The Positively Charged COOH-terminal Glycosaminoglycan-binding CXCL9(74–103) Peptide Inhibits CXCL8-induced Neutrophil Extravasation and Monosodium Urate Crystal-induced Gout in Mice*

Background: Chemokines, such as CXCL8 and CXCL9, drive leukocyte migration to an inflammation site. Results: CXCL9(74–103), derived from CXCL9, lacks leukocyte-attracting activity but competes with CXCL8 for GAG binding and inhibits neutrophil migration in two murine acute inflammation models. Conclusion: Through inhibition of chemokine-GAG interaction, CXCL9(74–103) blocks neutrophil migration. Significance: CXCL9(74–103) may be a lead molecule for development of anti-inflammatory agents. The ELR−CXC chemokine CXCL9 is characterized by a long, highly positively charged COOH-terminal region, absent in most other chemokines. Several natural leukocyte- and fibroblast-derived COOH-terminally truncated CXCL9 forms missing up to 30 amino acids were identified. To investigate the role of the COOH-terminal region of CXCL9, several COOH-terminal peptides were chemically synthesized. These peptides display high affinity for glycosaminoglycans (GAGs) and compete with functional intact chemokines for GAG binding, the longest peptide (CXCL9(74–103)) being the most potent. The COOH-terminal peptide CXCL9(74–103) does not signal through or act as an antagonist for CXCR3, the G protein-coupled CXCL9 receptor, and does not influence neutrophil chemotactic activity of CXCL8 in vitro. Based on the GAG binding data, an anti-inflammatory role for CXCL9(74–103) was further evidenced in vivo. Simultaneous intravenous injection of CXCL9(74–103) with CXCL8 injection in the joint diminished CXCL8-induced neutrophil extravasation. Analogously, monosodium urate crystal-induced neutrophil migration to the tibiofemural articulation, a murine model of gout, is highly reduced by intravenous injection of CXCL9(74–103). These data show that chemokine-derived peptides with high affinity for GAGs may be used as anti-inflammatory peptides; by competing with active chemokines for binding and immobilization on GAGs, these peptides may lower chemokine presentation on the endothelium and disrupt the generation of a chemokine gradient, thereby preventing a chemokine from properly performing its chemotactic function. The CXCL9 peptide may serve as a lead molecule for further development of inhibitors of inflammation based on interference with chemokine-GAG interactions.

The ELR ؊ CXC chemokine CXCL9 is characterized by a long, highly positively charged COOH-terminal region, absent in most other chemokines. Several natural leukocyte-and fibroblast-derived COOH-terminally truncated CXCL9 forms missing up to 30 amino acids were identified. To investigate the role of the COOH-terminal region of CXCL9, several COOH-terminal peptides were chemically synthesized. These peptides display high affinity for glycosaminoglycans (GAGs) and compete with functional intact chemokines for GAG binding, the longest peptide (CXCL9(74 -103)) being the most potent. The COOHterminal peptide CXCL9(74 -103) does not signal through or act as an antagonist for CXCR3, the G protein-coupled CXCL9 receptor, and does not influence neutrophil chemotactic activity of CXCL8 in vitro. Based on the GAG binding data, an antiinflammatory role for CXCL9(74 -103) was further evidenced in vivo. Simultaneous intravenous injection of CXCL9(74 -103) with CXCL8 injection in the joint diminished CXCL8-induced neutrophil extravasation. Analogously, monosodium urate crystal-induced neutrophil migration to the tibiofemural articulation, a murine model of gout, is highly reduced by intrave-nous injection of CXCL9(74 -103). These data show that chemokine-derived peptides with high affinity for GAGs may be used as anti-inflammatory peptides; by competing with active chemokines for binding and immobilization on GAGs, these peptides may lower chemokine presentation on the endothelium and disrupt the generation of a chemokine gradient, thereby preventing a chemokine from properly performing its chemotactic function. The CXCL9 peptide may serve as a lead molecule for further development of inhibitors of inflammation based on interference with chemokine-GAG interactions.
Chemokines constitute a group of small structurally related chemotactic proteins indispensable for the coordination of leukocyte migration during inflammation (1). In addition, they are essential factors in the control of leukocyte mobilization from primary or secondary lymphoid organs, in the generation of lymphoid follicles, and in the homing of leukocytes to specific organs and tissues (1). Upon tissue injury, sterile inflammation, or infection, chemokines are locally produced by tissue cells and resident leukocytes. Through their G protein-coupled receptors, which are expressed by specific leukocytes, chemokines induce the endothelial adhesion of leukocyte subtypes followed by their extravasation and directional migration toward the site of infection (2,3). In addition to receptor binding, a major role in chemokine activity has been allocated to interactions with glycosaminoglycans (GAGs), 4 such as heparin, heparan sulfate (HS), and dermatan sulfate (DS) (4,5). This interaction occurs between sulfated domains of GAGs and basic amino acid motifs of chemokines. The GAG binding motif on chemokines frequently takes the form XBBXBX or XBBBXXBX, in which B represents a basic and X represents any non-basic amino acid (6). In general, these GAG binding motifs are located at a site distant from the specific receptor-binding domain, often within the COOH-terminal region of the molecule. Although chemokine-GAG binding largely depends on electrostatic interactions, a certain degree of specificity has been ascribed to this interaction, suggested to be due to van der Waals and hydrogen bond forces (5,7).
