Interfering with the CCL2 – glycosaminoglycan axis as a potential approach to modulate neuroinflammation

Multiple Sclerosis, a chronic inflammatory demyelinating disease of the central nervous system, has been related with involves increased expression of monocyte chemotactic protein 1 MCP1-/CCL2. For exerting its chemotactic effects, chemokine binding to glycosaminoglycans (GAGs) is required and therefore this interaction represents a potential target for therapeutic intervention. Intending to engineer pro-inflammatory CCL2 towards an anti-inflammatory compound, we have designed a decoy variant, Met-CCL2 (Y13A S21K Q23R), embodying We have designed an anti-inflammatory decoy variant, Met-CCL2 (Y13A S21K Q23R), embodying increased affinity for GAGs as well as knocked-out GPCR activation properties. This non-signalling dominant-negative mutant is shown here to be able to displace wild type CCL2 from GAGs by which it is supposed to interfere with the chemokine-related inflammatory response. In vivo , the anti-inflammatory properties were successfully demonstrated in a murine model of zymosan-induced peritonitis as well as in an experimental autoimmune encephalomyelitis, a model relevant for multiple sclerosis, where the compound lead to significantly reduced clinical scores due to reduction of cellular infiltrates and demyelination in spinal cord and cerebellum. These findings indicate a promising potential for future therapeutic development.


Introduction
Chemokines are small secreted proteins that direct leukocyte trafficking from the lumen of blood vessels into the inflamed surrounding tissue [1]. Bound to glycosaminoglycans (GAGs, see below) on the endothelium, chemokines are retained at the site of action and are presented in their active conformation towards the attracted leukocytes [2]. Once activated via chemokine-specific GPCRs, the target blood cells migrate through the endothelial blood vessel cell layer to enter the site of tissue inflammation and to further induce and to boost the host immune response [3].
Monocyte chemoattractant protein-1 (MCP-1/CCL2) is a member of the chemokine-β family (CC chemokines), specifically activates monocytes and lymphocytes and is found in a variety of diseases that feature a monocyte-rich inflammatory component, such as atherosclerosis [4], rheumatoid arthritis [5] and congestive heart failure [6]. Most notably, there is also evidence that CCL2 plays a crucial role in the disease pathogenesis of Multiple Sclerosis (MS) [7,8].
MS is an inflammatory disease of the central nervous system (CNS). It is caused by infiltrating leukocytes damaging myelin and axons, ultimately leading to extensive and chronic neurodegeneration [9,10]. In this context, CCL2 is described to be responsible for CCR2-bearing leukocyte infiltration into the MS lesions of the CNS, but also for T-cell and monocyte migration within the CNS parenchyma. Moreover, CCL2 expression is altered depending on disease activity [11]. Particularly important is the reported observation that in active demyelinating as well as in chronic active MS lesions, reactive hypertrophic astrocytes are strongly immunoreactive for CCL2, suggesting a significant role for CCL2 in the recruitment and activation of myelin-degrading macrophages and thereby contributing to the evolution of MS [12]. Perivascular and parenchymal foamy macrophages do not express CCL2 protein and are likely to contribute to resolution of inflammation by inhibiting further lesion development and promoting lesion repair [13,14]. On this line of evidence it has recently been reported that in a murine model of experimental autoimmune encephalomyelitis, mice with conditional astroglial ablation of CCL2 showed a reduced severity of pathology due to less spinal cord axonal loss [15] (Moreno et al 2014).
Additionally, CCL2 significantly increases the permeability of the blood-brain barrier in vivo and thereby facilitates leukocyte migration during inflammation [16].
As mentioned above, chemokine activity in vivo was shown to be dependent on the interaction with glycosaminoglycans (GAGs) [2]. This means, that only the triple complex consisting of chemokine, chemokine-specific GPC receptor, and GAG co-receptor gives the fully functional entity for chemotaxis in vivo. These particular glycans are linear, negatively charged polysaccharides which consist of repeating disaccharide units of an amino sugar and a uronic acid [17]. Most of them are covalently attached to a core protein to form the so-called proteoglycans localized in the cell membrane or in the extracellular matrix [18]. The physiologically most relevant GAG types on cell surfaces and in the extracellular matrix are heparan sulfate and chondroitin sulfate which differ from one another in the disaccharide unit composition, the chain length and the degree of sulfation [19]. In fact, post-polymerization modifications, which include N-, 2-O, and 6-O sulfation as well as epimerization of GlcA into IdoA, provide high structural complexity within GAGs and allow the generation of particular oligosaccharide sequences that are supposed to be specific for their protein ligands, as it has been shown for antithrombin [20] and for basic fibroblast growth factor [21]. The sulfation patterns are cell type and tissue specific and are tightly regulated developmentally and pathophysiologically [22][23][24]. In the CNS GAGs/ proteoglycans can be found as ECM components and in the basement membrane [25,26]. Although specificity/ selectivity in the chemokine/GAG interaction has not been fully elucidated so far, the great variety in GAG structures on the one side and the reported different affinities for GAGs displayed towards chemokines on the other side, suggest that chemokines interact with disease specific GAG structures in vivo [27,28].
Applying a rational design approach based the CellJammer ® Technology [29] human proinflammatory CCL2 was turned into an anti-inflammatory chemokine decoy CCL2 was altered to an anti-inflammatory chemokine decoy protein with increased glycan binding affinity, as previously reported by Piccinini et al. [30] and was shown to limit neointima formation and myocardial ischemia/ reperfusion injury in mice [31].
In this letter we describe the investigation of the anti-inflammatory activity of Met-CCL2 (Y13A S21K Q23R) in animals using the mechanistic inflammatory model of zymosaninduced peritonitis as well as a disease relevant model of Multiple Sclerosis, the experimental autoimmune encephalomyelitis (EAE).

