Nematotoxicity of Marasmius oreades agglutinin (MOA) depends on glycolipid binding and cysteine protease activity

Fruiting body lectins have been proposed to act as effector proteins in the defense of fungi against parasites and predators. The Marasmius oreades agglutinin (MOA) is a Galα1,3Gal/GalNAc-specific lectin from the fairy ring mushroom that consists of an N-terminal ricin B-type lectin domain and a C-terminal dimerization domain. The latter domain shows structural similarity to catalytically active proteins, suggesting that, in addition to its carbohydrate-binding activity, MOA has an enzymatic function. Here, we demonstrate toxicity of MOA toward the model nematode Caenorhabditis elegans. This toxicity depends on binding of MOA to glycosphingolipids of the worm via its lectin domain. We show further that MOA has cysteine protease activity and demonstrate a critical role of this catalytic function in MOA-mediated nematotoxicity. The proteolytic activity of MOA was dependent on high Ca(2+) concentrations and favored by slightly alkaline pH, suggesting that these conditions trigger activation of the toxin at the target location. Our results suggest that MOA is a fungal toxin with intriguing similarities to bacterial binary toxins and has a protective function against fungivorous soil nematodes.     

De novo prediction of three-dimensional structures for major protein families

The family 10 xylanase from Streptomyces olivaceoviridis E-86 contains a (beta/alpha)(8)-barrel as a catalytic domain, a family 13 carbohydrate binding module (CBM) as a xylan binding domain (XBD) and a Gly/Pro-rich linker between them. The crystal structure of this enzyme showed that XBD has three similar subdomains, as indicated by the presence of a triple-repeated sequence, forming a galactose binding lectin fold similar to that found in the ricin toxin B-chain. Comparison with the structure of ricin/lactose complex suggests three potential sugar binding sites in XBD. In order to understand how XBD binds to the xylan chain, we analyzed the sugar-complex structure by the soaking experiment method using the xylooligosaccharides and other sugars. In the catalytic cleft, bound sugars were observed in the xylobiose and xylotriose complex structures. In the XBD, bound sugars were identified in subdomains alpha and gamma in all of the complexes with xylose, xylobiose, xylotriose, glucose, galactose and lactose. XBD binds xylose or xylooligosaccharides at the same sugar binding sites as in the case of the ricin/lactose complex but its binding manner for xylose and xylooligosaccharides is different from the galactose binding mode in ricin, even though XBD binds galactose in the same manner as in the ricin/galactose complex. These different binding modes are utilized efficiently and differently to bind the long substrate to xylanase and ricin-type lectin. XBD can bind any xylose in the xylan backbone, whereas ricin-type lectin recognizes the terminal galactose to sandwich the large sugar chain, even though the two domains have the same family 13 CBM structure. Family 13 CBM has rather loose and broad sugar specificities and is used by some kinds of proteins to bind their target sugars. In such enzyme, XBD binds xylan, and the catalytic domain may assume a flexible position with respect to the XBD/xylan complex, inasmuch as the linker region is unstructured.     

Structure of ricin B-chain at 2.5 A resolution

The heterodimeric plant toxin ricin has been refined to 2.5 A resolution. The B-chain lectin (RTB) is described in detail. The protein has two major domains, each of which has a galactose binding site. RTB has no regular secondary structure but displays several omega loops. Each RTB domain is made of three copies of a primitive 40 residue folding unit, which pack around a pseudo threefold axis. In each domain, galactose binds in a shallow cleft formed by a three residue peptide kink on the bottom and an aromatic ring on the top. At the back of the cleft, an aspartate forms hydrogen bonds to the C3 and C4 hydroxyls of galactose, whereas a glutamine bonds to the C4 alcohol, helping to define specific epimer binding. In addition to analyzing the sugar binding mechanism, the assembly of subdomain units around the pseudo threefold axis of each domain is described. The subdomains contribute conserved Trp, Leu, and Ile residues to a compact central hydrophobic core. This tight threefold binding probably drives the peptide folding and stabilizes the protein structure.     

