Structural Insight into the Function of Myelin Basic Protein as a Ligand for Integrin α_Mβ₂¹

Integrins are cell surface receptors that mediate cell-cell, cell-extracellular matrix, and cell-pathogen adhesion (1). Besides their key roles in development and homeostasis, integrins are crucial for the trafficking and functions of leukocytes (2). Integrins are composed of one α-chain and one β-chain; the family of integrins with a common β₂ chain includes α_Lβ₂ (also known as LFA-1 or CD11a/CD18), α_Mβ₂ (Mac-1, CD11b/CD18), and α_Xβ₂ (p150,95, CD11c/CD18), and these receptors support adhesion of leukocytes to target surfaces as well as migration into zones of inflammation. Expressed on phagocytes, the α_Mβ₂ receptor serves in the uptake of microbes and debris from apoptotic host cells through binding to proteolyzed complement factor C3 (iC3b) deposited on target surfaces (3). The binding of α_Mβ₂ to endogenous ligands as well as several microbial molecules points to a ligand selectivity typical of scavenger receptors (4). Nevertheless, the significance of α_Mβ₂ as a scavenger receptor is not established although expression of this receptor on the cell surface of all human myeloid leukocytes suggests an important role for α_Mβ₂ in a wide range of inflammatory responses.

Numerous physiological and pathophysiological processes leading to the change or loss of protein structure suggest that mechanisms for responding to protein unfolding play an important role in the body, which is highlighted by rapidly growing evidence that so-called intrinsically unfolded proteins are important disease factors in neurological disorders (5). Loss of protein-ligand structure increases their binding to α_Mβ₂ and α_Xβ₂ (6, 7, 8, 9), however, no link has been made between disease and the predilection of these receptors to bind unfolded proteins.

Myelin basic protein (MBP)³ is the major protein component of the lipid-rich myelin sheath surrounding nerve fibers (5) and stabilize the sheath structure via binding to the cytoplasmic surfaces of cell membrane. A model of the MBP three-dimensional structure showed that MBP assumes a conformation similar to an open C with a central β-sheeted region surrounded by large loop regions rich in charged amino acid residues, which make contact with the cell membrane (10). The model was supported by small angle x-ray scattering (SAXS) studies of MBP in the presence of lipids and detergents (11). By contrast, in water-rich (aqueous) media MBP has a poorly defined structure as revealed by circular dichroism (CD) spectroscopy (12) and SAXS (11). Blood or cerebrospinal fluid constitute aqueous environments, thus, the loss of myelin sheath integrity would expose MBP to aqueous media and is likely to change the MBP conformation. MBP is posttranslationally modified by deimination of arginyl residues to citrulin, which leads to a reduction in net charge of the molecule compared with unmodified MBP. Because the positively charged residues are critical in stabilizing cell membranes in the myelin sheath changes in deimination could be a part of multiple sclerosis (MS) pathogenesis as supported by the observation that nearly 50% of the MBP molecules from MS patients are deiminated compared with 20% from normal individuals (13, 14). MS is a disease that affects the CNS by an inflammatory “strike” against components of the myelin sheath causing large focal lesions of demyelinated nerve fibers. There is substantial evidence that MBP is a major autoantigen in MS with a direct link between increased disease activity and increased activity of T lymphocytes reacting with MBP (15). Therefore, it is of considerable interest that studies have suggested a link between the antigenicity and deimination of MBP (16), which apparently is affected by the reduced folding of deiminated molecules. However, the processes that regulate the immune response to unfolded protein species are not well understood.

Injection of purified MBP in several animal models can induce experimental autoimmune encephalitis (EAE) leading to histopathological and symptomatic similarities with MS in humans (17). Invading monocytes (18) and microglia cells (19) were demonstrated to act as terminal agents of the myelin sheath destruction in MS. During both MS and EAE the myelin lamellae become attached to coated pits on macrophage surface and are ingested by receptor-mediated phagocytosis (20, 21). Abs specific to α_Mβ₂ effectively block the phagocytosis of myelin particles by macrophages and microglia in vitro (22), and data from in vivo models suggest a key role of α_Mβ₂ in EAE (23). However, the important question what triggers the ability of phagocytic cells to recognize the myelin and the role of α_Mβ₂ in this process remains poorly understood.

In this study, we demonstrate that MBP is a potent and specific ligand for the α_Mβ₂ integrin. By studies of the MBP structure in solution, we identify the regions of MBP that undergo conformational change when the molecule is liberated from a lipid-rich environment. Glatiramer acetate (GA, also known as Copaxone) is a mixture of random copolymers of alanine, tyrosine, lysine, and glutamate, averaging 50 amino acid residues in length and designed to act as MBP mimic or decoy (24, 25). GA is one of the leading therapies for MS treatment (26), however, the detailed pharmaceutical mechanism of this drug remains controversial (26, 27). We show that the GA peptides are mimics of conformationally labile regions of MBP and strong antagonists of α_Mβ₂ binding to MBP. This is the first study linking the capability of integrins to bind a member of the group of intrinsically unstructured proteins with development of autoimmunity and disease.

Bovine MBP was purchased as 5 mg/ml stock solution in 10 mM MOPS (pH 7.0) (catalog no. 13-104; Upstate Biotechnology). Glatiramer acetate (804654B; Copaxone, Teva Pharmaceuticals) was obtained from the local pharmacy as 20 mg/ml solution stabilized with 40 mg/ml mannitol.

