Multiple sclerosis (MS) is an autoimmune–inflammatory disease of the central nervous system (CNS) with prominent demyelination and axonal injury. While most MS therapies target the immunologic response, there is a large unmet need for treatments that can promote CNS repair. LINGO-1 (leucine-rich repeat and Ig-containing Nogo receptor interacting protein-1) is a membrane protein selectively expressed in the CNS that suppresses myelination, preventing the repair of damaged axons. We are investigating LINGO-1 antagonist antibodies that lead to remyelination as a new paradigm for treatment of individuals with MS. The anti-LINGO-1 Li81 antibody,BIIB033, is currently in clinical trials and is the first MS treatment targeting CNS repair. Here, to elucidate the mechanism of action of the antibody, we solved the crystal structure of the LINGO-1–Li81 Fab complex and used biochemical and functional studies to investigate structure-function relationships. Li81 binds to the convex surface of the leucine-rich repeat domain of LINGO-1 within repeats 4–8. Fab binding blocks contact points used in the oligomerization of LINGO-1 and produces a stable complex containing two copies each of LINGO-1 and Fab that results from a rearrangement of contacts stabilizing the quaternary structure of LINGO-1. The formation of the LINGO-1–Li81 Fab complex masks functional epitopes within the Ig domain of LINGO-1 that are important for its biologic activity in oligodendrocyte differentiation. These studies provide new insights into the structure and biology of LINGO-1 and how Li81 monoclonal antibody can block its function.
LINGO-1 (leucine-rich repeat and Ig-containing Nogo receptor interacting protein-1; LERN1; LRRN6A) is a 581-amino-acid membrane-associated glycoprotein selectively expressed on neurons and oligodendrocytes in the central nervous system (CNS) (Mi et al., 2005). It contains a large extracellular domain with 12 leucine-rich repeat (LRR) motifs flanked by N- and C-terminal capping modules, one Ig domain of the I1 subtype, and a stalk region that is attached to a transmembrane region and a short cytoplasmic tail (Mi et al., 2004; Mosyak et al., 2006). LINGO-1 suppresses oligodendrocyte differentiation, thereby preventing axonal myelination (Mi et al., 2005, 2007, 2009). Blocking its function leads to robust myelination in vitro and in animal models of demyelination (Mi et al., 2005, 2007, 2009). The biological effects of LINGO-1 antagonists have been validated using small interfering RNA; a dominant negative construct lacking the cytoplasmic tail; a soluble LINGO-1–Fc fusion protein in which the LINGO-1 extracellular domain is genetically linked to the hinge region and Fc of human IgG1; and anti-LINGO-1 monoclonal antibodies (mAbs), Fabs, Fab2s, and polyethyelene glycol–Fabs (Mi et al., 2005, 2007, 2009; Ji et al., 2006; Fu et al., 2008; Pepinsky et al., 2010, 2011; Cen et al., 2013). Of these, antibodies were preferred because of their potency, selectivity, and duration of treatment effects due to long serum half-life (Aires da Silva et al., 2008; Kontermann, 2009; Pepinsky et al., 2011). The anti-LINGO-1 antibody Li81 was isolated using Fab phage display technology and engineered into a human IgG1 aglycosyl framework (Hoet et al., 2005; Pepinsky et al., 2011). We are currently investigating Li81 mAb (BIIB033) in clinical trials as a potential treatment to repair neuronal damage that occurs in the CNS of individuals with multiple sclerosis (MS) (Mi et al., 2013). The clinicaltrial.gov number for the phase 2 MS study is SYNERGY: NCT01864148.
The use of antibodies as drugs that target antigens expressed in the CNS is challenging because of poor exposure to the CNS following systemic administration due to the blood-brain barrier (Cirrito et al., 2003; Shen et al., 2004; Levites et al., 2006; Garg and Balthasar, 2009; Pepinsky et al., 2011). Current estimates are that only 0.1% of antibody drug levels in blood reach the CNS. Consistent with low levels reported by other laboratories, we determined that Li81 mAb levels in brain and spinal cord following systemic administration of the antibody in rats were 0.15% of blood levels, but despite the low CNS exposure, Li81 treatment led to the repair of damaged CNS neurons with a direct dependence of dose on the extent of CNS repair (Pepinsky et al., 2011). These findings provided clear evidence that remyelination that results from blocking of LINGO-1 function is directly linked to the binding of Li81 to LINGO-1. Cellular responses that are mediated by LINGO-1 result from direct binding of LINGO-1 to growth factor receptors to block their function. To date, four LINGO-1 signaling pathways have been elucidated (Mi et al., 2013; Lee et al., 2014). Structure-activity relationship studies suggest that the Ig domain of LINGO-1 is sufficient for its activity (Bourikas et al., 2010; Mi et al., 2010; this study). The crystal structure of the LINGO-1 ectodomain revealed that the protein self-associates to form a ring-shaped tetramer (Mosyak et al., 2006). In the structure, potential functional binding sites on the Ig domain are solvent-exposed, providing a model for how LINGO-1 functions. How the Li81 antibody blocks LINGO-1 function is unknown.
Here, to investigate the mechanism of action of the Li81 antibody, we solved the crystal structure of the LINGO-1 ectodomain–Li81 Fab complex and discovered that the binding of Li81 to LINGO-1 blocked contacts that allowed LINGO-1 to form the tetrameric state seen in the published LINGO-1 crystal structure. Li81 Fab treatment prevented the formation of LINGO-1 homo-oligomers in solution and on transfected cells expressing full-length LINGO-1, and led to a rearrangement in the quaternary structure of LINGO-1, which provides a model for how the antibody inhibits LINGO-1 function in oligodendrocytes. These studies provide a detailed assessment of domains within the LINGO-1 structure that are important for function and expand our understanding of how they contribute to the complex biology of LINGO-1.
