Structural Insights into Recognition of Triple-helical β-Glucans by an Insect Fungal Receptor

The innate ability to detect pathogens is achieved by pattern recognition receptors, which recognize non-self-components such as β1,3-glucan. β1,3-Glucans form a triple-helical structure stabilized by interchain hydrogen bonds. β1,3-Glucan recognition protein (βGRP)/Gram-negative bacteria-binding protein 3 (GNBP3), one of the pattern recognition receptors, binds to long, structured β1,3-glucan to initiate innate immune response. However, binding details and how specificity is achieved in such receptors remain important unresolved issues. We solved the crystal structures of the N-terminal β1,3-glucan recognition domain of βGRP/GNBP3 (βGRP-N) in complex with the β1,3-linked glucose hexamer, laminarihexaose. In the crystals, three structured laminarihexaoses simultaneously interact through six glucose residues (two from each chain) with one βGRP-N. The spatial arrangement of the laminarihexaoses bound to βGRP-N is almost identical to that of a β1,3-glucan triple-helical structure. Therefore, our crystallographic structures together with site-directed mutagenesis data provide a structural basis for the unique recognition by such receptors of the triple-helical structure of β1,3-glucan.


EXPERIMENTAL PROCEDURES
Protein Expression and Purification-N-terminal ␤1,3-glucan recognition domain of the Plodia interpunctella ␤GRP gene was amplified by PCR from the original gene. The PCR product (Tyr 1 -Thr 104 ) was cloned into a pCold-MBP-TRX vector with hexahistidine, maltose-binding protein (MBP), and thioredoxin (TRX) tags. This vector is capable of expressing a target protein at low temperature (15°C) using a cold shock promoter cspA (18). It was constructed by insertion of MBP, TRX genes, and a tobacco etch virus protease cleavage sequence into NdeI and BamHI sites of a cold shock vector pCold-I (TaKaRa Bio, Inc.). The construct was composed of the His 6 -MBP-TRX tag and ␤GRP-N with a tobacco etch virus protease cleavage site between them. For the expression of Bombyx mori ␤GRP-N, a DNA fragment encoding Tyr 1 -Ala 102 was cloned into pCold-MBP-TRX vector. Each expression plasmid was transformed into the Escherichia coli strain Rosetta2(DE3)pLysS (Novagen). The transformed cells were grown in LB medium at 37°C and induced with 0.1 mM isopropyl ␤-D-thiogalactoside (Wako) for 24 h at 15°C. The harvested Crystallization-Crystals of the N-terminal domain of ␤GRP were obtained by the sitting drop vapor diffusion method, in which 0.5 l protein solution was mixed with an equal volume of reservoir solution. Plodia ␤GRP-N crystals complexed with laminarihexaose (Seikagaku Corp.) were obtained using the following reservoir solution: 0.2 M lithium chloride, 20% (v/v) PEG 3350 and 5 mM laminarihexaose. The ligand-free crystals were grown in a reservoir solution containing 1.6 M ammonium sulfate, 0.1 M MES monohydrate (pH 6.5), and 10% (v/v) 1,4-dioxane. Bombyx ␤GRP-N crystals complexed with laminarihexaose were obtained in a buffer of 0.1 M ammonium sulfate, 0.1 M HEPES sodium buffer (pH 7.5), 10% (w/v) PEG 4000, and 5 mM laminarihexaose.
X-ray Data Collection and Structure Determination-The diffraction data were collected at the synchrotron radiation source at AR-NE3A, AR-NW12A, and BL5A in the Photon Factory (High Energy Accelerator Research Organization (KEK), Japan). The crystals were cryoprotected with a reservoir solution containing 25% glycerol (Bombyx ␤GRP-N) or 25% ethylene glycol (Plodia ␤GRP-N). The diffraction data were processed using HKL2000 (19). The crystal parameters are shown in Table 1. The structures of Bombyx and Plodia ␤GRP-N⅐laminarihexaose complexes and ligand-free Plodia ␤GRP-N were solved by the molecular replacement method using the program Molrep (20) with Drosophila ␤GRP-N (Protein Data Bank code 3IE4), the solved ligand-free Plodia ␤GRP-N and liganded Bombyx ␤GRP-N as search models, respectively. Model building was manually performed using programs XtalView/Xfit (21) and COOT (22). Refinement was carried out using the programs CNS1.1 (23) and REFMAC5 (24). The stereochemical quality of the final models was assessed by PROCHECK (25). The refinement statistics are summarized in Table 1.
