Duplicated Binding Sites for (1 3 3)- (cid:1) - D -Glucan in the Horseshoe Crab Coagulation Factor G IMPLICATIONS FOR A MOLECULAR BASIS OF THE PATTERN RECOGNITION IN INNATE IMMUNITY*

The horseshoe crab factor G, a heterodimeric serine protease zymogen, is activated by (1 3 3)- (cid:1) - D -glucan on fungal cell walls. The activation initiates the hemo-lymph-clotting cascade, a critical reaction for the defense against microorganisms. In the present study, we identified the domain responsible for the glucan recognition by factor G and characterized its interaction with (1 3 3)- (cid:1) - D -glucan and its derivatives. Among three do- mains in subunit (cid:2) of factor G, identified as the glucan-binding domain, was the COOH-terminal xylanase Z-like domain composed of two tandem-repeating units, each of which exhibits sequence similarities to the cellulose-binding domains of bacterial xylanases. Each of the single units bound to the glucan with lower affinities, and the association constant increased two orders with the tandem-repeating structure ( K a (cid:3) 8.0 (cid:4) 10 8 M (cid:5) 1 ). In addition to longer glucans, (1 3 3)- (cid:1) - D -glucan oligosaccha- rides

The innate immune system recognizes various pathogens with products of limited numbers of germ line-encoded genes via "pattern recognition" (1). The target molecules for "pattern recognition" are characteristic molecular patterns commonly found on the surface of microorganisms, but not on self (2). Recent studies on the mammalian innate immune systems have revealed that such pathogen-associated molecular patterns (PAMPs) 1 reside in several bacteria-derived molecules, including lipopolysaccharides (LPS), peptidoglycans, lipoproteins/lipopeptides, lipoteichoic acids, CpG DNA, and flagellins (3). Several toll-like receptors have been shown to be essential for the responses to these molecules. There is, however, no evidence for the direct binding of these cell-surface receptors with the microbial molecules. Molecules acting as pattern recognition receptors that directly recognize PAMPs and generate activation signals are still poorly understood.
Some invertebrate animals provide ideal systems for studies on innate immunity, because their defense systems depend solely on innate immunity. A type of hemocyte called granulocytes plays a major role in the innate immunity in horseshoe crabs, which are arthropods (4,5). Exposure of the hemocytes to LPS results in the exocytosis of the intracellular granules, followed by the activation of the hemolymph coagulation system resulting in gel formation. The series of reactions is very important in the defense system as well as hemostasis; the invaded microorganisms are engulfed in the hemolymph clot and finally killed by antibacterial substances released from the granules (6).
The LPS-mediated hemolymph coagulation is a cascade-type reaction composed of three serine protease zymogens, factor C (7), factor B (8), and proclotting enzyme (9), as well as a clottable protein, coagulogen (10). In addition to LPS, (133)-␤-Dglucans induce the clot formation of the hemocyte lysate (11,12). We have succeeded in the purification and characterization of factor G, which initiates the glucan-mediated clot formation (13). The purified factor G is a heterodimeric serine protease zymogen composed of the two non-covalently associated subunits ␣ (72 kDa) and ␤ (37 kDa). In the presence of nanogram quantities of (133)-␤-D-glucans, factor G is autocatalytically activated to an active serine protease, factor G , which then activates proclotting enzyme in the coagulation cascade. The amino acid sequences of the subunits deduced from their cDNA sequences showed that subunit ␤ is a serine protease zymogen and that subunit ␣ is a mosaic protein that contains three types of domains with similarities to bacterial polysaccharide-hydolases (14). The NH 2 -terminal portion of subunit ␣ is similar to Bacillus circulans ␤-1,3-glucanase A1 (the Gln A1-like do-* This work was supported by grants-in-aid for scientific research from the Ministry of Education, Science, Sports and Culture of Japan and the Japan Foundation for Applied Enzymology (to T. M.), by a Sasakawa Scientific Research Grant from the Japan Science Society (to Y. T.), and by grants from the Yamanouchi Foundation for Research on Metabolic Disorders (to T. M.), the Uehara Memorial Foundation (to T. M.), the Protein Research Foundation Peptide Institute (to T. M.), and the Ryoichi Naito Foundation for Medical Research (to T. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Present address: Laboratory for Proteolytic Neuroscience, Brain Science Institute, RIKEN, Wako-shi, Saitama 351-0198, Japan.
