAG binding form of RANTES is able to form non-functional heterodimers with endogenous RANTES, thereby acting as a dominant-negative inhibitor of the formation of heparan-sulfate-induced higher order oligomers. Without the formation of higher order oligomers the chemotactic activity of RANTES is seriously compromised (17). Third, transcytosis, the process by which chemokines secreted within the tissues are actively transported across the endothelial barrier, appears to be dependent on heparan sulfate binding (6).
Eotaxin (CCL11) was initially discovered in an attempt to identify molecules involved in the allergen-induced accumulation of eosinophils in the lung (19). Although other chemokines, such as RANTES, MCP-2, MCP-3, MCP-4 (CCL13), and MIP-1␣ (20,21), can recruit eosinophils into tissues, eotaxin has been identified as the most potent and specific eosinophil chemoattractant (19,(22)(23)(24). In contrast to many other chemokines, eotaxin signals exclusively through the seven transmembrane G-protein receptor, CCR3 (25,26). This, together with its eosinophil selectivity, has made eotaxin an attractive target for the development of drugs that inhibit eosinophil recruitment.
Despite the importance of eotaxin in eosinophil differentiation and recruitment being well established (27), little is known about the interaction between eotaxin and GAGs and the functional consequences of such an interaction. In this study we investigated the ability of various GAGs to bind to eotaxin and their role in protecting eotaxin from proteolysis and enhancing eotaxin-mediated eosinophil chemotaxis. Elucidating this association is crucial in understanding the factors that regulate the recruitment and homing of eosinophils. A reduction in eosinophil trafficking would alleviate many of the pathologies associated with allergic diseases, whereas the directed targeting of eosinophils to tissues could be harnessed to eradicate parasitic infections and tumor growth.
Mice-Male BALB/c wild-type and IL-5 transgenic mice were obtained from the Australian National University (ANU) Animal Services Division and housed in specific pathogen-free facilities at the John Curtin School of Medical Research, ANU. Mice were treated in accordance with ANU animal experimentation guidelines. Mice used in experiments were aged between 6 and 8 weeks.
Surface Plasmon Resonance-A Biacore 2000 instrument (Pharmacia Biosensor, Uppsala, Sweden) was used to measure the binding of eotaxin to immobilized bovine lung heparin and heparin sulfate. PBS/0.0005% Tween 20/10 mM EDTA, pH 6.2 or pH 7.2 as indicated, was used as a running buffer. Biotinylated heparin (1 g/ml) and heparan sulfate (5 g/ml) were immobilized onto flow-cells 2 and 4, respectively, of a CM5 sensor chip (Pharmacia Biosensor) as previously described (29). For the initial direct binding studies, decreasing concentrations of recombinant eotaxin (100 nM to 1.56 nM) in running buffer were injected at a flow-rate of 40 l/min for 5 min, and the association and dissociation phases were monitored. For the competitive binding studies, 50 nM recombinant eotaxin was co-incubated with decreasing concentrations of each GAG (500 nM to 15.6 nM) in running buffer, then injected at a flow-rate of 40 l/min for 7.9 min. The level of eotaxin binding in the absence of GAG was set at 100%, and the percent inhibition values for each GAG were calculated relative to this level. After each eotaxin injection and binding assay, the flow-cell was regenerated by injecting 10 l of 4 M NaCl at 5 l/min. Flowcells 1 and 3 were not coupled with GAGs and were used as blank control reference cells for flow-cells 2 and 4 coupled with heparin and heparan sulfate, respectively. Binding curves were analyzed using the BIAevaluation (version 4.0) program (Pharmacia Biosensor).
Enzyme-linked Immunosorbent Assay-Speciality GAGbinding ExpranEx plates (Plasso, Sheffield, UK) were used to measure the direct binding of eotaxin to various GAGs according to the manufacturer's instructions. Briefly, the wells of a 96-well ExpranEx plate were coated with GAG (25 g/ml) in PBS at room temperature overnight, washed with PBS/0.05% Tween 20 (PBST), then blocked at 37°C for 1 h with 3% (w/v) bovine serum albumin diluted in PBS. Blocking solution was removed, recombinant mouse eotaxin (3-48 nM) was added to the wells, and the mixture was incubated for 2 h at room temperature then washed with PBST. Bound eotaxin was detected using goat anti-mouse eotaxin antibody (AF-20, R&D Systems, Minneapolis, MN), followed by a polyclonal rabbit anti-goat antibody conjugated to horseradish peroxidase (Chemicon). Plate-bound peroxidase was detected using 2,2Ј-azinobis(3ethylbenzothiazoline-6-sulfonic acid) peroxidase substrate (Kirkegaard and Perry Laboratories Inc., Gaithersburg, MD) and measuring the absorbance at 405 nm on a Thermomax microplate reader. Data were analyzed using SoftMax Pro software (Molecular Devices, Sunnyvale, CA).