Chemokine binding to GAGs has been proven to be indispensable for chemokine activity in vivo (8 -16). Abrogated GAG binding (tested using chemokine variants with mutations in their GAG binding motif) results in abrogated activity in vivo, although receptor binding and chemotactic activity in vitro are seldom affected (9 -11). In addition, reduced chemokine-induced leukocyte migration is seen in mice with disturbed HS synthesis (8,13,14).
Based on in vitro data, this role for GAG binding in chemokine functioning in vivo can be explained by its importance in chemokine gradient formation, cis-or trans-(on endothelial cells) presentation of chemokines, chemokine transcytosis across venular endothelial cells, and stabilization of oligomer formation and protection against enzymatic degradation (1,8,9,11,(17)(18)(19)(20). More recently, in vivo, chemokines were shown to localize within capillary venules in a GAG-dependent way (13,18). Moreover, in zebrafish, fluorescently labeled CXCL8 was found to form an extracellular gradient in tissue that decays within a distance of 50 -100 m from the producing cells and was immobilized on the local venous vasculature (12). Both gradient formation and immobilization were dependent on extracellular HS proteoglycans. Analogously, endogenous HS-dependent gradients of CCL21 were detected within mouse skin (21). In all of the above cases, destruction of these gradients abolished directional leukocyte migration. Indeed, these data support the hypothesis that chemokine production at sites of inflammation results in the generation of GAG-mediated chemokine gradients in the extracellular matrix and presentation by the GAGs on the endothelial cell wall, thereby preventing their diffusion and degradation and retaining high local concentration of the produced chemokines (22). These chemokine gradients continuously drive the adhesion, extravasation, and tissue infiltration of specific leukocyte subsets.
Chemokine activity goes far beyond the mere induction of leukocyte migration. Chemokines have also been shown to be key players in embryonic development and angiogenesis, the formation of new blood vessels from pre-existing ones, both in embryonic (physiological) and adult (pathological) conditions (23,24). ELR ϩ CXC chemokines, such as CXCL6 and CXCL8, clearly show angiogenic properties, whereas ELR Ϫ CXC chemokines, including CXCL4/PF4 (platelet factor 4) and CXCL9, are angiostatic molecules. Based on their impact on leukocyte infiltration and angiogenesis, a role in tumor biology is conceivable (25,26). Fine-tuning chemokine activity is critical to control inflammation in time, space, and intensity (27). Besides other regulatory mechanisms, such as tightly regulated gene expression and the presence of atypical receptors, posttranslational modification undeniably controls chemokine activity and thereby leukocyte accumulation and inflammation (28 -32). Specific and limited processing of chemokines (truncation, glycosylation, or citrullination), executed by enzymes that are co-expressed under inflammatory conditions, can alter their biological activity due to increased or decreased receptor binding/signaling, changed receptor specificity, or altered GAG binding affinity (28). By up-or down-regulating chemokine activity, these enzymes determine the shape and speed of the immune response, triggering rapid leukocyte mobilization upon infection, but also efficiently dampening inflammation upon resolution of infection.
The secreted ELR Ϫ CXC chemokine CXCL9/Mig (monokine induced by interferon-␥) consists of 103 amino acids and attracts activated T lymphocytes, preferentially of the Th1 phenotype, and NK cells through activation of the CXC chemokine receptor 3 (CXCR3) (33). Like the other CXCR3 ligands CXCL10 and CXCL11, the production of CXCL9 is induced by the Th1 cytokine interferon (IFN)-␥. In contrast to CXCL10 and CXCL11 and to almost all other chemokines, the COOHterminal region of CXCL9 is remarkably long and basic with 50% positively charged amino acids. Although these CXCR3 ligands show a high degree of redundancy in vitro, this apparent redundancy seems to be absent in vivo (34). This is probably due to differential temporal and spatial expression patterns and different binding characteristics toward their shared receptors.
This study aims to obtain more insight into the contribution of this unique COOH-terminal region of CXCL9. We have explored the possible role of the COOH-terminal region by chemically synthesizing COOH-terminal CXCL9 peptides. We show that a COOH-terminal CXCL9 peptide, cleaved from natural CXCL9, competes with CXCL8 for GAG binding and inhibits CXCL8-induced neutrophil extravasation in vivo. Moreover, this peptide was able to neutralize neutrophil influx into the joint in monosodium urate (MSU) crystal-induced gout, a model for human gout in mice.
Purification and Identification of CXCL9 Isoforms from PBMC-and Fibroblast-conditioned Medium-Natural human CXCL9 was purified to homogeneity in a four-step purification procedure (38). Briefly, CXCL9 was detected by an enzymelinked immunosorbent assay (ELISA) and purified by adsorp-tion to controlled pore glass and heparin-Sepharose affinity chromatography. The eluted material was desalted through dialysis with a 6.5 kDa cut-off membrane before Mono S cation exchange (GE Healthcare) chromatography. Final separation was performed by reversed-phase high performance liquid chromatography (RP-HPLC) on a 2.1 ϫ 220-mm Brownlee C8 Aquapore RP-300 column (PerkinElmer Life Sciences). Of the RP-HPLC effluent, 0.7% was split on-line to an ion trap mass spectrometer (Bruker Daltonics, Bremen, Germany). Profile spectra were collected, averaged within the UV absorption ( ϭ 214 nm) peaks containing CXCL9 immunoreactivity, and deconvoluted to determine the average relative molecular mass (M r ). This M r was compared with the theoretical average M r of intact or truncated CXCL9. In addition, the NH 2 -terminal sequence of the isolated CXCL9 forms was determined by Edman degradation (Procise 491 cLC sequencer, Applied Biosystems, Foster City, CA).