Preparation of Mutant Proteins-Single or double amino acid substitutions of Plodia ␤GRP-N were performed by a PCRbased site-directed mutagenesis technique. The mutant proteins were expressed and purified using the same methods as for wild-type ␤GRP-N. CD spectra of these mutant proteins showed that their overall structure was similar to wild-type (data not shown).
Isothermal Titration Calorimetry-Laminarin binding experiments of wild-type and mutant Plodia ␤GRPs were conducted on an Auto-ITC instrument (MicroCal, Inc.) at 25°C. Laminarin from Laminaria digitata was purchased from Sigma. In a typical isothermal titration calorimetry (ITC) experiment, the cell was filled with 200 l of 50 M protein solution dissolved in 10 mM HEPES, 150 mM NaCl (pH 7.4). The protein sample was titrated with successive injections (2 l) of laminarin solution (0.67 mM as a triplex) in the same buffer. Reference experiments were performed by injecting laminarin solution into buffer. The experimental data were fitted using the program Origin (version 7.0, OriginLab).

RESULTS
Overall Structure of ␤GRP-N-After extensive screening, we were able to crystallize the N-terminal ␤-glucan-binding domains (␤GRP-N) of the moths P. interpunctella and B. mori in the presence of laminarihexaose and then solved the crystal structures at 2.20 and 2.05 Å resolution, respectively (  Table 1). We also obtained crystals of the ligand-free ␤GRP-N of Plodia and solved its structure at 1.58 Å resolution ( Table 1). The Plodia and Bombyx ␤GRP-N structures each have an immunoglobulin-like ␤-sandwich fold composed of two antiparallel ␤-sheets containing three and five ␤-strands (concave ␤-sheet, ␤1-2 and ␤5; convex ␤-sheet, ␤3-4 and ␤6 -8). There is little structural difference between liganded and unliganded forms of Plodia ␤GRP-Ns (root mean square deviation ϭ 0.49 Å for superimposed 98 C␣ atoms). In addition, the overall structure of Plodia ␤GRP-N is almost identical to the liganded Bombyx ␤GRP-N (root mean square deviation ϭ 0.35 Å for superimposed 99 C␣ atoms) and unliganded crystal (Drosophila) or NMR (Bombyx) structures (root mean square deviation ϭ 0.60 and 1.25 Å for superimposed 96 and 85 C␣ atoms, respectively) (16,17). Taken together, these results indicate that ␤GRP-Ns assume a common structural fold irrespective of the binding of ligand.
Structure of Laminarihexaoses-The crystal of laminarihexaose-bound Plodia ␤GRP-N belongs to space group P2 1 2 1 2 1 with two proteins and two laminarihexaoses per asymmetric unit. Two laminarihexaose helices are related by noncrystallographic 4-fold screw symmetry around crystallographic 2 1 screw axis, thus forming a pseudoquadruplex structure (supplemental Fig. S2A). The crystal of laminarihexaose-bound Bombyx ␤GRP-N belongs to space group P4 1 with two proteins and two laminarihexaoses per asymmetric unit. Two laminarihexaose helices are related by crystallographic 4-fold screw symmetry around the 4 1 screw axes respectively, forming pseudoquadruplexes structures (supplemental Fig. S2B). In both Plodia and Bombyx complexes, three (chains A, B, and C) of four laminarihexaoses interact with one ␤GRP-N molecule (structures of the Plodia complex shown in Fig. 2A and of Bombyx in supplemental Fig. S3A). The structure of the laminarihexaoses can be compared with a model of righthanded triple-helical ␤-glucan derived from x-ray fiber diffraction (2). In the case of the laminarihexaoses, interstrand hydrogen bonds are observed between O2 atoms (Fig. 2B, top, and supplemental Fig. S3B, top). The O2 atoms of Glc-1(C), Glc-2(C), and Glc-3(C) make hydrogen bonds with the O2 atoms of Glc-4(B), Glc-5(B), and Glc-6(B), whereas the O2 atoms of Glc-1(B), Glc-2(B), and Glc-3(B) bond with those of Glc-4(A), Glc-5(A), and Glc-6(A), respectively. These hydrogen-bonding patterns are very similar to those in the ␤-glucan structure, which is also stabilized via interchain hydrogen bonds between O2 atoms (2) (Fig. 2B, bottom, and supplemental Fig. S3B, bottom). In addition to the interstrand hydrogen bond interactions, the helical structure of laminarihexaoses is formed by intrastrand hydrogen bonds between their O4(i) and O5(iϩ1) atoms, as is also observed in the ␤-glucan structure, and by water moleculemediated hydrogen bonds between