Rationally Evolving MCP-1/CCL2 into a Decoy Protein with Potent Anti-inflammatory Activity in Vivo*

Leukocyte recruitment from the blood into injured tissues during inflammatory diseases is the result of sequential events involving chemokines binding to their GPC receptors as well as to their glycosaminoglycan (GAG) co-receptors. The induction and the crucial role of MCP-1/CCL2 in the course of diseases that feature monocyte-rich infiltrates have been validated in many animal models, and several MCP-1/CCL2 as well as CCR2 antagonists have since been generated. However, despite some of them being shown to be efficacious in a number of animal models, many failed in clinical trials, and therapeutically interfering with the activity of this chemokine is not yet possible. We have therefore generated novel MCP-1/CCL2 mutants with increased GAG binding affinity and knocked out CCR2 activity, which were designed to interrupt the MCP-1/CCL2-related signaling cascade. We provide evidence that our lead mutant MCP-1(Y13A/S21K/Q23R) exhibits a 4-fold higher affinity toward the natural MCP-1 GAG ligand heparan sulfate and that it shows a complete deficiency in activating CCR2 on THP-1 cells. Furthermore, a significantly longer residual time on GAG ligands was observed by surface plasmon resonance. Finally, we were able to show that MCP-1(Y13A/S21K/Q23R) had a mild ameliorating effect on experimental autoimmune uveitis and that a marginal effect on oral tolerance in the group co-fed with Met-MCP-1(Y13A/S21K/Q23R) plus immunogenic peptide PDSAg was observed. These results suggest that disrupting wild type chemokine-GAG interactions by a chemokine-based antagonist can result in anti-inflammatory activity that could have potential therapeutic implications.

Chemokines are small secreted proteins that function as messengers by orchestrating activation and directional migration of specific subtypes of leukocytes from the blood stream into injured tissues (1)(2)(3). They exert their specific functions through interactions with G protein-coupled receptors on the one hand and with cell-surface proteoglycans on the other hand, which we therefore call chemokine co-receptors to differentiate them from the "classical" chemokine GPC receptors. Proteoglycans are commonly defined as silent receptors, although evidence is accumulating that shows that proteoglycans binding chemokines exhibit non-classical downstream signaling in endothelial cells. 2 Within the CC family of chemokines, MCP-1/CCL2 (monocyte chemoattractant protein-1) specifically recruits and activates CCR2-positive cells. Thus, MCP-1 is highly induced in a variety of diseases that feature monocyterich cellular infiltrates, such as in atherosclerosis (4,5), congestive heart failure (6,7), rheumatoid arthritis (8 -10), inflammatory bowel disease (11), and uveitis (12)(13)(14). Unlike many other chemokines, MCP-1/CCL2 binds exclusively to CCR2, and the importance of this chemokine-receptor pair in mouse models of inflammatory diseases, including multiple sclerosis, atherosclerosis, and uveitis, has been confirmed (12)(13)(14)(15)(16)(17). It has, for example, been shown that knock-out mice lacking MCP-1/ CCL2 are unable to recruit monocytes and T cells to inflammatory lesions (18 -21). LDL-receptor/MCP-1-deficient and apoB-transgenic/MCP-1-deficient mice showed considerably less lipid deposition and macrophage accumulation throughout their aortas compared with the wtMCP-1 3 strains (22,23). Finally, CCR2 knock-out mice were found to be resistant to experimental autoimmune encephalomyelitis (16).
Many attempts to interfere with MCP-1/CCL2 activity have been made by applying several therapeutic strategies. Neutralizing mAbs against both MCP-1/CCL2 and CCR2 have been tested in vivo, where they exhibited an ameliorating effect on mouse crescentic glomerulonephritis and on restenosis in nonhuman primates (24,25). Truncation or modification of the N terminus yielded receptor antagonists of MCP-1-(9 -76), which failed to reduce the severity of the spontaneous onset of arthritis in lpr mice (26). These mutants also lacked efficacy in a mouse model of atherosclerosis (27,28). MCP-1/CCL2 oligomerization is known to be required for activity in vivo (29); thus, the abrogation of oligomerization has been investigated for its potential in anti-inflammatory activity. For instance, the mutant form CCL2(P8A) has been engineered, which is an obligate monomer and which was unable to recruit leukocytes into the peritoneal cavity and into lungs of ovalbumin-sensitized mice (30). Glycosaminoglycan (GAG) binding of chemokines has been shown to be essential to promote cell migration in vivo (29). Heparin and heparin derivatives exhibited mild anti-inflammatory effects in animal models as well as in humans (32). Although the true mechanism of action of heparin in these studies is not known, the inhibition of the interaction between chemokines and cell surface GAGs has been suggested as a viable therapeutic strategy. GAG binding-deficient variants of MCP-1 (29,31) and RANTES (29) have been generated in order to obtain mutant chemokines that would compete with their wild type counterparts on leukocyte G protein-coupled receptors, whereas their impaired GAG binding would result in anti-inflammatory activity. The antagonistic activity of these mutants in vivo has been related also to their oligomerization behavior, which differed from that of the wild type chemokines (33,34). These so-called GAG knock-out mutants represent the mirror image of the GAG knock-in mutants presented in this paper (see below), by which we aim to antagonize at the GAG binding site while impairing the G protein-coupled receptor activity of the chemokine.