Materials and Methods
LINGO-1–Fc produced by fusing the extracellular portion of human LINGO-1 (residues 1–499) to the hinge and Fc region of human IgGl was expressed in Chinese hamster ovary (CHO) cells and purified from clarified and filtered cell culture medium on recombinant protein A-Sepharose Fast Flow (GE Healthcare, Chalfont St. Giles, Buckinghamshire, UK) at a loading density of 6 mg of fusion protein/ml resin. The LINGO-1–Fc was eluted from the column with 25 mM NaH2PO4 (pH 2.8) and 100 mM NaCl and neutralized by adding 0.05 volume of 0.5 M Na2HPO4 (pH 8.6). Peak protein-containing fractions were identified by absorbance at 280 nm, pooled, and dialyzed against phosphate-buffered saline (PBS). The preparation was 0.2-μm filtered, and divided into aliquots that were frozen on dry ice, and stored at −70°C. The protein concentration was calculated using an extinction coefficient of ε2800.1% = 1.04 using DU 800 Spectrophotometer NetA software (Beckman Coulter, Brea, CA), which compensates for baseline shifting by subtracting absorbance due to turbidity from the value at 280 nm. LINGO-1 ectodomain was generated from LINGO-1–Fc by digestion with chymotrypsin or endoproteinase Lys-C as indicated. For limited proteolysis with chymotrypsin, LINGO-1–Fc (2.5 mg/ml) in PBS was diluted with 2 volumes of 100 mM Tris-HCl (pH 7.5) and 10 mM CaCl2. Chymotrypsin (Roche Diagnostics, Indianapolis, IN) was added to a 1:500 enzyme/protein ratio (w/w) and incubated for 3 hours at room temperature. Phenylmethanesulfonylfluoride was added to a final concentration of 0.5 mM, and an additional 100 mM NaCl was added. The sample was loaded onto a protein A-Sepharose column (8 mg protein/ml resin) equilibrated in PBS. The flow-through fraction was pooled, concentrated using an Amicon Ultra-15 Centrifugal Filter device (10-kDa membrane; Millipore, Billerica, MA), and loaded onto a Superdex 200 size exclusion chromatography (SEC) column (GE Healthcare). Peak fractions were pooled and divided into aliquots that were stored at −70°C. The protein concentration was calculated using an extinction coefficient of ε2800.1% = 0.9. For limited proteolysis with endoproteinase Lys-C (Wako Chemicals, Richmond, VA), LINGO-1–Fc at 7.5 mg/ml in PBS was incubated for 2 hours at room temperature with a 1:1000 enzyme/protein ratio (w/w). Leupeptin was added to a final concentration of 0.2 mM, and an additional 100 mM NaCl was added. The sample was loaded onto a protein A-Sepharose column (8 mg protein/ml resin) equilibrated in 20 mM Na2HPO4 (pH 7.5) and 250 mM NaCl. The flow-through fraction was pooled and loaded onto a Ni-NTA Superflow column (Qiagen, Germantown, MD) at 3 mg protein/ml resin. The column was washed with 3 column volumes of 20 mM Na2HPO4 (pH 7.5) and 250 mM NaCl, and the LINGO-1 was step eluted with the same buffer containing 20 mM imidazole. The elution pool was concentrated using Amicon Ultra-15 Centrifugal Filter devices (10-kDa membrane) and loaded onto a Superdex 200 SEC column. Peak fractions were pooled and divided into aliquots that were stored at −70°C. Boundary amino acid sequences within LINGO-1-Fc are 1TGCPPR……SNQPGE499 for the amino and carboxyl termini of LINGO-1 ectodomain and 500VDKTHT……LSLSPG726 for the amino and carboxyl termini of IgG1 hinge and Fc.
For structural work, the LINGO-1–Fc-producing CHO cells were cultured in the presence of kifunensine to reduce the complexity of the carbohydrates (Yu et al., 2011). The same methods were used for purification of LINGO-1–Fc and generation of the ectodomain fragments for LINGO-1–Fc produced in the presence and absence of kifunensine. Truncated versions of LINGO-1 ectodomain containing the LRR domain alone (residues 1–382), Ig domain alone (residues 383–460), and Ig domain (residues 383–460) with the RKH sequence (residues 423–425) mutated to EKV were fused to the hinge and Fc region of human IgGl, expressed in CHO cells, and purified from culture medium on protein A-Sepharose as described above.
Li81 Fab Production.
The Li81 mAb was purified from clarified and filtered CHO cell culture medium on recombinant protein A-Sepharose Fast Flow. The Li81 mAb was dialyzed into 10 mM sodium acetate (pH 3.6). Li81 Fab2 fragment was generated from the mAb by digestion with pepsin at an enzyme/protein ratio of 1:500 for 4 hours at 37°C. Tris-HCl was added to 300 mM to quench the reaction and the sample dialyzed against 50 mM Tris-HCl (pH 8.9) and 50 mM NaCl. The sample was subjected to anion exchange chromatography on a Fractogel EMD TMAE (M) (40–90 μm) (Merck, Darmstadt, Germany) column equilibrated in 50 mM Tris-HCl (pH 8.9) and 50 mM NaCl. The column was washed with 50 mM Tris-HCl (pH 8.9) and 50 mM NaCl and the Fab2 eluted with 20 mM Na2HPO4 (pH 7.5) and 150 mM NaCl (PBS). Li81 Fab fragment was generated from the Fab2 by digestion with papain at an enzyme/Fab2 ratio of 1:300. Fab2 in 10 mM Na2HPO4 (pH 7.5), 5 mM EDTA (pH 7.5), 20 mM cysteine-HCl, and 20 mM NaOH was incubated with papain at 37°C for 2.5 hours and quenched with the thiol protease inhibitor E64. The sample was dialyzed against PBS.
Samples were subjected to SDS-PAGE on 4–20% Tris-glycine gradient gels (Invitrogen, Carlsbad, CA) and stained with Coomassie Brilliant Blue. Nonreduced samples were diluted with Laemmli nonreducing sample buffer and heated at 75°C for 5 minutes prior to analysis. Reduced samples were treated with sample buffer containing 2% 2-mercaptoethanol and heated at 95°C for 2 minutes. Electronic images of stained gels were acquired with the true color setting on a HP Scanjet G4050 scanner (Hewlett-Packard, Palo Alto, CA). For densitometry, the gel images were acquired and band intensities quantified using a GS-800 Calibrated Densitometer (Bio-Rad Laboratories, Hercules, CA) and the Quantity One 1-D Analysis software (Bio-Rad). LINGO-1 and Li81 concentrations in the LINGO-1–Li81 Fab complex were calculated by interpolation of measured densities from standard curves of densities of LINGO-1 and Li81 Fab standards, which were run on the same gel.
Analytical SEC and Light Scattering.
Size exclusion chromatography was carried out on a BioSep-SEC-S3000 column, 300 × 7.8 mm (Phenomenex, Torrance, CA), in 20 mM sodium phosphate and 150 mM NaCl (pH 7.2), using a flow rate of 0.6 ml/min on a Waters Alliance instrument (Waters 2790; Waters Corp., Milford, MA). The column effluent was monitored by UV detection at 280 nm. Static light scattering was synchronized with SEC and measured online using a Precision PD2100 Detector (Precision Detectors, Bellingham, MA). Molecular weights were calculated using Discovery 32 Light Scattering Analysis Software.
Solution-phase affinity measurements were performed on a Biacore 3000 instrument (Biacore AB, Uppsala, Sweden). Human LINGO-1 ectodomain was immobilized on CM5 chips using amine-coupling chemistry in Biacore buffer [10 mM HEPES (pH 7.2), 150 mM NaCl, 3.4 mM EDTA, and 0.005% surfactant P20]. Binding experiments were run in Biacore buffer containing 0.05% bovine serum albumin. For measuring LINGO-1–LINGO-1 interactions, serial 2-fold dilutions of the analyte human LINGO-1 ectodomain from 0.2 to 2 μM were used. For affinity of Li81 Fab, serial 2-fold dilutions of the Li81 Fab from 0.1 to 1 nM were tested. Data were analyzed with BIAevaluation 3.0 Software. Competition analyses of Li81 Fab with 1A7 were performed at a fixed concentration of 10 nM LINGO-1 ectodomain with 1 μM 1A7 Fab, or 25 nM and 1 μM 1A7 mAb, on CM5 chips coated with Li81 and 1A7 mAbs using the amine-coupling method.