Heparanase Treatment of Heparin-Bovine lung or porcine mucosal heparin (1 g, Sigma) was incubated for 24 h with 40 g of purified human platelet heparanase in 200 l of 50 mM sodium acetate buffer, pH 5.1, containing 50 g/ml bovine serum albumin (30). The cleaved products were filtered through a Centricon 10 microconcentrator (Millipore, Bedford, MA) and the filtrate lyophilized and then desalted using a PD-10 column (Amersham Biosciences) developed in water. Heparin in the column eluant was detected using dimethyl methylene blue (31) and the solution lyophilized, dissolved in water, and stored at Ϫ20°C as a 1.2 mg/ml solution. Cleavage of heparin was confirmed by SDS-PAGE (12% acrylamide), with heparin bands being detected by toluidine blue staining.
Plasmin Treatment of Eotaxin-Plasmin (20 M) was produced by mixing plasminogen (20 M) with urokinase plasminogen activator (20 nM) in 20 mM HEPES/0.14 M NaCl/2 mM EDTA/0.02% sodium azide, pH 7.4, and incubating at 37°C for 1 h. Plasmin activity was determined by a chromogenic assay using the plasmin substrate H-D-Val-Leu-Lys-p-nitroanilide (S2251, Chromogenix AB, Mölndal, Sweden). Eotaxin (40 ng/sample) was mixed with either a 2-fold (120 ng/sample) or 10-fold (600 ng/sample) molar excess of heparin in a final volume of 15 l of 50 mM Tris, pH 7.5, and incubated at 37°C for 20 min to ensure binding. Plasmin (100 ng/sample) was then added in 5 l of 50 mM Tris, pH 7.5, and the samples were reincubated at 37°C for 4 h. Protease degradation of eotaxin was visualized by Western blotting. Briefly, samples were mixed with an equal volume of reducing sample buffer, and 30 l was loaded onto a pre-cast 4 -20% polyacrylamide gradient iGel (Gradipore, Frenchs Forest, Australia), then transferred to an Immobilon P20 membrane (Millipore). Eotaxin was detected by probing membranes with a rabbit anti-mouse eotaxin antibody (AF-20, R&D Systems), followed by a polyclonal rabbit anti-goat antibody conjugated to horseradish peroxidase (Chemicon) and developed using ELC chemiluminescence (GE Healthcare, Little Chalfont, UK). Rainbow low molecular weight markers (GE Healthcare) were used to determine the molecular weight of different bands.
Cross-linking Studies-Eotaxin (500 nM) in the presence and absence of heparin (5 M) was mixed with BS 3 (24 M) or NHS (1 mM)/EDC (0.5 mM) in a final volume of 10 l of PBS, and incubated at 4°C for 2 h or 30 min, respectively. The reaction was stopped by the addition of 2ϫ SDS reducing buffer, and cross-linking was visualized by SDS-PAGE and Western blotting as described above.