O4(i) and O6(iϩ1) (Fig. 2C  and supplemental Fig. S3C). The inter-and intrastrand hydro-    AUGUST 19, 2011 • VOLUME 286 • NUMBER 33 gen bonds of the laminarihexaoses contribute to stabilize their highly ordered helical structure, an arrangement almost certainly induced in association with ␤GRP-N to mimic the triplex structure of the ␤-glucan. In fact, the similar bonding patterns in the quadruplex and triplex structures unexpectedly derive from an almost identical spatial arrangement of six proteinbinding residues in the two ligands (see below). Laminarihexaose-␤GRP-N Interaction-Ligand-binding of ␤GRP-N is attained through a convex ␤-sheet and a characteristic long loop between ␤3 and ␤4, and a short loop between ␤6 and ␤7 ( Fig. 1 and supplemental Fig. S1). Binding interactions in the proteins from both insect species are almost identical, suggesting a common binding mode for this entire protein family. Six Glc residues from three laminarihexaose chains, Glc-5(A), Glc-6(A), Glc-3(B), Glc-4(B), Glc-1(C), and Glc-2(C), are involved in binding to ␤GRP-N ( Fig. 2A), and eight amino acid residues including His 31 , Leu 41 , Asp 49 , Trp 76 , Tyr 78 (Phe 78 in Bombyx), Gly 83 , Gly 85 , and Arg 87 show extensive polar and nonpolar interactions (Fig. 3A and supplemental Fig. S4A). The structure around the ligand-binding site is almost identical in liganded and unliganded forms (Fig. 3B), indicating that the site is rigid, and the surface is preformed. The side chain of Asp 49 interacts simultaneously with laminarihexaoses B and C. The two carboxyl oxygens of the Asp 49 side chain bind to the Glc-4(B) O4, O6, and Glc-2(C) O6 atoms via hydrogen bonds. There is a hydrophobic patch, including Trp 76 and Tyr 78 (Phe 78 in Bombyx) alongside the ligand-binding site ( Fig. 3C and supplemental Fig. S4B), and it possibly forms part of the binding site or plays a role in the initial encounter with a longer, structured ␤1,3-glucan. Most of the residues involved in binding to the laminarihexaoses are conserved among invertebrate ␤GRP-Ns (Fig. 4).

Recognition of Triplex ␤-Glucan by Insect Receptor
Verification of Interaction in Solution-To verify that these intermolecular interactions occur in solution and with ␤-glucan, we performed a mutational analysis of Plodia ␤GRP-N. A series of single/double mutants was prepared (H31A, L41A, R48A, D49A, W76A, Y78A, R87A, H31A/R48A, and R48A/ R87A), focusing on the residues located on the ligand-binding surface (Fig. 3). We examined the affinities of wild-type and mutant ␤GRP-Ns to laminarin, a soluble ␤1,3-glucan with a triple-helical structure (5), using isothermal titration calorimetry (supplemental Fig. S5 and summarized in Fig. 5). The concentration of laminarin solution is calculated as a triplex. The binding constant between wild-type Plodia ␤GRP-N and laminarin is 2.2 ϫ 10 6 M Ϫ1 , in agreement with the binding constant of Drosophila ␤GRP-N to laminarin (2.1 ϫ 10 6 M Ϫ1 ) (16). The apparent stoichiometry (n ϭ 1.5) observed in the ITC experiment may reflect the insufficient triple-helical structure of some part of laminarin due to its heterogeneity in length and linkage (26). Mutations D49A, W76A, and H31A/R48A abol-ished binding and H31A, L41A, R87A, and R48A/R87A decreased the affinity, indicating that His 31 , Leu 41 , Arg 48 , Asp 49 , Trp 76 , and Arg 87 are important for the binding to triplex ␤-glucan in solution. The essential nature of Asp 49 is seen in the extensive hydrogen bonds between two laminarihexaoses and the Asp 49 side chain (Fig. 3A). The importance of Trp 76 has been reported previously in Drosophila ␤GRP-N (Trp 77 in Drosophila) (16). Trp 76 directly interacts with two laminarihexaoses through hydrophobic interactions and builds the ligand-binding surface through extensive hydrophobic contacts with ligand-binding residues, including His 31 , Leu 41 , Tyr 78 , and Arg 87 (Fig. 3A). Double mutations of Arg 48 and His 31 or Arg 87 decrease the affinity for laminarin. Although Arg 48 does not directly interact with laminarihexaoses, the potential rotamers of the side chain without crystal packing constraints are within hydrogen-bonding distances to the laminarihexaose(s). Taken together, we suggest that the Arg 48 side chain does in fact interact with ␤-glucan in solution. We constructed a model of ␤GRP-N⅐triplex ␤1,3-glucan complex based on our crystal structure ( Fig. 6 and supplemental Fig. S6). In this model, the relative spatial arrangement of six binding Glc residues from the ␤GRP-N⅐laminarihexaoses complex is almost identical with that of the corresponding Glc residues of the ␤1,3-glucan, and there are no significant steric clashes between ␤GRP-N and ␤1,3-glucan (2). From these observations, we conclude that the type of interaction between ␤GRP-N and laminarihexaoses in the crystal is applicable to that between ␤GRP-N and triplex ␤1,3-glucan in solution.