LINGO-1 ectodomain (14 μM) alone in 20 mM sodium phosphate, 150 mM NaCl, and 35 mM HEPES (pH 7.5) in the presence of 28 μM Li81 Fab or within the SEC-purified complex was incubated for 70 minutes at room temperature with 1 mM bissulfosuccinimidyl suberate (BS3) (Thermo Scientific, Waltham, MA). The reaction was stopped with 50 mM ethanolamine (pH 8.0). LINGO-1 ectodomain was preincubated with Fab for 30 minutes at room temperature prior to treatment. For cross-linking on cells, stable CHO cells expressing hemagglutinin (HA) tag full-length LINGO-1 (residues 1–581 with HA tag engineered at the N terminus) were maintained at 37°C in Alpha plus Eagle’s minimum essential medium containing 10% fetal bovine serum and 400 μg/ml Geneticin (gentamycin; Invitrogen). Cross-linking was performed on six-well plates. Cells were seeded at 4 × 105 cells/well and incubated in medium without gentamycin for 2 days. Cells were washed twice with 2 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM CaCl2, 1 mM MgCl2, and 25 μg/ml human serum albumin, then treated with 10 μg/ml Li81 Fab for 15 minutes at room temperature and cross-linked at 4°C for 1 hour with 1 mM BS3. The reaction was stopped with 50 mM ethanolamine (pH 8.0). Cells were lysed in 1 ml of radioimmunoprecipitation assay (RIPA) lysis buffer [50 mM Tris HCl (pH 7.2), 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 10 mM MgCl2, and 5% glycerol], and samples were clarified by centrifugation in an Eppendorf centrifuge. Cell supernatants were incubated with 15 μl of anti-HA Affinity Matrix (Roche), the resin was washed three times with 0.3× lysis buffer, and the immunoprecipitates were analyzed by Western blotting using horseradish peroxidase–conjugated rat anti-HA antibody for detection. Similar results were generated when cells were treated with Li81 Fab for 3 hours at 37°C prior to cross-linking (data not shown).
Crystallization and Data Collection.
SEC-purified LINGO-1 ectodomain (5.6 mg) produced from LINGO-1–Fc cultured in the presence of kifunensine and digested with chymotrypsin was added to 4.2 mg of Li81 Fab. The sample was incubated for 1 hour at room temperature, then concentrated to 2 ml in an Amicon Ultra-15 Centrifugal Filter device and applied to a 1.6 cm × 60 cm Superdex 200 column. Fractions containing the LINGO-1–Li81 Fab complex were pooled, dialyzed against 10 mM HEPES (pH 7.5) and 150 mM NaCl, and 0.2-μm filtered. The complex was concentrated to 4.5 mg/ml. The LINGO-1–Li81 Fab complex was crystallized by the nanodroplet vapor diffusion method at a temperature of 297 K by mixing 200 nl of 4.3 mg/ml LINGO-1–Li81 Fab solution [10 mM Tris (pH 8.0) and 150 mM NaCl] with 200 nl of the reservoir solution containing 36% pentaerythritol propoxylate (5/4 PO/OH), 0.2 M sodium thiocyanate, and 0.1 M HEPES (pH 7.0). LINGO-1–Li81 Fab crystals were transferred to an oil-based cryoprotectant (66.5% paratone, 28.5% paraffin oil, and 5% glycerol) prior to harvesting and flash freezing in liquid nitrogen.
Diffraction data were collected at a wavelength of 0.9793 Å at the Advanced Photon Source on the Lilly Research Laboratories Collaborative Access Team beamline (Argonne, IL). The data set was collected at 100 K using a MARmosaic 225 detector. Data were integrated, reduced, and scaled using HKL2000 (Otwinowski and Minor, 1997). The LINGO-1–Li81 Fab crystal was indexed in the hexagonal space group P6522.
Structure Determination and Refinement.
The LINGO-1–Li81 Fab structure was determined to 3.23-Å resolution by molecular replacement using the apo Li81 Fab structure and LINGO-1 structure (Protein Data Bank code 2ID5) as the search models with the program Phaser (McCoy, 2007). The Li81–LINGO-1 model was manually built with Coot (Emsley and Cowtan, 2004). Structure refinement was performed using REFMAC (Vagin et al., 2004). The progress of the model refinement was monitored by cross-validation Rfree, which was computed from a randomly assigned test set comprising 5% of the data. The final stages of refinement used TLS refinement (Winn et al., 2001) with anisotropic motion tensors refined for each of the LRR and Ig domains of LINGO-1 and each domain of Li81 Fab. The final model includes one Fab molecule (heavy chain residues 1–221, light chain residues 1–214) and LINGO-1 molecule (residues 3–477). No electron density was observed for residues 1–2 of LINGO-1. The final R factor is 19.2%, with an Rfree factor of 25.6%. Refinement statistics are summarized in Table 1. Coordinates and structure factor data of the LINGO-1–Li81 Fab structure have been deposited with the Research Collaboratory for Structural Bioinformatics protein data bank (Protein Data Bank code 4OQT).
Hydrogen bonding, salt bridges, and van der Waals contacts were identified with the program CONTACTS (Collaborative Computational Project, Number 4, 1994). Buried surface areas were calculated with Areaimol with a 1.4-Å probe radius and standard van der Waals radii. Analysis of models for the macromolecular assemblies of the LINGO-12Li81 Fab2 tetrameric complexes were evaluated using the PISA algorithm (Krissinel and Henrick, 2007). Shape complementarity was calculated as described by Lawrence and Colman (1993). Figures were prepared with PyMOL (Schrödinger LLC, Portland, OR).
Dynamic Light Scattering.
The hydrodynamic radius of LINGO-1 in 50 mM Tris-HCl (pH 7.5) and 150 mM NaCl was measured at protein concentrations from 1.25–30 μM with dynamic light scattering (DLS) using a DynaPro Plate Reader II (Wyatt Technology, Santa Barbara, CA). Samples were filtered through a 0.1-μm filter, transferred to a 384-well plate, and analyzed using a 5-second acquisition time at 20°C; 10–20 acquisitions were averaged per sample. Cross-linked LINGO-1 and LINGO-1–Li81 Fab complex were analyzed using the same procedure. The Optimization Calculator tool in the DYNAMICS software package (Wyatt Technology) was used to determine the minimum protein concentration needed to obtain a good correlation function. The lowest recommended concentration for LINGO-1 was 1.18 μM.
Oligodendrocyte Differentiation Assays.
Enriched populations of oligodendrocyte precursors (A2B5+) isolated from the forebrain of female Long-Evans postnatal day 2 rats were grown in culture in high-glucose Dulbecco’s modified Eagle’s medium containing fibroblast growth factor/platelet-derived growth factor (PeproTech, Rocky Hill, NJ) (10 ng/ml). To assess differentiation of the rat A2B5+ progenitor cells into mature myelin basic protein (MBP)-positive oligodendrocytes, A2B5+ cells were plated into 24-well culture plates in fibroblast growth factor/platelet-derived growth factor-free growth medium supplemented with N2, and treated with Li81 antibody or soluble LINGO-1 reagents for 72 hours. The cells were lysed in RIPA buffer. The lysates were clarified by centrifugation, and then boiled in Laemmli sample buffer and subjected to SDS-PAGE on a 4–20% gradient gel. Proteins were transferred to nitrocellulose membranes in a Trans-Blot Cell (Bio-Rad) at 30 V for 2 hours at 4°C in 50 mM Tris-HCl, 400 mM glycine, 0.1% SDS, and 20% methanol. MBP expression was analyzed by Western blotting using a mixture of anti-MBP antibodies SMI-94 and SMI-99 (Calbiochem, San Diego, CA) and horseradish peroxidase–conjugated donkey anti-mouse antibody (Jackson ImmunoResearch, West Grove, PA) for detection. The same blot was probed with anti-actin antibody A5060 (Sigma-Aldrich, St. Louis, MO) as an internal control. For coculture studies, A2B5+ oligodendrocytes were added to dorsal root ganglion (DRG) neuron cultures as previously described (Mi et al., 2005). The cells were cultured for 14 days in the presence of the indicated treatments, then lysed with RIPA buffer and analyzed for MBP expression by Western blotting as described above.