** To whom correspondence should be addressed: Tel.: 81-92-642-6103; Fax: 81-92-642-6103; E-mail address: tmuta@mailserver.med. kyushu-u.ac.jp. main). In the middle of the molecule are three tandem-repeating units of 47 amino acids that show partial sequence homologies to carbohydrate-binding proteins, such as Streptomyces lividans xylanase A, Rarobacter faecitabidus protease I, Oerskovia xanthineolytica ␤-1,3-glucanase, and ricin (the Xln A-like domains). Two tandem repeats with 126 amino acids are present in the COOH-terminal domain that exhibits homologies to Clostridium thermocellum xylanase Z (the Xln Z-like domain). Because the activation of factor G is highly sensitive and specific to (133)-␤-D-glucan, it is utilized as a diagnostic reagent to detect fungal infections by measuring a trace amount of (133)-␤-D-glucan (15).
In addition to the horseshoe crab hemolymph coagulation, (133)-␤-D-glucan and its derivatives have also been known to induce activation of several defense reactions in other organisms. In insects and crustaceans, the glucans trigger the prophenoloxidase-activating system, one of the major defense systems in these animals (16). The glucan-mediated activation of defense reactions is not limited in invertebrate animals; they elicit anti-tumor and anti-microbial activities in mammalian immune systems (17,18) and induce the production of phytoalexins in plants (19,20). Because (133)-␤-D-glucans are major components of the cell wall of fungi, such reactions induced by the glucans are likely to be important for the defense against fungi. Several (133)-␤-D-glucan-binding proteins involved in these defense systems were reported in various organisms, such as (133)-␤-D-glucan-binding protein from a crustacean and insects (21,22) and ␤-glucan-elicitor-binding protein from plants (23). However, none of these proteins have known enzymatic activities, and hence the molecular basis of the glucan recognition leading to the generation of the activation signals remains elusive.
The horseshoe crab factor G is the only molecule that has been demonstrated to be directly activated by the glucan. Because this activation can be reconstituted in vitro with the purified protein and the glucan (13), it could provide a good model for investigating how proteins recognize and respond to the glucans. In this report, we identify the glucan-binding domain in factor G and characterize its interaction with (133)-␤-D-glucan. Factor G recognizes and binds to a (133)-␤-D-glucoside linkage through the COOH-terminal domain with a tandem-repeating structure. Multivalent binding of factor G to a simple characteristic structure on the glucan chain exemplifies a simple but specific recognition of PAMPs by pattern recognition receptors.
Expression of Subunits ␣ and ␤ in Insect Cells-cDNAs encoding the entire coding region of subunits ␣ and ␤ were subcloned in the baculovirus transfer vector pVL1392 (BD PharMingen, San Diego, CA). The transfer vectors were co-transfected into Sf9 or Sf21 insect cells with a modified Autographa californica nuclear polyhedrosis virus DNA. Resultant virus pools were collected 4 days later and were plaque-purified and amplified. The insect cells were infected with the recombinant virus and were harvested 72 h after the infection. The cells were homogenized in 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 1 mM diisopropyl fluorophosphate using a glass homogenizer. The homogenized cell lysate was centrifuged, and the resulting supernatant was used for the polysaccharide-binding analyses.
GST fusion proteins were expressed in the Escherichia coli strain BL21(DE3)/pLysS and purified according to the manufacturer's protocol. After dialysis against 20 mM Tris-HCl (pH 8.0) containing 150 mM NaCl (TBS), the fusion protein was digested with human thrombin, and the digest was passed through small columns of glutathione-Sepharose 4B and benzamidine-Sepharose 6B to remove GST and thrombin. The NH 2 -terminal sequences of the obtained proteins were confirmed by gas-phase sequencers, models 473A and 477A (Applied Biosystems). The protein concentrations used for kinetic analysis were determined by amino acid analysis with a Hitachi L-8500 automatic analyzer.
Immobilization of Polysaccharides with AF-Amino Toyopearl 650M-One gram of suction-dried AF-Amino Toyopearl 650M was suspended in 2 ml of 0.2 M K 2 HPO 4 containing 0.3 g of each polysaccharide (laminarin, mannan, and xylan). After the addition of 0.2 g of NaCNBH 3 , the suspension was incubated at 60°C overnight. Then the gel was acetylated to block remaining free amino groups. As a control, the resin was similarly treated without polysaccharide.