The KIM127 mouse mAb (28) reacting with an integrin activation-specific epitope in the β₂ chain (CD18) was purified on a protein G column (29) from the supernatant of a hybridoma culture with cells bought from the American Type Culture Collection (LCG-Promochem AB). The ICRF 44 mouse mAb reacting with the α_M (CD11b) chain was bought from Sigma-Aldrich (C0551). An IgG1 Ab (X0931) from DakoCytomation was used as istotypic control Ab.

The α_Mβ₂, α_Lβ₂ and α_Xβ₂-expressing K562 cell lines and the untransfected parental K562 cell lines cells have been described earlier (30, 31) and were a gift from Dr. T. A. Springer (Harvard Medical School, Boston, MA).

Human mononuclear cells (MNCs) were prepared from the buffy coat of centrifuged donor blood, prepared by the Blood Bank at Aarhus University Hospital Skejby. The buffy coat was depleted of erythrocytes by Ficoll-Paque PLUS (GE Health Care Bioscience AB) gradient centrifugation. MNCs were stored as frozen in culture medium (RPMI 1640 with 2 mM Gln, 10 U/ml penicillin, 10 μg/ml streptomycin, and 10 mM HEPES (pH 7.4)) supplemented with 15% (v/v) FBS and 10% DMSO (Sigma-Aldrich). Monocytes were isolated from MNCs prepared as described above by using the Dynal Monocyte Negative Isolation Kit according to the manufacturer’s protocol (113.10; Dynal Biotech ASA) and stored as frozen as described for the MNCs. In general, cell viability of the MNCs and monocytes was >90% after thawing.

Bovine MBP was diluted to a concentration of 0.5 mg/ml with PBS in a total volume of 1.0 ml, and dialyzed against PBS. The sample was mixed with 17 mM biotinyl-N-hydroxy-succinimide and incubated for 4 h at room temperature followed by dialysis against PBS. Neutravidin-coated fluorescent polystyrene particles (FluoroSpheres; Invitrogen Life Technology) were conjugated with MBP by mixing beads adjusted with PBS to a final concentration of 232 μg/ml and biotinylated MBP at final concentration of 23 μg/ml, followed by incubation for 1 h at room temperature. The particles were pelleted at 8,000 × g for 10 min and washed twice in 1 ml binding buffer (10 mM HEPES-KOH (pH 7.4); 150 mM NaCl; 5 mM KCl; 1 mM MgCl₂; 1.8 mM CaCl₂; 5 mM d-glucose) and 100 mg/ml HSA (Octalbin, Octapharma AB) followed by pelleting. The particles were resuspended and incubated in 250 μl binding buffer/HSA for 30 min before use.

MNC were gated into monocytes or lymphocytes by staining with FITC-conjugated anti-CD45 Ab and RPE-conjugated anti-CD14 Ab (MultiMix; DakoCytomation). In a parallel experiment, the human mononuclear cells were washed twice in 2 ml of binding buffer/HSA with centrifugation at 230 × g for 5 min and resuspended in 100 μl at a concentration of 0.5 × 10⁶ cells/ml. The cells were incubated either with MBP-conjugated particles or particles without MBP at a final concentration of 2.3 × 10⁷ per ml for 2.5 h at 37°C. The cells were then washed once in 2 ml binding buffer/HSA and analyzed by flow cytometry (FC 500; Beckman Coulter). The binding of beads to K562 cell lines were tested similarly to the MNCs.

The centrifugation-based cell adhesion assay was conducted essentially as described (9). Polystyrene microtiter plates with V-formed wells (3897; Costar, Corning) were coated at 37°C for 20 min in 20 mM Tris-HCl (pH 9.4); 150 mM NaCl with MBP, GA, or HSA diluted to the indicated concentrations. The plates were blocked in PBS with 0.05% (v/v) Tween 20 at 37°C for 20 min. All plastic tubes used for cell handling were preblocked using PBS with 0.05% (v/v) Tween 20. Cells were fluorescently labeled by incubation with 2′,7′-bis(2-carboxyethyl)-5(6)-carboxyfluorescein acetoxymethyl ester (BCECF · AM, 14560; Sigma-Aldrich) at 37°C for 15 min. The cells were washed twice with binding buffer, resuspended in the same buffer to 0.5 × 10⁶ cells/ml.

For cell adhesion assay with the K562 cell lines, the integrins were activated for ligand binding by addition of 1 mM MnCl₂. In the case of purified MNCs or monocytes activation of integrin ligand binding was conducted by adding 5 μg/ml KIM127 mAb (or isotypic control Ab) and incubating the cells for 30 min at 37°C before application to the adhesion assay. For testing the influence of a function-blocking mAb to α_M on the adhesion of monocytes to MBP these cells were incubated in the presence of the ICRF44 mAb at 5 μg/ml (or an isotypic control mAb at the same concentration) and the KIM127 Ab.