Negative Stain Electron Microscopy.
A preparation of SEC-purified endoproteinase Lys-C–generated LINGO-1 ectodomain–Li81 Fab complex was cross-linked with 1 mM BS3 and SEC-purified. The protein was adsorbed on ultrathin carbon, 400-mesh copper grids (Ted Pella, Inc., Redding, CA) and stained with 0.75% uranyl formate. Micrographs were recorded on a JEOL 1200EX electron microscope. To compare electron microscopy (EM) data to X-ray crystal structure models, a set of two-dimensional projections were generated from the three-dimensional X-ray maps on a coarse angular grid with ∼15° spacing using the SPIDER software package (Shaikh et al., 2008). A 20-Å resolution cutoff was applied to the high-resolution models before projection to mimic the lower-resolution EM images.
Biochemical Properties of LINGO-1 and LINGO-1–Li81 Fab Complex.
Anti–LINGO-1 Li81 (BIIB033) antibody is in clinical trials for treatment of MS. To understand the molecular basis of antigen recognition, we crystallized a Fab fragment of the Li81 monoclonal antibody (Li81 Fab) in complex with LINGO-1 ectodomain. LINGO-1–Fc was used as a source of LINGO-1 ectodomain. For these studies, LINGO-1–Fc was cultured in CHO cells in the presence of kifunensine to reduce glycan complexity, and the purified protein was subjected to limited digestion with chymotrypsin to release the LINGO-1 ectodomain from the Fc. The chymotryptic fragment (amino acids 1–478) contained the entire LRR and Ig domains of LINGO-1, the same version that had been previously crystallized by Mosyak and coworkers in 2006. A schematic summarizing structural features of the chymotryptic fragment of the ectodomain is shown in Fig. 1A. LINGO-1–Fc was also produced from CHO cells grown in the absence of kifunensine to assess the biochemical characteristics of LINGO-1 containing complex, sialylated glycans routinely found on mammalian proteins. For these studies, LINGO-1 ectodomain was produced from LINGO-1–Fc following digestion with endoproteinase Lys-C. The endoproteinase Lys-C fragment contained the entire chymotryptic fragment of LINGO-1 plus 10 amino acids of the stalk region ending with amino acid residue 488. Figure 1B shows an analysis of key reagents by SDS-PAGE. LINGO-1–Fc migrated under reducing conditions as a single band (>95% pure) with an apparent mass of 100 kDa (lane 2 without, lane 3 with kifunensine). Under nonreducing conditions, LINGO-1–Fc migrated with an approximate mass of 200 kDa, consistent with the dimeric structure of the fusion protein that is stabilized by disulfide bonds in the IgG hinge (lane 4). The chymotryptic fragment of the LINGO-1 ectodomain fragment had an apparent mass of 70 kDa (lane 6 without, lane 7 with kifunensine). The sample produced from LINGO-1–Fc cultured in kifunensine (lane 7) showed a sharper gel band than the corresponding fragment from LINGO-1–Fc produced using standard culture conditions (lane 6), consistent with the reduced glycan heterogeneity. The endoproteinase Lys-C fragment of the LINGO-1 ectodomain also had an apparent mass of 70 kDa (lane 5). When the endoproteinase Lys-C (lane 5) and chymotryptic (lane 7) fragments were mixed with Li81 Fab and the complexes were SEC-purified and analyzed by SDS-PAGE under nonreducing conditions, we obtained the preparations shown in lanes 8 and 9, respectively. In the presence of SDS, the complexes completely dissociated into free LINGO-1 and Fab. Quantification of band intensities by densitometry revealed equimolar associations of LINGO-1 and Fab in the complex. Densitometry data for the sample shown in lane 8 are provided in Supplemental Fig. 1. The batches of the endoproteinase Lys-C LINGO-1 ectodomain fragment, Li81 Fab, and SEC-purified LINGO-1–Li81 Fab complex shown in lanes 5, 8, and 10 were used in all the biochemical studies presented in the article.
Biochemical studies shown in Fig. 2 were used to assess the oligomerization state of LINGO-1. SEC analysis of the LINGO-1 ectodomain showed concentration-dependent changes in molecular weight consistent with formation of reversible oligomers, which were stabilized into dimer and tetramer following cross-linking (Fig. 2, A and B). Apparent molecular masses measured by light scattering ranged from 85 to 120 kDa in the absence of cross-linker to 160 and 230 kDa for cross-linked dimer and tetramer (Fig. 2, A and B). The presence of cross-linked dimer and tetramer in the SEC elution fractions was confirmed by SDS-PAGE (Supplemental Fig. 2). The dependence of size on the LINGO-1 concentration in the absence of cross-linking is as was expected for a protein in a rapidly exchanging equilibrium and is indicative of low-affinity interactions (Winzor and Scheraga, 1963; Zoltowski and Crane, 2008). Biacore studies established a more quantitative measure of the low-affinity LINGO-1–LINGO-1 interactions observed by SEC (Fig. 2C) and revealed an apparent KD of 0.6 μM for binding of LINGO-1 to LINGO-1. Both association and disassociation of oligomers observed by Biacore occurred in seconds, which accounts for the rapid exchange characteristics seen by SEC. During the 15–20 minutes LINGO-1 migrates on the SEC column, interconversion between monomer and oligomer states can occur 1000 times. This leads to concentration-dependent elution as single peaks intermediate between monomer and oligomer where the weight average migration is based on the equilibrium state. Concentration-dependent oligomerization was also observed by DLS (Fig. 2D). With and without cross-linking the hydrodynamic radius approached 6.5 nm, consistent with the formation of a tetramer. The oligomer state remained constant at concentrations >10 μM, which is in close agreement with the DLS value of ≥16 μM reported by Mosyak et al. (2006) in which a LINGO-1 ectodomain preparation lacking complex glycans had been characterized. DLS could not be used to investigate monomer-dimer conversion that occurs at low LINGO-1 concentrations because 1 μM LINGO-1 was the lower limit of detection of the method (see Materials and Methods). Cross-linking studies with CHO cells expressing full-length human LINGO-1, as assessed by SDS-PAGE/Western immunoblotting, were used to study LINGO-1–LINGO-1 interactions in its more relevant membrane-bound form (Fig. 2E). Full-length LINGO-1 migrated on SDS-PAGE gels with an apparent molecular weight of 80 kDa in the absence of cross-linker, consistent with the predicted size of the protein (Fig. 2E). After cross-linking, two prominent forms were detected with apparent masses of 200 and 400 kDa, consistent with the formation of dimer and tetramer. Cross-linking was very efficient, as evidenced by the near complete disappearance of the non–cross-linked LINGO-1 (80-kDa band). The large proportion of cross-linked 400-kDa product indicates that membrane-associated LINGO-1 can assemble into a tetramer, which supports the model for oligomerization based on the published crystal structure of the LINGO-1 ectodomain (Mosyak et al., 2006).