In Vitro Chemotaxis Studies-Eosinophils were harvested from the peritoneal cavity of BALB/c IL-5 transgenic mice by washing the cavity three times with Hanks' balanced salt solution, and purified by sorting on a BD FACS Vantage (BD, Franklin Lakes, NJ) based on forward and side scatter parameters and polarized light. The purity of the eosinophil enriched population was Ͼ95% as determined by differential staining with May-Grunwald Giemsa. The 3-m polycarbonate membrane of the upper wells of a 96-well multiscreen MIC plate (Millipore) was coated with fibronectin (50 g/ml, 37°C, 30 min, Sigma). Following incubation, the plate was washed three times with RPMI 1640 medium (Invitrogen) to remove unbound fibronectin. Purified eosinophils (1 ϫ 10 5 cells/well) in 50 l of RPMI/4% fetal calf serum were added to each of the upper wells of the multiscreen plate. 125 l of agonist in RPMI/4% fetal calf serum was added to each receiver well, the plate was assembled, and the mixture was incubated for 4 h at 37°C in a humidified atmosphere of 5% CO 2 . At the end of the assay, the upper wells were removed from the receiver plate, and the solution was aspirated to remove non-migrated cells. A known quantity of Calibrite calibration beads (BD Biosciences) was added to each bottom well before harvesting to allow determination of the cell number. 125 l of 5 mM EDTA in PBS was added to the receiver wells, the plate was reassembled, and the mixture was incubated for 30 min at 37°C with 5% CO 2 . The plate was then shaken gently to dislodge any migrated cells from the underside of the polycarbonate membrane and any cells adhered to the FIGURE 1. Binding affinity of eotaxin for bovine lung heparin. A, the binding of eotaxin to heparin and heparan sulfate was analyzed by SPR. Biotinylated heparin and heparan sulfate were immobilized onto a CM5 sensor chip. Decreasing concentrations of recombinant eotaxin (100 nM to 12.5 nM) were then injected. Binding to heparin and heparan sulfate is shown in response unit (RU) values for each concentration of eotaxin. B, the kinetics of eotaxin binding to, and dissociation from, immobilized biotinylated heparin was measured using SPR. Decreasing concentrations of recombinant eotaxin (100 to 1.56 nM) were injected in duplicate, and the association and dissociation phases were measured. The curves were used to calculate the kinetics of the reaction using BIAevaluation software. Curves shown are representative of duplicate samples. In both studies PBS/10 mM EDTA/0.05% Tween 20, pH 6.2, was used as the running buffer. receiver well. The detached cells were combined with the solution initially removed from the receiver wells and resuspended in 2% paraformaldehyde. The number of migrated cells in each well was quantified by flow cytometry as a function of the ratio of Calibrite beads to live cells. Cell migration was calculated as the percentage of input cells.
Air-pouch Studies-Air-pouch studies to determine eotaxininduced migration of eosinophils in vivo, with and without heparin, were performed in wild-type BALB/c mice as previously described (32,33). Briefly, mice were anesthetized with isoflurane, the dorsal skin was shaved, and 150 l of air, followed by 100 l of eotaxin (1-5 pmol/site), heparin (5-10 pmol/site), a combination of both, or control vehicle (Hanks' balanced salt solution/0.01% bovine serum albumin) in the same syringe, was injected subcutaneously. After 6 h, the mice were sacrificed, the skin around the air-pouch excised, and the dorsal skin membrane was transferred to a slide. Slides were fixed in methanol before staining with chromotrope to determine the level of eosinophilia. Eosinophils/field were determined by counting 10 fields of view (400ϫ magnification).
Statistical Significance-Statistical significance was measure using the Mann-Whitney test performed using InStat (GraphPad Software, San Diego, CA).

Characterization of the Interaction between Eotaxin and Various
GAGs-Many chemokines have been shown to bind to GAGs, in particular heparin and heparan sulfate, but with varying affinity (1)(2)(3).
Recently it has been suggested that the specificity with which individual chemokines bind various GAGs may fine-tune chemokine function and provide another level of regulation of chemokine action (34). We used SPR to measure the interaction of the eosinophil specific chemokine, eotaxin, with heparin and heparan sulfate. At both pH 6.2 and 7.2 soluble recombinant murine eotaxin bound to immobilized bovine lung heparin in a concentrationdependent manner, but did not bind at any concentration (12.5-100 nM) to immobilized heparan sulfate (pH 6.2 data are in Fig. 1A; pH 7.2 data not shown). Kinetic analysis of the association and dissociation phases of the sensorgrams using the BIAevaluation software revealed that eotaxin binds to heparin with a high affinity, exhibiting an equilibrium binding constant (K A ) of 8.14 ϫ 10 7 M and an equilibrium dissociation constant (K D ) of 1.23 ϫ 10 Ϫ8 M at pH 6.2 (Fig. 1B). Following correction for mass transfer effects, binding of eotaxin to heparin fitted a Langmuir (1:1) binding model, indicating that one molecule of eotaxin binds to one molecule of heparin. Similar kinetics were observed for the binding of eotaxin to immobilized heparin at pH 7.2, exhibiting a K A of 3.66 ϫ 10 7 M and a K D of 2.73 ϫ 10 Ϫ8 M (data not shown).