Spatial distribution of cytoplasmic domains of the Mg-transporter MgtE,in a solution lacking Mg,revealed by paramagnetic relaxation enhancement

MgtE is a prokaryotic Mg²⁺ transporter that controls cellular Mg²⁺ concentrations. We previously reported crystal structures of the cytoplasmic region of MgtE, consisting of 2 domains, that is, N and CBS, in the Mg²⁺-free and Mg²⁺-bound forms. The Mg²⁺-binding sites lay at the interface of the 2 domains, making the Mg²⁺-bound form compact and globular. In the Mg²⁺-free structure, however, the domains are far apart, and the Mg²⁺-binding sites are destroyed. Therefore, it is unclear how Mg²⁺-free MgtE changes its conformation to accommodate Mg²⁺ ions. Here, we used paramagnetic relaxation enhancement (PRE) to characterize the relative orientation of the N and CBS domains in the absence of Mg²⁺ in solution. When the residues on the surface of the CBS domain were labeled with nitroxide tags, significant PRE effects were observed for the residues in the N domain. No single structure satisfied the PRE profiles, suggesting that the N and CBS domains are not fixed in a particular orientation in solution. We then conducted ensemble simulated annealing calculations in order to obtain the atomic probability density and visualize the spatial distribution of the N domain in solution. The results indicate that the N domain tends to occupy the space near its position in the Mg²⁺-bound crystal structure, facilitating efficient capture of Mg²⁺ with increased intracellular Mg²⁺ concentration, which is necessary to close the gate.     

Structure of N-acetyl-beta-D-glucosaminidase (GcnA) from the endocarditis pathogen Streptococcus gordonii and its complex with the mechanism-based inhibitor NAG-thiazoline

The crystal structure of GcnA, an N-acetyl-β-d-glucosaminidase from Streptococcus gordonii, was solved by multiple wavelength anomalous dispersion phasing using crystals of selenomethionine-substituted protein. GcnA is a homodimer with subunits each comprised of three domains. The structure of the C-terminal α-helical domain has not been observed previously and forms a large dimerisation interface. The fold of the N-terminal domain is observed in all structurally related glycosidases although its function is unknown. The central domain has a canonical (β/α)₈ TIM-barrel fold which harbours the active site. The primary sequence and structure of this central domain identifies the enzyme as a family 20 glycosidase. Key residues implicated in catalysis have different conformations in two different crystal forms, which probably represent active and inactive conformations of the enzyme. The catalytic mechanism for this class of glycoside hydrolase, where the substrate rather than the enzyme provides the cleavage-inducing nucleophile, has been confirmed by the structure of GcnA complexed with a putative reaction intermediate analogue, N-acetyl-β-d-glucosamine-thiazoline. The catalytic mechanism is discussed in light of these and other family 20 structures.     

Tertiary structure and carbohydrate recognition by the chitin-binding domain of a hyperthermophilic chitinase from Pyrococcus furiosus

A chitinase is a hyperthermophilic glycosidase that effectively hydrolyzes both α and β crystalline chitins; that studied here was engineered from the genes PF1233 and PF1234 of Pyrococcus furiosus. This chitinase has unique structural features and contains two catalytic domains (AD1 and AD2) and two chitin-binding domains (ChBDs; ChBD1 and ChBD2). A partial enzyme carrying AD2 and ChBD2 also effectively hydrolyzes crystalline chitin. We determined the NMR and crystal structures of ChBD2, which significantly enhances the activity of the catalytic domain. There was no significant difference between the NMR and crystal structures. The overall structure of ChBD2, which consists of two four-stranded β-sheets, was composed of a typical β-sandwich architecture and was similar to that of other carbohydrate-binding module 2 family proteins, despite low sequence similarity. The chitin-binding surface identified by NMR was flat and contained a strip of three solvent-exposed Trp residues (Trp274, Trp308 and Trp326) flanked by acidic residues (Glu279 and Asp281). These acidic residues form a negatively charged patch and are a characteristic feature of ChBD2. Mutagenesis analysis indicated that hydrophobic interaction was dominant for the recognition of crystalline chitin and that the acidic residues were responsible for a higher substrate specificity of ChBD2 for chitin compared with that of cellulose. These results provide the first structure of a hyperthermostable ChBD and yield new insight into the mechanism of protein-carbohydrate recognition. This is important in the development of technology for the exploitation of biomass.     