Design, Expression and Purification
The CCL2 decoy Met-CCL2 (Y13A S21K Q23R) was designed on the basis of human mature CCL2 (1-76) (Swiss-Prot # P13500) by introducing 3 mutations in order to abolish receptor binding (Y13A) as well as to increase the affinity for its glycan ligand with additional positive amino acids (S21K Q23R) on both sides of the known GAG binding motif.
Furthermore, Met-CCL2 constructs were generated in the background of a M64I mutation, which alters neither binding nor activity, however improves homogeneity of the mutants by eliminating the possibility of methionine sulfoxide species on position 64 [32]. The leading Nterminal Met residue is a result of the recombinant protein expression in E. coli. It was not cleaved off after expression since it did not compromise the heparan sulfate binding properties of the protein [30].
Met-CCL2 and Met-CCL2 (Y13A S21K Q23R) were expressed and purified as described previously [30]. During the purification process, good lab standard practices were applied to reduce endotoxin contents of protein samples by working with endotoxin-free plastic ware and by rinsing glass ware and chromatographic equipment with concentrated NaOH.

Animal experiments
Animal care and handling procedures were performed in accordance with the European guidelines and all the experiments received prior approval from the local ethics committees. and data from one experiment are presented.

MOG-Induced Experimental Autoimmune Encephalomyelitis (EAE)
Ten-week old, female C57BL/6 mice (Charles River, NL,) were group housed under a 12:00 h light /dark cycle, provided with food and water ad libitum and allowed to acclimatize for at

Statistical analysis
All data are reported as means + standard errors of the means (SEM). Pharmacokinetic profiles of Met-CCL2 (Y13A S21K Q23R) were evaluated with WinNonLin using noncompartment model. For the zymosan-induced peritonitis, statistical analysis was performed using ANOVA followed by Dunnett's multiple comparison using GraphPad Prism software.
The EAE data were analyzed using the Kruskal-Wallis test followed by the Mann-Whitney Utest. The significance of differences between the treatment groups in the time-dependent outcome parameters (daily clinical score and body weight) were tested using repeatedmeasures ANOVA, followed by LSD post hoc. Analysis was performed using SPSS17 for Windows. In both cases, the significance level was set at p<0.05. Significance is reported as follows: * p<0.05; ** p<0.01; *** p<0.01.

ELISA-Like Competition Assay
Since our mutant proteins were designed in a way to displace the corresponding wild type chemokine from its GAG co-receptor, we have developed a novel competition assay which gives IC50 values derived from displacement curves rather than Kd values obtained from bimolecular binding isotherms [33]. In our assay we have tried to mimic the glycocalyx of cell surfaces by coating heparan sulfate onto specially prepared microtiter plates. We then added

Zymosan-Induced Peritonitis
In this study Met-CCL2 (Y13A S21K Q23R) at a concentration of 40 µg/kg i.v. was directly compared to the immunosuppressant dexamethasone at a dose of 1 mg/kg s.c. While the latter reduced the total cell infiltrate (data not shown), Met-CCL2 (Y13A S21K Q23R) effects were specifically observed on a particular subset of monocytes expressing F4/80 as well as Gr1 (Ly6C and Ly6G) marker, and considered to be a pro-inflammatory subset of monocytes ( Figure 3A R5 and 3C) , while not affecting the number of residential macrophages (F4/80 high /Gr1 -ve , Figure 3A R3-4 and 3B), that are thought to contribute to the pathology resolution [34].