Most chemokines bind to both heparin and heparan sulfate (1). The observation that eotaxin binds exclusively to heparin and not to heparan sulfate, led us to investigate the interaction of eotaxin with other GAGs by performing a series of competitive inhibition experiments ( Fig. 2A). As expected, soluble heparin completely inhibited eotaxin (50 nM) binding to immobilized heparin with an IC 50 of 25.5 nM (Table 1). In contrast, none of the other GAGs tested completely inhibited eotaxin binding, even at the highest concentration (500 nM) tested. Nevertheless, weak inhibitory activity was observed with chondroitin sulfate E Ͼ heparan sulfate Ͼ chondroitin sulfate B, The level of eotaxin binding in the absence of GAG was set at 100%, and the percent inhibition values for each GAG were calculated relative to this level. PBS/10 mM EDTA/0.05% Tween 20, pH 6.2, was used as the running buffer. B, the direct binding of eotaxin to different immobilized GAGs as measured by enzyme-linked immunosorbent assay, with increasing concentrations of eotaxin (3 to 50 nM) being added to GAG-coated wells of a 96-well ExpranEx plate.
whereas chondroitin sulfates A and C were inactive. Additionally, the direct binding of eotaxin to GAGs was measured by enzyme-linked immunosorbent assay using EpranEx GAGbinding plates (Fig. 2B). Consistent with the SPR analysis, eotaxin exhibited the strongest binding to immobilized heparin, followed by chondroitin sulfate E and chondroitin sulfate B, but did not bind to immobilized heparan sulfate or chondroitin sulfates A and C. Coating of wells with biotinylated heparin and heparan sulfate demonstrated equal binding of the GAGs to the wells, suggesting that this was not the reason for lack of eotaxin binding to the heparan sulfate-treated wells (data not shown).
Collectively, these data suggest an unusually high preference of eotaxin for binding to heparin compared with other GAG molecules.
Previous studies have shown that the chemokines MIP-1␣ and IL-8 have a higher affinity for heparin molecules with increasing chain length (2). Fig. 3A and Table 2 show that this is also the case for murine eotaxin, the ability of soluble heparin to inhibit the binding of eotaxin to immobilized heparin being proportional to heparin chain length, with a large decrease in inhibitory activity being seen with heparin preparation Յ6.7 kDa. Consistent with this finding, heparanase cleavage of 12.5-kDa bovine lung (BL) and 12.5 kDa porcine mucosal (PM) heparin resulted in a 3-fold and 40-fold reduction, respectively, in inhibitory activity based on IC 50 values (Fig. 3B and Table 2), with heparanase cleaving BL and PM heparin to fragments of ϳ6 kDa and ϳ4.5 kDa, respectively. Thus these data are consistent with the binding affinity of eotaxin for heparin being highly dependent upon heparin chain length (Fig. 3).
Heparin Protects Eotaxin from Protease Degradation-The interaction between chemokines and GAGs has been shown to modulate several physiological functions required for chemokine action in vivo. These include chemokine oligomerization (3,17,18) and protection of chemokines from proteolysis (13,14). We therefore examined whether the interaction of eotaxin with heparin modulated eotaxin function.
A number of proteases such as plasmin, elastase, and cathepsin G are present in inflammatory sites. Thus, heparin was examined for its ability to protect eotaxin from proteolysis by these proteases. Initial studies using SDS-PAGE and Western blotting revealed that eotaxin is highly sensitive to elastase, cathepsin G (data not shown), and plasmin digestion (Fig. 4, lane 2). A 10-fold molar excess of BL heparin was able to completely protect eotaxin from plasmin degradation (Fig. 4,  lane 3), with there being no reduction in eotaxin band density compared with the non-plasmin-treated eotaxin control (lane 1). Although binding affinity for eotaxin is similar, BL heparin offered slightly more protection from plasmin degradation than PM heparin, possibly due to slightly higher levels of sulfation in BL heparin. At a 2-fold molar excess, both BL and PM heparin were able to partially protect eotaxin from plasmin degradation. Heparanase treatment of BL and PM heparin resulted in reduced protease protection of eotaxin in accordance with the reduced bind- FIGURE 3. Binding affinity of heparin for eotaxin is dependent on heparin size. A, different molecular weight heparins were analyzed by SPR for their ability to inhibit the heparin-eotaxin interaction. B, BL and PM heparin, both before and after heparanase treatment, were analyzed by SPR for their ability to inhibit the interaction of eotaxin with immobilized heparin. The level of eotaxin binding in the absence of GAG was set at 100%, and the percent inhibition values for each heparin concentration were calculated relative to this level. PBS/10 mM EDTA/0.05% Tween 20, pH 6.2, was used as the running buffer. ing affinity of heparanase-treated heparin for eotaxin shown in Fig. 3B, i.e. heparanase-treated PM heparin had a lower protective ability than heparanase-treated BL heparin. The ability of heparanase digestion, to reduce the ability of heparin to bind and protect eotaxin from proteolysis, suggests a possible role for heparanase as a physiological regulator of eotaxin function in vivo. Interestingly, heparin was also able to protect eotaxin from degradation by elastase and cathepsin G (data not shown). However, we have previously reported that heparin directly inhibits elastase and cathepsin G action (29). Indeed, further analysis showed that heparin inhibition of cathepsin G and elastase activity, rather than direct binding to eotaxin, was the mechanism by which heparin protects eotaxin from cathepsin G and elastase proteolytic degradation. It should be noted that plasmin activity was not inhibited by heparin (data not shown).