How to Switch Off a Histidine Kinase: Crystal Structure of Geobacillus stearothermophilus KinB with the inhibitor Sda

Entry to sporulation in bacilli is governed by a histidine kinase phosphorelay, a variation of the predominant signal transduction mechanism in prokaryotes. Sda directly inhibits sporulation histidine kinases in response to DNA damage and replication defects. We determined a 2.0-Å-resolution X-ray crystal structure of the intact cytoplasmic catalytic core [comprising the dimerization and histidine phosphotransfer domain (DHp domain), connected to the ATP binding catalytic domain] of the Geobacillus stearothermophilus sporulation kinase KinB complexed with Sda. Structural and biochemical analyses reveal that Sda binds to the base of the DHp domain and prevents molecular transactions with the DHp domain to which it is bound by acting as a simple molecular barricade. Sda acts to sterically block communication between the catalytic domain and the DHp domain, which is required for autophosphorylation, as well as to sterically block communication between the response regulator Spo0F and the DHp domain, which is required for phosphotransfer and phosphatase activities.     

Crystal Structures of the CBS and DRTGG Domains of the Regulatory Region of Clostridiumperfringens Pyrophosphatase Complexed with the Inhibitor, AMP, and Activator, Diadenosine Tetraphosphate

Nucleotide-binding cystathionine β-synthase (CBS) domains serve as regulatory units in numerous proteins distributed in all kingdoms of life. However, the underlying regulatory mechanisms remain to be established. Recently, we described a subfamily of CBS domain-containing pyrophosphatases (PPases) within family II PPases. Here, we express a novel CBS-PPase from Clostridium perfringens (CPE2055) and show that the enzyme is inhibited by AMP and activated by a novel effector, diadenosine 5′,5-P1,P4-tetraphosphate (AP₄A). The structures of the AMP and AP₄A complexes of the regulatory region of C. perfringens PPase (cpCBS), comprising a pair of CBS domains interlinked by a DRTGG domain, were determined at 2.3 Å resolution using X-ray crystallography. The structures obtained are the first structures of a DRTGG domain as part of a larger protein structure. The AMP complex contains two AMP molecules per cpCBS dimer, each bound to a single monomer, whereas in the activator-bound complex, one AP₄A molecule bridges two monomers. In the nucleotide-bound structures, activator binding induces significant opening of the CBS domain interface, compared with the inhibitor complex. These results provide structural insight into the mechanism of CBS-PPase regulation by nucleotides.     

Crystal Structure of the GalNAc/Gal-Specific Agglutinin from the Phytopathogenic Ascomycete Sclerotinia sclerotiorum Reveals Novel Adaptation of a β-Trefoil Domain

A lectin from the phytopathogenic ascomycete Sclerotinia sclerotiorum that shares only weak sequence similarity with characterized fungal lectins has recently been identified. S. sclerotiorum agglutinin (SSA) is a homodimeric protein consisting of two identical subunits of ∼ 17 kDa and displays specificity primarily towards Gal/GalNAc. Glycan array screening indicates that SSA readily interacts with Gal/GalNAc-bearing glycan chains. The crystal structures of SSA in the ligand-free form and in complex with the Gal-β1,3-GalNAc (T-antigen) disaccharide have been determined at 1.6 and 1.97 Å resolution, respectively. SSA adopts a β-trefoil domain as previously identified for other carbohydrate-binding proteins of the ricin B-like lectin superfamily and accommodates terminal non-reducing galactosyl and N-acetylgalactosaminyl glycans. Unlike other structurally related lectins, SSA contains a single carbohydrate-binding site at site α. SSA reveals a novel dimeric assembly markedly dissimilar to those described earlier for ricin-type lectins. The present structure exemplifies the adaptability of the β-trefoil domain in the evolution of fungal lectins.     