Although several chemokines and chemokine receptors have been targeted successfully in animal models of immune-mediated diseases, species specificity of small antagonists and neutralizing mAbs complicate the use of these compounds in animal models. Hence, the development and the assessment of efficacy of some antagonists may be delayed. Additionally, the low oral bioavailability of some of the compounds may also prevent broad clinical applications. Up to now, the data on the effects of chemokine blockade in patients are still limited and unsatisfying. A trial with an anti-CCL2 mAb in patients with rheumatoid arthritis resulted in neither clinical nor immunohistological improvement, and administration of the highest dose may have been associated with worsening of arthritis (35). The effect of chemokines (or chemokine receptor) blockade seems, furthermore, to be dependent on the time of inhibitor administration. Administration of anti-CCR2 mAb or a small molecule antagonist of MCP-1 before collagen-induced or adjuvans-induced arthritis onset resulted in disease amelioration, whereas CCR2 blockade after disease onset aggravated the clinical and histological scores of experimental arthritis (36 -38). Similarly, treatment with MCP-1-(9 -76) (39), which inhibits MCP-1 binding to CCR2, prevented the complete Freund's adjuvant (CFA)-mediated exacerbation of the spontaneous development of arthritis in MRL-lpr mice but was less efficacious if administered after disease onset (26,37).
We have generated new MCP-1/CCL2 mutants by site-directed mutagenesis, aiming to modulate the function of the protein in vitro and in vivo. Assuming that the chemokine-GAG interaction is indispensable for chemokine activity in vivo (29), we identified solvent-exposed residues that, when mutated into basic amino acids, increased the protein affinity to the MCP-1/ CCL2-specific GAG ligands. We additionally mutated position 13 into alanine, which abolished signaling through CCR2. From all of these dominant-negative mutants, Met-MCP-1(Y13A/ S21K/Q23R) showed the highest GAG binding affinity and complete knock-out of CCR2 signaling in vitro. In addition, we show here that it considerably reduced ocular monocyte tissue infiltration in a rat experimental autoimmune uveitis model (EAU). In a separate study, we have recently shown that this mutant exhibits strong inhibitory activity also in a murine model of myocardial infarction and restenosis. 4 Although only a few residues have been mutated, the results suggest that Met-MCP-1(Y13A/S21K/Q23R) acts as a protein-based GAG antagonist and that the mechanism for its inhibitory activity arises from the dual modifications, as we anticipated recently (40).

EXPERIMENTAL PROCEDURES
Generation of Human MCP-1 Mutants for Expression in Escherichia coli-All MCP-1 mutant constructs were generated in the context of MCP-1 M64I, referred to as WT*. WT* behaves identically to wild type in binding and activity assays. This alteration in the primary structure improves the purity and homogeneity of the mutants by eliminating the formation of species containing methionine sulfoxide at position 64 (41).
The gene for WT* MCP-1 was constructed by standard gene synthesis techniques with optimal codon usage for expression in E. coli and was cloned into a pCP116 expression vector (42). Features of the vector include a T7 promoter for high level, isopropyl 1-thio-␤-D-galactopyranoside-inducible expression of the gene in E. coli and an ampicillin resistance marker for selection in E. coli.
Each mutation was introduced employing the QuikChange II site-directed mutagenesis method (Stratagene) according to the manufacturer's instructions. The mutagenesis primers were synthesized by Invitrogen.