Biochemical characteristics for binding of the Li81 antibody to LINGO-1 were also determined (Fig. 3). Biacore provided accurate measurements of binding affinities of the Li81 mAb and Fab for LINGO-1 (see Fig. 3A for Fab data). KD values were ≤20 pM for binding of Li81 mAb and ≤50 pM for the binding of the Li81 Fab to the LINGO-1 ectodomain. Similar values were also measured for their binding to LINGO-1–Fc. When Li81 Fab was added to the LINGO-1 ectodomain and analyzed by SEC equipped with light scattering, the LINGO-1 ectodomain–Li81 Fab complex eluted with molecular weight of 250 kDa (Fig. 3B), twice the anticipated mass of 120 kDa assuming a 1:1 LINGO-1–Fab complex. Varying the ratio of LINGO-1 to Li81 Fab over a wide range of concentrations had no effect on the assembly of the 250-kDa complex (Fig. 3C). Furthermore, in sharp contrast to concentration-dependent changes observed for LINGO-1 alone in Fig. 2A, there was no change in its association state when the purified 250-kDa complex was analyzed by SEC over a range of concentrations from 2 μM to 3 nM (data not shown). A concentration of the complex of 3 nM was the lowest that could be evaluated by SEC due to limits of detection of the method using absorbance at 280 nm for detection. By labeling the complex with Alexa-488 and using fluorescence detection, we were able to extend the lower limit of quantification of the SEC method by 100-fold. Based on the stability data during SEC analysis, we can infer an EC50 value for dissociation of the complex of ∼0.5 nM. Consistent with size measurements by SEC/light scattering, DLS analysis of the complex revealed a hydrodynamic radius of 6.95 nm, which is slightly larger than the 6.5-nm value observed for the LINGO-1 homotetramer.
Cross-linking studies were used to further assess the impact of Li81 Fab treatment on the assembly of LINGO-1 oligomers. Figure 3D shows an analysis of cross-linked LINGO-1 ectodomain alone and cross-linked SEC-purified LINGO-1–Li81 Fab complex by SDS-PAGE. In the absence of Fab, cross-linking led to the formation of LINGO-1 dimer and tetramer, in agreement with SEC results (Fig. 2B). The profile of cross-linked adducts closely resembled the data for cross-linked full-length LINGO-1 on cells shown in Fig. 2E, indicating that protein-protein interactions of LINGO-1 in solution and on cells are similar. SDS-PAGE analysis of the cross-linked LINGO-1–Li81 Fab complex revealed a prominent diffuse band at an intermediate position between LINGO-1 dimer and tetramer and the disappearance of the tetramer band observed in the cross-linked LINGO-1 sample without Fab treatment (Fig. 3D). The formation of the cross-linked dimer was also impacted, as evident from the decrease in the intensity of the cross-linked dimer band following Li81 Fab treatment. Although the cross-linking analysis was very sensitive to changes in the oligomerization state of LINGO-1, the analysis by SDS-PAGE did not allow us to determine the stoichiometry of LINGO-1 and Fab in the complex. On the basis of size by SEC/light scattering, hydrodynamic radius by DLS, equimolar stoichiometry determined by SDS-PAGE, and data from the crystal structure and EM analysis of the complex discussed below, we infer that the complex is a tetramer containing two copies each of the LINGO-1 ectodomain and of the Fab (LINGO-12Li81 Fab2). The EM analysis of the cross-linked SEC-purified complex revealed a very homogeneous preparation in which ≥95% of the protein had assembled into a single LINGO-12Li81 Fab2 tetrameric state. Quantification of the cross-linked samples seen in Fig. 3D by densitometry revealed that in the SEC-purified complex ≥90% of the LINGO-1 had been cross-linked, as evident from reduction in free LINGO-1, whereas only 30% of the Li81 Fab was cross-linked, as evident from reduction in free Li81 Fab. The lower extent of reaction for the Li81 Fab can account for the complex banding pattern observed by SDS-PAGE, as we would expect to see products that contain LINGO-12Li81 Fab2, LINGO-11Li81 Fab2, LINGO-12Li81 Fab1, LINGO-11Li81 Fab1, LINGO-12, free LINGO-1, and free Li81 Fab that are released from the cross-linked complex following treatment with SDS.
When cells expressing full-length LINGO-1 were treated with Li81 Fab, cross-linked, and analyzed by SDS-PAGE with Western blot detection (Fig. 2E), Li81 Fab treatment again prevented the formation of the tetramer observed in the absence of Fab. Instead, bands migrating at the positions of LINGO-1 monomer, dimer, and intermediate between dimer and tetramer were detected. The absence of LINGO-1 tetramer in cross-linked LINGO-1 following Li81 Fab treatment indicates that Li81 interferes with the ability of the protein to form a homotetramer. Fab treatment of the cells over a range of concentrations from 1.2 to 40 μg/ml and with incubation times ranging from 0.25 to 48 hours did not lead to a change in surface levels of LINGO-1 (Supplemental Fig. 3). The larger percentage of LINGO-1 monomer in the cross-linked LINGO-1–Li81 Fab sample on cells (Fig. 2E) versus in solution (Fig. 3D) may have been caused by a drop in the effective concentration of BS3 due to reaction of the cross-linker with the large pool of amine-containing lipids in the membrane. From titration studies of BS3 with LINGO-1–Li81 Fab complex in solution, we found that a decrease in the BS3 concentration from 1 to 0.6 mM could account for the change in the profile (data not shown).
Structure of the LINGO-1–Li81 Fab Complex.
The structure of the LINGO-1–Li81 Fab ectodomain complex was solved to 3.23-Å resolution by molecular replacement using the coordinates of LINGO-1 as the search model (Mosyak et al., 2006). After placing the LINGO-1, it was possible to locate the Li81 Fab in a subsequent molecular replacement search. There is a single copy of the LINGO-1–Li81 Fab complex in the asymmetric unit. Clear electron density was observed throughout the complex, with the exception of two N-terminal residues of LINGO-1 outside the binding interface. The structures of the LRR and Ig domains of LINGO-1 are essentially the same as in previous studies (Mosyak et al., 2006); however, the positioning of the Ig domain relative to the LRR is rotated by ∼15°. The Li81 Fab binds to a discontinuous epitope located at the convex face of the LRR domain at LRRs 4–8 (Fig. 4A). A 2Fo-Fc electron density data map showing electron density for side chains at the LINGO-1–Li81 Fab interface is shown in Supplemental Fig. 4. There are 12 residues from LINGO-1 and 20 residues from the Li81 Fab that are within 4 Å of the binding partner. The epitope is formed by residues Tyr122, Gln125, Arg146, Gly150, Asn152, His174, His176, Lys197, Arg198, Leu199, Tyr200, and Arg201 of LINGO-1 and contains a combination of hydrophobic, hydrophilic, and positively charged residues. The paratope of the Li81 Fab has 44% contributed by the heavy-chain variable domain and 56% by the light-chain variable domain. The key interactions made by the Fab are exclusively from the complementarity-determining regions (CDRs), with the major specificity-determining contacts coming from CDRs H2, H3, L1, and L3, though all of the CDRs contribute to binding. Cooperatively they form 18 direct antibody-antigen hydrogen bonds, 10 salt bridge interactions, and extensive hydrophobic contacts (Fig. 4B). Upon Li81 Fab binding, about 796 Å2 of solvent-accessible surface is buried on each side of the LINGO-1–Li81 Fab interface with a shape complementarity statistic of Sc = 0.74. Collectively these metrics are all consistent with the high-affinity binding observed. The binding site of the Fab overlaps with contact regions in LINGO-1 seen in the tetramer structure (Fig. 4, C and D), and therefore binding of the Fab would be expected to prevent the formation of the LINGO-1 tetramer.