Structural analysis of the PP2C phosphatase tPphA from Thermosynechococcus elongatus: a flexible flap subdomain controls access to the catalytic site

The homologue of the phosphoprotein PII phosphatase PphA from Thermosynechococcus elongatus, termed tPphA, was identified and its structure was resolved in two different space groups, C222₁ and P4₁2₁2, at a resolution of 1.28 and 3.05 Å, respectively. tPphA belongs to a large and widely distributed subfamily of Mg²⁺/Mn²⁺-dependent phosphatases of the PPM superfamily characterized by the lack of catalytic and regulatory domains. The core structure of tPphA shows a high degree of similarity to the two PPM structures identified so far. In contrast to human PP2C, but similar to Mycobacterium tuberculosis phosphatase PstP, the catalytic centre exhibits a third metal ion in addition to the dinuclear metal centre universally conserved in all PPM members. The fact that the third metal is only liganded by amino acids, which are universally conserved in all PPM members, implies that the third metal could be general for all members of this family. As a specific feature of tPphA, a flexible subdomain, previously recognized as a flap domain, could be revealed. Comparison of different structural isomers of tPphA as well as site-specific mutagenesis implied that the flap domain is involved in substrate binding and catalytic activity. The structural arrangement of the flap domain was accompanied by a large side-chain movement of an Arg residue (Arg169) at the basis of the flap. Mutation of this residue strongly impaired protein stability as well as catalytic activity, emphasizing the importance of this amino acid for the regional polysterism of the flap subdomain and confirming the assumption that flap domain flexibility is involved in catalysis.     

Crystal structure of Streptomyces olivaceoviridis E-86 beta-xylanase containing xylan-binding domain

Xylanases hydrolyse the beta-1,4-glycosidic bonds within the xylan backbone and belong to either family 10 or 11 of the glycoside hydrolases, on the basis of the amino acid sequence similarities of their catalytic domains. Generally, xylanases have a core catalytic domain, an N and/or C-terminal substrate-binding domain and a linker region. Until now, X-ray structural analyses of family 10 xylanases have been reported only for their catalytic domains and do not contain substrate-binding domains. We have determined the crystal structure of a family 10 xylanase containing the xylan-binding domain (XBD) from Streptomyces olivaceoviridis E-86 at 1.9 A resolution. The catalytic domain comprises a (beta/alpha)(8)-barrel topologically identical to other family 10 xylanases. XBD has three similar subdomains, as suggested from a triple-repeat sequence, which are assembled against one another around a pseudo-3-fold axis, forming a galactose-binding lectin fold similar to ricin B-chain. The Gly/Pro-rich linker region connecting the catalytic domain and XBD is not visible in the electron density map, probably because of its flexibility. The interface of the two domains in the crystal is hydrophilic, where five direct hydrogen bonds and water-mediated hydrogen bonds exist. The sugar-binding residues seen in ricin/lactose complex are spatially conserved among the three subdomains in XBD, suggesting that all of the subdomains in XBD have the capacity to bind sugars. The flexible linker region enables the two domains to move independently and may provide a triple chance of substrate capturing and catalysis. The structure reported here represents an example where the metabolic enzyme uses a ricin-type lectin motif for capturing the insoluble substrate and promoting catalysis.     