Polysaccharide Binding Assay-Sample protein (3 g) was mixed with 20 l of 50% (v/v) suspension of the polysaccharide-immobilized resin in 1 ml of TBS containing 0.05% Tween 20 at 4°C for 2 h with gentle agitation. After centrifugation, supernatants were separated, and the gels were washed three times with TBS containing 0.05% Tween 20 and three times more with TBS. Proteins bound to the gel or trichloroacetic acid precipitates of the supernatant were dissolved with the 2ϫ SDS-PAGE sample buffer (0.125 M Tris-HCl (pH 6.8), 14% glycerol, 4% SDS, and 0.01% bromphenol blue) in boiling water for 3 min and then subjected to SDS-PAGE. Proteins were visualized by Western blotting or Coomassie Brilliant Blue staining.
BIAcore Analysis-Five microliters of laminarioligosaccharides (25 nmol) in water was incubated at 90°C for 2 h with 5 l of Biotin-LC-Hydrazide (Pierce, 50 nmol), in 30% acetonitrile (25). Ten sets of the reaction mixtures were directly injected onto the surface of the streptavidin-coated sensor chip SA (BIAcore AB) at a flow rate of 5 l/min for 5 min with BIAcore 1000 (BIAcore AB).
Samples in 10 mM HEPES (pH 7.4), 150 mM NaCl, and 0.05% Tween 20 at a concentration of 10 nM to 5 M were passed over the surface of the sensor chip at a flow rate of 2 l/min. The interaction was monitored at 25°C as the change of surface plasmon resonance response. After 5 min of monitoring, the same buffer was introduced onto the sensor chip in place of the protein solution to start the dissociation. At the end of each cycle, regeneration of the chip was accomplished by washing away the surface-bound protein with 4 l of 50 mM H 3 PO 4 . Both the association rate constant (k ass ) and the dissociation rate constant (k diss ) were obtained from the surface plasmon resonance signal binding data and calculated using the BIA-Evaluation software version 2.1 (BIAcore AB). The association constant (K a ) was subsequently determined by k ass /k diss .
Competitive Inhibition Assay of the Factor G Activation-For the competition experiments, purified factor G was activated by curdlan at 37°C for 20 min in the presence of various concentrations of a recombinant protein or oligosaccharide in 200 l of 0.1 M Tris-HCl (pH 8.0) and 0.5 mg/ml bovine serum albumin. The amidolytic activity of the activated factor G was measured after the addition of 20 nmol of t-butyloxycarbonyl-␥-benzyl-L-glutamyl-glycyl-L-arginine 4-methyl-coumaryl-7-amide (Boc-E(OBzl)GR-MCA, Peptide Institute Inc., Osaka) as a substrate, as described previously (13).

Identification of the (133)-␤-D-Glucan-binding Domain of
Factor G-Although our previous studies have demonstrated that factor G is activated by (133)-␤-D-glucans (13), it remains to be determined whether it forms a stable complex with the glucans. We have developed a polysaccharide-binding assay utilizing polysaccharide-immobilized matrices to evaluate the (133)-␤-D-Glucan Recognition by the Horseshoe Crab Factor G glucan-binding abilities of proteins (Fig. 1). Different types of polysaccharides were coupled with a hydrophilic vinyl polymerbased resin, Toyopearl 650M and were incubated with purified factor G. After extensive washing, proteins bound to the resin were subjected to SDS-PAGE, followed by Western blotting with anti-factor G-subunit ␣ and ␤ antibodies, respectively (Fig. 1B). Both subunits were found to be bound with the laminarin (133)-␤-D-glucan)-and the xylan ((134)-␤-D-xylan)coupled resins. The binding appeared to be specific, because neither of the subunits was bound with the mannan ((132)-, (133)-, and (136)-␣-D-mannan)-immobilized resin nor with the control resin that was similarly treated without polysaccharides. The intact subunits were found in the unbound fractions of these resins, indicating that they were not degraded during the incubation.
To determine which subunit is responsible for the glucan binding, we used each subunit individually expressed in insect cells using the baculovirus expression system. Western blotting of each of the subunits ␣and ␤-expressing insect cell lysates indicated that the expressed subunits had the same mobility as that of the purified protein on SDS-PAGE (data not shown). The glucan-binding ability of each subunit in the cell lysate was analyzed as shown in Fig. 1B. Subunit ␣ specifically bound to the laminarin-and the xylan-coupled resins as the purified protein, whereas subunit ␤ did not bind to any of them (Fig.  1C). These results clearly indicate that factor G yields a stable complex through subunit ␣ with laminarin and xylan.