One hundred μl of the cells were added per microtiter well and incubated at 37°C for 20 min. To study the influence of the temperature on the binding to GA, this incubation step was also conducted at 5°C and 23°C. As a control, wells coated with iC3b were also included and the cellular binding to these surfaces at 5°C, 23°C, and 37°C were compared. The plates were centrifuged subsequently at 16 × g, 65 × g, 145 × g, 258 × g, and 717 × g for 5 min in a Beckman GPR centrifuge (Beckman Coulter) with each spin at room temperature. The fraction of binding cells was estimated from the fluorescence signals in ligand-coated wells (F_coated) and in detergent-blocked wells without ligand (F_uncoated) read in a fluorescence concentration analyzer (Victor 3; Wallac Oy) according to the following formula: Cell Binding (%) = 100% × (F_uncoated − F_coated)/F_uncoated.

Cellular binding was measured in triplicate wells in 2–5 independent experiments and stated as mean values ± SEM.

Random GA-like sequences were prepared by the random letter sequence generator (http://www.dave-reed.com/Nifty/randSeq.html), which was fed with a pool of letters corresponding to the amino acid composition of GA (42 A, 34 K, 14 E, and 10 Y); a total of 100 random sequences were generated, each with a length of 50 letters (amino acids) corresponding to the average length of the GA peptides. These GA-like sequences were aligned pair-wise with the 18.5 kDa isoform of human MBP (Swiss Prot entry P02686–5) using the DIALIGN alignment program at Bielefeld Bioinformatics Server (32) and the result presented as the alignment frequency for each amino acid in MBP is as follows: alignment score (%) = 100% × (alignment occurrences/number of alignments).

The distribution of the random GA-like sequences, the MBP, and as well as three fragments in a charge-hydrophobicity phase space was determined as described by Uversky et al. (33). In brief, by using the “ProtScale” algorithm at the Swiss Institute of Bioinformatics server (http://www.expasy.org/tools/protscale.html) for each peptide or protein, the hydrophobicity of each amino acid was calculated by the Kyte-Doolittle approximation (34) using a window size of five amino acids and normalizing values on a scale from 0 to 1. The mean hydrophobicity (<H>) was calculated by dividing the sum of the normalized hydrophobicities by the number of residues in polypeptides. The mean net charge (<R>) for each peptide or protein was calculated by dividing the total number of negatively (Asp + Glu) and positively (Arg + Lys) charged residues by the total number of residues. The values of <H> and <R> for histones and α-synuclein were taken from Uversky et al. (33). As a computational approach for predicting the presence of unstructured regions in MBP we applied the DisEMBL algorithm (35)(http://dis.embl.de/) using default parameters.

An α_M I domain mutated to favor the open, ligand binding conformation by replacing Ile-316 with glycine (36) was expressed and purified as described earlier (9). A CM-4 chip with a surface layer of carboxymethylated dextran for use with the BIAcore 3000 instrument (GE Healthcare Europe GmbH) was coupled with either GA or MBP through primary amine groups with BIAcore’s kit (BR-1000–50, GE Healthcare). The I domain was diluted in sterile-filtered and degassed running buffer containing 150 mM NaCl, 20 mM Tris-HCl (pH 7.4) with 1 mM MgCl₂ to concentrations of 2, 3, 4, 7, and 10 μM and applied to the chip surface in running buffer with a flow rate of 5 μl/min. As a control, the I domain was also applied to the chip surface at a concentration of 10 μM in the presence of 20 mM EDTA. For each run, the signal from a flow cell with a surface coupled with ethanol amine was subtracted from the sensorgrams obtained from flow cells with coupled protein ligands. Between runs, the chip was stored in dry and cold conditions.

Synchrotron radiation circular dichroism (SRCD) spectra were collected on beamline UV1 at the ASTRID storage ring (ASTRID at University of Aarhus, Denmark). MBP or GA was diluted into 20 mM sodium phosphate (pH 7.4), 100 mM NaF to a concentration of 1 mg/ml. Spectra for baseline subtraction were recorded on 2 mM MOPS for MBP or on 2 mg/ml mannitol for GA, both in 20 mM sodium phosphate (pH 7.4), 100 mM NaF, or in this buffer supplemented with 20 or 50% (v/v) trifluoroethanol (TFE). Samples were measured in 100-μm path-length Suprasil cells (Hellma GmbH). All sample and baseline spectra were collected 3 times with 1 nm step size and 3 s dwell time. For the temperature effect studies the GA spectra were collected by increasing the temperature stepwise from 5 to 85°C with 5°C increments followed by cooling of the sample to 25, 20, and 5°C. The spectra were averaged, baseline subtracted, and mildly smoothed with a Savitzky-Golay filter using the CDtool software (37). The fraction of secondary structures was calculated from these spectra using the Selcon3 algorithm and the SP175 protein CD spectra (175–260 nm) reference set (38).

Bovine MBP at concentrations of 2.5 mg/ml and 1.25 mg/ml in 10 mM MOPS, (pH 7.0) with 150 mM NaCl was used in SAXS experiments in a laboratory-based instrument at the University of Aarhus either in the presence of 20% (v/v) TFE or in plain buffer as described elsewhere (39). The data treatment was done with locally customized software (C. L. P. Oliveira and J. S. Pederson, unpublished study). The first step of data analysis was conducted with the indirect Fourier transformation method developed by Glatter (40) in a slightly different implementation (41). For the full ab initio dummy chain model a sequence with the same length of the MBP protein was modeled as dummy residues. For the modeling of the loop regions, the modeling was conducted assuming the presence of a β sheet with five strands in the central core of the molecule (10). Regions of the MBP outside the central β sheet were modeled with dummy residues. The modeling of missing loops was done by the program package BUNCH (42); other software packages used are described elsewhere (39).