Structure of Rhamnose-binding Lectin CSL3: Unique Pseudo-tetrameric Architecture of a Pattern Recognition Protein

The crystal structure of the l-rhamnose-binding lectin CSL3 was determined to 1.8 Å resolution. This protein is a component of the germline-encoded pattern recognition proteins in innate immunity. CSL3 is a homodimer of two 20 kDa subunits with a dumbbell-like shape overall, in which the N- and C-terminal domains of different subunits form lobe structures connected with flexible linker peptides. The complex structures of the protein with specific carbohydrates demonstrated the importance of the most variable loop region among homologues for the specificity toward oligosaccharides. CSL3 and Shiga-like toxin both use Gb₃ as a cellular receptor to evoke apoptosis. They have very different overall architecture but share the separation distance between carbohydrate-binding sites. An inspection of the structure database suggested that the pseudo-tetrameric structure of CSL3 was unique among the known lectins. This architecture implies this protein might provide a unique tool for further investigations into the relationships between architecture and function of pattern recognition proteins.     

Crystal structure of crotoxin reveals key residues involved in the stability and toxicity of this potent heterodimeric β-neurotoxin

The crystal structure of crotoxin, a potent presynaptic neurotoxin from Crotalus durissusterrificus, was solved at 1.35 Å resolution. It shows the architecture of the three disulfide-linked polypeptide chains (α, β, and γ) of the acidic subunit CA noncovalently complexed with the basic phospholipase A₂ (PLA₂) subunit CB. The unique structural scaffold of the association of the CA and CB subunits indicates that posttranslational cleavage of the pro-CA precursor is a prerequisite for the assembly of the CA-CB complex. These studies provide novel structural insights to explain the role of the CA subunit in the mechanism of action of crotoxin. The crystal structure of the highly toxic and stable CA₂CBb complex crystallized here allows us to identify key amino acid residues responsible for significant differences in the pharmacological activities of the two classes of crotoxin complexes. In particular, we show that critical residues Trp31 and Trp70 of the CBb subunit establish intermolecular polar contacts with Asp99 and Asp89, respectively, of the β-chain of CA₂ and contribute to the stability and toxicity of the CA₂CBb complex. These interactions also lead to decreased PLA₂ activity by partially blocking substrate access to the catalytic dyad and by masking several interfacial binding surface residues important for PLA₂ interaction with phospholipids.Identification of the binding interface between the CA subunits and the CB subunits of crotoxin is important for the structure-based design of antineurotoxic inhibitors. Since crotoxin displays numerous physiological functions, including antitumoral properties, knowledge of its three-dimensional structure will be useful for the understanding of these diverse effects.     

The molecular structure of epoxide hydrolase B from Mycobacterium tuberculosis and its complex with a urea-based inhibitor

Mycobacterium tuberculosis (Mtb), the intracellular pathogen that infects macrophages primarily, is the causative agent of the infectious disease tuberculosis in humans. The Mtb genome encodes at least six epoxide hydrolases (EHs A to F). EHs convert epoxides to trans-dihydrodiols and have roles in drug metabolism as well as in the processing of signaling molecules. Herein, we report the crystal structures of unbound Mtb EHB and Mtb EHB bound to a potent, low-nanomolar (IC₅₀ ≈ 19 nM) urea-based inhibitor at 2.1 and 2.4 Å resolution, respectively. The enzyme is a homodimer; each monomer adopts the classical α/β hydrolase fold that composes the catalytic domain; there is a cap domain that regulates access to the active site. The catalytic triad, comprising Asp104, His333 and Asp302, protrudes from the catalytic domain into the substrate binding cavity between the two domains. The urea portion of the inhibitor is bound in the catalytic cavity, mimicking, in part, the substrate binding; the two urea nitrogen atoms donate hydrogen bonds to the nucleophilic carboxylate of Asp104, and the carbonyl oxygen of the urea moiety receives hydrogen bonds from the phenolic oxygen atoms of Tyr164 and Tyr272. The phenolic oxygen groups of these two residues provide electrophilic assistance during the epoxide hydrolytic cleavage. Upon inhibitor binding, the binding-site residues undergo subtle structural rearrangement. In particular, the side chain of Ile137 exhibits a rotation of around 120° about its C^α-C^β bond in order to accommodate the inhibitor. These findings have not only shed light on the enzyme mechanism but also have opened a path for the development of potent inhibitors with good pharmacokinetic profiles against all Mtb EHs of the α/β type.     