Subunit ␣ of factor G consists of ␤-1,3-glucanase A1 (Gln A1)-like, the xylanase A (Xln A)-like, and the xylanase Z (Xln Z)-like domains (Fig. 1A) (14). To dissect the glucan-binding domain in subunit ␣, we separately expressed the three types of domains in the subunit in bacteria and examined their binding abilities to polysaccharides, as shown in Fig. 1. The Gln A1-like and the Xln A-like domains did not bind to any of the resins (Fig. 2). On the other hand, the expressed Xln Z-like domain specifically bound to the laminarin-immobilized resin. In contrast to the full-length subunit ␣, none of the domains bound to the xylan-immobilized resin.
Because the Xln Z-like domain is composed of two tandemrepeating units with 87% sequence identity (14), we next expressed each repeating unit (amino acid residues 387-524 (designated Xln Z-1) and 525-654 (Xln Z-2)) in this domain and examined its binding ability to laminarin. Even expressed as a single-repeating unit, both fragments bound to laminarin as the Xln Z-like domain, which contains the tandem-repeating units (Fig. 2). Thus, the Xln Z-like domain carries two independent glucan-binding sites.
Competitive Inhibition of Factor G Activation by the Glucanbinding Domain-To evaluate the biological significance of the glucan-binding abilities of the domain, we examined the effects of each domain on the activation of factor G induced by (133)-␤-D-glucan. Factor G was activated by curdlan, a linear (133)-␤-D-glucan, in the presence or absence of 100-fold molar excess of the domains. Following activation, we measured the amidase activity of the activated factor G to estimate the extent of the activation (Fig. 3A). In the absence of the domains, factor G was efficiently activated by curdlan and hydrolyzed a synthetic peptide substrate. Neither the Gln A1-like domain nor the Xln A-like domain showed any effect on the activation. On the other hand, the Xln Z-like domain strongly inhibited the expression of the amidase activity of factor G induced by curdlan. The inhibition was dose-dependent: 50-fold molar excess of the Xln Z-like  1. Glucan binding of factor G. A, the domain structure of the zymogen factor G (14). B and C, polysaccharide-binding abilities of factor G and its subunits. Purified factor G (B) or insect cell lysate containing recombinant subunit ␣ or subunit ␤ was incubated, respectively, with laminarin (G)-, xylan (X)-, and mannan (M)-immobilized resins or control resin (C). The gel-bound (Bound) and unbound (Unbound) materials were subjected to 12.5% SDS-PAGE, followed by immunoblotting with each of the anti-subunits ␣ and ␤ antibodies. See "Experimental Procedures" for details.
(133)-␤-D-Glucan Recognition by the Horseshoe Crab Factor G domain over curdlan inhibited 86.3% of the factor G activation (Fig. 3B). This domain did not affect the amidase activity of the activated factor G (data not shown). These results strongly suggested that the Xln Z-like domain inhibited the activation of factor G by competitively binding to the glucan.
In contrast to the Xln Z-like domain containing the tandemrepeating units, neither the single-repeating units of the Xln Z-like domain alone (Xln Z-1 or Xln Z-2) nor the combination of both (Xln Z-1 ϩ Z-2) inhibited the factor G activation at 100fold molar excess over the ␤-glucan (Fig. 3A). This suggested that the single-repeating units of the Xln Z-like domain have a weaker affinity to (133)-␤-D-glucan than the tandem-repeat structure, which was demonstrated by the following kinetic analyses of the binding.