The pCP116 plasmids containing the MCP-1 mutant genes were transformed into E. coli strain XL1-Blue Supercompetent (Stratagene). Prior to expression, the sequence of each construct was verified by DNA sequencing (GeneXpress).
Protein Expression and Purification-The pCP116 MCP-1 mutant constructs were transformed into the BL21 Star TM (DE3) E. coli strain (Invitrogen). Starting cultures were prepared and used for protein expression. Cultures were grown in 3-liter Erlenmeyer flasks under shaking at 37°C in LB broth containing 150 g/ml ampicillin to an A 600 of 0.6. Protein expression was induced by the addition of 0.5 mM isopropyl-␤-D-thiogalactopyranoside. Cells were incubated with shaking for an additional 3-5 h and harvested by centrifugation for 15 min at 6,500 ϫ g. Each g of wet cell pellet was resuspended in 4 ml of solution buffer, containing 10 mM KH 2 PO 4 , pH 7.5, 5 mM MgCl 2 , and disrupted by sonication on ice. The lysate was then cleared by centrifugation for 30 min at 20,000 ϫ g at 4°C. The inclusion body pellets were solubilized in 10 ml of solution buffer, containing 10 mM KH 2 PO 4 , pH 7.5, 6 M guanidinium hydrochloride per g of wet cell pellet under stirring for 1 h at room temperature. After centrifugation for 30 min at 20,000 ϫ g at 4°C, the supernatant was dialyzed against 10 mM KH 2 PO 4 , pH 7.5, at 4°C. The precipitant was spun down, and the solution was loaded on a SP-Sepharose high performance column (Amersham Biosciences). Mutants were eluted with a linear gradient from 10 mM KH 2 PO 4 , pH 7.5, 1 mM EDTA to 10 mM KH 2 PO 4 , pH 7.5, 1 mM EDTA, 1 M NaCl over 60 min with a flow rate of 1 ml/min; the MCP-1 mutants eluted at 0.6 -0.8 M NaCl. Peak fractions were pooled and purified by reversed-phase HPLC on a C18 column. The mutants were eluted with a nonlinear gradient: from 10 to 40% acetonitrile (0.1% trifluoroacetic acid) in 5 min, from 40 to 60% acetonitrile (0.1% trifluoroacetic acid) in 20 min, and from 60 to 90% acetonitrile (0.1% trifluoroacetic acid) in 5 min with a flow rate of 0.8 ml/min. Proteins eluted at 48 Ϯ 5% acetonitrile and were refolded by loading them on a SP-Sepharose high performance column as described above. Peak fractions were dialyzed against PBS and concentrated to 1 mg/ml.
Purity and identity of the MCP-1 mutants were confirmed by silver-stained SDS-PAGE and nano-HPLC electrospray ionization-tandem mass spectrometry, respectively. The bacterial endotoxin content of the protein samples was determined by the limulus amebocyte lysate test (BioWhittaker, Cambrex).
Fluorescence Spectroscopy-Steady-state fluorescence measurements were performed on a PerkinElmer Life Sciences LS50B fluorometer, as described previously (43), and analyzed using the program Origin (Microcal Inc., Northampton, MA). The temperature was maintained at 22°C during all experiments by coupling to an external water bath.
Isothermal Fluorescence Titration (IFT) Experiments-The emission spectra of 0.5 M wtMCP-1/CCL2 and mutant solutions in PBS were recorded over the range of 310 -390 nm upon excitation at 282 nm. The excitation and emission slit widths were set at 15 and 20 nm, respectively, and a 290-nm cut-off filter was inserted into the emission path to avoid stray light. The spectra were recorded at a speed of 200 nm/min. The addition of GAG aliquots was followed by an equilibration period of 1 min before the next spectrum was recorded. After background subtraction, the spectra were integrated, and the normalized mean changes in fluorescence intensity (Ϫ⌬F/F 0 ) obtained from three independent experiments were averaged and plotted against the volume-corrected concentration of the added ligand. The resulting binding isotherms were analyzed by non-linear regression to an equation describing a bimolecular association reaction as described elsewhere (44).