Conserved Hydrophobic Clusters on the Surface of the Caf1A Usher C-Terminal Domain Are Important for F1 Antigen Assembly

The outer membrane usher protein Caf1A of the plague pathogen Yersinia pestis is responsible for the assembly of a major surface antigen, the F1 capsule. The F1 capsule is mainly formed by thin linear polymers of Caf1 (capsular antigen fraction 1) protein subunits. The Caf1A usher promotes polymerization of subunits and secretion of growing polymers to the cell surface. The usher monomer (811 aa, 90.5 kDa) consists of a large transmembrane β-barrel that forms a secretion channel and three soluble domains. The periplasmic N-terminal domain binds chaperone-subunit complexes supplying new subunits for the growing fiber. The middle domain, which is structurally similar to Caf1 and other fimbrial subunits, serves as a plug that regulates the permeability of the usher. Here we describe the identification, characterization, and crystal structure of the Caf1A usher C-terminal domain (Caf1A_C). Caf1A_C is shown to be a periplasmic domain with a seven-stranded β-barrel fold. Analysis of C-terminal truncation mutants of Caf1A demonstrated that the presence of Caf1A_C is crucial for the function of the usher in vivo, but that it is not required for the initial binding of chaperone-subunit complexes to the usher. Two clusters of conserved hydrophobic residues on the surface of Caf1A_C were found to be essential for the efficient assembly of surface polymers. These clusters are conserved between the FGL family and the FGS family of chaperone-usher systems.     

Directly observed hydrogen bonds at calcium-binding-sites of calmodulin in solution relate to affinity of the calcium-binding

Molecular tuning to calcium-binding in the EF-hand motif of holo-calmodulin was studied in solution by NMR ^h3J_NC′ H-bond couplings. In the N-terminus lobe of holo-calmodulin, the glutamate crucial for Ca²⁺ coordination has network of H-bonds weaker than inferred from the X-ray crystal structure. This glutamate at position 12 appears shifted away from the Ca²⁺ preferred coordination, which can explain the lower affinity of the calcium-binding to the N-terminus with respect to C-terminus EF hands.     

Structure of the mosquitocidal toxin from Bacillus sphaericus

The catalytic domain of a mosquitocidal toxin prolonged by a C-terminal 44 residue linker connecting to four ricin B-like domains was crystallized. Three crystal structures were established at resolutions between 2.5A and 3.0A using multi-wavelength and single-wavelength anomalous X-ray diffraction as well as molecular replacement phasing techniques. The chainfold of the toxin fragment corresponds to those of ADP-ribosylating enzymes. At pH 4.3 the fragment is associated in a C(7)-symmetric heptamer in agreement with an aggregate of similar size observed by size-exclusion chromatography. In two distinct crystal forms, the heptamers formed nearly spherical, D(7)-symmetric tetradecamers. Another crystal form obtained at pH 6.3 contained a recurring C(2)-symmetric tetramer, which, however, was not stable in solution. On the basis of the common chainfold and NAD(+)-binding site of all ADP-ribosyl transferases, the NAD(+)-binding site of the toxin was assigned at a high confidence level. In all three crystal forms the NAD(+) site was occupied by part of the 44 residue linker, explaining the known inhibitory effect of this polypeptide region. The structure showed that the cleavage site for toxin activation is in a highly mobile loop that is exposed in the monomer. Since it contains the inhibitory linker as a crucial part of the association contact, the observed heptamer is inactive. Moreover, the heptamer cannot be activated by proteolysis because the activation loop is at the ring center and not accessible for proteases. Therefore the heptamer, or possibly the tetradecamer, seems to represent an inactive storage form of the toxin.     