Kinetic Analysis of Binding Using the BIAcore System-For more quantitative analysis, we further examined the binding of the domains of factor G to (133)-␤-D-glucan using the BIAcore system. Because curdlan is heterogeneous in length (degree of polymerization (d.p.) ϭ ϳ500) and has low solubility, we used a short linear (133)-␤-D-glucan preparation with a number-average molecular weight of 6800 (d.p. ϭ ϳ42), which was reported to have the ability to activate factor G (24). This watersoluble (133)-␤-D-glucan was first derivatized with biotin and fixed onto the surface of a streptavidin-immobilized sensor chip. When 20 nM of the purified factor G was injected onto the glucan-immobilized sensor chip, it bound to the chip time-and dose-dependently, and it dissociated slowly after being washed with buffer (data not shown). No specific binding was detected with a sensor chip without the glucan (data not shown). We also observed the binding of the recombinant Xln Z-like domain, whereas it was not detected with the Gln A-1-like domain and the Xln A-like domain, even at a higher concentration (5 M). Their binding parameters were obtained from the sensorgrams with different concentrations of the ligands (Table I). The association constant (K a ) of the Xln Z-like domain (8.03 ϫ 10 8 M Ϫ1 ) with the glucan was even higher than that of purified factor G (1.51 ϫ 10 8 M Ϫ1 ), supporting the correct folding of the recombinant protein. In addition to the Xln Z-like domain, the single-repeating units, Xln Z-1 and Xln Z-2, also bound to the glucan-immobilized chip as in the glucan-binding assay with the glucan-immobilized resin (Fig. 2). However, their K a values were approximately two orders lower than that of the Xln Z-like domain (Table I), as predicted from the competition assay (Fig. 3A).
Factor G is activated by various types of (133)-␤-D-glucan, but shorter glucans containing less than 7 glucose residues did not activate factor G at all (13). We next examined with laminarioheptaose, a linear (133)-␤-D-glucan containing 7 glucose residues, whether such shorter glucans also bind to factor G. The Xln Z-like domain as well as the purified factor G bound to the laminarioheptaose-immobilized chip (Fig. 4A). Neither the Gln A1-like domain nor the Xln A-like domain bound to the shorter glucan as expected. Their K a values of factor G or Xln Z-like domain for the shorter glucan (6.43 ϫ 10 7 and 3.47 ϫ 10 8 M Ϫ1 , respectively) indicated that the binding was only slightly reduced by shortening the glucan (Table II). On the other hand, when the binding of the single repeats, Xln Z-1 or Xln Z-2, was analyzed, neither of the single-repeating units showed specific binding to the shorter glucan under the same conditions as the experiments with longer glucans (Fig. 4B). When the (133)-␤-D-glucan was further truncated to tetrasaccharide (laminaritetraose) or disaccharide (laminaribiose), both purified factor G and the Xln Z-like domain bound to even the shortest (133)-␤-D-glucan, a glucose-disaccharide with a (133)-␤-D-glucoside linkage (laminaribiose), with the K a values of 1.58 ϫ 10 7 and 5.77 ϫ 10 7 M Ϫ1 , respectively. Neither factor G nor the Xln Z-like domain bound to glucose monomers. Their K a values against the laminaritetraose were in between those with laminariheptaose and laminaribiose (Table II).
Competitive Inhibition of Factor G Activation by Short Oligosaccharides-The analysis described above demonstrated that factor G or the Xln Z-like domain interact with even short oligosaccharides that do not have the ability to activate factor G. The differences between the binding constants for factor G activation-competent longer glucans (d.p. ϭ ϳ42) and incompetent shorter glucans (laminarioheptaose) were within 3-fold. Thus, we next investigated the effects of such shorter oligosaccharides upon the activation of the zymogen factor G induced by the longer glucans. When the zymogen factor G was prein-

(133)-␤-D-Glucan Recognition by the Horseshoe Crab Factor G
cubated with the short oligosaccharides, the factor G activation by curdlan was dose-dependently inhibited (Fig. 5). Neither glucose, cellobiose ((134)-␤-D-glucan), nor gentiobiose ((136)-␤-D-glucan) inhibited activation, even at the higher concentrations (Ͻ10 6 -fold molar excess of curdlan). As shown in Fig. 5, the longer oligosaccharides inhibited the activation at a greater level of efficiency than the smaller ones, which is consistent with the affinity determined by the BIAcore experiments. Thus, the shorter glucans act as a competitive inhibitor in factor G activation by the longer glucans by binding to the glucan-binding site of factor G.

DISCUSSION
In the present study, the horseshoe crab factor G was shown to form a stable complex with (133)-␤-D-glucan on fungal cell walls. Even after binding to the glucan, subunits ␣ and ␤, which are associated by one or more non-covalent bonds (13), were held together (Fig. 1B). Accordingly, after factor G was activated by fungi, the active protease (subunit ␤) would be kept associated on the surface of the fungus through (133)-␤-D-glucan and subunit ␣. Thus, the protease is prevented from diffusing throughout the hemolymph, which would cause unnecessary or unfavorable activation of hemolymph clotting at any sites outside of a local inflammatory region.