Calcium Mobilization by CXCL9 Peptides through CXCR3-The capacity of the CXCL9 peptides to induce signaling through human CXCR3, resulting in an increase of the intracellular Ca 2ϩ concentration or desensitization toward CXCL9 stimulation, was tested on CXCR3-transfected CHO cells, as described previously (41). In the case of preincubation, CXCL9 was incubated with CXCL9(74 -103) for 10 min prior to the addition to cells.
Binding of CXCL9(74 -103) to Cellular GAGs and Competition with CXCL8 -Binding of biotinylated CXCL9(74 -103) to CHO/CXCR3A, CHO/CXCR3B, and CHO/CXCR4 cells was assessed by flow cytometric analysis. To inhibit the sulfation of cellular GAGs, CHO cells were cultured in the presence of 100 mM sodium chlorate (NaClO 3 ). After 24 h, cells were detached with phosphate-buffered saline (PBS) enriched with 0.02% (w/v) EDTA for 5-10 min at 37°C, washed with culture medium, and resuspended after 1 h in ice-cold assay buffer (PBS ϩ 2% (v/v) FCS). Subsequently, cells (3 ϫ 10 5 ) were labeled with biotinylated CXCL9(74 -103) at different dilutions for 30 min on ice. After washing, cells were incubated with streptavidin-allophycocyanin (BD Biosciences) for 30 min on ice in the dark. Finally, cells were washed three times with icecold assay buffer and analyzed using a FACSCalibur flow cytometer (BD Biosciences). In parallel, to confirm inhibition of sulfation of cellular GAGs, part of the cells were stained with a mouse monoclonal anti-human HS antibody (Immunosource, Schilde, Belgium), washed with assay buffer, incubated with a secondary phycoerythrin-labeled goat anti-mouse antibody (BD Biosciences), and analyzed by flow cytometry. Statistical analyses were performed using the Mann-Whitney U test. Analogously, binding of 300 nM biotinylated CXCL8 in the presence or absence of CXCL9(74 -103) was compared on CHO/CXCR4 cells.
Competition of the COOH-terminal CXCL9 Peptides for Chemokine Binding to GAG-coated Plates-The ability of the COOH-terminal CXCL9 peptides to compete for GAG binding with the inflammatory chemokines CXCL8, CXCL11, or CCL2 was evaluated on GAG binding plates (BD Biosciences), which adsorb GAGs without modification to retain the protein-binding characteristics. Heparin (25 g/ml; Iduron, Manchester, UK) or HS (25 g/ml; Iduron), diluted in standard assay buffer (100 mM NaCl, 50 mM sodium acetate, 0.2% (v/v) Tween 20, pH 7.2) were coated overnight at room temperature on the plasmapolymerized surface of GAG binding plates. After three wash steps with standard assay buffer, the plates were blocked with blocking buffer (standard assay buffer enriched with 0.2% (w/v) gelatin) for 1 h at 37°C. Blocking buffer was discarded, and dilutions of CXCL9 peptide combined with recombinant human CXCL8(1-77) (PeproTech), recombinant human CXCL11(1-73) (PeproTech), or recombinant biotinylated CCL2(1-76) (ProtAffin Biotechnologie AG) (in blocking buffer) were added in duplicate and incubated for 2 h at 37°C. Unbound peptide and chemokine were removed by three washes with standard assay buffer. Bound CXCL8 or CXCL11 was incubated with biotinylated polyclonal rabbit anti-human CXCL8 or CXCL11 (PeproTech), diluted in blocking buffer, followed by detection with horseradish peroxidase-labeled streptavidin. Bound biotinylated CCL2 was directly detected using streptavidin-HRP (40). Subsequently, the plates were washed, the peroxidase activity was quantified by adding a horseradish peroxidase substrate solution, and conversion of 3,3Ј,5,5Ј-tetramethylbenzidine supplemented with 0.004% (v/v) H 2 O 2 was measured at 450 nm using a spectrophotometer. Percentage inhibition of chemokine binding to GAG was calculated as follows.

GAG-binding Peptide Blocks Neutrophil Migration in Vivo
Statistical analyses (comparison with the corresponding excess of CXCL9(74 -103)) were performed using the Mann-Whitney U test. ELISA-like Assay for Hetero-oligomerization of CXCL9(74 -103) with CXCL8 -Hetero-oligomerization of CXCL9 peptide with CXCL8 was quantified on high protein binding ELISA plates (Corning Inc., Tewksbury, MA). In brief, 1 g/ml or 500 ng/ml human recombinant CXCL8(1-77) (PeproTech) diluted in PBS was coated overnight at 4°C on 96-well plates. Plates were washed three times with assay buffer (0.5% (v/v) Tween 20 in PBS) and blocked at 37°C with assay buffer enriched with 0.1% (w/v) casein. Subsequently, a dilution series of biotinylated CXCL9 peptide was added and allowed to interact with CXCL8. After washing, biotinylated CXCL9 peptide was detected by peroxidase-conjugated streptavidin. Peroxidase activity was quantified by measuring the conversion of 3,3Ј,5,5Ј-tetramethylbenzidine at 450 nm.