By use of flow cytometry and cell adhesion assays, we investigated which subsets of human PBMCs were able to bind MBP. MNCs were incubated with fluorescently labeled beads coated with MBP and analyzed by flow cytometry. The lymphocyte and the monocyte populations were gated according to forward and side scatter. Cells were also stained with Abs to CD14 and CD45 validating the gates by confirming the CD14^neg/CD45^bright and CD14^pos/CD45^bright expression patterns characterizing lymphocytes and monocytes, respectively (data not shown).

The analyses of the MNCs binding to beads are shown in Fig. 1, A–D. The cells were analyzed without prior activation of integrin-ligand binding as this caused a high loss of monocytes in the assay. Activation of integrin binding by application of 1 mM Mn²⁺ selectively depleted the monocytes from the cells analyzed by the flow cytometer (data not shown), probably because this cell type is lost through binding to protein adsorbed on the plastic surfaces of the flow cytometer. In addition to α_Mβ₂, monocytes carry the α_Xβ₂, which binds strongly to certain denatured protein species on plastic surfaces (9); activation by Mn²⁺ would also activate this receptor and this way promote loss of monocytes in the flow cytometer. In each histogram, two gates were added. One gate (shown in the upper part of histograms) was set to include all cells shown in the histogram to quantify cells positive for binding to beads (cells negative for the fluorescence signal were excluded from histograms). The fluorescence intensities were quantified in individual peaks representing the number of beads bound per cell. The second gate (in the lower part of the histograms) was added to distinguish between cells with minimal fluorescence signal, i.e., those having bound only one bead, and cells with higher fluorescence. The monocytic subset of MNCs bound several beads with MBP per cell as judged from the multiple spikes at high fluorescence intensity (Fig. 1,B). This fraction of cells constituted 9.3 ± 2.5% (average ± SD for two independent experiments) of the monocytes (Fig. 1,B) compared with 5.6 ± 0.6% for beads without MBP (Fig. 1,A). The lymphocytic subset of MNC showed no major differences in the fluorescence profiles for beads with (Fig. 1,D) or without MBP (Fig. 1,C). We included human serum albumin (HSA) in the buffers for the flow cytometric assays. Native albumin (i.e., not denatured) is a weak ligand for α_Mβ₂ integrin (6, 7, 8) and was consequently used for blocking of nonspecific interaction between the cells and the polystyrene beads. Furthermore, the relevance of probing the binding between beads with MBP and cells in the presence of HSA was to show that this binding occurred even with this protein added, which simulates the environment in the body fluids with high concentrations of HSA. Nevertheless, it is possible that a significant fraction of the background binding observed in the flow cytometry assay (Fig. 1 A) is also contributed by binding of α_Mβ₂ to HSA.

To better quantify the percentage of MNCs and monocytes capable of binding MBP, we applied these cells to cell adhesion assays with the stringency of the assay regulated by centrifugational force (Fig. 2 A). In these experiments, MNCs bound poorly to the MBP-coated polystyrene surfaces in the presence of Mg²⁺ and Ca²⁺, while the further addition of Mn²⁺ to activate integrin ligand binding increased the binding to ∼6%. This percentage of binding cells agreed well with the percentage of monocytes in the preparation of MNCs. To test the role of β₂ integrins in cell adhesion to MBP, we also activated the binding by applying the mAb KIM127, which recognizes an epitope in the β₂ chain (CD18). As for integrin activation through the addition of Mn²⁺, application of the KIM127 Ab induced binding by ∼10% of the MNCs.

A similar experiment was conducted with purified monocytes. The percentage of cell adhesion induced by the addition of Mn²⁺ or the KIM127 was very similar with a maximum at ∼20% (Fig. 2,B). In the absence of integrin activation, i.e., in buffer with only Ca²⁺ and Mg²⁺ (Fig. 2,B) or with a control Ab added (Fig. 2,D), no cell binding was observed. Monocytes carry the integrin α_Mβ₂, while this receptor is not quantitatively found on resting lymphocytes; we hypothesized that this receptor was involved in the binding of MBP. To test the role of α_Mβ₂ integrin in the binding of monocytes to MBP, we applied a function blocking Ab (ICRF44) recognizing the ligand-binding domain in the α_M chain (43). Binding of monocytes to MBP was abolished in the presence of this Ab (Fig. 2,C), while a control Ab had no such effect (Fig. 2 D).

We tested whether α_Mβ₂ was capable of binding MBP in an assay using a K562 cell line derived from a human chronic myelogenous leukemia, expressing recombinant α_Mβ₂. The K562 cell line has no endogenous expression of β₂, which allows for a direct analysis of the contribution to adhesion by the recombinant expression of these integrins by comparison with untransfected K562 cells. The percentage of bead-binding cells at 34.5 ± 5.6% in the assay with α_Mβ₂-expressing cells (Fig. 1,E) was higher than for receptorless K562 cells, where 18.5 ± 6.6% of the input cells were positive for bead binding (Fig. 1,G). Efficient binding of beads with MBP by the α_Mβ₂/K562 cells required addition of Mn²⁺ (Fig. 1,E), which is a known activator of integrins in ligand binding. The α_Mβ₂/K562 cell binding in buffer with only Mg²⁺ and Ca²⁺ was at the level of background (Fig. 1, F–H). Combined with the studies on monocytes, these investigations show that activation of the integrin receptors significantly strengthens the interaction between MBP and α_Mβ₂. This is also consistent with the observations made in the flow cytometric assays with MNCs (Fig. 1, A–D) where only a low absolute percentage of monocytes bound MBP-conjugated beads in the absence of integrin-activating compounds.