Heparin Does Not Promote Eotaxin
Oligomerization-It has recently been suggested that oligomerization of several chemokines is facilitated through their interaction with GAGs, and this is essential for maximal chemokine activity (3,17). Previous studies have shown that at physiological pH human eotaxin exists in equilibrium between a monomer and dimer (35). To examine if heparin could promote eotaxin oligomerization we employed the cross-linking agents BS 3 , an amine-reactive cross-linker, or NHS and EDC, which is used to cross-link carboxyl groups to primary amines. Fig. 5 demonstrates that native eotaxin does form oligomers, which can be cross-linked by both BS 3 (lane 4) and EDC (lane 6). However, the presence of heparin did not enhance oligomerization (compare lanes 4 with 5, and 6 with 7), and, in fact, tended to reduce oligomerization.

Effect of the Heparin-Eotaxin Interaction on Eosinophil Chemotaxis in Vitro and in Vivo-An in vitro
Transwell chemokine assay was used to assess the effect of the eotaxin-heparin interaction on eosinophil migration in a soluble chemokine environment. In the absence of any added agonist, 4.1% of eosinophils migrated through the fibronectin-coated membranes, presumably induced by fetal calf serum-associated chemokines and chemokinetic migration (Fig. 6). Exogenous eotaxin at 25 nM induced an increase of ϳ2-fold in the number of migrating eosinophils (i.e. 9.0%), whereas addition of 62.5 nM eotaxin increased cell migration ϳ8-fold to 31%. However, addition of either a 2-fold or 10-fold molar excess of heparin relative to eotaxin did not significantly increase eosinophil migration above that observed with eotaxin alone. Heparin alone also did not induce eosinophil migration above background levels (Fig. 6).
Thus, heparin had no effect on the migration of eosinophils toward soluble eotaxin in vitro using a standard Transwell system. The interaction between GAGs and chemokines has been suggested to facilitate the formation of fixed (haptotactic) chemokine gradient within the extracellular matrix in vivo. It is possible that the interaction between heparin and eotaxin mod-   5 and 7), was cross-linked with BS 3 (lanes 4 and 5) or EDC (lanes 6 and 7) to examine if heparin aided the formation of eotaxin dimers or oligomers. Monomeric (lane 1) and dimeric (lane 2) eotaxin are shown as controls. Eotaxin multimerization was visualized by SDS-PAGE and Western blotting. Rainbow low molecular weight markers were used to determine the molecular weight of different bands. ulates this phenomenon. Thus, to establish the physiological significance of the eotaxin-heparin interaction, it was important to determine the effect of heparin on eotaxin-mediated eosinophil chemotaxis in vivo. Eosinophil chemotaxis in vivo was assessed using an air-pouch assay. Because we were interested in the effect of heparin on haptotactic migration, we examined the level of eosinophil recruitment into the air-pouch membrane rather than the air-space within the pouch, as has been reported previously (32,33). Using this model, the optimal concentration of eotaxin inducing eosinophil recruitment has previously been determined to be 5 pmol (33). At this concentration, co-administration of eotaxin and a 2-fold molar excess of heparin (10 pmol) resulted in a significant increase in the number of eosinophils recruited into the tissue, compared with eotaxin alone (Fig. 7A). Furthermore, when a concentration of eotaxin (1 pmol) was administered that induced negligible eosinophil recruitment, the addition of a 2-fold molar excess of heparin (2 pmol) resulted in substantial eosinophil recruitment (Fig. 7B). It should be noted that the administration of 10 pmol of heparin produced a small but not significant eosinophil influx, an effect not observed when 2 pmol of heparin was injected. This significant increase in eosinophil migration induced by eotaxin in the presence of heparin is clearly evident upon histological examination, with large numbers of eosinophils recruited into the air-pouch membrane in the group co-administrated with eotaxin and heparin (Fig. 7C, panel iv).