Surface Plasmon Resonance Analysis-Binding of wtMCP-1/ CCL2, Met-MCP-1(Y13A/S21K), and Met-MCP-1(Y13A/ S21K/Q23R) to unfractionated heparan sulfate was investigated on a BIAcore 3000 instrument (BIAcore AB, Uppsala, Sweden). The immobilization of the biotinylated GAG onto a streptavidin-coated CM4 sensor chip was performed according to an established protocol, described recently (45). The actual binding interactions were recorded at 25°C in PBS, pH 7.4, containing 0.01% (v/v) P20 surfactant (BIAcore AB). 4-min injections of different protein concentrations at a flow rate of 30 l/min were followed by 10-min dissociation periods in buffer and a pulse of 1 M sodium chloride for complete regeneration. The maximum response signals of protein binding to the GAG surface, corresponding to the plateaus of the respective sensorgrams, were used for Scatchard plot analysis and the calculation of equilibrium dissociation constants (K d values). The experiments were all carried out at a protein concentration of 600 nM, which was chosen on the basis of a series of ligand binding experiments performed at several protein concentrations rang-ing from 50 nM up to 1.2 M. Using the BIAevaluation software, k off values were calculated by separately fitting the dissociation phases according to the 1:1 Langmuir interaction model.
Chemokine Displacement Assay-wtMCP-1 (PeproTech) was labeled with fluorescein isothiocyanate (Sigma). For this purpose, 500 g of protein (in PBS) and a 4-fold molar excess of fluorescein isothiocyanate were mixed and incubated at room temperature under moderate shaking (200 rpm) for 2 h. Free fluorescein isothiocyanate was removed by centrifugation through ZEBA desalting columns (Pierce), and final protein concentration was determined photometrically. Streptavidincoated 96-well plates (Nunc) were washed with NaP i buffer (10 mM sodium phosphate buffer, mono-and dibasic, 50 mM NaCl, pH 7.0) and coated with 100 l of a 500 nM biotinylated heparan sulfate (HS) solution. This step was performed at room temperature under moderate shaking (incubation time ϭ 1 h). The plates were washed three times with NaP i buffer. In each well, 100 l of fluorescently labeled wtMCP-1 (500 nM) was placed and incubated for 30 min under moderate shaking. The plates were washed again, and finally 100 l of the respective competitor at a concentration ranging from 1 nM to 50 M was added to the wells in an interval of 20 s to maintain an incubation time of 5 min in every well. After incubation, 95 l were taken off and applied to untreated black 96-well plates (Greiner), and fluorescence (excitation 485 nm, emission 535 nm, integration time 1 s) was measured with a Beckman Coulter DTX800 instrument.
Cells-The THP-1 monocytic cell line was purchased from the European collection of cell cultures and grown in RPMI 1640 medium (Sigma) containing 10% fetal calf serum, 2 mM glutamine, 100 units/ml penicillin, and 100 units/ml streptomycin at 37°C and 5% CO 2 .
Primary monocytes were obtained from whole blood by collection with a Vacuette system (Greiner) containing K 2 EDTA diluted (1:1, v/v) with HBSS(Ϫ) (PAA), followed by centrifugation through Ficoll Paque Premium (GE Healthcare). The interface between plasma and Ficoll (containing mononuclear cells) was taken carefully and washed twice with HBSS(Ϫ) to remove platelets. Then the pellet was resuspended in HBSS(ϩ) to a concentration of 2 ϫ 10 6 cells/ml (calculated to the quota of monocytes). The cells were used immediately after preparation.
Chemotaxis Assay-The ability of wtMCP-1/CCL2 and mutants to mediate the migration of freshly prepared human blood-derived monocytes was investigated using a 48-well Boyden chamber system (Neuroprobe) equipped with 5-m polycarbonate polyvinylpyrrolidone-free membranes. Wild type MCP-1/CCL2 and MCP-1 mutant dilutions ranging from 5 to 20 or 40 nM in PBS were placed in the lower wells of the chamber in triplicates, including wells with buffer alone. 50 l of mononuclear cell suspension in HBSS(ϩ) at 2 ϫ 10 6 cells/ml was placed in the upper wells. After a 1-h incubation period at 37°C and 5% CO 2 , the upper surface of the membrane was washed with PBS. The migrated cells were fixed in methanol and stained with Hemacolor (Merck). The migrated cells of five ϫ400 magnifications/well were counted, and the mean of three independently conducted experiments was backgroundcorrected and plotted against the chemokine concentration.