Cell lines transfected with constructs expressing α_Mβ₂, α_Lβ₂, or α_Xβ₂ integrins were compared with the parental K562 cell line for their adhesion to MBP. In the presence of Mn²⁺, only the α_Mβ₂-expressing cells bound the MBP-coated surface better than the control cell line (Fig. 2,E). Addition of EDTA abolished the α_Mβ₂-expressing cell binding while buffer with Mg²⁺ and Ca²⁺ supported a weaker binding of these cells than with Mn²⁺ added (Fig. 2 F). No adhesion was detected to the HSA-coated surface under the condition of our assay suggesting that albumin was a significantly weaker ligand for α_Mβ₂ than MBP (data not shown).

GA has been suggested to mimic certain properties of MBP as an Ag (24, 25). MNCs adhered strongly to GA-coated surfaces in the presence of divalent cations but not in the presence of EDTA (Fig. 3,A). We tested whether GA, like MBP, was a ligand for α_Mβ₂. The α_Mβ₂-expressing cells adhered well to the GA-coated surfaces with adhesion peaking at 4 μg/ml coating concentration (Fig. 3,B) whereas neither the α_Xβ₂- nor α_Lβ₂-expressing cells bound the GA-coated surfaces above the level of untransfected cells. Without Mn²⁺ in the binding buffer or in the presence of EDTA, the adhesion of α_Mβ₂-expressing cells was reduced to background level (Fig. 3, C and D). The concentration of GA required to inhibit α_Mβ₂-expressing cell adhesion to MBP-coated surfaces by 50% (IC₅₀) was 0.1 μg/ml, corresponding to a total peptide concentration of ∼20 nM (Fig. 3,E). The IC₅₀ for MBP-mediated inhibition of cellular adhesion to GA-coated surfaces (Fig. 3 F) was 22 μg/ml (1.2 μM). By contrast, addition of 50 μg/ml (1 μM) HSA to the binding buffer had no effect on the cell adhesion to neither GA nor MBP (data not shown).

GA was also capable of inhibiting the binding of monocytes to MBP; preincubation of the cells with this compound at a concentration of 5 μg/ml completely quenched binding (Fig. 2 C).

We investigated the sequence similarity between the GA peptides and MBP by in silico generation of 100 GA peptide-like sequences followed by alignment of the peptides with the human MBP sequence for the 18.5 kDa isoform. Each peptide was aligned individually with the MBP sequence (Fig. 4,A); the alignment score for each MBP residue was calculated and plotted as shown in Fig. 4,C. The results revealed a much more frequent alignment of the GA-like peptides with MBP than would be expected from the variability of the 100 random sequences. The GA peptides aligned mainly within three distinct regions of the MBP primary structure (Fig. 4, A and C) that form the most hydrophilic regions of MBP according to the algorithm (34) by Kyte and Doolittle (data not shown). Approximately 30% of the GA-like peptides aligned with the N-terminal end of MBP (residues 2–25), 20% with a segment in the middle of the MBP primary structure (residues 51–76), and 50% with the C-terminal end of the protein (residues 127–164).

To compare the structural content of the MBP sequence in the regions aligned with the GA-like sequences, we plotted the disorder probability for each residue of MBP calculated with the DisEMBL algorithm (35). At least two of the peaks in alignment score (labeled with red and green in Fig. 4,C) coincided with a peak in disorder probability. A similar finding was made from comparison with the model of the MBP three-dimensional structure by Ridsdale et al. (10). In this model, the MBP structure carries a central β sheet with five strands. The aligned parts of MBP constituted the least structured regions in this model with the exception of a small overlap with the sequence forming the β1 strand in the model by Ridsdale et al. (10) (Fig. 4, C and D). Finally, we compared MBP and the GA-like peptides according to the method described by Uversky et al. (33) by showing these species on a charge-hydrophobicity space. In this plot, a linear relations ship between the mean net charge (<R>) and mean hydrophobicity (<H>) divides the space into two compartments mainly occupied by natively unfolded proteins and folded proteins, respectively, as shown in Fig. 4,B. The majority of the GA-like peptides occupied the compartment corresponding to the natively unfolded proteins, while a smaller fraction (6%) had properties allocating them to the compartment corresponding to folded proteins. MBP, as reported earlier (5), as well as the three fragments of this molecule with the highest alignment score (Fig. 4, A and C), also falls into the compartment with natively unfolded proteins. For comparison, we also showed the position of the histones H1 and H2B.1 as well as α-synuclein. The histones as well as α-synuclein were located at the edge of the area covered by the GA-like peptides (Fig. 4 B).