We identified the Xln Z-like domain, located at the COOH terminus of subunit ␣ as the (133)-␤-D-glucan-binding site of factor G. Among three types of domains in the glucan-binding subunit ␣, only the Xln Z-like domain bound to the glucan (Fig.  2) and competitively inhibited the glucan-mediated activation of factor G (Fig. 3). The K a value for the recombinant Xln Z-like domain (8.03 ϫ 10 8 M Ϫ1 ) was comparable to, or somewhat higher than, that of purified factor G (1.51 ϫ 10 8 M Ϫ1 ) (Table I), supporting our conclusion that the domain is the primary glucan-binding site of the protein. The higher K a of the Xln Z-like domain than that of the purified factor G is mostly due to higher association rate constant (k ass ), suggesting steric hindrance of the binding site or slower diffusion of the intact protein with a larger molecular mass in solution.
The Xln Z-like domain shows partial sequence similarities with polysaccharide-hydrolases isolated from various bacteria, such as xylanases A, B, U, V, and Z from Clostridium thermocellum (26 -28), xylanase A from C. stercorarium (29), xylanase D from Bacillus polymyxa (30), cellulase B from Cellvibrio mixtus (31), and ␣-1,6-mannanase from B. circulans (32). Based on its primary structure, these domains homologous to the Xln Z-like domain are classified into family VI of the cellulose-binding domain (CBD) (33). Some proteins contain tandem-repeating CBDs such as factor G, whereas the others have a single CBD. Some of the family VI CBDs have been shown to bind to xylan and/or cellulose with different affinities (26,27,31,34). Because the biochemical characteristics of many CBDs have not yet been extensively analyzed, some of them may have affinity for (133)-␤-D-glucan. Although this type of CBD has not been found in eukaryotes except for factor G, its discovery in the horseshoe crab and its functional importance in the defense system imply that it might be present in other animals as a functional unit for detecting fungal (133)-␤-D-glucan. Despite recent extensive studies on the role of toll-like receptors in response to bacterial products, molecules involved in the responses to fungi are poorly understood in mammalian innate immunity.
In addition to (133)-␤-D-glucan, subunit ␣ also bound to (134)-␤-D-xylan (Fig. 1). The physiological significance of this binding is currently unknown, however, because xylan does not activate the zymogen factor G efficiently (13). In contrast to the binding to (133)-␤-D-glucan, the binding of the recombinant Xln Z-like domains to the xylan-immobilized resin was not observed under the condition where the full-length subunit ␣ bound to it (Fig. 2). Accordingly, more precise conformation in the subunit or coordinated interaction with other domains in subunit ␣ appears to be required for the rigid binding.
In contrast to the Xln Z-like domain with the double CBD, neither single CBD (Xln Z-1 or -2) exhibited any effect on factor G activation, although both of them bound to the glucan (Figs. 2 and 3). The reason for this apparent conflict was explained after the quantitative analysis of the glucan binding: K a of the single CBD was approximately two orders lower than that of the double CBD, thus indicating that the single CBD does not form a stable complex with the glucan in solution (Table I). The analysis using BIAcore allowed for direct comparisons of the association and dissociation rate constants between the single and double CBDs. Their k ass values are nearly equal between the single and the double CBDs, whereas the dissociation rate constant (k diss ) for the single CBD is ϳ100-fold larger than that for the double CBD. These results indicate that, although both single and double CBDs associate with the glucan at the same rate, the single CBD dissociates from it more quickly. The difference of the binding of the two types of proteins could be compared with the difference of the stability between a unicycle and a bicycle (Fig. 6). Bicycles are more stable on the ground than are unicycles because of the two wheels, corresponding to the two binding sites of the double CBD. Because of the two binding sites, the glucan binding of the double CBD is constituted with equilibrium of four different states (I-IV in Fig. 6), in three of which the two substances are bound (states II, III, Amidolytic activity of activated factor G was measured by a peptidyl substrate, as described under "Experimental Procedures." The extent of the activation of factor G is shown as the percent relative to that in the absence of the laminarioligosaccharides. The increased affinity of genetically engineered double CBD has also been reported for Trichoderma reesei cellobiohydrolases (35). Surprisingly, factor G and the recombinant Xln Z-like domain interacted with short (133)-␤-D-glucan oligosaccharides that do not have the ability to activate the zymogen factor G (Table II, Fig. 5). Although K a became 10-fold lower than that for the longer glucan with a d.p. of ϳ42, factor G still bound to (133)-␤-D-glucan disaccharide (laminaribiose) but not to glucose monomers. Thus, the minimum requirement for the binding is a (133)-␤-D-glucosidic linkage. The decrement of K a values with the shorter oligosaccharides is mainly due to the increment of k diss values, indicating that multiple binding sites present on longer (133)-␤-D-glucans inhibit the dissociation of the proteins. In contrast to the double CBD, the single CBD (Xln Z-1 and Xln Z-2) did not exhibit measurable binding to the oligosaccharides. The clear differences between the single and double CBDs for binding to the oligosaccharides further support that the two tandem CBDs significantly stabilize the binding. The quick transition from state II to III should occur, because stabilization by the double CBD was observed even with the disaccharide, which contains only one binding site and is, therefore, unable to create the complex in state IV (Fig. 6).