Calcium Mobilization Assay-THP-1 cells were loaded with Fura in Krebs-Ringer buffer containing 1 mg/ml bovine serum albumin and 3 M Fura-2-acetoxymethylester for 30 min at 37°C, washed with PBS, and kept on ice. Just before use, the cells were resuspended at 2 ϫ 10 6 cells/ml in Krebs-Ringer buffer at 37°C and equilibrated for 10 min. The fluorescence emission at 495 nm was monitored upon excitation at 340 nm at 37°C under low stirring. Calcium mobilization was induced by the addition of 20 nM chemokine. The minimum fluorescence was determined after the addition of 1.8 mM EGTA, 14 mM Tris-HCl, and 0.1% Triton X-100, and the maximum fluorescence was determined after the subsequent addition of 10 mM HCl and 5 mM CaCl 2 .
Induction of EAU-Uveitis was induced in Lewis rats, 6 -8 weeks of age (Janvier, Le-Genest, St-Isle, France). All animal experiments were approved by the Review Board of the Government of Oberbayern. Treatment of animals confirmed to the Association for Research in Vision and Ophthalmology Statement on the use of Animals in Ophthalmic and Vision Research. Antigen peptides PDSAg and R14 (Biotrend, Cologne, Germany) correspond to the sequence of bovine S-Ag (amino acids 342-355, FLGELTSSEVATEV) and human IRBP (amino acids 1169 -1191, PTARSVGAADGSS-WEGVGVVPDV), respectively. Rats were subcutaneously immunized with 15 g of peptide in CFA as described (46). MCP-1 variant Met-MCP-1(Y13A/S21K/Q23R) (100 g/rat) was injected intraperitoneally daily, from immunization until indicated. The time course of the disease was determined by daily examination with an ophthalmoscope. Uveitis was graded clinically as described by de Smet et al. (53). Histological grading of cryosections from rat eyes was performed as previously described (46).

RESULTS
Engineering of MCP-1/CCL2 Mutants-Four novel MCP-1/ CCL2 mutants have been rationally engineered to identify an effective, novel strategy to antagonize GAG-related inflammatory responses. The GAG binding domain of MCP-1/CCL2 was recently mapped and was found to consist of arginine 18 and 24; lysine 19, 49, and 58; and histidine 66 as key residues for the chemokine-GAG interactions (31). A tetrameric structure has been proposed as the fundamental quaternary structure of MCP-1/CCL2 in the heparin-bound state. The GAG binding hot spots create a large positively charged surface that encircles the tetramer and, in principle, can interact with an extended GAG chain (31). The receptor binding and activation site has been described and shows tyrosine 13 at the N terminus as a key signaling residue (31, 41, 48 -49).
The strategy applied here aimed at increasing the GAG binding affinity while knocking out receptor activation. Our approach takes advantage of the significant overlap between the receptor and the GAG binding domains of MCP-1, thereby avoiding deletions or excessive mutations, which could lead to a total structural collapse and to higher immunogenicity. To increase GAG binding affinity, residues with solvent-exposed areas Ͼ30% and proximal to the mapped GAG binding site have been carefully selected and replaced with basic amino acids (see Fig. 1). The N-terminal methionine residue, retained by the protein due to E. coli expression, was not cleaved after bacterial expression because it did not compromise the binding affinity of the proteins to HS, as demonstrated by isothermal titration experiments 5 but rather increased the apparent binding affinity to heparin as has been shown elsewhere (31). Moreover, the N-terminal methionine was found to reduce the binding affinity of MCP-1/CCL2 for CCR2 on THP-1 cells (49) and the chemotactic potency of the chemokine, which is ϳ300-fold lower than for the wild type (48). In addition, in order to fully knock out receptor activation of MCP-1, we have mutated the tyrosine residue at position 13 to alanine.
GAG Binding Characterization-HS together with chondroitin sulfate (CS) are the most frequently observed GAG molecules involved in physiological and pathological processes taking place on the cell surface and in the extracellular matrix (50). Moreover, HS exhibits a much higher structural heterogeneity compared with the well known heparin and is thus expected to show higher sequence-related selectivity for specific protein ligands. Recently, it has been shown that MCP-1/CCL2 binds to a heparin octasaccharide/11SO 3 and heparin octasaccharide/ 12SO 3 (51). However, the precise structure of the natural GAG is not yet known, although the above mentioned positions of sulfation seem to be prerequisites for binding. We have used commercially available unfractionated HS to increase the probability of having the MCP-1-specific GAG sequence contained in our ligand preparation. Taking advantage of the tryptophan residue at position 59 of MCP-1/CCL2, an intrinsic fluorophore that is sensitive to conformational changes, the MCP-1-GAG interactions have been quantified by IFT. Changes of the (normalized) tryptophan fluorescence emission (⌬F/F 0 ) upon the addition of HS were plotted against increasing ligand concentrations, and from the resulting binding isotherms, the apparent dissociation constants (K d values) were calculated based on a bimolecular interaction. The K d values found by these exper-iments are summarized in Fig. 2a. Replacement of Ser 21 and Gln 23 by lysine and/or arginine residues, respectively, successfully increased the apparent affinity to HS, whereas the additional substitution of Val 47 by lysine did not further enhance the affinity.