Experimental evidence for the similarity between GA and MBP was investigated by the surface plasmon resonance (SPR) technique, which can establish the binding kinetics of macromolecular interactions. The data presented above showing that a mAb to the α_M I domain could block the interaction with MBP (Fig. 2, C and D) suggested that this domain is capable of forming a crucial contact with MBP. Therefore, 2–10 μM solutions of α_Mβ₂ I domain were injected over immobilized GA or MBP in SPR assays. Robust binding was observed for both ligands in the presence of Mg²⁺ (Fig. 5, A and C) while EDTA reduced the binding by 80%. Obtained binding isotherms were analyzed using the algorithm developed by Svitel et al. (44) as described by Vorup-Jensen et al. (9). Such analysis resolves heterogeneous interactions into a set of separate binding sites differing in their on and off-rates, which is represented as a two-dimensional plot of K_D and the associated k_off values. The distribution of binding sites was remarkably similar for the two ligands: binding to both GA (Fig. 5,B) and MBP (Fig. 5 D) was largely characterized by two types of binding sites with a similar K_D at 10⁻⁴ M, but differing in their k_off at 0.1 s⁻¹ and 0.008 s⁻¹. In the case of the binding of the I domain to MBP, there was a third type of binding with K_D at 10⁻⁶ M and k_off at 0.01 s⁻¹.

The similarity between the GA and MBP from the point of view of their structural order was tested by SRCD spectroscopy to determine their secondary structure content in aqueous and in membrane-mimicking environments at physiological ionic strength (Fig. 5, E and F). MBP carried β-sheeted regions in physiological buffer constituting 35% of the secondary structure while α-helical regions took only a small fraction (11%) of the secondary structure (Fig. 5,G). NMR data demonstrated recently that presence of TFE created an environment that mimics the cellular membrane with regard to structural properties of MBP (45). In our case, addition of 20% (v/v) or 50% (v/v) TFE increased the content of α-helical regions in MBP to 30%. The content of β-sheeted regions in MBP in the presence of TFE at ∼20% agreed well with the MBP structure model by Ridsdale et al. (10) (Fig. 5, E and G). Importantly, similar conformational changes could be induced by TFE in the structure of GA peptides. Addition of TFE markedly changed the structure of GA peptides: the content of α-helical regions in the peptides was more than two times higher and the fractions of, in particular, unstructured loop regions and β-sheets were correspondingly reduced (Fig. 4, F and G).

The SRCD spectra of GA in plain buffer were collected over a temperature range between 5 and 85°C. The CD signal from the GA solution decreased continuously with an increase in temperature (Fig. 6, A and B). The α-helical content of GA decreased from 50 to 20% within the temperature interval from 5 to 40°C (Fig. 6,B). This observation was used to examine the effects of GA secondary structure changes on the capability of these peptides to bind α_Mβ₂. As shown in Fig. 6,D, binding of α_Mβ₂-expressing cells to the GA-coated surface was highly influenced by the temperature with essentially no binding at 5°C, some binding at 23°C, and a very strong binding at 37°C (Fig. 6,D). Application of varying centrifugation force in the assay highlighted even more that at 37°C the GA was a much better ligand than at 5 and 23°C. Contrary to GA-coated surface, the adhesion of the α_Mβ₂ and α_Xβ₂-expressing cells to a ligand for both receptors iC3b-coated surface was robust, and no major differences in adhesion were observed in assays at 5, 23, or 37°C (Fig. 6 C). Thus, our data indicate that the temperature-dependent binding of α_Mβ₂ to GA was a consequence of properties of the GA, and that only largely unfolded GA is a strong ligand for α_Mβ₂ integrin.

SAXS is used for the determination of the structure of soft material in solution in terms of averaged particle sizes or shapes. Previous SAXS studies demonstrated that lipid-bound MBP had a three-dimensional structure similar to the one proposed by Ridsdale et al. (10) while in the absence of lipids, this protein was less compact and highly flexible (11). Our experiments showed that average gyration radius (R_G) of MBP in plain buffer at 59 ± 3 Å was significantly larger than in the presence of TFE where R_G was 25 ± 1 Å (Fig. 7,G). This finding was confirmed by the shape of the distance distribution function p(r), which suggested that MBP possessed a globular shape in membrane-mimetic environment and a larger, more elongated structure in aqueous environment (Fig. 7,G). In a first approach to model the structure of MBP, an unconstrained dummy chain model was used to fit the scattering data. In this procedure, a sequence of connected spheres (one for each amino acid residue) was used to represent the protein structure. Using a simulated annealing optimization procedure, the dummy backbone was changed by translation and rotation of the dummy residues in such a way that the final structure gave the best fit to the scattering data (Fig. 7,F). Another modeling approach was based on the results from SRCD, which suggested the presence of β-sheeted regions in physiological buffer (Fig. 5,G). Therefore, in the search for models of the structure of MBP in solution, we applied the constraint that a β-sheeted region similar to the β-sheeted region of the MBP model by Ridsdale et al. (10) should be included in the structures of MBP in both plain and TFE-containing buffer. The remaining parts of the structure were modeled using dummy residues in a similar way as described above, and an in-depth structural analysis of the SAXS data for the model with included a β-sheeted region was performed (Fig. 7, A, B, and F). As expected from the R_G determinations and p(r) functions (Fig. 7,G), MBP assumed a highly compact conformation in the presence of TFE where the loop regions of the Ridsdale et al. (10) MBP model (Fig. 7,D) were closely folded together with the central β-sheeted region (Fig. 7, A and C). By contrast, in plain buffer MBP had a more loose structure with a more elongated shape than the Ridsdale model (Fig. 7, B and E).