Neutrophil Migration in Vitro and in Vivo-The chemotactic activity of CXCL8(6 -77) in combination with CXCL9 peptides for neutrophilic granulocytes in vitro was compared using a 48-well Boyden chamber-based chemotaxis assay, and migrated neutrophils were counted microscopically, as described previously (42). In vivo blood vessel extravasation of circulating neutrophil granulocytes to the knee cavity was examined by injection of endotoxin-free recombinant CXCL8(1-77) (PeproTech) or vehicle (10 l of 0.9% NaCl) in the tibiofemural articulation of C57Bl/6 mice. A group of mice received intravenously CXCL9 peptide (100 g; 100 l) at the same time as CXCL8 or saline. Endotoxin concentrations were evaluated with the Limulus amoebocyte lysate test (Cambrex, East Rutherford, NJ). For preparation of MSU crystals, 1.68 mg of uric acid was dissolved in 500 ml of a 0.01 M NaOH solution heated to 70°C. The pH of the solution was kept between pH 7.1 and 7.2. Once dissolved, the solution was filtered through a sterile syringe filter with a 0.22-m pore size. The sterile solution was stirred for 24 h at room temperature to form the crystals, the supernatant was discarded, and the MSU crystals were washed with ethanol. Subsequently the size of the crystals was reduced by sonication, dried, and resuspended in PBS. PBS or 100 g of crystals (in 10 l) were injected in the tibiofemoral knee joint. A group of mice received intravenous CXCL9 peptide (10, 30, or 100 g in 100 l) or vehicle. The cells present in the tibiofemoral articulation were harvested 3 h after CXCL8 or MSU crystal administration by washing the cavity with 10 l of 3% (w/v) bovine serum albumin in PBS and diluted in 90 l of 3% (w/v) bovine serum albumin in PBS. The number of total leukocytes was determined by counting leukocytes in a Neubauer chamber after staining with Turk's solution. Differential counts were obtained from cytospin preparations by evaluating the percentage of each leukocyte type on a slide stained with May-Grunwald-Giemsa. Percentages of cells present in the joint lavages were statistically compared between CXCL9 peptide-treated and untreated mice using the Mann-Whitney U test. The experimental in vivo protocols (involving the use of laboratory animals) were reviewed and approved by the Animal Ethical Committee of KU Leuven or the University of Minas Gerais, and Belgian and European guidelines concerning the handling of laboratory animals were followed. Mice were sacri-ficed by injection of an overdose of Nembutal prior to the collection of joint lavages.

Results
Purification and Identification of Natural COOH-terminally Truncated CXCL9 -Previously, induction experiments showed that high amounts of CXCL9 were produced by PBMCs and fibroblasts upon stimulation with IFN-␥ and dsRNA (35). Furthermore, in the case of fibroblasts, synergistic induction of CXCL9 was observed when fibroblasts were stimulated with IFN-␥ and dsRNA or LPS, resulting in massive (ϳ1 g/10 6 cells) secretion of CXCL9. In contrast, in the case of PBMCs, LPS clearly inhibited IFN-␥-induced CXCL9 production. In order to investigate posttranslational modifications on CXCL9, natural CXCL9 was purified from the conditioned medium of PBMCs stimulated with dsRNA and IFN-␥ and from fibroblasts stimulated with dsRNA, IFN-␥, and LPS. Secreted proteins were separated, combining heparin affinity, cation exchange and RP chromatography. Analysis by mass spectrometry of those fractions containing CXCL9, as determined by ELISA, showed that these fractions did not contain any intact CXCL9; however, several COOH-terminally truncated isoforms were present (i.e. CXCL9  (Table 1). Interestingly, all of these CXCL9 isoforms lost their characteristic cationic COOH-terminal region. To illustrate, Fig. 1 shows a RP-HPLC profile of specific CXCL9-containing fractions selected after heparin affinity and cation exchange chromatographic purification of the conditioned medium from PBMCs stimulated with dsRNA and IFN-␥. ELISA data revealed the presence of CXCL9 in RP-HPLC fractions 49 -53. More detailed analysis of the proteins present in these fractions by mass spectrometry disclosed the presence of two different proteins with a relative M r of 9019.40 and 8718.35, which correspond to the theoretical relative molecular masses of CXCL9(1-81) and CXCL9(1-78), respectively.