A Model for the Structure of the LINGO-12Li81 Fab2 Complex.
The crystal structure revealed two possible assemblies for the LINGO-1–Li81 Fab complex based on crystal symmetry (Fig. 5A). In Model 1, the N-cap of the LRR from one LINGO-1 molecule is bound to the C-cap of the LRR on a second molecule of LINGO-1, which leads to a buried surface area of 595 Å2 at dimer interface, whereas in the Ig to LRR version (Model 2), the buried surface area at dimer interface is nearly twice that at 1012 Å2. Figure 5B shows the LINGO-1–LINGO-1 contact sites from the two models and from the homotetramer structure overlaid onto the LINGO-1 protomer structure. Despite the large interfaces and their close proximity in the structure, none of the same contact sites are used for binding in the three structures. The most striking difference between the models is that the Ig domain is exposed in Model 1 and buried in Model 2. A series of biochemical and functional studies were designed to distinguish between these possibilities. First, we used limited proteolysis as a probe as proteolytic susceptibility can be a very sensitive method for detecting changes in structure (see, for example, Wen et al., 1996). Endoproteinase Lys-C selectively cleaves LINGO-1 within the stalk region at Lys488 at low enzyme concentrations (Fig. 1B, lane 5), but at higher concentrations and longer incubation additional cleavage occurs at two sites within the Ig domain, Lys392 and Lys424. The identity of these cleavage sites was first determined by mass spectrometry but can be readily monitored by SDS-PAGE. As shown in Fig. 6, when the LINGO-1 ectodomain was treated with endoproteinase Lys-C, ∼80% of the protein was selectively cleaved at the sites within the Ig domain to generate the 55-kDa fragment. The 55-kDa fragment, which contains the entire LRR domain, was resistant to proteolysis by endoproteinase Lys-C under cleavage conditions used. In contrast, when the LINGO-1–Li81 Fab complex was treated with endoproteinase Lys-C, only 15% of the LINGO-1 was cleaved at these sites and there was no accumulation of the 55-kDa band at higher enzyme concentrations. This result supports the structure seen in Model 2, in which the Ig domain is buried and not accessible to the enzyme, as we would anticipate more extensive cleavage at the 55-kDa site if Model 1 were used.
We further investigated the structure of the LINGO-1–Li81 Fab complex using the noncompetitive 1A7 antibody that binds within the Ig domain of LINGO-1 as a probe (Supplemental Fig. 5). The 1A7 binding epitope contains residue Arg386, noted in Fig. 4A, as it is the single R386H amino acid change that leads to the 100-fold-lower affinity of 1A7 for rat versus human LINGO-1 (data not shown). Residue Arg386 is in the solvent-facing ABDE sheet of the Ig domain that is prominently exposed in the published LINGO-1 crystal structure and far removed from the Li81 binding site (Fig. 7A). Biacore analysis confirmed that the Li81 and 1A7 mAbs were noncompetitive, as simultaneous binding of both antibodies to the LINGO-1 ectodomain was observed (Fig. 7B). Consistent with the presentation of the Ig ABDE surface seen in the LINGO-1 crystal structure, 1A7 bound to the cross-linked tetramer, producing a shift in the SEC elution profile of cross-linked LINGO-1 from 270 to 460 kDa (Fig. 7C). In contrast, when the LINGO-1–Li81 Fab complex was treated with 1A7, there was no shift in mobility, indicating that 1A7 was unable to bind the complex (Fig. 7D). Thus, 1A7 is capable of binding to the binary Li81 Fab–LINGO-1 1:1 complex used in the Biacore assessment, but not to the 2:2 complex generated by solution binding of Li81 to LINGO-1, further supporting the structure seen in Model 2. 1A7 was the first anti-LINGO-1 antagonist antibody we generated (Fu et al., 2008; Mi et al., 2009), but was later replaced with the higher-affinity Li81 mAb when it became available. 1A7 Fab forms a 1:1 complex with the LINGO-1 ectodomain where it directly binds the Ig domain of LINGO-1 to block function, which is quite distinct from the mechanism of action of Li81, where the assembly of the 2:2 LINGO-1–Li81 Fab complex causes the ABDE surface of the Ig domain to be buried, thereby masking the 1A7 binding epitope (discussed below). In the same study, the binding of Li81 Fab to the cross-linked LINGO-1 homotetramer was also analyzed (Supplemental Fig. 6). Whereas the binding of 1A7 was very efficient and from the shift in mass we determined that the LINGO-1 homotetramer bound four 1A7 Fabs, the homotetramer bound only one Li81 Fab. The inefficient binding despite the higher affinity of Li81 for LINGO-1 than 1A7 is consistent with the overlap in contact sites for LINGO-1–Li81 Fab and LINGO-1–LINGO-1.
Third, we used oligodendrocyte progenitor cell (OPC) differentiation as a functional readout to probe the structure of the LINGO-1–Li81 Fab complex. The biology of LINGO-1 pathway modulators has been studied extensively using in vitro OPC differentiation and OPC/DRG coculture assays. LINGO-1–Fc treatment in both assays leads to dramatic morphologic changes and increased expression of MBP that are characteristic of OPC maturation (Mi et al., 2005). Previously we showed that the tripeptide sequence RKH within the LINGO-1 Ig domain contributed to the binding of LINGO-1 to Nogo receptor interacting protein-1 (NgR1) (Mi et al., 2010). Here we expanded that analysis to investigate if the RKH sequence is also responsible for the activity of LINGO-1 in the OPC/DRG coculture assay. Figure 8A shows three new constructs that were generated containing the LINGO-1 LRR domain alone, Ig domain alone, and Ig domain in which the RKH sequence was replaced with EKV by mutagenesis. Figure 8B shows that the LINGO-1 Ig domain–Fc fusion protein containing the RKH sequence promoted MBP expression, whereas the same construct in which the RKH sequence was mutated was inactive. The same result was obtained using a cyclic peptide containing either RKH or the mutated sequence. The cyclic peptide containing RKH promoted MBP expression, and the corresponding EKV-containing peptide did not. The observations that the Ig domain–Fc construct was equipotent with the full ectodomain version whereas the LRR-alone construct had minimal activity indicate that the Ig domain is a key determinant for the OPC differentiation activity of LINGO-1 and that the RKH tripeptide is important for the response. The activity of the LINGO-1–Li81 Fab complex was evaluated in the 3-day OPC differentiation assay out of concern that the complex might dissociate or degrade during the longer 14-day incubations used in the coculture study. When the LINGO-1–Li81 Fab complex was assayed in the OPC differentiation assay, there was a >10-fold reduction in activity seen with LINGO-1 ectodomain or Li81 Fab alone (Fig. 8C). The absence of a response at the 6-μg/ml dose of complex, which contains efficacious amounts of 3 μg/ml each of the Fab and LINGO-1 ectodomain, indicates that the complex is very stable as dissociation to free Fab and LINGO-1 would lead to MBP expression. This finding also supports Model 2, where the complex would lead to the RKH motif being buried. Together these studies reveal that Li81 binding induces a rearrangement in the oligomeric state of LINGO-1 from the homotetramer seen in the crystal structure in the absence of the Fab, in which the Ig domain is exposed, to our heteromeric LINGO-12Li81 Fab2 complex, where the Ig domain of LINGO-1 is buried, which was inactive when the complex was exogenously added to the OPC differentiation assay. The ability of the Li81 binding to inactivate LINGO-1 by assembling the LINGO-12Li81 Fab2 tetramer is a novel mechanism of action that is distinct from typical blocking mechanisms used by most antagonist antibodies.