DISCUSSION
The biological effect of ␤-glucan in relation to its tertiary structure is one of the fundamental immunological issues (7). Although many efforts have been made to reveal the structurefunction relationship (4,12), there is still no consensus on the  basic requirements of the biological activity of ␤-glucan. This is in part due to insufficient characterization of the tertiary structure of ␤-glucan in association with its counter receptor. Here, we reveal the structural basis of triplex ␤-glucan recognition by ␤GRP. Three strands of the triplex ␤-glucan can simultaneously interact with particular amino acid residues on the convex ␤-sheet. It clearly explains why ␤GRP binds specifically to the ␤-glucan triplex structure and not to shorter glucose oligomers, which cannot assume a triple-helical conformation in solution.
In a typical carbohydrate-protein interaction, the actual region of contact between the carbohydrate and the proteins involves only one to three monosaccharide residues (27). As a consequence, carbohydrate-binding proteins tend to be of relatively low affinity. To mediate biologically relevant interaction, many carbohydrate-binding proteins are oligomeric and achieve high affinity binding by making multiple interactions with multivalent ligands (27,28). Another possible strategy to gain the affinity is to have an extended carbohydrate-binding site. ␤GRP seems to adopt the latter strategy. A unique interaction mode is attained using an extensive binding surface of the ␤GRP-N protein. The unexpectedly large contact area of ␤GRP-N reaps a benefit of high affinity binding to triple-helical ␤-glycan using one carbohydrate recognition domain. Large binding platform of ␤GRP-N is constituted by hydrophobic and charged residues mostly on the convex ␤-sheet. A solvent-exposed hydrophobic patch is formed by Leu 27 , Leu 41 , Trp 76 , Tyr 78 (Phe 78 in Bombyx), and Ile 80 , thus contributing to the binding to ␤-glucan. Glucose rings are obliquely oriented onto the side chains of Trp 76 and Tyr 78 and seem to exhibit weak hydrophobic interactions. There are no parallel stacking interactions between glucose rings and aromatic side chains as often observed in protein-carbohydrate interactions (27,29). This is well explained from the structure of triple-helical ␤-glucan, in which hydrophobic surface of the glucose ring is not exposed to solvent and losing an ability to have a parallel stacking interac-tion with aromatic ring. Polar and charged residues (His 31 , Arg 48 , Asp 49 , and Arg 87 ) are surrounding this hydrophobic patch and involved in the interaction with O4 and O6 atoms of glucose residues. The residues, Asp 49 (␤4) and Trp 76 (␤6) are positioned to interact with more than one ␤-glucan strand, thus contributing to stabilize the ␤-glucan helix. The essential feature of these residues are supported the ITC experiment, in which mutation of each residue effectively eliminate the binding to laminarin. Takahasi et al. (17) reported the solution structure of the GNBP3 N-terminal domain and proposed the ␤-glucan binding surface as a concave ␤-sheet composed of ␤1-2 and ␤5 strands. On the other hand, our crystallographic and mutagenesis studies show the binding surface to be convex ␤-sheet composed of ␤3-4 and ␤6 -8. The discrepancy may be due to the different sources of laminarin used in the studies. Little structural rearrangement of the side chains were observed upon binding to laminarihexaoses. In common with lectins, ␤GRPs have preformed a carbohydrate-recognition site, which can accommodate the ␤-glucan ligand. This feature seems to minimize the energetic penalty paid upon binding to carbohydrate ligands (29).