Structural Evidence for the Functional Importance of the Heme Domain Mobility in Flavocytochrome b2

Yeast flavocytochrome b₂ (Fcb2) is an l-lactate:cytochrome c oxidoreductase in the mitochondrial intermembrane space participating in cellular respiration. Each enzyme subunit consists of a cytochrome b₅-like heme domain and a flavodehydrogenase (FDH) domain. In the Fcb2 crystal structure, the heme domain is mobile relative to the tetrameric FDH core in one out of two subunits. The monoclonal antibody B2B4, elicited against the holoenzyme, recognizes only the native heme domain in the holoenzyme. When bound, it suppresses the intramolecular electron transfer from flavin to heme b₂, hence cytochrome c reduction. We report here the crystal structure of the heme domain in complex with the Fab at 2.7 Å resolution. The Fab epitope on the heme domain includes the two exposed propionate groups of the heme, which are hidden in the interface between the domains in the complete subunit. The structure discloses an unexpected plasticity of Fcb2 in the neighborhood of the heme cavity, in which the heme has rotated. The epitope overlaps with the docking area of the FDH domain onto the heme domain, indicating that the antibody displaces the heme domain in a movement of large amplitude. We suggest that the binding sites on the heme domain of cytochrome c and of the FDH domain also overlap and therefore that cytochrome c binding also requires the heme domain to move away from the FDH domain, so as to allow electron transfer between the two hemes. Based on this hypothesis, we propose a possible model of the Fcb2·cytochrome c complex. Interestingly, this model shares similarity with that of the cytochrome b₅·cytochrome c complex, in which cytochrome c binds to the surface around the exposed heme edge of cytochrome b₅. The present results therefore support the idea that the heme domain mobility is an inherent component of the Fcb2 functioning.     

Crystal Structure of the HA3 Subcomponent of Clostridium botulinum Type C Progenitor Toxin

The Clostridium botulinum type C 16S progenitor toxin contains a neurotoxin and several nontoxic components, designated nontoxic nonhemagglutinin (HA), HA1 (HA-33), HA2 (HA-17), HA3a (HA-22-23), and HA3b (HA-53). The HA3b subcomponent seems to play an important role cooperatively with HA1 in the internalization of the toxin by gastrointestinal epithelial cells via binding of these subcomponents to specific oligosaccharides. In this study, we investigated the sugar-binding specificity of the HA3b subcomponent using recombinant protein fused to glutathione S-transferase and determined the three-dimensional structure of the HA3a-HA3b complex based on X-ray crystallography. The crystal structure was determined at a resolution of 2.6 Å. HA3b contains three domains, domains I to III, and the structure of domain I resembles HA3a. In crystal packing, three HA3a-HA3b molecules are assembled to form a three-leaved propeller-like structure. The three HA3b domain I and three HA3a alternate, forming a trimer of dimers. In a database search, no proteins with high structural homology to any of the domains (Z score > 10) were found. Especially, HA3a and HA3b domain I, mainly composed of β-sheets, reveal a unique fold. In binding assays, HA3b bound sialic acid with high affinity, but did not bind galactose, N-acetylgalactosamine, or N-acetylglucosamine. The electron density of liganded N-acetylneuraminic acid was determined by crystal soaking. In the sugar-complex structure, the N-acetylneuraminic acid-binding site was located in the cleft formed between domains II and III of HA3b. This report provides the first determination of the three-dimensional structure of the HA3a-HA3b complex and its sialic acid binding site. Our results will provide useful information for elucidating the mechanism of assembly of the C16S toxin and for understanding the interactions with oligosaccharides on epithelial cells and internalization of the botulinum toxin complex.