Interaction of CXCL9(74 -103) with Soluble GAGs-Given the high number of positive charges (more than 1 in 2 amino acids), one could presume a GAG binding function for the COOH-terminal region of CXCL9. Therefore, the affinity of the longest of the synthesized peptides (i.e. CXCL9(74 -103)) for LMWH, HS, and DS was determined by isothermal fluorescence titration (Fig. 2). Using this technique, TAMRA fluorescence quenching was measured upon interaction of NH 2 -terminally TAMRA-labeled CXCL9(74 -103) with LMWH, HS, or DS. The resulting binding isotherms were analyzed by nonlinear regression according to a bimolecular association reaction (43), which gave K d values of 61.21, 4.76, and 4.38 nM, respec- AUGUST 28, 2015 • VOLUME 290 • NUMBER 35

GAG-binding Peptide Blocks Neutrophil Migration in Vivo
tively. Interestingly, CXCL9(74 -103) shows 12-fold higher affinity for HS compared with LMWH. It can be concluded that CXCL9(74 -103) binds with high but different affinity to these glycosaminoglycans.
CXCL9(74 -103) Interacts with Cellular GAGs-To confirm these results on cellular GAGs, CXCL9(74 -103) was also sitespecifically biotinylated at the NH 2 terminus (to avoid modification of the lysine side chains and interference of biotin groups with GAG binding), and binding of biotinylated CXCL9(74 -103) to CHO cells was evaluated by flow cytometric analysis (Fig. 3). As reported, CXCL9(74 -103) is not able to bind to CXCR3 (44). Therefore, CHO/CXCR3A and CHO/CXCR3B cells were used to evaluate the binding of CXCL9(74 -103) to GAGs. As an additional control, a cell type with an irrelevant receptor for CXCL9 and the peptide, namely CHO/CXCR4 cells, was used. Flow cytometric analysis demonstrated that CXCL9(74 -103) binds dose-dependently to CXCR3A, CXCR3B, and CXCR4/CHO cells (Fig.  3B). Because CXCL4 is known to be one of the strongest GAG binding chemokines (37,45), we also evaluated the interaction of an NH 2 -terminally biotinylated COOH-terminal peptide of CXCL4, namely CXCL4 , with cellular GAGs as a reference peptide. Fig. 3B shows that CXCL4(47-70) binds dose-dependently to GAGs on CXCR3A, CXCR3B, and CXCR4/CHO cells but with lower affinity than CXCL9(74 -103). To ensure that binding was GAG-mediated, CHO cells were treated with NaClO 3 to reduce the expression of HS. NaClO 3 treatment significantly reduced the expression of HS on the CHO cells (on average 75%) and consequently the binding of CXCL9(74 -103) and CXCL4   (Fig. 3A).
COOH-terminal CXCL9 Peptides Compete with CXCL8, CXCL11, and CCL2 for Binding to Heparin, HS, or Cellular GAGs-The main human neutrophil attractant, CXCL8, has been demonstrated to be a relatively strong heparin-binding molecule (46). Using GAG binding plates in an ELISA-like assay, the ability of CXCL9(74 -103) to inhibit binding of CXCL8 to heparin was evaluated (Fig. 4A). As expected, given its highly positively charged nature, CXCL9(74 -103) clearly competed with CXCL8. If the molar concentration of CXCL9(74 -103) exceeded the concentration of CXCL8 by a factor 10, CXCL8 binding to heparin was reduced by 50%. Interestingly, the shorter COOH-terminal CXCL9 peptides CXCL9(79 -103), CXCL9(82-103), and CXCL9(86 -103) were less potent competitors. To achieve a 50% reduction of CXCL8 binding, a 100-fold (at least) excess of the shorter CXCL9 peptides was needed. Interestingly, CXCL9(74 -103) competed with CXCL8 for heparin binding with a similar potency as intact CXCL9(1-103). Additionally, the capacity of CXCL9(1-78) to compete for GAG binding was evaluated. A 10-fold molar excess of CXCL9(1-78) did not inhibit the binding of CXCL8 to heparin, suggesting that this CXCL9 isoform lost most of its capacity to bind to GAGs. GAG binding competition between CXCL9(74 -103) and CXCL8 was confirmed on CHO cells (Fig.  4B). Indeed, the binding of biotinylated CXCL8 to cellular GAGs was reduced by 50% in the presence of a 10-fold excess of CXCL9(74 -103).
To evaluate whether GAG binding competition was restricted to the ELR ϩ CXC chemokine CXCL8, binding competition assays were also performed with the ELR Ϫ CXC chemokine CXCL11, the most potent CXCR3 ligand, and the CC chemokine CCL2, a strong monocyte attractant (Fig. 4, C and  D). CXCL9(74 -103) clearly competed with both chemokines for GAG binding. Again, CXCL9(74 -103) turned out to be the most potent CXCL9 COOH-terminal peptide. These results indicate that CXCL9(74 -103) may compete with a broad range of chemokines belonging to different chemokine subfamilies.

Overview of the CXCL9 isoforms that were purified from the conditioned medium of PBMCs and fibroblasts and of the complementary COOH-terminal CXCL9 peptides synthesized to investigate their potential anti-inflammatory function
For clarity, basic amino acids were underlined, and amino acids 11-59 were left out of the amino acid sequences.