In this study, we show that α_Mβ₂ binds strongly to MBP, which is a major autoantigen in MS. Other reports have implicated a role of α_Mβ₂ in the pathogenesis of demyelinating diseases, most recently by studies of EAE in mice, where deficiency in α_Mβ₂ protects against development of symptoms (23). Nevertheless, the molecular mechanisms supporting the involvement of α_Mβ₂ in demyelinating diseases are unclear. Our findings now indicate a targeted role of α_Mβ₂ in this type of pathogenesis. MBP is well known to be a promiscuous ligand forming weak contacts with many receptors. However, we found that only α_Mβ₂ integrin binds MBP, whereas other β₂ integrins in direct comparisons did not bind this molecule. Furthermore, MBP appears to be a stronger ligand than previously reported ligands for α_Mβ₂, such as albumin. Osteopontin, a member of the family of small integrin binding proteins and with a potential role in MS, did not show any binding either to α_Mβ₂ under the conditions of our assay (data not shown).

We used the MS drug GA as a model tool to study the interaction between MBP and α_Mβ₂. By contrast to the binding of GA to MHC class II molecules, which was found of to be of low specificity (46), we now show that GA binds strongly to the α_Mβ₂ integrin, while other β₂ integrins are not capable of supporting similar strong binding. This clearly argues that binding of GA to α_Mβ₂ is relatively specific. GA potently blocks the interaction between α_Mβ₂ and MBP at a concentration of 1 μg/ml, which is close to the concentration of 4 μg/ml in 5 liters of body fluids following administration of the recommended pharmaceutical dosage of 20 mg per day. Our data provide the first indication that α_Mβ₂ integrin is a target in GA treatment and that the affinity for GA is sufficient for a role of α_Mβ₂ in the pharmacology of this drug. With the many roles of α_Mβ₂ in the immune system, e.g., as a complement receptor and receptor for ICAM-1, the properties of GA as a ligand for α_Mβ₂ could also contribute to modulation of the immune response in scenarios not involving MBP. This is supported by the influences of GA in models of graft-vs-host disease and inflammatory bowel disease (26). In a recently published study, GA was found to shift the cytokine production of monocytes toward a more anti-inflammatory profile in a MBP-independent fashion (47); α_Mβ₂ is highly expressed on monocytes, and we speculate that GA binding to α_Mβ₂ could be a contributing factor consistent with other observations that ligand binding to this receptor may regulate cytokine production (48).

Alignments of random GA-like peptides with MBP demonstrated that the most frequently aligned MBP segments are located in relatively hydrophilic and least-structured regions of MBP. Analysis of the GA-like sequences according to the method described by Uversky et al. (49) supported by SRCD measurements on GA clearly indicates that these peptides, despite the large variation in primary structure, nearly all share a common trait in being unstructured. Consequently, we concluded that, despite the randomness of GA peptides, the GA mimics the least-folded regions of MBP.

The significance of this observation was investigated further by SPR and SRCD techniques to compare the functional and structural properties of GA and MBP. The observation that only α_Mβ₂ but not α_Lβ₂ or α_Xβ₂, binds GA and MBP suggests that binding sites are located within the α_M chain. The I domain of α_M chain mediate binding of the receptor to a number of ligands (50). X-ray crystallography of the isolated domain (51) and other I domains (50) allowed to visualize how the Mg²⁺-ion chelated by the I domain mediates a crucial contact to glutamate side chain in ligands. This is only possible in the so-called open conformation of the I domain (50). The requirement of Mn²⁺ for efficient binding of the intact α_Mβ₂ receptor to GA and MBP indicates a type of conformational regulation of integrin binding, which involves transmission of conformational changes through the receptor (52) and ultimately leads to the conformational changes in the I domain that enables ligand recognition. We cannot exclude secondary effects of the application of Mn²⁺ to the experiments, although MBP was found not to bind Mn²⁺ (53). To test the influence of integrin conformation by another strategy, we showed that binding by MNCs as well as monocytes was also induced by the addition of a mAb (28) that activates β₂ integrins for ligand binding; this finding clearly suggests that integrin conformational changes are required for the α_Mβ₂ receptor to bind MBP and that the primary effect of Mn²⁺ like the Ab is to favor the integrin ligand-binding conformation.

The critical role of the I domain in the binding of α_Mβ₂ to MBP was further shown in the cell adhesion assay with monocytes where a function-blocking mAb to the α_M I domain abrogated the binding of these cells to MBP. The isolated α_M I domain, mutated to favor the open conformation (36), bound robustly to GA and MBP in SPR assays. Consistent with the structural evidence that the GA peptides mimic the ligand binding sites for α_Mβ₂ in MBP, the types of binding sites for the α_M I domain found on GA-coupled or MBP-coupled surfaces were very similar. The K_D values for the major binding sites were in the range of 10⁻⁴ M, which comes close to the K_D of 3 × 10⁻⁴ M for the α_M I domain binding to soluble glutamate (9), suggesting that the binding to GA and MBP mainly is constituted by the interaction of the I domain with carboxyl groups. This is supported by the requirement for unfolding of the peptides to obtain strong binding to the receptors since unfolding decreases sterical hindrance in forming contact between the glutamate side chains and the integrin I domain. In the case of integrin α_Xβ₂, it has been suggested that the relatively positively charged residues surrounding the metal-ion dependent adhesion site in the α_X I domain explain the ability of this domain to bind polyanionic molecules, such as heparin and polyglutamate (9). The α_M I domain does not interact equally strongly with such ligands and appears to have more negatively charged residues surrounding the metal-ion dependent adhesion site. However, whether the α_Mβ₂ receptor in general would react with positively charged species has not been investigated although our finding of the binding to MBP and GA suggests such a possibility.