DISCUSSION
Almost all chemokines interact with heparin and heparan sulfate, and this interaction has been shown to modulate a variety of functions associated with the induction of cell migration in vivo. In this report, we have shown that eotaxin interacts selectively with heparin, with a high affinity (K D ϭ 2.73 ϫ 10 Ϫ8 M, pH 7.2). This indicates that eotaxin has a comparable affinity for heparin as RANTES (K D ϭ 3.21 ϫ 10 Ϫ8 M, pH 7.4 (36)), previously the chemokine reported to have the highest binding affinity for heparin (34).
Unlike several chemokines studied previously, such as MCP-1, IL-8, and MIP-1␣ (2), eotaxin does not bind to heparan sulfate or any other GAG except chondroitin sulfate E. Interestingly, the only other chemokine reported to exhibit such as high level of discrimination in binding to different GAGS is RANTES (2), another eosinophil chemokine that signals though CCR3, although not exclusively as is the case for eotaxin. The GAG binding profiles of the other CCR3 ligands, MCP-3 and MCP-4, have not been examined.
The high level of discrimination of eotaxin binding between different GAGs suggests a unique role for heparin in modulating the functions of eotaxin-mediated cellular chemotaxis. Indeed, the in vivo studies reported here demonstrate that heparin can enhance eotaxin action in vivo, especially at sub-optimal levels of eotaxin, inducing an eosinophil influx not seen with eotaxin or heparin alone. Although heparin has been reported to inhibit eotaxin binding to its chemokine receptor, CCR3, and inhibit eosinophil chemokinesis and chemotaxis of a eosinophil cell line in vitro (37), we did not observe a reduction in eotaxin-mediated eosinophil chemotaxis in vitro when  eotaxin was co-administered with heparin. However, the study by Culley et al. that reported heparin-mediated inhibition of eosinophil chemokinesis in vitro used very high concentration of heparin, representing almost a 1000-fold molar excess compared with the heparin employed in our study. At such high concentrations heparin may interfere with eotaxin binding to CCR3. This inhibition, however, does not occur at much lower concentrations of heparin, with these low heparin concentrations actually enhancing eotaxin-mediated eosinophil recruitment in vivo.
The enhancement of eotaxin action in vivo observed in our study is most likely a result of heparin protecting eotaxin from proteolysis. Direct binding to heparin protected eotaxin from degradation by the protease plasmin. Interestingly, plasmin has previously been shown to play a role in eotaxin-mediated eosinophil chemotaxis (38). In vitro, plasminogen was required for eosinophils to migrate through Matrigel, an artificial extracellular matrix, toward eotaxin, presumably because of a role for plasmin in degrading the basement membrane. In related studies, eotaxin autocrine signaling within eosinophils has also been suggested to promote the release of proteases required for eosinophil migration through basement membranes and the extracellular matrix (39). We have shown that heparin was also able to indirectly protect eotaxin from elastase-and cathepsin G-mediated proteolysis by inhibiting protease action, as previously reported for IL-8 (14). This would suggest that the high affinity interaction of eotaxin with heparin confers protection from various proteases and allows eotaxin to persist in tissues as an active chemokine.
Heparin is synthesized by mast cells and released, bound to tryptase, during mast cell degranulation (40). Mast cells and eosinophils are both components of the innate arm of the CD4 ϩ Th2-mediated immune responses, and co-localize in many Th2-mediated inflammatory reactions that occur during allergy and parasitic infections. A number of mast cell-derived mediators has been shown to prime or activate eosinophil responses, whereas eosinophil-derived mediators can also directly affect mast cells responses (41)(42)(43)(44)(45)(46)(47)(48)(49)(50)(51). In fact, mast celldeficient mice develop less eosinophilia than wild-type mice in models of allergic airways disease (AAD) (52). Interestingly, the only other GAG found to bind eotaxin, chondroitin sulfate E, is also released from mast cells upon degranulation (53).