To investigate the general GAG specificity of our MCP-1/CCL2 mutants, we performed, in addition to the HS binding experiments, IFT experiments with dermatan sulfate (DS) as ligand. DS (or CS-B) belongs to the chondroitin sulfate GAG family in which the GlcA residues are epimerized into IdoA. Interestingly, the affinities detected for all MCP-1/CCL2 proteins were in general lower for DS compared with HS (see Fig. 2c); just the Met-MCP-1(Y13A/S21K/V47K) mutant displayed similar affinities for both GAG ligands. The relatively highest affinity for both GAG ligands was found for the Met-MCP-1(Y13A/ S21K/Q23R) mutant.
Those mutants carrying basic amino acids only at positions 21 and 23 were further investigated by surface plasmon resonance (SPR) analysis with respect to their HS binding properties. Scatchard plots and SPR sensorgrams of Met-MCP-1(Y13A/ S21K) and Met-MCP-1(Y13A/ S21K/Q23R) in comparison with wtMCP-1 are shown in Fig. 3, a and b. Both methods applied, IFT and SPR, identified Met-MCP-1(Y13A/ S21K/Q23R) as the variant with highest HS affinity compared with the wild type chemokine (Figs. 2b  and 3, a and b). In fact, the K d values of Met-MCP-1(Y13A/S21K/Q23R) obtained by IFT and SPR analysis are extremely similar, 156 and 152 nM, respectively. Our rational protein design approach thus led to an increased GAG binding affinity by a factor of 5, which was confirmed by two independent methods.
MCP-1/CCL2 Displacement Experiments-Because the natural situation in which our MCP-1/CCL2based mutants will encounter their GAG ligand will be a GAG chain occupied by the wild type chemokine, we performed chemokine dis-placement experiments using fluorescently labeled wild type MCP-1/CCL2 as protein prebound to HS and the MCP-1/ CCL2 mutants as competitors (see Fig. 4). The for the other MCP-1/CCL2 mutants. Similarly to the affinity measurements, the competition experiments revealed that the mutant with the most site-directed changes did not turn out to have the best displacement characteristics and that the mutant with only one additional basic amino acid (Met-MCP-1(Y13A/ S21K)) exhibited very strong displacement characteristics.
Calcium Mobilization Assay and Chemotaxis Assay-The knock-out of the G protein-coupled receptor domain and the consequent abrogation of downstream signaling via CCR2 has been verified by investigating the ability of the MCP-1/CCL2 variants to induce calcium mobilization from THP-1 cells and to recruit primary monocytic cells in a standard chemotaxis assay in comparison with wtMCP-1/CCL2 (primary monocytes could not be used in the calcium mobilization assays because of the insufficient number of CCR2(ϩ) cells derived from whole  red in c, e, g, i). The models, based on the MCP-1/CCL2 crystal structure (Protein Data Bank entry 1DOK), were created using the software WebLabViewerPro (Molecular Simulations Inc.).
The effect of our engineered mutants on wtMCP-1/CCL2induced chemotaxis was investigated in order to estimate the potential of mutant binding to CCR2. In Fig. 7, we have shown the chemotaxis of human blood-derived monocytes by a constant concentration of MCP-1/CCL2 (20 nM) in the presence of increasing concentrations of Met-MCP-1(Y13A/S21K/Q23R). Interestingly, the mutant inhibited monocyte chemotaxis significantly at all concentrations investigated here, implying a residual ability of the engineered protein to bind to CCR2. This has been independently confirmed in a radioligand CCR2 displacement assay. 4 Due to the insufficient number of CCR2(ϩ) cells derived from human blood, we were not able to study the inhibitory effect of Met-MCP-1(Y13A/S21K/Q23R) on MCP-1/CCL2-induced calcium mobilization.  Effect of Met-MCP-1(Y13A/S21K/Q23R) on the Induction of Uveitis and Oral Tolerance-Uveitis is an inflammatory autoimmune disease of the immune-privileged inner eye and one of the major causes of blindness in industrialized countries. EAU in Lewis rats is mediated by CD4 ϩ T cells with specificity for retinal antigens, which secrete cytokines and chemokines to recruit other leukocytes that constitute the inflammatory infiltrates in uveitis. Here T cell-mediated inflammation of the inner eye has been induced by active immunization with a peptide of the retinal soluble antigen (PDSAg), which is also thought to be an autoantigen in the human disease (46,47). Inflammation and subsequent tissue damage are induced by autoreactive Th1 lymphocytes and propagated by recruited mononuclear cells, including monocytes and macrophages. Particularly, the chemokines MCP-1 and RANTES are thought to play a key role in the recruitment of effector cells in uveitis (12)(13)(14).