The short oligosaccharides with fewer than seven glucose residues do not have the ability to activate factor G (13). The present study showed that these oligosaccharides also bind to factor G. The differences between the binding constants for factor G activation-competent longer glucans (d.p. ϭ ϳ42) and incompetent shorter glucans (d.p. ϭ 7) were within 3-fold. Therefore, it is unlikely that longer glucans have a specific conformation-dependent structure that is required for recognition by factor G. The observation that the short oligosaccha-rides competitively inhibit the activation by the longer glucan provides further evidence for the binding of factor G with these short oligosaccharides. These findings further support our previous model on the factor G activation, in which the activation requires an intermolecular interaction between two factor G molecules on a single (133)-␤-D-glucan chain (13). The shorter oligosaccharides are incapable of inducing the factor G activation, primarily because they are too short to function as a template for the interaction of two factor G molecules (Fig. 6B).
Thus, factor G contains two binding sites for the (133)-␤-Dglucosidic linkage between glucose residues of the glucan, which are multiply present on a single (133)-␤-D-glucan strand. Because the binding between the single CBD and the disaccharide are below the detection limit, the single interaction units have weak affinity. However, factor G and the glucan form a stable complex, because multiple binding sites are present on both of the molecules. These interactions between multivalent molecules are reminiscent of the interaction between selectins and their glycosylated ligands (36). Weak and multivalent interactions between the lectin domain of selectin and carbohydrate chains on the ligand allow "rolling" of the blood cells, which is essential for the initiation of inflammation (37,38). Large k ass and k diss values for the interaction between each binding unit of factor G (Xln Z-1 or Xln Z-2) and the glucan, as found in the interaction between selectins and their ligands (39,40), suggest that "sliding" of factor G molecules on the glucan strand may occur. Such "sliding" would increase the possibilities of the collapse between factor G molecules, which is essential for the activation of factor G.
In summary, we identified the (133)-␤-D-glucan-binding site on factor G and characterized the binding. This is the first study to provide the molecular basis for the defense mechanism responding to (133)-␤-D-glucan found in the cell surface of fungi. The minimum structure for the recognition by factor G is a (133)-␤-D-glucosidic linkage. The weak binding of the single binding unit is stabilized by multiple interactions between the two tandem binding sites on factor G and multiple (133)-␤-Dglucosidic linkages on the glucan. The binding itself is not sufficient for the activation of factor G, and sufficiently long glucans, which would be pathologically more important, are required for the activation to concentrate factor G molecules and to provide a template allowing the interactions between factor G molecules. Therefore, this protein functions as a biosensor for the longer (133)-␤-D-glucan present on pathogenic fungi. As found in the recognition of (133)-␤-D-glucan by factor G, the multivalent recognition of a small characteristic structure on the biological key molecules may be one of the principles for pattern recognition in innate immunity. FIG. 6. Schematic illustrations of the glucan binding of factor G with double CBDs in comparison with a bicycle with two wheels. A, the glucan binding of double CBDs is more stable than that of a single CBD, just as a bicycle is more stable than a unicycle. Because of the two binding sites, the binding between the protein with double CBDs and glucan is composed of an equilibrium of four different states (I-IV). B, a model for the inhibitory effects of the short glucan on the factor G activation. A long glucan allows the binding of multiple factor G molecules on a single chain, which is essential for the activation (13). On the other hand, a short glucan binds to factor G but does not have enough length to make the activation complexes that allow the collapse of factor G molecules, and it thus functions as a competitive inhibitor for the activation. See text for details.