In a previous study, mast cells were shown to be required for eosinophil infiltration in response to eotaxin injection into air pouches of sensitized mice (54). This could be explained by the eotaxin-heparin interaction and the protection it offers from proteolysis. It was recently reported that the mast cell protease, ␤-tryptase, can degrade both eotaxin and RANTES, the primary eosinophil chemokines, which both signal through CCR3 (55). In this case the addition of heparin was not able to protect these chemokines, and, in fact, increased proteolysis. Heparin is required for the formation of stable ␤-tryptase tetrameters, essential for protease activity (56). Thus, it is possible that ␤-tryptase acted as a competitive inhibitor of chemokine-heparin binding, facilitating chemokine proteolysis. However, the observation that the murine equivalent of ␤-tryptase, mouse mast cell protease-6, only binds heparin and is active at pH 6.0 not pH 7.4 (57), suggests that the cleavage of eotaxin and RAN-TES by tryptase may only occur at lower pH in vivo. This response would be important in the control of chronic inflammation, when it would be beneficial for the neutralization of chemokines by proteolysis to reduce infiltrating leukocytes. Thus we would suggest that during chronic inflammation, mast cell tryptase may negatively regulate eosinophilia by destroying eotaxin and RANTES chemokine gradients. Similarly, the observation reported here that heparanase cleavage of heparin greatly reduces heparin affinity for eotaxin, suggests that heparanase may also play a role in the negative regulation of eotaxin-mediated eosinophilia.
In addition to protease protection, the interaction of chemokines with heparin and heparan sulfate has been shown to promote chemokine oligomerization, which in the case of RAN-TES, MCP-1 and MIP-1␣ appears to be essential for function in vivo (17,18). Our results demonstrate that heparin did not increase murine eotaxin oligomerization. This is consistent with previous studies of human eotaxin employing mass spectrometry (58). Recently, it has been suggested that heparin may promote heterodimerization of eotaxin with MCP-1 (CCL2) (59), although the functional significance of this observation, remains to be elucidated. MCP-1 signals through CCR2 and results in the influx of neutrophils and the induction of Th1-like inflammatory responses. In contrast, eotaxin only signals through CCR3, but can act as a CCR2 antagonist (60). Thus the formation of MCP-1/eotaxin dimers may block the action of MCP-1 thereby skewing inflammatory responses toward a type 2 phenotype.
These results will have important implications for understanding and directing eosinophil recruitment in various disease settings. Eosinophils have been shown to play an important role in the pathophysiology of asthma and allergic airways disease. Controlling the influx of eosinophils into the asthmatic lung has therapeutic potential. Previous studies have examined the role of heparin in controlling allergen-induced airway eosinophilia in guinea pigs (61)(62)(63). Unfractionated heparin was shown to reduce rather than enhance eosinophil infiltration into the lung and bronchoalveolar lavage fluid. These previous studies did not elucidate the mechanism responsible for the reduction in eosinophil inflammation induced by heparin but report that low molecular weight heparins and chemically modified heparins without anticoagulant activity exhibited more potent anti-inflammatory effects (62,64). This suggests that the mechanism is not related to chemokine modulation, because many studies, including our current research, have shown that chemokines have a much lower affinity for low molecular weight heparins. Thus, if soluble heparin is "mopping" up chemokines and inhibiting binding to heparan sulfate and/or eosinophil chemokine receptors to reduce allergen-induced inflammation, low molecular weight heparin should be less, rather than more potent at inhibiting inflammation.
From the findings reported here, we would suggest that heparin helps to promote, not reduce, eosinophil recruitment, by enhancing eotaxin action through the inhibition of proteolysis and possibly the formation of a haptotactic chemokine gradient within the extracellular matrix. The potential exists for further investigation into heparin as an anticancer therapeutic to increase eosinophil homing into tumors (65,66) or, similarly, in the treatment of parasitic infections. Furthermore, disruption of the heparin-eotaxin interaction, possibly through employment of heparanase or other competitive binding inhibitors, may provide therapeutic opportunities for pathologies associated with eosinophilia such as asthma and allergic airways disease.