GAG-binding Peptide Blocks Neutrophil Migration in Vivo
erophilic interactions can be formed, and these interactions can dramatically affect a chemokine's biological activity. For example, CXCL8 and CXCL4 tend to dimerize, and this interaction decreases the potency of CXCL4 to inhibit proliferation of endothelial cells (47). This way, hetero-oligomerization might embody another degree of chemokine function regulation in a given in vivo chemokine environment. To exclude the possibility that CXCL9(74 -103) reduces CXCL8 binding to GAGs through direct interaction with CXCL8, hetero-oligomerization of CXCL9(74 -103) with CXCL8(1-77) was tested by isothermal fluorescence titration. When CXCL8 was titrated to the TAMRA-labeled CXCL9 peptide (600 nM), no significant interaction could be observed (Fig. 5). 10% quenching and late saturation of TAMRA fluorescence is indicative of a very weak and/or a rather nonspecific interaction between the two molecules. Analogously, no direct interaction between the peptide and CXCL8 was detected in an ELISA-like assay, in which the binding of biotinylated CXCL9(74 -103) was measured on a 96-well plate coated with CXCL8 (data not shown). As a positive control, the COOH-terminal CXCL4 peptide (residues 47 to 70) and EGF were tested in parallel and displayed clear heterooligomerization, as reported recently (37).

CXCL9(74 -103) No Longer Signals through CXCR3 and Does Not Influence in Vitro Chemotaxis of Neutrophils in Response to
CXCL8 -The longest of the synthesized CXCL9 peptides (i.e. CXCL9(74 -103)) has been reported to be no longer able to bind to CXCR3, the cognate receptor for intact CXCL9 FIGURE 1. Identification of COOH-terminally truncated isoforms of CXCL9 from leukocyte-derived conditioned medium. PBMCs were stimulated with a combination of dsRNA and IFN-␥ for 120 h. Conditioned medium was purified using a four-step chromatographic procedure. The results of the final C8 RP-HPLC separation are depicted in A. Proteins were eluted in an acetonitrile gradient in 0.1% (v/v) TFA and were detected on-line by UV absorption ( ϭ 214 nm) and by ion trap mass spectrometry. UV absorption (dotted line) and total ion chromatogram (representing the sum of all ions reaching the detector of the mass spectrometer; solid line) are shown in A. Subsequently, CXCL9 immunoreactivity of the RP-HPLC fractions was determined by ELISA (histograms; A). The mass spectra of CXCL9-positive fractions were more closely analyzed. B, averaged mass spectrum for molecules eluting in CXCL9-containing fractions 49 -53, displaying the intensity of the detected ions with their specific mass/charge ratios. In order to calculate the relative molecular mass (M r ) of the proteins corresponding with these ions, Bruker deconvolution software was used (inset on the right in B). The experimentally determined M r of the proteins is compared with the theoretical M r of CXCL9 isoforms indicated in the inset at the top of B. mAU, milliabsorbance units.
To exclude any interference of CXCL9(74 -103) with the interaction of CXCL8 with its receptor, the potency of CXCL9(74 -103) to affect CXCL8-induced neutrophil migration in vitro in a Boyden chamber chemotaxis assay, assumed to be a GAG-independent process, was evaluated (Fig. 6C). Even at a dose of 300 ng/ml, CXCL9(74 -103) did not influence the chemotactic potency of CXCL8 (1.5-15 ng/ml). No statistical differences on chemotactic responses were observed when neutrophils were stimulated with CXCL8 in the absence or presence of CXCL9 peptide. In addition, a cytotoxic effect of CXCL9(74 -103) on neutrophils was excluded in a lactate dehydrogenase cytotoxicity assay (data not shown).

CXCL9(74 -103) Lowers CXCL8-and MSU-induced Neutrophil Recruitment to the Joints-Based on its in vitro competition
for GAG binding combined with its abrogated G protein-coupled receptor-dependent activity, CXCL9(74 -103) might possess promising immunosuppressive properties. By competing with chemokines for GAG binding, this peptide might lower the chemokine's capacity to recruit leukocytes to a site of inflammation. Therefore, the potential anti-inflammatory capacity of CXCL9(74 -103) was evaluated in two mouse models of acute inflammation (Fig. 7). Because CXCL8 is the main human neutrophil attractant and an abundant chemokine in the arthritic joint (48), CXCL8 was used as a prototypic inflammatory chemokine. In a first acute inflammation model, injection of 1 g of CXCL8 into the tibiofemoral articulation of C57BL/6 mice results in a massive influx of leukocytes (Fig. 7A). This massive influx into the knee cavity was clearly reduced upon concomitant intravenous injection of 100 g of CXCL9(74 -103) (a reduction from ϳ43 ϫ 10 4 cells to 14 ϫ 10 4 cells). As shown by differential counting of the different leuko-  cyte subtypes, the infiltrated leukocyte population predominantly consists of neutrophils. The extravasation of these neutrophils is inhibited by 92% by intravenous injection of CXCL9(74 -103) and explains the reduced general influx of leukocytes (Fig. 7B). Second, the inhibitory activity of the CXCL9(74 -103) peptide was evaluated in a model of gout in mice. In this model, injection of MSU crystals induces an inflammasome-dependent production of CXCL1/2 chemokines, which are necessary for CXCR2-mediated recruitment of neutrophils and neutrophil-dependent inflammatory hypernociception (49,50). As seen in Fig. 7, C and D, injection of MSU crystals in the knee joint of mice induced significant recruitment of leukocytes, mostly neutrophils. Injection of 100 g of the CXCL9 peptide intravenously just before the injection of MSU crystals greatly inhibited (87%) leukocyte and neutrophil infiltration in this gout inflammation model. In contrast, the lower doses tested (i.e. 10 and 30 g) were not able to reduce MSU-induced inflammation.