Discussion
Met-CCL2 (Y13A S21K Q23R) shows higher GAG binding affinity compared to Met-CCL2 but shows impaired CCR2 receptor activation [30]. Here we show that, in addition, the mutant is able to displace the wild type chemokine efficiently from HS chains (see Figure 1).
Since it is proposed that CCL2 in vivo is mainly displayed to approaching monocytes/macrophages in a GAG-bound form, the competitive potency (expressed in the IC50 value) of a CCL2 mutant is much more relevant parameter than the direct binding of chemokine to its GAG ligand. This relates directly to the proposed mode of action of the CCL2 mutant, namely that it acts like a protein-based GAG antagonist.
The proposed anti-inflammatory effect of the mutant was subsequently tested in vivo in a mechanistic model of zymosan-induced peritonitis [35] to allow initial assessment of the dose in vivo and anti-inflammatory/anti-migratory properties, and only then in an animal model of multiple sclerosis, EAE. the experimental autoimmune encephalomyelitis (EAE).
In the zymosan-induced peritonitis model we could show that we are able to retain selectivity for the specific cell populations in vivo, thereby selectively inhibiting the recruitment of Gr1 and F4/80 double positive cells, considered to be newly recruited inflammatory monocytes, while not affecting macrophagesthe macrophage number. Moreover, this finding implies that the mutant has found its appropriate endothelial GAG target and gained the energetically preferred active form.
MOG-induced EAE in mice is recognized to be a good experimental model of human MS pathogenesis [7,36,37], and in the present study MOG   The observed effect could not be due to inhibition of bone marrow cell mobilization, as has been reported for inhibitors of the GPC receptor CCR2 [38,39], since we have already shown that Met-CCL2 (Y13A S12K Q23R) does not bind or activate, CCR2 [30]. Surprisingly, Met-CCL2 (Y13A S12K Q23R) activity on the different parameters measured was not dosedependent, with the lowest dose administered being the most effective. This is not due to a non-linear exposure, since the PK profile for the protein seems to be linear in this range (at least in term of AUC). The binding affinities of chemokines to GAGs are known to not be dose dependentbe not-dose dependent [40] with loss of affinity at too high concentrations, therefore, even if it seems unlikely from the plasma concentration measured during the PK assessment, we cannot exclude that at the two higher doses the binding equilibrium at the GAG target was moving toward dissociation. Alternatively, it is possible that the apparent differences observed between doses are due to the limited number of animals used in the experiment, where in which even the survival of one more animal in a treatment group may affect the overall clinical score profile, suggesting that plateau activity is already achieved at a dose of 40µg/kg. This aspect should be further analyzed using a broader range of doses and in a time-course experiment, with direct assessment of cell type infiltrates in the CNS of the animals by either immunohistochemistry, or FACs analysis.
However, the clinical development of a protein requesting daily administration is not foreseeable (and already difficult to assess preclinically, since twice a day chronic treatment is not feasible in mice), and strategies to extend its exposure, such as by conjugation with carrier proteins (e.g. human serum albumin) are currently under evaluation [41] and will be further tested in MS animal models.
What we can state from the current data is that the dose of 40 µg/kg Met-CCL2 (Y13A S12K Q23R) was, in both the experimental peritonitis and in the EAE model, at least as effective as dexamethasone in reducing inflammation and consequent demyelination. It has to be considered that dexamethasone was administered in this study at a dose highly effective in mice, but toxic if translated to a human dose.
In conclusion, we have confirmed that the engineered increased GAG binding affinity of Met-CCL2 (Y13A S21K Q23R), which was measured in vitro, can be translated into antiinflammatory activities in vivo. Our approach of targeting glycans for interfering with CCL2 signalling in CNS inflammation is completely unique, and opens a possible new avenue for therapeutic intervention that may be applicable to MS treatment.