Analysis of SRCD spectra revealed that in a membrane-like environment an α-helical structure in GA was induced similarly to the structural changes in MBP. This finding together with the alignment of the GA-like peptides with MBP suggested that GA peptides mimic the TFE-induced response of the hydrophilic, unstructured parts of MBP molecule thought to contact with the cytoplasmic side of myelin sheath membranes. To ascertain that such folding of the loop regions indeed occurs in MBP exposed to a membrane-like environment, we determined the solution structure of MBP in plain buffer and in presence of TFE. By searching for MBP structures with a central β-sheet included, we modeled the solution structure of MBP in plain buffer in a way that fitted well with both the recorded SAXS spectrum and the distribution of secondary structure determined by SRCD. The model of MBP in plain buffer indicates that the molecule assumes a highly elongated shape with the loop regions pointing away from the β-sheeted region. By contrast, in TFE the MBP assumes a highly compact conformation where the regions forming loops in the absence of TFE closely fold together with the central β-sheet region. SAXS analysis does not allow for the assessment of the exact structure of these regions in presence of TFE, however, their compactness combined with the SRCD data suggest that at least some parts of these regions possess an α-helical structure. The functional significance of these conformational changes in MBP emerged from comparison with GA, which we showed were similar to these flexible MBP regions. Only the unfolded GA peptides are ligands for α_Mβ₂; therefore, in a similar manner the lack of structure in these regions of MBP in aqueous environment is a critical determinant for its recognition by α_Mβ₂.

Based on our findings, we would like to propose a new model for the sequels of receptor recognition in MS pathogenesis (Fig. 8). In the intact myelin sheath the MBP, in a compact conformation, is enclosed inside the oligodendrocyte cell and interacts with the opposing membranes (5). However, there are factors capable of upsetting the integrity of this environment. As pointed out above, previous studies have indicated that deimination of arginyl residues in MBP appear to be associated with MS (13) and the deimination is a likely source of loss of integrity in the myelin sheath membrane through the pivotal role of MBP in structuring the sheath (14). Furthermore, the deimination also contributes to structural instability of the MBP, which may also be relevant for the interaction with α_Mβ₂ as well as in other proinflammatory mechanisms (16). However, because the MBP applied in our study is a heterogenous mixture of molecules with regard to the modifications of arginyl residue (5), it is not clear at present whether some of these form are better binders of α_Mβ₂ than others; yet, the reasonably strong binding observed in our assays suggests that a high fraction of the MBP preparation is capable of supporting this interaction. Another factor could be oligodendrocyte apoptosis, which is a primary event of myelin sheath damage as studies suggest (54). At the onset of the pathogenic events the oligodendrocyte membrane is damaged allowing the contents of the cell, including MBP, to leak out into the more aqueous surrounding environment. The ensuing conformational changes in MBP expose unfolded loop regions that are potent ligands for the α_Mβ₂ integrin. This enables α_Mβ₂-carrying phagocytic cells to recognize and bind the MBP molecules, either in a soluble form, which leads to the MBP presentation for autoreactive T cells and triggers the autoimmune cascade, or associated with the damaged myelin sheath, which leads to phagocytosis and removal of damaged myelin. Several studies have already shown that microglia, invading monocytes, and macrophages are present in the MS lesions and are capable of binding myelin sheath fragments in a α_Mβ₂-dependent fashion. The MS drug GA is a specific ligand for α_Mβ₂ integrin due to its lack of secondary structure at physiological temperature. We now suggest that one of the pharmaceutical mechanisms of this drug in MS is likely to be a highly efficient blocking of proinflammatory cells binding to MBP through α_Mβ₂. The broad albeit safe capability to attenuate the immune response by GA is entirely consistent with our suggestion that this drug represents a novel class of scavenger receptor antagonists that works by blocking the α_Mβ₂-mediated immune cell recognition of unfolded protein species.

This is the first report of integrin binding to an intrinsically unstructured protein, which is known to play a significant role in disease. We speculate that the recently observed influence of GA in Alzheimer’s disease models (55, 56), where the α_Mβ₂-expressing microglia cells are known to play a role, reflects a similar underlying pathogenesis that involves unfolded protein species and their interaction with α_Mβ₂ integrin. The broad albeit safe capability to attenuate the immune response by GA is entirely consistent with our suggestion that this drug represents a novel class of scavenger receptor antagonists that works by blocking the α_Mβ₂-mediated immune cell recognition of unfolded protein species.

We are indebted to Dr. Nathan Astrof for inspiration and Dr. Per Höllsberg for a critical reading of a draft version of the manuscript. We thank Dr. Jesper Reinholdt for help with protein purification and Bettina W. Grumsen for technical assistance. T.V.-J. dedicates this paper to the memory of Inger Mathilde Elise Studier.

The authors have no financial conflict of interest.