Intraperitoneal injection of Met-MCP-1(Y13A/S21K/Q23R) from the day of immunization until the end of the experiment on day 22 had a mild ameliorating effect on PDSAg-induced uveitis (Fig. 8, A and B). In contrast, uveitis after immunization with peptide R14 was slightly deteriorated by treatment with Met-MCP-1(Y13A/S21K/Q23R) (data not shown). This observation is in concordance with our previous findings described for treatment with Met-RANTES (53). The effect of Met-MCP-1(Y13A/S21K/Q23R) on oral tolerance was tested by daily oral application of the Met-MCP-1(Y13A/S21K/Q23R) variant during the period of oral tolerization with peptide PDSAg and prior to immunization with PDSAg-CFA (Fig. 9, A and B). We observed only a small effect on oral tolerance in the group fed with peptide and concomitantly with Met-MCP-1(Y13A/S21K/Q23R) compared with the group that had received only oral peptide (Fig. 9A). Oral application of only Met-MCP-1(Y13A/S21K/Q23R) prior to immunization with PDSAg had a slightly ameliorating effect on the clinical course of the disease (representing the inflammation in the anterior chamber of the eye) (Fig.  9B), as observed for the injection of Met-MCP-1(Y13A/S21K/Q23R) after immunization with PDSAg (Fig. 8B).

DISCUSSION
Inflammation is a complex response to injury and infection and represents a protective attempt by the organism to remove the harmful trigger and initiate the healing process (52). However, impaired regulation of inflammation can lead to  extensive tissue damage, the pathological consequences of which include a variety of autoimmune diseases, such as rheumatoid arthritis and multiple sclerosis, but also non-immune diseases, such as atherosclerosis, ischemic heart disease, and cancer. Leukocytes are critically involved in the initiation and maintenance of inflammation by secreting cytokines and chemokines that enhance cell recruitment and activation. MCP-1/ CCL2 is highly induced in many inflammatory conditions and is responsible for recruiting mononuclear cells to the site of injury by signaling through CCR2. In fact, MCP-1/CCL2 and CCR2 null mice, which are unable to recruit monocytes and T cells to inflammatory lesions, show less lipid deposition and macrophage accumulation throughout their aortas in an atherosclerosis model (54), are resistant to experimental autoimmune encephalomyelitis (16), and are strongly protected against endotoxin-induced uveitis (13).
Given the importance of MCP-1/CCL2 in human disease, various therapeutic strategies have been developed that target the activity of this CC chemokine in vivo at several levels. Many attempts by several laboratories have been made to antagonize the chemokine oligomerization ability as well as the chemokine-receptor and GAG co-receptor interactions. These approaches include an obligate MCP-1/CCL2 monomer (P8A-CCL2) with impaired oligomerization ability (30), neutralizing anti-CCL-2 and anti-CCR2 monoclonal antibodies (24,25), modified MCP-1/CCL2 receptor antagonists, low molecular weight inhibitors of CCR2 (26 -28, 36 -38), and GAG binding-deficient mutants (29,31). Nevertheless, clinical trials conducted thus far with MCP-1/CCL2 antagonists have been unsuccessful due to limitations such as species specificity of small antagonists and neutralizing mAb, low oral bioavailability of some compounds, and the time-and interventiondependent effect of chemokine or chemokine receptor blockade (36 -38).