Discussion
Posttranslational modification constitutes an important mechanism of regulating the activity of inflammatory chemokines (27). Frequently, chemokines seem to be subject to NH 2terminal truncation (28). In contrast, COOH-terminal truncation is only occasionally observed (51)(52)(53)(54)(55)(56)(57). The consequences of COOH-terminal truncation are less clear and seem to be rather limited to modification of GAG binding properties (58).
Upon stimulation with the Th1 cytokine IFN-␥ and the TLR3 ligand dsRNA, mimicking an inflammatory situation, both PBMCs and diploid fibroblasts produce substantial amounts of CXCL9 (35). Interestingly, the amount of natural CXCL9 produced by these cells is significantly (10 -100-fold) higher than the amount of the 3-fold more potent CXCR3 ligand CXCL11 (data not shown). In the current study, chromatographic purification of the natural human CXCL9 and evaluation of its molecular mass revealed that under the tested inflammatory conditions, all produced CXCL9 was processed COOH-terminally into CXCL9(1-74), CXCL9(1-75), etc. (Table 1). Indeed, under these conditions, natural CXCL9 seems to lose its unique and characteristic COOH-terminal region by which it differs from the other CXCR3 ligands and most other chemokines. The enzymes responsible for the COOH-terminal truncation of CXCL9 in these in vitro cell cultures have not been identified. However, plenty of enzymes have been described to COOHterminally cleave CXCL9 and may be involved in the formation of such isoforms (see above). Because CXCL9(1-73) has been shown to be a less potent T lymphocyte chemoattractant compared with intact CXCL9, these COOH-terminal truncations might be crucial to down-regulate CXCL9 activity in a timely manner. Such posttranslational processing under inflammatory conditions might constitute a negative feedback loop. This COOH-terminal region may either play a role in receptor binding or be important for the cis-presentation of CXCL9 on GAGs to its receptor. In general, GAGs are not an absolute requirement for receptor binding and signaling in vitro, but for some chemokines, they have been shown to enhance the sensitivity of a cell to chemokine stimulation (61).
Given the high number of positive charges, one could presume a GAG binding function for the COOH-terminal region of CXCL9. Furthermore, the CXCL9(74 -103) peptide was reported not to bind to CXCR3 (44). Keeping this in mind, the following hypothesis was formulated. Do these COOH-terminal CXCL9 peptides, through competition with active chemotactic chemokines for GAG binding, reduce chemokine immobilization and presentation, and do they thereby inhibit chemokine-induced leukocyte migration? Because this region tends to be often removed, it might play a role on its own during an inflammatory reaction. Our data show that the longest of the synthesized peptides (CXCL9(74 -103)) clearly interacts with both soluble and cellular GAGs with K d values in the low nanomolar range. Evaluating heparin binding of the most potent human neutrophil chemoattractant CXCL8 in the presence of the chemically synthesized CXCL9 peptides CXCL9(74 -103), CXCL9(79 -103), CXCL9(81-103), and CXCL9(86 -103) showed that the longest peptide, CXCL9(74 -103), clearly reduced CXCL8 binding to heparin, with an efficacy similar to that of intact CXCL9. Furthermore, reducing the length of the COOH-terminal CXCL9 peptide gradually decreased the capacity to compete with CXCL8. These data stress the importance of amino acids 74 -78 in the heparin binding characteristics. However, the inability of CXCL9(1-78) to compete with CXCL8 for heparin binding suggests that amino acids 74 -78 are necessary but not sufficient for GAG binding. CXCL9(74 -103) indeed contains two typical heparin binding motifs (BBXB and XBBBXXBX) ( Table 1) (i.e. Lys 75 -Lys 78 and Lys 85 -Lys 89 ), and both might play an important role in heparin binding of CXCL9.
Competition between CXCL8 and CXCL9(74 -103) was also confirmed on cellular GAGs. Furthermore, competition was not limited to CXCL8 and could be extended to CXCL11 and CCL2 (representatives of different chemokine subclasses).
The possibility that CXCL9(74 -103) modulates GAG binding properties of CXCL8 through hetero-oligomerization was excluded by both isothermal fluorescence titration and an ELISA-like assay, showing no direct interaction between CXCL9(74 -103) and CXCL8. Therefore, the results seen in the GAG binding competition assays can be interpreted as direct competition for GAG binding between CXCL9(74 -103) and CXCL8 and should not be the result of chemokine hetero-oligomerization reducing the affinity of CXCL8 for GAGs.
Based on the necessity of chemokine binding to GAGs for exerting biological functions in vivo, an anti-inflammatory action for CXCL9(74 -103) through GAG binding competition was evaluated in two murine models of acute inflammation, characterized by neutrophil extravasation. In both the CXCL8 model and the less chemokine-specific gout model, intravenous injection of the CXCL9 peptide efficiently inhibited neutrophildependent inflammation. The fact that the in vitro chemotactic potency of CXCL8 for neutrophils, which is assumed to be based on a GAG-independent soluble chemokine gradient, was not affected by the presence of the CXCL9 peptide confirms the presumption that the inhibitory effect of CXCL9(74 -103) is GAG-and not receptor-mediated.