Transmission electron microscopy provided further evidence of the arrangement of the Fab and LINGO-1 in the LINGO-1–Li81 Fab complex. The EM image shown in Fig. 9A revealed a highly homogeneous preparation of the complex with an average particle size of ∼15 nm, consistent with the predicted ∼18 nm for either model based on the crystal structure. Different orientations of the complex were evident in the micrograph, making it amenable to more in-depth analysis. Reference projections generated from 90°C transpositions of the two crystal structure models are shown in Fig. 9, B and C. Manual inspection of the raw EM particles and the reference projections allowed us to correlate EM data with the crystallographic data. The three most represented particles in the EM image match projections a–c from Model 2. Conversely, the distinctive projections expected from Model 1 are absent in the micrograph. The high representation of structures a–c in the EM image support biochemical and functional analysis showing that LINGO-1 and Li81 Fab are organized in the complex shown in Model 2. Approximately 30% of the particles could not be assigned to any of the shown structures. When particle shapes were interpreted using the more extensive collection of 15° projections (24 per model), over 95% of the shapes were classified as Model 2 and none supported Model 1 (data not shown). The structure of the LINGO-12Li81 Fab2 complex was confirmed by elucidating its three-dimensional volume by EM (Supplemental Fig. 7). The EM structure was produced by subtomogram averaging (single-particle tomography) to a resolution of ∼30 Å and agrees with the Model 2 complex observed in the crystal structure (0.5 Fourier Shell Correlation criterion).
Li81 mAb (BIIB033) is a fully human anti-LINGO-1 antibody in clinical trials for the treatment of MS. Here we used extensive biochemical, biophysical, and functional studies to investigate the mechanism of action of the antibody. The crystal structure of the LINGO-1–Li81 Fab complex revealed that the antibody bound to the convex surface of the LRR domain within LRRs 4–8. All of the LINGO-1–LINGO-1 contacts that contributed to the tetrameric structure in the absence of Fab were lost in the LINGO-1–Fab complex, thus revealing that Li81 binding interfered with the ability of LINGO-1 to oligomerize. Indeed, when transfected cells expressing full-length LINGO-1 or soluble LINGO-1 were cross-linked alone or in the presence of the Fab, we found that Li81 Fab treatment prevented and/or disrupted the formation of LINGO-1–LINGO-1 oligomers. The high affinity of the Li81 for LINGO-1 and conversely the low affinity of LINGO-1 for itself would drive the disruption of LINGO-1 oligomers. Whereas the paper by Mosyak et al. (2006) demonstrated tetramer formation of LINGO-1 ectodomain in solution, our study is the first to show the presence of LINGO-1 tetramers on cells expressing full-length LINGO-1. The ability of LINGO-1 to form tetramers on transfected cells in the absence of ligands or coreceptors indicates that oligomer formation is an intrinsic property of LINGO-1.
The oligomeric state of LINGO-1 in solution is dynamic. Monomeric, dimeric, and tetrameric states were readily detected using SEC, DLS, SDS-PAGE, and cross-linking analysis, consistent with published results (Mosyak et al., 2006). A KD value of 0.6 μM for LINGO-1 dimerization was determined by Biacore, and tetramer was observed by DLS at concentrations of >10 μM. In contrast, only a single species was observed in the presence of Li81 Fab. An unexpected feature of the LINGO-1–Li81 Fab complex was that it assembled into a tetramer containing two Fabs and two LINGO-1 subunits rather than the expected 1:1 association. Higher-order binding was also apparent using the Li81 mAb, but because each Fab forms a 2:2 association with LINGO-1 and each mAb contributes two Fabs for binding, bivalent binding through the mAb produced a heterogeneous series of complexes with apparent masses by SEC/light scattering of 500, 1000, and 2000 kDa (data not shown). Although the 500-kDa product is consistent with the assembly of a 2:4 mAb:LINGO-1 complex, we were unable to definitively identify the stoichiometry of this or of the higher mass forms.
The crystal structure revealed two possible arrangements of the LINGO-1–Li81 Fab subunits. Biochemical and functional studies support Model 2, in which the LRR-Ig domain contacts stabilize the structure and the LINGO-1 Ig domain is buried. In addition, we used the PISA algorithm (Krissinel and Henrick, 2007) as an interactive tool to explore the protein interfaces used in the two models. PISA analysis also predicted that the LINGO-1–LINGO-1 interface of Model 2 would be more stable in solution based on the more extensive buried surface area (595 and 1012 Å2 for Models 1 and 2, respectively) and stabilization by 10 hydrogen bonds. The Model 2 interface extends beyond LINGO-1–LINGO-1 contacts, with a significant contribution coming from a secondary interface between the LINGO-1 Ig domain and the constant region of the Fab (494-Å2 buried surface area), which presumably participates in the formation and/or stabilization of the higher-order complex. The 15° rotation in the LINGO-1 Ig domain observed in the LINGO-1–Li81 Fab complex is likely to be a structural prerequisite for the assembly observed in Model 2. Neither of the LINGO-1 dimeric assemblies observed in the LINGO-1–Li81 Fab structure is related to the ring-shaped LINGO-1 tetramer since the specific LRR-LRR and LRR-Ig contact points are different than those that stabilize the LINGO-1 tetramer (Figs. 4D and 5B); thus they are not expected to be intermediate conformations prior to activation. The LINGO-1–LINGO-1 contact points observed in the homotetramer and LINGO-1–Li81 Fab complex (Figs. 4D and 5B) may also contribute to interactions between LINGO-1 and its ligands and coreceptors. This type of situation where the same interface can be used for multiple interactions such as ligand binding and multimer formation has been reported for other LRR-containing proteins (Nose et al., 1992; Karaulanov et al., 2006; Seabold et al., 2008; Kajander et al., 2011).
A striking feature of LINGO-1 is the high protein sequence identity across species (human and rat LINGO-1 share 99.5% identity). In contrast, LINGOs 1–4 share only 44–61% identity. Despite low sequence identity, the predicted structures for LINGOs 2–4 are similar as all five major structural elements (LRR, Ig, stalk, transmembrane region, and cytoplasmic tail) are conserved. Many of the contact sites that contribute to LINGO-1–LINGO-1 binding are preserved in LINGOs 2–4, and one would thus infer that the dynamic oligomerization features we studied with LINGO-1 may apply to LINGOs 2–4. Similarly, the six N-linked glycosylation sites present within the LRR of LINGO-1, which restrict surfaces available for interactions between LINGO-1 and itself and its binding partners, are conserved. Whereas most LRR proteins use the concave surface for ligand binding and/or dimerization (Bella et al., 2008; Kajander et al., 2011), N-linked glycans on amino acid residues Asn254 and Asn302 in LINGO-1 block this surface and only the convex face is available for binding. We anticipate this will be true for LINGOs 2–4.