The structural basis of triplex ␤-glucan recognition by a ␤-glucan binding protein can now be understood. The interaction mode elucidated in this study will provide the basis for the understanding of structure/activity relationship of other ␤-glucan-binding proteins such as Dectin-1 found in vertebrates (10). Trp 221 and His 223 residues of mouse Dectin-1 are the key residues for the binding to ␤-glucan and for Toll-like receptor 2-mediated cellular activation (30). Although the topology of Dectin-1 carbohydrate recognition domain is different from that of ␤GRP-N (31), these residues may interact with triplehelical ␤-glucan by the non-stacking mode that is observed in the ␤GRP-N⅐laminarihexaose complex.
Full-length ␤GRP consists of a well conserved N-terminal domain and a C-terminal ␤1,3-glucanase-like domain (13,15). The N-terminal domain of ␤GRP plays a critical role for the detection of pathogen. In contrast, the C-terminal glucanaselike domain does not have either glucanase activity or affinity with the ␤1,3-glucan (13,15). The functional role of the C-terminal domain is still not established; however, recent evidence suggests that this domain is required for the activation of Toll signaling pathway by recruiting a downstream modular serine protease (32). Upon infection of fungi into hosts, ␤GRP recognizes the ␤1,3-glucan of the cell wall via ␤GRP-N. In this ␤GRP⅐␤1,3-glucan complex, C-terminal domain of ␤GRP seems located apart from the bound ␤1,3-glucan because the C terminus of ␤GRP-N is opposite to the ␤1,3-glucan-binding surface. It is tempting to speculate that the accumulation and clumping of ␤GRP on the fungal cell wall will expose its C-terminal domains to recruit and active the downstream signaling molecule. The crystal structure of SPN48, one of the serine protease inhibitors (serpins) from Tenebrio molitor, has been determined recently (33). SPN48 has a putative basic heparinbinding site and heparin inhibit a serine protease, Spätzle-processing enzyme, which produces a Toll receptor ligand, Spätzle by bridging SPN48 and the serine protease. Roh et al. (34) demonstrated that all the components essential for the recognition of ␤-glucan and pro-Spätzle were bound to a heparin-immobi-FIGURE 6. Structural model for binding between ␤GRP-N and triplex ␤1,3-glucan. Surface model of Plodia ␤GRP-N is shown and residues involved in binding to laminarihexaoses are colored orange. In the ␤GRP-N⅐laminarihexaose complex, interacting Glc residues are shown as ball-andstick models. Triple-helical ␤1,3-glucan is shown as thin stick models: chain A, yellow; chain B, green; chain C, cyan. lized column. In the ␤GRP-N structure, basic residues (Lys 7 , Lys 52 , Lys 54 , Arg 57 , Arg 61 , and Arg 63 in Plodia) are clustered opposite to the ␤1,3-glucan-binding surface (supplemental Fig.  S7). It is expected that the ␤GRP⅐␤-glucan complex may recruit the heparin or signaling/inhibiting components through the basic patch. Further experimental evidences are necessary to understand the signal transduction and inhibition mechanism.
Polysaccharides such as ␤-glucan recently have been shown to act as potent immunomodulating agents (35). Because the anti-tumor and anti-infective activities of ␤1,3-glucan (35,36) appear to be conformation-dependent, an analysis of ␤-glucanprotein interaction together with our structural information would be useful in the pharmaceutical field. From a diagnostic viewpoint, the presence of ␤-glucan in blood and sterile body fluid is a good marker because mammalian systems cannot produce it. High level of ␤-glucan was reported in the patients with infectious diseases such as invasive aspergillosis (37). Our findings may help the development of a ␤-glucan detection assay for diagnosis and therapeutic monitoring of certain infectious diseases.