Induction experiments show that the amount of CXCL9 produced by fibroblasts, endothelial cells, and peripheral blood mononuclear cells can clearly exceed the amount of CXCL11, which is active at lower concentrations (35,62). Therefore, COOH-terminal cleavage might lead to the generation of quite high local concentrations of CXCL9(74 -103), which compete with intact functional chemokines for GAG binding. As such, enzymatic cleavage of the COOH-terminal region of CXCL9 might contribute to the resolution of inflammation because it not only directly reduces the activity of CXCL9 but also indi-rectly, through competition for GAG binding, lowers the activity of other chemokines.
The strength of the peptide in vivo might be based on the fact that the peptide is chemokine-derived because it might therefore inherit the chemokine-intrinsic specificity for certain glycosaminoglycan sequences expressed by endothelial cells under inflammatory conditions. The exact composition of the GAGs on endothelial cells and in the extracellular matrix has been shown to depend on the type and location of the cells and the pathophysiological state of the tissue (63,64). Further in vivo investigation is necessary to evaluate the efficacy of the peptide in specific inflammation models (with different GAG profiles)  1-103). A, the CXCR3-dependent signaling capacity of CXCL9(74 -103) was evaluated in CHO cells transfected with human CXCR3 loaded with the ratiometric Ca 2ϩ -binding molecule Fura-2. The intracellular Ca 2ϩ concentrations ([Ca 2ϩ ] i , in nM) were calculated from the continuously measured fluorescence intensity ratio of Fura-2 (510 nm) in a LSB50 luminescence spectrophotometer. The cells were stimulated with either 266 ng/ml intact CXCL9 (as a positive control) (left) or 1600 ng/ml CXCL9(74 -103) (right). Potential desensitization of CXCR3 by CXCL9(74 -103) was assessed by stimulating the cells with 1600 ng/ml CXCL9(74 -103) 100 s prior to stimulation with 266 ng/ml intact CXCL9 (right). The response of the cells to stimulation with intact CXCL9 is compared with the response of the cells with prior stimulation with buffer (left). B, CHO cells transfected with human CXCR3 and loaded with Fura-2 were stimulated with 266 ng/ml CXCL9(1-103) upon prior incubation (10 min) with either buffer (left) or 10 g/ml CXCL9(74 -103) (right). C, the neutrophil chemotactic activity of CXCL8 (6 -77) in the presence of different concentrations of CXCL9(74 -103) was evaluated using 48-well plate Boyden chambers. The mean ϩ S.E. (n ϭ 8) chemotactic index, depicted on the y axis, is calculated by dividing the amount of migrated cells in response to CXCL8 by the amount of spontaneously migrated cells (in response to buffer). A statistical comparison to evaluate the effect of CXCL9(74 -103) was performed using the Mann-Whitney U test.
to uncover whether it shows a certain specificity for specific pathological conditions.
Although chemokines are clearly beneficial in the battle against infectious organisms and during wound healing after tissue injury, excessive on-going expression of chemokines has been associated with inflammatory disorders, characterized by high levels of leukocyte infiltration (65,66). Therefore, interfering with the chemokine system might constitute an attractive way of treating such diseases. Interestingly, viruses and ticks have been shown to successfully use this strategy to evade the immune system of the host (67). These organisms encode homologs of chemokines and chemokine receptors or proteins that are able to bind chemokines and thereby control their activity. Therapeutic intervention in the chemokine system has long been focused on the development of chemokine and chemokine receptor antagonists (68). However, the use of these agents in clinical trials, in general, remained unsuccessful. More recently, it has been suggested that interruption of chemokine-GAG interactions might represent an innovative, useful, and effective way of interfering with chemokine action (69). Indeed, in the case of the recently developed CellJammer approach, mutant chemokines are being engineered with increased GAG binding affinity (leaving the specificity unchanged) and abrogated G protein-coupled receptor signaling, creating antagonists for the wild type chemokines through competition for GAG binding. Until now, this approach has been successfully applied to generate antagonists for CXCL8 and CCL2 (43, 70 -72). Indeed, CXCL8-and CCL2-based decoy proteins were shown to moderate inflammation in different mouse models (ischemia/reperfusion, renal allograft rejection, antigen-induced arthritis, experimental autoimmune uveitis, MOG 35-55 -induced chronic experimental autoimmune en-cephalomyelitis), supporting the potential of this strategy. In view of this, the CXCL9(74 -103) might be a lead molecule for the generation of therapeutic peptides that compete with functional chemokines for GAG binding, and its efficacy confirms the applicability of GAGs as a therapeutic target.
Author Contributions-V. V., J. V. D., P. P., and A. M. designed the study. V. V., P. P., and A. M. wrote the paper. All authors analyzed and interpreted the data. V. V., R. J., D. B., N. K., N. B., I. R., F. A. A., P. P., and A. M. performed the experiments. R. J., D. B., F. A. A., and M. M. T. were involved in the mouse inflammation model studies. N. K. and A. J. K. performed isothermal fluorescence titration assays and GAG binding competition assays. V. V., A. M., and P. P. synthesized and purified CXCL9 and performed mass spectrometry and Edman degradation, GAG binding assays, oligomerization studies, and signaling and chemotaxis assays. All authors corrected and approved the final version of the manuscript.