With this in mind and prompted by the unmet need for a treatment effective in acute and transient inflammatory responses as well as in chronic inflammatory diseases, we targeted the GAG binding affinity of MCP-1/CCL2 and its receptor activation ability to rationally convert the proinflammatory chemokine MCP-1/CCL2 into a protein with anti-inflammatory activity. In this work, taking into account the previously mapped receptor activation and GAG-binding sites on MCP-1/ CCL2, we retained the N-terminal methionine after recombinant protein synthesis in E. coli, and we mutated the tyrosine 13 to alanine to impair receptor activation and signaling. More importantly, novel mutagenesis studies were performed to identify residues that, if replaced by the basic amino acids lysine and/or arginine significantly increased the GAG binding affinity. The rationale behind this strategy was to create MCP-1/CCL2 antagonists that compete with endogenous MCP-1/CCL2 for GAG binding while being unable to activate its cognate GPC receptor, which is highly expressed on monocytes during inflammatory conditions.
The domain structure of MCP-1/CCL2 is characterized by a partial overlap between the receptor activation site and the GAG-binding domain, which means that modifications of one site have mutual effects on the other. Four mutants have therefore been generated that have in common (a) a methionine residue at the N terminus that has previously been shown to reduce the MCP-1/CCL2 chemotactic potency (48) and to increase the apparent GAG binding affinity (31) and (b) an alanine residue at position 13 to further impair receptor activation. Positions 21, 23, and 47 showed solvent-exposed areas above 30% and were found to be located in close proximity to A, clinical and histological uveitis scores after oral tolerance induction. Groups of rats were orally pretreated as indicated before uveitis was induced by immunization with PDSAg in CFA. PBS or 100 g of mMCP-1 was fed daily from day Ϫ7 to Ϫ1 prior to immunization. PDSAg (200 g) was fed at days Ϫ7, Ϫ5, and Ϫ3. Columns show mean maximal uveitis scores Ϯ S.E. of all eyes (n ϭ 8). B, time course of clinical uveitis after oral tolerance induction. PBS or 100 g of Met-MCP-1(Y13A/S21K/Q23R) was fed daily from day Ϫ7 to Ϫ1 prior immunization. PDSAg (200 g) was fed at days Ϫ7, Ϫ5, and Ϫ3. Mean clinical uveitis scores from all eyes (n ϭ 8) per group Ϯ S.E. are shown for each day of clinical observation.

Evolving MCP-1/CCL2 into an Anti-inflammatory Decoy Protein
the mapped GAG binding site. These sites have been mutated either individually or in combination with lysine or arginine. The mutations S21K and Q23R, individually as well as in combination, increased the GAG binding affinity significantly and contributed positively to abrogate receptor activation. A relatively similar but absolutely different affinity pattern for the mutants was found when the two major GAG ligands, HS and CS, were compared (see Fig. 2), revealing CS as a low affinity ligand of MCP-1/CCL2. In fluorescence-based MCP-1/CCL2 displacement experiments, only the mutant Met-MCP-1 (Y13A/S21K/Q23R/V47K) showed almost no improvement in its IC 50 value with respect to displacing wtMCP1/CCL2 from the HS ligand.
Therefore, based on its overall molecular characteristics, the mutant Met-MCP-1(Y13A/S21K/Q23R) was identified as our lead compound based on its strongly impaired GPC receptor activity, its 87% disabled directional migration potency toward primary monocyte cells in the Boyden chamber assay, its completely knocked out intracellular calcium mobilization activity, and its highest GAG binding affinity. Isothermal fluorescence titrations and surface plasmon resonance experiments performed with the natural HS ligand gave for this mutant K d values of 156 and 152 nM, respectively.
The potential antagonistic activity of Met-MCP-1(Y13A/ S21K/Q23R) has been proven in a rat model of EAU. The mutant, administered from the time point of immunization during the entire course of the disease, decreased significantly the disease score in comparison with the animals treated only with PBS (see Fig. 8, A and B), thus protecting the animals from the development of severe inner eye inflammation. Overall, the animals treated with Met-MCP-1(Y13A/S21K/Q23R) showed a milder and delayed inflammatory response compared with PBS-treated animals during the 22-day kinetic of disease.
In summary, we propose here a novel strategy to antagonize MCP-1/CCL2-mediated inflammatory response characterized by an excessive monocyte infiltration. Competition with wtMCP-1 for the cell surface GAGs is proposed to be the mechanism of action of Met-MCP-1(Y13A/S21K/Q23R). Assuming an up-regulation of GAG co-receptors only at sites of inflammation and a selective interaction between MCP-1/CCL2 and its cognate GAG ligand, our approach may have significant advantages over therapeutic strategies applied thus far.