Structure-function studies using OPC differentiation to identify functional epitopes within the LINGO-1 ectodomain have confirmed and extended our understanding of how the LRR and Ig domains contribute to the biology of LINGO-1. Most significantly, we discovered that the Ig domain is a potent pathway antagonist and that the Ig-containing EKV mutant is inactive, revealing the importance of the RKH tripeptide to LINGO-1-dependent oligodendrocyte biology. One explanation for the RKH-mediated activity is that the soluble Ig domain competes for the binding of full-length LINGO-1 for formation of its signaling complex. Consistent with this notion, Bourikas et al. (2010) showed that a construct containing the Ig domain, stalk region, transmembrane region, and cytoplasmic tail but lacking the LRR region retained the activity of full-length LINGO-1 in inhibiting OPC differentiation. The same construct also supported homotypic LINGO-1–LINGO-1 binding (Stein and Walmsley, 2012). How the LRR domain contributes to the biology is less clear, but perhaps it contributes to oligomerization to increase avidity for signaling partners and/or directly binds and increases the affinity of the interactions. Mosyak et al. (2006) provided a detailed discussion of potential contributions of oligomer formation to LINGO-1 biology. The low-level activity seen in the OPC/DRG coculture assay with the LRR-alone construct (Fig. 8B) could reflect either of these scenarios. Recently, Jepson et al. (2012) published results showing that LINGO-1 forms intercellular contacts and can self-associate in trans, which is consistent with OPC/DRG coculture data we previously published showing that loss of LINGO-1 on either cell type alone or both promoted myelination (Lee et al., 2007). How Li81 at a molecular level affects the ability of LINGO-1 to interact with itself in cis (intracellular) and trans (intercellular) remains to be determined.
Of known ligands and coreceptors of LINGO-1 (Mi et al., 2013), the binding to NgR1 and p75 to form the LINGO-1–NgR1–p75 signaling complex that regulates neuronal survival through RhoA activation has been most studied (Mi et al., 2004; Mosyak et al., 2006; McDonald et al., 2011; Saha et al., 2011). Epitope mapping data showed that the LINGO-1 LRR- and Ig-alone constructs bound NgR1 with low affinity (EC50 values of 120 and 60 nM, respectively, versus 6 nM for the intact ectodomain), that the RKH-to-EKV mutation within the intact ectodomain led to a 20-fold reduction in EC50, and that an LRR-only construct of NgR1 bound LINGO-1 more weakly than longer versions of NgR1 containing the stalk region (Mi et al., 2010). Together these studies reveal that LINGO-1–NgR1 interactions use contacts over large surface areas of both proteins and that the RKH within the Ig domain contributes to the binding. Bourikas et al. (2010) also used a domain truncation analysis to map binding to p75 and showed that the LRR domain of LINGO-1 did not contribute but that the Ig domain and stalk region were sufficient. The apparent KD of ∼1 μM for monovalent binding of LINGO-1 ectodomain to soluble NgR1 deduced by Biacore (Mosyak et al., 2006) is of similar affinity to the binding of LINGO-1 to LINGO-1 we determined (Fig. 2). Higher affinities were reported using cell surface binding and enzyme-linked immunosorbent assay methods where avidity can lead to higher apparent affinities (Mi et al., 2004, 2010; Shao et al., 2005). Other studies have shown that Troy can substitute with p75 to form a LINGO-1–NgR1–Troy signaling complex (Park et al., 2005; Shao et al., 2005). Recently, Ahmed et al. (2013) showed that AMIGO3, a distantly related LRR-Ig protein involved in CNS axon growth inhibition, also forms NgR1–p75 and NgR1–Troy signaling complexes. While there is little sequence identity between LINGO-1 and AMIGO3, structural data for AMIGOs 1–3 suggest the importance of the Ig domain in ligand binding (Kajander at al., 2011).
Three notable differences were observed in the structure-activity relationship studies using NgR1 binding and OPC/DRG coculture readouts. First, the intact ectodomain had a 10-fold-higher apparent affinity for NgR1 than the Ig domain–only version, but was equipotent in the cell-based assay. Second, LINGO-1 Ig domain constructs containing the wild-type RKH sequence had a 2-fold-higher apparent affinity for NgR1 than the EKV mutant, but produced a 100-fold increase in activity in the cell-based assay. Third, LINGO-1 full ectodomain constructs containing the wild-type RKH sequence had a 10-fold-higher apparent affinity for NgR1 than the same reagent with the EKV mutation, but only a 2-fold-higher activity in the cell-based assay. There is no simple explanation that can account for all these differences. Because the signaling partner that drives OPC differentiation is unknown, one possibility is that LINGO-1-binding epitopes for this receptor and NgR1 could be different, which can only be addressed when the relevant OPC receptor(s) are identified. Although our studies provide structural evidence that Li81 binding disrupts contacts used in the assembly of LINGO-1 oligomers, further studies are needed to assess the effect of Li81 treatment on the heteromeric complexes that form between LINGO-1 and its other signaling components (Mi et al., 2013; Lee et al., 2014).
LRR and Ig domains are common structural motifs used in protein-protein binding (McEwan et al., 2006; Bella et al., 2008). An interesting feature of the LINGO-1 structure is the large number of LINGO-1–LINGO-1 contact sites in the tetramer structure and their ability to reassemble with alternative contacts in the LINGO-1–Li81 Fab complex. Studying low-affinity LRR-Ig protein-protein interactions that use an avidity component for function is challenging. The data we have generated for LINGO-1 should provide a framework for further studies designed to understand the molecular interactions between LINGO-1 and its coreceptors.
The authors thank Greg Thill, Joe Amatucci, and Shelly Martin for cell line selection and generating conditioned medium; Allan Capili, Lee Walus, David Mo, Sheng Gu, Dingyi Wen, and Craig Wildes for biochemistry support and helpful discussion; Alexey Lugovskoy for generating molecular models of the RKH epitope and LINGO-1–1A7 complex; Hernan Cuervo for preparation of the synthetic peptides; Dyax Corp. (Burlington, MA) for their antibody discovery efforts behind the identification of Li81; Kasim Sader, Lingbo Yu, and Erik Franken at FEI for generating the three-dimensional EM structure of the LINGO-12Li81 Fab2 complex; and Ajay Verma and Diego Cadavid, Biogen Idec, for critical reading of the manuscript. Use of the Advanced Photon Source was supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357. Use of the Lilly Research Laboratories Collaborative Access Team (LRL-CAT) beamline facilities at Sector 31 of the Advanced Photon Source was provided by Eli Lilly & Company, which operates the facility.
Participated in research design: Pepinsky, Arndt, Quan, Gao, Quintero-Monzon, Mi.
Conducted experiments: Pepinsky, Arndt, Quan, Gao, Lee, Quintero-Monzon, Mi.
Contributed new reagents or analytic tools: Pepinsky, Arndt, Quan, Gao.
Performed data analysis: Pepinsky, Arndt, Quan, Gao, Lee, Quintero-Monzon, Mi.
Wrote or contributed to the writing of the manuscript: Pepinsky, Arndt, Gao, Quintero-Monzon, Mi.
- bissulfosuccinimidyl suberate
- complementarity-determining regions
- Chinese hamster ovary
- central nervous system
- dynamic light scattering
- dorsal root ganglion
- electron microscopy
- leucine-rich repeat and Ig-containing Nogo receptor interacting protein-1
- leucine-rich repeat
- monoclonal antibody
- myelin basic protein
- multiple sclerosis
- Nogo receptor interacting protein-1
- oligodendrocyte progenitor cell
- phosphate-buffered saline
- radioimmunoprecipitation assay
- size exclusion chromatography
- Copyright © 2014 by The American Society for Pharmacology and Experimental Therapeutics